Assessing the Effect of Batterer Program Completion on Reassault: An Instrumental Variables Analysis

Abstract

The main objective of this paper is to investigate the utility of using the instrumental variables (IV) method to estimate batterer program efficacy, i.e., the program effect among batterers who complete batterer programs. This method takes account of possible confounding due to unmeasured traits of compliers and non-compliers. A structural model is used to estimate the impact of program completion on reassault using instrumental variables (IV) regression. Data on 640 batterers enrolled at three batterer programs are used. Results obtained from IV regression are compared with those obtained from a more traditional regression analysis. The results indicate that usual regression methods yield estimates of program effect that may be biased due to confounding by unmeasured batterer characteristics. Unfortunately, IV estimates may be unreliable due to failure of some of the assumptions on which they are based. If equations are adequately identified by the non-linear functional form used to estimate them, then IV results indicate that among a very select group of batterers, program completion significantly reduces the probability of reassault. The implications of confounding due to program non-compliance, program non-enrollment and attrition for future evaluations of batterer programs are discussed.