Unemployment, gender and mental health: the role of the gender regime

Authors


Address for correspondence: Mattias Strandh, Department of Sociology, Umeå University, Umeå 901 87, Sweden
e-mail: mattias.strandh@soc.umu.se

Abstract

Existing research suggests that gender differences in the effect of unemployment on mental health are related to the different positions and roles that are available for men and women in society and the family; roles that are connected with their different psychosocial and economic need for employment. The aim of this article is to analyse the role of gender in the relationship between unemployment and mental wellbeing in Sweden, representing a gender regime with a similar need for employment among women and men, and Ireland, representing a gender regime in which the need for employment differs between women and men. The results, based on longitudinal data from the two countries, show that unemployment was more negatively related to mental health among men than among women in Ireland, while men and women were equally affected by unemployment in Sweden. Factors related to the family and economic situation, as well as gendered selection into the unemployment population, explains the difference in mental health between unemployed men and women in Ireland. The overall conclusion is that the context has a major influence on the relationship between unemployment, gender and mental health.

Introduction

The conclusions from previous research on the relationship between unemployment and mental health have been clear. Although there are selection effects, as mental health affects job chances (see, for instance, Claussen et al. 1993, Mastekaasa 1996), a great body of high-quality longitudinal research has shown unemployment to be negative for mental wellbeing (see, for instance, Andersen 2009, Clark 2003, Murphy and Athanasou 1999, Strandh 2000a). The adverse influence of unemployment on mental health does, however, seem to differ depending on social characteristics. Variables such as the economic situation, social class, age, gender, marital status, duration of unemployment, previous unemployment experience, ethnicity and work involvement have all been found to influence the effect of unemployment on mental health (see, for instance, Andersen 2009, McKee-Ryan et al. 2005, Nordenmark and Strandh 1999).

The role of gender in the association between unemployment and mental health has been discussed in classical literature (Jahoda 1982) and has been received with considerable interest in modern research; however, results to date have tended to be inconclusive. Even meta-analyses have come to differing conclusions; for instance, McKee-Ryan et al. (2005) find that unemployed women have worse mental health and Paul and Moser (2009) find that men are more distressed by unemployment than women. In this article we discuss the role of gender in the association between unemployment and mental wellbeing from the perspective of theory and previous findings. This leads us to suggest that gender has different meanings for the unemployment experience depending on the context. The levels of equality in the labour market and in the family will here have real consequences for the way in which gender influences the association between unemployment and mental health. We investigate this using representative longitudinal data from Sweden and Ireland, two countries that represent different gender relations and gender regimes.

Unemployment, mental health and gender

The theoretical understanding of the relationship between unemployment and mental health has generally taken its starting point in the function, value of, or the need for employment. This approach was pioneered by Marie Jahoda (1982), who identified five necessary psychological functions (time structure, social contacts, participation in collective purpose, regular activity, and status and identity) which, she argued, could be satisfied fully only by employment. An alternative way to understand the negative relationship between unemployment and mental health has been to analyse the degree to which the unemployment situation limits the agency of the individual. The focus on limitations of the unemployment situation here, among other things, leads to an interest in poverty (Fryer 1992). Syntheses of these two approaches have been found fruitful, and the sociological, psychosocial and economic need for employment model combines psychosocial functions and agency perspectives into a conceptual model. The association can here be understood from the psychosocial and economic need for employment. These needs are, however, socially defined and can thus vary between individuals and groups of unemployed people. The model predicts that unemployed people who have both a weak economic need and a weak psychosocial need for employment (for instance, unemployed individuals living with a well-paid partner and strongly involved in activities not directly connected to employment) should not perceive unemployment as problematic and they may adapt relatively well to their new situation. On the other hand, the combination of strong psychosocial and economic need makes the likelihood of poor mental health higher (Nordenmark and Strandh 1999, Strandh 2000a).

The theoretical understanding of the relationship between unemployment and mental health has clear implications for how we understand their differential relationship for men and women. Theoretically, it should not be gender per se that decides the individual consequences of unemployment. Differences between men and women must instead be understood from the perspective of structurally different positions in the family, the labour market and society. These differences in position lead to differences in the economic and psychosocial importance of employment and could explain a differential association between unemployment and mental health (Nordenmark 1999). This view puts the emphasis squarely on the social context and the roles of men and women for understanding gendered differences in the outcome.

