Estimating causal crowding out effects with observational, nonexperimental data is difficult. Consider the following experiment as a potential alternative: A local public radio station receives a large federal grant to help fund station operations. You suggest that the station run a small experiment. The station sends out a promotional email to households in the station listening area, randomizing the recipients of the email into two groups-one group receiving an email promoting the new federal grant, and the other receiving an email highlighting the great work the station has done recently but omitting any mention of the grant. You then track donations over the next 6 months from these two groups. How could this experiment provide high-quality evidence on the size of crowd-out?