ll.- IllPo . It. I On 9:; run-“r ah.»- [Io-.00.?! . I . I I n . .. ...I .. . .I IV I I I? I. .. I -II I . I . . I .. . .,. .I . I ~. I is .I I - I I.n . . I-I V m... . .I I . u I - .I I I .. I- . I . . I .. I c . I- II . .II, V .1. .l I L I. I II n v DI .I . .IIIIIII . . w .I II. I - I I. I III. .I.. . . , In II?” I. 'I II ‘I 1'! Ll ',VI 7 I I II I. v. I I I I- I... I I I I l n . at I w I ,I III .mL I x I I .I . III II I». I I I} I III . I I I”..- I”. I I . I. I o .a II VHDI. . II 1 I l I III ,I' I I II I I II.I.Is I luI I I n. , I V IN. II . I , II\: I “I . 7 .I an. I . I. v.1 . I VI "VII .I I «I .. ,v III .I I Ifiv‘ ‘ I I . II II I .._I I II n . I . . I I.’ I I xv: I. I. . . V I I . ll... _ v I. . I . I. IIOl . I. I .II V I I III I I I I» II LI I, u . ‘ I II IIIII I I) W his. I I I ILII v.1.‘|IIII. . II I n I. II I . IIIIII I . ., II I I, I I. I .III I I I I I I I . IIIIIII I ..II I III I I I I In)“ I» IN IIIII. . t I I I I .Ii .. II I Qt! I I I. In II IV. I OtI-Iul .I I u .l . vb I .III. .. \I .I v I I I I I}: I . I I Io; I.II 1.. I II I I .0. LI rat I I I I I . .. . . .I . 3., I. I .. II..-EIII-.--. EnItI I I [2,. I I I . I . I 1.3.-- rIuII r I It It 91... .3... . I. I .I I. I II. «:.III JIJ....u..-.nR...P...~h IIIwIIWII .It ..... . JoIII.., .II. EIhIIIIhk .-I -IIII fr!!! 3...). I II. II: . I I I V“ . . ... II DWI III IQI. II I I . .. I I I I I I III I D I, I IIIII'I III I'll‘. l 1! ill" I [II I r I III l’bIQ—IS‘IVI MICHIGAN STATE UNIVERSITY LIBRARI s ' > I IIIIIIIIIII i'IIIIIf III ‘ HER-ARV 3 1293 90.582 0406 F5- Michigan State University I III This is to certify that the dissertation entitled Absence Intentions: A Construct Validation Study presented by Joseph John Martocchio has been accepted towards fulfillment of the requirements for PhoD. degreein LoI-R. %/Ja/%}/m P1 .2). Major professor DaW'éy/e/7d’7 MSU is an Affirmative Action/Equal Opportunity Institution 0-12771 —_.__—.—._. PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or betore due due. ,DATE DUE DATE DUE DATE DUE MSU Is An Afflrmdlve Action/Equal Opportunity Infiltution ABSENCE INTENTIONS: A CONSTRUCT VALIDATION STUDY BY Joseph John Martocchio A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY School of Labor and Industrial Relations 1989 ABSTRACT ABSENCE INTENTIONS: A CONSTRUCT VALIDATION STUDY By Joseph John Martocchio Results of employee absenteeism research have been disappointing because weak effects have undermined theoretical expectations that work-related attitudes (e.g., job satisfaction, organizational commitment) and voluntary absence behavior are strongly related. Results of recent meta-analyses indicated that each of these work-related attitudes has accounted for less than four percent of the variance in voluntary absenteeism. Based on a review of the employee absenteeism literature, it is suggested that weak effects may be the result of failure to examine whether behavioral intentions are the most immediate antecedents of absence behavior. As in the employee turnover literature, an absence intentions measure is proposed based on theoretical arguments that intention is 6 din??? .antengdgny .nfingnavigr-117ishbein. e..-’ _. ‘fiflfi ... _......A. 1967; Triandis, 1977) and that stronger effects found in the employee turnover literature are because of the incorporation of a turnover intentions measure (Steel & Ovalle, 1984). Thus, the primary objective of this dissertation is to assess the degree of construct validity of an absence intentions measure. The review of the literature also revealed that measurement unreliability and violations of key assumptions Joseph John Martocchio underlying least squares correlation/regression may be responsible for weak effects. A secondary objective is to conduct exploratory analyses to estimate whether these measurement and methodological issues might be responsible for weak effects. Based on analyses of 440 employees (176 blue collar- unskilled and 264 white collar-clerical) employed in a diversified financial services organization, a sufficient degree of construct validity of an absence intentions measure was demonstrated. As,predicted, absence intentipns significantlywpredictedlyoluntarymabsenteeism. Contrary to prediction, absence intentions significantly predicted involuntary absenteeism. These results lend support to the idea that the voluntary-involuntary distinction is artificial in terms of an individual's control over absenteeism (Hackett & Guion, 1985). One would not expect an individual’s intentions to be absent to be related to absence that is beyond personal control. Measurement and methodological issues were ruled out as alternative explanations to weak effects in the study. Implications for research and theory, alternative explanations, implications for practice, and directions for future research were discussed. Copyright by JOSEPH JOHN MARTOCCHIO 1989 .(m'nn To Rose and Joe (Ma and Dad): 'Iheir generosity and faith have enabled metoreadithestaxsandlivemdrean's. In loving memory of Papa: His respect for knowledge lives on. mom I am a very privileged person; this dissertation is only a physical manifestation of my status. I have been able to attain the education of my choice which has provided me with more wonderful options than I will ever be abletopursue. Altlnghlhavemrkedveryhanitwardneetingmy objectives, I have many people to whan I am grateful for different reasons. At Michigan State University, I thank Mike Moore, my chairperson and advisor, for providing me with an opportunity to do well as an aspiring scholar, sharing his knowledge with me, and allowing me to make a contrthtion to the School of Labor and Industrial Relations. Also, Mike’s interest inresearchonemployee absenteeismwas contagious, thus, this dissertatim. Rich Block and Sandy Gleason provided me with continuous financial suHJort and funding for professional conferences. They also taught us to appreciate other apsects of the industrial relations field. John Holla'beck will always be an excellent role model as I progress in my career. He instilled in me the inportance of rigor and precision. He was always available for advice and taught me what "go for the gold" means. I am very grateful to Linda Jackson, Dan Ilgen, Neal Scl'nnitt, and Ellen Kossek for their advice. 'Ihanks, Annette Bacon and Nita Berridge for making me laugh during sane tough mnerrts. At General Electric Aerospace, I will always be indebted to Jerry caplm. Jerry provided me with the qportunity to practice sane of what I learned. Even though Jerry knew I would ultimately pursue an academic career, he made many contributions to my professional developnent. Most importantly, Jerry made it possible for me to conduct my dissertation research in a "real" catpany with "real" people. Thank you, Phyllis Sprague vi forbeirgsomthun. Welamghedsomxzhsooften, yetwealwaysgotour work done. Half of the employees at GE-Valley Forge probably think we are rnts....d'1 well, that’s life” I am very grateful to Wiley Harris who provided funding for this dissertationaswellasaccesstohisenployees. 'Ihanksalsogotolynette Griffin, Georgietta Harris, Anne Marie Hcoey, Mike Herrin, Bob Hubregson, Jerry Pavelich, Debbie Harris, Ianny Branner, Karl Kiracofe, and Dennis Raab fortheirwillingmsstointemlptuieirmsysdxedtflestohelpneneetmy objectives in a timely fashim. Special thanks to Lisa Shoop for her patience and help. John Hollenback (no, not John Hollenbeck fran Michigan State), director of personnel research worked diligently with me to market my proposalarrirallythesmortofotherswithinthecatpany. Of course, I have been very fortunate for the friends I have made during nystayatMichiganStateaswell asmyfamily. 'IhankymImneO'Ieary, Don Flachs, Ellen Whitener, and Wayne Cooley for your friendship, caring, support, and unique inputs into this dissertation. Last, but certainly never least, Ithankmyparents. whammdmxcorriitionallyencmragedneto puraie my education. Because of their unselfishness, the quality of my life will be infinitely better. I will always be grateful. Joe Martocchio East Iansing, Michigan May 7 , 1989 vii CHAPTER CHAPTER CHAPTER CHAPTER TABLE OF CONTENTS ONE: INTRODUCTION AND OVERVIEW Importance of the Topic Research Strategy Key Assumptions and Limitations Contributions Outline of this Dissertation TWO: LITERATURE REVIEW Review of Theoretical Models Review of Measurement Issues Review of Methodological Issues THREE: RESEARCH MODEL AND HYPOTHESES Construct Validity: An Overview A Nomological Network for an Absence Intention Construct Absence Intentions Absence Behavior Job Satisfaction Organizational Commitment Personal Characteristics "Weak" Effects: Theory-Based or Methods— Based FOUR: METHOD The Organization Under Study Power Analysis Sample Data Collection Procedure Operationalization of Variables Data Analysis Procedure viii 10 11 12 14 14 24 32 35 36 37 38 58 59 68 7O 73 76 76 78 78 80 85 94 CHAPTER CHAPTER LIST OF FIVE: RESULTS Item and Scale Analyses: Absence Intentions (Hypothesis 1) Item and Scale Analyses: Attitudes Toward Absenteeism Item and Scale Analyses: Subjective Norms Absence Behavioral Measures Properties of Sample Distributions of Variables Results of Analyses by Hypothesis Hypothesis 2 Hypothesis 3 Hypothesis 4 Hypothesis 5 Hypothesis 6 Hypothesis 7 Hypothesis 8 Hypothesis 9 Hypothesis 10 Hypothesis 11 Hypothesis 12 Hypothesis 13 Hypothesis 14 Hypothesis 15 Hypothesis 16 Hypothesis 17 Hypothesis 18 Hypothesis 19 Hypothesis 20 Hypothesis 21 Hypothesis 22 Hypothesis 23 Hypothesis 24 Summary SIX: DISCUSSION Summary Implications for Research and Theory Alternative Explanations Implications for Practice Suggestions for Future Research REFERENCES APPENDIX A: Questionnaire ix 101 101 114 123 128 130 133 133 134 136 140 144 149 150 152 153 155 156 157 159 163 164 165 166 169 171 172 174 175 176 177 179 179 183 190 193 198 201 214 1. 2. 3. 4. 5. 7. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. LIST OF TABLES Dissertation Study Variables Demographic Profile of Final Sample Intercorrelation of Dissertation Study Variables Internal Consistency Reliability Estimates Results of Factor Analysis of Intention Items Reliability Coefficients for Intention Factors Means, Standard Deviations and Intercorrelation of Intention Items Means, Standard Deviations and Intercorrelations of Intentions Factors Item-Scale Correlations: Break From Work Items Item-Scale Correlations: Community and Religious Items Item-Scale Correlations: Family Responsibility Items Results of Factor Analysis of Belief Items Reliability Coefficients for Belief Factors Means, Standard Deviations, and Intercorrelations of Belief Items Item-Scale Correlations: Motivating Outcome Items Item-Scale Correlations: Deterrent Outcome Items Results of Factor Analysis of Subjective Norm Items Reliability Coefficients for Subjective Norm Items Means, Standard Deviations, and Intercorrelations of Subjective Norm Items Means, Standard Deviations, and Intercorrelations of Absence Behavioral Measures Sample Distributions of Variables Hierarchical Regression Results for Absence Intentions Predicted from Attitudes and Norms T-Tests Comparing Blue Collar and White Collar Workers on Absence Attitudes, Norms, Intentions, and Behavior Regression Results for Voluntary Absenteeism Predicted Absence Intentions Regression Results for Involuntary Absenteeism Predicted from Absence Intentions Regression Results for Testing Absence Intentions as a Mediating Factor Between Attitudes, Norms, and Behavior 39 81 86 87 103 106 108 110 111 112 113 115 117 119 121 122 124 126 127 129 132 134 136 140 144 146 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. Regression Results for Voluntary Absenteeism Predicted from Job Satisfaction T-Tests Comparing Correlations Between Absence Intentions-Absence Behavior and Job Satisfaction-Absence Behavior Regression Results for Absenteeism Predicted from Perception of Ability to Find A Job T-Tests Comparing Correlations Between Absence Intentions-Absence Behavior and Finding a Job- Absence Behavior Regression Results for Absenteeism Predicted From Kinship Responsibilities T-Tests Comparing Correlations Between Absence Intentions-Absence Behavior and Kinship Responsibilities-Absence Behavior Regression Results for Testing Job Involvement as a Mediating Factor Between Job Satisfaction and Absence Behavior Regression Results for Voluntary Absenteeism Predicted from Organizational Commitment T-Tests Comparing Correlations Between Absence Intentions-Absence Behavior and Organizational Commitment-Absence Behavior Regression Results for Testing Organizational Commitment as a Direct Predictor of Absence Behavior Zero-Order Correlations: Kinship Responsibilities and Primary Income Producer With Measures of Absence Behavior Partial Correlations: Kinship Responsibilities and Primary Income Producer Role with Measures of Absence Behavior Controlling for Sex Regression Results for Voluntary Absenteeism Predicted from Work Ethic T-Tests Comparing Correlations Between Absence Intentions-Absence Behavior and Work Ethic- Absence Behavior Comparison: Hypothesized vs. Actual Relationship Between Absence Intentions and Studied Variables xi 150 152 153 155 157 159 161 163 165 167 170 171 172 174 178 LIST OF FIGURES 1. Ajzen and Fishbein's Theory of Reasoned Action as a Basis for Predicting Voluntary Absenteeism xii 40 CHAPTER ONE INTRODUCTION AND OVERVIEW This chapter provides an introduction and overview of the entire dissertation. The introduction and overview include: (a) why it is of importance to investigate absence inten- tions, (b) the research strategy, (c) the key assumptions and limitations of this research, (d) contributions of this research, and (e) an outline of the subsequent chapters in the dissertation. Importance of the Topic Introduction We know that voluntary absenteeism is pervasive across organizational types. In addition, its costs to organizations are exorbitant (Steers & Rhodes, 1984). Unfortunately, our understanding of employee absenteeism has been limited because of weak effects or "small" effects (Cohen, 1977). In this case, weak effects are low correlations between measures of employee absenteeism and measures of relevant work-related attitudes (e.g., job satisfaction). Weak effects may be attributed to incomplete theoretical development, measurement issues, methodological issues, or all three (Cook & Campbell, 1979). A recent meta-analysis of the correlates of involuntary 1 2 and voluntary absenteeism (Farrell & Stamm, 1988) analyzed effects sizes between psychological factors, demographic factors, work environment factors, and organization-wide factors: and, absence frequency and absence duration. With the exception of job involvement and absence history, Farrell and Stamm (1988) found correlations each with a magnitude less than .20. Alternatively, each factor explained less than four percent of the variance in employee absenteeism. Thus, despite expectations in the literature that attitudes and absence behavior are strongly related, Farrell and Stamm (1988) concluded that these relationships are weak. In addition, despite theoretical expectations that job satisfaction and voluntary absenteeism are strongly related (e.g., Vroom, 1964), the results of empirical research suggest that the relationship between job satisfaction and voluntary absenteeism was weak. Again, less than four percent of the variance in absenteeism was explained by job satisfaction. This conclusion is based on meta-analyses by Hackett and Guion, 1985; and Scott and Taylor, 1985. W One potential reason for weak relationships between absenteeism and work-related attitudes is a failure to consider whether the relationship may be mediated by a cognitive antecedent (Fichman, 1988; Hackett & Guion, 1985). Recently, Fichman (1984: 1988) argued that absence is a 3 function of two motivation processes that include: (a) the reward value of attending work which decreases over time, and (b) the attractiveness of non-work alternatives which increases in strength with no compensating consummatory activity. In other words, an individual chooses to switch between attending work and being absent from work when the perceived utility of one behavior exceeds the other. Although these forces are presumed to be cognitively determined, Fichman (1988) did not empirically examine whether a cognitive factor mediates the relationship between attitudes and behavior. Consistent with Fichman's (1988) assumptions, Hackett and Guion (1985) also assume that voluntary absenteeism implies volition. Thus, an individual makes a conscious choice whether to attend work on any given day. Hackett and Guion (1985) suggest that one’s decision to be absent from work may result from consideration of the costs and benefits of missing work on any given day. In summary, Fichman's (1988) work is characteristic of empirical research in absenteeism. The focus is on predicting absence behavior from work-related attitudes rather than examining whether absence behavior is preceded by cognitions about the behavior. Hackett and Guion’s (1985) suggestions require research designs that facilitate assessment of expectancy and valence for a variety of potential outcomes of missing work. 4 MeasurementZMethodological Issues In addition to theoretical issues which may be associated with weak effects, measurement and methodological issues may also be responsible for weak effects. These issues include: (a) low measurement reliability; and, (b) non-normal sampling distributions. Hackett and Guion (1985) found that the levels of reliability for voluntary and involuntary absence measures have been below conventional levels in the behavioral sciences (Nunnally, 1978). Measurement unreliability can attenuate correlation coefficients (Cook & Campbell, 1979). This is also associated with decreased statistical power (Cohen, 1977). Also, the use of least squares analytic methods are problematic because absence events tend to be skewed. In other words, most people in a given sample have low absenteeism rates, whereas only a few are absent often; many are not absent. This may result in attenuated correlations or weak effects because least squares methods are based on the assumption that the marginal distribution of behavior is approximately normal. Consequently, researchers have possibly made false substantive conclusions because they relied on correlations with magnitudes that are possibly smaller because of measurement artifacts. fiummsrx The discussion presented three potential reasons for weak effects in employee absenteeism research. The first reason, a theoretical one, suggests that researchers have failed to consider whether a cognitive factor mediates the relationship between work-related attitudes and absence behavior. Attitudes and absence behavior may be related only indirectly. The second reason, a measurement one, suggests that weak effects may be the result of low measurement reliability. The third reason, a methodological one, suggests that weak effects may be the result of the violation of a key assumption underlying ordinary least squares regression techniques, namely, that the marginal distribution of the variables is normal in shape. It is possible that weak effects may be attributed to any number of these reasons. Research to date has not addressed all these issues in any one investigation of employee absenteeism. The Extent of Advance in Absenteeism Research Although researchers have recognized the importance of studying employee absenteeism, they have generally ignored the theoretical and methodological issues highlighted above. A computerized search of Psychological Abstracts spanning from 1982-1986 shows that the number of doctoral dissertations and published studies examining employee absenteeism indexed each year has remained relatively 6 steady, totalling 162 (i.e., 79 doctoral dissertations and 83 published studies) over this period. However, a detailed examination of those studies' abstracts reveals that little advance in the study of employee absenteeism has been made with respect to addressing the theoretical and methodological issues. Only three papers proposed new direction in absenteeism theory (Fichman, 1984, 1988: Johns & Nicholson, 1982; Nicholson & Johns, 1985). Only six studies directly addressed some of the measurement and methodological problems. Two of these papers examined the construct validity of voluntary versus involuntary absence indexes as well as addressed some of the measurement issues (Chadwick- Jones, Nicholson, & Brown, 1982; Hackett & Guion, 1985). Only two directly addressed methodological issues related to distributional problems associated with absence sampling distributions (Hammer & Landau, 1981: Watson, Driver, & Watson, 1985). One study examined the extent to which the effects of sampling error and measurement unreliability attenuated the work attitude-absenteeism correlation (Terborg, Lee, Smith, Davis, & Turbin, 1982). Finally, one study raised the issue of internal validity of the absenteeism research (Clegg, 1983). This assessment of the absenteeism literature suggests the need to consider different approaches to understanding voluntary employee absenteeism. Certainly, a call for 7 research that examines the plausibility of cognitive antecedents to absence behavior exists (Hackett & Guion, 1985). In addition, it is also important to assess whether weak effects are the result of measurement unreliability, and violations of key ordinary least squares assumptions because research has shown that failure to address these issues may undermine theoretical expectations (e.g., Hammer & Landau, 1981: Watson, et al., 1985). The Research Strategy Unlike the existing body of absenteeism research, this dissertation is on construct validity, rather than on the substantive validity issues. There are three objectives. The first objective is to investigate whether the problem of weak effects may be overcome by using a measure of absence intentions. This objective was stimulated by employee turnover research, which has suffered many of the same theoretical problems as absenteeism research. Impetus for the study of behavioral intentions has resulted, in part, from theoretical arguments that have singled them out as the most direct and immediate cognitive antecedents of overt behavior (Ajzen & Fishbein, 1980: Fishbein, 1967: Fishbein & Ajzen, 1975: Triandis, 1977). Many of the turnover models (e.g., Mobley, 1977; Steers & Mowday, 1981) now incorporate behavioral intentions in the turnover process. Steel and Ovalle (1984) noted that intentions to resign are universally regarded as the 8 culmination of one's decision to leave an organization. Turnover intentions represent an intermediate link between attitudes about the consequences of leaving the organization as well as subjective norms regarding leaving the organization and behavior. In a review and meta—analysis of research on the relationship between behavioral intentions and employee turnover, Steel and Ovalle (1984) found that the weighted correlation between behavioral intentions and employee turnover was .50. In addition, they found that intentions were more predictive of attrition than overall job satisfaction, satisfaction with the work itself, or organizational commitment. Despite the logical appeal of an intentions construct in employee turnover research and its contribution in explaining variance in turnover, absenteeism researchers have not studied whether employees may use a similar decision process that precedes voluntary absenteeism. After all, not being in the workplace involves choice when the reasons are within one's control. Under instances of control, it is unlikely that a person stays home from work without making some sort of conscious decision. Based on this rationale, a conceptual definition of absence intentions will be proposed. A nomological network (i.e., an exploratory conceptual framework relating constructs and observables to each other with specific _I L‘.» .- 3}. w. 'I ”*Cl-n In. f-it-s... 'lffio'if’ .4 1 9 hypotheses) will also be specified to assess the degree of construct validity of an absence intentions measure. Construct validation requires answering two questions: (a) How well does the instrument measure the individual trait or individual characteristic of interest? In this case, it is absence intentions; (b) Does the absence intentions measure relate to other factors as suggested by theory and empirical research? The second objective of this dissertation is to examine whether measurement unreliability is an alternative explanation for weak effects. As discussed, meta-analytic estimates of correlations between work-related attitudes (i.e., job satisfaction, organizational commitment) and voluntary absenteeism suggest that less than four percent of the variance in voluntary absenteeism has been accounted for by these factors. This percentage of variance is consistent with Cohen's (1977) convention for "small" effect sizes. Specifically, using Cohen’s (1977) standards for "small," "medium," and "large" effect sizes, do effect sizes increase beyond "small" when corrections for measurement unreliability are made? The third objective of this dissertation is to examine whether violation of least squares assumptions is an alternative explanation for weak effects. This will be achieved by examining the distributions of the variables under study and using the appropriate procedures when 10 nonnormality is salient. Then, the question will be asked, do effect sizes increase beyond "small" when corrections for measurement unreliability are made? Key Assumptions and Limitations The dissertation assumes that voluntary absenteeism is the result of an individual's conscious choice. Recent thought (Johns and Nicholson, 1982) suggests that voluntary absenteeism is the result of habit or proneness. Results of empirical research that operationalized habit as significant predictions of absenteeism behavior for one period based on a previous period (Breaugh, 1981: Garrison & Muchinsky, 1977) have been mixed. The nomological network includes widely studied constructs in employee absenteeism research. These include job satisfaction, organizational commitment, job involvement, and protestant work ethic. An assumption is made that measures of each construct is valid. As will be discussed, one’s attitudes toward absenteeism is grounded in expectancy theory. Expectancy theory has inherent limitations related to understanding the nature of behavioral choice. Simon (1956) criticized expectancy- based, behavioral choice paradigms, indicating that limited human cognitive resources do not allow one to rationally maximize outcomes. An individual’s limited cognitive resources and lack of access to all pertinent information results in an individual making "satisficing" choices 11 (i.e., less than optimal) rather than choices that are rational and maximize outcomes (Simon, 1958). Contributions This dissertation will assess the degree of construct validity of an absence intentions measure which is expected to provide an important mediating link between attitudes and voluntary absence behavior. However, failure to show that absence intentions is an important mediating link (assuming that plausible alternative explanations can be ruled out) would also be useful to researchers in industrial psychology and human resource management. Failure to identify absence intentions would suggest the need to pursue research that addresses absenteeism as a latent trait (Johns & Nicholson, 1982). Practitioners may also benefit from the results of this dissertation. To the extent that absence intentions is an important mediating factor, then employers will be able to understand absenteeism propensity within their organizations. While absence behavior data provide a post hoc view of what has happened, absence intentions may be useful for predicting behavior. Furthermore, the use of absence intentions will provide practitioners with greater understanding into the psychological correlates of behavior. By predicting when absenteeism is likely to occur rather than analyzing when it has occurred, employers will be able to develop proactive absence management programs that are yr '4'"! - Ivv‘q- - o ercfit‘."" I 'J..o~ "ii. - :Y‘_9D.'I fl.‘-v‘~“ '7“ tr‘ Wit. . l" .51 i: ~' ' no" _=u'.‘*' ‘”" ' '. J- v.37.- _,,.,1 - . J." C‘ .. F" 12 more responsive to employee and employer needs than before. Specifically, Ajzen and Fishbein’s (1980) theory suggests that behavioral change is ultimately the result of changes in beliefs. Thus, in order to influence behavior, management should provide information (i.e., absenteeism policies) that will produce changes in their beliefs. If absence intentions are identified, then it will be possible to determine the degree to which it is under attitudinal or normative control. This information will help management target information. This research is not designed to assess the meaning and interpretation of absenteeism to each individual. However, to the extent that factors related to absence intentions can be identified, it will provide a basis to identify differences in factors underlying absence motivation rather than simply mean differences in absence motivation. Outline of This Dissertation This dissertation contains six chapters. Chapter one, thus far, has generally described the rationale underlying the research objectives and the potential contributions of this dissertation. Chapter two includes a literature review of the employee absenteeism literature, focusing on theoretical, measurement, and methodological issues. The purpose of this discussion is to highlight the need to adopt alternative ‘ways of thinking and empirically examining employee 13 absenteeism. Chapter three presents a nomological network to guide construct validation assessment of absence intentions. The network is grounded in Fishbein’s (1967) theory of behavioral intentions and also includes theoretically meaningful work and non-work factors. Chapter four discusses the method of investigation. This includes the organization under study, subjects for this dissertation, the data collection procedure, the Operationalization of variables, and method of data analysis for each hypothesis. Chapter five shows the results of the data analysis for each of the hypotheses. Descriptive and inferential statistics have also been included. Inferential statistics were used to assess whether a hypothesis is statistically significant at an a priori alpha level. Chapter six contains a summary of these findings followed by a implications for theory. Alternative explanations were presented. Implications for practice followed.’ Finally, directions for future research conclude this chapter. CHAPTER TWO LITERATURE REVIEW Introduction Chapter One suggested deficiencies in our approach to studying voluntary employee absenteeism. The purpose of this chapter is to review the relevant literature. Specifically, this chapter will review relevant theories of absenteeism, as well as critical measurement and methodological issues in absenteeism research. ABSENTEEISM: THEORY, MEASUREMENT AND METHODOLOGY Review of Theoretical Models Hill and Trist (1953) Hill and Trist (1953) conceptualized employee absenteeism as withdrawal from stressful work situations. Withdrawal can take the form of accidents, turnover, and unexcused absence and results from dissatisfaction and conflict in obligations between one's work and non-work domains. The employee-employer relationship determines how the conflict is expressed. Hill and Trist (1953) discussed three phases of the employee-employer relationship. The first phase occurs during the period of initial employment. During this period, individuals are not fully 14 15 aware of the workplace norms (i.e., how much absence is tolerated and what individuals can "get away with"). In order to avoid conflict with the employer, withdrawal takes the form of turnover. The second phase is a period during which time norms have been learned. Workers are still not completely satisfied with their work situation. Absence is highest during this period which is termed "differential transit." The third period reflects an employee’s adjustment. Over time, individuals who determined that there was a poor fit between themselves and the employment situation have left. Thus, there is a relatively low level of withdrawal among remaining employees. For organizations that prohibit unsanctioned absence, employees cope with the workplace situation by exhibiting "legitimate" absenteeism. Specifically, individuals complain of minor illness that is psychosomatic in origin to legitimize absenteeism (Hill & Trist, 1953). Hill and Trist (1953) assumed that all forms of withdrawal stem from dissatisfaction with the work situation. They ignored the possibility that absenteeism may reflect proactive behavior instrumental in meeting non- work needs or wants. Hill and Trist (1953) adopted a group unit of analysis, making the assumption that the "cause" of withdrawal is the same for each employee. 16 W Gibson (1966) posited a model of absence behavior based upon concepts of the need-oriented individual and the goal- oriented organization joined together by a multi-faceted psychological contract. Absence from work implies an inability, inappropriateness, or an unwillingness to work with few exception (e.g, illness, funeral). Individual-level factors that shape an employee’s decision to breach the employment contract include work identification, status, sex, age, length of service, whether one’s belief system is either cosmopolitan or local (Gouldner, 1958). Gibson (1966) directly addressed the relationship between each of these factors with voluntary absenteeism. In each case, voluntary absenteeism is assumed to be the outcome of an adjustment process. Gibson’s (1966) model does not explicitly take into consideration the role of choice in behavior. As with Hill and Trist’s (1953) model, Gibson (1966) assumed that absenteeism is the result of negative affect toward work and the employer. Gibson (1966) did not acknowledge the possibility that absenteeism may reflect a choice to pursue more desirable activities outside the workplace. c olson 1977 Nicholson (1977) proposed an individual-level theory of absenteeism. He conceptualized absence events in terms of their avoidability along a continuum labeled ’A-B 17 Continuum.’ On either end of the A-B Continuum are Type ’A' absences (i.e., unavoidable) and Type ’B’ absences (i.e., avoidable). Key to the A-B Continuum model is an individual’s motivational state. Nicholson (1977) defined motivation to be absent as "the degree to which an employee is dependent upon the regularities of organizational life." (246) Similar to previous conceptualizations, motivation mainly depends upon the way one’s needs are satisfied through the employment relationship rather than factors outside the workplace. Nicholson (1977) suggested four broad categories of variables that affect motivation: personality traits, orientation to work, job involvement, and the employment relationship. Unfortunately, Nicholson's (1977) thoughts about each category of variables were not well-developed. Steers and Rhodes (1978) Steers and Rhodes (1978) advanced a series of propositions suggesting that an individual’s attendance is directly influenced by attendance motivation and ability to attend. An underlying premise of their model is that an employee's motivation to attend work represents the primary influence on actual attendance provided that the employee has the ability to attend. They hypothesized that attendance motivation and voluntary absenteeism are inversely related. In addition, ability to attend is 18 positively related to involuntary absenteeism (Steers & Rhodes, 1978). Motivation to attend is a function of the job situation (i.e., job scope, job level, role stress, work group size, co-worker relations, and advancement opportunities) and key pressures to attend (i.e., economic and market conditions, incentive/reward systems, work group norms, personal work ethic, and organizational commitment). The job situation and employee values and expectations interactively determine job satisfaction. Then, job satisfaction combines with various other pressures to attend to determine attendance motivation. Ability to attend depends upon: (a) illness and accidents, (b) family responsibilities, and, (c) transportation problems. Steers and Rhodes (1978) explicitly applied the expectancy-based motivation concept to the conceptualization of absenteeism. However, they addressed absenteeism as a choice to attend based on work factors. They ignored additional factors outside the workplace that may be relevant to one’s choice 39; to be in the workplace when scheduled. Furthermore, they focussed on motivation to be in the workplace instead of motivation not to be in the workplace when scheduled. Subsequent empirical research based on the model (e.g., Watson, 1981) has resulted in disappointing results. Brooke 19 (1986) found several problems with the model and operationalization of variables that have likely resulted in these disappointing results. Other problems include: (a) lack of evidence demonstrating the construct validity of the measures, (b) imprecision with respect to the dependent variable in the model, (c) lack of theoretical basis for assuming causality among variables, and, (d) criterion deficiency (Brooke, 1986). Chadwick-Jones. Nicholson. §1Brown (1982) Chadwick-Jones, Nicholson, & Brown (1982) proposed a social exchange theory of absenteeism based on two assumptions: (a) the interdependency of members of work organizations, and (b) social exchange between employers and employees. Social exchange between employers and employees is developed in, or revealed by, a pattern of behavior in the work situation that includes absenteeism. Absenteeism does not result in punishment or other negative consequences provided it is within the accepted, ongoing pattern of social exchange. Johns and Nicholson (1982) Johns and Nicholson (1982) provided several counter- propositions to beliefs and assumptions about absenteeism as stimuli to theory development. As with Chadwick-Jones et al. (1982), they addressed the absence phenomenon as a social one rather than motivational at the individual level 20 of analysis. Proposition . Absence events are phenomenologically unique. Because generalized predictions about absence behavior have been unreliable, they believe that a more contingent approach is needed which considers how specific contextual conditions affect individual absence episodes. Proposition . Absence is not the result of conscious choice but is the result of non-volitional forces. Individual differences in absenteeism will be due to variations in personal "habit strength," which is a function of one's degree of compulsive attachment to work, and external constraints on perceived decision making freedom. Proposition . Absence is a temporal behavior and continually subject to dynamic change. They reason that absence involves time reallocation between work and non-work based on salient priorities. Therefore, changes in absence behavior should not be viewed as "unwanted error variance." Absence represents non-work behavior and is subject to major causal influences from variables that transcend the workplace. Proposition 4. The meaning of absenteeism is complex. Absenteeism does not always reflect deviant behavior or escape from oppressive work situations. For example, Rosse and Hulin (1985) and Staw and Oldham (1978) have provided evidence to suggest that absenteeism also serves as a maintenance function for individual adjustment. 21 Prgposition . Absence occurs throughout the workplace, thus, research should reflect this phenomena. For example, most absenteeism research has focused only on blue collar absenteeism (Steers & Rhodes, 1984), yet, Johns and Nicholson (1982) maintain that white collar workers and managers also need time off from work. By examining absence behavior across hierarchical levels in the organization, researchers may better understand the extent to which absence has common costs, causes, and controls at different levels of the organization. Pighpan (1984.19881 Fichman (1984, 1988) proposed a dynamic model of attendance to help answer how long an individual will continue to attend work without an absence interruption. Unlike other theories of absenteeism, Fichman (1984) proposed that absenteeism could be more efficiently predicted based on the timing of the switch from absence to attendance. Understanding when absenteeism would most likely occur depends upon both the reward value of work activities and the reward value of non-work activities foregone while engaged in work activities. Fichman (1984; 1988) suggests that the dynamic component of attendance is captured in two motivational processes that operate simultaneously. The first motivational process is that the reward value of attending work gradually decreases to some asymptotic level. The 22 second motivational process is that the attractiveness of non-work alternatives increases in strength with no compensating consummatory activity. Consequently, at some point, the attractiveness of non-work activity will be greater than work activity, in which case an absence will occur. The dynamics of attendance is captured by the hazard rate h(t) of being absent from work. Fichman (1988) defined the hazard rate as the instantaneous rate of going from work attendance to work absence in a unit of time given that the individual has been in attendance until time t. The attendance hazard model was tested using three absence types for a sample of 465 union coal miners (Fichman, 1988). Voluntary absences were defined as absences categorized as discretionary, or contract days, discretionary holidays, graduated vacation days, and miscellaneous paid absences. Semi-voluntary absences included excused and unexcused unpaid absences. Involuntary absences included absences due to on-the-job injuries, illnesses, wildcat strikes, and off-the-job injuries. Fichman (1988) found support for the hazard model of attendance, suggesting that people will switch from one activity to another with some regularity. In addition, he also found that the hazard rate for an individual to end an absence is a function of the time since the last absence. 23 finnmsrx Early models of employee absenteeism specified absenteeism as the result need deprivation. Absenteeism was considered only from the vantage of the employer. With rare exception, voluntary absenteeism was negatively connotated. The world was defined in terms of the workplace. More recent models specified absenteeism as the result of an adjustment process. Still, the behavior was looked upon negatively by employers. Again, the world was defined in terms of the workplace. Steers and Rhodes (1978) introduced the expectancy- based notion of motivation as key to understanding voluntary absenteeism. Like other models, it addressed motivation as attraction to the workplace, and the lack thereof. The direction of causality was assumed to be uni-directional. Furthermore, non-work factors were not included as part of one’s choice to attend or to be absent. Only work-related factors were believed to affect levels of attendance. Finally, Fichman (1988) provided evidence to indicate that people may have cognitions about absenteeism and attendance. His work also suggests that absenteeism may reflect a choice to pursue activities with greater subjective utility outside the workplace. 24 Review of Measurement Issues W Gaudet (1963) reported that at least 41 different measures of absenteeism have been used in the literature. Despite the quantity of absenteeism measures, researchers (e.g., Chadwick-Jones, Brown, Nicholson, & Sheppard, 1971; Fox & Scott, 1943: Gibson, 1966: Hammer, Landau, & Stern, 1981; Steers & Rhodes, 1978) acknowledge two types of absenteeism: "involuntary" and "voluntary." Involuntary absenteeism is hypothesized as due to factors beyond the control of the employee. Voluntary absenteeism reflects conscious choice by employees for reasons within an individual’s control. One widely used purported measure of involuntary absenteeism is the Time Lost Index. Two widely used purported measures of voluntary absenteeism are Frequency Index and the Attitudinal Index. Time Lost Absence Measures. Chadwick-Jones et al., (1971) defined Time Lost as the number of days lost over a specified period for any reason other than organization- sanctioned leave. Several researchers (Chadwick-Jones et al., 1982; Fox & Scott, 1943; Yolles, Carone, & Krinsky, 1975) believe that involuntary absences are caused by serious illnesses that result in longer-term absences, that is, illness that removes the potential of choice is unlikely to be brief or to occur frequently. Ezegpepgy of Absence Measures. The Frequency Index is 25 defined as the number of absences in a specified period, regardless of duration, excluding holidays and workdays (Chadwick-Jones, et al., 1971). Fox and Scott (1943) reasoned that voluntary absences are reflected in the frequency of absence in contrast to the duration of absence spells because voluntary absences are likely to be short in duration. In addition, Vroom (1964) discussed voluntary absence from a valence-instrumentality-expectancy perspective, thus treating absence as choice. Huse and Taylor (1962) defined the Attitudinal Index as the number (i.e, frequency) of one-day absences. Froggatt (1970) and Taylor (1967) maintain that such short-term absence are voluntary in nature because absences based on the Attitudinal Index were not correlated with illness in excess of three days duration. Other purported measures of voluntary absence used in the absence literature include: (a) Lateness- the number of instances of lateness in any week: (b) Other Reasons- the number of days lost in a week for any reason other than holidays, rest days, and certified sickness (Chadwick-Jones et al., 1971): (c) Blue Monday Index- the number of individuals absent on Monday minus the number of individuals absent on Friday for any week; (d) Worst Day Index- the difference score between number of individuals absent on any week's ”best" and "worst" days (excluding holidays and rest days). The Blue Monday Index and Worst Day Index purport to 26 measure voluntary absence (Chadwick-Jones & et al., 1971). However, only Argyle, Gardner, and Cioffi (1958) provide support for the Worst Day Index as a measure of voluntary absence. Rgligbilitx The number of published absenteeism studies (workplace samples) ranges in the hundreds. The number of reliability coefficients for absenteeism measures, however, falls significantly short of the total number of studies. Hackett and Guion (1985) identified 27 reliability coefficients for the Frequency Index. This distribution had a mean of .51 and a standard deviation of .18. Twenty-nine reliability coefficients were found for the Time Lost Index. This distribution had a mean of .66 and a standard deviation of .28. Nineteen coefficients were found for the Attitudinal Index. This distribution had a mean of .41 and a standard deviation of .21. These mean reliability estimates are below conventional levels in the behavioral sciences (Nunnally, 1978). In addition, there also is a high degree of variability among the individual reliability estimates. Atkin and Goodman (1984) and Landy, Vasey and Smith (1984) suggested that the choice of time period on which absence measurement is based affects the level of reliability. For example, during relatively short periods, absences are more likely the result of accidents or illnesses associated with individuals 27 whose absence-taking behavior is in some way different from that of their peers. Atkin and Goodman (1984) also suggested that key aspects of absence policy (e.g., whether sick leave must be used within a fixed period or whether it can be cashed in or carried over into another period) and seasonal conditions (e.g., adverse weather) may also affect stability of absence measures. Among the reliability coefficients identified by Hackett and Guion (1985), different definitions of psychometric reliability were represented. The most common include test-retest reliability and, split-half reliability. The test-retest reliabilities estimate the measure’s stability over two periods. Split-half reliabilities reflect a measure’s internal consistency within one specified period. One implication of low reliability for absenteeism research is the problem of weak effects (Cook & Campbell, 1979), which may result in false substantive inferences because of failure to reach statistical significance. Weak effects will result in low statistical power (Cohen, 1977). Given that statistical power is a function of effect size, alpha level, and sample size, then the only immediate remedies would be to increase the probability of Type I error beyond the conventional .05 level (e.g., .10 or greater), or to increase sample size in order to achieve statistical significance. Although these adjustments may 28 enhance the probability of achieving statistical significance, they are unrelated to addressing the issue at a substantive level. Terborg et al. (1982) directly examined the issue of weak effects related to the job satisfaction and absence frequency relationship as well as the organizational commitment and absence frequency relationship. Data were collected on job satisfaction and organizational commitment from 242 sales employees in six retail stores. Evidence of differences were found on variable means, standard deviations, reliabilities, and predictive validities. When criterion unreliability, predictor unreliability, range restriction, sampling error, criterion contamination and deficiency: and differences in factor structures between different measures of similar constructs were controlled, little variability in validities remained. Their findings suggest that differences in these relationships may be the result of measurement and methodological deficiencies rather than theoretical deficiencies. Cppstppct Validity Construct validity refers to the extent to which a measure reflects the construct it purports to assess. Empirically, this is evidenced by both convergence and discriminability (Campbell & Fiske, 1959). Convergence refers to the extent to which different measures of similar constructs (or items purported to measure a particular 29 factor of a multi-factorial construct) correlate highly. Discriminability refers to the extent to which similar measures of differing constructs (or items purported to measure different factors of a multi-factorial construct) do not correlate highly. For example, assuming Time Lost does measure involuntary absence, then Time Lost should correlate more highly and positively with other known measures of involuntary absence than with known voluntary absence measures. Ideally, involuntary absence measures and voluntary absence measures should correlate zero if each measure possesses distinct factors. To date, two research efforts (Chadwick-Jones et al., 1982; Hackett & Guion, 1985) have examined the construct validity of the Attitudinal Index, Frequency Index, and Time Lost Index. Chadwick-Jones et al. (1982) assessed the extent of construct validity of several measures using the Multi-Trait-Multi-Method approach (Campbell & Fiske, 1959). Based on data collected over a one-year period for 21 distinct organizations, tentative support for the validity of the Attitudinal Index and Frequency Index as measures of voluntary absence, and, Time Lost Index as a measure of involuntary absence was found. Specifically, correlations between the Attitudinal Index and Frequency Index were more positively related than those between the Frequency Index and Time Lost Index. Caution, however, is warranted when 30 interpreting the correlation between the Frequency Index and Attitudinal Index because the former necessarily, by definition, includes the latter. This spurious overlap results in an "artificially" high correlation coefficient. Without a measure of internal consistency reliability for each absence measure along which to compare inter- measure correlations, it is difficult to assess whether any one of the measures is factorially complex (e.g., whether the Frequency Index taps into at least one other construct in addition to "voluntary" absenteeism or whether "voluntary" absence in itself is multi-factorial). In a second construct validity assessment, Hackett and Guion (1985) used exploratory factor analysis to determine whether a variety of absence measures reduced to a smaller number of sets tapping into distinct constructs (i.e., "voluntary" and "involuntary" absenteeism). An analysis was conducted separately for data sets (Chadwick-Jones et al., 1971; Huse & Taylor, 1962: Nicholson & Goodge, 1976) which contained absence data from a single sample, over two consecutive years. For each of the three data sets, Hackett & Guion (1985) found that both the Attitudinal Index and Frequency Index loaded highly on the same factor whereas the Time Lost Index loaded highly on a separate factor. Inter-factor correlations, however, were modest in magnitude for the Huse and Taylor (1962) and Chadwick-Jones et al. (1971) data 31 sets. This suggests that these constructs share some common factor variance. For the Nicholson and Goodge (1976) data, on the other hand, inter-factor correlations were near zero. Summsrx This review of measurement issues suggests that possibly false substantive inferences about the relationship between voluntary absenteeism and work-related attitudes may be the result of measurement unreliability. Despite the reliability issue, two recent investigations found tentative support for the involuntary-voluntary absenteeism distinction. These findings should be tempered with warnings by Hammer and Landau (1981). They suggested that researchers should not blindly accept the Frequency and Time lost as specific indexes of voluntary and involuntary absenteeism, respectively, because of contamination in raw data. One source of contamination is classification errors. Specifying the "cause" of an absence event is a difficult choice. A cause may depend upon any of the following factors: (a) the organization's absence policy, (b) the employee’s absence history, (c) the grievance history of the organization, (d) negotiations between employee and management, (e) political factors, and (f) rules used by the clerk who records the absence events (Atkin & Goodman, 1984). Conceptual ambiguity is likely minimized when the stated reasons for absenteeism are noted rather than 32 accepting blindly the Frequency and Time Lost Index (Hammer & Landau, 1981). Furthermore, use of an absence intentions measure may be useful in assessing the validity of reported classifications because one would not expect a cognition to be related to involuntary behavior. It is logical to expect that a cognition would be related to voluntary behavior as will be discussed in greater detail in Chapter Three. Review of Methodological Issues In addition to problematic measurement issues in absence research are methodological problems which have restricted our understanding of the absence phenomenon. These problems have also caused us to question the validity of some of our conclusions about absenteeism. One key methodological problem that may be associated with weak effects is the violation of statistical conclusion validity. Stapistical Conclusion Validity Statistical conclusion validity addresses the problem of drawing potentially incorrect conclusions about the relationship between two variables based on the violation of assumptions for probability distributions in statistical hypothesis testing (Cook & Campbell, 1979). Landy, et al. (1984) and Watson et al. (1985) noted that absenteeism is a low-base-rate event. This distributional property is because most people in a sample have low absenteeism rates, whereas only a few are absent often: many are not absent 33 (Hammer & Landau, 1981). Hammer and Landau (1981) identified three implications. When the marginal distribution of the absenteeism variable is nonnormal, some of its sample statistics (e.g., mean and variance) may differ significantly from the population values. Consequently, there is a loss of statistical power (Cohen, 1977). As discussed earlier, there also is a problem of weak effects. Second, as a result of low absence rates, recorded absences depart from the normal distribution, an assumption on which valid correlation and regression analyses are based (Hays, 1981). With rare exception (e.g., Fichman, 1988), most studies of absenteeism use correlation and regression models in hypothesis testing. With a skewed, truncated distribution, the value of the correlation coefficient is attenuated. The more disparate the marginal distributions are from one another, the lower is the maximum value of r (Hays, 1981). Again, the problem of weak effects is evident. Third, correlation and regression models assume linearity, normality in the marginal distributions, and homogeneity of variance in the conditional distributions of the dependent variable (i.e., homoscedasticity) around points of the independent variable(s) (Hays, 1981). Because skewness and leptokurtosis can result from hetero- scedasticity, a nonnormal sample distribution indicates that 34 the assumption of homoscedasticity is violated when correlation and regression models are used in data analyses (Cohen & Cohen, 1983). The implications for data analysis in absenteeism research are inaccurate estimates of the standard error of the regression weights, which can make significance tests meaningless or the prediction of negative absenteeism values that are meaningless (Hammer & Landau, 1981). §EEE§£1 The review of methodological issues suggests the need to examine the distributional properties of the absenteeism variable. Two alternate ways to manage nonnormality include transforming nonnormal variables to approximate normality, or removing influential outlier variables (Watson, et al., 1985). Because the former does not always result in normality, the latter approach may be necessary. A strategy for managing potential nonnormality will be discussed in Chapter Four. CHAPTER THREE RESEARCH MODEL AND HYPOTHESES Introduction The purpose of this chapter is to provide a framework to guide the assessment of the degree of construct validity of an absence intentions measure. This involves proposing a conceptual definition of absence intentions. In addition, construct validation requires specifying a nomological network (Cronbach & Meehl, 1955). The objective of construct validation is to develop a measure of an individual characteristic for its own sake as well as to determine how well an instrument measures this characteristic (Ghiselli, Campbell, & Zedeck, 1981). Construct validation requires answering the question how well does the instrument measure the individual trait or characteristic of interest? In this case, the construct is absence intentions. This nomological network will be grounded in Ajzen and Fishbein’s (Ajzen & Fishbein, 1980; Fishbein & Ajzen, 1975) theory of behavioral intentions. In addition, it will also include theoretically meaningful predictors that have been widely examined in employee absenteeism research. First, an overview of construct validity will be provided. Then, each 35 36 component of the nomological network will be presented. In line with Ghiselli et al. (1981), the specification of the nomological network will include: (a) identifying the meaning of the absence intentions construct, (b) how absence intentions is related to other constructs, and (c) how absence intentions is related to specific observable behaviors. Construct Validity: An Overview A construct represents a hypothesis that a variety of behaviors will correlate with one another in studies of individual differences (Nunnally, 1978). Construct validity is defined as representing the correspondence between a construct and the operational procedure to measure or manipulate that construct (Schwab, 1980). The process of assessing the validity of a construct is indirect, yet, it represents the link between psychometric properties of a measure to the theoretical essence underlying that measure. Schwab (1980) identified key issues related to construct validation. First, Schwab (1980) suggests that defining the construct is critical and should include: (a) relating the meaning of the construct; (b) specifying the level of analysis: (c) specifying whether the construct is unidimensional: and (d) indicating whether the construct is likely to be stable over time. Also critical to assessment of construct validity is the nomological network that Cronbach and Meehl (1955) have 37 defined as "the interlocking system of laws which constitute a theory." (p. 290) Specifically, Cronbach and Meehl (1955) indicate that the laws in a nomological network may relate (a) observable properties or quantities to each other: or (b) theoretical constructs to observables: or (c) different theoretical constructs to one another. The nomological network is useful in assessing the extent of convergence and discriminability (Campbell & Fiske, 1959) that are essential in ruling out alternative hypotheses. Convergence means that evidence from different sources gathered in different ways all indicates the same or similar meaning of the construct (Campbell & Fiske, 1959). Discriminability means that one can empirically differentiate the construct from other constructs that may be similar, and that one can identify what is unrelated to the construct (Campbell & Fiske, 1959). Specifying inter-construct linkages in a nomological network is also important for two reasons. Inter-construct linkages can provide clarification of the construct under consideration (Messick, 1975). Specification of inter- construct linkages can guide the analysis for assessing the degree of construct validity (Schwab, 1980). A NOMOLOGICAL NETWORK FOR.AN ABSENCE INTENTIONS CONSTRUCT The nomological network, which will be proposed to guide construct validation of the absence intention construct, includes absence intentions, voluntary absence 38 behavior, involuntary absence behavior, subjective norms regarding absenteeism, attitude toward absenteeism, job satisfaction, job involvement, organizational commitment, situational moderators, and personal characteristics. Table 1 shows the hypothesized relationships for each variable with absence intentions. These predictions are based on theory. The following discussion will detail the factors specified in Table 1 and the nature of the expected relationships. Absence Intentions An Introduction to Behavioral Intentions depyiew. Fishbein (1967) argued that a person’s attitude toward an object influences the overall pattern of his/her responses to the object. One's attitude, however, does not necessarily predict any given action. Instead, a single behavior is determined by the intentions to perform the behavior in question. Figure 1 outlines the Ajzen and Fishbein (1980) theory of behavioral intentions. The theory is based on the assumption that people use available information in a reasonable and rational way to arrive at a behavioral decision. Social behavior is pp; controlled by unconscious motives or overpowering desires. Thus, behavior is not capricious. Because simply predicting behavior from intentions does not provide much insight into the nature of human behavior, it is important to understand the determinants of behavioral intentions. 39 .aawamm Hms\mfls you Hmcumm mEOOCw humaflum may ma Hmsofl>fion cm umsumn3 ocm mmwuflaflnflmsommmu Qflsmax no mmunfluw>oo haco who .um>m3on .ommmmmmo mum3 xmm mmmoamfim ocm mom mmzoamfim m A+v Emflmmusmmnd annucnao> Amflsmcofluoamu ocv Emfimmucmmn< humussao>cH mmmmeMIAmmmMNmmmm A+v mmfiuflaflnwmcommmm mflcmcflm any umcumm maoosH xumfiflum oxmm mmhoamsm mmo< mmwoamam muouomm can an 0800 AIV oflnum xuo3 A+V Emfimmucmmnd uson< msuoz m>fluomnnzm any ucmauwfifiou Hmc0flpmuwcmmuc A+v amwmmucmmnd Ans :ofluomumflpmm non uson< mwosunuuc Any usefim>ao>cH moo mcofiusmucH mocmmnd mmmmmmmummcnvmwfiumq mmdnnwun> hosum sowunuummnwo H OHAUB 533.592 23.—...?» ”5.3.3...— .3..— £25 < 3. 5:2... 3:83: .0 38.; Refinish sea son? — 952% c.2338— 3:25:— Efloousamaa 31239.53." ESE—.9» ”553:8 38:23 2.3a 5.5: 9583.5 £3.2— 9:35.52 0 - ‘ x x 4 332.2. .53: 3.33: 32.—26> 2. 3 53:3:— Efieossomaa 568353.. 33.5.3 33.5.3 .6 mouse—.333 ...—26. 35:2 32.5 £3.3— 7 w s 41 Specifically, a person’s intention is in turn a function of his/her attitude toward performing the behavior and his/her subjective norm (Ajzen & Fishbein, 1980). One’s attitude toward performing the behavior is proposed to be a multiplicative function of the individual's beliefs about the consequences of performing the behavior or the object's attributes, and the individual's evaluations of those consequences. Mathematically, it is stated as follows: n AB = z bi*ei where i=1 bi is the belief that performing behavior B leads to consequence or outcome 1: ei is the person’s evaluation about whether the outcome i is good or bad: and, n is the number of beliefs the person holds about performing behavior B. The underlying theoretical base of the attitudinal component is expectancy theory (Vroom, 1964). hi is equivalent to the expectancy construct in expectancy theory. e1 is equivalent to the valence construct in expectancy theory. The second component of the theory deals with the influence of the social environment on behavior. The subjective norm is the person's perception that most people who are important to him/her think he/she should or should not perform the behavior in question. One’s subjective norm 42 regarding behavior is proposed to be a function of the person’s beliefs about what specific, important others think the person should do, weighted by the person’s motivation to comply with others. Mathematically, it is stated as follows: n SN= Z bi*mi where 1=1 b1 is the normative belief that individuals within the reference group of individual i thinks he/she should or should not perform the behavior B; ‘i is one’s general tendency to accept the directives of each individual in the given reference group or individual (i.e., motivation to comply); and, n is the number of relevant referents. Together, these two factors combine in an additive fashion to determine behavioral intentions I. Mathematically, the theory is stated as follows: I= (AB)w1 + (SN)w2 where V1 and wz are derived empirically. Fishbein (1967) indicated that these weights are expected to vary with the kind of behavior that is being predicted, with the conditions under which the behavior is to be performed, and with the person who is to perform the behavior. Atpitudinal and Behavioral Entities Ajzen and Fishbein (1977) maintain that attitudinal and behavioral entities may be viewed as consisting of four 43 different elements: (a) the action, (b) the target at which the action is directed, (c) the context in which the action is performed, and (d) the time at which it is performed. Three major factors have been found to influence the magnitude of the relationship between intentions and behavior (Ajzen & Fishbein, 1977; Fishbein & Ajzen, 1975): (a) the degree to which intentions and behavior correspond in their levels of specificity: (b) stability of the intentions, and (c) the degree to which carrying out the intention is completely under the person's volitional control. Intentions-Behavior Correspondence. The first factor requires that attitude and normative components in the nomological network be assessed for the same time, context, target and action in order to enhance correspondence between attitude, intentions, and absence behavior. The role of time has not been directly addressed in absenteeism theory. Time, however, has been an issue in measuring absenteeism behavior with respect to stability of the absence phenomenon (e.g., Landy, et al., 1984). This will be discussed shortly. Time may also be relevant to the extent that different variables or combinations of variables exert a differential influence on absenteeism as a function of stages in organizational socialization (Van Maanen & Schein, 1979). Clearly, the treatment of time in absence theory and measurement is a topic which requires additional research. 44 The context issue suggests the need to consider attitudinal and behavioral factors that consistently relate to a particular situation. In the case of absenteeism, job involvement, organizational commitment, job satisfaction, and protestant work ethic, for example, should be related to intentions to be absent from the work context. Spability pf Intentions. The longer the time interval between measurement of intentions and observation of behavior, the greater the probability that the individual may obtain new information or that certain events will occur which will change his/her intentions. Organizational Behavior researchers have examined attitudes such as job satisfaction, organizational commitment, and job involvement cross-sectionally. However, there is evidence to suggest that organizational commitment is relatively stable over time (Porter, Steers, Mowday, & Boulian, 1984). Although there is evidence for the transitory nature of satisfaction (Porter, et al. 1974: Smith, Kendall, & Hulin, 1969), Waters and Roach (1973) found that overall satisfaction and absence frequency were consistently related over a period of two years. Schneider and Dachler (1978) assessed the stability of the Job Descriptive Index (JDI) over a 16-month period for a diverse sample of 847 Atlantic coast utility employees. Results showed reasonable stability coefficients (average r= .57) and also indicated that the five satisfaction scales (i.e., work, pay, 45 promotion, supervision, co-workers) had retained their relative independence over time. More recently, Staw and Ross (1985) examined whether job satisfaction is consistent within individuals, showing stability both over time and across situations. A longitudinal analysis over five years for over 5,000 males between the ages 45 and 59 showed significant stability of attitudes over a five-year period. The analysis also indicated significant cross-situational consistency when individuals changed employers and/or occupations. Gerhart (1987) questioned the validity of the Staw and Ross (1985) findings for several reasons. First, the Staw and Ross (1985) sample contained men in a limited age cohort that is less likely to experience changes in the work situation. Second, given the high positive correlation (r= .84) between the component parts of both pay and occupational status in the Staw and Ross (1985) sample, the change scores used in their study may have been unreliable given a formula by Guilford (1954, p. 394). In summary, Gerhart (1987) concluded that measurement problems preclude accurate assessment of the predictive power of dispositional factors. There has been limited research on the stability of absence behavior. There is, however, some evidence for stability. For example, Breaugh (1981) and Keller (1983) found that previous absenteeism predicts future absenteeism. 