INTRODUCTION

Some of the most interesting and important questions in criminology are causal in nature: Does neighborhood disadvantage influence an individual’s risk of criminal involvement? Does drug treatment work? Do changes in the threat of punishment deter crime? Distinguishing between causation and mere correlation is central to developing, testing, and refining our theories about the determinants of criminal behavior, most of which build on basic facts about how different constructs are causally related to one another. The distinction between statistical correlations and causal relationships is also of more than just academic interest, as most policy decisions hinge on causal questions. Getting the wrong answers to these questions leads to misguided policies that divert resources away from more effective alternatives, and sometimes even impose direct harm on society as well.

Although everyone in empirical criminology recognizes the challenges to valid causal inference created by the threat of omitted variables, the field remains divided about the most constructive way to proceed. One camp is often viewed as having adopted a purist, “randomized-trial-or-bust” perspective – a group that Sampson (2010) calls “randomistas.” Another camp consists of those we would call “research pluralists,” who are happy to consider findings from any sort of research design on a case-by-case basis.