Effect of economic crisis on suicide cases: An ARDL bounds testing approach

Abstract

Background:

Suicide events observed in various groups, community or countries, especially in the periods of economic recession. It is thought that suicide cases increase when people’s income decreases dramatically and they lose their jobs.

Aim/Objective:

In this study, it was aimed to investigate whether the 2008 economic crisis had any effect on suicides in the United States.

Methods:

Autoregressive distributed lag method was used. For the purpose of the study, the number of suicide-related deaths was taken as the dependent variable, while unemployment rates and 2008 economic crisis were taken as independent variables.

Findings:

The short-term and long-term relationships obtained within the scope of the study indicated that the 2008 economic crisis had a statistically significant effect on suicide cases in the United States.

Results and Conclusion:

It can be said that the results of this study are consistent with the information which emphasizes that economic crises increase suicide cases in the literature.

Keywords

Suicide economic crisis ARDL

Introduction

Suicide events observed in various groups, community or countries throughout the history of mankind have been a problem that is not only related with the community and mental health professionals, but it is also a phenomenon with economic and cultural aspects. The act of suicide, which happens a lot or rarely in every society, is perceived as abnormal behavior by some societies (Deniz, Ersöz, İldeş, & Türkarslan, 1995). Every year, around 800,000 individuals commit suicide and more people attempt suicide. Suicides are the tragedies that affect families, societies and countries and leave long-term permanent effects on the people that are left behind (World Health Organization (WHO), 2018). Durkheim (2005) defines suicide as fatal acts that are directly or indirectly performed by an individual and caused due to positive or negative behaviors. Suicide has various reasons, yet is performed in many ways. In addition, it is obvious that suicide is not a characteristic of a disease (Durkheim, 2005).

Although there is a significant correlation between suicide and mental illnesses (depression and illnesses due to alcohol use) in high-income countries, suicide mainly occurs due to high stress circumstances such as financial issues, separation from spouse or family, severe pain caused by illnesses and chronic diseases. In addition, there is a strong correlation between suicide behaviors and events such as conflict, disaster, violation, abuse and social exclusion (WHO, 2018).

The studies regarding suicide have gained momentum with the studies of Durkheim, and according to the observations, there has always been an increase in suicide rates during the economic crises. In 2008, the United States and European countries had entered into a long-term economic crisis as a result of economic activity slowdown together with the rigid increases in the unemployment rates (Dolls, Fuest, & Peichl, 2012). The increase in suicide rates during the economic crisis periods is highly associated with the fear of unemployment and unemployment itself. Scientific studies also argue whether economic crisis really has an impact on the increase in suicide cases (Laanani, Ghosn, Jougla, & Rey, 2015; O’Connor & Pirkis, 2016).

Italian doctor Morselli indicated that agricultural problems, poverty and economic crises that continue for many years progressively increase the mental illness rates, which in turn increase suicide cases (Morselli, 1882). Many western countries have experienced increases in unemployment rates and suicide cases during the Great Depression that occurred at the end of 1920s and in the early 1930s. In the early 1990s, with the disintegration of the Soviet Union, the member countries (Russia, Belarus, Ukraine and Baltic countries) had entered into severe recession and major social transformation phase. During that period, there had been significant increases in suicide cases in Russia, Belarus, Ukraine and Baltic countries (O’Connor & Pirkis, 2016).

This study aimed to identify whether the 2008 economic crisis that was originated in the United States and that affected many countries had any influence on the suicides that occurred in the United States. Within this context, the method of autoregressive distributed lag (ARDL) was utilized.

Method

Within the framework of study objective, the number of deaths by suicides in the United States (in 100,000 deaths) was taken as the dependent variable, while the unemployment rates and 2008 crisis were considered as independent variables. The data set of this study was taken from OECD health and economic data (OECD, 2019a; OECD, 2019b). The descriptions and abbreviations of study variables are given in Table 1.

Table 1.

Study variables.

