“…Second, by exploiting the time-series and cross-sectional aspects of our data, the panel VAR allows us to model additional complexities than if we had either only time-series or cross-sectional data. Like many static and dynamic panel data models, we are able to control for unobserved characteristics across states that simultaneously determine income inequality and crime using state fixed effects, while the time-series aspect allows us to include sufficiently long lags of the endogenous variables, thereby practically eliminating concerns about endogeneity (Koop and Korobilis, 2016). Third, the time period for estimation, before taking lags, differencing, and other standard data transformations into account, is 1960-2015 (T = 56) for the fifty states and DC.…”