(. Ii 1 ”5' 1 ' I, # J . #7, I , , I ’ m rm;— , :7”— 11.3.3.1..5” um... .Jr... r. z . 21.1.... V .. :.. ....xi 71....1; . . . .2 . . 91..., . ..:..»..1..;a..1...u 5:1. . .3 12.15.: ,. 17.: 1., 3.171.... I. .. . 1hr...“ 21.1,. .. . £1.51 .3131}. 33.5713.) .5 1: . , 3...... . . 1.71.5.1... 1:71.. (4 1. It. . 1.11. .17: 3.2. ..x..:..:. t . » 11...)...1; 3 \\ Titiiittut\\\\\\\\\\\\\\\\i\\ii\tit\\\\\\\ Z 9 2 N 4’7 This is to certify that the dissertation entitled AN EXPLORATION OF THE DETERMINANTS OF ORGANIZATIONAL-LEVEL TURNOVER presented by Mary L. Doherty has been accepted towards fulfillment of the requirements for Ph.D. degree in Psychology .7 ‘ y ,_ cM¢flwfl Major professor Date /0/// 1/919 MSU is an Affirmative Action/Equal Opporrum'ty Institution 0» 12771 LIBRARY Mlchtgan State University ____________________. PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. DATE DUE DATE DUE DATE DUE I 5! .. 'l-éc! 5 I; up,” .‘-?'i ‘5‘. ,"a '7 *4 .35 c - ‘ 5 fl 3 MSU Is An Affirmative Action/Equal Opportunity Institution c:\clrc\datedue.pm3—p.1 AN EXPLORATION OF THE DETERMINANTS OF ORGANIZATIONAL—LEVEL TURNOVER by Mary Lynne Doherty A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 1990 e-y. 3:? é4fi~ 63‘? ABSTRACT AN EXPLORATION OF THE DETERMINANTS OF ORGANIZATIONAL LEVEL—TURNOVER By Mary L. Doherty Turnover research has traditionally focused on the variables that have an effect on an individual’s decision to leave an organization. The present study examines turnover at the organizational level of analysis, which focuses on the determinants of organizational turnover rates. A model of organizational—level turnover was proposed. The measures of the model components were obtained from principals and teachers in 188 schools who completed questionnaires in 1987. Turnover data was later obtained from the principals of the schools. The proposed model was subjected to a LISREL analysis. The results of the analysis suggested that the proposed model did not adequately represent the turnover process at the Organizational level. However, the original analysis and a revised analysis indicated that the relationships between turnover and a measure of the labor market, average teacher salary, and the percentage of teachers that b910nged to a union were significant and negative. Future research directions are suggested- ACKNOWLEDGEMENTS The completion of this dissertation represents the end of life as a graduate student, and the beginning of my career as an Industrial Psychologist. So, with feelings of both joy and sadness, I take this opportunity to thank the people who helped and supported me during this project. First I would like to express my appreciation to my chairperson, Neal Schmitt, for always having the time to read (and re—read) my drafts, answer my questions, and for guiding me through this process. I thoroughly enjoyed working with and learning from Neal during my tenure as a graduate student. I would also like to thank my committee members, Dan Ilgen, Mike Lindell, and Kevin Ford for their help and SUpport during this process, and throughout my years at MSU. I appreciate all of their comments and suggestions. My friends and fellow students also provided me with support and encouragement. I would specifically like to thank Sharon Wachter for her assistance and support during the last five years. Her calming demeanor and friendship helped me to get past the graduate school hurdles. Thanks also go to Jeff and Laurie Vancouver who often listened and responded to my questions and concerns over dissertation issues. Their friendship and help were deeply appreciated. Most importantly, I would like to thank my family for their love and encouragement. Even though my parents were living in Ireland during most of this dissertation process, their long-distance support was invaluable. I share this accomplishment with them. iv TABLE OF CONTENTS LIST OF TABLES . . . . . . . . LIST OF FIGURES . . INTRODUCTION . . . . . . . . . Importance of Turnover . . Consequences of Turnover . Innovation . . . . . . Formalization . . . . . . Communication . . . . . . Economic Consequences Turnover as a Criterion of Effectiveness Summary of the Importance of Turnover Individual Level Turnover Measurement of Turnover The Use of Turnover Intent Voluntary vs. Involuntary Turnove Functional Turnover Economic Correlates . . . . Employment Perceptions Labor Market Conditions . Organizational Correlates Unionization . . . . . Page xi 18 20 20 Work—Related Correlates . . Salary . . . . . . . . . Satisfaction . . . . . . Organizational Commitment Job Performance . . . . . Personal Variables . . . . Demographic Variables . . Met Expectations . . . Turnover Intent . . . . . Summary and Critique . . . Climate . . . . . . . . . Leader-Subordinate Relations Organizational—Level Turnover Previous Research . Economic Correlates . . . Labor Market Conditions Organizational Correlates Organizational Size Supervisor Turnover . . . Average Salary . . . - Union Presence . . Organizational Performance Organizational Climate . Work-Related Correlates . - Satisfaction . . . . . . vi Page 22 22 23 24 25 27 27 28 29 29 30 31 33 34 39 39 41 41 43 45 46 48 49 57 De Hypo METHOD Sa Me Meas Le pendent Variable . . . . Turnover Rate . . . . . . theses . . . . . . . . . . mple . . . . . . . . . Principals . . . . . . . . Teachers . . . . . . asures . . . . . . . . . . Organizational Size . . . . Organizational Performance Supervisor Turnover . Average Teacher Salary . Union Presence . . . . . School Climate Satisfaction . . . . . Labor Market Conditions . . Turnover Rate . . . . . urement Issues . . . . . . vels of Analysis Issues . . Composition Modeling Aggregation . . . . . . Perceptual Agreement Specification of Levels of Measurement and Analysis . . . . . Page 58 58 60 61 61 62 62 62 62 64 65 65 65 65 67 68 69 70 7O 71 72 74 75 Method Bias . . . . . . . . . . . Unmeasured Variables Analysis . . . . . . . . . . . . . . . RESULTS Development of the Teacher Self—Report Measures . . . Climate . . . . . . . . . . Satisfaction . . . . . Descriptive Statistics of the Model Components . . . . . . . Measurement Issues Common Method Variance Levels of Analysis Issues Analysis of Potential Moderators LISREL Analysis of the Model Model Specification Description of Paths in the Model Lambda Y (LY) Lambda X (LX) Beta (BE) Gamma (GA) . . . . . . . . Phi (PH) Psi (PS) Theta Epsilon (TE) Theta Delta (TD) . . . . . . . viii Page 76 78 79 86 86 86 88 88 94 94 100 101 101 106 106 107 107 108 108 108 109 109 109 Results of the LISREL Analysis Fit Indices . . . . Chi Square . . . Goodness of Fit and Adjusted Goodness of Fit Indices Root Mean Square Residual Squared Multiple Correlation Values Residual Matrix Summary of the Results of the Hypotheses ' Alterations to the Model (RMSR) Description of the Revised Model DISCUSSION . . . . . . Summary of Results Implications of Study The Turnover Proces Implications of Specific Paths in Teacher Salary — Professional — Unemployment Union Presence — Organizational Performance Supervisor Turnover - Satisfaction S Turnover Turnover 0 Size in the Turnover Process Climate and Satisfaction in the Turnover Process Practical Implications Levels of Analysis ix Climate o . the Model a o 1 Page 110 110 110 110 111 111 113 115 116 118 121 121 122 122 124 124 125 126 126 130 130 131 Page Limitations of Study . . . . . . . . . . . . . . 133 Sample . . . . . . . . . . . . . . . . . . . . 133 Turnover . . . . . . . . . . . . . . . . . . . 133 Satisfaction . . . . . . . . . . . . . . . . . 134 Unemployment Measures . . . . . . . . . . . . . 134 Future Research Directions . . . . . . . . . . . 135 Leader Subordinate Relations . . . . . . . . . 136 Satisfaction . ; . . . . . . . . . . . . . . . 136 Climate . . . . . . . . . . . . . . . . . . . . 136 ‘Training . . . . . . . . . . . . . . . . . . . 137 Societal Variables . . . . . . . . . . . . . . 138 Individual Performance . . . . . . . . . . . . 138 Levels of Analysis . . . . . . . . . . . . . . 139 Summary . . . . . . . . . . . . . . . . . . . . . 139 APPENDIX A Principal Items . . . . . . . . . . . . 140 APPENDIX B Teacher Items . . . . . . . . . . . . . 143 APPENDIX C Average Teacher Salaries for 1987 . . . 147 APPENDIX D Professional Unemployment Measure . . . 149 APPENDIX E Teacher Demand Ratings . . . . . . . . 151 APPENDIX F Letter to Principal 155 \PPENDIX G Construct Labels associated with Etas and Ksis - - - - ' ' - 158 APPENDIX H Results of the LISREL Analysis 159 182 REFERENCES TABLE TABLE TABLE TABLE TABLE FABLE PABLE TABLE ‘ABLE ABLE ABLE ABLE ABLE tBLE LIST OF TABLES Correlates that are often found in Turnover Studies . . . . . . . . Demographic Characteristics of Principal Sample . . . . . . Structural and Measurement Equations for the Proposed Model . . . . . . Climate Construct Intercorrelations . Descriptive Information for Model Variables . . . . . . . . . . . . . . Zero-Order Correlations between Model Variables . . . . . . . . . . Partial Correlations between Model Variables . . . . . . . . . . . . . . . Unrotated Factor Matrix of Climate and Satisfaction Items . . . . . . . . . . Rotated Factor Matrix of Climate and Satisfaction Items . . . . . . . . Rotated Factor Matrix of Climate and Satisfaction Scales . . . . . . . . Comparison of Individual-Level and Organizational—Level Zero Order Correlations . . . . . . . . . . . . Regression Analysis with Teacher Demand Ratings as the Moderator Variable Regression Analysis with Professional Unemployment as the Moderator Variable . . . . . . . . . . . . . . Structural Equations and Multiple Squared Correlations of the Endogenous Variables . xi Page 11 63 81 87 89 90 93 96 97 98 102 103 104 112 FIGURE FIGURE FIGURE FIGURE FIGURE LIST OF FIGURES Mobley’s Model of Individual Level Turnover . . . . . . . . Proposed Model of Organizational- Level Turnover Graphic Representation of the Proposed Model . Results of the Analysis of the Proposed Model . . . Results of the Analysis of the Revised Model . . . xii Page 38 83 114 120 INTRODUCTION According to Steers and Mowday (1981), over 1000 empirical studies of turnover had been conducted prior to 1981, as well as thirteen review articles. Turnover research has focused on identifying the reasons why individuals choose to leave an organization. Recently, researchers have emphasized the importance of also examining turnover at the organizational level (Baysinger & Mobley, 1983; Bluedorn, 1982b; Terborg & Lee, 1984). The study of organizational turnover focuses on the determinants of turnover rates of organizations. The turnover rate of an organization is often defined as the percentage of workers who voluntarily leave the organization during a specified period of time. Although the concept of organizational level turnover has not received much attention in the literature, it is important to study because of the effect that the turnover rate of an organization has on many other aspects of the organization. According to Bluedorn (1982b), the consequences of turnover on the organization is one of the most salient issues in turnover research. While the consequences of turnover are important, the identification 2 of the determinants of organizational turnover is also critical if we are to understand and learn how to deal with organizational turnover. This introduction is divided into four sections. The first section provides more detail on why organizational turnover is important to study. The next two sections focus on turnover at the individual and organizational levels, respectively. Within the organizational turnover section, the proposed model is presented, and levels of analysis issues are discussed. Also in this section, further support is presented for how the study of organizational level turnover may add to our understanding of the turnover construct. The final section of this introduction consists of the hypotheses associated with the proposed model. Importance of Turnover Consequences of Turnover Several researchers have discussed the effect that turnover has on various organizational processes. Research suggests that organizational turnover has an impact on processes such as innovation, formalization, and communication (Bluedorn, 1982b; Muchinsky & Morrow, 1980; Price, 1977). The economic consequences of turnover have also been a concern in the literature. The effect of rganizational level turnover on each of these processes 3 and the economic implications of turnover is briefly described. Innovation. There is some evidence to suggest that a higher rate of turnover may result in a higher degree of innovation. It has been suggested that new employees bring fresh ideas into the organization which in turn helps an organization continue to grow (Grusky, 1959; Muchinsky & Morrow, 1980; Staw, 1980). This proposed effect of turnover on innovation is based on the results of a few empirical studies, so further work is needed (Muchinsky & Morrow, 1980; Price, 1977). Bluedorn (1982b) suggests that turnover may result in greater innovation when replacement employees are hired from outside of the organization, into top—level positions. Formalization. Turnover may be related to the degree of formalization in an organization in that higher rates of turnover can result in a reliance on formal rules and orms. When turnover rates are high, the rules of the rganization have to be more explicit so new workers or ew management can begin to learn these rules, and be able 0 function within the organization (Muchinsky & Morrow, 980; Price, 1977). For instance, Carlson (1962) states hat newly—hired school superintendents focus on rocedures and making rules early in their tenure. This as also found in a very different setting, a prison camp Grusky, 1959). Grusky (1959) reported that one new 4 supervisor initiated 52 new rules for the prisoners. However, according to Bluedorn (1982b), the ability of the organization to enforce formal rules decreases when turnover is very high. Thus, Bluedorn (1982b) contends that the relationship between turnover and formalization is an inverted U—shaped one; thus when turnover rates are extremely high, formalization will decrease. Communication. The communication networks in an organization may also be affected by the rate of turnover. It is suggested that when the amount of turnover is high the quantity of communication increases, but the quality of the communication decreases (Bluedorn, 1982b). The amount of communication increases because more socialization and training of new employees will have to ake place. However, the quality declines because urnover interferes with the links in the communication etworks. Higher amounts of turnover result in gaps in he networks because individuals who were previously part f the network have left the organization (Bluedorn, 982b). Economic Consequences. It also seems likely that urnover has an impact on various economic aspects of the rganization. Several researchers argue that there are ome positive effects of turnover, including a decrease in ayroll and benefit costs, and an increase in the pportunities for movement or promotions in the 5 organization (Dalton & Todor, 1982; Muchinsky & Morrow, 1980; Staw, 1980). However, negative consequences of turnover are also discussed in the literature. It is costly to recruit, hire, place, and train new employees. In fact, Macy and Mirvis (1976) found that it could cost the organization five times what that employee earns a month or more to hire and train his/her replacement (Lawler, 1981). Productivity may also decrease as new workers are adjusting to the job, although some researchers suggest that productivity may increase if the new employees have higher motivation levels or are more skilled than the previous employees (Muchinsky & Morrow, 1980; Staw, 1980). Turnover as a Criterion of Effectiveness The turnover rate of an organization is also one important aspect of organizational effectiveness models. The history of organizational effectiveness research has been complex in that a number of different models of organizational effectiveness have been proposed in the literature (Cameron & Whetton, 1983; Seaton, 1984). Turnover could serve as a criterion of effectiveness in most of these models depending on the type of problem under study (Goodman & Pennings, 1977b). Cameron and Whetton (1983) assert that no one model or approach to organizational effectiveness is better than another, because the approach selected should depend on the 6 situation and the factors involved. Furthermore, these researchers suggest that the different models of organizational effectiveness exist because of the variety of ways in which an organization can be Conceptualized. For instance, some researchers look upon organizations as entities attempting to obtain goals (Goodman & Pennings, 1977b), while other individuals use a different framework to understand organizations (e.g., concept of social contracts, Keeley, 1980). Goodman, Atkin and Schoorman (1983) contend that one single theory will probably never be developed because researchers cannot agree on the definition of the effectiveness construct. However, many seem to agree that organizational effectiveness is an abstract construct that is defined by the researchers and by the situation (Cameron & Whetton, 1983; Goodman & Pennings, 1977b; Steers, 1977). Summary of the Importance of Turnover There are both practical and theoretical reasons to study organizational level turnover. From a practical standpoint, it would be useful for management to understand what causes organizational turnover. Turnover may be very costly to some organizations, and the identification of the determinants of turnover would be important to those who were interested in reducing the turnover rate in their organization. 7 The study of turnover also has theoretical implications. The concept of turnover is related to many other organizational constructs, several of which have been discussed previously. For instance, Bluedorn (1982b) considers turnover a disruption in the open systems perspective of input—throughput—output, because it has an effect on aspects of the organization such as communication networks and productivity. Moreover, Staw (1980) suggests that a high rate of turnover will be costly in an organization where the work of some employees is dependent upon that of other employees. Finally, some empirical evidence of the effect of turnover on other organizational constructs has been found in a recent study by Mueller and Price (1989). These researchers examined the effect of work unit turnover on integration, centralization promotion opportunities, instrumental communication, job satisfaction, and behavioral commitment (i.e., turnover intent). The results of the study indicated that turnover had a negative effect on instrumental communication and behavioral commitment. VTurnover can also be considered one criterion of organizational effectiveness, thus understanding the determinants of turnover may be useful to researchers of organizational effectiveness. This study is also theoretically important because it addresses several levels of analysis issues that have been of interest in T . 8 the literature recently, such as aggregation and composition modeling. These issues are discussed in a later section of this paper. In the next section, the results of individual—level turnover research will be reviewed briefly, followed by a discussion of organizational—level turnover. Individual Level Turnover Individual—level turnover has been the primary focus of the turnover literature. An examination of organizational—level turnover therefore should begin with an understanding of what occurs at the individual level. One of the most recent reviews of individual—level V! turnover states that turnover is generally thought to be a function of negative job attitudes combined with an ability to secure employment elsewhere" (Steers & Mowday, 1981, p. 237). Steers and Mowday also state that factors other than job attitudes will affect turnover. Research on individual-level turnover has often attempted to identify the specific determinants of turnover. A number of causal models have been proposed and examined in the literature (e.g., Bluedorn, 1982a; Mobley, Griffeth, Hand, & Meglino, 1979; Price, 1977; Steers & Mowday, 1981). The most widely used individual— level model of turnover has been a comprehensive one proposed by Mobley et al. (1979), and is presented in Figure 1. This model includes organizational, individual, l'_"—————_'—— —————— T ‘Figure ) Mobley's Node] of Individual Level Turnover Organizational Goals-Values Policies Practices Rewards Job Content Supervision Work Group Conditions Climate Size Job—Related Perceptions Expectations re: Individual Occupational Personal Hierarchical Age Level Tenure Skill Level Education Status Professionalism Interests Personality Socio- economic . Family Resp. Aptitude Individual Values Economic Unemployment Vacancy Rates Advertising Levels Recruiting Levels Word of mouth Communication Labor Market Perceptions Present Job 1. Expectancies re: Expectations Re: Alternative Jobs future job outcomes 2. Expectancies re: keeping job Satisfaction ‘ rac ion- Attraction— Expected Expected Utility Utility Present Job Alternatives Centrality of non—work values: Beliefs re: non-work consequences of quitting; Contract Constraints Alternative of withdrawal behavior forms Intentions to Search- Intentions to Quit Expectancies re: future job outcomes Expectancies re: attaining alternative Immediate vs. Delayed Gratification Turnover Behavior Impulsive behavior; Specificity & time between measures L_._._.____._.____..____.______.___.. L.._._-_ _ _,_____ _._ W: """'T m l aw" u. (D (‘3 u . .H no 4.. u .'- :1 l | o.‘ x ‘ .1; v-' .4‘ ! wal“ .;.LlG on _ iii.) a“: "11" . :v'~ ...:" .,: E5 p. ‘~' V I t 4 u. u I.‘\ 1—7 .. 10 and economic variables, as well as individual perceptions, and nonwork variables. The model proposed by Mobley et al.(1979) is much too complex to be examined in any one study, but can be thought of as a general framework which includes the types of variables that may affect the turnover process. However, this model has stimulated the interest of researchers over the years, and as a result, a large number of studies have examined various aspects of the model. This model has also been used by researchers to further refine the turnover model, and to develop other related models. For example, subsequent revisions of the model added the concept of organizational commitment to the process, and Steers and Mowday (1981) later included job involvement, job performance, and efforts to change the present situation. While the Mobley model includes organizational level variables (such as organizational climate and organizational size), all tests of the model have been conducted at the individual level. The following review focuses on the types of variables that are most often found in studies of turnover. Table 1 consists of a list of variables, which have been shown to have the strongest relationships with the turnover construct (Cotten & Tuttle, 1986). The variables in Table 1 have been categorized into four 11 Table 1 Correlates that are often found in Turnover Studies Economic Correlates Employment Perceptions Labor Market Conditions Organizational Correlates Unionization Work—related Correlates Salary Satisfaction Organizational Commitment Job Performance Personal Correlates Demographic Met Expectations Behavioral Intentions An 0”." "a; “.1 .‘V‘F W.“ 'n (I) (D I—4 L1 - H- D—f- 12 groups: economic, organizational, work-related, and rmrsonal. Each of the variables listed will be reviewed in subsequent sections. However, before reviewing this research, a discussion of the measurement of the turnover construct would be appropriate because several measurement issues should be understood when conducting or reviewing turnover research. Measurement of Turnover The measurement issues that should be considered when doing research on turnover include: (1) the use of turnover intent as a proxy for actual turnover, (2) the type of turnover measure that will be used, and (3) whether individual-level turnover is thought to have positive or negative outcomes. The Use of Turnover Intent. Several researchers have used turnover intent as a proxy for actual turnover (e.g., Marthn 1979; Werbel & Bedeian, 1989). In order to discuss this measurement issue, it is important to know that turnover intent and actual turnover are moderately to highly correlated. The correlation between turnover intent and turnover has ranged from .19 to .71 in a number (Histudies (e.g., Arnold & Feldman, 1982; Hom, Griffeth, & Sellaro, 1984; Lee & Mowday, 1987; Miller, Katerberg, & Hulin, 1979). There are a number of potential reasons for the differences found in the correlations between turnover - 9 3kg! .9"; hut in ou'L 13 intent and turnover. First, the items used to assess the turnover intention construct were very different in the various studies. Second, the amount of time between the hfitial questionnaire and obtaining the turnover data ranged from six months to one year. Finally, at least one smudy did not distinguish between voluntary and involuntary turnover (Arnold & Feldman, 1982). These cfifferences could account for the range of correlations that have been reported. However, even when turnover intent and turnover are highly related (e.g., .71), these constructs are not identical, and researchers that attempt to predict turnover intent are studying only part of the turnover process, not turnover itself. Voluntary vs. Involuntary Turnover. Another distinction in the literature is whether turnover is self- initiated or organizationally-initiated. Voluntary turnover is defined as "individual movement across the. nmmbership boundary of a social system which is initiated km the individual" (US Bureau of Labor Statistics 1966, p.1, cited in Price, 1977). In contrast, involuntary turnover is initiated by the organization, and would include individuals who have been fired or laid off. Additionally, involuntary turnover includes those individuals who have retired or died. According to Price (1977), voluntary turnover is usually studied in turnover 3259 «In J) Il'J .‘ I! in] 14 research, and most of the studies that will be reviewed used voluntary turnover as their criterion. Functional Turnover. ‘0ne recent View in the turnover literature is that not all turnover is negative (Abelson & Baysinger, 1984; Dalton & Todor, 1982; Hollenbeck & Williams, 1986; Mobley, 1982b; Porter & Steers, 1973). Dalton, Todor, and Krackhardt (1982) contend that the negative consequences of turnover have been exaggerated. It is also argued that turnover can have positive consequences for the organization in that new individuals will increase innovation and cause technological change (Dalton & Todor, 1979). A distinction has been made between turnover that is functional and turnover that is dysfunctional. Functional turnover includes those individuals who leave who are considered to be poor performers. In contrast, dysfunctional turnover occurs when an organization loses good performers. The importance of this distinction is that the effect of the turnover of the good performers on the organization is different from the effect of poor performers leaving. From the group of people that leave an organization each year, there will be some individuals that management does not want to lose, and others that will not be missed. Dalton et al. (1982) suggest that two subjective methods of classifying an individual as a functional leaver or a dysfunctional leaver might be «r: '15" q;“ i. 15 rmmformance ratings or whether the organization would rehire the person if given the chance. Other more objective measures might be productivity or sales measures. According to Mobley (1982b), most organizations do not consider the performance or ability level of those who leave. In addition, many researchers also do not make the distinction between good and poor performers (Mobley, 1982b). Staw and Oldham (1978) also suggest that a reconsideration of the dependent variables used so often urindustrial organizational research is necessary. The three dependent variables mentioned were task performance, absenteeism, and turnover. The utility of these variables to different groups in the organization is one focus of Staw and Oldham’s argument. For instance, while nmuagement might consider all turnover as dysfunctional, the workgroup losing a poor performer would regard the turnover of that individual as functional. Most of the studies that will be reviewed in the following sections do not distinguish between functional and dysfunctional turnover. Economic Correlates Two economic constructs that have often been found in studies on turnover are employment perceptions and labor nmrket conditions. These constructs are theoretically 16 -similar, but the first is subjective while the second is a more objective measure. Employment Perceptions. Employment perceptions refer to the degree to which an employee perceives that other job opportunities are available, or the utility of searching for another job (Steers & Mowday, 1981). Eupirical support for the relationship between employment perceptions and turnover has been mixed. A recent review listed twenty-one studies that had examined the relationship between employment perceptions and turnover; only eight of the studies reported a significant correlation (Steel & Griffeth, 1989). Additionally, the significant correlations that were reported were small, most below .20. Even with the lack of support found for the effect of employment perceptions, researchers believe that this construct is an important variable in the turnover process, but its effect has been limited by the existence of other variables or methodological problems (Hulin, Roznowski & Hachiya, 1985; Steel & Griffith, 1989). Hulin et a1. (1985) provided three explanations for the laCk of support for the relationship between employment perceptions and turnover. First, they suggest that the composition of the work force may be dependent on the existing economy. That is, individuals who voluntarily work part-time, or drift from one job to I qu- .Lu '- .