Variables	Abbreviation	Logarithm	Logarithmic difference
Dependent variables
Suicide numbers	sui	Lsui	dLsui
Independent variables
Unemployment rate	unemp	Lunemp	dLunemp
2008 economic crisis	2008 crisis	–	–

This study is a time series analysis with ARDL method. In the time series analyses, logarithmic transformation is first applied on the variables to normalize the data and harmonize the contradictory values in the data set. Since the possibility to observe spurious regression is high in the time series analyses, the data on variables were tested in terms of stationarity afterwards. Spurious regression occurs as a result of trend similarities among variables in time although there is not any statistically significant relationship between the variables in reality. In the spurious regression models, a high level coefficient of determination (R²) is generated although there is not any significant effect of independent variables on the dependent variables. This is a misleading situation, and the stationarity of data should be checked prior to time series analysis (Granger & Newbold, 1974; Gujarati, 2009). Due to the relevant reasons, the logarithm of study variables was taken before the analysis and they were checked in terms of stationarity. Pursuant to ARDL analysis, all of the variables covered under the established model should have the stationarity at the level of (I(0)) or first degree (I(1)) (Pesaran, Hashem, & Smith, 2001).

Error correction model (ECM) was performed to identify the timeframe that is required for the elimination of deviation in the dependent variables and to reflect the short-term effect of independent variables on the dependent variables. In ECM, a lagged term of model residuals generating the long-term relationship among the variables (ECM (–1)) is taken, and the ECM term coefficient shows how much of disequilibrium that occurred in the short term can be fixed in the long run (Akel & Gazel, 2014). The error correction term should be negative and statically significant under ECM (Saçık & Karaçayır, 2015). The ECM that was used under the study was created with the least squares technique.

The study was conducted with the data from the years of 1990–2016 (27 years) taken from the database of Organization for Economic Cooperation and Development (OECD). EViews 9 software was used for the analyses conducted for the study with 95% confidence level.

Findings

In this study, the stationarity of variables was tested with the Augmented Dickey–Fuller (ADF) test prior to ARDL analysis, and the variables were found to have no stationarity. Therefore, the study variables that were applied with logarithmic transformation were converted into stationary by taking their differences from the first level. The data from all variables were identified to be stationary in terms of their difference on the first level (Table 2). Hence, this model can be considered to be in compliance with ARDL analysis.

Table 2.

Values of p before and after ADF test.

Variables	p values before ADF test			p values after ADF test (first difference)
Variables	Intercept	Trend and intercept	None	Intercept	Trend and intercept	None
Lsui	.6368	.8677	.8368	.0350	.0115	.0029
Lunemp	.0633	.0585	.4670	.0118	.0434	.0007

ADF: augmented Dickey–Fuller.

After the variables were made stationary, the lag lengths of study were identified with vector autoregressive model (VAR). The maximum lag length of ARDL analysis was selected as 2 since 2 lag length was the most preferred lag length among the Likelihood Ratio (LR) analysis, Final Prediction Error (FPE), Akaike Information Criterion (AIC), Schwarz Information Criterion (SC) and Hannan-Quinn Information Criterion (HQ) models generated after VAR (Table 3).

Table 3.

VAR model results.

Lag	Log L	LR	FPE	AIC	SC	HQ
0	55.10021	–	2.61e–06	–4.341684	–4.194428	–4.302617
1	85.23027	50.21676	4.53e–07	–6.102522	–5.513495	–5.946253
2	106.5895	30.25887*	1.69e–07*	–7.132456*	–6.101659*	–6.858985*

VAR: vector autoregressive model; LR: likelihood ratio; FPE: final prediction error; AIC: Akaike information criterion; SC: Schwarz information criterion; HQ: Hannan–Quinn information criterion.

Chosen Lag Length according to the criteria on the column.

AIC was determined as model selection criteria in ARDL analysis. Pursuant to the analysis results, the most suitable model is 1, 0, 0. Table 4 provides the results from model. According to Table 4, one lagged dLint dependent variable and no lagged versions of independent variables were selected as optimum model. It is considered that the model is statistically significant as F statistics of model has p value <.05; it has normality assumption since it has p value >.05 from Jarque–Bera Test; it does not have any heteroscedasticity problem as F statistics has p value >.05; it does not have any serial correlation problem as F statistic from Breusch–Godfrey Serial Correlation LM Test has p value >.05 and finally it does not have any building error as F statistic of Ramsey Reset Test has p value >.05. Moreover, the model is considered not to have any spurious regression problem as the difference between Durbin–Watson value (1.95) and R² value (0.30) is high.

