PLACE IN RETURN BOX to remove this checkout from your record. To AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE 5/08 K'IProj/Acc8-Pres/CIRC/DateDue nndd A MULTILEVEL EXAMINATION OF THE ANTECEDENTS OF PROCEDURAL. DISTRIBUTIVE, AND RESTORATIVE JUSTICE EXPECTATIONS AND RELATED OUTCOMES IN THE CONTEXT OF STRONG PREFERENTIAL SELECTION IN SOUTH AFRICA By Lauren Jill Ramsay A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Psychology 2009 ABSTRACT A MULTILEVEL EXAMINATION OF THE ANTECEDENTS OF PROCEDURAL, DISTRIBUTIVE, AND RESTORATIVE JUSTICE EXECTATIONS AND RELATED OUTCOMES IN THE CONTEXT OF STRONG PREFERENTIAL SELECTION IN SOUTH AFRICA By Lauren Jill Ramsay This quasi-experimental study expands our understanding of justice theory dimensions and their antecedents and outcomes by examining attitudes and experiences in a sample of potential job applicants in South Africa. This work extends existing four- dimension justice theory by proposing and validating a fifth dimension of justice ~ restorative justice. South African workplaces operate under labour law that specifies strong preferential selection to promote the hiring of individuals from previously disadvantaged groups. The current study uses this existing selection context as a naturally occurring manipulation of strong preferential treatment in selection to empirically test a multilevel model of justice expectation antecedents, the dimensions of justice expectations, and related outcomes. Results of this study indicate that restorative justice as a distinct justice dimension beyond procedural, distributive, interpersonal, and informational dimensions, is supported. Further, three group-level constructs — group belief in white guilt, group culture of support for affirmative action, and group expectation of restorative justice — were examined. Group-level means on these constructs contributed through cross-level effects in building individual justice expectations. A variety of individual-level antecedents of justice expectations were also examined, with social influence regarding unfairness and African centrality being shown to have important roles here. Outcomes of interest included intention to withdraw from the selection process, emigration intention. entrepreneurial intention, and education intention, all of which are of interest in a country seeking to build a strong economy and a stable, working population. Group differences in three of these four outcomes were observed. Copyright by LAUREN JILL RAMSAY 2009 I»: ACKNOWLEDGMENTS I am grateful to Neal Schmitt for his steady support: over the many years it has taken to complete this project. His counsel, and comments provided by Ann Marie Ryan. Dan Ilgen, and Steve Kozlowski have been invaluable. My thanks go also to Don Conlon for serving as a Dean’s Representative for my final defense. This project would not have been completed had it not been for the ongoing support provided by my husband, Brad Zigler, along with companionship from Braxy and Panzer, and the emotional and logistical support provided by numerous family members. colleagues, and friends on several continents. Michigan State University research assistant Ryan Starkweather assisted in the early stages of this project. Deon Meiring. now at the University of Stellenbosch, helped arrange early stage interviews. William Jiang, Asbjorn Osland, Nancy da Silva, and Meg Ryan at San José State University worked to make my final data collection trip possible. Tom and Ginny Zigler moved in to take care of the pets during my trip. Desi Halse, Dawn and Allan Chaplin, Roni Ramsay. Shayne Ramsay, and Grant and Aileen Ramsay provided logistical support. Research assistants Holly Pittam and Kate Coleby at the University of KwaZulu—Natal and Dominique Brand and Michelle Roberts at the University of Cape Town assisted with participant recruiting and administration. Also, Philip and Rosemary Ramsay, Jaclyn Jensen, Isabel Botero, Hannah-Hanh Nguyen, Chi Dang, and Jailza Pauly provided ongoing US-based cheerleading. Their support has been invaluable. TABLE OF CONTENTS ’ LIST OF TABLES ......................................................................................................... ix LIST OF FIGURES ....................................................................................................... xii INTRODUCTION ........................................................................................................... 1 Paper Overview ................................................................................................... 9 Socio-Political Context ................................................................................ 9 Model ....................................................................................................... l 1 Method ...................................................................................................... l I Results and Conclusions ............................................................................ 12 Socio-Political Backdrop ................................................................................... l3 Inequalities Between Races ........................................... l5 Inequalities Between Genders ................................................................... 20 Labor History 22 Outcomes .................................................................................................. 27 THEORETICAL FRAMEWORK ................................................................................. 34 Expectations and Justice Theory ................................................................ 35‘ Expectations ............................................................................... 35 Justice Theory ............................................................................. 37 Distributive Justice ..................................................................... 39 Procedural Justice ....................................................................... 40 Restorative Justice ...................................................................... 42 Group Characteristics ................................................................................ 54 Group Level Beliefs and Culture ............................................................... 56 Group Belief in White Guilt ....................................................... 56 Culture of Support for Affirmative Action .................................. 59 Group-Level and Cross-Level Relations ...................................... 6] Group Justice Expectations ......................................................... 66 Individual Differences .............................................................................. 70 Race Centrality ........................................................................... 7| Existing Beliefs and Values ....................................................................... 73 Support for Affirmative Action ................................................... 73 Preference for Merit Principle ..................................................... 74 Remote Direct Experience ........................................................................ 76 Educational Integration ............................................................... 77 Indirect Experience .................................................................................. 78 Social Guidance .......................................................................... 79 Proximal Direct Experience ....................................................................... 80 Current Rejection/Acceptance ..................................................... 80 Outcomes ................................................................................................. 81 Withdrawal Intention .................................................................. 82 vi Emigration Intention ................................................................... 87 Entrepreneurial Intention ............................................................ 89 Education Intention ..................................................................... 90 Summary .................................................................................................. 91 METHOD ..................................................................................................................... 92 Sample and Study Design .................................................................................. 92 Procedure ........................................................................................................... 96 Data Cleaning and Handling of Missing Data ............................................ 97 Power ........................................................................................................ 98 Confidentiality ......................................... 100 Ethics ...................................................................................................... 100 Measures ......................................................................................................... 101 Group Beliefs ......................................................................................... .102 Group Culture ......................................................................................... 103 Group Justice Expectations ..................................................................... 104 Individual Differences ............................................................................. 105 Existing Beliefs/Values ........................................................................... 107 Remote Direct Experience ....................................................................... 108 Indirect Experience ................................................................................. 109 Proximal Direct Experience ..................................................................... 1 10 Individual Justice Expectations .............................................................. .1 10 Outcomes ............................................................................................... .1 15 Analyses .......................................................................................................... 1 18 RESULTS ................................................................................................................... 122 Descriptive Statistics,.Reliabilities, and Intercorrelations ................................. 122 Relations Between Individual-Level Justice Expectations ........................ 129 Individual-Level Justice Expectations and Group—Level Antecedents 130 Group-Level Antecedents and Individual-Level Antecedents .................. 133 Relations Between Individual-Level Antecedents .................................... 134 Withdrawal Intentions ............................................................................. 134 Emigration Intentions .............................................................................. 134 Entrepreneurial Intentions. ....................................................................... 135 Education Intentions ................................................................................ 135 Relations Between Intentions .................................................................. 136 Demonstrating Restorative Justice ................................................................... 136 General Restorative Justice ..................................................................... 136 Procedural Restorative and Distributive Restorative Justice .................... .141 Common Method Variance .............................................................................. 144 Scale Invariance across Gender and Race ......................................................... 145 Gender ........................................ .145 Race ....................................................................................................... .148 Multivariate ..................................................................................................... 1 53 Individual Justice Expectation Levels ...................................................... 154 Group Culture ......................................................................................... 163 vii Group-Level Relations ............................................................................ 169 Individual Justice Expectations as the Dependent Variable ...................... 170 Intentions as the Dependent Variable ....................................................... 197 THEORETICAL AND PRACTICAL IMPLICATIONS .............................................. 217 Race and African Centrality ................................................................... .219 Support for Affirmative Action ............................................................... 221 Preference for Merit Principle ................................................................. 222 Educational Integration ........................................................................... 223 Social Guidance ...................................................................................... 224 Race and Gender 225 Group Belief in White Guilt227 Group Culture of Support for Affirmative Action .................................... 229 Group Restorative Justice Expectation .................................................... 229 Withdrawal Intention ............................................................................... 230 Emigration Intention ............................................................................... 231 Entrepreneurial Intention ......................................................................... 233 Educational Intention .............................................................................. 233 Study Limitations and Future Research ............................................................ 233 Alternative Models of Normative Antecedents of Intentions .................... 233 Alternative Models of Antecedents of Emigration Intention .................... 236 Power and Sampling ............................................................................... 237 Scales Used ............................................................................................. 237 Restorative Justice and Legal Claiming ................................................... 241 Causality ................................................................................................. 242 Conclusion ....................................................................................................... 244 APPENDICES ............................................................................................................ 249 Appendix A — Sample Questionnaire ............................................................... 249 Appendix B — Sample Advertisement ............................................................... 266 Appendix C - Informed Consent ...................................................................... 267 Appendix D - Correlation Matrices and Descriptive Statistics for Items Used in Confirmatory Factor Analyses ............................................................. 271 REFERENCES ............................................................................................................ 319 viii LIST OF TABLES Table 1 — Population Distribution .................................................................................. 13 Table 2 —- Households Receiving Services, 2007 ............................................................ 15 Table 3 — Percentage of Employees by Population Group and Occupational Category 18 Table 4 — Self-Declared Emigrants from South Africa ................................................... 28 Table 5 — Race and Gender of Participants ..................................................................... 93 Table 6 — Enrollment of students by Race and Gender ................................................... 94 Table 7 — Home Language ofParticipants ........................................... 95 Table 8 — Major Areas of Study ..................................................................................... 95 Table 9 — Sequence of Measures in Survey Form ......................................................... 1 17 Table 10 — Scale Descriptive Statistics and Reliabilities .............................................. 122 Table 11 — Correlation Matrix of Variables in Model ................................................... 125 Table 12 —— Comparison of A Priori Organizational Justice Factors ............................. 137 Table 13 — Factor Loadings and Uniquenesses for Five-Factor Justice Model .............. 139 Table 14 — Interrelationships Between Factors ............................................................. 141 Table 15 — Comparison of A Priori Organizational Justice Factors including Target Group Referent Questions ........................................................................................... 143 Table 16 — Interrelationships Between Justice Factors including Target Group Referent ....................................................................................................................... 143 Table 17 — Comparison of Models Across Gender ....................................................... 146 Table 18 — Comparison of Models Across Race ........................................................... 149 Table 19 — Multilevel Analysis Results for Each Individual-Level Justice Dimension - Variance in Intercepts .................................................................................................. 158 Table 20 — Justice Dimension Means Across Groups ................................................... 159 ix Table 21 -- Multilevel Analysis Results for Each Individual-Level Justice Dimension — Variance in Intercepts and Slopes ................................................................................ 160 Table 22 -- Effect of General Restorative Justice on Procedural and Distributive Justice ......................................................................................................................... 161 Table 23 — Group Culture Aggregation Indices ............................................................ 164 Table 24 — Multilevel Analysis Results for Procedural Justice — Significant Pooled Slope ......................................... 170 Table 25 — Within-Group Level-1 Coefficients for Procedural Justice ......................... 173 Table 26 — Multilevel Analysis Results for Distributive Justice -— Significant Pooled Slope ............................................................................................................... 174 Table 27 — Within-Group Level-1 Coefficients for Distributive Justice ........................ 175 Table 28 — Multilevel Analysis Results for Procedural Restorative Justice — Women - Pooled Slope ............................................................................................... 180 Table 29 — Multilevel Analysis Results for Distributive Restorative Justice — Women — Pooled Slope ............................................................................................... 183 Table 30 - Within-Group Level-1 Coefficients for Procedural Justice — Women ......... 186 Table 31 — Within-Group Level-1 Coefficients for Distributive Restorative Justice — Women ......................................................................................................... 187 Table 32 — Multilevel Analysis Results for Procedural Restorative Justice — Blacks — Pooled Slope ................................................................................................. 188 Table 33 — Multilevel Analysis Results for Distributive Restorative Justice — Blacks - Pooled Slope ................................................................................................. 191 Table 34 — Within-Group Level-1 Coefficients for Procedural Restorative Justice — Blacks ......................................................................................................................... 193 Table 35 — Within-Group Level-l Coefficients for Distributive Restorative Justice — Blacks ......................................................................................................................... 194 Table 36 - Multilevel Analysis Results for General Restorative Justice — Pooled Slope ............................................................................................................... 196 Table 37 - Multilevel Analysis Results for Withdrawal Intention with Justice Factors — Pooled Slope ............................................................................................................... 198 Table 38 — Multilevel Analysis Results for Withdrawal Intention with Justice Factors — Women — Pooled Slope ................................................................................ 199 Table 39 — Within-Group Level-1 Coefficients for Withdrawal ................................... 200 Table 40 - Multilevel Analysis Results for Withdrawal Intention with Justice Factors — Blacks — Pooled Slope .................................................................................. 201 Table 41 — Multilevel Analysis Results for Emigration Intention — Pooled Slope ......... 202 Table 42 — Within-Group Level-1 Coefficients for Emigration Intention ..................... 204 Table 43 — Multilevel Analysis Results for Emigration Intention with Justice Factors - Women — Pooled Slope ............................................................................................... 205 Table 44 — Multilevel Analysis Results for Emigration Intention with Justice Factors Blacks — Pooled Slope ................................................................................................. 206 Table 45 — Multilevel Analysis Results for Entrepreneurial Intention — Pooled Slope .. 207 Table 46 -— Multilevel Analysis Results for Educational Intention — Pooled Slope ........ 208 Table 47 - Summary of Hypotheses Tested ................................................................. 209 xi LIST OF FIGURES Figure 1 — General Organizing Heuristic with Individual and Group Level Factors ........ 10 Figure 2 — Private Medical Insurance Coverage by Population Group ............................ 17 Figure 3 — Unemployment Rate by Population Group and Gender, 2001 — 200722 Figure 4 — Conceptual Model showing Variables of Interest .......................................... 33 Figure 5 —Within-Group Slopes for General Restorative Justice and Procedural Justice ......................................... 162 Figure 6 — Within-Group Slopes for General Restorative Justice and Distributive Justice ...................................................................................................... 162 Figure 7 — Group Culture of Support for Affirmative Action ....................................... 167 Figure 8 — General Restorative Justice Expectation Culture ......................................... 168 Figure 9 — Group Belief in White Guilt ........................................................................ 169 Figure 10 — Within-Group Slopes for African Centrality and Procedural Justice .......... 171 Figure 11 - Within-Group Slopes for African Centrality and Distributive Justice ........ 176 Figure 12 - Within-Group Slopes for Social Influence and Distributive Justice ........... 177 Figure 13 - Within-Group Slopes for Educational Integration and Distributive Justice ......................................................................................................................... 178 Figure 14 — Within-Group Slopes for African Centrality and Procedural Restorative Justice — Women ......................................................................................................... 181 Figure 15 — Within-Group Slopes for Social Influence and Procedural Restorative Justice - Women ..................................................................................................................... 182 Figure 16 — Within-Group Slopes for African Centrality and Distributive Restorative Justice — Women ......................................................................................................... 184 Figure 17 — Within-Group Slopes for Social Influence and Distributive Restorative Justice — Women ......................................................................................................... 185 Figure 18 - Within—Group Slopes for African Centrality and Procedural Restorative Justice — Blacks ........................................................................................................... 189 xii Figure 19 — Within-Group Slopes for Social Influence and Procedural Restorative Justice - Blacks ...................................................................................................................... 190 Figure 20 — Within-Group Slopes for African Centrality and Procedural Restorative Justice — Blacks ........................................................................................................... 192 Figure 21 — Mean Emigration Intention by Group ........................................................ 203 Figure 22 — Mean Entrepreneurial Intention by Group ......................................... 208 xiii INTRODUCTION This project is an examination of the theoretical dimensions of justice expectations, as well as an exploration of multilevel antecedents of those justice expectations, and related outcomes, among potential job applicants in the context of strong preferential treatment practices. The study of justice in Organizational research has sought to understand the relationships primarily among four different dimensions of justice (procedural, distributive, interpersonal and informational), as well as their antecedents and related outcomes (Bell, Wiechmann, & Ryan, 2006; Colquitt, Conlon, Wesson, Porter, & Ng, 2001; Nowakowski & Conlon, 2005). The selection process has been one context of interest for justice research (Gilliland, 1993) and a substantial body of research on affirmative action in selection exists (see Crosby, 2004; Harrison, Kravitz. Mayer, Leslie, & Lev-Arey, 2006; Kravitz, et a., 1996). Strong preferential selection is an extreme form of affirmative action that attempts to increase diversity and redress historical disadvantages by explicitly weighting demographic status giving preference in hiring to those from previously disadvantaged groups even when the member of the disadvantaged group is somewhat less qualified for the position (Kravitz, 1995). I argue that the existing four-dimension justice theOry as it is applied to affirmative action is too simplistic and does not take into account the principles of duality and simultaneity and the possibilities of compromise (Dawes, 1994) that exist in preferential selection. That is, it lacks the capacity to recognize that when you are selecting a group to be a target group, you simultaneously make the determination of which group will be the non-target group. Also, in deciding which members of a group to select, you also are determining who will not be selected. We need a justice model that allows for the assessment of fairness beliefs regarding groups having concurrent but differing experiences in selection. The four factors validated in current justice research are distributive (the fairness of outcome distributions or allocations), procedural (the fairness of procedures used to determine outcome distributions or allocations), interpersonal (the degree to which people are treated with politeness, dignity, and respect by authorities or third parties involved in executing procedures or determining outcomes). and informational (explanations provided to people that convey information about why procedures were used in a certain fashion or why outcomes were distributed in a certain fashion) (Colquitt, Conlon, Wesson, Porter, & Ng, 2001). The present study makes an important contribution to organizational justice theory by expanding this four-factor frame of organizational justice by proposing and testing a fifth organizational justice factor — restorative justice. That is, the fairness of outcome distributions or allocations to target groups that have suffered inequality in the past. Further, I argue that it is necessary to examine procedural and distributive justice of specific target allocations, rather than just procedural and distributive justice of the selection process. This framing supports a more sophisticated understanding of justice attitudes across and within different race and gender groups and is meaningful in the context of strong preferential selection. (While data on interpersonal and informational components Of justice are captured in this study. along with procedural and distributive dimensions, to demonstrate that restorative justice is distinct, for the sake of limiting the scope of this project I do not consider specific antecedents and outcomes of interpersonal and informational justice in this project. My emphasis is on restorative, procedural and distributive issues.) Further, by recognizing job applicants as part of a hierarchically nested system (Kozlowski & Klein, 2000) where applicants are members of race and gender groups. I examine the interplay of individual-level differences and group-level factors and their contribution to justice expectations in this context of preferential selection. As noted by Harrison, et a1. (2006), there is a dearth of cross-level modeling of relationships in affirmative action research, and cultures or socialization experiences may explain variance in attitudes. The current study furthers our understanding of the effects of the environment and individual factors in justice. Also, following the call for research that identifies the antecedents of individuals’ expectations of justice (Bell, Ryan. & Wiechmann, 2004; Bell, Wiechmann, & Ryan, 2006) this project seeks to improve our understanding of the factors that contribute to the building of justice expectations in selection. Harrison, Kravitz, Mayer, Leslie, and Lev-Arey (2006, p. 1029) state their ambivalence about the continued use of strong preferential treatment in the affirmative action research domain, arguing that such preferences are “almost always illegal" and that research serves to “perpetuate the public belief that affirmative action involves the abandonment of merit as an employment principle.” I believe that research in this area is sorely needed, and present two important reasons for engaging in research in precisely this area. First, strong preferential selection is a workplace reality. While strong preferential selection may be less common in the US (see Jain, Sloan, & Horwitz, 2003; Kravitz. 2008), it is used elsewhere. Strong preferential selection is currently not only legal in South Africa, but it is required there in most cases (Government Gazette, 1998'). Preferential treatment is also in use to address inequality in India and Malaysia (Jain, Sloane, & Horwitz, 2003), and research in applied psychology needs to take an expansive view. To ignore the reality that there are many areas in the world where there remain major differences in opportunity for men and women, and for people of different race. ethnic or social class groups is remiss. I anticipate that preferential treatment will continue to be regarded as an important solution for remedying imbalance, as is the case in South Africa (see “Affirmative action to stay,” 2007). Therefore, we should focus affirmative action research in precisely this area. Second, we need to understand the issues of duality and simultaneity (Dawes, 1994) in preferential selection. The belief that affirmative action means the abandonment of use of the merit principle in selection denies the possibility and importance of the application of the merit principle for within-group selection decisions. In a context where certain groups have been historically disadvantaged in a way that makes members of those groups less well qualified for a position, preferential selection based on demographic characteristics and selection based on merit can co-exist if we focus on within-group merit. We need to understand how people view within-group process. If we operate on the assumption that preferential selection is likely to be a tool in use in circumstances where there have been structural imbalances, as I believe we should. and consider within-group views, we will be better positioned to make relevant recommendations about how, within the limitations of a preferential system, process may be changed to facilitate the most positive outcomes. Therefore, I propose a multi-level model centered around justice expectations on the two established dimensions of most relevance, procedural and distributive justice, along with the proposed dimension, restorative justice. Based in part on the model proposed by Bell, Ryan, and Wiechmann (2004), individual characteristics examined here as antecedents of justice expectations include individual differences, beliefs/values. remote direct experience, and indirect experience. I expand the framework of Bell, et al. to include group-level antecedents of individual-level expectations. Race and gender groups in South Africa have shared socio-political history (Ross, 1999; Terreblanche. 2002; Thompson, 1990), and a current shared status as ostensible beneficiaries/non- beneficiaries under affirmative action (Government Gazette, 1998). At a group level I explore group belief in white guilt and group culture of support for affirmative action. Group belief in white guilt is the race/gender group’s shared belief that whites are guilty of having received benefits due to their race and that they should feel guilty about such benefits. Group culture of support for affirmative action is the race/gender group’s shared attitude of support for affirmative action. I also consider the effect of group membership itself. Responding to Harrison, et al.’s (2006) call for multi—level research, this study improves our understanding of how group membership, group culture, and individual variables independently and in concert shape justice expectations in the context of strong preferential selection. As noted by Shapiro and Kirkman (2001), the potentially negative effects of injustice are not only precipitated by unfair events —— negative attitudes and behaviors are also likely as a result of anticipating injustice. If justice expectations are malleable. organizations may be able to enhance the fairness expectations of prospective employees. and generate more positive outcomes. Key outcomes such as application intentions. attendance/withdrawal, recommendation intentions, and self-handicapping may be S“! a,“ affected by justice perceptions (see Bell, Ryan, & Wiechmann, 2004). Furthermore, exposure to the selection process may have negative psychological effects such as a loss of self-esteem, lowering of core self-evaluation, and a decrease in psychological wellbeing (see Anderson, 2004). Justice expectations have been tied directly to outcomes (Bell, Wiechmann, & Ryan, 2006; Shapiro & Kirkman, 1999), and I attempt to explicate the links between justice expectations and relevant applicant outcomes here by examining intention to withdraw from the selection process, emigration intention, entrepreneurial intention, and intention to study further beyond the present degree. These are important outcomes in any country seeking to build a strong economy and a stable, working population. Affirmative action policies are controversial (Crosby, Iyer & Sincharoen, 2006; Kravitz, 2008), and strong preferential selection generates the most negative attitudinal reactions (Harrison, et al., 2006). Nevertheless, such polices are implemented where a societal redistribution is regarded as necessary (Jain, Sloan, & Horwitz, 2003), such as in post-apartheid South Africa, where the 1998 Employment Equity Act (Government Gazette, 1998) requires that organizations move to hire applicants from previously disadvantaged groups. South Africa today provides a superb research environment within which to test theories of justice: there are multiple clearly defined cultural groups (Christopher, 2002) and a pre-existing policy that mandates preferential selection (Government Gazette, 1998). There is ongoing debate regarding the appropriateness of this workplace affirmative action (for example, see “De Klerk on South Africa.” 2005; “Fears over South Africa’s exodus”, 2006) and the practice of strong preferential selection has precipitated a variety of expectations and reactions among job applicants (Herholdt & Marx, 1999; Ramsay, 2005). This allows for a study that provides a complement to related Western literature which has been built largely on laboratory studies (e.g., Evans, 2003; Heilman & Alcott, 2001; Heilman, Battle. Keller, & Lee, 1998; Kravitz, 1995) and expands our understanding of justice and preferential selection beyond the relations shown by Harrison, Kravitz, Mayer, Leslie, and Lev-Arey (2006). As noted by Truxillo, Steiner and Gilliland (2004), in extant research on applicant reactions, few applicants indicate that they have been treated unfairly, and this may be a function of the lack of extreme unfairness in field sample research published to date. The samples may be biased in that perhaps only organizations With fair procedures examine these issues publicly, or the litigious nature of American society, where much of the research takes place, may lower the base rate of blatantly unfair practices. As there is variability in views on strong preferential selection in South Africa (Herholdt & Marx. 1999), this context provides an excellent opportunity to examine potential applicant reactions where there may be a high enough base rate on feelings of injustice that the relations between justice expectations and outcomes will be made clearer (see discussion in Truxillo, et al., 2004). Social changes do affect support for policies that address racial inequality. For example, Kaiser, Drury, Spalding, Cheryan, and O’Brien (2009) showed a reduction in support for policies that address racial inequality after the election of President Obama. Conducting a study while past socio-political inequity in South Africa (see Terreblanche, 2002) remains salient is crucial to capture the psychological processes of interest. While an improved understanding of potential applicant perspectives would be useful in informing theory regarding justice dimensions, their multi-Ievel antecedents. and applicant reactions, there is also practical value. Organizations seeking to attract and maintain an ethnically diverse, healthy and productive workforce would benefit from knowledge about how justice expectations are built so that they may plan procedures to facilitate positive expectations in the selection process and positive outcomes. Further. while South Africa provides the backdrop for the proposed study, many countries are taking steps to increase the diversity of their job incumbent populations where there has been previous inequality (Jain, Sloane, & Horwitz, 2003). By capturingjustice expectations and their links to important outcomes in the context of strong preferential selection, this study may serve to inform future policy both in South Africa and in other countries seeking to implement compensatory selection systems. As noted by Chan and Schmitt (2004), the research agenda in the arena of applicant reactions should address (among other issues) determinants of applicant reactions, dimensions of applicant reactions, criterion outcomes of applicant reactions, and changes in applicant reactions over time. This project, in concert with a longitudinal follow-up study, follows this agenda. As noted by Gilliland (1993), we need research that links applicant fairness perceptions to preemployment outcomes to help establish the importance of applicant perceptions on outcomes. My model links individual-level antecedents, group-level antecedents, individual justice expectations, and relevant outcomes, to build a parsimonious framework for justice expectations that remains meaningful in the context of strong preferential selection. 1 test specified hypotheses using data collected in South Africa from a sample of potential job applicants —- people who were in their final year of an undergraduate degree prOgram, who have some interest in a job, and who have a reasonable possibility of applying (see Ryan. Horvath & Kri ska. 2005). My research agenda is to answer the following major questions: Is an additional. distinct justice factor — restorative justice — supported? What are the relevant individual- level and group-level antecedents of justice expectations within the context of strong preferential selection, and how do they interact in influencing individual-level outcomes? What are the effects of justice expectations on positive and negative outcomes in this context? Paper Overview To provide the reader with an idea of the flow of this paper, I provide a pre’cis of what is to come. Soda-Political Context In 1994 South Africa held its first non-racial electiOn where South Africans ofall races could participate in the formal political process. This was the first of many steps to move away from the former Nationalist Party policies of separate development for different race groups (apartheid) (Ross, 1999). Another step to help redress racial imbalances was the implementation of the Employment Equity Act (Govermnent Gazette, 1998). This Act pushes employers to use strong preferential selection in hiring so that those from previously disadvantaged groups will be moved into positions that were previously primarily the purview of whites, and white men in particular (Ross. 1999; Terreblanche, 2002). This process has resulted in strOng reactions among job applicants of all races, and reactions have been linked to problems such as continued emigration, but also to positive outcomes such as increased entrepreneurial intentions and plans to pursue further education (Crush, Peberdy, & Williams, 2006; Herholdt & Marx. 1999; Ramsay, 2005). .32 3:223 {Suck ESQ 38.5 35» 333.355. 23.: 82.23.»: MEnEUMKQ chmzmb 3:222: 732 96.5 coucogoomm mcocfiuoaxo 3038898 985 _ 2%; moucotoaxo «09:6:— 828:0ng 82% 08on mos—$3023 wEExm moo=0co§c asses toga. mo $2.2. BEE—o 9.20 9.20 62.0383— 10 Model A general organizing heuristic is provided in Figure 1. In brief, 1 present a multilevel model where I link individual-level antecedents of justice, the establishment of group- level beliefs and restorative justice expectations, and individual-level justice expectations on three dimensions in explaining important individual-level outcomes. The organizing frame is based in part on the model of justice expectations proposed by Bell, Ryan. and Wiechmann (2004), but goes beyond their structure in that I recognize potential job applicants as part of a hierarchical multilevel structure, and therefore integrate individual- level and group-level constructs. At an individual level I focus on four groupings of direct antecedents of justice expectations. These are individual differences, existing beliefs and values, remote direct experiences, and indirect experiences. At a group level I examine group belief in white guilt as a contributor to group culture of support for affirmative action. Group support for affirmative action and group expectations regarding restorative justice are presented as having a cross-level effect on individual expectations. The group-level and individual-level predictors contribute to variance in justice expectations on three justice dimensions (procedural, distributive, and restorative). The expectations are linked to important outcomes (withdrawal intention, emigration intention, entrepreneurial intention, and education intention). Existing theory and research are reviewed and are used to build and support the model described here, and to frame specific hypotheses. ' Method I conducted a paper-based survey of the attitudes of potential job applicants for jobs for new graduates in South Africa. Surveys were administered on two campuses in a 11 classroom setting. Participants were final year students recruited via flyers and class announcements. They received a R100 voucher (about $12) in exchange for their participation. Data from each campus were merged into one dataset of N = 327. The final total stratified sample was comprised of men and women from each of four population groups; black African, coloured, Indian, and white, as well as a small number of people of other race groups. This sampling plan allowed sufficient power to confirm the factor structure of justice expectations and to test single-level and multilevel hypotheses. Results and Conclusions Results of this study indicate that restorative justice as a distinct justice dimension is supported. Further, three group-level constructs — group belief in white guilt, group culture of support for affirmative action, and group expectation of restorative justice — were demonstrated. These group-level cultures contributed through cross-level effects in building individual justice expectations. A variety of individual-level antecedents of justice expectations were also examined, with social influence regarding unfairness and African centrality being shown to have important roles here. Outcomes of interest included intention to withdraw from the selection process, emigration intention, entrepreneurial intention, and education intention, all of which are of interest in a country seeking to build a strong economy and a stable, working population. Group differences in three of these four outcomes were observed. Having offered a brief overview of my project, I now provide the substantive content that provides the full framework, analyses, results and conclusions. First, to provide the reader with sufficient background regarding South Africa to follow the model in context, I offer a short socio-political review. 12 In this paper I will use the general term black to refer to black Africans. coloured people, and Indians, consistent with South Africa’s Employment Equity law (Government Gazette, 1998). Although the census classification for Bantu people was “Black” for censuses of 1980, 1985 and 1996, and “African/Black” for 1996 and 2001 (Christopher, 2002), and “African” in the Employment Equity law, for clarity I use the term, black African, as South Africans of all races are being referred to as African (e.g.. see Vestergaard, 2001). To be consistent with cited laws and publications in South Africa I use South African/British English spelling rather than that of American English (e. g., labour versus labor) in this paper and survey form, where appropriate. Socio-Political Backdrop South Africa has a population of roughly 49.3 million and the racial breakdown is provided in Table 1 (mid-2000 estimate, Statistics South Africa, 2000; mid-2009 estimate, Statistics South Africa, 2009). Table 1 Population Distribution Population Millions Millions Percentage of Group (2000) (2009) Population (2009) Black African 33.88 39.14 79.3 White 4.52 4.47 9.1 Coloured 3.80 4.43 9.0 Indian/Asian 1.09 1.27 2.6 Between 1948 and 1994 South Africa was run unde'r an oppressive political system of separate development for different racial groups (apartheid) where the minority 13 white community dominated the country’s access to resources (Ross, 1999; Terreblanche. 2002). Under the Nationalist government, maintaining the purity of ethnic groups was regarded of critical ideological importance. Thus, through the passing and implementation of policies such as the Population Registration Act of 1950, a Prohibition of Mixed Marriages Act, and others, heterosexual intercourse across the color line was limited. One’s racial classification defined one’s access to residential areas (where black Africans were expected to reside in primarily rural reserves or homelands), education. and job opportunities, among other resources (Ross, 1999; Terreblanche, 2002; Thompson, 1990). Over the years it became increasingly clear that this separate development was highly unequal development, and the dramatic gap between whites and non-whites was evident in wealth, education, work opportunity, participative government and access to resources. After many years of political tension and violence, the country held its first non-racial democratic elections in 1994 (Ross, 1999; Terreblanche, 2002). Apartheid left behind entrenched social problems such as racism, poverty, disease and unemployment (Ross, 1999; Thompson, 1990). The country is battling violent crime and has limited resources to raise the living standards for its population (South African Police Services. 2007; Statistics South Africa, 2006a; Statistics South Africa, 2008a). Table 2 shows the proportion of households receiving basic services (Statistics South Africa, 2008a). 14 Table 2 Households Receiving Services, 2007 Percentage of Service Population Access to piped water at dwelling or site 71.3 Municipality services for refuse removal 61.0 Connected to mains electricity supply 81.5 Bucket toilet or no sanitation 8.3 The unemployment rate is estimated at 23.5% (Statistics South Africa, 2009), and this number would exceed 30% if “discouraged workseekers” who have given up looking for jobs were also included in the calculation of the unemployment rate. With a high unemployment rate it is not surprising that there is a high risk of experiencing incidents of crime (see Statistics South Africa, 1998 and 2006a). For example, the number of murders in South Africa over a twelve-month period, 2007/2008, was 18,487 (South African Police Services, 2008). By comparison, the US Department of Justice reports 16,929 cases of murder or nonnegligent manslaughter for 2007 (US. Department of Justice - Federal Bureau of Investigation, 2008), where the US population is more than six times larger than that of South Africa. Inequalities Between Races While a political democracy has now been established, there remain huge inequalities between race groups. As noted by the General Secretary of the Congress of 15 South African Trade unions (COSATU), Zwelinzima Vavi, in a recent speech (Vavi, 2007), “We have gained political freedom yet economic power is firmly in the hands ofa white minority, while poverty, unemployment and inequality are still rife among the black majority.” During the apartheid era a hierarchy of advantage developed among the different race groups, and that hierarchy remains, with whites being the most advantaged, and black Africans the least advantaged; the apartheid government favored coloureds and Indians over black Africans (Thompson, 1990). This legacy hierarchy can still be seen in access to a variety of resources. For example, Figure 2 shows the differences in private medical insurance coverage by population group for the years 2002 through 2007 (Statistics South Africa, 2008a). Recognizing the relative advantage over black Africans experienced by coloured and Indian people, Ware (2002, p. 135-136) notes that the “project of racial classification may have had the desired effect of holding those defined as white at the opposite end of the spectrum from those defined as black, but this did not mean that the other racially defined groups did not serve equally important functions in holding that system in place." 16 .8 8 3 8 .8 3 s .8 8 8 Percentage of Group wlth Coverage 8 8 Figure 2 Private Medical Insurance Coverage by Population Group Schooling was strictly segregated under apartheid government, and the legacy unequal education systems for different race groups remain (Asmal & James, 2001). The access of black Africans to effective education suffered most of all races because of very limited resources being assigned to that function. In the mid-1970s the government provided to black Africans only 6.5 percent of the amount granted to each white student, with this being raised to just under twenty percent over the course of the next decade (Ross, 1999). As noted by Ross, by the 1990s roughly 200 black Africans per year passed mathematics at the matriculation level — the final year of high school — hardly sufficient to produce a dynamic black African labor force. Furthermore, many blacks chose to eschew education in favor of fighting for liberation (Ross, 1999). Schooling was compulsory for whites, but not for non-whites, and white schools were superior to those of non-whites. The quality gap was greatest for black African schools and less so for the coloured and Indian schools (Thompson, 1990). Access to jobs was controlled according to race, with certain operations being reserved for whites (Nattrass & Seekings, 2001; Ross, 1999; Terreblanche, 2002). Also. the Government declared that coloureds should be given preference over black Africans, but not over whites (Ross, 1999). As noted by Ross regarding apartheid society (p. 200- 201), “Wherever White encountered Black, White was boss and Black was servant.” While the government had stopped reserving certain job categories for whites by 1986 (Thompson, 1990), this did not immediately remedy the imbalance. Table 3 shows the 1994 distribution of employees by race and occupational category according to the Labor Force Survey of 2004 (see Department of Labour, n.d). The relatively high proportion of Indian and white employees in the higher-skilled occupations compared to black African and coloured employees is apparent. (As a reference point, recall that black Africans comprise 79% of the population, coloured 9%, Indian 2.6%, and white 9.1% (mid-2008 estimate, Statistics South Africa, 2008a). Table 3 Percentage of Employees by Population Group and Occupational Category Black African Coloured Indian White Legislators 24.4 8.8 9.7 5 7.1 Professionals 42.2 6.2 7.7 43.9 Technicians 53.0 11.2 5.2 30.6 Clerks 43.6 17.7 8.1 30.6 18 Table 3 continued. Service 69.6 11.0 3.9 15.5 Skilled agricultural 51.2 1 1.4 0.2 3 7.1 Craft 66.2 15.2 2.2 16.4 Plant operators 77.8 13.9 , 4.1 4.2 Elementary occupations 76.1 20.3 0.8 2.8 Furthermore, the economic system gave whites a monopoly on skilled jobs. and there were dramatic inequities between white and black wage rates. For example. in 1982, white average annual wages in private manufacturing and construction were 4.4 times higher than were wages for black Africans (Thompson, 1990). Inequality across race groups was exacerbated by residential separation. As population density increased in the reserves, agricultural economies there began collapsing under the strain. This reduced quality of life for those living in the rural reserves and increased the gap in wealth between those prospering in the labor market and those relegated to the rural areas, accentuating the divisions created by apartheid (Nattrass & Seekings, 2001; Ross, 1999). While some black professionals, businesspeople and the political elite now earn enough to move to formerly white suburbs, most black South Africans live in settlement patterns similar to those established under apartheid (Ramphele, 2001). Participation in apartheid-era national government varied by race group. with coloureds and Indians being included in a revised parliamentary system only in 1984 — l9 black Africans remained excluded on the grounds that they had their own governance in the homelands (Thompson, 1990). It was not until 1994 that the country held a democratic election for citizens of all races. Thus, in the apartheid era, one’s life experiences were driven in very large part by one’s race group membership, and although political participation and government has since changed, there remain legacy differences in life experiences between race groups. Inequality Between Genders Not only are there historical and current differences for people of different race groups, there have been and remain many differences in the experiences of men and women. Sexism among both white and black males did not vanish with a change in government (Ross, 1999), and white male dominance of the economy remains (Ramphele, 2001). For example, women’s average wages are lower than are those of men of the same race, for all four race groups (see Nattrass & Seekings, 2001). Littrell and Nkomo (2005, p. 563) provide a description of the experience of women: “This historical racial division was accompanied by patriarchy with women of all races subordinate to males. Women of all races were primarily expected to be homemakers and were legally classified as “minors”. However, when they did work, black, Coloured and Asian women worked primarily in domestic and unskilled factory work while white women were employed in administrative and female occupations (e.g. nurses, teachers)” While women had certain shared experiences in this patriarchal society, some experiences are shared only by those falling within certain race and gender subgroups. For example, the women’s suffrage movement in South Africa reflected the drive of one specific race and gender subgroup; that of white women (Walker, 1979). As noted by Simons, and acknowledged by Walker, black women were busy with other priorities — fighting alongside their men for full civic rights. The success of the movement resulted in 20 the vote for white women in South Africa in 1930 (Walker, 1979). Further, the extension of pass laws to black African women in 1961 (Thompson, 1990) is one example of policy targeting one specific racial and gender subgroup at a different point in time, making the experience of that subgroup distinct. Also, in response to black male migrant labor. many black African women were heads of households (Thompson, 1990). Black African and coloured women have historically worked as domestic help in white households. and face the stereotypes associated with this history (Littrell & Nkomo, 2005). Department of Labor data reflecting numbers reported by large employers for employment equity reporting for 2004 (Department of Labour, n.d.) show that the distribution of employees by race, gender and occupational category continues to show inequality among race and gender groups. For example, elementary occupations were dominated by black African men and women, who are followed in frequency by coloured men and women. Plant operator positions were dominated by black African and coloured men. Compared to men, clerical positions were dominated by women of the same race. but the race hierarchy of advantage was reflected by white women dominating over Indian women, who in turn dominated over coloured women, who dominated over black African women. Indian men were the most likely group of men to be engaged in clerical positions. Professional positions were held most ofien by white men and white women while legislator positions were dominated by white men. These data demonstrate ongoing differences in dominance of certain occupational categories by people of different race and gender subgroups. Unemployment rates also vary by race and gender (Statistics South Africa. 2008) with race differences in unemployment demonstrating the hierarchy of race advantage 21 noted earlier. Women are more likely to be unemployed than are men of the same race. Unemployment rates are shown in Figure 3. 45 ~ ~ 40 ..._,___...__....__ .__.._,,,_ ,_ WWW.“ .. ._ .. ,.... . 2.. 35 - E 30 - 2001 2 2002 E 25 2003 g 2004 3 20 , .2005 E II: 2006 g 15 ‘ .2007 a" 10 - 5 0 . ,3“ O ‘30 Women Figure 3 Unemployment Rate by Population Group and Gender, 2001 - 2007 Thus, race and gender subgroups have unique socio-political histories. Labor History The history of the labor movement in South Africa contributes to our understanding of today’s workplace issues. I begin at the Industrial Conciliation Act of 1924, which elevated the status of white workers while black workers had few protections. This act was designed to bring labor into the system of state government. and promote collective bargaining, but was relevant only to whites as blacks were not recognized as employees. This served to provide further voice for whites. but none for 22 blacks (Finnemore, 1999). This made more concrete the trend of wealth being focused in the hands of the whites. As industry developed and more blacks became part of the workforce, black industrial unions began to appear, and by the late 19203 there were a number of non- racial unions made up of workers of all races fiom the factory floor, growing to 1 19 non- European unions by 1945 (Finnemore, 1999). The Nationalist Party came into power in South Africa in 1948, and instituted the policy of separate development known as apartheid. Their Suppression of Communism Act of 1950 resulted in banning and arrests of many trade union leaders, and led to the continued suppression of black workers (Finnemore, 1999). However, the Black workers could not be held at bay forever, and in 1973 the country saw widespread Black labour strikes over wages. That the strikes were coordinated without formal unions made management negotiation with the workers difficult. The government reacted with the Bantu Labour Relations Regulation Act which provided a dispute resolution avenue within the employing organization, but these were regarded by workers as “toy telephones” (Finnemore, 1999). The informal organization of Black workers grew outside the system, and the government was pressured to respond. The Wiehahn Commission reported in 1979 that reforms were necessary, and made recommendations that included freedom of association for workers of all races and migratory status. The government was very cautious in its actions relating to the commission’s report, and made amendments to the Industrial Conciliation Act which legitimized some black unionization. Subsequently, Black unions flourished. and the Labour Relations Act of 1981 provided blacks full union rights but still no political 23 power. The Congress of South African Trade Unions (COSATU) formed in 1985. creating an amalgam of unions, and the trade union movement continued to grow as blacks used this as an avenue for political voice, restricted from other avenues of participation. The increasing politicization of unions led to government efforts at control that included leaders of unions being subjected to various police actions in the apartheid era. However, moving to create change post-apartheid, the new multi-racial South African government developed a Reconstruction and Development Programme (RDP). and established the National Economic and Development and Labour Council (NEDLAC). Their goals were to build a productive and unified country. Another means of supporting the achievement of these goals was the establishment of the Council for Conciliation Mediation and Arbitration (CCMA), and the subsequent passing of the new Labour Relations Act (1996), which resulted in better standards to support consistency. equity and social justice in employment. There were two other interesting developments. First, between 1994 and 1998. attendance at union meetings dropped, as reported in a survey of members of COSATU (Wood, 2001). In 1994, 60% of study respondents reported that they had contributed in the election of shop stewards, dropping to 48% in 1998. While these participation numbers are high compared with the rest of the western world, they are a significant drop for South Africa. It may be the case that as other avenues of voice are created, the need to participate in union organization is minimized. Second, the Employment Equity Act (Government Gazette, 1998), which identifies at least 20 forms of discrimination, was applied, further protecting workers and creating better guidelines for workplace 24 consistency. This new culture of a political democracy embraced a more consensus- based approach to labor relations (Finnemore, 1999). The Employment Equity Act The Employment Equity Act (Government Gazette, 1998), which specified affirmative action for blacks, was designed to help remedy the imbalance in employment where most skilled and management jobs were held by whites and most unskilled jobs were held by non-whites. The purpose of the act is: (a) promoting equal opportunity and fair treatment in employment through the elimination of unfair discrimination; and (b) implementing affirmative action measures to redress the disadvantages in employment experienced by designated groups, in order to ensure their equitable representation in all occupational categories and levels in the workforce. The designated ostensible beneficiary groups are: blacks (including black Africans, coloured people and Indians). women of all races, and the disabled. I refer to these groups as target groups. Designated employers are required to comply with the Act, and those who contravene the Act are faced with fines (Government Gazette, 1998). Under this policy, a person may be regarded as suitably qualified for a job as a result of any one of, or any combination of: formal qualifications; prior learning; relevant experience; the capacity to acquire, within a reasonable time, the ability to do the job (Government Gazette, 1998). The last in this list of criteria is of critical importance as companies are now pressed to accept candidates based on race, for example, where a non- white candidate who has potential for development may be preferred over a white candidate who may already have demonstrated experience. While the need to create a system that facilitates the employment of those from previously disadvantaged groups 25 was essential, the new system does push organizations to discount traditional qualifications, as by sticking to traditional qualifications as the primary determinant across applicants, adverse impact for previously disadvantaged groups would be likely to continue. Degrees of Blackness The careful racial categorization into four major population groups is different from the white/non-white dichotomy emphasized in many other countries, and has led to specific post-apartheid problems. For example, although the new government is anxious to move away from the apartheid-era racial categorizations. (see Christopher, 2002). unless these categorizations are kept salient, it is not possible to provide compensation to those who suffered under apartheid-era policies. Clarity with regard to racial categorization is essential in this context. For example, the discussion of affirmative action by Herbert (1994, p. vii) begins with a statement, “Whenever the word ‘black’ is used in this book with regard to persons, unless otherwise stated, it refers to those individuals who are not white.” Such a statement may appear superfluous to those who are not aware of the historical context of specific racial categorization in South Africa. Current affirmative action law in South Africa uses the dichotomous black/white categorization as used by Herbert. The black category includes black Africans, coloured people and Indians. However, such a crude dichotomous black/white distinction causes problems when there had historically been a more specific categorization. For example. Indians, who held a relatively advantaged position compared to other non-white groups in the apartheid era (Thompson, 1990; Ware, 2002), are now able to benefit from affirmative action laws for blacks. Some Africans argue that Indians are ‘not black 26 enough’ as they were historically relatively advantaged, and should not be able to benefit from these new laws (Ramsay, 2005). Some coloured people believe that they are not “black” enough, and therefore anticipate barriers to career entry (Stead, Els, & Fouad. 