Using the list of registered voters in Iowa and Michigan, a total of 60,000 households with listed phone numbers were randomly assigned to be called; the corresponding control group contains 1,846,885 randomly assigned households with listed phone numbers. Because a handful of small counties in the Michigan subsample did not provide 2002 voter records, we removed 1565 observations, bringing the treatment group total to 59,972 and the control group total to 1,845,348. The voter file also contained a large number of names without phone numbers.
The congressional districts of each state were divided into ‘‘competitive’’ and ‘‘uncompetitive’’ strata. Within each stratum, households containing one or two registered voters were randomly assigned to treatment and control groups. For two-person households, just one representative from each household was assigned to treatment or control; if there was another voter in the household, he or she was ignored for purposes of calling and statistical analysis.
Get-out-the-vote (GOTV) phone calls
Vote in the 2002 mid-term elections