Table 4.

ARDL test results.

Dependent variable: dLsuiModel selection method: Akaike information criteriaSelected model: ARDL (1, 0, 0)
Variables	Coefficient	t	p
dLsui (–1)	0.063679	0.266866	.7922
dLunemp	0.021673	0.669794	.5103
2008 crisis	0.024660	2.100030	.0440
Constant	–0.004742	–0.808645	.4278
R ²	0.308809	Durbin–Watson Statistics	1.950249
Adjusted R²	0.210067	Ramsey Reset Test (F statistics)	0.1315
p value (F statistics)	0.047448	Jarque–Bera Normality Test	0.7675
		ARCH Değişen Variance Test (F statistics)	0.6921
		Breusch–Godfrey Otokorelasy on LM Test (F statistics)	0.8370

ARDL: autoregressive distributed lag; ARCH: autoregressive conditional heteroscedasticity; LM: Lagrange multiplier.

In addition to the aforementioned diagnostic tests, the variable data were analyzed with Cusum and Cusum-Square (SQ) tests to identify whether such data have structural breaks (Figure 1), where pursuant to Figure 1, Cusum-SQ graph shows a slight deviation. According to Yakışık and Çetin (2014), such small breaks do not deteriorate the stability of model provided that it is again in the confidence interval and is not continuous. Therefore, the model established under this study is considered not to have any structural breaks.

Figure 1.

CUSUM and CUSUM-SQ graphs.

Pursuant to the study model, the independent variable of 2008 crisis has a significant effect on the dependent variable of dLint. Pursuant to this model. the independent variables explain 30% of the change in the dependent variable (R² = 0.30).

Table 5 provides the results on the long-term effect of independent variables on the dependent variable. Hence, the 2008 crisis has statistically significant and positive effect on suicide cases in the long term (p < .05).

Table 5.

ARDL long term model (1, 0, 0).

Dependent variable: dLsui
Variables	Coefficient	t	p
dLunemp	0.023147	0.692550	.4962
2008 crisis	0.026337	2.640618	.0153
Constant	–0.005064	–0.842918	.4088

ARDL: autoregressive distributed lag.

Table 6 gives the ECM results, which reflect that the 2008 crisis has statistically significant and positive effect on suicide cases in the short term (p < .05). In addition, the one-lagged version of error terms is negative and statistically significant, which indicates that in the short term, 7.3% of deviations from the number of deaths from suicides can be eliminated (p < .05).

Table 6.

ARDL error correction (short term) model (ECM) (1, 0, 0).

Dependent variable: dLsui
Variables	Coefficient	t	p
dLunemp	0.026156	0.821032	.4209
2008 crisis	0.026859	2.876291	.0090
ECM (–1)	–0.073033	–2.305094	.0320
Constant	–0.005253	–0.938118	.3588
R ²	0.309525	Durbin–Watson Statistic	1.761918
Adjusted R²	0.210886	p value (F statistic)	0.000976

ARDL: autoregressive distributed lag; ECM: error correction model.

Pursuant to the short-term and long-term coefficients generated within the scope of the study, it is possible to say that the 2008 economic crisis has a statistically significant effect on suicide cases in the United States; hence, the results of this study are consistent with the literature as the economic crises increase the suicide cases.

