foumal of Occupational and Organisational Psychology (2000), 73, 501-510 Printed in Great Britain 501© 2000 The British Psychological SocietyPsychosocial job strain and productivity inhuman service workers: A test of thedemand-control-support modelMaureen F. Dollard*School of Psychology, University of South Australia, Whyalla Campus, AustraliaHelen R. WinefieldDepartment of Psychiatry, University of Adelaide, Adelaide, AustraliaAnthony H. WinefieldSchool of Psychologf, University of South Australia, Adelaide, AustraliaJan de JongeDepartment of Social and Organi^tional Psychology, Utrecht University, Utrecht, The NetherlandsThe aim of the study was to test the main and interactive effects of the keydimensions of the demand—control-support model in predicting levels of strain(specifically emotional exhaustion, depersonalizadon and job dissatisfaction) andfeelings of productivity and competency (personal accomplishment) in a multi-occupational sample of human service workers (A^= 813). Controlling fordemographics, negative affectivity (NA), and quadratic terms, structural equationanalyses showed some support for the additive iso-strain hypothesis: jobscombining high demands, low control and low support produced the lowest levelsof satisfaction in workers. High demands and low supports only were associatedwith high depersonalizadon, and high emotional exhaustion. Support was alsofound for the additive active learning hypotheses: jobs combining high demandsand high control produced the highest levels of personal accomplishment. Thestudy supports job redesign interventions for improving worker well-being andproductivity.International research efforts continue in an attempt to reduce the human andeconomic costs of work stress. Job demand—control (DC) theory argues that theorigins of work stress are situated primarily in the structural or organizationalaspects of the work environment rather than in personal attributes or demographics(Karasek, 1979). The expanded three dimensional demand-control-support (DCS)*Rcquests for reprints should be addressed to Dr Maureen Dollard, School of Psychology, University of SouthAustralia, Whyalla Campus, Nicolson Avenue, Whyalla, South Australia, 5600 (e-mail: [email protected]).502 Maureen F. Dollard et al.model predicts that workers with jobs combining high demands, low control,and low support from supervisors or co-workers are at the highest risk forpsychological or physical disorders (iso-strain hypothesis) (Johnson & Hall, 1988).There has been some contention in the literature that the strain hypothesis issupported only when interaction effects are shown between the work dimensions.Empirical tests of the DC model have shown that large-scale multi-occupationalstudies have tended to provide support for interaction effects between demand andcontrol predicting strain (de Jonge & Kompier, 1997; Schnall, Landsbergis, &Baker, 1994). Smaller scale studies of the DC model in single-occupational sampleshave found primarily main effects of demands and control (e.g. Hurrell & McLaney,1989; Perrewe & Anthony, 1990; Spector, 1987). Epidemiological studies providethe most support for the core assumptions of the DCS model (Amick et ai., 1998;Theorell et ai., 1998). However multiplicative interaction terms (i.e. demands xcontrol X support) were not assessed in these studies and de Jonge and Kompier(1997) have observed that the interaction hypothesis is not often supported inepidemiological studies.Recent reviews of the DCS model (de Jonge & Kompier, 1997; Kristensen, 1995)have argued that the framework is appropriate for further empirical investigation.Specifically the active-passive dimension of the model according to Theorell andKarasek (1996), has been under-utilized in research: 'certainly patterns of activecoping behavior could affect the progression of disease development' (p. 10). Fewstudies have examined the active learning hypothesis (cf. de Jonge & Kompier,1997), that jobs combining high demands and high control would lead workers toexperience feelings of competence and productivity, and accomplishment. Mosthave found empirical support (e.g. Dollard & Winefield, 1998; Karasek, 1981;Landsbergis, Schnall, Deitz, Friedman, & Pickering, 1992) with the exception ofMeijman, Ulenbelt, Lumens, and Herber (1996).The aim of the present study was to test the two main hypotheses (as above)using structural equation modelling to predict differences in self-reported levelsof strain and productivity within a relatively large sample with well-definedoccupational groups. It was reasoned that such a sample would provide morevariability on each of the work dimensions and increase the likelihood ofuncovering multiplicative interactions shown previously (cf. Kristensen, 1995,1996).MethodParticipantsParticipants were employees {N= 1229) of a public sector welfare agency: social workers (A'= 244),psychologists {N=T), youth workers (A'^= 140), community support workers (N=7\), financialcounsellors (A'^= 42), administrative staff (A^= 156), project staff (N= 54), and managers {N= 46).Measures/materialsDemographics assessed were: age, sex, living with partner, education level (from primary school topostgraduate quantifications), and shift work.Psychosocial fob strain and productivity in human service workers 503Negative affecHvity. A 10-item version of the State-Trait Anxiety Inventory (Spielberger, Gorsuch,Lushene, Vagg, & Jacobs, 1983) was used to measure negative affectivity (as recommended by Watson& Clark, 1984), as a potential confounder of the stressor-strain relationship.Work environment. The measures used were 9-item subscales of the Work Environment Scale (WES){Moos, 1986).Demands. The work pressure subscale measured the degree to which the pressure of work andtime urgency dominate the work environment, e.g. 'it is very hard to keep up with yourworkload'.Control. The autonomy subscale measured the extent to which employees are encouraged to beself-sufficient and to make their own decisions.Support. The peer cohesion subscale, which measures the extent to which employees ate friendlyand supportive of one another, e.g. 'staff often talk to each other about their personal problems'was combined with the the supervisor support subscaie, which assesses the extent to whichmanagement is supportive and encourages employees to be supportive of each other, e.g. 'staffdiscuss their personal problems with supervisors'.Strain. The emotional exhaustion subscale (9