Gender roles on the labour market and in the family have a strong influence on how researchers understand the consequences of unemployment. Female identity is seen to be less connected with employment and the female income as a secondary income (with poorer working conditions) in the family (Hakim 1991, 1995, Jahoda 1982). Paul and Moser (2009: 266) express this idea as follows: ‘masculine identity is intricately linked to having a job in Western societies and is severely threatened by unemployment’. The expectations of such gendered implications of unemployment potentially leading to gendered mental health outcomes of unemployment are also supported by the classical empirical studies. Studies of the consequences of unemployment from the 1930s and 1940s generally found that men suffered more from being unemployed (Jahoda et al. 1971, Komarovsky and College 1973). In these studies female suffering was largely understood from the perspective of the family, where economic hardship and a discouraged breadwinner affected the dynamics and functioning of the family.

More recent research into unemployment, gender and mental health is, however, much more inconclusive. Over the years there have been a substantial number of studies that seem to support the idea that women suffer less from the unemployment experience than men (see, for instance, Artazcoz et al. 2004, Hakim 1995, Harpaz 1989, Theodossiou 1998, de Vaus and McAllister 1991). Yet there have also been studies in recent decades indicating that unemployed women miss employment as much as men do and as a consequence suffer to the same degree as unemployed men (see, for instance, Ensminger and Celentano 1990, Leeflang et al. 1992, Nordenmark 1999, Hammarström et al. 2011, Thomas et al. 2005). These contradictory findings are echoed in the two large-scale meta-analyses in the field of gender, unemployment and health. In one of these, McKee-Ryan et al. (2005) find, contrary to what could be expected from theory, that unemployed women have if anything worse mental health and lower life satisfaction than unemployed men. In the other Paul and Moser (2009) found that men were substantially more mentally distressed by unemployment than women.

The divergent findings do, however, seem to fit two different contexts. Studies that find empirical support for a gendered association between employment and health on average tend to be older or from countries with lower female labour force participation. This differentiation also appears to be valid for the meta-studies in the field (Hammarström et al. 2011). These differences might be crucial for understanding both the divergent findings between these two studies and in the research field in general.

The theoretical foundations for expecting gender differences in the relationship between unemployment and mental health are based on assumptions of the different roles and social positions of men and women. Gender roles do, however, differ across time and space. Female labour force participation in the Nordic countries, for instance, has been very high for a long time, while the female labour force participation in countries such as Spain, Germany or Ireland has historically been substantially lower. There are also differences in what female labour force participation typically means. Even if families with more than one income are becoming more common across Europe, women’s labour market position still differs between countries and from that of men in many respects. Women generally work fewer hours than men, earn less than men, have worse career opportunities and are more often found in atypical or part-time employment (Tijdens 2002).

The differences in female labour market participation are intimately connected to differences in the institutional setting that affect the life chances of men and women. These differences have been seen to emanate from the different countries’ breadwinner models (Lewis 1992, Sainsbury 1999) connected with the gender division of paid and unpaid work in the family and society. Gendered typologies of the institutional context have been further developed through the concept of defamilisation, which focuses on individuals’ potential for independence from their family and relatives through paid labour and state intervention (Bambra 2004, Esping-Andersen 2009). Given the understanding that the relationship between gender, unemployment and mental health is related to differences in the roles and conditions of men and women, these contextual differences between countries should be important for this relationship. There is little reason to expect gender to influence the relationship in the same way in contexts that are characterised by different gender relations.

A labour market where female participation is not encouraged, and female engagement in housework encouraged, might offer the assumed alternative roles to employment for women. This could reduce the normative pressure to take up employment and thus the psychosocial need for employment in women as compared to men. In a similar manner, the strong male breadwinner model means that cohabiting couples in which the man becomes unemployed typically lose their only, or their major, source of labour market income. We would therefore expect the economic need for employment to be greater among unemployed men than among unemployed women. This should, however, not be the case in a dual breadwinner context where the institutional setting supports defamilisation. With employment rates relatively similar to that of men, and expectations of continuous labour market careers, women in such a context should face relatively similar psychosocial and economic pressures as men when unemployed. The consequence should be small gender differences in the relationship between unemployment and mental health. Several leading researchers have also suggested this is a necessary route for understanding the relationship (Hammarström et al. 2011, McKee-Ryan et al. 2005).

Despite the substantial amount of previous research on unemployment, gender and mental health, there has been very little research investigating the role of the regime context. This study uses longitudinal data on initially unemployed samples to investigate the relationship between unemployment, gender and mental health in Sweden and Ireland. These two countries represent radically different institutional settings and are at opposite ends of the spectrum when it comes to the level of defamilisation (Bambra 2004, 2007). These differences are clear from Figure 1 where women’s labour market participation in Ireland and Sweden are compared with an average of EU15 from 1990 to 2004. Sweden stands out with a high proportion of women in paid employment, while Ireland has a lower participation rate than the EU15 average up to the end of the period.