-r ~u¢ ’v v 1‘1! 17 another may be temporarily attracted to a full—time pmsition during the times when there are many jobs to be filled. When these workers decide to leave their jobs, the decisions may not be based on the usual reasons (e.g., other opportunities, dissatisfaction with job), but their termination decision may be based on the desire to return to temporary work. A second suggestion is that the effect of employment perceptions on turnover intent is indirect, and mediated by job satisfaction. Most studies have examined the direct relationship between perceptions of job opportunities and turnover (e.g., Arnold & Feldman, 1982; Michaels & Spector, 1982; Miller, Katerberg & Hulin, 1979). The third suggestion is that employment perceptions may influence turnover directly, not indirectly through turnover intent. Steel and Griffeth (1989) have also suggested three possible explanations for the lack of support for the employment perceptions — turnover relationship in the literature, but their focus is on methodological problems. The first issue is that researchers limit the potential of this relationship because they usually sample only one job in one organization in one region and at one point in time. Steel and Griffeth (1989) contend that if researchers would expand.their samples to include more jobs or regions of the country, the variance in the employment perceptions would increase. A second issue is ". a". 7‘: .b n A] ‘.~ ‘t ,l‘. 18 that researchers often do not acknowledge the effect of the turnover base rate on their results. Thus, the results of turnover studies will be affected by the amount of variance in the turnover construct. Finally, Steel and Griffeth (1989) along with Griffeth and Hom (1988) show that the employment perceptions construct is operationalized differently across studies, and suggest that this lack of consistency may have an effect on the relationship between this construct and turnover. Steel and Griffeth (1989) also note that a number of studies use only one item to measure employment perceptions which could serve to decrease the reliability of this construct. Labor Market Conditions. The condition of the labor market was a construct suggested by the Mobley et al. (1979) model. Individual-level research suggests that the effect of labor market conditions on turnover intentions is mediated by individuals’ perceptions of the availability of other jobs (Lee & Mowday, 1987). In addition to the perceptual measure described above, a literature review by Muchinsky and Morrow (1980) supports the inclusion of an objective measure of labor market conditions. They suggest that economic conditions have a strong impact on turnover, and cite a number of studies to SUpport this contention. For instance, Woodward (1975/76) has discussed a framework that could be used to understand how the labor market influences turnover. This framework, 19 called the push-pull approach, consists of two classes of factors. Push factors are those that originate within the organization that lead to problems (e.g., increased dissatisfaction and decreased commitment) which push the employees away from the organization. Pull factors occur outside the organization (e.g., demand for labor) and entice employees away fromtheir present organizations. Woodward (1975/76) asserts that one main cause of higher turnover rates during times of low unemployment is an increase in the variety of jobs that are available to workers. In addition to a variety of jobs, there is an increase in certain types of jobs that may be attractive to workers such as daywork or better working conditions. Another potential cause of increased turnover rates vmen unemployment is low is that the standards used in selection are lowered (Woodward, 1975/76). As an example, Woodward (1975/76) examined scores from selection tests over a two-year period for one organization. The results indicated that the percentage of recruits who had obtained lower percentile scores on the selection tests increased during months of lower unemployment. However, the results roported were percentages and no significance tests were conducted, thus the results are only suggestive. Research has suggested that labor market conditions “my not have a direct effect on turnover, but may moderate the relationship between satisfaction and turnover. .. ,. u; Ante I A nli rt 20 humminsky and Morrow (1980) suggest that fewer people will leave their jobs during periods of high unemployment, thus the relationship between satisfaction and turnover will be deflated. However, when there are more jobs available in the work force, more individuals who are not satisfied with their present position will leave, and the correlation between satisfaction and turnover will be higher. Two meta-analytic studies have examined this relationship. The first, conducted by Shikiar and Freudenberg (1982), indicated that the satisfaction- turnover relationship was strongest when the unemployment rate was high, which is just the opposite of what Muchinsky and Morrow (1980) had proposed. However, Carsten and Spector (1987) found a number of nmthodological problems with the Shikiar and Freudenberg (1982) study, and after correcting for these problems, replicated the meta-analysis and found that their results did support the hypothesis of Muchinsky and Morrow (1980). Qrganizational Correlates Unionization. The only organizational correlate that has been examined consistently in the literature in relation to turnover is the presence of a union. This construct has been examined as both a direct and an indirect predictor. Researchers have hypothesized that unionization leads to longer tenure and more job security because of the system set up by unions to deal With 'I,’ '_ an; I." u' ’-—I L, 21 'Inpblems (Farber, 1980; Wales, 1970). For instance, Freeman (1980) noted that 99% of the major organizations that are unionized in the United States have collective tmrgaining contracts that include grievance procedures. HOwever, only 30% of the non-unionized organizations that tmlong to the Bureau of National Affairs Personnel Fblicies Forum report having any formalized grievance IHpcedures. Thus, the union system would directly affect turnover. Empirically, Farber (1980) showed that out of a sample of 944 people, non-unionized individuals were more likely to quit than were unionized workers. Farber (1980) did not indicate how many of the workers in the sample kmre unionized. A second study also resulted in unionized workers having lower quit rates (Wales, 1970). Unionization has also been hypothesized to be an indirect predictor of turnover mediated by satisfaction. This hypothesis is called the "exit—voice tradeoff" which means that employees who are dissatisfied with working conditions do not have to leave the organization because they have a voice in their union, and can use their union to help them solve problems. The presence of a union is thought to have a negative affect on satisfaction, because unions often make workers more aware of company deficiencies, and union jobs may be more unpleasant than non-unionized jobs (Borajas, 1979; Farber, 1980). Borajas (1979) reported that the presence of a union did have a 22 direct negative effect on job satisfaction. Additionally, Freeman (1980) found that both satisfaction and the presence of a union were related to turnover rates. Workers who were highly satisfied were more likely to stay with their organization than were dissatisfied workers, and individuals who belonged to a union were more likely to stay than were non-unionized workers. However, Freeman’s (1980) results were reported as percents, and are only suggestive. WorkeRelated Correlates Four work-related constructs that are often found in turnover research are salary, satisfaction, organizational commitment, and job performance. Each of these constructs are discussed below. Salary. Salary is usually considered an indirect predictor of turnover intent, mediated by satisfaction, in models of turnover. According to Lawler (1981), employees are satisfied or dissatisfied with pay for several reasons. Satisfaction with pay is influenced by the amount received, as well as the amount that employees think they should receive. Satisfaction with pay is also affected by a comparison between one’s job and salary with what other employees do in their jobs and the amount of salary that they receive. If employees become too dissatisfied with their salary, they will consider leaving their organization. Price (1977) cited a number of 23 studies that found a negative relationship between salary and turnover, and Steers and Mowday (1981) contend that salary should affect job attitudes which in turn will influence turnover intent. Support for this relationship has been demonstrated. For instance, Motowidlo (1983) found that although amount of pay and pay satisfaction are highly related (r = .51, p < .01) and pay satisfaction and turnover intent are significantly correlated (r = .48, p < .01), amount of pay and turnover intent are not related (r : .21, ns). Another study examined this relationship somewhat differently. Hom, Griffeth, and Sellaro (1984) measured perceptions of inequity, job satisfaction, and thoughts of quitting. Thoughts of quitting precedes turnover intent in their model, but the two constructs are similar in content. Their analyses indicated that the inequity perceptions construct was predictive of job satisfaction, and job satisfaction was predictive of thoughts of quitting. Satisfaction. In most models of turnover, satisfaction is thought to be related to turnover intent which in turn influences actual turnover. It is possible that satisfaction also has a direct effect on turnover decisions, but empirical evidence suggests that the relationship is an indirect one. Research has consistently shown that the relationship between 24 satisfaction and turnover intent is significant and negative (e.g., Arnold & Feldman, 1982; Jackofsky & Slocum, 1988; Michaels & Spector, 1982; Parasuraman, 1982). The measurement of satisfaction has varied across studies from general scales to facet scales to intrinsic and extrinsic scales of satisfaction, but a significant negative relationship between satisfaction and turnover intent is observed consistently. Qgganizational Commitment. The effect of organizational commitment on turnover is also thought to be mediated by turnover intent (e.g., Lee & Mowday, 1987). The relationship between commitment and turnover intent is consistently significant and negative (e.g., Arnold & Feldman, 1982; Mowday, Koberg, & McArthur, 1984; Parasuraman, 1982). One problem with the commitment construct is that some researchers include items that are similar to turnover intent items when measuring organizational commitment (e.g., Angle & Perry, 1981; Welsch & LaVan, 1981). An example of an item that is similar to both scales is "It would take very little change in my present circumstances to cause me to leave this organization" (Angle & Perry, 1981, p. 5). The inclusion of these items confuses the interpretation of observed relationships between commitment and turnover intent. 25 There has been some disagreement over whether commitment leads to satisfaction or satisfaction is predictive of commitment. Williams and Hazer (1986) conducted a series of path analyses to address this question. The results of their study indicated that for two different samples (community mental health center and insurance company employees), the model that included a causal link from satisfaction to commitment fit the data better than did a similar model that included a causal link from commitment to satisfaction. However, a later study also examined this relationship using path analysis and concluded that both models were supported (Farkas & Tetrick, 1989). Both Williams and Hazer (1986) and Farkas & Tetrick (1989) contend that commitment and satisfaction are related, but that the direction of any causal influence between the two constructs cannot be determined. Farkas and Tetrick (1989) suggest that another explanation for the strong relationship between these two constructs is that they are not completely distinct operationally. Further research designed to examine this relationship is required. Job Performance. The role that this construct plays in the turnover process is unclear at the present time. It was initially proposed as a predictor of affective variables (Steers & Mowday, 1981), but has also been examined in the literature in several other ways. Lee and 26 ' Mowday (1987) did examine job performance as a predictor of affective constructs and found that job performance was Inedictive of organizational commitment and job involvement, but not of job satisfaction. In contrast, Jackofsky and Slocum (1988) showed that job performance was related to a measure of extrinsic satisfaction, but not intrinsic satisfaction. Performance is also seen as a direct precursor of turnover (Stumpf & Dawley, 1981). These researchers reported that two performance indices were significant predictors of turnover even after the variance due to demographics and an absenteeism measure was removed. One of these studies did address the issue of functional versus nonfunctional turnover, and found that performance was lower for those individuals that left the organization (Dreher, 1982). However, the other studies that examined the role of performance did not examine this issue. Using a different conceptualization of the role of performance, Spencer and Steers (1981) examined performance as a moderator of the satisfaction—turnover relationship and found a significant interaction between performance and satisfaction. Finally, a recent study examined the relationship between performance and turnover intent, moderated by age (Werbel & Bedeian, 1989). These researchers found a significant main effect for performance as well as a significant interaction between P. . 27 ; performance and age. One major problem with the research on the role of job performance in the turnover process is that researchers ignore the work conducted by others on the role of job performance. It would be useful if researchers would discuss how their results could be compared to other work in this area. The existence of job performance in the turnover process seems to be important, but the actual function of the construct is unclear. Personal Variables 1 Personal variables that are often thought to influence turnover decisions include demographic variables, met expectations, and behavioral intentions. Research on each of these types of variables will be discussed in the next sections. Demographic Variables. The demographic variables that are often found in turnover research are age, reducation, marital status, gender, number of dependents, and tenure. Each of these variables has been found to be significantly related to turnover or turnover intent in some studies, but the variables that are most consistently related are age, education, and tenure (Arnold & Feldman, 1982; Martin, 1979; Mitchell, 1981; Parasuraman, 1982; Spencer & Steers, 1981). The effect of age on turnover intent is sometimes mediated by satisfaction or commitment (e.g., Martin, 1979; Michaels & Spector, 1982; Williams & Hazer, 1986), but direct negative relationships with 28 turnover and turnover intent have also been observed (Martin, 1979; Parsuraman, 1982). Education is most often hypothesized to have a direct positive relationship with turnover and turnover intent, while tenure is reported to have a direct negative relationship (Mitchell, 1981; Parasuraman, 1982; Spencer & Steers, 1981). Met Expectations. The expectations of employees are thought to influence affective responses in the turnover process (Bluedorn, 1982b; Lee & Mowday, 1987; Steers & Mowday, 1981). It has been suggested that when expectations are met, affective responses (e.g., satisfaction, commitment) are more positive and turnover decreases (Steers & Mowday, 1981). However, empirical support for this proposition has been mixed. Dugoni and Ilgen (1981) examined the effect of realistic job previews (RJP’s) on the expectations - satisfaction — turnover process. They found support for the relationship between satisfaction and turnover, but did not find a significant relationship between met expectations and satisfaction. Similarly, Reilly, Tenopyr, and Sperling (1979) reported in their study that the relationship between the use of RJP’s and turnover was not significant. In contrast, the results of several studies have supported the inclusion of met expectations in the turnover process. Lee and Mowday (1987) found that met 1 i 29 expectations were predictive of job satisfaction, commitment, and job involvement. Additionally, Hom, Griffeth, and Sellaro (1984) and McKemey and Sims (1977, 1980, cited in Bluedorn, 1982b) reported that expectancies were significant predictors of satisfaction. Turnover Intent. Intent to leave an organization is another variable that has been found to be related to turnover in numerous studies (e.g., Arnold & Feldman, 1982; Hom, Griffeth, & Sellaro, 1984; Michaels & Spector, 1982; Williams & Hazer, 1986). In fact, Bluedorn (1982b) reported that in 23 studies that had collected data on both turnover intent and turnover behavior, all 23 of the studies found a significant positive relationship between the two variables. Moreover, in 19 out of 20 studies, the intent to turnover construct was more predictive of turnover behavior than any other predictor. Summary and Critiqge It should be recognized that one integrated theory of turnover does not really exist. A number of different models of turnover can be found in the literature, and the focus of some of these models differ (e.g., met expectations, organizational commitment). Most of the studies in the literature are testing one aspect of these more detailed models proposed by researchers such as Mobley et al. (1979) and Steers and Mowday (1981). _ In '3‘ .. '1 .AK _-'fl.h_AI-" _'. 1n-» . . . 30 The above review has described the research found on the types of variables that are often studied in relation to individual-level turnover. Some of the constructs discussed are consistently related to turnover, while others are not. The variables that have been consistent correlates of turnover are salary, satisfaction, commitment, certain demographics, and turnover intent. Constructs that have been inconsistently related to turnover include employment perceptions, unemployment rate, presence of a union, job performance, and met expectations. This review was not meant to suggest that these variables are the only ones related to turnover, but they are the constructs that are most frequently found in turnover studies and turnover models. Other potential correlates of turnover have been examined in a few studies. These include personal correlates such as job involvement, and various nonwork variables (e.g., family size) (e.g, Lee & Mowday, 1987). The constructs reviewed above appear to be most representative of the turnover literature. Two variables that have received recent attention in the literature and merit some consideration are climate and leader-subordinate relations. Climate. Climate can be defined as the perceptions 0f individuals of their environment or work setting. The concept of climate was present in the original Mobley et 31 al. (1979) model, but attempts to test this model at the individual-level usually have not included climate as a direct or indirect predictor of turnover. One exception was a study conducted by Martin (1979). The results of this analysis indicated that the relationship between several climate dimensions (i.e., routinization, communication, distributive justice) and turnover intent was mediated by satisfaction. In addition, an examination of the climate literature indicates that the suggested relationship between climate and turnover has also been virtually ignored. However, two studies were located. In one study, the researchers found a significant relationship between a measure of climate and turnover intentions (Schneider & Bowen, 1985). In a second recent study, Jackofsky and Slocum (1988) examined the relationship between seven climate dimensions (e.g., supervisory style) and turnover intentions. In this longitudinal study, significant relationships between the climate dimensions and turnover intent over two time periods were reported. The results of these studies indicate that the climate construct should be considered as an important part of the turnover process. Leader—Subordinate Relations. This construct is often considered to be one of the many dimensions of climate. However, several researchers have focused their attention on this climate dimension, while not addressing 32 other dimensions. Early research indicated that leader behavior was related to turnover (Fleishman & Harris, 1962). The role of this construct in the turnover process was then ignored for a number of years. However, recent research has been particularly supportive of the notion that a subordinate’s perception of a leader’s consideration or supportiveness has an effect on whether or not the subordinate leaves the organization (Ferris, 1985; Graen, Liden, & Hoel, 1982). Furthermore, the results of one study showed that the relationship between leadership consideration and turnover was mediated by job satisfaction, organizational commitment and turnover intent (Michaels & Spector, 1982). These few studies lend support to the hypothesis that the relationship that subordinates have with their superiors may affect the turnover process. The review of the individual—level research suggests that many different types of variables can influence an individual’s decision to leave an organization. Some of these variables will also be important at the organizational-level of analysis, while others will be less important. For instance, Rousseau (1985) contends that research has shown that economic variables account for 70% of the variance in turnover at the unit level, While behavioral intentions and attitudes account for 70% of the variance at the individual level. Additionally, SOI 33 some of the variables that will be important at the organizational level will be different from those that were discussed earlier. Now that research relating to turnover at the individual level has been discussed, the focus of this paper will turn to organizational-level turnover. Organizational Level Turnover Roberts, Hulin and Rousseau (1978) maintain that examining an area of research at only one level of analysis may be misleading. An example provided by these researchers illustrates their contention. When this book was published in 1978, the only variable that appeared to be consistently related to organizational turnover rates was economic conditions. Since then, researchers have not consistently identified many other predictors of organizational turnover, so the following example is still appropriate. Roberts et al. (1978) suggested that a manager may want to reduce the high rates of turnover within subunits of an organization. One of the only suggestions that could be given to this manager would be to change the economic rewards, which an individual manager cannot often control. Roberts et a1. (1978) suggested that subunit turnover rates may also be related to several variables including management style and level 0f employee satisfaction. These authors contend that an area of research such as turnover is complex and should be 34 studied at different levels of analysis so that the constructs that are important at each level can be identified. Roberts, et al. (1978) conclude that "only in this way will we be able to determine the direct and indirect influences of environmental and organizational characteristics on individual behaviors" (p. 134). If the determinants of turnover at the organizational level are different than those at the individual level, then the identification of the variables important at the organizational level will broaden our understanding of the turnover construct. This section of the paper first reviews research conducted previously on organizational—level turnover. Then a model of organizational—level turnover is introduced, and the components of the model are discussed. The model includes organizational—level constructs that may influence the turnover rate of an organization. Furthermore, an attempt is made to include constructs that were important at the individual—level of analysis, if these constructs also exist at the organizational level. Brevious Research Early research on organizational turnover focused mainly on the rate of turnover in an organization. Price (1977) cited 53 studies that examined organizational turnover rates, defined as the percentage of employees that left an organization during a specified period of rv‘ 7—— 35 time. Most of these researchers were interested in identifying turnover rates for various types of organizations (e.g., manufacturing, mining, government). Several studies did explore the determinants of organizational turnover. For instance, one early examination of nursing personnel turnover by Levine (1957) indicated that the turnover rate of hospitals was affected by the size of the hospital, ownership (i.e., government,. 5 church, other), and whether the hospital had a school of I nursing. Other early studies of turnover rates that have examined the relationships between turnover rates and constructs such as labor market conditions and wage rates (e.g., Eagly, 1965; Wales, 1970) will be discussed in more detail in subsequent sections. More recently, researchers have begun to focus on the determinants of organizational turnover. Baysinger and Mobley (1983) asserted that an aggregated measure of turnover is important to the development of personnel policies. They were interested in understanding the "quit propensity" or turnover intention of the average employee in an organization. Baysinger and Mobley (1983) presented a model that included the costs and benefits of staying or leaving an organization, as well as the degree of job dissatisfaction due to on-the—job experience. The model Posits that individual factors, organizational factors, 7__— 36 and environmental factors each have an effect on the components of the model. Other researchers have attempted to identify the determinants of turnover in empirical studies. In a longitudinal study, Terborg and Lee (1984) found that they could reliably predict turnover in a sample of sales personnel over two data collection periods. The predictors identified were local economic activity, | average age, tenure, time in present position, and education. In contrast, they were not able to find reliable predictors for a management group. A second organizational-level study identified a relationship between organizational commitment, turnover intent, and turnover rate (Angle & Perry, 1981). However, there were some methodological problems with this study. Pfeffer and his colleagues have concentrated on the study of turnover using group demography as a predictor (McCain, O’Reilly, & Pfeffer, 1983; Pfeffer, 1983; Wagner, Pfeffer, & O’Reilly, 1984). This line of research contends that demography, or more specifically, the degree of tenure similarity within a group, has an effect on the turnover rate of an organization. The results of their first study in this area indicated that turnover was hiSher in academic departments where either a large number of faculty members entered a department at the same time, or there were large tenure gaps between professors 37 (McCain, et al., 1983). Later research focused on tenure gaps within a department, contending that departments with small tenure gaps will have stronger social ties, and consequently lower rates of turnover. Wagner et al. (1984) tested this hypothesis, and found that turnover rate was positively related to the magnitude of the tenure gap. The research in this area to date indicates that the demographic characteristics of a group may be a useful predictor of turnover. Studies that have examined individual level turnover, and the research at the organizational level led to the proposed model of organizational level turnover depicted in Figure 2. A number of organizational-level constructs can be found in Figure 2 including union presence, average teacher salary, organizational size, organizational performance, supervisor turnover, and turnover rate. Two constructs, organizational climate and satisfaction, are aggregates of individual-level responses, and arguments are made that both of these can also be considered organizational—level constructs. To demonstrate that these constructs can be considered at the organizational level, both theoretical support and statistical evidence are presented. Theoretical support for these constructs is discussed in the introduction. Statistical evidence such as obtaining perceptual agreement within an Outflvzam 036C .0 500852 sOtOIOF .covaom /\ 92m ice—«5:090 005305.500 acoEco.:>cw notation acuuaam mocafiaouhwa 0.55:0 .nco.