Frey and Suarez 2006 - GOTV and Bilingual Ballot Education Messages
Leaflets with GOTV messages and bilingual ballot awareness information in Spanish and English are not shown to be effective in mobilizing turnout.
Frey, Valerie A. and Santiago Suarez. "¡Mobilización Efectiva de Votantes! Analyzing the Effects of Bilingual Mobilization and Notification of Bilingual Ballots on Latino Turnout." Unpublished Manuscript. Institution for Social and Policy Studies, Yale University.
Valerie Frey and Santiago Suarez present results from a randomized field experiment designed to explore the mobilization effects of bilingual leaflets on voter turnout, particularly among Latinos. The authors conducted the experiment in the days leading up to the March 14, 2006 special election for a General Assembly Member in the state's 174th district. Frey and Suarez find suggestive evidence of a small but positive effect of bilingual leafleting on voting turnout among Latinos. On the contrary, results from the pilot study suggest a negative and statistically significant effect of bilingual ballot notification on voter turnout.
Electoral Context: This experiment was conducted in Philadelphia, PA, during the March 14, 2006 special election for a state General Assembly seat. The election was non-competitive -- the winning candidate won 89% of the vote on Election Day -- and produced low turnout.
Subject Population: Leafleting took place in three election wards in northeastern Philadelphia, immediately preceding the special election in March 2006. The election was chosen for its location in the northeastern United States, an unusual location among studies of Latino political participation and mobilization. The sample consisted of voters living in three electoral wards in the 174th state legislative district. After dropping all households with more than two registered voters, our final sample consisted of 15,550 registered voters.
Randomization Procedure: Frey and Suarez randomized treatment and control groups at the level of street block rather than at the level of individual voter. Randomizing at the street level promoted a more efficient use of time for leafleting. Further, the authors were able to randomize treatment areas and develop maps without a list of registered voters, which the local voter registration office withheld before the election. A single street block consisted of one street between two consecutive intersections. Randomization of the 455 streets resulted in 140 street blocks (or 24.5%) assigned to the treatment group, within which 65 were assigned to receive the additional treatment: the notification of bilingual ballots. Matching the streets to voter addresses placed 2,358 voters in the treatment group. Within them, 1,361 voters received notification of the availability of bilingual ballots.
Treatment: Two different treatment leaflets were designed by one of the authors as door hangers, which increased their visibility and produced minimal littering. The leaflets were double-sided and bilingual with a clear message in English and Spanish on each side: "VOTE." The message on the leaflets promoted community solidarity and civic duty, calling for the expression of the Latino voice in "our district." Each leaflet also included administrative information, such as the date and polling hours of the special election. A second treatment also included one line stating (in English and Spanish) "Ballots in Spanish are available at the polling place." This addition meant that the voters in the second treatment would receive the exact same flyer than those in the first treatment group would, but with an additional message telling the voters about the possibility of voting in Spanish.
Volunteers canvassed the designated streets hanging leaflets for over 12 hours during the weekend preceding the election. Canvassers worked in pairs and were proficient in Spanish and English. All canvassing was performed during the day, and canvassers were provided with both types of treatment leaflets to distribute along their routes. A shortage of canvassers resulted in only 67 blocks (and 37 with notification of bilingual ballots) receiving treatment during the leaflet canvassing.
Findings: The results measure both the increase in overall turnout and the increase in Latino turnout. Due to the variation in the number of registered voters in each block, the authors provide block-level weighted least squares analysis which offers a more efficient estimate of the intent-to-treat effect than OLS regression. The estimate of the combined treatment effect suggests that leafleting depressed turnout by 0.2 percentage points; however the estimate is not statistically significant (standard error: 0.9 percentage points). Controlling for whether or not voters received the additional bilingual ballot notification treatment, WLS estimates suggest that leafleting without notifying the voters on the block about the availability of bilingual ballots increased voter turnout by 1.9 percentage points (standard error: 13.1 percentage points).