Nickerson, Friedrichs & King 2006 - Partisan Door Hangers, Phone Calls & Canvassing
Partisan GOTV campaign tactics are estimated to produce mobilization effects similar to results reported in nonpartisan experiments. Partisan door hangers may be more effective than nonpartisan doorhangers in mobilizing voters.
Nickerson, David W., Ryan D. Friedrichs and David C. King. 2006. "Partisan Mobilization Campaigns in the Field: Results from a Statewide Turnout Experiment in Michigan." Political Research Quarterly, Vol. 59 (1): pp. 85-97.
Partisan Door Hangers, Phone Calls and Face-to-Face Visits
David Nickerson, Ryan Friedrichs and David King address the possible differences in voter responses to partisan mobilization efforts in contrast with the predominately non-partisan GOTV experiments reported in previous studies. The authors argue that there are two reasons why results from partisan efforts might differ from non-partisan mobilization initiatives. First, voters may respond differently to a partisan appeal. Partisan efforts do more than encourage turnout; they ask voters to support specific candidates or ballot measures. Providing a reason to vote, the authors argue, may boost turnout more than non-partisan messages. Alternatively, partisan messages may be less effective at mobilizing voters than non-partisan messages. Voters may be put off by the self-interest of partisan messages; may fear parties are seeking additional commitments of time and money; or may prefer the altruism of non-partisan initiatives. It may also be the case that partisan messages of persuasion may dilute mobilization messages.
To explore this question, the authors carry out experiments of three treatments in the 2002 Michigan gubernatorial election including door hangers with candidate information, partisan phone calls and personal visits from campaign volunteers. The experiments were conducted through the Michigan Democratic Party's Youth Coordinated Campaign in the14 state house districts where the Party felt that the youth vote could be most valuable. While the experimental design does not directly compare the effects of partisan and non-partisan messages delivered in parallel treatments within the same study, Nickerson, et al, draw comparisons between the results from these experiments and results from past nonpartisan experiments. The authors find that face-to-face canvassing and phone mobilization efforts bear no statistically distinguishable difference compared to previously-reported results of non-partisan efforts measured in past experiments. Partisan content door hangers, however, turn out to exhibit better mobilizing effects than in non-partisan efforts and are cost-competitive with more personal forms of contact.
Electoral Context: This experiment was conducted during the 2002 Michigan gubernatorial race. The political environment was that of a charged and closely contested race characterized by a lot of media attention and partisan and non partisan mobilization efforts. The race pitted "Republican Lieutenant Governor Dick Posthumus against Democratic Attorney General Jennifer Granholm. Granholm won with only 51% of the vote. Nearly 3.2 million voters - a record for a non-presidential year cast ballots." The partisan experiments in this article were part of a wider GOTV effort on behalf of Granholm.
Subject Population: Areas targeted for the experiment include fourteen state house districts in Michigan. Districts were selected according to three criteria: 1) size of Democratic or independent 18-35 year-old voting population; 2) the number of contested races on the ballot; and 3) capacity of local Democratic youth organizations. Subjects were drawn from a composite list of registered voters in Michigan. Registered voters at the same address and surname were grouped into households. The overall list contained 55,472 households and 70,591 voters.
Treatments: Thirteen sites completed the protocol for the door-hangers experiment; six completed the phone call experiment and four completed the face to face experiment. ("Five sites combined flyers and phone calls. Within the sites that used both flyers and phone calls, so few subjects (roughly 5000) were eligible to be in the treatment group for both, that the interaction effect [could] not be reliably estimated.")
Treatment 1 - Door Hangers with Candidate Information:
Randomization Procedure: For the door hanger experiment, geographic units based upon the first seven digits of a household's nine-digit zip code, were randomly assigned to the treatment group or the control group.
Treatment: The door-hanger experiment was the largest experimental aspect of the Youth Coordinated Campaign. Treatment and control groups for door hangers were assigned in geographic units based upon the first seven digits of households' nine-digit zip codes. Only units with 20 or more eligible households (with young voters) were selected as target areas. Targeted areas were randomly divided into treatment areas; sixty percent of the households were assigned to receive a door-hanger on one of the three days leading up to the election. The door hanger experiment was completed with 20,186 individuals in the treatment group and 13,336 individuals in the control group. Contact rates were high because no interaction is necessary to leave the flyer on the door; failure to treat could have been the result of an idiosyncratic factor such as a dog in the yard.