46 Despite evidence for high inter-period correlations for absenteeism, Landy et al. (1984) suggest that stability of absenteeism behavior should be conceptualized in terms of the interaction of the period of aggregation, the type of index (e.g., duration versus frequency), administrative categorization of absence events, and base rate of absence behavior. Although empirical research is limited, it appears that aggregation periods of at least three months are more stable than shorter aggregation periods (Landy, et al., 1984). Control of Intentions. Fishbein and Ajzen (1975) suggested that the correlation between intentions and behavior may be attenuated due to the occurrence of events outside one’s control or a person’s habits. Herman (1973) demonstrated that when an employee is free of situation constraints in choosing among behavioral alternatives, attitudes predict behavior. Herman’s findings are consistent with theoretical foundations in absenteeism. For example, Gibson's (1966) theoretical analysis suggests that the relationship between job satisfaction and absenteeism depends upon the terms of the psychological contract between the employee and the employer. Specifically, if the cost of an absence either in pay or in the legitimation procedure is greater than the gain from the absence, absenteeism would not be a reasonable behavior alternative. In addition, Smith (1977) found that, 47 despite a crippling snowstorm, work-related attitudes (e.g., identification with the company) predicted attendance on the day after the storm when the company attendance policy was relaxed. For another office of the same company located in a city where there was no snowstorm, work attitudes did not predict attendance on the same day. The research examining whether absence may be the result of habit is less conclusive than for the impact of situational constraints. It has been hypothesized that absenteeism is the result of proneness based on prior absenteeism predicting future absenteeism (e.g., Breaugh, 1981: Keller, 1983). Though these results are statistically significant, the findings across samples are equivocal. Garrison and Muchinsky (1977) found that 17 percent of the work force accounted for 90 percent of paid time lost whereas 31 percent of the workforce accounted for 90 percent of unpaid time lost. Their findings suggest that investigators should consider characteristics of the sample, such as domestic responsibilities, that may account for this phenomenon. Thus, in order to assess whether absence proneness is a plausible explanation, phenomenological and historical analyses are needed. The goal would be to determine whether absence occurs among the few more often than expected and whether these occurrences are characteristic of a particular set of individuals (Johns & Nicholson, 1982). 48 Absence Intentions Defined Absence intentions is defined as the extent to which one does not expect to be in the workplace over a specified time interval because of voluntary choice. Absence intentions is proposed as a multi-factorial construct. As suggested in the literature (e.g., Chadwick- Jones, et al., 1982), there are distinct causal factors underlying reasons for absence other than illness or civic responsibilities (e.g., jury duty). The literature suggests three distinct categories that represent agents of voluntary absenteeism (i.e., absence for which one may have absence intentions). Paetor 1: Pain-Avoidance. An individual will choose to be absent to minimize exposure to an unpleasant situation in the workplace. Researchers believe that voluntary absenteeism is inversely related to job satisfaction (e.g., Brayfield & Crockett, 1956: Johns, 1978; Muchinsky, 1977: Vroom, 1964). A post hoc explanation for this finding is grounded in a hedonistic philosophy. Employees will withdraw or be absent from a work situation that is not satisfying. In addition, Nicholson, Brown, and Chadwick- Jones (1976) suggested that an inverse relationship between absenteeism and job satisfaction would be evident based on common sense rationale that employees who are happy with their jobs will attend work more regularly and permanently than dissatisfied employees (Argyle, 1972). 49 Pector 2: Non-york Attraction. An employee may choose to be absent from work because of attraction to desirable non-work alternatives. Only recently has voluntary absenteeism for reasons involving attraction to non-work alternatives been considered (e.g., Goodman & Atkin, 1984; Staw & Oldham, 1978). Little systematic empirical research, however, has been conducted. One exception is research by Morgan and Herman (1976), who found that future absence frequency would be instrumental to the following goals: having a break from the routine, participating in family functions, and enjoying leisure activities. Pactor 3: Family Reeponsibilities. An individual may choose to be absent from work because of family responsibilities. These include: caring for dependent children, and caring for other family members in need. Steers and Rhodes (1978) categorized absence due to family responsibility as resulting from ability to attend (i.e., involuntary absenteeism). While it is recognized that family responsibilities tend to be high priorities, it is assumed that such absences are voluntary. Essentially, absence of this type is based on a decision to adjust to responsibilities. For example, taking care of an ill child can be the result of a decision process to have both parents remain home, to have one parent remain home, to have a friend or relative care for the child while the parents attend work, or have day care watch over the child while the 50 parents attend work. For the purposes of this study, involuntary absences (i.e., absences for which an individual is unlikely to have intentions) are the result of civic duty (e.g., jury duty), personal illness, medical leave, funeral attendance, disciplinary leaves, and natural disaster. This conceptualization is consistent with that adopted by others (Hackett & Guion, 1985; Hammer & Landau, 1981). HYPOTHESIS 1: Absence intentions will be multi-factorial: Pain-avoidance, Non-work Attraction, and Family Responsibilities.1 Attitudes Toward Being Voluntarily Absent Overview. As discussed, Fishbein's (1967) attitude component is quite similar to expectancy theory. The basic tenet of expectancy theory (Vroom, 1964) is that an individual chooses to perform certain acts on the basis of the strength of a subjective belief that the act will be followed by a given outcome and on the attractiveness (i.e., valence) of the outcome. Expectancy theory has three independent variables. The first variable, expectancy, is the belief that a particular behavioral act will be followed by a particular outcome. The second variable, valence, is a measure of the 1 Pretest results indicate that the three scales correlate moderately. Thus, hypotheses will be specified for overall intentions. See Chapter Four for more details. 51 individual's feelings (positive or negative) about a particular outcome. More specifically, valence refers to the anticipated satisfaction associated with an outcome, and is distinguished from the value of the outcome (Lewin, 1938). The third variable, instrumentality, is the relationship between outcomes, specifically, the degree to which the person sees the outcome in question as leading to the attainment of other outcomes. An individual’s motivation is represented by two mathematical equations. Valence is a monotonically increasing function of the algebraic sum of the products of the valences of all other outcomes and the person’s conceptions of the instrumentality of a specific outcome for the attainment of other outcomes. n vi = fj * €31 (Vk * Ijk) where Vj equals the valence of the outcome j; Vk equals the valence of outcome k; and Ijk equals the cognized instrumentality of outcome j for the attainment of outcome k. An individual's motivation to exert effort is determined by the multiplicative interaction of expectancy times the perceived valence for the successful attainment of an outcome. The force on a person to act is a monotonically increasing function of the algebraic sum of the products of the valences of all outcomes and the strength of a person’s expectancies that the act will be followed by the attainment 52 of these outcomes. n Fi = fi * Z * (Eij * Vj) where F1 equals the force to perform act i: Eij equals the strength of the expectancy that act i will be followed by outcome j; and, V5 equals the valence of outcome j. Vroom’s (1964) formulation of expectancy theory is based on a within-subjects analysis. Much research to date using expectancy theory utilize a between-subjects analysis (Mitchell, 1974). Specifically, each subject was asked to provide three types of information. First, the subject is asked to indicate the degree to which one level of effort leads to good performance. Second, the subject indicates the degree to which good performance leads to each of several outcomes. Third, the subject is asked to indicate the attractiveness of each outcome. A between-subjects analysis does not lend itself well to analyzing whether an individual will exhibit any variance in effort, behavioral intentions, behavior, or action given the information available to an individual with respect to the organizational situation, the outcomes offered by the organization, the goals or outcomes sought by the individual, the expectancies the employee has for attainment of his or her goals, and the value, importance or attractiveness of the outcomes and goals (Zedeck, 1977). c an The and Absenteeism. To date, only three 53 absenteeism studies (Morgan & Herman, 1976; Rousseau, 1978; Youngblood, 1984) used an expectancy theory framework. While each of these studies explicitly incorporated work- and non-work-related factors as correlates of absenteeism, the studies were designed using a between—subjects rather than a within-subjects analysis. Consequently, little insight into choice was gained. Morgan and Herman (1976) found that the perceived achievement of off-the-job outcomes was a better predictor of absence frequency than perceived on-the-job consequences for 60 janitors and furnace operators in an automobile parts foundry. In addition, non—work activities often take precedence over work. This may be the case when costs associated with absenteeism are minimal. Furthermore, if work is a major source of involvement for the individual or the penalties for nonattendance great, absenteeism will be minimal. Rousseau (1978) correlated perceptions of work and non- work dimensions (based on Hackman and Oldham’s, 1975 Job Diagnostic Survey), and absence frequency. She found significant negative correlations for both work and non-work perceptions with frequency of absence for a sample of 97 product engineers, assemblers and quality control operators in an electronics organization, and 42 broadcasting technicians, reporters, and administrative personnel in a radio station. Rousseau (1978) also found that work and 54 non-work dimensions were significantly positively correlated, lending tentative support for the spillover hypothesis (Wilensky 1960). Finally, Youngblood (1984) tested the linkage between the relative attachments to work and non-work domains and absenteeism. Non-work attachment was measured as an employee's internal value of non-work time. Work attachment was measured using job-scope perceptions and satisfaction. Absenteeism was measured using frequency and duration episodes. Youngblood (1984) found that the value of non- work time was more strongly associated with absence duration, whereas work attachment was more consistently correlated with absence frequency for a sample of 406 full- time, unionized employees in a public utility. Sppjective Norms and the Absence Culture The absence culture notion was first discussed by Hill and Trist (1953) to explain how absences of different types are related to different phases in the employees' job tenure. As discussed in Chapter Two, as an employee learned the rules and norms regarding absence behavior, over time they ultimately learned to substitute unsanctioned for sanctioned absenteeism (viz., Hill & Trist, 1953). Comparative research (Chadwick-Jones, et al., 1982), however, has shown that the absenteeism profiles of individual organizations are distinctive. Thus, whether individuals substitute unsanctioned for sanctioned 55 absenteeism is organization-specific. Johns and Nicholson (1982) indicate that, even though there is no research that conclusively demonstrates the normative constraints on absence behavior, there are reasons to suggest why normative constraints may be operating. First, because absenteeism can be observed by individuals throughout an organization (e.g., coworkers, bosses), it is likely to be susceptible to potential normative sanctions. Second, research suggests that absenteeism may be associated with small group experience and organization (e.g., Mann, Indik, & Vroom, 1963; Mayo & Lombard, 1944). Nicholson and Johns (1985) believe that how culture is enacted is dictated by a psychological contract (Ofori- Dankwa, 1987; Schein, 1978) that emerges from interaction and communication. Schein (1978) defined the psychological contract as the set of unwritten reciprocal expectations between an individual employee and the organization. In other words, the psychological contract is a psychological mechanism by which collective influence is translated into individual behavior. Culture can affect absenteeism in several ways (Nicholson & Johns, 1985). First, culture may exert an effect on the level and patterning of absenteeism for a collectivity of workers through a shared understanding of how much absenteeism is tolerated. Second, culture may operate through social information processing mechanisms 56 (Salancik & Pfeffer, 1978). Specifically, employees may observe absence behavior and then adopt a pattern or level of absenteeism that reflects these observations. Finally, culture may operate indirectly to facilitate or constrain the extent to which individual level variables effectively influence absenteeism. An absence culture is shaped by the values and beliefs of the larger society as well as by the beliefs that are shared by virtue of membership in a particular organization or subunit (Nicholson & Johns, 1985). Regarding the former point, Steers and Rhodes (1984) reported that there are significant cross-national differences in time lost absenteeism ranging from one percent in Switzerland; to three percent in the United States; to 14 percent in Italy. Although these are aggregated data which do not rule out possible differences in satisfaction or motives, Nicholson and John’s (1985) theory suggests that cultural differences may account for the wide range in absenteeism across countries. In addition to the societal dimension of culture is the organizational dimension. Organizations or subunits within organizations may foster more or less distinctive beliefs about absenteeism, assumptions about employment, and conceptions of self-control. Less salient absence cultures invoke more subtle social influence and permit individual differences to have a greater impact on absence. Highly 57 salient absence cultures tend to be homogenous and often involve norms regarding attendance behavior. Based on this discussion about attitudes and subjective norms, the following hypothesis is suggested: HYPOTHESIS 2: Attitudes about absenteeism and subjective norms favoring absenteeism will independently predict absence intentions. In other words, individuals who hold positive attitudes about missing work (i.e., individuals who evaluate the outcomes of absenteeism as good) will be more likely to have absence intentions than will individuals who do not hold positive attitudes about missing work (i.e., individuals who evaluate the outcomes of absenteeism as bad). In addition, individuals who indicate that relevant individuals believe that absenteeism is acceptable will be more likely to have absence intentions than will people who do not indicate that relevant individuals believe that absenteeism is acceptable. Finally, technology and social ecology are critical determinants of cultural salience (Nicholson & Johns, 1985). Technology denotes the way work relations are structured to meet organizational goals whereas social ecology refers to the physical distribution of workers with various personal characteristics in the workplace. Nicholson and Johns (1985) maintain that blue collar technologies offer few incentives for employees to make their boss look good by showing up for work regularly. Blue collar workers often 58 are not provided with feedback about how their efforts contribute to organizational goals. For white collar workers, the reverse may be true because each successive level has a fair degree of control over the promotion opportunities and other privileges of lower levels. White collar workers are also more likely to interact with relatively higher management levels than blue collar workers. Thus, white collar workers may better understand how their efforts contribute to the overall organization. Gibson (1966) notes that the net effect may be a subjective norm among white collar workers that favors good attendance. Based on this discussion, the following hypothesis is suggested: HYPOTHESIS 3: Subjective norms favoring absenteeism will be greater for blue collar workers than for white collar workers. Absence Behavior In Chapter Two, involuntary and voluntary absenteeism were reviewed. Involuntary absenteeism was conceptualized as missing work, when scheduled, for reasons beyond one's control. Voluntary absenteeism, on the other hand, was conceptualized as missing work, when scheduled, for reasons yiphip one's control. In addition, construct validity assessments of measures of involuntary and voluntary absenteeism (viz., Chadwick-Jones, et al., 1982; Hackett & Guion, 1985) provided support for the involuntary and 59 voluntary distinction. Based on these findings and the discussion of the Ajzen and Fishbein (1980) model of behavioral intentions, the following hypotheses are suggested: HYPOTHESIS 4: Absence intentions will predict voluntary absenteeism. HYPOTHESIS 5: Absence intentions will not predict involuntary absenteeism. HYPOTHESIS 6: When the effect of absence intentions is controlled, attitudes favoring absenteeism and subjective norms favoring absenteeism will not predict voluntary absenteeism. Job Satisfaction Job satisfaction may be defined as a pleasurable or positive emotional state resulting from the appraisal of one's job or job experiences (Locke, 1983). Job satisfaction emphasizes the specific task environment where an employee performs his or her duties (Mowday, Steers, & Porter, 1979). Although daily events have been hypothesized as affecting the level of job satisfaction one experiences, reflecting more immediate reactions to specific and tangible aspects of the work environment (Mowday, et al., 1979), evidence for the stability of job satisfaction was noted (Staw & Ross, 1985). Waters and Roach (1971) reported significant negative correlations between absenteeism and overall job satisfaction, satisfaction with work, and satisfaction with coworkers. They found no relationship between absenteeism 60 and satisfaction with pay. Later, Waters and Roach (1973) replicated their study with the same female clerical sample. They failed to find a significant negative relationship between absenteeism and satisfaction. However, no significant relationship was found between absenteeism and satisfaction with pay. While Waters and Roach (1973) failed to replicate their earlier results, Nicholson et al. (1976) failed to find consistent relationships between job satisfaction and voluntary absenteeism within four samples from the same industry. They reported no systematic relationships between absenteeism and the job satisfaction facets of pay, promotions, supervision, and coworkers. Based upon their results, Nicholson et al. (1976) concluded "the common view of absence as a pain-reductive response on the part of the worker to his experience is naive, narrow, and empirically unsupportable." (735) Ilgen and Hollenback (1977) discussed two possible theoretical explanations for low, and often, mixed findings pertaining to the weak job satisfaction-voluntary absenteeism relationship. Both explanations assume that low job satisfaction is an antecedent of absenteeism behavior. The first explanation attributes the weak relationship to external constraints that influence absenteeism and do not allow the individual to behave in line with his/her satisfaction. The second adopts an expectancy theory 61 framework. Specifically, job satisfaction provides only one force to attend. The correlation between job satisfaction and absenteeism should be low because the decision to be absent from work is only partially related to one's job satisfaction. Despite the conflicting findings, meta-analysis of the job satisfaction-voluntary absenteeism relationship revealed an inverse relationship between these variables. Although the coefficients ranged in the .205, a significant relationship was revealed despite possible differences in sample characteristics, and situational characteristics which were not controlled. HYPOTHESIS 7: Job satisfaction will predict (inversely) voluntary absenteeism. HYPOTHESIS 8: The correlation between absence intentions and voluntary absenteeism will be greater than the correlation between job satisfaction and voluntary absenteeism. Sippational Moderators Although these meta-analyses yielded negative job satisfaction-voluntary absence relationships, the magnitude of the relationships were low. As noted, Ilgen and Hollenback (1977) discussed two theoretical explanations which suggest why the magnitude of observed correlations between absenteeism and job satisfaction may be low. First, low job satisfaction is an antecedent of absenteeism behavior. Advocates of this position argue that the low 62 observed correlation is due to external constraints which influence absenteeism and do not allow the individual to behave in line with his/her satisfaction (Herman, 1973). These external constraints operate as moderators of the job satisfaction-behavior relationship. Economic and job-market eonditions. Economic and job- market conditions often place constraints on an employee’s abilities to change jobs, thus, there may be increased pressure to maintain a good attendance record for fear of losing one’s job (Steers & Rhodes, 1978). Evidence suggests there is a close inverse relationship between changes in unemployment levels within a given geographical region and subsequent absence rates (Behrend, 1953; Crowther, 1957; Markham, 1985). An individual may not be aware of the objective indexes of unemployment rate. Also, how a particular unemployment index is interpreted (i.e., healthy economy versus unhealthy economy or abundance of job opportunities versus scarcity of job opportunities) may vary across individuals. Thus, one’s perception of the likelihood of attaining a similar job will be used. Based on this rationale, the following hypotheses are suggested. HYPOTHESIS 9: Perception of the likelihood of being able to attain a similar job in the area will predict voluntary absenteeism. 63 HYPOTHESIS 10: The correlation between absence intentions and voluntary absenteeism will be greater than the correlation between voluntary absenteeism and perception of the likelihood of being able to attain a similar job in the area. In addition to economic and job-market factors, organizational control policies are expected to affect voluntary absence behavior. Three categories of control policies have been examined with respect to absenteeism. These include positive-reinforcement programs, negative- reinforcement programs, and mixed—consequence-reinforcement programs. Positive-Reinforcement Programe. Schmitz and Heneman’s (1980) review of the literature on positive-reinforcement attendance control programs revealed their effectiveness in reducing absenteeism. Reinforcers found to be effective included bonuses (Grove, 1968: Hackman & Lawler, 1971; Orpen, 1978; Panyan & McGregor, 1976), bonus points (Baum & Menefee, 1979), participation in a lottery (Stephens & Burroughs, 1978), and food credit (Kent, Malott, & Greening, 1977). Negative-Reinforcement Programs. Unlike positive reinforcement programs, there have been far fewer studies using negative-reinforcement programs (Durand, 1985). Steers and Rhodes’ (1984) review of the literature suggests that negative-reinforcement programs can be effective in reducing absenteeism among chronically absent workers. 64 Mixed-Consegpence Programs. Nicholson (1976) used punishment on the absences of hourly workers in a food- processing factory. Progressive punishment for absenteeism in excess of 5 days over a 15-month period included verbal warnings, written warnings, and dismissal. Absenteeism was reduced. Kempen and Hall (1977) found significant decreases in absenteeism for 7,500 production workers in two plants after the intervention was implemented. Absenteeism did not decrease significantly for comparison groups during the same period. The intervention provided nonmonetary privileges to reinforce good and improving attendance along with progressive disciplinary warnings for excessive and worsening absenteeism. Kuzmits (1981) found a significant reduction in mean levels of absence occurrences, tardiness, part-day absences, and no-calls, but there were no significant changes in absence days or in absence days per occurrence. Under this system, employees received one point for each absence occurrence regardless of duration and partial points for tardiness and part-day absences. Also, employees who failed to report their absence by a designated time were given two points per incident. Harvey, Rogers, and Schultze (1983) found a 45.5 percent decrease in total sick leave used and a 55 percent reduction in total sick leave paid over the previous year. 65 Under this program, employees who were not absent for four weeks received a bonus of four hours' pay. Casual absence was penalized by nonpayment of the first eight hours of absence. In summary, although the literature suggests that positive-reinforcement programs are useful in reducing absenteeism, Steers and Rhodes' (1984) suggest several cautions when interpreting the results. First, in most studies, voluntary and involuntary absences were included in the baseline measure of absenteeism. Thus, the impact of the reinforcement programs would be understated. Second, most studies failed to use control groups to enhance ruling out alternative explanations. Finally, the time period in the post-intervention period was low, leaving unanswered whether these reinforcement programs might have a long term effect on absence behavior. Based on this discussion, the following hypothesis is suggested. HYPOTHESIS 11: The use of an absence control policy will moderate (i.e., lower) the relationship between job satisfaction and absence intentions. Pole Influences on Absenteeism The second theoretical rationale for an inverse job satisfaction-absenteeism relationship is also based on the assumption that job satisfaction is an antecedent of absence behavior (Ilgen & Hollenback, 1977). Advocates of this position argue that the low observed correlation is due to 66 other factors which create a force for attendance. Based on Graen's (1969) extension of Vroom's (1964) expectancy theory model, external and internal role pressures, under low situational constraints, are salient factors in one’s decision to attend. For example, Steers and Rhodes (1978) suggest that individuals with relatively high family responsibilities such as taking care of the children will more likely be absent. In this case, it is assumed that an individual does not have others to whom he/she can delegate such responsibilities (e.g., spouse). Thus, depending on whether kinship responsibilities are salient at a particular time, the following hypotheses are proposed. HYPOTHESIS 12: Kinship responsibilities will predict voluntary absenteeism. HYPOTHESIS 13: The correlation between absence intentions and voluntary absenteeism will be greater than the correlation between kinship responsibilities and voluntary absenteeism. er Invelvement as a Mediator In addition to rationale presented by Ilgen and Hollenback (1977), an alternative hypothesis that may also more accurately relate the job satisfaction-voluntary absenteeism relationship is one advanced by Cheloha and Farr (1980). They believe that the relationship between job satisfaction and employee absenteeism is not direct. Instead, they suggested that job involvement mediates the 67 absenteeism-job satisfaction relationship. Job involvement is the psychological importance of work to the individual. Lodahl and Kejner (1965) defined job involvement as “the degree to which a person's work performance affects his self-esteem." (p. 25) Job involvement, unlike job satisfaction, is not a reaction to one's job. Instead, it reflects one’s attachment to the work. The relationship between job involvement and voluntary absenteeism has been well-established. Patchen (1960) found significant negative involvement-absence correlations with full day absences for both engineers and steam plant personnel. Saal (1978) reported a significant, negative job involvement-absence correlation for 218 manufacturing employees. In a study that examined relationships between absenteeism and both job involvement and job satisfaction, Hackman and Lawler (1971) found the frequency of absence significantly related to job involvement for 298 telephone company employees. Breaugh (1981) found an inverse relationship between absence frequency and job involvement for a sample of 112 research scientists. Hammer et al. (1981) found that absence frequency and job involvement were inversely related for a sample mainly including production workers who had collectively bought their employing organization. For time lost and other involuntary measures of absence, results are evenly split between negative (Beehr 68 & Gupta, 1978; Blumberg, 1980; Cheloha & Farr, 1980; Saal, 1978) and nonsignificant relationships (Breaugh, 1981: Hammer, et al., 1981: Siegel & Ruh, 1973). Cheloha and Farr (1980) examined the relative influence of the Job Description Index (JDI) facets of job satisfaction measures and job involvement on absenteeism for a sample of state government employees. Zero-order correlations revealed that both job satisfaction and job involvement were inversely related to absence frequency, but job involvement was more consistently related. Partial correlational analysis for the satisfaction-absence relationship, holding job involvement constant, and involvement-absence relationship, holding job satisfaction constant, revealed that job involvement was related to absenteeism but job satisfaction was not. The authors analyzed data in Saal's (1978) study using partial correlation analysis and found similar results. It was suggested that the relationship between job satisfaction and absenteeism may be mediated by job involvement. HYPOTHESIS 14: Job satisfaction will not predict voluntary absenteeism when the effect of job involvement is controlled. Organizational Commitment Organizational commitment has been defined in terms of behaviors which represent sunk costs in an organization where an individual forgoes alternative courses of action 69 and chooses to link him/herself to the organization (e.g., Salancik, 1977). In addition to the behavioral conceptualization is one which addresses commitment as an attitude. March and Simon (1958) noted that organizational commitment encompasses an exchange relationship in which individuals attach themselves to the organization in return for certain rewards or payments from the organization. Studies examining commitment and employee absenteeism have adopted the following definition of the construct: (a) a strong belief in and acceptance of the organization's goals and values; (b) a willingness to exert considerable effort on behalf of the organization; and (c) a strong desire to maintain membership in the organization (Porter, et al., 1974). Highly committed persons are more likely to be motivated to attend in order to facilitate organizational goal attainment (Mowday, Porter, & Steers, 1982). This motivation may exist even if the employee does not enjoy the tasks required by the job. On the other hand, if an employee’s primary commitments are in other areas, such as family, home, a hobby, or sports, he or she would experience less internal pressure to attend (Morgan & Herman, 1976). Support for the former proposition was found (Hammer et al. 1981: Mowday, et al., 1979: Steers, 1977) where commitment and absenteeism were inversely related. Hammer et al. (1981) found an inverse relationship 70 between absence frequency and organizational commitment for a sample of 112 non-supervisory workers who had participated in a community "buyout" of their library furniture manufacturing organization. Mowday et al. (1979) and Steers (1977) found an inverse relationship between organizational commitment and absenteeism (measure not specified) for a sample of 569 clerical employees in public agencies, and for a sample of 119 scientist and engineers employed within an independent research laboratory. HYPOTHESIS 15: Organizational commitment will predict (inversely) voluntary absenteeism. HYPOTHESIS 16: The correlation between absence intentions and voluntary absenteeism will be greater than the correlation between organizational commitment and voluntary absenteeism. Personal Characteristics Age and Tenure Qualitative reviews (Porter 8 Steers, 1973; Rhodes 1983) of the literature found no consistent relationship between age and voluntary absenteeism across studies. A recent meta-analysis of the age-voluntary absenteeism relationship (Martocchio, 1987), however, suggests that age and frequency of absenteeism are inversely related. Despite the mixed findings across studies, the meta-analysis results are consistent with rationale (though it does not directly test posited factors) that voluntary absenteeism will be less frequent among older employees because they tend to 71 value and expect perquisites associated with their seniority. Also, older employees tend to be more committed to work likely because of family responsibilities (e.g., Baumgartel 8 Sobol, 1959; Cooper 8 Payne 1965; Nicholson, Brown, 8 Chadwick-Jones, 1977). HYPOTHESIS 17: Age will be inversely correlated with voluntary absenteeism. HYPOTHESIS 18: Organizational commitment will predict voluntary absenteeism when the effect of age is controlled. Family Size Currently, it is believed that female absenteeism increases with family size. Beatty and Beatty (1975) and Naylor and Vincent (1959) found support for this relationship. On the other hand, it is believed that male absenteeism decreases with family size. This relationship has been supported (Martin, 1971; Nicholson et al., 1977). The former conclusion was attributed to the belief that women traditionally assume responsibility for child care (e.g., Steers and Rhodes, 1978). The latter conclusion was based on the belief that males assume the primary income producer role for their families (Mowday, et al., 1982). Expectations about the family size-voluntary absenteeism relationship are based on gender-based stereotypes. sex A meta-analysis by Martocchio (1987) revealed a zero bi-variate relationship for females for the age and 72 voluntary absenteeism relationship and a significant inverse relationship for males. As with family size, gender-based stereotypes were relied upon to explain sex differences in voluntary absenteeism. However, in research which examined family size- and sex differences in voluntary absenteeism, neither a measure of kinship responsibilities nor income producing role was assessed. HYPOTHESIS 19: Kinship responsibilities (positively) and primary income producer role (inversely) will each be correlated with voluntary absenteeism when the effect of sex is controlled. ngk Ethic Research on the "work ethic" has shown substantial variance across employees in the extent to which they feel morally obligated to work (Blood, 1969: Rokeach, 1973). Empirical evidence indicates there is a positive relationship between work ethic and propensity to attend work (Feldman, 1974; Goodale, 1973; Ilgen 8 Hollenback, 1977: Searls, Braucht, 8 Miskimins, 1974). These studies concluded that attending work is an important aspect of life regardless of the nature of the job. HYPOTHESIS 20: Work ethic will predict (inversely) voluntary absenteeism. HYPOTHESIS 21: The correlation between absence intentions and voluntary absenteeism will be greater than the correlation between work ethic and voluntary absenteeism. 