~¢w.cau..0 3:025:93qu IW again—0.35". >Eocoua< + 0.3... 00:002.“. 595:2.» £230.33...» . :25.- I + 5.0.2250 ecu—om I 525:5... 7.8.3.2 Loan... I. oun..o>< 5023095 350:5... _w>®.._. .wcozmficmeO *0 EDGE vowoaonm N 959....— 39 organization and the use of composition modeling are discussed later in the method section. The final model component, labor market conditions, is included as an environmental or economic predictor. Arguments supporting the inclusion of each of these constructs in the model are presented in the following sections. The signs above the paths in Figure 2 indicate the hypothesized direction of the relationships in the model. Economic Correlates Labor Market Conditions. Several studies at the organizational level of analysis have examined the relationship between labor market conditions and turnover. Eagly (1965) reported a correlation of —.84 between the quit rate and unemployment over a period of 31 years (1931 to 1962). More recently, Terborg and Lee (1984) examined the relationship between objective measures of labor market conditions and turnover, and found that the labor market indicators (i.e., monthly local unemployment figures for nonfarm workers, help wanted index in the Conference Board periodical) were significant predictors of turnover (Terborg & Lee, 1984). Although research at both the individual and organizational level of analysis have found that labor market conditions are correlated with subsequent turnover, Muchinsky and Morrow (1980) caution researchers against assuming that economic conditions will have the highest 40 correlations with turnover. They contend that differences have been found in turnover rates within industry and within location, suggesting that other factors also have an effect on turnover rates. The research discussed above suggests that the labor market has a direct effect on turnover rate. Individual— level research has also suggested that labor market conditions may moderate the relationship between satisfaction and turnover rate (Carsten & Spector, 1987). However, this relationship has not been examined at the organizational level. In fact, none of the studies that examined the turnover construct at the organizational level included all three of these variables (i.e., labor market conditions, satisfaction, and turnover rate). Conceptually, it makes sense that the relationship between satisfaction and turnover rate would be stronger when unemployment is low, and jobs are more easily found. The literature supports a direct path from labor market conditions to turnover rate, and this is the link that is being hypothesized, but the use of labor market conditions as a moderator variable will also be investigated in this study. One other link that could be proposed is one between labor market conditions and supervisor turnover. However, supervisor turnover is being conceptualized as a construct that takes place over a number of years. The condition of 41 the labor market has the potential to change over time; thus the link between labor market condition and supervisor turnover is not being hypothesized. Organizational Correlates Organizational Size. Findings concerning the relationship between organizational size and turnover at both the individual level and the organizational-level of analysis have been inconsistent. For instance, Terborg and Lee (1984) found that the relationship between . organizational size and turnover rate approached significance for two samples over one time period. However, size was unrelated to turnover for the same two samples during another time period. Furthermore, Price (1977) stated that the relationship between organizational size and turnover has been found to be negative, positive, and nonsignificant. Because of the inconsistency of the relationship between size and turnover, a direct link between these variables is not being proposed. However, an indirect relationship between organizational size and turnover, mediated by climate and satisfaction, will be examined in this study. Larger organizations have been characterized as more bureaucratic, compliant, alienated, suspicious, and as- having employees who are less committed (March & Simon, 1958; Payne & Pugh, 1976). Furthermore, Joyce and Slocum (1979) suggest that size is one variable that can have a p01. a.) “k 42 potential effect on the way the climate of an organization is perceived. For instance, George and Bishop (1971) suggested that an increase in the size of schools results in lowered amounts of participatory decision making and authority. Empirically, the relationship between organizational size and climate was examined by Indik (1965). Indik (1965) examined four mediation models, three of which included a measure of organizational climate. The three measures of organizational climate were the perceived amount of communication with other workers, the perceived amount of higher-level interpersonal control, and the perceived degree of internal coordination of activities. Indik (1965) found that size had a negative effect on the amount of communication within the organization (which included items regarding freedom to discuss personal problems with superior and the amount of information provided and received by the employee). He also found that an increase in the size of the organization resulted in less coordination of activities within the organization. The relationship between size and amount of higher-level interpersonal control was not significant. Although size did not affect the interpersonal control climate measure, this study does provide support for the hypothesis that size may have a negative effect on climate perceptions. 7—7 I 4 3 One assumption being made in this proposal is that organizational size has an effect on some objective structural variables, while other structural variables, such as specialization, will be fairly similar across organizations because of the sample being used in this study (i.e., teachers in secondary schools). Although this study is not measuring objective structural variables, two perceptual variables that are usually 1 labeled as structural are being studied. More specifically, it is hypothesized that two climate dimensions (participation in decision making and degree of authority) along with other climate dimensions will mediate the relationship between organizational size and affective constructs. Supervisor Turnover. Research at the individual level suggests that the relationship an employee has with a supervisor is an important predictor of satisfaction in the turnover process (e.g., Williams & Hazer, 1935). A similar measure at the organizational—level of analysis might be obtained by averaging the perceptions of subordinates within an organization about their relationships with their supervisors. One assumption underlying this type of measure is that supervisors treat all of their subordinates alike (Graen & Cashman, 1975). However, research conducted on the vertical dyad linkage model has shown that a supervisor’s relationship with 44 subordinates differs from one subordinate to another (Dansereau, Cashman, & Graen, 1973; Dansereau, Graen, & Haga, 1975). In other words, subordinates within the same work group report differential treatment from the work group supervisor (Dansereau, Cashman, & Graen, 1973; Dansereau, Graen, & Haga, 1975). Thus, the relationship between leaders and their subordinates appears to be a dyadic one (Ferris, 1985; Graen, Liden, & Hoel, 1982), and this research indicates that an averaged measure of leader-subordinate relations may not be the most appropriate measure to use at the organizational-level of analysis. A measure that might be used in place of averaged leader-subordinate relations responses is the frequency of supervisor turnover. This measure would not be isomorphic with the individual-level construct, but would most likely have an effect on the leader-subordinate relations within an organization. For instance, it would be difficult to form any type of stable relationship with a supervisor if the person in that position was constantly changing. In addition, Price (1977) asserts that a high degree of managerial turnover results in higher amounts of formalization, or making organizational norms (as the new manager perceives them) more explicit. A workplace in Which the rules or procedures are often changing could become confusing which in turn could lead to decreased 01 ‘0 LL) 45 satisfaction. Research discussed earlier provided support for the contention that supervisor turnover leads to an increase in new rules and procedures (Carlson, 1962; Grusky, 1959). It should be noted that some of the variables in Figure 2 may also have had a similar effect on supervisor turnover. However, these links are not being considered in this proposal. Average Salary. The proposed model suggests that salary has a direct effect on turnover. Katzell, Barrett and Parker (1961) examined the relationship between turnover and wage rate for a group of 72 divisions within an organization. The authors of this group—level analysis found a significant negative correlation between wage rate and turnover (r:—.32). Additionally, these researchers looked at the relationship between wage rate and a number of satisfaction questionnaire items (n = 47 items) within division. They again found significant correlations between wage rate and 25 of the satisfaction items (mean r = -.32). Finally, the satisfaction items were not found to be related to turnover at the division level. Wales (1970) examined the relationship between salary and turnover at the organizational-level of analysis. He used a model that combined cross-sectional and time-series data to examine the influence of different variables on turnover rates in a sample of eighteen industries. The results indicated that there was a significant negative _T 46 relationship between salary and turnover rates. Other significant correlates of turnover rates were unemployment rate, unionization, age, and gender. Wage rate was most highly related to turnover in this study. The proposed model in Figure 2 started with the relationship between salary and turnover being mediated by satisfaction, because researchers have suggested that this relationship should be mediated by job attitudes (Lawler, 1981; Steers & Mowday, 1981). However, the empirical evidence at the division and organizational levels does not support this contention. In fact, both of the aforementioned studies provide support for a direct link between salary and actual turnover. Union Presence. There has not been much research on ~the relationship between union presence and turnover rate to date in the literature, and the research that does exist has produced mixed results. Three studies have been conducted that examine this relationship, and each one has found different results. Wales (1970) reported that there was a relationship between the two constructs, but that the relationship was moderated by unemployment rate. Longest and Clawson (1974) found that unionization was a direct negative predictor, and accounted for a significant amount of variance in turnover rates, and finally, Newton, Betcherman, and Leckie (1981) reported a nonsignificant relationship between unionization and turnover rate. 47 One possible reason for these mixed findings is that a different measure of unionization was used in each study. Wales’ (1970) measure of unionization was the percentage of individuals within an organization that were unionized. The measure used by Longest and Clawson (1974) was the existence of a grievance procedure, and Newton et al. (1981) used a dichotomous measure of unionized versus not unionized. Results of research on the effect of union presence in the turnover process have been inconsistent, so it may be necessary to explore the role of this construct in the model. However, in the present study, I am hypothesizing that union presence will have an indirect effect on turnover, mediated by satisfaction. This relationship was supported by research at the individual level of analysis. As mentioned earlier, the individual—level research suggests that the effect of union presence on satisfaction is negative, because of the type of job involved, and the fact that unions heighten the employees’ awareness of unsatisfactory working conditions. However, because of the sample being used in this study (i.e., secondary school teachers), it is hypothesized that the presence of a union will have a positive effect on satisfaction for the following three reasons. First, the type of job is unlikely to be a factor because the job is fairly similar across the sample. Additionally, it would SB 3'0 BE 48 seem that teachers would be aware of unsatisfactory working conditions whether or not a union was present. This may also be true in other occupations, although Borajas (1979) and Farber (1980) predicted the presence of a union would result in workers becoming more aware of poor working conditions. Finally, the presence of a union may make the teachers feel that their jobs are more secure, their pay is more equitable, and that they have the ability to take action to improve working conditions. Organizational Performance. Organizational performance is another exogenous variable that is proposed to have a direct effect on turnover. Organizational performance can be defined as the level of performance that the participants of the organization have achieved. Some empirical support for this contention was located in a study in which the researchers examined the relationship between turnover and organizational demography in a top management sample of 31 Fortune 500 companies (Wagner, Pfeffer, & O’Reilly, 1984). They included a measure of organizational performance as a control variable in this study. This measure was average return on investments (ROI) corrected so that industry type would not affect the ROI. When entered into a regression analysis, this measure of performance was found to be a significant predictor of turnover rate. It is possible that the relationship between performance and turnover may be [‘6‘ CG ,rr 49 reversed; that turnover may predict organizational performance, but this hypothesis will not be considered in this research. Organizational Climate. The history of the climate construct has been one fraught with debate. Two of the topics of debate include the definition of organizational climate, and whether climate and satisfaction represent the same construct. This section will attempt to address these topics, and then explains how organizational climate fits into the proposed model. Individual climate is typically defined as a person’s perceptions of a specified context, such as the work group or the organization, in which he/she is a participant (Rousseau, 1988). However, the definition of organizational climate has generated considerable controversy. James and his colleagues conceptualize organizational climate as an aggregate of individual-level climate perceptions (James, 1982; James, Joyce & Slocum, 1988). When perceptual agreement is demonstrated, the aggregate of these individual-level perceptions represents a "shared assignment of meaning" (James et al., 1988, p. 129), or organizational climate. In contrast, Glick (1985; 1988) conceptualizes organizational climate as an organizational-level construct (similar to structure), and not a perceptual construct (Rousseau, 1988). Glick (1985) contends that individuals may have inaccurate perceptions T—____7 5O 3 of the organization’s climate, and suggests that the organization should be the unit of analysis. Furthermore, Glick (1985) argues that instead of just aggregating individual responses, a sample should include different types of organizations, and researchers should use multiple sources of information to ensure that the measurement of climate is valid. Rousseau (1988) indicated in her review of the literature that Glick’s understanding of organizational climate does not fit with other models that conceptualize organizational climate as a cognitive construct. Climate conceptualized at the organizational level as a cognitive construct reflects the shared perceptions of the organizational members of the context or setting in which they work (Rousseau, 1988). According to Rousseau (1988), the assumption underlying the aggregation of individual-level climate responses is that units or organizations have different climates. Empirically, one can demonstrate that organizations have unique climates by examining the between-unit differences of aggregated climate responses (Rousseau, 1988). Joyce and Slocum (1984) discuss a different method to use when aggregating climate perceptions. They refer to aggregated climate perceptions as "collective climate". They suggest clustering individuals on the basis of the similarity of their responses to the climate measure, or pro. r) ,__J c: \he sat 51 profile similarity. This means that the criteria used to cluster individuals is agreement on climate perceptions. When using this method, employees do not have to be in the same work group or division to be clustered together; the main criterion is perceptual agreement. This conceptualization of climate is somewhat different from that of Schneider (1987) and Rousseau (1988) who suggest that because units within an organization (e.g., work groups or divisions) share common experiences, they will have similar climate perceptions. However, Rousseau (1988) also notes that employee interactions are thought to be important in collective climate formation. At least two studies have been conducted using this clustering technique, and the results of the studies provide support for this technique (Jackofsky & Slocum, 1988; Joyce & Slocum, 1984). When defining the organizational climate construct, one also must consider whether the construct is a general summary measure of what occurs in the organization, or as a multi-dimensional construct. At the individual—level of analysis, some researchers have used summary measures of organizational climate (e.g., Drexler, 1977), but most have used a variety of climate scales (e.g., supervisor suPport) to measure the construct (e.g., Kozlowski & Doherty, 1989; Schneider & Snyder, 1975). Drexler (1977) noted that he used a summary measure of climate because of T—7 52 the high intercorrelations between the various climate scales. A summary measure of climate was used in the only current organizational—level study that included climate as a predictor (Terborg & Lee, 1984). One current View is that the use of various scales is more informative because it provides the research with more precision (Rousseau, 1988). Additionally, the definition of a construct should include a statement of how the construct differs from other similar constructs. Guion (1973) expressed concern that organizational climate and satisfaction were both measuring the same construct. Later individual—level studies suggested that climate and satisfaction are distinct concepts if they are conceptualized and measured as separate constructs (LaFollette & Sims, 1975; Schneider & Snyder, 1975; Woodman & King, 1978). Joyce and Slocum (1979) noted that although the measurement of these two constructs may at times be similar, the conceptualization of the two constructs are not the same. The climate of a work group or organization should be conceptualized as a description of the activities that occur within the work area. In contrast, satisfaction is the affective response to that work area. One other question that might be asked is why We would expect organizations to have different climates. Schneider (1987) would argue that climates differ across 53 organizations because of the individuals within the organizations. In fact, Joyce and Slocum (1979) state that "organizations have climates in the same way that individuals have personalities" (p. 333). The contention of Schneider (Schneider, 1987; Schneider & Reichers, 1983) is that the climate within an organization forms through an attraction-selection—attrition (ASA) process. More specifically, individuals choose organizations or jobs to which they are attracted. In addition, organizations attempt to select individuals who seem to have goals and attitudes that are similar to those the organization expects. The ASA approach also suggests that when individuals are hired who do not have the same beliefs and views about the organization, or do not fit into the work group, they leave the organization. As Schneider (1987) stated "... people who don’t fit an environment well will tend to leave it" (p. 442). Furthermore, Kozlowski and Hults (1987) assert that individuals within an organization will have similar climate perceptions because of the interaction and socialization processes that take place within the organization. Now that the definition of organizational climate has been discussed, the next section will address how the construct fits into the proposed model. First the composition of the proposed climate construct will be discussed, followed by an attempt to justify the proposed % 54 paths in the model as well as one path that is not being proposed. As mentioned earlier, researchers have usually conceptualized the climate construct as consisting of a number of different dimensions. However, the dimensions used in research often differ from one study to another. There are a number of dimensions that could be of interest in the present study. The specific dimensions that are being used in the present study are coworker competence, degree of autonomy, participation in decision making, physical environment, and student behavior. Each of these dimensions will be briefly defined in terms of the sample being studied (secondary school teachers) before discussing the way in which they relate to other constructs. Coworker competence is defined as the degree to which teachers help students learn. The degree of autonomy consists of the teachers’ perceptions of the degree to which a supervisor’s approval is necessary in order for a decision to be made. The participation dimension assesses the degree to which teachers perceive that they participate in school decisions. The physical environment is defined as the safety and neatness of the school, and the final dimension, student behavior, assesses the way in which students behave and work. Each of the climate dimensions are hypothesized to have an effect on satisfaction. Furthermore, three of the 1 55 dimensions (coworker competence, autonomy, and participation) have been studied in the climate literature, and have been shown to be related to affective variables (e.g., Pritchard & Karasick, 1973; Jackofsky & Slocum, 1988). The physical environment and student behavior dimensions are particularly important because of the sample under study (i.e., secondary schools). The environment and the student behavior would be important to. teachers who interact with students on a regular basis. 1 In fact, Newman (1977) did find a significant relationship between some physical environment dimensions and satisfaction at the individual-level of analysis. Student behavior as a climate dimension has not been specifically studied as a correlate of affective variables, but it is similar to other climate dimensions that have been studied (e.g., supportiveness, peer relations). In Figure 2, the five climate dimensions are represented as indicators of organizational climate. As mentioned earlier, the inclusion of the climate construct in turnover research at the individual level has occurred only recently. The results of this research has suggested that climate is an important part of the turnover process. At the organizational level, it has been hypothesized that turnover will be influenced by different climate dimensions such as centralization and participation (Mobley, 1982a; Price, 1977; Terborg & Lee, 56 1984). One study was found in which the researchers did examine the relationship between organizational climate and turnover rate, but their findings were mixed. Their results were different for different samples (management and sales personnel), and the results also differed across time periods for the same samples (Terborg & Lee, 1984). Research at the individual—level of analysis, along with the inconclusive results reported by Terborg and Lee (1984) suggests that the proposed climate dimensions may also make a contribution at the organizational—level of analysis. Thus, it is hypothesized that organizational climate will have an indirect effect on turnover rate, mediated by satisfaction. One other possible link that should be considered in this study is the path between organizational climate and organizational performance. Researchers have reported different levels of relationship between climate and performance at the individual and organizational levels. At the individual level, some researchers have found that climate was related to performance (e.g., LaFollette & Sims, 1975). However, Jackofsky and Slocum (1988) examined this relationship over two time periods and found no relationship between the two constructs at time 1. However, a significant relationship was reported during the second measurement period. 57 At the unit or organizational level, empirical evaluation of this relationship has also produced inconsistent results. Pritchard and Karasick (1973) found that climate was related to subunit performance. In contrast, Schneider and Snyder (1975) reported that climate was not strongly related to production data. The results of a third study indicated that the relationship between organizational performance data and climate was negative (Heller, Guastello & Aderman, 1982). Because of the generally discouraging results of studies of this relationship, no path between climate and organizational performance is being hypothesized. Work—Related Correlates Satisfaction. Although employee satisfaction is usually perceived to be an individual—level variable, the aggregation of satisfaction to the organizational level is based on the assumption that the affective reactions of employees in general to various aspects of their work place differ across organizations. For instance, employees as a group in one organization could report being more or less satisfied than employees in another organization. James, et al. (1988) argue that an organization cannot have an attitude. This statement may be understood to mean that researchers should not aggregate affective constructs such as satisfaction, although other 58 researchers do discuss organizational-level affective constructs (e.g., Angle & Perry, 1981; Staw, 1980). However, aggregated affective variables are conceptualized in this paper, not as an organizational attitude, but as a description of the feelings that employees have in general in one organization as compared to employees’ attitudes in general in another organization. The relationship between job satisfaction and turnover has not yet been examined at the organizational— level of analysis, but has been studied at the group or division level. Two studies that have examined this relationship at the group level reported that the relationship was nonsignificant (Kerr, Koppelmeier, & Sullivan, 1951; Katzell, Barrett, & Parker, 1961). However, the literature clearly supports this relationship at the individual—level of analysis. Furthermore, the group—level findings are based on only two studies, neither of which attempted to evaluate statistically whether the constructs could be considered group—level constructs. Thus, this study includes a path from average satisfaction to turnover. Dependent Variable Turnover Rate. The only variable in the proposed model that has not been discussed is the turnover rate. Muchinsky and Tuttle (1979) stress the importance of reporting the type of turnover measure used in research 59 because of the large number of turnover measures available; Gaudet (1960) describes 25 turnover measures. There are measures which'focus on length of service, voluntary turnover, dismissals, the number of new workers hired, and individuals who stay with the organization (Gaudet, 1960; Price, 1977). A description of the various types of turnover measures can be found elsewhere (Gaudet, 1960; Price, 1977). 