Journal of Occupational and Organizational Psychology (1993). 66, 277-284 0 1993 The British Psychological Society
Printed in Great Britain 277
Mental health and work performance: Results of a longitudinal field study
Thomas A. Wright* Managerial Sciences Department, University of Nevada, Reno, NV 89557-401 6, USA
Douglas G. Bonett Department of Management and Marketing, University of Wyoming, Lararnie, WY 8207 1-3275, USA
Dennis A. Sweeney City and County ofSan Francisco, San Francisco, C A 94127, USA
Results from a two-year longitudinal field study supported the hypothesized positive relationship between mental health and subsequent work performance. Relative (retest correlational analysis) and absolute (change in mean level) stability analyses established mental health as a consistent and stable trait. The importance of dispositional (trait) explanations for organizational behaviours are discussed.
Dysfunctional mental health represents a serious cost for industry in terms of both human and financial maladaptive consequences (Cooper & Marshall, 1976). For instance, depres- sion, loss of self-esteem, hypertension, alcoholism and drug consumption have all been shown to be related to dysfunctional mental health (Ivancevich & Matteson, 1980).
Since much research has examined mental health as a dependent variable Uamal & Mitchell, 1980), considerable attention has been devoted to identifying the determinants of employee mental health (Ostell & Divers, 1987; Warr, 1990). For instance, the semi- nal work of Kornhauser (1965) recognized that the determinants of mental health are found in both work and non-work environments. More recent studies have investigated such work factors as job involvement, work overload, role conflict and person-nviron- ment fit (Cooper & Marshall, 1976; Furnham & Schaeffer, 1984; Gechman & Wiener, 1975; Martin, 1984), non-work factors such as family and health issues (French & Caplan, 1970), and the interaction of work and non-work variables on mental health (Sekaran, 1985).
Research considering the consequences of mental health is less prevalent and research examining the work-related behavioural consequences (i.e. work performance, absen- teeism and turnover) of mental health is quite limited (Wright & Bonett, 1992). * Requests for reprints.
278 Thomas A. Wright, Douglas G. Bonett and Dennis A. Sweeney Although such variables as alcoholism and drug abuse have been related to declines in work performance (Quick & Quick, 1984), the direct relationship between mental health and work performance has yet to be shown empirically.
The present research is the first to formally test the relationship between mental health and work performance. Following the well-stated admonition of Kelloway & Barling (1991) for more longitudinal field research on mental health and its correlates, the pre- sent study examined mental health as a predictor of subsequent work performance over a two-year period. The study hypothesis is:
Study hypothesis: A positive relationship exists between mental health and subsequent work performance.
The prominence of dispositional (trait) explanations of organizational behaviours has been well documented (Hoppock, 1935; Schneider, 1987; Staw, 1986). However, much research presuming to establish dispositional or trait influences has been criticized on methodological grounds (Newton & Keenan, 1991). A second objective of the present study was to provide a thorough methodological basis for establishing mental health as a stable trait.
Mental health as a trait While mental health has been treated as both an outcome and predictor variable in the literature, it has usually been viewed as a consistent and stable trait (Wright & Bonett, 1992). In this vein, the present study adopted Kornhausers (1965) conceptualization of mental health as a consistent and stable attribute or trait. Specifically, mental health is defined by Kornhauser as, indicators which, taken in combination, are representative enough of the important parts of a persons total behavior to provide useful estimates as to the satisfactoriness of this general level of psychological and social functioning (p. 17). More recently, Warr (1987, 1990) defined this view of mental health as context-free; a global construct not tied to any particular contextual situation.
Despite the wide use of conceptualizing such variables as mental health as traits, researchers such as Davis-Blake & Pfeffer (1989) and Newton & Keenan (1991) have noted that research investigating the effects of dispositional or trait variables is often plagued by methodological problems. Specifically, they have noted that much trait research is based solely on correlational analysis. Newton & Keenan (1991) further argued that significant correlations cannot, solely at face value, be taken as sufficient proof of vari- able stability because the correlations reveal only the relative, not absolute, positions of individuals in a group (p. 781). That is, high retest correlations do indicate a similarity in score ranking. However, they cannot be used, by themselves, to confirm the absence of absolute change, or stability. Thus, Newton & Keenan (1991) noted that correlational analysis should always be supplemented with an assessment of possible changes in mean variable levels to infer the presence or absence of dispositional effects. In this study, tests for both relative rankings and absolute position were used to establish mental health as a stable trait.
