Analyses Age-Periode-Cohorte : comparaison et explication des ruptures de cohortes du suicide

This paper assesses the stability of cohort effects in suicide rates age-period-cohort comparative analyses. Despite improvements in the APC methodology (Yang and al), twoproblems remain. The first one is the problem of linear versus fluctuant components of APC effects. No method can disentangle the linear components; anyway, the "detrended"fluctuations can be identified in an APCD model. The second problem relates to the durability of cohort effects over life course that can be either permanent or temporary: an APC-H a test "hysteresis" model is proposed. The suicide rates in sixteen WHO countries (periods 1970-1974 to 2005-2009 / ages 20-24 to 60-64) show the diversity of cohort dynamics. The models detect contrasted regimes of suicidity: for example, Finland and the Austria are more cohort-flat but Spain, Italy, Australia and the U.S. apparently show deep contrasts between cohort-specific suicide rates. Anyway, these cohort fluctuations are not stable over life course in Spain and in Australia (resilience) but are clearly remaining strong over life course in the U.S. and in Italy (hysteresis)