Suicide and business cycles: new empirical evidence

Abstract

This paper provides an empirical test for the hypothesis that suicides are related to economic determinants. More precisely a hypothesis is tested that changes in the suicide rate are determined by changes in the expected growth rate of income. This hypothesis is tested with an error-correction model which also takes into account various demographic and socioeconomic variables. Empirical results with Finnish data covering the period 1878–1999 provide strong support for this hypothesis.

Notes

¹ The level shift dummy could be rationalized by prohibition (in 1920–1933) and by various restrictions on civil liberties, which e.g. showed up in the outlawing of the communist party in 1924 and 1929. When the model is estimated for 1925–1940 only, or alternatively for 1878–1924 and 1941–1999 (see Table 3) the results (except for the constant term) were very similar to the full sample period, which led to the belief that the period 1925–1940 merely represents some sort of level shift outlier in the data.

² Although there seemed to be no compelling reason, the long-run relationship is also estimated using the Fully Modified Phillips–Hansen Estimator. The results were pretty well in line with the OLS results reported in Table 1. Thus, the following coefficient estimates were obtained: constant: 0.193, D1: 6.213, Age: 0.999, Age ²: −0.018 and Lf0 −0.383.

³ The linearity property was tested using a simple threshold model specification in terms of the Hodrick–Prescott output gap (which was threshold variable). Estimation results did not, however, lend much support to this kind of non-linear behaviour.