73 weak Effects: Theory-Based or Methods-Based? As discussed in Chapters One and Two, the question was raised whether weak effects are the result of theoretical deficiencies methodological deficiencies. The dissertation model includes hypotheses that address whether one's intentions to be absent are responsible for weak effects (i.e., a theoretical consideration). The following hypotheses address whether (a) measurement unreliability is responsible for weak effects, and, (b) whether violations of ordinary least squares distributional assumptions are responsible for weak effects. Measurement unreliability and least squares violations often result in attenuated correlations, which underestimate the value of the effect size which is dictated by theory. Consequently, inadequate statistical power (Cohen, 1977), in turn, will likely result in failure to achieve statistical significance. Very often, this leads researchers to conclude that weak effects are evident and question the validity of their theoretical assumptions. It is expected that, correction for unreliability and the use of procedures to minimize violations of ordinary least squares correlation/regression will result in effect sizes that exceed Cohen's (1977) definition of "small" effects for tests of the significance of a product moment correlation. It should be noted that these are not conventional 74 hypotheses designed to test whether an effect is statistically significant or differences between two effects are statistically significant. Instead, these are merely exploratory guides to assess whether methodological and measurement issues have an impact on effect size. Thus, the following "hypotheses" are suggested. HYPOTHESIS 22: The correlation coefficient squared for the job satisfaction - voluntary absenteeism correlation will exceed 4% after corrections are made for measurement unreliability, and adjustments made for distributional properties of the absenteeism variable. HYPOTHESIS 23: The correlation coefficient squared for the job involvement - voluntary absenteeism correlation will exceed 4% after corrections are made for measurement unreliability, and adjustments made for distributional properties of the absenteeism variable. HYPOTHESIS 24: The correlation coefficient squared for the organizational commitment - voluntary absenteeism correlation will exceed 4% after corrections are made for measurement unreliability, and adjustments made for distributional properties of the absenteeism variable. Summary An absence intentions construct was proposed as a sensitive link between attitudes and voluntary absence behavior. It was developed in response to both theoretical and methodological deficiencies characteristic of employee absenteeism research. In addition, a nomological network was proposed to guide the assessment of construct validity 75 of an absence intentions construct. The network was grounded in Ajzen and Fishbein's (1980) theory of behavioral intentions. CHAPTER FOUR METHOD Introduction This chapter discusses the organization under study, the sample of subjects for this dissertation, the data collection procedure, the Operationalization of variables discussed in Chapter Three, and method of data analysis. The Organization Under Study Overview The financial services subsidiary of a Fortune 500 corporation was secured for this study. Specifically, the auto auction business and the retailer financial services business provided access to its employees and its employee absenteeism data. The auto business operates 17 auto auctions nationwide (employees from only five of these locations nationwide participated). Across these locations, nearly 600,000 used cars and light trucks were sold by manufacturers, dealers, and leasing companies during 1987. Each auction secures large quantities of vehicles that they recondition for re- sale to auto dealers and auto leasing companies. The majority of employees at each auction are blue collar-unskilled employees. The next largest group is 76 77 comprised of white collar-clerical employees. Finally, small management teams direct the operation of each auction. The individuals employed within the selected auctions under study were not unionized. The retailer financial services business manages the credit card operation for several large retailers nationwide. This group is involved in all aspects of credit card management from establishing accounts to billing customers for their purchases. The majority of employees are white collar-clerical. There also is a management team whose members direct the operation of this business. The employees were not unionized. Absenteeism policy Absenteeism policies were similar across businesses. Each employee was allotted a specified number of sick days and personal days per calendar year based on length of service for which they were paid. Sick-day and personal-day allowances do not carry-over from one year to the next. Employees did not have to provide medical certification to justify paid sick days. Days absent beyond the allotted sick and personal days were unpaid (excluding specified holidays and scheduled vacation). 78 Power Analysis Approximately 200 employees per occupation group were needed to achieve statistical power equal to .80 (Cohen, 1977). This analysis was based on a test for the difference between two independent means, alpha equal to .05, one-tail. Cohen’s (1977) "small" effect size estimate was used as a conservative estimate because there currently are no empirical estimates of these effects. Although other statistical tests were conducted, a comparison of each one suggests that the test for the difference between two independent means requires a greater sample. Compared to the other tests, this test involves twice the amount of sampling error variance because each sample makes a unique contribution to total sampling error. Sample A total of 675 employees were surveyed. Among these, 635 individuals provided usable survey responses. This represents a response rate of 94%. The blue collar- unskilled sample consists of 275 employees from five auto auctions. The white collar-clerical sample consists of 360, 185 employees from the auto auctions, and 175 employees from the retailer financial service business. The following contains a discussion of the sample characteristics. The total sample (i.e., the 635 employees who provided completed surveys regardless whether they identified themselves) contains 368 females and 245 males. 79 The majority of blue collar employees (72%) are male. The majority of white collar employees (84%) are female. The mean age for the sample was 33.5 years (§Q=12.5). There was a significant age difference (p (586) = -1.98, p<.04) between the blue collar (M=32.3, §P=12.8) and white collar employees (M=34.4, §Q=12.2). Thirty-eight percent of the total sample completed high school. Half of the blue collar employees completed high school. Nearly half of the white collar employees had some college education, but no degree. White collar employees, on average, were more educated than blue collar employees (3K.2 (5, N=613) = 125.42, p< .001). Fifty-three percent of the sample were married at the time of survey administration. Forty-eight percent of the blue collar employees were married whereas 56% of the white collar employees were married. More white collar employees, on average, were married at the time of survey administration (x 2 (1, N=611) = 3.92, p_< .04). The average length of service in the present organization was 4.0 years (§Q= 5.1) for the total sample. For blue collar employees, the average length of service was 2.2 years (§Q=3.1). For white collar employees, the average length of service was 5.2 years (§P=5.9). White collar employees had a significantly greater length of service than blue collar employees (p (538) = 7.46, p< .001). As will be discussed below, the research design 80 required that attitudinal and demographic data be linked with absenteeism activity for the three-month period immediately following collection of those data. Of the total sample, 507 employees provided valid identification for this purpose. This includes 233 blue collar-unskilled employees and 273 white collar-clerical employees. The final sample consists of those employees who had not resigned before the end of the three-month period for whom absenteeism data for the period were available. This includes 440 employees, 176 of whom are blue collar- unskilled employees and 264 of whom are white collar- clerical employees. Table 2 shows the demographic profile of this sample. Data Collection Procedure The theoretical base for this dissertation required a longitudinal research design. Two separate waves of data collection were necessary. The first wave of data collection took place between October 28, 1988 and November 21, 1988. Specifically, this data collection effort involved collection of demographic and attitudinal factors using a structured questionnaire (Appendix A). The second wave of data collection took place during February, 1989. This data collection effort involved collection of individual records which contain activity of involuntary and voluntary absence behavior for the three-month period following collection of the survey data. 81 Table 2 Demographic Profile of Final Sample variable Blue Collar White Collar Total Sex: Female 53 219 272 Male 121 41 162 Age: Average years 32.6 34.8 33.9 Standard deviation 12.1 12.3 12.5 Race: Black 21/11.9% 87/33.0% 108/24.5% Caucasian 130/73.9% 155/58.7% 285/64.8% Hispanic 9/5.1% 10/3.8% 19/4.3% Asian 6/3.4% 5/1.9% 5/1.1% American Indian 1/0.6% 2/0.8% 8/1.8% Education (highest level completed): Some grade school 2/1.1% -------- 2/0.5% Some high school 37/21.0% 4/1.5% 41/9.3% High school 85/48.3% 75/28.4% 160/36.4% Some college 42/23.9% 129/48.9% 171/38.9% College degree 5/2.8% 55/20.8% 60/13.6% Organizational tenure: Average years 2.3 5.2 4.0 Standard deviation 3.3 5.9 5.1 N 176 264 440 82 Time 1 Several group meetings between the researcher and employees (20 to 50 persons per group) were held on site in the employee cafeteria. Before handing out the survey, employees were told that the purpose of the research was to help the researcher complete his thesis which will "look at" the relationships between work attitudes and absenteeism. Confidentiality of responses was promised. Only the researcher would have access to each person’s completed survey. Because the research design required that survey responses be linked with absenteeism activity, employee identification was necessary. Asking individuals to provide information about their absence attitudes and intentions may result in faking, one aspect of social desirability bias, because management policy explicitly discourages absenteeism. Thus, to have the sponsor require its employees participate in the study might threaten the validity of the results. An employee would likely feel uncomfortable about providing candid responses if they perceived they were forced to give information that could ultimately threaten their sense of job security. To minimize the chance that employees would provide fake responses to questions, participation in this study was on a yglgppery basis. The following informed consent policy was implemented and discussed before survey administration. 83 It contains three options. OppIon 1. Individuals were told they could remain in the room during survey administration, but were not required to complete the survey. At the end of the survey session, they would be able to leave without identifying themselves. thion 2. Individuals were told they could complete the survey, but were not required to identify themselves. At the end of survey administration, they would hand in a completed, anonymous survey. Option 3. Individuals were told they could complete the survey and provide identification on the form so that their survey responses could be linked up with their absenteeism activity. The informed consent policy resulted in the following response-types. Six percent (N=40) of the 675 employees chose not to identify themselves and not to complete the survey (option 1). Nineteen percent (N=128) of the employees chose not to identify themselves, but chose to complete the survey (option 2). Seventy-five percent (N=507) of the employees chose to provide valid identification and complete the survey (option 3). Individuals, regardless whether they identified themselves, indicated that they would not likely be absent. Specifically, individuals who identified themselves (M=4.9, §Qs1.3) indicated that they would less likely be absent than those who did not identify themselves (M=5.2, §Q=1.2). 84 Although this mean was statistically significant (p (632) = -3.0, p<.001), an analysis of the variance for each group’s response patterns indicated that response tendencies were not significantly different (P (1, 632) = 1.24, pe). This finding, in conjunction with the survey administration procedure, suggests that social desirability is not a likely explanation for these findings because those who chose not to identify themselves did not place their jobs in jeopardy. Thus, one would expect their responses to be frank. To further assess whether the two groups were similar, a comparison of demographic characteristics was made. There were no differences in sex composition (jK'Z (l, N=615) = 2.62, pe). There were no age differences (p (587) = -.29, pe). There were no education differences (LXLZ (5, N=614) = 6.00, pe). There were no marital status differences (DC 2 (1, N=612) s .80, pe). There were no differences in length of service between the groups (p (590) = -1.12, pe). The researcher actively participated in survey administration by pacing the completion of the survey. Specifically, for each section, he read the instructions to increase the chance that participants understood the purpose of each section. He also was available to answer any questions or respond to comments participants had while completing the survey. 85 11199—2. At the end of the three-month period following survey administration, the organization provided the following absenteeism data for each employee who identified themselves on the survey. These included whether the employee was absent for each work day of the specified period. If absent, the reason provided by the absence monitoring system was also collected. Because reasons for absenteeism were specified in the recording systems, absence frequency and time lost were calculated for both involuntary absenteeism and voluntary absenteeism. Operationalization of variables Table 3 shows the inter-correlations between each pair of variables for the final sample who identified themselves on the survey and for whom absenteeism data were available (for the three-month period following survey administration). 86 .So.vm .2. m M 689m .3. B 3. up “move .2. 3 mo. um .mmawzwfia .828 3238 .838 3340 .883 gnaw gong £5028 gflusgoo gm uguuflo >Hw>§ no: 8 8.38.39, honflm fin; 88883 @0988 gown—Lg mat/32H.” gflwgoo an.» «o 69.5% 055 .888. 8888 so 62 $868.: 38> .82 85 cm 388 $338., 38 .888 38.583 l 8 3.. 3: 3n 3.. 3: on: no: a? 8: mo 8 mo 8 0538868 .3 I: so 3 3T 3.. S. a? 3 3.. R: we as 3 am «3 .3 In no mo 8 no: 8 8 no: «o B 3 mo 3 88 £8 .2 u... 2 3 mo: 8.. 8: mo 8 8 8: 8n 8 mg .8 ul mm mo 2 mo ma mo 8: mo: «on 3.. 83 9:849 .S In mo 2 8 ma 2 no: 8 «on «T 858.8842, .3 ..l we no: 8 as 8: mo: 8 S: 82 8:495 .m ...l 8. 3 mm 8.. Nos 8.. 3.. 8.858884% .8 I. 3 mm B: 3.. m? a? 88 m>8n8 .n I: m... «on in 3.. mm: 8A8... mung .m In «T 8.. mm: am: 838.15 88mg .m In an as on 08 x83 .4 In mm mm 88% .95 .m In mm 838838 son . m ..l 8385 has A flvafifidoammswmemmu wing 33858888858883 mOHnNH 87 Table 4 shows the means, standard deviations, and internal consistency reliability estimates for the appropriate scales based on the final sample. Table 4 Internal Consistency Reliability Estimates Standard Coefficient Variable Heana Deviation Alpha Job involvement 4.4 .84 .85 Job satisfaction 4.8 .96 .90 Organizational commitment 4.4 .83 .80 Work ethic 5.3 .98 .74b Absence intentions 2.7 1.17 .78 Absence attitudes —1.4C 22.71 .78 Subjective norms 1.0 3.08 .56 Involuntary- frequency .9 1.11 --- Involuntary- time lost 1.2 1.89 —-- Voluntary- frequency .6 1.09 --- Voluntary- time lost .7 1.37 --— Kinship duties .7 .61 .72 5 Job involvement through subjective norms based on 7-point scale: 1=low on measure; 7=high on measure; absenteeism measured in number of workdays; kinship based on composite score of four items discussed below. b Corrected using the Spearman-Brown formula. c Negative value indicates attitudes, on average, that do not favor absenteeism. 427$N$440. 88 o vo vement Job involvement is the psychological importance of work to an individual. Specifically, job involvement is the degree to which a person's work performance affects his/her self-esteem (Lodahl & Kejner, 1965). Job involvement was measured with Lodahl and Kejner’s (1965) 20-item job involvement scale. The job involvement scale contains a seven-point likert scale ranging from strongly disagree to strongly agree. This instrument has demonstrated adequate psychometric properties with internal consistency reliability estimates (i.e., coefficient alpha) ranging from .72 to .89 (Lodahl & Kejner, 1965). Based on all completed surveys (N=635), coefficient alpha was .85 for this study. Job §atisfaction Job satisfaction is defined as a pleasurable or positive emotional state resulting from the appraisal of one’s job or job experiences (Locke, 1983). Job satisfaction was measured with the short form of the Minnesota Satisfaction Questionnaire (MSQ) (Weiss, Dawis, England & Lofquist, 1967). The MSQ contains items with two sub-scales: intrinsic satisfaction (12 items) and extrinsic job satisfaction (8 items). The MSQ contains a seven-point likert scale that ranges from very dissatisfied to very satisfied. The manual for the MSQ reports median internal consistency reliability coefficients of .86 for intrinsic satisfaction and .90 for extrinsic satisfaction. The 89 extrinsic and intrinsic sub-scales correlate .60 (Weiss et al., 1967). Thus, total score can be used as a general satisfaction index. Based on all completed surveys, coefficient alpha was .91 for this study. Organizational Commitment Organizational commitment is defined as (a) a strong belief and acceptance of the organization’s goals and values; (b) a willingness to exert considerable effort on behalf of the organization; and (c) a strong desire to maintain membership in the organization (Mowday, et al., 1979). Organizational commitment was measured with the Mowday et al., (1979) Organizational Commitment Questionnaire (OCQ). The OCQ contains 15 items. Each question has a seven-point likert scale ranging from strongly disagree to strongly agree. The OCQ has demonstrated adequate reliabilities with internal consistency reliability ranging from .82 to .93 (Mowday, et al., 1979). Based on all completed surveys, coefficient alpha was .86 for this study. The following three scales were designed to measure the components of Ajzen and Fishbein's (1980) theory of behavioral intentions. Because no scales were available, initial items were developed based on my review of the absence literature. The scales were pretested on a group of employees within the sponsoring organization. The objectives included assessing the factor structure of an 90 absence intentions scale, and, having the participants critique the content of the scales. Specifically, participants were asked to indicate whether any of the items were irrelevant. Also, participants were asked to add items that were relevant. W Based on the framework specified in Chapter Three, a set of eight intentions items was developed. Pretest results indicate that the three hypothesized factors (i.e., pain-avoidance, non-work attraction, and family responsibilities) were not supported with factor analysis. Pretest results indicate that corrected item-total correlations ranged between .39 and .62. Respondents are asked to rate whether they'd likely be absent from work during the next three months for each stated reason. All statements are qualified with the phrase "absenteeism excluding holidays, vacation, illness, and civic duties." A seven point likert scale ranging from highly likely to highly unlikely was used. Coefficient alpha was .78 for these items based on the total sample. All item pairs correlated significantly (p<.05). Attitudes About Absenteeism To assess these attitudes, participants were asked to provide (a) beliefs about whether a potential outcome of being absent would likely occur, and (b) an evaluation of 91 these outcomes (i.e., good - bad). The former was assessed using a seven point likert scale ranging from highly likely to highly unlikely. The latter was assessed using a seven point likert scale ranging from good to bad. Fifteen potential outcomes of missing work formed the basis for assessing attitudes. As specified by Ajzen and Fishbein (1980), each belief score was multiplied by its respective evaluation. The attitude score was the sum of the 15 products. Subjectiv§,Norm§ To assess subjective norms, participants were asked to indicate (a) their beliefs about what specific important individuals think he/she should do regarding absenteeism, and (b) the extent to which he/she listens to each person regarding work-related issues. The former was assessed using a seven point likert scale ranging from strongly agree to strongly disagree. The latter was assessed using a seven point likert scale ranging from never to always. Six relevant individuals formed the basis for assessing subjective norms. As specified by Ajzen and Fishbein (1980), each belief score was multiplied by the respective motivation to comply score. The subjective norm score was the sum of the six products. 893 Employee age was based on one's self-report. Specifically, how old were you (years) on your last 92 birthday? Organizational Tenure Organizational tenure, in months, was based on one’s self-report. Sex Sex was based on self-report. Kinship Responsibilities This was indexed by a composite score based on the number of children under age 7 living at home, and the number of individuals (excluding children) who are in need and place demands on one's time on a regular basis and whether the respondent was married. Level of Education Respondents were provided a list that reflects different levels of education and asked to mark the one which indicates their level of education. Work Ethic Work ethic was measured with the four pro—ethic items based on Blood’s (1969) eight-item Protestant Work Ethic scale. Wanous (1974) reported an internal consistency reliability of .70 for the four pro-work ethic items when the Spearman-Brown formula was used. Based on all completed surveys, coefficient alpha was .75 (corrected using the Spearman-Brown formula) for the four pro-work ethic items. 93 W This was assessed with a single item which asks whether the employee is the primary wage earner in his/her family. Involuntary Absenteeism Involuntary absenteeism was based on individual's self- report reasons for absenteeism that fall into the following categories: personal illness, death in the family, jury duty, and voting. Involuntary absence behavior was assessed for a three month period using a frequency index and time lost Index. The frequency index is defined as the number of absence for the specified reasons during scheduled work days, regardless of length of each occurrence, for the three month period following survey administration. The time lost index is defined as the total number of scheduled work days absent for the specified reason during the three month period following survey administration. The unit of analysis is the work-day (i.e., a continuous eight-hour span). Information on absenteeism for half-days (i.e., a continuous four—hour span) was known. Vgluntary Absenteeism Voluntary absenteeism was based on individual's self- report reasons for absenteeism that fall into the following categories: unexcused absence (i.e., no reason stated and absenteeism not acknowledged as legitimate by the organization), and personal days. Voluntary absence behavior 94 was assessed for a three month period using a frequency index and time lost Index. The frequency index is defined as the number of absence for the specified reasons during scheduled work days, regardless of length of each occurrence, for the three month period following survey administration. The time lost index is defined as the total number of scheduled work days absent for the specified reason during the three month period following survey administration. The unit of analysis is the work-day (i.e., a continuous eight-hour span). Information on absenteeism for half-days (i.e., a continuous four-hour span) was known. Neither scheduled holidays nor vacation days was included as occurrences of absenteeism. Data Analysis Procedure Five principle methods of analysis were used in this dissertation. The first method was a correlation analysis. The second method was a multiple regression analysis. The third method was a t-test for the difference between independent means. The fourth method was a test for the difference between dependent correlations. The final method was a test for the difference between independent correlations. The criterion of significance was alpha equal to .05 for all inferential statistics. The data analysis for each hypothesis will be discussed. Data analyses for each hypothesis specifying regression analysis were conducted twice: for variables whose 95 distribution was not normal in their original form, and for these variables after the square root transformation was applied. The distribution of all continuous variables were examined using measures of skewness and kurtosis. In normal distributions, both measures are zero. When the measure of skewness is positive, the distribution is skewed to the left. When the measure of skewness is negative, the distribution is skewed to the right. When the measure of kurtosis is greater than zero, the distribution is more highly peaked than the normal distribution (i.e., leptokurtic). When the measure of kurtosis is less than zero, the distribution is flatter (i.e., platykurtic). Kendall and Stuart (1958) indicate that when the measure of skewness is approximately 2 and the measure of kurtosis is greater than 5, then departure from normality is extreme. Because departure from normality was found for the measures of absence behavior, an analysis of studentized residuals (Weisberg, 1980) was performed to assess whether there were outliers in the space of the criterion (i.e., absence behavioral distributions). Because the residual is assumed to have a normal distribution with a mean of zero, then about 99% of the residuals should lie within three standard deviations of the mean. Therefore, any residual with an absolute value greater than three is an outlier and should be carefully examined to explain why there is a poor 96 fit (Stevens, 1984). Mahalanobis Distance was calculated to determine whether there are outliers in the space of the predictors. Critical values (Barnett & Lewis, 1978) were used to determine whether Mahalanobis distance for each case was an outlier at alpha of .05. Finally, Cook’s Distance (Cook, 1977), a measure of the change in the regression coefficients that would occur if this case was omitted, was calculated. It was used to identify whether any cases that were outliers on both predictor space and criterion space were influential in affecting the regression equation. Cook and Weisberg (1982) indicate that a Cook's Distance of approximately 1 would be considered sufficiently large, indicating an influential case. Although none of the outlier cases were influential, the marginal distributions of the absence behavioral measures were different than the marginal distributions of the attitudinal factors. Thus, square root transformations (Box & Cox, 1964) were applied to the absence distributions to induce normality and to minimize the disparity between the marginal distributions of the behavioral and attitudinal variables, respectively. fiypgthesis 1: Exploratory factor analysis with varimax rotation was used to assess the extent to which absence intentions is multi-factorial. Principal components 97 analysis was used. Criterion for factor extraction was an eigenvalue with a value greater than or equal to one. Based on pretest results, however, it was expected that the items will correlate highly, thus allowing the use of an overall absence intentions score based on the sum of eight intentions items. Hypothesis 2: Hierarchical regression analysis was used to predict absence intentions. Theory (viz., Ajzen and Fishbein, 1980) dictates the following order of entry: attitudes about absenteeism followed by subjective norms. Hierarchical regression analysis permitted assessment of whether each variable significantly predicted the dependent variable with variance due to other independent variables controlled. In addition, the interaction between attitudes and norms was entered following subjective norms in order to assess whether a joint effect of attitudes and norms predicted absence intentions beyond each factor. It was expected that attitudes and norms would make significant contributions to explaining intentions, but the interaction term would not be significant. Hypothesis 3: A one-tailed t-test for independent means was used. flyppthesis 4: Hierarchical regression analysis was used to predict voluntary absenteeism. To rule out job satisfaction and organizational commitment as alternative explanations, the order of entry was: job satisfaction, and 98 organizational commitment as a set followed by absence intentions. It was expected that the regression weight for absence intentions would achieve significance and contribute a significant amount of variance. Hypothesis 5: Regression analysis was used to predict involuntary absenteeism. It was expected that the regression weight for absence intentions would npr achieve statistical significance. Hypothesis 6: Hierarchical regression analysis was used. The order of entry was: absence intentions, attitudes favoring absenteeism, and subjective norms favoring absenteeism. It was expected that the regression weights for attitudes and subjective norms, respectively, would npr achieve statistical significance. Hypothesis 7: Regression analysis was used. Hypothesis 8: A one-tail test for the significance of the difference between dependent correlations (Cohen 8 Cohen, 1983) was conducted. Hypothesis 9: Regression analysis was used. Hypothesis 10: A one-tail test for the significance of the difference between dependent correlations (Cohen 8 Cohen, 1983) was conducted. Hypothesis 11: This hypothesis was not tested. Hypothesis 12: Regression analysis was used. Hypothesis 13: A one-tail test for the significance of the difference between dependent correlations (Cohen 8 99 Cohen, 1983) was conducted. Hypothesis 14: Hierarchical regression analysis was used. The order of entry was job involvement then job satisfaction. It was expected that the regression weight for job satisfaction would not achieve statistical significance. Hypothesis 15: Regression analysis was used. Hypothesis 16: A one-tail test for the significance of the difference between dependent correlations (Cohen 8 Cohen, 1983) was conducted. Hypotheses 17: A zero-order correlation was calculated. Hypothesis 18: Partial correlation analysis was used. Age was controlled. It was expected that the coefficient would be statistically significant. Hyppthesis 19: Partial correlation analysis was used. Sex was controlled. It was expected that the coefficients would be statistically significant. Hyppthesis 20: A zero-order correlation was calculated. Hypothesis 21: A one-tail test for the significance of the difference between dependent correlations (Cohen 8 Cohen, 1983) was conducted. Hypotheses 22 to 24: First, the correlation between the square-root transformed absence measures was used for each effect size. Then, using the formula for correction 100 for attenuation (Nunnally, 1978, p. 220), it was expected that the corrected correlation squared would exceed 4%. The effect size was corrected for attenuation in both variables. For each attitudinal factor, coefficient alpha found in this dissertation was used. For the absence measures, the average inter-period stability coefficients for the frequency and time lost indexes (Hackett 8 Guion, 1985) were used. CHAPTERFIVE RESUDTS Introduction This chapter discusses the results of the data analyses. First, the factor structures of absence intentions, attitudes regarding absenteeism, and subjective norms will be examined. Second, the results of each hypothesis will be presented. Finally, a summary of the results will be presented. Item and Scale Analyses: Absence Intentions (Hypothesis 1) Factor Structure Assessment Based on the exploratory conceptual framework presented in Chapter Three, factor analysis with varimax rotation was used to determine whether the intentions items reduced to three interpretable factors. Hypothesis 1 specified that factor analysis would produce three orthogonal factors. The three factors that were proposed are (a) family responsibilities (items 1-take care of dependent children; 3-attending to family responsibilities other than children: 6-spending leisure time with family), (b) non-work alternative attraction (items 2-hobby/1eisure activity; 4- religious commitments; 5-other community activities), and 101 102 (c) pain-avoidance (item 7-relief from dissatisfying work situation; 8-general intention item). To further assess the validity of the pain-avoidance factor, the correlation between the proposed pain-avoidance factor and job satisfaction was calculated. It is expected that this factor will be significantly and inversely related to job satisfaction. Factor analysis, when limiting the number of extracted factors to three, resulted in three interpretable factors. The principal components method was used. However, the resulting factors do not correspond to the hypothesized factors specified in Chapter Three. Thus, hypothesis 1 was not supported. Table 5 shows the rotated factor matrix. Overall, these factors accounted for 68.7% of the variance in these data. The correlation between the proposed pain avoidance scale and job satisfaction scale was r (427) = -.25, p<.001. With item 7 alone, the correlation was r (425) = -.28, p<.001. With item 8 alone, the correlation was r (395) = -.15, p<.001. Hactor 1: Items 2, 6, 7 and 8 formed a factor combining items specifying attraction to non-work alternatives such as leisure time as well as the item specifying one's intention to get away from a dissatisfying work situation. Overall, this factor may be labeled as absence to take a break from work. 103 Table 5 Results of Factor Analysis of Intention Items Factor Communality Loadings Estimates Item 1 2 3 Factor 1: Break from WOrk 8. General intention .76 -.13 .22 .64 2. Hobby/leisure .75 .30 .06 .65 6. Family leisure .70 .51 -.02 .75 7. Dissatisfying work situation .67 .28 .20 .58 Factor 2: Community and Religious Activities 4. Religious commitments .01 .82 .26 .74 5. Community activities .43 .73 -.02 .72 Factor 3: Family Responsibilities 1. Dependent children .06 .09 .90 .82 3. Family responsibilities .48 .16 .58 .59 Percent of common variance 44 13 12 N=428. 