Patterns of arrest in domestic violence encounters: Identifying a police decision-making model

Arrest patterns of police officers in domestic violence events were explore d for a stratified random sample of domestic violence incidents (from polic e reports) in two police districts in Boston for the calendar year 1993. Th e initial analytic strategy used was the chi-square automatic interaction d etector, which conducts segmentation modeling useful for identifying intera ction effects among a predefined set of variables. The interaction effects were then entered into several logistic regression models to generate odds ratios in the predictions of arrest. Results showed that risk to the victim is the most important decision-making criterion for officers, and that off icers use variable pieces of information in deciding when to make arrests i n domestic violence events based an the different levels of risk. (C) 1999 Elsevier Science Ltd. All rights reserved.