(This is a sample cover image for this issue. The actual cover is not yet available at this time.) This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright Author's personal copy The Journal of Socio-Economics 40 (2011) 723–731 Contents lists available at SciVerse ScienceDirect The Journal of Socio-Economics journal homepage: www.elsevier.com/locate/soceco Socio-economic determinants of suicide in Japan Antonio R. Andrés a,b,∗ , Ferda Halicioglu c , Eiji Yamamura d a Aarhus University, Institute of Public Health, Bartholins Allé 1, 8000 Aarhus C, Denmark Associate researcher, Institute of Economic Analysis & Prospective Studies (IEAPS), Al Akhawayn University, Ifrane, Morocco c Department of Economics, Yeditepe University, 34755 Istanbul, Turkey d Seinan Gakuin University, Department of Economics. Fukuokashi Sawaraku Nishijin 6-2-92, 814-8511, Japan b a r t i c l e i n f o Article history: Received 12 April 2011 Received in revised form 12 August 2011 Accepted 30 August 2011 JEL classification: C22 I12 Keywords: Cointegration Suicide Time series Japan a b s t r a c t Japan has the highest suicide rates among the OECD countries and this public health problem seems to be accelerating in over the recent decades. Investigating and understanding the suicidal behaviour is of crucial importance to society and health policy makers. Such an investigation could provide with useful information for those responsible in formulating the national policies on suicide prevention. This study estimates dynamic econometric models for total, male and female suicides in Japan for the period of 1957–2009. Using the ARDL approach to cointegration, we find that the associations of suicide with sociological factors (divorce and fertility rates) were stronger than those with economic factors (per capita GDP and unemployment) for females. © 2011 Elsevier Inc. All rights reserved. 1. Introduction Suicide is a very serious public health problem. The World Health organization (henceforth, WHO) estimates that worldwide there are approximately one million of deaths from suicide each year and 20 times this number of people have attempted suicide. According to many medical professions, suicide is considered to be the result of depression and other psychiatric disorders (Mann et al., 2005). Although Japanese life span is the longest in the world, it has nevertheless one the world highest suicide rates with nearly 33,000 people killing themselves in 2009. According to statistical data from the WHO, Japan, in 2004, reports the highest suicide rate with 24 per 100,000 people among the OECD countries. From 1995 to 2009, the total suicide rate increased from 17 to 25 per 100,000 people.1 Suicide is also associated with substantial economic costs (with particularly health care costs). In particular, Chen et al. (2009a) suggested that the costs associated with suicides were around 197 million USD in 2006 alone even if indirect costs such as psychological counseling expenditure were not taken into account. ∗ Corresponding author. E-mail address: ara@folkesundhed.au.dk (A.R. Andrés). 1 Data source is as follows. Periods 1955–2004: Statistics Bureau, Ministry of Internal Affairs and Communications (2006). Historical Statistics of Japan Volume 1 (New Edition). Tokyo: Japan Statistical Association. Periods 2005–2009: National Police Agency. http://www8.cao.go.jp/jisatsutaisaku/link/keisatsutyo.html (accessed 16.06.10). 1053-5357/$ – see front matter © 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.socec.2011.08.002 In comparison, there have been European studies highlighting the enormous costs of completed suicides. For instance in Ireland (Kennelly et al., 2005), the total cost has been shown to be 2.04 million Euros and in Scotland 1.88 million Euros (McDaid et al., 2007). In some Japanese media, the total costs of suicide and depression was reported to be about 2.7 trillion Yen in 2009 (available at http://search.japantimes.co.jp/cgi-bin/nn20100908a2.html). Prevention of suicide has been integral part of the Japanese public health agenda. The Japanese Government aimed to reduce the annual incidence of suicide and for this purpose implemented “the Basic Act of Suicide Prevention (jisatsu taisaku kihon hou)” in 2006. In addition, to the role of government, informal social ties regarded as social capital is also thought to play an important role in preventing suicide in Japan (Yamamura, 2010). In fact, community based suicide prevention programs were introduced in Akita prefecture (see Motohashi et al., 2004). For making the policy effective, it is important to ascertain how and why suicide rate of Japan is so high based on empirical analysis. Apart from the interest in describing and explaining suicidal behaviour, employing rates of suicide as a societal well-being indicator has several advantages. First, suicide rates are a more reliable and objective indicator of well-being compared to self-reported well-being measures (such as life satisfaction or self-reported happiness). Second, suicide rates do not have the common problems associated with survey data on self-reported well-being. Selfreported measures are often challenged on the basis of reliability and validity (see for an excellent discussion, see Bertrand and Mullainathan, 2001). It has been also shown that there is a high Author's personal copy 724 A.R. Andrés et al. / The Journal of Socio-Economics 40 (2011) 723–731 correlation between suicide and subjective well-being at individual and aggregate level (for instance, Koivumaa-Honkanen et al., 2001). Unlike as self-reported measures, suicide data is the kind of data that is more prone to make cross country comparisons. Using self reported data comparisons are still difficult because of problems with interpersonal comparisons of utility. Recently, an American study concluded that the determinants of well-being are the same determinants of suicide (Daly and Wilson, 2009). Despite of its importance, and a growing concern for the factors driving suicide mortality, suicide in Japan has received little attention. Although there have been some recent attempts, mainly using panel data techniques, in this direction (e.g. Chen et al., 2009b; Yamamura, 2010). Traditional theories of suicide (Durkheim, 1951; Hamermesh and Soss, 1974) have been tested using time series data for a large number of countries (e.g. Yang, 1992; Yang and Lester, 1990; Yang et al., 1992; Chuang and Huang, 1996; Platt and Hawton, 2000; Stack, 2000; Chang et al., 2010). Some researchers have investigated the socioeconomic determinants of suicide using time series data for Japan (Yamasaki et al., 2005, 2008). There are however few studies which employ causality or cointegration framework to investigate the causality between suicide and its socioeconomic determinants. A recent study is that of Inagaki (2010) who employs