Unemployment, furlough, short hours working and mental health

Psychological research on work and mental health has a long tradition of revealing negative mental health effects of unemployment (see the review by Wood and Burchell, Reference Wood, Burchell and Lewis2018), comparing mostly on the comparisons between being employed and being unemployed. Research shows that at the onset of the COVID-19 unemployment negatively impacted mental health (Li and Wang, Reference Li and Wang2020). However, during the COVID-19 pandemic the labour market policy measures initiated by many countries have created another new group – the furloughed employees that continued to be paid their full or partial wages without workingFootnote ¹ for their main employer. These measures have also expanded the number of employees whose working hours were reduced. The theoretical perspectives discussed below suggest distinct mental health implications of belonging to one of these four employment status groups – out of paid work, working reduced hours, being furloughed, and continuing to work full-time during the COVID-19 pandemic, an issue that has been so far neglected in the research.

A large body of research, collected well before the COVID-19 pandemic, shows that unemployment brings multiple negative short and longer-term individual and societal consequences, contributing to poverty and social inequality. These include social exclusion, financial losses, lower living standards, and a decline in the well-being, physical and mental health of unemployed people and their families (McKee-Ryan et al., Reference McKee-Ryan, Song, Wanberg and Kinicki2005; Paul and Moser, Reference Paul and Moser2009; Wood and Burchell, Reference Wood, Burchell and Lewis2018). Although some people with worse mental health are more likely to become unemployed, robust longitudinal studies and meta-analyses also suggest that becoming unemployed also leads to a decline in mental health (Jefferis et al., Reference Jefferis, Nazareth, Marston, Moreno-Kustner, Bellan, Svab, Rotar, Geerlings, Xavier, Goncalves-Pereira, Vicente, Saldivia, Aluoja, Kalda and King2011; McKee-Ryan et al., Reference McKee-Ryan, Song, Wanberg and Kinicki2005; Paul and Moser, Reference Paul and Moser2009; Wanberg, Reference Wanberg2012). The effect sizes are larger than most other common stress factors such as divorce, and (unlike many other stressors) the effects of unemployment hardly wear off if an individual remains unemployed for longer periods of time (Burchell, Reference Burchell2011). A meta-analysis suggests that in Western countries unemployment affects both men and women, but the effect tends to be stronger for men and for blue-collar workers (Paul and Moser, Reference Paul and Moser2009). There are, of course, some individual differences (McKee-Ryan et al., Reference McKee-Ryan, Song, Wanberg and Kinicki2005) – nevertheless, many of the findings on the average effects of unemployment on mental health have been replicated so widely across time and across countries that they can be stated with little controversy (Paul and Moser, Reference Paul and Moser2009). Therefore, we can hypothesise that

The theoretical reasoning that explains the dependency of mental health on employment for mental health is slightly more controversial than the empirical evidence. Marie Jahoda’s socio-environmental model of employment (Jahoda, Reference Jahoda1982) is the most influential of the theories, even in contemporary debates (Selenko et al., Reference Selenko, Batinic and Paul2011). Jahoda claimed that in addition to the manifest reason for working – wage – it was the accidental or latent consequences of working that were associated with the psychological benefits of working for mental health. According to her, employment is more than a source of income, supplying several latent socio-psychological benefits such as providing time structure, collective purpose and social contacts, identity and activity. The loss of these benefits due to unemployment damages both well-being and mental health. Some studies suggest the loss of latent benefits is the most important (Winkelmann and Winkelmann, Reference Winkelmann and Winkelmann1998); others find that income has the largest negative impact (Ervasti and Venetoklis, Reference Ervasti and Venetoklis2010; Paul and Batinic, Reference Paul and Batinic2010).

Many other more recent psychological models can be seen as refinements of Jahoda – for instance, by adding to the list of psychological benefits (Warr, Reference Warr1987), making more nuanced differentiations between good jobs and bad jobs (Warr, Reference Warr, Kahneman, Diener and Schwarz1999), adding individual differences in psychological and economic needs (Nordenmark and Strandh, Reference Nordenmark and Strandh1999), or proposing to expand understanding of workers’ well-being to reflect the broad importance of work in human life (Budd and Spencer, Reference Budd and Spencer2015). Fryer (Reference Fryer1986, Reference Fryer and McLaughlin1992) emphasises the importance of employment in empowering individuals to plan their lives, with the unemployed and workers in precarious jobs being deprived of this agency.

The current limitation of this line of theoretical reasoning and body of evidence is that it does not reflect more contemporary policy developments, especially furlough introduced during the COVID-19 pandemic. Recent research by Kamerāde et al. (Reference Kamerāde, Wang, Burchell, Balderson and Coutts2019), performed in the UK and replicated in all EU countries (Kamerāde et al., Reference Kamerāde, Buchell, Wang, Balderson and Coutts2021) suggested that the employment threshold for good mental health was about one day a week – above that, there was little difference to individuals’ wellbeing regardless of whether they worked eight hours or 48 hours a week – the mental health varied little, and in all categories the mental health was markedly better than those with zero hours a week, either due to unemployment or to economic inactivity.

