You are here

Article Text

Article menu

Download PDFPDF

Other topics
Do macroeconomic contractions induce or ‘harvest’ suicides? A test of competing hypotheses
  1. Alison Gemmill1,
  2. April Falconi2,
  3. Deborah Karasek2,
  4. Terry Hartig3,
  5. Elizabeth Anderson2,
  6. Ralph Catalano2
  1. 1Department of Demography, University of California, Berkeley, Berkeley, California, USA
  2. 2School of Public Health, University of California, Berkeley, Berkeley, California, USA
  3. 3Institute for Housing and Urban Research, Uppsala University, Uppsala, Sweden
  1. Correspondence to Alison Gemmill, Department of Demography, University of California, Berkeley, Berkeley, California 94720, USA; alisongemmill{at}gmail.com

Abstract

Background Researchers often invoke a mortality displacement or ‘harvesting’ mechanism to explain mortality patterns, such that those with underlying health vulnerabilities die sooner than expected in response to environmental phenomena, such as heat waves, cold spells and air pollution. It is unclear if this displacement mechanism might also explain observed increases in suicide following economic contraction, or if suicides are induced in persons otherwise unlikely to engage in self-destructive behaviour. Here, we test two competing hypotheses explaining an observed increase in suicides following unemployment—induction or displacement.

Methods We apply time series methods to monthly suicide and unemployment data from Sweden for the years 2000–2011. Tests are conducted separately for working age (20–64 years old) men and women as well as older (aged 65 years and older) men and women.

Results Displacement appeared among older men and women; an unexpected rise in unemployment predicted an increase in suicides 6 months later, followed by a significant decrease 8 months later. Induction appeared among working age men, but not among working age women; an unexpected rise in unemployment predicted an increase in suicides 4–6 months later.

Conclusions Displacement and induction both appear to have operated following unexpected labour market contractions in Sweden, though with different population segments.

  • SUICIDE
  • UNEMPLOYMENT
  • TIME-SERIES

https://doi.org/10.1136/jech-2015-205489

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Introduction

    Much research asks whether macroeconomic conditions affect human health and well-being. While those studying the relationship debate the direction and magnitude of the association between the economy and somatic health, most agree that psychological health positively correlates with economic indicators.1–3 In particular, suicide appears to increase during or soon after economic downturns.4–9 Mechanisms explaining this observed increase, however, remain poorly understood.

    The economic antecedents of suicide have been the subject of much research dating to at least the 19th century.10 Theory and literature suggest that two pathways may connect economic shocks to suicides. In the mortality induction pathway, economic contraction leads to increased psychological vulnerability in a population, which, in turn, induces suicide in persons otherwise unlikely to engage in self-destructive behaviour.7 This hypothesis implies a temporary increase above the endemic level of suicides.

    In the mortality displacement pathway, a contracting economy ‘capitalises’ on latent or underlying psychological vulnerability in the population, triggering suicides that would have occurred eventually.7 ,11 As with the induction pathway, one would observe a temporary increase in the number of suicides relative to some baseline, but this surplus would then be followed by a drop below the expected number of suicides. Economic contraction may, therefore, concentrate or ‘harvest’ suicides that would have happened later had the fear or experience of economic adversity not accelerated the process by which self-destructive behaviour emerges.

    While no study has explicitly pitted these competing mechanisms against each other, individual-level risk factor research suggests that job loss may induce suicide.4 ,7 ,12 ,13 Consistent with mortality induction, job loss reportedly increases the risk of suicide after controlling for the mental health status of study participants prior to economic contraction.4 A recent review of the literature, however, questions the validity of the mental health measurements in such studies.7 Support for the induction hypothesis therefore remains equivocal.

    To the best of our knowledge, studies invoking the mortality displacement pathway have been limited to investigations of the effect of heat waves, cold spells and air pollution.14 ,15 None of the work has focused on suicide as an outcome and the economy as an ambient stressor. The mortality displacement research uses, moreover, shorter units of time (ie, days or months) than the work studying the association between the economy and suicide, which typically uses annual data.