2004). Discontent Although the African National Congress (ANC) was the historical black liberation party. as the current ruling party they face tremendous challenges in satisfying a population still desperately lacking in many basic services. In some areas this discontent has led to violence including rioting over living conditions (see “South Africa discontent spreads". 2009). The global economic downturn has affected South Africa and the country is facing a recession that will make it particularly difficult for populist president Jacob Zuma to meet election promises. Outcomes South African today is seeing several patterns of behavior that may be related to the current system of affirmative action. A series of interviews held with employment equity experts in South Africa (Ramsay, 2005) and other literature provide information about possible outcomes linked to the current selection policies. First, some job applicants appear to be withdrawing altogether. Interviewees reported that some organizations are seeing far fewer whites applying for positions as they expect not to be considered due to race, and that white males are feeling alienated (Ramsay, 2005). This apparent sense of resignation is also reflected in the establishment of the Come Home Campaign, a joint project of the trade union, Solidarity. and the Company for Immigration. The Come Home Campaign’s aims are: 27 “To assist as many skilled expatriates to return to South Africa as possible; To retain and locally involve as many would-be skilled emigrants as possible; To engage skilled people still resident in South Africa who have withdrawn mentally and “emigrated” inwardly in the public debate and life once more.” (Come Home Campaign, n.d.) The Come Home Campaign has been acknowledged by the South African Government (Fraser-Moleketi, 2007, p. 21) as an important initiative: “The return, retention and involvement of skilled South African citizens are of cardinal importance to the South African economy. The launch of the Come Home Campaign was prompted by this necessity.” F raser-Moleketi’s response reflects the concern that there are South Africans who are withdrawing without actually leaving the country. Second, emigration is continuing (Crush, Peberdy, & Williams, 2006; Statistics South Africa, 2005), and has risen post-apartheid (see Bailey, 2003; “Fears over South Africa’s exodus”, 2006; Statistics South Africa, 2005). Table 4 shows the number of self- declared emigrants from South Africa between 1990 and 2003 (Statistics South Africa. 2005). Table 4 Self-Declared Emigrants from South A flica Year Count 1990 4,722 1991 4,256 1992 4,289 1993 8,078 1994 10,235 1995 8,725 28 Table 4 continued. 1996 9,708 1997 8,725 1998 9,031 1999 8,478 2000 10,262 2001 12,260 2002 10,890 2003 16,165 While emigration was high in 1994, the year of South Africa’s first non-racial election, emigration rates for 2000, 2001, 2002, and 2003 all surpassed what was recorded for 1994. It should be noted that these numbers appear to be a dramatic underestimate based on comparisons with data provided by countries that are receiving former South Africans as permanent residents or citizens (see Statistics South Africa. 2005). The outflow of academic staff from South African universities is concerning, with those at the peak of their academic careers leaving for other countries (see Waddington. 1999). However, it is not only academia that is noticing this trend. Around one third of skilled/professional emigrants are from education or the humanities, while legislative. executive and managerial people account for 29 percent of the skilled emigrants (Bailey, 2003). According to Statistics South Africa (2003), about half of those who declare 29 themselves emigrants when leaving South Africa are professional or managerial people. Intention to emigrate may be more likely among those who have a university qualification and/or social contacts abroad (Miller, Haskell, & Thatcher, 2002). A study of the views of skilled South Africans conducted by the Africa Institute of South Africa and the Southern African Migration Project in 2002 showed that both blacks and whites had given emigration some thought, and that twenty percent of their respondents reported that it was highly likely that they would emigrate (Waller, 2006). Opposition to affirmative action was overwhelmingly the case for whites in the Waller study; opposition to the policy was true for 20% of the blacks (Waller, 2006). According to Crush, et al., (2006), local economic and political circumstances have contributed to emigration intentions, while new job opportunities abroad have also played a part in some sectors. Furthermore, affirmative action has been cited as a reason why expatriate South Africans are unwilling to return (see Peta, 2005, and Waller, 2006). As recognized in The Economist (“Between staying and going”, 2008) there are both push and pull factors that appear to be contributing to emigration, and the brain drain is damaging to South Africa -— the country can ill afford the loss of its best-trained people. Third, entrepreneurship may be viewed as a feasible alternative to employment with a company or other organization. Interviewees reported that some job applicants. and white males in particular, might feel pressed to start their own business where they believe that they will not be considered for employment due to race and gender (Ramsay. 2005). Although South Africa lacks entrepreneurially-minded people relative to other countries the explanation offered for this gap has been a lack of education (South African Institute for Entrepreneurship, 2006). However, those with education. especially those 30 who feel that they are unlikely to be employed under current affirmative action policy. may be highly likely to try their hand at starting a business. Fourth, people are choosing to pursue further education, with enrollment in tertiary education rising from around 863,000 students in 2002, to around 960,000 in 2005 (Statistics South Africa, 2006a). Under the Nationalist government, government positions were largely reserved for Afrikaners, many of whom with limited education (Innes, 1993). Under the new dispensation individuals in this group no longer have this “safe” employment option. According to interviewees, this has pushed young whites. especially white males, to pursue further education or to start their own businesses (Ramsay, 2005). Fifth, exposure to the selection process may affect job applicant wellbeing, and individuals may be experiencing lowered core self-evaluation (see Anderson, 2004). There exists the perception that affirmative action is tokenism and that unqualified people are put into senior positions and will fail (Herholdt & Marx, 1999; Ramsay. 2005). This pressure for those from previously disadvantaged groups to succeed creates a high-stress environment for target group members. Steele (1991) has argued that affirmative action for blacks generates a kind of demoralization that is an internalization of the doubts of others about their ability. In a laboratory study, Heilman, Simon, and Repper (1987) showed a negative relationship between preferential selection based on gender and women leaders' self-perceptions and self-evaluations. In another laboratory study Heilman and Alcott (2001) showed that being aware that others viewed them as being selected based on sex led to participants’ negative self-regard when uncertain about their ability regarding a task. and produced negative affect. Further, non-target group members may feel devalued. This 31 new world order for whites may be perceived as white male oppression. and according to Allport (1954), group oppression may lower self-image. This context provides the backdrop for the theoretical framework and hypotheses discussed below. As expectations and justice are central to my model, 1 will now review related research and theory and follow that with a discussion of the group-level and individual-level factors that act as antecedents in the model. I will discuss justice factors and how they are distinct as a central issue in this project, and then provide discussion regarding outcomes of interest. I offer specific testable hypotheses interspersed throughout the literature review. 32 9.: \ @2893. 38m coast: 3.: a 36:25 xv 38:6:— coswosem mm: A \ E63285 : : 9,3898% .3: Ezzocoaobcm cm: :: 3:35me Al 2.: ll :ocfiwBE 55:25 i. 356805 f Ecocmoscm cotmuwmfim .mGOfiSOOQxO A / mt moocotoaxo 82% 89:3— :OMEBE 833. 3:239: mm: RBECEMB 2039898 .2636:— MCMN 2&0:th 868:5 . R N: E: :86 How. 82886qu . 2t 3: . . 3: cases 63:25.9? 2.: A E: 2: Sm tongsm mm: : E 2: / E: a: E: seamen? : m x 3: m: 2: a: bzabeoo e _ m R: 4: Q: 4: 38m _o>o_ 563%:— wm: mt mm: NI / menaceobG 3:232: 3053035 R: ER: @380 5260331 oomwmsm comuom o>fim§$mw 6.586me 8m Comma. Al 53m 323. 9.86 we 638:6 380 E «oxen 9.80 695383: $26383. 22.8830 380 \1 toga—3 mo 89:20 9.80 mtg—on 95.0 3.325% $33.8; MEESQ None: 333.0269 V ouzwwnw 33 THEORETICAL FRAMEWORK There are several factors that contribute to my theoretical model and these are reflected in the general organizing heuristic shown in Figure 1 provided earlier. Specific variables of the model are laid out in Figure 4, and specific hypotheses are identified later in the paper. This general heuristic is based in part on the model of Bell, Ryan, & Wiechman (2004). I address four groupings of antecedents of individual-level justice expectations. These include individual differences, existing beliefs and values, remote direct experiences, and indirect experiences. At the race/gender group level I examine group beliefs as an antecedent of group culture, which in turn has a cross-level effect on individual expectations, as do group expectations. At the individual level I explore the relations between justice expectations along different justice dimensions, and as expectations of justice may affect how people react to organizational events (see Bell. Wiechmann, & Ryan, 2006), expectations are viewed as directly linked to the outcomes of interest in this context. As discussed earlier, affirmative action target group membership status plays a hugely important role in selection decisions in South Africa. Further, race and gender group members have a shared socio-political history. Therefore, I expect that there may be distinct variation in expectations and reactions between race and gender groups. I also anticipate that individual-level antecedents may moderate the relationship between group membership and justice expectations. Thus my hypotheses specified below, while capturing relations across pooled target groups, also specifically address mean differences in individual-level constructs between target groups, slope differences in 34 predictor-criterion relationships between target groups, and group-level predictor- criterion relations. I will now address theory and published research findings relevant to supporting this framework for the research questions of interest. Expectations and Justice Theory “Relations among races and between sexes, assertions of individual rights, and demands for equality in distributions of society’s benefits. . .constitute the fundamental social problems of our era,” according to Goldman (1979), p. 3. Goldman could not have put South Africa’s current position more succinctly. The individual’s expectation of justice is an important issue in a climate where selection decisions are being made based on race and gender, in an effort to correct societal imbalances. Expectations I posit that justice expectations, or expectancies, are built regarding the selection process in South Africa, and affirmative action in particular, although these expectations may never have been articulated. Thus, theories of psychological contract (see Anderson & Schalk, 1998; Tekleab & Taylor, 2003) underlie the conceptualization of j ustice expectations addressed here. Psychological contract is a set of expectations that are created in the mind (Levinson, Price, Munden, Mandl, & Solley, 1966). These expectations capture beliefs about the obligations in a relationship between two parties. where each individual comes to the relationship with their own set of expectations, and many of those expectations may have nothing to do with anything that has been indicated or not indicated by the employer. The individual’s expectations may be established before they even apply for a job with the company. 35 As noted by Dabos and Rousseau (2004), expectations are shaped by influences outside the employment relationship of interest. These influences could include values. motives, experiences at work, and broader social norms. This is consistent with the notion that there could be different psychological contracts for different groups (e.g., Guzzo. Noonan, & Elron, 1994; Thomas & Anderson, 1998). Levinson, et al. (1966) discuss normative expectations of how individuals and work groups develop schemas for what they should expect from one another, and they note that these normative expectations vary across groups (e.g., type of industry, type of job, level of responsibility) and also vary according to the personal experience of the individual, as well as across seniority and gender. Thomas, Au, and Ravin (2003) identify a number of propositions that reflect the cultural differences in expectations and there is some evidence that cultural differences do exist with respect to views on commitments, however, we do not have a clear understanding of the development and change of the psychological contract across cultures. Many hiring organizations do not offer explicit statements about what job applicants should expect in the hiring process, yet applicants hold distinct expectancies regarding what will take place in their engagement with an organization. I suspect that in the South African context, the government mandate that organizations must remedy past imbalances (Government Gazette, 1998), and media coverage regarding that policy (e.g.. “Affirmative action to stay”, 2007) create a universal expectation that organizations will be making selection decisions based on target group membership status. Individuals may make judgments about the selection system in South Africa through heuristic processing. where, in a case such as this, they have limited information and the information that is 36 available from the government and the media may prime processing based on categories (see Chaiken, 1987; Kulik & Perry, 1994). Also, they may not be motivated to engage in systematic processing (Chaiken, 1987). Nacoste (1994) argues that the view of affirmative action policy is an affirmative action schema, where the individual may use the “average” process in affirmative action as the benchmark for what is to be expected. Media and government reports in South Africa would facilitate a common schema. Research such as that of Kravitz and Platania (1993) has shown that people may hold common beliefs about affirmative action process. They indicate that respondents. who were not given specific information about affirmative action policy, thought quota hiring likely. Thus, I posit that South Africans applying for jobs may hold a shared conceptualization of what affirmative action means for the selection process, and that in general they should expect quota hiring based on target group membership; however. this view will vary depending on hiring practices that they see around them. As articulated by Bell, Ryan, and Wiechmann (2004), expectations have received little attention in organizational justice research, yet have important implications for understanding applicant perceptions. As shown by Bell, Wiechmann, and Ryan (2006), applicants do hold justice expectations, and those expectations can be directly linked to outcomes such as job acceptance and recommendation intentions. Justice Theory Making a contribution to justice theory is an important goal of this research project, and justice is a central component of my theoretical model. I now provide a brief review of relevant literature on justice. However, justice-related studies are also 37 addressed later in this paper, along with the discussion of the specific hypotheses regarding the various antecedents of justice expectations and outcomes related to justice. As noted by Ambrose and Arnaud (2005), justice had historically been broken down into two dimensions, then three (distributive justice, procedural, and interpersonal justice), however, more recently Colquitt (2001) validated the four-factor justice model that includes informational justice. Colquitt, Conlon, Wesson, Porter, and Ng (2001) provide the following definitions of the four factors: Distributive Justice: “the fairness of outcome distributions or allocations“ p. 425. Procedural Justice: “the fairness of the procedures used to determine outcome distributions or allocations.” p. 425. Interpersonal Justice: “reflects the degree to which people are treated with politeness, dignity, and respect by authorities or third parties involved in executing procedures or determining outcomes.” p. 427. Informational Justice: “focuses on the explanations provided to people that convey information about why procedures were used in a certain fashion or why outcomes were distributed in a certain fashion.” p. 427. While interpersonal and informational components of justice are important in our overall understanding of justice, I believe that they are not as closely related to issues of restorative justice as are procedural and distributive justice. Therefore, while I do capture data on interpersonal and informational justice to show that restorative justice is distinct from extant justice dimensions, for the sake of limiting the scope of this project to something manageable, I do not consider specific antecedents and outcomes of interpersonal and informational justice in this project. That is, the full dimensionality of justice is captured to support the empirical demonstration of dimension distinctiveness, but the reader’s attention will be drawn to the literature regarding distributive and procedural justice issues, discussed below, with regard to the antecedents and consequences of the justice dimensions of interest. An additional organizational justice 38 dimension, restorative justice, is proposed as an important factor that is tied to procedural and distributive justice in contexts in which it is necessary to remedy past imbalances. and as such, relevant antecedents and outcomes for that dimension are also addressed. Distributive Justice Distributive justice relates to the fairness of the actual outcome: that is. the distribution or allocation (Colquitt, Conlon, Wesson, Porter, & Ng, 2001). Distributive justice as captured by Colquitt (2001) using items from Leventhal (1976) is an indirect measure of justice. That is, it captures fairness criteria by looking at underlying criteria of fairness, rather than asking specifically about fairness. For example. distributive justice items ask about whether the outcome reflects the effort put into the work. or the contribution made (see Colquitt, 2001). In the case of the items used by Colquitt. the underlying issue being tapped is equity (rewards are distributed in accordance with the contribution of the recipient; see Leventhal, 1976). As the items used by Colquitt are similar in format to those commonly used in the justice literature (e.g., McFarlin & Sweeney, 1992; Moorman, 1991; Niehoff& Moorman, 1993; Price & Mueller. 1986) and Colquitt’s scales are now being adapted for use (e.g., Bell, Wiechmann, & Ryan. 2006'). I view the standard framing of distributive justice in the literature as equity-based with that equity exchange being based on merit (e. g., contribution and performance). This is a useful general framing for justice in selection, and is the dominant conceptualization according to Colquitt, Greenberg, and Zapata-Phelan (2005), but this framing does not provide altogether meaningful information in the context of strong preferential selection under affirmative action where the distribution would be based on demographic characteristics and yet could still be construed as fair, even though this would be a “rule 39 violation” (see Ployhart and Ryan, 1998). As noted by Arvey and Sackett (1993, p. 194). the “presumption that selection should be based on the principle of meritocracy is not shared by everyone, and it is violated in many instances.” Under affirmative action policy in South Africa (Government Gazette, 1998), race and gender would typically be critical factors in determining who would be awarded the desirable outcome of being hired. Therefore, I argue that distributive justice as currently framed in the literature is insufficient to explain justice under strong preferential selection where there are demographic characteristics that would be a key factor in decisions. F olger (1986) posits that in referent cognitions theory there are three conditions that link to resentment about the decision-making process. These include high referent. low likelihood of amelioration, and low justification. Those who are members of non- target groups and are not getting hired due to affirmative action priorities in South Africa could easily imagine themselves being hired (high referent). As affirmative action is part of current labour law, they may have little hope that things will change (low likelihood of amelioration). Finally, if they do not believe affirmative action is a good policy they will believe that the hiring decision should have gone differently (low justification). Thus this model would lend support for non-target frustration where there is limited support for affirmative action. Procedural Justice Procedural justice aspects of process include, for example, the organization having valid reasons for their hire choice, treating candidates decently, and having a fair selection procedure for determining those to be turned down (see discussion in Greenberg, 1990). According to Colquitt, Greenberg, and Zapata-Phelan (2005). the 40 concept was first introduced by Thibaut and Walker (1975) who found that those who had greater control in an exchange (in their study, resolving disputes) perceived the process as more fair. Specifically, process control was identified as important. Further. it is helpful to have process rules clearly specified. Leventhal (1980) provides six rules for fair procedure. These include consistency, bias suppression, accuracy, correctability. representativeness, and ethicality. While these underlying rules are evidenced in the procedural justice scale items of Colquitt (2001) and are important in our understanding on justice expectations, these rules do not necessarily fit the context of strong preferential selection under affirmative action, where some would argue that selection based on demographics is biased and immoral, yet others would construe selection decisions under such a system as procedurally fair. Therefore, I believe that we need to reconsider the dimensions of justice and expand beyond the extant four-factor model of Colquitt. et al.. (2001) in the context of strong preferential selection. While some researchers criticize studies that continue to slice justice into more specific dimensions (e.g., Ambrose & Arnaud, 2005), I believe that further specification serves to build our understanding in certain contexts. While parsimony is desirable. the current theory of justice as a four-factor construct does not address the case of societal corrections and a redistribution of resources to group members according to special rules for group distributions. I therefore propose a fifth dimension, restorative justice. Dawes (1994; p. 223), in his discussion of the principles of affirmative action programs, argues that affirmative action programs don’t work because they claim that “we can affirmatively affect the probability of allowing members of target groups to be selected without affecting the probability that individual members outside the group will 41 not be selected.” He emphasizes that there are both group and individual decisions that are made, as deciding which group to select simultaneously precipitates decisions about which individuals to accept or reject. Dawes suggests that an appropriate approach to affirmative action would be to recognize the duality and simultaneity principles by attempting to minimize unfairness, as the system can’t be fair to everyone at once. Published research regarding affirmative action attitude reflects the overarching view of affirmative action as fair depending on race, gender, and self-interest, for example (Harrison, et al., 2006). Yet, such research does not capture the possibility of individual recognition of duality and simultaneity, where affirmative action may be valued as fair as long as it can minimize within-group unfairness, in a context where redistribution is necessary. Therefore, I propose that we consider an additional justice dimension, that of restorative justice, to facilitate understanding of these principles. I believe that there is a need to recognize the place of societal redistribution based on group need, along with fair process within group when distributions are being made based on group membership, in understanding fairness. I will argue that theory distinguishes restorative justice from other justice dimensions. it can be operationalized and measured independently, and it has different effects and different antecedents (see discussion of Four Key Questions based on Schwab (1980) and Bobocel and Homval (2001) in Ambrose and Arnaud, 2005). Restorative Justice Restorative justice (de Gruchy, 2002) is a “form of justice that has to do with healing relationships, when reconciliation becomes part of the justice equations.” Roche (2006) argues for a broader recognition of restorative justice beyond the fairly narrow 42 field in which the topic is addressed, and this study provides a clear response to that call. Criminal justice research frames the idea of restorative justice as occasions where the perpetrator of a crime must offer to engage in some compensatory behavior that benefits the victim, as a way to restore imbalance (see Bazemore, Elis, & Green, 2007). Such a concept is not tested in current US-based justice models in the industrial/organizational psychology literature, yet there may be the potential to generalize the concept to workplace justice in selection under preferential treatment, where the victims (members of previously disadvantaged groups) benefit from compensatory behavior (preferential treatment in selection) at the perceived expense of the perpetrator (whites who have been beneficiaries in a social imbalance). Steiner and Gilliland (2001) do note the recognition of special needs groups within certain cultures, where it is regarded as appropriate to provide advantages to certain groups who do have special needs. For example, they state that, “from a self-interest perspective, minority groups in a given country may feel that equality and special needs are important to correct long-term effects of discrimination against their group.” p.131. In South Africa, discrimination has been the case for the majority group (black Africans) and two minority groups (coloured people and Indians) (see Terreblanche, 2002). This distribution based on a societal correction could be viewed as moral and as such may fit what Goldman, Slaughter, Schmitt, Wiley, and Brooks (2008) call a deontic-based need fulfillment. Feather (2fl08) has similarly proposed capturing distributions based on entitlement. I propose that restorative justice differs from need-based distributive organizational justice in that restorative justice captures a desire for general societal-level redistribution. The redistribution is based on an overall societal correction and a social 43 restructuring plan. It is not a function of individual—level need, even though it must be implemented at the individual level for the societal correction to take place. Research that has been conducted on need-based distributions tends to focus on the neediness of the individual recipient (e.g., Giacobbe-Miller, Miller, & Victorov, 1998; Murphy-Berman & Berman, 2002; Pratto, Tatar, & Conway-Lanz, 1999; Steiner, Trahan, Haptonstahl, & Fointiat, 2006) and frames need and merit as alternative choices (e. g., Murphy-Berman & Berman, 2002; Pratto, Tatar and Conway—Lanz, 1999). We are not currently capturing the possibility of duality and simultaneity, or distributions based on a broad social redistribution requirement rather than an individual need. The moral principle of correction of past wrongs is a key component in a justice model when considering affirmative action. Affirmative action programs are built on the premise that redistribution is needed to correct some inequality (see Dawes, 1994). According to de Gruchy (2002), “Covenanting implies accepting responsibility for the past and committed participation in its healing, sharing together in the task of restoring justice in the present, and keeping hope alive for greater reconciliation in the future.“ (p. 6) He adds, “The sacrament of penance provides an important analogy for what is required in overcoming social and political alienation and fostering national reconciliation. Indeed, I would suggest that the sacrament of penance is the sacrament of restorative justice.” (p. 101) Goldberg (1997, p.67) pays careful attention to the issue of the past, where historical context is an important contributor to understanding and developing a sound moral standpoint and related justice perceptions. He defines the principle of compensation; “An individual harmed in violation of his rights should be restored by the 44 perpetrator of the injury to the position he would have occupied, had the injury not occurred.” As noted by Goldberg, this is not a payment because of an unpleasant experience, such as are the benefits accorded to military veterans. This is simply a correction of a specific injury, and in South Africa such injury was clearly inflicted upon groups by apartheid rule (Ross, 1999; Terreblache, 2003; Thompson, 1990). I define restorative justice as the fairness of outcome distributions or allocations to target groups that have suffered inequality in the past. In the South African context. those groups are identified under affirmative action policy as blacks (including black African, coloureds, and Indians), women, and the disabled (Government Gazette, 1998). I propose that a measure of restorative justice expectations should capture the expectation that distributions are being made to members of these target groups, and in this study, race and gender groups are the target groups of interest. Hypothesis 1: A five-factor justice model that includes Colquitt ’s (2001) dimensions as well as restorative justice expectations will be supported. Understanding simply that general redistributions are being made is insufficient to fully capture the issues of justice and the principles of duality and simultaneity. We must also capture differences in views of handling of different target groups as some groups are considered as being in greater need than are others (see earlier discussion of South Africa’s hierarchy of advantage). Thus, I believe that it is also important to consider specific circumstances of distributive and procedural justice, where there is a redistribution to specific target groups. I suggest that capturing distributive restorative justice as the fairness of outcome distributions or allocations within target groups as an alternative to the general distributive justice of the selection system. As race and gender 45 groups are the target groups of interest in this study, I examine distributive restorative justice for women and blacks (including black Africans, coloureds and Indians) to better understand the distributive justice of distributions within these designated beneficiary groups. We also need to understand the justice of how decisions are made for distribution within target groups. There may be expectations that some groups are being exposed to more appropriate process than are others, and I argue that general procedural justice of the selection system will not capture sufficient information. The examination of procedural issues for within-group distribution decisions acknowledges the duality and simultaneity principles noted by Dawes (1994). I propose also capturing procedural restorative justice; the fairness of the procedures used to determine outcome distributions or allocations within target groups. One criticism of affirmative action is that within- group selection decisions are based on tribalism, former exile status, or whom you know. rather than merit (Madi, 1997), and the procedural component within groups is important. As race and gender target groups are the groups of interest in this study, I examine procedural restorative justice for women and blacks. While affirmative action policy has been operational in South Africa for over a decade, there remain those who are frustrated that unemployment for non-whites remains higher than for whites, and that the pace of hiring and promotion of non-whites at the individual level is too slow (see Statistics South Africa, 2008; Vavi, 2007). These views have articulated by Vavi (2007), the leader of the Congress of South African Trade Unions (COSATU) most of whose members are black Africans. At the same time. whites, and white men in particular, express concern that they are not being considered 46 due to race and gender, and are assuming that jobs are being assigned based on target group membership (Herholdt & Marx, 1999; Ramsay, 2005). White females are reportedly frustrated as they expect to be supported by affirmative action policy but find themselves placed after blacks (Ramsay, 2005) and white men have a particularly negative view as they view themselves as being discounted on two fronts; race and gender. Thus, black Africans may view redistributions as insufficient and expect that restoration is not taking place, while whites, and white men in particular, may view redistributions as prevalent, and expect that restoration is taking place. Indian men and women, who have had better access to education than have black Africans or coloureds. are being placed in leadership positions, and I anticipate that they would see the prevalence of redistributions taking place. The hierarchy of advantage and differing access to work experience for the different race/ gender groups suggests that those from race/gender groups that have had the most advantage and the best access to experience. and yet are target group members, are likely to see greater levels of redistributions to their own group, so View redistributions as taking place. Hypothesis 2: Restorative justice expectations will vary significantly across race/gender groups with black A fiicans having the lowest expectations of restorative justice, and white men having the highest expectations of restorati ve justice. The concept of relative deprivation (Runciman, 1966) provides an explanation for differing justice perceptions as the result of a process of social comparison. Those who are not being hired for jobs make a comparison between themselves and those who have been hired. Those who are qualified for positions, yet expect that they are being turned down on the basis of race or gender, will be presented with their relative deprivation. hence a sense of injustice. This is important in expectations of distributive justice. as 47 distributive justice is based on principles of equity and merit, and in procedural justice. which is based on consistency, freedom from bias, and ethicality. As job applicants make assumptions about how jobs are being allocated, and the groups to which they are assigned, I anticipate that there will be race and gender group differences in the expectation of distributive justice and procedural justice. Because of the hierarchy of advantage in South Africa, and differing access to work experience based on race and gender, I anticipate that individuals make assumptions about the level of skill, experience and capacity to contribute that other applicants have, based on race and gender. and thus expectations regarding procedural and distributive justice are a function of group membership. Blau (1964) acknowledges the role of expectations in fairness, where individuals anticipate a distribution that is a fair exchange. Social norms are taken into account in building these expectations, and in terms of procedural justice, those norms include consistency, freedom from bias, and ethicality (see Colquitt, 2001). As found by Kravitz and Klineberg (2000), membership in a target group is an important factor in justice perceptions. They showed that in their sample in the US, blacks were the least likely to view affirmative action policies as giving unfair advantage to minorities and women. That is not to say that blacks in the US necessarily view preferential treatment as fair. Cropanzano, Slaughter, and Bachiochi (2005) showed that in a US sample of black engineering students, tiebreak and preferential treatment policies were rated as neutral with regard to their procedural and distributive justice. According to interviewees in South Africa (Ramsay, 2005). there is a perception among blacks that blackness should give you a free passport to success and that blacks 48 expect to be placed and promoted simply because of race. Thus, target group members may view current process favorably (as being free of bias against target group members. and ethical) and have higher expectations of procedural justice than will members of non- target groups. Those from non-target groups may allow their expectation that jobs will go to someone of another race to color their perception of the entire selection process. deeming it procedurally unfair, even though it may be unbiased when selecting candidates within a race group. ' Hypothesis 3: Distributive justice and procedural justice expectations will vary significantly across race/gender groups, with white men having the lowest expectations of procedural and distributive justice, and white women having the highest expectations of procedural and distributive justice. Expectations of procedural and distributive justice among black Africans will be lower those of coloureds, which will be lower than those of Indians. The relationship between restorative justice expectations and procedural and distributive justice expectations is expected to vary across groups. For example. restorative, procedural and distributive justice are expected to converge for black African men, yet be inversely related for white men. Black African men who believe that a redistribution is not taking place would have low expectations of procedural and distributive justice. Indian women, however, are likely to see higher levels of redistribution, which would likely be to their own group, and feel satisfied that the process is fair and unbiased. White men who believe that there is a high level of redistribution taking place will assume that people with less skills and experience are getting the jobs, and that the distribution is taking place in a biased and unethical way. Hypothesis 4: The relationship between restorative justice expectations and procedural and distributive justice expectations will vary significantly across race/gender groups with white men having an inverse relationship between restorative justice expectations and procedural and distributive j ustice. white white women, black Africans, coloureds and Indians will show a positive 49 relationship between restorative justice expectations and expectations of ' procedural and distributive justice. We know from existing research that affirmative action programs that are the most prescriptive are the programs that are more likely to generate negative attitudinal responses (Harrison, Kravitz, Mayer, Leslie, & Lev-Arey, 2006). While these meta- analytic results are helpful in summarizing the findings of existing literature. I propose research that drills down in a more specific way with regard to justice. Harrison. et al.. choose to combine studies that capture fairness perceptions as the dependent variable and those that capture attitude toward to affirmative action as the dependent variable, arguing that although these two constructs are not identical, they are “powerfully bound up in one another.” While the majority of the studies in their meta-analysis used attitude toward affirmative action as the dependent variable, I question the appropriateness of lumping the two constructs together. My concern is supported by their finding that there were larger relationships between race and gender and attitude toward affirmative action than was the case for the relationships between race and gender and fairness as the dependent variable. These constructs are conceptually distinct, and for us to be able to understand how people react under affirmative action policy, I view it as essential that we tease apart issues of support for a policy and issues of perceptions of justice regarding process. Harrison, et al., (2006) do acknowledge with their study limitations that they recommend that researchers use attitude rather than fairness measures in future affirmative action research. I posit that not only is separation of these two constructs necessary for theoretical development; it also is essential to provide a basis on which practical recommendations regarding affirmative action process can be built. 50 Although they are examining affirmative action attitudes rather that justice attitudes, Bell, Harrison, and McLaughlin’s (2000) results are informative. In testing Fishbein and Ajzen’s (1975) theory of reasoned action they demonstrate that demographic factors have unique effects on affirmative action attitude beyond those explained by beliefs and evaluations. Harrison, et al. (2006) show that both the race effect and the gender effect on attitude toward affirmative action are moderated by program prescriptiveness. Specifically, the effect of race and gender was especially potent under the most prescriptive preferential treatment. South African research has concluded that women are more positive than are men about employment equity (van Rensburg & Roodt, 2003; van Rensburg & Roodt, 2005). This is not surprising considering that women have target status. Thus, considering the prescriptiveness of the South African policy, and that demographics contribute to variance beyond that explained by other antecedents, I anticipate race and gender effects. In research on qualifications, fairness ratings in selection have been linked to qualifications, where the selection of candidates with superior qualifications is regarded as more fair (see review in Kravitz, Harrison, Turner, Levine, Brannick. Denning. Russell, Conard, and Bhagat, 1996). Under-qualified candidates may be selected in South Africa, where there has been historic inequality in access to education and certain work opportunities, as there is a policy of affirmative action that gives a positive weight to demographic status. The recognition that target group members in South Africa may be unable to compete on equal footing with those from previously advantaged groups is reinforced by the government policy that includes “potential” as a sufficient qualification for selection (Government Gazette, 1998). While less-qualified candidates may be chosen 5] based on demographic characteristics, this does not necessarily mean that merit has been put aside entirely. Also, it is still possible to choose the best-qualified candidate within a target group. I view recognizing this possibility as key to improving our understanding of the multiple dimensions of justice, and to informing practice regarding selection process under the real constraints of strong preferential selection. First, white women have historically experienced relative privilege when compared to other target groups. As discussed earlier, they have had access to better schools and work opportunities (Statistics South Africa, 2006a; Thomson, 1990). As a result, they are likely fairly well equipped to compete on traditional selection criteria such as education, skills and to a lesser extent, experience. Given their status, there would be little reason to ignore traditional merit criteria in selection decisions when selecting white women. However, there may be the expectation that traditional criteria are being overlooked for white women, to be more inclusive of black women. Therefore, I anticipate that procedural restorative and distributive restorative justice expectations about the treatment of women will vary across groups. Similarly, as black Africans have historically had the worst access to education and challenging work, I anticipate that both women and men from these groups are more likely to be selected based on potential, or on other characteristics such as social connections. Therefore, I anticipate that procedural restorative and distributive restorative justice expectations regarding treatment for blacks in black. African women and men will be moderate to low, with more variability than that seen for other race groups. I anticipate that coloureds and Indians are more likely to see distributions to members of their group 52 based on merit, and have higher expectations regarding procedural and distributive justice. Procedural restorative justice and distributive restorative justice are expected to vary across race and gender groups. Those who are target group members. yet are also members or groups lower on the hierarchy of advantage, are more likely to anticipate that the procedure and distribution is unfair because within the target groups, preference may be given to Indians first and then coloureds, rather than to black Africans, although all are target groups. Similarly, preference within women as a target group may be given to Indian women over coloured or black African because they have had relatively better access to experience and education and over white women because the organization is striving to redistribute to blacks. Distributive restorative and procedural restorative justice will be measured for treatment of two target groups: blacks and women. Blacks and women are categories that are designated as beneficiary groups under affirmative action law (along with the disabled, who fall beyond the scope of this study). Hypothesis 5 .° Distributive restorative and procedural restorative justice expectations for treatment of blacks will vary significantly across race and gender groups with black A fiican men having the lowest expectations of distributive restorative and procedural restorative justice, lower than the expectations of coloureds, which will be lower than those of Indians. Indian women will have the highest expectations of justice. Hypothesis 6: Distributive restorative and procedural restorative justice expectations for treatment of women will vary significantly across race and gender groups, with black African women having the lowest expectations of distributive restorative and procedural restorative justice, lower than the expectations of coloureds, which will be lower than those of Indians. W hitc women will have the highest expectations of justice. Having addressed the dimensions of justice that are relevant in this study. I now direct the reader’s attention to group characteristics and individual-level factors that may 53 contribute to our understanding of justice expectations. First, I discuss group _ characteristics. Group Characteristics While individual-level constructs have been the primary focus of the justice literature, group-level issues are also important and research in this area is needed. The South Africa context provides an excellent opportunity to examine group-level constructs and their cross-level implications. South Africa’s historic formalized segregation by race (Ross, 1999; Thompson, 1990) has meant that those of the same race group had commonality with others of their race in terms of access to resources, with whites having the greatest access, followed by the Indian, coloured and black African groups. in order of descending levels of access (Statistics South Africa, 2006a; Thomson, 1990). Similarly, women have had less access to certain job categories and resources than have men (Department of Labour, n.d.; Statistics South Africa, 2006a; Thomson, 1990). creating a shared experience within gender group. As I have discussed in the Socio- Political History section of this paper, each race and gender sub-group in South Africa had a unique position in society, creating a shared experience with others of the same race and gender subgroup. Although South Africa’s apartheid-era Population Registration Act was repealed in the 19905, theoretically ending the legal basis for race classification. people continue to identify themselves by these categorizations (Christopher. 2002). The development of group identity as an outcome of administrative categorization has been shown elsewhere (Kertzer & Arel, 2002), and South Africans appear to hold strong racial identities (Christopher, 2002; Vestergaard, 2001). Population group categorizations and gender are currently an important part of policies of redress in South Africa. where 54 designated beneficiary status for affirmative action is based on race, gender, and disability status (Government Gazette, 1998) as these groups are identified by the government as being underrepresented in the workplace. In this context, and for the purposes of this project, race and gender subgroups are necessary and appropriate grouping characteristics for considering group-level constructs. 1 posit that the homogenous social and economic context for race and gender groups created by shared status under apartheid policies and now under the new dispensation has meant that race/gender groups in South Africa have shared experiences and along with that, have developed shared attitudes and perceptions (see Klein & Kozlowski, 2000, on shared team properties; Campbell, 195 8, on common fate and proximity). Experiences under these policies can be viewed as goings on and events that allow collective constructs to emerge and be maintained (see Morgeson & Hofman. 1999). Social information processing within race and gender groups may be one mechanism by which these shared views develop (see Klein & Kozlowski, 2000) and with South Africa’s historic racial segregation and different gender roles, individuals have had greater exposure to within-group interaction. According to Morgeson and Hofman (1999, p. 256), “only through interaction does a construct acquire meaning and structure. As a result, interaction allows collective constructs to emerge and to be sustained, and it is this interaction that exerts an influence on organizational members.” With South Afn'ca’s dramatic differences in experience across race and gender groups. I expect that within-group interaction has led to the building and maintaining of shared perspectives, and as individuals continue engagement with others of their race and gender, I anticipate strong group effects. 55 Group Level Beliefs and Culture 1 will now offer details regarding the first two of three specific collective constructs that I posit exist, and plan to demonstrate in the South African context; group belief in white guilt and group culture of support for affirmative action. We know that organizations in South Africa are operating under labour laws specifying affirmative action (Government Gazette, 1998). Nonetheless, not all groups may support such policy. I expect that a group-level culture of support for affirmative action may exist, and that such a culture is built in part on the group’s shared belief in white guilt. I therefore intend to capture group-level referent views and aggregate responses to those measures of group belief in white guilt and group culture of support for affirmative action to a group level. This will allow us to better understand the group-level relations between beliefs about white guilt and group culture of support for affirmative action, and well as the cross-level effects of group characteristics as they influence individual-level justice expectations and related outcomes. I use the term culture rather than climate in line with the distinction discussed by James, Choi, Ko, McNeil, Minton, Wright, and Kim (2008) as I am capturing’normative beliefs and shared expectations measured through aggregated group—referent scales. Group Belief in White Guilt I define group belief in white guilt as a shared belief in a broad sense of guilt and responsibility on the part of whites regarding the social inequality between whites and non-whites. According to de Gruchy (2002, p. 194), there is such a thing as collective guilt; “Collective guilt, whether it is the collectivity of a colonial power, Nazi leadership. the apartheid government cabinet or the wider collectivity of those who kept them in 56 power is, it seems to me, an undeniable reality.” Collective guilt has been of interest to American and European researchers (e. g., Doosje, Branscombe, Spears, & Manstead. 1998; Iyer, Leach and Crosby, 2003; Swim & Miller, 1999) and these ideas can be extended to the context of post-apartheid South Africa. While their research has shown that it is possible to capture the group members’ feelings of group-based guilt (e.g.. whites’ feelings of white guilt by Swim and Miller, 1999, and experimental-group members’ feelings of in-group guilt by Doosje, et al., 1998), l posit that group members also hold beliefs about the level of guilt of out-groups, and that these beliefs are shared by group members. That is, a culture develops that supports a shared belief. Further, that culture affects individual attitudes and behaviors. Murithi (.2006) outlines five stages of the peacemaking process found among Ubuntu societies: acknowledging guilt, showing remorse and repentance, asking for and giving forgiveness, and paying compensation or reparation as a prelude to reconciliation. As noted by F olger and Cropanzano (2001), accountability and blameworthiness are fundamental to justice, and l posit that in the South African context, an acknowledgement of white guilt is an important aspect of the justice context, where compensation and reparation are being made in part through affirmative action in employment. Steele (1990) frames white guilt as a group culpability, which serves to empower blacks. He shares the story of an incident where a friend tested his own black power by attempting to make white guilt salient in a white man in a hotel men‘s room. The story demonstrates the black man’s belief in white guilt, and it is this assignment of guilt to a group that can be accomplished by either in or out-group members, although research has 57 tended to focus on in-group assignment (e.g., Swim and Miller, 1999; Klandermans, Werner, & van Doom, 2008). Information is available about the existence of past and present inequities between races and men and women (Statistics South Africa, 2006a; Terreblanche, 2002), and government policy regarding affirmative action is explicit about the need to correct these imbalances (Government Gazette, 1998), so it should come as no surprise to any South African that whites and males have held a position of tremendous privilege. The recognition that the relatively advantaged position of Indian and coloured people compared to black Africans (Statistics South Africa, 2006a; Terreblanche, 2002) helped to maintain the suppression of black Africans (see Vron, 2002, regarding Gandhi’s place in South Africa) may serve to lower belief in white guilt among members of Indian and coloured groups as they may assign some share of responsibility to their own groups. US-based research on gender differences in white guilt have shown mixed results. and on the whole, levels of guilt have been low, yet in two of their four studies, Swim and Miller (1999) indicated gender differences in white guilt, with women more likely to report white guilt. In an examination of the antecedents of group guilt, Mallett and Swim (2007) found an interaction between justifiability and responsibility that occurred only for whites and women. According to Hoffman (1994), people do not need to be transgressors to feel guilt. We know that women and men are different in many ways. and Chao and Moon (2005) argue for gender-as-culture, highlighting the appropriateness of gender as a demographic tile in the mosaic of tiles that will help us to understand culture. In a context where women are faced with similar social constraints, gender may also be an associative grouping. The South African context, with its clear inequality between race 58 and gender grbups, and relatively recent history of legislated inequality (Terreblanche, 2002), may allow clearer gender differences to be observed. Thus, I anticipate that belief in white guilt may vary across race and gender groups. Culture of Support for Affirmative Action The culture of support for affirmative action is expected to vary across race and gender groups. Self-interest is one explanation for why those who are beneficiaries support a policy. For example, research involving perceptions of banding in selection has shown that when banding is understood as a type of affirmative action, African Americans show greater support for banding as fair than when it is not viewed as a type of affirmative action. The same relationship was not true for whites (Truxillo & Bauer. 1999). Levi and Fried (2008) demonstrated that white and blacks differed in their level of support for strong preferential hiring of minorities, with whites holding negative views and blacks holding a slightly positive view. Weathers and Truxillo (2008) showed that Asian Americans are more likely than are white Americans to perceive that Asians should benefit from affirmative action. Klandermans, Werner, and van Doorn (2008) showed that in a sample of white South Africans, gender was associated with support for affirmative action, with women being more supportive. As members of target race/ gender groups have shared potential to benefit from affirmative action policy, I believe that there is likely a culture that supports a shared view of support for affirmative action in race/ gender groups. The strong philosophy of Ubuntu (which literally translated means. I am because we are) is a core component of black African culture. Murithi (2006, p. 25) notes that this philosophy “highlights the essential unity of humanity and emphasizes the importance of constantly referring to the principles of empathy, sharing and cooperation 59 in our efforts to resolve our common problems.” Even the Zulu greeting, Sanihona. the equivalent of hello, really means, we (I and my family) see you (you and your/amily). This acknowledgement of oneself as more than just an individual, but rather as a representative of a community, demonstrates the recognition of the self as a member of a group. Steiner and Gilliland (2001), in their discussion of cultural factors that may contribute to justice reactions in selection, note that in most western cultures, selection decisions are based on hiring the most deserving candidate, with the most deserving being the one most likely to perform well -- predictive validity in selection is regarded as preferable. However, South Africa is built of many cultures, and that of black Africans in particular has been identified as communitarian (Trompenaars & Hampden-Turner. 1998). In considering a workplace selection decision, I speculate that more collectivist groups may feel more positively about the broader benefit to society when previously disadvantaged people are hired, yet more individualistic cultures are less satisfied with a process that benefits others but denies opportunity to their own. As South Africa’s race groups vary in their collectivism, I anticipate that the group culture of support for affirmative action will vary across race/gender groups. The emergence of this culture and consensual view within groups is likely supported by members’ shared status under affirmative action policy as well as by interactions within the group (see Kozlowski and Klein, 2000). 60 Group—Level and Cross-Level Relations Next, I address the relationship between group belief in white guilt and group culture of support for affirmative action. Steele (1990, p. 79) promotes white guilt as a driving force in policy change in the US: “I believe that in the sixties the need for white redemption from racial guilt became the most powerful, yet unspoken, element in America’s social-policy- making process, first giving rise to the Great Society and then to a series of programs, policies, and laws that sought to make black equality and restitution a national mission. Once America could no longer deny guilt, it went after redemption, or at least the look of redemption, with a vengeance.” Doosje et al. (1998, p. 873) note that feelings of guilt “result in tendencies to make up for the wrong that an individual has done to another.” They showed that making salient a group’s history of mistreatment (in the case of their study, Dutch mistreatment of Indonesians) was linked to collective guilt. Further, they demonstrated that people were more willing to compensate for the behavior of their group if their behavior had been unfair to another group. Similarly, researchers have fecused on the opportunity for reconciliation where perpetrators of crimes take responsibility for their actions and engage with victims to restore justice. Tangney (2001) has found the moral emotion of guilt can predict acceptance of responsibility for crimes and lower aggression. This model of perpetrator guilt and compensatory behavior toward victims as a means of building positive outcomes in the community is a useful framework within which to consider social change and restorative justice. If people hold the belief that whites should have a sense of guilt and responsibility for past inequity, they may be more likely to support compensatory policies such as affirmative action. Swim and Miller (1999). in a series of studies at the individual level, showed that white guilt is associated with more positive views regarding affirmative action. Similar results were offered by Klandermans. 