Figure 1.

Proportion of women employed aged 15–64 as a percentage of the total female population in that age group. Data source: Eurostat, New Cronos data

Given these contextual differences between Sweden and Ireland, we expect gender to affect the relationship between unemployment and mental health differently in the two countries. Ireland fits the classic assumptions of a different need for employment between men and women, while Sweden represents a context where the psychosocial and economic need for employment should be relatively similar between men and women. One would therefore expect the relationship between unemployment and mental health to be gendered in Ireland but not in Sweden. Our first hypothesis is thus:

H1. Unemployment has a stronger negative connection with mental health among men than among women in Ireland but there will be no difference between men and women in Sweden

The theoretical underpinning of the first hypothesis is the assumption that the expected gendered association in Ireland is a function of the different psychosocial and economic need for employment between men and women, something that is not present in Sweden. If this is the case, variables that could be related to the different structural positions of men and women, such as family situation and economic situation, should be able to explain the difference in mental health between unemployed men and women in Ireland, while they should not affect the difference between unemployed men and women in Sweden. Given data restrictions it is only partially possible to investigate this but it allows for a second hypothesis:

H2. Variables related to the family and economic situation explain all or part of the difference in mental health between unemployed men and women in Ireland; however, they will not affect the difference between men and women in Sweden

Data and variables

In order to be able to investigate the hypotheses one needs access to comparable data for unemployed samples from different gender regimes as well as data that are longitudinal and provide information on mental wellbeing. The data from Sweden come from the long-term unemployment project (LUP), a nationally representative longitudinal data set based on a random sample of 3500 unemployed and recently unemployed people,1 drawn from the register of all unemployed individuals (HÄNDEL) at the beginning of 1996. In conjunction with the sampling an initial telephone interview was carried out where 2590 individuals responded (74 per cent), of whom 1806 were unemployed at the time of the first interview. A second telephone interview was carried out a year later, directed only at the 1806 individuals who were unemployed and who had responded at the first interview. This time 1415 individuals responded (78 per cent). Additional individual-level register information such as age, sex, citizenship, education and labour market status (openly unemployed, in a labour market programme and not unemployed) was drawn for the entire sample (both respondents and non-respondents) from HÄNDEL. The non-responses do not seem to be structured in a way that makes the data set particularly problematic (see Nordenmark 1999 for a detailed analysis).

The Irish data come from the Living in Ireland panel survey (LII). The LII survey included the Irish component of the European Community household panel survey and began in 1994. The respondents were followed up every year from 1994 up until 2001. The LII was based on a two-stage clustered random probability sample with an initial sample of 4048 households and 9904 individuals (a 63 per cent response rate at the individual level). Following this, the attrition was heaviest in 1995 but continued to occur up until the final year used in this article, 2001. In 1995, 89 per cent of the original households (3584) and 86 per cent of the original individuals (8532) were re-interviewed, although some households and individuals were re-recruited in subsequent years. By 1999 the number of households had been reduced to 2378 and that of individuals to 5451, so it was decided to supplement the sample. This led to the addition of 2661 households and 3527 individuals in 2000, giving a total sample of 5027 households and 8056 individuals. In order to create a data set that was comparable with the Swedish LUP data we selected all individuals who were found to be unemployed or in active labour market programmes in 1994, 1996 and 2000 (to achieve a large enough sample). The information was collected on these individuals one year later (1995, 1997 and 2001). Some respondents remained unemployed between the selected years. For these individuals only the first occurrence of unemployment was selected in order to avoid double counting. The resulting total Irish sample of initially unemployed is 1499 individuals, 1105 (74 per cent) of whom were available at follow up. As the base population for weighting was not clear, data are used unweighted.

Looking at the characteristics of the two samples in Table 1, there are substantial differences between them. These differences are very much what could be expected when drawing a sample of unemployed people from these two countries at the end of the 1990s. The differences reflect on the one hand the recent and long-term unemployment history of the two countries. At the time of the sampling Sweden had relatively recently had its first encounter with mass unemployment since the Great Depression. An economic and financial crisis during the first part of the 1990s had led to a dramatic increase in unemployment across a broad spectrum of the population. This led to a relatively heterogeneous unemployed population that reflected the labour force reasonably well. Ireland, at the time of the sampling, had long experienced high unemployment rates connected with entrenched unemployment as well as a more selected group of unemployment individuals. This is reflected in the Swedish respondents on average having a more even age distribution, being better educated, having been unemployed for a shorter period and being more likely to have a partner than the Irish respondents.