    Here, we determine whether mortality induction or displacement better fits the response of suicides to economic contraction, using 144 months of Swedish data that include the recent global recession. More specifically, we estimate the response of suicides to unexpected changes in unemployment for working-age men and women as well as older men and women. Both mortality induction and displacement predict a positive association between unemployment and the incidence of suicide. The mortality displacement argument, however, also predicts that a negatively signed association will appear after a positively signed coefficient.

    We estimate the age-specific and gender-specific response of suicides to unemployment because research suggests that these groups may respond differently to environmental stressors.16 Studies of gender differences in risk factors for suicide found, for example, higher odds of suicide in unemployed males but not females.17 ,18 The effects of economic contraction may, moreover, extend to those not in the labour force. Older individuals, for example, exhibit higher rates of suicide compared to younger persons, as well as higher prevalence of self-reported hopelessness and suicidal ideation.19 The ‘miasma’ associated with economic decline may exacerbate these risk factors for suicide, not least through the loss of stable employment by which individuals and families commonly meet their needs.

    Methods

    Data: We use data from Sweden because epidemiologists judge Swedish suicide counts relatively reliable.20 ,21 We obtained these counts by gender and age for the months January 2000–December 2011 from the Swedish National Board of Health and Welfare.22 We constructed four time series from these data—monthly suicide counts among working age (ie, 20–64 years old) and older (ie, aged 65 years and older) men and women. We extracted Swedish unemployment data for the same months from an on-line resource.23

    Consistent with Galton's definition of empirical correlation,24 our test turns on whether the observed number of monthly suicides differs, as predicted by either induction or displacement, from the values expected under the null hypothesis of no association between unemployment and suicide. Such tests typically assume that the mean of the observed values is the expected value of the dependent variable under the null hypothesis. Suicide, however, exhibits trends, cycles (eg, seasonality) and a tendency to remain elevated or depressed after high or low values. These patterns, referred to collectively as ‘autocorrelation’, complicate hypothesis tests because the mean of an autocorrelated series is not its expected value. No one would expect, for example, that the next observed value of a series previously exhibiting secular trend would equal its mean.

    Since Fisher’s seminal work on crop yields,25 researchers have responded to the autocorrelation problem by estimating the value of a series at time t as a function of the value observed at time t-n. This approach ‘models’, for example, seasonality in monthly suicides by estimating the expected value at month t as a function of the observed value at month t-12. The residuals (ie, differences between the expected and observed values) from a model that specifies autocorrelation in a series will exhibit a mean of 0 and no autocorrelation and thereby meet the assumptions of statistical testing. The analyst can, therefore, add independent variables suggested by theory to the model to determine if their coefficients differ from zero in the hypothesised direction. This strategy not only complies with Galton's and Fisher's logic, but also precludes type I errors (ie, false rejection of the null) that arise from shared trends, seasonality, or other forms of autocorrelation.

    Our test proceeded by applying the following steps to the four gender-by-age groups:

    1. We implemented Fisher’s strategy using the method developed by Box and Jenkins,26 to identify and model autocorrelation in the monthly incidence of suicide for each of the test groups. The method, Autoregressive, Integrated, Moving Average (ARIMA) modelling, has been used by epidemiologists27 to identify which of a very large family of mathematical expressions best predicts measurements made over time. The method assumes that the measurements passed through an unobserved filter that imposed autocorrelation on them. The approach looks for patterns in the data that narrow the search to likely filters and then applies estimates of ‘fit’ to identify the best fitting. These filters, or models, express autocorrelation in terms of integration (ie, trends or cycles), moving averages (ie, a sequence of high, low or oscillating values after an outlier), or autoregression (ie, regression to the mean after high or low values). These models yield expected values that approximate what would have occurred if the circumstances that produced earlier values remained unchanged. The residuals of the model estimate the degree to which monthly suicides among Swedes in the age-by-sex-test group fell above or below expected levels.

