Article Text

Download PDFPDF

Employment status and suicide: the complex relationships between changing unemployment rates and death rates
  1. Paul S F Yip1,
  2. Eric D Caine2
  1. 1The University of Hong Kong, Hong Kong, China
  2. 2The University of Rochester, Rochester, New York City, USA
  1. Correspondence to Dr Paul SF Yip, The Hong Kong Jockey Club Centre for Suicide Research and Prevention and Department of Social Work and Social Administration, The University of Hong Kong, Pokfulam, Hong Kong, China; sfpyip{at}


Background Existing studies have described a strong correlation between unemployment rates and suicide rates, but the exact mechanisms through which they may interact with one another remain unknown.

Method This study examined the complex relationships between suicide rates and both regional unemployment rates and individual employment status during times of economic recession (2000–3) and recovery (2003–6) in Hong Kong.

Results Despite the strong correlation (0.86) between the unemployment rates and suicide rates for 2000–6, the rates of suicides within the employed and unemployed groups moved in the opposite direction from the overall population trend. That is, the suicide rate among the unemployed decreased during economic recession and increased during recovery.

Conclusion It is important to be able to distinguish precisely between population-level concepts, such as rates, and individual-level characteristics, such as employment status, when considering the development of evidence-based suicide prevention strategies.

  • Decomposition analysis
  • employment
  • epidemiology FQ
  • health
  • Hong Kong
  • suicide rate
  • suicide SI
  • unemployment

Statistics from

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.

Numerous studies have tied unemployment to suicide, either by examining the correlation of unemployment rates to suicide rates1–3 or exploring the role of being unemployed among individuals who have killed themselves. Being unemployed is viewed as a major risk factor for suicide by multiple authors.4–10 There is debate among others as to whether being unemployed itself is the driving risk at the individual level. The vast bulk of unemployed people never kill themselves, and many studies have shown that the ‘unemployment effect’ disappears, or is greatly diminished, when co-varying depression status in large scale analyses.11 Furthermore, Beautrais et al11 showed that when taken together the population attributable risk for depression greatly outweighs the impact of unemployment at the time of death for those who killed themselves.

This debate is confounded by a number of considerations. Whereas unemployment alone may not be necessary or sufficient to cause suicide, it may contribute to the context in which suicide is more common; for example, it may precede the onset of depressive thoughts, symptoms, or syndromes, acting as a promoting effect or trigger.12 13 Alternatively, unemployment may reflect long-standing dysfunction associated with psychopathology, itself a marker, when the unemployed state reflects drift into a less effective social role.1 14 Finally, and perhaps most important, discussion in the literature may be conflating consideration of ‘being unemployed’, as an individual characteristic (ie, representing a person's job or employment status), with unemployment as an indicator of larger economic forces at play in the community (ie, the unemployment rate).4

It is essential first to disentangle these distinctive perspectives before being able to appreciate the potential contributions of both sociological and psychopathological approaches to understanding the factors that contribute to suicide. We report in this paper an examination of the relationship of increasing and decreasing unemployment rates on suicide rates in Hong Kong. Using available data, we then begin the process of examining the contributions to the overall community burden of suicide from the population groups (‘pools’) of employed and unemployed people. In this way we seek to begin a process of developing a more nuanced appreciation of the interplay of forces that are associated with self-inflicted death.


Data sources

Most of the data sources have been described previously. In brief, the study was based upon data from the Coroner's Court of Hong Kong for suicides, aged 20–59 years, which comprised the great bulk of the region's working age population, and drawn from a period of economic downturn (2000–3) that was followed by a time of recovery (2004–6).

Following the 10th edition of the International Classification of Diseases and Related Health Problems,15 we classified as suicide all reportable deaths with an external cause code in the range from X60 to X84. Collected variables included age, gender, marital status, employment status and other information such as manner of death; these were separated from individual identifiers before any analyses. Given the nature of the information used, based on analyses of de-identified data and community statistics, it was exempted for human subjects review by the institutional review board of the University of Hong Kong.