61 Werner, and van Doom (2008) in a sample of white South Africans. While these studies were conducted capturing the perspectives of white participants regarding their own feelings of group guilt, as noted earlier, I believe that beliefs regarding white guilt should be captured from people of all races, as non-whites also hold opinions regarding the guilt of whites, and that shared beliefs should be captured in a justice model. Acknowledgement of white guilt may serve as an explanation for why affirmative action is necessary and appropriate. Justification has been linked to more positive attitudes to affirmative action (Aberson, 2003). While collective guilt and support for compensatory action or entitlement have been captured in prior research (e.g., Boeckmann & Feather, 2007; Doosje, Branscombe. Spears & Manstead, 1998; Klandermans, Werner, and van Doom, 2008; Swim & Miller. 1999), these constructs have been measured and conceptualized at the individual level, yet, as noted earlier, I anticipate that cultures within race/gender groups develop for each of these constructs. While I focus on the relationships between these group-level constructs, my logic is driven by relationships that have been demonstrated at the individual level. That is, white guilt at an individual level has been linked to greater willingness to compensate for the group’s behavior (Doosje, et al., 1998) and to positive attitudes toward affirmative action (Swim & Miller, 1999). My first group-level hypothesis is based on the inclination toward reparation where there is a sense of guilt: Hypothesis 7: Group belief in white guilt will be positively related to group culture of support for aflirmative action. As individuals are nested within their race/gender group, I expect that they are individually influenced by the culture of their group, and I therefore use those group-level factors as cross-level predictors of individual-level expectations. Perceptual measures of 62 the environment, that is, measures capturing the culture of the group, can be linked to behavior at the individual level. Kozlowski and Klein’s (2000) discussion regarding top- down process with organizations having a direct effect on the behavior of individual employees can be generalized to groups such as race and gender groups, where the culture of the group will directly influence members of the group. People do not operate in a vacuum — they are influenced by their context (e.g., Hofmann, Morgeson, & Gerra. 2003; Seibert, Silver, & Randolph, 2004). Social information processing theory (Salancik and Pfeffer, 1978) may help to explain this link in that individuals do not operate entirely independently of their social context. Rather, they are adaptive and take into account cues from their social environment. As noted by Rousseau (1985) contexts structure behavior. and the attributes and function of the context can be used to understand individual behavior. Chao and Moon (2005) propose the mosaic as a framework for understanding multiple cultural identities at individual and group levels. I view group culture of support for affirmative action as one tile in the cultural mosaic that influences the individual. while other individual-level factors also play a part. Constructs at higher levels can exert independent influences (Kozlowski & Klein, 2000); this is a top-down contextual influence. The individual takes into account the group’s culture of support for affirmative action in building his or her individual expectations regarding the selection process. However, although a person may be nested within a group, and influenced by that group. they also have their own individual differences, beliefs and values and experiences (see Bell, Ryan, & Wiechmann, 2004) that contribute to their sense-making. Thus. both 63 group-level and individual-level tiles must be part of the mosaic (see Chao & Moon. 2005). The normative group culture of support for affirmative action may add to the explanation of variance in an individual’s justice expectations. That is, in addition to individual level predictors, there will be a cross-level inference (James & Williams. 2000). As noted by Crosby, Iyer and Sincharoen (2006) studies have varied widely in how they have operationalized attitude to affirmative action, and are only sometimes capturing fairness. According to Harrison, Kravitz, Mayer, Leslie and Lev-Arey (2006). perceived fairness and attitudes to affirmative action are so closely related that they may be grouped together when evaluating attitudes to affirmative action. While such an approach may be convenient for methodological purposes, it does little to further our theoretical development of issues around affirmative action, and I believe that a critical path for theoretical development in this area is to separate key components and examine their relations. I will therefore discuss specific hypotheses regarding group culture of support for affirmative action and its relationship with individual-level procedural. distributive and restorative justice expectations. Individuals who are nested within groups with a culture that lacks support for affirmative action will view the process used to determine outcome distributions as unfair. However, those who lie within groups with a culture of greater support for affirmative action will be influenced by that positive culture regarding affirmative action and will not make such negative judgments about the fairness of process. Thus, I anticipate a greater cross-level effect on individual procedural justice expectations for members of groups with a culture that lacks support for affirmative action than will be 64 the case for members of groups that have a culture of support for the policy. That is, the culture is not as relevant to the building of justice expectations when the culture is one of support -— the link is discounted in the minds of those who fall within such a context. Hypothesis 8: Group culture of support for affirmative action will add to the variance explained in individual procedural justice expectations, and group culture of support for afiirmative action will be more strongly related to procedural justice expectations for members of groups where there is a culture of lack of support for affirmative action. Individuals who are nested within groups with a culture that lacks support for affirmative action will view the fairness of outcome distributions as unfair. A criticism raised by opponents of affirmative action is that it is a system that violates the merit principle, and distributive justice as currently framed in the literature (see Colquitt. 2001) captures the fairness of outcome distributions based on items that tap effort and performance. That is, there is an underlying assumption that the distribution should go to the best candidate for the distribution to be perceived as fair. Thus, I anticipate that individuals who are nested within groups that have a culture that lacks support for affirmative action be more likely to hold expectations of distributive injustice. However. those who lie within groups with a culture of greater support for affirmative action will be influenced by that positive culture regarding affirmative action and will not make such strongly negative judgments about the fairness of outcome distributions. That is, the culture is not as relevant when it is a culture of support for affirmative action. Hypothesis 9: Group culture of support for affirmative action will add to the variance explained in individual distributive justice expectations, and group culture of support for aflirmative action will be more strongly related to distributive justice expectations for members of groups where there is a culture of lack of support for aflirmative action. 65 While published research on justice fails to address the problem of restorative justice, such an issue is crucial to our understanding of justice when dealing with strong preferential selection. As noted above, individuals nested within groups are influenced by the culture of support for affirmative action. With regard to restorative justice expectations, I don’t predict any strong links between group culture of support for affirmative action and the individual’s restorative justice expectation. However, I do anticipate that the individual builds expectations regarding procedural restorative and distributive restorative dimensions. Specifically, for members of groups with a culture lacking in support for affirmative action, I anticipate low procedural restorative and distributive restorative expectations regarding the selection of members of the various target groups. Hypothesis 10: Group culture of support for affirmative action will add to the variance explained in individual procedural restorative and distributive restorative justice expectations, and group culture of supportfor affirmative action will be more strongly related to procedural restorative and distributive restorative justice expectations for members of groups where there is a culture of lack of support for affirmative action. Further, I anticipate that the effect of group belief in white guilt on individual justice expectations is fully mediated by group culture of support for affirmative action. Hypothesis I 1: The effect of group belief in white guilt on individual justice expectations is fitlly mediated by group support for affirmative action. Group Justice Expectations As noted by Ehrhart (2004), there has been an abundance of research on fairness at the individual level (see Colquitt, Conlon, Wesson, Porter, & Ng, 2001). but less on justice at the unit level. The current proposal seeks to further our understanding of j ustice cultures, specifically with regard to restorative justice expectations. I expect that 66 race/gender groups hold shared expectations regarding restorative justice. That is. l posit that a culture for restorative justice expectations has emerged within race/ gender groups. Levinson et al. (1966) discuss normative expectations of how individuals and work groups develop schemas for what they should expect from one another, and they note that these normative expectations vary across groups. I view this emergence of group norms as a function of holding shared status as affirmative action targets or non-targets. having a shared socio-political experience and social contact with one another (see Kozlowski & Klein, 2000; Klein & Kozlowski, 2000; Morgeson & Hofman, I999). Mossholder, Bennett, and Martin (1998) measured procedural justice context by aggregating individual justice perceptions regarding their workplace. Naumann and Bennett (2000) framed aggregated justice perceptions as a justice climate. Ehrhart (2004) used an aggregation of individual perceptions of justice to capture procedural justice climate, and Simons and Roberson (2003) show the usefulness of aggregate justice perceptions. Similarly, I intend to capture race/gender groupjustice expectations regarding restorative justice by aggregating individual responses to questions (which have a group referent) regarding restorative justice expectations. As individuals consider who it is that they see being hired, their frame of reference is likely to be those around them. For example, whites are likely to see whites not being hired, and conclude that priority in hiring is being given to non-whites. Black Africans are likely to see black Africans not being hired and conclude that priority is not being given to previously disadvantaged groups, or is being given to Indian and coloured people rather than to black Africans. With this in mind I expect that white men may have the highest mean level of restorative expectations, and be the most consistent as a group. while black 67 African men may have the lowest level of restorative justice expectations, and be less consistent as a group. Individuals do not operate in a vacuum. They are engaged with those around them, and it is this interaction that exerts an influence on group members (Morgeson & Hofman, 1999). Information from the group contributes to the building of the individual‘s schemas. Thus, I anticipate that individual perspectives are influenced by the group culture. That is, group restorative justice expectations are expected to have a cross-level influence on individual-level justice expectations. Existing research has demonstrated cross-level effects of justice climate. For example, Mossholder, Bennett, and Martin (1998) found that procedural justice context contributed to 'the explanation of individual- level satisfaction. Similarly, Naumann and Bennett (2000) found that group-level justice climate contributed to the explanation of individual-level helping behavior and Ehrhart (2004) demonstrated an effect of procedural justice climate on organizational citizenship behaviors. I anticipate that a normative group expectation regarding restorative justice may add to the explanation of variance in an individual’s justice expectations. That is. there will be a cross-level inference (James & Williams, 2000). More specifically, I expect that group-level restorative justice expectations contribute to the forming of individual-level restorative justice expectations. The individual will take into account the group culture for restorative justice expectations in building an individual expectation regarding what is happening. That is, in Chao and Moon’s (2005) terminology, this group-level tile of the cultural mosaic contributes to the individual-level understanding. (One is not part of the other — group-level restorative justice expectations are based on the aggregation of responses of items that have a group 68 referent. It is not the aggregation of the individual-level items with an individual-level referent.) Hypothesis 12: Group-level restorative justice expectations will be positively related to individual-level restorative justice expectations. Further, I expect that group-level restorative justice expectations contribute to the explanation of individual-level procedural and distributive justice expectations. That is. the extent to which one’s group believes that a redistribution is taking place affects the individual’s expectations regarding these two dimensions of justice expectations. However, the nature of the relationship between group-level restorative justice expectations and individual procedural and distributive justice expectations will vary across groups. Based on self-interest (see Harrison, et al., 2006) members of target groups who anticipate restorative justice (that is, expect that there will be will distributions to target groups) view the entire process as procedurally and distributivel y fair. Specifically, members of affirmative action target groups will demonstrate a positive relationship between group-level restorative justice expectations and procedural and distributive justice expectations of the selection process. Members of non-target groups will demonstrate a negative relationship between restorative justice and expectations of procedural and distributive justice of the selection process. Hypothesis 13: The relationship between group-level restorative justice expectations and individual-level procedural and distributive justice expectations will vary across groups such that group-level restorative justice expectations will be positively related to individual-level procedural and distributive justice expectations in target group members, and negatively related in non-target group members. Beyond that explained by the aggregated group-level cultures identified above. race and gender group membership may add to the explanation of variance at the 69 individual level. As a myriad of unmeasured social differences between races, and between men and women, may explain justice expectations, race and gender dummy variables will be included in my analyses to examine whether there is any additional variance explained by group membership. While I have identified cross-level effects here, I have not yet addressed circumstances under which these effects are moderated by individual-level antecedents. I also expect possible person by situation interactions. The interplay of individual and group-level antecedents of expectations is an important component of this study. and hypotheses regarding such interactions are offered after I have laid out main effect hypotheses for the individual-level factors. I will now discuss the individual-level contributors: individual differences, beliefs/values, remote direct experiences. and indirect experiences. I also note proximal direct experiences as a potential control variable. Individual Differences Recent research on employment equity in South Africa has demonstrated that fairness perceptions of employment equity do vary across personal characteristics such as ethnicity, gender and age (van Rensburg & Roodt, 2004; van Rensburg & Roodt. 2005). However, the attitudes being measured have not always been clearly defined constructs. where attitudes to employment equity may encompass several constructs. Thus, existing results offer little guidance. Thus, age will be captured as potential control variable, and I do not offer specific hypotheses regarding this variable. As fear of crime is an additional reason why many people leave the country (“Fears over South Africa’s exodus,” 2006). I measure fear of crime for use as a control variable in predicting emigration intention. 70 Other potential control variables that may relate to emigration intention are having friends who have settled abroad, being dissatisfied with conditions within South Africa. and believing that quality of life would be better overseas (see Miller, Haskell. & Thatcher, 2002). The individual difference of primary interest here is race centrality. and I now discuss hypotheses regarding this individual difference. Racial Centrality The centrality of one’s race group identification may be an important contributor in the building of expectations. Tajfel and Turner’s (1986) social identity theory offers that self-image stems in part from group membership. Yet there are varying levels of identification within groups. In a country where there have been dramatic changes in levels of integration and interaction amongst race groups, it is to be expected that there are varying levels of racial identification. According to de Gruchy (2002, p. 31). “The struggle over identity, and the attempt to construct a common South African identity within which other identities are recognized and respected, is at the heart of national reconciliation. Indeed, it is often said that the ‘national question’ has to do with ‘who is an African’ within the South African context, a question that does not have to do so much with ethnicity as it does with commitment and building relationships.” Sellers, Rowley. Chavous, Shelton, and Smith (1997) demonstrate factors of black identity and the Centrality Scale of their Revised Multidimensional Inventory of Black Identity can be usefully adapted for the evaluation of identity in South Africa. In a study examining the effect of group identity on compensatory behavior. Doosje, Branscombe, Spears, & Manstead (1998) showed that those who were low 71 identifiers with their group were more willing to accept group-based guilt when presented with ambiguous information, which was linked to compensatory behavior. The authors‘ explanation for this effect is that the low identifiers were less defensive than high identifiers, and thus more willing to accept responsibility. People high in white identification may be more likely to suggest in-group variability (e.g., saying “not all whites supported apartheid”) (see Doosje et al., 1998) and may therefore see less group based guilt. Thus I expect that white high identifiers may have low expectations for procedural and distributive justice in this strong preferential selection context while I expect that those with high black centrality have high procedural and distributive justice expectations. Hypothesis 14: The relationship between race centrality and distributivejustice expectations will vary across race groups such that there will be a positive relationship between race centrality and distributive justice expectations for race- based target groups and a negative relationship for race-based non-target groups. Hypothesis 15: The relationship between race centrality and procedural justice expectations will vary across race groups such that there will be a positive relationship between race centrality and procedural justice expectations/or race- based target groups and a negative relationship for race-based non-target groups. African centrality is of increasing importance as South African culture changes and people begin to identify more with their nation than with their race. Strength in a shared identity may help to move the country forward. I posit that a strong African identity is positively related to procedural and distributive expectations. and that a strong African identity moderates the effect of a culture of lack of support for affirmative action on procedural and distributive expectations. That is, where'there is low African identity. there is a strong relationship between culture and justice expectations, but when there is 72 high African identity, the relationship between culture and justice expectations is weakened. Even though an individual may be a member of a group that does or does not support affirmative action, their individual focus on membership in the broader African community, and the need to support the development of that broader community. tempers the effect of the culture on justice expectations. Hypothesis 16: African centrality will moderate the effect of culture of support for afiirmative action on procedural justice expectations such that high African identity weakens the relationship between culture and justice expectations. Hypothesis 17: African centrality will moderate the effect of'culture of support. for affirmative action on distributive justice expectations such that high African identity weakens the relationship between culture and justice expectations. Existing Beliefs and Values Bell, Ryan and Wiechmann (2004) present existing beliefs as an antecedent of justice expectations. These beliefs may tap a variety of cultural and values differences. I incorporate two areas of beliefs into my conceptual modell These include individual support for affirmative action and a preference for merit principle. Support for A flirmative Action While target status is expected to be important in explaining an individual’s expectation of distributive justice, the view that a societal correction is necessary may also be relevant; it may be that the broader social good is more important than the equity of the outcome for the individual. This new norm may predominate, and the need for allocation on this basis may be perceived as fair. Whites may cognitively distort and shift their focus from personal benefits (having a job) to non-monetary societal outcomes (correcting the wrongs of apartheid) as a way to restore the illusion of balance in the equity equation (see discussion in Folger and Cropanzano, 1998). According to Gilliland (1993, p. 698), “situational factors may increase the salience of the alternate distribution 73 rules of equality and needs.” Adams’ equity theory (1965) captures fairness as a calculation of inputs and outputs, where an imbalance would lead to feelings of distress. In the case of selection in South Africa, to balance the input-output equation, individuals may change their cognitions regarding the process, or modify their input behavior. Those changing their cognitions may embrace the societal need for affirmative action, to balance the equation. In South Africa there is a broad understanding on a societal level that affimrative action is needed, and it is socially expected that one agrees with the policy - it's the "right thing” to do. However, on an individual level, opinions do vary, and those who disagree with affirmative action will have expectations that differ from those of people who support affirmative action. Those who support affirmative action are most likely to see the current strong preferential selection framework as making appropriate distributions of jobs. They are also more likely to expect the selection process to be procedurall y fair than would those who view preferential selection as an inappropriately biased process. Hypothesis [8: Support for aflirmative action will be positively related to procedural j ustice expectations. Hypothesis I 9: Support for aflirmative action will be positively related to distributive justice expectations. I Preference for Merit Principle Broad South African culture has a dominant pattern of achievement orientation (Trompenaars & Hamden-Turner, 1998) nevertheless affirmative action requires that selection decisions use group affiliation as a primary consideration for hire. Those who are highly achievement oriented may find this process most frustrating. However. it is not only the process of selecting by race or gender that those high in achievement orientation 74 detest — it is also the choosing of candidates based on cronyism, nepotism and favoritism. This is highlighted as of particular concern when the successful candidate has less merit than another applicant of the same race (Madi, 1997; Ramsay, 2005). According to interviewees (Ramsay, 2005), blacks express concern about being labeled as affirmative action hires and worry about the resultant stereotyping. Both whites and blacks express concern about tokenism and that ill-equipped individuals are being pushed into senior positions where they are bound to fail. One of the major issues raised by affirmative action in South Africa is the fear that hiring and promotion based on race will lower standards in the workplace, reduce competitiveness in the global market, and weaken the local economy. Innes (1993) notes that while many people worry about the lowering of standards in selection decisions under affirmative action, decisions prior to affirmative action were not always based on criteria that related to qualification standards. For example, he notes that in the English-speaking business world in South Africa, it has been important to have gone to the ‘right’ school and lived in the ‘right’ neighborhood, and white Afrikaners have been promoted in the civil service simply because that was in the interests of the Nationalist party. Someone with a preference for merit principle may also be more likely to be concerned with procedural issues, expecting the top candidate to be hired rather than having selection be based on the old school tie. As noted by Kravitz (1995), if people believe that selection decisions should be based on merit, using demographic status as a criterion for employment decisions will be seen as procedurally unfair. This would also mean that the distribution would be viewed as unfair. As the current system is one of 75 selection based on affiliation, I anticipate low expectation for procedural and distributive justice where individuals have a strong preference for merit principle. Hypothesis 20: Preference for merit principle will be negatively related to procedural justice expectations. Hypothesis 21: Preference for merit principle will be negatively related to distributive justice expectations. The preference for merit principle may be useful in demonstrating the need to capture procedural restorative justice, and distributive restorative justice (that is. procedural and distributive justice expectations within specific demographic assignments). The preference for merit principle should show a stronger relationship with the expectations for justice within the specific assignment, such as for blacks. or for women. When a redistribution is taking place based on group membership, we can still have merit-based decisions within-group, although between-group decisions are not based on merit, but rather on group membership. Hypothesis 22: Preference for merit principle will be more strongly related to procedural restorative justice expectations than to procedural justice expectations. Hypothesis 23: Preference for merit principle will be more strongly related to distributive restorative justice expectations than to distributive justice expectations. Remote Direct Experience While differences in existing beliefs based on culture are a key component of setting justice expectations, direct and indirect experiences should also be included in a model of justice reactions (see Bell et al., 2004). Direct experiences may have a greater weight in the development of j ustice expectations than other variables. I posit that direct 76 experiences may be remote or proximal, with remote direct experiences being in the more distant past, and proximal direct experiences being very recent. Educational Integration Remote direct experiences include experience of integration in education. South Africa’s apartheid history included racially segregated schooling which, in concert with the Group Areas Act, allowed individuals of different race groups to grow up with only very limited exposure to people of other races. After the dismantling of apartheid, government schools were opened to all races and this increased the social contact across race groups at an individual level. I posit that this normalizing of relations across the color line meant increased opportunity for social contact, and with that greater appreciation of one another as equals. Allport argued that prejudice may be reduced by changing the way that people conceived of category memberships (see Gaertner & Dovidio, 2005), and I posit that exposure to integrated education may have served to shift focus from race group membership to a broader integrated group (perhaps school identity). Following the logic of Tajfel (1969) as described by Gaetner and Dovidio. when people are categorized into the new grouping, actual differences between members in the grouping may be perceptually minimized, making people seem more similar. In the South African context that would mean people of different races who experience integrated schooling and become categorized as part of the broader collective would seem more similar than before. According to Allport, a reduction in prejudice between groups results when the groups have equal status, common goals, intergroup cooperation. and the support of authorities, law or custom (see Pettigrew & Tropp, 2005). These conditions should theoretically be in place in the school system, supporting a reduction in prejudice. 77 As demonstrated by Pettigrew and Tropp in their meta-analysis, greater intergroup contact is associated with lower levels of prejudice. Therefore I expect that those who experienced integrated schooling may be less supportive of policy used after graduation that maintains decision-making based on race. Thus I anticipate that those who experienced more integrated education will have lower expectations of procedural and distributive justice. Hypothesis 24: Educational integration will be negatively related to procedural justice expectations. Hypothesis 25: Educational integration will be negatively related to distributive justice expectations. Educational integration may buffer the effect of the broader community‘s view. That is, if an individual has experienced integration, their view that people should all be evaluated on equal footing may moderate the effect of the group's culture of support for affirmative action. Waddy (2004) quotes a student at Cape Town High School speaking to the local newspaper, saying that the time has come to end affirmative action as we talk of equality, but equality is not reality when you see people being given jobs based on race . Hypothesis 26: Educational integration will moderate the effect of group culture of support for affirmative action on procedural justice expectations such that those with an integrated education will show a weaker relationship between culture and procedural justice expectations. Hypothesis 27: Educational integration will moderate the effect of group culture of support for aflirmative action on distributive justice expectations such that those with an integrated education will show a weaker relationship between culture and distributive justice expectations. Indirect Experience Subjective norms guide behaviors, and I now address social guidance as an influence on justice expectations. 78 Social Guidance Subjective norms may become established where views or behaviors are encouraged or discouraged by people who are important to an individual (see discussion in Burgess, 2002). This is a process of social guidance, where people of influence offer information, advice or guidance (Vaux, Riedel & Stewart, 1987). These salient individuals may include one’s friends, family and co-workers (Goldman, 2001 ). There has been little systematic research conducted on the social guidance provided by others in the setting of justice expectations (see Goldman, 2001), and the influence of social information on job applicant perceptions has been neglected (see Ryan & Ployhart. 2000). In the present study, the form of social guidance being considered is the respondents’ perception of whether friends, family, or co-workers made suggestions regarding the fairness of the job application process. I expect that such guidance contributes to the building of justice expectations in this context. Goldman found that social guidance was related to legal-claiming in a sample of terminated workers. I view that social guidance as providing support for the individual’s sense of justice violation. and I posit that job applicants’ justice expectations are based in part on such social information. Social information processing theory (Salancik and Pfeffer, 1978) captures the idea that individuals do not operate entirely independently of their social context. Rather, they are adaptive and take into account cues from their social environment. Salancik and Pfeffer also link the ambiguity of the environment to the use of social information. As South Africa’s laws create a selection framework that can be viewed as good for the whole, but possibly bad for individuals, applicants may depend on additional 79 social information when generating justice expectations in this context. I expect that social guidance influences individual expectations of procedural and distributive justice. Hypothesis 28: Social guidance — unfair - will be negatively related to procedural justice expectations. Hypothesis 29: Social guidance — unfair - will be negatively related to distributive justice expectations. Proximal Direct Experience Current Rejection/A cceptance Interactions between decision outcomes and justice have been shown in extant research (see Ryan & Ployhart, 2000). For example, in a study by Gilliland (I994). perceived fairness on both procedural and distributive dimensions was greater among those who were selected than those who were rejected. In a study of job applicants by Bauer, Maertz, Dolen, and Campion (1998), outcome favorability and procedural justice perceptions were related to attitudes and test fairness perceptions, with procedural justice explaining incremental variance after outcome favorability was controlled. Justice expectations would be modified in reaction to an event, but as a strong source of information, an event would also have a direct effect on outcomes. Heilman and Alcott (2001) demonstrated that study participants who were aware that others viewed them as selected on the basis on gender (rather than merit) had negative self-regard when uncertain about their skill level, and experienced negative affect. Heilman, Simon and Repper (1987) showed that women chosen under preferential selection had lower self- perceptions and self-evaluations. While these studies are not directly relevant for the current project, as I am dealing with people who have probably not yet begun their formal job search, they are important for understanding that a control measure may be necessary. and for the longer term perspective, as I plan to follow up with the respondents as a 80 longitudinal study. Thus, I capture whether or not respondents have experienced a recent event — current rejection or acceptance by a hiring organization. Outcomes This study seeks to link different dimensions of justice to the important outcomes in the context of strong preferential selection. These include withdrawal intention, emigration intention, entrepreneurial intention, and education intention. Individuals’ expectations of justice can directly influence how they react to situations. When people are faced with unfamiliar situations they make predictions about what will happen and these expectations can be directly linked to attitudinal and behavioral outcomes. As noted by Olson, Roese and Zanna (1996, p. 211), “The concept of “expectancy” forms the basis for virtually all behavior. Expectancies can be defined as beliefs about a future state of affairs. As such. expectancies represent the mechanism through which past experiences and knowledge are used to predict the future. Every deliberate action we take rests on assumptions (expectancies) about how the world will operate/react in response to our action.” In the selection context, a high-stakes circumstance involving uncertainty. higher expectations regarding justice have been linked to greater test-taking efficacy and motivation, intention to accept the job, and less psychological withdrawal and negative affect (Bell, Wiechmann, & Ryan, 2006). The expectations may provide a mental map of how the world works. According to Shapiro and Kirkman (1999), the expectation of justice (they use the term anticipatory injustice) is used as an information source. They demonstrate that those who held higher expectations of distributive justice showed higher organizational commitment and lower turnover intentions and resistance to change. Thus. individuals do appear to think and behave in a way that is consistent with their expectations of justice, even if they have not yet experienced an event that is evaluated as 81 just or unjust. According to Olson, Roese, and Zanna (1996, p. 220), “one general behavioral consequence of expectancies is that individuals will act in ways that are consistent with the particular content of the expectancies.” In South Africa’s selection context, outcomes of importance are withdrawal from the selection process, emigration intention, entrepreneurial intention, education intention. and lowered core self-evaluation. Such outcomes need to be understood as organizations are making difficult decisions and risk alienating components of the workforce if decision-making is not managed in a way that is viewed as just across the broad continuum of applicants. Further, according to Anderson (2004), companies in several countries are able to be sued by job incumbents who have been harmed by negligent human resource processes. South African labour law already views applicants as employees in terms of their rights to due process from the organization. Applicants are able to seek recourse through the Commission for Conciliation Mediation and Arbitration (CCMA). Further, I believe that industrial/organizational psychologists have an ethical responsibility to evaluate whether selection processes under their purview are having deleterious effects. Withdrawal Intention As discussed earlier, South Africa appears to be facing a sense of resignation and a withdrawal from the applicant pool among some potential job applicants (Fraser- Moleketi, 2007; Ramsay, 2005). Withdrawal from the application process would limit the opportunity of organizations to select from a broad pool of the best candidates, and withdrawal may be linked to low justice expectations. As noted by Gilliland ( l 993). we should expect that decisions to apply for a job will be influenced by distributive and 82 procedural fairness. In US-based research, applicants who have lower expectations of procedural and distributive justice have been shown to be more likely to withdraw psychologically from the selection process, and were also less likely to intend to accept the job or recommend the job to others (Bell, Wiechmann, & Ryan, 2006). I posit that those who have low expectations of justice in South Africa may react similarly and withdraw from the selection process, not bothering to apply for jobs. While a strong relationship between fairness perceptions and self-selection out of the hiring process has not been demonstrated in American research (see Ryan & Ployhart, 2000), the US job market is very different from that in South Afiica, and I anticipate that such links nevertheless exist. For example, Ryan, Sacco, McFarland and Kriska (2000) study the reactions of police applicants in a US sample. They found that there was little relationship between fairness and actual withdrawal (self-selecting out of the selection process). However, in examining the mean levels of fairness across groups in their study. which was captured on a five-point scale, it appears that on the whole, the process was regarded as fair. Mean levels of fairness across groups (selection status, race, sex) ranged from 3.70 (those who failed at Time 1) to 3.93 (those who failed at Time 2), and the largest standard deviation was .65 (those who failed at Time 2). I posit that the context for US applicants is very different from that faced by South African applicants. First, there may be far greater variability in perceptions of fairness in the South African selection context (see Madi, 1997). As noted by Truxillo, Steiner and Gilliland (2004,). in extant research on applicant reactions, few applicants indicate that they have been treated unfairly, and this may be a function of the lack of extreme unfairness in US field sample research. US applicants are not faced with as extreme affirmative action protocols. and 83 are probably more likely to be faced with selection processes with demonstrated validity based on the Uniform Guidelines on Employee Selection Procedures (1978). Second, US applicants are not competing in a job market where the narrow unemployment rate is 23% (Statistics South Africa, 2008). This narrow unemployment rate in South Africa is calculated based on actual job-seekers and does not include “discouraged work-seekers” who are persons “who want to work and are available to work but who say that they are not actively looking for work.” There are almost as many discouraged work-seekers as there are unemployed people, so a broad definition of unemployment that includes both groups would be dramatically higher. These characteristics may leave the South African job applicant with far less hope of success, increasing the likelihood that they will withdraw if the have concerns about the process. Interviewees in Ramsay (2005) reported that there are individuals who don‘t bother to apply for jobs, or don’t try to do well in the application process, because they don’t expect to be considered due to their race or gender giving them non-target status. This issue has also been raised by Herholdt and Marx (1999; p. 69-70). who state that among those who are “peripheral to” affirmative action (that is, non-target group members), there is growing tendency “to stop competing for higher positions as they believe these positions will in any event be filled by affirmative action beneficiaries.“ Even high school students have reported that they perceive race and gender to be barriers to career development for non-target group members in South Africa (Stead & Els, 2004). Schmit and Ryan (1997), through a qualitative examination of a sample of applicants for police positions in the US, observed that some of those who withdrew indicated that a 84 reason for withdrawal was their being upset about perceived hiring practices (e.g., minority preferences). This reaction was not limited to non-target group members — some target group members viewed preferential treatment for target groups as upsetting. Ryan. Horvath and Kriska (2005), in a study of potential job applicants for firefighter positions in the US, observed that members of under-represented demographic groups were more likely to withdraw from the selection process than were members of majority groups. although their study does not fully explain why. In that study, the organization had previously been required to hire more African-Americans and women, but at the time of the study was hiring members of those two groups at rates lower than their presence in the applicant population. Konrad and Linnehan (1995), in a study of managers, found that attitudes to identity-conscious human resource processes were more positive among white women and people of color than among white men, consistent with a self-interest rationale. They also found that all three demographic groups held positive attitudes regarding identity- blind interventions. Konrad and Linnehan argue that women and people of color support identity-blind processes, as these processes do not stigmatize them as incompetent. l posit that members of target groups in South Africa recognize that a redistribution based on demographics is necessary, but that they would still want to be selected based on merit within the target group, and that the view that procedures are not fair within target group (low procedural restorativejustice and distributive restorative justice) would trigger negative reaction. 85 Higher expectancies of self-efficacy in a task have been linked to higher levels of intrinsic motivation in the task (see Bandura, 1982), and perceived behavioral control in a situation influences intention to perform (see Aj zen, 1985). Those who are convinced that they will not be considered will withdraw. Specifically, I anticipate that those who have low expectations on procedural and distributive justice dimensions will be inclined to withdraw from the selection process. Hypothesis 30: Procedural justice expectations will be negatively related to withdrawal intention. Hypothesis 31: Distributive justice expectations will be negatively related to withdrawal intention. Parker, Baltes and Christiansen (1997) showed that relationships between perceived support for affirmative action or equal opportunity, justice perceptions and career development opportunities were moderated by group membership. Withdrawal intention is a rational response in light of the belief that there will not be career opportunities. As noted by Nowakowski and Conlon (2005), individual differences including gender have been shown to moderate the relationship between justice perceptions and outcomes. I posit that those who anticipate that there will be a distribution to members of target groups, and who are not target group members. will intend to withdraw. Hypothesis 32: Restorative justice expectations will be positively related to withdrawal intention for non-target group members, and negatively related to withdrawal intention for target group members. However, as noted above in relation to Konrad and Linnehan’s (1995) argument that target group members prefer identity-blind procedures because those do not 86 stigmatize them as incompetent, it is not only restorative justice expectations that are relevant. We need to drill down more specifically and consider the possible effects of procedural restorative and distributive restorative justice. While someone may have high restorative justice expectations, and be a member of a target group, that may not necessarily drive continued commitment to apply for jobs. As identified by interviews in Schmitt and Ryan (1997), there may be members of target groups who view the process of giving preference based on race to be unfair, prompting their withdrawal intention. If women view the procedural and distributive restorative justice of the application process for women as low, they may withdraw, even though they are target group members. For example, black African women who feel that they are going to be passed over in favor of white women because of ongoing discrimination against blacks may be inclined to withdraw. Such process would be viewed as unethical and biased against certain groups of women. Similarly, black Afiican women who feel that they are going to be passed over in favor of those who have political connections even through they have better skills and more experience than other women may choose to withdraw from the process. Hypothesis 33: Procedural restorative and distributive restorative justice expectations will be negatively related to withdrawal intention. Emigration Intention Emigration from South Africa is continuing (Bailey, 2003; Crush, Peberdy. & Williams, 2006; “Fears over South Africa’s exodus”, 2006; Statistics South Africa. 2005), and emigration may be in part a result of frustration with current select practices. One study of skilled South Africans showed that both blacks and whites had given 87 emigration some thought, and that twenty percent of the respondents reported that it was highly likely that they would emigrate (Waller, 2006). Opposition to affirmative action was overwhelmingly the case for whites in this study; opposition to the policy was true for 20% of the blacks (Waller, 2006). Over the last twenty years there has been a “white flight” from South Africa, but this loss of expertise may be exacerbated by the Employment Equity Act. There are reports of dramatic changes in the population: from 1995 to 2005, there was a 17% population increase in black Africans, 15% increase in coloured people, 9% increase in Asians, and a 16% fall in whites (“Fears over South Africa’s exodus”, 2006). While many factors have contributed to this mass loss of expertise from South Africa, the affirmative action requirements of the Employment Equity Act leave many whites feeling frustrated and hopeless (see National Research Foundation, n. d.). Many expatriate South Africans indicate a feeling of frustration with the perceived injustice of a system that discriminates against them based on race and gender. For example, consider this excerpt of a letter from a mechanical engineer to the editor of Focus magazine (University of Natal, 1999, p. 4): “As part of focusing on its people capital the country needs to drop the race card and with it affirmative discrimination. How much chance would I have ifI were now to apply for a position of head of an Engineering School at a technikon or university in South Africa? I am both male and white, and that puts me at the bottom of the heap. Yet New Zealand is happy to have me as head of Engineering at Christchurch Polytechnic, simply for the skills that I bring. South Africa is destroying its people capital, and with them its prospects for the future.” A tone of self-righteousness hostility pervades many of the published communications from South African e’migrés. Folger and Cropanzano (1998, p. l 9) note anger as socially targeted resentment, a response to a social injustice that “arouses the urge to punish those who do not adhere sufficiently to the accepted codes of conduct 88 governing self-interest.” The belief that selection process is being managed in a fashion that is counter to these codes may be driving emigration. Complete withdrawal from the South Afiican economy through emigration is problematic — South Africa needs educated, skilled citizens to remain in the country. It appears that the flight is primarily white, thus I posit that non-target group members who expect the selection process to be unfair on procedural and distributive dimensions will be more inclined to emigrate. Hypothesis 34: Procedural justice expectations will be negatively related to emigration for non-target group members. Hypothesis 35: Distributive justice expectations will be negatively related to emigration for non-target group members. Hypothesis 36: Restorative justice expectations will be positively related to emigration for non-target group members. However, it is apparent that target group members are also giving thought to emigration (Waller, 2006). This may be a reflection of strong merit principles, combined with the belief that there are problems with within-group decisions. That is, that jobs for target groups are perhaps being assigned based on biases and connections rather than being distributed in a consistent way based on skills and contribution (see Madi, 1997). Thus, I posit that a preference for merit principle combined with low procedural restorative and distributive restorative expectations for one’s own group may precipitate emigration intention. Hypothesis 3 7: Preference for merit principle combined with low procedural restorative and low distributive restorative justice expectations for own in-group will be positively related to emigration intention for target group members. Entrepreneurial Intention One positive outcome for the country has been that many of those who feel they will not be hired under current affirmative action policies are contributing their skills via other avenues; starting their own businesses (Ramsay, 2005). As noted by Olson. Roese 89 and Zanna (1996), expectancies form the basis for action because individuals‘ expectations serve as a mental map for what will happen, and thus guide behavioral choices. Those who anticipate that they will not be hired under the current political dispensation may be more likely to pursue starting their own business. This taking action may be explained by Ajzen and Fishbein (1980) with their theory of reasoned action. where entrepreneurial activity is likely recognized as a practical strategy by the individual who is not a target group member, where preferences based on demographics are in place, and the subjective norm may be very supportive of the behavior. Hypothesis 38: Restorative justice expectations will be positively related to entrepreneurial intention for non-target group members. Education Intention Some individuals believe that under current labour law and economic conditions they must get the highest level of qualifications available to increase the likelihood of their finding a job. White males who historically would have found it easy to get a job with limited education due to government protections (Ross, 1999; Terreblanche, 2002) are now being pushed to find alternatives and some choose to improve their education as one strategy (Ramsay, 2005). Choosing to return to school to improve one’s qualifications is a positive outcome for South Africa as a whole as it increases the level of skills available in the economy. This is particularly important when large numbers of skilled workers have emigrated. For all groups, discovering that one is not as easily hired as one might have expected may prompt individuals to return to school. Hypothesis 39: Restorative justice expectations will be positively related to education intention in non-target groups. 90 Summary In summary, my model is thus: Personal characteristics such as socio-economic status influence an individual’s justice expectations. Also contributing to the establishment of expectations are existing beliefs and values (support for affirmative action and preference for merit principle, group culture of support for affirmative action. group culture of white guilt, and group expectation of restorative justice). Further variance in expectations is explained by remote direct experiences (educational integration) and indirect experiences (social guidance). The relationship between these antecedents and justice expectations (general restorative, procedural, distributive. procedural restorative for women and blacks, and distributive restorative for women and blacks) may be moderated by race and gender group membership. These justice expectations reflect what the applicant anticipates in their interaction with the prospective employer. These expectations are directly related to the outcomes of interest; withdrawal intention, emigration intention, entrepreneurial intention, and education intention. 91 METHOD Sample and Study Design My goal was to capture the attitudes of educated South Africans who are potential job applicants. South Africans who are in their final year of a university degree program. who have some interest in a job after graduation, and who have a reasonable possibility of applying for a job comprised the population of interest for this study. A total of 327 students who were in their final year of study at one of two South African universities -- the University of Cape Town and the University of KwaZqu-Natal — took part in the study. This study has a quasi-experimental design that takes advantage of an existing contrast between race and gender groups; these groups have specific target status under South Africa’s labour law. Thus, my sample is a purposive, stratified sample capturing black African men, black African women, coloured men, coloured women, Indian men. Indian women, white men, and white women. This purposive sampling across race and gender was to facilitate appropriate responses for cross-group analysis. Although not a formally random sample, my aim was to capture the breadth of responses on presumptively important dimensions (see Shadish, Cook, & Campbell, 2002, p. 23). Such sampling is critical as these groups had differing levels of advantage under the Nationalist Party political dispensation (Ross, 1999; Terreblanche, 2002), and race and gender are affirmative action target status identifiers (Government Gazette, 1998). More specifically, the sample of 327 participants was comprised of 165 men and 162 women, and most participants (96.9%) were from South Africa’s four major race groups (Black African, Coloured, Indian and White), while there was a small group of 92 people who were of other races. These included respondents who indicated that they were Asian/Chinese, of mixed parentage (biracial), Hispanic, or did not specify. Attempts were made to balance the number of cases in each of the cells of the design, but perfect balance was not achieved due to the characteristics of the student population and insufficient effective targeted recruiting for the coloured group. Table 5 shows the race and gender of participants. Table 5 Race and Gender of Participants Frequency Percent Black Male 50 49.0 African Female 52 51.0 Total 102 100.0 Coloured Male 14 37.8 Female 23 62.2 Total 37 100.0 Indian Male 46 58.2 Female 33 41.8 Total 79 100.0 White Male 50 50.5 Female 49 49.5 Total 99 100.0 Other Male 5 50.0 Female 5 50.0 Total 10 100.0 Table 6 shows the headcount enrollment of on-campus students, by race and gender, in 2006, at the institutions of interest. Although for this study I did not manage to attract as many participants from the coloured group as would have been desirable. the proportion of coloured participants (11%) was higher than the proportion of coloured students enrolled (7%). 93 Table 6 Enrollment of Students by Race and Gender Black Coloured Indian/ White Total' Women Men African Asian UCT 5,971 2,91 1 1,733 9,493 21,224 10,736 10,488 count _ UCT 28.1 13.7 8.2 44.7 100 50.6 49.4 % UKZN 13,995 868 12,255 5,464 32,621 17,886 14,735 count . UKZN 78.2 2.7 37.6 16.7 100.0 54.8 45.2 % UCT + 20,072 3,795 14,034 15,018 54,045 28,727 25,318 UKZN count _‘ UCT + 37.1 7.0 26.0 27.8 100.0 53.2 46.8 UKZN % lThe totals in the table equal total male plus total female. Because students whose population group was unknown are not included here, the counts for black African, coloured, Indian/Asian and white students may not sum to the total. Of the 312 participants who indicated their year of birth, all were born between 1975 and 1991, with more than 75% of these respondents being born between 1986 and 1989. As the first non-racial democratic election in South Africa was held in 1994. the majority of these participants would have been between five and eight years old at the end of the apartheid era, and would therefore have attended high school during the African National Congress’ administration. With 1988 as the modal year of birth. the modal age was about 20 years old. The most common home language amongst the respondents was English (62.6%). followed by Zulu (19.6%). Sixteen of the respondents indicated speaking more than one 94 language at home. Only one participant did not indicate their home language. Table 7 shows the home languages spoken by the respondents. Table 7 Home Language of Participants Language ‘Frequency Percent Afrikaans 14 4.3 English 204 62.6 isiNdebele 3 0.9 isiXhosa 14 4.3 isiZulu 64 1 9.6 Sepedi 3 0.9 Sesotho 4 1.2 Setswana 2 0.6 siSwati 5 1.5 Tshivenda I 0.3 Xitsonga 28l 8.6 Other W , 14 W 4.3 Teaasaaa:5;: 326‘ We T‘Sixteen respondents indicated speaking mt; than one language at home. While most of the participants indicated only one major area of study. some indicated two majors. Almost half of the respondents were studying Humanities while almost 40 percent were studying science, engineering or technology. Table 8 shows the major areas of study of the participants. Table 8 Major Areas of Study WField . W lFWrequency jPercent SEER} engineering and technology j 127‘ 39.2 Business and management 1 511 15.7 Education 1 4| 1 .2 HtrmanitiesW W ,_ W _W , , W 'WW -155, 47.8 Total respondents l , 324 , 1040* fThirteen respondents indicated mbre thanbfiei major field offstudy. 95 Respondents indicated the racial composition of their class in their final year of high School. The percentages accounted for by each race were used to calculate a heterogeneity index. The mean of the heterogeneity index was 0.42, and. the standard deviation was 0.25. A high heterogeneity index would be 0.80 (e.g., from a school with equal proportions of black African, coloured, Indian, white, and other race group students). A low heterogeneity index would be 0 (e.g., from a school with all white students). On average, schools were moderately heterogeneous. Procedure The opportunity to participate in the study was advertised via flyers, announcements in class, and e-mail to students at two universities. Interested participants were asked to sign up for a classroom-based data collection session in advance via a Web page. Data collection took place in classrooms on the university campus, where an English-language paper-based survey was administered to the participants. Participants were given 3 R100 Woolworths gift voucher (about $12) for their participation. This amount was sufficient to provide a small incentive to spend 30-45 minutes participating in the study, yet not large enough that the applicant would feel unable to turn down the opportunity. (Woolworths in South Africa is a high-end chain; this is not the same chain as that which existed in the US.) Although English is not the home language of all respondents, English is the lingua franca of business in South Afi'ica, and the language of instruction at the campuses from which the samples are drawn; final year university students would all be conversant in English. 96 Data Cleaning and Handling of Missing Data Where respondents had indicated two different choices for an item, the data point for the item was handled as missing. Where one respondent had modified the scale to reflect higher scores (e.g., 10, as in very, very, very strongly agree), the closest point on the existing scale was used (i.e., 5). Where one respondent had skipped Item 1 10 (male or female), the missing response was replaced by logical deduction as they had indicated later in the survey (Item 126) that they were female. Two respondents whose race could not be identified from the survey form due to missing data (Item 11 1) were coded as being of “Other” race. For months looking for a job (Item 41), where a fraction of a month was reported by the respondent, these data were rounded up to the next whole number of months. Items 1 through 38, 47 through 54, 59 through 108, 113 through 123, and 125 were items that are used to make up scales or are categorical pieces of information. These items showed very few missing data points. The frequency of missing data for these items ranged from one missing data point to five missing data points out of a possible 327 data points. Of these items, 38 had only one missing data point, 18 items had two missing data points, five items had three missing data points, and one item had four missing data points. As the proportion of missing data is very small (less than .3 percent) I do not anticipate data replacement affecting interpretation of my results. For these scale items. missing data were replaced with the mean value for the item within race and gender group. For categorical variables, the modal value for the item within race and gender group was used to replace the missing data. 97 For Item 124, addressing the racial composition of class in high school, five respondents skipped the item altogether, and one respondent indicated a proportion for one group only, effectively skipping the item. As responses to this item are used to calculate Blau’s Index of Heterogeneity (school integration in this study), these missing data points for school integration were replaced with the mean value of school integration within race and gender group. Thus, this replacement was needed for 1.8% of the respondents. Thirteen of the respondents who did provide responses to Item 124 did so with errors in their calculations. Their percentages of different groups, which should sum to 100%, summed to 90% (four respondents), 95% (one respondent), 98% (one respondent), 105% (one respondent), or 110% (six respondents). For these respondents, the fraction making up each group was recalculated, converting the value given as a proportion of the respondent’s sum total, to a value that is a proportion of 100 as the total. These corrected fraction values were then used to calculate the relevant school integration values. Items 55 through 58 address co-workers in current or previous jobs. There were 28 missing data points for each of these four items, and no data replacement was conducted as these may be a legitimate non-response if the respondent does not have co- workers. Power The total sample (N = 327) used for the purposes of demonstrating the justice factors was of a sufficient sample size to achieve a stable factor solution for justice expectations using confirmatory factor analysis (see Kline, 2005; MacCallum, Widaman. Zhang, and Hong 1999). This sample containing women (n = 162) and men (n = 165.) 