'Muchinsky and Tuttle (1979) report that some form of a yearly percentage rate is usually found in turnover research. The proposed research uses what Price (1977) labels the instability rate which is the number of workers who have voluntarily left the organization divided by the total number of individuals who were employed at the beginning of the measurement period. This measure will be used for two reasons. The focus of this research is to understand the determinants of organizational turnover, thus all individuals who have voluntarily left should be included in our measure, not the workers that were fired or just the newcomers. More importantly though, the use of this measure will make the results more readily comparable to others that have used this type of measure (e.g., Terborg & Lee, 1984). Although researchers have made a distinction between functional and dysfunctional turnover, this distinction is still exploratory. This study will focus on turnover BE [—1 r‘r‘ 60 ignore the functional and for the present, frequency, it would be interesting to aspect of turnover. However, evaluate the relationships proposed in Figure 2 and a measure of functional turnover in a later study. Hypotheses The relationships suggested in Figure 2 represent the major hypotheses of this proposal. Each of the hypothesized relationships in Figure 2 will be presented beginning with the constructs that are shown to have direct effects on actual turnover. H1: Labor market conditions, salary, satisfaction each will have a direct negative effect performance, and on turnover rate. H2: Climate will have an indirect effect on turnover. Satisfaction will mediate the indirect relationship The relationship between climate and turnover. between climate and satisfaction will be positive. Satisfaction will mediate the relationship between The relationship H3: union presence and turnover. between union presence and satisfaction will be positive. Satisfaction will mediate the relationship between The relationship H4: supervisor turnover and turnover. between supervisor turnover and satisfaction will be negative. H5: Organizational size will have an indirect effect on turnover mediated by organizational climate and satisfaction. The relationship between organizational size and climate will be negative. H6: Labor market conditions may moderate the relationship between satisfaction and turnover rate. When labor market conditions are positive (unemployment rate is the relationship between satisfaction and low), turnover will be stronger than when labor market conditions are negative (unemployment rate is high). METHOD The data for this study were collected as part of a larger research project. In 1985, a conceptual model was developed to examine the determinants of school effectiveness by the National Association of Secondary School Principals (NASSP). An initial pilot study was conducted to examine the model (Schmitt & Ostroff, 1987), and the results from the pilot were used to revise the initial measures. The data for the present study along with other information were collected to examine how well the revised instruments measured the specific model components and to test the NASSP school effectiveness model. The results of the larger project can be found in Schmitt and Doherty (1988). Sample Three hundred and sixty four schools from 36 states and Canada were used in the Schmitt and Doherty (1988) study. These schools were contacted to provided further information for the present study. From that group of schools, 188 of the schools provided usable information, and were used in the present study. There were 41 schools (21.8%) that were junior-high schools, 126 (67.0%) that were at the senior-high level, and 21 schools (11.2%) that 61 62 reported having both junior and senior-high students. The school principal and teacher samples as well as the questionnaires used are described in the following sections. Principals. The principal from each of the 188 schools was asked to respond to a two-part questionnaire. The first part of the questionnaire was in multiple—choice format, and the second part of the questionnaire consisted of open—ended questions. The information that was obtained from the principal questionnaires is the size of the organization, organizational performance, supervisor turnover, and average teacher salary. These items can be found in Appendix A. Demographic information on the principals can be found in Table 2. Teachers. The teacher sample consisted of 7,691 teachers from the 188 schools. The number of teacher respondents per school ranged from 12 to 86 with a mean number of teachers being 51.08 (standard deviation = 26.14). Teachers were not asked to provide demographic information, but were asked to respond to questions regarding school climate and satisfaction. Each of these items grouped by scale can be found in Appendix B. 63 Table 2 Demographic Characteristics of Principal Sample1 Gender Male Female Ethnic Status Asian American Black Hispanic White Tenure Less than 1 year More than 1 year, but less than 2 More than 2 years, but less than 3 More than 3 years, but less than 4 More than 4 years, but less than 5 More than 5 years, but less than 8 8 or more years 1 Sample Size = 188 92.0% 8.0% 3.7% .5% 94.7% 9.0% 11.2% 13.8% 6.9% 6.4% 16.6% 36.2% 64 Measures Organizational Size. Principals provided this information by responding to an open-ended question which asked the principals to report the number of students enrolled in their school. A second measure of organizational size was computed by dividing the number of students in the school by the number of full-time equivalent teachers to get a measure of the student— teacher ratio in the school. Organizational Performance. This variable was operationalized by using student achievement measures on standardized tests. The principal was asked to report test information by grade on reading comprehension, arithmetic, and science for grades 6 through 12. For each grade level in the school, the principal was asked to provide the type of test used (e.g., California Achievement Test), the actual score for the grade, and whether the score reported was an average percentile or an average normalized curve equivalent score (NCES). When the data were received, all of the NCES scores were converted into percentile scores. The three achievement variables were combined into a composite measure of performance since the intercorrelations between the variables were relatively substantial (r > .78). 65 Additionally, research has shown that student achievement is affected by social economic status (SES). Thus, a measure of SES was first partialled out of the relationships with the student achievement measure. The SES variable used was the percentage of students in the school that received free or reduced-price lunches. After lunch was partialled out of the relationships between the three achievement measures, the intercorrelations between these variables was still greater than .74. Supervisor Turnover. This measure was obtained by asking the principal to report the number of principals or headmasters that have served in the school in the past decade. Average Teacher Salary. This was obtained with an open-ended question that asked the principal to report the average teacher salary in the school for the year that the data were collected. Average teacher salaries ranged from 18,800 to 44,000 by state across the country, so this measure was corrected by area of the country. This was done by assigning the average state salary to each school in the sample. The average state salary was then partialled out of the relationships between salary and other constructs. Appendix C contains the average teacher salaries by state. 66 Union Presence. This measure was obtained by asking the principal to report the percentage of the teachers in the school that were unionized. This percentage was used as the measure of union presence. School Climate. Teachers reported their perceptions of five dimensions of school climate. The dimensions used were coworker competence, physical environment, student behavior, participation in decision making, and degree of autonomy. If these climate dimensions are highly correlated, they will be combined into one measure of organizational climate. The items in the first three dimensions were taken from a larger group of items used in the Schmitt and Ostroff (1987) pilot study. Most of the items in this larger group of climate items were obtained from the NASSP Climate Survey (Kelly, Glover, Keefe, Halderson, Sorenson & Speth, 1986). As mentioned earlier, the results of the Schmitt and Ostroff (1987) study were used to revise the measures used in the survey in this study. Schmitt and Ostroff (1987) had used ten climate dimensions to assess school climate, and found that the climate dimensions were highly intercorrelated. In an attempt to reduce the number of items in the survey, they selected the climate item from each dimension that had the highest item—total correlation with its dimension. These items were used in the present study to measure school 67 climate. However, upon examination of the items chosen by Schmitt and Ostroff (1987), I felt the items could be grouped into three dimensions by item content (i.e., coworker competence, physical environment, and student behavior). The competence dimension consists of four items regarding how teachers help students learn. The physical environment dimension consists of two items regarding the safety and neatness of the school. The third dimension, student behavior, includes three items which asked about the way students behave and work. Teachers were asked to respond to these items by stating the degree to which most individuals in the school or community would agree with the items. The possible responses ranged from strongly disagree (1) to strongly agree (5). The last two climate dimensions were also used in the Schmitt and Ostroff (1987) study. The participation in decision—making dimension included four items which assessed the frequency of teacher participation in four types of school decisions (Hage & Aiken, 1967). The coefficient alpha for this scale was .76 in the pilot study (Schmitt & Ostroff, 1987). The possible responses for the scale ranged from never (0) to always (4). The final climate dimension, degree of autonomy, consisted of five items used to determine the teachers’ 61’ (I) 68 perceptions of the degree to which a supervisor’s approval is necessary in order for a decision to be made (Hage & Aiken, 1967). Schmitt and Ostroff (1987) reported finding a coefficient alpha of .91 for this scale. Possible responses ranged from definitely false (0) to definitely true (4). Satisfaction. The satisfaction items were chosen by Schmitt and Ostroff (1987) in the same way they selected the climate items. In the pilot study, nine scales were used to assess various satisfaction dimensions. In an attempt to be more parsimonious, the best items from these scales were chosen. The result was a nine—item scale that assessed the teachers’ satisfaction with different aspects of the school. The scale responses ranged from very dissatisfied (1) to very satisfied (5). Labor Market Conditions. There are a variety of ways that labor market conditions can be operationalized for this sample. The best measure would be teacher unemployment by state, but attempts to locate this type of measure were unsuccessful. Instead, two other measures were selected. The first was a measure that was obtained from a publication entitled "Geographic Profile of Employment and Unemployment, 1987" (U. 8. Bureau of Labor Statistics, 1988). In this table, employment and unemployment statistics were provided for different 00( (I: 69 occupational groups. The group that was appropriate for teachers was the professional specialty group. The numbers reported in Appendix D are unemployment rates (in thousands) for the professional specialty group. A second measure that was used is an annual report written by James N. Akin (1989) which provided supply/demand ratings for both secondary and elementary school teachers by region of the country. This measure was obtained by sending surveys to 502 teacher placement officers in December of 1987. The placement officers were asked to indicate their perceptions of current and future teacher employment opportunities. Out of the 502 surveys, 247 (49%) were returned. The continental United States were divided into nine areas, and Hawaii and Alaska were considered separate areas. The responses of the placement officers were collapsed within an area of the country. No information was provided on the number of placement officers responding in each of the areas of the country. Additionally, interrater reliability analyses were not reported. Thus, there is no indication of the degree of agreement between placement officers within the areas of the country. The supply/demand ratings can be found in Appendix E. If the school was a junior high school, the secondary elementary school ratings were used. However, if the school was a senior high school or a combined 70 junior-senior high school, the average ratings at the bottom of the table were used. Thus, the school was assigned one of ten ratings depending on the level of the school and the area of the country in which the school was located. Turnover Rate. As mentioned in the introduction, there are a number of different measures of turnover. The measure chosen in the proposed research for reasons discussed earlier is the number of workers who had left the organization divided by the total number of individuals who were employed at the beginning of the measurement period. In the fall of 1987, principals provided information on the number of teacher full-time equivalents (FTEs) employed at their schools. The measure of actual turnover was obtained from principals. In early 1990, principals were asked to provide information on the number of teacher FTEs who were employed in the fall of 1987, and who had subsequently left the school (during the 1988—89 school year). The principals were asked to include only those teachers who voluntarily left the school. Because it is being suggested that the various constructs be used to predict turnover rates, there should be a period of time that intervenes between the collection of the predictor data and the collection of the criterion. This measure 71 can be found in Appendix F. The measure also asked principals to provide ratings of their teachers’ skills and abilities, but these ratings were not used in this research. Measurement Issues Three major issues that need to be addressed in a study such as this are levels of analysis, potential method bias, and the effect of unmeasured variables on the evaluation of the proposed model. Each of these issues is discussed. Levels of Analysis Issues Levels of analysis research often includes variables that have been aggregated to represent a higher-level construct. When researchers do aggregate variables to a higher level, they should provide support for the conceptualization of a variable as a higher—level construct. Composition Modeling. One method that provides evidence that constructs can be considered higher level variables is the use of composition modeling. A composition model must include constructs that exist at more than one organizational level. In other words, constructs that can be measured at one level seem to exist in a similar form at another level. Ostroff and Kozlowski (1986) contend that when developing a composition model, U! 72 adjacent levels in the organizational hierarchy should be used (e.g., individual and group) in the model. In other words, levels should not be omitted from the hierarchy. However, organizations do exist (e.g., small businesses, pre-college schools) in which the individuals at the group level and those at the organizational level are the same persons. When describing the degree of similarity between a construct at one level and the construct at another level, researchers use the term isomorphic. It is necessary to satisfy three criteria in order to demonstrate the isomorphism of different forms of a construct. First, a construct that exists at more than one level may be considered the same construct if the definitions are similar (James, 1982). Second, the relationships between potential isomorphic constructs at one level must be similar to the relationships between the same constructs at another level (Rousseau, 1984). For example, if the relationship between climate and satisfaction at the individual level is a moderate, positive correlation, then the relationship between climate and satisfaction at the unit level must also be moderate and positive in order for both climate and satisfaction to be considered isomorphic constructs. The final criterion is only important when aggregation is necessary. According to James (1982), 73 perceptual agreement must be demonstrated before a higher level construct may be considered to be equivalent to a lower level construct. Aggregation and perceptual agreement are integral issues in levels of analysis research so each of these is discussed in further detail. Aggregation. One potential problem when attempting to develop a composition model is that often lower level variables must be aggregated to form higher level variables because of the nature of the constructs. For instance, the usual method used to find out the satisfaction level of a group of people is to aggregate individual satisfaction responses. However, researchers have argued that "global (indivisible) data" are more appropriate to use than aggregate data (Roberts, et al., 1978; Rousseau, 1985). An example of a global satisfaction variable may be an item that requires a unit supervisor to report his or her perception of the satisfaction of the unit. Researchers suggest that aggregate data may be ambiguous and misleading because it is not obtained directly from the level to which it is attached (Roberts, et al., 1978; Rousseau, 1985). However, because of the nature of some variables, the most representative and practical method to use when examining a higher-level variable (e.g., satisfaction) is aggregation. Asking a supervisor for his/her perception of ii if FF} H L 1‘ \- 74 of the satisfaction level in the group can be very misleading if the supervisor’s perception is incorrect, or if there is reason for them to distort their responses. To illustrate, Kerr (1947) obtained a departmental morale rating from the personnel manager and another joint rating of departmental morale from a vice president and the union president of a company. Kerr (1947) then correlated these ratings with the turnover rate of the department. The results of his analysis indicated that the correlation between the personnel manager’s rating and turnover was -.02, while the correlation between the joint rating and turnover was —.44. These results show that different individuals in a company can have different perceptions of departmental morale. Even though some variables, such as satisfaction, may be best examined by aggregating individual responses, care must be taken to ensure that perceptual agreement exists, and that the data is not contaminated by problems such as method bias. In this study, the variables that are aggregates of individual- level constructs are climate and satisfaction. Rousseau (1985) also contends that when constructs Will be aggregated to a higher level, the items used to measure those constructs should be worded at the level to Which they will be aggregated. For instance, when aggregating climate to the group level, respondents should 75 be asked about group perceptions of climate. In the present study, the climate items were asked at the organizational level, but the satisfaction items were asked at the individual level. Perceptual Aggeement. According to James, et al. (1988), if employees within an organization have similar perceptions of a construct (e.g., climate), aggregation is possible because this agreement "implies a shared assignment of meaning" (p. 129). An aggregated construct is a measure of the responses of the employees in general. The aggregation of these individual—level variables also indicates that the researcher is contending that various units or organizations can have differing amounts or levels of these constructs. In essence, organizations as a whole have employees with different degrees or perceptions of climate and satisfaction that range along a continuum. Researchers that examine levels of analysis issues argue that before a construct is aggregated, agreement within the organization must be demonstrated (James, 1982)v This agreement within an organization is an indication that employees share similar perceptions regarding the construct for which agreement has been demonstrated. One way to assess agreement within an organization is to compute eta squared which compares the within-organizational variance to the 76 between—organizational variance of the construct. For example, Zohar (1980) compared the variance of climate scores within factories to the variance of scores between factories to determine if the climate scores within the factories were homogeneous. The results of his analysis of variance were significant, and he concluded that the climate perceptions of workers within each factory were similar. While researchers have not specified an acceptable level of eta squared, James (1982) stated that the median eta squared reported in the literature is .12. For each aggregated variable in the model, an eta squared was calculated to assess agreement. §Q§Qification of Levels of Measurement and Analysis. There are two types of levels that should be specified in any study in which level is an issue. The level of measurement refers to the level from which the data were Obtained. The level of analysis refers to the level to Which the data are assigned for analysis (RObePtS et 31" 1978; Rousseau, 1985). In this study, the unit 0f analysis is the school. However, the unit of measurement for the teacher scales is the individual. It was argued earlier that the teacher scales could be aggregated to the Organizational level because hypotheses about the effect Of these variables on organizational—level turnover rates ' ' ' h se scales are conceptually meaningful. Additionally, t e 77 an be aggregated if individuals within an organization have similar perceptions or feelings (James, et al., 1988). The teacher scales that were aggregated were the climate dimensions and satisfaction. Method Bias When measuring more than one construct that is perceptual, method bias is one potential problem. There are three steps that can be used to increase the possibility that method bias has not inflated the relationship between the perceptual constructs. The first two steps occurred during the construction of the measures. The third step took place during the construction and analysis of the data set. The first step used was to ensure that the instructions and responses were clear so that the teachers understood how they should think about each question when responding. For instance, three of the climate scales stated that a description of the school from the standpoint of most individuals in the school or community was the desired response. In contrast, the directions to the satisfaction scale asked the respondents to report their feelings abOUt various aSpects of the school. The second step that was used to attempt to control method bias is that several of the variables are quasi- 1 Objective in that the data was collected from severa r") [(7 78 sources. Specifically, the principals provided information on supervisor turnover, average salary, organizational performance, organizational size, and actual turnover, while teachers provided information on climate and satisfaction. Furthermore, one variable (labor market conditions) was obtained from a third source (e.g., library sources). Finally, the items in all of the scales were subjected to a principal-axis factor analysis. This analysis provided information on the possibility that method bias exists. The factor analysis should yield several factors, one for each scale. If the factor analysis should result in one global factor, this will provide evidence that the subjects did not distinguish between the various constructs (Harman, 1967; Podsakoff & Organ, 1986; Kozlowski & Doherty, 1989). Method bias could be cited as one cause for subjects not being able to differentiate between constructs. Unmeasured Variables. Another problem with testing a model such as the one proposed is that of unmeasured variables (Billings & Wroten, 1978; James, 1980). Billings and Wroten (1978) assert that in studies they have reviewed, missing variables often may have been responsible for relationships found between other constructs. James 79 (1980) contends that a missing variable is not the problem, but rather that the degree to which the missing variable biases the estimates of the existing path coefficients is problematic. A missing variable need not seriously bias the existing paths. According to James (1980), a missing variable will not bias the results of a path model if the variable meets certain criteria. First, if the missing variable is highly correlated with a measured variable, the results will not be affected. Second, bias will not occur if a missing variable does aCcount for variance in the dependent variable, but is unrelated to the other predictors in the model. This type of missing variable problem will result in the model explaining less variance in the dependent variable, but the other existing path coefficients will not be affected. The development of causal models of turnover at the organizational level is a fairly recent addition to the turnover literature. Therefore, it is difficult to be sure that all of the important variables have been included in the model. However, one variable missing from the proposed model that has been found to be related to turnover in the literature is tenure or tenure gap. The fact that this variable is missing should not bias the other path coefficients because the relationship between tenure and turnover would have been a direct relationship. 80 Additionally, the correlations between tenure and the other predictors should be low or nonsignificant. Thus, the existing paths in the model will not be affected by this missing variable, but the amount of variance explained in the dependent variable will be lower than if tenure had been included. Analysis A number of basic statistics, such as means, standard deviations, intercorrelations between variables, and reliabilities, were computed to ensure that there are no coding or data definition problems and that the scales measure the intended constructs. Two other preliminary analyses are necessary before the hypotheses can be tested. Three of the climate dimensions were formed on a conceptual basis. A factor analysis was conducted to determine if the climate items loaded on their respective scales. The second set of analyses that were needed were the computation of eta squares for the various teacher scales to assess agreement and the appropriateness of computing aggregate statistics. Before testing the full model in Figure 2, moderated regression was used to find out if the condition of the labor market moderates the relationship between turnover intent and turnover rate. If the results indicate that the labor market does moderate this relationship, then a 81 series of regressions will be used to test the model in Figure 2. If the moderated regression results are not significant, then the SPSSX program, LISREL (Joreskog & Sorbom, 1986), will be used to test the model. LISREL is a program that can be used to estimate the path coefficients in a group of linear structural equations (Joreskog & Sorbom, 1986). In order to illustrate how the proposed model was evaluated, the model is presented using matrix equations (see Table 3). Table 3 includes both the structural equation and the measurement equations. Figure 3 provides a graphic representation of the matrix equations. However, the proposed model consists of constructs that are represented by single indicators. With the exception of organizational climate and organizational performance, each construct in Figure 3 is being measured by one item or one scale. Thus, the measurement model cannot be tested by LISREL, but the measurement model equations are included for completeness. The reliability of the measures was assessed prior to testing the structural components of the model. The first matrix equation in Table 3 represents the structural equation model, which provides information about the causal relationships between the latent variables and describes the unexplained variance. The first part of the equation represents the paths between 82 Table 3 Structural and Measurement Equations for the Proposed Model Structural Model Equation - lst Equation CLM SAT AT CLM O O O CLM :2 SAT ,9. 0 0 SAT AT 0 82 0 AT ._. _l _. _, SZ SAL PRF STO UP LMC CLM Y, O O 0 0 O 82 SAT 0 o 0 Y4 Y5 o SAL AT 0 Y2 Y3, o 0 Y6 PRF STO a UP r1: I LMC ”£2 — '— is L J 1 CLM = Climate, SAT 2 Satisfaction, AT = Actual Turnover, 82 = Organizational Size, SAL = Salary, PRF : Organizational Performance, STO = Supervisor Turnover, UP = Union Presence, LMC = Labor Market Conditions. 