Method and results Thirty-three human services supervisory personnel participated in this two-year study. All subjects were employed within the same department and are representative of the
Mental health and work performunce 279
organizations population. The majority (84.8 per cent) of the subjects are male and all had a four-year college degree. The sample included all subjects who remained within the department, performed the same job over the two-year period, and for whom measures of both mental health (Tl , T2) and work performance (T3) were available.
Data were collected at three time periods (Tl , T2, T3), each separated by one year. Mental health data were collected on the job at T1 and T2. Performance data were col- lected at T3. Complete measures of mental health and performance were available for all 33 subjects.
Mental health In the present study, mental health was assessed with the Index of Psychological Well- Being (Hochstim, 1970). The Index follows Kornhausers (1965) attempt to ascertain an individuals general level of mental health. Mental health is viewed as a global construct, not tied to any particular contextual situation.
The Index measures the reported frequency (never? sometimes? often?) of ways that people feel (in general). The eight items involve an individuals positive (three items) and negative (five items) feelings. The eight items are: on top of the world, very lonely or remote from other people, particularly excited or interested in something, pleased about having accomplished something, bored, depressed or very unhappy, so restless you couldnt sit long in a chair, and, vaguely uneasy about something without knowing why.
Following Berkman (1971) and Wright & Bonett (1992), weights of 0, 1 and 3, respec- tively, were assigned to the responses, never, sometimes and often. These weights were summed into positive and negative scores, ranging from 0 to 9 for positive feelings and 0 to 15 for negative feelings. Following Berkman (197 l), these positive and negative values were combined into an overall Index of Psychological Well-Being score. These composite Index scores ranged from 1 to 7. Specifically, a score of 1 indicates all or almost all neg- ative feelings, a score of 4 indicates about as many positive as negative feelings, while a score of 7 indicates all or almost all positive feelings.
Thus, the composite Index score is a measure of the general balance between the rela- tive frequencies with which the given positive and negative feelings are typically experi- enced by the respondents. Cronbachs coefficient alphas for the positive feeling data were .76 at T1 and .71 at T2, and .69 at T1 and .68 at T2 for the negative feeling results.
The means of mental health at times 1 and 2 were 3.42 and 3.64, respectively. The standard deviations of mental health at times 1 and 2 were 1.41 and 1.52, respectively. A test for the equality of correlated means found that the two means are not significantly different (t(32) = 1.19, n.s.). Furthermore, a test for the equality of correlated variances did not reveal any differences (t(31) = 0.596;p = .5558). These findings indicate that the two mental health scores satisfy the definition of a strictly parallel measurement model (Kristof, 1963) and confirm that the best estimate of the mental health trait is the average of the two mental health scores.
Work performance Following the procedure of Wright & Bonett (1993), a panel of experts composed of three departmental supervisors highly familiar with both the work as well as workforce, unan-
280 Thomas A . Wright, Douglas G. Bonett and Dennis A. Sweeney
imously identified and confirmed several relevant performance dimensions for the human services personnel sampled. The three performance dimensions identified for use in this study were: goal emphasis, team building and work facilitation. All employees ultimately reported to the top-ranking administrative officer in the department. At T3, this top- ranking officer rated each employees work performance for the preceding one-year period on each of the three dimensions. These three dimensions were measured using a five-point scale ranging from never to always regarding the extent to which the employee empha- sized a particular dimension. Ratings on these three dimensions were summed to form a composite measure of performance for each employee. Cronbachs coefficient alpha for the performance measure was .86.
Findings Table 1 reports the means, standard deviations and the intercorrelations for the variables. The primary analysis of mental health for the study hypothesis was conducted using the mean scores obtained by averaging the T1 and T2 mental health scores. As noted above, these mental health scores satisfy the definition of a strictly parallel measurement model (Kristof, 1963), providing the empirical justification for the creation of these mean mental health scores.