Table 1. Selected characteristics of unemployed individuals at time 1 (t1) by country and sex (mean of GHQ and percentages)
  Sweden (n)Ireland (n)
Women (804)Men (946)Women (488)Men (1011)
GHQ-12 (mean)8.347.8110.6411.06
Gender45.954.132.667.4
Single40.245.977.957.4
Partner59.854.122.142.6
Children aged under 18 in the household50.333.725.433.4
Age –24 years21.417.951.831.4
Age 25–34 years31.228.323.024.4
Age 35–44 years22.220.113.918.8
Age 45–54 years14.119.48.615.7
Age 55 + years11.014.22.79.7
Non-native citizen10.912.44.35.4
Compulsory education31.533.445.671.3
Secondary education58.258.140.622.5
Tertiary education10.38.513.86.2
Unemployed–12 months54.256.047.130.7
Unemployed 12–23 months19.620.014.212.5
Unemployed 24–47 months17.918.310.215.1
Unemployed 48+  months8.45.728.541.7
Making ends meet with ease26.731.628.518.4
 with some difficulty18.015.536.337.1
 with difficulty31.428.620.722.1
 with great difficulty23.824.314.522.4

On the other hand, these differences also reflect the difference in female labour force participation rates over the life course in the two countries. The lower proportion of women in the Irish sample corresponds to the lower labour force participation rate among Irish women. There are also substantial differences between the unemployed women and the unemployed men in the Irish sample; differences that are not present in the Swedish sample. The major differences between men and women in the Irish sample are that the women are younger, more likely to be single, less likely to be long-term unemployed and better educated than the men. All in all this would seem to fit the earlier discussion of differences in gender regimes. Generally, more discontinuity in the female labour market career and a stronger male breadwinner system, for instance, should be connected with female labour force participation (and hence the female unemployed population) being younger and more single. The first explanation for the differences proposed is less problematic as it is not directly related to, and should not have an impact on, the issues and mechanisms we investigate. The second explanation, however, is more important. It means that the unemployed Irish women probably are a more selected group in relation to their peers as regards labour market involvement than Swedish unemployed women. This has to be kept in mind when interpreting the results and is thus a weakness in the data. It also means, however, that the hypotheses will be tested using data that in the case of Ireland are biased against finding support for our hypothesis.

Variables

The central issue in this article is how gender affects the relationship between unemployment and mental health. Many measures of mental health are available, some specific to dimensions such as depression or anxiety and others that are more general. This study makes use of the 12-item version of the general health questionnaire (GHQ). This is a short and well-validated measure of psychological distress (Goldberg and Williams 1988) that is probably the most widely used screening scale for mental disorders (Goldberg et al. 1997). As the GHQ measures psychological distress that deviates from the individual’s usual condition it is possible that extended periods of distress would not be measured. Respondents do, however, actually seem to compare themselves to a pre-distressed state, and the GHQ-12 has been widely used in labour market studies as a relatively robust indicator of general mental health. The 12 items on the GHQ-12 (presented in Table A1 of the Appendix) have four alternative answers and are recoded and summed to provide a simple scale running from 0 to 36, where 36 is the highest level of psychological distress and 0 the lowest.

The distribution of the GHQ-12 in Table 1 does, however, suggest that there might be problems of comparison between the two countries. The GHQ scores are substantially lower in Sweden than in Ireland. This could to some extent be explained by the differences in characteristics of the two samples discussed above. It is, however, also likely that the country averages are influenced by cultural or historical factors that make absolute comparisons across countries problematic. To guard against this we have taken the precaution of standardising the GHQ variables in each country using a Z-score procedure and normalised each to have a mean of 50 and a standard deviation of 10. This ensures that the distributions across country are more or less identical and makes cross-country comparisons simpler. The methodology limits country comparisons to group differences within countries and prohibits the pooling of data. This is not a limitation in the study as the focus of the hypotheses is on comparative within-country comparisons between men and women.2

The central independent variable of the article is gender (man or woman). Besides this variable the article uses as controls a number of variables that have previously been shown to be of importance for the relationship between unemployment and mental health. These variables, which can be seen in Table 1, include cohabitation status (single or having a partner), children under 18 in the household for which the respondent is the guardian (yes or no), age (in five categories: –24, 25–34, 35–44, 45–54 and 55+), citizenship (native or non-native), education (compulsory, secondary or tertiary), months in unemployment (months) and perceived economic strain in the form of how easily the household is making ends meet (easily, some difficulty, difficulty or great difficulty). In addition to this the article also uses the employment status at time 2 (t2) (unemployed, re-employed or outside the labour market3) for longitudinal analysis.

