Public and private sector work stress: Workers compensation, levels of distress and job satisfaction, and the demand-control-support model

Abstract

Proportionately higher rates of work stress compensation claims in the public sector, and the media's portrayal of stress in the public sector as an epidemic has fuelled the commonly held notion that stress levels are higher in the public sector. However, reports from the research literature are somewhat contradictory. Using a heterogeneous community sample of public (N = 84) and private (N = 143) sector employees, we found no difference between sectors on levels of stress (psychological distress, job satisfaction). Using the Demand-Control-Support (DCS) model to operationalise psychosocial risk and the work stress process we found no difference on levels of risk by sector, however public sector workers reported higher levels of control. For psychological distress we found clear support for a DC interaction effect. Further support for an interactive DCS by sector (4 way) effect was found, with social support operating differently between the sectors. For job satisfaction a DCS main effects model was found in both sectors. The results also showed important differences in stress levels between gender and job categories. The study challenges the commonly held notion that work stress is only a public sector phenomenon.

The aims of this paper are to determine whether differences in levels of work stress actually exist between the public and private sectors, and to identify factors that contribute to levels of work stress in public and private sector workers. This was achieved by embedding the analyses of workplace variables and their contribution to work stress within the theoretical framework of the Demand Control Support (DCS) model of work stress (Johnson, 1986, 1991; Johnson & Hall, 1988; Karasek & Theorell, 1990).

To explicate these aims we provide a brief overview of the public and private sectors in Australia, followed by a discussion of theoretical perspectives for possible differences/similarities in the experience of stress and stress related compensation claims between sectors. Finally, a brief review of the DCS model and some methodological issues raised in the literature are discussed.

Public vs. Private sector

Approximately 20% of Australian employees work in the public sector and 80% in the private sector (Australian Bureau of Statistics (ABS), 1998, 2000). The public sector is predominately concerned with regulating and maintaining social and economic conditions. It embodies a consistent system of rules and regulations determining the nature and structure of work to be done (Fogarty, Machin, Albion, Sutherland, Lalor & Revitt, 1999). In contrast, the private sector is less encumbered by rules, regulations, and restrictive work practices. It is motivated and dependent on making profits, which requires employees to be multi-skilled and creatively resourceful (ABS, 1998; Aungles & Parker, 1992; Martin & Parker, 1997; Pirie, 1988).

Two theoretical viewpoints regarding differences and similarities between the sectors have been distinguished (Scott & Falcone, 1998). The generic viewpoint argues that all organisations are essentially the same with respect to organisational processes such as management, functions, values, and decision making principles. They are driven by decisions that succumb to cost/benefit analysis and competing inputs in one form or another. Increased work pressure due to changing employment patterns fuelled by cost cutting and competition (i.e., downsizing, outsourcing, subcontracting and globalization) increasingly affect both sectors (Koukoulaki, 2002). Although end goals may differ between sectors, the processes of achieving them are fundamentally alike (Choudry, 1989; Lau, Newman & Broedline, 1980; Stillman, 1988).

Thus one would expect to find no differences between the experiences of stress across the two sectors. In contrast, the core viewpoint (DeSantis & Durst, 1996; Freeman, 1996; Mohanty & Mishra, 1998; Zeffane, 1994) argues that fundamental differences (e.g., bureaucratisation and market forces) exist between sectors that have different implications in the general orientation of the workplace including workers' psychological health and well-being (Scott & Falcone, 1998). For example, private sector workers have more access to economic gains, increasing their incentive to be more productive. Furthermore, different legal and political constraints are placed on government rather than private organisations that lead to perceptions of lower job satisfaction and lower job control in the public sector (Scott & Falcone, 1998). Braden and Hyland (1993, p. 14) reported that notable discrepancies between public and private sector employee compensation costs were attributable to “the large variation in the work activities and occupational structures of the two sectors”. Thus from a “core” perspective one should find differences in the experience of work stress across sectors.

Stress compensation claims and job sectors

Work stress results in widespread negative effects for workers, organisations, communities and the economy. Australian occupational health and safety law generally allows that a disability consisting of an illness or disorder of the mind caused by stress is compensable if and only if the stress arising out of the employment was a substantial cause of the disability. The economic cost of such claims is high, estimated at AUD $200 million every year (National Occupational Health and Safety Commission, 2003).

The cost of public and private sector stress claims is often debated in the Australian media reflecting stakeholder positions and concerns. Earlier research by Lewig and Dollard (2001) showed work stress in the Australian newsprint media from 1997 – 98 was represented as an epidemic, as an economic problem and as a public sector phenomenon. There is reason to believe that what is cited in the media influences readers' behaviour. Consequently, more claims could be expected from the public sector. During the period of the study, the public sector claim rates for some Australian states were higher but not all. For example, claim rates for South Australian public sector employees were higher than for the private sector, however time lost and average cost per claim in the private sector exceeded that in the public sector. This circumstance could arise where stress is equally an issue in the private sector but workers delay action until a critical point, perhaps due to less job security in the private sector (Dollard & Walsh, 1999).

