Article Text

Download PDFPDF

Suicide rates among working-age adults in South Korea before and after the 2008 economic crisis
  1. Chee Hon Chan1,2,
  2. Eric D Caine3,4,
  3. Sungeun You5,
  4. King Wa Fu6,
  5. Shu Sen Chang1,2,
  6. Paul Siu Fai Yip1,2
  1. 1Department of Social Work and Social Administration, University of Hong Kong, Hong Kong, Hong Kong
  2. 2Centre for Suicide Research and Prevention, University of Hong Kong, Hong Kong, Hong Kong
  3. 3Injury Control Research Center for Suicide Prevention and Department of Psychiatry, University of Rochester Medical Center, Rochester, New York, USA
  4. 4VA Center of Excellence for Suicide Prevention, Canandaigua, New York, USA
  5. 5Department of Psychology, Chungbuk National University, Cheongju, Republic of Korea
  6. 6Journalism and Media Studies Centre, University of Hong Kong, Hong Kong, Hong Kong
  1. Correspondence to Professor Paul Yip, Centre for Suicide Research and Prevention, HKJC Multidisciplinary Research Building, 2nd Floor, No. 5 Sassoon Road, Pokfulam


Background Multiple studies have shown that macroeconomic factors are associated with changes in suicide rates. We investigated how changes in economic conditions associated with the recent economic crisis in South Korea influenced suicide rates among working-age adults.

Methods Time-series analyses were performed to examine the temporal associations of national unemployment rates and sex-employment-specific suicide rates in South Korea from 2003 to 2011, with particular attention to the increases of suicides that occurred during the recessionary period that began in 2008. We also compared the relative risk of suicide among different occupations.

Results National unemployment rates were positively associated with suicide rates among employed and unemployed men and women, with a 2-month to 3 month lagged period. Significant increases of suicide rates among working-age adults during the recession were detected in most of the subgroups stratified by age, sex and employment status. Forty-three per cent of the increase of suicides was derived from the employed population. Compared with workers in elementary occupations, the relative risk of suicide for mangers increased by threefold during the recessionary period. Among those who were employed, half of the increases in suicides occurred among clerks and workers involved in sales and services.

Conclusions Changes in macroeconomic conditions are tied to population-level suicide risks for employed and unemployed persons. However, these associations vary depending on sex, employment status and occupational roles. In advance of future economic crises, it is important to develop prevention initiatives intended to reach the diverse populations potentially exposed to the adverse effects of sudden economic disruptions.


Statistics from


Macroeconomic factors have been shown repeatedly to be associated with significant changes in suicide rates when measured at a population level. Multiple studies have tied rising unemployment rates to short-term increases in suicide rates.1–5 At the individual level, job loss or insecurity has also been viewed as key contributors to increased suicide risk.6 A recent study from Hong Kong found that, at a time when unemployment increased, a greater proportion of suicides were from employed adults rather than the unemployed.5 Those findings underscored, as well, the need to separate population-level indicators and individual employment status. Among the employed, suicide risk has been shown to differ across occupations.7 Recent investigations have further highlighted that fluctuating labour market conditions may disproportionally influence suicide rates in different occupational groups.8

Following the recovery from the 1997 Asian financial economic crisis, South Korea experienced sustained economic growth until 2008. Its economy shrank beginning in late 2008, continuing for three consecutive quarters; the unemployment rate began to rise during the first quarter of 2009, which continued into 2010. While the suicide rate of South Korea had declined from 2005 to 2008, it increased sharply to 33.8 per 100 000 people in 2009.9 By 2012, the suicide rate remained at a historic high level, 31.7 per 100 000 people, the highest level among the Organisation for Economic Co-operation and Development countries and above neighbouring regional countries (eg, China, Japan). In light of previous studies that described a link between macroeconomic forces and suicide rates in South Korea,10–12 we decided to explore in greater depth the timing and impact of changing unemployment rates on workers—employed and unemployed, and stratified by sex, age and occupation.