a Vector Autoregressive (VAR) model. But this methodology has several shortcomings. First, this methodology requires the set of variables to be split into exogenous and endogenous variables. Second, the variables should be integrated of order 1. This study aims at contributing to the empirical studies of Japanese suicide by applying a relatively new time series cointegration technique known as the Auto Regressive Distributed Lag (ARDL) bounds testing procedure. The ARDL approach to cointegration is preferable to other conventional cointegration procedures (Engle and Granger, 1987). One of the reasons for preferring the ARDL approach to cointegration it is that overcome the problem of potential endogeneity of some regressors and serial correlation, which might lead to biased estimates of the cointegrating coefficients. Another reason is that this technique does not require pre-testing for the order of integration of the underlying time series. Moreover, the results from this approach to cointegration are more robust in presence of small samples (such as in this study) than in other cointegration techniques. Finally, as opposed to multivariate cointegration techniques such as Johansen and Juselius (1990), it allows the cointegration relationship to be estimated by ordinary least squares (OLS) once the lag order of the model is chosen. In addition to studying the total suicides, we also analysed male and female suicides separately, as the underlying determinants of suicide could differ between the sexes (e.g. Andrés, 2005; Chuang and Huang, 2007; Yamamura, 2010). Understanding the gender differences might be also important in informing appropriate policy formulations. The remainder of this paper is organized as follows. The next section presents the socio-economic situation of Japan relating to the suicides. Section 3 describes our empirical model and methodological approach. Section 4 displays our empirical results along with some discussions. Section 5 is the conclusion. Fig. 1. Changes of per capita GDP. what follows, we begin with a simple description of the potential socio-economic factors affecting suicidal behaviour. As shown in Fig. 1 illustrating changes of real per capita income, Japan has experienced the rapid economic growth in the post war period and became among the most developed countries. Japanese people enjoyed the rise in income and are thought to be satisfied with this life style change accompanied with economic growth. Concerning the growth rate of real per capita GDP, it drops constantly and to below zero several times after 1990s. We can see from Fig. 2 that the unemployment rate has been also low level until mid1990s, however, exceeded 3% after mid-1990s. This seems to reflect the depression period after 1992 when the prosperity of the “bubble economy” (from mid 1980 to the beginning of the 1990s) came to an end in Japan. In this period, number of business bankruptcies also steeply increased in this period because of macro level economic stagnation. In particular, it was difficult for owners of small and medium size enterprise to run business. Economic recession lead a lot of people to face the difficulty and suffer distress. Transition of divorce rate in Fig. 3 shares similarity with unemployment rate in the point that after entering the recession period divorce rate remarkably increased. The increase in divorce rate can be in part caused by the economic recession. Marriage leads couple to be integrated into the new social network, which is expressed as “when you get married, you get married for the people around you” (Brinton, 1993, p. 99). Hence, divorce seems to be more stigmatized in Japan than in the Western countries because of the greater importance of extended family and kinship ties in marriage (Ono, 2006). That is, people who encounter the economic difficulty more likely to experience divorce and so lose the psychological support from family and kinship ties. During the economic depression period, not only economic difficulty but also social stigma caused by divorce lead people to suffer from increase of distress in Japan. In Japan, over 60% of the individuals committing suicide were identified as depressive (Nakao and Takeuchi, 2006). 2. Review of the socio-economic situation of Japan Total life expectancy at birth of Japanese is 82 years old, which leads the world in longevity (WHO, 2006; Nakao and Takeuchi, 2006). However, suicide rate is obviously higher than other OECD countries, which becomes the one of major problem in the modern Japan society (Chen et al., 2009b). Japan’s suicide problem is very different from those of other OECD countries because the impact of the socioeconomic variables on suicide is greater in Japan than in other OECD countries (Chen et al., 2009b). To implement appropriate suicide prevention policies, it is important to ascertain how and why suicide rate of Japan is so high based on empirical analysis. In Fig. 2. Changes of unemployment rate (%). Author's personal copy A.R. Andrés et al. / The Journal of Socio-Economics 40 (2011) 723–731 Fig. 3. Changes of divorce rate (%). Fig. 4 shows that rates of suicide has obviously decreased from mid 1950s to 1970, and then slightly increased until mid-1990s for males and females. For example, in Korea with the similar socio-cultural background, the remarkable increase of suicide rate from 1997 to 1998 under the economic recession period (Khang et al., 2005). However, in case of Korea, not only male but also female suicide rate increased. It is surprising to observe that in the end of 1990s suicide rate of male has drastically increased whereas that of female was stable. According to Fig. 4, there was a marked rise in female suicide rates from 1997 to 1998 although its magnitude was smaller than for male suicide rates. Indeed, in late 1997, Hokkaido Takushoku Bank (one of major commercial banks) and Yamaichi Securities (one of Major securities) became bankrupt. Further 1998, The Long Term Credit Bank of Japan and the Nippon Credit Bank were nationalized, implying the old economic regime’s failure (Cargill, 2006). Taken together, these results imply that the problem of committing suicide became remarkably serious, especially for males. The question arises “why is there difference of suicide rate between male and female?”. An imbalance of increases in suicide appears to come from the different impact of various factors between males and females. As pointed out by Nakao and Takeuchi (2006), the most drastic increase has involved middle-aged males partly because most middle-aged males may be too busy to visit a clinic when they feel mental distress. Therefore, the case of Japan is suitable for examining how committing suicide depends upon gender and differences in the impact of socioeconomic factors. 