In sum, several dilemmas emerge when viewing compensation data as an indicator of actual levels of stress. The popular belief is that stress is a public sector phenomenon. Although the gap is closing, it is true that the number of claims in the public sector is higher. There is some evidence to indicate that the extent of the problem may be underestimated in the private sector using claims data. For both sectors, claims data can be unreliable in the sense that we know that many people who are equally stressed do not submit claims. Therefore, compensation data can significantly understate the number of work related injuries and diseases which actually occur (Worksafe, 1995). A better metric of the problem is to compare work environments directly, and compare levels of stress directly between sectors. The present study will do this using the stress measures of psychological distress and job satisfaction. Furthermore, the DCS model will be used to compare public and private sector work environment dimensions.

Stress and job sectors

In recent times both private sector and public sector workplaces have undergone, and continue to undergo, major restructuring in the form of de-layering, downsizing and task redesign. It has been hypothesised that public sector workers are twice as likely as private sector workers to experience these changes, and report feeling more stressed than their private sector peers when change occurs (Moorehead, Steele, Alexander, Stephen & Duffin, 1997). However, an earlier study by Otto (1980) found that private car plant workers were under greater pressure, encountered more alienating situations, experienced more stress and reported higher frequencies of psychological and physiological symptoms than government plant workers. Furthermore, a large scale Australian Work and Industrial Relations Survey (Mitchell & Mandryk, 1995) reported that improvement of occupational health and safety was ranked last as a reason to implement change in private sector work places.

It is tempting to hypothesise that levels of stress in the public sector are higher than in the private sector, but the empirical evidence is mixed and workers compensation statistics are ambiguous indicators of actual stress levels. The current literature has little specific evidence on the actual comparative levels of stress between the two sectors, or evidence about job design differences by sector that could throw some light on the complex issues raised above. This study was designed to directly compare the generic and core viewpoints of work in the public and private sectors using measures of psychological distress and job satisfaction.

The Demand-Control-Support model

The DCS model predicts that “high-strain work” occurs in working environments combining psychosocial risks of high psychological demands and low control. The model identifies three other types of work including active (high demands/high control), passive (low demands/low control), and low strain (low demands/high control). In all of these conditions social support at work is proposed to further offset the negative effects of high demands. The model proposes that low strain work is associated with better than average health status while active and passive work conditions are associated with an average risk of poor health. Workers are actively motivated and more satisfied with their work in “active” conditions of high demands and high control. However, active jobs may present sustained levels of high demands so that any possible benefits attained from an increased variety of tasks are undone by an increase in intense loads (Dollard & Winefield, 1998; Dollard, 1997; Johnson, 1986, 1991; Johnson & Hall, 1988; Karasek & Theorell, 1990; Vermeulen & Mustard, 2000). The practical implication of the model is that the adverse effects of high levels of demand can be moderated by high levels of control, in the presence of high social support at work (Johnson, 1991; Wall, Jackson, Mullarkey & Parker, 1996).

A recent review of high quality longitudinal research using the DCS model found strong support for a normal causal relationship; main effects for demands, control or support as separate variables. The research also found modest support for the interactive strain model (de Lange, Taris, Kompier, Houtman, & Bongers, 2003). Another study completed reliability and validity tests of the DCS scales and concluded that the data basically supported the assumptions of the model (Pelfrene, Vlerick, Mak, De Smet, Kornitzer & De Backer, 2001). The present study utilised the DCS model of work stress to operationalise core psychosocial risk factors and the process by which the factors combine to predict stress.

Operationalising stress: GHQ-12 and job satisfaction

The GHQ-12 is frequently used in research testing the DCS model as an indicator of stress. High levels of GHQ-12 are associated with high demands, low control and low support (Calnan, Wainwright & Almond, 2000; Dollard & Winefield, 1998). In recent findings, high levels of distress on the GHQ-12 were associated with higher stress claim submissions in a sample of 2112 Australian university staff. Two years later high levels of distress predicted self reported, stress related workers compensation claims (Zadow, Dollard, Boyd & Winefield, forthcoming).

Other research has shown significantly higher levels of psychological distress for those on stress leave. This relationship has been observed in samples of correctional officers (Dollard, 1997), in human service workers (Dollard, Winefield, Winefield & de Jonge, 2000) and longitudinally in nurses (Dollard, 1997). Stress compensation claims were not measured in the current study because the sample size (limited by resources) would not be large enough to predict workers compensation claims, as population claim rates are generally very low. Distress and job satisfaction were included as more accurate indicators of stress. Given the above evidence we can hypothesise that these indicators would predict workers compensation stress claims, or at least reveal the severity of the problem by sector.

