Abstract
This article is concerned with whether employees on temporary contracts in Britain report lower well-being than those on permanent contracts, and whether this relationship is mediated by differences in dimensions of job satisfaction. Previous research has identified a well-being gap between permanent and temporary employees but has not addressed what individual and contract specific characteristics contribute to this observed difference. Using data from the British Household Panel Survey, the article finds that a large proportion of the difference in self-reported well-being between permanent and temporary employees appears to be explained by differences in satisfaction with job security. Other dimensions of job satisfaction are found to be less important. In fact, after controlling for differences in satisfaction with security, the results suggest that temporary employees report higher psychological well-being and life satisfaction. This indicates that an employment contract characterized by a definite duration lowers individual well-being principally through heightened job insecurity.
Introduction
During the last two decades extensive labour market reforms have been undertaken around Europe in an effort by policy makers to enhance the flexibility and improve the performance of European labour markets. An aspect of these reforms has been the easing of the restrictions regulating the use of temporary employment contracts (OECD, 2004). So-called flexible contracts such as temporary work, fixed-term/contract work and agency work are now widespread across the European workforce. Britain has always been a relatively deregulated labour market by European standards and, thus, has been relatively unaffected by these recent labour market reforms. However, the prevalence of temporary forms of employment (according to Eurostat, some 1.5 million UK employees are in temporary employment today) has led to a growing interest among academics and policy makers in the impact of increased flexibility on employment outcomes and, importantly, the well-being of individual employees (Booth et al., 2002).
In principle, temporary employment can have both positive and negative aspects and, hence, consequences for workers. Flexible scheduling arrangements and other aspects of the daily work experience related to temporary work may be valued and preferred by some employees, whereas the insecurity and poorer working conditions associated with these contract types can have a negative impact on workers’ well-being (Blanchard and Landier, 2002; Carrieri et al., 2012). The relationship between temporary contracts and well-being, thus, will depend, among other things, on the voluntary or involuntary nature of such work, the specific type of the non-permanent contract, the institutional context and the overall labour market performance of the country of interest, as well as, importantly, the workers’ reactions and attitudes towards the conditions associated with temporary employment (Bardasi and Francesconi, 2004; Carrieri et al., 2012; De Cuyper and De Witte, 2005; De Witte and Näswall, 2003; Gash et al., 2007; Silla et al., 2005; Virtanen et al., 2005).
Using data from the British Household Panel Survey (BHPS), this article contributes to this literature by analysing the relationship between temporary employment status and four subjective well-being measures in Britain. These measures are frequently examined in related research and are meant to capture different concepts of well-being, namely the subjective (or self-assessed) health and the psychological (or mental) well-being of the workforce (Bardasi and Francesconi, 2004; Robone et al., 2011; Rodriguez, 2002). They include psychological distress and life dissatisfaction (both closely related to the concept of psychological/mental well-being), anxiety/depression (capturing an explicit and probably diagnosed, long-term mental health condition), and poor general health (which refers to subjective general health status that includes physical health issues as well). Previous research in this area typically establishes the link between contract type and well-being with the inclusion of temporary employment dummy variables in standard well-being equations. While this sheds some light on well-being variations between the different groups, it does not provide much information upon the origins of these differences, which can be informative for public and labour market policy. Our key contribution is not only to study the effects of temporary employment on subjective well-being but to try to understand the mechanisms behind this relationship, with a particular focus on the mediating role of self-reported job satisfaction. In particular, a range of dimensions of job satisfaction are examined, including overall satisfaction with the job, satisfaction with job security, total pay, hours of work and the actual work itself. Job satisfaction measures can be quite informative in this respect as they capture the workers’ reactions and attitudes towards the array of job characteristics associated with each type of contract and, therefore, allow for a summary subjective evaluation of the consequences of temporary employment (Hamermesh, 2001). It seems quite surprising that previous research has not investigated this mediating influence of job satisfaction, since a substantial body of research has focused on the tendency of those employees on temporary contracts to report lower satisfaction than permanent workers, especially in domains associated with job security (see, for example, Chadi and Hetschko, 2013; Green and Heywood, 2011; Origo and Pagani, 2009).
In order to gain a deeper understanding of this issue, our analysis of BHPS data employs a variety of modelling approaches, including a decomposition technique which allows us to quantify the contribution of job specific and demographic characteristics, as well as the different dimensions of job satisfaction, on well-being differentials between groups. Our analysis reveals that a large proportion of the difference in well-being between an employee on a temporary employment contract and a similar one on a permanent contract can be attributed to differences in their satisfaction with job security. Other facets of job satisfaction are found to contribute much less to the overall difference, as are individual and other job specific factors. This contribution to the literature is particularly relevant from a policy perspective, considering that contemporary labour market policy reforms throughout Europe have focused on the goal of ‘flexicurity’, i.e. the combination of labour market policies and measures that aim to enhance both labour market and organizational flexibility and the employment security of individual employees (Berglund et al., 2014; Chadi and Hetschko, 2013; Origo and Pagani, 2009).
The remainder of this article is structured as follows. The next section presents the background to the questions at hand and reviews the related literature. We then describe the longitudinal data source we use and provide some preliminary descriptive analysis. The following two sections develop the empirical methodology and present the results of the empirical analysis, while the last section provides a final discussion and conclusion.