98 also allowed me sufficient power to test for measurement invariance of measures across gender. Testing for measurement invariance of measures across race was conducted for the three race groups with reasonable sample sizes (black African n = 102, Indian n = 79. and white n = 99) although there is only sufficient power with these sample sizes to detect large effects (see Meade, Johnson, & Braddy, 2008). For subsequent hypothesis testing, the complete data set of 327 was then trimmed to include only those cases who were members of one of the four major race groups (black African, coloured, Indian, white), dropping those of other races, resulting in a final N = 317 for further hypothesis testing. This final data set was comprised of eight race/gender groups, with group size ranging from n = 14 in the case of coloured men to n = 52 in the case of black African men. The average group size was 40. For purposes of multilevel modeling, I used the power analysis software Optimal Design Version 1.76 to consider approximate power where power for these analyses depends upon the number of people in the groups (n), the number of groups (.1). the intra- class correlation (p), and the effect size (0) (see Spybrook, Raudenbusch, Liu, Congdon & Martinez, 2008). I have eight race/gender groups with between 14 and 52 individuals within each group. With an average of 40 people per group, with alpha at .05. and p of .03, there would be sufficient power (close to .80) to reject the null hypothesis in the case of strong effects. If I were to raise the alpha level to .10, which is reasonable considering that this research tests new concepts that have not been demonstrated elsewhere and as the consequences of Type I error are not dire here (see Hox, 2002), this would raise the power for demonstrating group effects such that there would be a power of .80 in the case of strong effects and a p of about .07. 99 Confidentiality Each participant was provided with a unique identification number. That number was marked on his/her survey form. Respondents returned the completed survey and a separate informed consent form along with contact details if they were willing to be contacted for longitudinal follow-up. To maintain confidentiality, names and contact information of participants are being stored separately from the returned survey forms and subsequent data file. Requests to participate in the second wave survey as part of the longitudinal study that will follow from the current proposed study will be marked with the respondent’s unique identification number. This will allow the matching of Time 1 (dissertation data) and Time 2 (follow-up non-dissertation project data) in the subsequent research study while keeping names and response data separate. Data collected for the study are kept confidential to the maximum extent allowable by law. Results are presented in an aggregated fashion so that individual respondents cannot be identified through the results. Ethics Research has been conducted in accordance with the principles and ethical standards of the United States American Psychological Association, the Psychological Society of South Africa, and the research procedure was designed to protect the participants from harm, and maintain their confidentiality (Aguinis & Henle, 2002). Ethical principles in research in the United States and South Africa are largely shared (Leach & Harbin, 1997). There are ethical problems with implementing research in developing countries where the participants do not fully understand the risks associated with their involvement in research. In this case, I worked with a sample of individuals 100 who are educated and urbanized. As final-year students in major universities using English as a language of instruction, they would have been able to read sufficiently well to understand the letter of invitation to participate and be well informed as to the nature of this study. Also, the money offered to participants in appreciation for their involvement was set at a level that does not create an undue draw to participate where they have limited economic resources. As noted by Ndebele (2005), western researchers have an ethical responsibility to share knowledge and generate capacity building in the regions where they conduct their research. Results of this study will be made available to individual participants who indicated an interest in receiving an electronic summary of results and will be posted on a publicly-accessible website. In multicultural research it is particularly useful to ask for feedback from collaborators within the culture to identify any problem areas as research progresses. The questionnaire was developed tapping current published research in South Africa, as well as knowledge gained from my own interviews conducted on this topic (Ramsay, 2005). Further, I sought institutional review board review and gained approval at both South African campuses. The survey form was also pilot tested with South African students. Measures The following measures were included in the questionnaire. The items shown here for each measure comprise the complete scale used for this study. First, I address the group-level factors, and second, the individual-level factors. Following are the group- level factors and their specific items. 101 Group Beliefs Group belief in white guilt. Group belief in white guilt is the race/ gender group’s belief that whites are guilty of having received benefits because of their race and should feel guilty about such benefits. Items here have been written to capture the group’s belief in a broad sense of guilt and responsibility on the part of whites in South Africa. Questions are based in part on the concepts captured by items for white guilt in America by Swim and Miller (1999), who reported alpha levels of .74 to .86 in their studies using a five-item scale. However, their focal point was the white individual’s own guilt, and the context addressed in some of their questions was America, "which was not appropriate for the present study. Therefore, all of the questions used here have been written specifically for this study. As I am working in the context of South Africa, and attempting to capture a broader shared belief in white guilt, I created items that captured views from respondents of all races regarding their beliefs about the guilt whites should feel. Items were worded to Responses were captured on a 5-point Likert-type scale ranging from I (strongly disagree) to 5 (strongly agree). Following are some questions regarding the beliefs and values of people of your race and gender. For example, if you are an Indian man, respond about the beliefs and values that you think Indian men have. People of my race and gender believe that whites should feel guilty that they have enjoyed benefits and privileges at the expense of non-whites. People of my race and gender believe that whites are guilty of having received benefits and privileges simply because of their race. People of my race and gender believe that whites should feel guilty about the past behavior of their race. 102 Group Culture Group culture of support for aflirmative action. Group culture of support for affirmative action is the race/gender group’s shared attitude regarding affirmative action. and was captured using a five-item scale. Items for the group-level predictor were similar to those for the individual-level construct but have group-level referents. The first three items are those from the six-item attitude scale of Kravitz and Platania (1993) that are appropriate in the South African context, with the third of these being modified for use here. Other items from Kravitz and Platania’s (1993) scale were not used as they were not appropriate in the South Africa context where organizations have greater pressure to have affirmative action programs. The final two items used in the present study were taken from the attitude scale by Parra (1991) who generated a long scale of 50 diverse items and who demonstrated alpha levels of .96 and .97 in different samples. Kravitz and P latania ( 1 993) demonstrated an alpha level of .86 for their six-item scale. All items were tailored to be appropriate for a group referent. Responses were captured on a 5-point L. i kert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). F 01 lowing are some questions regarding the beliefs and values of people of your race and gender. For example, if you are an Indian man, respond about the beliefs and values that you think Indian men have. l)e()ple of my race and gender believe that affirmative action is a good policy. All in all, people of my race and gender oppose affirmative action plans in industry. (reverse coded) People of my race and gender believe that the goals of affirmative action are 800d. PeOple of my race and gender believe that affirmative action does more harm than good. (reverse coded) Pe()ple of my race and gender believe that affirmative action is beneficial to Society. 103 Group Justice Expectations Group restorative justice expectations. Group restorative justice expectations is the race/gender group’s shared fairness expectation regarding the distribution or allocation to target groups that have suffered inequality in the past. Items here were written specifically for this study. The general framing as justice questions was designed to be similar to other justice items that are used in the present study to capture individual- level justice expectations (addressed below). However, although this general framing of items is similar to that used by Colquitt (2001) for other justice dimensions (procedural. distributive, interpersonal, informational) and captures some similar issues to eight items tapping attitudes to compensatory policy in an Affirmative Action Attitude scale (coefficient alpha of .83) (Swim and Miller, 1999), the items I use here are written specifically for this study and are distinctly different in content from published material. They were designed to capture the restoration of social balance by giving preference to those from previously disadvantaged groups in the South African context. Items have been framed as expectations of justice (see Bell, Wiechmann, & Ryan, 2006) but with group-level referents. Apart from the referents, items for the group-level predictor were similar to those for the individual-level construct. Responses were captured on a 5-point Likert-type scale ranging fi'om 1 (strongly disagree) to 5 (strongly agree). Thinking about the jobs for which you and your peers will be applying. the following items refer to the beliefs of people of your race and gender regarding company/organisation consideration of job applicants. For example. if you are an Indian man, respond about the expectations that you think Indian men have. People of my race and gender expect that special consideration in hiring will be provided to those from previously disadvantaged groups. People of my race and gender expect that an attempt will be made to correct past social injustices in hiring. People of my race and gender expect that jobs will go to people from affirmative action’s target beneficiary groups. 104 People of my race and gender expect that preference in hiring will be given to women, blacks, or the disabled. I will now address the individual-level measures. Individual Differences Individual differences in racial identification (specifically centrality) and age were captured. Racial identification (centrality). Race centrality is the extent to which being of a particular race group is central to the respondents’ definition of themselves. Items were based on four items selected from the Centrality Scale of the Revised Multidimensional Inventory of Black Identity (Sellers, Rowley, Chavous, Shelton, and Smith, 1997). Their ten-item scale showed reliability coefficients from .75 to .78 in different samples. A selection of four items that appeared to capture the breadth of the construct were selected as the length of the final survey form was a concern. Items were rewritten to be race- neutral to be appropriate for respondents from all of South Africa’s race groups. Responses were captured on a 7-point Likert-type scale ranging from 1 (strongly disagree) to 7 (strongly agree). Race centrality Overall, being of my race has very little to do with how I feel about myself. (reverse scored) In general, being of my race is an important part of my self-image. I have a strong sense of belonging to people of my race. Being of my race is an important reflection of who I am. Afiican centrality. An additional individual difference variable that was captured is African centrality. As there is a strong sense of being African, regardless of race. all participants were asked to answer the following questions about their African centrality: the extent to which being African is central to the respondents’ definition of themselves. 105 These questions parallel those used for Race Centrality as described above. but the referent here is African Centrality. African centrality ' Overall, being African has very little to do with how I feel about myself. (reverse scored) In general, being African is an important part of my self-image. I have a strong sense of belonging to Afiican people. Being African is an important reflection of who I am. Fear of crime. Fear of crime is the extent to which respondents worry about being victims of crime. Fear of crime was captured for use as a control variable in predicting emigration intention. The three items used here are based on those in a four-item scale (no alpha level reported) used by the British Home Office for community research that included worry about crime (Nicholas, Povey, Walker, & Kershaw, 2005). The selected items covered crimes that could reasonably affect most people in South Africa. That is. an item regarding having your car stolen was dropped from the scale and an item addressing having items stolen from your car was modified to be appropriate for those who do not own cars. Responses are captured on a 5-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). I am worried about mugging, rape, or physical attack by a stranger. I am worried about having my home broken into. I am worried about having something of mine stolen. Potential emigration factors. Additional control variables were based on three selected items from the 24-item Intention to Emigrate scale of Miller, Haskell. and Thatcher (2002) who report a coefficient alpha of .95 for their scale. These control items were chosen that captured push and pull factors that have been linked to emigration. Responses were captured on a 5-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). 106 I have friends who have settled abroad. I am dissatisfied with conditions within South Africa. I believe that my quality of life would be better overseas. Existing Beliefs/ Values Measures of existing beliefs and values include support for affirmative action and preference for merit principle. Support for aflirmative action. Support for affirmative action is the extent to which respondent’s hold attitudes of support for affirmative action policy, and was captured using a five-item scale. Items were written with individual referent to parallel the group-level referent items discussed earlier. As with the group-level items, the first three items were those from the attitude scale of Kravitz and Platania (1993) that are appropriate in the South African context, with the third of those being modified for use here. The final two items used in the present study were taken from the attitude scale by Parra (1991) who generated a long scale of 50 diverse items and who demonstrated alpha levels of .96 and .97 in different samples. Kravitz and Platania (1993) demonstrated an alpha level of .86 for their six-item scale. Responses are captured on a 5-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). I believe that affirmative action is a good policy. All in all, I oppose affirmative action plans in industry. (reverse coded) I believe that the goals of affirmative action are good. I believe that affirmative action does more harm than good. (reverse coded) I believe that affirmative action is beneficial to society. Preference for merit principle. Preference for merit principle is respondents’ preference for allocating outcomes on the basis of the distributive justice principle of merit, and was captured using a selection of four items from the 15-item Preference for Merit Principle Scale of Davey, Bobocel, Hing, and Zanna (1999) which covers a variety 107 of contexts. Only four items were used as the length of the final survey form was a concern. Items were chosen to capture the breadth of the preference for merit principle construct. Their coefficient alpha for the scale was .60 to .70 in different data collections. The second item was modified slightly. Responses were captured on a 5-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). In organisations, people who do their job well ought to rise to the top. It is wrong for an employer to give a job to someone they know without first advertising the job to other candidates. The effort a worker puts into a job ought to be reflected in the size of the raise he or she receives. Promotion decisions ought to take into account the effort workers put into their job. Remote Direct Experience Educational integration: Educational integration is the extent to which the high school that the respondent attended was racially integrated, and was captured through a single item created for this study. What was the racial composition of your class in your final year of high school? Please indicate the percentage accounted for by each race. Black African % Coloured % Indian % White 0/0 Other % ++++ TOTAL _100 % To generate an integration index for each individual’s school experience. Blau‘s (1977, p. 78) index of heterogeneity was used: heterogeneity = l — 2 [7,2 where p, = the fraction of the school making up each group. 108 As a potential control variable, respondents were also asked whether their high school was single-sex or co-educational. Indirect Experience Social guidance: Social guidance is the respondents’ perception of whether friends, family, or co-workers made suggestions regarding the fairness of the job application process. A social guidance scale regarding job application fairness was created for this study. These items are not framed with reference to a particular organization —- they are general. Responses were captured on a 5-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). Following the suggestion of Goldman (2001) the scale captured the views of friends, family, and coworkers. The focus here was social guidance regarding unfairness. My close friends have said that the job application process is unfair. My close friends have suggested that job applications will be unfairly handled. My close family members have said that the job application process is unfair. 11:2; :llgse family members have suggested that job applications will be unfairly My coworkers have said that the job application process is unfair. My coworkers have suggested that job applications will be unfairly handled. 109 Proximal Direct Experience Current rejection experience: I captured proximal direct experience in selection as a potential control variable through including current rejection experience. Current rejection experience is having been recently turned down for a job. The respondents were asked to indicate the number of times they have been turned down. If you have begun looking, how many times in the past 6 months have you been turned down by a company/organization (e. g., by letter, e-mail, telephone, or in person)? Individual Justice Expectations Justice expectations are the respondents’ expectations regarding the fairness of the application process, and were measured using items from Colquitt’s (200l) scales on his four dimensions of justice perceptions (procedural, distributive, interpersonal. informational). New items on the restorative justice dimension were created for this study. Similarly operationalized by Bell, Wiechmann, and Ryan (2006), justice expectations items were framed to capture what the individual is expecting in terms of justice, rather than their reaction to an event. Procedural justice expectations. Procedural justice expectations capture the fairness of procedures used to determine outcome distributions or allocation (see Colquitt, 2001). Items were based on selected justice perception items of Leventhal (1980) combined as a procedural justice scale as proposed by Colquitt (2001). Items were reframed as expectations of justice (see Bell, Wiechmann, & Ryan, 2006) and the focal point was changed from the self to applicants in general. Colquitt (2001) demonstrated reliability for the seven-item procedural justice scale of .78 to .93 in different samples. In an effort to keep the final survey form a reasonable length,,four items that captured the 110 breadth of the construct and that could reasonably be reframed as expectations were selected. Responses were captured on a 5-point Likert-type scale ranging from I (strongly disagree) to 5 (strongly agree). Thinking about the jobs for which you and your peers will be applying, the following items refer to company/organisation procedures used to make hiring decisions. To what extent do you agree: I expect that those procedures will be applied consistently across applicants. I expect that those procedures will be free of bias across applicants. I expect that those procedures will be based on accurate information across applicants. I expect that those procedures will uphold ethical and moral standards. Distributive justice expectations. Distributive justice expectations are the respondents’ expectations regarding the fairness of outcome distributions or allocations (see Colquitt, 2001). Items were based in part on those of Leventhal (1976). revised to be appropriate for a job applicant, and combined as a distributive justice scale. Items content was revised to tap issues of work experience, education, or, skills, and potential, where necessary for this study. Items were reframed as expectations of justice (see Bell. Wiechmann, & Ryan, 2006) and the focal point was changed from the self to applicants in general. Colquitt (2001) demonstrated reliability for his distributive justice scale of .92 to .93 in different samples. Responses were captured on a 5-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). Thinking about the jobs for which you and your peers will be applying. the following items refer to hiring decisions about applicants. To what extent do you agree: I expect that the hiring decision about applicants will reflect the effort that applicants have put into their application. I expect that the hiring decision about applicants will be appropriate for the w ork experience, education or skills that applicants have. I expect that the hiring decision about applicants will reflect the potential that applicants have to contribute to the organization. I expect that the hiring decision about applicants will be justified. given their performance in the application process. Ill Interpersonal justice expectations. Interpersonal justice expectations are the respondents’ expectations regarding the fairness of interpersonal treatment in the application process; the degree to which people will be treated with politeness. dignity. and respect by authorities or third parties involved in executing procedures or determining outcomes (see Colquitt, 2001). Items were based on those of Bies and Moag (1986), and combined as an interpersonal justice scale. Colquitt (2001) demonstrated reliability for his interpersonal justice scale of .79 to .92 in different samples. Items were reframed as expectations of justice (see Bell, Wiechmann, & Ryan, 2006) for applicants. Responses were captured on a 5-point Likert-type scale ranging from I (strongly disagree) to 5 (strongly agree). Thinking about the jobs for which you and your peers are applying, the following items refer to the handling of applicants. To what extent do you agree: I expect that the company/organisation will treat applicants in a polite manner. I expect that the company/organisation will treat applicants with dignity. I expect that the company/organisation will treat applicants with respect. I expect that the company/organisation will refrain from improper remarks or comments. Informational justice expectations. Informational justice expectations are the respondents’ expectations of fairness in the explanations provided to applicants that convey information about why procedures were used in a certain fashion or why outcomes were distributed in a certain fashion (see Colquitt, 200] ). Items were based on those of Bies and Moag (1986) and Shapiro, Buttner and Barry (1994) and combined as an informational justice scale (Colquitt, 2001). Colquitt (2001) demonstrated reliability for his informational justice scale of .79 to .90 in different samples. Items were ref ramcd as expectations of justice (see Bell, Wiechmann, & Ryan. 2006) for applicants. 112 Responses were captured on a 5-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). Thinking about the jobs for which you and your peers are applying, the following items refer to the handling of applicants. To what extent do you agree: I expect that the company/organisation will be candid in its communication with applicants. I expect that the company/organisation will explain the procedures thoroughly. I expect that the company/organisation’s explanations regarding the procedure will be reasonable. I expect that the company/organisation will communicate details in a timely manner. . I expect that the company/organisation will seem to tailor its communications to individual applicants’ needs. Restorative justice expectations. Restorative justice has three possible facets. The first is a general restorative dimension capturing the existence of preference for those from previously disadvantaged groups. The second is a procedural restorative dimension that captures fair process within a target group. The third is a distributive restorative dimension that captures appropriate distribution within a target group. General restorative justice expectations. General restorative justice is the respondents’ fairness expectation regarding the distribution or allocation to target groups that have suffered inequality in the past. Items here were written to capture the restoration of social balance by giving preference to those from previously disadvantaged groups. The items are the same as those for group restorative justice expectations discussed above, but are framed for the individual-level referent. Items were framed as expectations of justice (see Bell, Wiechmann, & Ryan, 2006). Responses were captured on a 5-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). Thinking about the jobs for which you and your peers will be applying the following items refer to company/organisation consideration of job applicants. Io what extent do you agree. 113 I expect that special consideration in hiring will be provided to those from previously disadvantaged groups. I expect that jobs will go to people from affirmative action’s target beneficiary groups. I expect that an attempt will be made to correct past social injustices in hiring. I expect that preference in hiring will be given to women, blacks, or the disabled. Procedural restorative justice expectations. Procedural restorative justice expectations are the respondents’ expectations regarding the fairness of procedures used to determine outcome distributions or allocations within target groups. Items were based on justice perception items of Thibaut and Walker (1975) and Leventhal (1980), and combined as a procedural justice scale as proposed by Colquitt (2001) showing alpha levels of .78 to .93. Items were framed as expectations of justice (see Bell, Wiechmann, & Ryan, 2006), and the focal point here was each of the two target groups of interest (women and blacks). Responses were captured on a 5-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). Thinking about the jobs for which you and your peers will be applying. the following items refer to company/organisation procedures used to make hiring decisions about [target group]. When selecting [target group], to what extent do you agree: I expect that those procedures will be applied consistently across [target group]. I expect that those procedures will be free of bias across [target group]. I expect that those procedures will be based on accurate information across [target fight that those procedures for [target group] will uphold ethical and moral standards. Distributive restorative justice expectations. Distributive restorative justice expectations are the respondents’ expectations regarding the fairness of outcome distributions or allocations within target groups. Items were based on those of Leventhal (1976), revised to be appropriate for a job applicant, and combined as a distributive justice scale as used by Colquitt (2001) with alpha levels of .92 to .93. Items were framed 114 as expectations of justice (see Bell, Wiechmann, & Ryan, 2006) for target groups (women and blacks). Responses were captured on a 5-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). Thinking about the jobs for which you and your peers will be applying, the following items refer to hiring decisions about [race/ gender target group]. To what extent do you agree: I expect the hiring decision about [race/gender target group] to reflect the effort that they have put into their application. I expect that the hiring decision about [race/gender target group] will be appropriate for the work experience, education or skills that they have. I expect that the hiring decision about [race/ gender target group] will reflect the potential that they have to contribute to the organization. I expect that the hiring decision about [race/gender target group] will be justified. given their performance in the application process. Outcomes Outcomes that were measured include the following. Withdrawal intention: Withdrawal intention is a retreat from applying for jobs. Applicant withdrawal intention was measured on a four-item scale developed for this study. Responses are captured on a 5-point Likert-type scale ranging from I ( .s'trongl 1' disagree) to 5 (strongly agree). There is no point in my applying for jobs. I don’t intend to apply for many jobs. I have taken steps toward applying for jobs. (reverse coded) I probably won’t put much effort into applying for jobs. Emigration intention: Emigration intention is the respondents’ thinking about and taking steps toward emigration, and was measured on a three-item scale using two items taken directly from the 24-item Intention to Emigrate scale of Miller, Haskill. and Thatcher (2002), along with one modified item. Their 24—item scale had a coefficient alpha of .95. As their scale was too long for the purposes of the present study. and captured a variety of dimensions, three items were selected that captured the focused 115 intention that I was seeking. Responses were captured on a 5-point Likert-type scale ranging from I (strongly disagree) to 5 (strongly agree). I frequently discuss emigration with people who are close to me. I have taken steps toward emigration. I am very likely to emigrate. Entrepreneurial intention: Entrepreneurial intention, or intent to start one‘s own business was measured on a three-item scale developed for this study. Responses were captured on a 5-point Likert-type scale ranging from I (strongly disagree) to 5 (.s'trrmglv agree). I intend to start my own business. I frequently discuss starting my own business with people who are close to me. I have taken steps to start my own business. Education intention: Educational intention, or intent to study further beyond the present degree, was measured on a three-item scale developed for this study. Responses were captured on a 5-point Likert-type scale ranging from I (strongly disagree) to 5 (strongly agree). I intend to improve my education beyond this degree. I frequently discuss getting further education beyond this degree with people who are close to me. I have taken steps toward studying further beyond this degree. Core self-evaluation: Core self-evaluation is a broad latent personality trait indicated by self-esteem, generalized self-efficacy, neuroticism, and locus of control (Judge, Erez, Bono, & Thoreson, 2003). The Core Self-Evaluation scale (Judge, et al.. 2003) was used to measure core self-evaluation. They showed an average reliability of .84 across five samples. Responses were captured on a 5-point Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). This measure was included primarily to 116 capture a baseline for core self-evaluation for purposes of examining change over time in a longitudinal study. I am confident I get the success I deserve in life. Sometimes I feel depressed. (reverse coded) When I try, I generally succeed. Sometimes when I fail I feel worthless. (reverse coded) I complete tasks successfully. Sometimes, I do not feel in control of my work. (reverse coded) Overall, I am satisfied with myself. I am filled with doubts about my competence. (reverse coded) I determine what will happen in my life. I do not feel in control of my success in my career. (reverse coded) I am capable of coping with most of my problems. There are times when things look pretty bleak and hopeless to me. (reverse coded) The order of the measures in the questionnaire was arranged to reduce the effect of making especially salient demographic variables on responses to attitudinal indicators. That is, demographic questions and racial identity-related questions were placed toward the end of the questionnaire. The sequence of measures is shown in Table 9. Table 9 Sequence of Measures in Survey F orm 1 Individual justice expectations (five justice dimensions) 2 Employment and job-seeking status and interests 3 Social influence (views of family, friends and co—workers) 4 Individual feelings, behavior and intentions (education intention, entrepreneurial intention, withdrawal intention, emigration intention, core self- evaluation) 5 Individual beliefs (preference for merit principle, affirmative action support) 6 Race/gender group beliefs (support for affirmative action, white guilt. restorative justice expectations) - 6 Demographics 7 Race centrality 8 Exposure to integration 9 Designated affirmative action beneficiary group status 10 Fear of crime 1 1 Background and experience 12 Qualitative items — reasoning behind outcome intentions, and comments about selection 117 Analyses First, to establish restorative justice as a distinct factor, I conducted confirmatory factor analysis and examined goodness-of-fit indices to evaluate whether the expanded justice model provides a better fit than does the established model. CF A is appropriate (see Lance & Vandenberg, 2002) as I argue that there are conceptual and theoretical distinctions between restorative justice and other justice dimensions as identified by Colquitt (2001 ), and I have determined the observed variables that should represent each latent variable. CFA was conducted using LISREL 8.71 (Joreskog & Sorbom. 2004). CF A was also used to demonstrate the value of considering more specific justice expectations — procedural restorative justice and distributive restorative justice. in the case of this study for women and for blacks. Scale invariance tests were conducted using multiple groups CFA (see Vandenberg and Lance, 2000) to examine factor loading invariance across gender and across the three race groups of sufficient sample size. For these analyses there are various indicators of model fit (Hu & Bentler. I999; Vandenberg & Lance, 2000). A chi-square ()6) that is statistically nonsignificant suggests a well-fitting model. A root mean square error of approximation (RMSEA) of 0 is ideal. with up to .06 being very good and up to .08 being acceptable. The nonnormed fit index (NNFI or TLI) is not affected by sample size, and a value of .95 shows excellent fit. with .90 as a lower bound of good fit. A comparative fit index (CFI) of .90 or higher is generally considered acceptable. A standardized root-mean-square residual (SRMR) of .08 indicates excellent fit, with an upper bound of .10 for good fit. Using the guide of 118 Meade, Johnson, and Braddy (2008), a CFI change of 0.002, when nested models are compared, was used as the cutoff indicating lack of invariance. Next, to test for the effects of individual-level predictors on individual-level dependent variables, as well as to consider the effect of group membership on the relationship between predictors and dependent variables at Level I, 1 simultaneously tested for significant pooled slopes and between-group variance in Level 1 intercepts and slopes. In using hierarchical linear modeling software HLM 6.06 (Raudenbush. Bryk. & Congdon, 2008) for these analyses, race and gender are simultaneously tested as moderators of the relationship between antecedents and expectations or intentions. rather than by using a product term as would be done in regression analysis. That is, to consider the contribution of the group level factors to the explanation of variance in individual- level dependent variables, I examined the significant effects of group culture of support for affirmative action and group expectations of general restorative justice on Level 1 dependent variables. In HLM terminology, estimation of parameters here are “fixed” effects and “random” effects (see Raudenbusch & Bryk, 2002). Fixed effects show the population regression intercept (1’00); the population regression coefficient for the regression of the dependent variable (e.g., procedural justice expectations) on the Level 1 predictor (e.g., social guidance) (ym); the population regression coefficient of the dependent variable on the Level 2 predictor (e.g., group culture of support for affi rmatix'c action) (70/); and the population regression coefficient for the interaction of Level I and Level 2 variables in predicting the dependent variable (y, 1). The random effects are the variance in intercepts, e. g., group mean differences in distributive justice expectations (T00); variance in slopes, e.g., race/gender groups have significant differences in slope for 119 the relationship between African centrality and procedural justice expectations ( 17,); and covariance between intercepts and slopes (1:01). In this study, significant fixed effects results for y”; would be showing a main effect for a relationship between a Level I predictor and a Level 1 dependent variable; the pooled slope. However, it should be noted that when groups have very different relationships between the predictor and dependent variable, such as one group having a relationship with a positive slope and another group having a negative slope, these differences in patterns for groups may result in a nonsignificant pooled slope. Therefore it is important to consider the random effects and examine the variance in slopes 1:]; across groups. There may still be a meaningful relationship between a predictor and dependent variable, but that relationship differs. or is moderated by, group membership. As I was examining the contextual effects of culture of white guilt, culture of affirmative action support, and culture for restorative justice expectations, 1 evaluated whether I should aggregate responses to capture the group level construct. To evaluate whether aggregation of responses to groups is appropriate, ICC( I ), ICC( 2) and rug values (see Klein & Kozlowski, 2000) were calculated for white guilt, support for affirmative action, and group-level restorative justice expectations. Greater between-group variance relative to total variance would be indicated by a significant F-test, and as an appropriate indication to support aggregation. Where ICC(2) values meet or exceed .7. this suggest differentiation in group means, indicating reliable between-group differences. An rug value of .7 was used as a rule of thumb indicator of within-group agreement. 120 To test the significance of purely group-level relations I examined the generalized least squares estimates of Level 2 regression parameters. To test for the fully mediated relationship between group beliefs and individual expectations, I followed the steps of Baron and Kenny (1986), attempting to demonstrate that direct relationships exist between all three of the constructs, but that when testing for the effect of group beliefs regarding guilt on individual expectations, controlling for group support of affirmative action, the effect should be zero. 121 First, I provide scale descriptive statistics, reliabilities and scale intercorrelations. RESULTS Second, confirmatory factor analysis results for demonstrating restorative justice are presented. Third, common method variance is addressed. Fourth, I provide results of scale invariance tests across gender and race before these scales are used in multivariate hypothesis testing. Fifth, results of multivariate analyses are presented. Descriptive Statistics, Reliabilities, and Intercorrelations Descriptive statistics and reliabilities for scales are provided in Table 10 and correlations among variables in the model are provided in Table 11. Table 10 Scale Descriptive Statistics and Reliabilities Mean SD Skew- Kurtosis (l ness Individual-Level Justice Expectations Procedural (4) 3.64 1.04 -0.32 -1.03 0.89 Distributive (4) 3.76 0.82 -1.04 1.02 0.80 General Restorative (4) 3.63 1.04 ' -0.59 -0.41 0.86 Procedural Restorative — Women (4) 3.73 0.91 -0.44 -0.42 0.86 Distributive Restorative —Women (4) 3.85 0.87 -0.63 -0.22 0.86 Procedural Restorative — Blacks (4) 3.65 1.03 -0.52 -0.54 0.89 Distributive Restorative — Blacks (4) 3.66 1.08 -0.69 -0.38 0.91 Group-Level Antecedentsl Mean Support for AA (5) 2.44 0.61 0.32 -l .34 0.9l Mean Belief in White Guilt (3) 3.93 0.79 -0.21 -l .33 0.80 Mean Restorative Justice (4) 3.76 0.17 -0.47 -1.07 0.77 Individual-Level Antecedents Racial Centrality (4) 3.19 0.93 -0.33 -0.35 0.76 African Centrality (4) 3.52 0.91 -0.58 0.06 0.8] Support for AA (5) 3.15 0.95 -0. l 7 -0.35 0.87 Preference for Merit Principle (4) 4.32 0.58 -l .08 2.83 0.56 Educational Integration 0.42 0.25 -0.55 -l . l 5 Social Guidance — Unfair (4) 3.26 0.87 , -0.13 -0.09 0.84 Intentions Withdrawal (4) 1.81 0.63 1 .06 0.92 0.55 Emigration (3) 2.99 1.10 -0.26 -0.63 0.83 Entrepreneurial (3) 3 .02 1.12 0.03 -0.79 0.86 Education (3) 4.06 0.78 -0.81 0.74 0.69 122 Table 10 continued. Note. AA = Affirmative Action. The possible range on scales was I —— 5. N = 317, which is the sample for later multivariate analyses. IAlthough mean, SD, skewness,-and kurtosis values for group- level antecedents are calculated after individuals were assigned the group mean scale score. coefficient alpha is calculated based on relevant original individual item responses. The number of items in the scale is shown in parentheses after the scale name where appropriate. Mean levels for individual justice expectations ranged from 3.63 to 3.76 (where 3 = Neither disagree nor agree, 4 = Agree, and 5 = Strongly agree). suggesting that on the whole, respondents tended to agree with the justice expectation statements provided: expectations were generally for fairness rather than unfairness. Skewness and kurtosis statistics for the individual justice measures did not reflect problematic distributions. The range of coefficient alpha for the individual justice expectations dimensions was a = 0.80 to a = 0.91. Viewing a reliability of a = 0.70 as acceptable, all of the individual justice expectation scales demonstrated high internal consistency reliability. Mean levels of the group-level constructs reflected that on average. respondents expected that members of their race and gender group were neutral on affirmative action (mean = 2.44), believed in white guilt (mean = 3.93), and expected restorative justice (mean = 3.76). Skewness and kurtosis statistics for the group-level antecedents did not reflect problematic distributions. Scale reliabilities for the group constructs were acceptable, ranging from a = 0.77 for general restorative justice to a = 0.91 for group culture of support for affirmative action. Mean levels for individual-level constructs showed respondents to be fairly neutral overall on race centrality (mean = 3.19), support for affirmative action (mean t 3.15), and social guidance (mean = 3.26). Respondents were slightly more likely to affirm their African centrality (mean = 3.52). A preference for merit principle was 123 affirmed by the respondents (mean = 4.32). Skewness and kurtosis statistics for the individual-level antecedents did not reflect problematic distributions, however. the preference for merit principle scale showed the greatest deviation from a normal distribution, with a kurtosis of 2.83. Scale reliabilities ranged from a = 0.56 (merit principle), which is low, to a = 0.87 (support for affirmative action). Responses regarding intentions showed that on average, respondents were unlikely to withdraw from the selection process (mean = 1.81), were neutral regarding emigration (mean = 2.99) and entrepreneurship (mean = 3.02), and were positive regarding seeking further education (mean = 4.06). Skewness and kurtosis statistics for intentions did not reflect problematic distributions. Scale reliabilities ranged from 0.55 (withdrawal intention) to 0.89 (entrepreneurial intention). The low reliability for preference for merit principle (a = 0.56) and withdrawal intentions (or = 0.55) is problematic as low scale reliability limits the power for showing significant relations between scales. 124 cm: mm:- Er o..- N: m 2. f.- omr ENE: - 852:0 Boom _m a m .- 8.- 8. 8.- 8.- 8.- 8.- 8. 55525 558538 on 8. 88. 8. 8. 88. 8. 8.- 8.- 58955 .52 2 2.. 8. 8. 8. 8. 2. 8.- :. 55< 5828.2 5.. e888 2 2. 8. 8.- 8. 8.- E. 8.- 8.- 5:850 52?. 2 8. 8. 8. 8. 8. 8.- :. 2. 2.858 88 2 hukmmumbfi: V ~m>mNuNB-~N~>§§~N 8. 2.-. 8.- 8. 8. 8.- 8.- 8. A _ H 2.252 8 u gas: .850 2 8.- 8.- 2.- 2.- 2.-. E. - 2.- 8.- . . 323 - 832 E 2.- 8. 8. 8. :. 8.- 8. 8. 585 - 8am 2 8. 8.- 8. 8.- 8. 8.- 8. 8. 5528 - 832 2 8. 8m. 8. 8. 8. 8.- 8. t. 582 52m - 8.2 : 8. 8. 8.- 8.- :.- 8. 8.- 8.- 825088 2,58an 885 :52 8. 8. 8. 3. 2. E. 2 .- :. 8. 85 85$ 5 55m 95.5 522 8 88.2 8. 2. :. :. 2.- 8. 8. << 8.2 e885 885 =52 w Eammgmbmu: V Exam-N 1%: QLmV 82 8. 8. 8. 8.- 8. om. 222m - 228952 525255 8 8.2 8. 8. cm.- 8. 8. 202m - 588952 55588 8 82 8. on- 8. 8. 58o? - 35.553. 25855 m 004 omf 0?. Nb. c0803 1 o>mwnuoumom HSUQOOum .v 8.2 8.- 8.- 2,5552 3550 m 82 8. 58355 N 8.. 38585 2 ate-QGBwQR-N math-:5 ~m>mfit~33§~>m~u=~ w 8 8 m 2. m N _ News: E 833.8: \e 8.23: 2833285 : 2an 125 3.- no. w _. S. No... 2.. Sr 2...- ewe-ED - 00:8sz 880m .m 2...- mo. mm. mo: 00. mm; mo: 3... 00:83:: Ecozeosum om v0. 00. mo.. 2. E .- oo. 00.- oo. 00:85ch €02 3 2. 8.- 8.- 8.- 8.- 84. 8. 8. 55< 55.502 5 e888 2 mm. 00.- mo. 2.. 3.- om. 2. 00. 5:05:00 SSE< 2 co; wor em; 2. 00.- mm. 00.- mm. bzmbcou 000% 0— 2203825- Nessa-3:83.55 co._ oo. 00.- 00. No. 5.- mo.- A _ n 2080..” no u 03>: 000000 2 .. 2: 02- 3..- 2V.- .. mm. 3.. . .. BE? - 000m 3 co; 3.- oer wmr vo. 5:05 - 000m 2 002 mm; mo. 3. 0050—00 - 000m N— 004 mm. 5. 5355. 2005 - 000m : co; 2.- mcoufiooaxm 0382083” 0:000 0002 o— oo._ :30 0:55 E mom—0m 000.5 :82 0 << 20.2 tongsm 00020 :32 w 2202000032 Neg-798.5 302m - 0308083” 0325855 b 8.005 - gums—080% 8500005 0 0080.3 - 0303080M 032:8:me m 5803 - 0388083“ 8060005 .0 0308083“ 120000 m 250355 N 3300005 2 etched-ommwm 0.0.535 Kama-3:35.85 3 m2 3 2 Q I 2 0 300: 2.2 830.20: \0 323$ 20250-200 00:03:00 I 033- 126 2: we. 2. mm.. #0. ENE: - 0000050 Eoom _m 8.. 8.- S.- 3- 802025380830 on of 2.- 8.- 5095:0052 2 8.. 0m. :23 0135025030030 M: 2: 0:20:00 58:2 2 00:05:00 0005 2 $203000Ew ~0>0~<§EPSE 8 n 01080-5 2 H 03>: 000000 2 . . 0:55 - 0005 3 00:05 - 0005 m 5 0050—00 - 0005 ~— 0005< x005 - 0005 2 m00m§000xm 030800005 @005 0002 3 :so 323 5 00:00 0:80 :82 0 << .05 tommsm 9.0.5 0002 w S§B000§V ~0>0--§0.5 8.8—m - 030800005 0330505 5 302m - 030800005 3.500005 0 00803 - 030008005 0.52:me m 0080? - 030800005 58200005 v 030800005 300000 m 0302:0005 N 70.500005 _ 02023.00me 0020-3- ~0>0N4§3>§=~ _N om o. M: E 3003 :5 0.03020; \0 H.503 20.230589 60:50:00 2 0305- 127 803300 03333:. 803. .55 03:30 05 m_ 5033 \- _ m n 2 .3. w Q3 EmoEcm8 08 00.. 08 3:00 .0 :35 880% 20:20:09 .032 8. :. 8.- 8. 8.- 8. 8. 8. 8:880 8 8._ 3. 8.- 8.- 8. 8.- 8. 8. 30888280 8 8._ 8. ,8. .8. 2. 8.- 8..- . 88880 8 8.2 . 2. 8. 8.- 8.- 8.- 328:3 8 8:02:85 88 mm 88 _m 88 8_ 8_ 8. 8.- 8. 8.- 8. 8.- 8. 8.- 2. 80880 8 8. :- 8- 8. 8.- S. 8.- 8. 8882880 8 8. 8. 2. 2. 2.- :.- 8.- S.- 88880 8 8.- 2.- 8. 8.- 8.- 8.- S. 8.- 388:3 8 thwntmufiN 8 2 S 2 S : 2 8 2. 8.- 8.- 8.- 8.- 8. 8. 8. 8:853 8 8. 8. :. :. S. 8.- 8. :. 35058.85 8 t.- 8.- 8. 8.- 8.- 8.- 8. 8.- 88880 8 8.- 8.- 8. 8.- 8.- 8. 8.- 8.- 328:3 8 mic-ztmstN 8 8 0 m v m 8 _ 3303 8: 803558 \0 8.23: 20:59:09 60:58:00 : 0.3-w 128 Relations Between Individual-Level Justice Expectations The sizes of correlations between individual-level justice dimensions range from r = .32 to .69, suggesting that the justice dimensions are related yet distinct, as anticipated with Hypothesis 1. General restorative justice is negatively related to the other justice dimensions suggesting that on the whole, where there is the belief that preferential selection will be taking place based on demographic group membership, expectations of procedural and distributive justice will be lower. Relations Between Group-Level Antecedents Mean group belief in white guilt was very strongly related to mean group support for affirmative action (r = .97), as was anticipated with Hypothesis 7. Group belief in white guilt showed a small negative relationship with group general restorative justice expectations (r = -.11). Group differences in beliefs regarding white guilt were anticipated, and group belief in white guilt was very strongly positively related to race for black Africans (r = .81) and very strongly negatively related to race for whites (r = -.88). Similarly, mean group support for affirmative action was very strongly positively associated with race for black Africans (r = .90), and very strongly negatively related for whites (r = -.76). There was also a negative relationship between group support for affirmative action and race for Indians (r = -.15). Group general restorative justice expectations was strongly negatively associated with race for Indians (r = -.58) and positively associated with race for whites (r = .38) and black Africans (r = .32). Race and gender group differences were anticipated, but based on the correlations shown. race is more important than is gender in relation to differences in these group-level constructs. 129 Individual-Level Justice Expectations and Group-Level Antecedents Procedural justice expectations were moderately negatively associated with race for whites (r = -.26), and moderately positively associated with race for black Africans (r = .17). Procedural justice expectations were moderately positively associated with mean group support for affirmative action (r = .24) and mean group belief in white guilt (r = .25). These patterns for procedural justice are consistent with what was anticipated in part of Hypothesis 3 and in Hypothesis 8 and 11. Counter to what was anticipated in part of Hypothesis 3 and Hypothesis 9 and 1 l- distributive justice expectations did not show strong relationships with group-level antecedents. The largest relationships here were with race for whites (r = -.13), and mean group belief in white guilt (r = .11). Further, general restorative justice expectations did not show strong relationships with group-level antecedents. The largest relationships here were with race for whites (r = .14), mean group belief in white guilt (r = -.10) and mean group support of affirmative action (r = -.10). These patterns for race combined with relations between justice dimensions are consistent with what was anticipated in Hypothesis 4. Counter to what was anticipated with Hypothesis 12, there was only a marginally significant relationship between general restorative justice at the individual level and the group-level restorative justice expectation. Procedural restorative justice - women was moderately strongly negatively related to race for whites (r = -.19) and showed a smaller and positive relationship for Indians (r = .11). Procedural restorative justice -——- women showed small relationships with mean group belief in white guilt (r = .14), mean group support for affirmative action (r = .11), and mean group restorative justice expectations (r = -.1 1). Distributive restorative 130 justice — women was moderately strongly negatively related to race for whites (r = -.15). Distributive restorative justice — women showed small relationships with mean group belief in white guilt (r = .13), mean group support for affirmative action (r = .1 1). These patterns are consistent with those anticipated in Hypothesis 10 for procedural and distributive restorative justice — women. Procedural restorative justice - blacks was moderately strongly negatively related to race for whites (r = -.15). Procedural restorative justice — blacks showed relationships with mean group belief in white guilt (r = .14), and mean group support for affirmative action (r = .12). Distributive restorative justice - blacks was moderately strongly negatively related to race for whites (r = -.20) and positively for black Africans (r = .20). Distributive restorative justice - blacks showed moderately strong relationships with mean group belief in white guilt (r = .24), and mean group support for affirmative action (r = .23). Distributive restorative justice — blacks was negatively related to gender for women (r = -.10). These patterns are consistent with those anticipated in Hypothesis 10 for procedural and distributive restorative justice — blacks. These relations for distributive restorative and procedural restorative justice and race are consistent with Hypothesis 5, but the lack of a relationship for gender with most of the justice dimensions is not. Individual-Level Justice Expectations and Individual-Level Antecedents Procedural justice expectations were moderately strongly negatively related to social guidance regarding unfairness (r = -.20), consistent with Hypothesis 28, and showed a small positive relationship with support for affirmative action (r = .1 1) consistent with Hypothesis 18, and race centrality (r = .12), consistent with Hypothesis 131 15. The lack of relationship between African centrality and procedural justice is inconsistent with Hypothesis 16. The lack of a relationship between preference for merit principle and procedural justice is inconsistent with Hypothesis 20. Distributive justice expectations were moderately strongly negatively related to social guidance regarding unfairness (r = -.16), consistent with Hypothesis 29, and showed a small positive relationship with race centrality (r = .1 1), consistent with Hypothesis 14. The absence of a significant relationship between distributive justice expectations and educational integration is inconsistent with what was anticipated in Hypothesis 27. The lack of relationship between African centrality and distributive justice is inconsistent with Hypothesis 17. The lack of a relationship between support for affirmative action and distributive justice is inconsistent with what was anticipated in Hypothesis 19. The lack of a relationship between preference for merit principle and distributive justice is inconsistent with Hypothesis 21. General restorative justice expectations were moderately strongly positively related to social guidance regarding unfairness (r = .15) and support for affirmative action (r = .18), and showed a small positive relationship with African centrality (r = .10). Procedural restorative justice — women showed a small negative relationship with social guidance regarding unfairness (r = -.12). Distributive restorative justice women (r = - .19), procedural restorative justice — blacks (r = -.17), and distributive restorative justice blacks (r = —.22) all showed moderate negative relationships with social guidance regarding unfairness. Distributive restorative justice — blacks also showed a moderate positive relationship with support for affirmative action (r = .25). 132 Group-Level Antecedents and Individual-Level Antecedents Group culture of support for affirmative action was strongly positively related to individual support for affirmative action (r = .43) and race centrality (r = .34) and showed a small positive relationship with African centrality (r = .13). Group culture of support for affirmative action was strongly negatively related to educational integration (r = -.33) and social guidance regarding unfairness (r = -.20). Hypothesis 27 reflected the anticipation that group culture of support for affirmative action would be associated with distributive justice, but that relationship was only marginally significant. Group belief in white guilt was strongly positively associated with individual support for affirmative action (r = .42) and race centrality (r = .37). Group belief in white guilt was strongly negatively associated with educational integration (r = -.33) and showed a moderate negative relationship with social guidance regarding unfairness (r = -.21). Group general restorative justice expectations showed a small relationship with African centrality ('r = .13). That group general restorative justice expectations showed no significant relationship with procedural and distributive justice expectations is inconsistent with what was anticipated in Hypothesis 13. Race for black Africans was associated with several individual-level antecedents. These include positive relationships for support for affirmative action (r = .40). race centrality (r = .28), and African centrality (r = .20). There were negative relationships between race and educational integration (r = -.32) and social guidance regarding unfairness (r = -.17) for black Africans. Being coloured was negatively associated with African centrality (r = -.19) and preference for merit principle (r = -.10). Being Indian was negatively associated with African centrality (r = -.10) and positively associated with 133 preference for merit principle (r = .10). Race was strongly negatively related to race centrality (r = -.34) and support for affirmative action (r = -.31) for whites. Being white was positively related to educational integration (r = .28) and social guidance regarding unfairness (r = .18). Relations Between Individual-Level Antecedents Race centrality was strongly positively related to African centrality (r = .33) and more moderately to support for affirmative action (r = .18). Race centrality was negatively related to educational integration (r = -.24). Support for affirmative action w as positively associated with African centrality (r = .26) and negatively related to social guidance regarding unfairness (r = -.28) and educational integration (r = -. 14). Preference for merit principle was positively associated with social guidance regarding unfairness (r = .16). Withdrawal Intentions Withdrawal intentions were largely unrelated to individual justice expectations. which is inconsistent with what was anticipated in Hypotheses 30, 31. 32, and 33. but showed a small positive relationship with group general restorative justice expectations (r = .10). Withdrawal showed a positive relationship with social guidance regarding unfairness (r = .15). Emigration Intentions Emigration intention was largely unrelated to individual justice expectations. which is inconsistent with what was anticipated in Hypotheses 34, 35, and 36, although there were relations with race. Emigration intention was positively related to race for whites (r = .15) and Indians (r = .10), and negatively related for black Africans (r = -.1 l) and coloured 134 people (r = -.19). Emigration intention was negatively related to group culture of support for affirmative action (r = -.l7) and group belief in white guilt (r = —.17). Emigration intention was strongly negatively related to individual support for affirmative action (r = -.30) and moderately negatively related to African centrality (r = -.18). Emigration intention was moderately positively associated with social guidance regarding unfairness (r = .23) and showed small relationships with preference for merit principle (r = .13) and educational integration (r = .12). Entrepreneurial Intentions Entrepreneurial intention was positively related to individual procedural justice expectations (r = .11), procedural restorative justice — women (r = .10), distributive restorative justice — women (r = .11), procedural restorative justice - blacks (r = .1 1). distributive restorative justice -— blacks (r = .20), yet was unrelated to general restorative justice expectations as was anticipated in Hypothesis 38, although were associations with race. Entrepreneurial intention was positively associated with race for black Africans (r = .17) and negatively associated with race for whites (r = -.19) and Indians (r = -.17). Entrepreneurial intention was also positively associated with group culture of support for affirmative action (r = .18) and group belief in white guilt (r = .21). Education Intentions Education intentions were largely unrelated to individual justice expectations- which is inconsistent with what was anticipated with Hypotheses 24, 25, and 39. Education intentions showed a small positive relationship with group culture of support for affirmative action (r = .10) and group belief in white guilt (r = .10). 135 Relations Between Intentions Withdrawal was negatively related to education intention (r = -.19). Emigration intention showed small positive relationships with entrepreneurial intention (r = .14) and education intention (r = .11). Entrepreneurial intention was positively associated with education intention (r = .21). The multivariate analyses are next reported to address specific hypotheses. Demonstrating Restorative Justice General Restorative Justice Using the full sample (N = 327), confirmatory factor analysis (C F A) was conducted using LISREL 8.71 (Joreskog & Sorbom, 2004). This was to assess the dimensionality of justice expectations (Colquitt, 2001; Colquitt & Shaw, 2005; Lance & Vandenberg, 2002), using maximum likelihood estimation to evaluate whether a fifth dimension of justice (restorative justice) is empirically supported. CF A is appropriate (see Lance & Vandenberg, 2002) as I argue that there are conceptual and theoretical distinctions between restorative justice and other justice dimensions as identified by Colquitt (2001), and I have determined the observed variables that should represent each latent variable. The correlation matrix and standard deviations of observed variables used for this analysis are provided in Appendix D. Resultant indicators of model fit are shown in Table 12. 