3. Literature review 3.1. International experience Sociologists have played an important role in providing the theory of suicide. Durkheim (1951) viewed the suicide as a soci- Fig. 4. Changes of rate of suicides. Note: Number of total suicides per total population (100,000), number of male suicides per male population (100,000), number of female suicides per female population (100,000). 725 ological phenomenon. He argues that suicide is related to both social integration and social regulation. Economists claim that suicide involves rational economic decision making. Hamermesh and Soss (1974) were the first to provide an economic theory of suicide. According to their economic model an individual decides to commit suicide when the discounted expected lifetime utility remaining to him falls below some threshold level. This model also predicts that suicide rates would increase with age, unemployment and decrease with income (Hamermesh and Soss, 1974). Recently, Suzuki (2008) incorporates the concept of income uncertainty within the model of Hamermesh and Soss (1974). These approaches (sociological and economics) motivate many of the control variables included in a variety of econometric studies of macro level determinants of suicide. According to the Hamermesh and Soss’s model, the higher future expected income is, the higher is the expected utility; thus, living is relatively more attractive than committing suicide, and a higher income should lower suicide rates. However, Durkheim postulates that higher income levels increase independence (the opposite of social integration) and might lead to a higher suicide rate. Along this line, Lester (1996) and Unnithan et al. (1994) state that economic development increases rates of suicide. Both the existing economic and sociological theories are inconsistent, and they do not permit a determination of whether income or economic growth may have a positive or negative effect on suicide. Durkheim (1951) suggests that changes in income are more likely to be relevant for suicide than the absolute level of income. The empirical evidence for the effect of income on suicide is mixed, however. Though some empirical studies indicate that suicide rates have a positive association with income (e.g. Hamermesh, 1974; Jungeilges and Kirchgässner, 2002; Viren, 1999), there are many others suggesting the opposite effect (e.g. Andrés, 2005; Brainerd, 2001; Neumayer, 2003; Chuang and Huang, 1997, 2007; Minoiu and Rodríguez, 2008; Altinanahtar and Halicioglu, 2009; Andrés and Halicioglu, 2010). Others have reported an insignificant effect of income on suicide (Ruhm, 2000; Cuellar and Markowitz, 2006). The significant negative correlation effect seems to be stronger for men than for women Qin et al. (2003). Another economic variable that has received a lot of attention is the unemployment rate. Unemployment implies less economic opportunity, lowering an individual’s expected income and therefore increasing the likelihood of a person’s committing suicide. The unemployment rate is often used as a proxy variable for economic hardships and lifetime earnings, because measuring an agent’s lifetime income is not easy in practice (Koo and Cox, 2008). But unemployment might be also associated with factors such as depressive episodes, anxiety, and loss of self-confidence that might lead directly to suicide. Much of the empirical literature reports a positive relationship, associating higher unemployment with higher suicide rates (for example, Brainerd, 2001; Ruhm, 2000; Chuang and Huang, 1997, 2007; Lin, 2006; Andrés, 2005; Koo and Cox, 2008; Minoiu and Rodríguez, 2008). Furthermore, the impact of unemployment might also differ across gender. In particular, male suicide rates are significantly affected by unemployment, but female suicide rates are not (Chuang and Huang, 1997). As mentioned above, Durkheim (1951) indicates that suicide is influenced by other factors. These factors relate to the way in which individuals are integrated into a social group that is regulated by norms and conventions. This sociological approach predicts that lower levels of social integration and regulation are associated with higher societal suicide rates. From this perspective divorce and fertility rates can be viewed as indicators of social integration. Divorce can be also a traumatic event for the individuals involved as well as for other affected parties, and it might lead individuals toward isolation and reduced poor psychological well-being. Thus, higher divorce rates might be expected to have a positive correlation with Author's personal copy 726 A.R. Andrés et al. / The Journal of Socio-Economics 40 (2011) 723–731 suicide rates. Another explanation is that a divorced individual has lower utility than a married one because marriage has a mercenary value (Becker, 1974). Koo and Cox (2008) also suggest that divorced people have less utility than married people and therefore they are more likely to commit suicide. Several studies have reported a positive association between divorce and suicide (e.g. Andrés, 2005; Chuang and Huang, 1997, 2007; Kunce and Anderson, 2002; Lester, 1996; Neumayer, 2003). Also, some papers show that the male suicide rate is more sensitive to divorce than the female suicide rate (e.g. Koo and Cox, 2008; Andrés, 2005; Yamamura, 2010; Neumayer, 2003). Again, endogeneity concerns are relevant here, as divorce might be also related to mental health problems. It should be also noted that this variable might capture the influence of diverse societal problems. Durkheimian arguments of social integration suggest that increased fertility rates should be associated with lower levels of suicide, as the presence of children promotes social and family ties. By increasing social integration, these factors lower the likelihood of a person’s committing suicide. Empirical research has documented the existence of a protective effect of fertility against suicide (e.g. Andrés, 2005; Neumayer, 2003; Chuang and Huang, 2007). However, some studies like Chen et al. (2009b) and Lester (1995) show that the birth rate has either a positive impact or no impact on suicide rates. One possible explanation for the latter result is that childcare may put excessive strain on a parent or be too much of an economic burden, thus leading to suicidal behaviour (Chen et al., 2009b). Endogeneity issues might be relevant here, as better functioning people are more likely to have children. Lastly, the gender differences in suicide represent a double puzzle: Whilst rates of suicide are far higher among males, females have higher rates of non-fatal attempts. This suggests there may be different responses by males and females to the control variables used in the formal analysis. In light of the gender differential in suicidal behaviour (e.g. Minoiu and Rodríguez, 2008; Altinanahtar and Halicioglu, 2009; Andrés, 2005; Yamamura, 2010), we run separate