104 Communality estimates indicate that 65.2% of the total variance in item 2 was accounted for by the combination of all common factors. For item 6, 75.3% of the total variance was accounted for by the combination of all common factors. For item 7, 57.6% of the total variance was accounted for by the combination of all common factors. Finally, 64.1% of the total variance in item 8 was accounted for by the combination of all common factors. Each item has a factor loading of at least .30 which is consistent with conventions in the behavioral sciences (Nunnally, 1978). This factor accounted for 44.1% of the variance in these data. Internal consistency reliability for these items was .75. Hggror 2: Items 4 and 5 formed a factor combining items specifying commitment to religious and community activities. This factor may be labeled absence for community and religious activities. Communality estimates indicate that 74.1% of the total variance in item 4 was accounted for by the combination of all common factors. For item 5, 71.9% of the total variance was accounted for by the combination of all common factors. This factor accounted for 12.9% of the variance in these data. Internal consistency reliability for these items was .62. Eactor 3: Items 1 and 3 formed a factor which combines items specifying fulfillment of family responsibilities. Communality estimates indicate that 82.2% of the variance 105 was accounted for by the combination of all factors. For item 3, 58.9% of the total variance was accounted for by the combination of all common factors. This factor accounted for 11.7% of the variance in these data. Internal consistency reliability for these items was .26. The low internal consistency estimate is to be expected in light of the magnitude of the eigenvalue for this factor (.93). The eigenvalue is often used as an index of how many factors one should rotate and interpret. Kaiser (1958) has proven that the internal consistency of a factor with an eigenvalue of 1.0 is zero. Also, 1.0 is the amount of variance accounted for by a single item (r2 = 1.0). Thus, if a factor has an eigenvalue less than 1.0, this means it accounts for less variance than a single item (N. Schmitt, personal communication, Psychology 818, 1987). Although factor analysis produced one sufficiently rglinplg factor, further analyses are needed to determine whether all items should be used as a single intentions scale or whether the reduced item pools should be used separately as a measure of different aspects of absence intentions. Specifically, three additional statistical item analyses were performed to examine this issue. These include assessment of internal consistency reliability, the correlation between each of the factors derived from factor analysis, and item-total correlation analysis. The decision whether to use separate factors or the overall item pool is 106 based on the results of all three analyses. Internal Consistency Reliability Table 6 shows the internal consistency reliability estimates for the overall set of absence intentions items, factor 1, factor 2, and factor 3, respectively. The results suggest that the overall set of items is more internally consistent than the individual factors. However, because reliability is a function of both inter-item correlation and number of items, it is possible that the estimate for the overall set is highest because it contains the most number of items. Table 6 Reliability Coefficients for Intentions Factors Coefficient Alpha Factor 1: Break from Work .75 (4 items; n=428) Factor 2: Community and Religious .62 Activities (2 items: n=428) Factor 3: Family Responsibilities .26 (2 items; n=428) All Items .78 (8 items; n=428) Table 7 shows the inter-item correlations between each of the intentions items. The average inter-item correlation for the overall scale was (r=.35). The average inter-item 107 correlations for factors 1, 2, and 3 were r=.48, r=.43, and r:.35, respectively. The similarity of the average inter- item correlations suggests that the differences in level of reliability are due to the number of items in factors 2 and 3. On the basis of this analysis, only the overall set of items and factor 1 meet the criterion for internal consistency reliability in basic research (Nunnally, 1978). 108 .Hoo.vm .mumsuo Haw «mo.vm« .msmufl umnuo sou pmewzw4~¢ “a Emu“ Hafiz mcoflumamuuoo you mnmwzwsam .usmmnm on on >mefla hammflnuh uucmmnm on on mamxwass >Hnownua m III mm. hm. mm. ha. mm. me. 5H. m.H h.~ coflucmusa Hmumcmw .m III mm. 44. mm. 88. m8. mm. o.~ m.~ cofiumsuam xuos msflammwummmflo .5 III mm. on. H8. mm. «8H. 0.H N.N musmwma %HflBMh .9 III m8. mm. 2V. 8:. min aim mmfiufirwuom muwssfifioo .m III pm. mm. on. m.a m.m mucmsufissoo msOfloflHmm .8 III mm. mm. m.a o.m mmfiuflafln Iflmsommmu aaflamm .m III mm. 8.H o.m musmflwa \snnom .m III m.H m.m cmuoaflso usoosmmwo .H m h m m 8 m m a an a! HmuH mamuH msowusmusH no mGOMHMHOHHOOkusH can msowuww>wa phwpcdum .msmm: h OHAMB 109 Correlation Between Factors Table 8 shows the inter-scale correlations between the factors derived from factor analysis. The correlation between scales was moderate, although statistically significant. The relationship between factor 2 and factor 3 was r (426) = .36, p<.001. The relationship between factor 1 and factor 3 was r (427) = .47, p<.001. The relationship between factor 1 and factor 2 was r (426) = .50, p<.001. These correlations represent the relationship between two factors that are imperfectly measured, thus, undermining the theoretical relationship between two hypothetical factors. Ghiselli et al. (1981) recommend that correction for attenuation be made when assessing the relationship between factors in construct validity assessment. An objective of construct validity assessment is to estimate the theoretical relationship between two constructs that are imperfectly measured (Nunnally, 1978). Thus, the corrected correlation between factor 1 and factor 2 was r (426) = .90, p<.001. The corrected correlation between factor 1 and factor 3 was r (426) > 1.0. The corrected correlation between factor 1 and factor 2 was r (426) = .73. The correction for the relationship between factor 1 and factor 3 produced a nonsensical estimate that may occur when the degree of measurement unreliability is high (Nunnally, 1978). Nevertheless, the high percentage of measurement error present in factor 3 suggests that the measured correlation 110 including factor 3 is likely to be quite attenuated. Because the reliability estimate for the entire scale is similar in magnitude to the corrected correlations between factors, there is little discriminability among factors. Thus, the overall set of items can be used to form a homogenous absence intentions scale. Table 8 Means, Standard Deviations and Intercorrelations of Intentions Factors Factor M SD 1 2 3 Factor 1: Take a Break 2.42 1.35 --- Factor 2: Community/ Religion 2.21 1.30 .50 --- Factor 3: Family Respon- sibilities 4.07 1.77 .47 .36 --- 438$N$439. p<.001 for all coefficients. Itgn-Scale Correlations Another criterion for scale homogeneity is whether an item correlates more highly with its own scale (corrected for spurious item-total overlap) than with another scale. In other words, an item should be grouped with a scale with which it correlates most highly to the extent that it makes sense from a content perspective. Table 9 shows the 111 corrected item-scale correlations for factor 1 compared with the item-scale correlations for factor 2, factor 3, and the overall set of items. Overall, this comparison indicates that all items forming factor 1 correlate more highly with the overall set of items than with itself, factor 2 or factor 3. One exception is item 8, the general intention item, which correlates slightly more highly with factor 3 than with factor 1 or the overall set of items. However, there is no substantive difference between r=.38 and r=.34. Table 9 Item-Scale Correlations: Break From Work Items Factor All Itemsa Item la 2 3 8. General intention .34 .23 .38 .34 2. Hobby/leisure .62 .41 .34 .63 6. Family leisure .62 .54 .37 .66 7. Dissatisfying work situation .54 .40 .40 .57 aCorrected for spurious overlap. N=428. p<.001. Table 10 shows the corrected item-scale correlations for factor 2 compared with the item-scale correlations for factor 1, factor 3, and the overall set of items. Item 4 112 correlates slightly more highly with factor 2 than with the overall set of items. Item 5 correlates slightly more highly with factor 1 than with the overall set of items. However, there are no substantive differences between each pair of correlations. Table 10 Item-Scale Correlations: Community and Religious Items Factor All Itemsa Item 1 2a 3 4. Religious commitments .33 .45 .31 .40 5. Community activities .56 .45 .31 .54 aCorrected for spurious overlap. N=428. p<.001. Table 11 shows the corrected item-scale correlations for factor 3 compared with the item-scale correlations for factor 1, factor 2, and the overall set of items. Even though item 1 correlates more highly with factors 1 and 2 than with its own factor or the overall set of items, the similar internal consistency reliability estimates for factor 1 and the overall set of items suggests that it may be included in either set, though the magnitude of the correlation is low. However, corrected item-total 113 correlations of at least .15 are within conventions expected for homogenous tests (N. Schmitt, Personal communication, Psychology 818, 1987). Table 11 Item-Scale Correlations: Family Responsibilities Factor All Itemsa Item 1 2 3a 1. Dependent children .26 .21 .15 .15 3. Family responsibilities .50 .35 .15 .52 aCorrected for spurious overlap. N=428. p<.001. W In summary, these analyses suggest that the overall set of items form a more homogenous and psychometrically sound scale than do the individual factors. This conclusion is further supported by the insufficient level of reliability for factors 2 and 3. Furthermore, because the magnitude of the corrected correlations between factors are similar to the magnitude of the reliability estimates, there is little discriminability among the factors. Given these results, using the overall set of items will provide an overall absence intentions measure that incorporates various reasons for intending to be absent. Also, because the maximum level 114 of validity between a predictor and criterion is the level of reliability, using the overall set of items will enhance predictability. Thus, weak effects will be minimized. Item and Scale Analyses: Attitudes Toward Absenteeism F c Structure Assessment Exploratory factor analysis with varimax rotation was used to assess the factorial structure of attitudes toward absenteeism based on 15 belief statements about the potential outcomes of voluntary absenteeism. The principal components method was used. Based on Morgan and Herman's (1976) findings that people perceive outcomes of voluntary absenteeism as either motivating or deterrents, this analysis limited the number of extracted factors to two. This analysis resulted in two interpretable factors that are consistent with Morgan and Herman's (1976) research. Table 12 shows the rotated factor matrix. Overall, these factors accounted for 40.6% of the variance in these data. Factor 1: Items 1, 2, 6, 7, 9, 10, and 14 formed a factor that is consistent with Morgan and Herman's (1976) set of motivating outcomes of voluntary absenteeism. Communality estimates indicate that 38.4% of the total variance in item 1 was accounted for by the combination of all common factors. For item 2, 36.2% of the total variance was accounted for by the combination of all common factors. For item 6, 52.0% of the total variance was accounted for by 115 Table 12 Results of Factor Analysis of Belief Items Factor Communality Loadings Estimates Item 1 2 Factor 1: Motivating Outcomes 7. Relief from job .76 .12 .59 6. Things I enjoy .72 -.06 .52 10. Get Away from co—workers .68 .16 .48 14. Get Away from supervisor .68 .16 .49 1. Time with friends .62 .00 .38 2. Time for family .60 .04 .36 9. Meet obligations .59 .15 .37 Factor 2: Deterrent Outcomes 12. Fired from company .03 .76 .58 13. Loss of promotion opportunities -.02 .72 .52 11. Loss of benefits .16 .67 .48 3. Discipline by superv1sor .06 .65 .42 8. Demoted -.05 .55 .30 5. Loss of pay .30 .49 .33 4. Upset co-workers .14 .41 .19 15. Heavier workload .03 .27 .07 Percent of Common Variance 25 15 (n=428) 116 the combination of all common factors. For item 7, 59.0% of the total variance was accounted for by the combination of all common factors. For item 9, 36.7% of the total variance was accounted for by the combination of all common factors. For item 10, 48.2% of the total variance was accounted for by the combination of all common factors. Finally, 48.9% of the total variance in item 14 was accounted for by the combination of all common factors. Internal consistency reliability for these items was .79. Factor 2: Items 3, 4, 5, 8, 11, 12, 13, and 15 formed a factor that is consistent with Morgan and Herman’s (1976) deterrent outcomes of voluntary absenteeism. Communality estimates indicate that 42.3% of the total variance was accounted for by the combination of all factors. For item 4, 18.7% of the total variance was accounted for by the combination of all common factors. For item 5, 32.8% of the total variance was accounted for by the combination of all factors. For item 8, 29.8% of the total variance was accounted for by the combination of all common factors. For item 11, 48.1% of the total variance was accounted for by the combination of all factors. For item 12, 57.9% of the total variance was accounted for by the combination of all factors. For item 13, 51.6% of the total variance was accounted for by the combination of all factors. Finally, for item 15, 7.2% of the total variance was accounted for by the combination of all common factors. 117 Internal consistency reliability for these items was .72. Although factor analysis produced two reliable factors, further analyses are needed to determine whether all items should be used collectively as a single attitudes toward absenteeism scale or whether each reduced item pool should be used as measures of different aspects of attitudes about absenteeism. These include assessment of internal consistency reliability, the correlation between each of the factors derived from factor analysis, an item-total correlation analysis. The decision whether to use separate factors or the overall item pool is based on the results of all three analyses. Internal Consistency Reliability Table 13 shows the internal consistency reliability estimates for the overall set of attitude items, factor 1 and factor 2, respectively. The magnitude of the coefficients for each set of items is approximately similar. Table 13 Reliability Coefficients for Belief Factors Coefficient Alpha Factor 1: Motivating Outcomes .79 (7 items: n=428) Factor 2: Deterrent Outcomes .72 (8 items: n=428) All Items .78 (15 items: n=428) 118 Table 14 shows the intercorrelation between each pair of belief items. The average inter-item correlation for the motivating outcome items was r = .33. For the deterrent item, r = .24. For the overall set of items, r = .19. 119 38.3 .3. m m 30.3 .3. 3 3. 3 “86m .3. 3 SIM ("Hmong 8023.385 2283: 8:53.83 a . 3533380 5383.938 g 633.3 8083 1.580 m 80QO 3 map ".3920 mama—mu .8893 gonna “8.50.50 8.39% "302 .II no em mo ma OH OH 50 no mo 3 2 No 00 MOI ¢.N m.m G .3 II. NH ma 0H mm mm HO! Hm mm mm ma «H mm on ¢.N m4 2:: II mm mm «.0 ma NN ma 00! ma om mm v0 00 m.N NJ G .MH ll N¢ 0H ma mm ma MOI hm .vH O¢ 3 3 m.N bé O .NH I ma ma Nv ma ma vm 0H Hm OH ma o.N ¢.m Q .HH .II Hm mo wm on HN ma HH mm .vm m.N mé 2 .OH ll mo me hm Hm MH no on mm o.N w.m 2 .m I No No 3” mo ma oo 50 H.N m.m Q .w I Nv ....N OH NH em mm N.N ozv 2 .5 I. ha 00 00 mm m¢ .....N H.m Z .0 I am mm 3” MH m.N OJ. 0 .m II on no mo N10. ozv G .v ll md 00 H.N azv Q .m ll mm o.N 03v 2 .N l h...” m.m S .H medium—“NH HH OH m w h w m .v n N H a an 5H EH33HO§D§§E§§.§ :83 120 Correlation Between Factors The correlation between factor 1 and factor 2 was moderate, although statistically significant (; (428) = .28, p<.001). When both factors were corrected for attenuation, the coefficient increased to r=.37. Thus, the two factors are not orthogonal. Item-Scale Correlations Another criterion for scale homogeneity is whether an item correlates more highly with its own scale (corrected for spurious item-total overlap) than with another scale. Table 15 shows the corrected item-scale correlations for factor 1 compared with the item-scale correlations for factor 2 and the overall set of items. Each item for factor 1 correlates more highly with factor 1 than with factor 2 or the overall set of items. The item-total correlations for factor 1 items with the overall set of items, however, meet conventions for homogenous tests (N. Schmitt, personal communication, Psychology 818, 1987). Furthermore, the differences in magnitude between the item-factor 1 correlations and their respective item-all items correlations were not substantive. 121 Table 15 Item-Scale Correlations: Motivating Outcome Items Factor All Itemsa Item 1a 2 7. Relief from job .64 .20 .52 6. Things I enjoy .52 .08 .36 10. Get Away from co-workers .55 .23 .49 14. Get Away from supervisor .53 .24 .48 1. Time with * friends .44 .11 .32 2. Time for family .45 .14 .39 9. Meet obligations .46 .22 .44 aCorrected for spurious overlap. N=428. *Q<.010; others, Q<.001. Table 16 shows the corrected item-scale correlations for factor 2 items compared with the item-scale correlations for factor 1 and the overall set of items. With the exception of items 5 and 15, factor 2 items correlate more highly with factor 2 than with factor 1 or the overall set of items. However, the item-total correlations for factor 2 with the overall set of items meet conventions for homogenous tests. Furthermore, with the exception of item 12 (fired from the company), the difference in magnitude 122 between its correlation with factor 2 and its correlation with the overall set of items was not substantive. Table 16 Item-Scale Correlations: Deterrent Outcome Items Factor All Itemsa Item 1 2a 12. Fired from company .15 .54 .20 13. Loss of promotion opportunities .10 .51 .38 11. Loss of benefits .23 .49 .44 3. Discipline by supervisor .13 .45 .38 8. Demoted to a lower job .04b .33 .24 5. Loss of pay .30 .37 .43 4. Upset co-workers .16 .36 .34 15. Heavier workload .07b .20 .20 aCorrected for spurious overlap. not significant; others, p<.010. N=428. 5111111113: 2 In summary, these analyses suggest that the overall set of items can be used as either a single scale or as two separate ones. The significant correlation between factor 1 and factor 2 suggests that individuals who perceive a 123 particular set of outcomes as motivating also perceive the other set of outcomes as deterrents. Together both sets of items represent a range of potential outcomes of voluntary absenteeism that is more meaningful than a restricted range of either motivating or deterrent outcomes when formulating attitudes about voluntary absenteeism. Thus, given the results of the previous statistical analyses and theoretical considerations, the entire set of items will be used as the basis for attitudes about voluntary absenteeism. Item and Scale Analyses: Subjective Norm Factor Structure Assessment Exploratory factor analysis with varimax rotation was used to assess the factorial structure of items developed to assess subjective norms about absenteeism. This analysis resulted in two interpretable factors. Table 17 shows the rotated factor matrix. Overall, these factors accounted for 51.2% of the variance in these data. Factor 1: Factor 1 is comprised of items that represent individuals who are referents in forming norms about absenteeism within the workplace, but whose relationship with the respondent is not solely a formal workplace relationship. These individuals include: family, friends, spouse or significant other, and yourself. It may be titled non-work affiliates. Internal consistency reliability was .62. 124 Table 17 Results of Factor Analysis of Subjective Norm Items Factor Communality Loadings Estimates Item 1 2 Factor 1: Non-work Affiliates 2. Family .78 -.08 .61 1. Friends .75 .10 .58 6. Yourself .63 .21 .44 5. Spouse/Significant Other .58 .08 .35 Factor 2: WOrk Affiliates 3. Co-Worker .03 .76 .58 4. Supervisor .13 .71 .52 Percent of Common Variance 34 18 N=428. Communality estimates indicate that 60.8% of the total variance in the family item was accounted for by the combination of all common factors. For the friend item, 57.8% of total variance was accounted for by the combination of all common factors. For the spouse/significant other item, 34.6% of the total variance was accounted for by the 125 combination of all common factors. For the item addressing yourself, 43.7% of the total variance was accounted for by Factor 2: Factor 2 is comprised of two items representing individuals who are members of the organization: co-workers and supervisor. It may be titled work affiliates. Communality estimates indicate that 52.3% of the total variance in the supervisor item was accounted for by the combination of all factors. For the co-worker item, 57.7% of the total variance was accounted for by the combination of all factors. Internal consistency reliability was .20. Internal Consistency Reliability Table 18 shows the internal consistency reliability coefficients for factor 1, factor 2, and the overall set of items. These coefficients are below conventional levels in the behavioral sciences (Nunnally, 1978). The low coefficients may be due to the low average inter-item correlation that may be the result of error in content sampling or the number of items. 126 Table 18 Reliability Coefficients for Subjective Norm Factors Coefficient Alpha Factor 1: Work Affiliates .62 (4 items; n=428) Factor 2: Deterrent Outcomes .20 (2 items; n=428) All Items .56 (6 items; n=428) For factor 1, the average inter-item correlation was moderate (; = .31). For factor 2, the correlation between the two items was low (E = .13). The average inter-item correlation for the overall set of items was also low (; = .18). Table 19 shows the means, standard deviations, and intercorrelations of the subjective norm items. Internal consistency reliability estimates are coefficients of equivalence (Cronbach, 1947) that measure the extent of a single group factor shared by all individuals. In this case, it would be consistency in terms of who believes voluntary absenteeism is acceptable or unacceptable among the six individuals represented by these items. The low magnitude of these coefficients, in conjunction with the low average inter-item correlations, suggest that these items do not form the basis of subjective norms that is consistent across individuals. Overall, the respondent may perceive that the relevant six individuals feel that absenteeism is either acceptable or unacceptable. 127 .ucmmnm on on usmccommmu now manmummoom ma aw 0>mwamn «ucomnw on 0» ucmocommmu you manmummoom mw NH m>ofiamn no: moon somummua . «W00 H VM¥¥% HO V {fi Mow... .mmevzvmav mmoc comummnh M nun «44mm. ..«em. 4..»m. «ca. «wwma. m.H H.m uammuso» .6 nu: «.«mm. «4.0m. Ho. «.«ma. o.m m.n umsuo .cmflm \mmsomm .m In: «44mv. «4*ma. «ca. m.H m.m mccwflum .e nu: mo. no. o.m m.m sangum .m nu: «.ma. m.H m.m mumxuozuoo .m III m.a m.~ Homfl>ummom .H w m e m m H om a: amuH mflmuH lHoz O>Huomnnflm no mGOHHMHOHHOOHmucH 0:8 msowumfi>wn cumvcmum .mcdfll ma OHAMB 128 However, which particular individuals feel that absenteeism is acceptable or unacceptable may vary across respondents. Nevertheless, these items provide a basis for indicating each individual's perceptions about whether relevant others (e.g., supervisor, spouse) feel that voluntary absenteeism is acceptable. 31111111132! The results of these analyses suggest that subjective norms do not form an internally consistent, homogenous scale. In other words, the nature of acceptability of voluntary absenteeism varies across individuals. The low internal consistency reliability estimates are likely the result of inadequate content sampling. In other words, perhaps other individuals not included in this list may be relevant in the formation of norms about voluntary absenteeism. For example, the low reliability coefficient for the work affiliate scale suggests that this scale may be conceptualized as "peers of the workplace" such as oo- workers versus "superiors of the workplace" such as management. Absence Behavioral Measures As discussed in Chapter Two, research has suggested two types of absenteeism: involuntary and voluntary. Based on the data received from the sponsoring organization, four absence indexes were calculated: involuntary-frequency, involuntary-time lost, voluntary-frequency, and voluntary- 129 time lost. Table 20 shows the correlations between each pair of absence measures. Clearly, there is a high degree of discriminability between the self-report reasons of absenteeism that fall into involuntary and voluntary categories. These results are consistent with findings by Chadwick-Jones, et al. (1982) and Hackett and Guion (1985). Table 20 Means, Standard Deviations and Intercorrelations of Absence Behavioral Measures Measure M3 so 1 2 3 4 1. Involuntary- frequency .9 1.1 .10b 2. Involuntary- time lost 1.2 1.9 .78*** .07b 3. Voluntary- frequency .6 1.1 .13** .09* .34b 4. Voluntary- time lost .7 1.4 .12** .08 .92*** .33b a measured in work days. average interperiod stability coefficients based on three one-month periods; N=206. N=328. **Q<.05. ***p<.01. p<.001. Table 20 also shows the estimates of reliability for each of the absence measures. The coefficients reported 130 reflect stability of the absence behavior based on the average interperiod correlation among absences for each of the three one-month periods studied. Within each category of behavior, there is no substantive difference in the magnitude of the stability coefficients between frequency and time lost measures. Therefore, analyses will be provided for each measure. Properties of Sample Distributions of variables As discussed in Chapter Two, weak effects may be the result of violations in the assumptions underlying correlation and regression analyses. To the extent that the marginal distribution between two variables differs, then biased results may be likely. Heteroscedasticity may be responsible for leptokurtosis and skewness in the marginal distribution. Thus, a nonnormal sample distribution indicates that the violation of heteroscedasticity has occurred. Measures of skewness (g1) and kurtosis (92) will be examined for all attitudinal variables and absence behavioral measures to assess the extent of nonnormality in the marginal distribution of each continuous variable. Kendall and Stuart's (1958) criteria will be used to assess whether each sample distribution is normally distributed. They suggested that when 91 approaches 2 and g2 exceeds 5, there is considerable skewness and leptokurtosis. Table 21 shows sample statistics for the variables used 131 in this study. The statistics include the mean, median, mode, standard deviation, range, skewness, and kurtosis. A review of these statistics reveals that voluntary- frequency of absence, voluntary-time lost, and involuntary- time lost sample distributions are consistent with Kendall and Stuart’s (1958) criteria for nonnormality of the marginal distribution. Although the sample distribution for involuntary-frequency of absence does not meet both criteria, it is noticeably skewed to the left. The results for the absence behavioral measures are consistent with Hammer and Landau’s (1981) investigation of absence distributional properties. The positive values of both skewness and kurtosis indexes suggest that the noted sample distributions are both low-base rate and leptokurtic (i.e., more highly peaked than normal distributions). Of noteworthy interest are the properties of the absence intentions and attitudes toward absenteeism distributions. Compared with the skewness and kurtosis indexes of the absence behavior measures, the skewness and kurtosis indexes are substantively closer to zero for the attitudes about absenteeism and absence intentions distributions. These findings suggest that attitudes about absenteeism and absence intentions are distributed approximately normal. In general, not everyone who intends to be absent is absent. More will be said about this in Chapter Six. 132 .ov¢WZWva 44.54 mo.m 06.44 om.4 oo. om. 4~.H .H.uu>umu:sao>cH m~.~ 44.4 oo.m NH.H oo. 66.4 mm. .vmuuumumucsao>cH NH.>H pm.m oo.~a 04.4 oo. oo. 45. umoa msfluusumucsao> 6m.44 so.m oo.m 04.4 oo. oo. mm. scamswmuunmumucsao> 46.: mm. sm.m >4.H oo.H sm.~ mn.~ mcoflucmucfl mocmmna 04.m >4.H nm.4m mo.m >4. om. mm. mayo: m>fluomnnnm mm. mm.: oo.>ma H>.- oo. oo. ~4.Ha mmusufluum mocmmn< oo.~u 46. 06.4 om. oo. 00. m4. umcumm moms .sflum 44.4 46.4 m~.m 44. oo. no. we. sueaflnflmcoommu :«M 40.: m4.u om.m mm. 00.4 m~.m m~.m carom xuoz mm. mm.u o4.m mm. o~.m mm.4 mm.4 cofiuomumflumm non 4m.u Hm.n mm.4 mm. o~.4 64.4 mm.4 unmauflssoo HMGOfiuwuficmmuo mm.u m~.u oo.m mm. mH.4 64.4 nm.4 newsm>ao>cfl non mm Hm mwmouhsfi mmwsbmxm mmcmm am moo: cowoml :80! macawum> moanwwum> mo mcowusnfluumwo «Haida HN OHQMB 133 Summer! Overall, these findings suggest that the absence behavioral measure and subjective norm measures violate the assumption of heteroscedasticity. Thus, the use of ordinary least squares regression may result in constricted effect size estimates. Diagnostic statistics to detect outliers will be calculated. Also, an assessment of whether any outliers are influential (Stevens, 1984) will also be made. If influential outliers are found, a decision whether to delete them or conduct appropriate data transformations will be made in order to enhance the statistical conclusion validity of the analyses. Results of Analyses by Hypothesis Hypothesis 1 The results of hypothesis 1 were presented earlier in the Item and Scale Analyse : Absence Intentions section. Hypothesis 2 Hypothesis 2 stated that attitudes favoring absenteeism and subjective norms favoring absenteeism will independently predict absence intentions. Table 22 shows the results of the hierarchical regression analysis. 134 Table 22 Hierarchical Regression Results for Absence Intentions Predicted from Attitudes and Norms Predictor Ba R2 (change) Attitude favoring absenteeism .26* ' .10 Subjective norms .27* .08 Interaction term .06 .00 R2 . 18 §B=Beta, the standardized regression coefficient. p<.001. The results indicate that for the combination of attitudes and subjective norms favoring absenteeism, 32 was .18 for absence intentions. Attitudes favoring absenteeism (Beta = .26, p<.001) and subjective norms (Beta = .27, p<.001) significantly predicted absence intentions. The interaction term (Beta = .06, as) was not statistically significant. Both attitudes favoring absenteeism and subjective norms favoring absenteeism independently predicted absence intentions. Thus, Hypothesis 2 was supported. Hypothesis 3 Hypothesis 3 stated that subjective norms favoring absenteeism will be greater for blue collar workers than for white collar workers. Table 23 summarizes the results. On 135 average, both blue collar workers (5 = 1.1, SD = 3.4) and white collar workers (M = .8, SD = 2.8) have norms that favor absenteeism. Although the magnitude was greater for blue collar workers than white collar workers, this difference was not statistically significant (t (321) = .95, _§). Thus, hypothesis 3 was not supported. The 95% confidence interval (.705 M g 1.56) for the mean blue collar norms contained only positive values which suggests that blue collar worker norms favor absenteeism. The 95% confidence interval for the mean white collar norms (.545 M g 1.12) contained only positive values which suggests that white collar norms favor absenteeism. Finally, an analysis of each group's response patterns (indexed by standard deviation) shows that there was wider variation in subjective norms about absenteeism among blue collar workers than white collar workers. This difference was statistically significant (F = 1.47, p<.005). Table 23 summarizes the mean values for the blue collar and white collar workers on attitudes regarding absenteeism, subjective norms, and absence intentions. Overall, blue collar workers had attitudes and intentions favoring voluntary absenteeism that were significantly greater than those attitudes held by white collar workers (with the exception of subjective norms). In addition, blue collar workers exhibited significantly greater voluntary and involuntary absenteeism than white collar workers (with the 136 exception of absenteeism measured by the involuntary-time lost index). Table 23 T Tests Comparing Blue Collar and White Collar Workers on Absence Attitudes, Norms, Intentions and Behavior H Factor Blue White t p Attitudes favoring absenteeism 3.86 -4.94 4.04a .001 Subjective norms 1.13 0.83 0.95b ns Absence intentions 3.03 2.53 4.47C .001 Voluntary-frequency 0.84 0.48 3.02d .003 Voluntary-time lost 1.00 0.56 2.99e .003 Involuntary-frequency 1.11 0.72 3.53f .001 Involuntary-time lost 1.43 1.12 1.769 ns agf = 436; ha; = 321; °_; = 437; d_; = 241; eg; = 240 g; = 314; 90: = 419. Hypothesis 4 Hypothesis 4 stated that absence intentions will predict voluntary absenteeism. Results will be presented for both measures of voluntary absenteeism: frequency and time lost. Table 21 indicated that the sample distributions of voluntary absenteeism depart significantly from normality. In addition, the distributions for voluntary 137 absenteeism differ from the absence intentions distribution which is approximately normal. Thus, the results of regression analysis will be presented for the data in their original form as well as for the absenteeism data after square root transformations were applied. The distributions of the absence measures were examined for significant outliers in the space of the criterion and predictors. Studentized residuals were examined (Weisberg, 1980) for outliers in the criterion space. Five cases fell outside three standard deviations for the frequency index. Six cases fell outside three standard deviations for the time lost index. Mahalanobis distance (Q2) was calculated to assess the extent of outliers in the space of the predictor. Only one case exceeded the critical value for a significant outlier (p2 = 28.21, p< .05). Stevens (1984) indicated that outliers are likely to bias regression results when a case is significantly aberrant in both the predictor and criterion space. Specifically, when Cook’s Distance (CD) for a particular case is greater than one, the case should be considered as a candidate for deletion. None of Cook’s Distances exceeded one. The greatest value for Cook's Distance was .13 for the frequency index and .15 for the time lost index. Based on these analyses, a decision was made to apply a square root transformation to each absence distribution, 138 which is most likely to minimize skewness and kurtosis in truncated distributions with zero values (Box a Cox, 1964). After the transformations were applied, skewness (g1) was 1.01 and kurtosis (92) was .07 for the frequency index. For the time lost index, g1 was 1.15 and g2 was .57. These transformed distributions approximate normality more closely than the untransformed distributions. Table 24 shows the results of the regression analyses for hypothesis 4. Hierarchical regression analyses were used. Job satisfaction and organizational commitment were entered into the equation first followed by absence intentions in order to rule them out as alternative explanations to absence intentions. The regression coefficients for both organizational commitment and job satisfaction were not statistically significant for any of the analyses. Absence intentions significantly predicted voluntary-frequency that was not transformed (Beta = .11, p<.02). The 95% confidence interval excluded zero (.02 S B 5 .20). Absence intentions marginally predicted voluntary- time lost that was not transformed (Beta = .09, p<.07). Its 95% confidence interval contained zero. Using the same strategy, absence intentions did not significantly predict either transformed measure of voluntary absenteeism. Overall, hypothesis 4 was supported for the untransformed frequency index of voluntary absenteeism. 