Discussion and Conclusion

While individual characteristics are the biggest reason for suicides, recent empirical studies indicate that the residences and socio-economic characteristics of individuals have an effect on suicide. Hence, the residence country or region of individuals may have positive or negative effects on the mental health of individuals (Evans, 2003; Patel et al., 2010; Roux & Mair, 2010; Shim et al., 2014; WHO, 2008, 2014). Within this context, there are a number of studies under the literature regarding the correlation between the socio-economic indicators and suicide numbers. In the related studies, factors such as poverty, income, socio-economic status, unemployment, population density, rural/urban population and financial and economic crisis are taken as socio-economic indicators (Agerbo, Sterne, & Gunnell, 2007; Burrows, Auger, Gamache, St-Laurent, & Hamel, 2011; Burrows, Auger, Roy, & Alix, 2010; Ceccherini-Nelli & Priebe, 2011; Chang et al., 2011; Cheung, Spittal, Williamson, Tung, & Pirkis, 2014; Ferretti & Coluccia, 2009; Kim, Jung-Choi, Jun, & Kawachi, 2010; Middleton, Sterne, & Gunnell, 2006; Milner, Mcclure, & De Leo, 2012; Murali & Oyebode, 2004; Razvodovsky & Stickley, 2009; Stark, Hopkins, Gibbs, Belbin, & Hay, 2007). For this study, the 2008 economic crisis in the United States and unemployment rate were taken as socio-economic indicators, and this study analyzes the effect of such indicators on suicides in the United States.

The research reflected that the 2008 economic crisis in the United States has long- and short-term effects on suicides such that the number of suicides has increased after the economic crisis. The reason for such a situation is considered to be originated due to the negative effect of crisis on the mental health of individuals and the decrease in the future expectations of individuals. Similar to this study, the study by Reeves et al. (2012) conducted in the United States showed that there has been an increase in the whole country after the economic crisis of 2008. Similarly, the study by Çıraklı and Yıldırım (2019) conducted in Turkey indicated that economic crisis has positive effect on suicide cases. Huikari, Miettunen and Korhonen (2019) conducted a study on 21 OECD countries which detected a significant increase in suicide cases during the economic crisis period. The study by Chang, Stuckler, Yip and Gunnell (2013) concluded that the economic crisis in 2008 has an effect in increasing suicide cases. Considering the short- and long-term effects of crises, the policies and practices developed to mitigate the negative effects of economic crises should cover not only the crises periods but also a long-term period that would be applicable post-crises. In addition, the number of suicide cases may be significantly diminished with the implementation of such policies where it is possible to foresee the economic crises.

This research also analyzed the short- and long-term effect of unemployment on suicide cases. Yet, the unemployment rate was found to have no effect on suicides in the short and long run. Similarly, the study by Gajewski and Zhukovska (2017) reflected that unemployment does not have any effect on suicide cases in the countries where social state is enforced. Unlike such studies, the study by Lin and Chen (2018) conducted in the United States showed that unemployment has asymmetrical effect on suicides. The study by Andrés and Halicioglu (2010) performed regarding Denmark and the long-term effect of unemployment on suicides reflected that unemployment has statistically significant positive effect on suicides. The study by Barr, Taylor-Robinson, Scott-Samuel, Mckee and Stuckler (2012) conducted in England emphasized that the two out of five suicide cases increasing among men are caused by unemployment. Finally, the study by Baumbach and Gulis (2014) indicated that unemployment has much higher effect in the countries with lower socio-economic levels.

The studies reflected that suicide is a multi-dimensional fact that is affected by various factors (Platt, Arensman, & Rezaein, 2019; Ren et al., 2019). However, this study is only about the effects of 2008 economic crisis and unemployment on suicide cases in the United States. Hence, future studies should assess the economic indicators as well as the effects of social indicators on suicide cases.

One of the limitations of this study is that it only covers the years of 1990–2016 and is only about United States since there are not any data from the previous years and from other countries. Hence, it is recommended that future studies should cover additional countries for longer periods. Since the study data are taken from the OECD database, similar studies should be conducted with the data from other national and international organizations, and the results should be compared accordingly.

Finally, it is crucial to develop economic, social and psycho-social support programs and to provide required supports for the families in need so that the negative effects of economic crises can be mitigated on the individuals and their families. Hence, the quality and quantity of mental health services should be enhanced during the economic crises, and public awareness activities should be conducted. In addition, preventive strategies toward improving the social awareness on the suicide concept should be developed and implemented.

Footnotes

Funding

The author(s) received no financial support for the research, authorship and/or publication of this article.

ORCID iD

Birol Yetim

References

Agerbo

Sterne

J. A.

Gunnell

D. J.

(2007). Combining individual and ecological data to determine compositional and contextual socio-economic risk factors for suicide. Social Science & Medicine, 64, 451–461.

