For counties w/ orders past 2020, rearrests reduce over time. Initial stat sig increase on rearrests starts to decrease after 4 months (Aug 2020). Over 1st year of order, avg effect was not stat sig (see Figure 6a). However, notably, rise in felony rearrests does not subside. 7/n

The increase in rearrests is driven by felony offenses, but we do not find any evidence that there is an increase for violent felonies, a concern regularly raised during this period by some (see Figure 5a). 6/n

Use a triple diff regression using staggered timing of county adoption and revocation and comparing offenses eligible for these orders vs not. Implementation of these orders (Mar-Apr 2020) sig increased prob and the number of rearrests within 30 days of an initial arrest. 5/n

Statewide mandate was Apr-June 2020, but counties could implement early and keep after. Majority of CA had an order until July 2022. Unlike most "bail reform" measures, no shift to electronic monitoring or intensive case mgmt b/c it occurred in a pandemic emergency setting. 4/n

There’s also a sharp increase in robberies and murders. Nearby cities experience declines in arrest with no change in crime, suggesting that the increase may be a response to the incident itself. I rule out other potential explanations, including investigations into the law enforcement agency. 7/n

After a high-profile police killing, arrests for low-level offenses continually plummet in the community where the death occurred for at least 1.5 years while arrests for more serious offenses do not change. Public scrutiny of police seems to be a driving factor in the reduction in arrests. 6/n