Conrad Miller from MIT finds in his job market paper that US affirmative action regulation introduced from 1979 onwards had substantial effect on the black share of employees, also after deregulation. The exogenous variation comes from “changes in employers’ status as a federal contractor” and the fact that it was only federal contractors who were subject to these regulations. To get at the full dynamic effect of the regulation, Miller does not stop at comparing employers when they switch contractor status, but exploits also variation in when the firms are contractors for the first or the last time. In this way he can estimate whether there is a (persistent) causal effect also after a firm has lost his status as a federal contractor (has become “deregulated”). 

The event study results are striking: 

Figure 2 Event studies, from Miller 2014 The persistent effect of temporary affirmative action

The effect is quite small – becoming a contractor on average increases an establishment’s black share of around 0.15 percentage points per year – but the key point is that it persists, even when the firm is no longer is a contractor. There is much more in the paper, including a proposed explanation in terms of employers being induced to improve their screening procedures for potential employees.