models for males and females. Although the cause of these differences has not been sufficiently investigated (Yamamura, 2010). In sum, the formal literature provides ambiguous results on the ways socioeconomic factors relate to male and female suicide rates. The existing literature has not come to a firm conclusion about the correlates of suicide. This is due to different countries employed in the empirical analysis, more points of the time, and the statistical techniques employed (time series/cross-section analysis). Nevertheless, of all the variables considered, the results corresponding to social factors such as divorce and fertility seem to be more robust than those related to economic factors such as unemployment and income. Nonetheless, the socio-economic control variables used in this paper appear to be among the relatively important determinants. employing a panel data approach by using Japanese data and OECD data analyse to what extent suicide in Japan is different from suicides in other countries. Inagaki (2010) using time series focuses on the link between income inequality and suicide. He finds a positive relationship between income inequality proxied by the Gini index and suicide rates. Kuroki (2010) is the most recent paper using Japanese data at municipality level. He provides evidence that unemployment has a positive significant effect on male suicide rates and that this effect differs across age groups, in particular, the largest effect is found in the 55–64 age group. He also finds a negative effect of unemployment on female suicide rates. That is, higher unemployment is associated with lower female suicide rates. They conclude that the impact of socioeconomic factors on suicide in Japan is greater than in OECD countries. Lastly, Yamamura (2010) using panel data at prefectural level suggests that social capital and divorce have an impact on suicide rates and that these effects are different between males and females. This leads us to anticipate that sociological factors plays more critical role on determining on suicide rates than other countries. Unlike previous studies of suicide in Japan, this work employs a new recently methodological approach using time series data to examine how suicide is related to socio-economic factors in Japan as in the short as well as in the long run. This approach is more robust in presence of small samples, and allows us to account for potential endogeneity of the variables included in the empirical model. Endogenity issues might lead to misleading results in past empirical studies. 5. Model and methodology Following the empirical literature on suicide (for an extensive review of the literature, see Lester and Yang, 1997), we form the following long-run relationship between suicide, per capita income, unemployment rate, divorce rate, and fertility variables in linear form as: where the subscript t indexes time period with t = 1957, . . ., 2009; j indexes each suicide with j = 0 (total), 1 (male), and 2 (female); st is suicide rate; yt is per capita real income; ut is the unemployment rate; dt is the divorce rate; ft is the fertility rate; and εt is the classical error term. All variables are in their natural logarithms which allow us interpreting the estimated coefficients as constant elasticities. Recent advances in econometric literature dictate that the longrun relation in Eq. (1) should incorporate the short-run dynamic adjustment process. It is possible to achieve this aim by expressing Eq. (1) in an error-correction model, known as the Engle–Granger’s (1987) approach. 4. Japanese experience st,j = b0 + Although, the epidemiological literature has explored the risk factors of suicide in Japan (e.g. Yamasaki et al., 2008; Motohashi et al., 2004), there are a few studies exploring the determinants of suicide in Japan from an economic perspective (Watanabe et al., 2006; Koo and Cox, 2008; Akechi et al., 2006; Chen et al., 2009b; Yamamura, 2010; Inagaki, 2010). Watanabe et al. (2006) using prefecture level data find that unemployment rate and personal bankruptcy are positively associated with suicide rates. Koo and Cox (2008) using time series data find that the relationship between unemployment and suicide is significantly positive for males and females. Akechi et al. (2006) shows that there is an inverted U shape between alcohol consumption and suicide employing prefecture level data between 1953 and 1986. Chen et al. (2009b) (1) stj = a0 + a1 yt + a2 utj + a3 dt + a4 ft + εt m1  b1i,j st−i,j + i=1 + m4  i=0 b4i dt−i + m2  i=0 m5  b2i yt−i + m3  b3i ut−i,j i=0 b5i ft−i + εt−1 + t (2) i=0 where  represents change, is the speed of adjustment parameter, and εt−1 is the lagged error term, which is estimated from the residuals of Eq. (1). The Engle–Granger method requires that all variables in Eq. (1) are integrated of order one, I(1), and that the lagged error term is integrated order of zero, I(0), in order to establish a cointegration relationship. If some variables in Eq. (1) are non-stationary, we may use a new cointegration method. This procedure is known as ARDL approach to cointegration of Pesaran Author's personal copy A.R. Andrés et al. / The Journal of Socio-Economics 40 (2011) 723–731 et al. (2001) that combines Engle–Granger two steps procedure into one by replacing εt−1 in Eq. (2) with its equivalent from Eq. (1). εt−1 is substituted by linear combination of the lagged variables as in Eq. (3): n1 st,j = c0 +  n2 c1i,j st−i,j +  i=1 + n4  n3 c2i yt−i + i=0 c4i dt−i + i=0 n5   c3i ut−i,j i=0 c5i ft−i + c6 st−1,j + c7 yt−1 i=0 (3) + c8 ut−1,j + c9 dt−1 + c10 ft−1 + vt To obtain Eq. (3), one has to solve Eq. (1) for εt and lag the solution equation by one period. Then, this solution is substituted for εt−1 in Eq. (2) to arrive at Eq. (3). Eq. (3) is a representation of the ARDL approach to cointegration. Pesaran et al. (2001) approach to cointegration has some methodological advantages in comparison to other single cointegration procedures. They are as follows: (i) endogeneity problems and inability to test hypotheses on the estimated coefficients in the long-run associated with the Engle–Granger (1987) method are avoided; (ii) the long and short-run parameters of the model in question are estimated simultaneously; (iii) the ARDL approach to testing for the existence of a long-run relationship between the variables in levels is applicable irrespective of whether the underlying regressors are purely I(0), purely I(1), or a combination of the two; (iv) the small sample properties of the bounds testing approach are far superior to that of multivariate cointegration, as argued in Narayan (2005). The bounds-testing procedure is based on the F- or Waldstatistics, and this is the first stage of the ARDL cointegration method. Accordingly, a joint significance test that implies no cointegration hypothesis, (H0 : c6 = ...... = c10 = 0), against the alternative hypothesis, (H1 : at least one of c6 to c10 = / 0), should be performed for Eq. (3). The F-test used for this procedure has a non-standard distribution. Thus, Pesaran et al. compute two sets of critical values for a given significance level with and without a time trend. One set assumes that all variables are I(0), and the other set assumes that they are all I(1). If the computed F-statistic exceeds the upper critical bounds value, then the H0 is rejected. If the Fstatistic falls into the bounds, then the test becomes inconclusive. Lastly, if the F-statistic is below the lower critical bounds value, it implies no cointegration. Once a long-run relationship has been established, Eq. (3) is estimated using an appropriate lag-selection criterion. At the second stage of the ARDL cointegration procedure, it is also possible to obtain the ARDL representation of the error-correction (EC) model. To estimate the speed with which the dependent variable adjusts to independent variables within the bounds-testing approach, following Pesaran et al. (2001), the lagged-level variables in Eq. (3) are replaced by ECt−1 as in Eq. (4): st,j = ˛0 + k1  ˛1i,j st−i,j + i=1 + k4  i=0 ˛4i dt−i + k2  i=0 k5  ˛2i yt−i + k3  i=0 ˛5i ft−i + ECt−1 + t Table 1 Unit root results. Variables ADF PP ERS st,0 st,1 st,2 yt ut,0 ut,1 ut.2 dt ft st,0 st,1 st,2 yt ut,0 ut,1 ut,2 dt ft 2.66 2.85 2.35 1.50 2.58 2.51 2.63 3.14 1.89 4.07* 3.97* 4.15* 2.79 4.33* 4.54* 4.19* 3.29* 6.64* 2.21 2.43 1.98 1.23 2.94 3.05* 2.62 2.59 2.28 6.24* 5.96* 6.67* 3.08* 5.16* 5.65* 5.12* 3.98* 12.3* 1.46 1.43 1.78 0.31 1.53 1.54 1.52 2.56* 1.67 4.11* 3.71* 3.81* 2.52* 3.73* 3.66* 4.20* 2.28 5.44* Notes: Sample levels are 1958–2009 and differences are 1959–2009. The critical values for ADF and PP with a constant and without a trend at the 5% level of significance are 2.91. The critical value for ERS with a constant and without a trend at the 5% level of significance is 2.29. All test statistics and critical values are expressed in absolute terms for convenience. Rejection of unit root hypothesis is indicated with an asterisk.  stands for first difference. 6. Results Annual data over the period 1957–2009 were used to estimate Eq. (3) by the ARDL cointegration procedure of Pesaran et al. (2001). Variable definitions and sources of data are provided in Appendix. To implement the Pesaran et al. (2001) cointegration procedure, one has to ensure that none of the explanatory variables in Eq. (1) is above I(1). In the presence of I(2) or higher variables, the computed statistics provided by Pesaran et al. (2001) are not valid. Consequently, the implementation of unit root tests in the ARDL approach is necessary to ensure that none of the variables included in the model is integrated of order 2 or beyond. Three tests were used to test unit roots in the variables: Augmented Dickey–Fuller (henceforth, ADF) (1979, 1981), Phillips–Perron (henceforth, PP) (1988), and Elliott–Rothenberg–Stock (henceforth, ERS) (1996). Unit root tests results are displayed in Table 1. The conditions for applying the ARDL bounds testing approach are satisfied. In other words, all variables included in the model are either I(0) or I(1). Table 2 The results of F and W tests for cointegration. 95% LB (4) i=0 A negative and statistically significant estimation of  not only represents the speed of adjustment but also provides an alternative means of supporting cointegration between the variables. 95% UB 90% LB 90% UB  Panel A: The assumed long-run relationship: F/W (s0 y, u0 , d, f ) F-statistic 5.30 W-statistic 26.51 3.10 4.35 15.52 21.75 2.60 3.74 13.01 18.70  Panel B: The assumed long-run relationship: F/W (s1 y, u1 , d, f ) F-statistic 6.54 W-statistic 32.72 ˛3i ut−i,j 727 3.10 4.35 2.60 3.74 15.52 21.75 13.01 18.70  Panel C: The assumed long-run relationship: F/W (s2 y, u2 , d, f ) F-statistic 3.46 W-statistic 17.31 3.10 4.35 2.60 3.74 15.52 21.75 13.01 18.70 If the test statistic lies between the bounds, the test is inconclusive. If it is above the upper bound (UB), the null hypothesis of no level effect is rejected. If it is the below the lower bound (LB), the null hypothesis of no level effect cannot be rejected. Author's personal copy A.R. Andrés et al. / The Journal of Socio-Economics 40 (2011) 723–731 728 Table 3 ARDL cointegration results. Regressor Coefficient Standard error T-ratio Panel A: Estimated long-run coefficients using the ARDL approach for aggregate suicide model: ARDL (1,1,1,0,1) selected based on the Akaike Information Criterion, 1957–2009 Dependent variable st,0 yt −0.4106* 0.1397 2.9385 0.2024 0.1922 1.0533 ut,0 0.8832* 0.3990 2.2134 dt −0.7003** 0.3352 1.7134 ft 1.4435 0.8130 Constant 0.3437 Panel B: Error correction representation results Dependent variable st,0 yt −0.9085* 0.0586 ut,0 dt 0.3709* ft −0.0575 ECt−1 −0.4199* Diagnostic tests R̄2 0.41 RSS 0.12 0.3787 0.0095 0.1301 0.1222 0.1057 2.3988 0.6116 2.8501 0.4712 3.9714 F-statistic DW-statistic 8.89* 1.94 2SC (1) 2N (2) 0.06 19.26 2FF (1) 2H (1) 0.73 2.49 RSS stands for residual sum of squares. T-ratios are in absolute values. 2SC , 2FF , 2N , and 2H are Lagrange multiplier statistics for tests of residual correlation, functional form mis-specification, non-normal errors and heteroskedasticity, respectively. These statistics are distributed as chi-squared variates with degrees of freedom in parentheses. The critical values for 2 (1) = 3.84 and 2 (2) = 5.99 are at 5% significance level. ***Significance at 10% level. * Significance at 1% level. ** Significance at 5% level. Visual inspections of the variables in logarithm show no structural breaks. Eq. (3) is estimated in two stages. In the first stage of the ARDL procedure, the long-run relationship of Eq. (1) was established in two steps. First, the selection of the lag length on the firstdifferenced variables for Eq. (3) was obtained from unrestricted Vector Autoregression (VAR) by means of Akaike Information criteria (AIC) and the Schwarz Bayesian Criterion (SBC). The results suggest the optimal lag length as 2, but this stage of the results is not presented here to conserve space. Second, a bound F-test was applied to Eq. (3) in order to determine whether the dependent and independent variables are cointegrated in each model. The results of the bounds F-testing are reported in Table 2. From Table 2, it can bee seen that the computed F statistics are above the upper bound values in the cases of total and male suicides models’ thus, implying cointegration relations. The ARDL cointegration procedure was implemented to estimate the parameters of Eq. (3) with maximum lag-order set to 2, which is selected on the basis of AIC, SBC and R̄2 selection criteria. This stage involves estimating the long-run and short-run coefficients of Eqs. (1) and (2). The summary ARDL results