Experimental Design
The National Voter Fund purchased a database with information of 3.7 Million African-American voters, of which approximately 1 million voters, across ten states, were chosen as sample for the trial. 98.3 percent (approximately 963,500) were randomly assigned to receive mailings and phones calls, while the remaining 1.7 percent was randomly assigned to receive nothing (control).
Phone calls were made to the treatment group voters starting four weeks before, and including on, the November 7 election day. Voters were attempted to be called up to five times before election day, although the precise mix of calls and mail varied by state.
The direct mail was targeted to young voters who had abstained from voting in recent elections. Prior to random assignment, one third of the sample received pre-experimental mailers to weed out bad addresses. These mailers were sent with equal probability to treatment and control groups thus should not lead to bias in the findings. The experimental mailer focused on the issue of discrimination, featuring a white cop pulling over a black motorist, in order to get-out-the-vote.
Voter turnout data was collected from public records.