    2. We applied the same Box-Jenkins methods to monthly unemployment rates to estimate residuals that gauge the degree to which each of the months differed from its expected value.

    3. We specified the test equation by adding the residuals from the unemployment model estimated in step 2 to each of the four models of age and gender-specific suicide estimated in step 1. We specified the unemployment residuals occurring in the same month as suicides as well as for the 12 previous months to ensure we allowed a substantial period in which to detect any delayed associations. Our general test equation was, therefore, as follows: Embedded Image

    Yt is the number of gender and age-specific suicides in Sweden during month t; c is a constant; Xt is the residual for month t of the best fitting Box-Jenkins model of the monthly unemployment rate in Sweden; ω1 through ω13 are estimated coefficients; θ is the moving average parameter; φ is the autoregressive parameter; B is the ‘backshift operator’ that yields the value of the series it conditions at time t-p for the autoregressive parameter or t-q for the moving average parameter; at is the residual for month t.

    If bad economic times displace suicides, at least one estimated ω would significantly exceed 0 (ie, positively signed and at least twice its SE) and precede at least another ω significantly less than 0. One or more positively signed and significant ω’s with no following negatively signed and significant ω would imply that bad economic times in Sweden did not displace suicides otherwise expected to occur, but, consistent with induction, increased suicides above the number otherwise expected.

    Results

    Step 1, deriving the expected values of suicides for each of the four age-by-gender groups yielded the series shown in figures 1⇓⇓–4. Monthly suicide counts among working age and older men ranged, respectively, from 20 to 54 (mean=37.00) and from 5 to 27 (mean=14.35). Suicides among working age and older women ranged from 8 to 32 (mean=19.24) and from 1 to 12 (mean of 5.64), respectively.

    Figure 1
    Figure 1

    Observed and expected number of monthly suicides among working age (20–64) males in Sweden (2000–2012).

    Figure 2
    Figure 2

    Observed and expected number of monthly suicides among older (65 and over) males in Sweden (2000–2012).

    Figure 3
    Figure 3

    Observed and expected number of monthly suicides among working age (20–64) females in Sweden (2000–2012).

    Figure 4
    Figure 4

    Observed and expected number of monthly suicides among older (65 and over) females in Sweden (2000–2012).

    Figure 5 shows the expected and observed monthly unemployment rate for Sweden during the test period. As expected, the effects of the global recession appear in the later months of the series. The rate ranged from 4.6% to 9.8% (mean=6.9%).

    Figure 5
    Figure 5

    Observed and expected monthly unemployment rate in Sweden (2000–2012).

    Table 1 shows the results of estimating the test equation shown in step 3 above for the four age-by-gender groups. The findings appear to support induction among working age men, but displacement among older persons of both genders. Among working age men, unemployment at month t positively correlated with suicides at months t+4 and t+6. Among older men and women, unemployment at month t positively correlated with suicides at month t+6. Eight months after unexpected changes in unemployment, however, suicides moved opposite the shift seen at 6 months. An unexpected rise in unemployment, in other words, predicted an increase in suicides 6 months later among older Swedes but the increase may have ‘harvested’ suicides that would otherwise have occurred 8 months after the given month.

    View this table:
    Table 1

    Estimated parameters (SE in parentheses) for models predicting monthly suicides from unemployment in Sweden (n=144 months beginning 1/2000)

    No association appeared between unemployment and suicide among working-age women. In addition to the displacement pattern noted above among older women, a positive association 1 month after unexpected changes in unemployment also appeared. This pattern implies that if displacement was, in fact, at work in older women, it could have been from month 8 after an unexpected rise in unemployment to the first and sixth month following the rise.

    We assessed the robustness of our results with several additional analyses. First, we transformed both suicides and unemployment rates into their natural logarithms to ensure that variation in variability over time did not induce our results. The findings did not change.