with some diagnostic tests for total suicides, male suicides, and female suicides are presented in Tables 3–5, respectively. The overall empirical results appear to be rather satisfactory. First, income enters negatively in the regressions for overall, male, and female suicides. The long-run elasticity of suicide with respect to income is highest in the case of male suicides. This is −0.54, suggesting that one per cent increase in per capita income will decrease the number of male suicides by 0.54% whilst other factors remain constant. The long-run income elasticities with respect to total and female suicides are −0.41 and −0.36, respectively. This finding implies that males are more vulnerable Table 4 ARDL cointegration results. Regressor Coefficient Standard error T-ratio Panel A. Estimated long-run coefficients using the ARDL approach for male suicide model: ARDL (1,1,0,0,0) selected based on the Schwarz Bayesian Criterion, 1957–2009 Dependent variable st,1 0.1297 4.1785 yt −0.5420* ut,1 0.0133 0.1755 0.7581 1.1635* 0.4030 2.8870 dt 0.0456 0.1716 0.2661 ft 2.3613** 1.2274 1.9239 Constant Panel B. Error correction representation results Dependent variable st,1 yt −1.0050* 0.0560 ut,1 dt 0.4896* 0.0192 ft −0.4208* ECt−1 Diagnostic tests R̄2 0.47 RSS 0.15 0.3704 0.0812 0.1382 0.0712 0.0944 F-statistic DW-statistic 2.7128 0.6890 3.5428 0.2697 4.4551 10.6* 2.11 2SC (1) 2N (2) 0.34 18.10 2FF (1) 2H (1) 1.31 2.41 RSS stands for residual sum of squares. T-ratios are in absolute values. 2SC , 2FF , 2N , and 2H are Lagrange multiplier statistics for tests of residual correlation, functional form mis-specification, non-normal errors and heteroskedasticity, respectively. These statistics are distributed as chi-squared variates with degrees of freedom in parentheses. The critical values for 2 (1) = 3.84 and 2 (2) = 5.99 are at 5% significance level. ***Significance at 10% level. * Significance at 1% level. ** Significance at 5% level. Author's personal copy A.R. Andrés et al. / The Journal of Socio-Economics 40 (2011) 723–731 729 Table 5 ARDL cointegration results. Regressor Coefficient Standard error T-ratio Panel A. Estimated long-run coefficients using the ARDL approach for female suicide model: ARDL (1,0,0,0,1) selected based on the R-Bar Squared Criterion, 1957–2009 Dependent variable st,2 yt −0.3677* 0.2491 18.0085 ut,2 0.3431 1.7628 0.0197** 0.8844 0.6168 0.2570 dt 0.4241 2.0300 ft 0.4974* −0.6752 2.8321 0.2384 Constant Panel B. Error correction representation results Dependent variable st,2 yt −0.0869*** 0.0046 ut,2 0.2091** dt ft −0.1197 −0.2364* ECt−1 Diagnostic tests R̄2 0.21 RSS 0.13 0.0565 0.0818 0.1127 0.1251 0.0899 1.5367 0.0570 1.8552 0.9564 2.6297 F-statistic DW-statistic 3.98* 1.84 2SC (1) 2N (2) 0.48 22.41 2FF (1) 2H (1) 1.10 0.16 RSS stands for residual sum of squares. T-ratios are in absolute values. 2SC , 2FF , 2N , and 2H are Lagrange multiplier statistics for tests of residual correlation, functional form mis-specification, non-normal errors and heteroskedasticity, respectively. These statistics are distributed as chi-squared variates with degrees of freedom in parentheses. The critical values for 2 (1) = 3.84 and 2 (2) = 5.99 are at 5% significance level. * Significance at 1% level. ** Significance at 5% level. *** Significance at 10% level. to income loss than females. However, the magnitude of this effect is rather minimal. Second, unemployment rates are positively and significantly associated with female suicides. The long-run partial elasticity of suicides with respect to unemployment rates is 0.01, indicating that a 1% rise in unemployment rates will trigger an increase in female suicides by about 0.01%. Although there seems to be the almost same impact exists in the case of the male suicides but that is not statistically significant. Hence one argues broadly that the impact of male and female unemployment rates on suicides is identical. Gender seems to have no special effect on a suicide decision, when an individual becomes unemployed. Third, divorce rates are positively correlated with suicides but are statistically insignificant in the case of female suicides. Male population appears to be suffering more as a result divorce since the long-run elasticity of divorce rate with respect to male suicides is 1.16, suggesting that a 1% increase in divorce rates will rise the male suicides by 1.16% which is the stronger determinant of suicide in the entire analysis. Finally, we find a statistically significant negative association between fertility rates and suicides only in the case of the total suicides. Thus, a 1% rise in the total fertility rates will drop the total number of suicides by 0.70 whilst the other explanatory factors are constant. The long-run elasticities of suicides in respect to fertility for male and female suicides appears to be in wrong signs. In regards to the relative magnitude of the explanatory variables in this study, the fertility rate seems to be the second most important factor in explaining suicides, followed by real per capita income and unemployment rates. Tables 3–5 also report the coefficients of coefficients of ECt−1 the error correction model. All coefficients of ECt−1 are statistically significant and have the negative expected sign in all models. This situation provides further confirmation for cointegration relationships between variables of total and male suicides models as well as suggesting an alternative means of long-run relationship in the case of female suicide model. The magnitude of the speed of equilibrium is relatively low, since their values are less than 0.5. The lowest error correction coefficient appeared in the female regression model, which means that about 25% of disequilibrium is corrected every year. As the suicide is a longterm phenomenon, the short-run elasticities will have no real impact in policy designing therefore we are not evaluating them further. 