Sample and data analyses

We collected South Korean mortality data coded X60 to X84 (intentional self-harm) and Y10 to Y34 (event of undetermined intent) in the ICD-10 from January 2003 through December 2011 from the National Statistical Office of Korea (NSO). Within this period, there were 117 300 suicides and 19 452 undetermined deaths. For each death, we obtained information on the date of death, age, sex, employment status and the type of occupation. The categorisation of occupation obtained from NSO includes: manager and senior officials; professionals; clerks; sales and service workers; agriculture, fishery, forestry; craft and related trades; process, plant, and machine operatives; and elementary workers. South Korean national unemployment, population and occupation statistics were obtained from the NSO website. NSO conducts the census for the South Korean population every 5 years (ie, 2005 and 2010), and it is the official government body to collect employment statistics.

Given the nature of the information used, based on analyses of deidentified data and population statistics, this study was exempted from ethical review by the Human Research Ethics Committee for Non-Clinical Faculties, The University of Hong Kong.

Statistical analyses

The analysis was organised in three parts. First, we used autoregressive (AR) time-series modelling to ascertain whether there was a temporal association between the Korean monthly national unemployment rates and Korean monthly suicide rates, specific to sex and employment status. Since changes of economic conditions may have a lagged effect on suicide rates, we explored the association with a 0-month to 6 month lagged unit. Given that previous research had described that three high profile celebrity suicides were related to increases in the suicide rate in March 2005, February 2007, October 2008 and December 2008,13 we used a dummy variable in the AR models to adjust for these effects. Also, seasonality of the series was adjusted.

Second, we performed an interrupted time-series analysis to investigate whether there was significant increase of Korean sex×age×employment suicide rates after the onset of the 2008 global financial crisis. Similar to previous studies,1 ,2 we decided to use the national unemployment rate as the indicator of the economic environment; as expected, the Korean unemployment rate modestly lagged the start and the finish of the period of gross domestic product contraction (typical of its status as a ‘lagging indicator’). The period between January 2003 and December 2008 in which the unemployment rate remained relatively stable was marked as the ‘precrisis period,’ whereas the period with an abrupt increase of unemployment, between January 2009 and December 2010, was considered as the ‘recessionary period.’ In all models, we used step functions taking the precrisis period as reference to assess the increase of the suicide rate during the recessionary period. The effect of celebrity suicides and seasonality of the series were adjusted.

To determine the numerical increase of suicides during the recessionary period, we used this equationEmbedded ImageEmbedded ImageEmbedded Image ΔSijk denotes the numerical increase of suicides in a specific age(i), sex (j) and employment status (k) group, r is the estimated increase of suicide rate of an age-sex-employment-specific group from the AR model, and P is the size of the corresponding population. Estimates of the increased suicide rate (r) from the AR models take into account the autocorrelation of the residual in the series. Hence, using it to proxy the numerical increase should be more stringent than direct computation from count data based on simple ordinary least-square extrapolation.

Third, we stratified suicide counts based on occupation types, and computed the relative risks (RRs) for each occupation during the precrisis and the recessionary periods. Also, we assessed the numerical increases of suicides stratified by occupation during the latter period.

In both time-series analyses, we focused on the working-age population, defined as ages 20–59 years. Individuals younger than 20 years or 60 years old and above were considered as students and retirees, respectively. Based on the employment statistics from NSO, approximately 90% of the employed population in the study period fell in the 20–59 year-old age range. The adequacy of all time-series models was checked by visual inspection of the autocorrelation function and partial autocorrelation function to ascertain absence of autocorrelation in the residuals. Box's test with multiple lag time was performed. For the exploration of occupation-specific suicide risk, we included populations who were labelled as employed by the NSO. Crude age-sex-employment-specific or occupation-specific suicide rates are calculated by dividing the number of suicides of the specific group by the corresponding population. All analyses were performed in R. Statistical significance was set at 0.05.


