# A dynamical systems analysis of criminal behavior using national longitudinal survey of youth data

## TL;DR

This paper uses a mathematical model to study how criminal behavior changes over time and how different factors affect crime rates and racial disparities.

## Contribution

The paper introduces a new dynamical systems model to analyze criminal behavior and its long-term effects using longitudinal survey data.

## Key findings

- First-time arrest rates increase long-run crime for all subgroups, but not repeat offender arrests.
- Black men's crime levels are most sensitive to initial flows into crime and rehabilitation.
- Black women with no arrest history are more likely to desist from crime than other subgroups.

## Abstract

Building on previous work on the spread and sustenance of crime, we construct and analyze a dynamical systems model of criminal involvement, arrest, desistance, and rehabilitation to be estimated empirically using interviews in the National Longitudinal Survey of Youth. We examine how marginal increases in flows between states interact to decrease or increase the long-run level of crime, and whether this varies by subgroup. We study how observed racial disparities along certain pathways interact to generate macro-level disparities in criminal involvement as measured by arrest and self-report. Finally, we discuss the implications of the model for a broader policy debate on crime control and for competing explanations of the Black-White gap in criminal involvement. We find, among other conclusions, that marginal independent increases in first-time arrest rates (but not arrest rates for repeat offenders) increase long-run crime for all subgroups; that long-run crime levels for Black men are most sensitive to initial flows into crime and arrest and to rehabilitation; and that among people with no arrest history, Black women are significantly more likely than other subgroups to desist the following year.