Before testing the hypothesized relationship between mental health and work perfor- mance, it was necessary to establish the trait nature of the variable, mental health. To infer a dispositional or trait effect for mental health, tests for both relative (correlational analysis) and absolute (change in mean level) stability were undertaken (Newton & Keenan, 1991). Support was obtained for the relative stability (test-retest correlation of .76, p = .OOOl) of mental health. The level of absolute stability was assessed using a matched t test procedure. As further noted above, the results of this analysis suggested the existence of absolute stability over time for the trait measure of mental health.
We now report the results of our study hypothesis. As predicted, a positive relationship between mental health and subsequent work performance was established (t(3 1) = 2.32, p = .0136; adjusted R2 = .12), with a Pearson correlation of .38. These one-tailed findings provide support for the study hypothesis.
Table 1. Means, standard deviations and Pearson correlations (one-tailed tests)
Variables M S D 1 2 3 4 5 6 7
1. Time 1 mental
2. Time 2 mental
3 . Average mental
4. Composite work
- - ~ _ _ _ _ ~ ~
health 3.42 1.41 - .76**** .93**** .23 .22 .16 .24
health 3.64 1.52 - .94**** .48** .38* .47** .42**
health 3.53 1.37 - .38* .33* .34* .36*
performance 10.21 2.50 - 89**** 39**** ,88**** 5. Goal emphasis 3.48 0.87 - 69**** ,70**** 6. Work facilitation 3.45 1.03 - .66**** 7. Team building 3.27 0.91 -
Mental health and work performance 281
Separate analyses established the relationship between mental health and each subse- quent work performance dimension. Specifically, significant positive relationships between mental health and goal emphasis (431) = 1.921, p = .0320); between mental health and work facilitation (t(31) = 2.012,p = .0265), and between mental health and team building (t(31) = 2.123,p = .0210) were obtained with Pearson correlations of .33, .34 and .36, respectively. These one-tailed findings provide further support for the study hypothesis.
Additional results are reported regarding the relationship between mental health con- sidered separately at T1 and T2, and composite work performance at T3. The relationship between mental health at T1 and subsequent work performance approached significance (t(31) = 1.314,p = .0992), while a significant relationship between mental health at T2 and subsequent work performance (t(31) = 3.065, p = .0023) was obtained. Coupled with the significant findings regarding average mental health, these one-tailed findings provide further support for the hypothesized relationship between mental health and subsequent work performance. Table 1 reports the correlations between each measure of mental health, composite performance, and each separate performance dimension.
Discussion and implications
As predicted in the study hypothesis, mental health was positively related to subsequent work performance. While many studies have examined mental health as an outcome vari- able (Cooper & Marshall, 1976), longitudinal field research investigating the work- related consequences of employee mental health is quite limited (Kelloway & Barling, 1991). The present study was the first to formally examine the role of mental health as a predictor of subsequent work performance.
Furthermore, confidence in the findings is increased by the establishment of both rel- ative (retest correlational analysis) and absolute (change in mean level) stability for the mental health measure. Previous research on dispositional variables relied primarily on the use of correlational analysis to establish relative stability. However, several authors (Davis-Blake & Pfeffer, 1989; Newton & Keenan, 1991) indicated that establishment of relative stability alone is insufficient proof of dispositional or variable stability. The use of both techniques, and the agreement in results, provide support for the view of mental health as a consistent and stable trait.
The small sample size makes a detailed examination of the pattern of association between mental health and work performance difficult. Thus, while interesting results were observed in this study, i t remains for future research using additional subjects in a variety of settings to establish the generalizability of the finding to other employee group- ings.
However, certain tentative conclusions can be drawn based upon a comparison with other samples which used the Index of Psychological Well-Being to measure mental health. For instance, a comparative norm was found in a study by Belloc, Breslow & Hochstim (197 1). Their sample included working and non-working adults from the same geographic area as the present sample. While roughly two-thirds of the subjects in both samples reported about as many positive as negative feelings, almost 15 per cent of the Belloc et al. sample reported all or almost all negative feelings exclusively, compared to none in the present sample.
