The length of unemployment predicts mortality, differently in men and women, and by cause of death: A six year mortality follow-up of the Swedish 1992–1996 recession

https://doi.org/10.1016/j.socscimed.2012.01.034Get rights and content

Abstract

This study examines the relationship between the total amount of accumulated unemployment during the deep Swedish recession of 1992–1996 and mortality in the following 6 years. Nearly 3.4 million Swedish men and women, born between 1931 and 1965 who were gainfully employed at the time of the 1990 census were included. Almost 23% of these individuals were unemployed at some point during the recession. We conduct a prospective cohort study utilizing Cox proportional hazard regression with a mortality follow-up from January 1997 to December 2002. We adjust for health status (1982–1991), baseline (1991) social, family, and employer characteristics of individuals before the recession. The findings suggest that long-term unemployment is related to elevated all-cause mortality for men and women. The excess mortality effects were small for women and attributable to a positive, linear increase in the hazard of alcohol disease-related mortality and external causes-of-death not classified as suicides or transport accidents. For men, the excess hazard of all-cause mortality was best represented by a cubic, non-linear shape. The predicted hazard increases rapidly with the shortest and longest accumulated levels of unemployment. However, the underlying pattern differed by cause-of-death. The cancer, circulatory, and alcohol disease-related analyses suggest that mortality peaks with mid-levels of accumulated unemployment and then declines with longer duration unemployment. For men, we observed a positive, linear increase in the hazard ratios associated with transport and suicide mortality, and a very steep non-linear increase in the excess hazard ratio associated with other external causes of death that were not classified as suicide or transport accidents. In conclusion, mortality risk increases with the duration of unemployment among men and women. This was best described by a cubic function for men and a linear function for women. Behind this pattern, different causes-of-death varied in their relation to the accumulation of unemployment.

Highlights

► The length of unemployment predicts mortality, differently in men and women, and by cause of death: A six year mortality follow-up of the Swedish 1992–1996 recession. ► Long-term unemployment is related to elevated all-cause mortality for Swedish men and women. ► This was best described by a cubic function for men and a linear for women. ► Behind this pattern, different causes-of-death varied in their relation to the accumulation of unemployment.

Introduction

The central disagreement in most individual level studies of unemployment and mortality has focused on whether the relationship between job loss and death is causal or spurious (Gerdtham & Johannesson, 2003; Martikainen, 1999; Martikainen, Mäki, & Jäntti, 2007; Martikainen & Valkonen, 1996, 1998; Stewart, 2001; Sullivan & von Wachter, 2009).1 Many have argued that health selection or other unmeasured characteristics of individuals explain a substantial part of the association (Martikainen & Valkonen, 1996, 1998; Martikainen et al., 2007) or, at least some part of it (Stewart, 2001). However, the evidence considered together also supports the interpretation that unemployment may cause excess mortality among those who lose their jobs in periods of deep recession when prior poor health or other individual level characteristics are less likely to be a major cause of unemployment.

Prevailing macro-economic conditions also influence the duration of unemployment. Here, we examine whether the length of unemployment (i.e., the total amount of accumulated joblessness during the recession) predicts excess mortality in a dose–response fashion. It is probable that the duration of an unemployment spell partially determines whether the loss of control and associated stress is acute, chronic, or a combination of both. Material consequences of long-term unemployment are also likely to be more severe and long-lasting than those of short-term unemployment.

There are few recent or past studies that have examined the duration of unemployment and the relationship spell length has to an excess risk of death. A Danish study suggested that the mortality risk of unemployment increased as spell duration increased for both men and women under the age of 61 (Osler et al., 2003). A Swedish cohort study that calculated gender and age adjusted mortality rate ratios found higher mortality for men who experienced long-term unemployment. Mortality rate ratios were particularly elevated among younger cohorts of men but not for unemployed women (Stefansson, 1991). A Canadian community level study suggested that a higher rate of community wide, long-term unemployment was associated with higher community level mortality and lower life expectancy (Safaei, 2008).