Methods

In order to investigate the hypotheses the article uses cross-sectional data to look at gender differences in mental health among the unemployed as well longitudinal data in order to investigate the change in mental health when exiting unemployment. An ordinary least squares regression is used for the cross-sectional analysis, with variables entered stepwise in order to see how the effect of gender on mental health is affected by the control variables in the respective countries. This technique is not, however, suitable for the analysis of repeated measures where data from the two time points can be expected to be correlated. For instance, we can expect substantial variation between individuals when it comes to the normal level of mental health. This variation in the normal level of mental health might, if not taken into account, muddle the effect of the change variable re-employment or leaving the labour market. In order to deal with this issue the article uses a repeated-measures linear mixed-models approach with random intercepts. A linear mixed model is an extension of the general linear model that allows error terms and random effects to have correlated variability. The repeated measures mean that the model analyses within-subject effects on repeated observations, and the random intercept is here an assumption that individuals have different intercepts (which was found to be statistically supported in analyses not presented in the article).

Results

Psychological distress among the unemployed

Table 2 shows the standardised GHQ-12 at time 1 (t1) when all respondents were unemployed. Variables are entered stepwise in Table 2, where model 1 includes only gender, model 2 adds variables related to the family situation and their significant interactions with gender and model 3 includes all control variables. Model 1 indicates that while unemployed women in Sweden have worse mental health than unemployed men, unemployed men in Ireland have worse mental health than unemployed women.

Table 2. Ordinary least squares regression of standardised GHQ-12 at time 1 (t1)
 SwedenIreland
Model 1Model 2Model 3Model 1Model 2Model 3
BSEBSEBSEBSEBSEBSE
Woman (man = 0)1.560.521.530.461.730.48−1.740.58−0.080.690.300.81
Partner (single = 0)
 Partner  −2.530.57−1.370.54  3.060.93−0.261.24
 Woman*partner        −3.791.28−0.881.68
 Children (no = 0)  1.020.57−1.200.58  2.060.853.411.14
Age (19–24 = 0)
 25–34 years    0.100.71    0.600.86
 35–44 years    1.160.83    3.451.10
 45–54 years    0.580.85    3.431.21
 55+    −1.160.96    5.211.45
Citizenship (native = 0)
 Other    6.790.84    −0.391.38
Education (compulsory = 0)
 Secondary    −0.830.54    −0.070.75
 Tertiary    0.480.90    0.601.19
 Months unemployed    0.070.02    0.000.00
Making ends meet (ease = 0)
 Some difficulty    4.730.71    0.480.89
 With difficulty    5.430.62    0.770.99
 With great difficulty    8.770.69    4.361.01
 Intercept49.300.3550.340.4644.580.8550.570.3448.530.4547.181.00
 Adj. R20.01 0.02 0.19 0.01 0.04 0.09 

When family-related variables are included in model 2 the statistical effect of gender remains in Sweden but disappears in Ireland. Looking first at the Swedish case in model 2, those who are living with a partner have lower levels of psychological distress than singles. Having children is (with borderline significance) connected to higher levels of psychological distress. None of these variables, however, influences the statistical effect of gender. Turning to the Irish case in model 2 the relationship between gender and mental health not only becomes insignificant but disappears completely when the family-related variables are entered. Cohabitation status is strongly related to psychological distress but in the opposite direction from what was found in Sweden. Unemployed individuals with partners suffer from higher levels of psychological distress than those who are single. There is, however, a strong significant interaction between gender and cohabitation status where the combination of being a woman and having a partner is connected with better mental health. This indicates that the negative relationship does not exist for women. Finally, having children is strongly related to worse mental health in Ireland.

These findings appear to support our second hypothesis. The difference in mental health between unemployed men and women in Sweden is not connected to differences in family position or a difference in the importance of family position. In Ireland, on the other hand, lower psychological distress among unemployed women than among unemployed men can be understood from differences in family position, and partly from the different implication this position has for men and women. Being unemployed and having a partner is strongly connected with worse mental health for men but is not for women. The findings fit the theoretical assumptions of the article about the similarity between the available roles of unemployed men and women in Sweden but the dissimilarity between them in Ireland. This should in turn differentially affect mental health through similar or differential psychosocial and economic need for employment among men and women in the two countries.

