Determinants of youth suicide: the Easterlin-Holinger cohort hypothesis re-examined

American Journal of Economics and Sociology, The,  April, 1998  by Donald G. Freeman

• E-mail
• Print
• Link

Hamermesh and Soss (1974) proposed an expected lifetime utility model of suicide. If an individual's expected discounted lifetime utility, [Z.sub.i], falls below the individual's taste for living, [b.sub.i], the individual commits suicide. [Z.sub.i] is a function of the individual's age and expected permanent income, in their empirical tests, the authors use income growth (based on historical trends) and unemployment as proxies for permanent income. The authors find, as expected, that higher unemployment rates are associated with higher suicide rates--irrespective of age--but that income growth has mixed effects, conditioned on age. For older subgroups, suicide is negatively related to income growth but, paradoxically, for younger groups the relationship is reversed. Hamermesh and Soss admit no simple explanation for this finding but suggest that postponement of consumption for education may have a bearing. Adolescents (ages 15-19 and below) were not among the groups studied by Hamermesh and Soss.

Ahlburg and Schapiro (1984) combine relative cohort size with economic cycles, divorce rates, labor force participation rates for women, fertility rates and real income measures to predict suicide rates. They conclude that relative cohort size is a positive predictor of youth suicide rates, but their regression results for youths (ages 15-24) appear to be influenced by strong serial correlation in the residuals, and the coefficient of the cohort variable is statistically insignificant for young males, who comprise the majority of youth suicides. The indirect effects of cohort size (i.e., through unemployment, divorce rates, etc.) for younger males are large and highly significant, however, leading the authors to conclude that "as relative cohort size continues to fall throughout the remainder of this century, we expect that [female and younger male] suicide rates will continue to fall, perhaps to historic lows" (p. 103).

Yang (1992) found mixed effects for per capita income growth, with coefficient sign, depending on race and gender. The unemployment rate was an unambiguously positive predictor of suicides, but the significance of the coefficient was sensitive to race and gender.

III

The Model and the Variables

In this section, we develop a model that combines the factors from previous work with new elements that may be of significance to adolescent behaviors: single-parent families and poverty rates for children.

Kosky (1983), in a study of 70 psychiatric adolescents, found that 60 percent of suicidal children lived with one parent but 82 percent of nonsuicidal children lived with two parents. Kosky also found that suicidal children tended to be from lower income families.

Households headed by single parents, although growing-as a share of total households, are still a distinct minority: Twenty-two percent of U.S. households were headed by single parents in 1994, 17.5 percent by women and 4.5 percent by men. Single-parent families, on average, have significant economic disadvantages. A disproportionate number live in poverty: fifty-seven percent of families with children living below the poverty line are headed by single parents. In 1994, households with children headed by single parents had median income just one-half that of two-parent households (Bureau of the Census, 1996).