نابرابری درآمد و نرخ خودکشی در ژاپن : مدارک و شواهد از هم انباشتگی و LA-VAR

Using time series techniques, this paper examines the relationship between the suicide rate and income inequality in Japan. Since both the suicide rate and income inequality (Gini coefficient) in Japan are integrated of order one for the sample period 1951–2007, the existence of cointegration is a prerequisite for the successful modeling of their relationship. The Durbin-Hausman test shows that the suicide rate is cointegrated with income inequality and the unemployment rate. The dynamic ordinary least squares (DOLS) and fully modified ordinary least squares (FMOLS) methods demonstrate that income inequality and the unemployment rate are positively and significantly related to the suicide rate in Japan, and there is evidence supporting the parameter stability of our suicide model. Furthermore, the lag-augmented vector autoregression (LA-VAR) approach shows that there exists unidirectional Granger-causality from income inequality to the suicide rate. Hence, the fluctuations in Japan's suicide rate are partially explained by income inequality.

Since the early 1990s, the suicide rate in Japan has increased remarkably, escalating from 16.1 (per 100,000 people) in 1991 to 25.4 in 1998—an approximate 60% rise within the short span of 8 years. Furthermore, in 2000, the suicide rate in Japan was the highest among the developed countries, excluding Russia and the central and eastern European countries. The World Health Organization (WHO) reported that the overall suicide rate in Japan was 24.1 in 2000; this figure was higher than thos of most other developed countries such as Canada (11.7), France (18.4), Germany (13.5), Italy (7.1), the U.K. (7.2), and the U.S. (10.4). These facts have generated considerable concern among Japanese policy makers (e.g., Cabinet Office 2007). However, the determinants of suicide in Japan are still a matter of debate. To illustrate the nature of the suicide problem in Japan, it is important to provide theoretical explanations of its determinants. In this regard, Daly, Wilson, and Johnson (2007) and Daly and Wilson (2009) point out the importance of interpersonal income comparisons. They argue that suicide data can be used as a direct measure to assess utility, and their theoretical model predicts that an individual commits suicide when his/her current and expected future utility is less than the perceived utility of committing suicide. If individuals compare themselves to others to determine their own utility, a deterioration of their relative position may lead to a decrease in their perceived utility and, consequently, to an increase in their suicide risk. In this respect, interpersonal income comparisons seem to be a useful measure for examining the determinants of suicide. Specifically, if an individual’s utility does indeed depend on relative income as Daly, Wilson, and Johnson (2007) and Daly and Wilson (2009) demonstrate empirically, then an increase in the income of an individual has a negative impact on others since it deteriorates their relative positions. In other words, an increase in income inequality may lead to an increase in suicide risk for the less well-off individuals in the community. 1 Hence, from this perspective, it seems reasonable to suppose that income inequality is one of the determinants of suicide. This paper addresses the question of whether income inequality in Japan affects the suicide rate. In the literature on the relationship between income inequality and suicide, little attention has been paid to the case of Japan. The prevention of suicide is an important issue in Japan, since it has one of the highest suicide rates in the world. In this light, an investigation of the relationship between income inequality and suicide in Japan seems significant; moreover, such an investigation could provide useful information to those responsible for formulating the Japanese policy on mental health care.This paper examines the relationship between the suicide rate and income inequality in Japan by using time series techniques. As Figure 1 indicates, the suicid rate in Japan has been on the rise since the 1970s. In the late 1990s, the suicide rate rose to its highest level since the 1970s, and it has remained at that level after 2000. The time series fluctuations in the suicide rate clearly describe the recent suicide conditions in Japan, and therefore, an investigation of the sources of these time series fluctuations may be beneficial. For example, the Gini coefficient and unemployment rate (a measure of economic conditions) appear to exhibit an upward trend similar to that of the suicide rate (Figure 1).2 In particular, all three variables increased during the same period in the mid-1990s. These facts suggest the possibility that both income distribution and labor market conditions affect the time series fluctuations in the suicide rate. Therefore, the use of time series techniques could facilitate our understanding of the determinants of suicide in Japan. 3 We examine the cointegration of the relationship between the suicide rate and income inequality in Japan. The cointegration property is important because it can provide greater insights into this relationship. For example, let us assume that the suicide rate and its determinants are integrated of order one. The existence of cointegration between them implies that they have common trends; in this case, we can estimate a regression equation for the suicide rate, which would be indicative of a long-run equilibrium relationship. However, if the suicide rate is not found to be cointegrated with its determinants, then the relationship can be regarded as a spurious regression, which gives rise to serious problems in terms of statistical inference (e.g., Granger and Newbold 1974). Therefore, to ensure that the regression results we obtain are not spurious, we need to test for the existence of cointegration. In other words, we can reasonably state that the existence of cointegration is a prerequisite for the successful modeling of the relationship between the suicide rate and income inequality in Japan, since both are likely to be integrated of order one. This paper contributes to the literature on the relationship between suicide and income inequality in three respects. First, it uses the Durbin-Hausman test (Choi 1994), which is a residual-based test for cointegration, to examine the relationship. This test is more powerful in finite samples than the other residual-based tests for cointegration, such as the augmented Dickey-Fuller and Phillips-Perron tests (e.g.Choi 1994). Hence, the Durbin-Hausman test can provide more accurate results for cointegration in cases with small samples as in the case of this study. Second, we pay careful attention to the cointegration property of the suicide model that we estimate. More specifically, we use the fully modified ordinary least squares (FMOLS; Phillips and Hansen 1990) and dynamic ordinary least squares (DOLS; Stock and Watson 1993) methods. These methods correct for endogeneity and serial correlation in cointegrating regressions, thereby providing unbiased estimates of the cointegrating coefficients. In addition, we examine the parameter stability of the suicide model by using the SupF and MeanF tests (Hansen 1992), which can be applied to the FMOLS estimator. Third, we apply the lag-augmented vector autoregression (LA-VAR) approach (Toda and Yamamoto 1995) to test for Granger-causality between the suicide rate and income inequality in Japan. The LA-VAR approach allows us to test for Grangercausality among time series variables in levels without taking their integration and cointegration properties into consideration. In this respect, it is more advantageous to use the LA-VAR approach rather than the standard VAR and vector error correction approaches when examining Granger-causality between the suicide rate and income inequality. This paper is organized as follows. The next section briefly reviews the literature on the relationship between income inequality and suicide. Section III explains the methodologies used in the paper. Section IV describes the data. Section V reports the empirical results. Section VI presents the conclusion.

