Personality and Individual Differences 35 (2003) 537-557

Personality and Individual Differences 35 (2003) 537–557

quality of present-day working life, it is important to find individual stress resistance resources which modify stress-health relationships.

In recent stress research, it has become clear that the tendency to perceive job circumstances as stressful depends in part upon the characteristics of the individual. Individuals exposed to the same environmental conditions may express remarkably different psychological, physical, and behavioral reactions on account of different personality characteristics (see Kahn & Byosiere, 1992). In particular, self-esteem has been the most extensively investigated personal resource in the work context (Brockner, 1988; Locke, McClear, & Knight, 1996; Tharenou, 1979).

The present study contributes to the stress-buffering literature by analysing whether self-esteem and optimism modify the impact of harmful psychosocial work stressors on well-being. Specifically, the aim was to explore the main and interactive effects of psychosocial work stressors (i.e. time pressures at work, lack of control, job insecurity, and poor organizational climate), self-esteem and optimism on employee well-being (i.e. job satisfaction, emotional exhaustion, mental distress and physical symptoms) for a sample of Finnish male and female employees in a 1-year follow-up study.

In stress literature, personality has been considered to affect the stress process in five different ways (Cox & Ferguson, 1991; Kivimaki, 1996). First, personality may influence stress reactions by modifying the appraisal of stressors. For example, a person with high self-esteem and opti­mism is more likely to view a stressful work situation as challenging rather than threatening. Optimism, which is defined as a generalized expectation of positive experiences and outcomes throughout one’s life (Scheier, Carver, & Bridges, 2001), makes a person apprise a stress situation in a more positive light. Optimists make more extensive use of a variety of coping strategies and have better physical and psychological health than pessimists (Carver & Scheier, 1999; Chang & Farrehi, 2001; Scheier & Carver, 1992; Scheier et al., 2001). By contrast, individuals lacking in optimism and self-esteem have been shown to experience greater negative stress and to use more withdrawal and passive forms of coping to manage stressful events (Aspinwall & Taylor, 1992; Harju & Bolen, 1998; Scheier et al., 2001; Scheier, Weintraub, & Carver, 1986).

Second, personality factors may effect stress reactions independently of stressors. Studies have found that self-esteem, which refers to an individual’s general sense of his or her value or worth (Rosenberg, 1979; Locke et al., 1996), is negatively associated with all three burnout dimensions (e.g. exhaustion, depersonalization and reduced personal accomplishment), but especially with emotional exhaustion (Carmel, 1997; Fothergill, Edwards, Hannigan, Burnard, & Coyle, 2000; Janssen, Schaufeli, & Houkes, 1999; Kinnunen, Mauno, Natti, & Happonen, 1999; Rosse, Boss, Johnson, & Crown, 1991). Furthermore, self-esteem is shown to be associated negatively with depressive symptoms (Schonfeld, 2000) and positively with job satisfaction (Abraham, 1999; Carmel, 1997; Frone, 2000; Saks & Ashforth, 1997).

Third, work stressors and stress reactions may modify personal factors. For example, it has been found that psychosocial work stressors and burnout reduce self-esteem (Golembiewski & Aldinger, 1994; Kivimaki & Kalimo, 1996). Fourth, work stressors may cause differences in per­sonality factors, which may mediate the quantity and the quality of stress reactions. Self-esteem has been found to mediate the effects between unemployment and psychological health (Kokko & Pulkkinen, 1997; Pearlin, Lieberman, Menaghan, & Mullan, 1981) and between stress and expectancy of success (Abel, 1996).