Background
Several studies have examined the relationship between contract type and the well-being of employees in Britain and across Europe (see Virtanen et al., 2005, for a detailed review). The typical (mainly cross-sectional) finding is that employees on temporary employment contracts record lower levels of subjective health and psychological well-being than comparable employees on traditional, full-time permanent contracts (Quesnel-Vallée et al., 2010; Silla et al., 2005; Virtanen et al., 2005). However, studies endeavouring to identify causal relationships between different contract types and subjective well-being using panel data have in general found a weak or no negative impact of atypical employment on the well-being of workers (e.g. Bardasi and Francesconi, 2004; Rodriguez, 2002).
There are some important reasons why the relationship between temporary contracts and well-being may not be unambiguously negative. On the one hand, temporary work can be desirable for employees that want to have more flexibility and an independent control over their working schedule, while others may consider it as a necessary stepping-stone towards a more integrated position in the labour market (Bardasi and Francesconi, 2004; Booth et al., 2002; Virtanen et al., 2005). On the other hand, temporary jobs are associated with higher job insecurity and increased unemployment risk and are more likely to be characterized by poorer working conditions and wage penalties relative to permanent jobs (Booth et al., 2002). Indeed, an examination of the studies most closely related to ours and which examine and report the type-of-contract effect on well-being using the same data as we do, suggests that the empirical evidence is quite mixed.
Rodriguez (2002) uses British data from the BHPS and German data from the German Socio-Economic Panel and finds no statistically significant relationship between atypical employment and subjective general health status in Britain (for Germany, though, the results point to a negative association). Using the same dataset, Bardasi and Francesconi (2004) report negative effects of temporary employment only on job satisfaction, while their estimates for psychological well-being and subjective general health status are statistically insignificant in the majority of their model specifications. In contrast, Taylor (2006) reports that holding a casual or seasonal temporary employment contract has a detrimental effect on psychological well-being, whereas fixed-term contracts are found to have a positive effect. Finally, Robone et al. (2011) report some negative effects of contractual conditions on subjective general health and psychological well-being, although these depend on the working time preferences, the family situation and the employability of the survey respondents.
A parallel body of literature has also examined the relationship between temporary employment and job satisfaction, with the majority of studies finding a negative relationship (Chadi and Hetschko, 2013; De Witte and Näswall, 2003; Green and Heywood, 2011; Origo and Pagani, 2009). In particular, De Witte and Näswall (2003), based on deprivation, psychological contract and job stress theories, provide a series of arguments on why those on temporary contracts may have lower job satisfaction than permanent employees. More specifically, temporary employees may belong to a peripheral group of employees who are not considered to be ‘part of the corporate family’ (De Witte and Näswall, 2003: 152), characterized also by lower wages and worse working conditions. Moreover, the temporarily employed may feel that the ‘psychological contract’ is unbalanced in favour of the employer, while temporary contracts are also associated with more job stressors related to work content, conditions and relations than permanent ones. Perhaps one of the most important consequences for workers on temporary contracts is a heightened feeling of job insecurity (Burchell, 1999; De Cuyper and De Witte, 2005; De Witte and Näswall, 2003; Green, 2003; Sverke et al., 2006). Temporary employment can be thought as a more insecure form of work since, by definition, a temporary contract has a limited time span which is likely to create concerns about job continuation through a renewal of the contract or an upgrading to a permanent one (De Witte and Näswall, 2003).
In this context, measures of different domains of job satisfaction (with security, pay, hours and the work itself) can be useful in capturing the subjective conceptualization of these negative consequences of temporary employment which may impact more widely on worker well-being. For instance, Burchell et al. (2002) find evidence of a strong relationship between job insecurity feelings and stress, while job insecurity has also been linked with constraints on the planning of current and future life activities, more work–family conflict and a deterioration of family life that can lead to greater psychological distress (Burchell, 1999; Gash et al., 2007; Scherer, 2009).
In this article, we therefore contribute to the literature by using four subjective well-being measures as the outcomes of interest and estimating the contribution of differences in several dimensions of job satisfaction to the well-being gap between temporary and permanent employees. We also discuss in detail the contribution of other individual and contract specific characteristics to the overall well-being gap.
Data and descriptive analysis
The data used for the empirical analysis are from the British Household Panel Survey (BHPS) and cover the years 1991–2008 (Waves 1–18). BHPS is a nationally representative survey of more than 5000 households and approximately 10,000 individuals in Great Britain, sampled in 1991 and followed since then. The BHPS contains data at the individual and household level covering household composition, housing characteristics, education and training, health, labour market status and job characteristics, and values and opinions on social and political matters.
