Is it time for the use of pair-matching in all randomized controlled trials of crime and violence prevention? A review of the research

Pair-matching in randomized controlled trials (RCTs) has received increased attention in criminology, the social sciences more generally, and medicine and public health, with a growing body of research demonstrating the design's benefits over simple RCTs. We carry out a review of matched-pair RCTs compared with simple RCTs to address a somewhat provocative yet fair question for evaluation research on crime and violence prevention interventions: Is it time for the use of pair-matching in all RCTs? At the heart of this question is the ability of the design to most efficiently and robustly compare like with like, thereby, improving confidence in observed effects of intervention trials. Several key findings emerge from the review. First, it is inadequate to examine or discuss RCTs as a single, uniform evaluation design. Here, the key organizing construct is the unit of allocation: individuals; groups of individuals (or clusters); and geographical places. Second, the advantages vastly outweigh the disadvantages for the use of matched-pair RCTs compared to simple RCTs, and most of the advantages hold for all three units of allocation. Third, pair-matching can be used with rather small samples (>= 6 units) in clusterbased trials without compromising statistical power or degrees of freedom; less is known about individual- and place-based trials. Fourth, pair-matching cannot be used with some types of RCTs (e.g., cross-over) and is less amenable in other contexts (e.g., RCTs that enroll and randomize individuals on a rolling basis). Implications for evaluation research and public policy are discussed.

Automated summary

Pair-matching in randomized controlled trials has gained attention in various fields and has been shown to have more advantages compared to simple RCTs. This review highlights that RCTs cannot be treated as a single evaluation design, and the advantages of pair-matching hold for different units of allocation. Moreover, pair-matching can be used with small samples in cluster-based trials without compromising statistical power or degrees of freedom. However, its applicability may vary in different types of RCTs. The implications for evaluation research and public policy are discussed.