    We used the outlier detection and adjustment routines of Chang et al28 to determine if extreme values inflated our CIs and led to false acceptance of the null for lags of unemployment other than those found statistically significant in our test. Results did not change.

    We deleted statistically non-significant predictors from the models shown in table 1 and re-estimated the equations to determine if any of our results assumed associations that could have arisen by chance at the p<0.05 level (2-tailed test). Only the association for working age men 4 months after changes in unemployment fell below the significance criterion.

    Discussion

    Using monthly data from Sweden, we find support for the mortality induction and displacement explanations of observed increases in suicide following economic contraction. Among working-aged males (ie, ages 20–64), larger than expected increases in suicide occur 4–6 months after unexpected increases in unemployment, suggesting an induction of suicides. Among men and women 65 years and older, however, we find evidence of mortality displacement, such that an unexpected rise in unemployment predicts an increase in suicides 6 months later, followed by a significant decrease 8 months later. Therefore, an increase in unemployment above expected levels may have ‘harvested’, among older individuals, suicides that would have otherwise occurred later without an increase in unemployment.

    The current study finds no association between suicides among working age females and unemployment. This finding appears consistent with earlier reports that males, but not females, commit suicide in response to economic contraction,16–18 and that men react more than women to unemployment, retirement and occupational instability.29 We also note, however, that other individual-level studies find no difference by gender in the suicide–unemployment relationship.4 ,12 ,13 It is possible that these differences arise from differences in labour market composition among the countries studied. The fact that women disproportionately work in Sweden's public sector,30 which provides more security during recessions than does employment in the private sector, may, for example, contribute to the gender difference we found. Males may also respond more to unemployment because they have, at least historically, assumed the ‘breadwinner role’ defined by success in the labour market.31

    Our finding of mortality displacement among older persons may seem counterintuitive given that they participate less frequently in the labour market than do younger men and women. Durkheim's typology of suicide, however, suggests a post hoc explanation.32 A downturn in the economy could harvest ‘fatalistic suicides’ among older persons who endemically and increasingly fear the loss of independence and must confront decreases in the value of their assets or retirement funds. Economic contraction may also displace ‘altruistic suicides’ among older persons who fear they burden their children even in good times. The shock of witnessing job or income loss among children may ‘harvest’ altruistic suicides that would otherwise have occurred later.

    Readers may intuit that Sweden’s traditionally high levels of public welfare programming should buffer workers and retired persons from the stress of economic contraction. We note, however, that Sweden has reduced its social programming.33 ,34 We further note that research has reported that the incidence of clinically treated depression among older persons appears tied to the labour market experiences of younger persons.35

    We used monthly, rather than yearly, counts of suicide to allow us to test for the temporally compensating associations implied by displacement. Annual data, used in most studies of the association between economic indicators and suicide, would mask displacement because the net effect of unemployment on suicide would appear null. Monthly data, however, introduce the complexity of lagged relationships. Explaining the observed lag between unexpected changes in unemployment and suicide leads to post hoc speculation about mechanisms that tie labour market dynamics to pathology in individuals. As demographers and epidemiologists, we probably know too little of the clinical literature to offer compelling explanations, but we note that much work suggests and reports such latency; that literature offers two explanations for these lags. The first, based on Durkheim’s seminal work,10 assumes that economic contraction not only increases the incidence of hopelessness but also induces ‘anomie’ in the population, thereby weakening the vigilance by which society enforces its disapproval of self-destructive behaviour.36 These processes presumably incubate before manifesting in the incidence of suicide. The second argument assumes that welfare programmes protect the unemployed, but that such programmes do not support them indefinitely or without demoralising stigma.4 ,34 Pathology in this explanation emerges after the rise in unemployment, and suicidal ideation may come even later. Both these arguments predict the lags we found. Those lags, moreover, appear consistent with individual level research reporting that suicidal ideation emerges mostly among those who have been unemployed more than 6 months and have reported previous anxiety and demoralisation.37

    This study contributes to the literature by empirically asking whether displacement or inducement better explains the observed association between economic indicators and the incidence of suicide. Our results from Sweden find support for both mechanisms when the analyses include a period of 1 year beyond unexpectedly high unemployment.