Temporal associations between unemployment rate and sex-employment specific suicide rate

Figure 1 displays the temporal association between the Korean national unemployment rate and the sex-employment-specific suicide rates with 0-month to 6 month lagged units. Among the unemployed population, the Korean national unemployment rate was positively and significantly associated with the unemployed suicide rate of men and women, with a 2-month and a 3-month lagged unit (unemployed men: 2-month lagged β=1.21, 95% CI 0.14 to 2.29; 3-month lagged β=2.54, 95% CI 1.68 to 3.40; unemployed women: 2-month lagged β=0.44, 95% CI 0.06 to 0.83; 3-month lagged β=0.61, 95% CI 0.21 to 1.01). For the employed population, although the associations were somewhat weaker, significant associations also were noted between the national unemployment rate and the employed suicide rate of men and women, with a 2-month lagged unit (employed men: β=0.26, 95% CI 0.06 to 0.46; employed women: β=0.15, 95% CI 0.02 to 0.28) and a 3-month lagged unit (employed men: β=0.33, 95% CI 0.13 to 0.53; employed women: β=0.20, 95% CI 0.06 to 0.35). Examining the association of sex-specific unemployment rates and sex×employment specific suicide rates, we detected that the sex-specific unemployment rate was significantly tied to employed-sex-specific and unemployed-sex-specific suicide rates, with either a 2-month lagged unit or a 3-month lagged unit (see online supplementary appendix 1). The increase in the male unemployment rate was associated with a larger increase in suicide rates for men and women in the employed and unemployed groups, as compared with increased unemployment among women.

Figure 1

Temporal association of the national unemployment rate and the (lagged) sex-employment-specific suicide rate. The x-axis denotes the lagged unit of the association. The y-axis shows the beta of the association between the national unemployment rate and the sex-employment-specific suicide rate. The upper and lower bounds of the error bars show the 95% CI of the association.

Increase of suicide in the recessionary period after financial crisis

Table 1 summarises the increase of suicide rates and the numerical increases of suicides during the recessionary period (January 2009 to December 2010). Suicide rate increases were relatively higher among unemployed than employed groups. Nonetheless, significant increases were detected among all subgroups, except for employed men and women in their 50s, and unemployed men in their 30s. Among women of all age groups, suicide rate increases were relatively higher among the unemployed than the employed, and the numerical increases of suicides among unemployed women were greater than the employed. Among men, the increase of suicide rates was also higher in unemployed groups than the employed groups.

Table 1

Increase of suicide rates and numerical increase of suicides during the recessionary period (January 2009 to December 2010)

Although the increase in suicide rates among men aged 30–49 years was higher among unemployed persons, the absolute numerical contribution to the suicide burden was greater among the employed in these age groups. For the entire group ages 20–59 years, the increase of suicides among men was greater among the employed (n=1431, 51%) than the unemployed (n=1376, 49%); for women, the increase in suicides reflected a greater contribution from members of the unemployed group (n=1525, 67%) than the employed (n=768, 33%). Combining suicides of both sex, the employed group accounted for 43% of the increase in deaths.

Occupation-specific suicide trends among the employed

Occupational stratified suicide trends for the employed population are shown in figure 2. The suicide trends of all occupation groups during the precrisis period appeared to be either stable or steadily decreasing. Table 2 summarises findings of the differential occupational suicide risk and the annual suicide cases of each occupation during the precrisis and the recessionary period. In both periods, the suicide risks for workers in elementary occupations were used as the reference. We observed a relatively greater suicide risk among agriculture, fishery and forestry workers (RR=5.52, 95% CI 4.93 to 6.18), and among clerks (RR=1.35, 95% CI 1.19 to 1.53), and sales and service workers (RR=1.34, 95% CI 1.19 to 1.51) during the precrisis period. The RR of suicide for the latter two groups increased during the recessionary period, and their numerical increase of suicides was greatest among all occupational groups, accounting for 53% of the total increase. In contrast, the suicide risk of the agriculture, fishery and forestry decreased during the recessionary period (accounting for 5% decrease of total suicides), although these groups continue to present the greatest risks across all occupations Managers and senior officials (RR=0.75, 95% CI 0.57 to 0.99), and professionals (RR=0.77, 95% CI 0.67 to 0.89), had a relatively lower suicide risk before the recession; of particular note, the risk increased by more than threefold for managers and senior officials during the recession (RR=2.81, 95% CI 2.22 to 3.55), but remained low among professionals (RR=0.81, 95% CI 0.67 to 0.98).