When all the available control variables are entered in the final model, model 3, there is some further support that this is the case. If we look at Sweden, the difference between men and women remains, and it is clear that unemployed women in Sweden have higher levels of psychological distress than unemployed men. The importance of family-related variables, however, decreases. The relation between having a partner and psychological distress is roughly halved, although it still remains significantly related to lower psychological distress. The variable, children, is even more strongly affected and switches direction, with having children being connected with lower levels of psychological distress. These changes are connected with the economic situation of the household, which has a very strong relationship with psychological distress among unemployed people in Sweden. The initial findings were caused by unemployed people with partners being less economically strained while unemployed people with children were economically more strained. Looking at Ireland we see no change from the non-relationship between gender and mental health found in model 2. Having children is also, contrary to the case in Sweden, not affected by the introduction of the additional controls. Cohabitation status, however, and the interaction between cohabitation status and gender, becomes insignificant. This change is caused by the role of age and economic situation, where women who are cohabiting are both more economically strained and younger than men who are cohabiting. This would suggest that the better mental health among women in Ireland is related to the differential positions and roles of men and women, but also to the more selected labour market participation of women in Ireland.

If we compare these different gender gaps among unemployed individuals in Sweden and Ireland with what is known about the gender gap in mental health in general, it is clear that our Swedish finding does fit the general pattern, while the Irish finding does not. Mood disorders and anxiety have been found to be more common among women than men across time and space (Seedat et al. 2009). This has been found to be the case for both Sweden and Ireland, where women in general have been found to suffer from psychological distress, sleeping disorders and anxiety to a greater extent than men (Tedstone Doherty et al. 2008, Clevenpalm and Karlsson 2010), although Ireland actually has been found to be among the very few countries with non-significant gender differences in depression scores (Van de Velde et al. 2010).

Change in psychological distress upon re-employment

With these known general differences between men and women in mind the cross-sectional analysis of standardised GHQ-12 at t1 appears to support hypothesis 1. Given the limitations of cross-sectional analysis for investigating causal assumptions, these analyses need to be complemented with a longitudinal analysis of the relationship. If hypothesis 1 is correct there should be similar findings on the change in mental health between t1 and t2. Table 3 shows the longitudinal analyses of change using mixed models for repeated measures. The analysis is done in three models where the initial model is empty, which allow us to see the change in goodness of fit (−2LL) when other variables are entered. Model 2 includes re-employment or leaving the labour market, which represents a change in status as well as the fixed variable4 gender, and model 3 additionally includes interactions between these changes in labour market status and gender.

Table 3. Mixed models standardised GHQ-12 t1 and t2 (repeated measures, random intercept)
  SwedenIreland
Model 1Model 2Model 3Model 1Model 2Model 3
BSEBSEBSEBSEBSEBSE
Change in labour market status
 Re-employment  −4.350.35−4.490.48  −4.510.47−5.530.61
 Outside  −2.430.49−1.910.76  −0.120.87−1.721.46
Fixed variable
 Woman (man = 0)   1.010.42 1.040.46  −1.060.47−1.650.51
Interaction
 Re-employment*woman     0.300.72     2.570.97
 Outside*woman    −0.860.99     2.651.82
 Intercept48.90.2149.50.3049.50.3148.80.2249.90.2950.10.29
 −2LL215302136721364160461595015935

When comparing models 1 and 2 in Table 3 by looking at the drop in −2LL we can see that the inclusion of change in labour market status and gender improves the model. This is because the variables included have strong and significant regression coefficients in both countries in Table 3, model 2. Re-employment leads to very strong and significant drops in the level of psychological distress in both countries. It is thus clear that re-employment in both countries has the same positive relationship with mental health. Leaving unemployment for something outside the labour market, however, is not related with the same change in the two countries. In Sweden it is significantly connected with improved mental health, while no such relationship is found in Ireland. Previous research has shown that different non-employment exit routes from unemployment have radically different consequences for mental health where, for instance, a return to the normal educational system can be very positive and exit because of illness can be very negative (Strandh 2000b). As there is no information in this data set on what the labour market exit implies, and the number who do so in Ireland is about half as large as in Sweden, we should thus not read too much into this difference. There could be, and indeed probably are, systematic differences in what these exit routes imply in the two countries.