Job satisfaction has been used as an indicator of job stress in past research. Job satisfaction is often influenced by factors specific to both public and private work environments including long working hours, working for a large organisation, and lack of promotion opportunities (Clark, 1996). Heslop, Smith, Metcalfe, Macleod, and Hart (2002, p. 1590) define job satisfaction as “the discrepancy between what an individual expects, needs or values about their job compared with how much of this the job actually delivers”. There is some evidence that morale (satisfaction) is more important in organisational and job withdrawal behaviour (e.g., claim submissions) than distress (Dekker & Schaufeli, 1995).

Job satisfaction shows a significant (negative) correlation with psychological distress as evident in a recent study of Victorian police officers (Work & Stress Research Group, 2005). In other Australian research GHQ-12 was found to have a significant negative association with job satisfaction in a study of correctional officers and nurses. Furthermore, job satisfaction correlated negatively to stress leave in nurses cross sectionally and longitudinally (Dollard, 1997). Therefore, job satisfaction was used as another indicator of stress in the present study.

Methodological issues

One of the reasons for the lack of consistent support for the interactive strain hypothesis predicted by the DCS model may relate to the socioeconomic status of workers. Unlike psychological theories, sociological theories of stress, such as Labor Process theory, take account of socio-political and broader class relations to understand factors affecting the experience of work (Peterson, 1999). The notion of control, for example, is extended beyond the immediate work environment to include the socio-political processes in which work is carried out. Many theorists have suggested that participants from low socioeconomic status (SES) living conditions may over-represent job characteristics of high demands, low control, and low support (de Jonge & Kompier, 1997; de Jonge, Breukelen, Landeweerd & Nijhuis, 1999; Schnall, Landsbergis & Baker, 1994; Spector, 1987). It is therefore possible that interaction effects of demand, control, and support and subsequent levels of stress may be explained by SES. This study will assess for the possible influence of socioeconomic conditions, as well as sample within a range of varying socioeconomic conditions.

Braden and Hyland (1993) found that white collar workers (professional/administrative) in the private and public sector had higher compensation costs than blue collar workers (manual/labour work). On a more specific level, compensation for white collar workers was 40% higher in government than in private industry, while the difference for blue collar workers was 14%. Therefore, we will assess differences between white collar and blue collar workers and their influence in the present study.

The existence of curvilinearity in one or more variables can confound the potential of finding linear interactions. Several researchers advocate the importance of controlling for curvilinearity while testing for interactions (Aiken & West, 1991; de Jonge & Kompier, 1997; Dollard, Winefield, Winefield, & de Jonge, 2000; Judd, McClelland & Culhane, 1995; Warr, 1990). However, most studies testing interactions have failed to do this. In two recent studies of DCS interactions curvilinear effects were found to impact on predictions of emotional exhaustion and job satisfaction, although the effect of curvilinearity was minimal (Dollard et al., 2000; Mansell & Brough, 2005).

The aim of this research was to test the DCS model in private and public sector workers to ascertain levels of work stress and factors contributing to work stress while taking account of specific demographic variables and possible curvilinear effects. A central goal of this study was to compare the generic and core viewpoints of public and private sector work environments in regard to their impact on work stress and job satisfaction. In addition, this study was designed to test the predictions of the DCS model in regard to the predictors of psychological distress in the public and private sectors. Specifically it was hypothesised that:

1.	There will be a DC interaction effect in both sectors (hypothesis 1).
2.	There will be a DCS interaction effect in both sectors. More specifically, high levels of control will moderate the relationship between demands and stress measures under conditions of high support (hypothesis 2).

Method

Sample and procedure

We used a random sample based on a range of occupations, stratified by socioeconomic level. Communities were selected using 1996 Australian Census databases to ascertain median weekly incomes. For ease of access, low and medium income areas were chosen. From a total of 362 Adelaide residential areas, low income earners ($300 – 499 pw, coded as low SES) comprised 40% and medium income earners comprised 38% ($500 – 699 pw, coded as medium SES) of the population. Every third household in these areas was approached and one respondent in each household, 18 years or older and employed, was asked to complete the survey before the researcher returned to collect the surveys two days later.

The community-based approach was used to reduce the possibility of demand characteristics in the work environment that may produce biased responses, and to ensure greater access to workers often not afforded in organisational based research. Surveys were delivered at mixed times to maximise responses. From the total pool of 300 surveys distributed 240 participants responded (227 were useable) giving a response rate of 80%. Table I displays the sample characteristics for both private and public sector workers.

Table I. Sample characteristics of public and private sector workers

Characteristics	Public (N = 84)	Private (N = 143)
Gender
Male	43	84
Female	41	59
Age
<30	17%	25%
31 – 50	66%	60%
>51	17%	15%
Median income ($)	625.00	500.00
SES
Low	37%	48%
Medium	63%	52%
Education
Secondary	55%	71%
Undergraduate	25%	18%
Postgraduate	20%	11%
Occupational category
Blue collar	19 (23%)	47 (33%)
White collar	65 (77%)	96 (67%)
Length of employment
5 years or less	56%	61%
>5 years	44%	39%

Independent measures

Demographic variables

Participants were asked to self report gender, age, marital status, number of dependents, status collar (blue collar, white collar), educational qualifications, occupation, weekly income, sector (public vs. private), hours worked, responsibility for other workers (supervision), job duration, and customer (public) contact.

