This study is motivated by the potential problem of using observational data to draw inferences about treatment outcomes when experimental data are not available. We compare two statistical approaches, ordinary least-squares (OLS) and instrumental variables (IV) regression analysis, to estimate...