139 Ghiselli et al. (1981) recommend that correction for attenuation be made when assessing the relationship between factors in construct validity assessment. An objective of construct validity assessment is to estimate the theoretical relationship between two constructs that are imperfectly measured (Nunnally, 1978). The corrected correlation between absence intentions and the untransformed measure of voluntary-frequency was p (428) =.21. Thus, after correcting for measurement unreliability, absence intentions accounted for 4.4% of the variance in voluntary absence frequency. 140 Table 24 Regression Results for VOluntary Absenteeism Predicted from Absence Intentions Criterion Ba R2 Voluntary-frequency ** (not transformed) .11 .01 Voluntary-time lost (not transformed) .09* .01 Voluntary-frequency (transformed) .08 .01 Voluntary-time lost (transformed) .06 .01 aB=Beta, the standardized regression coefficient. *p<.07. **p<.02. Hypothesis 5 Hypothesis 5 stated that absence intentions will not predict involuntary absenteeism. Results will be presented for both measures of involuntary absenteeism: frequency and time lost. Table 21 indicated that the sample distributions of involuntary absenteeism depart significantly from normality. In addition, the distributions for involuntary absenteeism differ from the absence intentions distribution which is approximately normal. Thus, the results of regression analysis will be presented for the data in their original form as well as for the absenteeism data after square root transformations were applied. The distributions of the absence measures were examined 141 for significant outliers in the space of the criterion and predictor. Studentized residuals were examined (Weisberg, 1980) for outliers in the criterion space. Three cases fell outside three standard deviations for the frequency index. No cases fell outside three standard deviations for the time lost index. Mahalanobis distance (Q2) was calculated to assess the extent of outliers in the space of the predictor. Only one case exceeded the critical value for a significant outlier (D2 = 28.21, p< .05). Stevens (1984) indicated that outliers are likely to bias regression results when a case is significantly aberrant both in the predictor and criterion space. Specifically, when Cook’s Distance (CD) for a particular case is greater than one, the case should be considered as a candidate for deletion. None of Cook’s Distances exceeded one. The greatest value for Cook’s Distance was .04 for the frequency index and .20 for the time lost index. Based on these analyses, a decision was made to apply a square root transformation to each absence distribution, which is most likely to minimize skewness and kurtosis in truncated distributions with zero values (Box & Cox, 1964). After the transformations were applied, skewness (g1) was .40 and kurtosis (92) was -1.22 for the frequency index. For the time lost index, 91 was .73 and Q; was -.24. These transformed distributions approximate normality more closely 142 than the untransformed distributions. Table 25 shows the results of the regression analyses for hypothesis 5. Hierarchical regression analyses were used. Job satisfaction and organizational commitment were entered into the equation first followed by absence intentions in order to rule them out as alternative explanations to absence intentions. The regression coefficients for both organizational commitment and job satisfaction were not statistically significant for any of the analyses. Absence intentions significantly predicted involuntary absenteeism for all measures. After controlling for the effects of organizational commitment and job satisfaction, the following regression coefficients were found: (Beta = .24, p<.001) for the untransformed frequency index; (Beta = .20, p<.001) for the untransformed time lost index; (Beta = .22, p<.001) for the transformed frequency index; (Beta = .21, p<.001) for the transformed time lost index. Each of the 95% confidence intervals excluded zero: (.14 g B 5 .32) for the untransformed frequency index; (.16 5 B IA .47) for the untransformed time lost index, (.07 < B IA .19) for the transformed frequency index, and (.08 IA B S .22) for the transformed time lost index. Thus, hypothesis 5 was pp; supported. Ghiselli et al. (1981) recommend that correction for attenuation be made when assessing the relationship between 143 factors in construct validity assessment. An objective of construct validity assessment is to estimate the theoretical relationship between two constructs that are imperfectly measured (Nunnally, 1978). Corrected correlations for the untransformed measures were calculated. For the frequency index, the correlation was ; (428) = .82. Thus, absence intentions accounted for 67% of the variance in involuntary absenteeism after corrections were made for measurement unreliability. For the time lost index, the correlation was r (428) = .81. Thus, absence intentions accounted for 66% of the variance in involuntary absenteeism after corrections were made for measurement unreliability. 144 Table 25 Regression Results for Involuntary Absenteeism Predicted from Absence Intentions Criterion Ba R2 Involuntary-frequency (not transformed) .24* .05 Involuntary-time lost * (not transformed) .20 .04 Involuntary-frequency (transformed) .22* .05 Involuntary-time lost (transformed) .21* .04 aB=Beta, the standardized regression coefficient. *p<.001. Hypothesis 6 Hypothesis 6 examines whether absence intentions mediates the relationship between attitudes and norms about absenteeism, and voluntary absence behavior. Specifically, when the effect of absence intentions is controlled, attitudes favoring absenteeism and subjective norms favoring absenteeism will not predict voluntary absenteeism. Both attitudes favoring absenteeism (; (438) = .33, p<.001) and subjective norms favoring absenteeism (E (438) = .33, p<.001) were significantly related to absence intentions. As in hypotheses 4 and 5, absence intentions significantly predicted voluntary and involuntary absence behavior. Based on the discussion of the absence distributions, analyses will be presented for both untransformed and 145 transformed distributions. In addition, because intentions were significantly related to measures of involuntary absenteeism, results for analyses using these measures will also be presented. Table 26 shows the results of the regression analyses. To assess whether absence intentions is a mediating factor, it was entered first into the equation followed by attitudes and subjective norms, respectively. The results varied based on the type of measure and whether the measure was transformed. This hypothesis was supported for the untransformed frequency and time lost indexes of voluntary absenteeism. The respective partial correlations were statistically significant: frequency index (p; (438) = .07, p<.05); time lost index (p; (438) = .07, p<.05). Although the attitude regression coefficients for the untransformed voluntary- frequency measure (Beta = .12, p< .02) and the untransformed voluntary time-lost measure (Beta = .12, p<.01) were statistically significant, their respective 95% confidence intervals contained zero. Thus, when the effect of absence intentions was controlled, attitudes favoring absenteeism and subjective norms did not predict voluntary absence behavior, but absence intentions did. The attitudes and subjective norms regression coefficients were not statistically significant for both transformed and untransformed indexes of involuntary 146 absenteeism with one exception. The subjective norms regression coefficient for the untransformed involuntary- time lost index was statistically significant (Beta = .01, p< .03). The 95% confidence interval, however, contained zero. Thus, when the effect of absence intentions was controlled, attitudes favoring absenteeism and subjective norms did not predict involuntary absence behavior. Table 26 Regression Results for Testing Absence Intentions as a Mediating Factor Between Attitudes, Norms and Behavior Predictor Ba R2 (change) Criterion: Voluntary-frequency (not transformed) Absence intentions .08 .01 Attitude favoring absenteeism .12** .01 Subjective norms favoring absenteeism -.02 .00 R2 . 02 Criterion: VOluntary-time lost (not transformed) Absence intentions .04 .01 Attitude favoring absenteeism .12*** .01 Subjective norms favoring absenteeism .00 .00 R2 . 02 (table continued on next page) 147 Table 26 (continued) Predictor B R2 (change) Criterion: Involuntary-frequency (not transformed) **** Absence intentions .26 .05 Attitude favoring absenteeism -.02 .05 Subjective norms favoring absenteeism -.09 .06 R2 .05 Criterion: Involuntary-frequency (transformed) Absence intentions .24**** .04 Attitude favoring absenteeism -.04 .00 Subjective norms favoring absenteeism -.06 .01 R2 . 05 Criterion: Involuntary-time lost (not transformed) **** Absence intentions .23 .04 Attitude favoring absenteeism -.02 .00 Subjective norms favoring absenteeism .01* .01 R2 . 05 148 Table 26 (continued) Predictor B R2 (change) Criterion: Involuntary-time lost (transformed) Absence intentions .23**** .04 Attitude favoring absenteeism -.03 .00 Subjective norms favoring absenteeism -.07 .00 R2 .04 aB=Beta, the standardized regression coefficient. p<.03. *::p<.02. ****p<.01. p<.001. 149 Hypothesis 7 Hypothesis 7 stated that job satisfaction will significantly predict (inversely) voluntary absenteeism. In other words, dissatisfied workers are likely to be absent more than satisfied workers. Based on the discussion of the absence distributions, analyses will be presented for both untransformed and transformed distributions. In addition, because intentions were significantly related to measures of involuntary absenteeism, results for analyses using these measures will also be presented. Table 27 shows the results of the regression analyses. None of the regression coefficients were statistically significant. Whether an individual is absent from the workplace is unrelated to his/her job satisfaction. Thus, hypothesis 7 was not supported. 150 Table 27 Regression Results for Vbluntary Absenteeism Predicted from Job Satisfaction Criterion Ba R2 Voluntary-frequency (not transformed) -.04 .00 Voluntary-frequency (transformed) -.06 .00 Voluntary-time lost (not transformed) -.04 .00 Voluntary-time lost (transformed) -.06 .00 Involuntary-frequency (not transformed) .00 .00 Involuntary-frequency (transformed) .00 .00 Involuntary-time lost (not transformed) .01 .00 Involuntary-time lost (transformed) .00 .00 aB=Beta, the standardized regression coefficient. Hypotggsis 8 Hypothesis 8 stated that the correlation between absence intentions and voluntary absenteeism will be greater than the correlation between job satisfaction and voluntary absenteeism. Because involuntary absenteeism was related to intentions, analyses for these measures will also be provided. Table 28 shows the correlation coefficients and t statistics. 151 The correlation between absence intentions and behavior was statistically significant for each measure of absenteeism: for voluntary-frequency (B (438) = .11, p< .05); for voluntary-time lost (p (438) = .09, p< .05); for involuntary-frequency (; (438) = .23, p< .001); for involuntary-time lost (p (438) = .19, p< .001). The correlation between job satisfaction and absence behavior for each measure, however, was not statistically significant. There were only significant differences comparing correlations between absence intentions-absence behavior and job satisfaction-involuntary absence behavior: for the involuntary frequency index (p (437) = 3.95, p<.05); for the involuntary time lost index (p (437) = 3.06, p<.05). As stated, hypothesis 8 was not supported, however, absence intentions is a significant predictor of voluntary absenteeism compared to job satisfaction. Thus, absence intentions is more strongly related to voluntary absenteeism than job satisfaction. 152 Table 28 T Tests Comparing Correlations Between Absence Intentions- Absence Behavior and Job Satisfaction—Absence Behavior Correlation Absence Intentions- Job Satisfaction- Absence Behavior Absence Behavior ta p VOluntary-frequency .11* -.04 1.18 ns VOluntary-time lost .09* -.04 .84 ns Involuntary-frequency .23** .00 3.95 .05 Involuntary-time lost .19* .01 3.06 .05 a g;=437. *p<.05. ** p<.001. Hypothesis 9 Hypothesis 9 stated that the perception of the likelihood of being able to attain a similar job in the area will predict voluntary absenteeism. Results will be presented for both transformed and untransformed measures of voluntary and involuntary absenteeism (table 29). None of the regression coefficients for the independent variable were statistically significant. These 153 results suggest that one’s perception of being able to attain a similar job in the area was unrelated to absence behavior. Thus, hypothesis 9 was not supported. Table 29 Regression Results for Absenteeism Predicted from Perception of Ability to Find a Job Criterion Ba R2 Voluntary-frequency (not transformed) -.05 .00 Voluntary-frequency (transformed) .00 .00 Voluntary-time lost (not transformed) -.05 .00 Voluntary-time lost (transformed) .00 .00 Involuntary-frequency (not transformed) .00 .00 Involuntary-frequency (transformed) -.02 .00 Involuntary-time lost (not transformed) .00 .00 Involuntary-time lost (transformed) -.01 .00 aB=Beta, the standardized regression coefficient. Hypgthesis 10 Hypothesis 10 stated that the correlation between absence intentions and voluntary absenteeism will be greater than the correlation between voluntary absenteeism and perceptions of the likelihood of securing a similar job in 154 the area. As discussed earlier, the correlations between absence intentions and absence behavior were statistically significant. The correlation between one's perception of finding a similar job and absence behavior for each measure, however, was not statistically significant. Table 30 shows the results. There were only significant differences comparing correlations between absence intentions-absence behavior and finding a job-involuntary absence behavior: for the involuntary frequency index (p (437) = 3.59, p<.05): for the involuntary time lost index (p (437) = 2.78, p<.05). As stated, hypothesis 10 was not supported, however, absence intentions is a significant predictor of voluntary absenteeism compared to one's belief about being able to find a similar job. Thus, absence intentions is more strongly related to voluntary absenteeism than to one's perceptions of finding a similar job in the area. 155 Table 30 T Tests Comparing Correlations Between Absence Intentions- Absence Behavior and Finding a Job—Absence Behavior Correlation Absence Intentions- Finding a Job- Absence Behavior Absence Behavior ta p Voluntary-frequency .11 -.05 .92 ns Voluntary-time lost .09* -.05 .61 ns Involuntary-frequency .23** .00 3.59 .05 Involuntary-time lost .19* -.01 2.78 .05 a g;=437. *p<.05. **p<.001. Hypothesis 11 Hypothesis 11 stated that the use of an absence control policy will moderate (i.e., lower) the relationship between job satisfaction and absence intentions. This hypothesis could not be tested because there were no differences in absence control policies across the locations studied. 156 Hypothesis 12 Hypothesis 12 stated that kinship responsibilities will predict voluntary absenteeism. In other words, individuals whose kinship responsibilities are high will exhibit more voluntary absenteeism than individuals whose kinship responsibilities are low. Results will be presented for untransformed and transformed measures of absence behavior (table 31). The regression weights for the measures of voluntary absenteeism only were statistically significant: untransformed voluntary-frequency measure (Beta = .10, p< .02): untransformed voluntary-time lost measure (Beta = .11, p< .01); transformed voluntary-frequency measure (Beta = .11, p< .01); transformed voluntary-time lost measure (Beta = .12, p< .01). The 95% confidence interval for each of the voluntary measures excluded zero: untransformed voluntary-frequency measure (.035 B 5.37): untransformed voluntary-time lost measure (.045 B 5.46): transformed voluntary-frequency measure (.025 B 5.22); transformed voluntary-time lost measure (.025 B 5.24). Thus, hypothesis 12 was supported. 157 Table 31 Regression Results for Absenteeism Predicted from Kinship Responsibilities Criterion Ba R2 Voluntary-frequency * (not transformed) .10 .01 Voluntary-frequency (transformed) .11** .01 Voluntary-time lost (not transformed) .11** .01 Voluntary-time lost (transformed) .12** .01 Involuntary-frequency (not transformed) -.08 .01 Involuntary-frequency (transformed) -.07 .00 Involuntary-time lost (not transformed) -.07 .00 Involuntary—time lost (transformed) -.07 .00 aB=Beta, the standardized regression coefficient. :p<.02. * p<.01. Wis—1.; Hypothesis 13 stated that the correlation between absence intentions and voluntary absenteeism will be greater than the correlation between kinship responsibilities and voluntary absenteeism. Because involuntary absenteeism was related to intentions, analyses for these measures will also be provided. Table 32 shows the correlation coefficients 158 and t statistics. As discussed earlier, the correlation between absence intentions and behavior was statistically significant for each measure of absenteeism. For both voluntary indexes of absenteeism, the magnitude of the coefficients between kinship responsibilities and absence behavior were identical and statistically significant (p (438) = .11, p<.01). Individuals with greater kinship responsibilities will have more frequent and longer durations of voluntary absenteeism than individuals with lesser kinship responsibilities. For the involuntary frequency index, the relationship between kinship responsibilities and absence behavior was statistically significant but not in the expected direction (1 (438) = -.08, p<.05). The same relationship for the involuntary time lost index was not statistically significant. There were only significant differences comparing correlations between absence intentions-absence behavior and kinship responsibilities-involuntary absence behavior: for the involuntary frequency index (5 (437) = 2.32, p<.05): for the involuntary time lost index (5 (437) = 1.84, p<.05). Hypothesis 13 was not supported, however, absence intentions was significantly related to voluntary absenteeism. In this case, though, kinship responsibilities were related similarly to voluntary absence behavior as absence intentions. 159 Table 32 T Tests Comparing Correlations Between Absence Intentions- Absence Behavior and Kinship Responsibilities-Absence Behavior Correlation Absence Intentions- Kinship Responsibilities- Absence Behavior Absence Behavior ta p VOluntary-frequency .11 .11 0.00 ns Voluntary-time lost .09* .11* -.30 ns Involuntary-frequency .23** -.08* 2.32 .05 Involuntary-time lost .19* -.07 1.84 .05 a g;=437. *p<.05. ** p<.001. Hypothesis 14 Hypothesis 14 stated that job satisfaction will not predict voluntary absenteeism when the effect of job involvement is controlled. In other words, the regression coefficient for job satisfaction should not reach statistical significance. Based on the discussion of the absence distributions, analyses will be presented for both untransformed and transformed distributions. In addition, 160 because job involvement was significantly related to measures of involuntary absenteeism, results for analyses using these measures will also be presented. Table 33 shows the results of the regression analyses. To assess whether job involvement is a mediating factor, it was entered first into the equation followed by job satisfaction. This hypothesis was supported for the transformed and untransformed indexes of voluntary absenteeism. Specifically, after the effect of job involvement was controlled, job satisfaction did not significantly predict voluntary measures of absenteeism, but was significantly related to job involvement (5 (438) = .58, p<.001). With the exception of both transformed voluntary measures, the regression weights for job involvement were statistically significant: for the untransformed voluntary- frequency measure (Beta = -.14, p<.01); for the untransformed voluntary-time lost measure (Beta = -.12, p<.03); for the untransformed involuntary-frequency measure (Beta = -.17, p<.005); for the transformed involuntary- frequency measure (Beta = -.14, p<.01); for the untransformed involuntary-time lost measure (Beta= -.16, p<.005); for the transformed involuntary-time lost measure (Beta = -.16, p<.01). The 95% confidence intervals for these coefficients excluded zero. Thus, hypothesis 14 was supported. 161 Table 33 Regression Results for Testing Job Involvement as a Mediating Factor Between Job Satisfaction and Absence Behavior Predictor Ba R2 (change) Job Job Job Job Job Job Job Job Criterion: VOluntary-frequency (not transformed) involvement -.14** .01 satisfaction .04 .00 R2 . 01 Criterion : Voluntary-frequency (trans formed) involvement -.10 .01 satisfaction .00 .00 R2 . 01 Criterion: VOluntary—time lost (not transformed) involvement -.12* .01 satisfaction .03 .00 R2 . o1 Criterion: VOluntary-time lost (transformed) involvement -.09 .01 satisfaction -.01 .00 R2 .01 162 Table 33 (continued) Predictor B R2 (change) Criterion: Involuntary-frequency (not transformed) Job involvement -.17*** .02 Job satisfaction .10 .00 R2 .02 Criterion: Involuntary-frequency (transformed) Job involvement -.14** .01 Job satisfaction .07 .00 R2 .01 Criterion: Involuntary-time lost (not transformed) *** Job involvement -.16 .02 Job satisfaction .10 .00 R2 .02 Criterion: Involuntary-time lost (transformed) Job involvement -.16** .02 Job satisfaction .08 .00 R2 .02 aB=Beta, the standardized regression coefficient. **p<.03. ***p<.01. p<.005. 163 Hypothesis 15 Hypothesis 15 stated that organizational commitment will predict (inversely) voluntary absenteeism. In other words, individuals whose commitment to the organization is high will have less voluntary absenteeism than individuals whose commitment to the organization is low. In addition, because intentions were significantly related to measures of involuntary absenteeism, results for analyses using these measures will be presented. Table 34 shows the results. Table 34 Regression Results for VOluntary Absenteeism Predicted from Organizational Commitment Criterion Ba R2 Voluntary-frequency (not transformed) .00 .00 Voluntary-frequency (transformed) .00 .00 Voluntary-time lost (not transformed) .00 .00 Voluntary-time lost (transformed) -.02 .00 Involuntary-frequency (not transformed) .00 .00 Involuntary-frequency (transformed) .00 .00 Involuntary-time lost (not transformed) .00 .00 Involuntary-time lost (transformed) -.03 .00 §B=Beta, the standardized regression coefficient. 164 None of the regression coefficients were statistically significant. Thus, hypothesis 15 was not supported. Hypothesis 16 Hypothesis 16 stated that the correlation between absence intentions and voluntary absenteeism will be greater than the correlation between organizational commitment and voluntary absenteeism. Because involuntary absenteeism was related to intentions, analyses for these measures will also be provided. Table 35 shows the correlation coefficients and t statistics. As discussed earlier, the correlation between absence intentions and behavior was statistically significant for each measure of absenteeism. Consistent with the results of hypothesis 15, none of the correlations between organizational commitment and absence behavior were statistically significant. There were only significant differences comparing correlations between absence intentions-absence behavior and organizational commitment-involuntary absence behavior: for the involuntary frequency index (5 (437) 3.99, p<.05); for the involuntary time lost index (5 (437) 2.58, p<.05). As stated, hypothesis 16 was not supported. Nevertheless, given that organizational commitment and voluntary absenteeism are unrelated, and that absence intentions and voluntary absenteeism are significantly related, absence intentions is a stronger predictor of absence behavior than 165 organizational commitment. Table 35 T Tests Comparing Correlations Between Absence Intentions- Absence Behavior and Organizational Commitment- Absence Behavior Correlation Absence Intentions— Organizational Commitment- Absence Behavior Absence Behavior ta p V01untary-frequency .11 .00 1.53 ns VOluntary-time lost .09* .02 1.19 ns Involuntary-frequency .23** .00 3.99 .05 Involuntary—time lost .19* -.04 2.58 .05 9*g;=437. p<.05. **p<.001. Hypothesis 17 Hypothesis 17 stated that age will be inversely related to voluntary absenteeism. In other words, older individuals will have less absenteeism than younger individuals. In addition, because intentions were significantly related to measures of involuntary absenteeism, results for analyses using these measures will also be presented. The 166 correlations between age and absenteeism for each measure were statistically significant: for voluntary-frequency (; (438) = -17, p<.001); for voluntary-time lost (; (438) = - .14, p<.002); for involuntary-frequency (; (438) = -.16, p<.001), for involuntary-time lost (5 (438) = -.12, p<.008). Thus, hypothesis 17 was supported. Hypothesis 18 Hypothesis 18 stated that organizational commitment will significantly predict voluntary absenteeism when the effect of age is controlled. Based on the discussion of the absence distributions, analyses will be presented for distributions that were not transformed and distributions that were transformed. In addition, because age was significantly related to measures of involuntary absenteeism, results for analyses using these measures will also be presented. Table 36 shows the results of the regression analyses. To assess whether age is spuriously related to absenteeism, age was entered first into the equation followed by organizational commitment. This hypothesis was not supported for any of the absence measures. After the effect of age was controlled, organizational commitment was not related to absenteeism. The regression coefficients for age were statistically significant in all analyses. In addition, the confidence intervals for age excluded zero. Thus, these results suggest that age is a direct predictor 167 of absenteeism rather than a covariate as hypothesized in Chapter Three. Hypothesis 18 was not supported. Table 36 Regression Results for Testing Organizational Commitment as a Direct Predictor of and Absence Behavior Predictor Ba R2 (change) Criterion: Voluntary-frequency (not transformed) Age -.17*** .03 Organizational commitment .02 .00 R2 .03 Criterion: VOluntary-frequency (transformed) *** Age -.18 .03 Organizational commitment .00 .00 R2 .03 Criterion: Voluntary-time lost (not transformed) Age -.15** .02 Organizational commitment .03 .00 R2 .02 Criterion: Voluntary-time lost (transformed) _ 17*** Age .03 Organizational commitment .00 .00 R2 .03 168 Table 36 (continued) Predictor B R2 (change) Criterion: Involuntary-frequency (not transformed) Age -.16*** .03 Organizational commitment .01 .00 R2 .03 Criterion: Involuntary-frequency (transformed) Age -.17*** .03 Organizational commitment .01 .00 R2 .03 Criterion: Involuntary-time lost (not transformed) * Age -.11 .01 Organizational commitment -.02 .00 R2 .01 Criterion: Involuntary-time lost (transformed) Age -.15*** .02 Organizational commitment -.01 .00 R2 .02 a B=Beta, the standardized regression coefficient. p<.01. p<.002. * ** **p<.001. * 169 Hypothesis 19 Hypothesis stated that kinship responsibilities (positively) and primary income producer role (inversely) will each be correlated with voluntary absenteeism when the effect of sex is controlled. Table 37 shows the zero-order correlations between kinship responsibilities and the four measures of absenteeism. It also contains the zero-order correlations between primary income producer role and the four measures of voluntary absenteeism. As noted in the discussion of hypothesis 13 results, kinship responsibilities were significantly related to voluntary absenteeism. It also contains the partial correlations between primary income producer role and the four measures of voluntary absenteeism controlling for sex. Sex was related to the frequency index of voluntary absenteeism (p (420) = -.08, p<.10), the time lost index of voluntary absenteeism (p (420) = -.08, p<.10), and the frequency index of involuntary absenteeism (p (420) = -.08, p<.10). In other words, females exhibited greater voluntary absenteeism than males. In addition, whether one is the primary income producer for their family is unrelated to absence behavior. 170 Table 37 Zero-Order Correlations: Kinship Responsibilities and Primary Income Producer Role with Measures of Absence Behavior V01.- V01.- Invol.- Invol.- Freq. T.L. Freq. T.L. Kinship * ** * responsibilities .11 * .11 -.08 -.07 Primary income producer role .00 .02 .00 -.04 *:p<.05. p<.Ol. Table 38 shows the partial correlations between kinship responsibilities and the four measures of absenteeism controlling for sex. When the effect of sex was controlled, the partial correlation coefficients for kinship responsibilities remained statistically significant. There was also no change in magnitude of the coefficients when the effect of sex was controlled: for voluntary-frequency measure (pp (438) .11, p<.01); for voluntary-time lost measure (pp (438) .11, p<.01); for involuntary-frequency measure (pp (438) -.08, p<.05). For the partial correlations between primary income producer role and absence behavior, there was a slight reduction in magnitude. The partial correlations for these variables remained statistically insignificant. Hypothesis 19 was supported for kinship responsibilities, but not for 171 primary income earner role. Table 38 Partial Correlations: Kinship Responsibilities and Primary Income Producer Role with Measures of Absence Behavior Controlling for Employee Sex VOl.- VOl.— Invol.- Invol.- Freq. T.L. Freq. T.L. Kinship ** * responsibilities .11** .11 -.08 -.07 Primary income producer role .02 .04 -.03 -.07 *p<.05. **B<.o1. Hypothesis 20 Hypothesis 20 stated that work ethic will predict (inversely) voluntary absenteeism. In other words, individuals whose work ethic is high will have less absenteeism than those whose work ethic is low. Because intentions were significantly related to measures of involuntary absenteeism, results for analyses using these measures will also be presented. Table 39 shows the results of the regression analyses. The results suggest that work ethic was not related to any of the measures of absence behavior. Thus hypothesis 20 was not supported. 172 Table 39 Regression Results for VOluntary Absenteeism Predicted from WOrk Ethic Criterion Ba R2 Voluntary-frequency (not transformed) -.03 .00 Voluntary-frequency (transformed) -.03 .00 Voluntary-time lost (not transformed) -.03 .00 Voluntary-time lost (transformed) -.03 .00 Involuntary-frequency (not transformed) -.03 .00 Involuntary-frequency (transformed) -.02 .00 Involuntary-time lost (not transformed) -.05 .00 Involuntary-time lost (transformed) -.04 .00 aB=Beta, the standardized regression coefficient. Hypothesis 21 Hypothesis 21 stated that the correlation between absence intentions and voluntary absenteeism will be greater than the correlation between work ethic and voluntary absenteeism. Because involuntary absenteeism was related to intentions, analyses for these measures will also be provided. Table 40 shows the correlation coefficients and t statistics. 173 As discussed earlier, the correlation between absence intentions and behavior was statistically significant for each measure of absenteeism. Consistent with the results of hypothesis 20, none of the correlations between work ethic and absence behavior were statistically significant. There were only significant differences comparing correlations between absence intentions-absence behavior and work ethic-involuntary absence behavior: for the involuntary frequency index (5 (437) 3.22, p<.05); for the involuntary time lost index (5 (437) = 2.23, p<.05). As stated, hypothesis 21 was not supported. Nevertheless, given that work ethic and voluntary absenteeism are unrelated, and that absence intentions and voluntary absenteeism are significantly related, absence intentions is a stronger predictor of behavior than work ethic. 