Looking at gender, one can see in Table 3, model 2 that gender has the same differential relationship with mental health in the two countries as found in the cross-sectional analysis when the samples were all unemployed. Women in Sweden experience significantly higher levels of psychological distress than men, while women in Ireland experience significantly lower levels of psychological distress than men. In order to see whether the gender differences actually are related to a differential meaning of unemployment and re-employment, however, we have to look at the interaction between gender and change in labour market status. When the interaction terms are included in model 3 there is no such interaction in Sweden. The association between re-employment and leaving the labour market do thus seem to be similar for both men and women, and the conclusion that can be drawn from this is that the higher level of psychological distress among women found in Tables 1 and 2 is not related to their unemployment situation. In Ireland, on the other hand there is a strong interaction between gender and re-employment. For Irish women there is a reduction in the level of psychological distress when they are re-employed, but this improvement is only about half of what men experience. Given the increased importance of gender when the interaction term is included, the statistical effect seems to be related to unemployed women having lower levels of psychological distress in unemployment. Further, in Table 3, model 3 there is a similar, but statistically insignificant effect of leaving the labour market, which seems to be worse for women than for men. As discussed above, the low number of unemployed people in the Irish sample who leave the labour market and our lack of knowledge about the content of such a move mean that one should not draw any conclusion about it.

Psychological distress among the re-employed

The conclusions that we can draw from Table 3 about the relationship between unemployment, gender and mental health in the two countries do strongly support hypothesis 1. In Sweden the difference in mental health between women and men appear to exist equally among both unemployed and re-employed people. Re-employment is also equally positive for both genders. In Ireland women appear to have better mental health than men when they are unemployed, but they do not get the same strong positive boost of mental health from re-employment as men do. That this is the case can be further demonstrated through looking at the simple bivariate statistical effects of gender on the change in standardised GHQ between t1 and t2 and standardised GHQ at t2 among re-employed people only in Table 4. Doing this has the additional advantage of limiting the analyses to a group of initially unemployed individuals that, given their equal success, should be more similar in the two countries than a general sample of the unemployment stock (where the composition is related to the different labour market histories of the two countries).

Table 4. Ordinary least squares regressions of gender and change in standardised GHQ between t1 and t2 as well as standardised GHQ t2, re-employed respondents only (Sweden n = 513, Ireland n = 375)
  SwedenIreland
Model 1 Change t1 to t2Model 2 GHQ- t2Model 1 Change t1 to t2Model 2 GHQ- t2
BSEBSEBSEBSE
  1. Bold lettering represents significance < 0.05 or lower.

Woman (man = 0)0.950.891.260.652.371.140.920.52
Intercept−3.700.5844.680.43−5.040.7444.420.52
R2 adjusted0.00 0.01 0.02 0.00 

Looking firstly at the change in psychological distress for respective country in Table 3, model 1 we see that the conclusions from Table 3 hold also when the analysis is limited to re-employed people. The strongly negative intercept show that level of psychological distress between t1 and t2 in both countries is decreased in re-employed people. There is further no significant difference in this improvement between men and women in Sweden, while women in Ireland have a significantly lower improvement in their level of psychological distress than men. Turning to the actual level of psychological distress among re-employed individuals in Table 3, model 2 we find in Sweden the same pattern as previously found, where women have a significantly higher level of psychological distress. In Ireland, on the other hand, we can see that the lower level of psychological distress among women found in the unemployed sample has disappeared. Looking at the re-employed group we find no significant difference between men and women, and the coefficient indicates that, if anything, women have higher levels of psychological distress. This pattern would seem to fit previous findings on the gender gap in mental health in the general population in both countries and it seems that the gender gap in mental health has reverted to what is normally the case in our re-employed Irish sample.

Conclusions

The main advantage of this study is the longitudinal design with follow ups of unemployed samples in two countries. Thus, we avoided the common problem of statistical power in unemployment research. However, the samples of initially unemployed individuals also bring with it methodological limitations. There is no comparison group of employed people except for those being re-employed within the samples. Samples originally containing both employed and unemployed people could have contributed to a deeper analysis of the reasons behind the various gender differences in the two countries. Another main methodological issue is the comparability of the samples in the two countries. In the Irish sample young women (without childcare responsibilities) and long-term unemployed individuals were overrepresented. Important variables (like education and unemployment) were self-reported in the Irish sample but retrieved from register data in the Swedish sample. Despite these limitations, we claim that our analyses represent an important contribution to research into the relationship between unemployment, gender and mental health.

This is a research area that, although it is present in classic unemployment research, has tended to produce inconclusive results in recent decades. Existing theory suggests that differences in the influence of unemployment on mental health are related to the different positions and roles that are available for men and women in society and the family; roles that are connected with differential psychosocial and economic needs for employment. In line with some previous researchers, the suggestion is that these differences are not static but can change and vary between contexts. This should produce a situation where the relationship between unemployment, gender and mental health can vary over time in one country and could vary between countries that are characterised by different gender relations.