Work environment measures

Variables consistent with the DCS model were used. The Work Environment Scale (WES; Moos, 1994) was used to measure demands, control, and support. The WES subscales are consistent with measures used in previous studies of the DCS model (e.g., Dollard et al., 2000). Each subscale contained nine items with a true/false response format.

The demands subscale measures the degree to which the pressure of work and time urgency dominate the work environment and includes items such as “There is constant pressure to keep working” and “There always seems to be an urgency about everything”. The Cronbach alpha reliability (internal consistency) was .72.

The autonomy (control) subscale measures the extent to which employees are encouraged to be self-sufficient and to make their own decisions. Items include “Staff have a great deal of freedom to decide how they work” and “Supervisors encourage staff to rely on themselves when a problem arises”. Alpha coefficient was .70.

The support subscales measure peer cohesion and supervisor support and include items such as “Staff go out of their way to make a new staff member feel comfortable” and “Supervisors usually compliment staff who do something well”. The alpha coefficient for the combined scale was .77.

Dependent measures

Measures of stress

The 12-item GHQ-12 (Goldberg,1978) was used to measure non-psychotic psychological distress. Although the factor structure of the GHQ-12 has been debated in the literature (Kalliath, O'Driscoll & Brough, 2004) it is a widely used and empirically validated indicator of mental health in occupational studies, particularly in Australian populations (e.g., Banks, Clegg, Jackson, Kemp, Stafford, & Wall, 1980; Cotton, Dollard, de Jonge, & Whetham, 2003; Payne & Morrison, 1999).

The GHQ-12 includes items such as “Have you recently lost much sleep over worry” and “Have you recently felt constantly under strain” on a Likert scale with typical responses: “not at all”, “no more than usual”, “rather more than usual”, and “much more than usual”. Scores range from 0 to 3. The alpha reliability was .91.

Job satisfaction was measured using the Job Satisfaction scale (Warr, Cook & Wall, 1979). The two subscales of extrinsic and intrinsic satisfaction in this measure are often found to be highly correlated (Furnham, Petrides, Jackson & Cotter, 2002), and past research has predominately used the whole scale as a single measure of job satisfaction (Winefield, Gillespie, Stough, Dua & Hapuararchchi, 2002). The scale comprises 15 items which asks participants to rate their level of satisfaction with aspects of their work environment including “physical working conditions”, “rate of pay”, and “job security” on a scale from extremely dissatisfied (1) to extremely satisfied (7). The alpha reliability was .92.

Statistical treatment

Moderated hierarchical regression analyses were used to examine demand, control, and support main effects and their interactions on mental health (GHQ-12) and job satisfaction. This statistical technique is widely used in the work stress literature testing the DCS model (e.g., de Jonge, Dollard, Dormann & Le Blanc, 2000; Dollard & Winefield, 1998; Dollard et al., 2000; Lobban, Husted & Farewell, 1998; Payne & Morrison, 1999). Z scores for the independent variables were calculated (Aiken & West, 1991) and interactions (cross-product of terms) were computed by multiplying the main effects (e.g., demands × control).

The hierarchical arrangement was constructed to first test interactions after accounting for main effects and to give every possibility of finding an effect. In the first step of the analysis demands, control, and support were entered to test the original model. This was followed by sector, coded as 0 = public sector, and 1 = private sector. The interactions terms; demand × control (D × C), demand × support (D × S), control × support (C × S), demand × control × support (D × C × S), demand × control × support × sector (D × C × S × Sec) were entered at the third step. Finally, possible confounds were entered to ascertain their impact on the effects noted. Demographics, age, gender, educational level, SES, and collar (blue/white collar occupation) were entered at the fourth step.

Independent samples t-tests were performed on work environment and dependent variable measures to assess differences between public and private sector workers. The rejection rate for all analyses was set at p = .05.

Results

Means, standard deviations and Pearson correlations

Table II shows the means and standard deviations of the work environment measures, and dependent measures for public and private sector workers. Results of the independent t-test using the Levene test for equality of variances revealed no significant differences between sectors on levels of psychological well-being and job satisfaction, thus providing support to the generic viewpoint. Public sector workers reported having significantly more control than private sector workers, t(225) = 2.30, p < .05. There were no other significant differences between sectors on work environment variables providing partial support for the generic viewpoint.

Table II. Means, standard deviations, and t-tests results of work environment, and dependent measures for public and private sector workers

	Private Sector (N = 143)		Public Sector (N = 84)
Measure	M	SD	M	SD	t
Job Demands	6.09	2.25	6.10	2.40	.03
Job Control	5.62	2.34	6.33	2.03	2.30*
Social Support	9.87	3.95	10.34	3.81	.88
Peer Support	5.27	2.13	5.40	1.92	.44
Supervisor Support	4.59	2.35	4.94	2.45	1.05
Job Satisfaction	69.93	17.24	71.05	14.88	.52
Psychological Distress	12.25	5.76	11.75	6.28	.59
*p < .05.