The BHPS asks individuals to self-report whether their current job is on either a permanent or a non-permanent contract, thus identifying temporary employment on this basis. It is also possible to partition the sample of non-permanent employees into two further groups of individuals: (1) those holding a seasonal, agency temporary or casual job, and (2) those under contract for a fixed period or for a fixed task. This is a standard practice in the related literature (Bardasi and Francesconi, 2004; Booth et al., 2002; Taylor, 2006) and, as is shown below, there are substantial differences between the job characteristics of these two groups that would render the use of an overall ‘temporary contract’ variable inappropriate. For the remainder of the article those in the latter group are referred to as fixed-term workers and those in the former as casuals. 1 The sample used for the subsequent analysis is restricted to the original BHPS sample covering Great Britain and to employees that are below the state pension age (16–59 for women, 16–64 for men) and gave a valid response to being on either a permanent or non-permanent contract. The final sample consists of 60,058 person-year observations. These correspond to 57,567 person-year observations for permanent employees, 1310 for fixed-term employees and 1181 for casuals. 2
To explore the association between employment contract and well-being, information is used from four questions routinely used in empirical analyses of this type (see Bardasi and Francesconi, 2004; Madden, 2010; Robone et al., 2011; Rodriguez, 2002; Taylor, 2006). These are the following (see also Table 1):
Psychological distress. This uses the General Health Questionnaire (GHQ, 12-point measure) asked at each wave of the BHPS. The GHQ is widely used especially in the medical literature as an indicator of minor psychiatric morbidity and psychological distress (Madden, 2010). The GHQ has 12 items which have a four-point (from 0 to 3) scoring system that ranges from a ‘better/healthier than normal’ option, through a ‘same as usual’ and a ‘worse/more than usual’ to a ‘much worse/more than usual’ option. Higher scores correspond to lower well-being (higher psychological distress). 3
Poor general health. Respondents were asked at each wave (except for 1999, when this question was substantially changed) of the BHPS: ‘Compared to people of your own age, would you say your health over the last 12 months on the whole has been: excellent, good, fair, poor or very poor?’ Maintaining the same range, we construct a five-point scale that is increasing in poor general health.
Anxiety/depression (mental health condition). Respondents are asked at each wave of the BHPS: ‘Do you have any of the health problems or disabilities listed on this card?’ An option is ‘Anxiety, depression or bad nerves, psychiatric problems’. Responses are binary and take the value 1 if an individual suffers from a health problem related to anxiety or depression and 0 if not.
Life dissatisfaction. In Waves 6–10 and 12–18 respondents were asked: ‘How dissatisfied or satisfied are you with your life overall?’ Responses were given on a seven-point Likert scale ranging from ‘not satisfied at all’ to ‘completely satisfied’. Maintaining the same range, we rescale the variable so that it is increasing in life dissatisfaction.
Description and sample means for selected variables.
Notes: Observations refer to the final model for GHQ (see Table 2); asterisks refer to results from two-tailed t-tests of the null hypothesis that the difference between the temporary and the permanent mean is equal to zero (* rejected at 0.10, ** at 0.05, *** at 0.01).
Apart from being standard practice in some of the related literature (e.g. Bardasi and Francesconi, 2004), we also use these four measures in order to check the robustness of our results, keeping in mind the fact that these variables are also related to somewhat different underlying concepts of the well-being of the employees in our sample: psychological distress and life dissatisfaction are closely related to the concept of psychological or mental well-being, anxiety/depression is closely related to an explicit, and probably diagnosed, long-term mental health problem/condition, while poor general health refers to subjective general health status that should also include physical health issues as well. Confirming this, the pairwise correlations between these four dependent variables are generally low. The strongest correlation, as expected, is between psychological distress and life dissatisfaction (0.47), with the remaining correlation coefficients being below 0.30. Consequently, we proceed by analysing these four aspects of well-being separately.
Table 1 summarizes the distribution and variation of well-being indicators among the sample groups described earlier in the section as well as other potential mediating influences on well-being that have been discussed in the literature. From the descriptive information provided in Table 1, it is evident that there is a large well-being differential between contract types. In particular, those on permanent contracts have almost exclusively the lowest means on all four well-being indicators. This is followed by those employees on fixed-term contracts and then by those on casual contracts. T-tests are also performed for the difference in means between contract types. Comparing casuals to those on permanent contracts, the differences are highly significant for all four well-being measures, with casuals reporting much lower well-being. For fixed-term employees the differences are smaller and not statistically significant for anxiety/depression and life dissatisfaction. Moreover, while they report significantly higher psychological distress, they also appear to have better general health status than permanent workers. The remainder of this article seeks to identify the underlying mechanisms behind these observed raw differences in the data.
Looking at the sample means reported in Table 1 and Appendix Table A1, it is worth noting that those on casual contracts are more likely to be younger, female, single and, consequently, less likely to have an employed spouse/partner than permanent employees. They are also more likely to hold a second job, to work fewer normal or overtime hours and to be lower paid. Importantly, casuals are much less likely to have promotion prospects in their jobs, less likely to receive bonus payments or annual pay increments, and less likely to be members of an employer-provided pension scheme than permanent employees. The differences in these latter job characteristics and working conditions are especially large. On the other hand, fixed-term workers tend to be younger, female and better educated than permanent employees, while they also work shorter hours overall. There is also no significant difference in terms of the hourly wages of permanent and fixed-term contracts. Fixed-term employees also have lower promotion opportunities and are less likely to be members of employer pension schemes. Both flexible worker types are more likely to expect either a better or worse year financially, while the majority of permanent employees expects a financially similar immediate future. Moreover, from Appendix Table A1 we can see that fixed-term workers are concentrated in high-skilled occupations (professional, associate professional and technical) and are more likely to work in the public or other non-profit sector and, especially, in education or health and social work (almost 45% of fixed-term workers are employed in these two major industry groups). On the other hand, casual employees are more likely to work in personal and protective services or as plant and machine operatives (see also Booth et al., 2002). This descriptive evidence shows that employees on fixed-term contracts hold on average higher quality jobs than casuals. 4 Some of the above differences are expected to contribute to some extent to the differences in well-being mentioned earlier.
