Door-to-door canvassing is found to be an effective means of mobilizing Latino voters in a variety of electoral contexts.

Michelson, Melissa R. 2005. “Meeting the Challenge of Latino Voter Mobilization” The Science of Voter Mobilization. Special Editors Donald P. Green and Alan S. Gerber. The Annals of the American Academy of Political and Social Science. vol. 601: 85-101.

Melissa R. Michelson presents the results from four field experiments examining the effects of door-to-door canvassing on turnout among Latino voters. Michelson notes that turnout among Latino voters has historically been low, and for this reason, mobilization efforts have recently begun to target Latino communities. Previous research about the effectiveness of mobilization campaigns among Latino voters relied exclusively upon survey analysis. The first experiment examines the effects of canvassing by bilingual Latino canvassers with civic duty and ethnic solidarity messages on turnout. The second experiment examines the effect on turnout among voters in the Latino community canvassed by Latino canvassers compared with voters canvassed by non-Latino canvassers. The third experiment compares the effects of partisan versus nonpartisan messages. The fourth experiment explores the importance of the message delivered in canvassing efforts. Overall, results from Michelson’s field experiments demonstrate that door-to-door canvassing is effective in mobilizing Latino voters and is a cost effective mechanism for substantially increasing turnout in these communities.

Experiment 1: Dos Palos, California, 2001

Electoral Context: This canvassing experiment was conducted in the weeks leading up to the November 6, 2001, school board election of the Dos Palos–Oro Loma Unified School District in Central California.

Subject Population: The subject population consisted of 2,775 registered voters in Dos Palos.

Randomization Procedure: 1,709 of the 2,775 individual registered voters were randomly assigned to the treatment group and targeted for mobilization.

Treatment: During the two weekends prior to the election, thirty bilingual students from California State University, Fresno, visited each of the individuals in the treatment group and handed out a one-page flyer to each contacted voter that included the names and contact information of the candidates and the location of the voter’s polling place. Some treatment group voters were randomly assigned to receive a mobilization appeal emphasizing civic duty and the remaining treatment group members received a mobilization message encouraged them to vote as a tool for ethnic solidarity. Canvassers successfully contacted 76.7% of the voters on the treatment list.

Findings: Michelson analyzes the effects of the treatment on turnout among the following subgroups of the sample: Latino Democrats, Latino non-Democrats, non-Latino Democrats, and non-Latino non-Democrats. Analysis of Intent-to-Treat effects suggests that canvassing had the largest effect on Latino Democrats: a 7.1 percentage point (Standard Error: ) increase among recipients of the civic duty message and 10.6 percentage points among recipients of the ethnic solidarity message.

Experiment 2: Fresno, California, 2002

Electoral Context: This experiment was conducted in the weeks leading up to the November 5, 2002 general gubernatorial election in Fresno, California.

Subject Population: The subject population consisted of a list of 3,000 young registered Latino voters (between the ages of 18 and 25).

Randomization Procedure: Individuals from the list were randomly assigned to equal-sized treatment and control groups.

Treatment: Eighty student canvassers from California State University, Fresno, representing Latino, African American, Anglo and Asian ethnicities, canvassed in the two weekends prior to the election. Canvassers were divided into forty pairs matched by ethnicity (either two Latinos or two non-Latinos) and assigned to a portion of the list of registered voters. Treatment list voters were randomly assigned to receive either a civic duty or an ethnic group solidarity mobilization message. Canvassers also delivered flyers with information about the candidates and voters’ polling places. 50% of the treatment list was contacted. Latino canvassers were more successful at contacting Latino voters regardless of canvassing experience.

Findings: Voters in the treatment group turned out at rate 2.4 percentage points higher than voters in the control group (p= .013, Fisher’s exact test). There was no distinct difference in turnout among voters who received the civic duty message relative to voters who received the ethnic solidarity message. Many subjects had never voted before. Among this group, voters that were contacted by non-Latino canvassers turned out at a rate of 8.1 percent compared to 5.5 percent in the control group, and the voters in this group that were contacted by Latino canvassers turnout out at 8.6 percent in the treatment group relative to 8.0 percent in the control group. Among subjects that had previously voted, individuals assigned to non-Latino canvassers turned out at a rate of 27.3 percent in the treatment group relative to 13.3 percent in the control group and individuals assigned to Latino canvassers turned out at a rate of 35.0 percent relative to 9.7 percent in the control group.

Experiment 3: Fresno, California, 2003

Electoral Context: This experiment was conducted during the October 7, 2003, gubernatorial recall election in Fresno, California. Turnout was expected to be high in this election due to the unusual nature of the recall.

Subject Population: The mobilization effort targeted precincts with a large number of young voters (ages eighteen to thirty) generating a list of 6,715 registered voters.

Randomization Procedure: Individual voters were randomly assigned to treatment or control groups. The treatment group was then randomly divided into two groups, one which was assigned to receive a partisan message versus a non-partisan message.

Treatment: One hundred student canvassers were allowed to choose between partisan and nonpartisan canvassing. Partisan canvassers were assigned to canvass voters of their preferred parties. Nonpartisan canvassers were not told the partisan affiliation of the names on their lists. Canvassers contacted voters between September 28 and October 6. Canvassers successfully contacted 33.7 percent of the treatment list.

Findings: Results of the experiment suggest no statistically significant turnout between voters in the control and treatment groups or between the partisan and nonpartisan messages.

Experiment 4: Maricpa County, California, 2003

Electoral Context: This experiment was conducted during a special election in Maricopa Coutny, California, on November 4, 2003. The ballot included county-wide Proposition 414, a measure to establish a special hospital district. Voter turnout was expected to be low.

Subject Population: The Association of Community Organizations for Reform Now (ACORN) supported the measure on the basis that much of their Latino membership relied on the hospital’s service. ACORN sought to mobilize Latino voters in twenty-nine low to middle income precincts

Randomization Procedure: Fifteen percent of individual voters were randomly assigned to the control group. It was decided that the treatment group would be expanded to include all members of households which contained members of both the control and treatment groups. The final treatment group included 5,216 individuals and the control group consisted of 545 individuals.

Treatment: Between the dates of Monday October 13 and Tuesday November 4, twenty-eight paid canvassers including full-time ACORN community organizers and inexperienced locals contacted targeted lists. The first phase consisted of voter identification between October 13 and October 30 in which canvassers delivered memorized scripts about the hospital’s uncertain future and the need for its presence in the community. Voters were asked if they could be counted on to vote yes, and if they did so were asked to sign a pledge sheet. Canvassers contacted voters who pledged support a second time and tried to reach uncontacted voters on subsequent visits. At the end of the voter identification phase, 75.2 percent of the targeted list had been contacted. The mobilization phase of the campaign began on November 1. Canvassers contacted all voters that were indicated they would vote “yes” or were undecided three times using a script that reminded voters of the importance of the hospital and emphasized the anticipated low rate of turnout in the election. Contact rates were not measured with the exception of a rough tally in which ACORN estimated having contacted nearly all targeted subjects at least once in the GOTV phase.

Findings: For the purpose of analysis, one and two member households are separated. Logistic regressions within each type of household confirm the randomness of assignment to treatment and control groups. Results demonstrate that the ACORN effort produced large mobilization effects in both one-voter and two-voter households. In one-voter households, the intent-to-treat effect is 8.5 percentage points with a standard error of 1.8. In two-voter households, the intent-to-treat effect of 14.1 percentage points with a robust clustered standard error of 3.6 percentage points.