Pearson's correlations are presented in Table III. SES determined by living location (based on income) was not correlated with any outcome variables and therefore was omitted from hierarchical regression analyses. Taking both samples together, collar was correlated with income r(227) = .22, p < .01 and SES r(227) = .30, p < .01 and the outcome variables; psychological distress r(227) = .14, p < .05 and job satisfaction r(227) = .16, p < .05, and was included in subsequent hierarchical regression analyses as an SES related measure.

Table III. Pearson intercorrelations for private (above diagonal) and public sector workers

Variable	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21
1. Age		.17*	−.41***	.19*	.20	−.06	−.04	−.09	.18*	.20*	.07	.30***	−.12	−.03	.10	.14	.14	.23**	.29***	.16	−.08
2. Gender	.05		−.16	.21*	.38***	−.12	.12	−.04	−.13	.48***	.31***	−.11	−.23**	−.28**	−.16	−.03	−.11	.18*	.07	.01	−.14
3. Marital status	−.14	−.27*		−.36***	−.28***	.31***	.01	.04	−.07	−.25**	−.23**	−.24**	.13	−.06	−.09	−.22**	−.18*	−.13	−.20*	−.11	.12
4. Dependants	.09	.13	−.25*		.21*	−.18	.00	.01	−.01	.19*	.36***	.09	−.02	.05	−.12	−.02	−.08	.15	.02	−.01	.05
5. Income	.15	.38***	−.42***	.22*		−.33***	−.17*	.10	.12	.67***	.42***	.22**	.01	.17	.00	.07	.04	.16*	.13	.14	−.05
6. SES	.12	−.19	.17	−.30**	−.23*		.02	.05	−.10	−.32***	−.33***	.02	.03	−.27**	−.09	−.20*	−.17*	−.06	−.25	−.15	.04
7. Secondary	−.18	−.07	.20	−.21	−.51***	.10		−.74***	−.54***	−.08	−.09	−.03	−.19*	−.28**	−.20*	.11	−.17*	−.14	−.14	−.12	.01
8. Graduate	.11	.12	−.12	.18	.26*	.01	−.63***		−.16	.08	.08	−.13	.21*	.23**	.11	.09	.12	.06	.08	.17*	.04
9. Postgraduate	.09	−.04	−.12	.06	.35**	−.14	−.55***	−.29**		.01	.02	.21*	.02	.39**	.14	.04	.10	.12	.10	−.03	−.07
10. Hours	.27*	.34**	−.22*	.07	.46***	−.18	−.40***	.30**	.17		.30***	.18*	−.03	.09	−.05	−.02	−.04	.36***	.22**	.06	.03
11. Responsibility	.07	.28**	−.25*	.17	.26*	−.09	−.22*	.22*	.04	.22*		.17*	−.05	.30***	−.08	.07	−.00	.12	.17*	.20*	.00
12. Job exposure	.36**	−.05	−.06	−.16	−.11	−.04	.10	−.09	−.02	.18	−.13		.02	.09	.08	.03	.06	.05	.10	.14	−.10
13. Customer contact	−.05	.11	.06	.07	−.03	−.26*	.03	.05	−.10	.05	.18	.02		.24**	.14	.05	.10	.11	.11	.20*	−.01
14. Collar	−.07	.15	−.28**	.17	.31**	−.29**	−.18	.31**	.45***	.26*	.17	−.11	.08		.15	.06	.12	.11	.17*	.11	.23**
15. Peer cohesion	−.06	−.12	.05	−.22*	−.13	.08	.01	−.09	.07	.04	.09	.19	−.30**	−.01		.55***	.86***	−.19*	.38***	.36***	−.14
16. Supervisor support	−.02	.13	−.07	−.01	.14	−.06	−.13	.17	−.02	.06	.11	.14	−.20	.05	.51***		.89***	−.19*	.53***	.47***	−.22**
17. Social support	−.04	.02	−.01	−.12	.02	.00	−.07	.06	.02	.06	.12	.19	−.28**	.02	.83***	.90***		−.22**	.52***	.47***	−.21*
18. Demands	.08	.10	.04	.10	.22*	−.29**	−.16	.20	−.01	.29*	.08	−.00	.09	.22*	−.13	−.18	−.19		.04	−.12	.20*
19. Control	.07	.25*	−.19	.06	.31**	−.12	−.14	.15	.02	.24*	.19	−.03	−.08	.14	.28**	.59***	.52***	.01		.52***	−.13
20. Job satisfaction	−.02	−.12	−.12	−.00	.09	−.01	−.20	.16	.07	.00	.11	−.03	−.07	.21	.31**	.47***	.46***	−.15	.39***		−.30***
21. Psych. distress	.25*	−.13	.06	.09	−.13	−.15	.05	.01	−.08	.07	−.11	.26*	.14	.10	−.25*	−.21	−.26*	.35**	−.17	−.39***
*p < .05, **p < .01, ***p < .001.