The BHPS also contains a number of items concerning facets of job satisfaction which are used throughout this article as key mediating influences on well-being. Within the BHPS responses for job satisfaction questions are given on a seven-point Likert scale ranging from ‘not satisfied at all’ to ‘completely satisfied’. All dimensions of job satisfaction available in the BHPS are used: (1) overall job satisfaction, (2) satisfaction with pay, (3) satisfaction with job security, (4) satisfaction with hours worked and, lastly, (5) satisfaction with the work itself. 5 Each job satisfaction aspect is correlated strongly and negatively with life dissatisfaction and, to a lesser extent, with psychological distress. In contrast, job satisfaction aspects appear weakly correlated with our other two well-being indicators.
Concerning the differences between contract types, Table 1 illustrates that casuals have lower levels of job satisfaction than those on permanent contracts. The t-tests are highly significant, confirming the differences in each case. For fixed-term workers, the differences in job satisfaction aspects from the permanent group are much less clear-cut. In fact, fixed-term workers appear to enjoy higher satisfaction both with hours and with the work itself. The key difference between the permanent and temporary groups, however, is the difference in satisfaction with security. For permanent contracts the mean level of satisfaction with job security is 5.4, while it is around 3.9 for fixed-term and casuals (there is no significant difference between the two temporary types). More specifically, permanent employees have approximately a 39% and a 38% higher mean level of satisfaction with job security than casuals and fixed-term workers, respectively. For the remaining dimensions of job satisfaction, permanent employees only have a modest satisfaction advantage over casual employees, ranging between 2% and 7%.
The above differences are similar to those reported by Green and Heywood (2011) and Booth et al. (2002). Consequently, the next section of this article is concerned with whether differences in the levels of satisfaction (and, especially, satisfaction with job security) between contract types are important factors in explaining why permanent and temporary workers report significantly different levels of well-being.
Multivariate regression results
Using our four measures of well-being as our dependent variables, this section reports the empirical results from multivariate regression analysis. All our ordered measures are treated as cardinal and the models are estimated by ordinary least squares (OLS). The same is done for the anxiety/depression (mental health condition) regression which is interpreted as a linear probability model. This modelling strategy was mainly chosen in order to be able to perform the detailed linear decomposition (see next section). Additionally, results from fixed effects OLS regressions are presented. A fixed effects specification is not possible with non-linear models. It should be noted though that all results are qualitatively very similar to those obtained from non-linear ordered and binary response models. 6
All regression models include as controls the range of covariates described in Table 1 and Appendix Table A1. These are standard socio-demographic and job-related controls that are consistently used in the literature. The variables for working hours, managerial-supervisory status, promotion opportunities in current job, bonus or profit-share payments, membership in an employer-provided pension scheme, annual increments, place of work and working in rotating shifts are used to control for differences in working conditions that are likely to affect well-being (Robone et al., 2011).
Pooled OLS estimates
We begin by pooling the data across the three contract types and estimate well-being equations with dummy variables included to identify the influence of the contract type. Table 2 presents the results for this procedure. These are all linear regressions estimated by OLS, with the standard errors clustered by individual to account for intra-group correlations. In each case comparisons are made between contract types while controlling for heterogeneity among individuals using standard control variables. We also add sequentially to the right-hand side of our well-being equations a single facet of job satisfaction as a key mediating influence on the relationship between contract type and our subjective well-being measures. These results are reported in columns (2) to (6). Column (1) reports estimates of the effects of the employment contract on well-being without controlling for variations in aspects of job satisfaction between the groups. Our estimates in column (1) are therefore our baseline estimates. While it is recognized that the dimensions of job satisfaction are likely endogenous in equations (2) to (6), we are not interested directly in the coefficient of the job satisfaction controls but in the effect of their inclusion on the contract-type coefficients (Green and Heywood, 2011).
The effect of different aspects of job satisfaction on mental well-being (pooled OLS).
Notes: All models include controls for gender, age, age squared, number of cigarettes smoked per day, marital status, number of children, household size, log of household income, whether spouse/partner employed, education, housing tenure, financial expectations for year ahead, union coverage and membership, usual normal hours worked per week and its square, usual paid overtime hours, usual unpaid overtime hours, managerial-supervisory status, holding a second job, promotion opportunities in current job, pay includes bonus or profit-share, member of employer-provided pension, pay includes annual increments, place of work, working in rotating shifts, occupation, industry, sector, firm size, job tenure and its square, log of hourly wage, region and survey year. Cluster-robust standard errors in parentheses. * Coefficient significant at 0.10, ** at 0.05, *** at 0.01.