Fifth, personality factors may influence stress reactions by moderating the relationship between stressors and stress reactions. Conceptually, a moderator is a variable which alters the direction

A. Ma¨kikangas, U. Kinnunen /Personality and Individual Differences 35 (2003)537–557 539

or strength of the relationship between two other variables (see Baron & Kenny, 1986; Parkes, 1994). Empirical research has provided some support for the moderating effects of self-esteem, although the findings have been inconsistent (see Cohen & Edwards, 1989; Locke et al., 1996). The theoretical viewpoint for studying self-esteem as a moderator has its roots in Brockner’s (1983, 1988) plasticity hypothesis. Brockner (1983, 1988) suggested that low self-esteem indivi­duals, i.e. lacking self-confidence and certainty of one’s own beliefs and behaviors, are generally more susceptible to environmental events than those with high self-esteem. Thus, low self-esteem employees are more prone to regard social cues as guides for appropriate action and more dependent on others’ evaluations than high self-esteem employees. A number of studies have supported Brockner’s presumptions and found that low self-esteem employees are more strongly influenced by role conflict, role ambiguity, role overload, peer group interaction and supervisory support than those of high self-esteem employees (Elangovan & Xie, 1999; Fernandez, Mutran, & Reitzes, 1998; Ganster & Schaubroeck, 1991; Jex & Elacqua, 1999; Mossholder, Bedeian, & Armenakis, 1981, 1982; Pierce, Gardner, Dunham, & Cummings, 1993; Wiener, Muczyk, & Martin, 1992).

Optimism has been found to moderate the relationship between daily hassles and health out­comes (e.g. symptoms of physical illness, feelings of exhaustion, burnout and loss of self-esteem; Fry, 1995), hassles and physiological symptoms (Lai, 1996), perceived stress and depression (Sumi, Horie, & Hayakawa, 1997). Furthermore, three-way interactions have been found between ratings for optimism, social support, and stress on physical and psychological well-being (Sumi, 1997). Thus, individuals who reported higher optimism and social support tended to report better well-being, regardless of their reported stress.

Despite the growing interest in stress moderators, the gender differences have remained under-researched. So far, almost all existing moderator studies have been cross-sectional and have concentrated solely on either men or women and used samples of certain occupational groups or university students (see e.g. Fry, 1995; Ganster & Schaubroeck, 1991; Pierce et al., 1993; Sumi, 1997; Sumi et al., 1997; Wiener et al., 1992) Therefore, knowledge about their actual gender dif­ferences is lacking. Gender specificity in moderator studies should be considered significant, because the same moderator variables may have different impact on males and females (Sherman & Walls, 1995).

In this study, well-being was viewed from positive (job satisfaction) and negative (emotional exhaustion, mental distress and physical symptoms) standpoints. There is a large body of empirical evidence to suggest that psychosocial work stressors are likely to increase strain and impair well-being at work. It has been shown that perceived workload and time pressures are strongly and consistently related to burnout (Abel & Sewell, 1999; Houkes, Janssen, de Jonge, & Nijhuis, 2001; Lee & Ashforth, 1996). In addition, lack of social support at work has been found to increase levels of burnout (Houkes et al., 2001; Peeters & Le Blanc, 2001) and depression (Frone, 2000). On the other hand, good social relations can be a source of positive personal outcomes, i.e. job satisfaction (Moyle, 1998; Sargent & Terry, 2000). Control at work (e.g. the extent to which employees are able to make their own decisions about work) also increases the level of job satisfaction (Carayon & Zijlstra, 1999; Jimmieson, 2000) and decreases the level of burnout (see Burke & Richardsen, 2001). Job insecurity is associated with emotional exhaustion (Mauno & Kinnunen, 1999), poor physical and mental health, low job satisfaction and higher levels of turnout intentions (Hellgren, Sverke, & Isaksson, 1999).

540 A. Ma¨kikangas, U. Kinnunen /Personality and Individual Differences 35 (2003)537–557

Taken as a whole, the present study focuses on the possible moderator effects of self-esteem and optimism between several work stressors and well-being at work in a 1-year follow-up study. Many of the previous stress moderator studies have been cross-sectional with small samples and, in addition, few conclusions can be drawn regarding their associations with gender differences. Consequently, in contrast with earlier studies, the present study has unique features that may provide valuable additional insights into the relationship between work stressors and stress reac­tions. For one, it is based on longitudinal data which enables a better analysis of cause–effect relations, and secondly it takes gender differences into account.