Regression analyses of the DCS model

Psychological distress

Hierarchical regression analysis revealed significant main effects for control and a significant interaction effect for demands and control with respect to psychological distress (see Table IV for Beta values, and significance). This provides support to hypothesis 1.

Table IV. Hierarchical regression analysis predicting psychological distress

Variables	Additive Model (Work environment) (Step 1)	Additive Model (Sector) (Step 2)	Interactive Model (Step 3)	Demographics (Step 4)
Standardised Regression Coefficients (Beta)
Work Environment
# Demands (D)	.24***	.24***	.30***	.33***
Control (C)	−.07	−.09	−.13	−.11
Support (S)	−.12	−.12	−.12	−.13
# Sector (Sc)		−.09	−.07	−.07
# Interactions
D×C			−.21*	−.18*
D×S			.20*	.14
C×S			.04	.04
D×C×S			.06	.06
D×C×S×Sc			−.26*	−.28*
# Demographics
Gender				−.16*
Δ R^2	.01***	.01	.06*	.05
Adjusted R^2	.08	.09	.12	.16
# Represents a new step in the regression.

The analysis reveals a significant four-way interaction effect of demands, control, support and sector. In particular the DC interaction held under conditions of high support in the private sector and low support in the public sector. The adjusted R square for the DCS model was .16. These results show mixed support for the second hypothesis.

An interaction effect was also found between demands and support, however this lost significance when demographics (age, gender, education level, SES, and status collar) were entered in the final step. The significance of gender at the final step indicates that the interaction between demands and support may be mediated by gender. For instance we found that women reported significantly higher demands t(225) = −2.31, p < .05 than men.

Job satisfaction

Hierarchical regression analysis revealed significant main effects for demands, control and support with respect to job satisfaction lending support to a DCS main effects model as no significant interaction effects were noted (see Table V). No main or interaction effects were noted for this sector. The addition of demographics to the model at the final step showed that being a white collar worker was associated with higher levels of job satisfaction. The adjusted R square for the DCS model was .47.

Table V. Hierarchical regression analysis predicting job satisfaction

Variables	Additive Model (Work environment) (Step 1)	Additive Model (Sector) (Step 2)	Interactive Model (Step 3)	Demographics (Step 4)
Standardised Regression Coefficients (Beta)
Work Environment
# Demands (D)	−.18**	−.18**	−.26***	−.30***
Control (C)	.29***	.30***	.32***	.30***
Support (S)	.43***	.43***	.43***	.43***
# Sector (Sc)		.04	.04	.05
# Interactions
D × C			.05	.06
D × S			−.02	−.04
C × S			−.04	−.04
D × C × S			.13	.14
D × C × S × Sc			.02	.01
# Demographics
Collar				.12*
Δ R^2	.46***	.00	.01	.02
Adjusted R^2	.46	.46	.46	.47
# Represents a new step in the regression.

The relationship of collar to job satisfaction was further investigated and showed that white collar workers are involved in active jobs, reporting significantly higher demands t(225) = −2.23, p < .05 and significantly higher control t(225) = −2.62, p < .05 than blue collar workers.

Interpreting DCS and sector interactions

Based on criteria recommended by Aiken and West (1991) and Parkes (1990) for plotting interactions, graphical representations are illustrated in Figures 1 and 2 to enable interpretation of the interactive effects for private sector workers. All variables not involved in the interaction were held constant by multiplying their unstandardised coefficients with their mean score (0). Conditions of the variables involved in the interaction were formulated for low (−1 SD) and high (+1 SD) conditions.

Figure 1. Interaction between demands and control

Figure 1 illustrates the demand and control interaction and that the effect of demands on levels of psychological distress is conditional upon levels of job control. In conditions of high demands, psychological distress decreases as levels of control increase (hypothesis 1).

Figures 2 and 3 show the four way interaction effect of demands and control under conditions of low and high support in the public sector. In accordance with the DCS model, under conditions of low support, as levels of control increase the relationship between demands and psychological distress is reduced (see Figure 2). Under conditions of high support, the relationship between demands and distress is reduced at low levels of control. However, the relationship between demands and distress increases when control is high (Figure 3).

Figure 2. Interaction between demands and control under conditions of low support (public sector)

Figure 3. Interaction between demands and control under conditions of high support (public sector)

Figures 4 and 5 show the interaction effect of job demands and control on psychological distress in the private sector under conditions of high and low support. Under conditions of high support, an increase in demands is paralleled by an increase in levels of distress under conditions of low control. Under high control the relationship is significantly reduced (see Figure 4). When support is low, there is a positive correlation between demands and distress at high and low levels of control. However, this positive correlation is stronger at high levels of control (see Figure 5).