This is investigated using longitudinal data on initially unemployed samples from two countries that represent different gender relations and regimes: Sweden, which represents a strongly defamilised context with high female labour market participation and Ireland, which represents a weakly defamilised context with a historically low female labour market participation. Given these contextual differences, the suggestion is that Ireland fits the classic theoretical assumptions of a differential need for employment between men and women, while Sweden does not. This gave the first hypothesis; unemployment has a stronger negative connection with mental health among men than among women in Ireland but there will be no difference between men and women in Sweden, which was supported. Unemployed women in Ireland were found to have lower levels of psychological distress than men and showed less improvement from re-employment than men. In Sweden unemployed women were initially found to have higher levels of psychological distress than men. There was, however, no difference in the reduction of the level of psychological distress upon re-employment between men and women, and the higher level of psychological distress displayed by women in the unemployed sample was still present upon re-employment.

The second hypothesis was that the expected differences between men and women in Ireland were related to differences in economic and psychosocial need of employment. These differences should be related to the different structural positions of men and women, such as their family and economic situation. This gave the second hypothesis; variables related to the family and economic situation explain the whole or part of the difference in mental health between unemployed men and women in Ireland; however, they will not affect the difference between men and women in Sweden, which was supported overall. Other variables did not affect the statistical relationship in Sweden. In Ireland the lower level of psychological distress among unemployed women disappeared when controlled for family position. An interesting finding was that the role of having a partner was different for unemployed men and women in Ireland. Having a partner was related to increased psychological distress among unemployed men but not unemployed women. This difference would appear to support the assumptions about differential roles of men and women leading to their differential needs of employment and a differential relationship between unemployment and mental health. After the introduction of control variables this relationship could be explained by the economic situation of the household and the age of the respondent.

The overall conclusion that can be drawn from the findings in this article is that the contextual situation does matter for the relationship between unemployment, gender and mental health. It is the difference in roles and positions between men and women on the labour market and in the family that produce gendered differences in the relationship between unemployment and mental health. In a situation where gender relations are characterised by a relative similarity of the roles of men and women there will be no gender difference in the relationship between unemployment and mental health. In a situation where there are substantial differences, however, there will be a differential relationship.

This is caused by two mechanisms, both related to consequences of the gender regime. The first mechanism is related to the differential economic and psychosocial need of employment between men and women caused by a male breadwinner system. If female labour market participation is not encouraged (and female engagement in housework is encouraged), and the woman’s income typically is the secondary income, women will tend to have both a lower psychosocial and an economic need of employment than men. The second mechanism is related to the more selected labour market participation among women than among men caused by the male breadwinner system. If women but not men leave the labour market when they become parents the female labour force will on average be younger, more often single, more often childless and have less economic responsibility than the male labour force. These are all characteristics that might have implications for how damaging an unemployment experience can be.

Footnotes

  • 1

     In this study individuals are regarded as unemployed when they are registered at an employment office.

  • 2

     All analyses in the article have also been made using the non-standardised GHQ-12. As could be expected, the results and conclusions from these analyses turned out to be the same as those using the standardised GHQ-12.

  • 3

     The ‘outside the labour market’ group is heterogeneous and contains very different statuses. It had to be collapsed into one category due to a lack of information in the Irish data. In the Swedish data the content of the category was: 50 per cent in education, 21 per cent in retirement, 12 per cent on parental leave, with 8 per cent ill and 8 per cent unknown inactivity. Given this and the possibility of systematic differences between the samples for this category (an indication that this is the case is that the category is much larger in the Swedish sample; 298 as compared to 103 individuals), the results for this category should be interpreted with caution.

  • 4

     A fixed variable is here a variable that cannot vary between data points and that is assumed to have the same effect at both points in time.

Appendix

Table A1 Items included in the general health questionnaire; psychological distress

  1. Note Each item assesses the severity of a mental problem over the last few weeks using a 4-point Likert scale from 0 (never) to 3 (always). Positive items are corrected from 0 (always) to 3 (never) resulting in a combined index 0–36 where higher scores indicate worse mental health.

You have lost much sleep over worry
You feel that you are playing a useful part in things
You feel constantly under strain
You feel that you could not overcome your difficulties
You feel capable of making decisions about things
You are able to concentrate on whatever you are doing
You are able to face up to your problems
You feel unhappy and depressed
You have been losing confidence in yourself
You are thinking of yourself as a worthless person
You feel reasonably happy, all things considered
You are able to enjoy your normal day-to-day activities

Ancillary