Looking firstly at columns (1) of Table 2, those on casual employment contracts have significantly higher psychological distress and life dissatisfaction scores than employees on permanent contracts. For fixed-term employees the coefficients are positive in both cases but not statistically significant at conventional levels. As illustrated by the descriptive evidence in the previous section, the differences in well-being and other covariates between fixed-term and permanent employees are much less pronounced than those between casuals and permanent employees, possibly due to the higher quality jobs that fixed-term workers hold relative to casuals. Hence, we find no evidence of a gap in well-being between fixed-term workers and permanent employees once we control for a number of socio-demographic, job and working conditions variables. Another important observation from the results in columns (1) is that there is no statistically significant relationship between flexible employment and either poor general health or mental health condition. This is in line with our argument in the previous section that the concept of psychological well-being is better captured by our variables for psychological distress and life dissatisfaction. On the other hand, poor general health status is more closely related to health than psychological well-being (as it also covers physical health), while the mental health condition variable identifies long-term and, probably, formally diagnosed psychiatric problems. 7
The subsequent columns (2) to (6) control in turn for different facets of job satisfaction. All satisfaction coefficients are large and significantly negative, confirming the strong association between job satisfaction and well-being. In column (3) satisfaction with job security is included. This causes a dramatic change in the coefficients reported in the baseline estimates. In particular, if the differences in job satisfaction associated with security across the contract types are controlled for, fixed-term workers and casual employees are less likely than permanent employees to report psychological distress, with the former relationship being statistically significant at the 1% level. In contrast, including the other dimensions of job satisfaction as controls does not change the conclusions drawn from our baseline estimates. Consequently, it can be argued that differences in psychological distress between contract types are mediated by differences in satisfaction with job security and not by other aspects of job satisfaction that may be associated with the type of contract. For life dissatisfaction, the inclusion of job satisfaction covariates reveals a similar picture. That is, controlling for all dimensions of job satisfaction except for satisfaction with job security, yields similar results to the baseline model; however, controlling for differences in job security feelings suggests that both fixed-term workers and casuals have lower life dissatisfaction. Both these findings are statistically significant at the 1% and 5% level respectively. For the other two well-being indicators, the inclusion of the job satisfaction variables does not cause any substantial change in the baseline estimates. The exception is, again, when satisfaction with security is added. For the poor general health variable, both flexible employment coefficients become significantly negative, indicating a better health status for people in flexible contracts once job security considerations are taken into account. For mental health condition, the two coefficients are now negative, although still statistically insignificant.
To summarize, after controlling for satisfaction with security, all temporary contract dummies appear to acquire a negative coefficient which points to a positive association between temporary employment and well-being. This reflects aspects of this type of contract that are beneficial for individual employees, such as greater scheduling flexibility, or a voluntary sorting in such jobs (especially in fixed-term contracts) which may be considered as stepping-stones towards a permanent contract (Booth et al., 2002; Carrieri et al., 2012; Green and Heywood, 2011).
The associations between the other covariates and well-being are now briefly discussed. 8 Psychological distress depicts an inverted U-shaped relationship with age and is substantially higher for females and those with low financial expectations. Noticeably, psychological distress is also higher for employees who work more unpaid overtime hours and for the better educated, while, importantly, it is considerably lower for employees with promotion prospects and those whose pay includes annual increments. Life dissatisfaction also has an inverted U-shaped relationship with age and is higher for the better educated. The latter finding may be related to Clark and Oswald’s (1996) argument which suggests that the negative relationship between higher education and job satisfaction may be due to education raising aspiration targets. Life dissatisfaction is also markedly higher for individuals who work longer hours (normal or unpaid) and the lower paid. Consistent with the findings for psychological distress, employees with promotion prospects and whose pay includes annual increments report much lower life dissatisfaction. For poor general health the results suggest that a high score is less likely for males, the better educated, those with promotion prospects and the better paid. Finally, anxiety/depression is again less likely for males, employees with promotion prospects and those who hold a second job.
Fixed effects estimates
The pooled cross-sectional estimates just presented may not reflect the true impact of temporary employment on individual well-being. They may simply reveal that low well-being types seek (or are only hired on) flexible employment contracts or that unobservable individual characteristics such as talent, motivation or attitudes towards work predict both the type of employment contract and worker well-being. Consequently, a further strategy to control for these possibilities is to re-estimate the models described above using a fixed effects (or within) estimator. These estimates identify the effect of contract type on well-being by individuals who transition into and out of temporary employment and examine the corresponding changes in well-being. Previous studies using the same data and that control for individual worker fixed effects (Bardasi and Francesconi, 2004; Green and Heywood, 2011) suggest that well-being is largely unaffected by switches into and out of temporary employment.
However, one issue not addressed by Bardasi and Francesconi (2004), and only touched upon by Green and Heywood (2011), is that changes in contract type status often occur simultaneously with employer or job changes (Chadi and Hetschko, 2013). In fact, the majority (around 68%) of contract changes in our sample are job transitions. As such there are likely to be cross-firm heterogeneities, such as differences in the working environment or employer pressures exerted on workers that may be correlated with well-being. If these factors are not controlled for because the relevant data are unavailable, fixed effects estimates will attribute these effects to the type of contract. Although in our models we control for various job characteristics and working conditions variables, the BHPS does not contain thorough indicators that would adequately capture these cross-firm differences. One important implication related to this issue is that those individuals observed as leaving their permanent jobs and entering into temporary employment may have unusually poor permanent jobs (Green and Heywood, 2011). If this is the case, the effect of contract type change on individual well-being will be downwardly biased when using fixed effects. This is because poor quality permanent employment is likely to influence both the employment transition and overall well-being. Ideally, we would like to control for this by identifying individuals who remain in the same job but change only their contract type; unfortunately, these numbers within the BHPS are quite small. To investigate this issue further, we compared the job satisfaction of people in permanent employment who never become temporarily employed with those currently in permanent employment who subsequently become temporarily employed. We found very strong evidence that those permanent employees who will be in temporary employment in the future have substantially lower levels of job satisfaction than those who will not. These results lead us to conclude that previous studies on the causal influence of employment contract on well-being are likely to be downward biased. 9
Although the above discussion highlights the potential limitations of fixed effects modelling on estimating the relationship between contract types and well-being, Table 3 presents these estimates. First, what is interesting from these results is that even after controlling for unobserved individual heterogeneity, all facets of job satisfaction appear to have strong influences on all well-being measures. The coefficients are smaller than those presented in Table 2 but a similar conclusion can be drawn, i.e. that job satisfaction is positively related to individual well-being. We now turn our attention to the baseline estimates in columns (1). Casuals have higher psychological distress and life dissatisfaction scores than those in permanent contracts, with the former relationship being significant at the 5% level. For fixed-term workers the coefficients are negative but not statistically significant at the conventional levels. Again, the subsequent columns (2) to (6) control in turn for different facets of job satisfaction. First, for life dissatisfaction the inclusion of satisfaction with security causes a dramatic change in the coefficients reported in the baseline estimates. Casuals and fixed-term workers are now both less likely to report life dissatisfaction, with the latter relationship being statistically significant at the 5% level. The inclusion of other facets of job satisfaction, though, does not affect our baseline estimates. The results for our psychological distress estimates are very similar with those in Table 2. Again, only the inclusion of satisfaction with security causes a substantial change to the coefficients of interest. For poor general health and anxiety/depression, when controlling for fixed individual effects and satisfaction with security, fixed-term workers report significantly higher (at the 10% level) well-being than permanent workers. This is not the case when other aspects of job satisfaction are controlled for. In contrast, the coefficients for casuals in the poor general health models are significantly negative across all fixed effects specifications, while they are insignificant in the anxiety/depression ones.