Table 2

Relative risks (RRs) and the numerical increases of suicides among the employed population in the precrisis and the recessionary periods

Figure 2

Suicide trends among the employed population stratified by occupation types.


The relationship between unemployment statistics and suicide is complex.5 In keeping with past studies from South Korea,11 ,12 we found when using data regarding its working-age population further supports linking changes of macroeconomic conditions, as measured by the increasing national unemployment rate, and increasing suicide rates among specific populations. While increasing unemployment has been viewed as the prelude to an upswing in suicides, the timing of such changes has been less clear. In this study, changes of the South Korea national unemployment rate preceded 2-month to 3-month lagged increases in suicide rates among employed and unemployed populations. The increases in suicides among employed and unemployed men and women were most closely tied to changes in the unemployment rate of men, which in South Korea may have greater overall community impact. Such nuances require further quantitative and qualitative study. Data from this study and past reports underscore that recession-related increases in suicide are not solely a matter of unemployment, despite the robust statistical link between unemployment rates and suicide rates. Job insecurity can be a major stressor among those who have killed themselves.14 Corporate downsising and labour market restructuring during economic contraction may create additional work stress and a sense of job insecurity among the employed (eg, increasing workload and uncertainty about the future), which may contribute to a risk for depression and an adverse impact on mental health.15 ,16 Others may experience a dramatic loss of personal wealth when related changes occur in the residential housing market (called in the USA, ‘being under water,’ when deflated housing values fall below the residual mortgage value). In many nations, persons may remain employed but are either transferred to part-time roles or furloughed intermittently. Thus, maintaining employment does not necessarily spare many from the stresses of macroeconomic turmoil. What specific factors contribute to such vulnerability—individually and collectively? As important, what protects the vast majority of financially distressed people from killing themselves? Is it qualities within the person alone or are there other social circumstances that add to the mix? Persons living alone and having relatively low income, in addition to suffering clinically significant depression, are suggested to bore a higher risk.14 However, we know of no comparably detailed data drawn from socially diverse countries that might begin to offer preliminary insights beyond speculation.

In line with previous reports on Korean working population,17 ,18 we found that managers and senior officials, as well as professional workers, carried a lower risk of suicide during the precrisis period. Members of specific occupations, particularly those in lower-paid roles and job categories (semiskilled and non-skilled workers) and occupations with ready access to lethal methods of suicide (eg, farmers) apparently experienced greater vulnerability.17 ,19–21 During the economic downturn, nearly all occupations had an increased suicide risk (except the agricultural workers and professionals). The relative increase was greatest among managers. Our findings underscore that suicide vulnerability is not limited solely to persons on the lower rungs of the economic ladder, a result reinforced by a recent report from Japan that described increased suicide trends among managers and professional workers in the midst of its longstanding economic stagnation.8 Overall, half of the increase in suicides occurred among clerks, and sales and service workers, individuals stationed at the lower end of the pay scale and the ones most vulnerable to job loss. We caution, however, that efforts to interpret trends in national suicide rates must appreciate the dynamic nature of the interactions between societal, community and individual processes. For example, South Korea was in the midst of efforts to limit access to lethal agricultural pesticides,22 ,23 which may help explain the relative stability among agricultural workers during the recession.