83 Table 3 (cont’d) Structural and Measurement Eguations for the Proposed Model Measurement Model Equations (Endogenous Variables - 2nd Equation) {— -H »-—- — —— --— — . —- Y1 1 o 0 ‘fl‘ 8‘ Y2 : f 0 1 O “a + Ea Y3 0 O 1 Y1}, LE3 _. _J _‘ __ _ a _‘ (Exogenous Variables — 3rd Equation) _ .fl _. _1 __ _1 _. ._ X1 1 o o 0 0 0 ‘5‘ 0“ X2 0 1 o 0 0 0 E2 J2 X3 0 0 1 o 0 o E} + J3 X4 - o o o 1 o 0 Z4 J, X5 0 o 0 o 1 0 E5 is 0 O 1 £6 6 _-_~—~1._—~--:——_-__ «_.__‘_ _-.i_~h.__. H__—.--_—. - 84 mm / 0.6m 595:2; 0:50.330 no 503:5: oocaEtoebon .aco:¢u.caubo m? 0:60— .03000... acovgw /\ 0.2m .dco:a~_cau..0 .m: Cr cocoaonEOO acoficostEw / Gama-2.0 .aco3nw.cao..0 Eco—2 fen. Deceden— GF aux NQ 9.32.230 aorta—2 .3an ns).-com «conga—m d/ co_~an_o.:¢& >Eoco~:( NM . QE 1V! , 00:003.". cozoaaozdw m? COED Jr >53¢m 525:2; 099.22 boo—>505.” 05 a0 cozwucmwoacmm 2:936 m 9:9“. 85 the endogenous variables in the model (e.g.,é3 1 is a path coefficient between climate and satisfaction). The relationship between the exogenous and the endogenous variables is represented in the second part of the equation (e.g.,‘Y’l is a path coefficient between size and climate), and the last part is the error associated with this equation (e.g.,‘C 1). The second and third equations in Table 3 represent the measurement equation model, and link the latent constructs to the variables that are being used to measure the constructs. The second equation represents the relationship between the endogenous latent constructs (.n' to 713) and their observed indicators (\fl to ¥5), and the error ( E, t£>;3) associated with the relationship. The last equation in this Table is similar to the second, but represents the relationship between the exogenous constructs (‘3, to E5 ) and their observed indicators ( X, to XL. ) plus error (0“, 1;on ). The LISREL program was used to evaluate the structural equation model (Hayduk, 1987; Schmitt.& Bedeian, 1982; Williams & Hazer, 1986). Several parts of the LISREL output were used to evaluate the proposed model. The fit indices and the chi— Square statistic were examined to find out how well the model fits the data. Low values for these indices indicate that the data is well represented by the model. 86 The residual matrix was also be used to evaluate the fit of the model. The residual matrix should consist of low values indicating that the original correlation matrix, and the correlation matrix reproduced by the LISREL program were not very different. RESULTS Development of the Teacher Self-Report Measures Climate. The climate items were combined to create the five climate dimensions which are worker competence, environment, student behavior, participation, and autonomy. The reliabilities of the dimensions are reported in the diagonal of Table 4. These reliabilities are high, with the exception of the environment dimension (.58). As mentioned earlier, these dimensions would be combined into one summary climate measure if the dimension intercorrelations were high. An examination of the correlations in Table 4 indicates that the five dimensions cannot be combined into one measure. The correlations between some of the dimensions are very low. For instance, the correlation between the environment and Participation dimensions is .06, and the correlation between the autonomy and student behavior dimensions is .16. However, the correlations do suggest that the dimensions can be combined into two climate measures. The first measure could be labeled environmental climate, and 87 ercmp Envron Stdbeh Partic Autonm 88 Table 4 Climate Construct Intercorrelationsl ercmp2 Envron Stdbeh Partic Autonm (.87)3 .41 (.58) .59 .48 (.86) .42 .05 .35 (.78) .22 .08 .16 .53 (.95) 1 Sample size for all variables is 188 2 ercmp Stdbeh Autonm worker competence; Envron : environment; student behavior; Partic : participation; autonomy. 3 Reliability coefficients are presented in parentheses in the diagonal. 89 consists of dimensions that assessed the teachers’ perceptions of their surroundings. The dimensions included in this measure were worker competence, environment, and student behavior (mean r = .49). The second measure of climate could be called degree of latitude climate which included the participation and autonomy dimensions (r : .53). These dimensions assessed the degree to which teachers could participate in policy decisions or make their own decisions without consulting their supervisors. The reliabilities of these two summary climate measures were computed, resulting in reliability coefficients of .85 for the first measure, and .90 for the second measure. The intercorrelation between the two climate measures was .30. Satisfaction. The nine teacher satisfaction items were combined to form a summary measure of satisfaction. The reliability of this measure was .78. Descriptive Statistics of the Model Components Table 5 presents the means, standard deviations, and the ranges for the variables that are included in the Organizational-level turnover model. The construct intercorrelations and reliabilities can be found in Table 6- Many of the correlations between the predictors and ' ' ‘ 'ons turnover are very low. The only Significant correlati found in the table are between turnover and the number of 90 Table 5 Descriptive Information for Model Variables1 Variable2 Mean SD Range Climl 3.67 0.287 2.54 — 4.61 Clim2 1.30 0.227 0.84 — 1.99 Satis 3.26 0.293 2.48 - 4.17 Trate 0.03 0.045 0.00 — 0.24 Ratio 16.74 4.129 5.77 - 29.13 Studs 876.64 525.863 117 — 2541 Prins 1.10 1.025 0 — 6 Tcsal 27226.57 5719.312 15,250 - 42,000 Union 0.77 0.330 0.00 — 1.00 Dmand 3.17 0.300 2.40 — 3.56 Unemp 2.07 0.570 0.60 - 2.80 Perfm 66.38 14.406 6.0 - 96.8 1 Sample size for all variables is 188 Clim2 = Degree 2 Climl = Environmental climate measure, . . Of Latitude Climate measure, Satis = Satisfaction, . Trate = Turnover Rate, Ratio 2 Student - Teacher Ratio, Studs = Number of Students, Prins = PrinCipal Turnover, Tcsal : Average Teacher Salary, Union = Union Presence, Dmand : Teacher Demand Ratings, Unemp = Unemployment 3: Rating Perfm = Organizational Performance Zero-Order Correlations1 91 Table 6 between Model Variables2 Clim1 Clim2 Satis Trate Ratio Studs Clim1 (.85)3 Clim2 .298 (.90) Satis .654 .331 (.78) Trate .058 —.118 -.074 --- Ratio .115 —.029 —.054 -.136 --- Studs .127 -.130 .012 —.237 .557 —-- Prins .118 —.026 -.154 .109 .073 -.006 Tcsal .009 .021 .131 —.298 .137 .328 Union .115 .138 -.040 — 271 .033 -.008 Dmand .135 -.133 —.181 052 .100 .234 Unemp .038 -.029 .067 - 167 .445 .199 Perfm .509 .099 .287 - 020 —.110 ~.045 r — .127, p < .10 r - .154, p < .05 1 Sample size for all correlations is 188 2 Clim1 : Environmental climate measure, Clim2 = Degree of Latitude Climate measure, Satis = Satisfaction, Trate = Turnover Rate, Ratio : Student - Teacher Ratio, Studs = Number of Students, Prins = Principal Turnover, Tcsal = Average Teacher Salary, Union 2 Union Presence, Dmand : Teacher Demand Ratings, Unemp = Unemployment Ratings, Perfm = Organizational Performance Reliability coefficients are presented in parentheses in the diagonal. The variables with dashes in the diagonal are one—item variables. 92 Table 6 (Cont’d) Zero—Order Correlations1 between Model Variables2 Prins Tcsal Union Dmand Unemp Perfm Prins ———3 Tcsal —.028 ——— Union -.146 .405 —-- Dmand .167 —.190 -.276 --- Unemp .127 .300 .144 .071 -—- Perfm -.112 .101 .019 -.061 —.110 -—- r = .127, p < .10 r : .154, p < .05 Sample size for all correlations is 188 Clim1 : Environmental climate measure, Clim2 = Degree of Latitude Climate measure, Satis = Satisfaction, Trate : Turnover Rate, Ratio 2 Student - Teacher Ratio, Studs = Number of Students, Prins : Principal Turnover, Tcsal : Average Teacher Salary, Union : Union Presence, Dmand : Teacher Demand Ratings, Unemp = Unemployment Ratings, Perfm : Organizational Performance Reliability coefficients are presented in parentheses in the diagonal. The variables with dashes in the diagonal are one—item variables. 93 students, average teacher salary, unionization, and the unemployment measure. Similarly, there also high correlations between some of the predictors. For instance, most of the relationships between the climate, satisfaction, and performance measures are significant. Additionally, teacher salary is related to the number of students, unionization, and both of the labor market variables. The low zero-order correlations indicate that Some of the paths proposed in the model will not be supported. It also was necessary to compute partial correlations for the relationships between two of the variables with the other constructs. First, the percentage of students receiving free or reduced-price lunches was partialled out of the relationships between organizational performance (achievement) and the other constructs to control for social economic status. Additionally, the variance attributable to average state salary was partialled out of the relationships between average teacher salary and the other variables. These partial correlations are presented in Table 7. A comparison of the numbers in Tables 6 and 7 indicate that the partial correlations are not very different from the zero—order correlations. The one exception is between the unemployment measure and teacher salary. When the variance due to average state salary is 94 Table 71 Partial Correlations between Model Variables Perfm2 Tcsal3 Clim1 .427 -.028 Clim2 .101 .018 Satis .223 .113 Trate .019 . -.199 Ratio —.180 .007 Studs —.176 .325 Prins —.086 -.O77 Tcsal .104 -—- Union .044 .287 Dmand —.096 -.175 Unemp -.236 -.159 Perfm ——— .104 r z . 15, p < .05 1 Clim1 : Environmental climate measure, Clim2 : Degree of Latitude Climate measure, Satis : Satisfaction, Trate : Turnover Rate, Ratio 2 Student - Teacher Ratio, Studs = Number of Students, Prins : Principal Turnover, Tcsal = Average Teacher Salary, Union = Union Presence, Dmand = Teacher Demand Ratings, Unemp = Unemployment Ratings, Perfm 2 Organizational Performance 2 These correlations represent the relationships between organizational performance (Achievement) and the other variables controlling for lunch (SES). 3 These correlations represent the relationships between teacher salary and the other variables controlling for average state salary. Lunch and average state salary have both been partialled out of the relationship between organizational performance and teacher salary. 95 partialled out of this relationship, the correlation changes from .30 to -.159. Before describing the analysis of the model, two measurement issues will be discussed. Measurement Issues Common Method Variance. The relationship between climate and satisfaction was examined to determine if common method variance (CMV) was a possible explanation for the relationship. As mentioned earlier, the questionnaire was constructed so that the instructions and item responses were clearly written so that subjects could differentiate between the two constructs in an attempt to avoid this problem. After collecting the data, factor analyses were used to look for evidence of CMV. There are no standard rules that can be used to examine a data set for CMV. Podsakoff and Organ (1986) suggest that the unrotated factor solution should be examined to determine if CMV exists. According to these authors, if only one factor is present in the unrotated solution or if a general factor is responsible for a majority of the variance in the results, this suggests that CMV is present. However, the rotated solution should also be examined for CMV. For instance, if the factor analysis resulted in climate items loading on one scale, and the satisfaction items loading on another scale, this would indicate that subjects were able to differentiate between ti of factor in the fol In t1 climate at This anal: over one. factor ma' Table 8 31 large amO' account f analysis loaded on Satisfact dimension factor 1 represent dimensior dimenSiOr factor 4, participe Satisfact faCtor dc Th8} 96 between the climate and satisfaction constructs. A series of factor analyses were conducted, and will be described in the following sections. In the first analysis, the items that made up the climate and satisfaction constructs were factor analyzed. This analysis resulted in six factors with an eigenvalue over one. Table 8 presents the results of the unrotated factor matrix. The variance percentages at the bottom of Table 8 suggest that one general factor does account for a large amount of variance, but the other factors also account for variance. Examination of the rotated factor analysis indicates that most of the items in the matrix loaded on factors that could be interpreted as a satisfaction construct, or one or more of the climate dimensions (see Table 9). The items that loaded on factor 1 are the autonomy dimension items. Factor 2 represents the environment and student behavior climate dimensions. The items from the worker competence dimension load on factor 3, and satisfaction items load on factor 4. The fifth factor can be interpreted as the articipation dimension, and although two of the atisfaction items load on the final factor, this sixth actor does not represent any of the dimensions. There is another way to examine the data for CMV. In his second analysis, the computed climate and Unrotated ] FAC' SAT1 SATZ SAT3 SATA SAT5 SATS SAT? SAT8 SAT9 PARTl PARTZ PART3 PARTA AUT01 AUTOZ AUTOS AUTOA AUTOS ENV1 ENV2 SBEH1 SBEH2 SBEH3 VCMPI WCMPZ WCMP3 WCMP4 . . . . - o - c n O o O O D "‘ III "' (I) m m U. 07 0' 4‘- H U. o c on ’1. % Var2 3 SATl t Dartic items; SBEHB wOrkel % Var 97 Table 8 .tated Factor Matrix of Climate and Satisfaction Items1 FACTOR FACTOR FACTOR FACTOR FACTOR FACTOR 1 2 3 4 5 6 .52 -.15 .50 -.06 .39 -.16 .15 .26 .11 .69 —.09 .44 .41 .03 .17 .58 .16 .26 .60 .26 .23 —.35 .39 .10 .53 .10 .31 .16 .33 —.26 .54 .16 .05 .10 .22 -.51 .62 .39 —.04 .18 .06 -.06 .46 .36 .46 .13 —.29 .17 .56 .10 .34 .29 .20 -.07 1 .42 -.19 -.56 .10 .34 .13 '2 .30 —.28 -.41 .25 .36 .25 ‘3 .63 -.38 —.26 —.07’ .10 .14 '4 .66 —.27 -.30 .02 .01 .16 '1 .57 —.64 .11 -.08 —.09 .02 I2 .60 —.65 .12 —.08 —.13 —.01 '3 .61 —.66 .17 .00 —.15 .04 )4 .58 -.68 .13 —.03 —.22 —.01 )5 .53 -.69 .14 —.O5 -.19 —.05 53 35 13 -.14 — 26 30 44 .46 46 —.11 —.20 07 [1 67 .36 — 24 —.22 .06 05 '2 58 .41 — 06 —.46 03 .32 3 .60 .31 —.13 -.42 .05 .19 1 64 .29 —.33 26 — 30 -.27 2 66 .35 - 34 .13 — 22 — 24 3 .59 .16 —.34 .12 -.29 —.22 4 64 .31 — 14 —.O7 — 03 — 19 r2 30.9 15.1 8.1 6.7 5.0 4.6 AT1 to SAT9 are satisfaction items; PART1 to PART4 are articipation items; AUTOl to AUTOS are autonomy tems; ENV1 and ENV2 are environment items; SBEHl to BEH3 are student behavior items; WCMPl to WCMP4 are orker competence items. Var : Percent of variance accounted for by each factor. Rotated Fa FACT SAT1 SAT2 -. SAT3 SATA SAT5 SAT6 SAT? - SAT8 SATQ PART1 PARTZ PART3 PART1 AUTOI Atioz AUT03 AUT04 AUTOS ENV1 ENV2 SBEHl SBEHZ SBEH3 WCMPI WCMPZ wanes WCMP4 1—400 “Rm 0 o a o c . O . - - - - . a x . . H p—J r—A . . CJI OI H n o r/x IA m SATl 1 Part1: items SBEH3 wOl‘ke of th 98 Table 9 .otated Factor Matrix of Climate and Satisfaction Items1 FACTOR FACTOR FACTOR FACTOR FACTOR FACTOR 1 2 3 4 T1 .36 .19 -.10 .74 * —.03 .03 .T2 —.12 —.01 .10 —.03 * .05 .87 T3 .13 .01 .09 .30 x .20 .69 T4 .07 .69 .00 .53 x .08 —.03 T5 .13 .11 .18 .71 x 01 .14 T6 .07 .05 .48 .63 x 04 -.08 T7 -.02 .34 .50 .35 x 09 .27 T8 .12 ' .39 .21 .20 * - 43 50 T9 .18 .14 .18 .59 x — 01 .38 RT1 .17 .12 .23 .04 .76 x .01 RT2 .19 .00 .04 .04 .72 x .18 RT3 .56 .26 .20 .08 .47 x .00 RT4 .51 .27 .30 .03 .45 * 10 T01 .84 * .07 .04 .13 .13 —.01 T02 .88 x .06 .07 .14 .10 -.01 T03 .90 x 04 .05 .14 .08 .09 T04 .92 x .01 .09 .09 .05 .02 T05 .89 x —.03 .06 .10 .04 —.03 v1 .13 .63 x .25 —.01 -.19 .29 V2 .03 .53 x .18 .27 —.47 .27 EH1 .04 .64 * .46 .17 .21 —.01 EH2 .02 .88 x .19 .05 .07 .00 EH3 .08 .77 x .26 .10 .13 —.08 MP1 .10 .14 .88 x 08 .06 18 MP2 .06 .27 .83 * 11 .09 10 MP3 .18 .16 .75 x 01 .09 06 MP4 .07 .40 .58 x 27 .05 - 01 SATl to SAT9 are satisfaction items; PART1 to PART4 are participation items; AUTOl to AUTOS are autonomy items“ ENVl and ENV2 are environment items, SBEHl to SBEH3’are student behavior items; WCMPl to WCMP4 are Worker competence items. These items were part of the scale that loaded on each of the factors. satisfacti results of indicate 1 dimensions behavior) dimension: factor. 1 structure satisfact Anot method va climate 3 Construct if the pa SPECifica measure 5 COrrelatj this is 1 Vas pTeS( between . Satisfac. Organiza lS Simll organize However ’ 99 atisfaction scale scores were factor analyzed. The esults of the rotated factor matrix (see Table 10) ndicate that satisfaction, and three of the climate imensions (worker competence, environment, and student ehavior) load on one factor, while the other two climate imensions (participation and autonomy) load on a second actor. Finally, one last analysis examined the factor tructure of the two summary climate measures and atisfaction, and resulted in only one global factor. Another method can also be used to determine if ethod variance may account for the relationships between limate and satisfaction. The correlations between these )nstructs and other variables can be examined to find out f the pattern of correlations are similar. More Jecifically, if the correlations between one climate aasure and other variables are similar to the )rrelations between satisfaction and the other variables, 118 is further evidence that a method variance problem 18 present. As an example, consider the correlations atween the organizational performance, climate, and Ltisfaction measures in Table 6. The correlation between ‘ganizational performance and the first climate measure 1 Similar to the relationship between satisfaction and "ganizational performance (i.e., -509 V5' '287)' . ' ate )Wever, the correlation between the second 011m Rotated F Satisfaci Autonomy Particip: Worker C Student Environm 100 Table 10 Rotated Factor Matrix of Climate and Satisfaction Scales FACTOR FACTOR 1 2 Satisfaction .80 .21 Autonomy .07 .84 Participation .21 .86 Worker Competence .72 .34 Student Behavior .76 .21 Environment .84 —.14 measure a] the other This patt similarit To 8 of the va satisfact two clima satisfact ngg were usec and the s Organ12a1 COmputed individu, by Schoo sums of to deter school. the envi 1atitude mentione Eta Squa of theae 101 measure and organizational performance was not similar to the other two correlations (i.e., .099 vs. .509 and .287). This pattern of correlation similarity (or lack of similarity) exists throughout Table 6. To summarize, method bias probably accounted for some of the variance between the first climate measure and satisfaction. In contrast, the relationships between the two climate measures, and the second climate measure and satisfaction were most likely not affected by method bias. Levels of Analysis Issues. Two statistical methods were used to examine whether the two climate constructs and the satisfaction construct could be considered arganizational-level constructs. Eta squares were computed by conducting analyses of variance at the individual-level of analyses for each of these constructs 3y school. To derive the eta squares, the between—group sums of squares were divided by the total sums of squares LO determine the amount of variance attributable to the i0h001. These analyses resulted in eta squares of .27 for :he environmental climate construct, .15 for the degree of .atitude climate construct, and .19 for satisfaction. As lentioned earlier, James (1982) reports that the median (ta square reported in research is .12, so the magnitude f these eta squares was high. The data are consistent with the r at the org A 88( be consid« the corre analysis. correlati similar t the indi\ computed the sampj correlat magnitud The resu Satisfac the indi hypotheg moderat. tUPnove examine are pre UREmp10 tuI‘1'10Ve 102 ith the notion that the three constructs are meaningful t the organizational level. A second way to determine if these constructs could e considered organizational—level variables is to examine ,he correlations between the constructs at both levels of .nalysis. As can be seen, in Table 11, the patterns of :orrelations at the individual-level of analysis are imilar to those found at the organizational level. At ,he individual level of analysis, the correlations were omputed using only those teachers in the 188 schools in .he sample. According to Rousseau (1984), the orrelations at one level must be similar in direction and iagnitude for the constructs to be considered isomorphic. 'he results of these analyses suggest that the climate and .atisfaction constructs can be considered isomorphic to ,he individual—level variables. Analysis of Potential Moderators. Earlier, it was .ypothesized that the labor market variables might loderate the relationship between satisfaction and .urnover. Moderated multiple regression was used to xamine these relationships. The results of the analyses re presented in Tables 12 and 13, and indicate that the nemployment measures do not moderate the satisfaction — urnover relationship. In fact, these analyses show that Comparis< Lam Climatel Climatel ClimateE 1 The Were Clin Lati Table 11 Comparison of Individual—Level and Organizational-Level Zero—Order Correlations1 Variables2 Ind. Level Org. Level Climatel & ClimateZ .232 .298 (n:7513) (n=188) Climatel & Satisfaction .527 .654 (n:7527) (n=188) Climate2 & Satisfaction .398 .331 (n27653) (n=188) The data used for the individual—level correlations were the teachers in the 188 schools in the sample. Climatel : Environmental Climate; ClimateZ 2 Degree of Latitude Climate Regre Variables in regres equation Step 1: Demand R2 Satisfact Step 2: Demand X \ B-wei been 104 Table 12 Regression Analysis with Teacher Demand Ratings as the Moderator Variable 'ariables Multiple B—weight R2 F of n regression R Change Change equation :tep 1: Demand Ratings -.141 latisfaction .08 —.15 .007 .65 ltep 2: lemand X Satis .12 .04 .006 1.19 B-Weights are those reported after all variables have been entered into the equation. Regres; Variables in regres equation Step 1: Professic Unemplc SatisfaC‘ StEp 2: tnemp X \ B-wei been 105 Table 13 Regression Analysis with Professional Unemployment as the Moderator Variable 'ariables Multiple B-weight R2 F of n regression R Change Change ‘quation tep 1: ’rofessional Unemployment .051 atisfaction .18 .03 .03 .05 tep 2: ‘nemp X Satis .19 —.02 .005 .32 B—weights are those reported after all variables have been entered into the equation. none of ti for a sig The the hypot This prog are hypot the spec Sev to the L correlat tYpioall asserts correla‘ covaria Correla affect However PEsultE Complet Converfi indete] it are pePfOF. 106 lone of the predictors included in these analyses account for a significant amount of variance in turnover rates. LISREL Analysis of the Model The computer program LISREL was used to test the hypothesized structural model and the hypotheses. This program is used to evaluate the relationships that are hypothesized to exist between the latent variables in the specified model. Several different types of data can be used as input to the LISREL analysis, such as a covariance matrix, a :orrelation matrix, or raw data. A covariance matrix is typically used as input (Bentler, 1980). Cudeck (1989) asserts that the results of a LISREL analysis when a correlation matrix is used may be different than when a zovariance matrix is used as input. The use of a :orrelation matrix instead of a covariance matrix may affect the standard errors and the tests of significance. lowever, the use of a covariance matrix in this model resulted in an error, which stopped the computer from :ompleting the analysis. The program was unable to :onverge to reach a solution because of various .ndeterminancies in the data matrix. The data or parts of .t are such that various matrix operations cannot be )erformed. The 1 was SPSSX edition. model can A correla As mentic for state this mat] The includin union pr Variable Size, nt moderate indicatt teacher bOth in meaSUre were 00 Al endogen C1imate rate. 107 The statistical package used to examine this model IS SPSSX LISREL VI (Joreskog & Sorbom, 1986) fourth lition. The program that was written to analyze the )del can be found on the first three pages of Appendix H. correlation matrix was used as input to the analysis. ; mentioned, some partial correlations which controlled )r state salary and socioeconomic status were included in 118 matrix instead of the zero—order correlations. >del Specification There were seven exogenous variables in the model, lcluding organizational performance, principal turnover, iion presence, teacher salary, two labor market triables, and organizational size. The two indicators of .ze, number of students and student — teacher ratio, were )derately correlated (r : .56), and were considered to be 1dicators of the same underlying construct. Although aacher demand ratings and professional unemployment were )th intended to be measures of the labor market, the two easures were uncorrelated (r : .07). Thus, the measures :re considered as separate labor market constructs. Also included in the model were four latent idogenous variables which included the two measures of ,imate, along with measures of satisfaction, and turnover Lte. The correlation between the two climate measures was low (I separate n W Each be briefl; describe matrices a LISREL causal pa values p1 analyses labels t] matrices provided be descr 169 and 0f zeroe that re? estimatt at 1.00 USed to two Var PEflect 108 was low (r = .30), so these measures were considered as :eparate measures of climate. )escription of the Paths in the Model. Each of the matrices used in the LISREL program will >e briefly described. Specifically, these matrices lescribe the model presented earlier in Figure 3. Pattern latrices and matrices of start values must be specified in t LISREL analysis. The pattern matrices described the :ausal paths in the model, while the matrices of start 'alues provide the computer with values with which the .nalyses are started. Appendix G provides the construct .abels that correspond to the etas and ksis found in these Iatrices. The program used to evaluate Figure 3 is )rovided in Appendix H, but only the pattern matrices will >e described here. These matrices can be found on pages .69 and 170 of Appendix H. The pattern matrices consist bf zeroes and numbers. The zeroes are used to indicate .hat relationships between certain variables were not :stimated by the program. These values were usually fixed Lt 1.00 or .00. The non—zero numbers in the matrix are lsed to indicate that a factor loading or a path between Lwo variables was to be estimated. The pattern matrices 'eflect the hypotheses implicit in Figure 3. Some order to involved modifica1 hypothes the foll- Lam were con construe single i LE! variable underly' a facto Constru indicat Scaling Separat Be endoger Paths 1 Satisf; g to the SEVen 109 Some modifications to the original model were made in rder to achieve an estimable model. These modifications nvolved aspects of the measurement model. None of the 3difications involved any of the structural coefficients ypothesized earlier. Modifications will be described in me following sections. Lambda Y gLY). This matrix specifies which variables are considered to be measures of the same underlying anstructs. Each of the measures were considered to be ingle indicators of the underlying constructs. Lambda X (LX). This matrix also specifies the iriables that were considered to be measures of the same mderlying constructs. The value of one in this matrix is factor loading of number of students on the size )nstruct. The factor loadings of the other size idicator, student-teacher ratio, were fixed at 1.00 as a 3a1ing factor. The other six exogenous variables are aparate indicators of the six latent constructs. Beta 3B 2. This matrix specifies paths between the idogenous variables. In this case, the program estimated aths from the climate measures to satisfaction, and from itisfaction to turnover rate. Gamma {G 3. The paths from the exogenous constructs 3 the endogenous variables are specified in this matrix. even paths are specified. These paths include one from size to < turnover performa turnover Eh; between model, t and sale associat unemplog presence in prel these c were in SUbStan BE bEtWeep reSidue COl‘l‘elz residuE were b( the Fe? Satisf: estima' 110 ize to climate, two from union presence and principal urnover to satisfaction, and four from organizational erformance, salary, and the two labor market measures to urnover rate. Phi gPH}. The PH matrix specifies the relationships etween the residuals of the exogenous variables. In this odel, the residuals associated with organizational size nd salary were assumed to be correlated as were those ssociated with organizational size and professional nemployment, union presence and salary, and union resence and teacher demand ratings. Modification indices 1 preliminary analyses suggested that the residuals of iese constructs were related. While some other residuals are interrelated, these interrelationships were not ibstantively meaningful. Psi PS . This matrix specifies the correlations etween the residuals of the endogenous variables. The ESiduals of the two climate measures were assumed to be Drrelated. Additionally, the climate and satisfaction esiduals should be correlated, because these constructs are both measured with self—report instruments. However, 1e relationship between the second climate measure and itisfaction was not estimated in order to achieve an ;timable model. Ihgtg observed e The error: not estim: an estima' Ihgt errors as The only student-t two error model, ar Variables 111 Theta Epsilon (TE). The errors associated with the .erved endogenous variables are found in this matrix. : errors associated with the endogenous variables were ; estimated, and were fixed at zero in order to achieve estimable model. Theta Delta jTD). This last matrix includes the rors associated with the observed exogenous variables. e only variables for which error was not estimated were udent-teacher ratio and union presence. Again, these 0 errors were fixed at zero to achieve an estimable del, and because errors of measurement in these two .riables were thought to be minimal or zero. Results of the LISREL Analysis The results of this analysis can be found in Appendix starting on page 173. Several sections of these results re used to evaluate the hypothesized model. t Indices Chi Sguare. When evaluating a model with LISREL, a nsignificant chi square indicates that the model is a 0d fit, whereas a significant chi square suggests that e model does not fit the data well. This analysis sulted in a chi square of 199.87 with 50 degrees of eedom (see page 175 in Appendix H). The chi Square was gnificant (p < .000), which by itself, indicates that e model does not fit the data. However, researchei dependent when the : (1980) su evaluate sample si 999d The goodr to which covarian< size. T] the GFI These me hear one rule-of- Consider Of this ”38 .783 mOdel d' 39_ a measu (Joresk than .0 aSSOCia the mod 112 archers recognize that the chi square statistic is ndent on sample size, and will often be significant the sample size is large (Bentler, 1980). Bentler 0) suggests that other methods should be used to uate a model, such as an index that is not affected by >le size, (e.g., the residual matrix). Goodness of Fit and Adjusted Goodness of Fit Indices. goodness of fit index (GFI) is a measure of the degree Jhich the model accounts for the observed variance- ariance matrix. This measure is independent of sample a. The adjusted goodness of fit index (AGFI) adjusts GFI for degrees of freedom (Joreskog & Sorbom, 1986). se measures can range from zero to one, with a value r one indicating a good fit of the model. While any e-of—thumb is arbitrary, a value above .90 is usually sidered evidence of relatively good fit. The results this analysis showed that the GFI was .865 and the AGFI .789 (see page 175 of Appendix H), indicating that the el did not fit the data as well as might be desired. Root Mean Square Residual fRMSR). This fit index is easure of the mean residual variances and covariances reskog & Sorbom, 1986). A value near zero (i.e., less n .05) would suggest a good fit of the model. The RMSR ociated with this model was .112, again indicating that model did not fit the data well. These fit of the not fit, < & Sorbom, examining squared m and the p Squared M Tabl squared a three enc the pr0pt Variable Suggest was very The between I‘epl‘Odm Values _ been a . 