136 Table 12 Comparison of A Priori Organizational Justice Factors Structure XZ df NNFI CFI SRM RMSF, RMSEA R A confidence interval l-factor 1624.14 189 0.83 0.84 0.11 0.17 (0.17: 0.18) 2-factor 1338.86 188 0.86 0.87 0.12 I 0.16 (0.16; 0.17) 3-factor 975.70 186 0.90 0.91 0.12 0.14 (0.13: 0.15) 4-factor 714.66 183 0.93 0.94 0.11 0.11 (0.1 l : 0.12) S-factor 348.36 179 0.98 0.98 0.06 0.06 (0.05: 0.06) Note. All )Xvalues are significant at p < .001 . df= degrees of freedom; NNFI nonnormed fit index; CFI = comparative fit index; SRMR = standardized root-mean-square residual; RMSEA : root mean square error of approximation. N : 327. Model 1 was a one-factor model with all items loading on a generaljustice dimension. Model 2 was a two-factor model capturing distributive and procedural dimensions only. Items designed to capture restorative justice were assigned to the distributive dimension while items from traditional interpersonal or informational justice scales were assigned to the procedural dimension. Model 3 was a three-factor model capturing distributive, procedural and interactional justice. Items designed to capture restorative justice were assigned to the distributive dimension while items from traditional interpersonal or informational justice scales were assigned to the interactional dimension. Model 4 was a four-factor model capturing distributive, procedural. interpersonal and informational justice, where restorative justice items were still assigned to the distributive dimension. Model 5 was five-factor model assigning restorative justice items to a distinct restorative justice dimension. The five-factor model with a general 137 restorative justice dimension shows the best model fit of the five models tested, and that fit is excellent. The five-factor model provided a change in CFI of .04 when compared with the next best fitting model, the four-factor model. The CF I change suggests a practically meaningful change where the value is higher than .002 (Meade. Johnson. & Braddy, 2008). Following the model of existing reactions research (e. g., Aquino, Lewis. & Bradfield, 1999; Bell, Wiechmann, & Ryan, 2006), if common method variance is extreme, I would expect that testing for a one-factor model (common method) will produce good fit, while a multi-construct model will demonstrate a worse fit. As shown. the multi-factor models demonstrate better fit. While the one-factor model did not show an acceptable level of fit as indicated by the nonnormed fitindex (NNFI) of 0.83. comparative fit index (CFI) of 0.84, standardized root-mean-square residual ( SRM R) of 0.11, and root mean square error of approximation (RMSEA) of 0. 1 7, other models demonstrated better fit. As the more specific dimensionality of justice was recognized in the models, the model fit improved. The five-factor model with procedural. distributive, informational, interpersonal, and general restorative dimensions produced the best fit with excellent fit evidenced by multiple indices (NNFI = 0.98. CFI = 0.98. SRMR = 0.06. and RMSEA = 0.06). Evidence for convergent validity was demonstrated by strong factor loadings of the appropriate items on the prescribed factors. Factor loadings and uniquenesses for the five-factor model are shown in Table 13. Most factor loadings were well above 0.60. The item with the lowest factor loading (0.56) and largest uniqueness (0.69) was on the informational justice dimension, “I expect that the company/organisation will seem to tailor its communications to individual applicants’ needs.” 138 Table 13 Factor Loadings and Uniquenesses for F ive-F actor Justice Model Factor Item Loading Unique- DCSS l I expect that those procedures will be applied consistently across applicants. I expect that those procedures will be free of bias across applicants. I expect that those procedures will be based on accurate information across applicants. I expect that those procedures will uphold ethical and moral standards. 0.78 0.92 0.74 0.82 0.39 0.16 0.45 0.32 I expect that the hiring decision about applicants will reflect the effort that applicants have put into their application. I expect that the hiring decision about applicants will be appropriate for the work experience, education, or skills that applicants have. I expect that the hiring decision about applicants will reflect the potential that applicants have to contribute to the organization. I expect that the hiring decision about applicants will be justified, given their performance in the applicatiormrocess. 0.66 0.76 0.74. 0.68 0.56 0.42 0.46 0.53 I expect that the company/organisation will treat applicants in a polite manner. I expect that the company/organisation will treat applicants with dignity. I expect that the company/organisation will treat applicants with respect. 139 0.74 0.93 0.92 0.45 0.13 0.16 Table 13 continued. I expect that the 0.60 0.64 company/organisation will refrain from improper remarks or comments. 4 I expect that the 0.62' 0.62 company/organisation will be candid in its communication with applicants. I expect that the 0.78 0.39 company/organisation will explain the procedures thoroughly. I expect that the 0.71 0.50 company/organisation’s explanations regarding the procedure will be reasonable. I expect that the 0.73 0.47 company/organisation will communicate details in a timely manner. I expect that the 0.56 0.69 company/organisation will seem to ' tailor its communications to individual applicants’ needs. 5 I expect that special consideration 0.81 0.34 in hiring will be provided to those from previously disadvantaged groups. [expect that preference in hiring 0.81 0.34 will be given to women, blacks, or the disabled. I expect that an attempt will be 0.70 0.50 made to correct past social injustices in hiring. I expect that jobs will go to people 0.82 0.33 from affirmative action’s target beneficiary groups. Table 14 shows the pattern of interrelationships between justice dimensions. Although some strong relationships exist, the justice constructs do appear to be distinct. The strongest relationships are between informational and procedural (r = .69). 140 informational and interpersonal (r = .67), and between distributive and procedural justice '- (r = .65). I would argue that the constructs are distinct unless the correlations are above .80. Restorative justice is negatively related to the other justice dimensions. Table 14 Interrelationships Between Factors __g Factor 1 2 3 4 5 1. Procedural 1.00 _. 2. Distributive 0.65 1.00 3. Interpersonal 0.54 0.38 1.00 4. Informational 0.69 0.57 0.67 1.00 5. General Restorative -0.48 -0.41 -0.34 -0.46 1.00 Procedural Restorative and Distributive Restorative Justice Having demonstrated that general restorative justice is a distinct justice factor. I next sought to demonstrate the importance of capturing more specific distributive and procedural justice expectations in the context of affirmative action — expectations regarding procedural and distributive justice when a distribution is being made to a target group; procedural or distributive restorative justice. Thus [test a series of models to examine whether model fit is improved by identifying differing factors that reflect possible differences in views on how specific target groups are handled. Items included in the analysis are items from the following scales: Distributive Justice. Procedural J ustice. Distributive Restorative Justice for Women, Procedural Restorative Justice for Women. Distributive Restorative Justice for Blacks, and Procedural Restorative Justice for Blacks. Confirmatory factor analysis results for model testing are shown in Table 15. First. a one- 141 factor model with all items assigned to one general justice dimension was tested. Second. a two-factor model was tested. There, Distributive Justice, Distributive Restorative Justice for Women, and Distributive Restorative Justice for Blacks items were assigned to a distributive factor while Procedural Justice, Procedural Restorative Justice for Women, and Procedural Restorative Justice for Blacks items were assigned to a procedural justice factor. The third model tested was a six-factor structure with Distributive Justice items assigned to a distributive dimension, Procedural Justice items assigned to a procedural dimension, Distributive Restorative Justice for Women items were assigned to a distributive restorative justice for women dimension, Procedural Restorative Justice for Women items were assigned to a procedural restorative justice for women dimension. Similarly, Distributive Restorative Justice for Blacks items were assigned to a distributive restorative justice for backs dimension, and Procedural Restorative Justice for Blacks items were assigned to a procedural restorative justice for blacks dimension. The 6-factor model demonstrated excellent fit (NNFI = 0.96. C F I = 0.96, SRMR = 0.05, RMSEA = 0.10), and better fit than the previous models. The change in CFI was .04, suggesting practically meaningful change. 142 Table 15 Comparison of A Priori Organizational Justice Factors including Target Group Referent Questions Structure df NNFI CF I SRMR RMSEA RMSEA confidence interval l-factor 2220.87 252 0.88 0.89 0.09 0.17 (0.17; 0.18) 2-factor 1806.52 251 0.91 0.92 0.08 0.16 (0.15; 0.16) 6-factor 914.61 237 0.96 0.96 0.05 0.10 (0.09; 0.10) Note. All X2 values are significant at p < .001. df= degrees of freedom; NNFI = nonnormed fit index; CFI = comparative fit index; SRMR = standardized root-mean-square residual; RMS EA 2 root mean square error of approximation. N = 327. Table 16 shows the pattern of interrelationships between justice dimensions including the target group referent. Table 16 Interrelationships Between Justice Factors including Target Group Referent _ _ Factor 1 2 3 4 5 6 1. Procedural 1.00 “7* O _ - 2. Distributive 0.65 1.00 3. Procedural Restorative - 0.77 0.60 1.00 Women 4. Distributive Restorative - 0.61 0.72 0.79 1.00 Women 5. Procedural Restorative - 0.66 0.45 0.76 0.60 1.00 Blacks . 6. Distributive Restorative - 0.56 0.65 0.58 0.74 0.70 1.00 Blacks There are strong relationships between several of these justice factors. For example, procedural restorative for women and procedural (r = .77), distributive restorative for women and procedural restorative for women (r = .79), procedural 143 restorative for blacks and procedural restorative for women (r = .76), and distributive restorative for blacks and distributive restorative for women (r = .74). These strong relationships are to be expected in such closely related constructs, but are not so high that one would argue that these factors are not distinct. The results of the first set of confirmatory factor analyses confirm Hypothesis 1. A five-factor justice model that includes Colquitt’s (2001) dimensions as well as restorative justice expectations was supported. Supplemental analyses demonstrate that there is value in capturing justice expectations regarding procedural restorative justice and distributive restorative justice for specific target groups (in this case. women and blacks). Common Method Variance While I have demonstrated that a multi—factor justice model provides a better fit than does a one-factor model, which suggests that there is not a strong method factor. it is accepted that where perceptions are measured at the same point in time, they are likely to be more strongly related than when measured with a temporal gap (Hauscknecht, Day. & Thomas, 2004). To address this common method variance concern I note the variety of measures bearing varying levels of relation to each other. Further, I have shown through confirmatory factor analyses that items loaded on the anticipated factors, resulting in good model fit. Nevertheless, following the example of Hofrnann and Stetzer (1996). I attempted to identify common method effects through group-level analyses (i.e., the relationship between group belief in white guilt and culture of support for affirmative action). For this purpose, each of the race/gender groups was randomly split into two sub- samples (sub-samples A and B within each race/gender combination). The correlation 144 was then examined in three ways; independent variables from A with dependent variables from B (resultant r = .34, p < .01), independent variables from B with dependent variables from A (r = .45, p < .01), and finally, the total sample (r = .57, p < .01). Although there are similar strong relationships between the two group cultures in all three tests, the difference across these three analyses may indicate some possible common method effects (see Hofmann & Stetzer, 1996) and I am therefore cautious in my interpretation of relationships found. However, beyond the scope of this dissertation study, I plan to conduct a follow-up study with these same respondents after one year. allowing a temporal delay before their reactions are measured again. Scale Invariance Across Gender and Race Scales to be used in model testing should be first examined to establish whether people of different group are conceptualizing the captured constructs in a similar way so that results can be appropriately interpreted. Gender I tested scale measures for invariance across gender (see Vandenberg and Lance. 2000) using multiple groups confirmatory factor analysis. In Model 0, fit was examined separately in each gender group. Each model tested separately for women and men showed at least one index reflecting acceptable model fit, with most models showing good fit as indicated by high NNFI and CFI values and low SRMR values. If there was bad fit at this stage, subsequent multiple group analyses would not be necessary. Next. two multiple group models were tested for each measurement invariance test. Model 1 specified the same pattern of factor loadings and uniquenesses across groups. and allowed for unique parameter estimates. That is, one indicator on the factor was set to 145 1.00 and the error variances of indicators, error variances of latent variables, and covariances of latent variables (where appropriate) were freely estimated. In Model 2 for each construct, the unstandardized factor loadings were constrained to be equal across gender. As Models 1 and 2 are nested, change statistics from Model 1 to Model 2 are provided. As recommended by Cheung and Rensvold (2002) and Meade, Johnson. and Braddy (2008), I consider multiple indicators of meaningful change in model fit. Table 17 provides the fit indices for the different models. Table 17 Comparison of Models Across Gender Structure X2 p df NNFI CFI SRMR RMSEA RMSEA confidenc e interval Individual-Level Justice Expectations (seven-factor model) MOWOmen 777.29 0.000 329 0.96 0.964 0.07 0.09 (0.08:0.09) MOMen 745.79 0.000 329 0.95 0.953 0.06 0.09 (0.08; 0.09) M1 M0 1523.08 0.000 658 0.95 0.959 0.06 0.09 (0.08; 0.09) M2510 1550.69 0.000 679 0.95 0.959 0.07 0.09 (0.08; 0.09) Comparing M2110 to MlMg, A x2 = 27.61, A df= 21, and A CFI = 0.000. Race Centrality (one-factor model) MOWomen 13.52 0.001 2 0.86 0.954 0.05 0.18 (0.09; 0.28) MOMen 5.09 0.079 2 0.95 0.985 0.03 0.09 (0.00; 0.20) MIMG 18.61 0.001 4 0.90 0.968 0.03 0.14 (0.08: 0.22) MZMG 21.09 0.004 7 0.95 0.969 0.04 0.10 (0.05; 0.16) Comparing MZMG to MlMg, A )(2 = 2.48, A df= 3, and A CFI = 0.001. A fiican Centrality (one-factor model) .. MOWomen 8.91 0.012 2 0.93 0.977 0.04 0.15 (0.06:0.25) MOMen 3.17 0.205 2 0.99 0.997 0.02 0.06 (0.00: 0.18) M1 W 12.08 0.017 4 0.96 0.987 0.02 0.11 (0.04: 0.19) MZMG 15.56 0.030 7 0.98 0.987 0.04 0.09 (0.03: 0.15) Comparing M2MG to mm. A j = 3.48, A df= 3. and A CF] = 0.000. Supportfor A flirmative Action (one-factor model) MOWomen 17.35 0.004 5 0.96 0.981 0.03 0.12 (00610.19) MOM-m 19.27 0.002 5 0.94 0.969 0.04 0.13 (0.03;0.19) MlMg 36.62 0.000 10 0.95 0.976 0.04 0.12 (0.08:0.17) 146 Table 17 continued. MZMG 37.80 0.001 14 0.97 0.979 0.05 0.10 (0.06: 0.14) Comparing MZMG to MlMg, A x2 = 1.18, A df= 4, and A CFI = 0.003. Merit Orientation (one-factor model) MOWomen 0.50 0.781 2 1.11 1.000 0.02 0.00 (0.00; 0.21) MOMen 5.26 0.072 2 0.91 0.968 0.04 0.10 (0.00; 0.21) MI W 5.76 0.218 4 0.96 0.988 0.04 0.05 (0.00; 0.14) MZMG 7.53 0.376 7 0.99 0.996 0.04 0.02 (0.00; 0.10) Comparing MZMG to MlMg, A X2 = 1.77, A df= 3, and A CFI = 0.008. Social Guidance (three-factor model) Mowomen 15.44 0.017 6 0.95 0.979 0.03 0.10 (0.04; 0.17) MOMen 23.22 0.001 6 0.94 0.975 0.02 0.14 (0.09; 0.21) MIMG 38.66 0.000 12 0.94 0.977 0.02 0.13 (0.08; 0.17) MZMG 47.94 0.000 15 0.94 0.971 0.04 0.12 (0.09; 0.16) Comparing MZMG to M1 M0, A X2 = 9.28, A df= 3 (change significant at p= .05). and A CF I = -0.006. Group Belief in White Guilt and Support for Affirmative Action (two-factor model) MOW-men 36.24 0.010 19 0.98 0.986 0.04 0.07 (0.02; 0.1 1) MOMelrl 35.82 0.011 19 0.97 0.982 0.03 0.08 (0.04; 0.1 l) MIMG 72.06 0.001 38 0.98 0.984 0.03 0.07 (0.04; 0.10) M2 MG 84.154 0.000 44 0.98 0.981 0.05 0.07 (0.05; 0.10) Comparing M2... to MlMg, A x2 = 12.09, A df= 6, and A CFI = -0003. Group General Restorative Justice Expectations (one-factor model) MOWomen 30.76 0.000 2 0.59 0.863 0.09 0.30 (0.21; 0.40) MOMen 19.66 0.000 2 0.78 0.925 0.07 0.23 (0.14; 0.33) M1 MG 50.41 0.000 4 0.69 0.896 0.07 0.27 (0.20; 0.34) M2 MG 53.47 0.000 7 0.82 0.896 0.08 0.20 (0.15; 0.25) Comparing MZMG to M1MG, A pi = 3.06, A df= 3, and A CFI = 0.000. Intentions (four-factor model) MOWomen 133.83 0.000 59 0.86 0.892 0.09 0.09 (0.07; 0.1 l ) MOM-m 128.71 0.000 59 0.88 0.906 0.08 0.09 (0.07; 0.1 1) MI W 265.30 0.000 124 0.88 0.902 0.08 0.08 (0.07; 0.10) MZMG 274.27 0.000 133 0.88 0.902 0.09 0.08 (0.07; 0.10) Comparing MZMG to MIMG, A pf = 8.97, A df= 9, and A CFI = 0.000. Note. df= degrees of freedom; NNF I = nonnormed fit index; CFI = comparative fit index: SRMR = standardized root-mean-square residual; RMSEA = root mean square error of approximation. N a 327 (with n = 162 women and n = 165 men) for all multiple group analyses except for Social Guidance analyses, where N = 299 (with n = 151 women and n = 148 men). 147 For individual-level justice expectations, race centrality, African centrality. group general restorative justice expectations, and intentions, the invariance tests did not indicate meaningful lack of invariance across gender when considering CFI and chi- square change from Model 1 to Model 2. For support for affirmative action the change in CFI was 0.003, which just exceeds Meade, Johnson and Braddy’s (2008) recommended cutoff of 0.002, suggesting meaningful change. However, there was not a significant change in chi-square. Similarly, for merit orientation the change in CF I of 0.008 exceeded the recommended cutoff, yet the change in chi-square was not significant. Also. group belief in white guilt and support for affirmative action showed no significant change in chi-square but a small CFI change of 0.003. Social guidance showed a significant chi square change (A x2 = 9.28, A df= 3; p = .05) as well as a CFI change of 0.006. It appears that there is minimal evidence of lack of invariance across gender for the models tested here, with the biggest concern being with social guidance. However. as acknowledged by Meade, et al. (2008), lower standards may be accepted if the data are being used for correlational purposes such as in the present study, which is not a high stakes context. Thus, I will proceed with subsequent hypothesis testing using all of the measures discussed here. Future examination of the structure of the social guidance scale within larger groups of men and women may be informative regarding reasons for potential noninvariance. Race I also tested scale measures for invariance across the three largest race groups in the sample (black African, n = 102), Indian, n = 79, and white, n = 99) which would have at least marginal sample size for potentially detecting lack of invariance (Meade. 148 Johnson, & Braddy, 2008). For testing invariance across race I followed the same procedure as that used for testing invariance across gender. First, the model was tested separately in each race group (Model 0), and then multiple group analysis was conducted with a comparison of Model 1 and Model 2. Model 1 specified the same pattern of factor loadings and uniquenesses across groups, and allowed for unique parameter estimates. In Model 2 for each construct, the unstandardized factor loadings were constrained to be equal across race. As Models 1 and 2 are nested, change statistics from Model 1 to Model 2 are provided. Results of these analyses are shown in Table 18. Table 18 Comparison of Models Across Race Structure )8 p df NNFI CFI SRMR RMSEA RMSEA confidence interval Individual—Level Justice Expectations (seven-factor model) MOBrack 654.51 0.000 329 0.90 0.913 0.08 0.10 (0.09;0.11) African Mon-d...- 564.19 0.000 329 0.94 0.949 0.08 0.09 (0.07;0.10) Mow... 709.63 0.000 329 0.95 0.955 0.07 0.10 (0.09;0.11) MIMG 1928.33 0.000 987 0.94 0.944 0.07 0.10 (0.09;0.10) MZMG 1994.43 0.000 1029 0.94 0.943 0.08 0.09 (0.09.010) Comparing MZMG to MIMG, A X = 66.10, A df= 42 (12.-mm. significant at [7 = .05), and A CFI = 0.001. Race Centrality (one-factor model) MOBiack 10.50 0.005 2 0.84 0.946 0.06 0.20 (0.09;0.33) African MOM-an 2.16 0.339 2 1.00 0.998 0.03 0.04 (0.00; 0.23) MOWhi-e 9.94 0.007 2 0.73 0.909 0.08 0.20 (0.09; 0.33) M1 M0 22.60 0.001 6 0.86 0.953 0.08 0.17 (0.10; 0.25) MZMG 37.39 0.000 12 0.89 0.927 0.12 0.15 (0.09; 0.20) Comparing MZMG to M1140, A X2 = 14.79, A df= 6 (xzdifference significant at p = .05), and A CFI = -0.026. African Centrality (one-factor model) MOBtack 1.76 0.415 2 1.00 1.000 0.02 0.00 (0.00; 0.19) African 149 Table 18 continued. MOI-“ii... 6.1 5 Table 18 continued. MOM-tic 7.72 M 1 MG 1 5.63 MZMG 30.36 0.046 0.021 0.016 0.002 2 2 6 12 0.92 0.93 0.95 0.95 0.974 0.975 0.982 0.966 0.04 0.04 0.04 0.13 0.16 0.17 0.13 0.12 (0.01; 0.32) (0.06; 0.31) (0.05; 0.21) (0.06; 0.18) Comparing M2...G to mm... A x2 = 14.73, A df= 6 (fa-a.,“... significant at p = .05). and A CFI = -0.016. Support for A flirmative Action (one-factor model) MOBiacr- 33.86 0.000 5 0.79 0.897 0.08 0.24 (0.16; 0.32) African Mom...- 15.45 0.009 5 0.84 0.919 0.07 0.16 (0.07; 0.25) MOWhite 5.82 0.324 5 1.00 0.998 0.03 0.04 (0.00; 0.15) MIMG 55.13 0.000 15 0.89 0.947 0.03 0.17 (0.12; 0.22) M2 MG 63.48 0.000 23 0.93 0.947 0.06 0.13 (0.09; 0.18) Comparing MZMG to M1MG, A / = 8.35, A df= 8, and A CFI = 0.000. Merit Orientation (one-factor model) MOBlack 0.46 0.795 2 1.24 1.000 0.02 0.00 (0.00; 0.12) African Mop-d...- 4.98 0.083 2 0.75 0.916 0.06 0.14 (0.00; 0.30) MOWhite 4.14 0.126 2 0.87 0.957 0.05 0.10 (0.00; 0.25) Ml MG 9.58 0.143 6 0.90 0.966 0.05 0.08 (0.00; 0.17,) M2370 14.17 0.290 12 0.97 0.979 0.05 0.05 (0.00; 0.12) Comparing MZMG to MIMG, A X2 = 4.59, A df= 6, and A CFI = 0.013. Social Guidance (three-factor model) MOBlack 31.84 0.000 6 0.84 0.935 0.04 0.23 (0.15; 0.30) African MOM-an 5.75 0.452 6 1.00 1.000 0.03 0.00 (0.00; 0.15) MOWhite 15.07 0.020 6 0.92 0.968 0.04 0.12 (0.03; 0.20) M1 M0 52.66 0.000 18 0.91 0.964 0.04 0.15 (0.10; 0.20) M2MG 66.68 0.000 24 0.92 0.956 0.06 0.15 (0.10; 0.19) Comparing M2MG to M1 M0, A )8 = 14.02, A df= 6 ()(7 difference significant at p = .05), and A CFI = -0.008. Group Belief in White Guilt and Support for A flirmative Action (two-factor model) MOBlack 48.1 I 0.000 African MOIndt-m 37.88 0.006 MOW“;-e 13.26 0.825 M 1 MG 99.25 0.000 19 19 19 57 0.88 0.87 1.03 0.93 150 0.921 0.914 1.000 0.954 0.08 0.10 0.04 0.04 0.13 0.1 l 0.00 0.09 (0.09; 0.17) (0.05; 0.16) (0.00; 0.06) (0.06; 0.12) Table 18 continued. 119.15 0.000 MZMG 69 0.93 0.945 0.07 0.09 (0.06; 0.1 1) Comparing MZMG to MIMG, A x2 = 19.90, A df= 12, and A CFI = -0009. Group General Restorative Justice Expectations (one-factor model) MOBlack African MZMG 21.86 10.00 25.54 57.40 70.56 0.000 0.007 0.000 0.000 0.000 2 2 2 6 12 0.7 0.38 0.59 0.62 0.78 0.898 0.794 0.863 0.873 0.856 0.07 0.09 0.11 0.11 0.11 0.33 0.22 0.32 0.30 0.22 (0.22; 0.45) (0.09; 0.37) (0.21; 0.45) (0.23; 0.37) (0.17; 0.28) Comparing M2MG to M1 pp, A )6 = 13.16, A df= 6 (imp... significant at ,7 = .05), and A CFI = -0017. Intentions (four-factor model) MOBlack 119.24 0.000 59 0.82 0.866 0.10 0.10 (0.07; 0.12) African " MOI-mm. 112.49 0.000 59 0.78 0.832 0.10 0.10 (0.06; 0.13) MOWhite 109.36 0.000 59 0.87 0.901 0.09 0.09 (0.07; 0.12) MIMG 348.63 0.000 189 0.85 0.875 0.09 0.09 (0.08; 0.1 l) M2110 387.67 0.000 207 0.84 0.859 0.12 0.09 (0.08; 0| 1 ) Comparing MZMG to M1 M0, A )6 = 39.04, A df= l8 (idin‘crcncc significant at p = .01), and A CFI = -0.016. Note. df= degrees of freedom; NNFI = nonnormed fit index; CFI = comparative fit index: SRMR = standardized root-mean-square residual; RMSEA = root mean square error of approximation. N I 280 (with n = 102 black African, n = 79 Indian and n = 99 whites) for all multiple group analyses except for Social Guidance analyses, where N = 255 (with n = 89 Black African, n = 71 Indian, and n = 95 Whites). For support for affirmative action the invariance tests did not indicate meaningful lack of invariance across race groups when considering CF I and chi-square change from Model 1 to Model 2. Individual-level justice expectations showed a significant change in chi-square (A )8 = 66.10, A df= 42; p = .05) only. Race centrality showed a significant change in chi-square (A )8 = 14.79, A df= 6;p= .05) and a change in CFI of 0.026. African centrality showed a significant change in chi-square (A )(7 = 14.73. A df= 6: /) 2 .05) and a change in CFI of 0.016 across race groups. Merit orientation showed a CH 151 change of 0.013 while group belief in white guilt and support for affirmative action showed a CFI change of 0.009. Social guidance showed a significant chi-square change (A X2 = 14.02, A df= 6; p= .05) as well as a CFI change of 0.008. Group general restorative justice showed a significant change in chi-square from Model 1 to Model 2 (A )8 = 13.16, A ray: 6; p= .05) combined with a CFI change of 0.01 7. Intentions showed a significant chi-square change (A )(2 = 39.04, A df= 18; p=n .01) as well as a CFI change of 0.016. Although some of the changes in CFI across race groups here exceed Meade, Johnson and Braddy’s (2008) recommended cutoff of 0.002, Meade et al. acknowledge that their cutoff was based on analysis of two groups, and it is unclear whether this cutoff is appropriate for more than two groups. As I have three race groups, and the context for the research results here is not a high stakes one, I will accept lower standards and proceed with using all of the measures discussed above in subsequent hypothesis testing. Future examination of the structure of the scales that have indicated potential noninvariance here across race groups would be recommended within larger samples for each race. A careful examination of fit of the model within each group may facilitate better understanding of where differences lie. As we have multiple home languages spoken by participants in this study, home and language and race are related, and the survey was administered in English, some level of lack of invariance across race is not surprising. However, there may be interesting cultural differences in meaningfulness of these constructs (such as race centrality and African centrality. for example) in different race groups that should be explored in future research. 152 Multivariate To retain the maximum number of cases for subsequent analyses and hypothesis testing, the co-worker facet of social guidance was dropped. Twenty-eight participants had not responded to co-worker guidance items, presumably because they did not have co-workers, and as responses regarding the influence of friends and family were highly correlated, the latter two facets are combined to form the Social Guidance scale. Thus. N available for subsequent hypothesis testing is 327, however, of those participants, ten (five women and five men) did not identify themselves as being members of one of the four major race groups in South Africa. Because many of the subsequent hypotheses are related to major race group membership, and I have no specified hypotheses for participants who are of other races, responses from these ten participants are not considered in the subsequent modeling. Thus, N = 317 for subsequent hypothesis testing. Hierarchical linear modeling (HLM) was used to assess the mixed-determinant model with individuals nested within relevant race and gender groups as HLM is well suited for such research questions (Hofmann, Griffin, & Gavin, 2000). My model has three of the stated characteristics of hierarchical linear models (Hofmann, Griffin, & Gavin, 2000): my lowest level (individuals) is nested within higher-level groups (race/gender); the individuals are influenced by higher-level characteristics (group membership and group norms); and the outcome variables are measured at the individual level. To consider the effect of group membership on the relationship between predictors and dependent variables at Level 1, 1 simultaneously test for significant between-group variance in Level 1 intercepts and slopes. In using HLM for these analyses, race and gender are simultaneously tested as moderators of the relationship between antecedents 153 and expectations, rather than by using a product term as would be done in regression analysis. To consider the contribution of the group level factors to the explanation of variance in individual-level dependent variables, I will examine the incremental variance accounted for by the group-level factors over and above the individual-level factors. I have used a 2-level hierarchical model where at the individual level, N = 317, and at the group level there are eight race/gender combination groups. I have only eight groups and some of those have a small 71, which limits power for demonstrating group effects. and considering that this research tests new concepts that have not been demonstrated elsewhere, and as the consequences of Type I error are not dire here, I consider p < .10 as reasonable, which raises the power for demonstrating group effects. Individual Justice Expectations To test Hypotheses 2, 3, 5, and 6, and demonstrate between group variance for individual-level justice expectations, a fully unconditional HLM model for each of the individual-level justice factors was run. That is, the model specified (in equation format): Level-l Model is Y)!- = [30]- + rij and Level-2 Model is 601- = m; + up]. Multilevel random coefficient regression results for each of the individual-level justice dimensions are shown in Table 19, demonstrating variance in intercepts. TIN). Mean levels on each dimension for each group, along with effect sizes. are shown in Table 20. The individual-level justice dimensions that showed significant between-group variance are Procedural (with between-group variance of 6%), Procedural Restorative —~ Women (3%) and Distributive Restorative — Blacks (9%). These dimensions saw the largest differences in mean levels when the group mean is compared to that of a reference 154 group, in this case, that of white men. (White men are an appropriate reference group here as they are not affirmative action beneficiary targets.) As expected, white men had the highest level of general restorative justice expectation (mean level = 3.98). While I had anticipated that black Africans would have the lowest expectations on this dimension, it was coloured women (mean = 3.27. d = - 0.66) who had the lowest level of general restorative justice expectation, followed by black African women (mean = 3.44, d = -0.50). Variance in intercepts for general restorative justice was not significant. Hypothesis 2, that restorative justice expectations will vary significantly across race/gender groups with black Africans having the lowest expectations of restorative justice, and white men having the highest expectations of restorative justice, was not supported. White women had the lowest expectation of distributive justice (mean = 3.47, d = -0.33), counter to what was anticipated. Coloured men (mean = 3.68) were the next lowest group. The groups with the highest expectation of distributive justice were coloured women (mean = 3.97, a' = 0.28) and Indian women (mean = 3.92. d = 0.22). They were followed by black African men (mean = 3.82), black African women (mean = 3.81) and Indian men (mean = 3.80), all at roughly the same level. However. variance in intercepts for distributive justice was not significant. Black African and Indian women showed the highest level of procedural justice expectations (3.93 and 3.77 respectively), with large respective d-values of 0.76 and 0.74. The group with lowest procedural justice expectations, as anticipated, was white men (mean = 3.21). However, white women also showed low mean procedural justice expectations (mean = 3.29), counter to what was anticipated. 155 Hypothesis 3 stated that distributive justice and procedural justice expectations will vary significantly across race/gender groups, with white men having the lowest expectations of procedural and distributive justice, and white women having the highest expectations of procedural and distributive justice. Also, it was stated that expectations of procedural and distributive justice among black Africans will be lower those of coloureds, which will be lower than those of Indians. Thus, Hypothesis 3 was partially supported for procedural justice. Black African men (mean = 3.84, d = 0.27) and Coloured men (mean = 3.83. d = 0.26) showed the highest levels of expectation regarding procedural restorative justice for blacks. White women showed the lowest expectation on this dimension (mean = 3.29. d = -0.25). For distributive restorative justice for blacks, black African men had the highest expectations (mean = 4.12, d = 0.47), counter to what was anticipated. Hypothesis 5 stated that distributive restorative and procedural restorative justice expectations for treatment of blacks will vary significantly across race and gender groups with black African men having the lowest expectations of distributive restorative and procedural restorative justice, lower than the expectations of coloureds, which will be lower than those of Indians. Indian women will have the highest expectations of justice. Thus, Hypothesis 5 was partially supported for distributive restorative justice for blacks. Both Indian women (mean = 4.09, d = 0.67) and coloured women (mean = 4.03. d = 0.61) showed high expectations of procedural restorative justice for women. Counter to what was anticipated, white women (mean = 3.49) were the women’s group with the lowest expectations on this dimension. Indian women (mean = 4.03. a’ = 0.42) showed the highest level of expectation regarding distributive restorative justice for women. C ounter 156 to what was anticipated, white women showed the lowest expectation on this dimension (mean = 3.62). Hypothesis 6 stated that distributive restorative and procedural restorative justice expectations for treatment of women will vary significantly across race and gender groups, with black African women having the lowest expectations of distributive restorative and procedural restorative justice, lower than the expectations of coloureds, which will be lower than those of Indians. White women will have the highest expectations of justice. Thus, Hypothesis 6 was not supported. 157 ood cod \- ofio co. _ m 3005 - 0308083” 03:8:me Ed _N.o \- :.o 00.0 3005 - 03800832 8500005 56 omd .8 mod ww.w c0803 -. 0380083“ 0350885 mod mod n w 2 .o m _ .5 00803 - 0380080m 8.500005 mod N _ .o n v _ .o em . _ _ 0308083“ 30:00 56 mmd N- wod 24.0 0380885 cod cod 5 mmd wwwm 8800005 00§_8> 030>d Xe QM. NR 0080085 :03 -000300m 80.0225 5 0050.208 1 =08§§Q 00.55- 3884008332005 «00m .5\ 830.0% 35028 3833: 3 033- 158 000%: $30.23 m go 0205 0 010003030 bmseha _ mo 020> 0 88.. 0mg: 2:00 00:08 8.00m £029.28 :283200 :8 anew 00:20.3: 05 mm 0>00m :0E 0:55 .0002 85502 8.? 8.0- -:.o ._ Rd- :5 :3 :0 0 03 Sm S: ta 00m 02 NE. 03. E 38.0-3523: 30350: 8:232 8.0- 20 NS 80 00¢ 20 NS 0 02 00m 000 w: ta. :2 :2 03 E 38.0-2528: 350820 85502 Ed- 03 as 80. 8o 02 $5 0 ohm N3 gm 8.0 03“ 2: 3m 3m 3 8603-802930 2:350: 8:202 8.0 02 So mod 6.0 02 moo 0 3m 9% Ra. 8.0 02 8+ 93 a; 2 5503020238035820 8552 mg- 3.0- 02- 000. 00.? 92- one- 0 03 :0 a; 03. Ra Rm 000 3% E 3:298: .2050 8552 m2- So mg 80. 00¢ 20 3o 0 0.2 5% 23 N3. 03 SM gm :3. 3 96330: 85502 85 03 Ed and 8.0 $0 03 0 30 an Sm Sm RM 32 £0 30 E 350880 :02 :0803 :02 :0803 :02 :0803 :02 :0803 223 323 5:3 E2: 8528 8328 53:2 98:2 0.020 0320 0&396 0.0.3000 05002 220:0ENQ 00203. cm 030... 159 To test Hypothesis 4, and demonstrate between group variance in Level 1 slope. I ran an HLM model examining the relationships between general restorative and procedural expectations, and general restorative justice and distributive justice expectations. That is, the model specified (in equation format): Level-1 Model is Y” = [$0, + [3,] (General Restorative Justice)+ rij and Level-2 Model. is [30]- : m; + u”, and :81} = 710 + u U- Table 21 shows results for each of the two individual-level justice dimensions of interest here. This model showed that the relationship between general restorative justice expectations and procedural justice expectations (T; I) did vary significantly across groups (X) = 19.36, df= 7, p < .01) as was true for the relationship between general restorative justice expectations and distributive justice expectations (X7 = 37.39. df= 7. p < .01 ). Table 21 Multilevel Analysis Results for Each Individual-Level Justice Dimension — Variance in Intercepts and Slopes Between- Unit Dimension )(2 SD df P-value Variance Procedural Intercept 22.62 0.20 7 0.00 0.04 Slope 19.36 0.19 7 0.01 0.04 Distributive Intercept 8.99 0.06 7 0.25 0.00 Slope 37.39 0.23 7 ' 0.00 0.05 160 Table 22 shows the intercepts and slopes for each of the groups, for both procedural and distributive justice. For general restorative justice and procedural justice. the groups with the steepest negative slopes are Indian men and women and white men and women. These same groups showed the strongest negative relationships between general restorative justice and distributive justice. Table 22 Eflect of General Restorative Justice on Procedural and Distributive Justice Procedural Distributive Intercept Slope Intercept Slope Black African Women 0.31 0.10 0.09 0.18 Back African Men 0.22 -0.16 0.06 0.09 Coloured Women 0.11 -0.26‘ 0.1 l -O.27 Coloured Men 0.16 -O.23 -0.07 -0.08 Indian Women 0.27 -0.53 0.16 -O.45 Indian Men -O.15 -0.59 -0.0l -0.37 White Women -0.31 -O.47 -0.24 -O.51 White Men -O.27 -O.48 0.10 -O.36 Note. Variables were grand mean centered. Figure 5 provides a graphic representation showing" the within-group slopes for general restorative justice and procedural justice, where Indian men show the steepest negative slope, followed by Indian women, white men, and white women. Figure 6 provides a graphic representation showing the within-group slopes for general restorative justice and distributive justice, where white women show the steepest negative slope. followed by Indian women, Indian men, and white men. 161 4.50 . o \ a . 2 : i! 3.50 e g . 5. . Low (-1 SD) High (+1 SD) General Restorative Justice -0- Black African Female -I- Black African Male +Coloured Female +Coloured Male -II- Indian Female -0- Indian Male -i—Whlte Female — White Male Figure 5 W it‘hin-Group Slopes for General Restorative Justice and Procedural Justice 4.50 5 O 8 3 . fl . g 3.50 1 3 E o 2.50 5 Low (-1 30) High (+1 SD) General Restorative Justice -0- Black African Femde -l- Black African Male + Coloured Female -n- Coloured Male -I- Indian Female -0- Indian Male d-White Female -— White Male Figure 6 Within-Group Slopes for General Restorative Justice and Distributive Justice 162 For procedural justice the difference in intercepts and slopes was supported. and for black African women the slope was positive, and negative in all other groups. For distributive justice the difference in intercepts and slopes was supported, and for black African men and women the slope was positive, and negative in all other groups. Hypothesis 4, that the relationship between restorative justice expectations and procedural and distributive justice expectations will vary significantly across race/ gender groups with white men having an inverse relationship between restorative justice expectations and procedural and distributive justice, while white women, black Africans. coloureds and Indians will show a positive relationship between restorative justice expectations and expectations of procedural and distributive justice, was partially supported for each of the two justice dimensions of interest here. Group Cultures As labor law in South Africa specifies racial and gender target groups for affirmative action, I have been conducting analyses using membership in race/gender groups as grouping variables. Individuals within the same group are set up for shared experience in the selection process as they have a shared target or non-target status. They also have a shared history in terms of their degree of privilege under apartheid government. As discussed earlier, I posit that cultures with regard to belief in white guilt. support for affirmative action, and restorative justice expectations may exist. Studying the role of the group context on the actions and attitudes of individuals makes explicit the group characteristics that contribute to variance in the dependent variable. While the disadvantage of using this context information is that it has less predictive power. it is useful in that it provides explicit information about the group-level factors contributing to 163 the model (F irebaugh, 1979). This is relevant in the case of my model, where I am interested in understanding group norms as predictors of individual-level reactions. Thus. the group mean for each of the group-level factors will be entered into the equation in HLM analyses. It could be that the social norm of the group is as powerful as, or even more powerful than, the individual belief in predicting reactions. Aggregation. As I am examining the contextual effects of culture of white guilt. culture of affirmative action support, and culture for restorative justice expectations. I first evaluated whether it is appropriate that I should aggregate responses to capture the group level construct. For purposes of this evaluation, ICC(l), I C C(2) and r“.g values were calculated for belief in white guilt, group culture of support for affirmative action. and group-level restorative justice expectations. Table 23 shows the results of these analyses. Table 23 Group Culture Aggregation Indices Group Culture F ICC ( 1 ) [CC (2) rwg Range r“.g Mean Support for Affirmative 43.71* .52 .98 .68 — .85 .79 Action General Restorative 2.21 * .03 .55 .60 — .82 .74 Justice Expectation ' Belief in White Guilt 23.07”“ .36 .96 .25 — .62 .38 *Significant at p < .01. ICC(l) shows that there is a group effect on all three of the constructs of interest. That is, there is greater between-group variance relative to total variance, as indicated by a significant F-test, and I view this as an appropriate indication to support aggregation (see Klein & Kozlowski, 2000). ICC(2) values of .98 for support for affirmative action 164 and .96 for belief in white guilt certainly exceed the .7 rule of thumb often used as a cutoff value, and I am satisfied that there appears to be differentiation in group means and reliable between group differences here, so aggregation is appropriate (see Klein & Kozlowski, 2000). Further, using a value of .7 as a rule of thumb indicator of agreement (see Klein & Kozlowski, 2000), the average rwg values for support for affirmative action and general restorative justice demonstrate within-group agreement on these constructs. however, relying on the uniform null distribution may yield an inflated estimate of agreement (Bliese, 2000). Therefore I will regard these rwg values as an upper bound. While there are rules of thumb for these indicators, these are not hard and fast rules, and the literature includes studies where in some cases these levels were not met. but the authors aggregated their data and conducted their subsequent analyses all the same (e.g., Ehrhart, 2004; Hofmann & Jones, 2005). In the present study, individual cut— off points are not met in every case. Although for general restorative justice the ICC (l ) is low (.03) and the ICC(2) value (.55) does not meet the rule of thumb cutoff. there does appear to be within-group agreement (although limited difference in group means) so I will proceed with aggregation for group general restorative justice expectations. While the rwg values for belief in white guilt (average of .3 8) do not meet the rule of thumb cutoffs, in considering the pattern of ICC(l), ICC(2) and rwg results, I believe that the overall pattern for belief in white guilt suggests that aggregation is supported because there is differentiation in group means, even though there is not high within-group agreement. Although there may not be a strong shared view within the group. the different means within groups remain relevant as there are meaningful group differences in the mean levels. Thus, although I continue with subsequent analyses having aggregated 165 responses on these data, interpreting the conclusion to aggregate as a confirmation that this is a group construct would be overreaching. Further examination of this issue will be necessary. Regardless, where there are low aggregation values, relationships with the aggregated variables will be attenuated, providing less powerful tests of hypotheses regarding relationships with these variables. Figure 7 shows the level of culture of support for affirmative action within each group. The height of the columns indicating support for affirmative action parallel a hierarchy of historic disadvantage — on the far left are black African women and men, who have experienced the greatest historic disadvantages and the greatest support for affirmative action (although not high), followed by coloured and Indian people, who have been relatively more advantaged than black Africans, but less advantaged than whites. Whites have experienced the greatest advantage and a lackgof support for affirmative action. 166 5.00 a 8 _8 ‘8 ( E 5 3.00 3? i (1:) N O O i i i E E . i i i j l l l i i 1.00 _. ’ y ., 7 . . Black Black Coloured Coloured Indian Indian White White Male African African Female Male Female Male Female Female Male Figure 7 Group Culture of Support for A ffirmative Action Figure 8 shows the level of the general restorative justice expectation culture within each group. The height of the columns indicating the expectations of general restorative justice do vary across groups, but the group differences are small. and all groups have an expectation that there will be general restorative justice. 167 5.00 l l l E l * ; l g . 1.00; ~ -- . . - ~ w * Black Black Coloured Coloured Indian Indian White White Male African African Female Male Female Male Female Female Male to a 8 8 General Restorative Justice Expectation re 8 Figure 8 General Restorative Justice Expectation Culture Figure 9 shows the level of group belief in white guilt within each group. The height of the columns indicating belief in white guilt, like support for affirmative action. parallel a hierarchy of historic disadvantage — on the far left are black African women and men, who have experienced the greatest historic disadvantages and who show extremely high levels of belief in white guilt, followed by coloured and Indian people. who have been relatively more advantaged than black Africans, but less advantaged than whites, and who show moderate levels of belief in white guilt. Whites have experienced the greatest advantage and toward neutral regarding belief in white guilt. 168 5.00 » l l 4.00 ., i =5 l ' 0 o l a: l .C l g 3.00 l, .E l .5 E 8 i 2.00 l i l l Black Black Coloured Coloured Indian Indian White White Male African African Female Male Female Male Female Female Male Figure 9 Group Belief in White Guilt Group-Level Relations Having aggregated the appropriate data to the group level, next I tested the significance of purely group-level relations by examining the generalized least squares estimates of Level 2 regression parameters. Specifically, to test Hypothesis 7. that group belief in white guilt will be positively related to group culture of support for affirmative action, Level 2 regression was conducted. Group belief in white guilt was strongly associated with group culture of support for affirmative action (unstandardized beta = .74. t(325) = 71.80, p < .01), supporting Hypothesis 7. 169 Individual Justice Expectations as the Dependent Variable Next, I examined mixed models as necessary for hypothesis testing related to each justice expectation as an outcome. First, results for procedural justice are provided. Procedural justice. Table 24 provides the fixed effects results and Table 25 provides the random effects results for this dependent variable. Hypotheses 8. 9, and l(). were tested, demonstrating the effects of a group culture of support on level-l outcomes. That is, the model specified (in equation format): Level-l Model is Y1]- = [3,)!- + r),- and Level-2 Model is fioj = Yon + 701 (Group culture of support for affirmative Action) + 10),. Table 24 Multilevel Analysis Resultsfor Procedural Justice — Signyicant Pooled Slime __ Standard P- Dimension Coefficient Error T-ratio (if value 0.41 0.14 3.00 0.03 Y0] Culture of Support for AA -0.04 0.47 -0.08 0.94 yng Restorative Justice Culture ' 0.07 0.07 0.89 0.40 ym Race Centrality -0.01 0.12 -0.09 0.93 ygn African Centrality -0.05 0.09 -0.55 0.60 730 Support for AA -0.06 0.13 -0.50 0.63 740 Preference for Merit Principle - -0.26 0.09 -2.84 0.03 m Social Guidance - Unfair 0.04 0.09 0.46 0.66 no Educational Integration Note. Variables were grand mean centered. Hypothesis 8 stated that group culture of support for affirmative action would add to the variance explained in individual procedural justice expectations, and group culture of support for affirmative action will be more strongly related to procedural justice 170 expectations for members of groups where there is a culture of lack of support for affirmative action. Group culture of support for affirmative action (34;, = 0.41, t = 3.00, p < . 05) explained significant variance in procedural justice, but the slopes did not vary significantly across groups. Thus, Hypothesis 8 was partially supported. Although there was not a significant pooled slope, the relationship between African centrality and procedural justice varied significantly across groups (x2 = 26.10. df= 7, p < .01). Figure 10 shows the within-group slopes for African centrality and procedural justice. The strongest negative slope was for white women, where high African centrality was linked to low procedural justice. In contrast, coloured women with high African centrality indicated higher procedural justice expectations. 4.50 0 .9 ti 3 2 8 3 g 3.50 5 o. . 2.50 3 » Low (-1 SD) High (+1 SD) African Centrality -0- Black African Female -I- Black African Male + Coloured Female -N- Coloured Male -II- Indian Female -O- Indian Male -'-White Female —- White Male Figure 10 W ithin-Group Slopes for African Central ity and Procedural Justice 171 Hypothesis 15 states that the relationship between race centrality and procedural justice expectations would vary across race groups, however, there were not significant group differences in these slopes and Hypothesis 15 was not supported. l'lypothcsis l6 states that African centrality would moderate the effect of culture of support for affirmative action on procedural justice expectations. To test this hypothesis the interaction term was added to the model, but no significant interaction effects were found, and Hypothesis 16 was not supported. Hypothesis 18, that individual support for affirmative action would be positively related to procedural justice expectations, was not supported. Hypothesis 20, which stated that preference for merit principle would be negatively related to procedural justice expectations, was not supported. Hypothesis 24. that educational integration will be negatively related to procedural justice expectations. was not supported. Hypothesis 26, that educational integration would moderate the effect of group culture of support for affirmative action on procedural justice was not tested as there was not a main effect for each variable. Hypothesis 28, that social guidance ~ unfair — would be negatively related to procedural justice expectations was supported (rm = - 0.26, z = -2.84, p < . 05). 172 .8328 :38 2me 203 333:; .292. Be one- Ed mod- 33. Rd 8o mod 8:235 1222835 :8. on? So. «no- one- 25. . So 25. .323- 8525308 :.o- 86 ”we- 36 26. «no 86. mg- 2285 :52 E 8:235 :.o- 2 .o one- 26 «we 35 one. N2- <<é e253 5 _ .o- one- 33 so. So was 86. 86 £250 50:2 :.o- 86 So- :3 8d- sod Ed 86 E828 8mm as- 23. 8.? go o; m; 25 Rd 385:: 2&2 0_NEOHH 01w: o—NEUD 0E2 vacuum 262 OEEOK Gommdoemm 223 233 565 565 8530 8:530 §E< €022 seam 0.85 82.3% Nexzheocxm SKwEmBmefimcD TESQ mzcgbémfitx mm 2an 173 Distributive justice. Next are results for distributive justice. Table 26 provides results for the pooled slope, and Table 27 provides within-group coefficients. Hypothesis 9 stated that group culture of support for affirmative action would add to the variance explained in individual distributive justice expectations, and group culture of support for affirmative action would be more strongly related to distributive justice expectations for members of groups where there is a culture of lack of support for affirmative action. There was no effect for culture of support for affirmative action on distributive justice. thus Hypothesis 9 was not supported. Table 26 Multilevel Analysis Results for Distributive Justice — Significant Pooled Slope Standard P— Dimension Coefficient Error T-ratio elf value rm Culture of Support for AA -013 0.09 -1 .43 5 0.21 ya; Restorative Justice Culture 025 0.28 0. 89 5 04] 710 Race Centrality 0.09 0.09 1.06 7 0.33 anAfricau Centrality 0.00 0.11 0.01 7 0.99 730 Support forAA -0.08 0.06 -1.31 7 0.23 ym Preference for Merit Principle 0.05 0.10 0.54 7 (”)0 Km Social Guidance - Unfair _0. 18 0.08 -2.18 7 0.07 ya; Educational Integration 0.04 0.09 0.49 7 0.64 Note. Variables were grand mean centered. 174 622:3 Son Baum 22$ 3388> $3.2... 032m xoflm omd- emd- wed 5d mmd Cd 52o mmd 20:83:: Ecocmosnm ., med- wed- Ed- 36. N. _ uo. mmd mod Cd- :flcDi 8:3sz 300m mmd- Nod- mod- ovd N _ .o omd 56 «md 2905.5 :52 e8 032$on :d- 8.0 o _ .o- nod mmd Ed de- mod << 8% tomasm ood- mmd- :6 de- wed- med vmd Rd- b28300 GSE< :6 wmd :6- 3‘6 5.0- mmd- m _ .o- vmd bzmbcou comm Nod mod mod vod- m _ .o- de mmd cod- E0885 23g Bmfiom 232 29:0,; 032 0380““ 232 oREom comma—2:5 853 833 SE3: .865 35200 8.5200 Son< SNQE< 35:5 mésfihflQ xcsxfltfiugeob :23 onxDéNSSA mm 2an 175 Although there were not significant pooled slopes at p < .05 for the relationship between any of the antecedents and distributive justice, social guidance did show a marginally significant effect (y50 = -0.18, t = -2.18, p < .10). There were significant group differences in slopes for African centrality ()f = 34.78, df= 7, p < .01). social guidance (X2 = 16.15, df= 7, p < .05) and educational integration ()8 = 14.98. a’f: 7. p < .05). Figure 11 shows the within-group slopes for African centrality and distributive justice, where coloured women and Indian men with high African centrality appear more likely to expect distributive justice, whereas white women with high African centrality have lowered distributive justice expectations. 4.50 0 .9 ‘65 3 2 E 3 E 3.50 2.50 é . . Low (-1 SD) High (+1 SD) African Centrality -0- Black African Female -I- Black African Male -- Coloured Female -N- Coloured Male + Indian Female -0- lndian Male -¢-White Female - White Male Figure 1 1 W ithin-Group Slopes for African Centrality and Distributive Justice 176 Figure 12 shows the within-group slopes for social influence and distributive justice, where white women and men with high social influence regarding unfairness have lower expectations for distributive justice. 4.50 ‘ \ — K ‘ \. ‘_ g w‘ I 3 . ‘7 ,3 3.50 ‘5 % ,n 5 .92 D 2.50 i Low (-1 SD) High (+1 SD) Social Influence - Unfair -0- Black African Female + Black African Male -I- Coloured Female -N- Coloured Male -II- Indian Female -0- Indian Male -0- White Female -—White Male Figure 12 Within-Group Slopes for Social Influence and Distributive Justice Figure 13 shows the within-group slopes for educational integration and distributive justice where for most group there was convergence regarding distributive justice with high educational integration, although white women showed lower distributive justice expectations than other groups when having high integration. and Indian men showed higher distributive justice expectations than other group when having high integration. 177 4.50 a Distributive Justice on 8 2.50 , ., . , , Low (-1 30) High (+1 SD) Educational Integration -0- Black African Female + Black African Male + Coloured Female -*- Coloured Male + Indian Female -0- Indian Male -I-White Female — White Male Figure 13 W ithin-Group Slopes for Educational Integration and Distributive Justice Hypothesis 13 stated that the relationship between group-level restorative justice expectations and individual-level procedural and distributive justice expectations would vary across groups. As there were no significant group differences in slopes. Hypothesis 13 was not supported. Hypothesis 14 stated that the relationship between race centrality and distributive justice expectations would vary across race groups, however, there were not significant group differences in these slopes and Hypothesis 14 was not supported. Hypothesis 17 stated that African centrality would moderate the effect of culture of support for affirmative action on distributive justice expectations. As there were not significant main effects for the relevant variables, the interaction term was not added. Hypothesis 19, which stated that support for affirmative action would be positively 178 related to distributive justice expectations, was not supported. Hypothesis 21, that preference for merit principle would be negatively related to distributive justice expectations, was not supported. Hypothesis 25 stated that educational integration would be negatively related to distributive justice expectations. Although there was no significant pooled slope for educational integration, there were significant group differences in slope, with only white men and women showing negative (yet non- significant) coefficients for educational integration. Hypothesis 25 was not supported. Hypothesis 27 stated that educational integration would moderate the effect of group culture of support for affirmative action on distributive justice expectations. As there were not significant main effects for the relevant variables, the interaction temi was not added. Hypothesis 29 stated that social guidance — unfair — would be negatively related to distributive justice expectation. Social guidance only showed a marginally significant pooled slope. There were significant group differences in slope, with black African males and coloured females showing positive coefficients, while all other groups showed negative relationships — the direction anticipated. 179 Procedural restorative justice — women. There was no significant pooled slope for culture of support for affirmative action on procedural restorative justice — women (see Table 28). Nevertheless, there were group differences in slopes for African centrality (X7 = 34.29, df= 7, p < .01) and social influence of = 15.09, df= 7. p < .05). Table 28 Multilevel Analysis Results for Procedural Restorative Justice — Women Pooled Slope Standard Dimension Coefficient Error T-ratio aff P-value ya, Culture of Support for AA -0.03 0.12 -O.29 5 0.78 ya; Restorative Justice Culture -O.34 0.43 -0.78 5 0.47 YmRace Centrality -0.03 0.07 -039 7 0.71 Y20 African Centrality 0.01 0.12 0.07 7 0.95 Y30 Support for“ -0.06 0.06 -095 7 0.38 fig Preference for Merit Principle 0.01 0.09 0.12 7 0.91 7'50 Social Guidance - Unfair -O.14 0.10 . -1.36 7 022 1’60 Educational Integration _0_04 0_07 -0.64 7 0_ 54 Note. Variables were grand mean centered. 180 Figure 14 shows the within-group slopes for African centrality and procedural restorative justice - women, where high African centrality is tied to divergent levels of procedural restorative justice — women. For coloured women and Indian men, high African centrality was linked to a greater expectation of procedural restorative justice - women, whereas white women and coloured men with high African centrality have lowered procedural restorative justice expectations - women. 4.50 3.50 Procedural Restorative Justice - Women 2.50 5 . . 4 . . Low (-1 SD) High (+1 SD) African Centrality -0- Black African Female + Black African Male + Coloured Female -N- Coloured Male -I- Indian Female -0- Indian Male -1-White Female -— White Male Figure 14 W ithin-Group Slopesfor African Centrality and Procedural Restorative Justice - ”"0an 181 Figure 15 shows the within-group slopes for social influence regarding unfaimess and procedural restorative justice - women, where high social influence regarding unfairness is tied to lower expectation regarding procedural restorative justice — women for three groups; coloured and white men and white women. 4.50 c 0 1 g a a: 3 _8 i .2 a O i > 5 :3 3.50 e, E 5 e 1 3 a 95 E i 2 a. 2.50 - .. Low (-1 SD) High (+1 SD) Social influence - Unfair -0- Black African Female -I- Black African Male -- Coloured Female -N- Coloured Male -II- lndian Female -0- Indian Male -*- White Female —White Male Figure 15 W ithin-Group Slopes for Social Influence and Procedural Restorative Justice - ll’omen 182 Distributive restorative justice — women. As shown in Table 29, there was no significant pooled slope for the effect of culture of support for affirmative action (or other antecedents) on distributive restorative justice — women. There was a marginally significant slope for social guidance (y50= -0.20, t= -2.05, p < .10). Table 29 Multilevel Analysis Results for Distributive Restorative Justice — Women — Pooled Slope Standard Dimension Coefficient Error T-ratio elf P-value yo; Culture of Support for AA 0.02 0.10 0.21 5 0.84 yo; Restorative Justice Culture -O.15 0.31 -O.50 5 0.64 m Race Centrality -0.06 0.07 -0.84 7 0.43 7’20 African Centrality 0.10 0.12 0.89 7 0.41 3’30 Support forAA -0.08 0.06 -1.36 7 0.22 Y4!) Preference for Merit Principle 0.12 0.09 1.37 7 0.21 fig Social Guidance - Unfair _0.20 0.10 _2.05 7 0.08 m) Educational Integration _0.07 0.07 _0_95 7 0.38 Note. Variables were grand mean centered. 183 There were group differences in slopes for African centrality of = 33.81, df= 7, p < .01) and social influence (X7 = 15.14, df= 7, p < .05). There were marginally significant group differences in the race centrality slope (X7 = 12.20, df= 7, p < .10). Figure 16 shows the within-group slopes for African centrality and distributive restorative justice - women, where high African centrality is tied to higher expectation regarding distributive restorative justice — women for coloured women and Indian men. In contrast, high African centrality was associated with him distributive restorative justice expectations — women in the case of Indian and white women and white men. 4.50 3.50 Distributive Restorative Justice - Women 2.50 ~ — ~ 1 Low (-1 SD) High (+1 SD) African Centrality -0- Black African Female + Black African Male + Coloured Female -M- Coloured Male -Il- lndian Female -0- Indian Male -*-White Female -—White Male Figure 16 Within-Group Slopes for Afiican Centrality and Distributive Restorative Justice - Women Figure 17 shows the within-group slopes for social influence regarding unfairness and distributive restorative justice - women, where high social influence regarding 184 unfairness is tied to lower expectation regarding distributive restorative justice — women for coloured men, white women, and white men. 4.50 3.50 Distributive Restorative Justice - Women 2.50 4 . Low (-1 SD) High (+1 SD) Social Influence - Unfair -0- Black African Female + Black African Male + Coloured Female -N- Coloured Male -II- lndian Female -0- indian Male -t-White Female — White Male Figure 17 Within-Group Slopes for Social Influence and Distributive Restorative Justice - ll’omen Table 30 and Table 31 show the within-group coefficients for procedural restorative justice — women distributive restorative justice — women, respectively. 185 .Uuhoucou cam—.2 GEN-hm 0.53 m0_£m_.=w> 62C? mdd Ed- ddd mod. N _ d- 2d 8d- 2d. 8:535 deceased Ned- ded- 2d- 2d.- Ed- 2d 2d odd- . due: - 8525 38m 8d. 8d ad ad Rd- 3 d 3d- 8d eased “to: 5 Seed-ed 8d- 5d. Rd- Ed 3d mmd Ed- mmd- 5:8 225% 8d- ddd- 5d m _ d- edd- Qd Ed 8d £350 525. 8d. :d :d- and odd. Ed- Ed. 8d 3:888 83d 2d- 8d. 8d 3d ded- ddd 2d 3d asses 2m: 0380“— 032 Beach 032 295m 232 Bass-m commaoEE 223 323 565 sees 8328 descend game. got? seam 22:3: - 82.35 mean-52mm Nexzheoexk Le\&§.8m§meb TN / seem WSNV QxNQKb add cm £an 186 .3523 SEE Baum 0.53 832:; .222 :6- de- mod- duo-O nod- ood Ed mod- cotfim85 3:280st mmd- mod- .. 56. Ed mmd- Sud vod- Cd- awn-ED - eon-mgsmv Boom £6 ood N _ .o mod oed- mmd mod m _ .o 2305.5 E22 com mucus-«8m omd- vod- mmd- cod mod med wmd- Nod- << how tongsm ood- de- de cod- Dio- ood ovd id. 328300 Swot-Fm 3.0- Cd Rd- Ed Ed- omd- Cd- wfio ban-5:00 3mm 3.? mod. m fl .o mod 36. omd 26 mod 30885 222 228m 232 01“th 032 295m 232 oEEom 222585 353 83>? 563 53?: 35200 35200 gout?