Figure 4. Interaction between demands and control under conditions of high support (private sector)

Figure 5. Interaction between demands and control under conditions of low support (private sector)

The results reveal a classic demand-control interaction at high levels of support in the private sector and support for the second hypothesis, but at low levels of support in the public sector. This latter result is somewhat consistent with the iso-strain hypothesis (Karasek & Theorell, 1990), high levels of stress shown in jobs combining high demands, low control and low support. In sum the classic demand-control interaction is noted in either high or low work support conditions depending on the sector.

Testing for curvilinear effects

Further analyses were conducted given previous findings of a curvilinear relationship between job demands and control in predicting employee anxiety levels. However, some contention exists with regard to whether quadratic or linear interactions should be included in the final step of the regression model. We entered variables in the following order: linear main effects, quadratic main effects (the quadratic terms calculated by squaring the appropriate continuous variables, Mansell & Brough, 2005), linear interactions, and demographics based on recommendations from Dollard et al. (2000) and Judd et al. (1995) for predicting mental health. Results showed that when controlling for curvilinearity, linear interactions and main effects remained the same as reported in Tables IV and V for both psychological well-being and job satisfaction. However, the interaction effect of demands and support on psychological well-being was no longer significant. Similar to the study by Mansell and Brough (2005), this study shows little evidence for the contention that job characteristics may show curvilinear relationships with psychological well-being.

Discussion

In support of the generic viewpoint which argues that all organisations, public or private, are essentially the same (Choudry, 1989; Lau, Newman & Broedline, 1980; Stillman, 1988), the study found no evidence to suggest that the experience of work stress is more prevalent in the public sector. This finding is interesting given that public sector workers reported significantly more autonomy over their work, a variable that is considered to be a principal moderator of work stress according to the DCS model (i.e., control may have reduced the impact of demands on distress). Furthermore, the results also show the work stress processes between sectors vary, in other words the way DCS elements combine together varies seemingly because support plays a different role between sectors.

Psychological distress

In relation to psychological distress, the DCS model explained 16% of the variance in public and private sector workers. The regression also revealed a significant interaction effect for demand and control. In addition, the results indicated that psychological distress may be moderated by increasing worker control under conditions of high support (de Jonge, Mulder, & Nijhuis, 1999).

However, the four way interaction effect revealed a different picture between sectors. For private sector workers with low levels of control, the presence of high support appears to exacerbate the effect of high demands on levels of psychological distress. Conversely, the presence of high support reduces the impact of high demands on psychological well-being at high levels of control. This result supports the second hypothesis (similar to Dollard & Winefield, 1998, and Landsbergis, Schnall, Deitz, Friedman, & Pickering, 1992). This finding also supports Schaubroeck and Fink's (1998) argument that psychological distress from failing to cope with high demands may occur if either support or control are high and the other is low.

Different results were found for public sector workers. The second hypothesis was not supported as when support was high, high control seemed to exacerbate the positive relationship between demands and distress. This result is similar to Rodriguez, Bravo, Peiró and Schaufeli's (2001) finding that high levels of these factors caused higher job dissatisfaction in administration workers. The presence of high support is ordinarily expected to offset the negative effects of high demands (Dollard & Winefield, 1998; Dollard, 1997; Johnson, 1986, 1991; Johnson & Hall, 1988; Karasek & Theorell, 1990; Vermeulen & Mustard, 2000). On the other hand, under conditions of low support the classic demand-control interaction was observed. This is somewhat consistent with the iso-strain hypothesis (Karasek & Theorell, 1990); workers who experience conditions of high demands, low control and low support also show the highest levels of distress.

Job satisfaction

In relation to job satisfaction, the DCS model accounted for 47% of the variance. Specifically, main effects were found for demands, control and support, however no interaction effects were noted. The lack of significance for a correlation between sector and job satisfaction supports the generic viewpoint. The main effects of the present study follow the conclusions of past research (DCS main effects model) that people with high job satisfaction have low demands, high control and high support at work (Dollard & Winefield, 1998) rather than “active jobs”. In addition, being a white collar worker also contributes to levels of job satisfaction. T-tests revealed that white collar workers are more likely to experience “active” job conditions. This finding is in keeping with the notion posited by sociological theories that blue collar workers are less likely to derive a sense of mastery and control from their work than white collar workers (Peterson, 1999). Although SES showed no significant contribution, the significant result for status collar shows some small support for the role of SES related variables in the work stress process.

Work stress and stress compensation claims

Work stress compensation rates indicate that the public sector should theoretically show higher rates of distress and lower rates of job satisfaction. However the study raises questions about the relationship between actual work stress and stress compensation claims. Using better indicators of job stress the study found no difference in levels of job satisfaction and psychological distress between sectors. This discrepancy points to the possibility of other factor(s) influencing work stress compensation claims (e.g., media, sociopolitical factors, stigma). We only found one clear difference between sectors (autonomy) that could account for variation in claim levels, and this would predict higher distress in the private sector.

Even though we found that the way the DCS elements combined varied between sectors for psychological distress (but not for job satisfaction), this can not plausibly account for stress differences between sectors. The question of whether workplace variables such as demand, control and support play a larger role in stress compensation claims than personal levels of distress and satisfaction could be explored in future research.