These findings might be highly relevant to the immense labour market changes during the COVID-19 pandemic. Kamerāde et al. (Reference Kamerāde, Wang, Burchell, Balderson and Coutts2019) findings suggest that avoiding exclusion from paid work should be a top priority as a labour market policy. They also suggest that there may be a plausible way of doing this with relatively little damage to the mental health levels of the nation, through short-time working. In contrast to being unemployed, working reduced hours still provides the workers with some, albeit reduced, access to both manifest and latent benefits of paid work. We therefore could hypothesise that

The experience of being paid but having no work for several months (i.e. being furloughed) is a novel phenomenon for which we have no knowledge base, except for the somewhat relevant McKenna and Fryer’s (Fryer and McKenna, Reference Fryer, McKenna and Fineman1987; McKenna and Fryer, Reference McKenna and Fryer1984) small scale study of the temporary laid off male factory workers. This study found that compared to the men made unemployed, temporarily laid-off men had significantly better mental health. However, COVID-19 pandemic furlough was different in three ways. First, the laid off men in McKenna and Fryer’s study knew that they were laid off for seven weeks, while furloughed workers were uncertain about the length of their furlough as the end date for the scheme changed several times, often with a very short notice period. While many European schemes required employers receiving support not to make people redundant before a certain date, the UK furlough scheme was not limited by such constraints. Consequently, despite the protection offered by the scheme, half of the employees being furloughed felt that the scheme was not contributing adequately to job security, were concerned whether they would actually return to work after the furlough period or whether they would be made redundant (CIPD, 2020). Second, the laid-off men signed up as unemployed and received welfare benefits, while furloughed workers continued to be paid at least 80% of their salary. So being furloughed still provided the workers with the manifest benefit of income but limited access to most of latent benefits of working, such as social contacts, time structure, and it also was associated with some job insecurity. Third, the laid-off men engaged in a wide variety of activities in their newfound spare time such as travel, sport, and socializing. Many of these activities were banned during the COVID-19 lockdowns and many furloughed workers (more so women than men) had childminding and home educating forced upon them. Therefore, we hypothesise that

Gender differences

In the UK there are still marked gender differences in labour market experience (Kamerāde and Richardson, Reference Kamerāde and Richardson2018) and the division of household responsibilities (Kan and Laurie, Reference Kan and Laurie2018). Thus, it is important to explore whether the mental health impacts of (un)employment experience at the onset of COVID-19 vary between men and women. Drawing on role conflict theory, we argue that men are more likely to benefit from furlough, shorter working hours and other forms of employment than women at the onset of COVID-19.

The role conflict theory argues that when individuals have multiple and overlapping social role responsibilities, compliance with one social role may make it difficult to comply with the other (Greenhaus and Beutell, Reference Greenhaus and Beutell1985). This is especially the case when people have responsibilities at both work and family domains, which are thought to have different rules, thinking and behavioral patterns, and thus the role pressures from both domains are often mutually incompatible. At the onset of the COVID-19, virus containment measures, such as closing schools, restricting mixing of different households and government orders to work from home where possible, blurred the boundaries between work and family. Given the persistent gender inequalities in household division of labour, the blurred work-family boundaries could significantly increase women’s housework burden and mental stress during the pandemic. For example, women were not only expected to continue to work, but also needed to raise and educate their children at home and care for elderly without institutional and community support (Power, Reference Power2020; Sevilla and Smith, Reference Sevilla and Smith2020), as well as to provide emotional support to their partners and other family members (Chung, Reference Chung2020). During COVID-19 pandemic, the additional care work of women could worsen their mental health (Chandola et al., Reference Chandola, Kumari, Booker and Benzeval2020; Wade et al., Reference Wade, Prime, Johnson, May, Jenkins and Browne2021).

Therefore, it is expected that furlough, reduced working time and other forms of employment may not be able to help maintain women’s mental health level at the onset of the COVID-19 because they could exacerbate gender inequalities in the division of care responsibilities. Also, given women’s lower labour market status (Wang et al., Reference Wang, Coutts, Burchell, Kamerāde and Balderson2020), furlough and reduced working hours might have been more likely to induce job insecurity, job-related anxiety and stress in women than men and that could have overridden their protective benefits in terms of mental health. In contrast, men who had less involvement in additional care work created by the pandemic were less likely to experience inter-role conflicts and could obtain more mental health benefits from furlough, reduced working time and other forms of employment during COVID-19. Therefore, we hypothesise that

Variables

Changes in mental health - the key dependent variable – were measured using the General Health Questionnaire (GHQ-12), a validated scale widely used in the community or non-clinical settings to measure the levels of general psychiatric disorders (Goldberg and Williams, Reference Goldberg and Williams1988). GHQ-12 included 12 questions about respondents’ depressive symptoms, anxiety symptoms, sleeping problems confidence and overall happiness etc. They were measured on a four-point scale (0 ‘less than usual’, 1 ‘no more than usual’, 2 ‘rather more than usual’, and 3 ‘much more than usual’). The answers to the 12 questions were then summated to obtain a GHQ-12 Likert score (0-36) which was reversed so that a higher score indicated better mental health and calculated GHQ-12 mental health change between Wave 10 and April 2020, where a positive score denoted improved mental health, whereas negative scores denoted worsened mental health.