Strength and limitations

Mortality data in South Korea are collected either by the NSO or the National Police Agency. Suicide cases recorded by the former are based on death certificates and are reported primarily by families. It is possible that families deny or are reluctant to report suicide as the cause of death due to social stigma. Also, for suicides that did not have identified families (eg, homeless), the accurate manner of death may be not included in these databases. Prior work has suggested that the NSO may have underestimated the number of suicides, misclassifying them as deaths due to other causes (ie, car accident).24 ,25 We found that from 2003 to 2011, the period of this study, the number of undetermined deaths was stable, averaging about 2000 per year. Repeating our analyses, with undetermined deaths included as suicides, had no significant effect on our findings.

As noted in our prior work,5 population level unemployment rates serve as an indicator used to explore the effects of contextual economic factors, but they are only a distant approximation of individual-level financial conditions. Investigations must avoid the potential for ecological fallacies. Many people not in the workforce will be adversely affected as the overall increases in unemployment, and among the employed, many may experience fiscal hardships while retaining their jobs (eg, changing values of home or property; risk and fear of losing the job; unemployment of a family member). Mindful of these constraints, we also have sought to gain more information—albeit inferential—by using occupation-related characteristics to divide or subgroup men and women. In this work we are using employment status and occupation as the means to explore how the economic forces may be acting differentially during a time of sudden economic decline. Such an approach allows us to focus the lens somewhat more tightly, but admittedly we remain too distant to consider individual-level effects in the context of broad social disruption.

This was an observational study, and our findings must be viewed cautiously. By its nature we cannot interpret causation; however, our analyses showed consistently that suicide rates in Korean working-age adults lagged behind the fluctuations of unemployment rates. These findings reaffirm with greater precision the temporal association of these factors.26 When combined with prior findings,1 ,4 the evidence continues to grow in strength that economic stress serves as a specific contributory factor at the individual level. However, we would expect that for some this would be a more ‘distal’ event while for others it may serve as an acute precipitant.

Another limitation to our study derived from the inclusion of students, housewives and retired persons among the NSO label of ‘unemployed’. It is highly probable that they do not share similar risks with those who have recently lost jobs. Also, despite the use of the official mortality database, misclassification of manner of death remains likely. Finally, we cannot rule out unknown confounding factors whenever using large datasets such as these.


Our results again point to the powerful role of contextual factors in suicide. Currently, clinicians and research funding agencies heavily emphasise discovering factors within the person to explain these tragic outcomes. However, there is fertile ground to explore what makes some people more vulnerable, while most are protected in the face of commonly occurring, measurable social events. As well, knowing that economic disruptions are times of highly predictable increases in suicides, whatever the causal mechanisms, invites public health and clinical providers—as well as employers who may be ‘shedding’ their workers—to consider proactively what might be developed as interventions to enhance the well-being of those who inevitably are confronted with major life crises.

What is already known on this topic?

  • Previous studies using unemployment rates as indicators of nationwide economic conditions have repeatedly shown that, during the time of economic turmoil, suicide rates increase.

What this study adds?

  • Suicide rates in South Korea among working-age men and women lagged 2 months to 3 months behind the rise of national unemployment rate.

  • The suicide risks for the employed and the unemployed were adversely affected by the sudden economic downturn, highlighting the importance of understanding how broader contextual (economic) factors affect population-level measures.

  • Differential suicide risks were apparent across various occupations, further emphasising the dynamic, non-uniform effects of contextual factors.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors CHC, SSC, KWF and PSFY were involved in the study design and the analysis and interpretation of data. CHC, EDC and PSFY were involved in drafting the article. PSFY and EDC were involved in the final approval of the version to be published.

  • Funding This work was supported by General Research Grant (HKU 784012M) and DHHS/PHS/CDC Award 1 R49 CE002093.

  • Competing interests None.

  • Ethics approval Ethical review by the Human Research Ethics Committee for Non-Clinical Faculties, The University of Hong Kong.

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

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.