The present study aimed to examine the transitional trends of suicide rates among the ‘employed’ and ‘unemployed’ in the Hong Kong Special Administrative Region (SAR). In this context, it is important to note potential limitations of data sources as a caution when later interpreting results.

The denominators for our calculations, community data on employment and unemployment, were drawn from the General Household Survey (GHS) of the Census and Statistics Department of the Hong Kong government of SAR,16 which was used to estimate the size of the active working population, including the pools of employed and unemployed workers. Data regarding the employment status of those deceased by suicide were extracted from the death inquest report made available by the coroner's office of the Hong Kong SAR. As part of the formal investigative process conducted by the police for the coroner and required by law for all unnatural deaths in Hong Kong, family members (most often) provided information regarding the deceased's work status.17 Part-time employment equal to half time or more was coded as ‘employed’. Deceased homemakers, students and disabled individuals were considered as ‘economically inactive’, and as such, members of another group.

When calculating the ‘unemployed’ suicide rate, we thus counted those classified by the coroner as ‘unemployed’ for the numerator, and used the GHS figures for the denominator. The official unemployment rate relates to those people in the work force who are actively seeking jobs. Individuals who drop out of the active labour pool do not contribute to unemployment rates – rather they are classified as ‘economically inactive’ in the GHS, in common with international standards. This is not the same as ‘economically inactive’ used by the coroner. At the time data were collected by the Coroner's Office for this study, we had no method of ensuring that its method of classification mapped closely onto the GHS method. For example, we expect that the coroner would have labelled ‘unemployed’ a previously employed man who killed himself after becoming so depressed that he stopped looking for work. Unless that person had received disability status, the vernacular use of ‘unemployed’ would have prevailed rather than the GHS definition.

This type of potential misclassification would have had the effect of increasing the number of ‘unemployed suicides.’ Given the numbers involved, however, it would have had little overall impact on the denominator of ‘unemployed’ based on the GHS labour statistics, thus tending to overstate the suicide rate among unemployed workers.

Statistical analyses

To explore the factors associated with the transitional trends of suicide rates in Hong Kong during the study period (2000–6), we examined the changes in the employment profile of suicides. Unemployment figures and suicide rates for the period of economic recession (2000–3) and economic recovery (2004–6) are given. A decomposition method18 19 was then used to quantity the changes of suicide within the pools of the employment and unemployed (for details see appendix 1). Given our focus on employment status, suicides among economically inactive individuals, including retirees, housewives and students, were excluded for this analysis.


Table 1 shows that from 2000 to 2003, the number of employed in Hong Kong rose from 3 035 400 to 3 042 500, as the unemployment rate rose from 4.9% to 7.9%. The very minor growth in the absolute numbers of the employed pool, occurring at the same time as the unemployment rate rose, reflected continuing addition of new members to the overall workforce. Subsequently, the unemployment rate dropped 4.8% in 2006, while the work force increased to 3 236 300. The number of the unemployed increased from 145 100 in 2000 to 249 300 in 2003, before falling back to 154 900 in 2006 for the group of 20–59 year olds. The increase in the unemployed pool from 2000 to 2003 represented a compositional shift of workers who lost jobs; less visible, it also was likely that some of the previously unemployed stopped looking for work at the height of the recession, shifting into the economically inactive pool. Once the economy began to recover, these shifts would have reversed, ultimately leading to an increase of nearly 200 000 employed workers and a decline of nearly 100 000 in the unemployed. It is probable that entirely new individuals also entered the workforce. (Most of this would have come from local sources, as Hong Kong has heavily regulated the inmigration of workers from mainland China during the past decade.)

Table 1

Unemployment statistics and suicide rate for employed and unemployed for the periods of economic recession (2000–3) and recovery (2003–6), Hong Kong

Total suicides in the 20–59-year-old age group rose from 493 in 2000 to 704 in 2003, and then fell to 528 in 2006. In 2000, 228 of the suicides were employed and 265 were unemployed; 349 were employed and 355 were unemployed in 2003; and 225 were employed and 303 were unemployed in 2006. The suicide rate for the employed pool increased from 7.5 (per 100 000) to 11.5 between 2000 and 2003, and then decreased to 7.0 in 2006. Across the same period, the rates for the unemployed pool decreased from 182.6 in 2000 to 142.4 in 2003, before rising in to 196.3 in 2006. The ratio of the suicide rates, unemployed versus employed, fell from 24.3 to 12.4 between 2000 and 2003, before rising to 28.0 in 2006.