Akel

Gazel

(2014). The co-integration linkages between exchange rates and BIST industrial index: an ARDL bound testing approach. Erciyes University Journal of Faculty of Economics and Administrative Sciences, 44, 23–41.

Andrés

A. R.

Halicioglu

(2010). Determinants of suicides in Denmark: Evidence from time series data. Health Policy, 98, 263–269.

Barr

Taylor-Robinson

Scott-Samuel

Mckee

Stuckler

(2012). Suicides associated with the 2008-10 economic recession in England: Time trend analysis. British Medical Journal, 345, Article e5142.

Baumbach

Gulis

(2014). Impact of financial crisis on selected health outcomes in Europe. European Journal of Public Health, 24, 399–403.

Burrows

Auger

Gamache

St-Laurent

Hamel

(2011). Influence of social and material individual and area deprivation on suicide mortality among 2.7 million Canadians: A prospective study. BMC Public Health, 11(1), Article 577.

Burrows

Auger

Roy

Alix

(2010). Socio-economic inequalities in suicide attempts and suicide mortality in Québec, Canada, 1990-2005. Public Health, 124, 78–85.

Ceccherini-Nelli

Priebe

(2011). Economic factors and suicide rates: Associations over time in four countries. Social Psychiatry and Psychiatric Epidemiology, 46, 975–982.

Çıraklı

Yıldırım

H. H

. (2019). The impacts of economic crises on health behaviors of population: An ARDL bounds testing approach. Çankırı Karatekin Üniversitesi İİBF Dergisi, 9, 101–123.

10.

Chang

S. S.

Sterne

J. A.

Wheeler

B. W.

T. H.

Lin

J. J.

Gunnell

(2011). Geography of suicide in Taiwan: Spatial patterning and socioeconomic correlates. Health & Place, 17, 641–650.

11.

Chang

S. S.

Stuckler

Yip

Gunnell

(2013). Impact of 2008 global economic crisis on suicide: Time trend study in 54 countries. British Medical Journal, 347, Article f5239.

12.

Cheung

Y. T. D.

Spittal

M. J.

Williamson

M. K.

Tung

S. J.

Pirkis

(2014). Predictors of suicides occurring within suicide clusters in Australia, 2004-2008. Social Science & Medicine, 118, 135–142.

13.

Deniz

Ersöz

A. G.

Ildes

Türkarslan

(1995). A study on suicides and suicide attempts in the official records of the province of Batman during the period of 1995–2000. Journal of Social Policy Studies, 1.

14.

Dolls

Fuest

Peichl

(2012). Automatic stabilizers and economic crisis: U.S. vs. Europe. Journal of Public Economics, 96, 279–294.

15.

Durkheim

(2005). Suicide: A study in sociology. London, England: Routledge.

16.

Evans

G. W.

(2003). The built environment and mental health. Journal of Urban Health, 80, 536–555.

17.

Ferretti

Coluccia

(2009). Socio-economic factors and suicide rates in European union countries. Legal Medicine, 11, S92–S94.

18.

Gajewski

Zhukovska

(2017). Short-run and long-run effects of unemployment on suicides: Does welfare regime matter? European Journal of Public Health, 27, 1038–1042.

19.

Granger

C. W. J.

Newbold

(1974). Spurious regressions in econometrics. Journal of Econometrics, 2, 111–120.

20.

Gujarati

D. N.

(2009). Basic econometrics. New Delhi, India: Tata McGraw-Hill Education.

21.

Huikari

Miettunen

Korhonen

(2019). Economic crises and suicides between 1970 and 2011: Time trend study in 21 developed countries. Journal of Epidemiology and Community Health, 73, 311–316.

22.

Kim

M. H.

Jung-Choi

Jun

H. J.

Kawachi

(2010). Socioeconomic inequalities in suicidal ideation, parasuicides, and completed suicides in South Korea. Social Science & Medicine, 70, 1254–1261.

23.

Laanani

Ghosn

Jougla

Rey

(2015). Impact of unemployment variations on suicide mortality in Western European countries (2000-2010). Journal of Epidemiology and Community Health, 69, 103–109.

24.