7. Summary and conclusions This paper, from a socioeconomic point of view, investigates the determinants of suicides in Japan for the time span between 1957 and 2009. Unlike earlier studies, this paper employs a relatively recent econometric procedure, the ARDL approach to cointegration, which has been utilized to obtain the long-run elasticities of the suicides with respect to the total, male and female suicides. To our knowledge, this paper is the first paper to apply an ARDL approach to examine the determinants of suicide in Japan. This approach seems to have several potential advantages as it needs no a large number of observations to guarantee the robustness of the estimators and performance of the statistical tests. Furthermore, the choice of the suicide static model could influence the analysis. Individuals might respond with some delayed to changes in socioeconomic factors. In this case, suicides are explained by current and lagged differenced values of real per capita income, unemployment, and divorce rates. We show that in the long run, the divorce is the highest suicide cause and the Japanese men seem to be suffering particularly from this situation. The second most important determinant of the suicides in Japan is also a sociological factor, fertility rates. As expected, the female population are more affected with the decreasing level of fertility rates. Combining these two suicide causes, one may argue that sociological factors are more dominant than economic factors in the case of Japanese suicides, which is inconsistent with Chen et al. (2009b). This might be partly because that Chen et al. (2009b) uses the panel data of OECD countries to make a comparison between Japan and other OECD countries. For robustness check of this paper by comparing Japan and other OECD countries, it is required to use time series data of other OECD countries to conduct ARDL estimation in the future studies. Furthermore, the economic determinants of suicides in Japan appear to be moderate in magnitude and similar in both sexes indicating that male and female participation to work and sharing the burden of economic difficulties are almost the same. Our results support the existence of a long run relationship between socio-economic factors and suicides, regardless of gender. Finally, recommendations for suicide prevention are generally a combination of strategies targeting high-risk groups and strategies targeting a whole population. The findings of this study reveal that Author's personal copy 730 A.R. Andrés et al. / The Journal of Socio-Economics 40 (2011) 723–731 government policies should promote family cohesion and provide economic incentives to raise birth rates, as these policies will be effective in lowering suicide rates. Acknowledgements The authors are grateful to an anonymous referee for his/her useful comments and suggestions on an earlier version of this work. Appendix A. Appendix Data Definitions and Sources All data were collected online with the provided internet links below: st,j are crude suicide rates for total, male and females per 100,000 in logarithm. Source: Period 1955–2004: Statistics Bureau, Ministry of Internal Affairs and Communications (2006). Historical Statistics of Japan Volume 1 (New Edition). Tokyo: Japan Statistical Association. Period 2005-2009: National Police Agency. http://www8.cao.go.jp/jisatsutaisaku/link/keisatsutyo.html (accessed 16.06.10). yt is per real capita income in logarithm. Base year is 1990. Source: Period 1955–2003: Statistics Bureau, Ministry of Internal Affairs and Communications (2006). Historical Statistics of Japan Volume 1 (New Edition). Tokyo: Japan Statistical Association. Period 2004–2009: Cabinet office of Government of Japan. http://www.esri.cao.go.jp/jp/sna/qe101-2/gdemenu ja.html (accessed 16.06.10). ut,j are unemployment rates for total, male and females in logarithm. Source: Period 1955–2009: Statistics Bureau, Ministry of Internal Affairs and Communications. http://www.stat.go.jp/ data/roudou/longtime/03roudou.htm#hyo 1 (accessed 16.06.10). dt is divorce rate per 1000 in logarithm. Source: Period 1955–2003: Statistics Bureau, Ministry of Internal Affairs and Communications (2006). Historical Statistics of Japan Volume 1 (New Edition). Tokyo: Japan Statistical Association. Period 2004–2009: Ministry of Health, Labour, Welfare. http://www.mhlw.go.jp/toukei/saikin/hw/jinkou/suikei09/ index.html (accessed 16.06.10). ft is fertility rate per 1000 in logarithm. Source: Period 1955–2003: Statistics Bureau, Ministry of Internal Affairs and Communications (2006). Historical Statistics of Japan Volume 1 (New Edition). Tokyo: Japan Statistical Association. Period 2004–2009: Ministry of Health, Labour, Welfare. http://www.mhlw.go.jp/toukei/saikin/hw/jinkou/suikei09/ index.html (accessed 16.06.10). References Akechi, T., Iwasaki, M., Uchitomi, Y., Tsugane, S., 2006. Alcohol consumption and suicide among middle aged men in Japan. British Journal of Psychiatry 188, 231–236. Altinanahtar, A., Halicioglu, F., 2009. A dynamic econometric model of suicide in Turkey. Journal of Socio-Economics 38, 903–907. Andrés, A.R., 2005. Income inequality, unemployment, and suicide: a panel data analysis of 15 European countries. Applied Economics 3, 439–451. Andrés, A.R., Halicioglu, F., 2010. Determinants of suicides in Denmark: evidence from time series data. Health Policy 98, 263–269. Becker, G., 1974. A theory of marriage. In: Schultz, T.W. (Ed.), Economics of the Family. University of Chicago Press, Chicago, pp. 299–344. Bertrand, M., Mullainathan, S., 2001. Do people mean what they say? Implications for subjective survey data. American Economic Review 91, 67–72. Brainerd, E., 2001. Economic reform and mortality in the former Soviet Union: a study of the suicide epidemic in the 1990. European Economic Review 45, 1007–1019. Brinton, M.C., 1993. Women and the Economic Miracle: Gender and Work in Postwar Japan. University of California Press, Berkley. Cargill, T.F., 2006. Japan’s economic and financial stagnation in the 1990 and reluctance to change. In: Blomsrom, M., La Croix, S. (Eds.), Institutional Change in Japan. Routledge, London. Chang, S.S., Sterne, J.A.C., Huang, W.C., Chuang, H.L., Gunnell, D., 2010. Association of secular trends in unemployment with suicide in Taiwan, 1959–2007: a time series analysis. Public Health 124, 49–54. Chen, J., Choi, Y.J., Mori, K., Sawada, Y., Sugano, S., 2009a. Those who are left behind: an estimation of the number of family members of suicide victims in Japan. Social Indicators Research 94, 535–544. Chen, J., Choi, Y.J., Sawada, Y., 2009b. How is suicide different in Japan? Japan and the World Economy 21, 140–150. Chuang, H.L., Huang, W.C., 1996. A reexamination of sociological and economic theories of suicide: a comparison of the USA and Taiwan. Social Science and Medicine 43, 412–423. Chuang, H., Huang, W., 1997. Regional suicide rates: a pooled cross-section and time-series analysis. Journal of Socio-Economics 26, 277–289. Chuang, H., Huang, W., 2007. A re-examination of the suicide rates in Taiwan. Social Indicators Research 83, 465–481. Cuellar, A.E., Markowitz, S., 2006. Medicaid Policy Changes in Mental Health Care and Their Effect on Mental Health Outcomes. NBER Working Papers No. 12232. Daly, M., Wilson, D.J., 2009. Happiness, unhappiness and suicide: an empirical assessment. Journal of the European Economic Association 7, 539–549. Dickey, D.A., Fuller, W.A., 1981. Likelihood ratio statistics for autoregressive time series with a unit root. Econometrica 49, 1057–1072. Dickey, D.A., Fuller, W.A., 1979. Distributions of the estimators for autoregressive time series with a unit root. Journal of the American Statistical Association 74, 427–431. Durkheim, E., 1951. Suicide: A Study in Sociology. Free