Green 2004 - Election Day Mobilization Campaign in New Jersey
An election day mobilization campaign targeting registered voters between the ages of 18 and 25 is found to increase turnout, particularly among voters that had previously expressed an intention to vote.
Green, Donald P. 2004. “The Effects of an Election Day Voter Mobilization Campaign Targeting Young Voters.” Unpublished manuscript, Institution for Social and Policy Studies, Yale University.
Summary: This study fills a gap in the literature documenting the effects of voter mobilization initiatives on turnout among young voters. Donald P. Green measures the effect of an election day mobilization campaign targeting registered voters between the ages of 18 and 25 years that had previously been contacted by a GOTV campaign. Treatment group members that were successfully contacted by phone (conversation or voice mail message) and urged to vote on the day of the election voted at a rate 5.0 percentage points higher than the turnout rate of the control group. The effect of the election day phone campaign was strongest among voters that had previously expressed an intention to vote during the pre-experiment mobilization campaign.
Electoral Context: This experiment was conducted in the context of the New Jersey state legislative elections in 2003. New Jersey’s state elections are held in off-years. Elections that do not included a gubernatorial race, such as the 2003 statewide election, are characterized by low rates of voter turnout.
Subject Population: The subject population consisted of registered voters between 18 and 25 years of age and was carried out in 60 precincts with large concentrations of young people including campus and non-campus precincts.
Randomization Procedure: Prior to the experiment, a mobilization campaign was conducted using a decentralized network of organizers and recruiters. Four organizers recruited volunteer precinct captains for each target precinct. These volunteers were in turn responsible for contacting young voters residing in their precincts. The volunteer precinct captains, with some assistance from volunteer phone banks and canvasser teams, were able to contact approximately three thousand voters through the peer-to-peer phoning and door-knocking. A few hundred of these voters were contacted by both methods. When contact was made with voters, they were asked whether they intended to vote.
Individuals who were contacted in some way by the campaign (including those for whom messages were left, whether with housemates or on message machines) were randomly divided into treatment and control groups. Excluding two very small counties for which voter turnout data are unavailable, the treatment group consisted of 1,418 voters and the control group included 1,399 voters.
Treatment: The treatment intervention consisted of a GOTV campaign conducted on election day. Precinct captains and volunteers made calls throughout the day to voters in the treatment group using a script that inquired whether subjects intended to vote and urged the subjects as young voters to reverse the trend in declining turnout among the members of their age group. Volunteers reached approximately 38 percent of the treatment group with phone conversations and left voicemail messages with 39 percent of the treatment group resulting in a combined contact rate of approximately 77 percent.
Findings: Individuals assigned to the treatment group voted at a rate of 17.0 percent compared to the turnout rate of the control group which was 13.2 percent. The intent-to-treat effect of 3.8 percentage points is significant at the one percent level. Controlling for county of residence, pre-election voter disposition and turnout in previous elections, the intent-to-treat effect is 3.0 percentage points (standard error: 1.2 percentage points). The effect of actually being contacted by a caller or a voice mail message on the day of the election on turnout is a difference of 5.0 percentage points (standard error: 1.8 percentage points) between the treatment and control groups without covariates and 3.9 percentage points (standard error: 1.6 percentage points) with controls for county of residence, pre-election disposition and past turnout. Comparing the effect of the election day campaign across different pre-election dispositions, the author finds that the effect was the strongest among voters who had previously indicated to volunteers that they intended to vote in the election. Among this subgroup, turnout rises from 16.9% to 27.5%, a large statistically significant increase.