174 Table 40 T Tests Comparing Correlations Between Absence Intentions- Absence Behavior and Werk Ethic-Absence Behavior Correlation Absence Intentions— Work Ethic- Absence Behavior Absence Behavior ta p Voluntary-frequency .11* -.03 1.26 ns VOluntary—time lost .09 -.O3 .63 nS Involuntary-frequency .23** -.03 3.22 .05 Involuntary-time lost .19* -.05 2.23 .05 a g;=437. *p<.05. ** p<.001. Hypothesis 22 Hypothesis 22 stated that the correlation coefficient squared for the job satisfaction-voluntary absenteeism relationship would exceed 4% after corrections were made for measurement unreliability and adjustments made for distributional properties. Hackett and Guion's (1985) findings that the average interperiod correlation coefficient was .51 for the frequency index and .66 for the 175 time lost index were used as an estimate for stability for the absence measures. The uncorrected correlation for the frequency index was -.04, pg). The correlation for the frequency index was (r (I -.06) after the absence measure was transformed. After that correlation was corrected for unreliability, the corrected correlation for the frequency index was (p=-.09). When squared, job satisfaction accounted for less than 1% of the variance in voluntary absence frequency. The uncorrected correlation for the time lost index was (5: -.04, pg). The correlation for the time lost index was (; = - .07) after the absence measure was transformed. After that correlation was corrected for measurement unreliability, the corrected correlation for the frequency index was (;= -.09). When squared, job satisfaction accounted for less than 1% of the variance in time lost. Thus, neither measurement unreliability nor disparate marginal distributions accounted for weak effects. Hypothesis 23 Hypothesis 23 stated that the correlation coefficient squared for the job involvement-voluntary absenteeism relationship would exceed 4% after corrections were made for measurement unreliability and adjustments made for distributional properties. Hackett and Guion’s (1985) findings that the average interperiod correlation coefficient was .51 for the frequency index and .66 for the 176 time lost index were used as an estimate for stability for the absence measures. The uncorrected correlation for the frequency index was (; = -.12, p<.05). The correlation was (p= —.11) after the absence measure was transformed. The corrected correlation for the frequency index was (; = -.18). When squared, job involvement accounted for 3% of the variance in voluntary absence frequency. The uncorrected correlation for the time lost index was (5 = —.11, p<.05). The correlation was (p= -.10) after the absence measure was transformed. The corrected correlation for the time lost index was (; = - .15). When squared, job involvement accounted for 2% of the variance in time lost. Thus, neither measurement unreliability nor disparate marginal distributions accounted for weak effects. Hypothesis 24 Hypothesis 24 stated that the correlation coefficient squared for the organizational commitment-voluntary absenteeism relationship would exceed 4% after corrections were made for measurement unreliability and adjustments made for distributional properties. Hackett and Guion’s (1985) findings that the average interperiod correlation coefficient was .51 for the frequency index and .66 for the time lost index were used as an estimate for stability for the absence measures. An assessment could not be made for the frequency index because the zero-order correlation 177 between organizational commitment and absence frequency was zero. The uncorrected correlation for the time lost index was (5 = -.04, pg). The correlation was (5: -.03) after the absence measure was transformed. The corrected correlation for the time lost index was ( = -.04). When squared, organizational commitment accounted for less than 1% of the variance in time lost. Thus, neither measurement unreliability nor disparate marginal distributions accounted for weak effects. Summary Table 41 shows compares the hypothesized relationships with the actual relationships between absence intentions and various attitudinal, demographic, and behavioral factors. Overall, a degree of construct validity was assessed for an absence intentions measure. Most importantly, item and scale analyses indicated that absence intention items formed a homogenous and internally consistent scale. In addition, attitudes favoring absenteeism and subjective norms predicted absence intentions which, in turn, was a significant predictor of voluntary absence behavior. A detailed summary and discussion of these results will be presented in Chapter Six. 178 Table 41 Comparison: Hypothesized vs. Actual Relationshipsa Between Absence Intentions and Studied Variables variable Hypothesized Actual Atpitudinal Absence attitudes + + Subjective norms favoring absenteeism + + Job involvement - - Job satisfaction - 0 Organizational commitment - 0 Work ethic - 0 Demographic Employee age 0 - Primary income earner - 0 Kinship responsibilities + + Bepgvioral Involuntary absenteeism 0 + Voluntary absenteeism + + 3.. represents a significant inverse relationship; + represents a significant positive relationship; 0 represents no significant relationship. CHAPTER SIX DISCUSSION Introduction This chapter presents a discussion of the results of this dissertation research. This discussion contains an overall summary of the findings presented in Chapter Five compared to claims presented in Chapter One, implications for research and theory, implications for practice, and, suggestions for future research. Summary Despite the high financial costs of voluntary absenteeism to organizations, there is more speculation rather than knowledge about absenteeism (Steers & Rhodes, 1984). Less than four percent of the variance in absence behavior has been explained by work-related attitudes (i.e., weak effects). The purpose of this study was to investigate why the relationships between work-related attitudes (e.g., job satisfaction) and voluntary absence behavior were low despite theoretical expectations. Three potential reasons for weak effects were introduced. The primary objective of this study was to examine whether cognitions about being absent preceded absenteeism. 179 180 Also, were specific attitudes and specific norms about absence behavior more efficient predictors of absenteeism than general work attitudes (e.g., job satisfaction)? The focus was on the construct validity assessment of an absence intentions measure. Second, are small effect sizes the result of low measurement reliability? Third, are small effect sizes the result of disparate marginal distributions? Overall, a sufficient degree of construct validity was assessed for an absence intentions measure. Absence intentions mediated the relationship between attitudes and norms about absenteeism, and absence behavior. Methodological and measurement deficiencies were not alternative explanations for weak effects in this sample. Specifically, the findings of this study indicate the following: 1. Item and scale analyses revealed that absence intentions was a homogenous and internally consistent measure. Similar analyses revealed that attitudes toward absenteeism were multi-factorial. Individuals perceived outcomes of absenteeism as either motivating or as deterrents. Each of these factors was internally consistent. Finally, subjective norms were multi- factorial, based on respondents' expectations of how work affiliates and nonwork affiliates would likely perceive absenteeism, respectively. Neither factor nor the overall set of items was internally consistent which 181 suggests that who considers absence as either acceptable or unacceptable varied across respondents. The marginal distribution of absence intentions and other attitudinal variables closely approximated normality. Unlike absence behavior which has been shown to be a low base rate behavior, absence intentions demonstrated to be more prevalent among individuals studied than absence behavior. Reasons for this finding will be discussed in the next section. The marginal distributions of absence behavior departed significantly from normality. Square root transformations adjusted these distributions to approximate normality and were similar to the marginal distributions of other continuous variables studied. Adjusting for disparate distributions, however, did not enhance the percentage of variance explained. There was a high degree of discriminability between involuntary and voluntary self-report reasons for absence behavior. There were no significant differences between blue collar-unskilled workers and white collar-skilled workers on subjective norms. On average, both perceived norms that favored absenteeism. Blue collar workers, however had greater absence intentions, and exhibited more 182 voluntary and involuntary absence behavior than white collar workers. Those differences were statistically significant. Finally, blue collar workers had attitudes that favored voluntary absenteeism while white collar workers had attitudes that did not favor voluntary absenteeism. 6. Consistent with Ajzen and Fishbein (1980), attitudes about absence behavior and subjective norms favoring absenteeism were significant predictors of absence intentions. Furthermore, absence intentions was a significant predictor of voluntary absence behavior. Contrary to expectations, absence intentions significantly predicted measures of involuntary absence behavior. 7. Consistent with Ajzen and Fishbein (1980), absence intentions mediated the relationship between attitudes favoring absenteeism and voluntary absence behavior. Absence intentions also mediated the relationship between subjective norms and absence behavior. 8. Absence intentions was a stronger predictor of voluntary absence behavior than either job satisfaction or organizational commitment. Job involvement, however, was an equally efficient predictor of voluntary absence behavior. 183 9. Using Hackett and Guion's (1985) estimates for absence stability, correcting for unreliability in the absence measures as well as the attitudinal factors did not enhance the percentage of variance explained. Implications for Research and Theory Chapter Two described many "theories" of absenteeism sans cognition about being absent. The results of this study suggest that voluntary absenteeism is preceded by cognitions about the behavior. Specifically, attitudes about motivating and deterrent outcomes of voluntary absenteeism and subjective norms about absenteeism were strongly related to absence intentions. Absence intentions, in turn, was a significant predictor of voluntary absence behavior. Absence intentions mediated the relationship between attitudes favoring absenteeism and voluntary absence behavior as well as the relationship between subjective norms and voluntary absence behavior. These findings are consistent with researchers' (Fichman, 1988; Hackett & Guion, 1985) assumptions that voluntary absenteeism implies volition. In particular, Fichman (1988) argued that individuals evaluate the cost and benefits of absenteeism. To the extent that the perceived benefits outweigh the perceived costs, an individual will likely be absent. Fichman (1988) found that individuals switched regularly between absence from work and attendance at work. Although the behavior was assumed to be preceded 184 by cognitions about the behavior, Fichman (1988) did not directly assess whether cognitions preceded the behavior nor did he consider whether individuals evaluate the costs and benefits of absenteeism. This research suggests that individuals not only evaluate the outcomes of absenteeism as benefits (i.e., motivating outcomes) or costs (i.e., deterrent outcomes), but their evaluations also lead to intentions that significantly predict absence behavior. Although intentions significantly predicted absence behavior, this study also included several other hypotheses that linked work-related attitudes and characteristics to absence behavior. The purpose was part of the construct validity assessment of an absence intentions measure based on rationale that these factors would relate to absence intentions in the same way they relate to absence behavior. These hypotheses were based on current thought in the literature and empirical evidence. Not all these hypotheses were supported. For example, organizational commitment and job satisfaction were unrelated to voluntary absenteeism. An explanation for these findings is presented. Much absenteeism research is based on the assumption that a typical response to job dissatisfaction and low organizational commitment is absenteeism (e.g., Mowday, et al., 1982). Absenteeism, however, represents only one potential response to these conditions. Furthermore, absenteeism is a specific instance of withdrawal behavior 185 that represents one category of an adaptive response. For example, an individual may exhibit other responses to job dissatisfaction and low organizational commitment including turnover (e.g., Marsh & Mannari, 1977), lower job inputs (e.g., Oldham, Kulik, Ambrose, Stepina, & Brand, 1986); and, aggression and overt violence (Spector, 1978). Each measure of the behavior is likely to be characterized by a large amount of specific variance and a small amount of variance common to other measures of the same construct. Thus, one would not necessarily expect a strong relationship between job dissatisfaction or low organizational commitment and voluntary absenteeism. Given this rationale, the lack of support for these hypotheses does not necessarily provide evidence against the construct validity of an absence intentions measure. As expected, involuntary absence behavior and voluntary absence behavior were empirically unrelated. However, cognitions about voluntary absenteeism also were significantly related to involuntary measures of absence behavior. This finding is counterintuitive because one would not expect an individual’s cognition about absenteeism to be related to absence behavior that is beyond one's control. There are three plausible explanations for this finding. First, Hackett and Guion (1985) indicated that the voluntary-involuntary absence dichotomy may be artificial 186 and misleading. They suggested that the type of absence be conceptualized on a continuum of justifiability. For example, should an absence that is associated with a body temperature of 99° be labeled as involuntary or voluntary? Also, would the classification apply to all individuals? In virtually all absenteeism research, the reason for each absence occurrence is based on each absentee's self-report. Even though the majority of involuntary absenteeism in this study was attributed to sickness, perhaps the illness was not sufficiently serious to prevent an individual from attending work. In addition, some individuals may have claimed sickness as a legitimate excuse to justify an absence for other reasons that would not be acceptable to the employer (Hill & Trist, 1953). Thus, not all involuntary absenteeism was necessarily precipitated by factors beyond one’s control. The second explanation for intentions relating to involuntary absence behavior is based on external factors. Ilgen (1977) and Hammer and Landau (1981) suggested that specifying whether the cause underlying absenteeism is voluntary or involuntary is a difficult task. A cause may depend upon any of the following factors: (a) the organization’s absence policy, (b) the employee’s absence history, (c) the grievance history of the organization, (d) negotiations between employee and management, (e) political factors, and (f) rules used by the clerk who records absence 187 events (Atkin & Goodman, 1984). Factor (a) noted above is especially relevant here. The sponsoring organization provides each employee with a number of paid sick days per calendar year. An employee forfeits the right to the allotted paid sick days granted during a particular year if they are not used during the period of allotment. In other words, paid sick days do not "carry over" into the next calendar year. Thus, employees will likely be motivated to be absent because of illness in order to receive the benefits to which they are entitled. It is possible that some employees planned to use their entitled benefits. In this study, employee absenteeism was monitored mainly during November and December. It is possible that many of the involuntary absence occurrences identified in this sample were the response of some employees to claim their entitled benefit. Furthermore, because November and December are holiday months, it is possible that employees planned in advance to use some sick days for voluntary reasons such as holiday shopping with friends, or family gatherings. The third explanation for intentions relating to involuntary behavior is based on the psychological contract (Schein, 1972) concept. Schein (1972, p. 77) defines the psychological contract as the degree to which the employee's expectations of what the organization will provide him/her 188 and what he/she owes the organization match what the organization's expectations are of what it will give and get. To the extent that an employee perceives that the employer has breached the contract, he/she may justify using paid sick days for other reasons to restore a sense of equity in the employment relationship. Clearly, this research has provided evidence that undermines the construct validity of widely used, purported measures of voluntary and involuntary absenteeism. To the extent that involuntary absenteeism is the result of factors beyond one's personal control, then one’s attitudes and intentions about absenteeism that is within one's control should not be related to involuntary behavior. Researchers should no longer speculate about the psychological characteristics of absence behavior based on individual's self-report reasons. Instead, researchers should refer to the objective characteristics of the behavior. In this case, measures of "voluntary" absenteeism are absences that are neither sanctioned nor paid by the organization. Measures of "involuntary" absenteeism are absences that are sanctioned gpg paid by the organization. This study has provided support for the idea that voluntary absence behavior is preceded by intentions. This study also examined whether methodological and measurement shortcomings may be responsible for weak effects. Although the marginal distributions of the behavioral variables were 189 nonnormal and dissimilar to the attitudinal factors, square root transformations did not enhance the effect size between various work-related attitudes and absence behavior. In addition, correcting for measurement unreliability did not enhance the magnitude of the effect size estimates. Thus, based on this sample, neither nonnormal and disparate marginal distributions of absence behavior nor measurement unreliability was responsible for weak effects. Although absence intentions significantly predicted both voluntary and involuntary measures of employee absence behavior, the overall percentage of variance explained in behavior by intentions was 1.5% for the voluntary measures and 5.0% for the involuntary measures. Clearly the percentage of variance explained was not enhanced by the use of absence intentions. However, in this instance, the concept of variance explanation is misleading. Consistent with Abelson's (1985) argument, percent variance explained is a misleading index of the influence of systematic factors (in this case, intentions) when small influences (indexed by relatively small percentage of variance explained) cumulate to produce meaningful outcomes. Specifically, Mirvis and Lawler (1977) estimated that the cost of absenteeism to organizations was $26.4 billion in 1977. Thus, to the extent only 5% of total absenteeism is under normative and attitudinal control, then savings to organizations will be $1.32 billion annually. When this 190 dollar estimate is adjusted to reflect increases in wage and salary levels, the effect will be more pronounced. Despite the low percentage of variance explained in absence behavior by attitudes and subjective norms, the impact of intentions was significant in this sample. Specifically, for the three-month period examined in this study, individuals who intended to be absent were absent 22.60 times more often than individuals who did not intend to be absent. For the three-month period examined in this study, individuals who intended to be absent were absent 16.19 times more days than individuals who did not intend to be absent. Alternative Explanations The results of this dissertation provide a sufficient degree of support for the construct validity of an absence intentions measure. The findings suggest that voluntary absenteeism is preceded by cognitions about the behavior, however, potential alternative explanations to these findings will be discussed. Although intentions significantly predicted absence behavior, not all the variance in absence behavior was explained by absence intentions. Ajzen and Fishbein (1980) specified several conditions that may reduce the magnitude of the intention-behavior relationship. In this study, two conditions are in question: (a) stability of intentions, and (b) control of intentions. 191 Ajzen and Fishbein (1980) found that the longer the time interval between measurement of intentions and observation of behavior, the greater the probability that the individual may obtain new information or that certain events will occur which will change his/her intentions. Intentions and behavior were assessed for a three-month period to enhance the likelihood that there would be sufficient variance in absence behavior within the sample. Of course, it is possible that intervening factors occurred after the completion of the intentions questions which may have changed actual behavior from what was intended. In the future, researchers should vary the period for which intentions and behavior are assessed to compare the predictive efficiency of intentions. Ajzen and Fishbein (1980) suggested that the relationship between intentions and behavior may be attenuated due to the occurrence of events outside one's control. For example, Smith (1977) found that despite a crippling snowstorm, work-related attitudes predicted attendance on the day after the storm when the company attendance policy was relaxed. For another office of the same company located in a city where there was no snowstorm, work attitudes did not predict attendance on the same day. Thus, although individuals may have intentions to be absent, they will necessarily not be absent because of loss of pay or discipline by their supervisor. 192 In this sample for voluntary or unpaid absences, 68% of the sample who had intentions to be absent behaved in a manner that was inconsistent with their intentions while less than 1% of the sample who had no intentions to be absent behaved in a manner that was inconsistent with their intentions. For the involuntary or paid absences, 42% of the sample who had intentions to be absent behaved in a manner that was inconsistent with their intentions while 3% of the sample who had no intentions to be absent behaved in a manner that was inconsistent with their intentions. For both types of absences, these data lend indirect support for the stability and control issues. Finally, it was hypothesized that blue collar workers would have subjective norms that significantly favor voluntary absenteeism compared with white collar workers. On average, both groups had norms favoring absenteeism, however, the difference between groups was not significant. The discussion by Nicholson and Johns (1985) referred to blue collar-unskilled workers and white collar- professionals. In this study, blue collar-unskilled workers were included, but the white collar sample consisted of clerical employees. Clearly, the occupation range represents a continuum from unskilled to skilled, professional along which differences in norms likely exist. This study did not represent both ends of the continuum to which Nicholson and Johns (1985) were referring. Thus, the 193 range in norms studied was restricted, which may account for the nonsignificant difference in norms about voluntary absenteeism. Implications for Practice The high costs of absenteeism (Steers & Rhodes, 1984) as well as the deleterious effects of absenteeism on organizations (Goodman & Atkin, 1984) warrant the need for research that will shed light on ways to more effectively manage employee absenteeism. This dissertation provided support for the idea that absenteeism is under attitudinal and normative control. There are two key implications of these findings. The first implication deals with the psychological contract (Schein, 1972) and the explicit employment agreement between employee and employer. The second implication deals with structuring employee attitudes about absenteeism based on management and union expectations through conventional human resource management (HRM) practices. Each will be addressed in turn. The Employpent Relapipnship In the existing agreement between General Motors Corporation and the United Auto Workers (October 8, 1987), both parties recognized the principle that benefits are generated, earned, and funded by regular employment. One factor that determines an employee's entitlement to benefits is regular attendance at work. Both management and union 194 have a vested interest in effectively managing controllable absenteeism. From management’s perspective, excessive controllable absenteeism that is not sanctioned by policy will be harmful to the business in terms of cost, quality, and efficiency. From the union’s perspective, excessive controllable absenteeism that is not sanctioned by policy will undermine the credibility of its leadership as well as place undue burdens on the grievance procedure. Thus, employees who experience controllable absences in excess of a specified percentage of available hours will forfeit their entitlement to maximum benefits. Simply stated, excessive absenteeism that is associated with choice and not sanctioned either by the collective bargaining agreement or by management policy is a breach of the explicit employment agreement and psychological contract between the employee and the employer. Although the sponsoring organization does not maintain a policy that penalizes employees who are absent excessively for reasons other than those sanctioned by the organization, it clearly communicates its expectations (similar to General Motor’s and the UAW's) on attendance to its employees. This policy, in conjunction with the findings of this study, suggest critical implications for the fidelity of the employee-employer relationship. This study revealed that an employee's evaluation of the costs and benefits of being absent from the workplace is 195 associated with absence behavior whereas an employee’s commitment to the organization or satisfaction on the job is not associated with absence behavior. In other words, some employees contemplate being absent. Despite management's expectations that an employee will not experience excessive absenteeism for reasons other than those sanctioned by policy, this study indicated that some employees were unfaithful to the employer by using "sick" days for reasons other than personal illness. Not only did employees use some of their "sick" days for reasons other than personal illness--a breach of the contract in itself, but they also failed to report the actual reason for these absences. Clearly, the trust underlying the employee-employer relationship was undermined. This research was not designed to study the dynamics of the employee-employer relationship. Insufficient information is available to blame any one party as an instigator. Thus, finger-pointing would be unjustified. Attitude Structuring with HRM Practices While the first implication paints a negative picture of the employee-employer relationship, the results of this study suggest that management and unions may work jointly toward structuring employee attitudes. Specifically, in order to influence behavior, management should provide information to employees that will produce changes in their beliefs. Information about management's expectations on 196 absenteeism and attendance can be incorporated into many conventional human resource management practices. Management and union expectations about attendance and absenteeism can be incorporated into realistic job previews (Wanous, 1980) during the recruitment process. The purpose of the realistic job preview is to facilitate a dual matching process between (a) the applicant’s capabilities and job requirements, and (b) the individual applicant’s needs and the organizational requirements (Wanous, 1980). To the extent that management and union expectations about absenteeism and attendance are not consistent with the applicant's goals and expectations, then the applicant will likely self-select him/herself out of the situation. To the extent that management's expectations are consistent with the applicant’s goals and expectations, then the applicant will likely be more committed to the choice than to one that is made with external pressure. Consistent with the rationale for the use of realistic job previews, management and the union should incorporate information about absenteeism into the orientation program for new employees. Incorporating this information into orientation programs will reinforce management and union expectations about attendance and absenteeism. In addition, employees who might not have had realistic job previews with information on absenteeism should be included into the orientation program. 197 Absenteeism and attendance should be incorporated into measures of employee performance. Incorporating absenteeism and attendance in performance appraisal will likely focus an employee’s attention on that behavior. Not only does incorporating absenteeism in performance appraisal indicate that the level of behavior is important to the organization, but it is also likely to enhance acceptance of legitimate authority (Katz & Kahn, 1966). Management’s expectations about attendance and absenteeism may be reinforced on a regular basis through employee supervisors and union stewards. Supervisors represent the employer to most employees. Union stewards represent the union to most employees. Thus, counseling supervisors and union stewards to model union and management expectations about absenteeism and attendance may be effective. For example, Wexley and Nemeroff (1975) found that a decrease in absenteeism resulted from goal setting and feedback regarding role-playing exercises about different employees who wanted to take vacation at the same time, employees who were abusing coffee breaks, denying overtime to subordinates who insisted on it, and getting agreement among subordinates on how to maximize profits. Finally, this research suggested that individuals intend to be absent because of family responsibilities as well as stressful work situations. With this in mind, voluntary absenteeism may be reduced by offering employees 198 flexible work schedules. Research has shown that flexible work schedules have been effective in reducing voluntary absenteeism (Chadwick-Jones, et al., 1982). Stress management training should also be offered so that employees can use simple techniques, such as muscle relaxation, during stressful moments at work rather than being absent to cope with the stress. Lastly, research which examines the effectiveness of day care programs should be conducted in order to assess their impact on absenteeism. Suggestions for Future Research The results of this study have demonstrated, to a sufficient degree, the efficacy of absence intentions in understanding absence behavior. Although it has shed light on key issues in the absenteeism literature and about the employee-employer relationship, it has raised many more questions. Several suggestions for future research are offered. First, the results of this study should not be generalized to other occupations or organizations. Future research examining absence intentions in organizations located in different industrial sectors, organizations with different absence management policies, organizations with a unionized workforce, and assessment of intentions over varying time spans should be conducted. Second, because the link between absence intentions and absence behavior was established, policy-capturing research 199 may be a fruitful area of pursuit. Specifically, policy capturing research would provide insight into the cues and relative importance of cues that account for each employee's intention to be absent. In addition, policy capturing would allow comparison of employee's policy equations. This comparison will provide insight into whether the decision to be absent varies across employees. If so, how each employee's decision differs can be understood by analyzing the relative importance attributed to each cue. Third, the relationship between absence intentions and purported measures of involuntary absenteeism raises questions about what constitutes involuntary absenteeism. As discussed earlier, Hackett and Guion (1985) suggested that the voluntary-involuntary absence dichotomy may be artificial. Perhaps future research should focus on the justifiability of absenteeism. Rather than studying a large number of employees over a long time span, fewer employees should be studied for a shorter time span with the focus on assessing how each person justifies a particular absence immediately after its occurrence. This dissertation studied absence behavior as the only reaction to absence intentions. Organizational policy may remove much of the control an employee has over actual absenteeism. Perhaps future research should incorporate other behavioral responses to absence intentions which may more likely be under an employee’s control. One example 200 would be reduced job inputs. Finally, this study revealed that employment relationships may be characterized by infidelity. Some employees breached the employee-employer contract (explicit and psychological). Future research is needed to investigate what factors are likely to result in a breach of the contract and by whom. LIST OF REFERENCES Abelson, R. P. (1985). A variance explanation paradox: .When a little is a lot. Psychological Bulletin, 21, 129-133. Ajzen, I., & Fishbein, M. (1977). Attitude-behavior relations: A theoretical analysis and review of empirical research. Psychological Bulletin, B5, 888- 918. Ajzen, I., & Fishbein, M. (1980). Hnderstandipg attitpggs and predicting social behavior. Englewood Cliffs, NJ: Prentice-Hall. Argyle, M. (1972). The social psychology of work. Harmondsworth, Middlesex: Penguin. Argyle, M., Gardner, G., & Cioffi, F. (1958). Supervisory methods related to productivity, absenteeism, and labour turnover. Human Relations, ll, 23-40. Atkin, R. S., & Goodman, P. S. (1984). Methods of defining and measuring absenteeism. In P. S. Goodman, R. S. Atkin and Associates (Eds.), Abseptegism (pp. 47-109). San Francisco: Jossey-Bass Publishers. Barnett, V., & Lewis, T. (1978). Outliers in statistigal data. New York: Wiley. Baum, J. F., & Menefee, M. L. (1979). An experimental study of operant conditioning and absenteeism. Paper presented at the 39th annual meeting of the Academy of Management, Atlanta, GA. Baumgartel, H., & Sobol, R. (1959). Background and organizational factors in absenteeism. Bersopngl Bsychglogy, _1_2, 431-443 . Beatty, R. W., & Beatty, J. R. (1975). Longitudinal study of absenteeism of hard-core unemployed. Psychological Reports, BB, 395-406. Beehr, T. A., & Gupta, N. (1978). A note on the structure of employee withdrawal. Organizational Behavior gnd Human EgLformance, Bl, 73-79. Behrend, H. (1953). Absence and labour turnover in a changing economic climate. Occupational Bsychology, B1, 69-79. Blood, M. R. (1969). Work values and job satisfaction. Joupnal gf Applied Psycpology, 5;, 456-459. 201 202 Blumberg, M. (1980). Job switching in autonomous work groups: An exploratory study in a Pennsylvania coal mine. Academy of Management Journal, BB, 287-306. Box, G. E. P., & Cox, D. R. (1964). An analysis of transformations. Journal of thggRoval Statistical Society. Series B, BB, 211-252. Brayfield, A., & Crockett, W. (1955). Employee attitudes and employee performance. Ps cholo ical Bulletin, 52, 396- 424. Breaugh, J. A. (1981). Predicting absenteeism from prior absenteeism and work attitudes. Journal of Applieg Bgvcholoqv, BB, 555-560. Brooke Jr., P. P. (1986). Beyond the Steers and Rhodes model of employee attendance. Academy of Management Review, ll, 345-361. Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validation by the Multitrait-Multimethod Matrix. Psychological Bulletin, BB, 81-105. Chadwick-Jones, J. K., Brown, C., Nicholson, N., & Sheppard, C. (1971). Absence measures: Their reliability and stability in an industrial setting. Bersonnel Bsychology, BB, 463-470. Chadwick-Jones, J. K., Nicholson, N., a Brown, C. (1982). §ocial psychology of absenteeism. New York: Praeger. Cheloha, R. S., & Farr, J. L. (1980). Absenteeism, job involvement, and job satisfaction in an organizational setting. Bournal of Applied Psychology, BB, 467-473. Clegg, C. W. (1983). Psychology of employee lateness, absence, and Turnover: A methodological critique and an empirical study. Journal of Applied Psychology, BB, 88- 101. Cohen, J. (1977). Btatistical powet analysis for the pgpavigpal ggiences (Rev. ed.). New York: Academic Press . Cohen, J., & Cohen, P. (1983). Applied multiplg zggtgggiopzcorrelatign analysis to; the behaviopal ggienges. Hillsdale, NJ: Lawrence Erlbaum Associates. Cook, R. D. (1977). Detection of influential observations in linear regression. Iechnomgtrics, l2, 15-18. 