Far more numerous are studies suggesting that unemployment of longer duration affects mental health and health behavior more negatively (e.g., increased cortisol levels, more frequent bouts of depression, and more severe anxiety, more dinking and smoking) than unemployment of shorter duration or full employment (Iversen & Sabroe, 1988; Maier et al., 2006; Montgomery, Cook, Bartley, & Wadsworth, 1998; Stankunas, Kalediene, Starkuviene, & Kapustinskiene, 2006; Wadsworth, Montgomery, & Bartley, 1999). There is often an implied or stated assumption that the risks of mortality or negative health effects of unemployment vary according to the length of an individual spell or by the total amount of unemployment accumulated over the life course (Bartley, 1994; Bjorklund & Eriksson, 1998). However, almost without exception, researchers have utilized definitions of long-term unemployment that are categorical and almost exclusively dichotomous. Hence, threshold values that are used to distinguish between short and long-term spells vary widely in most studies. We reason that the excess mortality effects of unemployment can be better assessed by an interval level measure of unemployment duration. Finally, it is important to point out that our study examines how unemployment duration during a specific and unusually deep Swedish recession may cause mortality after that recession.

Job loss that is not resolved immediately results in the loss of income and can rapidly impose subsequent economic hardship (Hammermesh, 1989; Ruhm, 1991). Acute mental stress and material deprivation are likely to arise among many displaced workers after initial job loss, as uncertainty about immediate and future life prospects are heightened. This may have serious consequences among workers who are vulnerable, due to social or health circumstances. For example, some studies have suggested that mortality risks were elevated among workers who were younger and not close to retirement (Osler et al., 2003; Sullivan & von Wachter, 2009). However, these same studies also suggested that for those individuals who were near retirement age, were employed in less desirable jobs, or had quickly returned to work, the health and mortality effects of recently occurring job loss appeared to be small or non-existent (Osler et al., 2003; Strully, 2009; Sullivan & von Wachter, 2009).

For individuals who accumulate unemployment experience and cannot find re-employment quickly, it is probable that a state of acute stress and hardship transforms into a chronic state. Long-term unemployment is likely to be a burden both economically and psychologically. It has been suggested that job loss occurring during mass, firm level layoffs, or during deep recessions is less likely to engender a high level of social stigma (Martikainen & Valkonen, 1996). Loss of social status and accompanying shame may be less problematic for most because unemployment experience is more widespread. However, this interpretation could be modified to suggest that as unemployment duration increases, social stigma and shame are likely to increase as many friends and peers who share the initial experience return to work. Coupled with increased financial strain, depression, anxiety, or other mental health problems might become more severe if they pre-exist. There is empirical support from several longitudinal studies for the hypothesis that longer duration unemployment is not only unpleasant from a psychological standpoint but may also be related to unhealthy coping behavior including increased drinking and smoking.

For example, Iversen and Klausen (1981) found an increased use of medicine and reported psychological problems among a third of Danish shipyard workers who had been laid off for three to four months. At a scheduled one year follow-up, those with the longest duration unemployment reported more frequent medicine use. The frequency of psychological and psychosomatic symptom reporting was elevated as well. These findings are consistent with the hypothesis that longer unemployment duration could trigger and/or exacerbate stress and mental health problems. In related studies, Iversen and Sabroe, 1987, Iversen and Sabroe, 1988 examined shipyard closure over a later three year period and concluded that unemployment led to poorer mental well being but improved if re-employment followed. Perhaps more importantly, findings also suggested that the period prior to layoff had strong negative effects on mental outlook. The fear of job loss led to psychological problems rather than psychological problems leading to unemployment.

Other longitudinal studies have shown that unemployment was associated with increased drinking (Catalano, Dooley, Wilson, & Hough, 1993; Montgomery et al., 1998; Mossakowski, 2008), problem drinking (Montgomery et al., 1998); more frequent heavy drinking (Mossakowski, 2008) uptake of drinking (Gallo, Bradley, Siegel, & Kasl, 2001); a higher rate of smoking, a greater risk of beginning smoking (Hammarstrom & Janlert, 1994; Montgomery et al., 1998), and a lower probability of quitting smoking (Kriegbaum, Larsen, Christensen, Lund, & Osler, 2011). Importantly, all of these studies employed controls for alcohol and tobacco consumption prior to the period of unemployment, clearly strengthening inferences about the direction of causation.