Table 2 shows the results of the decomposition analysis, which indicated that 78.3% of the increase in suicide rate for the period 2000–3 was accounted for by the increase in the unemployment rate. From 2004 to 2006, 83.2% of the reduction in the suicide rate was accounted for by a reduction in the unemployment rate. With the results in table 1 it suggests that the suicide rate for the employed pool rose during a time of increasing unemployment and decreased during a time of economic recovery, whereas the rate fell for the unemployed pool during the period of economic decline and rose during the time of recovery despite the strong correlation (ρ=0.82) between the employment rate and suicide rate for the period (see figure 1).

Table 2

A decomposition analysis for change of suicide rates in respect to employment status and its suicide rate for the period 2000–6

Figure 1

The relationship between the rates of unemployment* and suicide among the working age population† in Hong Kong for the period 2000–6. *The rates of unemployment were obtained from the Hong Kong Census and Statistics Department.16 †The rates of suicide were reported by the Hong Kong Jockey Club Centre for Suicide Research and Prevention.


Our results present another demonstration of the close relationship between unemployment rates and suicide rates. We detected a robust relationship between increasing unemployment and increasing suicide, and conversely, between decreasing unemployment and the return of suicide rates to their previous level. This study benefited by having data to examine the ascension and decline of rates, which adds what we might term the ‘dynamic dimension’ to earlier findings.

At the same time, our findings present what might be viewed at first as a paradox. While we detected a robust association between changing unemployment rates and changing suicide rates in the overall workforce population, the rates of suicide within the employed and unemployed pools moved in the opposite direction to the ‘macro’ trends. That is, at a time when the unemployment rate was rising along with the overall suicide rate, the rate within the unemployed group was dropping. On the other hand, the suicide rate among the employed was rising. When the unemployment rate dropped, the rates returned to or beyond their baseline (year 2000) levels (in the case of the unemployed, to a level higher than in 2000).

How can this be explained? We suggest that at baseline the high rate within the unemployed group reflected the high percentage of ‘vulnerable’ people who were members of this relatively small group (approximate 145 000), in keeping with suggestions by many that unemployment status – especially in a robust economy – may serve as a marker of functional difficulties. While the much larger employed group (approximately 3 000 000) also contained vulnerable individuals, the proportion of the population they represented was relatively small, as one might infer from the baseline suicide rate. (We would note, however, that the total number of ‘vulnerable individuals’ among the employed probably surpassed the number among the unemployed, given the extreme differences in the sizes of the two pools.)

At the time of economic upheaval, as reflected in the increasing unemployment rate, there was a migration of approximately 100 000 workers from the employed to the unemployed pools, thus expanding the latter by more than two-thirds. While we can surmise that a disproportionate number of vulnerable individuals comprised those who transferred groups, the overall compositional effect was to dilute the concentration of vulnerable individuals among the unemployed. Therefore, although the suicides among the unemployed increased by 90 from 265 to 355, it reflected a diminished rate. On the other hand, the increase of 121 deaths from 228 to 349 among the employed occurred over a relatively stable denominator.

No doubt, the economic turmoil in Hong Kong during the early years of the 21st century stressed the employed members of the region (in addition to those who lost jobs), as evidenced by the increase in suicides within this group. Therefore, although there was a robust statistical relationship between the changing unemployment rate and the shifts in the suicide rate, the majority of the increase in deaths came from the employed. It is our view that, for Hong Kong during the study period, the ‘unemployment rate’ served as an indicator of economic (and social) upheaval faced by the entire community, irrespective of individual employment status.

Of note, the previously described uncertainty regarding the coroner's method of defining employment status would have tended to overstate the ‘unemployed suicides’. Therefore, more precise classification would probably have further accentuated the direction of our findings.

