Partisan phone calls and door-to-door canvassing carried out by a grassroots political network do not result in significant mobilization effects. Higher contact rates could improve estimates.

Green, Donald P. 2004. "Results from a Partisan Phone and Canvassing Mobilization Campaign in Pennsylvania Primary Election." Unpublished Manuscript. Institution for Social and Policy Studies, Yale University.

Donald Green reports results from a partisan mobilization campaign carried out in conjunction with a grassroots political network seeking to mobilize pro-choice Democratic women. The campaign tested two mobilization methods experimentally: door-to-door canvassing and phone calls. For both studies, of the total target list was assigned to a control group that received neither calls nor canvassing visits. Results indicate that the door-to-door canvassing contact rate was too low to estimate an effect. Results of the phone campaign suggest that the calls did not appear to increase turnout. Responses to questions in subsequent phone surveys indicate that subjects on the targeted lists may have been called so many times in the past as to have reduced the effectiveness of contact. It is possible, however, that although the phone campaign did not increase turnout, it may have persuaded some voters with regard to candidate selection.

Electoral Context: The campaign was carried out in the weeks leading up to the April 27, 2004, Democratic primary in a Congressional district in Pennsylvania. The purpose of the campaign was to bolster support for a Democratic Congressional candidate aiming to both mobilize turnout as well as persuade voters to decide in favor of the organization's preferred candidate.

Treatment 1: Door to Door Canvassing

Subject Population: The canvassing campaign targeted registered voters who had participated in one of the past three primary elections in 2000, 2002 and 2003, and women who had participated in the previous two general elections. This ambitious, targeted canvassing universe consisted of approximately 84,000 voters.

Randomization Procedure: Addresses were randomly assigned to treatment and control groups.

Treatment: Paid canvassers contacted voters and attempted to increase turnout among the Democratic candidate's supporters as well as persuade other voters of the candidate's merits. Only a small portion of targeted candidates were contacted. The target population consisted of 84,000 voters. 40,000 lived in precincts that were never canvassed; of the remaining 44,000, only 1,473 were reached by paid canvassers. Even when Green excludes the nine-digit zip codes in which no contact was attempted, 23,000 targeted voters remain, and the contact rate is still low at six percent.

Findings: Although there were hints in the data that the canvassing increased turnout, the low contact rate renders it impossible to make estimates with statistical precision.

Treatment 2: ID/ Callback Phone Campaign

Subject Population: The target universe for the phone calls consisted of voters with known phone numbers who had voted in at least one of the previous three primaries. More than 52,000 individuals met the criteria.

Randomization Procedure: Approximately seven percent of the targeted voters were randomly assigned to the control group.

Treatment: This phone campaign consisted of two calls. The first call was a voter identification call that was placed over a month before the election; callers asked voters about their candidate preference and the issue of most importance to them. The second call targeted respondents from the first call that indicated support for the organization's preferred candidate. The script for the second contact focused on the issue that the respondent had indicated was most important to him/her in the first phone call. The second round of calls was completed a week before the election. The phone bank completed first-round voter identification surveys with 22,000 individuals with a contact rate of 40%. Of the respondents 5,000 indicated support for the candidate advocated by the organization and 12,000 were undecided. The phone bank contacted 7,800 in the second round of calls.

Findings: The results do not indicate that the phone campaign increased turnout. Turnout in the control group was 53.1% and 53.0% in the treatment group. The difference in turnout between the treatment and control groups is not statistically significant from zero. Controlling for various demographic variables does not increase the power of the estimate. While there is no evidence that phone calls increased turnout, the results do not rule out the possibility that phone calls persuaded voters' choices.