The effect of different aspects of job satisfaction on mental well-being (fixed-effects OLS).
Notes: See notes in Table 2. Robust standard errors in parentheses. * Coefficient significant at 0.10, ** at 0.05, *** at 0.01.
The majority of associations between the well-being measures and the rest of our covariates in the pooled OLS models are largely robust to the inclusion of fixed effects. Briefly, psychological distress is higher for employees who work more unpaid overtime hours and for those with low financial expectations. Importantly, it is considerably lower for employees with no promotion prospects and those whose pay includes annual increments. Consistent with the above, life dissatisfaction is higher for those who work more unpaid overtime hours and the lower paid, while it is also lower for those with no promotion prospects. Again, poor general health and anxiety/depression are negatively associated with promotion opportunities.
To sum up, after controlling for differences in satisfaction with job security, becoming a casual and (mainly) a fixed-term employee has in general a positive influence on subjective psychological well-being. This is broadly in line with the conclusions drawn from the pooled OLS results.
Decomposing the well-being gap between temporary and permanent employees
In the previous sections we provided evidence that satisfaction with job security is a strong predictor of the difference in well-being between employees in permanent and temporary contracts and that this result is more clear-cut for the two variables more closely related to psychological well-being, namely psychological distress and life dissatisfaction. However, we have said nothing about its total contribution to the temporary/permanent well-being gap relative to the contribution of the other explanatory variables in the models. To enable a further understanding of the differences between permanent and temporary employees in the strength of the various factors entered in the regression models, a linear decomposition analysis is undertaken in this section.
When the outcome of interest is continuous and modelled using a linear regression, the Blinder–Oaxaca (Blinder, 1973; Oaxaca, 1973) decomposition technique is widely used. The standard Blinder–Oaxaca decomposition of the permanent/temporary gap in the average value of the outcome variable Y can be expressed as:
where
Table 4 provides the results of this decomposition analysis for the explained part of the well-being gap between temporary and permanent employees. For brevity, we only report results of the model which includes the facet of job satisfaction associated with job security as a control and we briefly comment in the end on the results from the other specifications. This is because satisfaction with job security represents the largest raw differential between the contract types, as shown in Table 1.
Blinder–Oaxaca decomposition of mean differences in well-being between permanent and temporary employees – contributions of selected variables.
Notes: Asterisks indicate whether the contribution of each variable to the mean difference in well-being is statistically significant at the 0.01 (***), 0.05 (**) or 0.10 (*) level respectively.
Panel A of Table 4 reports the well-being decompositions between fixed-term and permanent workers, while Panel B reports the decompositions for casuals compared to those in permanent employment. The upper part of each panel shows the mean well-being score for the employment contract subsamples. The differences in these average well-being scores are then shown, followed by the difference explained by all the explanatory variables of the model. The lower part then provides individual contributions to the well-being gap from selected differences in covariates along with indicators of their statistical significance. 11 Cluster-robust standard errors (not reported) used for the significance tests are calculated via the delta method (see Jann, 2008).
Starting our analysis with Panel A of Table 4, we notice that the differences in the group means of the well-being indicators between permanent and fixed-term workers tend to be relatively small compared with the mean differences observed in Panel B, something that we have already noted in the previous sections. Consequently, most of our discussion below is concentrated on the difference between permanent employees and casuals. As a brief comment on the results in Panel A, the small observable differences between the two groups can largely be explained by differences in satisfaction with job security. In fact, the size of the contributions of the satisfaction with job security variable is so large, that it appears that those on fixed-term contracts would have markedly higher well-being than permanent employees once we controlled for heterogeneity among individuals and differences in job security. Another observation is that differences in socio-demographic characteristics and financial expectations appear to be much more important than the differences in job characteristics in explaining the gaps between permanent and fixed-term employees, a direct result of the generally higher quality of these fixed-term jobs. The only difference in job characteristics that appears to be a consistently significant determinant of the well-being gap is the difference in promotion opportunities between the two contract types.