183 of size an had a g WOuld t 113 These fit indices provide information on the overall . of the model, but do not indicate why the model does fit, or which parts of the model do not fit (Joreskog iorbom, 1986). These questions were answered by 1mining other sections of the results, such as the Jared multiple correlation values, the residual matrix, 1 the paths in the model. uared Multiple Correlation Values Table 14 presents the structural equations and uared multiple correlation values associated with the ree endogenous variables. The R2 values indicate that e proportion of variance shared by the endogenous .riables and their predictors is low. These low values lggest that variance in none of the endogenous variables s very well explained. sidual Matrix The fitted residual matrix presents the difference tween the original correlation matrix, and one produced by the LISREL analysis (see page 182). The lues in this matrix would be near zero had the model en a good fit. The normalized residuals, found on page 3 of Appendix H, corrects these residuals for sample ze and scaling differences (Bollen, 1989). If the model d a good fit, almost all of the values in this matrix uld be below 2.0; any value above 2.0 is an indication Structur Environm Climate Degree c Latitude Satisfa Turnove ‘ 011 Of Tre StL Toe sz Ra* 114 Table 141 ructural quations and Multiple Squared Correlations of the Endogenous Variables Structural Equations Multiple R2 ironmental mate —,11 (Size) .012 :ree of ;itude Climate ~.03 (Size) .001 :isfaction .342 (Clim1) + .226 (Clim2) .382 — .077 (Prins) — .010 (Union) rnover --.2362 (Tcsal) + .021 (Dmand) .099 -.2082 (Unemp) + .001 (Perfm) -.027 (Satis) Clim1 : Environmental climate measure, Clim2 : Degree of Latitude Climate measure, Satis = Satisfaction, Trate = Turnover Rate, Ratio : Student - Teacher Ratio, Studs = Number of Students, Prins 2 Principal Turnover, Tcsal = Average Teacher Salary, Union : Union Presence, Dmand : Teacher Demand Ratings, Unemp : Unemployment Ratings, Perfm : Organizational Performance p < .05 of a prob that four further e appropria additions later in The examined were con SUpporte Pages 17 181 of A the mode hYpothes Obtainec diagram Signifi PPOfess that We influen indioat Signifj aSSOCiz The er: in the 115 a problem with the model. The matrix on page 183 shows it fourteen of the values were above 2.0, which is rther evidence that the model is not completely propriate. These values can also be used to suggest ditional paths in the model, but this will be discussed ter in a section on possible alterations to the model. The path diagram resulting from the analysis was amined to understand what aspects of the model ire confirmed, as well as the hypotheses that were not ipported. The path coefficients can also be found on tges 173 and 174 of Appendix H. T—values (pages 180 and 11 of Appendix H) were used to determine which paths in ie model were significant. Figure 4 presents the rpothesized model including the path coefficients tained from the LISREL output in Appendix H. The path agram in Figure A shows that the only paths that were gnificant were from teacher salary to turnover and from ofessional unemployment to turnover. The other paths at were hypothesized to have a direct or indirect fluence on turnover rates were nonsignificant. The dicator of organizational size that was estimated was gnificant. However, the magnitude of the error sociated with this measure was high, and significant. e errors that were associated with the other constructs the model were low and nonsignificant. The number of V 0530-“. w w oo. {00. oceanic—boo. >..a_dm _¢c0_—GN_CGULO Lozoaflr Foo . new... in. m. 00. OO. to>oc._= :0 on n a F hwof 2 u 2 w 6 1 l . wowf rNo. hhof «cuanEEocD vans—on .scoiofotl seconok ro.\—, moo. \—/ .mpo—z powocotn. o5 so w_w>_mc< or: W . 00. v G 0 OO. accenutn. . COED 00 .w 0:6! wof . 50:068. 00 «cow—5m 035:0 .a—coEco.:>cw a on. O? 00.: ONN 82m .ocozuflcduho Dunc—:0 w Ff «so. Gun—26.. uo wouawo OD. ..0>oc.:.c. scoff—onam 00. U6 wtzwwm v 9:9". nonsigni figure a discusse well. The would he constru: unemplo; satisfa found t profess Th would m measure Provide 82 Polati( superv- four. analys Satisf and tu Signif 117 .onsignificant hypothesized structural coefficients in the 'igure along with the other results that have been .iscussed indicate that the model did not fit the data ’ell. Summary of the Results of the Hypotheses The first hypothesis suggested that five constructs ould have a direct, negative effect on turnover. These onstructs were the teacher demand ratings, professional nemployment, organizational performance, salary, and atisfaction. The only two of these variables that were ound to be significant predictors were salary and rofessional unemployment. The second hypothesis suggested that satisfaction ould mediate the relationships between the climate easures and turnover. The structural coefficients rovide no support for this hypothesis. Satisfaction was also hypothesized to mediate the elationships between union presence and turnover, and upervisor turnover and turnover in hYPOtheseS three and our. These hypotheses were not confirmed by the LISREL nalysis. The fifth hypothesis suggested that climate and atisfaction would mediate the relationship between Size Dd turnover. Again, these relationships were not ignificant. Fin: hypothes: relation: regressi: variable Aft modifica modify t model. explorat indicate However, paths CE Figure 1 tUrnove] Suggest' mode}. Th. in Apps; could b the mod Square Value 0 indicat 118 Finally, the two labor market variables were hypothesized to moderate the satisfaction - turnover relationship. The results of moderated multiple regression reported earlier indicated that these two variables did not moderate this relationship. Alterations to the Model After examining the fit of the hypothesized model, modification indices, residuals and t—values were used to modify the model to attempt to increase the fit of the model. These additional analyses were considered exploratory, not confirmatory analyses. The t—values indicate which of the paths in the model were significant. However, the t—values can also be used to determine which paths can be deleted. The two structural coefficients in Figure 4 that were significant were from salary to turnover, and from professional unemployment to turnover, suggesting that all other paths could be deleted from the model. The modification indices, found on pages 176 and 177 in Appendix H, provide information about other paths that could be included in the model. If a path suggested by the modification indices is added to the model, the chi square statistic will be reduced by an amount equal to the value of the modification index. Thus, high values indicate places in the model where paths could be added if the path modifies most of substant indices logical union p a media literat suggest Organiz the fix also l< Collec‘ While - 1987. the ac knowle Climat T 0thEr of the were < aDove enVir, 119 the paths are conceptually meaningful. Many of the modification indices were moderate to high values, but most of the paths suggested by these indices were not substantively meaningful. However, the modification indices did suggest two additional paths that were logical. The first modification was a direct path from union presence to turnover rate, deleting satisfaction as a mediator. This relationship was suggested in the literature at the individual level. The second change suggested by the modification indices was that organizational performance might be a direct antecedent of the first measure of climate, not turnover. This path was also logical because the achievement test data that was collected was administered during the 1986-87 school year while the climate measure was obtained during the fall of 1987. Thus, the teachers should have known the results of the achievement tests by the fall of 1987, and this knowledge could have influenced their perceptions of climate. The normalized residuals can also be used to identify other paths that could improve the fit of the model. Both of the paths suggested by the modification indices that were discussed above also had normalized residual values above 2.0. In fact, the relationship between environmental climate and organizational performance had the hig' these 1 Tu modific The new enviror The twc perform satisfz indica‘ separa measur includ T revise Square AGFI j .829 1 OVeraI impro‘ model Signi that 120 the highest value in the matrix. Other paths suggested by these indices did not seem meaningful. Description of the Revised Model Two additional paths that were suggested by the modification indices are included in this revised model. The new paths are from organizational performance to environmental climate and from union presence to turnover. The two paths that were deleted were from organizational performance to turnover, and from union presence to satisfaction. The significance of the error term size indicator suggest that the two size indicators should be separate constructs. Conceptually though, the two size measures should be related; thus this change was not included in the revised model. 1 The fit indices of the model indicated that the revised model had a better fit with the data. The chi square decreased from 199.87 to 156.51. The GFI and the AGFI increased slightly from .865 to .890 and from .789 to .829 respectively. The RMSR decreased from .112 to .096. Overall, these indices suggest that the new paths slightly improve the fit of the model. The revised model is presented in Figure 5. This model resulted in more structural coefficients that were significant. As Figure 5 indicates, the same two paths that were significant in the previous analysis were also 121 \V % W0. v Q 0 0°. 00. woo. 3:035 humus. OOCODOLI >5. 0 no a ! .ozowom 88.52 22. .m .w 00. . 0.? . NrN.u . mor.u . on w . B6 QuGF-zo oo. 00. w .3:0Ec9..>cw .0:o:0u.:nu..0 .o>o:.= I on..| :o_.00—0:nw 0.05:0 occur—23.5.". 0uac=ua .0 00.000 .0c020N_c0uhO . «9.- can..- :9- «new. 00. oo. o tn .c0E>o_nE0:D Eco-oooaoaa 3.6600 ..m.;0¢0.r ..0>oc..:.r son—>50asw OO. .0002 UOE>¢E 0...: *0 m_w>_mc< $9: #0 wtzwwm m 9:9“. signifi paths f perform climate structt there t sugges‘ repres. 122 significant in this revised analysis. Additionally, the paths from union presence to turnover, from organizational performance to environmental climate, and from both climate measures to satisfaction have significant structural coefficients in this revised model. However, there were a number of paths that were not significant suggesting that the revised model was still not completely representative of the turnover process. Summer} Tl hypothe fit in( residu: not apj were s- Teache were b The re was he T organ, Organi other in thf two m, Satis. deman, indie DISCUSSION Summary of Results The results of the LISREL analysis indicated that the hypothesized model did not adequately fit the data. The fit indices, squared multiple correlation values, and the residual matrices all provided evidence that the model was not appropriate. However, two of the hypothesized paths were supported by the results of the initial analysis. Teacher salary and the professional unemployment measure were both found to be significant predictors of turnover. The relationship between turnover and these two predictors was negative. The constructs that were aggregated to the organizational level, including climate, satisfaction, and organizational performance, were not related to turnover. Other hypothesized relationships that were nonsignificant in this sample included relationships between size and the two measures of climate, principal turnover and satisfaction, union presence and satisfaction, and teacher demand and turnover. The results of the analysis of a revised model indicated that this model fit the data somewhat better than ti three a Organi: climat presen from o and th of the predic betwee lmpli< broad also Issue first base: 00nd, that that diff thos indi Same 124 than the hypothesized model. As indicated in Figure 5, three additional paths in the model were significant. Organizational performance was related to environmental climate, both climate measures to satisfaction, and union presence to turnover. Even though inclusion of the paths from organizational performance to environmental climate, and the climate measures to satisfaction improved the fit of the model, these paths did not improve the predictability of turnover because of the low correlation between satisfaction and turnover. Implications of Study This study is important for two reasons. It has broadened our knowledge of the turnover process, and has also provided information on levels of analysis issues. Issues related to the turnover process will be addressed ‘first. ‘ The Turnover Process. The hypothesized model was based, in part, on the findings of research that has been conducted at the individual—level of analysis. The fact that the hypothesized model did not fit the data suggests that the predictors of organizational turnover are different from those of individual—level turnover. Thus, those variables that account for variance in an individual’s decision to leave an organization are not the same variables that predict the percentage of employees that wi of time Ro account affecti varianc this st always the int organi: variabi variabi Salary Signif have a study ecOnom leve1, based leVel Organi A model, import of thi 125 that will leave an organization during a specified period of time. Rousseau (1985) suggested that economic variables account for much of the variance at the unit level, while affective variables and behavioral intentions account for variance at the individual level of analysis. In fact, this study showed that satisfaction, which is almost always found to be important in the turnover process at the individual level, was unrelated to turnover at the organizational level. Moreover, two of the three. variables that were related to turnover were economic variables (i.e., professional unemployment and teacher salary). It also could be argued that the third significant correlate of turnover, union presence, would have an effect on economic conditions in schools. This study provided support for Rousseau’s contention that economic variables are important at the organizational level. Furthermore, it does not appear that hypotheses based on the results of research done at the individual level provide an understanding of turnover at the organizational level. Although this study was not entirely successful in modeling the turnover process, this type of study is important. According to Terborg and Lee (1984), studies Of this type are needed in order to develop a model of turnover research' process. 19p the hype regardir explorec these ii of the This se hYpothe Vere di relati< addres: T that s turnoV Salar§ relat: the r. anaiy bEtwe but a direc relai 126 turnover that is comprehensive and multi—level, so that researchers have a better understanding of the turnover process. Implications of specific paths in the Model. Each of the hypotheses will be addressed in this section. Results regarding several of the specific relationships that were explored have implications for turnover research, and these implications will be discussed. Additionally, each of the paths that were nonsignificant will be discussed. This section addresses each of the relationships that were hypothesized earlier as well as the relationships that were discovered during the analysis of the model. The relationships that were found to be significant will be addressed first. Teacher Salary — Turnover. Hypothesis 1 suggested that salary would have a direct negative effect on turnover. This study confirmed this relationship between salary and turnover, and these results indicate that the relationship at the organizational level is different from the relationship reported at the individual level of analysis. At the individual level, the relationship between salary and turnover is mediated by satisfaction, but at the organizational level, salary was found to be’a direct predictor of turnover. The significant relationship found between teacher salary and turnover in this study atthe org al. (l96ll significar the relat negative, with lowe Prol also sug, direct n market c related researc' Althoug PTEdict rETatit and tu availa Unemp} Condii and t- left mark. Sat‘l 127 this study is similar to the findings of earlier research at the organizational and departmental levels. Katzell et al. (1961) and Wales (1970) found that salary was significantly correlated with turnover. In this study, the relationship between these two constructs was negative, indicating that higher salaries are associated with lower turnover rates. Professional Unemployment - Turnover. Hypothesis 1 also suggested that labor market conditions would have a direct negative effect on turnover. Measures of labor market conditions have been reported to be significantly related to turnover in previous organizational-level research (e.g., Eagly, 1965; Terborg & Lee, 1984). Although teacher demand ratings were not found to be predictive of turnover, a significant negative relationship was found between professional unemployment and turnover, indicating that when more jobs were available, turnover was higher. The professional unemployment measure was a general measure of labor market conditions. Thus, the relationship between this measure and turnover may have been affected by teachers who had left their teaching careers for other professions. Individual-level research also indicated that labor market conditions might moderate the relationship between satisfaction and turnover (Muchinsky & Morrow, 1980; CarS' that orga stud hypo pres rese mixe unio The sati unic The indi to; sug, rel: sim usi to‘ Wit uni uni Org 128 Carsten & Spector, 1987). The sixth hypothesis indicated that this relationship might also be found at the organizational level. However, the results of the present study did not support this hypothesis. Union Presence — Turnover. Satisfaction was hypothesized to mediate the relationship between union presence and turnover (Hypothesis 3). However, the research that has been conducted thus far has-produced mixed results, so it was stated earlier that the role of union presence in the turnover process would be explored. The results of the LISREL analysis showed that satisfaction did not mediate this relationship, but that union presence had a direct relationship with turnover. The relationship between these two constructs was negative indicating that when a lower percentage of teachers belong to a teachers’ union, turnover is higher. This finding suggests that the presence of a union would most likely be related to lower turnover in organizations that are similar to schools, but this result should be explored using other samples. Often, upper management are opposed to having a union started in their companies, or dealing with existing unions. This relationship between unionization and turnover suggests that the acceptance of unions in organizations might be one way to decrease organizational turnover. ( perf01 predh relat LISRE did p suppo from much itsel orgar resui find nega thes Kara diff resu thre all meas difi for hem was 129 Organizational Performance — Climate. Organizational performance was originally hypothesized to be a direct predictor of turnover (Hypothesis 1). However, this relationship was not significant. The statistics in the LISREL output suggested that organizational performance did predict climate, and a revision of the model did support this supposition. Although the significant path from organizational performance to climate did not add much to the prediction of turnover, the relationship itself was interesting. Previous research at the unit or organizational level of analysis has yielded inconsistent results. As mentioned earlier, researchers have reported finding a significant relationship (both positive and negative) as well as a nonsignificant relationship between these two variables (Heller et al, 1984; Pritchard & Karasick, 1973; Schneider & Snyder, 1975). It is difficult to explain with any amount of certainty why the results of these studies have been so different. The three studies mentioned above and the present study have all used different climate surveys, different performance measures, and different samples. Any or all of these differences (or one not cited) could have been the cause for the various findings. The zero-order correlation between the environmental climate measure and performance was substantial (r = .509, p < .05), and the t-value associ (t : f preser < betwet was h (Hypo Hazer were The t satis turn< subo Howe stud Alth rela turr leve doe sat Sam 8qu 130 associated with the path in the LISREL analysis was high (t = 5.27, p < .01). Thus, the relationship in the present study was a strong one. Supervisor Turnover — Satisfaction. The relationship between supervisor turnover and organizational turnover was hypothesized to be moderated by satisfaction (Hypothesis 4). At the individual level, Williams and Hazer (1986) showed that leader-subordinate relationships. were predictive of satisfaction in the turnover process. The hypothesized relation between supervisor turnover and satisfaction was based on the assumption that supervisor turnover could be used as a substitute for leader— subordinate relations at the organizational level. However, this assumption may have been incorrect. This study indicated that the relationship was not significant. Although supervisor turnover and leader-subordinate relations would almost always be related, supervisor turnover may not affect satisfaction at the organizational level in the same way that leader-subordinate relations does at the individual level. Another explanation may also account for not finding a relationship between supervisor turnover and satisfaction. The basis for this explanation is the sample used in the current study. Research has shown that supervisor turnover leads to an increase in the number of new ru (Carls changi satisf starts princi teach< teach teach degre super overe meaS\ betw: sign betw prov turr Vari hyp, Sch was cor 131 new rules and regulations that are put into effect (Carlson, 1962; Grusky, 1959). This increase in rules and changing procedures was hypothesized to lead to decreased satisfaction. It is possible that when a new principal starts at a school, any new rules implemented by the principal would not have much of an effect on the teachers’ routine. Thus, the overall satisfaction of teachers in schools would not be affected. Additionally, teachers may not interact with principals to the same degree as employees in other occupations, so the degree of supervisor turnover would not have as much of an effect on overall satisfaction. It would be interesting to use this measure in other samples to see if the relationship between supervisor turnover and satisfaction was significant in samples where there is more interaction between supervisors and their subordinates. Size in the Turnover Process. This study did not provide any new information about the role of size in the turnover process. Again, previous research between these variables has yielded inconsistent results. The hYpothesis in the present study that the size of the school would influence the teachers’ perception of climate was not supported (Hypothesis 5). The zero-order correlation between one of the two size constructs (number of students) and turnover was significant, but the relati appro; modiff not 81 const Satis effec media lHch SUPP‘ for ‘ are , issu of o Peas eSti Peri to € in 4 dEC‘ 132 correlation between the other size variable, student— teacher ratio, and turnover only approached significance (r = -.136, p < .10). This suggests that a direct relationship between size and turnover might be appropriate at the organizational level. However, the modification indices in the original LISREL analysis did not support the inclusion of a path between the size construct and turnover. Climate and Satisfaction in the Turnover Process. Satisfaction was predicted to have a direct negative effect on turnover (Hypothesis 1), and was predicted to mediate the relationship between climate and turnover (Hypothesis 2). These hypothesized relationships were not supported by the results of this study. Potential reasons for the all of the above mentioned nonsignificant findings are discussed later in the sections on levels of analysis issues and study limitations. Practical Implications. Identifying the determinants of organizational turnover is also important for practical reasons. It would be useful if organizations could estimate the percentage of people that will leave during a period of time. This estimate would enable organizations to include the selection and training costs of new hires in their budgets. Some organizations also may want to decrease their turnover rates. An attempt to decrease turnov emploi turnOi an or: they study allow cours other deter resui be u: turn the prod CORE tum Wen pro mea men lit We} ant 133 turnover would most likely result in the retention of some employees who are poor performers. However, if the turnover rate is extremely high, it may be worthwhile for an organization to retain some poor performers so that they also retain good performers. The results of this study suggest that organizations could raise salaries or allow unions to grow in order to decrease turnover. Of course, both of these 'solutions" may be undesirable for other organizational reasons. Further study of the determinants of organizational turnover would probably result in the identification of other variables that could be used by organizations