Bounding Disagreements About Treatment Effects: A Case Study of Sentencing and Recidivism

Empirical inference on treatment effects is a core objective of social science research. The conventional practice is to obtain point estimates of treatment effects using models that make strong and thereby controversial assumptions about treatment selection and outcomes. In this paper we obtain bounds under weak nonparametric assumptions and explore how the bounds vary with the assumptions imposed. This mode of analysis clarifies the source of common disagreements about the magnitudes and signs of treatment effects. We use a treatment question facing the juvenile justice system to showcase the value of the approach in empirical social science research. We compare the impacts on recidivism of the two main sentencing options available to judges: confinement in residential treatment facilities and diversion to nonresidential treatment.