282 Thomas A. Wright, Douglas G. Bonett and Dennis A. Sweeney This finding of relatively fewer employed workers reporting a preponderance of nega-
tive feelings is predicted by the healthy worker effect (Sheppard, 1982; Wright, 1984). That is, employed individuals, on average, experience less physical (Sheppard, 1982) and mental ill health (Isaksson, 1990) than non-working individuals. While the relationship between employment status and mental health is well established (Ostell & Divers, 1987), further research is now needed which will more closely examine individual and environmental correlates of mental health over time and across varied work settings. For instance, Wright & Bonett (1993) found that employees in the advancement phase of career development (Super, 1980) who changed jobs and/or careers (occupations) experi- enced improved mental health as a result of the jobkareer change.
An alternative explanation to the finding that mental health is positively related to work performance concerns how work performance is measured. For instance, although the dimensions of work performance were objectively derived and validated in the present study, these dimensions range from primarily task specific (i.e. goal emphasis), to pri- marily non-task specific (i.e. team building and work facilitation). An alternative expla- nation for the present finding is that jobs which emphasize non-task specific performance dimensions may result in performance being confounded with supervisory perceptions of employee mental health. In fact, this potential for confounding may be so widespread that it applies irrespective of the performance dimension.
Research does indicate that many jobs (i.e. service oriented) emphasize non-task specific performance dimensions in the appraisal process (Wright, 1992). For instance, Puffer (1987) noted that non-task behaviours, such as citizenship or prosocial behaviour, are relevant to the work context and can have positive implications for the organization. Thus, the appraisal process for many jobs may be primarily based on supervisory percep- tions of employee psychological well-being. If this alternative explanation is supported by future research, the role of trait variables (i.e. mental health) takes on added significance.
Currently, the relative importance of dispositional and situational influences on organ- izational behaviours is being debated within the discipline (Newton & Keenan, 1991; Staw, Bell & Clausen, 1986; Staw & Ross, 1985). For instance, George (1991) concluded that transitory mood or state measures are more important than trait measures in under- standing absence and prosocial behaviour. Staw et al. (1986) argued for a more disposi- tional approach. The identification of mental health as a trait in this study lends support to the importance of dispositional influences on organizational behaviours. Although our results favour the dispositional explanation, given the moderate strength of the correla- tion (.38), more research is needed to further delineate the relative merits of trait versus state explanations for work performance and other organizational behaviours such as employee turnover and absenteeism.
Acknowledgements The authors wish to thank Elizabeth M. Doherty, Kay D. Wright, Daulatram Lund, Michael West and the two anonymous reviewers for their helpful suggestions.
References Belloc, N. B., Breslow, L. & Hochstim, J. R. (1971). Measurement of physical health in a general population
survey. American Journal of EpiAtniology, 93, 328-336.
Mental health and work performance 283 Berkman, P. L. (1971). Measurement of mental health in a general population survey. American Journal of
Cooper, C . L. & Marshall, J. (1976). Occupational sources of stress: A review of the literature relative to coro-
Davis-Blake, A. & Pfeffer, J. (1989). Just a mirage: The search for dispositional effects in organizational
French, J. R. P. & Caplan, R. D. (1970). Psychological factors in coronary heart disease. lndustrial Medicine,
Epidemiology, 94, 105-1 11.
nary heart disease and mental ill health. Journal of0ccupational Psychology, 49, 11-28.
research. Academy ofManagement Review, 14, 385-400.