The aim of the present study was to investigate the main effects of psychosocial work stressors and personality characteristics and the moderating effects of these variables at Time 1 on well-being indicators at Time 2, after controlling for the effect of each well-being indicator at Time 1. We hypothesized that, compared with employees with high self-esteem and optimism, those with low self-esteem and optimism at Time 1 would experience higher levels of emotional exhaustion, mental distress and more physical symptoms as well as reduced levels of job satisfaction at Time 2 when work stressors are high at Time 1.

1. Method

1.1. Participants and procedure

The data for the study were obtained as part of a research project entitled ‘‘Economic Crisis, Job Insecurity and the Household’’. The random sample (n=1878) was selected from the files of the Finnish Population Register Center in 1999, and restricted to working aged people between 25 and 59 years. Two identical postal questionnaires were sent to the sample in the spring of 1999. One questionnaire was intended for the target individual and the other, when relevant, for his or her partner. In 2000 the questionnaires were sent, using the same procedure, to those individuals who had answered in 1999. Responses were received in 1999 from 851 target persons and 608 spouses/partners, and in 2000 from 655 target persons and 468 spouses/partners.

Because of the relatively low response rate in 1999 (45%), we compared the demographic characteristics of the initial and final sample in both years. Analysis of respondents versus non-respondents in 1999 revealed no significant differences with respect to gender, age, marital status and geographical location. However, in 2000 there were some differences concerning participants who were somewhat older and better educated (for more details, see Kinnunen, Natti, Happonen, Kalliolahti, Kelhala, & Mauno, 2000).

In the present study, data analysis was restricted to those who responded in both years (n=640) and to those who were employed in both years (n=457). Of the employed respondents, 225 were female and 232 were male and the majority of them were between 35 and 43 years old. Table 1 summarizes the demographic characteristics for the male and the female respondents separately. There were some significant differences between the sexes. The most notable was that nearly 60% of the men in the sample worked for a private employer, whereas over 40% of the women worked in the municipality sector. In addition, men were more often than women manual workers (42%) when women mostly were employed in lower non-manual occupations (e.g. nurse or secretary) (47%).

A. Ma¨kikangas, U. Kinnunen /Personality and Individual Differences 35 (2003)537–557

541

1.2. Measures

All the composite variables were created by averaging their respective items, and were scored so that a high score represents a higher level of the construct. The reliabilities (Cronbach alphas) turned out to be acceptable, so all the scales used in the present research were considered to have an adequate internal consistency. The original measures were translated from English into Fin­nish if Finnish versions did not already exist. They were checked by a professional translator.

1.3. Work stressors as predictor variables at Time 1

Time pressures at work were measured by the Quantitative Workload Inventory (Spector & Jex, 1998). This consisted of five items which assessed the amount or quantity of work in a job (e.g. ‘‘How often do you have to do more work than you can do well?’’; ‘‘How often does your job require you to work very fast?’’). The subjects responded on a five-point scale (1=not at all, 5=very often). Cronbach’s alphas were 0.80 and 0.84 for the men’s and women’s reports, respectively.

Lack of control was assessed through the scale derived by Jackson, Wall, Martin, and Davids (1993). However, of the original five scales reflecting the different dimensions of job control, we measured only two, which were control over timing (four items, e.g. ‘‘Do you set your own pace at work?’’) and over method (four items, e.g., ‘‘Can you choose the methods to use in carrying out your work?’’). Accordingly, the respondents rated the degree of control for each of these items on a five-point scale (1=not at all, 5=a great deal). In the present study, the eight items yielded a Cronbach alpha of 0.92 for the men and 0.93 for the women.

Job insecurity measure consisted of four items which assessed uncertainty of job continuity (e.g. ‘‘How certain are you about your job security in this company?’’ (Caplan, Cobb, French, van Harrison, & Pinneau, 1980)). The items were rated on a five-point scale (1=very certain, 5=very uncertain). Cronbach’s alphas for job insecurity were 0.77 for the men and 0.76 for the women.