Changes in employment status are our key independent variable. We combined information from questions about individuals’ employment status before COVID-19 pandemic (UKHLS Wave 10 from 2018 to February 2020) and at the onset of COVID-19 pandemic (April 2020) and created five categories: ‘Left/lost paid work’, ‘Furloughed under COVID-19 job retention scheme’, ‘Remained part-time employed (1-34 hours per week)’, ‘From full-time to part-time employed’, and ‘Remained full-time employed (35-48 hours per week)’. Detailed economic activities were only reported in UKHLS wave 10, whereas in COVID-19 study wave 1 we can only distinguish between employment, self-employment, and no paid work. We checked that those who were furloughed did not actually work any hours, as some research suggests some employees continued to work for the same employer while being on furlough (Adams-Prassl et al., Reference Adams-Prassl, Boneva, Golin and Rauh2020a), despite that violating the rules of furlough in place till July 2020¹.

We controlled for several demographic and socioeconomic factors identified in the COVID-19 survey dataset: age, whether living with a partner and presence of children in household (‘No children’, ‘Children aged 0-4’, ‘Children aged 5-11’, ‘Children aged 12-15’ and ‘Children aged 16-18’). In addition, we also controlled for logged changes in household income before and at the onset of the COVID-19 pandemic and for UKHLS Wave 10 interview year to consider the period effects. Finally, to explore whether housework responsibilities can mediate gender differences in employment mental health benefits at the onset of the pandemic, we also included housework hours per week at the onset of the COVID-19, changes in housework hours per week before and at the onset of the COVID-19 (a positive score indicates an increase in housework burden) and weekly hours spent on childcare and home schooling.

Analytic strategy

Using Chi-Squared test, independent samples T-test and paired samples T-test, we report the bivariate descriptive statistics of the main variables included in the study and sample characteristics. We compared mental health scores before and at the onset of the pandemic by the changes in employment status for the entire sample and separately for men and women.

To measure the dynamic impact of employment status changes due to COVID-19 pandemic on mental health we used change score analysis for men and women. Understanding Society as panel data offers a unique opportunity enabling observation and analysis of change and so that mental health outcomes are measured before and after a specific period, which in our case is before and at the onset of the COVID-19. To fit a change score model, we first calculated the difference in mental health between wave 10 and April 2020 as outcome and the difference in employment status as predictor variable. Next, we used Ordinary Least Squared regression analysis to investigate the relationship between changes in employment status and mental health while controlling for a wide range of demographic and socio-economic characteristics. As the measure of employment status is different before and at the onset of the COVID-19, we are not able to use fixed effects models to fully explore the within-individual variation. Instead, a less detailed employment status variable is tested with fixed effects model as a robustness check (see Table A4 in Appendix). Nevertheless, by maximizing the utilization of the available longitudinal data in the UK, using the change score model could facilitate a deeper understanding of the dynamic relationship between changing employment status and mental health, and reduce endogeneity problems.

Robustness checks

To explore the heterogeneity and ensure robustness of the results, we have conducted several robustness check analyses. First, we examined whether transition in home working status influences mental health changes for men and women. Consistent with our main findings, Table A1 shows that compared with those left/lost paid work, furlough and different forms of home working status can significantly improve men’s mental health but not women’s mental health. Further interaction analyses also suggest that the mental health benefits of various forms of employment transitioning into or out of home working are significantly larger for men than women. Second, instead of controlling for interview year in the UKHLS wave 10 survey, we have calculated the number of months elapsed from UKHLS wave 10 to COVID study wave 1 and control for it in the models in Table A2. Reassuringly, we find that this does not make any differences, suggesting that our models are robust to alternative variable specification. Third, Table A3 explores whether keyworker status can help explain women’s worse mental health status. However, the results show that keyworker status is not a significant predictor of mental health changes and its interaction with gender is not significant as well. This suggests that women’s worse mental health at the onset of the pandemic cannot be attributed to their keyworker status. Fourth, in order to ensure a more robust causal inference, Table A4 uses fixed effects models to examine a less detailed measure of employment status because the measure of furlough is only available in the COVID-19 study. Consistent with our main findings, it shows that compared with no paid work, part-time and full-time employment can significantly improve men’s mental health but not women’s mental health.