However, predicting the health effects of changed drinking and smoking habits as unemployment experience accumulates is likely to be complicated. For example, heavy binge drinking and chain smoking might immediately follow job loss among those who already drink and smoke. However, the extent to which these individuals could continue over the long-term while unemployed would partly depend on personal resources, state transfer payments, and the price of alcohol and tobacco. At some point dwindling resources for those who are jobless long-term might limit consumption, especially among those who are closest to their economic margin (Bartley, 1994). There is some evidence to support this idea with respect to drinking. Montgomery et al. (1998) found that among young men who had experienced unemployment of longer duration, only those men who experienced one to 36 months of accumulated unemployment had elevated risks of problem drinking. These risks were not elevated among those with total accumulated unemployment in excess of 37 months.

The bulk of reviewed studies examining unemployment duration and health effects focus on somatic or psychosomatic health while fewer examine mortality. Many of these studies have shown that health effects vary according to accumulated unemployment experience, both more recent and longer-term. However, it is unclear in the majority of studies how threshold values were chosen to distinguish between different levels of accumulated unemployment experience. It is also worth noting that the current literature does not suggest whether the mortality risks of job loss are distributed similarly for different causes of death over increasing levels of unemployment exposure.

Section snippets

Study population

This study included all Swedish men and women (n = 3,392,169) who were born between 1931 and 1965, and who were still alive on Dec 31st, 1996. All individuals were also gainfully employed at the time of the 1990 census. This limited the inclusion of individuals with a tenuous, pre-existing connection to the labor market and likely retirees.

Information about these individuals was extracted from The Work and Mortality Database (WMD) held at CHESS, Center for Health Equity Studies in Stockholm.

Methods

Our investigation examines several specific causes of death in addition to all-causes of mortality. The purpose of the analysis was to test whether there was a dose–response relationship between unemployment and predicted mortality. We specified linear, quadratic, and cubic models. The quadratic and cubic products of the accumulated unemployment measure (TPD) were computed so that a test for non-linearity in the hazard of mortality among individuals with longer duration unemployment could be

Males

The fully adjusted analysis suggests that the relationship between excess all-cause mortality and accumulated unemployment exposure was best represented by a cubic, non-linear, serpentine shape (Fig. 2).

Log likelihood ratio testing shows improved global model fit with the addition of quadratic and cubic unemployment duration terms (p ≤ .0001; p ≤ .046) (Table 4).

Shorter duration unemployment of up to 768 days (hazard ratio = 1.32; 95% C.I. = not calculated) was associated with a steep increase

Discussion

Unemployment of the longest duration is associated with the highest all-cause mortality risk for both men and women. The shape of this relationship is different for men and women, but in both cases there is nearly a continuous increase in mortality risk as unemployment experience increases. However, all-cause-mortality reflects the sum of risks for a number of specific causes-of-death. A separate inspection of each cause-of-death suggests that there is no general law that allows us to predict

Conclusion

The findings presented in this study provide additional support for a causal interpretation of the association between unemployment and mortality. The all-cause mortality effects of unemployment, especially long-term unemployment, were more pronounced among men than among women. For men, a larger number of specific causes were also involved. Men often have better jobs and receive higher compensation than women. Therefore, men and women may respond to job loss differently. Although female

Acknowledgments

The research was supported by a grant from Forskningsrådet för Arbetsliv och Socialvetenskap (FAS), project # 1017720. We also wish to acknowledge the two anonymous reviewers.

References (37)

  • M. Eliason et al.

    Does job loss shorten life?

    The Journal of Human Resources

    (2007)
  • P. Fagan et al.

    Cigarette smoking and quitting behaviors among unemployed adults in the United States

    Nicotine Tobacco Research

    (2007)
  • W.T. Gallo et al.

    The impact of involuntary job loss on subsequent alcohol consumption by older workers: findings from the health and retirement survey

    Journals of Gerontology Series B-Psychological Sciences & Social Sciences

    (2001)
  • A. Hammarstrom et al.

    Unemployment and change of tobacco habits – a study of young people from 16 to 21 years of age

    Addiction

    (1994)
  • D. Hammermesh

    What do we know about displacement in the United States?

    Industrial Relations

    (1989)
  • L. Iversen et al.
    (1981)
  • L. Iversen et al.

    Plant closure, unemployment and health: Danish experiences from the declining ship-building industry

  • L. Iversen et al.

    Psychological well-being among unemployed and employed people after a company close-down: a longitudinal study

    Journal of Social Issues

    (1988)
  • Cited by (0)

    View full text