Poor organizational climate was measured by a 10-item scale that concerned social support, feedback and frankness from supervisors (e.g. ‘‘I can get help and support from my supervisor’’) and co-workers (e.g. ‘‘People in this organization tend to be cool and aloof towards each other’’), and warmth of organizational climate (e.g. ‘‘This organization is characterized by a relaxed, easy­going working climate’’). The scale was a modification of items based on the study ‘Quality of Working Life in Finland 1977–1997’ (Lehto & Sutela, 1998; see Kinnunen & Natti, 1994; Litwin & Stringer, 1968). The alphas for the scale were 0.85 for the men and 0.87 for the women.

In addition to original articles, see for example, Dallner et al. (2000) and Mauno, Leskinen, and Kinnunen (2001) for validity-evidence.

1.4. Personality characteristics as moderator variables at Time 1

Self-esteem was measured by Rosenberg’s (1965) 10-item scale. This scale is a self-report measure of generalized feelings about the self. The self-esteem items (e.g. ‘‘I feel I have a number of good qualities’’; ‘‘At times, I think I am no good at all’’) were rated on a five-point Likert scale (1=totally agree, 5=totally disagree). The Cronbach alpha coefficients for the scale were 0.87 for the men and 0.88 for the women. Although the self-esteem scale is widely used there is only little

A. Ma¨kikangas, U. Kinnunen /Personality and Individual Differences 35 (2003)537–557 543

data available on its psychometric properties (see e.g. Bosson, Swann, & Pennebaker, 2000; Pullman & Allik, 2000).

Optimism was assessed through an abbreviated version of ‘‘The Revised Life Orientation Test’’ developed by Scheier, Carver, and Bridges (1994), for which also validity evidence exists. It con­sisted of six items (e.g. ‘‘If something can go wrong for me, it will’’; ‘‘In uncertain times, I usually expect the best’’). These statements were rated on a five-point Likert scale (1=totally agree, 5=totally disagree). The alphas for the scale were 0.73 and 0.72 for the men and the women, respectively. The findings of Scheier et al’s (1994) study supported the discriminant validity of the LOT-R(see also Burke, Joyner, Czech, & Wilson, 2000), although there has been discussion about the possible overlap between LOT (especially pessimism dimension) and neuroticism (first discussed by Smith, Pope, Rhodewalt, & Poulton, 1989).

1.5. Well-being indicators as control variables at time 1 and outcome variables at Time 2

Job satisfaction was measured with a two-item, five-point scale based on Hackman and Old-ham’s (1980) Job Diagnostic Survey. The items measured general job satisfaction (‘‘Generally speaking, I am very satisfied with this job’’ and ‘‘I am generally satisfied with the kind of work I do in this job’’). The Cronbach alphas for the scale at Time 1 were 0.85 and those at Time 2 were 0.87 for both sexes. For stability and validity-evidence of job satisfaction, see Dormann and Zapf (2001) and Renn and Swiercz (1993).

Emotional exhaustion at work was assessed by four items from the Maslach Burnout Inventory (Maslach, Jackson, & Leiter, 1996) capturing feelings of fatigue that develop as one’s emotional energies become drained at work (e.g. ‘‘I feel emotionally drained from my work’’). The subjects responded with five response alternatives (1=never, 5=always). The alphas for the scale at Time 1 were 0.88 for the men and 0.90 for the women and those at Time 2 were 0.88 for the men and 0.89 for the women. Studies employing different versions of the MBI have found good support for different forms of validity (see e.g. Maslach et al., 1996; Schaufeli, Bakker, Hoogduin, Schaap, & Kladler, 2001).