203 Cook, R. D., & Weisberg, S. (1982). Hgsigpals and inflpence ip tggtgssion. New York: Chapman & Hall. Cook, T. D., & Campbell, D. T. (1979). Quasi- gxpetimentation; Design & gnalysis issugs for field sgttings. Boston: Houghton Mifflin. Cooper, R., & Payne, R. (1965). Age and absence: A longitudinal study in three firms. Occupational Bsychology, BB, 31-43. Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. Bsychological Bulletin, 5;, 281-302. Crowther, J. (1957). Absence and turnover in the divisions of one company-1950-55. Occu ational s ho o , Bl, 256-270. Durand, V. M. (1985). Employee absenteeism: A selective review of antecedents and consequences. Journal of Qrggnizational Behavior Management, 1, 135-167. Farrell, D., & Stamm, C. L. (1988). Meta-analysis of the correlates of employee absence. Human Relations, Bl, 211-227. Feldman, J. (1974). Race, economic class, and the intention to work: Some normative and attitudinal correlates. Journal pf Appligd nggpology, BB, 179-186. Fichman, M. (1984). A theoretical approach to understanding employee absence. In. P. S. Goodman, R. S. Atkin and Associates (Eds.), Abse t e' , (pp. 1-46). San Francisco: Jossey-Bass Publishers. Fichman, M. (1988). Motivational consequences of absence and attendance: Proportional hazard estimation of a dynamic motivation model. Journal A lied Ps cholo , 1B, 119-134. L/Eishbein, M. (1967). Attitude and the prediction of “”' behavior. In M. Fishbein (Ed.), Readings in attitude tpggty apd measurement. New York: Wiley. ~Fishbein, M., & Ajzen, I. (1975). Belief, gttitude, ~v intgntion ang behavior; An ipttoduction tg theoty gpg peseatcp. Reading, MA: Addison-Wesley. Ford, R. N. (1969). Hotivatiop thtough tpe yotk itself. New York: American Management Association. 204 Fox, J. B., & Scott, J. R. (1943). Hpsenteeisp: Mapagement’s problem. Graduate School of Business Administration, Harvard University. Froggatt, P. (1970). Short-term absence from industry: I. Literature, definitions, data, and the effect of age and length of service. Britisp Bgurnal of Industrial Hedicine, B1, 199-210. Garrison, K. R., & Muchinsky, P. M. (1977). Evaluating the concept of absenteeism-proneness with two measures of absence. Personnel Bsychology, BB, 389-393. Gaudet, F. J. (1963). Bolving tpg problems of employgg absence. New York: American Management Association. Gerhart, B. (1987). How important are dispositional factors as determinants of job satisfaction? Implications for job design and other personnel programs. Journal of A lied Ps cholo , 1B, 366-373. Ghiselli, E. E., Campbell, J. P., & Zedeck, S. (1981). Measurement thgorv for thg behavioral gcignceg. San Francisco: W. H. Freeman and Company. Gibson, R. O. (1966). Toward a conceptualization of absence behavior of personnel in organizations. Administrativg Science Quarterly, ll, 107-133. Goodale, J. G. (1973). Effects of personal background and training on work values of the hard-core unemployed. Bournal of Applied Bsychology, B1, 1-9. Goodman, P. S., & Atkin, R. S. (1984). Effects of absenteeism on individuals and organizations. In P. S. Goodman, R. S. Atkin and Associates (Eds.), Absenteeism (pp. 276-321). San Francisco: Jossey-Bass Publishers. Gouldner, A. W. (1958). Cosmopolitans and locals: Toward an analysis of latent social roles: I. Administrativg Bgienge Quapterly, BB, 444-480. Graen, G. B. (1969). Instrumentality theory of work motivation: Some experimental results and suggested modifications. Bgurnal of Applied Psychology, BB, 1-25. Grove, B. A. (1968). Attendance reward plan pays. Persgppel Boutnal, Bl, 119-120. Guilford, J. P. (1954). Psychometric mgthods (2nd ed.). New York: McGraw-Hill. 205 Hackett, R. D., & Guion, R. M. (1985). A reevaluation of the absenteeism-job satisfaction relationship. nggnizational Behavio; gpd Human Decision Processes, 35. 340-381. Hackman, J. R., & Lawler, E. E., (1971). Employee reactions to job characteristics. ngrnal of Applied Psychology, 55, 259-286. Hackman, J. R., & Oldham, G. R. (1975). Development of the Job Diagnostic Survey. Boutnal of Applied Psychglogy, 59, 159-170. Hammer, T. H., & Landau, J. C. (1981). Methodological issues in the use of absence data. Bournal of Applied Psychology, BB, 574-581. Hammer, T. H., Landau, J. C., & Stern, R. N. (1981). Absenteeism when workers have a voice: The case of employee ownership. Bourngl of Applied Psycholggy, BB, 561-573. Harvey, B. H., Rogers, J., & Schultze, J. A. (1983). Sick, pay versus well pay: An analysis of the impact of rewarding employees for being on the job. Public Bersonnel Management Journal, lB, 218-224. Hays, W. L. (1981). Statigticg, (3rd ed.). New York: Holt, Rinehart and Winston. Hedges, J. N. (1973). Absence from work: A look at some national data. Monthly ngg: Review, 2B, 24-31. Herman, J. B. (1973). Are situational contingencies limiting job attitude-job performance relationships? Organizational Behavior and Human Performance, lB, 208- 224. Herzberg, F., Mausner, B., & Snyderman, B. S. (1959). The potiygtigp tg wotk. New York: Wiley. Hill, J. M., & Trist, E. L. (1953). A consideration of industrial accidents as a means of withdrawal from the work situation. Human Relations, B, 357-380. Huse, E. F., & Taylor, E. K. (1962). The reliability of absence measures. Journal of A lied Ps 010 , 1B, 159-160. Ilgen, D. R. (1977). Attendance behavior: A reevaluation of Latham and Pursell's conclusions. Bournal of Appligg 25321191292. i2: 230-233- 206 Ilgen, D. R., & Hollenback, J. H. (1977). The role of job satisfaction in absence behavior. Organizationgl Behavior and Human Performgpge, lg, 148-161. Johns, G. (1980). Attitudinal and nonattitudinal predictors of two forms of absence from work. Organizatiopgl ngaviot and Human Performapcg, BB, 431-444. Johns, G., & Nicholson, N. (1982). The meanings of absence: New strategies for theory and research. In B. M. Staw & L. L. Cummings (Eds.), Research in Organizational Behavior, 5, (pp. 127-172). Greenwich, CT: JAI Press. Kaiser, H. F. (1958). The varimax criterion for analytic rotation in factor analysis. Psygpopettikg, BB, 187- 200. Katz, D., & Kahn, R. L. (1966). The social psychology of organizations. New York: Wiley. Keller, R. T. (1983). Predicting absenteeism from prior absenteeism, attitudinal factors and nonattitudinal factors. Journal of A lied Ps cholo , BB, 536-540. Kempen, R. W., & Hall, R. V. (1977). Reduction of industrial absenteeism: Results of a behavioral approach. Joutpal pf Organizgtional Behaviot Hanagepent, 1, 1-21. Kendall, M. G., & Stuart, A. (1958). The a v nced theo of gtatistics. New York: Hafner. Kent, H. M., Malott, R. W., & Greening, M. (1977). Improving attendance at work in a volunteer food co-operative with a token economy. Journal of Organizational Hanaggment, 1, 89-98. Kuzmits, F. E. (1981). The impact of a legalistic control policy on selected measures of absenteeism behavior. In R. H. Kilmann & M. Jelinek (Eds.), The Bgstern Academy of Hgnggemept Ptocegdings of the Eighteenth Annual Meeting, Binghamton, NY. Landy, F. J., Vasey, J. J., & Smith, F. D. (1984). Methodological problems and strategies in predicting absence. In P. S. Goodman, R. S. Atkin and Associates (Eds.), Abse teeism, (pp. 110-157). San Francisco: Jossey Bass. Lewin, K. (1938). contribution to psychological theoty. Durham, NC: Duke University Press. 207 Locke, E. A. (1983). The nature and causes of job satisfaction. In M. D. Dunnette (Ed.), Handpook of lndusttial apg opganizatiopgl psychology, (pp. 1297- 1349). Chicago: Rand McNally. Lodahl, T. M., & Kejner, M. (1965). The definition and measurement of job involvement. Journal of Applied Bsychology, 52, 24-33. Mann, F. C., Indik, B. P., & Vroom, V. H. (1963). Ipe prgdpgtivity o; wgrk groups. Ann Arbor, MI: Survey Research Center, Institute for Social Research, University of Michigan. March, J. G., & Simon, H. (1958). Opganizations. New York: Wiley. Markham, S. E. (1985). An investigation of the relationship between unemployment and absenteeism: A multi-level approach. Academy of Management Journal, BB, 228-234. Marsh, R., & Mannari, H. (1977). Organizational commitment and turnover: A predictive study. Administrative Sgience Quarterly, BB, 57-75. Martin, J. (1971). Some aspects of absence in a light engineering factory. Occupational Bsycpology, BB, 77- 91. Martocchio, J. J. (1987). Age-related differences in employee absenteeism: A meta-analytic review. Unpublished manuscript, Michigan State University. Mayo, E., & Lombard, G. (1944). Teamwork and labor turpover in tpe aircraft industty in southern California. Cambridge, MA: Harvard University Press. Messick, S. (1975). The standard problem: Meaning and values in measurement and evaluation. Apetican Bgygpologist, B9, 955-966. Mirvis, P. H., & Lawler, E. E. III (1977). Measuring the financial impact of employee attitudes. Bournal pf Bppligd Psychology, BB, 1-8. Mitchell, T. R. (1974). Expectancy models of job satisfaction, occupational preference, and effort: A theoretical, methodological, and empirical appraisal. Win. 81. 1053-1077- 208 Mobley, W. H. (1977). Intermediate linkages in the relationship between job satisfaction and employee turnover. Bourpal of Applied Psychology, BB, 237-240. Morgan, L. G., & Herman, J. B. (1976). Perceived consequences of absenteeism. Bournal of Applied Bsychology, Bl, 738-742. Mowday, R. T., Porter, L. W., & Steers, R. M. (1982). lo ee-or 'o in e : he s holo o commitment- absenteeism and turnover. New York: Academic Press. Mowday, R. T., Steers, R. M., 8 Porter, L. W. (1979). The measurement of organizational commitment. Journal of Vocational Behavio , lB, 224-247. Muchinsky, P. (1977). Employee absenteeism: A review of the literature. Journal of Vocational Behavior, lB, 316- 340. Naylor, J. C., & Vincent, N. L. (1959). Predicting female absenteeism. Pergonnel Psychology, lB, 81-84. Nicholson, N. (1976). Management sanctions and absence control. Human Relations, BB (2), 139-151. Nicholson, N. (1977). Absence behavior and attendance motivation: A conceptual synthesis. Bgurnal of Hgnaggment Studies, lA, 231-252. Nicholson, N., Brown, C. A., & Chadwick-Jones, J. K. (1976). Absence from work and job satisfaction. Bournal of Applied Psyghglogy, Bl, 728-737. Nicholson, N., Brown, C. A., & Chadwick-Jones, J. K. (1977). Absence from work and personal characteristics. Boutnal pf Applied Psychology, BB, 319-327. Nicholson, N., & Goodge, P. M. (1976). The influence of social, organizational, and biographical factors on female absence. Journal of Hanagemept Studies, lB, 234- 254. Nicholson, N., & Johns, G. (1985). The absence culture and the psychological contract-Who's in control of absence? Agadepy of ngagement Review, lB, 397-407. Nunnally, J. C. (1978). Bsychomettic tpeoty. New York: McGraw-Hill. 209 Ofori-Dankwa, J. (1987). The effect of the psychological contract on employee job satisfaction, satisfactoriness, and tenure. Unpublished Doctoral Dissertation, School of Labor and Industrial Relations, Michigan State University. Oldham, G. R., Kulik, C. T., Ambrose, M. L., Stepina, L. P., 8 Brand, J. F. (1986). Relations between job facet comparisons and employee reactions. Organizational Behavior and Human Decisiop Brgcesses, BB, 28-47. Orpen, C. (1978). Effects of bonuses for attendance on the absenteeism of industrial workers. Journal of Organizational Behavio; Management, l, 118-124. Panyan, S. W., 8 McGregor, M. (1976). How to implement a proactive incentive plan: A field study. Persopnel Journal, BB, 460-462. Patchen, M. (1960). Absence and employee feelings about fair treatment. Personnel Psychology, lB, 349-360. Porter, L. W., 8 Steers, R. M. (1973). Organizational, work, and personal factors in employee turnover and absenteeism. Ps cholo 'cal Bulletin, BB, 151-176. Porter, L. W., Steers, R. M., Mowday, R. T., 8 Boulian, P. V. (1974). Organizational commitment, job satisfaction, and turnover among psychiatric technicians. Boutnal of Applied Psychology, BB, 603-609. Rhodes, 8. R. (1983). Age—related differences in work attitudes and behavior: A review and conceptual analysis. Psychological Bulletin, BB, 328-367. Rokeach, M. (1973). Ihe nature of human values. New York: Free Press. Rosse, J. G., 8 Hulin, C. L. (1985). Adaptation to work: An analysis of employee health, withdrawal, and change. Qrggnigational Behavior and Human Decision Btocesses, 15. 324-347. Rousseau, D. M. (1978). Relationship of work to nonwork. Bgurnal of Applied Psychology, BB, 513-517. Saal, F. E. (1978). Job involvement: A multivariate approach. Joutnal of Applied Psychology, BB, 53-61. 210 Salancik, G. R. (1977). Commitment and the control of organizational behavior and belief. In B. M. Staw, 8 G. R. Salancik (Eds.), New Bipggtions in Otganizatiopal Bepavior. Chicago: St. Clair Press. Salancik, G. R., 8 Pfeffer, J. (1978). A social information processing approach to job attitudes and task design. Administrative Science Quapterly, BB, 224-253. Schein, E. H. (1972). Or ani onal s cholo . Englewood Cliffs, N.J.: Prentice Hall. Schein, E. H. (1978). Career d nam cs: Matchin i d'v'dua gnd grganizational needs. Reading, MA: Addison-Wesley. Schmitz, L. M., 8 Heneman III, H. G. (1980). Do positive reinforcement programs reduce employee absenteeism? Bersonnel Administrator, BB(9), 87-93. Schwab, D. P. (1980). Construct validity in organizational behavior. In B. M. Staw, 8 L. L. Cummings (Eds.), Hesearch in Organizational Behaviop, B, (pp. 3-43). Greenwich, CT: JAI Press. Schwab, D. P., 8 Heneman III, H. G. (1970). Aggregate and individual predictability of the Two-Factor theory of job satisfaction. Personnel Bsychology, BB, 55-66. Scott, K. D., 8 Taylor, G. S. (1985). An examination of conflicting findings on the relationship between job satisfaction and absenteeism: A meta-analysis. Aggggpy of Management Journal, BB, 599-612. Searls, D. J., Braucht, G. N., 8 Miskimins, R. W. (1974). Work values and the chronically unemployed. Journal of A lied Ps holo , BB, 93-95. Siegel, A. L., 8 Ruh, R. A. (1973). Job involvement, participation in decision making, personal background and job behavior. Organizational Behavio; and Human Performance, 9, 318-327. Simon, H. A. (1956). Rational choice and the structure of the environment. Psychological Review, BB, 129-138. Smith, F. J. (1977). Work attitudes as predictors of specific day attendance. Journal of Applied Psychology, BB, 16-19. Smith, P. C., Kendall, L. M., 8 Hulin, C. L. (1969). The peasupgment of satisfactiop in work and retirement. Chicago: Rand-McNally. 211 Spector, P. (1978). Organizational frustration: A model and review of the literature. Bgrsonnel Psychology, Bl, 815-829. Staw, B. M., 8 Oldham, G. R. (1978). Reconsidering our dependent variables: A critique and empirical study. Academy of Management Journal, Bl, 539-559. Staw, B. M., 8 Ross, J. (1985). Stability in the midst of change: A dispositional approach to job attitudes. Boutpal of Appligg Psychology, 1B, 469-480. Steel, R. P., 8 Ovalle, N. K. (1984). A review and meta- analysis of research on the relationship between behavioral intentions and employee turnover. Bourpal of Applied Psycpolggy, BB, 673-686. Steers, R. M. (1977). Antecedents and outcomes of organizational commitment. dm 'st a iv c Quattetly, BB, 46-56. Steers, R. M., 8 Mowday, R. (1981). Employee turnover and post-decision accommodation processes. In L. Cummings and B. Staw (Eds.), Besearch in organizational behavior, B, (pp. 235-281). Greenwich, CT: JAI Press. Steers, R. M., 8 Rhodes, S. (1978). Major influences on employee attendance: A process model. Journal of Applied Psychology, BB, 391-407. Steers, R. M., 8 Rhodes, S. (1984). Knowledge and speculation about absenteeism. In P. S. Goodman, R. S. Atkin, 8 Associates (Eds.), Absenteeism, (pp. 229-275). San Francisco: Jossey-Bass. Stephens, T. A., 8 Burroughs, W. A. (1978). An application of operant conditioning to absenteeism in a hospital setting. Journal of Applied Bsycpology, BB, 518-521. Stevens, J. P. (1984). Outliers and influential data points in regression analysis. Bsycpglogical Bulletin, BB, 334-344. Taylor, P. J. (1967). Shift and day work: A comparison of sickness absence, lateness, and other absence behaviour at an oil refinery from 1962 to 1965. British Joutpgl of lndpstrial Medicine, BA, 93-102. 212 Terborg, J. R., Lee, T. W., Smith, F. J., Davis, G. A., 8 Turbin, M. S. (1982). Extension of the Schmidt and Hunter validity generalization procedure to the prediction of absenteeism behavior from knowledge of job satisfaction and organizational commitment. Journal of Applied Esycpolggy B1, 440-449. \Triandis, H. (1977). Intetpersonal behavior. Monterey, CA: Brooks/Cole. Van Maanen, J., 8 Schein, E. H. (1979). Toward a theory of organizational socialization. In B. M. Staw (Ed.), Research in Orgapizational Behaviop, l, Greenwich, CT: JAI Press. Vroom, V. H. (1964). Work and motivation. New York: Wiley. Vroom, V. H. (1966). Organizational choice: A study of pre- and post-decision processes. Organizational Behavior and Human Performance, l, 212-225. Wanous, J. P. (1974). Individual differences and reactions to job characteristics. Journal of Applied Psyghology, BB, 616-622. Wanous, J. P. (1980). Organizational entry: Regruitmept selectionl and socialization of newcomers. Reading, MA: Addison-Wesley. Waters, L. K., 8 Roach, D. (1971). Relationship between attitudes and two forms of withdrawal from the work situation. Journal of Applied Psychology, BB, 92-94. Waters, L. K., 8 Roach, D. (1973). Job attitudes as predictors of termination and absenteeism: Consistency over time and across organizational units. Journgl pf A lied P ChOlO , 51, 341-342. Waters, L. K., 8 Roach, D. (1979). Job satisfaction, behavioral intention, and absenteeism as predictors of turnover. Personnel Psychology, BB, 393-397. Watson, C. J., Driver, R. W., 8 Watson, K. D. (1985). Methodological issues in absenteeism research: Multiple absence measures and multivariate normality. Academy og Management Review, lB, 577-586. Weisberg, S. (1980). Applied linear tegression. New York: Wiley. 213 Weiss, D., Dawis, R., England, G., 8 Lofquist, L (1967). Manual for Minnesota Satisfaction Questionnaire (No. 22). Minneapolis: University of Minnesota, Minnesota Studies in Vocational Rehabilitation. Wexley, K. N., 8 Nemeroff, W. F. (1975). Effectiveness of positive reinforcement and goal setting as methods of management development. Journal of A lied Ps cholo , BQ, 446-450. Wilensky, H. L. (1960). Work, careers and social integration. International Social Science Journal, lB, 543-560. Yolles, S. F., Carone, P. A., 8 Krinksy, L. W. (Eds.) (1975). Absenteeism in industty. Springfield, IL: Charles C. Thomas Publisher. Youngblood, S. A. (1984). Work, nonwork, and withdrawal. Bournal of Applied Psychology, BB, 106-117. Zedeck, S. (1977). An information processing model and approach to the study of motivation. Organizational Behavior and Human Performance, lB, 47-77. APPENDIX A Sectim 1: Jcb Involvement mas: The following statements are oatments people have made or might mkeabcuttheirwork. Pleaseconsidereadmstatanentandcirclethemmber whid1bostrepresentsfl1ee3¢enttowhid1ymagreeordisagreewiththe statement. 'meratingschanefortlmestatemmtsislistedbelw. DoNOI‘ resporxlinawayymthinkothersbelieveywsrmldbehave. YGJRRESPQISESWIIL mmmmymlmmmmmmmammon OUTSIDE‘IHECIJMPANY. 1= Strongly Agree . 5= Slightly Disagree 2: Agree 4=Ne1ther Agree nor 6= Disagree 3= Slightly Agree Disagree 7= Strongly Disagree 1. I’ll stay overtime to finish a job, even if I’m not paid for it. 1 2 3 4 5 6 7 2. chcnnmeasureapersmprettywellbyhowgoodajdaheorshedoos. 1 2 3 4 5 6 7 3. The major satisfaction in my life canes fran my job. 1 2 3 4 5 6 7 4. For me, mornings at work really fly by. 1 2 3 4 5 6 7 5. I usually show up for work a little early, to get things ready. 1 2 3 4 5 6 7 6. 'memostiuportantthingsthathagpentomeinvolvemywork. l 2 3 4 5 6 7 7. Sanetimas I lie awake at night thinking ahead to the next day’s work. 1 ' 2 3 4 5 6 7 8. I'm really a perfectionist about my work. 1 2 3 4 5 6 7 9. IfeeldeprossedwmenIfailatsanethingcormectedwithmyjob. l 2 3 4 5 6 7 214 215 1= Strongly Agree % Slightly Disagree 2= Agree 4=Neither Agree nor 6= Disagree 3= Slightly Agree Disagree 7= Stm1gly Disagree 10. I have other activities more inportant than my work. 11. 12. 13. 14. 15. 16. 17. 1 2 3 4 5 6 7 I live, eat, andbreathemyjob. 1 2 3 4 5 6 7 Iwould probablykeepworking even ifIdidn’tneedthenoney. l 2 3 4 5 6 7 QuiteoftenIfeel likestayingtnnefrmworkinsteadofcaningin. 1 2 3 4 5 6 7 'I'one,myworkisonlyasmallpartofwho1am. 1 2 3 4 5 6 7 Iamverynuchinvolved personallyinmywork. 1 2 3 4 5 6 7 I avoid taking on extra duties and responsibilities in my work. 1 2 3 4 5 6 7 Iusedtobemreanbitiwsabartmymrktmnlamnow. 1 2 3 4 5 6 7 18. Most things in life are more important than work. 19. 20. 1 2 3 4 5 6 7 Iusedtocaremmabmtnymrk,h1tmwotherthingsaremreiuportantto m. l 2 3 4 5 6 7 Sanetimas I’dliketokickmyself forthemistakesImakeinmywork. 1 2 3 4 5 6 7 216 Sectim 2: (nlpany (Imminent mamas: Listedbelowareammberofstatanentsthatrepresentpossible feelings that people might have about the cmpany for which they work. Think abouthowymfeelabtheneralElectricnow. Oonsidereachstatanentand cimleflemmberwhidibestmpresemsfleextammmmywagreeordisagree withthestatanerrt. 'Iteratingsdiemefortl'mestatementsislistedbelow. Do Norrespondinawayyouthinkothersbelieveyoushouldanswer. YCIJRRESPONSES WHLBEKEPPGIJHIINIIALANDYGRIWWHLMBEREVEAIEDTOANYWEW ORCIJTSIIE‘H'IECIIIPANY. 1= Strongly Agree . 5= Slightly Disagree 2= Agree 4=Ne1ther Agree nor 6= Disagree 3: Slightly Agree Disagree 7= Strongly Disagree 1. Iamwillingtopatinagreatdealofeffortbeyondthatnormallyexpected inordertohelpthisccmpanybesucoassful. 1 2 3 4 5 6 7 2. Italkupthiscaupanytomyfriendsasagreatorganizationtoworkfor. 1 2 3 4 5 6 7 3. Ifeelverylittle loyaltytothisccmpany. 1 2 3 4 5 6 7 4. Iwouldaooeptalmostanytypeofjobassigrmerrtinordertokeepworkingfor thisccmpany. l 2 3 4 5 6 7 5. I find thatmyvaluesandthisccmpany's valuesarevery similar. 1 2 3 4 5 6 7 6. IamprouitotellothersthatIampar‘tofthiscarpany. l 2 3 4 5 6 7 7. Icmldjustaswellbemrkingforadifferentccmpanyaslongasthetype of workwas similar. 1 2 3 4 5 6 7 8. ‘Itfiscmpanyreallyinspirestheverybestinmeintlewayofjob performance. 1 2 3 4 5 6 7 217 1= Stmmgly Agree . 5= Slightly Disagree 2: Agree 4=Ne1ther Agree nor 6= Disagree 3= Slightly Agree Disagree 7= Strongly Disagree 9. Itmfldtakeverylittledxangeinmypresentcimmetamostocameneto leavethiscaupany. 1 2 3 4 5 6 7 10. IamextranelygladthatIdmsefluiscmpanytomrkforoverotherstas oasidering atthetime I joined. 1 2 3 4 5 6 7 11. There’s not toomuch tobe gainedby sticking with this oarpany indefinitely. 1 2 3 4 5 6 7 12. Often, I find it difficmlt to agree with this catpany’s policies on important matters relating to its employees . 1 2 3 4 5 6 7 13. Ireallycareaboutthefateofthiscatpany. 1 2 3 4 5 6 7 14. Formethisisthebestof all possible cmpanies forwhichtowork. 1 2 3 4 5 6 7 15. Decidingtoworkforthisoarpanywasadefinitemistakeonmypart. 1 2 3 4 5 6 7 l6. Ineversayanythingbadaboutmyself. 1 2 3 4 5 6 7 17. IamalwayshomstineverythingIdo. 1 2 3 4 5 6 7 18. 'Iherehavebeenoccasions onwhich I felt jealous of otherpeople. l 2 3 4 5 6 7 19. Icanranalbertimeswheanasatbarrassed. l 2 3 4 5 6 7 218 Sectim 3: Jd) Satisfactim mas:'niep.uposeoffl1esestatatentsistogiveymad1anoetotell howyoufeelaboutyourpresentjob, Matthingsyouaresatisfiedwitharriwhat things you are pot satisfied with. Read each statement carefully. Circle the mmzberwhichbestrepresentshmymfeel. Behonest. YCIJRRESPCNSESEBE KEPTGNFIDENITALANDYGRIDEWPTTYWELMBEREVEMEDTOANYGTEWTBENGZ OUTSIIE'IHECIMPANY. l= Ebctranely Satisfied 5= Slightly Dissatisfied 2= Satisfied 4= I can' t decide 6= Dissatisfied 3= Slightly Satisfied whether I am 7= Extremely Dissatisfied satisfiedornot Onmypresentjob,thisishowaeelabout: 1. Beingabletokeepmsyallthetime. 1 2 3 4 5 6 7 2. Thechancetoworkalone (irdeperrierrtly)mthejob. 1 2 3 4 5 6 7 3. Thediarcetododifferentthingsfruntimetotime. 1 2 3 4 5 6 7 4. 'mediancetobe"satebody"intheoam1mity. 1 2 3 4 5 6 7 5. The way my immediate supervisor handles his or her workers. 1 2 3 4 5 6 7 6. The ompetenoe of my supervisor innaking decisions. 1 2 3 4 5 6 7 7. Beingabletodothingsthatdon’tgoagainstmyconscience. 1 2 3 4 5 6 7 8. The way my job provides for steady employment. 1 2 3 4 5 6 7 9. 'mediancetodothirgsforotherpeople. 1 2 3 4 5 6 7 2: 3: Extremely Satisfied Satisfied Slightly Satisfied 219 4= I can't decide whetherIam 5= Slightly Dissatisfied 6= Dissatisfied 7= Extremely Dissatisfied satisfiedornot 10. 'Ihecharcetotellpeqalewhattodo. 1 2 3 4 5 6 7 11.'med1ancetodosarethingthatmakesuseofmyabilities. 1 2 3 4 5 6 7 12. Thewaycmpanypoliciesareput intopractioe. 1 2 3 4 5 6 7 13.Mypayarritheam1ntofmrkIdo. 1 2 3 4 5 6 7 14.Thed1amesforadvanoementmthisjob. l 2 3 4 5 6 7 15. Thefreedantousemyownjtfigment. 1 2 3 4 5 6 7 16.Thed1ancetotrymyommethodsofdoingthejob. 1 2 3 4 5 6 7 17. Theworkingconditions. 1 2 3 4 5 6 7 18. 'mewaymyoo-mrkersgetalongwitheachother. 1 2 3 4 5 6 7 19.1hepraiseIgetfordoingagoodjob. 1 2 3 4 5 6 7 20. The feeling of acoalplislmerrt I get frunthe job. 1 2 3 4 5 6 7 220 Sectim 4: Wad: Ethic mars: 'mefollowing statanerrtsareabouthowpeople might feelabout workingeneral. Pleasereadeach statanent carefully. Circlethemmberwhich bestdescribeswhetheryouagreeordisagreewitheadistatenent. Behonest. YMRESIUEESWELBEKEPPWIALANDYGRWWHLMEEREVEALED 'IUANYCNEWI'IHINORCUISIIE'IHECIMPANY. 1= Strongly Agree . 5= Slightly Disagree 2= Agree 4=Neither Agree nor 6= Disagree 3= Slightly Agree Disagree 7= Strongly Disagree 1. When the workday is finished, a person should forget his/her job and enjoy him/terself. 1 2 3 4 5 6 7 2. Hardworkmaloesaniniividualabetterperson. 1 2 3 4 5 6 7 3. Theprincipalpurposeofaperson’s job istoprovidehim/herwiththeneans for enjoying his/her free time. 1 2 3 4 5 6 7 4. Wastingtimeisasbadaswastingmey. 1 2 3 4 5 6 7 5. Merever possible, a person should relax and accept life as it is, rather than always striving for unreachable goals . l 2 3 4 5 6 7 6. A good indication of a person's worth is how well he/she does his/her job. 1 2 3 4 5 6 7 7. Ifallotherthingsareequal, itisbettertohaveajobwithalotof responsibility than ore with little responsibility. 1 2 3 4 5 6 7 8. Peoplewho "do things the easyway" are better off. 1 2 3 4 5 6 7 221 Sectims: mmmssirgm mas: Please read each statement carefully. Then, for each statement, circle the nunberwhidi represents your beliefs aboutmissing work. Be cmpletelyhonest. DoNOI'respondinawayymthirfltothersbelieveymstnild behave. YGJRRESPCNSESWIILBEKEPTCQIFIDENIIALANDYGJRIDENTITYWIILMBE REVEAIED'IOANYCNEWI‘IHINORCIJISIIE'IHECD‘IPANY. Forquestions 1thraigh7, please ratethechancethatyouwouldmissworkduring the next three months (excluding your holidays, vacation, illness, civic duties) on the following scales: 1= Very likely Miss Work 5= Smaahat Unlikely Miss Work 2= Likely Miss Work 4=Neither Likely 6= Unlikely Miss Work 3= W Likely Nor Unlikely 7= Very Unlikely Miss Work Miss Work 1. to take care of your dependent children in need (for example, an ill child. If you do not have diildren, leave blank) 1 2 3 4 5 6 7 2. when you feel like doing a hobby or leisure activity (NOT VDRK-REIATED MIX-RELATED, 0R WHY-RELATED) 1 2 3 4 5 6 7 3. attending to family responsibilities other than children 1 2 3 4 5 6 7 4. to fulfill religious ocmnitnerrts 1 2 3 4 5 6 7 5 . involvalerrt in cammity activities other than religion, family, or work 1 2 3 4 5 6 7 6. to spend leisure time with my family 1 2 3 4 5 6 7 7. to get relief from a dissatisfying work situation 1 2 3 4 5 6 7 222 8. I'll probably miss sale days from work during the rext three months (excluding holidays, vacation, illness, civic duties). 1 2 3 4 5 6 7 W AGREE SLIQ-II'LY NEITHER SLIGH'IX DISAGREE MK AGREE Am DISAQIEE DISAQEE 9. Pleaseestimatehmmenytimesymeiqaecttomissmrkdurirgtrerextthree months (excluding holidays, vacation, illness, civic duties). times 10. Please estimate the number of times you will likely slm up for work during the next three months based on total work days (excluding holidays, vacation, illness, civic duties). times 11. Doyoufeel astrorgobligationtoshowup formrkduringtherextthree months (excluding holidays, vacation, illness, civic duties)? 1 2 3 4 5 6 7 W AGREE SLIGH'LY NEITHER SLIGHTLY DISAGIEE W AGREE AGREE DISAGREE DISAGIEE 12. Oreofmygoalsdurirgtlerexttlmeemrmthsistohaveperfectatterdame at work (excluding holidays, vacation, illness, civic duties). 1 2 3 4 5 6 7 9mm AGREE SLIGHT! NEITHER SLIGHT! DISAGREE W AGREE AQEE DIME DISAGREE 223 Sectim6:lwao.IrTimeIsSpart mas. Please consider the eight activities below. Before respa'ding, takeafevmmentstoflrinkabatymrlifeingeremlardlnvtlnseactivities arepartofatypicalweek. Foreadistatement,circlethemmberwhid1 indicateshowoftenyouareinvolvedineadiactivity. YQJR RFSPCNSESEBE KEPPWALANDYGJRWNMBEREVEAIHDTOANYQIEWCR ansmnmmm. 1=NEVER 5=NGVAND1HEN 2=HARDLYEVER 4= AKIJTHAIF 6=USUAIIX 3=DUI'OFI'EN THETIME 7=AINAYS 1. Taking care of your dependent diildren in need (for example, an ill child. If you do not have children, leave blank). 1 2 3 4 5 6 7 2. Attending to family responsibilities other than children 1 2 3 4 5 6 7 3. Enjoying hobbies and leisure time mr related to your job, family, calamity activities, or religion 1 2 3 4 5 6 7 4. Spending time attending to work-related activities (doing your job or other supportive activities) 1 2 3 4 5 6 7 5. Fulfilling religious catmitments 1 2 3 4 5 6 7 6 . Involvement in catmmity activities other than religion, family, or work 1 2 3 4 5 6 7 7. Sperding leisure time with your family 1 2 3 4 5 6 7 8 . Taking time away frcm a dissatisfying work situation (excluding holidays, vacation) 1 2 3 4 5 6 7 224 Sectim 7: Attitnie Toward MissirgWork PartA 118M108. Please consider eachpairof words below. For eachpairof words, circletlenmberwhidirepresentshavyoufeelabartymrmissingwork (excluding illness, holidays, civic duties, vacation). Men answering think abortrmyoufeel,mthowymratployerorothersvmldfeel. Pleasebe honest. Domrarewereadiitaninawaythatyoubelieveotherswouldlikeyou toanswer. YCIJRRESRIGSESWHLBEKEPTCINFMITALANDYCIJRIDENTMEEQM BEREVEAIED‘IOANYCNEWI'IHENORQHSM'HEGMPANY. l. MESH 1 2 3 4 5 6 7 WISE 2. 0000 1 2 3 4 5 6 7 BAD 3. HAMUL 1 2 3 4 5 6 7 ENEFICIAL 4. HJNISHENG 1 2 3 4 5 6 7 REWARDING 5. FAVIRABIE 1 2 3 4 5 6 7 MW 6.VAIUABIE 1 2 3 4 5 6 7 MESS 225 m was: The following 15 items are outcomes of missing work. Sale are positive and sale are negative. For each statement, circle the number which irriicatestnwlikelyormilikelyeadiaitcanemayhappentoyouifyoumisswork. Then, cimlettemmberwhichrepresentshowyoufeelabcuteadioutoare. DoNOI' answereachiteminawaythatymbelieveotherswuildlikeywtoanswer. YCIJR mmmmmmmmmmmmmmm ANYWEWI’DHNGQGJTSIIE'IHEGMPANY. 1 2 3 4 5 6 7 Higily Likely Satewhat Neither Somewhat Unlikely Higily likely Lj kely Likely Nor Unlikel' Y [ha-MY Unlikely 1 2 3 4 5 6 7 Very Good Somewhat Neither Satewhat Bad Very Good Good Good nor Bad Bad Bad 1 . time with friends 1 2 3 4 5 6 7 Highly Highly Likely Unlikely l 2 3 4 5 6 7 Very Very Good Bad 2 . time for family 1 2 3 4 5 6 7 Highly Highly Likely Unlikely 1 2 3 4 5 6 7 Very Very Good Bad 3 . discipline by supervisor 1 2 3 4 5 6 7 Highly Highly Likely Unlikely 1 2 3 4 5 6 7 Very Very Good Bad 4 . upset coworkers 1 2 3 4 5 6 7 Highly Highly Likely Unlikely 1 2 3 4 5 6 7 Very Very 226 2 3 4 5 6 7 Higfly Likely Somewhat Neither Satewhat Unlikely Higily Likely Likely Likely Nor Unlikely Unlilmly Unlikely 2 3 4 5 6 7 Very Good Sanewhat Neither Salewhat Bad Very Good Good Good nor Bad Bad Bad loss of pay 1 2 3 4 5 6 Highly Highly Likely Unlikely l 2 3 4 5 6 7 Very Very Good Bad time for myself to 1 2 3 4 5 6 do things I enjoy Highly Highly Likely Unlikely l 2 3 4 5 6 7 Very Very Good Bad relief from the job 2 3 4 5 6 Highly Highly Likely Unlikely 1 2 3 4 5 6 7 Very Very Good Bad denoted to a lower job 1 2 3 4 5 6 7 Highly Highly Likely Unlikely l 2 3 4 5 6 7 Very Very Good Bad chance to meet other 2 3 4 5 6 obligations Highly Highly Likely Unlikely l 2 3 4 5 6 7 Very Very 227 1 2 3 4 5 6 7 Higily Likely Saneahat Neither Satewhat Unlikely Higily Likely Likely Likely Nor Unlikely Utililmly Unlikely l 2 3 4 5 6 7 Very Good Salewhat Neither Sarewhat Bad Very Good Good Good nor Bad Bad Bad 10 . to get away from 1 2 3 4 5 6 7 co-workers Highly Highly Likely Unlikely 1 2 3 4 5 6 7 Very Very Good Bad 11 . loss of benefits 1 2 3 4 5 6 7 Highly Highly Likely Unlikely 1 2 3 4 5 6 7 Very Very Good Bad 12 . fired fran the outpany 1 2 3 4 5 6 7 Highly H}9hly Likely Unlikely 1 2 3 4 5 6 7 Very Very Good Bad 13 . loss of pranotion 1 2 3 4 5 6 7 opportunities Highly Highly Likely Unlikely 1 2 3 4 5 6 7 Very Very 1 2 3 Higily Likely Sanewhat LiJQly Likely 1 2 3 Very Good Sarewhat Good Good 15 . heavier workload when return to work 228 Neither Likely Nor Unlikely 4 Neither Good nor Bad 1 2 Highly Likely 1 2 Very Good 1 2 Highly Likely l 2 Very Good 5 Somewhat Unlikely 5 Somewhat Bad 3 4 3 4 3 4 3 4 6 7 Unlikely Higny Unlikely 6 7 Bad Very Hal 5 6 7 Highly Unlikely 5 6 7 Very Bad 5 6 7 Highly Unlikely 5 6 7 Very Bad 16. Doyouagreethat itwillbeeasyforyoutogetanotherfull-tinejoblike theoneyouhaveintown? 1 2 3 STRONGLY AGREE m AGREE 4 5 SLIG'ITLY NEI'n-IER SHE-III! DISAGZEE 6 7 DIME SWIM DISAGREE 229 Sectim8: HaaYai'minkOUersFeelflmYmMissm memos. Please read each statement carefully. Then, for each statelent, cimleflemmbermidibestrepresertshavymfliirfltotrerswfllfeelabortywr missingmrkdiringtherextthreewnths. Becaxpletelyhonest. YCIIRRESPQEFS WHLBEKEPTGNHDBHIALANDYQRIDENTTTYWHLMBEREVEAIEDTOANYQJEW GRCIJ'ISIIE'IHECIMPANY. 1= Strongly Agree . 5= Slightly Disagree 2= Agree 4=Ne1ther Agree nor 6= Disagree 3= Slightly Agree Disagree 7= Strongly Disagree Part A 1. My supervisor would prdoably be upset if I missed work (excluding illness, civic duties, holidays, vacations). 1 2 3 4 5 6 7 SIRCNGLY SHINGLY AGREE DIM 2. My co-workers would not care if I missed work (excluding illness, civic duties, holidays, vacations). 1 2 3 4 5 6 7 SIKNGLY W ACREE DISACREE 3. Mariners of my family (parents, brothers, sisters, or other relatives other than a spouse) mild prefer me to take sate time off from work (ea-(cluiing illness, civic duties, holidays, vacations). 1 2 3 4 5 6 7 W SMLY AGREE DISAGREE 4. My friends think it is o.k. for me to take time off from work (excluding illness, civic duties, holidays, vacations). 1 2 3 4 5 6 7 W W AGREE DISAC-REE 5. My spouse or significant other would be upset if I missed work (excluding illress, civic duties , holidays, vacations) . 1 2 3 4 5 6 7 W SM! AGIEE 013m 230 6. I think it is o.k. for me to miss work scmetimes (excluding illness, civic duties, holidays, vacations). 1 2 3 4 5 6 7 emery 9mm AGREE DISAGREE Part—B mas: For each of the individuals listed below, please consider who is FDSPIIKELYtothinkymstmldtakesmetimeofffrumworkardwtnileAsr Immthimtymslmldtakesanetimeofffrummrkdurirguierextthree months. Rankordereachpersononthatbasisusing 1,2, 3,4, 5,6. Donor USEAGIVENWIVDRE‘IHANCNCE. Pleasedonotinclude holidays, illness, vacaticm, or civic duties as missing work when answering. 1=lbstLikelytothirflcywshouldtakesatetimeoff fromwork 2=Likelytothinkywshouldtakescmetimeofffrumwork 3=SmatiatLikelytothinkyoushouldtakesaretimeoffframwork 4=Salewhattmlikelytothinkyoushwldtakesatetimeoff franwork 5=Unlikelytothinkycmislmldtakesatetimeoffframmrk 6=lhstmlikelytothirfl