There are notable limitations to this type of ‘macro level’ study, in which we have examined neither community nor individual level factors that certainly must contribute to suicide.8 9 It is thus critical to disentangle the effects of a particular individual's employment status, which can serve as a marker of a person's current life circumstances and broader indicators of a community's economy. For suicide, the local economy serves as a broad context in which personal stories unfold. Such contexts can be distress-promoting or distress-reducing, and in turn we expect that they may prove either conducive or protective for adverse outcomes such as suicide. Our results suggest that we must view these as dynamic substrates, such that the nature of contributions from different subgroups of the overall population (eg, employed and unemployed) will change over time in a fashion driven by broad external events.

It is not inconsistent to view broad social or economic contexts as potentially important factors among the overall population forces that influence suicide among individuals, while appreciating findings such as those of Beautrais et al11 that the largest population attributable risk is depression rather than unemployment. This study underscores that an economically distressed environment affects suicides among the employed as well as the unemployed. Moreover, whatever the factors contributing to a person's broader social circumstances, there is nothing in our data to suggest that these change the specific psychological or psychopathological presentations during the moments immediately before death by suicide.

Fortunately, the vast majority of distressed people do not kill themselves. It is beyond the scope of this study to discern what protects most individuals while others succumb. These results underscore the need for future studies that are capable of examining these issues at population, community and individual levels. Furthermore, they reinforce the need to distinguish precisely between population-level concepts such as rates, from individual-level characteristics such as employment status.

What is already known on this subject

Existing studies suggested that a strong correlation exists between unemployment rates and suicide rates. However, the exact mechanism of how they interact with one another is not known.

What this study adds

Despite the strong correlation (0.86) between the unemployment rates and suicide rates for the whole period (2000–6), it was found that the rates of suicide within the employed and unemployed groups moved in the opposite direction to the population trend in periods of economic recession and recovery. The suicide rate among unemployed decreased in the period of economic recession and increased in the period of recovery. Therefore, it is important to be able to distinguish precisely between population-level concepts, such as rates from individual-level characteristics such as employment status, in devising an evidence-based suicide prevention strategy.


The authors would like to thank the two reviewers for constructive comments and Carmen Lee for data analysis support.

Appendix 1

Decomposition of differences between rates

Decomposition of differences between the suicide rates of the two periods 2000–3 and 2003–6

Define the original difference as Δ for the difference in time 1 and time 2.Δ=SRt1SRt2=iCit1·Mit1iCit2·Mit2=iCit1·Mit12+iCit1·Mit12iCit2·Mit22iCit2·Mit22+iCit1·Mit22iCit1·Mit22+iCit2·Mit12iCit2·Mit12=iCit1·[Mit1+Mit22]iCit2·[Mit1+Mit22]+iMit1·[Cit2+Cit12]iMit2·[Cit2+Cit12]=i(Cit1Cit2)·[Mit1+Mit22]+i(Mit1Mit2)·[Cit2+Cit12]=[difference in age composition]·[weighted by average age - specific mortality]+[difference in rate schedule]·[weighted by average age composition]=contribution of age compositional differences toΔ+contribution of rate schedule differences toΔWhere SRt1 is the suicide rate in time 1, Cit1 be the ith age composition in population in time 1, Mit1 is the suicide rate in the ith age group in the population in time 1.

The relevant information for computing the difference between 2003 and 2006 is enclosed for reference.

No of suicide deathsNo of totalSuicide rate
Employed2283 035 4007.5
Unemployed2651 45 100182.6
Total4933 180 50015.5
Employed3493 042 50011.5
Unemployed3552 49 300142.4
Total7043 291 80021.4
Employed2253 236 3007.0
Unemployed3031 54 900193.5
Total5283 391 20015.6



  • Funding This work was supported in part by the Research Grant Council of the Hong Kong Government (for PSFY) NIH grant D43TW005814 from the Fogarty International Center and NIH grant P20MH071897 from NIMH/NIDA (for EDC).

  • Competing interests None.

  • Patient consent Obtained.

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