Identifying Classes of Explanations for Crime Drop: Period and Cohort Effects for New York State

Abstract

This paper advances current understanding of the contemporary crime drop by focusing on the changes in the age distribution of arrests from 1990 to 2010. Using the New York State Computerized Criminal History (CCH) file, which tracks every arrest in the state, we apply standard demographic methods to examine age-specific arrest rates over time. We test whether the 25 % drop in the felony arrest rate can be best explained by period or cohort effects with special attention to how the phenomenon varies across crime types and regions within the state. Following the analytic approach of O’Brien and Stockard (J Quant Criminol 25(1):79–101, 2009), we fit the age–period–cohort (APC) model using the generalized inverse matrix, which creates an estimable model. We partition the model variation into each factor by subtracting the variation of the two-factor model from the variation of the three-factor model to provide a direct comparison of the two different classes of explanations for crime drop: period and cohort. Our analysis supports a cohort explanation over a period explanation. Controlling for the (substantial) variation due to age, the cohort effect accounts for twice as much of the remaining variation as the period effect. Specifically, the drop in arrest rates is concentrated in more recent birth cohorts across all ages. Although we found statistically significant age–period interaction effects for the younger age group (ages 16–20) in 1990 and 1995, the cohort effect was still a much stronger predictor of felony arrest rates than the period explanation, even with the age–period interaction. The current study reports that the overall drop in felony arrest rates from 1990 to 2010 is mostly due to decreased arrests among those who were born after 1970 rather than a universal drop across different age groups. We discuss but do not test two potential explanations—the legalization of abortion and the ban on leaded gasoline—for the underlying factors associated with a different criminal propensity among birth cohorts.