Instrumental variables estimators typically must satisfy monotonicity conditions to be interpretable as capturing local average treatment effects. Building on previous research that suggests monotonicity is unlikely to hold in the context of school entrance age effects, we develop an approach...

We employ a unique 10-year panel dataset from California to examine both the effect crime has on homelessness as well as the effect homelessness has on crime. Our main estimator accounts for endogeneity by incorporating dynamics, controlling for time invariant unobserved heterogeneity, and...