Lin

Y. H.

Chen

W. Y.

(2018). Does unemployment have asymmetric effects on suicide rates? Evidence from the United States: 1928–2013. Economic Research-Ekonomska Istraživanja, 31, 1404–1417.

25.

Middleton

Sterne

J. A.

Gunnell

(2006). The geography of despair among 15-44-year-old men in England and Wales: Putting suicide on the map. Journal of Epidemiology and Community Health, 60, 1040–1047.

26.

Milner

Mcclure

De Leo

(2012). Socio-economic determinants of suicide: An ecological analysis of 35 countries. Social Psychiatry and Psychiatric Epidemiology, 47, 19–27.

27.

Morselli

(1882). Suicide: An essay on comparative moral statistics. New York, NY: D. Appleton. Retrieved from https://archive.org/details/suicide00morsgoog

28.

Murali

Oyebode

(2004). Poverty, social inequality and mental health. Advances in Psychiatric Treatment, 10, 216–224.

29.

O’Connor

R. C.

Pirkis

(2016). The international handbook of suicide prevention. Chichester, UK: John Wiley.

30.

Organization for Economic Cooperation and Development (OECD). (2019a). Economy data. Retrieved from http://www.oecd.org/economy/

31.

Organization for Economic Cooperation and Development (OECD). (2019b). Health data. Retrieved from https://www.oecd.org/health/

32.

Patel

Lund

Hatherill

Plagerson

Corrigall

Funk

Flisher

A. J.

(2010). Mental disorders: Equity and social determinants. In Equity, social determinants and public health programmes (pp. 115–134). Retrieved from https://apps.who.int/iris/bitstream/handle/10665/44289/9789241563970_eng.pdf;jsessionid=BF70983600AC8830065FD3663CA3D9EC?sequence=1

33.

Pesaran

Hashem

Y. S.

Smith

R. J.

(2001). Bounds testing approaches to the analysis of level relationships. Journal of Applied Econometrics, 16, 289–326.

34.

Platt

Arensman

Rezaein

(2019). National suicide prevention strategies – Progress and challenges. Crisis, 40, 75–82.

35.

Razvodovsky

Stickley

(2009). Suicide in urban and rural regions of Belarus, 1990-2005. Public Health, 123, 27–31.

36.

Reeves

Stuckler

Mckee

Gunnell

Chang

S. S.

Basu

(2012). Increase in state suicide rates in the USA during economic recession. The Lancet, 380, 1813–1814.

37.

Ren

You

Zhang

Huang

Conner

B. T.

Sun

Lin

M. P.

(2019). Differentiating suicide attempters from suicide ideators: The role of capability for suicide. Archives of Suicide Research, 23, 64–81.

38.

Roux

A. V. D.

Mair

(2010). Neighborhoods and health. Annals of the New York Academy of Sciences, 1186, 125–145.

39.

Sacik

S. Y.

Karacayir

(2015). The financing of current account in Turkey: ARDL bound test approach. Selcuk University Journal of Social Sciences Institute, 33, 155–166.

40.

Shim

Koplan

Langheim

F. J. P.

Manseau

M. W.

Powers

R. A.

Compton

M. T.

(2014). The social determinants of mental health: An overview and call to action. Psychiatric Annals, 44, 22–26.

41.

Stark

Hopkins

Gibbs

Belbin

Hay

(2007). Population density and suicide in Scotland. Rural and Remote Health, 7(3), 672.

42.

World Health Organization. (2008). Closing the gap in a generation: Health equity through action on the social determinants of health: Final report of the commission on social determinants of health. Author. Retrieved from https://apps.who.int/iris/bitstream/handle/10665/43943/9789241563703_eng.pdf?sequence=1

43.

World Health Organization. (2014). Preventing suicide: A global imperative. Luxembourg. Retrieved from who.int/mental_health/suicide-prevention/world_report_2014/en/

44.

World Health Organization. (2018). Suicide. Retrieved from https://www.who.int/news-room/fact-sheets/detail/suicide

45.

Yakisik

Cetin

(2014). The impact of education, health and level of technology on economic growth: ARDL bounds test approach. Socioeconomy, 21, 169–186.