that want to estimate or decrease turnover. Perhaps the major implication is that attacking the turnover problem at the individual level may not produce the desired organizational-level consequences. Levels of Analysis. As mentioned earlier, the constructs that were aggregated were not related to turnover. Eta squares found for those variables which were aggregated suggest that there could have been a problem with the supposition that these constructs have meaning at the organizational level of analysis. As mentioned earlier, the median eta square reported in the literature is .12. The eta squares found in this research were .15 (degree of latitude climate), .19 (satisfaction), and .27 (environmental climate). Although the eta squares found resea withi betwe the i aggrt org aEE Fur 134 found were higher than the median level usually found in research, these figures indicate that there was much more within—organizational variance in these constructs than between—organizational variance. Thus, the differences at the individual level are averaged when the responses are aggregated. Furthermore, there is reason to believe that the aggregation of individual—level satisfaction responses would not necessarily be related to organizational turnover. We can conceptualize the relationship between employee satisfaction and turnover at the individual level as lining all of the employees in front of the door to the organization in order of their degree of satisfaction. The employees in the front of the line, closest to the door, would be those people who were the most dissatisfied. However, factors other than an employee’s dissatisfaction also play a role in the turnover decision, so those closest to the door may not be the employees that actually leave the organization. ‘ This line of reasoning is also true at the organizational—level of analysis. More specifically, aggregated satisfaction does not provide information on how many people would actually leave an organization. Furthermore, this relationship between aggregated satis the v crite const const consl is, I indi that leve not This asse not and aft dis 135 satisfaction and turnover rate would be lower because of the variance that is lost when satisfaction is aggregated. The results of this study suggest that satisfying the criteria that are used by researchers to ensure that constructs can be considered organizational-level constructs may not provide sufficient evidence that the constructs have meaning as organizational variables. That is, eta square statistics and similar correlations at the. individual and organizational levels may not guarantee that the constructs have meaning at the organizational level. The measure of organizational performance also may not have meaning at the organizational level of analysis. This measure was not a rating, so agreement could not be assessed. Additionally, the individual-level data were not available, so the correlations at one level could not be compared to the correlations at the second level. Thus, there is no evidence of the quality of this measure at the aggregate level. Limitations of the Study There were some potential problems with the sample and some of the measures used in this study that may have affected the results. Each of these problems will be discussed next. turnt infli inst migh teac thei resu anot errc prox affs tho: had ind EI‘I‘ COI‘ ana As Con Shc le\ 136 Sample. The factors that have an effect on the turnover rate of schools may be different from those that influence turnover in other types of organizations. For instance, a more stable measure of labor market conditions might have been found for another sample. Additionally, it may be more difficult for teachers to find other teaching jobs, so teachers that would otherwise leave their jobs end up staying with their school. Thus, the. results of this study could have been different had another sample been used. Turnover. It is impossible to know the amount of error that was included in the turnover measure that was provided by the principals. This measure would be affected by error if the principals did not include all those who left voluntarily when asked how many teachers had left. Conversely, if the principal included other individuals who did not leave the school voluntarily, error would have affected the measure. Satisfaction. The eta squared statistic and the correlational results presented earlier supported the analysis of this construct at the organizational level. As mentioned earlier though, Rousseau (1985) suggests that constructs that will be aggregated to a higher level should be worded so that the questions refer to the group level to which the construct will be aggregated. In this study level indiv relat organ apprc The n demar prof: that Howe ment thes cour in c liEI‘E mea tur I‘e} 137 study, the climate items were worded at the organizational level, but the satisfaction items were worded at the individual level. This may have affected the relationships between satisfaction and other organizational—level constructs. Unemployment Measures. It was difficult to identify appropriate indicators of the labor market for teachers. The most appropriate measure seemed to be the teacher demand ratings supplied by James Akin (1989), because the professional unemployment measure was a general measure in that it included many other professional careers. However, there were problems with the demand ratings. As mentioned earlier, there was no indication of how reliable these ratings were. Additionally, the areas of the country identified by Akin were extremely broad. In fact, in one area of the country (the southeast), ten states were combined. It is probable that some of the states (e.g., Kentucky and Florida) that were combined in the various country areas did not have similar hiring needs for teachers. Even though the teacher demand ratings seemed more appropriate, the professional unemployment measure had a higher and a significant correlation with turnover. It is unclear whether the nonsignificant relationship between the demand ratings and turnover was caused by the construction and possible unreliability of the d! just Futur deter impor const stag: and . Addi cons anal Seve will sim' dif ana EXP Sut ins Sul sul up 138 the demand measure, or whether the two constructs were just unrelated. Future Research Directions Researchers have just recently recognized that the determinants of turnover at the organizational level are important to understand. The study of the turnover construct at this level of analysis is still in the early stages. Models such as the one proposed can be modified and examined using different samples of employees. Additionally, researchers should attempt to identify other constructs that exist at the organizational level of analysis that may be important in the turnover process. Several constructs that researchers may want to consider will be discussed next. Some of these constructs are similar to those in the proposed model while others are different. Following that discussion, some levels of analysis issues will be considered. Leader Subordinate Relations. It would be useful to explore other measures that might represent leader- subordinate relations at the organizational level. For instance, a measure of leader effectiveness collected from subordinates might provide information about leader— subordinate relations. It also might be useful to ask Upper management about the organizations’ philosophy on how s1 direc of le relat satie meaSL at tl inst: part info orga In a file leve Clir this Fest Zoh USS 139 how supervisors should act (e.g., participative, directive). Finally, a variable that reflects the recency of leader turnover might provide information about the relationship between supervisors and their subordinates. Satisfaction. If there is a problem aggregating satisfaction to the organizational level, perhaps another measure that is similar to satisfaction exists naturally at the organizational level and could be identified. For instance, the percentage of employees that normally participate in nonmandatory work functions might provide information about how much the employees like the organization or the other employees in the organization. In a unionized organization, the number of grievances filed may also provide information about the satisfaction level of the employees. Organizational Climate. Although organizational climate was not predictive of organizational turnover in this study, the construct should be considered in other research before dismissing its potential importance. Zohar (1980) contends that the type of climate measure ised should be specific to a certain situation. Thus, certain climate measures may be important in some situations and unimportant in others. For instance, a climate for safety (Zohar, 1980) might be more important in a factory, and less important in a department store. may t Manag impo: emplt wort} empl: have the trai perc perc empl the the be 1 to 4 918: Her Var con the has in; 140 Training. The existence of training opportunities may be important in predicting organizational turnover. Management’s attitude toward training might be more important than the training itself. For instance, if employees realize that management considers training to be worthless or only good for providing skills that their employees should already have, training opportunities may have no impact on the turnover rate, or may even increase, the amount of turnover. So, if management considers training to be a demeaning process, the employees may also perceive it that way. On the other hand, if training is perceived as an opportunity, as a productive process that employees can use to hone their skills or learn new ones, the existence of training should have a negative effect on the turnover rate of an organization. However, it should be noted that a positive training experience can also lead to turnover if employees gain skills that are marketable elsewhere. Societal Variables. The only societal variables that were considered in the present study were the economic variables. Future research endeavors might include other constructs such as the cost of living, the crime rate, and the amount of entertainment in the area. These variables have not been considered in previous studies, but could be important in the turnover process. It is reasonable to hype livi area of] org the pex Whe QUt 3P 141 hypothesize that in some areas of the country, the cost of living is so high that employees of organizations in the area consider moving to another area that has a lower cost of living, and consequently leave the organizations. Individual Performance. One variable that was discussed earlier as having potential importance in this area is individual performance. Researchers have concluded that not all turnover should be considered as having negative consequences for the organization. The question that should be addressed here is whether the computation of organizational turnover should be weighted in some way to account for the degree to which an organization will be hurt by the loss of each individual that does leave. Several researchers would argue that the performance level of former employees should be considered when studying organizational turnover. However, this question can only be answered by conducting the appropriate research. Levels of Analysis. Future researchers in this area might also want to build a more complete composition model. If the data are available at the individual, group and organizational level, researchers would be able to compare the correlations between similar variables at different levels. In the present study, it would have been interesting to see if there had been a relationship betw that with Summ 00nd area man) comm pres sah tur ind had Hop COD 142 between satisfaction and turnover at the individual level of analysis. However, the teacher data were collected so that it was impossible to identify the specific teachers within a school. Summary In conclusion, additional research should be conducted on organizational turnover. It is an important area of study because turnover may have an effect on so many other organizational processes, such as communication netWOrks and formalization of rules. The present study did confirm that economic conditions, salary, and unionization all play an important role in the turnover process. Results regarding aggregated individual—level variables indicated that these constructs had little influence on organizational-level turnover. Hopefully, additional research in this area will contribute to a better understanding of the organizational turnover process. APPENDIX A Principal Items 0_rg_: Ax APPENDIX A Principal Items Organizational Size How many students are enrolled in your school? Supervisor Turnover How many principals or headmasters have served your school in the last decade? 0. one 5. six 1. two 6. seven 2. three 7. eight 3. four 8. nine or more 4. five Average Teacher Salary What is the average teacher salary in your school this year? Social Economic Status HOW many students in your school receive free or reduced-price lunches? 144 Organizational Performance In the tables below, please report standardized achievement test scores for all grades in your school from 6 to 12. Give average percentiles or average normalized curve equivalent scores (NCES). READING COMPREHENSION Grade Reading Type of Score Test Used, Test Level Comp. Reported: NCES Form and Score or percentiles Publication Date TOTAL MATH Grade Reading Type of Score Test Used, Test Level Comp. Reported: NCES Form and Score or percentiles Publication Date 145 Organizational Performance (continued) TOTAL Grade Level SCIENCE Reading Comp. Score Type of Score Test Used, Reported: NCES Form and or percentiles Publication Test Date APPENDIX B Teacher Items Chm “1L5 desi stx 146 APPENDIX B Teacher Items School Climate Choose the answer from the following scale that you think most people in your school and community would pick to describe your school. Use this scale for items 71 — 82. 1 : Most people would strongly disagree with this statement. 2 : Most people would disagree with this statement. 3 : Most people would neither agree nor disagree with this statement. 4 : Most people would agree with this statement. 5 2 Most people would strongly agree with this statement. 6 = I don’t know what most people think about this statement, or I don’t know whether this statement fits the school. Coworker Competence Teachers are patient and make extra efforts to help students. Teachers understand and meet the needs of each student. Students can get help and advice from teachers or counselors. Most classroom time is spent in learning activities. PhYSical Environment n the school building. Students and teachers are safe i The school is kept neat and attractive. Stude1 Stude Stude Stude Use t deois new How pron How ado, How ado 147 School Climate (Continued) Student Behavior Students work hard on their studies. Students are well—behaved. Students care about and respect each Participation in Decision Making other. Use the scale below to answer items 38—41 about your decision making role in the district/school. 4 2 Always 3 : Often 2 = Sometimes I : Seldom O : Never How frequently do you participate in new staff? How frequently do you participate in promotion of any of the professional How frequently do you participate in adoption of new policies? How frequently do you participate in adoption of new programs? the decision to hire decisions on the staff? decisions on the the decisions on the Degre Use t about Litt appr A pe quic Even Upf I he Any 148 School Climate (Continued) Degree of Autonomy Use the scale below to respond to the following statements about practices in your district/school. 3 : Definitely true 2 = true 1 : False O: Definitely false Little action can be taken here until a supervisor approves a decision. A person who wants to make his/her own decisions would be. quickly discouraged here. Even small matters have to be referred to someone higher up for a final answer. I have to ask my supervisor before I do almost anything. Any decision I make has to have my superior’s approval. Use desc scho The You You Stu The Ths y01 Cm pm 149 Teacher Satisfaction Use the scale below to select the answer that best describes how you feel about the following aspects of your school. 1 = I am very dissatisfied with this aspect of the school. 2 = I am dissatisfied with this aspect of the school. 3 = I am neither satisfied nor dissatisfied with this aspect of the school. 4 = I am satisfied with this aspect of the school. 5 = I am very satisfied with this aspect of the school. 6 : I don’t know how I feel about this aspect of the school, or I don’t know whether this statement fits my school. The administrators in your school. Your pay, fringe benefits, and other compensation. Your Opportunities for career advancement in your school or district. Student discipline and sense of responsibility. The school curriculum and your job duties. The competence, commitment, and level of cooperation of your fellow teachers. Community and parent support for your school and its programs. The availability and quality of school facilities, Supplies, and maintenance. The extent and quality of communication about school matters within the school and the district. APPENDIX C Average Teacher Salaries for 1987 (by State) 150 APPENDIX C Average Teacher Salaries for 1987 by State1 State Average Salary Alaska - 44,000 District of Columbia 33,800 New York 32,600 Michigan 31,500 California 31,200 Rhode Island 31,100 Minnesota 29,100 Connecticut 28,900 New Jersey 28,900 Maryland 28,700 Illinois 281400 Massachusetts 28,400 Wisconsin 28,200 Wyoming 27,700 27,500 Delaware Washington 27,500 Colorado 27,400 Pennsylvania 27,400 Hawaii 26,800 Oregon 26,800 United States 26,700 Arizona 26,300 Ohio 26,300 Nevada 26,000 Indiana 22,288 Virginia 2 , Texas 251300 Georgia 32,888 New Mexico 23,800 Florida 23,800 North Carolina ; - 23,600 Kansas 23,500 Alabama 23,500 Missouri 151 Average Teacher Salaries for 1987 by State (Cont’d) State Average Salary Utah 23,400 Montana 23,200 South Carolina 23,000 Tennessee 22,700 Iowa 22,600 Kentucky 22,600 Nebraska . 22,100 Oklahoma 22,100 North Dakota 21,800 Vermont 21,800 Idaho 21,500 New Hampshire 21,400 West Virginia 21,400 Louisiana 21,300 Maine 21,300 Arkansas 20,000 Mississippi 19,600 South Dakota 18,800 Information obtained from the Almanac of the 50 States (Hornor, 1989) APPENDIX D Professional Unemployment Measure AAAACCCDDFGHIIIIKKLMMMMMMMMNNNNNNN 152 APPENDIX D Unemployment Rate1 by State2 State Unemployment Rate Alabama 1.5 Alaska 4.0 Arizona 2.5 Arkansas 2.2 California 2.6 Colorado 2.8 Connecticut 1.5 Delaware 1.0 District of Columbia 3-0 Florida 2'0 Georgia 1-2 Hawaii 1-2 Idaho 3'0 Illinois 2-5 Indiana 1'7 Iowa 1'4 Kansas 1'8 Kentucky 0'6 Louisiana 3'8 Maine 1'2 Maryland 1.: Massachusetts 2'7 Michigan 2‘1 Minnesota 2‘1 Mississippi 1'0 Missouri 3:6 Montana 1.0 Nebraska 4'2 Nevada 1.7 New Hampshire 1.9 New Jersey 1.9 New Mexico 1.7 New York 1.5 North Carolina NOOOPRSCLTTL\\IVVIV.) l 153 Unemployment Rate by State (Cont’d) State North Dakota Ohio Oklahoma Oregon Pennsylvania Rhode Island South Carolina South Dakota Tennessee Texas Utah Vermont Virginia Washington West Virginia Wisconsin Wyoming Unemployment Rate 03 NNHwi—‘ONNNP—‘D—‘D—‘I—lr—‘Np—iy—L OWGOQQWOCDCOOOQO’D-wx] A value of one in this table represents one thousand people. Information obtained from the Geographical Profile of Employment and Unemployment, 1987, Labor Statistics, 1989. U. S. Bureau of APPENDIX E Teacher Demand Ratings APPENDIX E Teacher Demand Ratings lNTRcDUCTlOfi This publication contains the twelfth annual supply/demand report sponsored by the Association for School, College and University Staffing (ASCUS). During the twelve years some reports have been based on surveys of the total ASCUS membership, while others were based upon samples. This year. fer the twelfth annual report, questionnaires were again sent to all ASCUS member teacher placement offices. Five hundred and two questionnaires were mailed in December, 1987,and two hundred and forty—seven (492) were received in time and in condition to be used in this report. The basic portion of the survey instrument has remained relatively constant throughout the :uelve year period and this report contains material from previous years for comparison purposes. thank you goes to Rosie Ferris of Kansas State University for her ass stance :ouard the development of this report. Aux-«mu Mum-m. 1=NW'.( 2m,“ 3mm” “pom-r "Cu-r Plum/moan 5:50.01 Comm. 95w“: 7-54-13“; a-Moasc Ana-woe V’Mum 154 155 THE REPORT SUPPLY/DEMAND Teacher placement officers continue to report improved job markets for their candidates as compared to previous years. Of the placement officers responding. 572 indicated that the job market has been better or much better for elementary teachers and 582 indicated that it has been better or much better for secondary teachers, as compared to 222 year earlier. Compared to four years earlier. 762 of the respondents indicated an improved job market at the elementary level and 72% indicated improvement at the secondary level for their candidates. Fifty-six percent of the respondents indicated that they expected the job market to be improved for this years elementary level graduates and 512 expected improvement for secondary level candidates. The chart which follows summarizes the improving opportunities for teachers in the United States as described by the responding placement officers. Question: In general terms how available were employment opportunities for elementary and secondary teachers for the 1987-88 teaching year (last completed placement season) compared to those one year earlier? 2 fl Elementary: 2 6 Secondary: 9 22 much better 15 35 much better 48 116 better 43 100 better 37 88 same 37 87 same 6 14 worse 5 12 worse __9 __9 much worse __9 0 much worse 100 240 100 234 Question: In general terms how available were employment opportunities for elementary and secondary teachers for the 1987-88 teaching year (last completed placement season) compared to those four vears earlier (1983-84)? 2 0 Elementary: 2 0 Secondary: 37 80 much better 28 66 much better 39 86 better 44 103 better 14 30 same 17 39 same 9 20 worse 9 21 worse __1 __2' much worse __3 __2 much worse 100 218 l 0 234 Question: In general terms as compared to one year earlier. how do you expect employment opportunities to be for elementary and secondary teachers for the approaching 1988-89 teaching year (current placement season)? 2 0 Elementary: Z l Secondary: 8 20 much better 8 18 much better #8 113 better 43 100.5 better 42 99 some 47 110.5 same 2 5 worse 2 4 worse __2 __9 much worse __2 __9 much worse 100 237 100 233 156 TEACHE? SL‘FFLY/TEUéJQ BY FIELC AN? REC-10‘: Fegicn Alaska Hawaii 1 2 3 5 5 6 Field Agriculture -— NA 3.00 2.50 2.00 3.08 1.78 3.30 Art 3.00 .;63 1.13 1.85 1.93 2.56 2.20 3.48 Bilingual Ed. 3.00 3;:7 4.03 4.88 4.40 4.17 4.54 4.00 Business 4.00 455 2.57 2.83 3.05 2.72 2.38 3.13 Eorrfuter Science 3.00 3'3 3.33 3.15 4.40 3.85 3.71 3.67 Counselor-Elem. 5.03 fig 3.22 2.52 2.50 3.19 3.12 3.48 'c:.~:.sc::~:-:e;. 5.1: 34-. 3.3: :.s: :5: 3. 4 2.35 3.10 Data Processing 3.00 NA 3.33 4 1 3.50 3.77 3.25 3.31 priver Ed. 3.00 557 2.50 2.71 3.00 2.25 2.43 3.00 EZamantagy-Frimarf 4.00 NA 2.56 3.28 2.53 2.31 3.61 2.92 Elementagy-lntermecia:e 4.00 N5 2.56 3.22 3.00 2.45 3.56 3.00 'srgiish 2.00 32:. 2.90 3.42 3.00 3.20 3.53 3.0:. Health Education 3.00 K; 1.89 2.2 2.00 1.91 2.00 2.10 Home Economics 5.00 NA 1.73 2.00 2.30 2.04 1.58 2.47 Industrial Arts 3.00 NA 3.20 3 22 2.58 2. 1 2.40 3.63 3:;rnalis: 3.0? x; 2.07 2.73 2.23 3.13 2.55 3.20 Language. Hod.-French 3.03 NA 3.11 2.88 3.00 3.33 3.73 4.08 Langpage. Hod.—Cerma: 3.00 NA 2.89 2.57 2042 3.25 3.60 3.90 Language, Rod.:§gafiisc 3.0C X5 3.33 3.24 3.50 3.53 3.53 3.87 Library Science 3.03 fig 3.17 3.88 3.75 3.3 3.55 3.79 Mathematics 4.00 \4 3.22 4.05 ‘«.CJ 3.90 4.39 4.19 M:sic-1ns:rumental 4.0: 55 3.56 2.57 3.29 3.56 2.35 3.04 Music-Vocal 4 03 Kn 3.67 2.60 2.71 3.43 2.41 2.75 Physical Education 3 00 N“ 1.13 2.06 1.53 1.2 1.73 1.96 fsigpolggist (school) 4. 2' NA 3.40 3.31 3.20 3.67 3.23 3g§L Efience-Eiclogy 3.00 NA 2.78 3.39 2.43 3.30 3.59 3.86 Science-Chemistry 3.00 x, 3.33 4.-2 3.57 3.83 4.00 4.25 Sczence—Earth 3.03 fig 2.78 3.00 3.14 3.43 3.75 3.83 s.:;ence-ce:~.erai 3.00 .3 2.78 3.65 3.00 3.02 3.88 3.? Science-Physics 3.r3 SA 3.56 4.29 4.09 3.83 4.50 4.23 Sccial Sciences 2 C2 NA 1.44 1.5: 2.07 1.89 2.35 2.12 Social Vorker (school‘ —— NA 2.17 3.50 2.75 3.19 2.63 2.62 53eech 3.00 NA 2 3 2.80 2.75 2.94 2.73 2.94 §2§c.-Weaf Education 5.00 NA 3.20 4.00 4.25 4.24 3.80 3.71 ggec.-EDSPSA 5.03 NA 4.43 4.46 4.25 4.72 4.13 4.13 saga-Gmed 5.00 3:3 3.53 3.70 3.35 1.. 4 3.86 3.05 39ec.-LD 5.00 NA 4.38 4.50 4.50 4.41 4.20 4,23 §3§c.-HT 5.00 NA 4.38 4.50 4.50 4.23 4.13 4.09 §£§c.-Hulti Handi 5.00 NA 4.50 4.62 4.50 4.46 4.00 4.27 §£ec.—Readin£ 3.00 NA 3.44 3.75 3.42 3.24 3.87 3.40 §peech Path./Audio. 5.00 NA 3.43 . 4.00 4.20 4. 4 3.69 4.17 COMPOSITE 4.35 NA' 2.99 3.31 3.13 3.23 3.29 3.43 Regions are coded as follows: Alaska, Hawaii, l-Northuest, E-West, 3~Rocky Mountain, 4-Creat Plains/Midwest, S'SOUth Central, 6-Southeast. 7-Great Lakes. B-Hiddle Atlantic. 9-Northeast. Alaska and Hawaii are nOt included in the Continental United States totals. ASCUS Supply/Demand January, 1988 157 JAvaRy, 1958 REPORT 7 E 9 Continental United States 1988' 1987 1980' 1985 1984 .4 \O m N 1976‘ Ag Art Bil. Ed. Eus. Comp. Sci. Couns.-El. Couns.-Sec. Data Proc. Dr. Ed. El.-Frim. El.—1nter. English ealth Ed. Home Ec. 1nd. Arts Earth General Physics Soc. Sci. Soc. Kerk Speech Deaf Ed. ED/PSA C;fted LD HE rm Reading Sp./Aud. 3.21 3.40 3.35 3.28 3.29 3.38 3.36 3.19 3.20 -— COMP. 5 ' Considerable Shortage, 4 = Some Shortage. 3 ‘ Balanced, 2 = Some Surplus. l - Considerable Surplus From October. survey of Teacher Placement Officers James N. Akin Kansas State University *Hailings for the 1976. 1986 and 1988 reports included all teacher placement offices which were members of ASCUS. APPENDIX F Letter to Principal 158 APPENDIX F Letter to Principal Dear Principal: In the fall of 1987. your school participated in a project that was developed by researchers at the National Association of Secondary School Principals and Michigan State University. The project examined the relationship betWeen different aspects of the school (such as goals and school climate) and school effectiveness. I was responsible for compiling all of the data we received onto a computer system and analyzing this data. I am now working on my dissertation and would like to use this data to examine a different problem facing schools today, that of teacher turnover. I am attempting to identify some of the variables that have an effect on the rate of turnover in schools. In order to complete my dissertation, I need some further information from you. There are seven questions on the enclosed form asking about teachers that have left the school and whether or not your teachers are unionized. This task should not take more than five minutes to complete. It would be extremely helpful if you could provide this information to me, and return it to me in the enclosed envelope._ You Will notice a number at the top of the questionnaire. This identification number was used in the earlier Project, and will be used to match your responses on the form to the data your school provided earlier. 159 Letter to Principal (Cont’d) When my dissertation is completed, I would be happy to provide you with a summary of my results. Although the summary will not provide specific information on your school, it will provide information on the types of variables that influence teacher turnover based on all of the schools in the sample. The information that you provide will be treated with strict confidence. You indicate your voluntary agreement to participate in this study by completing and returning this questionnaire. If you have any questions, please contact me at (517) 353—5324 (office) or (517) 882—4623 (home). Thank you very much for your help on this project. Sincerely, Mary L. Doherty 160 Letter to Principal (Continued) 1. What percentage of your teachers belonged to a union during the 1987—1988 school year? 2. How many teacher full—time equivalent (FTEs) voluntarily left your school during or at the conclusion of the 1988—1989 school year (Do not include retirees or those who were asked to leave)? I would also appreciate your assessment of the teachers’ overall skills and abilities. If you would rather not provide these ratings, or feel that you did not know the teachers well enough to rate their skills, I would appreciate it if you would still provide the above information. All information that you provide will be treated confidentially. 3. Out of the teachers who have left, how many of them would you consider to be excellent teachers? 4. Out of the teachers who have left, how many of them would you consider to be better than average teachers? 5- Out of the teachers who have left, how many of them would you consider to be average teachers? 