    Those who make economic policy typically claim that they account the costs and benefits of options before choosing a course of action. Our research suggests that those making such claims cannot invoke mortality displacement as a justification for excluding suicide among costs accounted. Although the effect of rising unemployment on suicide among the elderly may sum to near zero, that on working aged men may not. We believe that our results support programming aimed at reducing the psychological pathology among job losers.38

    What is already known on this subject

    • Several individual-level and aggregate-level studies suggest that suicide appears responsive to economic downturns, but few studies have investigated mechanisms explaining this observed increase in suicides.

    What this study adds

    • This study contributes to the literature by empirically asking whether unexpectedly high unemployment appears to induce suicide in persons otherwise unlikely to engage in self-destructive behavior (ie, induction) or if a contracting economy ‘capitalises’ on latent or underlying psychological vulnerability, triggering suicides that would have occurred eventually (ie, displacement or ‘harvesting’).

    • We find that displacement and induction both appear to have operated following unexpected labour market contractions in Sweden, though with different population segments.

    References

    1. ↵
      1. Catalano R,
      2. Goldman-Mellor S,
      3. Saxton K, et al
      . The health effects of economic decline. Annu Rev Public Health 2011;32:431–50. doi:10.1146/annurev-publhealth-031210-101146
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      1. Burgard SA,
      2. Ailshire JA,
      3. Kalousova L
      . The great recession and health: people, populations, and disparities. Ann Am Acad Pol Soc Sci 2013;650: 194–213. doi:10.1177/0002716213500212
      OpenUrlAbstract/FREE Full Text
    3. ↵
      1. Ruhm CJ
      . Are recessions good for your health? Q J Econ 2000;115:617–50. doi:10.1162/003355300554872
      OpenUrlAbstract/FREE Full Text
    4. ↵
      1. Blakely TA,
      2. Collings SC,
      3. Atkinson J
      . Unemployment and suicide. Evidence for a causal association? J Epidemiol Community Health 2003;57:594–600. doi:10.1136/jech.57.8.594
      OpenUrlAbstract/FREE Full Text
    5. ↵
      1. Laanani M,
      2. Ghosn W,
      3. Jougla E, et al
      . Impact of unemployment variations on suicide mortality in Western European countries (2000–2010). J Epidemiol Community Health 2015;69:1039.
      OpenUrlAbstract/FREE Full Text
    6. ↵
      1. Reeves A,
      2. McKee M,
      3. Stuckler D
      . Economic suicides in the great recession in Europe and North America. Br J Psychiatry 2014;205:246–7. doi:10.1192/bjp.bp.114.144766
      OpenUrlAbstract/FREE Full Text
    7. ↵
      1. Goldman-Mellor SJ,
      2. Saxton KB,
      3. Catalano RC
      . Economic contraction and mental health. Int J Mental Health 2010;39:6–31. doi:10.2753/IMH0020-7411390201
      OpenUrlCrossRef
    8. ↵
      1. Zivin K,
      2. Paczkowski M,
      3. Galea S
      . Economic downturns and population mental health: research findings, gaps, challenges and priorities. Psychol Med 2011;41:1343–8. doi:10.1017/S003329171000173X
      OpenUrlCrossRefPubMed
    9. ↵
      1. Norström T,
      2. Gronqvist H
      . The great recession, unemployment and suicide. J Epidemiol Community Health 2015;69:110–16. doi:10.1136/jech-2014-204602
