“EVIDENCE‐BASED POLICY” SHOULD REFLECT A HIERARCHY OF EVIDENCE

Janeen Buck Willison and I agree on one important point: We need a lot more research on wraparound services and prisoner reentry programs more broadly. Perhaps it is not surprising that that is the conclusion drawn by two researchers, despite our different takes on existing evidence. But this is the beauty of social science: We can use theory and existing evidence to formulate new hypotheses, and test those new hypotheses in future studies. There is always more to learn. But as we review current and future evidence, it is important to acknowledge that some studies produce more accurate estimates than others. Interpreting evidence requires more than simply comparing the number of studies finding benefits with the number finding null or detrimental results. Results should be weighted based on where a study falls in a hierarchy of evidence: raw correlational analyses near the bottom, outranked by studies with rich control variables, then by studies using matched comparison groups, then by studies using natural experiments to avoid selection bias (e.g., studies using sound difference-in-difference, regression discontinuity, and instrumental variable designs), then randomized controlled trials (RCTs) at the top. As we move up the hierarchy, we are more sure that the methods 1