- Swot-fix seam seam 55$: - 85.3. eéuxeamx 3239525 SKmEeBmeeb Zea—e4 macaw-23$: fim oEm-r 187 Procedural restorative justice — blacks. There were no significant pooled slopes for the relationship between the various antecedents and procedural restorative justice — blacks (Table 32). Table 32 Multilevel Analysis Results for Procedural Restorative Justice - Blacks — Pooled Slope Standard Dimension Coefficient Error T-ratio df P-value 70! Culture of Support for AA -0.03 O. 12 -0.29 5 0.78 yo; Restorative Justice Culture -O.34 0.43 -O.78 5 0.47 710 Race Centrality -o.03 0.07 -o.39 7 0.71 Y20 African Centrality 0.01 0.12 0.07 7 0.95 Y30 Support for AA -0.06 0.06 -095 7 0.38 Y4” Preference for Merit Principle 0.01 0.09 0.12 7 0.91 Km Social Guidance - Unfair -014 0.10 -l.36 7 0.23 M50 Educational Integration _0.04 0.07 -0.64 7 O. 5 4 Note. Variables were grand mean centered. However, there were significant group differences in slope for African centrality (X7 = 34,29, df= 7,}; < .01) and for social influence (x2 = 15.09, df= 7,- P < 05)- 188 Figure 18 shows the within-group slopes for African centrality and procedural restorative justice - blacks, where high African centrality is tied to higher expectation regarding procedural restorative justice — blacks for coloured women and Indian men. In contrast, high African centrality was associated with lower procedural restorative justice expectations — blacks in the case of white women and coloured men. U) 'Q 4.50 E 8 '5 = 4 3 3 .9 a 2 <8 3 50 ._ I? o/ E i o. 250 L V _ V s. . , . . 1 Low (-1 SD) High (+1 SD) African Centrality -0- Black African Female + Black African Male -t- Coloured Female -N- Coloured Male -II- lndian Female -0- Indian Male -t-White Female — White Male Figure 18 W ithin-Group Slopes for A fiican Centrality and Procedural Restorative Justice - Blacks Figure 19 shows the within-group slopes for social influence regarding unfairness and procedural restorative justice - blacks, where high social influence regarding unfairness is tied to lower procedural restorative justice expectations — blacks in the case of white and coloured men and white women. 189 4.50 CD '5 1: g — in .§ 4.. 73' ‘.\"—" fi‘ 3 , 0 ‘i .> g ‘6 3.50 I g E E g t a 2.50 a, - . . .. Low (-1 SD) High (+1 SD) Social Influence - Unfair -0- Black African Female + Black African Male -t- Coloured Female -H- Coloured Male -Il- lndian Female -0- Indian Male *Miite Female —White Male Figure 19 W ithin-Group Slopes for Social Influence and Procedural Restorative Justice - Blacks 190 Distributive restorative justice — blacks. There were no significant pooled slopes for the relationship between the various antecedents and distributive restorative justice — blacks at p < .05 (Table 33). The pooled slope for social influence was marginally significant (y50= -0.22, t = -2.23,p < Table 33 .10). Multilevel Analysis Results for Distributive Restorative Justice — Blacks ~ Pooled Slope Standard Dimension Coefficient Error T—ratio P-value yo] Culture of Support for AA -0.01 0.14 -0.05 0.96 yo; Restorative Justice Culture 0.59 0.47 1.26 0.26 YmRace Centrality 0.01 0.09 0.14 0.90 720 African Centrality -o.04 0.13 -030 0.77 730 Support for AA 0.15 0.09 1.59 0.16 ym Preference for Merit Principle 0.09 0.12 0.73 0.49 y50 Social Guidance - Unfair _0.22 0.10 4.23 0.06 You Educational Integration -0.06 0.10 _0. 58 0. 5 8 Note. Variables were grand mean centered. There were significant group differences in slope for African centrality ()6 = 27.62, df= 7, p < .01). 191 Figure 20 shows the within-group slopes for African centrality and distributive restorative justice - blacks, where high African centrality is tied to higher expectation regarding procedural restorative justice — blacks for coloured women. In contrast, high African centrality was associated with lower distributive restorative justice expectations — blacks in the case of white women and coloured men. 9‘ or o so 8 Distributive Restorative Justice - Blacks 2.50 i Low {-1 SD) High (+1 SD) African Centrality -0- Black African Female + Black African Male -t- Coloured Female -N- Coloured Male -II- lndian Female + lndian Male -0—White Female —White Male Figure 20 Within-Group Slopesfor African C entrality and Distributive Restorative Justice - Blacks Table 34 and Table 35 show the within-group coefficients for procedural restorative justice — blacks and distributive restorative justice —— blacks, respectively. 192 028:8 SSE 052w who; 832:; .222. _N.0- 0Y0- No.0 ~00 0V0- n _ .0 00.0 NWO- :ouflwfifi Ecocmoswm 0m.0- 0m.0- mm.0- SWO! M00 0v.0- - :20 2.0- 58:3 - 352:6 32% h _ .0 000 0N0 mm .0 00.0- m 0 .0 m m .0 m _ .0- 290550122 80 8:80.080 00.0- 50.0- :0- 2.0 mmd- 8.0 00.0- .20- $.80 toamsm no.0- mmd- 5N0 0m.0- N10- 5 .0 2.0. 00.0- b=mbcoo 5854.. mod. 50.0 m0.0- M00 mm.0- 00.0 00.0 0m.0 5:50:00 8mm M00- 000- 00.0 00.0 00.0- 0m .0 3.0 00 .0 000835 032 0380.0 232 295m 232 00“th 2&2 0:“th :06585 BE? 853 9E2: :39: 00.5200 35030 SSE/x Son< x85 beam 833% - 85$. méexofimx BSEQEK notxfiammufimeb N423 o.§.5-=.§.§ 3H 2an 193 622:8 SEE 28% 20.5 333:; $32 d d d- 3d. 5d 8d 8d. 8d ddd d _ d- dosed-35 REESE 3d- 3d- 2d. Ed- :d- 8d. 2d Nod .55 - 8525 38m Rd mmd- odd- ad ad. Ed. 2d :d- 2385 E22 5 85505 8d mmd 5d 3d 8d 5d Rd- Sd 5:8 c253 odd- mmd- 3d dmd- Rd- Ed 8d 9d- €550 sac-2 ddd odd- Rd- dmd 2d- 8d- 8d- 3d E858 83d Ed fid- ddd odd- dmd- mmd ad mdd- 58.35 232 Edam-.0 032 295m 232 canon 032 £95m 85585 SE? 323 $65 565 8325 8328 53% 98:3. xosm fism £93m - 83-5. méuxoamm Saddam-ENG S\&§.§N\m¢b T384 mzoxb-=.§.§ m m 2an 194 Hypothesis 10 stated that group culture of support for affirmative action would add to the variance explained in individual procedural restorative and distributive restorative justice expectations, and that group culture of support for affirmative action would be more strongly related to procedural restorative and distributive restorative justice expectations for members of groups where there is a culture of lack of support for affirmative action. Hypothesis 10 was not supported. To demonstrate that the effect of white guilt on individual justice expectations is fully mediated by group culture of support for affirmative action, the relations between each of these two Level-2 variables and the Level-l justice‘ expectation were assessed. As group culture of support for affirmative action does not appear to be significantly related (p < .05) to general restorative justice, distributive justice, and procedural and distributive justice for women and blacks the focus here in on procedural justice expectations only. For procedural justice expectations, culture of support for affirmative action showed a significant pooled slope (ylo = 0.41, t= 4.08, p < .01), as did group belief in white guilt (ym = 0.33, t = 4.55, p < .01) when entered separately. As neither of these variables demonstrated significant group differences in slope in relation to procedural justice expectations, the mediation test is conducted using ordinary least squares regression. Group culture of support for affirmative action, entered alone, accounted for 5. % of the variance in procedural justice expectations. Group belief in white guilt, entered alone. accounted for 6.2% of the variance in procedural justice expectations. When group belief in white guilt was added to the regression analysis in a second block after group culture of support for affirmative action, the change in R2 was 0.006 (p = .17). Thus, Hypothesis 11, that the effect of group belief in white guilt on individual justice expectations is fully 195 mediated by group support for affirmative action was supported for procedural justice expectations. General restorative justice. There was a significant pooled slope for the effect of individual support for affirmative action on general restorative justice expectations (Km = 0.33, t = 4.49, p < . 01). There was a marginally significant slope for culture of support for affirmative action (Y0! = -0.30, t = -2.43, p < .10) and for social guidance ( y5(, = 0.19. t = 2.14,p < .10). These results are shown in Table 36. There were no significant group differences in slopes. Table 36 Multilevel Analysis Results for General Restorative Justice — Pooled Slope Standard Dimension Coefficient Error T-ratio clf P-value yo, Culture of Support for AA -0.30 0.12 -2.43 5 0.06 y"; "Restorative Justice Culture 0.14 0.41 0.33 5 0.76 710 Race Centrality -0.02 0.08 -0.26 7 0.80 720 African Centrality 0.04 0.09 0.47 7 0.65 9’30 Support for AA 0.33 0.07 4.49 7 0.00 ym Preference for Merit Principle 0.06 0.12 0.5] 7 0.63 196 Table 36 continued. y50 Social Guidance - Unfair 0.19 0.09 2.14 7 007 yégEducational Integration 0.01 009 ~ 0.14 7 0.89 Note. Variables were grand mean centered. Hypothesis 12 stated that group-level restorative justice expectations would be positively related to individual-level restorative justice expectations. Hypothesis 12 was not supported. Hypothesis 22 stated that preference for merit principle would be more strongly related to procedural restorative justice expectations than to procedural justice expectations and Hypothesis 23 stated that preference for merit principle would be more strongly related to distributive restorative justice expectations than to distributive justice expectations. Neither of these hypotheses was supported. Intentions as the Dependent Variable To test hypotheses related to outcomes of intention to withdraw from the selection process, emigration intention, entrepreneurial intention and education intention, each outcome was regressed onto the three major justice dimensions of interest as predictors. As there is conceptual overlap between procedural and distributive justice and procedural restorative and distributive restorative justice respectively, the relations for procedural restorative and distributive restorative justice for women and for blacks were examined separately. Withdrawal intention. Results of analyses for withdrawal intention regressed onto procedural, distributive and general restorative justice are presented in Table 37. 197 Table 37 Multilevel Analysis Results for Withdrawal Intention with Justice Factors — Pooled Slope Standard Justice Dimension Coefficient Error T-ratio df P-value rm General Restorative 0.02 0.05 0.46 7 0.66 720 Procedural 0.04 0.06 ~ 0.66 7 0.53 no Distributive -0.09 0.06 -1 .52 7 0.17 Note. Variables were grand mean centered. There were no significant pooled slopes for any of these three justice dimensions. Thus, Hypothesis 30, which stated that procedural justice expectations would be negatively related to withdrawal intention, was not supported. Similarly, Hypothesis 31. which stated that distributive justice expectations would be negatively related to withdrawal intention, was not supported. Further, there were no significant differences in Level-1 slopes across groups. Therefore, Hypothesis 32, which stated that restorative justice expectations would be positively related to withdrawal intention for non-target group members, and negatively related to withdrawal intention for target group members, was not supported. Table 38 shows the fixed effects results for analyses where withdrawal intention was regressed onto procedural restorative — women, distributive restorative justice — women, and general restorative justice. 198 Table 38 Multilevel Analysis Results for Withdrawal Intention with Justice Factors — Women — Pooled Slope Standard Justice Dimension Coefficient Error T-ratio df P-value 1’10 General Restorative 0.02 0.05 0.41 7 0.70 no Procedural restorative - -0,04 0,09 -0.46 7 0.66 women y30 Distributive restorative - -0,02 0,08 -026 7 08] women Note. Variables were grand mean centered. There were no significant pooled slopes. However, there were significant group differences in slope for procedural restorative justice - women ()(y = 16.47, df= 7, p < .05). Table 39 shows the within-group coefficients, where white men and Indian men both show meaningful negative coefficients for procedural justice — women. 199 3n- 33:5 623:8 52: 052w 0.53 mo_na_8> .802 5:83 0m.0 00.0- N00 3.0- 3.0. _0.0 00.0- 5.0- - 2,0888% 3:20:65 5an 0W0- m0.0 0N0- 0m.0 8.0 0m.0- 3.0 02.0- - 030.2930 3.50805 m 0 .0- 00.0 0 _ .0- 00.0 00.0 <00. m _ .0 m H .0 033.8330 03250 m0.0 m0.0 00.0- 00.0- m 0 .0- v0.0 n _ .0 n d .0- Eco-.85 2d: 2950 £32 DEEP-H 232 295m 232 Badman 88585 853 BE? 233 565 563 005200 005200 58:3. Swot-F. xoflm xoflm Nose-$00.0 xo\fl§.~o~\\mob 7333 “SEC-55¢: 0m 033- 200 Table 40 shows the fixed effects results for analyses where withdrawal intention was regressed onto procedural restorative — blacks, distributive restorative justice —— blacks, and general restorative justice. Table 40 Multilevel Analysis Results for Withdrawal Intention with Justice Factors ~ Blacks —— Pooled Slope Standard Justice Dimension Coefficient Error T-ratio df P—value 1’10 General Restorative 0.03 0.04 0.65 7 0_ 5 3 7'20 Procedural restorative - 0,05 0,05 1,10 7 031 blacks 730 Distributive restorative - -0,04 0,04 -1 .02 7 ()_34 blacks Note. Variables were grand mean centered. There were no significant pooled slopes for withdrawal intention and procedural and distributive restorative justice — blacks. There were no significant group differences in slopes. Thus, Hypothesis 33, that procedural restorative and distributive restorative justice expectations would be negatively related to withdrawal intention, was partially supported for procedural restorative justice — women. Emigration intention. Table 41 shows the fixed effects results for analyses where emigration intention was regressed onto procedural, distributive and general restorative justice. 201 Table 41 Multilevel Analysis Results for Emigration Intention — Pooled Slope Standard Justice Dimension Coefficient Error T-ratio df P-value }’]() General Restorative -0.12 0.05 -2.48 7 0.04 Y20Pt00edutai -0.04 -0.10 -0.41 7 0.69 Y30 Distributive 0.08 0.12 0.73 7 0.49 Note. Variables were grand mean centered. There was a significant pooled slope for the effect of general restorative justice on emigration intention (7m: -O.12, t = -2.48, p <. 05). There were significant group differences in intercept ()(7 = 27.47, df= 7, p < .01), and slbpe for procedural justice ()6 = 14.01, df= 7, p = .05). Figure 21 shows the mean levels in emigration intention across groups, while Table 42 shows the within-group coefficients. Coloured men showed the lowest emigration intention, indicating they did not intend to emigrate. Indians and whites of both genders showed the highest level of emigration intention. Hypothesis 34- which stated that procedural justice expectations would be negatively related to emigration for non-target group members, was not supported. Hypothesis 35, that distributive justice expectations would be negatively related to emigration for non-target group members, was not supported. Hypothesis 36, that restorative justice expectations would be positively related to emigration for non-target group members, was not supported. 202 5.00 1.00. I I l_ I I I I 2 Black Black Coloured Coloured Indian ' lndian White White Male African Afrimn Female Male Female Male Female Female Male 0’ P 0 0 0 0 Emigration Intention O 0 Figure 21 Mean Emigration Intention by Group 203 083:8 58:. 2de 083 moZEE> .282 0Y0- 000- 3” .0 m _ .0 3.0. m0.0- 3.0 3.0- 8:2:me 3.0 2.0. 3.0. mm.0- 0N0- £0 2 .0- 3.0 350820 00.0- 0 _ .0- mm .0- 0 _ .0- 0m .0- 00.0 mm.0- N00 aide-288% 388$ Nun-.0 mm.0 00.0 _m.0 m:- 0m.0- $0. 2.0- £886— 232 £280 232 228-0 232 288”— 232 228.0 comma—BED 80mg 83>» 0053 865 :32: 88200 023030 SSE-a Swot-F. x85 x85 =Qw-kN~=N =QNNQKMNEM K°\.MN:N.~U~\\NQD NINNQWN / Q: 0&0 1~A~V\§ 204 Table 43 shows the fixed effects results for analyses where emigration intention was regressed onto procedural restorative justice — women, distributive restorative justice - women, general restorative justice, and preference for merit principle. Table 43 Multilevel Analysis Results for Emigration Intention with Justice Factors -- Women Pooled Slope Standard Justice Dimension Coefficient Error T-ratio elf P-value y”; General Restorative _0_ 12 0.06 . _1 . 8 4 7 0. I 1 720 Procedural Restorative - 0.1 1 0.14 0.84 7 0.43 Women Y30 Distributive Restorative - -0,13 0,14 .093 7 0,38 Women Y40 Preference for Merit 0.24 0.13 1.90 7 0.10 Note. Variables were grand mean centered. There were no significant pooled slopes. In examining the random cf fects. there were marginally significant differences in Level 1 slopes for procedural restorative justice — women of = 13.12, df= 7, p < .10) and distributive restorative justice — women (f = 12.36, df= 7,p < .10). Table 44 shows the fixed effects results for analyses where emigration intention was regressed onto procedural restorative justice — blacks, distributive restorative justice - blacks, general restorative justice, and preference for merit principle. 205 Table 44 Multilevel Analysis Results for Emigration Intention with Justice Factors ~— Blacks — Pooled Slope Standard Justice Dimension Coefficient Error T-ratio 4f P-value ym General Restorative -0.12 0.06 -' -2.00 7 0.09 ygg Procedural Restorative - 0.09 0.09 1.02 7 0.34 Blacks y30 Distributive Restorative - -O.13 0.11 -l .15 7 029 Blacks ym Preference for Merit 0.23 0.1 1 1.99 7 0.09 Note. Variables were grand mean centered. There were marginally significant pooled slopes for general restorative justice / (1’10: -0-12, 1 = -2-00, 1) < .10) and preference for merit principle (1’30: 0.23. t = 1.99, p a .10). Results for random effects showed significant group differences in slope for distributive restorative justice — blacks (x2 = 14.16, df= 7, p < .05). Without significant main effects it is not necessary to test for interaction effects. Hypothesis 37, that preference for merit principle combined with low procedural restorative and low distributive restorative justice expectations for own in-group would be positively related to emigration intention for target group members, was not supported. 206 Entrepreneurial intention. Table 45 shows the fixed effects results for analyses where entrepreneurial intention was regressed onto procedural, distributive and general restorative justice. Table 45 \ Multilevel Analysis Results for Entrepreneurial Intention — Pooled Slope Standard Justice Dimension Coefficient Error T-ratio df P-value ym General Restorative _0.03 007 _0_3 8 7 0.7] 1’20 Procedural 0.08 0.10 0.87 7 0.41 no Distributive 0.03 0.10 0.25 7 0.8] Note. Variables were grand mean centered. There were no significant pooled slopes for entrepreneurial intention. In considering random effects, however, there were significant group differences in intercept (X7 = 24.35, df= 7, p < .01). Figure 22 shows the mean levels in entrepreneurial intention across groups. For black Africans, Indians and whites, men appeared more likely than women of the same race to hold entrepreneurial intentions, although all groups were fairly neutral regarding entrepreneurial intentions. Black African men showed the highest (yet modest) entrepreneurial intention. There were no significant differences in Level 1 slopes. Thus, Hypothesis 38, which stated that restorative justice expectations would be positively related to entrepreneurial intention for non-target group members. was not supported. 207 5.00 4.00 . l l t Entrepreneurial Intention 3.00 2.00 l l . l l l 1.00 i . .. ,. II I Black Black Coloured Coloured Indian lndian White White Male African African Female Male Figure 22 Female Male Female Male Female Mean Entrepreneurial Intention by Group Educational intention. Table 46 shows the fixed effects results for analyses where educational intention was regressed onto procedural, distributive and general restorative justice. Table 46 Multilevel Analysis Results for Educational Intention — Pooled Slope Standard Justice Dimension Coefficient Error T-ratio df P-value Y") General Restorative 0.01 0.03 0.30 7 0.77 720 Procedural 0.03 0.04 0.82 7 0.44 208 Table 46 continued. Y30 Distributive 0.05 0.09 0.61 7 0.56 Note. Variables were grand mean centered. There were no significant pooled slopes. In considering random effects. there were no significant group differences in intercept or slopes. Thus, Hypothesis 39. that restorative justice expectations would be positively related to education intention in non- target groups, was not supported. Table 47, below, provides a summary of the hypotheses tested and the extent to which hypotheses were supported. Table 47 Summary of Hypotheses Tested Hypothesis Support Size of Effect (ifavailable) Hypothesis 1: A five-factor justice model that Supported - Good Goodness of includes Colquitt’s (2001) dimensions as well as model fit for fit: restorative justice expectations will be expanded model NNFI = 0.98 supported. when conducting CFI = 0.98 confirmatory SRMR = 0.06 factor analysis. RMSEA = 0.06 Hypothesis 2: Restorative justice expectations will vary significantly across race/gender groups with black Africans having the lowest expectations of restorative justice, and white men having the highest expectations of restorative justice. Not supported —- Lack of significant between group variance in Level 1 intercepts. 209 Table 47 continued. Hypothesis 3: Distributive justice and procedural justice expectations will vary significantly across race/gender groups, with white men having the lowest expectations of procedural and distributive justice, and white women having the highest expectations of procedural and distributive justice. Expectations of procedural and distributive justice among black Africans will be lower those of coloureds, which will be lower than those of Indians. Partially supported - Significant between group variance in Level 1 intercept for procedural justice. Between group variance in procedural justice: 6% Hypothesis 4: The relationship between Partially Between restorative justice expectations and procedural supported - group variance and distributive justice expectations will vary Significant in Level 1 significantly across race/gender groups with between group slope for white men having an inverse relationship variance in Level procedural between restorative justice expectations and 1 slopes for justice: 13.7% procedural and distributive justice, while white restorative justice Between women, black Africans, coloureds and Indians expectations and group variance will show a positive relationship between both procedural in Level 1 restorative justice expectations and expectations and distributive slope for of procedural and distributive justice. justice. distributive justice: 10% Hypothesis 5: Distributive restorative and Partially Between procedural restorative justice expectations for supported - group variance treatment of blacks will vary significantly Significant in distributive across race and gender groups with black African men having the lowest expectations of distributive restorative and procedural restorative justice, lower than the expectations of coloureds, which will be lower than those of Indians. Indian women will have the highest expectations of justice. between group variance in Level 1 intercepts for distributive restorative justice for blacks. restorative justice for blacks: 9% 210 Table 47 continued. Hypothesis 6: Distributive restorative and procedural restorative justice expectations for treatment of women will vary significantly across race and gender groups, with black African women having the lowest expectations Partially supported - Significant between group variance in Level 1 intercepts for Between group variance in procedural restorative justice for of distributive restorative and procedural procedural women: 3% restorative justice, lower than the expectations restorative justice of coloureds, which will be lower than those of for women. Indians. White women will have the highest expectations of justice. Hypothesis 7: Group belief in white guilt will Supported - Variance in be positively related to group culture of Significant group culture support for affirmative action. generalized least squares estimate of Level 2 regression parameters. of support for affirmative action: 94% Hypothesis 8: Group culture of support for affirmative action will add to the variance explained in individual procedural justice expectations, and group culture of support for affirmative action will be more strongly related to procedural justice expectations for members of groups where there is a culture of lack of support for affirmative action. Partially supported - Significant variance accounted for by group-level factor, but no significant group differences in slope. Between- group variance in procedural justice accounted for by group culture of support for affirmative action: 96% Hypothesis 9: Group culture of support for affirmative action will add to the variance explained in individual distributive justice expectations, and group culture of support for affirmative action will be more strongly related to distributive justice expectations for members of groups where there is a culture of lack of support for affirmative action. Not supported —— Lack of significant variance accounted for by group-level factor. 211 Table 47 continued. Hypothesis 10: Group culture of support for affirmative action will add to the variance explained in individual procedural restorative and distributive restorative justice expectations, and group culture of support for affirmative action will be more strongly related to procedural restorative and distributive restorative justice expectations for members of groups where there is a culture of lack of sgiport for affirmative action. Not supported — Lack of significant variance accounted for by group-level factor. Hypothesis 11: The effect of group belief in white guilt on individual justice expectations is fully mediated by group support for affirmative action. Supported for procedural justice - No incremental variance accounted for by white guilt over and above group affirmative action support, after showing relation between all three constructs. Non- significant R2 change = 0.0006 Hypothesis 12: Group-level restorative justice expectations will be positively related to individual-level restorative justice expectations. Not supported — Lack of significant variance accounted for by group-level factor. Hypothesis 13: The relationship between group-level restorative justice expectations and individual-level procedural and distributive justice expectations will vary across groups such that group-level restorative justice expectations will be positively related to individual-level procedural and distributive justice expectations in target group members, and negatively related in non-target group members. Not supported — Lack of significant between group variance in Level 1 slopes. 212 Table 47 continued. Hypothesis 14: The relationship between race centrality and distributive justice expectations will vary across race groups such that there will be a positive relationship between race centrality and distributive justice expectations for race-based target groups and a negative relationship for race-based non-target groups. Not supported — Lack of significant between group variance in Level 1 $10pes. Hypothesis 15: The relationship between race centrality and procedural justice expectations will vary across race groups such that there will be a positive relationship between race centrality and procedural justice expectations for race-based target groups and a negative relationship for race-based non-target goups. Not supported —- Lack of significant between group variance in Level 1 slopes. Hypothesis 16: African centrality will moderate the effect of culture of support for affirmative action on procedural justice expectations such that high African identity weakens the relationship between culture and justice expectations. Not supported — Lack of significant between group variance'in Level I slopes. Hypothesis 17: African centrality will moderate the effect of culture of support for affirmative action on distributive justice expectations such that high African identity weakens the relationship between culture and justice expectations. Not supported —- Lack of main effect so significant between group variance in Level 1 slopes not tested. Hypothesis 18: Support for affirmative action will be positively related to procedural justice expectations. Not supported — Lack of significant pooled Level 1 slope. Hypothesis 19: Support for affirmative action will be positively related to distributive justice expectations. Not supported — Lack of significant pooled Level 1 slope. Hypothesis 20: Preference for merit principle will be negatively related to procedural justice expectations. Not supported - Lack of significant pooled Level 1 slope. Hypothesis 21: Preference for merit principle will be negatively related to distributive justice expectations. Not supported — Lack of significant pooled Level 1 slope. 213 Table 47 continued. Hypothesis 22: Preference for merit principle will be more strongly related to procedural restorative justice expectations than to procedural justice expectations. Not supported -— Lack of significant pooled Level 1 slope so no z-test conducted. Hypothesis 23: Preference for merit principle will be more strongly related to distributive restorative justice expectations than to distributive justice expectations. Not supported — Lack of significant pooled Level 1 slope so no z—test conducted. Hypothesis 24: Educational integration will be negatively related to procedural justice expectations. Not supported — Lack of significant pooled Level 1 slope. Hypothesis 25: Educational integration will be negatively related to distributive justice expectations. Not supported — Lack of significant pooled Level 1 slope. Hypothesis 26: Educational integration will moderate the effect of group culture of support for affirmative action on procedural justice expectations such that those with an integrated Not supported — Lack of significant main effect so interaction not education will show a weaker relationship examined. between culture and procedural justice expectations. Hypothesis 27: Educational integration will Not supported — moderate the effect of group culture of support for affirmative action on distributive justice expectations such that those with an integrated Lack of significant main effect so interaction not education will show a weaker relationship examined. between culture and distributive justice expectations. Hypothesis 28: Social guidance — unfair will be Supported - Variance in negatively related to procedural justice Significant pooled procedural expectations. Level 1 slope. justice expectations: 3.8% 214 Table 47 continued. Hypothesis 29: Social guidance - unfair will be negatively related to distributive justice expectations. Supported - Significant pooled Level 1 slope. Variance in distributive justice expectations: 2.3% Hypothesis 30: Procedural justice expectations will be negatively related to withdrawal intention. Not supported — lack of significant pooled Level I Skye. Hypothesis 31 : Distributive justice expectations will be negatively related to withdrawal intention. Not supported — Lack of significant pooled Level 1 slope. Hypothesis 32: Restorative justice expectations will be positively related to withdrawal intention for non-target group members, and negatively related to withdrawal intention for target group members. Not supported —- Lack of significant between group variance in Level 1 slopes. Hypothesis 33: Procedural restorative and distributive restorative justice expectations will be negatively related to withdrawal intention. Partially supported — Lack of significant pooled Level 1 slope but significant group differences on slope and negative relationship for white men and Indian men with Variance in withdrawal intention explained by procedural restorative justice - women: 1.8% regard to procedural justice - women. Hypothesis 34: Procedural justice expectations Not SUpp'OI'ted - will be negatively related to emigration for non- Significant target group members. 215 between group variance in Level 1 slopes but not a negative slope for white men. Table 47 continued. Hypothesis 35: Distributive justice expectations will be negatively related to emigration for non-target group members. Not supported — Lack of significant between group variance in Level 1 slopes, but there is a negative slope for white men. Hypothesis 36: Restorative justice expectations will be positively related to emigration for non- target group members. Not supported — Lack of significant between group variance in Level 1 slopes. Hypothesis 37: Preference for merit principle combined with low procedural restorative and low distributive restorative justice expectations for own in-group will be positively related to emigration intention for target group members. Not supported ~— Lack of significant main effect so interaction not tested. Hypothesis 38: Restorative justice expectations will be positively related to entrepreneurial intention for non-target group members. Not supported —- Lack of significant between group variance in Level 1 slopes. Hypothesis 39: Restorative justice expectations will be positively related to education intention in non-target groups. Not supported — Lack of significant between group variance in Level 1 slopes. 216 THEORETICAL AND PRACTICAL IMPLICATIONS My research agenda was to answer the following major questions: Is an additional, distinct justice factor — restorative justice — supported? What are the relevant individual-level and group-level antecedents of justice expectations within the context of strong preferential selection, and how do they interact in influencing individual-level outcomes? What are the effects of justice expectations on positive and negative outcomes 1 in this context? In this section I address how the results of this study answer each of these questions. Is an additional, distinct justice factor — restorative justice -— supported? In analyses to demonstrate the multi-dimensionality of justice, adding general restorative justice as a distinct factor such that the justice model is comprised of procedural, distributive, interpersonal, informational (see Colquitt, 2001), and general restorative justice dimensions, served to improve model fit, demonstrating excellent fit and a meaningful change from a four-factor, three-factor, two-factor, and one- factor model. A distinct general restorative justice factor was clearly supported. These findin gs are consistent with my argument that a distribution based on demographic group must be distinguished from distributions that are more typical in the justice literature —- equity- based distributions —— that is the dominant conceptualization according to C olquitt, Greenberg, and Zapata-Phelan (2005). Neither is this a distribution based on individual need. The focus is on a societal correction at the group level, although it must be implemented at the individual level. Further, the value of differentiating between procedural and distributive j ustice and the more specific procedural restorative and distributive restorative justice factors. 217 respectively, was demonstrated for specific target groups - blacks and women. Models that allowed the restorative justice components to be distinct had better model fit. and this was a meaningful improvement. Distinct procedural restorative and distributive restorative justice dimensions regarding the handling of blacks and the handling of women were supported. These findings are consistent with my argument that there are issues of duality and simultaneity in selection under strong preferential selection as different groups are handled differently, and to better understand expectations and reactions in this context. it is necessary to be specific about examining views and considering experiences within a target group, and establishing views regarding within-group handling rather than capturing only an overarching impression regarding the selection process. This specificity serves to improve our understanding of the relations among justice dimensions. their antecedents and outcomes. The relations between general restorative and procedural justice as well as between general restorative and distributive justice did vary across groups, with the general pattern indicating little difference in connection between low restorative justice and procedural or distributive justice among groups, but when genera] restorative justice expectations were high, some groups showed lowered procedural or distributive justice expectations. As the groups with the steepest slopes here are whites and Indians of both genders, this pattern may be explained by self-interest, in that those perceiving their group members to be less likely to be beneficiaries indicated lower procedural or distributive justice expectations. This suggests that although Indians of both genders and white women are designated beneficiary groups under current policy, potential applicants 218 may expect that in reality preference is being given to black African and coloured applicants. What are the relevant individual-level and group-level antecedents of justice expectations within the context of strong preferential selection, and how do they interact in influencing individual-level outcomes? First, I will discuss the individual-level antecedents of justice expectations. Race and African Centrality Pooled race centrality and justice slopes were not significant and did not vary significantly across groups. It appears that on the whole, race centrality does not play a meaningful role in justice expectations under preferential selection. However, Indian women and white women did show coefficients of 0.44 and 0.38, respectively, when looking at the within-group relationship between race centrality and procedural justice. It may be that some higher race identifiers in relatively advantaged groups do recognize their advantaged status and translate that into higher justice expectations through their support of the current redistributive policy, or it may be that these individuals expect that they personally will receive preferential treatment as target group members, and thus hold expectations for justice out of self-interest. Such issues would need to be examined in future research. African centrality and justice slopes capturing procedural or distributive issues varied across groups, where white men and women generally showed negative slopes. For example, high African centrality was related to low procedural justice for white men and women. These differences may be explained in part by frustration among high African centrality whites with the inconsistency that whites are treated as an out-group with regard to the black-white dichotomy and affirmative action policy. yet these whites 219 feel a strong affinity with the African community and view themselves as in—group. They may feel that as in-group members they should not be treated differently, yet they are. resulting in lower fairness perceptions. Gilbert (2005, p.64), in discussing the ambiguities of being of mixed race, notes that there is a need for models that explain mixed identities. He views being of mixed race as having the potential to “simultaneously transgress many people’s rigid perceptions of racialised differences, particularly in terms of insider/outsider. In other words, border experiences and developing identities challenge simple binary oppositions in some racialised discourses.” 1 view this observation as relevant for the border identity that is evolving for those who recognize the centrality of their African identity and are not black African. The border experience creates challenges for those non-black Africans who are high in African centrality. These may include moving within “local” and “foreigner” communities simultaneously, shifting identities according to the social context and adopting whichever identity is situational, living a border identity, or living a “foreigner” identity and making‘ forays into African communities (see Root’s, 1996, theory of four positions that mixed-race people could adopt). The ambiguity presented through this border identity may explain the negative African centrality — justice expectation slopes for whites. It is interesting to note that there are significant race group differences both in race centrality (F(3, 313) = 18.32, MSE = 0.75, p = 0.00) and African centrality (F(3,3 l 3) = 0.78, MSE =, p = 0.00). Black Africans and Indians responded positively regarding both their race centrality and African centrality. The black African mean for race centrality was 3.56 and for African centrality was 3.78. The Indian mean for race centrality was 3.38 and African centrality was 3.36. The response from coloured people 220 was neutral regarding both their race centrality (mean = 3.03) and African centrality (mean = 3.03). Whites, however, were a low neutral with regard to. race centrality (mean = 2.71), but positive regarding African centrality (mean = 3.55). This may generate specific challenges for whites and their identity, as from the outside their visible whiteness may to others make salient their racial identity, which they themselves don‘t view as central, and their African identity, which is more meaningful among these respondents, may be invisible unless they engage is specific activities to demonstrate it. Support for A fiirmative Action Pooled slopes for individual support for affirmative action and procedural or distributive justice or procedural restorative or distributive restorative justice ~— women or blacks were not significant in the HLM models tested. This suggests that respondents are not making the simplistic judgment that because they like something, it must be fair. Instead, they may be making a more careful assessment of what they think will truly happen in the selection process. This possibility is supported by the pattern of within- group coefficients for the relationship between support for affirmative action and distributive restorative justice — blacks. Indian women and Black African women show positive coefficients (0.48 and 0.44 respectively) while black African men showed a negative relationship, although group differences in slope were not statistically significant. Individual support for affirmative action was associated with general restorative justice expectations. This was not anticipated, and may be explained in part by a potential misunderstanding of survey questions among some participants. General restorative justice questions may have been interpreted as asking what ought to happen rather than 221 t what one anticipates actually will happen, which would explain the association with support for affirmative action. In the present study, on average, respondents indicated that they were neutral with regard to individual support for affirmative action (mean = 3.15. SD = 0.95). Kaiser. Drury, Spalding, Cheryan, and O’Brien (2009) demonstrated a decrease in support for social justice actions after the election of President Obama. Considering the overall neutral view toward affirmative action found here it may be that South Africans are showing lowered support for affirmative action as they have been operating as a non- racial democracy since 1994, and view this as evidence of a sufficiently level playing field. Preference for Merit Principle The unreliability of the preference for merit principle measure may explain the limited relationships observed. The measure captures items that address multiple dimensions, not all of which might be construed as true merit in this context. Effort. which is captured by some items, may be regarded as different from output in a setting where not everyone has the same level of skills and experience to produce output. The resultant unreliability may have attenuated the merit preference justice expectation relationships. Further, Son Hing, Bobocel, and Zanna (2002) demonstrated that the presence of workplace discrimination mitigated the relationship between preference for meritocracy and opposition to preferential treatment. That is, those who were high in preference for meritocracy showed a weakened relationship between merit preference and opposition to preferential treatment in the presence of discrimination. Similar processing may be taking 222 l‘J place in the present study, although the same specific relations were not being tested. It may be the case that in the present study, those who held a preference for the merit principle, and who perceive that there is ongoing discrimination in South African workplaces, reduced their opposition to affirmative action and thus raised their justice expectations. This may explain the limited relationships observed between preference for merit principle and justice expectations. Also, Aberson (2007) demonstrated that diversity valuation is related to support for affirmative action, and l speculate that diversity valuation may serve a function similar to that perception of discrimination in moderating the relationship between merit and justice expectations. An alternative explanation may be that individuals believe that organizations are choosing between equally qualified candidates, and that the process by which they are choosing under these circumstances does not violate the equity rule. Hence the limited relationships between preference for merit principle and justice expectations. Educational Integration In the present study, educational integration did not appear to play a meaningful role in explaining procedural or distributive justice expectations, nor did it serve as a moderator of the effect of group culture of support for affirmative action on individual justice expectations. In this sample, as blacks were less likely and whites were more to likely to have experienced integration in high school, it appears that the measure overlaps with race, as well as with support for affirmative action and race centrality. and as such does not explain unique variance in justice expectations. This may explain nonsignificant results for this variable. 223 Social Guidance At the individual level, social guidance regarding unfairness appears to play an important role in building justice expectations. Social guidance — unfairness was related to justice expectations, where having friends and family report that the process would be unfair was related to lower expectations of both procedural and distributive justice as well as distributive restorative justice for women and blacks. Social infomration processing theory explains how this social influence may affect justice expectations. There were group differences in most of the social influence — unfairness and justice slopes, with white women and men generally having the most negative perception of justice when there was a higher level of social influence indicating unfairness. Subjective norms may become established where views are encouraged or discouraged by people who are important to an individual (see discussion in Burgess, 2002). This process of social guidance, where people of influence offer information, advice or guidance (Vaux. Riedel & Stewart, 1987) may come through one’s friends, family and co-workers (Goldman, 2001). This difference in slopes may be explained if white women and men are more likely to have friends and family who are white, and as such, hold more negative perceptions regarding the context of strong preferential selection. Whites may be more dependent on social influences regarding this context as they may perceive greater ambiguity here than do members of most other groups. Whites men may recognize that as a group they are not beneficiaries under affirmative action policy. and should expect to he looked over in favor of a non-white or female candidates, yet at the same time. whites have the lowest unemployment rate of all groups, which is encouraging regarding their prospects. Unemployment rates vary across races and follow the hierarchy of advantage: 224 black Africans have the highest unemployment rate at 27.7%, the coloured unemployment rate is 19.1%, the Indian unemployment rate is 11.8%, while only 5.3% of whites are unemployed (Statistics South Africa, 2009). White women may recognize that as women they have affirmative action target group status and should be given preference, yet at the same time, as whites they may be looked over in favor of non-white candidates. This ambiguity may create a culture in which members of these groups are more likely to rely on social information processing. This behavior pattern would be consistent with Salancik and Pfeffer’s (1978) social information processing theory. Next, I address the group-level antecedents of justice expectations. Race and Gender Published research regarding afiirmative action attitude reflects the overarching view of affirmative action as fair depending on race, gender, and self-interest (Harrison. et al., 2006). l have demonstrated between-group variance in individual procedural justice expectations, procedural restorative — women, and distributive restorative ~— blacks. The group differences in these dimensions can be explained by self-interest. Groups that showed the highest level of procedural justice expectations were those that are likely being viewed as highly desirable targets — black women. White men, who are not targets. report the lowest level of procedural justice. While white womenrare officially targets by virtue of being women, their relatively low level of procedural justice expectations may suggest that they are being overlooked in favor of black women, if self-interest is the true driver of these views. Similarly, that black African and Coloured men show the highest levels of expectation regarding procedural restorative justice for blacks may suggest that Indian men expect other black groups to be favored, in line with the argument that 225 Indians are “not black enough”. Results for procedural and restorative justice for women show that white women held the lowest expectation, suggeSting that although they ostensible beneficiaries, they may not expect procedural and distributive fairness when compared to other women in the selection process. Black women of the more advantaged black groups (Indian and Coloured) have more positive expectations. That race/gender groups in South Africa have shared experiences and along with that, have developed shared attitudes and perceptions (see Klein & Kozlowski, 2000: Campbell, 1958) was partially supported by findings here. It appears that at least one collective construct, group culture of support for affirmative action, has emerged and is being maintained (see Morgeson & Hofman, 1999). Social‘information processing within race and gender groups may be one mechanism by which these shared views develop (see Klein & Kozlowski, 2000) as individuals in this context have had a great deal of exposure to within-group interaction. Group general restorative justice expectation showed within- group agreement, but only small mean differences across groups. Group belief in white guilt, on the other hand, demonstrated less within-group agreement, but large differences in group means. Future research will be necessary to explore whether these should be interpreted as collective constructs. Nevertheless, the use of group mean levels in the analysis in the present study remains valuable. . Analysis of these Level-2 constructs indicated there was greater between-group variance relative to total variance for each of the three cultures proposed. Culture of support for affirmative action aggregation indices showed the greatest between-group variance and also the greatest within-group agreement of the three Level-2 constructs. Self-interest is one explanation for why those who are beneficiaries support a policy. As 226 race/ gender groups have shared ostensible beneficiary status, they would share support for the policy. Group Belief in White Guilt Research in the area of group guilt has generally not addressed the issue of a shared belief within a group of the guilt associated with a specific out-group. Thus, this study extends the existing literature (e.g., Swim and Miller, 1999; Doosje, Branscombe. Spears, & Manstead, 1998; Klandermans, Werner, van Doom, 2008) by addressing both in-group guilt (for whites) and out-group guilt (for non-whites). I have demonstrated here that there do appear to be shared beliefs within groups regarding the guilt and responsibility on the part of whites, and that the level of belief in white guilt varies across groups. While Baumeister, Stillwell, and Heatherton (1994) argue that guilt is interpersonal in nature, I propose that the relationships necessary to trigger guilt do not need to be individual-level relationships. Guilt can be established at the group level. and is based on group experiences. For group belief in white guilt, there were larger between-group differences in means, yet less within-group agreement than was the case for the other group constructs explored here. Within-group variability may be a function of differing perceptions regarding social progress being made. Those for whom the change is a less important goal (e. g., more prejudiced whites) may look backward as see how much has changed since the apartheid era. Those for whom change is a more important goal (e.g., less prejudiced whites) may look forward and focus on what still has to be achieved. This may explain the within-group differences. An alternative explanation if within-group 227 differences may be differences in political ideology (see Klandermans, Werner, & van Doom, 2008). An examination of the between-group levels of belief in guilt shows a pattern consistent with the inverse of the hierarchy of advantage discussed earlier (Statistics South Africa, 2006a; Thompson, 1990). Whites show the lowest level of belief in white guilt, while black Africans, the most historically disadvantaged of the race groups. indicate the highest level of belief in white guilt. This pattern is consistent with a rational response within previously disadvantaged groups to observing their relative deprivation and with white denial of their own culpability. The application of equity theory would suggest that if whites felt that they had benefitted unfairly at the expense of others, they would compensate for this imbalance and create balance by feeling guilty. Baumeister. et al. (1994) frame guilt as a type of distress over inequity that is suffered by people who are overrewarded. However, here we do not have whites indicating that whites should feel guilt and responsibility. This implies that whites must believe that their relative privilege was not gained at the expense of others. The group differences in group belief of white guilt and responsibility are worrying in that all of the non-white groups view whites as guilty. That suggests that non-whites might expect reparation to balance inequity. Unless whites are committed to reparation, it is likely that relations between whites and non- whites will be difficult. According to Schnabel and Nadler (2008), being a perpetrator threatens one’s public moral image and fosters a greater need for social acceptance. Unless whites embrace the perpetrator role, they will probably not feel that their moral image is problematic nor seek greater social acceptance through reparation. 228 As noted by Baumeister, Stillwell, and Heatherton (1994), guilt motivates relationship-enhancing behaviors. As posited, levels of group belief in white guilt were associated with a group culture of support for affirmative action. This linking of guilt and compensatory action or entitlement has been shown at an individual level in other research (e.g., Boeckmann & Feather, 2007; Doosje, Branscombe, Spears & Manstead. 1998; Swim & Miller, 1999), but here I have demonstrated a group-level effect. However, while I have presented group belief in white guilt and group culture of support for affirmative action as conceptually distinct, and would argue that an examination of the content of the items is evidence of their being different‘constructs, the very high correlation between these two constructs could be a result of the participants not differentiating between the two. Group Culture of Support for A flirmative Action Cross-level effects of the culture of support for affirmative action were demonstrated, with group culture of support for affirmative action contributing to the building of individual procedural justice expectations and individual general restorative justice expectations. That is, individuals appear to be affected by the culture of shared beliefs within their race and gender group. Social information processing theory (Salancik and Pfeffer, 1978) argues that people are adaptive and take into account cues from their social environment. That is, people are influenced by their context (e. g.. Hofmann, Morgeson, & Gerra, 2003; Seibert, Silver, & Randolph, 2004). Group Restorative Justice Expectations While group general restorative justice expectation showed within-group agreement, there were not large differences between group means. The limited between- 229 group variance in group general restorative justice expectations implies that there is fairly high consistency across groups in people’s expectations, suggesting that most people are aware of the realities of how selection is being handled. This suggests that there may be a shared affirmative action schema in this context. The consistency of this schema across groups may be explained by individuals making judgments about the selection system through heuristic processing, where they have limited information and the information that is available from the government and the media may prime processing based on categories (see Chaiken, 1987; Kulik & Perry, 1994). My results supports Nacoste‘s ( 1994) argument for the view of affirmative action policy as an affirmative action schema, where the individual may use the “average” process in affirmative action as the benchmark for what is to be expected. Relations between group general restorative justice expectations and other variables were probably attenuated and thus provide less powerful tests of my hypotheses regarding these variables. What are the effects of justice expectations on positive and negative outcomes in this context? Withdrawal Intention Group differences in slope for procedural restorative justice - women and withdrawal intention were demonstrated, with the steepest negative slopes for white and Indian men. If white and Indian men perceive themselves as the least likely to be considered for positions because they are historically relatively advantaged. they may frame their views about what will happen to women in a way that parallels what they expect for themselves. That is, lower procedural justice. This may be explained by feelings of relative deprivation (Runciman, 1966) as the result of a process of social comparison. Those who see in—group people who are qualified for positions. yet expect 230 that they are being turned down on the basis of race or gender, will be presented with their relative deprivation, hence a sense of injustice. However, for most justice expectation dimensions there were not meaningful relationships with withdrawal. This may be explained in part by the fact that some of those who do intend to withdraw may be doing so because they are scholarship students who have an extant work commitment for the company that is supporting them. Also. some participants did not intend to apply for jobs as they intended to pursue graduate study. These possibilities were identified by some respondents in their responses to open- ended survey questions. Their behavior is not necessarily related to their justice expectations in these cases. Further, there was a limited range of responses for withdrawal intention, and it appears that although there may be the perception that some people will likely not be considered due to race, individuals still continue to apply. As indicated by Slaughter, Sinar, and Bachiochi’s (2002) US study of black engineering students, although there may be some form of preferential selection based on demographics, and the plan may be viewed as less fair, this does not altogether destroy job pursuit intentions, although the level of pursuit intention is lower than is the case for other selection systems. Emigration Intention With regard to emigration intention, there were significant differences in mean level of emigration intention across groups. The groups most likely to emigrate were white and Indian men and women. While these race groups were historically the most advantaged, it should be noted that they are also the race groups that may be more likely to have the right to live and work in other countries through descent. Having family 231 fl ‘ r members who have settled abroad, and fear of crime, entered as control variables, added to the variance explained in emigration intention. Preference for merit principle was tied to emigration intention when entered with procedural restorative and distributive restorative justice for blacks and women, with a preference for merit being positively associated with emigration intention. It is interesting to note that there was little variance on the fear of crime measure — the worry about crime was nearly universal. The mean scale score regarding worry about crime was 4.4 (maximum score = 5). Eighty-eight percent of the respondents indicated that they worry about violent crime. As a comparative reference point, results from a 2004/2005 community study in the United Kingdom (Nicholas, Povey, Walker. & Kershaw, 2005) indicated that worry about violent crime varied across regions, with a low of 10 percent of the respondents in the South West region reporting that they worry about violent crime, and a high of 26 percent saying the same in the London region. Folger (1986) posits that in referent cognitions theory there are three conditions that link to resentment about the decision-making process. These include high referent. low likelihood of amelioration, and low justification. Indiahs and whites may easily be able to imagine what it would be like to be hired (high referent), yet see themselves being turned down because of their relative advantage. As affirmative action law is unlikely to change soon (low likelihood of amelioration) and with a merit orientation they may believe that the hiring decision should have gone differently (low justification). this system would trigger resentment that may reveal itself in emigration intention. 232 Entrepreneurial Intention There were group differences in entrepreneurial intention, with white women being the least likely to hold entrepreneurial activity intentions, and black African men the most likely to do so. Apart from the coloured group, men were more likely than were women of the same race to hold entrepreneurial intentions» Gender differences in entrepreneurial intention are consistent with those shown in other studies (e. g.. Zhao. Seibert, & Hills, 2005, where women had lower levels of entrepreneurial intention than men), possibly as a result of masculine stereotypes being associated with this role (see Gupta, Turban, & Bahwe, 2008). The anticipated relations between justice expectations and entrepreneurial intention were not demonstrated. It may be the case that other factors not captured by this study are far more meaningful predictors of entrepreneurial intention (e.g., self-efficacy -- Zhao, Seibert, & Hills, 2005; attitudes toward entrepreneurship. perceived behavioral control, and subjective norms — Gird ‘& Bagraim, 2008). Educational Intention Educational intention was surprisingly high and this strongly positive response was almost universal. This may be a function of the weak South African economy and high unemployment rates, where pursuing further study may be both a stalling tactic in avoiding trying to compete in a flooded labor market, and a proactive approach in seeking further qualifications to become more competitive. Study Limitations and Future Research Alternative Models of Normative Antecedents of Intentions Consistent with Bell, Wiechmann, and Ryan (2006). I have demonstrated that applicants do hold individual justice expectations, and those expectations can be directly 233 linked to outcomes, in this case withdrawal intention and emigration intention in some groups. However, based on the results of work by other researchers, it may be appropriate to consider alternative models regarding the antecedents of intentions and to consider additional constructs that also provide subjective norms that may directly affect intentions. White, Charles, and Nelson (2008) examine affirmative action attitude and use the theory of reasoned action to explain direct links between subjective norms and intentions. They argue that there is a direct link between the belief among salient references that a person should or should not perform a behavior, and the intention to act. Bell, Harrison, and McLaughlin (2000) proposed a similar argument and showed direct links between the subjective norm and intention. I now test revised models with supplemental antecedents of intentions, including group culture of support for affirmative action and social influence regarding unfairness. Results of these supplemental antecedent analyses indicate that for withdrawal intention, when social influence and group culture of support for affirmative action were added to the model as supplemental normative influences, there were not significant pooled effects for either of these variables when entered with the procedural. distributive and general restorative justice factors, although group differences in slope for social influence regarding unfairness were shown (X7 = 14.70, df= 7. p < .05). Similarly. when entered with procedural and distributive restorative justice expectation — women and general restorative justice there were group differences in slope for social influence affecting withdrawal ()(7 = 12.19, df= 7, p < .10). This pattern emerged also when the two supplemental antecedent variables were entered with procedural and distributive restorative justice — blacks and general restorative justice of = 15.17. df= 7. p < .05). 