Policing and crime: dynamic panel evidence from California
利用加州26年数据构建动态面板，用Arellano-Bover/Blundell-Bond方法处理内生性，发现增加警力并未显著减少犯罪，且同期性偏差可能比以往认为的更小。
Abstract We exploit police and crime data from California over 26 years to construct a dynamic panel, which is then estimated using Arellano–Bover/Blundell–Bond techniques to address concerns about simultaneity, unobserved heterogeneity, and inertial effects in the policing-crime relationship. We find no evidence that increases in police staffing lead to meaningful reductions in crime through either deterrence or incapacitation. Estimates are not wholly supportive of a compelling relationship between prior criminal offending and current police staffing; providing suggestive evidence that, at least within our sample, simultaneity bias may be more modest in nature than has been previously supposed in prior studies.