Findings: In ten of the thirteen door hanger experiments, the treatment group voted at a rate higher than the control group. When all 13 sites are pooled together, the rate of turnout in the treatment group is 1.2 percentage points higher than in the control group (standard error: 0.6 percentage points). This intent to treat effect implies that 12 votes were generated as a result of every 1,000 flyers. Because of the high contact rate, the treatment on the treated effect differs little from the intent to treat effect (1.6 percentage point increase in turnout with a standard error of 0.7 percentage points) The authors control for demographic characteristics and past voter turnout and approximate a 1.3 percentage point (standard error: 0.8 percentage points) increase in turnout among the treatment group for which door hangers were actually delivered.
Assuming labor costs of $15/ hour, the authors estimate a cost of $29 per vote. (Cost estimates do not include list purchase or printing materials.) The authors find that partisan door hangers are as effective as face-to-face contact.
Treatment 2- Volunteer Phone calls:
Randomization Procedure: For the phone call experiment, households were randomly assigned to the control and treatment groups. Treatment group phone numbers were then randomly ordered so households not contacted could be rolled into the control group.
Treatment: Phone numbers were available for about half the homes; eighty percent of these homes were assigned to receive a call. The treatment to control ratio was 65% (10,547) to 35% (5,634). Members of the treatment group received a phone call prior to the election. The more limited number of phone experiments (six of fourteen districts) was due to the more labor intensive nature of phone calls as well as the difficulty in finding space available at call centers. Volunteer callers at the various participating sites followed strict protocol in placing calls. Subjects were counted as treated if the volunteer spoke with the person or was able to leave a message. Contact rates ranged between 43.9 and 54.3 percent across districts/ experiments.
Findings: In five of the six phone call experiments, voter turnout was higher than in the control groups. Pooling the six estimates together, the authors find that the intent to treat effect of the calls was 1.6 percentage points with a standard error of 0.8 percentage points. Including contact rates in the analysis, the treatment effect on the people who actually received the calls was about 3.2 percentage points (with a standard error of 1.7 percentage points). Controlling for demographic characteristics and past voter turnout, the authors estimate a 3.5 percentage point (with a standard error of 1.7 percentage points) boost in turnout among the individuals who were actually contacted by phone calls.
Assuming labor costs of $15/ hour, the authors estimate a cost of $24 per vote. (Cost estimates do not include list purchase or printing materials or multiple phone lines). Phone calls were the most cost effective of the three treatments.
Treatment 3 - Face to Face Canvassing:
Randomization Procedure: For the face-to-face canvassing experiment, geographic units based upon the first seven digits of a household's nine-digit zip code, were randomly assigned to the treatment group or the control group.
Treatment: Only units with 20 or more eligible households (with young voters) were selected as target areas. Targeted areas were randomly divided into treatment areas; two-thirds of the households were assigned to receive a knock on the door in the weeks leading up to the election. The YCC attempted to contact 4,894 households through canvassing, leaving 2,458 in the control group. The contact rate for the face-to-face canvassing experiment was only eight percent. Low contact rates were attributed to the facts that young people move frequently; are home less often than older people; and often live in apartment buildings with restricted access. Additionally, canvassers failed to attempt most of the doors listed; they were demoralized by the long walks between target households and rarely completed walk lists.
Findings: Because of the low contact rate, the standard errors are high and the estimated effects of face-to-face canvassing on turnout are only suggestive. The authors found results consistent with nonpartisan mobilization experiments. Two-stage least squares regression estimates of the effect imply that canvassing may have boosted turnout rates by 16.8 percentage points, but the high standard error of 15.9 indicates that the estimates are only suggestive. Controlling for demographics and past turnout rates, the estimate suggests an 18 percentage point boost in turnout, but a standard error of 14.8 percentage points renders it impossible to determine the effects of the canvassing with certainty.
Assuming labor costs of $15/ hour, the authors estimate a cost of $28 per vote. (Cost estimates do not include list purchase or printing materials.)