      OpenUrlAbstract/FREE Full Text
    10. ↵
      1. Durkheim E
      . Suicide: a study in sociology. New York: Free Press, 1897.
    11. ↵
      1. Jones SC,
      2. Forster DP,
      3. Hassanyeh F
      . The role of unemployment in parasuicide. Psychol Med 1991;21:169–76. doi:10.1017/S0033291700014756
      OpenUrlCrossRefPubMedWeb of Science
    12. ↵
      1. Lewis G,
      2. Sloggett A
      . Suicide, deprivation, and unemployment: record linkage study. BMJ 1998;317:1283–6. doi:10.1136/bmj.317.7168.1283
      OpenUrlAbstract/FREE Full Text
    13. ↵
      1. Kposowa AJ
      . Unemployment and suicide: a cohort analysis of social factors predicting suicide in the US National Longitudinal Mortality Study. Psychol Med 2001;31:127–38. doi:10.1017/S0033291799002925
      OpenUrlCrossRefPubMedWeb of Science
    14. ↵
      1. Zanobetti A,
      2. Schwartz J,
      3. Samoli E, et al
      . The temporal pattern of mortality responses to air pollution: a multicity assessment of mortality displacement. Epidemiology 2002;13:87–93. doi:10.1097/00001648-200201000-00014
      OpenUrlCrossRefPubMedWeb of Science
    15. ↵
      1. Saha MV,
      2. Davis RE,
      3. Hondula DM
      . Mortality displacement as a function of heat event strength in 7 US cities. Am J Epidemiol 2014;179:467–74. doi:10.1093/aje/kwt264
      OpenUrlAbstract/FREE Full Text
    16. ↵
      1. Gunnell D,
      2. Middleton N,
      3. Whitley E, et al
      . Why are suicide rates rising in young men but falling in the elderly?—a time-series analysis of trends in England and Wales 1950–1998. Soc Sci Med 2003;57:595–611. doi:10.1016/S0277-9536(02)00408-2
      OpenUrlCrossRefPubMedWeb of Science
    17. ↵
      1. Qin P,
      2. Agerbo E,
      3. Westergard-Nielsen N, et al
      . Gender differences in risk factors for suicide in Denmark. Br J Psychiatry 2000;177:546–50. doi:10.1192/bjp.177.6.546
      OpenUrlAbstract/FREE Full Text
    18. ↵
      1. Pompili M,
      2. Vichi M,
      3. Innamorati M, et al
      . Suicide in Italy during a time of economic recession: some recent data related to age and gender based on a nationwide register study. Health Soc Care Community 2014;22:361–7. doi:10.1111/hsc.12086
      OpenUrlCrossRefPubMed
    19. ↵
      1. O'Connell H,
      2. Chin AV,
      3. Cunningham C, et al
      . Recent developments: suicide in older people. BMJ 2004;329:895–9. doi:10.1136/bmj.329.7471.895
      OpenUrlFREE Full Text
    20. ↵
      1. Allebeck P,
      2. Allgulander C,
      3. Henningsohn L, et al
      . Causes of death in a cohort of 50 465 young men—validity of recorded suicide as underlying cause of death. Scand J Public Health 1991;19:242–7.
      OpenUrlAbstract/FREE Full Text
    21. ↵
      1. de Faire U,
      2. Friberg L,
      3. Lorich U, et al
      . A validation of cause-of-death certification in 1,156 deaths. Acta Med Scand 1976;200:223–8. doi:10.1111/j.0954-6820.1976.tb08223.x
      OpenUrlPubMedWeb of Science
    22. ↵
      Swedish National Board of Health and Welfare. [10/28/2014]. http://www.socialstyrelsen.se/english
    23. ↵
    24. ↵
      1. Stigler SM
      . Francis Galton's account of the invention of correlation. Stat Sci 1989;4:73–9. doi:10.1214/ss/1177012580
      OpenUrlCrossRef
    25. ↵
      1. Fisher RA
      . Studies in crop variation I. An examination of the yield of dressed grain from Broadbalk. J Agric Sci 1921;11:107–35. doi:10.1017/S0021859600003750
      OpenUrlCrossRef
    26. ↵
      1. Box G,
      2. Jenkins G,