234 but there was also a marginally significant pooled effect for social influence regarding unfairness (740 = 0.11, t = 1.90, p < .10). The groups with meaningful positive slopes were coloured women and white men. For emigration intention, when social influence and group culture of support for affirmative action were added to the model that contained general individual-level justice expectations and merit orientation, there was a significant pooled effects for social influence regarding unfairness (m; = 0.24, t = 2.68, p < .05) and there were group differences in slope for procedural justice ()6 = 14.51, df= 7, p < .05). When the same model was examined substituting procedural restorative and distributive restorative -—- women in place on general procedural and distributive justice expectations, the were direct effects for group culture of support for affirmative action (yo; = -0.45, t = -3.25. p < .05) and social influence regarding unfairness (y40= 0.25, t = 3.00. p < .05). as well as group differences in slopes for procedural restorative justice — women ()6 = 19.02. df= 7, p < .05) and distributive restorative justice — women (X2 = 14.08. df= 7. p < .05). The same model for procedural restorative justice — blacks and distributive restorative justice — blacks showed direct effects for group culture of support for affirmative action (yu, = - 0.74, t = -3.62,p < .05), general restorativejustice (y,0= -0.13, t = -1.95, p < .10), and social influence regarding unfairness (ym: 0.24. t= 2.80, p < .05) and there were group differences in slope for procedural restorative justice - blacks (x2 = 12.27. df= 7. p < .10). 235 For entrepreneurial intention, the additional antecedents did not show meaningful effects when general distributive and restorative justice dimensions were in the model. nor when procedural restorative and distributive restorative for women or blacks were included. However, group culture of support for affirmative action did add to the variance explained in educational intention (1’01 = 0.18, t = 2.34, p < .10) and there were group differences in slope for procedural justice ()6 = 13.38, df= 7, p < .10). When procedural restorative — women and distributive restorative — women were substituted in the model for educational intention, there was an effect of group culture of support for affirmative action on educational intention (yo, = 0.17, t = 2.34, p < .10). Similarly. when procedural restorative — blacks and distributive restorative — blacks were substituted in the model for educational intention, there was an effect of group culture of support for affirmative action on educational intention (1’01 = 0.18, t = 2.32, p < .10). The contribution of these supplemental variables suggests that further theoretical model refinement regarding the specification of the normative antecedents of withdrawal. emigration and educational intentions should be considered in future research as direct links between normative information and these intentions are evident. Alternative Models of A ntecedents of Emigration Intention While the model proposed and tested in the present. study does contribute to our understanding of emigration intentions in the South African context, it is clear from the pattern of relationships between variables here that further theoretical development is necessary in understanding the antecedents of emigration intentions. For example. the negative relationship between African centrality and emigration intention makes intuitive sense, where those who have a strong local identity are less likely to hold emigration 236 intentions. However, theoretical development regarding this and other direct relationships are necessary. Future research should consider the direct and interactive effects of support for affirmative action, African centrality, merit orientation, social influence. and fear of crime on emigration intention. Power and Sampling For greater power to detect multilevel effects it would be beneficial to have a larger number of participants within each group, although I was still able to detect several important effects in this study. A larger number of groups would not be feasible in this case as South Africa has only four major race groups. Because of the small sample sizes for coloured men and women, results regarding these two groups may be somewhat unstable. Also, as the sample here is drawn from those who are on campus reading flyers. or hearing the announcement in class, and respondents self-select to participate in the study, 1 have limited knowledge of how the respondents may differ from non-respondents within this population of potential job applicants. Furthermore, the strategic choice to pick two campuses that are in different regions to ensure the ability to capture all race groups, as well as the nature of the institutions as top-tier English-medium universities, may influence the characteristics of the sample. Results may be influenced by additional variables that were not captured in the study, and caution should be exercised in generalizing results. Scales Used The restorative justice scale was designed for this study based on theory. Although theoretical discussion and empirical results here demonstrate that restorative justice is distinct from other dimensions, it would need to also be validated in other 237 samples where a preferential selection policy is in place. This could include samples of those who have been on the job market for an extended period of time. or those who have been considered for promotion, training opportunities, or layoff. The context of the study may be South Africa, where restorative justice is based on distributions to women. blacks, and the disabled, or other counties, such as India, where restorative justice may be based on caste, for example. Demonstrating that the restorative justice construct is distinct from yet related to other more widely accepted dimensions of justice and that this is true in a variety of settings will help to establish restorative justice as a necessary facet of justice models in the preferential selection context. Furthermore, Cropanzano, Slaughter, and Bachiochi (2005), among other researchers, have shown three-way interactions between procedural, distributive and interactional justice. In their study of black engineering students, Cropanzano. et al. showed that when there was low procedural justice, the distributive and interactional justice term was significant. While the present study did not address the interactive effects of restorative justice with interpersonal and informational dimensions of j ustice. future research in that area may offer useful theoretical and practical developments to assist organizations in facilitating positive outcomes among applicants even where they do engage in preferential selection. Some scales have low reliability, lowering the likelihood of demonstrating relationships between scales. For example, preference for merit principle and withdrawal intentions demonstrated low coefficient alpha. These scales would need to be revised for use in future studies such as this. An examination of the specific items making up the preference for merit principle scale showed that removal of the merit principle scale Item 238 89 “It is wrong for an employer to give a job to someone they know without first advertising the job to other candidates” would raise the scale coefficient alpha to 0.61. Removal of the withdrawal intention Item 61 “l have take steps toward applying for jobs” (reverse coded) from the withdrawal intention scale would result in a coefficient alpha of 0.67. Further, where there are low aggregation values, relationships with the aggregated variables will be attenuated, providing less powerful tests of hypotheses regarding relationships with these variables. Direct comparison of the results of the present study with results of other studies are made difficult where scales were either written specifically for use in this survey. or were modified scales that are based on those used elsewhere. For example, the items used for the four established justice dimensions (Colquitt, 2001), although very similar to those in other published work, have been modified to be appropriate for those who are entering the job market. General restorative justice items were developed for this study. Race centrality was revised from a black centrality scale to being generic race centrality, and only a small selection of the modified original items were used. Because of these two changes it is difficult to directly compare results on this study with those of other researchers. Both race centrality and African centrality scales showed some evidence of a minimal lack of invariance across race groups, suggesting that people of different races might conceptualize some of the items differently. Similarly, preference for merit principle used only a small selection of items from an existing scale. The current scoring of the scale presents both effort and contribution as merit-based, and giving a job to someone you know as not merit-based, yet the low scale reliability in the present study suggests that it may not be appropriate (in the South 239 African context at least) to combine these facets. As noted above, the removal of Item 89, “It is wrong for an employer to give a job to someone they know without first advertising the job to other candidates,” would improve scale reliability to alpha of .61. l posit that giving a job to someone you know first does not necessarily imply that the choice was not based on merit — you may have a highly qualified friend. Further, in the context of the local philosophy of Ubuntu, it may be regarded as socially responsible to hire someone you know. The remaining items captured issues of effort (Items 90 and 91) versus contribution (Item 88), and having these two facets may be problematic in a context where there is an affirmative action policy that requires organizations to hire candidates with “potential to do the job” as many applicants from previously disadvantaged groups may not make the grade based on traditional selection criteria. In this context. workers may be especially attentive to who is actually succeeding and making a contribution. Effort without effect may be viewed differently. Members of different groups may conceptualize these issues differently. As there are several possible explanations regarding the limited relationships between preference for merit principle and justice expectations, including unreliability of measures, the presence of prejudice, and equal qualifications among candidates, future research is necessary to resolve how merit relates to justice in the context of preferential selection. The group belief in white guilt scale was built for this study based on individual- referent research. Similarly, the social guidance, withdrawal intention, education intention and entrepreneurial intention items were written for this study. This makes it difficult to use the results of other studies for benchmarking, but this does create a reference point for future research. 240 While the present study shows clear differences in group means for belief in white guilt, there is within-group variability that is not fully explored here. Future research that explains the within-group variability may help us to better understand support for compensatory systems. For example, political ideology and issues of identity may play a role here (see Klandermans, Werner, and van Doom, 2008). The evolving African identity has important implications for reactions to process in Africa. Studies that examine both the measurement of this self-identification and the effects of this identity on outcomes in South Africa and other African countries may be valuable in building understanding regarding intergroup relations and community effects. This is of particular import due to the rise in xenophobic violence in South Africa (see “Rainbow nation's outsiders live in fear”, 2009) and the need to ensure stability and a place of welcome for people of a variety of identities. Martins and Parsons (2007) demonstrated that women with higher gender identity centrality reacted differently with regard to diversity management programs than women with low gender identity centrality. Research on other identities, such as that on gender centrality, may serve as a useful reference point for structuring future research on the evolving African identity and its implications in the workplace. Restorative Justice and Legal Claiming Groth, Goldman, Gilliland, and Bies (2002) examine commitment to legal claiming as a function of attributions and social guidance. They show that commitment to legal claiming is associated with attributions that the organization is at fault, and that social guidance serves as a moderator of this relationship. When there is a lower level of external attribution, individuals appears to be more strongly influenced by others with 241 regard to legal claiming. Legal claiming literature may be informative in the context of preferential selection as those who feel maltreated under affirmative action policy where there is the expectation of general restorative justice may seek legal recourse. One example of such litigation that has generated broad media interest is the recent US Supreme Court consideration of a case of white firefighters who claim reverse discrimination (see “High court backs firefighters in reverse discrimination suit”, 2009). It is clear from the present study that there are expectations regarding general restorative justice, people vary in their support of affirmative action policy, and for some groups. social guidance plays an important role in justice expectations. While I did not examine litigation intention in this study as those who are prospective applications are likely not yet deep enough into the selection process to have had experiences that prompt litigation intention, future research that explores general restorative justice, support for affirmative action, and social guidance, and their links to litigation intention among those who have been through the selection process may be informative regarding individual differences and commitment to legal claiming under preferential selection. Causality There is an underlying implication of causality in the overall model as presented. However, the study as conducted does not allow conclusiohs regarding causality. Such directional links would need to be demonstrated in further studies. I do plan to conduct a follow-up study with these same respondents after a year, allowing a temporal delay before their reactions are measured again. This will limit attenuation in the relationship between predictors and outcomes that might be a consequence of common method variance, and will lend support to causal inferences. Furthermore. this would allow the 242 capture of data on additional variables that may be relevant. For example. Kanfer. Wanberg, and Kantrowitz (2001) demonstrated the importance of personality traits such as extraversion and conscientiousness in predicting job search behaviors. The present study does not capture these traits. The Time 2 data collection is beyond the scope of this dissertation — it is a follow-up study. I anticipate that most of the participants will be faced with decisions regarding their applications during this period, allowing me to better examine the effect of the salient direct experiences (getting a job/getting rejected for a job, going through the interview process, etc.). This will allow measurement of whether and how expectations change over time, and will also support the capture of additional information such as potential legal claiming. As noted by Chan and Schmitt (2004). change in applicant reactions over time is one of the most important but neglected areas of research into applicant reactions. My interest would be in intra-individual consistency. change after multiple events, and outcomes that relate to these trajectories. This dissertation study sets the groundwork for that longer-term study. 243 Conclusion Although Kravitz (2008) notes that most of the problems associated with affirmative action are with preferential selection, and that preferential selection should be eliminated, he does not acknowledge that there are many contexts where preferential selection is highly unlikely to end soon. It remains one of the few options available to dramatically change the demographics of the workplace in contexts such as South Africa where the playing field is far from level and dramatic change is needed. Considering that preferential selection remains a strategy that will very likely be used in such contexts. this study makes several theoretical and practical contributions. Figure 1 provided a general organizing heuristic for this study, where 1 proposed that there are both group-level and individual-level antecedents of justice expectations. and that these expectations are directly related to outcomes. At an individual level I focused on four groupings of direct antecedents of j ustice expectations. These are individual differences (race centrality and African centrality), existing beliefs and values (support for affirmative action and preference for merit principle), remote direct experiences (educational integration), and indirect experiences (social guidance regarding unfairness). Although not all anticipated, there were interesting results for individual differences that should drive future research. African centrality proved to play an important role in this context, with group differences in African centrality —justice expectation slopes for procedural, distributive, procedural restorative — blacks. distributive restorative — blacks, procedural restorative - women and distributive restorative — women expectations. Of the existing beliefs and values. individual support for affirmative action was related to general restorative justice expectations and a 244 preference for merit principle was related to procedural justice expectations. While both constructs are useful in this model, a more reliable measure of preference for merit principle should be used in future research. Remote direct experiences (educational integration) did not play a meaningful role, and I would drop this facet of the model from future framing. The indirect experience of social guidance regarding unfairness played a meaningful role in procedural and distributive justice expectations, and there were group differences in this slope. Social guidance appeared to be very important in explaining variance in justice expectations, and should receive greater attention in future research. At a group level I proposed examining group belief in white guilt as a contributor to group culture of support for affirmative action. This link was demonstrated. Future research with regard to white guilt might explore the within-group variability that was evidenced even though there were differences in mean levels across groups. I posited that group-level and individual-level predictors would contribute to variance in justice expectations on three justice dimensions (procedural, distributive, and restorative). Group support for affirmative action and group expectations regarding restorative justice were presented as having a cross-level effect on individual justice expectations. This was demonstrated to be true for climate of support for affirmative action on procedural justice expectations and general restorative justice expectations. However, group shared expectations of general restorative justice did not affect individual justice expectations as anticipated. There were mean group differences in individual justice expectations on several dimensions: procedural, distributive restorative — blacks and procedural restorative — women. There were also group differences in slope between certain justice expectation dimensions: general restorative justice and procedural justice, and general 245 restorative justice and distributive justice. The justice expectations were presented as being linked to important outcomes in this context (withdrawal intention. emigration intention, entrepreneurial intention, and education intention). The only meaningful anticipated links were the effect of procedural justice — women on withdrawal intention and general restorative justice expectation on emigration intention. However. there were group differences in levels of withdrawal, emigration and education intention, and that may be a practical focal point for future research. Direct links between normative factors such as group culture of support for affirmative action and social guidance regarding unfairness may be more appropriate than the fully mediated model that was proposed here. Also, this model did not include direct experiences such as being rejected. as most of the participants had not yet begun the application process. Future research should expand the individual antecedents to include direct experiences such as rejection. In summary, this study makes some important theoretical contributions. First. I have proposed an expansion of the current four—factor frame of organizational justice (Colquitt, 2001; Colquitt, Conlon, Wesson, Porter, & Ng, 2001), to include a fifth organizational justice factor; restorative justice that takes into account duality and simultaneity (Dawes, 1994). Such specification is necessary as the existing four-factor model is too simplistic for circumstances of strong preferential selection. Dawes suggests that an appropriate approach to affirmative action would be to recognize the duality and simultaneity principles by attempting to minimize unfairness, as the system can’t be fair to everyone at once. This study takes us a step closer to recognizing these principles by validating general restorative justice, and procedural restorative and distributive restorative justice for two target groups, women and blacks. This paper extends the 246 justice literature by capturing the possibility of distributions based on a broad social redistribution requirement rather than an individual need, and by considering perceptions regarding the handling of different target groups. This is different from examining the neediness of the individual recipient (e.g., Giacobbe-Miller, Miller, & Victorov. 1998: Murphy-Berman & Berman, 2002; Pratto, Tatar, & Conway-Lanz, 1999; Steiner, T rahan. Haptonstahl, & Fointiat, 2006). In the present study need and merit are not framed as exclusively alternative choices, as is the case in extant research (e.g.. Murphy-Herman & Berman, 2002; Pratto, Tatar and Conway-Lanz, 1999). Responding to Harrison, Kravitz, Mayer, Leslie, and Lev-Arey’s (2006) call for multi-level research, this study improves our understanding of how group membership. group cultures, and individual variables shape justice expectations in the context of strong preferential selection. Constructs at higher levels can exert independent influences (Kozlowski & Klein, 2000); this is a top-down contextual influence on individuals. In this case, I have demonstrated value in capturing group belief in white guilt, group culture of support for affirmative action, and group expectation of restorative justice. Importantly, I have demonstrated that both in-group members (whites) and out—group members (non-whites) build different views with regard to white guilt and responsibility. although there is variability within groups. I have also demonstrated that the group means for these variables have cross-level effects on individual justice expectations. However. individual-level factors also contribute to this sense-making (see Bell. Ryan. & Wiechmann, 2004). In this case I have demonstrated that social influence regarding unfairness and African centrality play important roles in justice expectations. Thus. both group-level and individual-level tiles must be part of the mosaic (see Chao & Moon. 247 2005). This work contributes to our understanding of specific value dimensions and the role of culture in justice, responding to Morris and Leung’s (2000) call for improving our understanding of the effects of specific cultural beliefs on justice perceptions. Also. I examine links between applicant expectations and outcomes (Anderson, 2004: Bell et al.. 2004) and identify group differences in withdrawal intention patterns, emigration intention, and entrepreneurial intention. By understanding differing justice expectations and resultant outcomes across cultural groups, organizations will be better equipped to manage and to meet the needs of their current and prospective employees. This may need to,be accomplished through aiding individuals in setting appropriate expectations and through implementing organizational processes that facilitate positive reactions. Understanding the effects of individual factors such as social influence and African centrality as well as group membership and group means on factors such as belief in white guilt, support for affirmative action, and restorative justice expectations may be a starting point for organizations and communities to consider ways in which they can help to build positive expectations among men and women of all races who are prospective job applicants. Hopefully this will result in applicants who are satisfied in=the hiring process experience. In the long term, this may improve organizational commitment of current employees. attraction of prospective employees, positive perceptions within the community. and skills retention in the country. 248 APPENDIX A Sample Questionnaire 249 Study of the Attitudes and Experiences of Potential Job Applicants in South Africa Thank you for participating in this study as a final-year student who will be in the market for a job in the next year. We will be asking you to respond to a series of questions about your expectations regarding the selection process that you will be facing. We will also ask about your background, experiences, and opinions. Please circle the appropriate number to mark your answer, or’fill in the answer where appropriate. Strongly Disagree Neither Agree Strongly disagree disagree agree nor agree 1 2 3 4 5 Thinking about the jobs for which you and your peers will be applying, the following items refer to company/organisation procedures used to make hiring decisions. To what extent do you agree: l. I expect that those procedures I 2 3 4 5 will be applied consistently across applicants. 2. I expect that those procedures I 2 3 4 5 will be free of bias across applicants. 3. I expect that those procedures 1 2 3 4 5 will be based on accurate information across applicants. 4. I expect that those procedures 1 2 3 4 5 will uphold ethical and moral standards. 250 Strongly Disagree Neither Agree Strongly disagree disagree agree nor agree I 2 3 4 5 Thinking about the jobs for Which you and year peers will be applying, the following items refer to hiring decisions about 44a": it: To: what extent do you agree: 5. I expect that the hiring decision I 2 3 4 5 about applicants will reflect the effort that applicants have put into their application. IJ! 6. I expect that the hiring decision I 2 3 4 about applicants will be appropriate for the work experience, education, or skills that applicants have. 7. I expect that the hiring decision 1 2 3 4 5 about applicants will reflect the potential that applicants have to contribute to the organization. 8. I expect that the hiring decision 1 2 3 4 5 about applicants will be justified, given their performance in the application process. 9. I expect that the hiring decision I 2 3 4 5 about applicants will reflect their need for the job. Thinking about the jobs for which yOu and your peers will be applying, the following items refer to company/organisation consideration of job applicants. To what extent do you agree: 10. I expect that special I 2 3 4 5 consideration in hiring will be provided to those from previously disadvantaged groups. 11. I expect that preference in l 2 3 4 5 hiring will be given to women, blacks, or the disabled. 12. I expect that an attempt will be 1 2 3 4 5 made to correct past social injustices in hiring. 13. I expect thatjobs will go to 1 2 3 4 5 people from affirmative action’s tar et beneficiar rou s. 251 Strongly disagree I Disagree Neither 2 disagree nor agree 3 Agree Strongly agree 4 5 Thinking about thejobs fiat which youmdyourpeers infill be applying, the following . item's referto melanin-80f. applimtsTomemnt do you agree: 14. I expect that the 1 company/organisation will treat applicants in a polite manner. 2 3 I 15. I expect that the l company/organisation will treat applicants with dignity. 16. i expect that the l company/organisation will treat applicants with respect. 17. i expect that the l company/organisation will refrain from improper remarks or comments. 18. I expect that the l company/organisation will be candid in its communication with applicants. 19. I expect that the l company/organisation will explain the procedures thoroughly. 20. I expect that the l company/organisation’s explanations regarding the procedure will be reasonable. 21. I expect that the l company/organisation will communicate details in a timely manner. 22. I expect that the l company/organisation will seem to tailor its communications to individual applieants’ needs. 252 Strongly Disagree Neither Agree Strongly disagree disagree agree nor agree 1 2 3 4 5 Thinking about the jobs for which you and your gem will be applying, the following items refer to «imprint/organisation i um used to make hiring decisions about 3959;; When selecting womaatowharexteatdeyemgree 23. I expect that those procedures I 2 3 4 5 will be applied consistently across women. 24. I expect that those procedures I 2 3 4 5 will be free of bias across women. 25. I expect that those procedures I 2 3 4 5 will be based on accurate information across women. 26. I expect that those procedures I 2 3 4 5 for women will uphold ethical and moral standards. 27. I expect that the hiring decision I 2 3 4 5 about women will reflect the effort ’ that they have put into their application. 28. I expect that the hiring decision I 2 3 4 5 about women will be appropriate for the work experience, education, or skills that they have. 29. I expect that the hiring decision I 2 3 4 5 about women will reflect the potential that they have to contribute to the organization. 30. I expect that the hiring decision I 2 3 4 5 about women will be justified, given their performance in the application process. 253 Strongly Disagree Neither Agree Strongly disagree disagree agree nor agree I 2 3 4 5 Thinking about the jobs for which you and yam peers will be applying, the following items refer to companylorg ' ‘ mad to make; hiring decisions about m (including black African, coldared, and Indian people). When selecting blacks. to what extent do you agree: 31. I expect that those procedures I 2 3 4 5 will be applied consistently across blacks. 32. I expect that those procedures I 2 3 4 5 will be free of bias across blacks. 33. I expect that those procedures I 2 3 4 5 will be based on accurate information across blacks. 34. I expect that those procedures I 2 3 4 5 for blacks will uphold ethical and moral standards. 35. I expect that the hiring decision I 2 3 4 5 about blacks will reflect the effort that they have put into their application. 36. I expect that the hiring decision I 2 - 3 4 5 about blacks will be appropriate for the work experience, education, or skills that they have. 37. I expect that the hiring decision I 2 3 4 5 about blacks will reflect the potential that they have to contribute to the organization. 38. I expect that the hiring decision I 2 3 4 5 about blacks will be justified, given their performance in the application process. 254 Next are some questions about your experience and interests. 39. How are you currently Full time Part time Not employed employed? 1 2 3 40. Have you begun looking for a Yes, I have I No. I have not IOb for after you graduate? begun looking ‘ begun looking I 2 (If No, SKIP TO QUESTION 45) 41. If you have begun looking, how long have you been looking for a job? months 42. If you have begun looking, how many jobs have you applied for in the past 6 months? jobs 43. Ifyou have begun looking, how many times in the past 6 months have you been turned down by a company/organisation (e.g., by letter, e-mail, telephone, or in person)? times 44. Ifyou have begun looking, how many times in the past 6 months have you been offered ajob? times 45. What types of jobs are of interest to you? (Please circle all that apply.) 1 Consulting 2 Education 3 Government 4 Industry 5 Non-profit 6 Other: 46. If there are organizations that I stand out to you as very attractive 2 places to work, please list your top 3 five. 4 5 255 Strongly Disagree Neither Agree Strongly disagree disagree agree nor agree I 2 3 4 5 Next are some questions regarding your close friends. 47. My close friends have said that I 2 3 4 5 the job application process is fair. 48. My close friends have I 2 3 4 5 suggested that my job applications will be fairly handled. 49. My close friends have said that l 2 3 4 5 the job application process is unfair. ' 50. My close friends have I 2 3 4 5 suggested that my job applications will be unfairly handled. Next are some questions regarding your close family members. 51. My close family members have I 2 3 4 5 said that the job application process is fair. 52. My close family members have I 2 3 4 5 suggested that my job applications will be fairly handled. 53. My close family members have I 2 3 4 5 said that the job application process is unfair. 54. My close family members have I 2 3 4 5 suggested that my job applications will be unfairly handled. Next are some questions regarding your co-workers in your current or previous jobs. 55. My co-workers have said that l 2 3 4 5 the job application process is fair. 56. My co-workers have suggested I 2 3 4 5 that my job applications will be fairly handled. 57. My co-workers have said that l 2 3 4 5 the job application process is unfair. 256 58. My co-workers have suggested I 2 3 4 5 that my job applications will be unfairly handled. Strongly Disagree Neither Agree Strongly disagree disagree agree nor agree I 2 3 4 5 Following are some questions about your feelings, behaviour, and intentions: 59. There is no point in my I 2 ' 3 4 5 applying forjobs. 60. I don’t intend to apply for many I 2 3 4 5 jobs. 61. I have taken steps toward 1 2 3 4 5 applying forjobs. 62. I intend to improve my I 2 3 4 5 education beyond this degree. 63. I frequently discuss getting I 2 3 4 5 fithher education beyond this degree with people who are close to me. «"4 64. I probably won’t put much I 2 3 4 S effort into applying forjobs. 65. I have taken steps toward l 2 3 4 5 studying further beyond this degree. 66. I intend to start my own I 2 3 4 5 business. 67. I frequently discuss starting my I 2 3 4 5 own business with people who are close to me. 68. l have taken steps to start my I 2 3 4 5 own business. 69. I am very likely to emigrate. l 2 3 4 5 70. I frequently discuss emigration l 2 3 4 5 with people who are close to me. 71. I have taken steps toward l 2 -. 3 4 5 emigration. 257 72. l have friends who have settled Yes, I have fiiends who No. I do not have abroad. friends have settled abI‘oad who have settled abroad I 2 73. l have family members who Yes, 1 have family No, I do not have have settled abroad. family members abroad members abroad I 2 Strongly Disagree Neither Agree Strongly disagree disagree agree nor agree I 2 3 4 5 Following are some questions about your feelings, behaviour, and intentions: 74. I am dissatisfied with I 2 3 4 5 conditions within South Africa. 75. I believe that my quality of life I 2 3 4 5 would be better overseas. 76. I am confident I get the success I 2 3 4 5 I deserve in life. 77. Sometimes I feel depressed. I 2 3 4 5 78. When I try, I generally succeed. I 2 3 4 5 79. Sometimes when I fail I feel I 2 3 4 5 worthless. 80. I complete tasks successfully. I 2 3 4 5 81. Sometimes, I do not feel in l 2 _ 3 4 5 control of my work. ' 82. Overall, I am satisfied with l 2 3 4 5 myself. 83. I am filled with doubts about I 2 3 4 5 my competence. 84. I determine what will happen in l 2 3 4 5 my life. 85. 1 do not feel in control of my I 2 3 4 5 success in my career. 86. I am capable of coping with l 2 3 4 5 most of my problems. 258 87. There are times when things I 2 3 4 5 look pretty bleak and hopeless to me. Strongly Disagree Neither Agree Strongly disagree disagree agree nor agree I 2 3 4 5 Following-are some , .. "I, gyeur befiefaand values. 88. In organisations, people who do I 2 3 4 5 their job well ought to rise to the top. 89. It is wrong for an employer to l 2 3 4 5 give a job to someone they know without first advertising the job to other candidates. 90. The effort a worker puts into a l 2 3 4 5 job ought to be reflected in the size of the raise he or she receives. 91. Promotion decisions ought to l 2 3 4 5 take into account the effort workers put into their job. Followingaresomequestions regardmgyow Motown"! affirmative action. 92. I believe that affirmative action I 2 3 4 5 is a good policy. 93. All in all, I oppose affirmative l 2 3 4 5 action plans in industry. 94. I believe that the goals of l 2 3 4 5 affirmative action are good. 95. I believe that affirmative action I 2 3 4 5 does more harm than good. 96. I believe that affirmative action I 2 3 4 5 is beneficial to society. 259 Strongly Disagree Neither Agree Strongly disagree disagree agree nor agree I 2 -. 3 4 5 Fellowing are some questions regaining. the hem and values of people Mm . _ I ll .. .mmt you think Indian men have 97. People of my race and gender 1 2 3 4 5 believe that affirmative action is a good policy. 98. All in all, people ofmy race I 2 3 4 5 and gender oppose affirmative action plans in industry. 99. People of my race and gender 1 2 3 4 5 believe that the goals of affinnative action are good. 100. People of my race and , l 2 3 4 5 gender believe that affirmative ‘ action does more harm than good. 101. People of my race and l 2 3 4 5 gender believe that affirmative action is beneficial to society. 102. People of my race and l 2 3 4 5 gender believe that whites should feel guilty that they have enjoyed benefits and privileges at the expense of non-whites. 103. People of my race and l 2 3 4 5 gender believe that whites are guilty of having received benefits and privileges simply because of their race. 104. People of my race and I 2 3 4 5 gender believe that whites should feel guilty about the past behavior of their race. 260 Strongly Disagree Neither Agree Strongly disagree disagree agree nor agree 1 2 3 4 5 Thinking about the jobs for which you and your peers will be applying, the following items refer to the beliefs of people of your race and gender regarding company/organisation consideration of job applicants. For example, if you are an Indian man, respond about the expectations that you think Indian men have. 105. People of my race and I 2 3 4 5 gender expect that special consideration in hiring will be provided to those from previously disadvantaged groups. 106. People of my race and l 2 3 4 5 gender expect that an attempt will be made to correct past social injustices in hiring. 107. People of my race and 1 2 3 4 5 gender expect that jobs will go to people from affirmative action’s targgt beneficiary groups. 108. People of my race and l 2 3 4 5 gender expect that preference in hiring will be given to women, blacks, or the disabled. Following are some demographic and identity questions. Your answers to these questions are very important as they allow us to look at results across different population groups. I 109. What year were you born? 110. Gender: Male Female I 2 111. Race: 1 Black African 4 White 2 Coloured 5 Other 3 Indian 261 112. Home language: I Afrikaans 7 Sesotho 2 English 8 Setswana 3 isiNdebele 9 siSwati 4 isiXhosa I0 Tshivenda 5 isiZulu ll Xitsonga 6 Sepedi 12 Other I Strongly Disagree Neither Agree Strongly disagree disagree agree nor agree I 2 3 4 5 Thinking of your race, please complete these questions: 113. Overall, being of my race I 2 3 4 5 has very little to do with how I feel about myself. 114. In general, being of my race I 2 3 4 5 is an important part of my self- image. 115. I have a strong sense of l 2 3 4 5 belonging to people of my race. 116. Being ofmy race is an I 2 3 4 5 important reflection of who I am. There are people of all races who view themselves as Afi'ican. Regardless of your race. please answer the following questions about being African. 117. Overall, being African has I 2 3 4 5 very little to do with how I feel about myself. 118. In general, being African is l 2 3 4 5 an important part of my self-image. I I9. I have a strong sense of l 2 3 4 5 belonging to African people. 120. Being African is an I 2 3 4 5 important reflection of who I am. 262 Strongly Disagree Neither Agree Strongly disagree disagree agree nor agree I 2 3 4 5 Next are some questions about crime. ' 121. I am worried about I 2 3 4 5 mugging, rape, or physical attack by a stranger. 122. I am worried about having I 2 3 4 5 my home broken into. 123. I am worried about having I 2 3 4 5 something of mine stolen. I Next arequesttons regardrng your backgrounds“ experEnees. 124. What was the racial composition of your class in your final year of high school? (Please indicate the percentage accounted for by each race.) Black African Coloured Indian White Other ++++ || TOTAL % % % o/O % I ()0 % 125. What was the gender composition of your class in your final year of high school? 1 Single sex (girls or boys only) 2 Co-educational (both girls and boys) 126. Do you fall into any of the groups that are designated to benefit from affirmative action? (Please mark all that apply to you.) I Black (includes black African. coloured. Indian) 2 Female 3 Disabled 263 127. What is you major field of study? I Science, engineering & technology (engineering. health science, life sciences, physical sciences, computer sciences and mathematical sciences) 2 Business & management (accounting. management, and all other business-related majors. such as marketing) 3 Education 4 Humanities (languages and literary studies. fine arts, music and the social sciences) q 128. If you do not intend to apply for jobs, please note the most important reasons behind this intention. 129. If you intend to emigrate, please note the most important reasons behind this intention. 130. If you intend to start your own business, please note the most important reasons behind this intention. 131. If you plan to study further beyond the present degree, please note the most important reasons behind this intention. 264 If you have any comments that you would like to make regarding your expectations or experience as ajob applicant in South Africa, please do so in the space provided. This is the end of this survey. Your answers are very important to us. Thank you for participating in this study. Please remain in your seat until the proctor calls you to hand in your survey. The proctor will provide you with a R100 Woolworths gift certificate in exchange for your completed survey. 265 APPENDIX B Sample Advertisement MICHIGAN STATE‘ U N I V E R SIT Y Are you a final-year undergraduate student? Will you be in the market for a job in the next year? We invite you to participate in an important research study on the attitudes and experiences of potential job applicants in South Africa today. As a study participant you will complete a paper-based survey regarding your attitudes and experiences. The survey should take about 30-45 minutes to complete. As a token of appreciation for your completing the survey we will give you a Woolworths gift certificate worth R100. Surveys will be administered on UCT campus on Thursday, March 26, Monday, March 30, and Tuesday, March 31, between 2:00 and 3:45 PM. To apply to take part in the survey, and sign up for a session, please visit www.understandingwork.com We hope to hear from men and women of all races. We will accept the first 200 participants who meet our sample needs. To be eligible to participate, you must be a final-year student at University of Cape Town, and be in the market for a job in the next year. Further information regarding this study can be found at www.understandingwork.com or by calling Lauren Ramsay at 076 716 0400. The major purpose of this project is to begin to investigate how individuals of different backgrounds perceive the hiring process, and hbw those backgrounds and that process might affect their attitudes, behavior or well-being. We will be asking participants to respond to a series of questions about their background and experiences and their observations and opinions about hiring processes in South Africa. Investigators: Neal Schmitt and Lauren Ramsay Department of Psychology Michigan State University East Lansing, MI 48824 USA 266 APPENDIX C Informed Consent 267 Research Participant Information and Consent Form STUDY OF THE ATTITUDES AND EXPERIENCES OF POTENTIAL JOB APPLICANTS IN SOUTH AFRICA You are being asked to participate in a research project. Researchers are required to provide a consent form to inform you about the study, to convey that participation is voluntary, to explain risks and benefits of participation, and to empower you to make an informed decision. You should feel free to ask the researchers any questions you may have. PURPOSE OF RESEARCH You are being asked to participate in a research study of the attitudes and experiences of potential job applicants in South Africa today. The major purpose of this project is to begin to investigate how individuals of different backgrounds perceive the hiring process. and how differences between people and groups, and expectations about the job application process, might affect individual attitudes, behavior or well-being. 0 In the entire study, 400 people from two universities are being asked to participate. ° You have been selected as a possible participant in this study because you indicated via our web-based sign up that you would like to take part. you are a final-year student, and you will potentially be in the market for ajob in the next year. . ° Your participation in this study will take about 30-45 minutes. ° If you are under 18, you cannot be in this study. WHAT YOU WILL DO In this project we will ask you to respond to a series of survey questions in a paper booklet. These questions are about your background and experiences and your observations and opinions about hiring processes in South Africa. 0 Participation in this study is voluntary and you are free to refuse to participate in this project or any part of the project. You may refuse to answer some of the questions and may discontinue your participation at any time. ' A summary of the study findings will be posted at www.understandingwork.cmn at the conclusion of this study. If you would like to receive the results of this study via e-mail, please provide your e-mail address in the space provided at the end of this form. 268 POTENTIAL BENEFITS You will not directly benefit from your participation in this study. However, your participation in this study may contribute to the understanding of the attitudes and experience of potential job applicants in South Africa. POTENTIAL RISKS The potential risks of participating in this study are some possible discomfort associated with potentially sensitive questions about race and gender. PRIVACY AND CONFIDENTIALITY lnforrnation about you will be kept confidential to the maximum extent allowable by law. Only the project team, and the Institutional Review Board if necessary, will have access to the survey responses. ' You will be assigned a unique identification number and the data file containing survey responses will not identify your name. The key linking names and identification numbers will be stored separately under lock and key. All electronic data will be stored securely on a password-protected computer. Data will be stored for five years. ' The results of this study may be published or presented at professional meetings. but the identities of all research participants will remain anonymous. All results of this study will be reported at a group level or anonymously so that no one will be able to identify a particular person. ' One year after the completion of this study you will be contacted via e-mail and invited to participate in a follow-up survey. The follow-up survey will cover similar questions, and your responses for the follow-up studies would be linked to your responses to this current study using your unique identification number so that we can see how your attitudes have changed over time. You would be compensated separately for taking part in that and any other follow-up studies related to this project. Your participation is voluntary. YOUR RIGHTS TO PARTICIPATE, SAY NO, OR WITHDRAW Participation in this research project is completely voluntary. You may change your mind at any time and withdraw. You may choose not to answer specific questions or to stop participating at any time. COSTS AND COMPENSATION FOR BEING IN THE STUDY You will receive compensation of a R100 gift certificate for submitting the completed survey. That is, if you skip a few questions, but have in essence completed the survey, you will receive a R100 certificate. However, if you complete only about half of the survey, you will receive only a R50 gifi certificate. There is no compensation provided if 269 you choose to withdraw from the survey, or for responding to fewer than about half of the questions. CONTACT INFORMATION FOR QUESTIONS AND CONCERNS If you have concerns or questions about this study, please contact the researchers, Prof. Neal Schmitt or Ms. Lauren Ramsay Department of Psychology, Michigan State University, 2628 Psychology Building. East Lansing, Michigan 48824, USA E-mail: schmitt@msu.edu. Phone in the US: +1 517-355-9563. Phone in South Africa: 076-716-0400. If you have any questions or concerns about your role and rights as a research participant, or would like to register a complaint about this study, you may contact, anonymously if you wish, the Director of MSU’s Human Research Protection Program, at 517-355-2180. Fax 517-432-4503, or e-mail irb@msu.edu or regular mail at 202 Olds Hall, MSU. East Lansing, MI 48824, USA. DOCUMENTATION OF INFORMED CONSENT Your signature below means that you voluntarily agree to participate in this research study. Signature Date Would you like to receive a summary of study results via e-mail? Cl Yes Cl No If yes, please provide your e-mail address: May we contact you via e-mail in the future to invite you to take part in follow-up studies? [3 Yes Cl No If yes, please provide your e-mail address: You will be given a copy of this form to keep. 270 APPENDIX D Correlation Matrices and Descriptive Statistics for Items Used in Confirmatory Factor Analyses 271 C onfirmatory Factor Analysis —— Validation of General Restorative Justice Descriptive Statistics for Items 1 to 8 and 10 - 22 Item Mean Std. Deviation V1 3.59 1.21 V2 3.33 1.38 V3 3.88 1.05 V4 3.78 1.16 V5 3.58 1.1] V6 3.89 1.03 V7 3.91 0.99 V8 3.70 1.01 V10 3.69 1.24 V11 3.51 1.32 V12 3.78 1.16 V13 3.46 1.28 V14 4.15 0.79 V15 4.19 0.75 V16 4.18 0.78 V17 4.02 0.95 V18 3.81 0.93 V19 4.02 0.97 V20 3.96 0.84 V21 3.86 1.00 V22 3.17 1.14 Note. N = 327. 272 00.0- m _ .0- 0 0 .0- and R0 0 _ .0 0m.0 0N0 0N0 Nm.0 «N0 mm> 2 .0- mm.0- g0- 0m.0 mm.0 mm .0 0N0 0V0 3.0 010 mm .0 _m> 3 .0- 0m.0- BNO- 0m .0 3.0 mm.0 7N0 5&0 3.0 3.0 3.0 0~> 3.0- 0N.0- 0N0- 0m.0 3.0 3.0 mm.0 3&0 0V0 00.0 3.0 0_> 00.0- m H .0- 3.0- $0 NN.0 3.0 w _ .0 - mm.0 0m.0 0m.0 Vm.0- . 00> 50.0- m _ .0- w _ .0- $0 0N0 mN.0 3.0 000 3.0 5.0 5.0 :> m _ .0- mm.0- 0N0- 0N0 N00 N~.0 NN.0 3.0 0m .0 3.0 0m.0 0:» S .0- 3.0. 0N.0- 0N0 0N0 2.0 3.0 3X0 0m.0 3.0 N00 m_> m0.0- h _ .0- v0.0- 8.0 mN.0 3.0 3.0 0m.0 3.0 2.0 0N0 3> No.0 mod $0 0N0- 0N.0- 0N0- mm.0- bar-.0- 3:0- vvd- _m.0- m_> 00._ $0 0m.0 50.0- 00.0- :0- m _ .0- N. _ .0- 2.0- 5.0- 00.0- m_> 00._ 00.0 0N.0- mm.0- 0N0- ~N.0- mm.0- 3:0- 0m.0- mN.0- :> 00.0 $0. 0N0- hm.0- vmd- mm.0- 3.0- $0. 8.0- 0 ~> 00.0 0m.0 0m.0 :0 000 and 000 3.0 w> 00._ 0m.0 0Y0 0m.0 0m.0 R0 30 n> 00. _ 0V0 3.0 2.0 0v.0 «no.0 0> 00.0 3.0 000 0V0 0m.0 m> 00.0 $0 3.0 N00 v> 00._ 00.0 cum .0 m> 00A mud m> 00A ~> N_> :> 0_> w> m> 0> m> v> m> N> ~> mm - 0 N ES 0 - N was: .5\»..EBV< 26.239269 273 nmm -. .2. .202 00. _ 2V0 v00 $0 3.0 3.0 mm.0 0N0 0m.0 0N0- mm> 00._ 0V0 3.0 3.0 Nm.0 vvd No.0 3.0 3.0- _m> 00.0 3.0 mm.0 000 no.0 3.0 mm.0 mm.0- 0m> 00._ 0V0 Vm.0 0V0 00.0 wmd mm.0- 0_> 00. N 5x0 Nvd 0V0 3.0 NN.0- w _ > - . 00; $0 3.0 3.0 NN.0- :> 00._ 00.0 00.0 ~00- 0_> 00. N 050 0N0- w _ > 00._ 2.0- E> 00; m_> N_> :> 0_> w> n> 0> m> v> m> N> _> NN> _N> 0m> 00> w_> C> 0_> mN> 3> m_> fiozttzoo mm - 0 N has m -N mEoN SNXESN 26.259290 274 C onfirmatory Factor Analysis — Validation of Procedural Restorative and Distributive Justice for Women and Blacks Descriptive Statistics for Items 1 to 8 and 23 - 38 Item Mean Std. Deviation V1 3.59 1.21 V2 3.33 1.38 V3 3.88 1.05 V4 3.78 1.16 V5 3.58 1.11 V6 3.89 1.03 V7 3.91 0.99 V8 3.70 1.01 V23 3.65 1.10 V24 3.53 1.23 V25 3.90 0.97 V26 3.87 1.03 V27 3.70 1.08 V28 3.93 1.03 V29 3.98 1.00 V30 3.82 1.01 V31 3.62 1.23 V32 3.53 1.29 V33 3.73 1.10 V34 3.75 1.14 V35 3.57 1.24 V36 3.66 1.27 V37 3.79 1.19 V38 3.64 1.14 Note. N = 327. 275 nm0 mm0 mm0 3.0 30 mm0 mm0 000 nm0 3H0 im0 wm> $0 000 mm0 0V0 $0 9.0 $0 nv0 No.0 Q0 3.0 nm> 3.0 30 $0 000 nm.0 0m0 mm0 no.0 0V0 0V0 nm0 0m> wm0 nm0 $0 $0 nm0 :0 Q0 :0 wm0 nm0 0N0 mm> 0Y0 Nm0 0V0 nm0 30 000 Nm0 nm0 $0 $0 $0 Vm> 00.0 0m0 3.0 0m0 Nm0 nN0 nm0 0V0 0m0 0V0 0V0 mm> 0m0 5.0 0m0 NNO nm0 mm0 vm0 Nm0 90 N00 30 Nm> $0 .0m0 0W0 NNO 2.0 $0 NNO. wm0 wm0 $0 000 Tm> mm0 0Y0 30 $0 000 0V0 3.0 $0 mm0 nm0 mm0 0m> mm .0 no.0 nv0 mm0 30 ~00 mm0 wm0 nm0 $0 mm0 0N> 30 Nm0 010 $0 mm0 0m0 mm0 0m0 wm0 $0 $0 0N> mm0 nv0 wm0 mm0 mm0 0m0 0V0 3V0 mm0 0m0 mm0 nm> N00 30 0V0 0m0 mm0 mm0 0N0 mm0 0m0 Nm0 0V0 0m> 00; 000 mm .0 0N0 mm0 3.0 {m0 0Y0 Vm .0 00.0 0V0 mm> 00._ 00.0 Nm0 000 0m0 30 $0 30 v0.0 30 vm> 00; Nm0 mm0 0m0 0N0 wm0 mm0 3.0 30 mm> 00.0 0m0 0m.0 nv0 000 wm0 wm0 000 w> 00.0 $0 $0 000 0m0 nm0 X0 n> 00._ 0V0 3.0 nm0 0V0 $0 0> 00.0 3.0 2.0 0V0 0m0 m> 00.0 N00 mn0 No.0 v> 00.0 00.0 $0 m> 00._ mn0 N> 00; _> mm> vm> mm> w> n> 0> m> v> m> N> _> mem- - MN has 0 -N mESN Sneaks: totontoU 276 nmm u .2... 2:2. vn0 0n.0 00.0 mm0 mm0 30 0m0 nm0 3.0 00.0 «.00 0m0 wm> 00; nn0 0n.0 N00 nm0 0V0 wm0 2&0 0m0 $0 .30 3&0 nm> 00._ vn0 00.0 0n.0 mm0 NWO 0V0 E0 30 $0 3.0 0m> 00._ 0m0 mm0 $0 3.0 n00 0V0 3.0 mm0 mm0 mm> 00.0 n00 $0 $0 0m0 $0 $0 000 $0 Vm> 00.0 0n.0 M00 vv0 N10 NWO 0m0 3.0 mm> 00; 0n.0 3.0 30 0m0 wm0 $0 mm> 00.0 wm0 mm0 wm0 30 N00 _m> - - 00.0 $0 $0 . - 30 ~60 0m> 00._ 0n.0 mm0 mm0 0N> 00. _ v00 mm0 wm> 00A 30 nm> 00._ 0N> mm> vm> mm> w> n> 0> m> v> m> N> _> nm> 0m> mm> Vm> mm> Nm> _m> 0m> 0m> wN> nm> 0m> 083538 0m - mm N38 0 - N was: 350.25% 29:33.50 277 C onfirmatory Factor Analysis — Invariance Across Gender Women: Descriptive Statistics for Justice Items Item Mean Std. Deviation V1 3.62 1.28 V2 3.34 1.42 V3 3.96 1.07 V4 3.95 1.14 V5 3.64 1.15 V6 3.83 1.08 V7 3.96 1.00 V8 3.62 1.06 V10 3.56 1.27 V11 3.43 1.34 V12 3.73 1.22 V13 3.40 1.30 V23 3.70 1.15 V24 3.59 1.23 V25 4.01 0.93 V26 3.94 1.09 V27 3.75 1.12 V28 3.94 1.07 V29 3.96 1.06 V30 3.83 1.06 V31 3.56 1.26 V32 3.44 1.29 V33 3.72 1.11 V34 3.75 1.19 V35 3.51 1.28 V36 3.55 1.30 V37 3.70 1.23 V38 3.51 1.20 Note. N: 162 278 cm0- 0.0- 0m0- 0m0- :0 :0 9.0 3.0 9.0 .30 mm0 0M0 wm> 3.0- m . .0- 0m0- wm0- 9.0 000 0.0 5.0 3.0 3.0 3.0 3.0 nm> 3.0- 0m0- 3:0- mm0- mm0 0.0 mm0 0n.0 0m0 5.0 0W0 9.0 om> mv.0- 30- 3.0- 5.620- 00.0 30 nm0 3.0 3.0 3.0 3.0 wm0 mm> $.0- N.0- 30- .m0- $0 30 ”V0 3.0 3.0 «$0 .060 2.0 Ym> 0v0- 0.0- 0N0- NNO- 0m0 ~60 0m0 0N0 3.0 000 3.0 30 mm> ~60- » . .0- mm0- wm0- BNO 0V0 mm0 0m0 3.0 5.0 .m0 :0 Nm> 3.0- 30- 0m0- wm0- mm0 E0 v~0 5N0 mmd 0.0 9.0 5.0 .m> 0N0- m . .0- om0- 0N0- 5.0 30 0V0 mm0 cm0 0N0 mm0 0N0 0m> 0m0- v00- wm0- 0N0- .m0 090 0n.0 3.0 3.0 3.0 .m0 3.0 0N> 3.0- 0.0- .09- 30. 0m0 . 5‘0 000 - «v.0 mmd $0 $0 3.0 wm> nm0- 000- 000. mN0- 5H0 0n.0 $0 3.0 mvd ©m0 hm0 3H0 hm> .N0- .00 0N0- m . .0- 30 ov0 hm0 .m0 000 .00 000 $0 0N> 000- .00- mm0- 30. 3.0 9.0 hm0 20 3.0 000 3.0 3.0 mm> 3.0- m . .0- mm0- ©m0- 30 :0 3.0 0W0 nm0 E0 30 Nm0 v~> mm0- 000- .00. mm0- 3.0 mm0 0n.0 0V0 5‘0 3.0 30 «m0 MN> 00.. 30 80 N00 0n.0- ~m0- 0N0- 0N0- mm0- Nm0- 0V0- hN0- m .> 00. . 00.0 000 000- 000- . .0- v . .0- N . .0- 500- N . .0- N00 N . > 00.. ~00 mm0- 0m0- EHO- 5N0- mm0- 5N0- vm0- mm0- . .> 00. . mm0- 5N0- .m0- 0N0- 0m0- mm0- vM.0- vm0- 0 .> 00. . v0.0 000 .m0 3.0 0m0 0V0 30 w> 00. . $0 3.0 3.0 9.0 9.0 0m0 h> 00. . .00 3.0 0V0 mm0 mv0 0> 00. . NV... .v0 30 mm0 m> 00.. $0 3.0 $0 v> 00.. N50 000 m> 00.. «50 m> 00.. .> m.> N.> ..> 0.> w> h> ©> m> v> m> m> .> .82»: 853. 2.23:. 29:39:00 23835 279 00.0 5.0 3.0 3.0 00.0 0m0 000 3.0 3.0 3.0 3.0 .30 wm> 0....0 v.00 0m0 0.0 0.0 .00 3.0 0V0 2.0 0W0 w..0 3.0 mm.) 00.0 0W0 .m0 N...0 2.0 .00 ~00 3.0 2.0 3.0 00.0 mm0 0m> m00 5.0 0m0 0...0 09.0 0m0 #00 ~00 0m0 0.0 000 5.0 mm> 00.. 05.0 N00 :0 3.0 00.0 Nm0 3.0 .00 R0 ..00 5.0 00> 00.. 3.0 ..00 3.0 .m0 3.0 2.0 $0 000 00 N00 mm> 00.. :0 nm0 000 2.0 0..0 mm0 00.0 0n.0 3.0 Nm> 00.. ..m0 wm0 3.0 .00 ..m0 2.0 mm0 000 .m> 00.. 00.0 00.0 .00 N00 5.0 00.0 3.0 0m> - 00.. 050 M00 - mm0 N00 ..00 0m0 .. 0N> 00.. .00 00.0 000 000 00.0 0N> 00. . 3.0 50.0 .00 N00 hN> 00.. 000 000 Nm0 0N> 00. . 050 M00 0N> 00.. :0 ..N> 00.. MN> 0.> N.> ..> 0.> w> N.> 0> m> ..> m> N> .> ..m> mm> Nm> .m> 0m> 0N> 0N> h 0N> 0N> ..N> MN> 0.2.5.28 Shut 8.33. 3:33 29:53.89 "5:83 280 00. . :0 00.0 0....0 00.. 2.0 v.50 00.. 0n.0 00.. wm> nm> 0m> m2 .. a ...3_.,,._. mm> m> 5m.) 0m.) mm> m> mm> mm> .m> 0m> 0N> 0N> BN> 0N> 0N> VN> MN> m_> N.> _ _> 0.> w> N.> 0> m> ..> m> N> .> wumzfittpu WENCN 90.3035 00.:ch!» SORBNNLKQU 23:83 281 Men: Descrigtive Statistics for Justice Items Item Mean Std. Deviation V] 3.56 1.14 V2 3.33 1.33 V3 3.8] 1.04 V4 3.62 1.16 V5 3.53 1.07 V6 3.95 0.97 V7 3.86 098 V8 3.78 0.95 V10 3.81 1.20 V11 3.59 1.29 V12 3.82 1.11 V13 3.52 1.27 V23 3.59 1.06 V24 3.48 1.22 V25 3.78 1.01 V26 3.81 0.98 V27 3.66 1.04 V28 3.92 1.00 V29 3.99 0.94 V30 3.811 0.96 V31 3.68 11.19 V32 3.62 1.29 V33 3.75 1.10 V34 3.75 1.10 V35 3.64 1.20 V36 3.76 1.24 V37 3.87 1.16 V38 3.76 1.06 Note. N = I65 282 all : |.|!Illl.n.-Li ‘lfllllllnllll ’- 0N0- 000- 0.0- 0n.0- R0 0n.0 0n.0 mm0 00.0 wm0 mm0 ..m0 my...) 3:0- 30. 0N0- wm0- mm0 0m0 2.0 :20 NWO 2.0 2.0 0n.0 nm> .m0- m_.0- mm0- mm0- 3:0 0n.0 5.0 0N0 00.0 3.0 ~40 ~00 0m> 0N0- 0..0- 0N0- wm0- 000 0.0.0 0m0 3.0 .30 .320 0m0 0.0 mm> 0m0- m . .0- .m.0- mm0- mm0 mm0 0.0.0 ..m0 000 3.0 mm0 3:0 0m> 3.0- 000- 0.0- 0..0- 0N0 Nm0 mm0 0N0 0n.0 $0 000 0n.0 mm> 0N0- ...0- 0.0- 0.0- 0.0 0.0 0.0 0.0 00.0 00.0 $0 9.0 mm> 0N.0- ..00- 00.0- 0. .0- 0. .0 0. .0 h . .0 h . .0 .30 0m0 3.0 00.0 .m> wmd- 00.0- 0.0- 3.10- mv0 NWO 00.0 3.0 00.0 30 N00 ~00 0m> 0N0- v . .0- 0 . .0- 3.0.- 0. .0 .m0 .~.0 0 . .0 .m0 0N0 0N0 .~.0 0N> NNO- m . .0- mm0- 0n.0- 3:0 80 00.0 N. . .0 00.0 mm0 $0 3.0 0N> 0N0- 0. .0- N. . .0- NNO- mm0 0N0 0N0 2.0 000 .m0 0n.0 0m0 nm> 0N0- m . .0- mm0- 0N0- 0N0 NNO 0N0 0n.0 ”V0 5.0 3.0 000 0N> ..m0- 0 . .0- 0 . .0- w . .0- 0N0 0N0 mmd :0 00.0 00.0 00.0 00.0 0N> mm0- .m0- 0N0- mm0- 0N0 0m0 0n.0 0N0 ”.00 2.0 M00 mm0 v~> mm.0- 00.0- N . .0- 0. .0- w . .0 m . .0 0N0 N. . .0 NN.0 ~N0 0V0 mm0 m~> 00. . 00.0 00.0 00.0 mm0- mN0- 0N0- NNO- wm0- 0n.0- 3.0- 0m0- m . > 00.. 00.0 00.0 00.0- 0.0- N..0- N..0- NNO- 0..0- N00- .N.0- N.> 00.. 3.0 NNO- 0.0- NNO- 3.0- .000. mm0- 0m0- 0N.0- . .> 00. . w ..0- 0.0.0- vm.0- h . .0- 0n.0- 0N0- mv0- 0N0- 0 . > 00. . 5.0 3.0 3.0 nm0 0V0 00.0 0N0 w> 00.. 3.0 Nm0 .00 mm0 0m0 0N0 ..> 00.. 0m.0 000 ..m0 000 3.0 0> 00.. 00.0 0N0 mm0 0N0 m> 00.. 00.0 05.0 .00 ..> 00.. .00 3.0 m> 00.. 3.0 m> 00.. .> m.> N.> ..> 0.> 0> >> 0> m> ..> m> m> .> at»: 8.53. 5.5.3. 20.239209 :32 283 3... 2... 3.... 3.... mm... .2. .2. 2... .2. mm... 8... 8... w; 2... 8... 5.... mm... 3.... 2.... 8... mm... 8... 3.... 8... mm... 2.» 2... 8... mm... 9.... 8... 8... 9.... 8... m2. 2... 8... 9.... 9...) S... 8... 3.... R... 8... 3.... mm... mm... 8... mm... 2... o... mm> 8.. 8... 8... E... 2... R... 8... R... 2.... m2. 8... mm... 35 8.. 8... 8... 8... 8... 8... 8... 3.... 8... 8... 8... m9 8.. z... 2.... 8.... 3.... 8... 8... 3.... mm... 2.... N; 8.. Q... E... 8... mm... .m... 9.... S... a... .9 8.. 8... .3. a... .2. 8... .2. .2. 8> - 8.. .3. :8 mm... a... 8.... R... . 8> 8.. 8... 8... mm... 3.... mm... 8> 8.. 8... 8... 2.... mm... R> 8.. 8... 3... 2.... om> 8.. S... 3.... m9 8.. a... ..9 8.. m9 m.> N.> ..> o.> ..> 3 o> m> ..> .9 ~> .> ..9 35 N9 .m> 8> 0N> ..9 R> 8> m9 ..~> m9 “50:22:39 WEN: @92033 .353: QQCBNNKKQD 2.9.). 284 M 'I‘a c. n 2.. 3:2. 00.. m..-0 ~00 000 «..> 00.. 00.0 00.0 mm.) 00.. ~00 0m> 00.. mm> vm> mm> mm> .m> 0m> 0N> 0N> 0N> 0N> mm> vm> mm> m.> N.> ..> 0.> w> ..> 0> m> ..> m> N.> .> wm> nm> 0m> mm> fim::.:=8 was: «5.22.0 £233. 5.263..er 23.). 285 Women: Descriptive Statistics for Race C entrality Items Item Mean Std. Deviation V1 13 2.37 1.22 V114 3.34 1.25 V115 3.41 1.17 V116 3.35 1.19 Note. N: 162. Women: Correlations for Race Centrality Items V113 V114 V115 V116 V113 1.00 V114 0.33 1.00 V115 0.13 0.62 1.00 V116 0.30 0.66 0.74 1.00 Note. N= 162. Men: Descriptive Statistics for Race C entralfl» Items Mean Std. Deviation V113 2.45 1.29 V114 3.36 1.25 V115 3.71 1.12 V116 3.50 1.30 Note. N= 165. Men: Correlations Lor Race Centrality Items V113 V114 V115 V116 V113 1.00 V114 0.24 1.00 V115 0.11 0.53 1.00 V116 0.26 0.61 0.68 '1 .00 Note. N = 165. 286 Women: Descriptive Statistics for Afi'ican C entralitx Items Item Mean Std. Deviation V117 3.09 1.18 V118 3.46 1.09 V119 3.54 1.10 V120 3.56 1.08 Note. N= 162. Women: Correlations for A fiican Centrality Items V117 V118 V119 V120 V117 1.00 V118 0.35 1.00 V119 0.20 0.72 1.00 V120 0.23 0.75 0.76 1.00 Note. N = 162. Men: Descriptive Statistics for African C entrality Items Mean Std. Deviation V117 3.08 1.29 V118 3.73 1.12 V119 3.76 1.12 V120 3.72 1.17 Note. N = 165. Men: Correlations for A fiican Centrality Items V117 V118 V119 V120 V117 1.00 V1.18 0.33 1.00 \ V119 0.21 0.73 1.00 V 120 0.30 0.86 0.78 1.00 Note. N= I65. 287 Women: Descriptive Statistics for A flirmative Action Items Item Mean Std. Deviation V92 2.98 1.19 V93 3.11 1.20 V94 3.54 1.03 V95 2.72 1.15 V96 3.04 1.10 Note. N -——- 162. Women: Correlations for A flirmative Action Items V92 V93 V94 V95 V96 V92 1.00 V93 0.64 1.00 V94 0.64 0.51 1.00 V95 0.72 0.68 0.49 1.00 V96 0.79 0.65 0.51 0.70 1.00 Note. N = 162. Men: Descriptive Statistics for Aflirmative Action Items Item Mean Std. Deviation V92 3.12 1.23 V93 3.18 1.25 V94 3.77 1.07 V95 2.83 1.22 V96 3.02 1.17 Note. N = 165. Men: Correlations for A flirmative Action Items V92 V93 V94 V95 V96 V92 1.00 V93 0.58 1.00 V94 0.56 0.34 1.00 V95 0.56 0.53 0.34 1.00 V96 0.70 0.49 0.49 0.65 I .00 Note. N = 165. 288 Women: Descriptive Statistics for Merit Items Item Mean Std. Deviation V88 4.46 0.75 V89 4.28 0.98 V90 4.17 0.93 V91 4.53 0.63 Note. N = 162. Women: Correlations for Merit Items V88 V89 V90 V91 V88 1.00 V89 0.03 1.00 V90 0.24 0.15 1.00 V91 0.21 0.13 0.38 1.00 Note. N = 162. Men: Descriptive Statistics for Merit Items Item Mean Std. Deviation V88 4.37 0.83 V89 4.05 1.13 V90 4.21 0.92 V91 4.43 0.74 Note. N = 165. Men: Correlations for Merit Items V88 V89 V90 V91 V88 1.00 V89 0.16 1.00 V90 0.31 0.28 1.00 V91 0.41 0.16 0.52 1.00 Note. N: 165. 289 Women: Descriptive Statistics for Social Guidance Items Item Mean Std. Deviation V49 3.55 1.00 V50 3.12 1.01 V53 3.42 1.03 V54 3.17 1.04 V57 3.23 0.89 V58 3.01 0.85 Note.N=151. Women: Correlations for Social Guidance Items V49 V50 V53 V54 V57 V58 V49 1.00 V50 0.64 1.00 V53 0.34 0.33 1.00 V54 0.40 0.50 0.69 1.00 V57 0.42 0.43 0.31 0.35 1.00 V58 0.30 0.50 0.29 0.42 0.60 1.00 Note. N = 151. Men: Descriptive Statistics for Social Guidance Items Item Mean Std. Deviation V49 3.38 1.08 V50 3.21 0.99 V53 3.22 1.15 V54 3.07 1.10 V57 3.25 1.02 V58 3.14 0.98 Note. N= 148. Men: Correlations for Social Guidance Items V49 V50 V53 V54 V57 V58 V49 1.00 V50 0.79 1.00 V53 0.60 0.48 1.00 V54 0.56 0.57 0.80 1.00 V57 0.45 0.43 0.45 0.44 1.00 V58 0.44 0.48 0.43 0.51 0.82 1.00 Note. N = 148. 