In Panel B, the differences in the group means on the well-being indicators are relatively large. For psychological distress the employment contract gap is 0.419, a 25% difference in percentage terms. Of this gap, 123% can be explained by differences in the covariates’ distribution, with the remaining small offsetting difference of 23% (–0.096) being due to the differences between the coefficients. A large proportion of the raw difference can be explained by the difference in job satisfaction with security distributions between contract types, as well as the different gender, household characteristics, annual increments and promotion prospects distributions. In particular, the higher average permanent employee satisfaction with job security in the sample explains 89% of the gap. For gender, annual increments in pay and promotion opportunities the percentages are 16%, 8% and 7%, respectively. For anxiety/depression the well-being gap between contract types is 0.018 (a 41% difference in percentage terms). Of this observed difference, 94% can be explained by the whole model, 42% can be explained by differences in satisfaction with security between the contract types and 18% by the difference in the hourly wage. Other large and statistically significant contributions to this particular health gap are those of gender (18%), age (–35%) and differences in promotion opportunities (9%). The negative contribution of age to the anxiety/depression gap is due to the fact that permanent employees are older on average than casuals and age has a positive effect on the probability of reporting a mental health condition, thus reducing this particular well-being gap.
For life dissatisfaction the raw gap is 0.165 and the model used explains approximately 177% of the gap. Again, of the observed raw difference between the groups, 137% can be explained by satisfaction with job security, 38% by marital status and −48% by the higher mean age for permanent workers. Since permanent employees work more hours and working hours are positively related with life dissatisfaction, this set of variables has a negative contribution (–22%) to the well-being gap as well. For our last indicator of well-being, poor general health, the observed difference is relatively small, but the results are consistent with the other measures. Satisfaction with security explains a large proportion of the well-being gap. In fact, for all well-being measures the unexplained part suggests that if those on casual contracts had identical socio-demographic, job and workplace characteristics to those in permanent employment, those on casual contracts would have no worse well-being.
It should be noted, finally, that decompositions using the other job satisfaction variables were also estimated. The conclusions are the same as those reported in the previous section: the security aspect of job satisfaction has the largest contribution to the permanent/temporary well-being gap. For example, overall job satisfaction explains 37% of the permanent/casuals gap in psychological distress (relative to 89% for the satisfaction with security aspect reported above) and 55% of the gap in life dissatisfaction (relative to 137% for the security aspect). Lower contributions are also estimated when using the other satisfaction aspects in our regression models.
Discussion and conclusions
This article has been concerned with the extent to which the difference between temporary and permanent employees in Britain in their self-reported levels of well-being can be explained by differences in observable characteristics and certain dimensions of job satisfaction. Previous research has established the link between temporary employment contracts and below average well-being in some detail (Bardasi and Francesconi, 2004; Virtanen et al., 2005). However, the important question of whether certain job characteristics and, in particular, certain dimensions of job satisfaction are likely to influence the well-being differentials between different contract types has not been addressed. This is particularly surprising given that a parallel literature has also examined the proposition that temporary employees tend to report lower job satisfaction, especially in domains associated with job security (Chadi and Hetschko, 2013; Green and Heywood, 2011).
By analysing data from the BHPS, this article finds evidence that individuals on temporary employment contracts, especially casuals, report lower well-being than their counterparts in permanent employment. The multivariate analysis shows that this is more the case for the well-being measures capturing psychological well-being (psychological distress and life dissatisfaction) and less so for those measures more closely related to subjective health status (anxiety/depression and poor general health). Consistent with these findings, temporary employees are found to have generally lower job satisfaction, with the difference between the contract types being especially large for satisfaction with job security. This latter finding, in turn, appears to explain a very large proportion of the difference in well-being between temporary and permanent employees. More specifically, we find that, if differences in satisfaction with security between contract types are controlled for, fixed-term workers exhibit significantly higher levels of well-being and casuals no worse than those workers in permanent employment. An important further finding is that controlling for any other aspects of job satisfaction (satisfaction with pay, hours, or the work itself) does not alter our baseline conclusion drawn from Table 1 that temporary employees are more likely to report lower well-being. Other variables, including some socio-demographics and working conditions like household income, promotion prospects in current job and existence of annual increments in pay, appear to explain a part of the well-being gap between contract types, although they are far less important than satisfaction with job security.
The promotion of the idea of ‘flexicurity’ among policy circles has meant that labour market reforms undertaken throughout Europe in the last decades have as their main aim to increase both labour market flexibility (mainly through the promotion of flexible employment contracts and less strict job protection legislation) and the employment and income security of individuals (Chadi and Hetschko, 2013; Origo and Pagani, 2009). However, the results presented here show that although increased flexibility associated with temporary contracts may offer reimbursements that are beneficial for individual well-being at the micro level, workers on these types of contracts in Britain suffer from a well-being penalty, at least in the short-run. This means that the gains from flexibility, such as more flexible scheduling arrangements and the opportunity to accumulate skills in order to secure a permanent contract in the future, cannot outweigh the costs in terms of psychological well-being that are mainly the result of greater dissatisfaction with security among temporary workers. Moreover, some recent evidence also suggests that it is not at all certain that the costs in terms of well-being due to increased job insecurity can be effectively reimbursed by alternative policies that enhance employability and income security instead (Berglund et al., 2014; Chadi and Hetschko, 2013; Silla et al., 2009). This in turn may have serious implications for the welfare state and the macro-economy (Burgoon and Dekker, 2010), through a greater political pressure for a more generous welfare state as the number of temporary workers increases, and an accompanying increase in health care or other welfare state costs. These unfavourable developments can be prevented by a different policy approach that takes into account the importance of satisfaction with job security for the well-being of individual employees.