5- Out of the teachers who have left, how many of them would you consider to be below average teachers. 7- Out of the teachers who have left, how7many of them would you consider to be poor teachers. If you would like a copy of the summary of resulSS Mentioned earlier, please write your name and ad ress below. Thank you again for your help. APPENDIX G Construct Labels associated with Etas and Ksis APPENDIX G Construct Labels associated with Etas and Ksis Eta or Ksi ETAl ETA2 ETA3 ETA4 KSIl KSIZ KSI3 KSI4 KSIS KSIS KSI7 Construct Label Environmental Climate Degree of Latitude Climate Satisfaction Turnover Organizational Size Union Presence Principal Turnover Teacher Salary Teacher Demand Ratings Professional Unemployment Organizational Performance (Achievement) APPENDIX H LISREL Program Used to Evaluate Hypothesized Model APPENDIX H LISREL Program Used to Evaluate Hypothesized Model ooooooo ooooooo ooooooo ooooooo ooooooo oooooo_ ooooooo A.uac munch ~unmp Hmnma HquQ Dunwm HLH>J Hmnxu hnxz vnwz mnxz vn>Z OE Eumwuu2MZDDZ4EEDD UHKmEDZ 25 «2 .s,_Fo—o A—Hmc o» «a 0000 AFHVV up «a .oooso.___ A—_o.c we «a oooooooooooosooosooo_soooooo A_~mmc In «a ____coo oooop —o 000009 000000— A_Hsc «o «a 0.00 00: ocoo ocoo fl_aec mm «a coooooo COCOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO .mou>n oom.ovw.. mm: motm maoommucoo “meats. oi» .w_nm_wm>m >LOEoE co mcu>c oom.ovm._ mcm mcwzp mum: ozw o no, om.u2e H2 <> um ma >H mm :0 O we. op.op.op.o_.o_.oo.o_.oo. o mop l0 Am.muwc c cc— /0 oh <2 0 mo, 11 0.0.0.0. o No, C.mu: o I: we <2 0 0c, m.o.o.o.m.o.m.m.m.m. 0 mm A_.~uo,c 0 mm ma <2 0 mm o.c.c.o.o.o.o.opo.o.o.o.m.o.o.o.m.O.opo.m.~.o_o.o.o—o.op 0 mm A,.Nummw 0 mm In <2 0 cm N.F.—.m.o.o.o. 0 mm o.o.o.o.N.m.o. 0 mm o.o.o.o.o.o.m. o _m o.o.o.o.o.o.~. o om A—.Nunv 0 mm <0 <2 0 mm 0.0.0.0. o no o.o.m.m. 0 mm 0.0.0.0. 0 mm 0.0.0.0. O we A..Nuvc 0 mm mm «2 0 mm 165 huuok H 00. Lump Hunma _unIa zunwm 0uu>4 .muX; nuxz vnmz mnxz an): 02 Am0. mm.- mw. 00. mo. m_.- 00.- m_. 00., 00.- mm. mm. ,0. 0N.- 0,. N0. 00.- 00.. u 21 Exn<2 mmpuoz N—HHZ do OJm m_.© ZO_maw> I _> 4 m a m h 4 m .0 ma; .\‘-..;_._'.! 166 0m0uzp Hz <> mm mm >h mm 00 A~.mumv 0? <2 A—.muq0 mp <2 AF.~¢0.0 ma <2 Ap.mu0m0 In «2 :.mu: 40 as. 167 OZ ZO~h—b<>Hmwo ameO hmm—L OZ mzonmmszwK mmmOum mohuqu mm) mwumozn ZO~k meZOU DZ< mmuz<~z<> mm> whuwumm 4 mpZNEOZ DthHu OZ wmh<§~hmw no MZOHP mw34<> I h mm) mzommw OKmwmmo mo umeDZ h qum<fim<> I ~mx no mmeDZ v mw4m<~m<> I (kw LO ammEDZ 0 mw4m<_a<> 1 x no awmiDZ v mw4m<_z<> I > no zmmEDZ m. mw4m<~u<> kDQZH no mwmijz .0 ZO_mam> I H> 4 w a m ~ 4 m~m>44<2< 0NN.0 050.0 00—.0: 0p—.0 0m..0| 0v0.0| 0—0.0 0m0.0| 000.0» 000., m~h44 (00244 .8. .8. .mx. 5... .3. .3. .9. 40m 40w (hm (hm (hm (kw (km (#0 2uama QEwZD 02420 44<2< 20~H 400240 (hm (hm «hm 4<2< zomhdhqmwm.0 . m . ._~.0__- .._.:___ __ .. _ . 172 N00.0 m. 100.0.11 mmem-ql 0202 02<20 00..0 00..mls- 02020 02420 000.0 v (hm m20~0<00w J I X 000.. whqah qumq~a<> I > 00..0 00..0 .(mu. 02.00 000.0 0.40. 000.0 m-m-mzlu- NNm.0 - p m < w mv0.0 N «“0024; .0001.000.. 202.x<2. mmN02.Nmm 0000.. m ~m>0 I x aOm mZO~F44wmmOU wJa~h402 0wa430m 000.. 000.. 000.. 000.. wk I > aOu MZOHh4memOu w4a~b402 003430w 0N.2.000 0>.N.mca N02 0. «N400 04 .mOmQV 00.00. m. 2000010 00 mummOwD 0m 10.3 wa400m:~IU . 00002 04013 010 300 Fun 00 Mmmz0ooo uO mmm0m4m2 176 N.0. 000.0 I N mx 000.0 000.0 mmN.0 nun.om N .mx 000. 004. wow. 50.. own. 000. 000. OPO*‘N"'LDO n '— m X 004.0 000.0 0 .mx 000.0 .mo.v N.o.o 0N..o 0 .m0 000. 000. .00. Q00. 00h. .Nv. 0N0. lOC‘GJF/QINOO , <1 G I l 1 u? D—( W: x I '— m00.. 440.0 000.0 m .mx 000.0 400.. mmO.N th.0 I m .mx 000. 000. 000. 0NN. 00N. 0N0. 0NN. mNN. OQ'JOQLDLOOF mI.m X 00.. v00. mum. v04. BLDCDN m.m.. 0N..N 0Nh.m 0.0.0 000.0 0 .mx h_m.N 000.0 000.0 mVN.N m me 000- 000. n0.. 000. uJNOOO .Nm. 000. mam. 0... 000. 000. 000. nvN. m OOPOOWW" .— W X 000.0 0N0.N 000.0 000.0 000.. Nmo.0 NIISixI‘ 0.0. mom. ©v.. 000. 0v.. 00N. 000. 00.. vv.. N _m OOOQOVNO X 00N.N N00.N 000.0 000.0 N 400 . ~m 00m. 000. 400. 000. .00. 000. XOOOOPO" 000.0 000.0 000.0 000.0 I .Ia . ~mx 400.0 000.N 00m.N 040.0 0.0.0 mmN.. 000.0 000.0 42240 . .mx v-NC'DVLD‘DY‘ v-NC'DQ «~qu x 400244 N00.. 000.. 000.0 000-0 . 4pm > 400240 mw0N02. .mx .mx .wx .mx .mx .mx .mx 4km 4km 4km 4km 4pm 4km 4km 4km 2000a 0202: 02420 JamuN m2.za 20.2: mODNm 0..40 0.4a» m~h4m N2.JU .2.00 20.h40.u.002 m.m>0 «om244 .mx .mx .mx .mx .mx ~mx .mx 4PM 4PM 4km 4km 4km 4km 4.0 4km 2uawa Q2023 02420 J4muh mz.aa ZO . ZD mOka Omh4a wp4ah m.h4m N2HJU .2H40 macamw 0&4024bm m0m>4424 20.h4»awmm.0 179 00.. .0..0 40..0 00..0 00..0 000.0 .N0.0 000.0 Enzma a2wZD 02420 J4m0h mz.ma ZOHZD monhm OHH4a 4h4w0 (hwlk 000.0 000.0 000.0 000.0 mh4mh mmk4m N2~40 .2NJU mam 4hth m00.0 000.0 000.0 000.0 mom.0 000.0 N.m.0 00..0 050.0 N0..O v 4.0 m 4pm N 4km . 4pm —N(')V 4km 4km 4km 4km 180 000.0 000.0 000.0 000.0 000.0 000.0 000.0 N .wx 000.0 000.0 000.0 000.0 000.0 0.0.5 0 me 000.0 000.0 000.0 www.ml 000.0 m .mx 000.0 000.0 0.0.m .00.. 4 .mx 000.0 000.0 000.0 0 .mx 000.0 000.0 N .wx 000.0 . .mx 5 .mx 0 .mx 0 .mx 4 me m me N .0: . .mx .IQ .N0.0 Nmm.Nn Nom.0 44..MI 000.0 000.0 000.0 4 4km 000.0 000.0 000.0 000.0 000..: 00..0: 000.0 m 4km 000.0 000.0 000.0 000.0 000.0 000.0 0.4.01 N 4km 000 0 000.0 000.0 000.0 000.0 000.0 m.m..| . 4km 5 .mx 0 .mx m .mx 4 .mx 0 .mx N wa . .0: 42240 000.0 000.0: 000.0 000.0 4 4pm 000.0 000.0 <0N.. m00.0 0 4pm 000.0 000.0 000.0 000.0 N 4km 000.0 000.0 000.0 000.0 . (bu v 400244 me44>Ik m.m>4424 ZOHH4bmwmm.0 181 N00.pi? 02420 NNm.0 .2me 5.0.0 000.0 000.0 20.20 law. 000.0 05N.N m 4.0 050.0 00050 000.0 N2H4U 000.0 000.0 050.0 N 450 000.0 0.540 45400 45015 000.0 .2H4U 000 45015 000.0 4 450 004.0 m 450 0.0.0 N 450 050.0 . 450 . 450 .mu 182 00..0: 050.0 000.0: :02420 000.0 coc.o 000.. 02420 00..0 00..0: 05..0: 400.0: 44mg. :co.0 000.0 :00.0 450.. :44m05 2.0Q 000.0: 00..0 05..0 000.0: 000.0 mZ.aQ 000. 000. 000. 000. 000. POOOO J) 040.0 04..0 .40.0: 4N0.0 00..0: 000.0 20.23 000.0 000.0 00N.0: 00N.0 000.0 000.. 20.23 00..0: 000.0: 00N.0 05N.0 0.0.0: 0.0.0: 000.0 moamm 000. 00N. 000. 000. 000. 000. 000. v’CDCDCJOCDCD o..4m 000.0 004. 000. 000. 000. 000. 000. 000. POOOOOO III04H4m m4430.mwa 0055.m m_o.o ONN.o oo,.o omc.o Nmo.o omo.o o.o.o: o.o.o mNo.o Nm_.o: om..o- on..o- omo.o o.,.o mNo.o ONo.o- mo..o mNo.o: omo.o- oN..o- NoN.o- Nmo.o- oc..o o...o- o...o- mmo.o m...o- mmo.o- oNo.o- woo.o- ooo.o ooo.o 0.0.0: 040.0: 4...0: 000.0: N.o.o ,oo.o o.o.o ooo.o ooo.o ooo.o w.4 5N0.0: 000.0 000.0 000.0 m 450 450 N00.0 550.0: 000.0 000.0 m .0: > N00.0 550-0: 000.0 000.0 0 .0x 450 000.0: 000.0: 0NN.0 000.0 000.0 N 450 000.0 000.0: 000.0 000.0 N .mx 000.0 000.0: 000.0 000.0 N .mx 40.N 000N 043 02.5:3Q0 0I5 000.0: N4m.0 000.0 000.. . 450 20 450 00 0500000 44505 000.0: N4m.0 000.0 000.0 . 450 20 450 00 0500000 44505 .00.0 440.0: 000.0: 0...0: . .0x 20 .02 00 0500000 44505 .00.0 440.0: 000.0:. 0...0: . .0x 20 .0x 00 0500000 44505 v-NC'JV v-NC'DV 000.3000 204000Q 015 05405 0.540 N2.40 .2.40 .X002. >5.4.04500 0000Q02405:A4500:.0*.450ml.0 00 0344>200.0 5000044 450 450 450 450 05405 0.540 N2.40 .2.40 450 450 450 450 0500000 44505 0.0>4424 20.5450000.0 L I ST OF REFERENCES LIST OF REFERENCES Abelson, M. A., & Baysinger, B. D. (1984). Optimal and dysfunctional turnover: Toward an organizational level model. Academy of Management Review, g, 331-341. Akin, J. N. (1989). 1988 Teacher Supply/Demand Report. Addison, IL: Association for School, College and University Staffing, Inc. Angle, H. L., & Perry, J. L. (1986). Dual commitment and labor—management relationship climates. Academy of Management Journal, _g, 31—50. Angle, H. L., & Perry, J. L. (1981). An empirical assessment or organizational commitment and organizational effectiveness. Administrative Sc1ence Quarterly, gg, 1—14. Arnold, H. J., & Feldman, D. C. (1982). A multivariate analysis of the determinants of job turnover. Journal of Applied Psychology, 81, 350—360. BaYSinger, B. D., & Mobley, W. H. (1983). Employee turnover: Individual and organizational analysis. In D. M. Rowland & G. R Ferris (Eds.), Research in personnel and human resource management, 1, 269-319. 185 186 Bentler, P. M. (1980). Multivariate analysis with latent variables: Causal modeling. Annual Review of Psychology, _l, 419-456. Billings, R. S., & Wroten, S. P. (1978). Use of path analysis in I/O psychology: Criticisms and suggestions. Journal of Applied Psychology, g3, 677-688. Bluedorn, A. C. (1982a). A unified model of turnover form organizations. Human Relations, §§, 135-153. Bluedorn, A. C. (1982b). The theories of turnover: Causes, effects, and meaning. In S. Bacharach (Ed.), Research in the soCiongy of organizations. Greenwich, Conn.: JAI Press, 75-128. Bollen, K. A. (1989). Structural equations with latent variables. New York, NY: John Wiley & Sons. Borjas, G. J. (1979). Job satisfaction, wages, and unions. Journal of Human Resources, 13, 21—40. Buchanan, B. (1974). Building organizational commitment: The socialization of managers in work organizations. Administrative Science Quarterly, 12, 533-546. Burton, J. F., & Parker, J. E. (1969). Interindustry Industrial variations in voluntary labor mobility. and Labor Relations, _g, 199-216. 187 Cameron, K. S., & Whetten, D. A. (1983). Qgganizational effectiveness. New York: Academic Press. Cameron, K. S., & Whetten, D. A. (1983). Organizational effectiveness: One model or several? In K. S. Cameron & D. A. Whetten (Eds.), Qgganizational effectiveness. New York: Academic Press. Cammann, C., Fichman, M., Jenkins, D., & Klesh, J. (1979). The Michigan Organizational Assessment Questionnaire. Unpublished manuscript. Ann Arbor, MI: University of Michigan. Campbell, J. P. (1977). On the nature of organizational effectiveness. In P. S. Goodman & J. M. Pennings (Eds.), New perspectives ongggapizational effectiveness. San Francisco, CA: Jossey—Bass Publishers. Carlson, R. O. (1962). Executive succession and organizational change. Danville, IL: Interstate Printers & Publishers, Inc. Carsten, J. M., & Spector, P. E. (1987). Unemployment, job satisfaction, and employee turnover: A meta- analytic test of the Muchinsky model. Journal of Applied Psychology, _;, 374—381. Cotton, J. L., & Tuttle, J. M. (1986). Employee turnover: nd review with implications for Academy of Management Review, _l, 55—70. A meta—analysis a research. 188 Cudeck, R. (1989). Analysis of correlation matrices using covariance structure models. Psychological Bulletin, L_§, 317-327. Dalton, D. R., & Todor, W. D. (1979). Turnover turned over: An expanded and positive perspective. Academy of Managgment Review, i, 225-235. Dalton, D. R., & Todor, W. D. (1982). Turnover: A lucrative hard dollar phenomenon. Academy of Management Review, 1, 212-218. Dalton, D. R., Todor, W. D., & Krackhardt, D. M. (1982). Turnover overstated: The functional taxonomy. Academyfof Managgment Review, 1, 117-123. Dansereau, F., Cashman, J., & Graen, G. (1973). Instrumentality theory and equity theory as complementary approaches in predicting the relationship of leadership and turnover among managers. Qgggnizational Behavior and Human Performance, IQJ 184—200. Dansereau, F., Graen, G., & Haga, W. J. (1975). A vertical dyad linkage approach to leadership within formal organizations. Organizational Behavior and 3, 46-78 Human Performance, Dreher, G. F. (1982). The role of performance in the turnover process. Academy_of Management Journal, 5, 137—147. 189 Drexler, J. A. (1977). Organizational climate: Its homogeneity within organizations. Journal of Applied Psychology, fig, 38—42. Dugoni, B. L., & Ilgen, D. R. (1981). Realistic job previews and the adjustment of new employees. Academy of Managgment Journal, 24, 579-591. Eagley, R. V. (1965). Market power as an intervening mechanism in Phillips Curve analysis. Economica, gg, 48-64. Farber, H. S. (1980). Unionism, labor turnover, and wages of young men. Research in Labor Economics, Q, 33—53. Farkas, A. J., & Tetrick, L. E. (1989). A three-wave longitudinal analysis of the causal ordering of satisfaction and commitment on turnover decisions. Journal of Applied Psycholggy, 13, 855-868. Ferris, G. R. (1985). Role of leadership in the employee withdrawal process: A constructive replication. Journal of Applied Psychology, 1Q, 777-781. Fleishman, E. A., & Harris, E. F. (1962). Patterns of leadership behavior related to employee grievances and turnover. Personnel Psychology, 1Q, 43-56. Franklin, J. L. (1975). Power and commitment: An empirical assessment. Human Relations, 28, 737~753. 190 Freeman, R. B. (1980). The exit-voice tradeoff in the labor market: Unionism, job tenure, quits, and separations. Qaarterly Journal of Economics, ag, 643-673. Gaudet, F. J. (1960). Labor turnover: Calculation and cost. American Management Association, Research Study 39. George, J. R., & Bishop, L. K. (1971). Relationship of organizational structure and teacher personality characteristics to organizational climate. Administrative Science Quarterly, lg, 467—475. Glick, W. H. (1985). Conceptualizing and measuring organizational and psychological climate: Pitfalls in multilevel research. Academy of Managament Review, lg, 601-616. Glick, W. H. (1988). Response: Organizations are not central tendencies: Shadowboxing in the dark, round 2. Academy of Management Review, 13, 133-137. _—. Goodman, P. S., Atkin, R. S., & Schoorman, F. D. (1983). On the demise of organizational effectiveness studies. In K. 8. Cameron & D. A. Whetten (Eds.), Qgganizational effectiveness: A comparison of multiple models. New York: Academic Press. 191 Goodman, P. S., & Pennings, J. M. (1977a). Nag perspectives on organizational effectiveness. San Francisco, CA: Jossey-Bass Publishers. Goodman, P. S., & Pennings, J. M. (1977b). Perspectives and issues: An introduction. In P. S. Goodman & J. M. Pennings (Eds.), New perapectives on organizational effectiveness. San Francisco, CA: Jossey-Bass Publishers. Graen, G. & Cashman, J. F. (1975). A role-making model of leadership in formal organizations: A developmental approach. In J. G. Hunt & L. L. Larson (Eds.), Leadership frontiers (pp. 143—165). Kent, OH: Kent State University. Graen, G. B., Liden, R. C., & Hoel, W. (1982). Role of leadership in the employee withdrawal process. Journal of Applied Psycholqu, g1, 868-872. Griffeth, R. W., & Hom, P. W. (1988). A comparison of several conceptualizations of perceived alternatives in turnover research. Journal of Orgapizational Behavior, 9, 103—111. Grusky, O. (1959). Role conflict in organization: A study of prison camp officials. Administrative Science Quarterly, ;, 452-472. 192 Guion, R. M. (1973). A note on organizational climate. Organizational Behavior and Human Performance, a, 120—125. Hage, J., & Aiken, M. (1967). Relationship of centralization to other structural properties. Administrative Science Quarterly, 12, 72—92. Harman, H. H. (1967). Modern factor analysis. Chicago, IL: University of Chicago Press. Hayduk, L. A. (1987). Structural equation modeling with I LISREL: Essentials and advances. London: The Johns Hopkins University Press. Heller, R. M., Guastello, S. J., & Aderman, M. (1982). Convergent and discriminant validity of psychological and objective indices of organizational climate. Psychological Reports, al, 183—195. Hellriegel, D., & Slocum, J. W. (1974). Organizational climate: Measures, research and contingencies. A9adsmx_2I_Maneg§msa£_lgsrnaly 11. 255-280- Hollenbeck, J. R., & Williams, C. R. (1986). Turnover functionality versus turnover frequency: A note on work attitudes and organizational effectiveness. Journal of Applied Psychology, _1, 606-611. 193 Hom, P. W., Griffeth, R. W., & Sellaro, C. L. (1984). The validity of Mobley’s (1977) model of employee turnover. Organizational Behavior and Human Performance, £3, 141—174. Hom, P. W., Katerberg, R., & Hulin, C. L. (1979). Comparative examination of three approaches to the prediction of turnover. Journal of Applied Psychology, _4J 280—290. Hornor, E. R. (Ed.) (1989). Almanac of the 50 states: Basic data profiles with comparative tables. Palo Alto, CA: Information Publications. Hulin, C. L., Roznowski, M., & Hachiya, D. (1985). Alternative opportunities and withdrawal decisions: Empirical and theoretical discrepancies and an 7, 233-250. integration. Psychological Bulletin, Iaffaldano, M. T., & Muchinsky, P. M. (1985). Job satisfaction and job performance: A meta—analysis. Psychological Bulletin, _1, 251-273. Ilgen, D. R., & Seely, W. (1974). Realistic expectations as an aid in reducing voluntary resignations. Journal of Applied Psycholagy, Q9, 452—455. Indik, B. P. (1965). Organizational size and member participation. Human Relations, 18, 339-350. 194 Jackofsky, E. F., & Slocum, J. W. (1988). A longitudinal study of climates. Journal of Organizational Behavior, 9, 319-334. James, L. R. (1980). The unmeasured variables problem in path analysis. Journal of Applied Psychology, aa, 415-421. James, L. R. (1982). Aggregation bias in estimates of perceptual agreement. Journal of Applied Psychology, £1, 219-229. James, L. R., Joyce, W. F., & Slocum J. W. (1988). Comment: Organizations do not cognize. Academy of Management Review, 13, 129-132. Joreskog, K. G., & Sorbom, D. (1986). LISREL: Analysis of linear structural relationships by the method of maximum likelihood. Mooresville, IN: Scientific Software, Inc. Joyce, W. F., & Slocum, J. W. (1979). Climates in organizations. In S. Kerr (Ed.), Organizational Behavior. Columbus, OH: Grid Publishing Inc. Joyce, & Slocum, (1984). Collective climates: Agreement as a basis for defining aggregate climates in organizations. Academy of Management Journal, _1, 721—742. 195 Katzell, R. A., Barrett, R. S., & Parker, T. C. (1961). Job satisfaction, job performance, and situational characteristics. Journal of Applied Psychology, gg, 65—72. Keeley, M. (1980). Organizational analysis: A comparison of organismic and social contract models. 5, 337-362. Administrative Science Quarterly, Kelly, E. A., Glover, , Keefe, J. W., Halderson, , Sorenson, , & Speth . (1986). School Climate Survey. Reston, VA: National Association of Secondary School Principals. ( Kerr, W. A. (1947). Labor turnover and its correlates. Journal of Applied Psycholagy, _1, 366—371. Kerr, W. A., Koppelmeier, G. J., & Sullivan, J. J. (1951). Absenteeism, turnover and morale in a metals fabrication factory. Occupational Psychology, 2Q, 60-55. Kozlowski, S. W. J., & Doherty, M. L. (1989). Integration of climate and leadership: Examination of a neglected issue. Journal of Applied Psychology, 14, 546—553. Kozlowski, S. W. J., & Hults, B. M. (1987). An exploration of climates for technical updating and performance. Personnel Psychology, 49, 539-564. 196 LaFollette, W. R., & Sims, H. P. (1975). Is satisfaction redundant with organizational climate? Organizational Behavior and Human Performance, L3, 257—278. Lawler, E. E. (1981). Pay and organization development. Menlo Park, CA: Addison-Wesley Publishing Company. Lee, T. W., & Mowday, R. T. (1987). Voluntarily leaving an organization: An empirical investigation of Steers and Mowday’s model of turnover. Academy of Management Journal, 39, 721—743. Levine, E. (1957). Turnover among nursing personnel in general hospitals. Hospitals, _l, 50—53, 138, 140. Longest, B. B., & Clawson, D. E. (1974). The effect of selected factors on hospital turnover rates. Personnel Journal, 53, 30—34. Macy, B. A., & Mirvis, P. H. (1976). A methodology for assessment of quality of work life and organizational effectiveness in behavioral-economic terms. Administrative Science Quarterly, 2;, 212—226. March, J. G., & Simon, H. A. (1958). Organizations. New York: Wiley. Martin, T. N. (1979). A contextual model of employee Academy of Management Journal, turnover intentions. 2, 313-324. 197 Martin, T. N., Price, J. L., & Mueller, C. W. (1981). Job performance and turnover. Journal of Applied Psychology, 66, 116-119. McCain, B. E., O’Reilly, C., & Pfeffer, J. (1983). The effects of departmental demography on turnover: The case of a university. Academy of Maaagement Journal, 26, 626—641. McKemey, D. R., & Sims, H. P. (1977). Personnel turnover: An expectancy theory approach. Paper presented at the 37 Annual Meeting of the Academy of Management, Orlando, FL. McKemey, D. R., & Sims, H. P. (1980). A cognitive-ordered model of turnover behavior. Working paper, The Pennsylvania State University. Michaels, C. E., & Spector, P. E. (1982). Causes of employee turnover: A test of the Mobley, Griffeth, Hand, and Meglino model. Journal of Applied Psychology, 61, 53—59. Miller, H. E., Katerberg, R., & Hulin, C. L. (1979). Evaluation of the Mobley, Horner, and Hollingsworth model of employee turnover. Journal of Applied Psychology, 66, 509-517. 198 Mitchel, J. O. (1981). The effect of intentions, tenure, personal, and organizational variables on managerial 4. turnover. Academy of Management Journal, 742—751. Mobley, W. H. (1982a). Employee turnover: Causes, consequences, and control. Menlo Park, CA: Addison-Wesley. Mobley, W. H. (1982b). Some unanswered questions in turnover and withdrawal research. Academy of Management Review, 1, 111-116. Mobley, W. H., Griffith, R. W., Hand, H. H., & Meglino, B. M. (1979). Review and conceptual analysis of the employee turnover process. Psycholagical Review, 66, 493-522. Mowday, R. T., Koberg, C. S., & McArthur, A. W. (1984). The psychology of the withdrawal process: A cross- validation test of Mobley’s intermediate linkages model of turnover in two samples. Academy of Management Journal, 21, 79-94. Mowday, R. T., Steers, R. M., & Porter, L. W. (1979). The Journal of rement of organizational commitment. measu Vocational Behavior, lg, 224-247. Motowidlo, S. J. (1983). Predicting sales turnover from pay satisfaction and expectation. Journal of Applied Psychology, 66, 484—489. 199 Muchinsky, P. M., & Morrow, P. C. (1980). A multidisciplinary model of voluntary turnover. Journal of Vocational Behavior, 17, 263-290. Muchinsky, P. M., & Tuttle, M. L. (1979). Employee turnover: An empirical and methodological assessment. Journal of Vocational Behavior, lg, 43-77. Mueller, C. W., & Price, J. L. (1989). Some consequences of turnover: A work unit analysis. Human Relations, 42. 389-402. Newman, J. E. (1977). Development of a measure of perceived work environment. Academy of Management Journal, 20, 520-534. Newton, K., Betcherman, G., & Leckie, N. (1981). The determinants of voluntary separation rates. Industrial Relations Journal, 16, 72-76. Ostroff, C., & Kozlowski, S. W. J. (1986). The utility of composition theories for elaborating models of organizational phenomena. Unpublished manuscript. Parasuraman, S. (1982). Predicting turnover intentions and turnover behavior: A multivariate analySis. Journal of Vocational Behaviog, _g, 111—121. Payne, R., & Pugh, D. S. (1976). Organizational structure and climate. In M. D. Dunnette (Ed.), Handbook of industrial and organizational psychology. Chicago, IL: Rand McNally. 200 Pfeffer, J. (1983). Organizational demography. Research in Organizational Behavior, 6, 299-357. Podsakoff, P. M., & Organ, D. W. (1986). Self-reports in organizational research: Problems and prospects. Journal of Management, g6, 531-544. Porter, L. W. (1968). The etiologygof organizational commitment: A longitudinal studyyof initial stages of employee—ogganization relationships. Unpublished manuscript. Porter, L. W., & Smith, F. J. (1970). The etiology of ( organizational commitment and managerial turnover. Unpublished paper. Irvine, CA: University of California. Porter, L. W., & Steers, R. M. (1973). Organizational, work, and personal factors in employee turnover and 0, 151-176. absenteeism. Psychological Bulletin, Price, J. L. (1977). The study of turnover. Ames, Iowa: The Iowa State University Press. Pritchard, R. D., & Karasick, B. W. (1973). The effects of organizational climate on managerial job performance and job satisfaction. Organizational Behavior and Human Performance, 6, 126—146. 201 Pugh, D. S., Hickson, D. J., Hinings, C. R., McDonald, K., Turner, C., & Lupton, T. (1963). A conceptual scheme for organizational analysis. Administrative Science Quarterly, 6, 289—315. Reichers, A. E. (1985). A review and reconceptualization of organizational commitment. Academy of Management Review, g6, 465-476. Reilly, R. R., Tenopyr, M. L., & Sperling, S. M. (1979). Effects of job previews on job acceptance and survival of telephone operator candidates. Journal of Applied Psycholagy, 6g, 218-220. Roberts, K. H., Hulin, C. L., & Rousseau, D. M. (1978). Developing an interdisciplinary science of organizations. San Francisco, CA: Jossey-Bass. Rousseau, D. M. (1984). Theories of level in organizational science. Paper presented at the 92nd Annual Convention of the American Psychological Association. Toronto, Canada. Rousseau, D. M. (1985). Issues of level in organizational research. Research in Organizational Behavior, 1, 1-37. Rousseau, D. M. (1988). The construction of climate in organizational research. In C. L. Cooper & I. Robertson (Eds.), International review of industrial and organizational psychology. 202 Schmitt, N., & Bedeian, A. G. (1982). A comparison of LISREL and two—stage least squares analysis of a hypothesized life-job satisfaction reciprocal relationship. Journal of Applied Paychology, 61, 806-817. Schmitt, N., & Doherty, M. L. (1988). NASSP study of measurement and model linkagayissues for the comprehensive assessment of school environments. Technical Report. E. Lansing, MI: Michigan State University. Schmitt, N., & Ostroff, C. (1987). Pilot study of measurement and model linkagg issues for the comprehensive assessment of school environments. Technical Report. E. Lansing, MI: Michigan State University. Schneider, B. (1987). The people make the place. Personnel Psychology, g6, 437-453. Schneider, B., & Bowen, D. E. (1985). Employee and customer perceptions of service in banks: Replication and extension. Journal of Applied Psychology, 70, 423—433. Schneider, B., & Reichers, A. E. (1983). On the etiology of climates. Personnel Psychology, 36, 19-39. 203 Schneider, B., & Snyder, R. A. (1975). Some relationships between job satisfaction and organizational climate. Journal of Applied Psychology, 66, 318-328. Seaton, F. W. (1984). A comparative analysis of current models of organizational effectiveness within the context of organization theopy. Unpublished Manuscript. Shikiar, R., & Freudenberg, R. (1982). Unemployment rates as a moderator of the job dissatisfaction-turnover relation. Human Relations, 66, 845-856. Spencer, D. G., & Steers, R. M. (1981). Performance as a moderator of the job satisfaction—turnover relationship. Journal of Applied Psychology, 66, 511-514. Staw (1980). The consequences of turnover. Journal of Occupational Behavior, 1, 253—273. Staw, B. M., & Oldham, G. R. (1978). Reconsidering our dependent variables: A critique and empirical study. Academy of Management Journal, 61, 539-559. Steel, R. P., & Griffeth, R. W. (1989). The elusive relationship between perceived employment opportunities and turnover behavior: A methodological or conceptual artifact? Journal of applied Psychology, 74, 846—854. 204 Steers, R. M. (1977). Antecedents and outcomes of organizational commitment. Administrative Science Quarterly, gg, 46-56. Steers, R. M. (1977). Qgganizational effectiveness: A behavioral view. Santa Monica, CA: Goodyear. Steers, R. M., & Mowday, R. T. (1981). Employee turnover and post-decision accommodation processes. Research in Qrganizational Behavior, g, 235-281. Stumpf, S. A., Dawley, P. K. (1981). Predicting voluntary and involuntary turnover using absenteeism and performance indices. Academy of Manggement Journal, 24, 148-163. Terborg, J. R. & Lee, T. W. (1984). A predictive study of organizational turnover rates. Academy of Management Journal, _1, 793-810. U. S. Bureau of Labor Statistics. (1988). Geographic profile of employment and unemployment, 1987. Washington, D. C.: U. S. Department of Labor. U. S. Bureau of Labor Statistics. (1966). Measurement of labor turnover (Unpublished study, U. S. Department of Labor). Wagner, W. G., Pfeffer, J., & O’Reilly, C. A. (1984). Organizational demography and turnover in top—management groups. Administrative Science Quarterly, 29, 74-92. 205 Wales, T. J. (1970). Quit rates in manufacturing industries in the United States. Canadian Journal of Economics, g, 123—139. Welsch, H. P., & LaVan, H. (1981). Inter—relationships between organizational commitment and job characteristics, job satisfaction, professional behavior, and organizational climate. flumgn Relations, 34, 1079—1089. Werbel, J. D., & Bedeian, A. G. (1989). Intended turnover as a function of age and job performance. Journal of Organizational Behavior, lg, 275—é81. Williams, L. J., & Hazer, J. T. (1986). Antecedents and consequences of satisfaction and commitment in turnover models: A reanalysis using latent variable structural equation methods. Journal of Applied Psychology, _l, 219—231. Woodman, R. W., & King, D. C. (1978). Organizational climate: Science or folklore? Academy of Management Review, g, 816—826. Woodward, N. (1975/76). The economic causes of labour turnover: A case study. Industrial Relations Joggggl, Q, 19-32. Zohar, D. (1980). Safety climate in industrial organizations: Theoretical and applied implications. Journal of Applied Psychology, gg, 96-102.