Age-period-cohort models applied to the participation of the labor force: searching for a parsimonious version

Abstract

This paper applies age, period and cohort (APC) models to the analysis of trends of the participation in the Brazilian urban labor force over the last two decades, using data from the PNAD demographic surveys. Rate models based on count models were estimated by Poisson regressions, where the logarithm of the economically active population is the dependent variable, controlled by the logarithm of the population at working ages, as a function of the independent variables of age, period and cohort. Over the period, the percentage of women on the labor market in Brazil increased substantially for all ages. Men, on the other hand, showed a stable pattern of entering and leaving the labor market, maintaining a nearly steady state. This means a prevalence of age effects among men. Among women, age effects are also the most important, but period and cohort effects are also extremely relevant. This differentiation of female cohorts regarding participation in the labor force is not attributable to temporal or age variations. Seeking parsimony of models using direct measures of period and cohort effects, educational level is the best indicator for both men and women. The differentiated educational experience of the cohorts is an important determinant of the variation in participation in the labor force, especially among women.