39,383-397. Furnham, A. & Schaeffer, R. (1984). Person-environment fit, job satisfaction and mental health. Journal of
Occupational Psycho logy, 57, 295-307.
time devoted to work. Journal ofApplied Psychology, 60, 521-523. Gechman, A. & Wiener, Y (1975). Job involvement and satisfaction as related to mental health and personal
George, J. M. (1991). State or trait: Effects of positive mood on prosocial behaviors at work. Journal of Applied
Hochstim, J. L. (1970). Health and ways of living. In I. J. Kessler, & M. L. Levin (Eds), The Community as an
Hoppock, R. (1935).]ob Satisfaction. New York: Harper. Isaksson, K. (1990). A longitudinal study of the relationship between frequent job change and psychological
Ivancevich, J. W. & Matteson, M. T. (1980). Stress and Work: A Managerial Perspective. Glenview, IL: Scott,
Jamal, M. & Mitchell, V .F. (1980). Work, nonwork, and mental health: A model and a test. Industrial
Kelloway, E. E. & Barling, J. (1991). Job characteristics, role stress and mental health. Journal of0cmpational
Kornhauser, A. (1965). Mental Health of the Industrial Worker: A Detroit Study. New York: Wiley. Kristof, W. (1963). The statistical theory of stepped-up reliability coefficients when a test has been divided
Martin, T. N. (1984). Role stress and inability to leave as predictors of mental health. Human Relations, 37,
Newton, T. & Keenan, T. (1991). Further analyses of the dispositional argument in organizational behavior.
Ostell, A. & Divers, P. (1987). Attributional style, unemployment and mental health. Journal of Occupational
Puffer, S. M. (1987). Prosocial behavior, noncompliant behavior, and work performance among commission
Quick, J. C. & Quick, J. D. (1984). Organizational Stress and Preventive Managmnt. New York:
Schneider, B. (1987). People make the place. Personnel Psychology, 40,437-453. Sekaran, U. (1985). The paths to mental health: An exploratory study of husbands and wives in dual-career
Sheppard, B. (1982). Mortality and stress survey of San Francisco probation officers. Pamphlet published by
Snedecor, G. W. & Cochran, W. G. (1980). Statistical Methods. Ames, IA: Iowa State University Press. Staw, B. M. (1986). Organizational psychology and the pursuit of the happy/productive worker. California
Staw, B. M., Bell, N. E. & Clausen, J. A. (1986). The dispositional approach to job attitudes: A lifetime
Staw, B. M. & Ross, J. (1985). Stability in the midst of change: A dispositional approach to job attitudes.
Super, D. E. (1980). A life-span, life-space approach to career development. Journal of Vocational Behavior, 16,
Warr, P. (1987). Work, Unemployment and Mental Health. New York: Oxford University Press. Warr, P. (1990). The measurement of well-being and other aspects of mental health. Journal of Occupational
Psyrhology, 76, 299-307.
Epidemiological Lboratoty, pp. 149-176. Baltimore: Johns Hopkins Press.
well-being. Journal of Occupational Psychology, 63. 297-308.
Relations, 19, 88-93.
into several equivalent parts. Psychometrika, 28, 12 1-1 38.
Journal of Applied Psychology, 76, 781-787.
Psychology, 60, 333-337.
salespeople. Journal of Applied Psychology, 72,6 15-62 1.
families. Journal of Occupational Psychology, 58, 129-1 37.
Local No. 2 1 AFL-CIO. San Francisco: AFL-CIO.
Management Review, 28,40-53.
longitudinal test. Administrative Science Quarterly, 31, 56-77.
Journal of Applied Psychology, 70,469-480.
PJychology, 63, 193-210.
284 Thomas A. Wright, Douglas G. Bonett and Dennis A. Sweeney Wright, T. A. (1984). Examination of the processes of cognitive appraisal, coping, and adaptation in an
Wright, T. A. (1 992). Work satisfaction, work performance, and organization tenure considered as predictors
Wright, T. A. & Bonett, D. G. (1992). The effect of turnover on work satisfaction and mental health: Support
Wright, T. A. & Bonett, D. G. (1993). Role ofemployee coping and performance in voluntary employee with-
organization field setting. Unpublished doctoral dissertation, University of California, Berkeley.
of employee turnover. Journal of Applied Businen Resmt-ch, 8 , 8 3 4 8 .
for a situational perspective. Journal ofOrganizationa1 Behavior, 13,603-61 5 .
drawal: A research refinement and elaboration. Journal of Management, 19, 147-161.
Received 8 April 1992; revised version received 1 G March 1993