290 Women: Descriptive Statistics for Group A flirmative Action and Group Guilt Items Item Mean Std. Deviation V97 3.04 1.398 V98 3.09 1.302 V99 3.34 1.227 V100 2.92 1.266 V101 3.04 1.213 V102 2.66 1.379 V103 3.07 1.354 V104 2.88 1.313 Note. N= I62. Women: C orrelations for Group Affirmative Action and Group Guilt Items __ V97 V98 V99 V100 V101 V102 V103 V104 V97 1.00 V98 0.65 1.00 V99 0.72 0.64 1.00 V100 0.69 0.69 0.62 1.00 V101 0.71 0.60 0.75 0.72 1.00 V102 0.45 0.41 0.33 0.40 0.35 1.00 v103 0.44 0.39 0.34 0.37 0.35 ‘ 0.69 1.00 V104 0.33 0.33 0.20 0.27 0.27 0.62 0.60 1.00 Note. N = 162. Men: Descriptive Statistics for Group A flirmative Action and Group Guilt Items Item Mean Std. Deviation V97 2.81 1.281 V98 3.01 1.176 V99 3.29 1.158 V100 2.85 1.218 V101 2.95 1.183 V102 2.88 1.504 V103 2.99 1.366 V104 2.99 1.304 Note. N = 165. 291 Men: Correlations for Group A flirmative Action and Group Guilt Items V97 V98 V99 V100 V101 * V102 V103 V104 V97 1.00 V98 0.55 1.00 V99 0.63 0.42 1.00 V100 0.69 0.70 0.65 1.00 V 101 0.73 0.55 0.65 0.78 1.00 V102 0.27 0.16 0.22 0.28 0.32 1.00 V103 0.27 0.14 0.21 0.19 0.26 0.63 1.00 V104 0.22 0.19 0.15 0.25 0.21 0.53 0.41 1.00 Note. N = 165. Women: Descriptive Statistics for General Restorative Justice Items Item Mean Std. Deviation V105 3.64 1.06 V106 3.79 0.86 - V107 3.71 0.96 V108 3.75 1.09 Note. N = 162. Women: Correlations for General Restorative Justice Items V105 V106 V107 V108 V105 1.00 V 106 0.54 1.00 V107 0.48 0.26 1.00 V108 0.49 0.28 0.66 1.00 Note. N = 162. Men: Descriptive Statistics for General Restorative Justice Items Item Mean Std. Deviation V105 3.73 0.98 V106 3.70 0.95 V107 3.79 0.96 V108 3.79 1.09 Note. N = 165. 292 Men: Correlations for General Restorative Justice Items V105 V106 V107 V108 V105 1.00 V106 0.50 1.00 V107 0.46 0.41 1.00 V108 0.47 0.35 0.71 1.00 Note. N: 165. Women: Descriptive Statistics for Intentions Items Item Mean Std. Deviation V59 1.42 0.72 ' V60 1.54 0.91 ' V61 2.48 1.07 V62 4.38 0.85 V63 4.33 0.89 V64 1.56 0.91 V65 3.51 1.19 V66 3.24 1.21 V67 2.95 1.28 V68 2.31 1.09 V69 3.35 1.34 V70 3.41 1.34 V71 2.34 1.15 Note. N = 162. 293 .N0. u 2. .202 .00 00.0 mm0 ..0 :0 0.0 .00 N00- 000- N00 0.0 0. .0 .0> 00. . N00 .m0 0 . .0 m00 . . .0 00.0 .00- .00- 000 00.0- N . .0 00> 00. . w . .0 000 N00 00.0 .00 000- 00.0 .00- 000- . .0 00> 00. . 000 000 00.0 00.0 00.0 00.0 00.0- 000 0. .0 00> 00. . 00.0 00.0 00.0 v . .0 . .0 000- 00.0- 00.0 00> 00. . NNO m . .0- 0 . .0 m . .0 M00- 00.0- .00 00> 00.. Q0- ¢m0 Nm0 000- 000- .000 00> 00.. NNO- 0.0- N..0 wm0 0N0 00> 00.. N00 00.0- 0 . .0- #00- m0> 00.. 0m0- .00 .00- N0> 00.. m..0 N..0 .0> 00.. 30 00> 00.. 00> 00> 00> 00> 00> 00> 00> 00> m0> m0> .0> 00> 00> 2:20 8.20.2225 xo\.m=c...5m.tc.v "5835 294 Men: Descriptive Statisticsjbr Intentions Items Item Mean Std. Deviation V59 1.48 0.85 V60 1.75 1.01 V61 2.39 1.03 V62 4.33 0.87 V63 4.26 0.93 V64 1.86 1.06 V65 3.50 1.19 V66 3.56 1.28 V67 3.38 1.33 V68 2.62 1.27 V69 3.15 1.30 V70 3.37 1.31 V71 2.35 1.16 Note. N: 165. 295 .-I .00. H2 .802 .00 000 .Vm0 00.0 :0 0N0 0. .0 m00 m00 v. .0- N. .0 0N0 .0> 00.. m00 00.0 m . .0 m00 m . .0 00.0 .N0 000 :0- 00.0 m . .0 00> 00. . 00.0 .00 N00- 3 .0 00.0 000 .000 N00 0 . .0 0 _ .0 00> 00.. 00.0 000 .N0 000 0.0 w . .0 :0- 000- m : .0 00> 00.. 00.0 00.0 00.0 .N0 v.0 0:0- 0.0- 00.0 00> 00. . 00.0 00.0 v. .0 0. .0 00.0- 0 . .0- 00.0 00> 00.. :0- mm0 .00 0 . .0- m. .0- :0 00> 00. . ~00 0 _ .0- w : .0 0V0 g0 V0> 00. . 00.0 .00- 000- 00.0- m0> 00. . 000- 30.. 000- N0> 00.. 30 0.0- .0> 00.. 0V0 00> 00.. 00> 00> 00> 00> 00> 00> 00> 00> m0> m0> .0> 00> 00> ES: ESEES x0\.m:e...£mtob ":02 296 C onfirmatory Factor Analysis — Invariance Across Gender Black African, Coloured and Indian: Descriptive Statisticflor Justice Items Black African Indian ‘White Mean SD Mean SD Mean SD Vl 3.82 1.09 3.72 1.13 3.18 1.30 V2 3.67 1.25 3.51 1.30 2.73 1.36 V3 4.01 1.09 3.87 1.06 3.66 1.08 V4 4.07 1.07 3.75 1.19 3.41 1.19 V5 3.67 1.13 3.67 1.20 3.40 1.09 V6 3.92 0.98 4.00 1.00 3.71 1.15 V7 3.94 0.91 3.91 1.04 3.79 1.12 V8 3.74 0.99 3.81 0.98 3.54 1.11 V10 3.58 1.22 3.62 1.15 4.01 1.21 V11 3.50 1.26 3.51 1.21 3.71 1.37 V12 3.83 1.11 3.67 1.12 3.86 1.17 V13 3.27 1.11 3.41 1.32 3.81 1.28 V23 3.69 1.00 3.89 1.06 3.38 1.16 V24 3.61 1.15 3.78 1.16 3.18 1.28 V25 3.93 0.94 3.92 0.98 3.75 1.02 V26 3.95 0.96 4.01 1.01 3.61 1.10 V27 3.82 1.13 3.76 1.03 3.51 1.08 V28 3.97 1.00 4.06 0.93 3.73 1.12 V29 4.06 0.93 4.10 0.93 3.81 1.10 V30 3.90 0.99 3.95 1.02 3.60 1.05 V31 3.57 1.17 3.66 1.28 3.58 1.22 V32 3.64 1.19 3.58 1.32 3.34 1.32 V33 3.87 0.97 3.85 1.09 3.37 1.21 V34 3.95 0.94 3.75 1.22 3.40 1.29 V35 3.95 1.06 3.53 1.30 3.26 1.31 V36 3.96 1.00 3.71 1.33 3.31 1.36 V37 4.12 0.96 3.70 1.31 3.46 1.26 V38 3.84 0.98 3.75 1.27 3.30 1.18 Note. Black African N = [02, Indian N = 79, White N = 99. 297 000 _00 E0 000- m _ 0- m _ 0- no.0- m00- :0 >00 30 >00 >00 00.0. 000- N00 00.— m _ 0 0N0 mm0 m _ .0 N. _ 0 30 500 000 000 _N0 0N0 500 20 000 M00 0N0 0V0 00.— 000 _00 m _ 0 000 000 000 N00- 000 ~00 000 v00 M00- M00 #00. 000 v00 mv0 0V0 00.— m00 000 v00 «00. N00 N00 000 .80 N00- _00 :0 000- 30 no.0 M00- 000 0V0 mm .0 m m .0 00.— 3.0 2.20 0n.0 wm0 0N0 000 E0 20 «m0 2 .0 NNO _N0 20 _N0 30 3.0 30 0.0 30 ”—0 00.— nm0 .940 mm0 9.0 _m0 :0 NNO 30 vm0 .ov0 ~N0 3.0 30 90 0N0 N. _ 0 v00- 0 _ 0 N00 000- Nm0 00. _ n _ 0 vm0 30 mm0 500 M00- 30 M00 m _ .0 30 30 30 m :0 20 30 30 £0 5N0 S0 .00 0N0 v00 00.— _m0 wm0 R0 30 _N0 5N0 cm0 ”—0 :0 30 £0 30 NNO 30 ”N0 h~0 000 0:0 30 N00 0n.0 30 R0 00.— v _ .0 o _ 0 mm0 0N0 0m .0 30 $0 900 0N0 0N0 Nm0 20 ”V0 0n.0 0V0 :0 500. 000 M00- 000- wm0 £0 £0 :0 00.— 000 mm0 0m .0 w _ 0 m m .0 0m0 0V0 3H0 0N0 Q0 30 0n.0 3.0 30 0V0 0n.0 000- 000 80 000- mm0 w _ 0 30 0V0 310 $0 00.— m _ 0 mm0 o _ 0 _ _0 3.0 ~m0 0m0 90 .N0 20 R0 30 ov0 3V0 3x0 _N0 80. 500 _00- ~00- 0N0 _N0 mN0 mm0 N50 $0 :0 00.— wm> nm> om> mm> Vm> mm> mm> _m> om> 0N> mm> h~> 0N> mN> vm> mN> m_> N.> :> 0_> w> n> o> m> v> m> N> ~> m_> N.> :> 0_> w> N.> c> m> v> m> N> _> 3:6: 835. RQESEEEQU "canine. mow—m— 298 wm0 :0 0.00 3.0 $0 wm0 0n.0 0N0 0m0 3:0 .20 3.0 wm> :m0 000 3.0 20 mm0 3.0 3.0 9:30 mm0 3H0 wm0 wm0 nm> 9.0 :m0 :30 BNO 30 9x0 000 500 0n.0 0m0 NNO mm0 0m> 9.0 0m0 Nm0 mm0 2‘0 5‘0 30 0V0 0n.0 3.0 :m0 NWO mm> 00. _ :30 30 3:0 0N0 0V0 vm0 0n.0 :0 3.0 0W0 wm0 um> 00.: 00.0 000 30 $0 3 .0 3:0 0m0 mm0 0v0 0N0 mm> 00. _ 0n.0 ©m0 :m0 vm0 0m0 $0 200 00.0 mm0 Nm> 00.: Nm0 0N0 «m0 0N0 mm0 0V0 3.0 000 :> 00. : 30 Nm0 :m0 :0 mm0 0m0 mm0 0m> 00.: :0 000 .. 3.0 mm0 R0 30 0N> 00.: N00 0V0 mm0 0N0 mm0 0N> 00. : $0 on .0 0V0 0N0 n~> 00. : v00 :0 wm0 0N> 00.: :00 2.0 mm> 00.: 0V0 vm> 00.: mm> M:> N:> ::> 0:> w> n> 0> m> v> m> - N> :> .3.5 mm> Nm> :m> 0m> om> wm> n 0N> mm> 032.586 «.53: 3.5.3. SNESEBEQU Eafib< saw:— 299 l a... \.~\_¢.\~\. .\\,\\..€~\\\.\\\v\~vq\.\\\. V .. 239...: < in 3.: _--——-—'— .HC_ 1. 2. 3:32.: 00. _ wm> 0w0 0m0 5.0 00.: $0 :00 00.: N00 00.: nm> 0m> mm> m> nm> om.) mm> am> mm> mm> _m> 0m> 0N> 0N> ::> 0N> mm> vm> mm> M:> N:> ::> 0:> w> N.> 0> m> v> m> m> _> 06:22:96 “.52: 3.5.35 Exampzfimtsb "=a9€:< sou—m 300 If .’.~M\~w\\ NV..V:\\»’I\~\I QK‘N .«N :> [I‘ll HEQW3=~ 0m0- w _ .0- 0m0- 30- mm0 3.0 0.0 0n.0 mm0 3.0 v.00 mm0 mm.) 3.0. 0.00. 30. wm0- 0V0 wm0 0m0 3.0 3.0 mm0 mm0 3.0 nm> 3x0- mm0- 3.0- 0:..0- 0:..0 $0 :0 0N0 «m0 0m0 9.0 om0 0m> wm0- :m0- :0- nm0- .90 a0 0N0 3.0 3.0 mm0 20 30 mm> $0- «m0- nv0- mm0- mm0 om0 mm0 w _ .0 0m0 mm0 3.0 mm0 wm> NNO- :00- «N0- 0_0- 0N0 NNO :0 0:0 20 mm0 wm0 0:0 mm> Q0- mm0- :v0- on0- 3.0 wm0 mm0 _ :0 3.0 0V0 :0 30 mm> mv0- 0:0- 3.0- mm0- :0 0:0 20 2.0 mm0 0W0 3.0 :0 :m> 0n.0- 000 3:0- 0N0- 0V0 ov0 v00 0V0 hi0 3.0 $0 mm0 0m> 0N0- wo0- NNO- NNO- mm0 mm0 NNO 30 ”N0 0N0 0N0 : :0 om> 0N0- w00- cm.0- wm0- 3.0 . 0N0 mm0 om0 mm0 - ~60 :0 3:0 wm> 320- 000- 3.? nm0- wm0 $0 3.0 $0 $0 0m0 ”v.0 mm0 nm> :v0- 0:.0- 3:0- 0N0- Nm0 :m0 :m0 :m0 000 Nm0 mm0 :m0 0N> mv.0- no.0- 5N0- NNO- 0N0 0N0 9.0 :m0 mm0 :m0 :m0 mm0 m~> 0V0- » : .0- :m0- 0 : .0- 0: .0 ~v0 5.0 0N0 NWO :0 v0.0 0m0 vm> 0m0- mm0- 0m0- h : .0- 0N0 000 :v0 :00 090 R0 30 $0 mm> 00. : Nm0 00.0 0n.0 VNO- :m0- ”v.0- mv0- Q0- nm0- $0- wm0- m : > 00. : 3.0 :m0 no.0 v : .0- m : .0- 30- 0N0- : :0- 0N0- m00- N : > 00.: v00 NVO- mm0- 0v0- 0m0- ov0- vv0- 0m .0- cm.0- : :> 00. : wm0- Vm0- Nm0- 0n.0- 0v0- :v0- mm0- 0m0- 0 : > 00.: 9.0 0m0 0m0 0N0 3:0 :00 mm0 w> 00.: 30 Q0 3.0 0V0 3.0 5N0 N.> 00. _ 0V0 3.0 30 :m0 #0 0> 00. : 30 :m0 :0 vm0 m> 00.: :0 2.0 500 v> 00.: $0 $0 m> 00.: v00 N> 00.: :> ::> m_> ::> 0:> ::> \.> 0> m> v> m> N> :> .25: 35:0 RQESEELSD "5:05 301 r> I- . I! I II . . 14- [If V».ds-\.~\:c. v .a ~s\.V\\ b. Ii) s\§.~\~,v\.‘v.\.\§. V h:lfl-I.U:~ E .- 7N> mm> 3.0 3.0 9.0 :0 $0 5.0 7.0 3.0 3.0 wm0 5N0 N. _ .0 wm> 0m0 3.0 3.0 0m0 3.0 00.0 9.0 3.0 :0 R0 0N0 _m.0 nm> N00 wm0 «m0 :0 0.00 N00 100 $0 500 3.0 .mv0 3.0 0m> 0.00 $0 :0 mm0 0m0 0V0 ~00 ~00 mm0 0N0 0N0 vm0 mm> 00._ $0 ~50 $0 3.0 0N0 CHO mm0 3.0 00.0 3.0 0V0 VT» 00. _ 00.0 _00 00.0 NNO wm0 _m.0 wm0 00.0 0m0 _m.0 mm> 00. _ 30 N10 50 _m0 0N0 «m0 000 mm0 0V0 Nm> 00._ wm0 :0 0n.0 0N0 0N0 5.0 30 3.0 _m> 00. _ V0.0 m00 N50 30 00.0 0m0 wm0 0m> 00.. 000 N00 mm0 R0 nm0 0n.0. ._ 0N> 00. _ N00 M00 E0 000 3.0 0N> 00. _ Vm .0 0V0 09.0 3.0 NN> 00._ 00.0 000 E0 0N> 00.0 05.0 00.0 0N> 00,— 5.0 0N> 00._ 0N> m_> N_> :> 0_> w> >> 0> m> v> m> N> _> um> mm> ~m> _m> 0m> 0N> w hm> 0N> 0N> vm> mN> fimztzzg was: 82$. SNESEEEQD ":30:— 302 05 n >3. .332. 00._ 05.0 #50 N50 00._ 05.0 05.0 00._ 05.0 00._ xm> 5m> 0m> mm> vm> mm> mm> _m> 0m> 0N> 0N> 5N> 0N> 0N> 0N> mm> m_> N_> :> 0.> w> 5> 0> m> v> m> N> _> wm> 5m> 0m> mm> 0332.058 2:»: 8.530 3530chth "5:0:— 303 $0- 0_.0- $.0- ...m.0- 5.0 000 .m0 N10 0.0 3.0 0m0 :0 wm> 500- ..m0- 0.0- mm0- ~00 wm0 wm0 0T0 mm0 0m0 3.0 0m0 5m> E0- 0520- mm0- «m0- mw0 5.0.0 3.0 .20 2.0 3.0 00.0 2.0 0m> 000- .m0- 0.0- 0N0- 3.0 5m0 00.0 0W0 0V0 000 0m0 3:0 mm> 5V0- 0N0- :0- mv0- mm0 v.00 00.0 N00 50.0 .m0 N00 000 vm> 0v0- 0 _ .0- 5N0- mm0- 0N0 0m0 0V0 0m0 3.0 50.0 .00 000 mm> 0m0- 5 . .0- 0N0- 0m0- Nm0 Nm0 mm0 3.0 3.0 0v0 00.0 000 mm> 0m0- m . .0- 0N0- 0N0- 0N0 mm0 0n.0 3:0 0n.0 3.0 3.0 wm0 .m> mv.0- .m0- mm0- mm0- 3.0 .00 .30 00.0 0.0 0n.0 mm0 00.0 0m> .m0- mm0- mm0- Nv0- 00.0 00.0 00.0 wm0 . 00¢ .30 30 N00 0N>. 00.0- 0n.0- 3.0- 00.0- 3.0 00.0 .00 .00 5m0 0m0 .00 Nm0 0N> 0m0- mm0- mv.0- mm0- wm0 0N0 5m0 00.0 0n.0 0n.0 5m0 5N0 5N> mm0- v.0- mm0- 0N0- mm0 .v0 000 0n.0 N00 000 0n.0 09.0 0N> wm0- 3.0- 0N0- 0m0- Nm0 mm0 0v0 0N0 00.0 00.0 00.0 $0 0N> mm0- Nm0- 2.0- 00.0- 5.0 5.0 00.0 Nm0 mm0 0n.0 N00 mm0 v~> 500- 0m0- wm0- 0V0- 0m0 5.0 .00 0m0 5.0.0 500 000 .m0 MN> 00.. M50 55.0 «(0.0 00.0- 5m0- ”v.0- mm0- 5V0- mm0- mm0- 0m0- m.> 00. . 00.0 .5.0 mm0- vm0- vM.0- 3H0- 5m0- w . .0- 0v0- mm0- N . > 00. . 00.0 00.0- mm0- .00- 0v0- 5V0- wm0- N00- 30- ..> 00. . 000- 9-20. wm0- 0m0- 3.0- Nm0- 00.0- 30- 0. > 00.. 00.0 .50 00.0 ~00 0m0 .00 3x0 ..> 00.. 55.0 000 00.0 000 .m0 000 5> 00.. 500 N00 .00 .00 00.0 0> 00. . ”v.0 0m0 .v0 0N0 m> 00.. .00 .50 00.0 v> 00.. 00.0 N00 m> 00.. 05.0 m> 00.. .> m.> N.> ..> 0.> ..> 5> 0> m> ..> m> N> .> .28.. 8.535. 359.288.2000 "3.55 304 00.0 000 00.0 00.0 _0.0 00.0 50.0 00.0 00.0 500 ~00 00.0 00> 05.0 000 00.0 00.0 00.0 00.0 00.0 00.0 00.0 00.0 ~00 .00 50> 05.0 _0.0 .00 00.0 00.0 ~00 00.0 00.0 00.0 500 00.0 00.0 00. 00.0 ~00 ~00 ~00 .00 00.0 00.0 00.0 00.0 ~00 00.0 00.0 00> 00. . 05.0 00.0 00.0 00.0 .00 .00 00.0 00.0 00.0 00.0 00.0 00> 00. . ~00 ~50 50.0 .00 00.0 ~00 00.0 05.0 00.0 00.0 00> 00.. .00 00.0 ~00 000 00.0 00.0 00.0 00.0 00.0 ~m> 00. . 00.0 00.0 00.0 00.0 000 00.0 00.0 00.0 _m> 00. . 55.0 05.0 .00 00.0 00.0 00.0 00.0 00> 00. . 05.0 00.0 00.0 00.0 ~00 00.0 0~> - 00._ .000 50.0 00.0 .00 00.0 0~> 00. . 00.0 .00 50.0 00.0 5~> 00. . 00.0 00.0 00.0 0~> 00. . 05.0 00.0 0~> 00.. 55.0 0~> 00.. 0~> 0.> ~.> ..> 0.> ..> 5> 0> 0> 0> m> ~> .> 00> 00> 0> .m> 00> 0~> 0~> 5~> 0~> 0~> 0~> 0~> 083.28 .423... 83.5. 35.420.08.889 "32>? 305 .2. u .< 2:2. 00. _ 000 «#0 0n.0 00._ ~00 ~50 00._ E0 00._ wm> nm> cm> mm> Vm> mm> ~m> _m> 0m> 0~> 0~> h~> 0~> m~> v~> m~> m_> ~_> __> 0.> m> N.> o> m> v> m> ~> _> wm> >m> ©m> mm> 025.228 VIE»: 85:0 SNEEEREQU "83>? 306 Black African, Coloured and Indian: Descriptive Statisticsfor Race C entraligy Items Black African Indian White Mean SD Mean SD Mean SD V113 2.30 1.33 2.57 1.28 2.33 1.13 V114 3.86 1.21 3.51 1.19 2.76 1.09 V115 4.11 1.02 3.73 1.07 2.97 1.05 V116 3.95 1.13 3.72 1.13 2.76 1.10 Note. Black African N = 102, Indian N = 79, White N = 99. Black African: Correlations for Race Centralitz Items V113 V114 V115 V116 V113 1.00 V114 0.24 1.00 V115 -0.01 0.64 1.00 V116 0.06 0.70 0.81 1.00 Note. N= 102 Indian: Correlations for Race C entrality Items V113 V114 V115 V116 V113 1.00 V114 0.23 1.00 V115 0.32 0.54 1.00 V116 0.43 0.52 0.72 1.00 Note. N = 79. White: Correlations for Race Centrality Items V113 V114 V115 V116 V113 1.00 V114 0.40 1.00 V115 0.04 0.26 1.00 V116 0.36 0.57 0.52 1.00 Note. N = 99. 307 Black African, Coloured and Indian: Descriptive Statistics jar A fiican Centrality Items Black African Indian White Mean SD Mean SD Mean SD V117 2.93 1.34 2.99 1.14 3.42 1.20 V118 3.97 1.09 3.48 1.07 3.66 1.03 V119 4.17 0.90 3.51 1.06 3.51 1.13 V120 4.06 1.06 3.48 1.08 3.63 1.08 Note. Black African N = 102, Indian N = 79, White N = 99. Black African: Correlations for African C entrality Items V117 V118 V119 V120 V117 1.00 V118 0.16 1.00 V119 0.09 0.68 1.00 V120 0.07 0.75 0.69 1.00 Note. N: 102 Indian: Correlations for A fiican Centrality Items V117 V118 V119 V120 V117 1.00 V118 0.29 1.00 V119 0.11 0.75 1.00 V120 0.11 0.87 0.84 1.00 Note. N = 79. White: Correlations for African C entralij/ Items V117 V118 V119 V120 V117 1.00 V118 0.53 1.00 V119 0.38 0.73 1.00 V120 0.43 0.80 0.81 1.00 Note. N = 99. 308 Black African, Coloured and Indian: Descriptive Statistics for A flirmative Action Items Black African Indian ‘White Item Mean SD Mean SD Mean SD V92 3.74 1.04 2.97 1.13 2.49 1.16 V93 3.71 1.09 2.91 1.16 2.83 1.28 V94 3.96 0.96 3.52 1.02 3.52 1.14 V95 3.46 1.09 2.63 1.13 2.26 1.05 V96 3.66 0.97 2.95 1.17 2.49 1.00 Note. Black African N: 102, Indian N = 79, White N = 99. Black African: Correlations for A ffirmative Action Items V92 V93 V94 V95 V96 V92 1.00 V93 0.40 1.00 V94 0.66 0.28 1.00 V95 0.53 0.64 0.42 1.00 V96 0.71 0.52 0.54 0.52 1.00 Note. N = 102. Indian: Correlations for A flirmative Action Items V92 V93 V94 V95 V96 V92 1.00 V93 0.55 1.00 V94 0.45 0.33 1.00 V95 0.35 0.37 0.13 1.00 V96 0.61 0.37 0.30 0.54 1.00 Note. N = 79. White: Correlations for A fi‘irmative Action Items V92 V93 V94 V95 V96 V92 1.00 V93 0.67 1.00 V94 0.60 0.49 1.00 V95 0.69 0.63 0.44 1.00 V96 0.71 0.59 0.52 0.69 1.00 Note. N = 99. 309 Black African, Coloured and Indian: Descriptive Statistics for Merit Items Black African Indian White Mean SD Mean SD Mean SD V88 4.40 0.81 4.52 0.68 4.43 0.73 V89 4.40 0.94 4.22 1.03 3.97 1.13 V90 4.14 0.96 4.34 0.73 4.17 0.97 V91 4.51 0.73 4.57 0.57 4.43 0.67 Note. Black African N = 102, Indian N = 79, White N = 99. Black African: Correlations for Merit Items V88 V89 V90 V91 V88 1.00 V89 0.15 1.00 V90 0.17 0.06 1.00 V91 0.29 0.19 0.28 1.00 Note. N: 102. Indian: Correlations for Merit Items V88 V89 V90 V91 V88 1.00 V89 0.04 1.00 V90 0.23 0.36 1.00 V91 0.32 0.12 0.48 1.00 Note. N = 79. White: Correlations for Merit Items V88 V89 V90 V91 V88 1.00 V89 0.13 1.00 V90 0.34 0.27 1.00 V91 0.36 0.06 0.48 1.00 Note. N = 99. 310 r"- Black African, Coloured and Indian: Descriptive Statistics for Social Guidance Items Black African Indian White Mean SD Mean SD Mean SD V49 3.36 1.15 3.44 1.02 3.65 0.93 V50 3.08 1.12 3.08 0.94 3.39 0.91 V53 3.13 1.11 3.25 1.08 3.56 1.10 V54 2.84 1.12 3.27 1.00 3.35 1.07 V57 3.17 1.05 3.28 0.91 3.33 0.89 V58 2.88 1.02 3.23 0.90 3.20 0.82 Note. Black African N = 89, Indian N = 71, White N = 95. Black African: Correlations for Social Guidance Items V49 V50 V53 V54 V57 V58 V49 1.00 V50 0.70 1.00 V53 0.42 0.42 1.00 V54 0.44 0.56 0.82 1.00 V57 0.49 0.41 0.66 0.55 1.00 V58 0.38 0.51 0.55 0.64 0.66 1.00 Note. N = 89. ’ lndian: Correlations for Social Guidance Items V49 V50 V53 V54 V57 V58 V49 1.00 V50 0.76 1.00 V53 0.56 0.43 1.00 V54 0.58 0.46 0.61 1.00 V57 0.60 0.62 0.45 0.48 1.00 V58 0.40 0.49 0.29 0.33 0.81 1.00 Note. N= 71. 311 White: Correlations for Social Guidance Items V49 V50 V53 V54 V57 V58 V49 1.00 V50 0.69 1.00 V53 0.47 0.40 1.00 V54 0.49 0.60 0.76 1.00 V57 0.19 0.26 0.08 0.26 1.00 V58 0.26 0.38 0.22 0.33 0.74 1.00 Note. N = 95. Black African, Coloured and Indian: Descriptive Statistics for Group A flirmative Action and Group Restorative Justice Items Black African Indian White Mean SD Mean SD Mean SD V97 4.18 0.80 2.77 1.13 1.81 0.84 V98 3.90 1.01 2.96 1.09 2.21 1.05 V99 4.13 0.94 3.03 1.05 2.75 1.15 V100 3.97 0.94 2.65 1.08 2.02 0.86 V101 3.95 0.92 2.81 1.11 2.18 0.89 V102 3.43 1.36 2.87 1.29 1.99 1.34 V103 3.63 1.27 3.16 1.31 2.28 1.13 V104 3.56 1.19 3.09 1.23 2.24 1.12 Note. Black African N = 102, Indian N = 79, White N = 99. Black African: Correlations for Group A flirmative Action and Group Restorative Justice Items V97 V98 V99 V100 V101 V102 V103 V104 V97 1.00 V98 0.46 1.00 V99 0.72 0.53 1.00 V100 0.38 0.68 0.45 1.00 V101 0.55 0.50 0.57 0.49 1.00 V102 -0. 12 0.01 -0.02 —0.01 -0.05 1.00 V103 0.11 0.09 0.20 0.10 0.15 0.49 1.00 V104 -0.15 -0.01 -0.11 -0.04 -0.05 0.63 0.44 1.00 Note. N = 102. 312 Indian: Correlations for Group A flirmative Action and Group Restorative Justice Items V97 V98 V99 V100 V101 V102 V103 V104 V97 1.00 V98 0.25 1.00 V99 0.62 0.25 1.00 V100 0.32 0.43 0.46 1.00 V101 0.47 0.14 0.54 0.57 1.00 V102 0.13 -0.08 0.13 0.00 0.11 1.00 V 103 0.06 -0.16 -0.01 -0.21 -0.01 0.69 1.00 V104 -0.06 -0.08 0.00 -0.16 -0.07 0.66 0.48 1.00 Note. N = 79. White: Correlations for Group Affirmative Action and Group Restorative Justice Items V97 V98 V99 V100 V101 V102 V103 V 104 V97 1.00 V98 0.37 1.00 V99 0.33 0.32 1.00 V100 0.52 0.48 0.54 1.00 V101 0.51 0.43 0.61 0.68 1.00 V102 0.23 0.13 0.21 0.30 0.27 1.00 V103 0.23 0.17 0.18 0.31 0.18 0.59 1.00 V104 0.15 0.17 0.10 0.25 0.11 0.24 0.32 1.00 Note. N = 99. Black African, Coloured and lndian: Descriptive Statistics for Group Restorative Justice Items Black African Indian White Mean SD Mean SD Mean SD V105 3.81 0.96 3.49 0.83 3.76 1.17 V106 4.08 0.84 3.44 0.90 3.71 0.88 V107 3.70 0.93 3.53 0.96 4.07 0.97 V108 3.79 1.06 3.56 1.11 3.94 1.14 Note. Black African N = 102, Indian N = 79, White N = 99. 313 Black African: Correlations for Group Restorative Justice Items V105 V106 V107 V108 V105 1.00 V106 0.69 1.00 V107 0.47 0.45 1.00 V108 0.60 0.52 0.68 1.00 Note. N= 102. Indian: Correlations for Group Restorative Justice Items V105 V106 V107 V108 V105 1.00 V106 0.34 1.00 V107 0.33 0.14 1.00 V108 0.30 -0.02 0.49 1.00 Note. N = 79. White: Correlations for Group Restorative Justice Items V105 V106 V107 V108 V105 1.00 V106 0.56 1.00 V107 0.58 0.32 1.00 V108 0.50 0.33 0.84 1.00 Note. N = 99. 314 Black African, Coloured and Indian: Descriptive Statistics for Intentions Items Black African lndian White Mean SD Mean SD Mean SD V59 1.50 0.84 1.43 0.83 1.45 0.72 V60 1.56 0.92 1.61 0.95 1.76 1.02 V61 2.29 1.05 2.38 1.03 2.53 1.01 V62 4.52 0.81 4.38 0.91 4.23 0.82 V63 4.39 0.94 4.29 0.95 4.29 0.75 V64 1.83 1.10 1.76 1.04 1.68 0.95 V65 3.54 1.24 3.57 1.15 3.39 1.19 V66 3.71 1.11 3.46 1.28 3.08 1.32 V67 3.50 1.20 3.39 1.40 2.78 1.31 V68 2.66 1.31 2.57 1.11 2.28 1.14 V69 3.08 1.37 3.46 1.23 3.46 1.30 V70 3.17 1.37 3.57 1.17 3.70 1.26 V71 2.21 1.12 2.51 1.05 2.53 1.26 Note. Black African N= 102, Indian N = 79, White N = 99. 315 V-q‘m-I-w .~0_ HZ .262 $0 80 0~0 ~00 _~0 w~0 20 80- M00 w00- _~0 3.0 :> 00._ R0 20 _~0 ~_0 20 I0 000 30 000 m00 n~0 0~> 00. _ o~0 m _ 0 3 .0 h _ 0 ~00 v00- v0.0 000 2 0 -0 00> 00._ mm0 $0 000 -0 :0 20 20 v~0 w~0 00> 00; 3.0 000 $0 20 000 :0 000 _~0 50> 00._ >~0 S0- ~_0 20 000 000 20 00> 00._ S0 m~0 ~m0 30. 000 m~0 mo> 00. _ 0_ 0- V0.0. 000 ~00 500 vc> 00.0 000 500- _ #0- m00 m©> 00._ 30- 000- #00 ~©> 00. 0 ~_ 0- 000- $> 00._ ~m0 00> 00._ 0m> 0~> 00> 00> $> 00> mo> vo> m©> ~o> $> 00> 0m> was: 25.2225 SNEEENESD 2.3;: sou—m 316 05 n 2 .202 0V0 3.0 0~0 0~ .0 m~0 0~0 00.0- 2.0 m00- 50.0 00.0 50 _5> 00. H 05.0 00.0- 50.0 ~ ~ .0 00.0 «.00 3 .0 - 0- m 0 .0 ~00 50.0 05> 00.0 0~ .0- v0.0 000 00.0 0— .0 m g .0 00.0 00.0 ~00 50.0 00> 00; 00 00.0 0m0 00.0.. 2 .0 000 :0- 00.0- 00.0 00> 00.0 ~00 20 ~00- 20 00.0 m00- 50.0 m00 50> 00. fl 0~0 00.0 0~0 2 .0 00.0 00.0 ~00 00> 00._ -0- 3.0 5m0 :0- 500- 000- 00> 00.0 0~0- ~m0- $0 000 30 v0> 00. _ ~50 m _ .0- ~00 000- m0> 00.0 0~0- ~ _ 0- 00 .0- ~0> 00.0 ~00 ~00- ~0> 00.0 00 .0 00> 00.0 00> 05> 00> 00> 50> 00> 00> 00> m0> ~0> ~0> 00> 00> as»: 28.2225 acsxmzctthob ":30:— 317 00 u 2 .202 00.0 50.0 0n.0 5 _ .0 0:0 0:0 3 .0 ~00 50.0 0~0- 50.0 5:0 _5> 00.: m50 5~0 v~0 0:0 000 00.0 -0 2.0 3.0- :0- :0 05> 00.: 0:0 50 00.0- 0:0 00.0- 50.0 00.0 :0- 000- 0:0 00> 00._ M50 000 -0 50.0.. 5:0 0:0 0~0- w _ 0.. m00 00> 00._ 050 S0 v00- :0 :0 v~0- 5 _ .0- ~00- 50> 00.: w _ .0 E .0- :0 w : .0 20- m~.0- 00.0 00> 00._ -.0- $0 0V0 ~:.0- 5:0- 000 00> 00. _ 000- : _ .0- 00.0 00.0 0 : .0 00> 00.: 55.0 -0- :m0- :m0- m0> 00.: :0- m~.0- w _ .0- ~0> 00._ _m0 000 ~0> 00.: mm0 00> 00.: 0m> 05> 00> 00> 50> 00> 00> 00> m0> ~0> _0> 00> 0m> .22me hfiQCtNNQN LQ\%=QNNU~kaQ-V "as? 318 REFERENCES 319 Aberson, C. L. (2007). Diversity, merit, fairness, and discrimination beliefs as predictors of support for affirmative-action policy actions. Journal of Applied Social Psychology, 37(10), 2451—2474. Aberson, C. L. (2003). Support for race-based affirmative action: Self-interest and procedural justice. Journal of Applied Social Psychology, 33(6), 1212-1225. Adams, J. S. (1965). Inequity in social exchange. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 2, pp. 267-299). New York, NY: Academic Press. Affirmative action to stay, says business body. (2007, September 11). Mail and Guardian. Retrieved September 11, 2007, from http://www.mg.co.za/articlepage.aspx?area=/breaking_news/breaking__news__busi ness/&articleid=3 1 8905 Aj zen, l. (1985). From intentions to actions: A theory of planned behavior. In J. Kuhl & J. Beckman (Eds.), Action-control: From cognition to behavior (pp. 11-39). Heidelberg, Germany: Springer. Allport, G. (1954). The nature of prejudice. New York, NY: Doubleday Anchor Books. Ambrose, M. L., & Arnaud, A. (2005). Are procedural justice and distributive justice conceptually distinct? In J. Greenberg and J. A. Colquitt (Eds), Handbook of organizational justice. Mahwah, NJ: Lawrence Erlbaum Associates, Publishers. Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103(3). 411-423. Anderson, N. (2004). Editorial — The dark side of the moon: Applicant perspectives, negative psychological effects (NPEs), and candidate decision making in selection. International Journal of Selection and Assessment, 12, 1-8. Anderson, N., & Schalk, R. (1998). The psychological contract in retrospect and. prospect. Journal of Organizational Behavior, 19, 637-647. Aquino, K., Lewis, M. U., & Bradfield, M. (1999). Justice constructs, negative affectivity, and employee deviance: A proposed model and empirical test. Journal of Organizational Behavior, 20(7), 1073-1091. Arvey, R. D., & Sackett, P. R. (1993). Fairness in selection: Current developments and perspectives. In Schmitt, N., & Borman, W. (Eds). Personnel Selection, 171—202. San Francisco: Jossey-Bass. 320 Aselage, J ., & Eisenberger, R. (2003). Perceived organizational support and psychological contracts: A theoretical integration. Journal of Organizational Behavior, 24(5), 491-509. Asmal, K., & James, W. (2001). Education and democracy in South Africa today. Daedalus, Winter, 185-204. Bailey, T. (2003). Skills migration. HRD Review, 2003, 234-255. Retrieved June 5. 2007 from http://www.hsrc.ac.za - Bandura, A. (1982). Self-efficacy mechanism in human agency. American Psychologist. 37, 122-147. Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic and statistical considerations. Journal of Personality and Social Psychology, 51, 1 173-1 182. Bauer, T. N., Maertz, C. P., Jr., Dolen, M. R., & Campion, M. A. (1998). Longitudinal assessment of applicant reactions to employment testing and test outcome feedback. Journal of Applied Psychology, 83, 892-903. Baumeister, R. F ., Stillwell, A. M., & Heatherton, T. F. (1994). Guilt: An interpersonal approach. Psychological Bulletin, 1 15(2), 243-267. Bazemore, G., Elis, L., & Green, D. L. (2007). The "independent variable" in restorative justice: Theory-based standards for evaluating the impact and integrity of victim sensitive process (part II). Victims & Oflenders, 2(4), 351-373. Bell, B. 8., Ryan, A. M., & Wiechmann, D. (2004). Justice expectations and applicant perceptions. International Journal of Selection and Assessment, 12, 24-3 8. Bell, B. S., Wiechmann, D., & Ryan, A. M. (2006). Consequences of organizational justice expectations in a selection system. Journal of Applied Psychology, 91, 455-466. Bell, M. P., Harrison, D. A., & McLaughlin, M. E. (2000). Forming, changing, and acting on attitude toward affirmative action programs in employment: A theory-driven approach. Journal of Applied Psychology, 85(5), 784-798. Between staying and going. (2008, September 25). The Economist. Retrieved July 27. 2009, from https://www.economist.com/displayStory.cfm?story_id=12295535 Bies, R. J., & Moag, J. F. (1986). Interactional justice: Communication criteria of fairness. In R J. Lewicki, B. H. Sheppard, & M. H. Bazerman (Eds.), Research on negotiations in organizations (Vol. 1, pp. 43-55). Greenwich, CT: JAl Press. 321 Blau, P. (1964). Exchange and power in social life. New York, NY: Wiley. Blau, P. (1977). Inequality and heterogeneity. New York, NY: The Free Press. Bliese, P. D. (2000). Within-group agreement, non-independence, and reliability: Implications for data aggregation and analysis. In K. J. Klein & S. W. J. Kozlowski (Eds.), Multilevel theory, research, and methods in organizations. San Francisco, CA: Jossey-Bass. Bobocel, D. R., & Holmvall, C. (2001). Are interactional justice and procedural justice different? Framing the debate. In S. Gilliland, D. Steiner, & D. Skarlicki (Eds). Research in social issues in management: Theoretical and cultural perspectives on organizational justice (pp. 85-110). Greenwich, CT: Information Age. Boeckmann, R. J., & Feather, N. T. (2007). Gender, discrimination beliefs, group-based guilt, and responses to affirmative action for Australian women. Psychology of Women Quarterly, 31, 290-304. Booysen, L. (2001). The duality in South African leadership: Afrocentric or eurocentric”. South A fiican Journal of Labour Relations. 25 (3&4), 36-64. Brodish, A. B., Brazy, P. C., & Devine, P. G. (2008). More eyes on the prize: Variability in white Americans' perceptions of progress toward racial equality. Personality and Social Psychology Bulletin, 34, 513-527. Brockner, J. & Wiesenfeld, B. M. (1996). An integrative framework for explaining reactions to decisions: Interactive effects of outcomes and procedures. Psychological Bulletin, 120, 189-208. Brocker, J. & Wiesenfeld, B. (2005). How, when, and why does outcome favorability interact with procedural fairness? In J. Greenberg and J. A. Colquitt (Eds), Handbook of organizational justice. Mahwah, NJ: Lawrence Erlbaum Associates. Publishers. Burgess, S. M. (2002). SA. tribes: Who we are, how we live, what we want/ram life. Claremont, South Africa: David Philip Publishers. . CC MA (2001). Preparation for Hearings Part 1. Retrieved June 6, 2007, from http://www.ccma.org.za CCMA (2006). Annual Report 2005/2006. Retrieved June 5, 2007, from http://www.ccma.org.za Campbell, D. T. (1958). Common fate, similarity, and other indices of the status of aggregates of persons as social entities. Behavioral Science, 3, 14-25. 322 Cascio, W. F. (2003). Changes in workers, work, and organizations. In W. C. Borman. Ilgen, D. R., & Klimoski, R. J. (Ed.), Handbook of psychology: Industrial and organizational psychology (Vol. 12, pp. 401-422). Hoboken, NJ: John Wiley & Sons, Inc. Chaiken, S. (1987). The heuristic model of persuasion. In M. P. Zanna, J. M. Olson, & C. P. Herman. (Eds.), Social influence: The Ontario Symposium on Personalin and Social Psychology (Vol. 5, pp. 3-39). Hillsdale, NJ. : Lawrence Erlbaum. Chan, D., & Schmitt, N. (2004). An agenda for future research on applicant reactions to selection procedures: A construct-oriented approach. International Journal of Selection and Assessment, 12, 9-23. Chao, G. T., & Moon, H. (2005). The cultural mosaic: A metatheory for understanding the complexity of culture. Journal of Applied Psychology, 90(6), 1 128-1 14. Christopher, A. J. (2002). “To define the indefinable’: Population classification and the census in South Africa. Area, 34.4, 401-408. Cohen, J ., Cohen, R, Aiken, S. G., & West, L. S. (2003). Applied multiple regression/correlation analysis for the behavioral sciences. Third edition. Mahwah, NJ: Lawrence Erlbaum Associates, Publishers. Colquitt, J. A. (2001). On the dimensionality of organizational justice: A construct validation of a measure. Journal of Applied Psychology, 86, 386-40. Colquitt, J. A. (2005). What is organizational justice? A historical overview. In J. Greenberg and J. A. Colquitt (Eds), Handbook of organizational justice. Mahwah. NJ: Lawrence Erlbaum Associates, Publishers. Colquitt, J. A., Conlon, D. E., Wesson, M. J., Porter, C. O..L. H., & Ng, K. Y. (2001). Justice at the millennium: A meta-analytic review of 25 years of organizational justice research. Journal of Applied Psychology, 86, 425-445. Colquitt, J. A., Greenberg, J., & Zapata-Phelan, C. P. (2005). What is organizational justice? A historical overview. In J. Greenberg and J. A. Colquitt (Eds), Handbook of organizational justice. Mahwah, NJ: Lawrence Erlbaum Associates. Publishers. Colquitt, J. A., & Shaw, J. C. (2005). How should organizational justice be measured? In J. Greenberg and J. A. Colquitt (Eds), Handbook of organizational justice. Mahwah, NJ: Lawrence Erlbaum Associates, Publishers. Come Home Campaign (n.d.). What is the Come Home Campaign? Retrieved on September 9, 2007, from http://www.comehome.co.za/ 323 n v—Aavir- -|_- Cropanzano, R, Slaughter, J. E., & Bachiochi, P. D. (2005). Organizational justice and black applicants’ reactions to affirmative action. Journal of Applied Psychology. 90(6), 1168 -1184. Crosby, F. (2004). Aflirmative action is dead: Long live afiirmative action. New Haven, CT: Yale University Press. Crosby, F., & Gonzalez-Intal, A. M. (1984). Relative deprivation and equity theories: Felt injustice and the underserved benefits of others. In R F olger (ed.), The sense of injustice: Social psychological perspectives (pp.l41-166). New York: Plenum Press. Crosby, F., Iyer, A., & Sincharoen, S. (2006). Understanding affirmative action. Annual Review of Psychology, 5 7, 585-611. Crush, J ., Peberdy, S., & Williams, V. (2006). International migration and good governance in the South African region (Migration Policy Brief No. 17). Cape Town, South Africa: Southern African Migration Project. Dabos, G. E., & Rousseau, D. M. (2004). Mutuality and reciprocity in the psychological contracts of employees and employers. Journal of Applied Psychology. 89(1). 52- 72. Davey, L. M., Bobocel, D. R, Hing, L. S. S., & Zanna, M. P. (1999). Preference for the Merit Principle Scale: An individual difference measure of distributive justice preferences. Social Justice Research, 12(3), 223-24. De Gruchy, J. W. (2002). Reconciliation: Restoring justice. David Philip Publishers: Claremont, South Africa. De Klerk on South Africa and global leadership. (2005, November 15). The SA Times, p. 3. Department of Labour (n.d.). Employment Equity Report 2004. Retrieved June 6, 2007, from http://www.labour. gov.za/useful_docs/doc_display . j sp?id= I 0295 De Vos, A., Buyens, D., & Schalk, R. (2003). Psychological contract development during organizational socialization: Adaptation to reality and the role of reciprocity. Journal of Organizational Behavior, 24(5), 537-559. Doosje, B., Branscombe, N. R., Spears, R, & Manstead, A. S. R. (1998). Guilty by association: When one’s group has a negative history. Journal of Personality and Social Psychology, 75, 872-886. Ehrhart, M. G. (2004). Leadership and procedural justice climate as antecedents of unit- level organizational citizenship behavior. Personnel Psychology, 5 7, 61-94. 324 Erez, M., & Earley, P. C. (1987). Comparative analysis of goal setting strategies across It“: cultures. Journal of Applied Psychology, 72, 65 8-665. Evans, D. C. (2003). A comparison of the other-directed stigmatization produced by legal and illegal forms of affirmative action. Journal of Applied Psychology, 88, 121- 13. Uniform Guidelines on Employee Selection Procedures (1978). Federal Register, 43, (166), 38297. Fears over South Africa’s exodus. (2006, October 6). BBC News. Retrieved October 7. 2006, from http://news.bbc.co.uk/go/pr/fi/-/2/hi/africa/5412892.stm Feathers, N. T. (2008). Perceived legitimacy of a promotion decision in relation to deservingness, entitlement, and resentment in the context of affirmative action and performance. Journal of Applied Social Psychology, 38(5), 1230- 1 254. F innemore, M. (1999). The history of labour relations in South Africa. In M. F innemore & R. van Rensburg (eds), Contemporary Labour Relations (pp.23-48). Durban. South Africa: Butterworths. ‘ Fishbein, M., & Ajzen, I. (1975). Belief attitude, intention and behavior. Reading, MA: Addison-Wesley. Flynn, F. J ., & Brockner, J. (2003). It’s different to give than to receive: Predictors of givers’ and receivers’ reactions to favor exchanges. Journal of Applied Psychology, 88(6), 1 034-1045. Folger, R (1986). Rethinking equity theory: A referent cognitions model. In H. W. Bierhoff, R. L. Cohen, & J. Greenberg (Eds.), Research in social relations (pp. 145-162). New York, NY: Plenum Press. Folger, R., & Cropanzano, R. (1998). Equity and distributive justice as outcome fairness. In R. Folger & R. Cropanzano (Eds.), Organizational justice and human resource management (pp. 1-24). Thousand Oaks, CA: Sage, Folger, R, & Cropanzano, R. (2001). Fairness theory: Justice as accountability. In J. Greenberg and R. Cropanzano (Eds.), Advances in organizational justice. Stanford, CA: Stanford University Press. Fraser-Moleketi, G. (2007). Consolidated response from South Africa on the case evidence on brain gain. Downloaded on September 9, 2007, from http://capacity.undp.org/index.cfm?module=Library&page=Document&Documen tID=6307 325 Gaertner, S. L., & Dovidio, J. F. (2005). Categorization, recategorization, and intergroup bias. In Dovidio, J. F ., Glick, P., & Rudman, L. A. (Eds.), 0n the nature of prejudice.“ Fifty years after Allport. Malden, MA: Blackwell Publishing. Gelfand, M. J ., Raver, J. L., & Ehrhart, K. H. (2002). Methodological issues in cross- cultural organizational research. In S. G. Rogelberg (Ed.), Handbook of research methods in industrial and organizational psychology (pp. 216-246). Oxford. UK: Blackwell Publishers. Giacobbe-Miller, J. K., Miller, D. J., & Victorov, V. I. (1998). A comparison of Russian and US. pay allocation decisions, distributive justice judgments, and productivity under different payment conditions. Personnel Psychology, 51, 137-163. Gilbert, D. (2005). Interrogating Mixed-Race: A Crisis of Ambiguity? Social Identities, 11(1), 55-74. Gilliland, S. W. (1993). The perceived fairness of selection systems: An organizational justice perspective. Academy of Management Review, 18(4), 694-734. Gilliland, S. W. (1994). Effects of procedural and distributive justice on reactions to a selection system. Journal of Applied Psychology, 79, 691 -701. Gilliland, S. W., Groth, M., Baker, R. C. IV, Dew, A. F ., Polly, L. M., & Langdon. J. C. (2001). Improving applicants’ reactions to rejection letters: An application of fairness theory. Personnel Psychology, 54, 669-703. Gilliland, S. W., Benson, L., III, & Schepers, D. H. (1998). A rejection threshold in justice evaluations: Effects on judgment and decision-making. Organizational Behavior and Human Decision Processes, 76, 113-131. Gird, A., & Bagraim, J. J. (2008). The theory of planned behaviour as predictor of entrepreneurial intent amongst final-year university students. South African Journal of Psychology, 38(4), 711-724. Goldman, A. H. (1979). Justice and reverse discrimination. Princeton University Press: Princeton, New Jersey. Goldman, B. M. (2001). Toward an understanding of employment discrimination claiming: An integration of organizational justice and social information processing theories. Personnel Psychology, 54(2), 361-386. Goldman, B. M., Slaughter, J. E., Schmitt, M. J., Wiley, J. W., & Brooks, S. M. (2008). Perceptions of Discrimination: A Multiple Needs Model Perspective. Journal of Management, 34(5), 952-977. 326 Government Gazette, Republic of South Africa, Vol. 400, no. 193 7. Cape Town, 19 October 1998. Greenberg, J. (1987). Reactions to procedural injustice in payment distributions: Do the means justify the ends? Journal of Applied Psychology, 72, 55-61. Groth, M., Goldman, B. M., Gilliland, S. W., & Bies, R. J. (2002). Commitment to legal claiming: Influences of attributions, social guidance, and organizational tenure. Journal of Applied Psychology, 87, 781-788. Guest, D. E. (1998). On meaning, metaphor and the psychological contract: A response to Rousseau (1998). Journal of Organizational Behavior, 1 9, 673-677. Gupta, V. K., Turban, D. B., & Bahwe, N. M. (2008). The effect of gender stereotype activation on entrepreneurial intentions. Journal of A pplied Psychology. 93(5). 1053-1061. Guzzo, R. A., Noonan, K. A., & Elron, E. (1994). Expatriate managers and the psychological contract. Journal of Applied Psychology, 79, 617-626. Harrison, D. A., Kravitz, D. A., Mayer, D. M., Leslie, L. M., & Lev-Arey, D. (2006). Understanding attitudes toward affirmative action programs in employment: Summary and meta-analysis of 35 years of research. Journal of Applied Psychology, 91(5), 1013—1036. Hauscknecht, J., Day, D., & Thomas, S. (2004). Applicant reactions to selection procedures: An updated model an meta-analysis. Personnel Psychology, 5 7. 639- 683. Heilman, M. E., & Alcott, V. B. (2001). What I think you think of me: Women’s reactions to being viewed as beneficiaries of preferential selection. Journal of Applied Psychology, 86, 574-582. Heilman, M. E, Battle, W. S., Keller, C. E., & Lee, R. A. (1998). Journal of Applied Psychology, 83, 190-205. Heilman, M.E., Block, C.J., & Lucas, J. (1992). Presumed incompetent? Stigmatization and affirmation action efforts. Journal of Applied Psychology, 77, 536-544. Heilman, M.E., Block, C.J., & Stathatos, P. (1997). The affirmative action stigma of incompetence: Effects of performance information ambiguity. Academy of Management Journal, 40, 603-625. 327 Heilman, M.E., Rivero, J.C., & Brett, JP. (1991). Skirting the competence issue: Effects of sex-based preferential selection on task choices of women and men. Journal of Applied Psychology, 76, 99-105. ‘ Heilman, M. E., Simon, M. C., & Repper, D. P. (1987). Intentionally favored. unintentionally harmed? Impact of sex-based preferential selection on self- perceptions and self-evaluations. Journal of Applied Psychology, 72. 62—68. Herbert, T. (1994). A flirmative action in the South African workplace: Allowing black eagles to soar. Salty Print: Cape Town, South Africa. Herholdt, W., & Marx, M. (1999). Employment equity in South Africa: A guide to affirmative action success. Prentice Hall South Africa: Cape Town, South Africa. High court backs firefighters in reverse discrimination suit. (2009, June 29). CNN Politics.com. Retrieved August 6, 2009, from http://www.cnn.com/2009/POLITICS/06/29/supreme.court.discrimination/ Hofmann, D. A., Morgeson, F. P., & Gerra, S. J. (2003). Climate as a moderator of the relationship between leader-member exchange and content specific citizenship: A safety climate exemplar. Journal of Applied Psychology, 88(1), 170-178. Hofmann, D. A., & Jones, L. M. (2005). Leadership, collective personality, and performance. Journal of Applied Psychology, 90(3), 509-522. Hofmann, D. A., & Stetzer, A. (1996). A cross-level investigation of factors influencing unsafe behaviors and accidents. Personnel Psychology, 49, 307-339. Hox, J. (2002). Sample sizes and power analysis in multilevel regression. In J. Hox, Multilevel analysis, techniques, and applications. Mahwah, MJ: Lawrence Erlbaum Associates. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55. ‘ Innes, D. (1993). Affirmative action: Issues and strategies. In D. Innes, M. Kentridge and H. Perold (eds.), Reversing discrimination: Affirmative action in the workplace (pp. 4-21). Cape Town: Oxford University Press. Iyer, A., Leach, C. W., & Crosby, F. J. (2003). White guilt and racial compensation: The benefits and limits of self-focus. Personality and Social Psychology Bulletin. 29. 117—129. Jain, H. C., Sloane, P. J ., & Horwitz, F. M. (2003). Employment equity and affirmative action: An international comparison. Armonk, NY: M. E. Sharpe. 328 James, L. R, Choi, C. C, K0, C. H. E., McNeil, P. K., Minton, M. K., Wright, M. A., & Kim, K. (2008). Organizational and psychological climate: A review of theory and research. European Journal of Work and Organizational Psychology, ]7( 1 ). 5-32. James, L. R. & Williams, L. J. (2000). The cross-level operator in regression, ANCOVA, and contextual analysis. In K. J. Klein & S. W. J. Kozlowski (Eds.), Multilevel theory, research, and methods in organizations. San Francisco, CA: Jossey-Bass. Joreskog, K., & Stirbom, D. (2004). LISREL 8.71 [Computer software]. Lincolnwood. IL: Scientific Software International, Inc. Judge, T.A., Erez, A., Bono, J.E., & Thoresen, C.J. (2003). The core self-evaluations scale: Development of a measure. Personnel Psychology, 56, 303-331. Kaiser, C. R., Drury, B. J., Spalding, K. E., Cheryan, S., & O’Brien, L. T. (2009). The ironic consequences of Obama’s election: Decreased support for social justice. Journal of Experimental Social Psychology, 45, 556—559. Kanfer, R., Wanberg, C. R., & Kantrowitz, T. M. (2001). Job search and employment: A personality-motivational analysis and meta-analytic review. Journal of A pplied Psychology, 86(5), 837-855. Kantor, P. (2001). CCMA: A commentary on the rules. Claremont, South Africa: Siber Ink. Kertzer, D. I., & Arel, D. (2002). Census and identity: The politics of race, ethnicity and language in national censuses. Cambridge, UK: Cambridge University Press. Klandermans, 8., Werner, M., & van Doom, M. (2008). Redeeming apartheid’s legacy: Collective guilt, political ideology, and compensation. Political Psychology. 29(3), 331-349. Konrad, A. M., & Linnehan, F. (1995). Race and sex differences in line managers‘ reactions to equal employment opportunity and affirmative action interventions. Group &* Organization Management, 20(4), 409-439. Kozlowski, S. W. J ., & Klein, K. J. (2000). A multilevel approach to theory and research in organizations. In K. J. Klein & S. W. J. Kozlowski (Eds.), Multilevel theory. research, and methods in organizations. San Francisco, CA: Jossey-Bass. Kravitz, D. A. (1995). Attitudes toward affirmative action plans directed at blacks: Effects of plan and individual differences. Journal of Applied Social Psychology, 25(24), 2192-222. 329 .IfifT '- Kravitz, D. A. (2008). The diversity-validity dilemma: Beyond selection — the role of affirmative action. Personnel Psychology, 61(1), 173-193. Kravitz, D. A., Harrison, D. A., Turner, M. E., Levine, E. L., Chaves, W., Brannick, M. T., Denning, D. L., Russell, C. J., & Conard, M. A. (1996). Affirmative action: A review of psychological and behavioral research. Bowling Green, OH: Society for Industrial and Organizational Psychology. Kravitz, D. A., & Klineberg, S. L. (2000) Reactions to two versions of affirmative action action among Whites, Blacks, and Hispanics. Journal of Applied-Psychology, 85. 597-611. Kravitz, D. A., & Platania, J. (1993). Attitudes and beliefs about affirmative action: Effects of target and of respondent sex and ethnicity. Journal of Applied Psychology, 78(6), 928-938. Kulik, C. T., & Perry, E. L. (1994). Heuristic processing in organizational judgments. In L. Heath, R. S. Tindale, J. Edwards, E. J. Posavac, F. B. Bryant, E. Henderson- King, Y. Suarez-Balcazar, & J. Myers (Eds.), Applications of heuristics and biases to social issues. New York, NY: Plenum Press. Lance, C. E. & Vandenberg, R. J. (2002). Confirmatory factor analysis. In F. Drasgow & N. Schm1tt (Eds) Measuring and analyzing behavior in organizations. San Francisco, CA: Jossey-Bass. Lance, C. E., Vandenberg, R. J., & Self, R. M. (2000). Latent growth models of individual change: The case of newcomer adjustment. Organizational Behavior and Human Decision Processes, 83 , 107-I4. Leventhal, G. S. (1976). The distribution of rewards and resources in groups and organizations. In L. Berkowitz & W. Walster (Eds.), Advances in experimental social psychology (Vol. 9, pp. 91-131). New York, NY: Academic Press. Leventhal, G. S. (1980). What should be done with equity theory? New approaches to the study of fairness in social relationships. In K. Gergen, M. Greenberg. & R. Willis (Eds.), Social exchange: Advances in theory and research (pp. 27-55). New York. NY: Plenum Press. Levinson, H., Price, C. R., Munden, K. J., Mandl, H. J., & Solley, C. M. (1966). Men. management, and mental health. Cambridge, MA: Harvard University Press. Levi, A. S., & Fried, Y. (2008). Differences between African Americans and Whites in reactions to affirmative action programs in hiring, promotion, training, and layoffs. Journal of Applied Psychology, 93(5), 1118 —1129. 330 Littrell, R. F., & Nkomo, S. M. (2005). Gender and race differences in leader behaviour preferences in South Africa. Women in Management Review, 20(8), 562-58. MacCallum, R. C., Widaman, K. F., Zhang, S., and Hong, S. (1999). Sample size in factor analysis. Psychological Methods, 4, 84-99. Madi, P. M. (1997). Black economic empowerment in the new South Africa: The rights and the wrongs. Randburg, South Africa: Knowledge Resources. Major, D. A., Kozlowski, S. W. J., Chao, G. T., & Gardner, P. D. (1995). A longitudinal investigation of newcomer expectations, early socialization outcomes, and the moderating effects of role development factors. Journal of Applied Psychology, 80, 418-431. Mallet, R. K., & Swim, J. K. (2007). The influence of inequality, responsibility and justifiability on reports of group-based guilt for ingroup privilege. Group Processes & Intergroup Relations, 10(1), 57-69. Martins, L. L., & Parsons, C. K. (2007). Effects of gender diversity management on perceptions of organizational attractiveness: The role of individual differences in attitudes and beliefs. Journal of Applied Psychology, 92(3), 865— 875. McFarlin, D. B., & Sweeney, P. D. (1992). Distributive and procedural justice as predictors of satisfaction with personal and organizational outcomes. Academy of Management Journal, 35, 626-637. Meade, A. W., Johnson, E. C., Braddy, P. W. (2008). Power and sensitivity of altemative fit indices in tests of measurement invariance. Journal of A pplied Psychology, 93(3), 568-592. Miller, K., Haskell, C., & Thatcher, A. (2002). The relationship between intention to emigrate and organisational commitment. South A frican Journal of Psychology. 32(3), 16-2. 8 Miller, D. T., & Tumbull, W. (1986). Expectancies and interpersonal processes. Annual Review of Psychology, 3 7, 233-56. Mkandawire, T. (1993). Problems and prospects of social sciences in Africa. International Social Science Journal, 45(1(135)), 129-14. Moorman, R. H. (1991). Relationship between organizational justice and organizational citizenship behaviors: Do fairness perceptions influence employee citizenship? Journal of Applied Psychology, 76, 845-855. 331 Morgeson, F. P., & Hofman, D. A. (1999). The structure and function of collective constructs: Implications for multilevel research and theory development. Academy of Management Review, 24(2), 249-265. Morris, M. W., & Leung, K. (2000). Justice for all? Progress in research on cultural variation in the psychology of distributive and procedural justice. Applied Psychology: An International Review, 49(1), 100-132. Morrison, E. W., & Robinson, S. L. (1997). When employees feel betrayed: A model of how psychological contract violation develops. The Academy of Management Review, 22(1), 226-256. Mossholder, K. W., Bennett, N., & Martin, C. L. (1998). A multilevel analysis of procedural justice context. Journal of Organizational Behavior, 19, 131-141. Murithi, T. (2006). Practical peacemaking wisdom from Africa: Reflections on Ubuntu. The Journal of Pan African Studies, 1(4), 25-34. Murphy-Berman, & Berman, J. J. (2002). Cross-cultural differences in perceptions of distributive justice: A comparison of Hong Kong and Indonesia. Journal (J'C‘ross- Cultural Psychology, 33(2), 157-17. Murray, B. (2002). Psychology tackles apartheid’s aftermath. Monitor on Psychology. 33 (1), 50-51. Nacoste, R. W. (1994). Policy schemas for affirmative action. In L. Heath, R. S. Tindale, J. Edwards, E. J. Posavac, F. B. Bryant, E. Henderson-King, Y. Suarez-Balcazar, & J. Myers (Eds). Applications of heuristics and biases to social issues. New York, NY: Plenum Press. Nattrass, N., & Seekings, J. (2001). “Two nations”? Race and economic inequality in South Africa today. Daedalus, Winter, 45-7. Nauman, S. E., & Bennett, N. (2000). A case for procedural justice climate: Development and test of a multilevel model. Academy of Management Journal, 43, 881-889. Nicholas, 8., Povey, D., Walker, A., & Kershaw, C. (2005, July). Home Office statistical bulletin: Crime in England and Wales 2004/2005. Downloaded on October 13. 2006, from http://www.homeoffice.gov.uk/rds/pdstS/hosb1 105.pdf Niehoff, B. P., & Moorman, R. H. (1993). Justice as a mediator of the relationship between methods of monitoring and organizational citizenship behaviors. Academy of Management Journal, 3 6, 527-556. Ndebele, P. (September, 2005). Reviewing and Conducting Multi-nation and Industry Sponsored Research: Challenges for African Institutions and Countries. Invited 332 presentation for the African Studies Center, Michigan State University, East Lansing, MI. Nowakowski, J. M., & Conlon, D. E. (2005). Organizational justice: Looking back, looking forward. International Journal of Conflict Management, 16, 4-29. Parker, C. P., Baltes, B. B., & Christiansen, N. D. (1997). Support for affirmative action. justice perceptions, and work attitudes: A study of gender and racial-ethnic group differences. Journal of Applied Psychology, 82(3), 376-389. Peta, B. (2005, February 26). Please come home: South Africa backs campaign to attract exiles. The (London) Independent. Downloaded on September 9, 2007, from http://findarticles.com/p/articles/mi_qn4158/is_20050226/ai_n1 1 827857 Pettigrew, T. F., & Tropp, L. R. (2005). Allport’s intergroup contact hypothesis: Its history and influence. In Dovidio, J. F., Glick, P., & Rudman, L. A. (Eds.), On the nature of prejudice: Fifty years after Allport. Malden, MA: Blackwell Publishing. Ployhart, R. E., & Ryan, A. M. (1998). Applicants’ reactions to the fairness of selection procedures: The effects of positive rule violations and time of measurement. Journal of Applied Psychology, 83, 3-16. Pratto, F., Tatar, D. G., & Conway-Lanz, S. (1999). Who gets what and why: Determinants of social allocations. Political Psychology, 20(1), 127-15. Price, J. L., & Mueller, C. W. (1986). Handbook of organizational measurement. Marshfield, MA: Pitman. Rainbow nation’s outsiders live in fear. (2009, May 28). BBC News. Retrieved August 4. 2009, from httpz/lnews.bbc.co.uk/2/hi/africa/8070919.stm Ramsay, L. J. (2005, May). Expectations and frustrations: Cross group reactions to workplace affirmative action in South Africa. Presented in I. Nikolaou, I. T. Robertson, & J. F. Salgado (Co-chairs), Organizational convivence: The role of employee selection and assessment. Symposium conducted at the 12th European Congress on Work and Organizational Psychology, Istanbul, Turkey. Ramphele, M. (2001). Citizenship challenges for South Africa. Daedalus, Winter, 1-17. Raudenbusch, S., & Bryk, A. (2002). Hierarchical linear models: Applications and data analysis methods. Thousand Oaks, CA: Sage Publications. Raudenbusch, S., Bryk, A., & Congdon, R. (2008). HLM 6. 06. [Computer softwarcl. Lincolnwood, IL: Scientific Software International, Inc. 333 Roche, D. (2006). Dimensions of restorative justice. Journal of Social Issues, 62(2), 217—238. Root, M. (1996). The multiracial experience: racial borders as a significant frontier in race relations. In M. Root (Ed.), The multiracial experience: Racial boundaries as the new Frontier. Thousand Oaks, CA: Sage Publications. Ross, R. (1999). A concise history of South Africa. Cambridge, UK Cambridge University: Press. Rousseau, D. M. (1985). Issues of level in organizational research: Multi-Ievel and cross- Ievel perspectives. Research in Organizational Behavior, 7, 1-3 7. Rousseau, D. M. (1995). Psychological contract in organizations: Understanding written and unwritten agreements. Thousand Oaks, CA: Sage Publications. Rousseau, D. M., & Tijoriwala, S. A. (1998). Assessing psychological contracts: Issues. alternatives, and measures. Journal of Organizational Behavior. 1 9, 679-695. Runciman, W. G. (1966). Relative deprivation and social justice. London: Routledge & Kegan Paul. Ryan, A. M., Horvath, M., & Kriska, S. D. (2005). The role of recruiting source informativeness and organizational perceptions in decisions to apply. International Journal of Selection and Assessment, I 3(4), 235-248. Ryan, A. M., & Ployhart, R. E. (2000). Applicants’ perceptions of selection procedures and decisions: A critical review and agenda for the future. Journal of Management, 26, 565-606. Ryan, A. M., Sacco, J. M., McFarland, L. A., & Kriska, S. D. (2000). Applicant self- selection: Correlates of withdrawal from a multiple hurdle process. Journal of Applied Psychology, 85(2), 163-179. ‘ Salancik, G. R, & Pfeffer, J. (1978). A social information processing approach to job attitudes and task design. Administrative Science Quarterly, 23, 224-253. Schmit, M. J., & Ryan, A. M. (1997). Applicant withdrawal: The role of test-taking attitudes and racial differences. Personnel Psychology, 50(4), 85 5-876. Schmitt, N. & Chan, D. (1999). The status of research on applicant reactions to selection tests and its implications for managers. International Journal of Management Reviews, I, 45-62. Shnabel, N., & Nadler, A. (2008). A needs-based model of reconciliation: Satisfying the differential emotional needs of victim and perpetrator as a key to promoting reconciliation Journal of Personality and Social Psychology, 94(1), 1 16 —132. 334 Schwab, D. P. (1980). Construct validity in organizational behavior. In B. Straw & L. 1.. Cummings (Eds.), Research in organizational behavior (Vol. 2, pp.3-43). Greenwich, CT: JAI Press. Seibert, S. E., Silver, S. R., & Randolph, W. A. (2004). Taking empowerment to the next level: A multiple-level model of empowerment, performance, and satisfaction. Academy of Management Journal, 4 7(3), 332-349. Sellers, R. M., Rowley, S. A. J., Chavous, T. M., Shelton, J. N., & Smith, M. A. (1997). Multidimensional Inventory of Black Identity: A preliminary investigation of reliability and constuct validity. Journal of Personality and Social Psychology. 73, 805-815. Sethi, S. P., & Williams, 0. F. (2000). Economic imperatives and ethical values in global business: The South A fi'ican experience and international codes today. Boston, Massachusetts: Kluwer Academic Publishers. Shadish, W. R, Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi- experimental designs for generalized causal inference. Boston, MA: Houghton Mifflin Company. Shapiro, D. L., & Kirkman, B. L. (1999). Employees’ reaction to the change to work teams: The influence of anticipatory injustice. Journal of Organizational Change Management, 12, 51-66. Shapiro, D. L., & Kirkman, B. L. (2001). Anticipatory injustice: The consequences of expecting injustice in the workplace. In J. Greenberg and R. Cropanzano, Advances in organizational justice. Stanford, CA: Stanford University Press. Simons, T., & Roberson, Q. (2003). Why managers should care about fairness: The effects of aggregate justice perceptions on organizational outcomes. Journal of Applied Psychology, 88(3), 432-443. Slaughter, J. E., Sinar, E. F., & Bachiochi, P. D. (2002). Black applicants’ reactions to affirmative action plans: Effects of plan content and previous experience with discrimination. Journal of Applied Psychology, 87(2), 333-344. Son Hing, L. S., Bobocel, D. R., & Zanna, M. P. (2002). Meritocracy and opposition to affirmative action: Making concessions in the face of discrimination. Journal of Personality and Social Psychology, 83(3), 493—509. South Africa discontent spreads. (2009, July 22). BBC News. Retrieved July 22. 2009. from http://news.bbc.co.uk/go/pr/fi/-/2/hi/africa/8163187.stm South African Advertising Research Foundation (2005). Living Standards Measure. Downloaded from http://www.saarf.co.za/ 335 South African Department of Labour. http://www.labour.gov.za South African Department of Labour (n.d.). Women in the South African Labour Market 1995 -— 2005. Retrieved June 6, 2007, from http://www.labour.gov.za South African Institute for Entrepreneurship (2006). South-Afiica: Alive with possibility? Retrieved September 13, 2007, from http://www.entrepreneurship.co.za/ South African Police Services (2008). Murder. Retrieved July 21, 2009, from http://www.saps.gov.za/statistics/reports/erimestats/ZOO8/crime_stats__2008.htm Spybrook, J ., Raudenbusch, S. W., Liu, X., Congdon, R., & Martinez, A. (2008). Optimal design for longitudinal and multilevel research: Documentationfor the “Optimal Design” software, last revised on March 12, 2008. Retrieved June 14, 2008. from http://sitemaker.umich.edu/group-based/optimal_design_sofiware Statistics South Africa (2005). Documented migration, 2003. Retrieved June 6. 2007. from http://www.statssa.gov.za Statistics South Africa (2003). General household survey, July 2002. Retrieved April 9. 2007, from http://www.statssa.gov.za l Statistics South Africa (2006a). General household survey, July 2005. Retrieved April 9. 2007, from http://www.statssa.gov.za Statistics South Africa (20083). General household survey, 2007. Retrieved July 27. 2009, from http://www.statssa.gov.za Statistics South Africa (2000). Mid-year estimates statistical release P0302. Retrieved July 27, 2009, from http://www.statssa.gov.za Statistics South Africa (2006b). Mid-year population estimates, South Africa 2006. Retrieved April 9, 2007, from http://www.statssa.gov.za Statistics South Africa (2009). Mid-year population estimates, South Africa 2009. Retrieved July 27, 2009, from http://www.statssa.gov.za Statistics South Africa (2008). Labour force survey, September 200 7. Retrieved March 27, 2008, from http://www.statssa.gov.za Statistics South Africa (2009). Statistical release P021 I Quarterly Labour Force Survey. Quarter 1, 2009. Retrieved July 27, 2009, from http://www.statssa.gov.za Stead, G. B., Els, C., & Fouad, N. (2004). Perceived career barriers among South African high school learners. South African Journal of Psychology, 34(2), 206-221. 336 Steele, S. (1990). The content of our character: A new visibn of race in America. New York: St. Martin’s Press. Steiner, D. D., & Gilliland, S. W. (2001). Procedural justice in personnel selection: International and cross-cultural perspectives. International Journal of Selection and Assessment, 9(1/2), 124-137. Steiner, D. D., Trahan, W. A., Haptonstahl, , D. E., & Fointiat, V. (2006). Thejustice of equity, equality, and need in reward distributions: A comparison of French and American respondents. Revue Internationale de Psychologie Sociale, 19(1), 49- 74. Stouffer, S. A., Suehman, E. A., De Vinney, L. C., Star, S. A.,& Williams, R. M., Jr. (1949). The American soldier: Adjustment during Army life, Volume 1. Clinton. MA: Colonial Press. Swim, J. K., & Miller, D. L. (1999). White guilt: Its antecedents and consequences for attitudes toward affirmative action. Personality and Social Psychology Bulletin. 25, 500-514. Taris, T. W., Feij, J. A., & Capel, S. (2006). Great expectations — and what comes of it: The effects of unmet expectations on work motivation and outcomes among newcomers. International Journal of Selection and Assessment, 1 4, 256-267. Tekleab, A. G., & Taylor, M. S. (2003). Aren't there two parties in an employment relationship? Antecedents and consequences of organization-employee agreement on contractual obligations and violations. Journal of Organizational Behavior, 24(5), 585-608. Terreblanche, S. (2002). A history of inequality in South Africa. Pietemiaritzburg. South Africa: University of Natal Press. Thibaut, J. & Walker, L. (1975). Procedural justice: A psychological analysis. Hillsdale, NJ: Lawrence Erlbaum Associates. Thomas, H. D. C., & Anderson, N. (1998). Changes in newcomers' psychological contracts during organizational socialization: A study of recruits entering the British Army. Journal of Organizational Behavior, 19(745-767). Thomas, D. C., Au, K., & Ravlin, E. C. (2003). Cultural variation and the psycholological contract. Journal of Organizational Behavior, 24(5), 451-471. Thompson, L. (1990). A history of South Africa. New Haven, CT: Yale University Press. 337 Triandis, H. C., Bontempo, R., Villareal, M., Asai, M., & Lucca, N. (1988). Individualism and collectivism: Cross-cultural perspectives on self-ingroup relationships. Journal of Personality and Social Psychology, 5 4. 323-338. Trompenaars, F., & Hampden—Tumer, C. (1998). Riding the waves of culture: Understanding diversity in global business (2nd ed.). New York, New York: McGraw-Hill. Truxillo, D. M., & Bauer, T. N. (1999). Applicant reactions to test score banding in entry-level and promotional contexts. Journal of Applied Psychology, 84, 322- 339. ‘ Truxillo, D. M., Steiner, D.D., & Gilliland, S. W. (2004). The importance of organizational justice in personnel selection: Defining when selection fairness really matters. International Journal of Selection and Assessment, 12, 39-53. Tumley, W. H., Bolino, M. C., Lester, S. W., & Bloodgood, J. M. (2003). The impact of psychological contract fulfillment on the performance of in-role and organizational citizenship behaviors. Journal of Management, 29(2), 187-206. Tumley, W., H., & Feldman, D. C. (1999). The impact of psychological contract violations on exit, voice, loyalty, and neglect. Human Relations, 52(7), 895-922. US. Department of Justice - Federal Bureau of Investigation (2008). Crime in the United States, 2007. Retrieved July 27, 2009, from http://www.fbi.gov/ucr/ciusZOO7/data/table_01 .html Vandenberg, R. J ., & Lance, C. E. (2000). A review and synthesis of the measurement invariance literature: Suggestions, practices, and recommendations for organizational research. Organizational Research Methods, 3, 4-69. Van Rensburg, K. J., & Roodt, G. (2004). Perceptions of employment equity and black economic empowerment as predictors of employee commitment. Paper presented at the Annual Conference of the Society for Industrial and Organizational Psychology of South Africa, Pretoria, South Africa. Van Rensburg, K. J., & Roodt, G. (2005). The perceptions of employment equity and black economic empowerment as predictors of union commitment. SA Journal of Industrial Psychology, 31, 55-64. Vaux, A., Riedel, S., & Stewart, D. (1987). Modes of social support: The social support behaviors. American Journal of Community Psychology, 15, 209-23 7. Vavi, Z. (2007). Address to Numsa 20th Anniversary 19 - 05 - 07. Retrieved June 6. 2007, from http://www.cosatu.org.za/speeches/2007/zv20070519.htm 338 Vestergaard, M. (2001). Who’s got the map?: The negotiation of Afrikaner identities in post-apartheid South Africa. Daedalus, Winter, 19—44. Waddington, K. (1999). The brain drain. Focus, 10(3), 8-9. Waddy, N. L. (2004). Affirmative action versus nonracialism in the new South Africa. African Issues, (Fall), 1-8. Waller, L. (2006). Migration of skills in South Africa: Patterns, trends and challenges (Migration Policy Brief No. 20). Cape Town, South Africa: Southern African Migration Project. Walker, C. (1979). The women’s suflrage movement in South Africa. Cape Town. South Africa: University of Cape Town Printing Department. Ware, V. (2002). Otherworldly knowledge: Toward a “language of perspicuous contrast.” In V. Ware and L. Back (Eds), Out of whiteness: Color, politics, and culture. Chicago, IL: The University of Chicago Press. Weathers, V. M., & Truxillo, D. M. (2008). Whites’ and Asian Americans’ perceptions of Asian Americans as targets of affirmative action: Journal of Applied Social Psychology, 38(1 1), 2737—2758. White, F. A., Charles, M. A, Nelson, J. K. (2008). The role of persuasive arguments in changing affirmative action attitudes and expressed behavior in higher education. Journal of Applied Psychology, 93(6), 1271—1286. Wood, G. (2001). Shop floor democracy in the Congress of South African Trade Unions. South A fi'ican Journal of Labour Relations, 25 (3&4), 4-21. Zhao, H., Seibert, S. E., & Hills, G. E. (2005). The mediating role of self-efficacy in the development of entrepreneurial Intentions. Journal of A pplied Psychology, 90(6). 1265—1272. 339