Mental distress was assessed through the General Health Questionnaire (Goldberg, 1972; Goldberg et al., 1997). The original GHQ-12 was designed to identify short-term changes in mental health (e.g. ‘‘Have you recently felt constantly under strain?; ‘‘Have you recently been feeling unhappy and depressed?’’). However, the original GHQ-12 contains two items which are related to self-esteem (‘‘Have you recently been thinking of yourself as a worthwhile person?; ‘‘Have you recently lost your self-confidence?’’) so we used only 10 items to avoid confusion of concepts. The Cronbach’s alpha coefficients for the 10-item mental distress scale at Time 1 were 0.89 for both the men and the women and those at Time 2 were 0.89 for the men and 0.87 for the women. For test–retest reliability and validity evidence, see Pevalin (2000) and Goldberg et al. (1997).

Physical symptoms were assessed by 10 questions taken from the Physical Symptoms Inventory (Spector & Jex, 1998). The sum variable covers a wide spectrum of physical symptoms that are known to be common stress symptoms (e.g. headache, nausea or backache). Respondents were requested to answer each item on the basis of their experiences over the previous 12 months, using a five-point scale ranging from 1 (never) to 5 (often/continuously). The alpha coefficients for the scale at Time 1 were 0.81 for the men and 0.83 for the women. Those at Time 2 were 0.77 for the men and 0.83 for the women. For validity evidence, see Spector and Jex (1998).

544 A. Ma¨kikangas, U. Kinnunen /Personality and Individual Differences 35 (2003)537–557

1.6. Data analysis overview

Hierarchical multiple regression analysis was used to examine the potential main effects of psychosocial stressors and personality factors, and the moderating effects at Time 1 on well-being outcomes at Time 2. The regression procedure followed Baron and Kenny’s (1986) guidelines. To minimize potential multicollinearity problems optimism and self-esteem (r=0.75 for men and r=0.73 for women) were analysed in different regression models.

Each well-being indicator at Time 2 (2000) was regressed on the antecedent sets at Time 1 in five steps as follows: (1) each well-being outcome measured at Time 1 (1999) was entered to control for its effect; (2) demographics (age, education, leadership position); (3) psychosocial work stressors at Time 1 (1999); (4) personality characteristic at Time 1 (1999) and (5) the inter­action terms between work stressors and personality characteristics. The magnitude of R² change at each step of the hierarchical regression analysis was used to determine the variance explained by each set of antecedents. The beta values reported were used to determine the effect of each variable in the antecedent sets on well-being.

2. Results

2.1. Descriptive results

The means, standard deviations and correlations of all the study variables are displayed separately for men and women in Table 2. Gender differences emerged in two of the variables studied. The women suffered more than the men from time pressures at work at Time 1 [t (449)=9.64, P <0.01] and phy­sical symptoms both at Time 1 [t (452)=32.3, P <0.001] and at Time 2 [t (451)=7.67, P <0.001].

In general, the correlations of the study variables between Time 1 and Time 2 were in the expected direction. The psychosocial work stressors and personality characteristics at Time 1 were moderately related to all well-being indicators at Time 2 (i.e. job satisfaction, emotional exhaustion, mental distress and physical symptoms). Most of the psychosocial work stressors at Time 1 correlated moderately with each other, as did the well-being indicators at Time 2. Of the demographic variables, education was related to self-esteem among men; those male employees with a higher education perceived their self-esteem higher than those with a lower education. In addition, the older female employees were less educated than the younger employees and the older male employees perceived more job satisfaction than the younger ones.

2.2. Main effects

After controlling for the outcome variable’s own effect at Time 1 (Step 1) and the demographics (Step 2), the entry of the psychosocial work stressors (Step 3) and personality characteristics (Step 4) revealed some support for the hypothesized main effects. As seen in Tables 3 and 4, the well-being outcomes at Time 1 (Step 1) explained a substantial proportion of the variance in all the well-being outcomes at Time 2, due to the fact that well-being seems to be relatively stable over time. Partly because of this, the full set of antecedent variables at Time 1 accounted for a notable proportion of variance in well-being indicators at Time 2 (24%–44%) for both genders.

Table 2