In this paper, we examine the relationship between the suicide rate and income inequality in Japan. We distinguish our work from the existing literature by using time series techniques such as the Durbin-Hausman test for cointegration, the DOLS and FMOLS methods, and a Granger causality test based on the LA-VAR approach. Since our focus is on Japan, an empirical analysis based on Japanese time series data serves our purpose. Our main findings can be summarized as follows: (i) the suicide rate, income inequality (Gini coefficient), and the unemployment rate are integrated of order one; (ii) the suicide rate is cointegrated with income inequality and the unemployment rate, and the parameter stability of the basic suicide model is supported; (iii) income inequality and the unemployment rate are positively and significantly related to the suicide rate; (iv) income inequality and the unemployment rate Granger-cause the suicide rate, but not vice versa; (v) the results showing the impact of income inequality are robust when we test a broader model (there, female labor force participation is negatively related to suicide rate); and (vi) the positive impact of income inequality is robust when we examine that effect separately by gender. As Förster and Mira d’Ercole (2005) indicate, income inequality in Japan has increased consistently since the 1980s. In 2007, the share of non-regular workers reached approximately a third of employees, and this labor market dualism has created a large segment of workers with lower wages, short-term job experience, and limited opportunities to enhance their human capital. These facts have raised serious concerns about equity (e.g., OECD 2008). Hence, it is likely that Japanese people tend to make interpersonal income comparisons. At the same time, the suicide rate in Japan appears to exhibit an upward trend similar to that of the Gini coefficient. Indeed, there is substantial evidence to support the argument that a significant relationship exists between the suicide rate and income inequality in Japan. On these grounds, it seems reasonable to conclude that the time series fluctuations in the suicide rate in Japan over the past 50 years can be partially explained by changes in income inequality. It is important to consider the reason why our results are different from those reported in several previous studies. For example, Andrés (2005), Leigh, and Jencks (2007), and Minoiu and Andrés (2008) report that income inequality does not affect suicide. A possible explanation for this inconsistency is that the determinants of suicide are different in Japan. Chen, Choi, and Sawada (2009) demonstrate this point by using a dummy variable technique. It is also important to note that the panel data analyses of Andrés (2005), Leigh and Jencks (2007), and Minoiu and Andrés (2008) do not include Japanese data, and therefore, the inconsistency is likely to occur. In other words, the suicide and income inequality data for Japan may be outliers compared to the data for the other countries. In this respect, our results are complementary to those reported in the previous studies. Finally, as to the implications of the paper’s results for Japanese mental health care policy, we can reasonably state that it is important to consider both income distribution and labor market conditions in order to reduce the overall suicide rate in Japan. For example, the mental health care policies for males and females in the lowest income group are likely to play an important role in the prevention of suicides in Japan. Likewise, careful attention should be paid to people who lose their jobs, particularly during periods of economic recession, since overall suicide in Japan is countercyclical.