Methodological considerations

The present study has a number of limitations. Firstly, the sample was relatively small in size (N = 227). However, the demographics of the workers in the sample were generally representative of the broader South Australian and Australian public and private sector populations in terms of gender, income, education, and occupational category (ABS, 2000; Commissioner for Public Employment, 2003). This increases the generalisability and veracity of the results. However public sector employees were overrepresented in this sample (36%) when compared to the percentage of public sector workers both in South Australia (12.4%) and nationally (20%), but this in itself would not have impacted on the results. It should be acknowledged that the self reporting of public/private sector status may have lead to some inaccuracy in categorisation. However there were no demand characteristics of the research to influence participants either way. Still a methodological improvement could be to use more “objective” indicators gather this information (e.g., organisation name; funding source; government/NGO/private status of employer). It is also possible that the differences observed between sectors may have reflected differences in the types of jobs held by respondents in each sector, rather than differences between the sectors themselves. The use of a random sample across a range of occupations would have guarded against this possibility to some extent. Nevertheless, collection of larger samples that allow stratification by occupation should be a priority for future research.

Second, the study did not measure actual rates of stress compensation claims within the sample of workers. The conclusions drawn in this study around work stress and compensation claims would be strengthened by measuring actual compensation rates in a larger sample of the same occupation types across job sectors.

Third, the DCS model was only able to explain 16% of the variance in public and private sector workers in relation to psychological distress. However, Rodriguez et al. (2001, p. 109) argue that although an amount of variance might be small, “this does not deny the theoretical relevance of the results and it does not mean the interaction effect has little substantive significance. The size of the effect is attenuated by measurement error and this is greatly exacerbated when variables are multiplied together from a cross product term as required to test interactions in regression analysis”.

Fourth, common method variance is widely noted as a problem for self report and cross sectional research. It is particularly problematic when independent measures induce affective responses and are then correlated with predominantly affective outcomes (i.e., psychological distress). However, the present study selected work environment scales that did not elicit explicit evaluative affect. Nevertheless the implications are the possibility of inflated main effects increasing the possibility of a Type II error that interaction effects do exist but were not found.

Finally, only job characteristics at a general level (DSC) were used in the present study. As mentioned earlier public sector work may be intrinsically more stressful because of, for example exposure to critical incidents in correctional work or police work. Although these aspects were not explored, research in these agencies tends to suggest the more routine organisational aspects of the job are more stressful than these “core” aspects (Dollard, 1997; Hart & Cotton, 2003).

Implications

The results of the present study highlight the role of support in the satisfaction and well-being of workers. By increasing support for workers in the presence of high control in the private sector, stress that is related to high demands at work can be moderated. In the private sector, it is also important to match levels of support and control to maximise an individual's resilience to the deleterious effects of work demands. Interventions focusing on these factors would have positive benefits for individuals and organisations through increasing well-being, subsequently reducing stress and stress compensation claims. However, the results showed that overly high support at work with high control and high demands in the public sector can also cause distress, possibly through an increase in responsibility and work load. Therefore, alternative stress interventions apart from increased control and support need to be explored and implemented. For example, the Effort Reward Imbalance model of work stress would recommend increasing rewards to offset demands (Siegrist, 1996).

The finding that white collar workers experience higher job satisfaction than blue collar workers suggests that further research needs to explore the differences between these occupational categories. Evidence from this and past studies (Peterson, 1999) indicates that this difference could be due to blue collar workers experiencing less control at work than white collar workers. Interventions focusing on control and autonomy could be used to increase job satisfaction for blue collar workers.

The results of this study and past studies have important implications for women at work. The finding that women experience more distress and higher demands at work than men is consistent with previous findings and compensation statistics (Karasek & Theorell, 1990; Pugliesi, 1995; Workcover NSW, 1998). This suggests that work stress interventions such as reducing/off-setting demands (see Muhonen & Torkelson, 2003), need to be improved and tailored for women.

Conclusion

This study sought to understand the factors underlying the levels of stress, and differential patterns of psychosocial job characteristics in the private and public sector that could lead to distress. Consistent with the generic viewpoint, the results indicated no difference in distress or satisfaction between sectors, nor on levels of psychosocial factors (except autonomy). This challenges the everyday notion that work stress is a public sector issue and reveals new possibilities for action in the private sector.

From a theoretical perspective, the results show that it is not simply a matter of correlations between job characteristics that explain the experience of distress between sectors. The present study shows that complex patterns of interactions between sectors need to be considered when contemplating causes of stress compensation claims and intervention for distress. However, the main effects of demands, control, and support operate similarly between the sectors for job satisfaction. Additionally certain demographics are important to consider: women are more likely to experience distress, and white collar workers more likely to experience job satisfaction. Future research should explore the role of support in work stress interventions and the causes of increased stress compensation rates.

Acknowledgments

We are grateful to Kerry Lewig, Research Assistant and Natalie Skinner, Research Associate of the Work and Stress Research Group, University of South Australia.