From Euclidean distances to APC models

In this paper we show that a recently developed method for the study of “cultural” differences, called DBS-EM, or Distance Between Strata estimated with the EM (Expectation Maximization) algorithm, can also be used to circumvent the difficulties posed by APC (or Age, Period, Cohort) models. The DBS-EM method produces an original measure of the distance (dependent variable) between any two subsets of observations (strata) within a sample, where the stratification variables can be interpreted as regressors. When these stratification variables are age, period, and cohort, what results is an APC model which, however, proves immune to the “intrinsic collinearity problem” (C = P−A). With a few limitations, to be sure, which are discussed in the article. In our application to Italian data over the years 1993–2013, age and cohort strongly shape cultural consumption, while cohort and period impact, but only up to a point, on political participation.