      3. Reinsel G
      . Time series analysis: forecasting and control. London: Prentice-Hall, 1994.
    27. ↵
      1. Zeger SL,
      2. Irizarry R,
      3. Peng RD
      . On time series analysis of public health and biomedical data. Annu Rev Public Health 2006;27:57–79. doi:10.1146/annurev.publhealth.26.021304.144517
      OpenUrlCrossRefPubMedWeb of Science
    28. ↵
      1. Chang I,
      2. Tiao GC,
      3. Chen C
      . Estimation of time series parameters in the presence of outliers. Technometrics 1988;30:193–204. doi:10.1080/00401706.1988.10488367
      OpenUrlCrossRefWeb of Science
    29. ↵
      1. Artazcoz L,
      2. Benach J,
      3. Borrell C, et al
      . Unemployment and mental health: understanding the interactions among gender, family roles, and social class. Am J Public Health 2004;94:82–8. doi:10.2105/AJPH.94.1.82
      OpenUrlCrossRefPubMedWeb of Science
    30. ↵
      1. Rubery J,
      2. Smith M,
      3. Fagan C, et al
      . Women and European employment. New York: Routledge, 2014.
    31. ↵
      1. Norström T
      . Leder arbetslöshet till fler självmord? Framtider 2009;3:18–20.
      OpenUrl
    32. ↵
      1. Marson SM,
      2. Powell RM
      . Suicide among elders: a Durkheimian proposal. Int J Ageing Later Life 2011;6:59–79. doi:10.3384/ijal.1652-8670.116159
      OpenUrlCrossRef
    33. ↵
      1. Lindbom A
      . Political partisanship and policy feedback: the Swedish welfare state after eight years of center-right government. In: Pierre J, ed. The Oxford handbook of Swedish politics. Oxford: Oxford University Press, 2015.
    34. ↵
      1. Paulsen R
      . Vi bara lyder: en berättelse om Arbetsförmedlingen. Stockholm: Bokförlaget Atlas, 2015.
    35. ↵
      1. Hartig T,
      2. Catalano R,
      3. Ong M, et al
      . Vacation, collective restoration, and mental health in a population. Soc Mental Health 2013;3:221–36. doi:10.1177/2156869313497718
      OpenUrlCrossRef
    36. ↵
      1. Pierce A
      . The economic cycle and the social suicide rate. Am Sociol Rev 1967;32:457–62. doi:10.2307/2091092
      OpenUrlCrossRefPubMedWeb of Science
    37. ↵
      1. Dooley D,
      2. Catalano R,
      3. Rook K, et al
      . Economic stress and suicide: multilevel analyses. Part 2: cross-level analyses of economic stress and suicidal ideation. Suicide Life Threat Behav 1989;19:337–51.
      OpenUrlPubMedWeb of Science
    38. ↵
      1. Vinokur AD,
      2. Price RH,
      3. Schul Y
      . Impact of the JOBS intervention on unemployed workers varying in risk for depression. Am J Community Psychol 1995;23: 39–74. doi:10.1007/BF02506922
      OpenUrlCrossRefPubMedWeb of Science

    Footnotes

    • Contributors AG helped plan the study and wrote the introduction and discussion sections. AF, DK and TH were critical in the thinking behind the study as well as made contributions to the introduction and discussion. EA prepared the data used in the analysis and provided editing of the manuscript. RC helped plan the study and wrote the methods and results sections. All the authors contributed to the design and writing of the manuscript. All the authors have approved the final version.

    • Funding This work was supported by the National Institute of Child Health and Human Development (grant number T32-HD007275) and the Robert Wood Johnson Health and Society Scholars Programme.

    • Competing interests None declared.

    • Provenance and peer review Not commissioned; externally peer reviewed.