In the immediate aftermath of the 2020 wave of civil unrest in the United States, triggered by the murder of George Floyd, calls for de-funding the police countered with more moderate proposals for “only throwing out the bad apples” could be heard. Similarly, following transition from authoritarian rule, new democracies have to reckon with members of their repressive and enforcement apparatus. One solution is to just disband enforcement agencies in their entirety and build new ones, along with institutions of the new democratic state. Another, competing approach, is to select the “bad apples” and fire them. The latter approach is one a new democracy may want to take when it is strapped for competent enforcement agents and when competent as opposed to predominantly loyal agents were indeed employed by the preceding autocratic regime.

This paper analyzes the costs and benefits of conducting what we call a thorough purge—understood as dismissing everyone up to a certain point in the law enforcement hierarchy—and a selective purge, where the competence and loyalty of agents is evaluated on a case by case basis. We propose a theory that builds on the loyalty-competency trade-off facing new democracies deliberating how to reform their security apparatus. Our model allows to make predictions about the causes as well as consequences of thorough as opposed to selective purges. We corroborate our theory using newly acquired data from the Institute of National Remembrance (IPN) in Poland on the operation of verification commissions in 49 sub-national regions in 1990.