CT (Bridgeport), OH (Columbus), MI (Detroit), MN (Minneapolis and St. Paul), NC (Raleigh)
Using official lists of voters gathered immediately after the close of registration, we compiled a database of registered voters’ names and addresses. Names of individuals residing at the same address were grouped into households, which were in turn grouped geographically into walk lists... We restricted our attention to households with fewer than five registered voters, and in two sites, Raleigh and St. Paul, the population was restricted to households with at least two voters.
The procedures by which subjects were assigned at random to treatment and control groups varied slightly across sites. Subjects in Detroit, Minneapolis, and St. Paul were stratified into walk lists before random assignment, while Bridgeport and Columbus were not, but in each of these sites subjects were assigned the same probability of receiving a treatment. In Raleigh, the proportion of subjects assigned to the treatment condition varied across walk lists.
Door-to-door canvassing treatment (see Supplemental materials for script)
Turnout rate in 2001 local elections
Donald P. Green, Alan S. Gerber, David W. Nickerson