Pooled results from eight voter mobilization volunteer phone call experiments indicate that brief volunteer phone conversations that are personal in tone can be effective in mobilizing voters. Volunteer phone campaigns can be cost-effective with door-to-door canvassing.

Nickerson, David. 2006. "Volunteer Phone Calls Can Increase Turnout." American Politics Research. Vol 34(3): 271-292.

David Nickerson presents pooled results from eight voter mobilization volunteer phone call experiments across six different cities during the 2001 presidential election and a year later, in anticipation of 2002 local elections.  The experiments were conducted in conjunction with Youth Vote, an organization seeking to mobilize voters between the ages of 18 and 30.  Nickerson argues that the quality of the conversations matter greatly in increasing turnout.  Contrary to previous studies, Nickerson finds that brief volunteer conversations can be effective in mobilizing voters if they are personal enough.  He also finds that volunteer phone campaigns are cost-comparative with more personalized methods such as door-to-door canvassing, at approximately $26 for each additional vote.

Youth vote targets voters between the ages of 18 and 30.  The six 2000 presidential election phone campaigns were carried out in the following cities:

Site 1: Albany, NY

Electoral Context: The State University of New York (SUNY) Albany campaign was carried out in the context of the November 2000 presidential election.

Subject Population: The population consisted of voters registered by student organizations on the State University of New York (SUNY) Albany campus (N= 1,122).  Respondents were new voters living on or near the college campus.

Randomization Procedure: Individual subjects on the target list of voters registered by the student organizations were randomly assigned to treatment and control groups.

Treatment: Treatment group members were each assigned to receive a phone call and control group members did not receive calls.  Calling began the day before the election.  Callers were not given a script; they were asked to introduce themselves; ask the person to vote; refrain from sharing opinions and finish the list before calling people twice.  Oversight was limited.  Callers spoke to or left messages for seventy-six percent of the subjects on the list.

Site 2: Stonybrook, NY

Electoral Context: The State University of New York (SUNY) Stonybrook campaign was carried out in the context of the 2000 presidential election.

Subject Population: The subject population consisted of voters registered by student groups on the State University of New York (SUNY) Stonybrook campus (N= 959). 

Randomization Procedure: Individual subjects on the target list of voters registered by the student organizations were randomly assigned to treatment and control groups.

Treatment:  Treatment group members were each assigned to receive a phone call and control group members did not receive calls.  Nearly all callers had previous phone advocacy experience.  Calls were made on the day before the election.  Callers were not given a script; were reminded to remain strictly nonpartisan; and were taught how to record call disposition.  Calls were made from volunteers’ dorm rooms, and oversight was, therefore, virtually nonexistent.  Callers spoke to or left messages for eighty percent of the subjects on the list.

Site 3: Eugene, OR

Electoral Context: The Eugene campaigns were carried out in the context of the 2000 presidential election.

Subject Population: The Eugene, OR, campaign targeted two distinct groups; one consisted of citizens registered by student groups (N= 1,202) and the other group was taken from a voter list purchased from a vendor (N= 1,960). This distinction provides a test of whether mobilization campaigns are unusually effective when a group tries to mobilize people it had registered during the preceding months.

Randomization Procedure: An independent control group was randomly extracted from both the student- generated list and the purchased list. The treatment groups for both lists were mixed together so that callers were unable to determine the source of the names.

Treatment: Calls were placed beginning six days before the election, so callers would have sufficient time to make two attempts to call everyone on the lists.  Most callers had past campaign experience.  Callers were provided a detailed script but were urged to develop their own style.  The calling took place in a large room with multiple lines and was supervised.  Half of the numbers on the purchased subject list were incorrect.  Volunteers spoke directly with or left messages for twenty-seven of the subjects on the purchased list and sixty-four percent of the student-harvested list.

Site 4: Boulder, CO

Subject Population Site 4 – The Boulder, CO, campaigns targeted two distinct groups; one consisted of citizens registered by student groups (N=1,094) and the other was taken from a list purchased from a vendor (N= 2,318).

Electoral Context: The Boulder, CO, campaign was carried out in the context of the 2000 presidential election.

Randomization Procedure: An independent control group was randomly extracted from both the student- generated list and the purchased list. The treatment groups for both lists were mixed together so that callers were unable to determine the source of the names.

Treatment: Calling took place between the Friday and Monday prior to the election.  The labor pool included college activists as well as individuals fulfilling court orders for community service. The training of new volunteers occurred each night prior to calling.  Calling took place in various offices with minimal supervision. During the course of the conversation, callers reminded respondents of the location of their polling places. Volunteers spoke directly with or left messages for forty percent of the subjects on the purchased list and sixty-four percent of the student-harvested list.

