Experimental Design
The researchers obtained 11,000 nine-digit zip codes from the Ohio Department of Taxation and randomly allocated 1,000 to treatment and 10,000 to control. Of these, some zip-codes were found to have no registered voters thus reducing treatment group to 837 and control group to 9,072 zip-codes respectively. Treatment group received a series of mailings before and after the federal election cycle while the control group received nothing. The total number of voters who received the mailing was 3,392.