Press, New York. Elliott, G., Rothenberg, T., Stock, J., 1996. Efficient tests for an autoregressive unit root. Econometrica 64, 813–836. Engle, R.F., Granger, C.W., 1987. Cointegration and error correction: representation, estimation and testing. Econometrica 55, 251–276. Hamermesh, D., 1974. The economics of black suicide. Southern Economic Journal 41, 188–199. Hamermesh, D.S., Soss, N.M., 1974. An economic theory of suicide. Journal of Political Economy 82, 83–98. Inagaki, K., 2010. Income inequality and the suicide rate in Japan: evidence from cointegration and LA-VAR. Journal of Applied Economics 13, 113–133. Johansen, S., Juselius, K., 1990. Maximum likelihood estimation and inference on cointegration with application to the demand for money. Oxford Bulletin of Economics and Statistics 52, 169–210. Jungeilges, J., Kirchgässner, G., 2002. Economic welfare, civil liberty, and suicide: an empirical investigation. Journal of Socio-Economics 31, 215–231. Kennelly, B., Ennis, J., O’Shea, E., 2005. Economic cost of suicide and deliberate self harm. In: Reach out National Strategy for Action on Suicide Prevention 2005–2014. Department of Health and Children, Ireland. Khang, Y.H., Lynch, J.W., Kaplan, G.A., 2005. Impact of economic crisis on causespecific mortality in South Korea. International Journal of Epidemiology 34, 1291–1301. Koivumaa-Honkanen, H., Honkanen, R., Viinamki, H., Heikkil, J., Kaprio, J., Koskenvuo, M., 2001. Life satisfaction and suicide: a 20-year follow-up study. American Journal of Psychiatry 158, 433–459. Koo, J., Cox, W.M., 2008. An economic interpretation of suicide cycles in Japan. Contemporary Economic Policy 26, 162–174. Kunce, M., Anderson, A.L., 2002. The impact of socioeconomic factors on state suicide rates: a methodological note. Urban Studies 39, 155–162. Kuroki, M., 2010. Suicide and unemployment in Japan: evidence from municipal level suicide rates and age specific suicide rates. Journal of Socio-Economics 39, 683–691. Lester, D., 1995. Explaining regional differences in suicide rates. Social Science and Medicine 40, 719–721. Lester, D., 1996. Patterns of Suicide and Homicide in the World. Nova Science Publishers, New York. Lester, D., Yang, B., 1997. The Economy and Suicide: Economic Perspectives. Nova Science Publishers, New York. Lin, S.J., 2006. Unemployment and suicide: panel data analyses. Social Science Journal 43, 727–732. McDaid, D., Halliday, E., McKenzie, M., MacLean, J., Maxwell, M., McCollam, A., Platt, S., Woodhouse, A., 2007. Issues in the Economic Evaluation of Suicide Prevention Strategies: Practical and Methodological Challenges. Personal Social Services Research Unit, LSE, London. Mann, J.J., Apter, A., Bertolote, J., Beautrais, A., Currier, D., Haas, H., Hegerl, U., Lonvquist, J., Malone, K., Marusic, A., Mehlum, L., Patto, G., Phillips, M., Rutz, W., Rihmer, Z., Schmidtke, A., Shaffer, D., Silverman, M., Takahashi, Y., Varnik, A., Wasserman, D., Yip, P., Hendin, H., 2005. Suicide prevention strategies: a systematic review. Journal of American Medical Association 294, 2064–2074. Minoiu, C., Rodríguez, A., 2008. The effect of public spending on suicide: evidence from US state data. Journal of Socio-Economics 37, 237–261. Motohashi, Y., Kaneko, Y., Sasaki, H., 2004. Community based suicide prevention programme in Japan using a health promotion approach. Health and Preventive Medicine 9, 3–8. Nakao, M., Takeuchi, T., 2006. The suicide epidemic in Japan and strategies of depression screening for its prevention. Bulletin of the World Health Organization 84, 492–493. Narayan, P.K., 2005. The saving and investment nexus for China: evidence from cointegration tests. Applied Economics 37, 1979–1990. Author's personal copy A.R. Andrés et al. / The Journal of Socio-Economics 40 (2011) 723–731 Neumayer, E., 2003. Socioeconomic factors and suicide rates at large-unit aggregate levels: a comment. Urban Studies 40, 2769–2776. Ono, H., 2006. Divorce in Japan: why it happens, why it doesn’t. In: Blomsrom, M., La Croix, S. (Eds.), Institutional Change in Japan. Routledge, London. Pesaran, M.H., Shin, Y., Smith, R., 2001. Bounds testing approach to the analysis of level relationships. Journal of Applied Econometrics 16, 289–326. Phillips, P.C.B., Perron, P., 1988. Testing for a unit root in time series regression. Biometrika 75, 335–346. Platt, S., Hawton, K., 2000. In: Hawton, K., van Heeringen, K. (Eds.), Suicidal Behaviour and the Labour Market, in the International Handbook of Suicide and Attempted Suicide. John Wiley & Sons, Ltd., Chichester, UK, pp. 309–384. Qin, P., Agerbo, E., Mortensen, P.B., 2003. Suicide risk in relation to socioeconomic, demographic, psychiatric, and familial factors: a national register based study of all suicides in Denmark. American Journal of Psychiatry 160, 765–772. Ruhm, C., 2000. Are recessions good for your health? Quarterly Journal of Economics 115, 617–650. Stack, S., 2000. Suicide: a 15-year review of the sociological literature. Part I: cultural and economic factors. Suicide and Life-Threatening Behavior 30, 145–162. Suzuki, T., 2008. Economic modeling of suicide under income uncertainty: for better understanding of middle-aged suicide. Australian Economic Papers 47, 296–310. Unnithan, N.P., Huff-Corzine, L., Corzine, J., Whitt, H.P., 1994. The Currents of Lethal Violence: An Integrated Model of Suicide and Homicide. State University of New York Press, Albany. 731 Viren, M., 1999. Testing the natural rate of suicide hypothesis. International Journal of Social Economics 26, 1428–1440. Watanabe, R., Furukawa, M., Nakamura, R., Okura, Y., 2006. Analysis of the Socioeconomic Difficulties Affecting the Suicide Rate in Japan. KIER Discussion Paper Series. No. 626. WHO, 2006. Mortality Country Fact Sheet 2006. Yamamura, E., 2010. The different impacts of socio-economic factors on suicide between males and females. Applied Economics Letters 17, 1009–1012. Yamasaki, A., Araki, S., Sakai, R., Yokoyama, K., Voorhees, A.S., 2008. Suicide mortality of young, middle aged and elderly males and females in Japan for the years, 1953–1996: time series analysis for the effects of unemployment, female labour force, young and aged population, primary industry and population density. Industrial Health 46, 541–549. Yamasaki, A., Sakai, R., Shirakawa, T., 2005. Low income, unemployment, and suicide mortality rates for middle-age persons in Japan. Psychological Reports 96, 337–348. Yang, B., 1992. Economy and suicide: a time series study of suicide in USA. American Journal of Economics and Sociology 51, 187–199. Yang, B., Lester, D., 1990. Time series analyses of the American suicide rate. Social Psychiatry and Psychiatric Epidemiology 25, 274–275. Yang, B., Lester, D., Yang, C.H., 1992. Sociological and economic theories of suicide: a comparison of the USA and Taiwan. Social Science and Medicine 34, 333–334.