Footnotes
Appendix
Descriptive statistics for other control variables.
| All | Permanent | Fixed-term | Casual/seasonal/agency | |
|---|---|---|---|---|
| Age squared (/100) | 15.792 | 15.901 | 14.157 | 12.314 |
| (8.772) | (8.746) | (8.795) | (9.132) | |
| Number of cigarettes per day | 3.914 | 3.889 | 3.276 | 5.811 |
| (7.771) | (7.766) | (6.898) | (8.627) | |
| Married or cohabiting | 0.744 | 0.750 | 0.651 | 0.551 |
| Widowed, divorced or separated | 0.074 | 0.074 | 0.064 | 0.078 |
| Never married (base category in models) | 0.182 | 0.176 | 0.285 | 0.371 |
| Outright house owner | 0.132 | 0.131 | 0.124 | 0.154 |
| House owner with mortgage | 0.688 | 0.692 | 0.623 | 0.530 |
| Rented house | 0.086 | 0.083 | 0.165 | 0.159 |
| Social housing (base category in models) | 0.095 | 0.094 | 0.089 | 0.157 |
| Union covered, not member | 0.188 | 0.182 | 0.374 | 0.262 |
| Union covered, member | 0.324 | 0.330 | 0.266 | 0.108 |
| Not union covered (base category in models) | 0.488 | 0.488 | 0.360 | 0.631 |
| Manager/foreman/supervisor | 0.393 | 0.402 | 0.215 | 0.114 |
| Holding second job | 0.092 | 0.089 | 0.168 | 0.148 |
| Work location – home | 0.011 | 0.011 | 0.016 | 0.009 |
| Work location – other | 0.073 | 0.070 | 0.131 | 0.117 |
| Work location – driving/travel | 0.082 | 0.082 | 0.052 | 0.070 |
| Work location – employer (base category in models) | 0.835 | 0.837 | 0.802 | 0.804 |
| Managers & administrators | 0.159 | 0.164 | 0.066 | 0.033 |
| Professionals | 0.110 | 0.106 | 0.312 | 0.083 |
| Associate professional & technical | 0.123 | 0.123 | 0.164 | 0.070 |
| Clerical & secretarial | 0.190 | 0.189 | 0.170 | 0.253 |
| Craft & related | 0.102 | 0.103 | 0.073 | 0.058 |
| Personal & protective services | 0.100 | 0.099 | 0.102 | 0.172 |
| Sales | 0.068 | 0.068 | 0.025 | 0.082 |
| Plant & machine operatives | 0.085 | 0.085 | 0.037 | 0.136 |
| Other occupations (base category in models) | 0.064 | 0.063 | 0.051 | 0.114 |
| Civil service | 0.047 | 0.048 | 0.034 | 0.022 |
| Local government | 0.147 | 0.143 | 0.293 | 0.179 |
| Other public | 0.085 | 0.083 | 0.188 | 0.053 |
| Non-profit | 0.031 | 0.031 | 0.068 | 0.020 |
| Private firm (base category in models) | 0.690 | 0.695 | 0.417 | 0.727 |
| Workplace size 1–49 | 0.467 | 0.464 | 0.454 | 0.582 |
| Workplace size 50–499 | 0.355 | 0.358 | 0.309 | 0.283 |
| Workplace size > = 500 (base category in models) | 0.178 | 0.178 | 0.237 | 0.136 |
| Tenure in years | 4.538 | 4.679 | 1.434 | 1.105 |
| (5.785) | (5.836) | (2.920) | (3.024) | |
| Tenure squared | 0.541 | 0.560 | 0.106 | 0.104 |
| (1.388) | (1.408) | (0.584) | (0.686) | |
| Agriculture & fishing | 0.008 | 0.008 | 0.005 | 0.016 |
| Mining & quarrying | 0.003 | 0.004 | 0.002 | 0.003 |
| Manufacturing | 0.204 | 0.208 | 0.118 | 0.135 |
| Electricity, gas & water | 0.011 | 0.011 | 0.010 | 0.009 |
| Construction | 0.038 | 0.038 | 0.040 | 0.029 |
| Wholesale & retail trade | 0.135 | 0.138 | 0.031 | 0.114 |
| Hotels & restaurants | 0.037 | 0.036 | 0.013 | 0.110 |
| Transport, storage & communication | 0.064 | 0.065 | 0.040 | 0.065 |
| Financial intermediation | 0.055 | 0.056 | 0.038 | 0.035 |
| Real estate & business activities | 0.101 | 0.099 | 0.104 | 0.151 |
| Public administration & defence | 0.081 | 0.082 | 0.070 | 0.045 |
| Education | 0.095 | 0.090 | 0.282 | 0.121 |
| Health & social work | 0.114 | 0.113 | 0.170 | 0.086 |
| Social & personal services | 0.035 | 0.034 | 0.047 | 0.053 |
| Private households & extra-territorial organizations (base category in models) | 0.006 | 0.005 | 0.003 | 0.011 |
| Observations | 60,058 | 57,567 | 1310 | 1181 |
Notes: Standard deviations for continuous variables in parentheses. Models also include controls for region and survey year.
Acknowledgements
The authors would like to thank Mark Bryan, Alex Bryson, Manos Matsaganis and an anonymous reviewer for valuable comments on earlier drafts of this article.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.