The 2001 local election campaigns were carried out in Boston, MA, and Seattle, WA and targeted people from the general population as opposed to targeting people exclusively between the ages of 18 and 30 as in the other experiments.

Site 5: Boston, MA

Electoral Context: The Boston experiment was carried out in the context of 2001 local elections. The mayoral election in Boston was a low salience election with one controversial issue on the ballot regarding  residential rents.

Subject Population: The Boston campaign targeted 7,055 randomly selected registered voters from a purchased list.  The only people on the list were active voters, so the list was heavily weighted with people who had voted in 2000. 

Randomization Procedure: Rather than dividing subjects into monolithic treatment and control groups, they were instead placed in a random order. Volunteers were then told to begin calling at the top of the list and gradually work their way down. The names toward the end of the list, for which contact was never attempted, constituted the control group because placement within the list is random. This protocol maintains statistical efficiency in the face of labor shortfalls.

Treatment: Local volunteers began calling respondents the Thursday prior to election day. Callers were drawn from a pool of local activists. The callers were young, and roughly two thirds had prior experience in phone canvassing. Calling took place in a single centralized room, so organizers could oversee every call. Sample scripts were provided and callers quickly developed individual variations on the script. Mayoral elections in Boston are generally uncompetitive, with low salience among voters. The most competitive issue on the ballot was a “livable city initiative” aimed at curbing rising residential rents. Callers were instructed not to endorse any particular candidate but were allowed to express an opinion on the livable-city initiative. Callers spoke with, or left messages for, forty-nine percent of the treatment group.

Site 6: Seattle, WA

Electoral Context: The Seattle experiment was carried out in the context of 2001 local elections.

Subject Population: The Seattle campaign targeted 44,000 registered voters whose name from the county list of voters could be matched to a household and phone number by a consumer research firm.  Callers contacted 1,420 individuals.

Randomization Procedure: Rather than dividing subjects into monolithic treatment and control groups, they were instead placed in a random order. Volunteers were told to begin calling at the top of the list and gradually work their way down. The names toward the end of the list for which contact was never attempted constituted the control group because placement within the list is random. This protocol maintains statistical efficiency in the face of labor shortfalls.

Treatment: Volunteers began calling the week prior to election day and managed to dial only 1,420 numbers. Callers were members of a campus Public Interest Research Group (PIRG) chapter. Almost every one of the callers possessed considerable experience in phone canvassing. A script was provided, but every caller with a little experience ignored the script and used an individual style. The calls were strictly nonpartisan throughout the campaign. Because all of the calls took place in a single office, there was considerable oversight of the calls by organizers. Callers were extremely successful in contacting people at the numbers dialed (73% were completed), but very few phone numbers in the treatment group were attempted. Thus, the contact rate for the Seattle experiment is a very low 9.6%

Findings: The eight experiments across six cities were small, so Nickerson presents pooled results.  The pooled results of the six experiments conducted during 2000 presidential elections in Albany, Stonybrook, Eugene, and Boulder indicate that being assigned to the treatment group increased the likelihood that subjects voted by 2.3 percentage points relative to the control group with a standard error of 1.0 percentage point.  Examining the effect of the calls among only those individuals in the treatment group who were successfully contacted by the volunteers suggests that a conversation with a volunteer caller increased the likelihood of voting by 4.5 percentage points with a standard error of 1.7 percentage points.  With regard to the 2001 election experiments, the low contact rate in Seattle rendered the Seattle experiment and pooled results of the 2001 experiments inconclusive. The results of the Boston experiment suggest that being assigned to the treatment group increased the likelihood that a subject voted by 1.6 percentage points relative to the likelihood that members of the control group voted.  However, the high standard error (1.6 percentage points) makes it impossible to say with reasonable certainty that the effect was not due to chance.  Adjusting for the contact rate, the associated increase in voting among individuals in the treatment group who were actually contacted was 2.9 percentage points with a standard error of 2.8.  When all eight experiments are pooled together, results suggest that being contacted by a volunteer phone bank increases the likelihood that a citizen will vote by 3.8 percentage points with a standard error of 1.4 percentage points.

Nickerson contends that volunteer phone calls are also a cost-effective method of mobilizing voters.  “If a volunteer caller can make 15 contacts per hour and is paid an hourly wage of $15, then volunteer phone banks require an investment of $26 to create one vote.”