Reintroducing “time” into the time series analysis of the police size‐crime relationship

Abstract

PurposeThe purpose of this article is to examine the causal relationship between crime rate measures (per 100,000 population) and police force size (full‐time employees per 100,000) within Milwaukee, Wisconsin. The data are annual, covering the years 1930 to 2004.Design/methodology/approachThe authors specify and estimate ARIMA and error correction models to examine the bivariate association between police force strength and total, property, and personal crime rates for a large, mid‐western city.FindingsConsistent with past research, the bivariate ARIMA analyses yield no evidence of a short‐term association between police force size and crime. However, the parameter estimates from error correction models indicate that changes in the level of crime have a longer‐term impact on police force strength.Research limitations/implicationsThis study focuses on a single municipality. Hence, before one can generalize to cities as a whole, the findings need to be replicated in other jurisdictions. Nonetheless, the findings do suggest that municipalities are more responsive to changes in the level of crime than prior ARIMA analyses seemed to indicate.Practical implicationsThe findings point to the conclusion that, when studying causal processes that operate over time, one must be careful not to remove long run information from the data in the attempt to control for the spurious effects of autocorrelation.Originality/valueThis paper represents the first attempt to apply error correction models to the examination of the longitudinal relationship between crime and police force size.