Door-to-door canvassing is found to increase turnout not only in the person directly contacted by the canvassers. In two-member households, the uncontacted member of the households is also mobilized by GOTV canvassing.

Nickerson, David W. 2006. "Is Voting Contagious? Evidence from Two Field Experiments." Paper prepared for presentation at the Annual Meeting of the Midwest Political Science Association, Chicago, IL, April 20-23.

David Nickerson examines the norm of voter turnout and the extent to which voting is contagious behavior within social networks, particularly within two-voter households. Existing interpersonal influence literature cannot "separate the unique effect of contagion from selection process, congruence of material interests or exposure to external forces." Current studies assume an atomistic voter with weak ties to others. For this reason existing GOTV experimental literature may understate the effectiveness (and cost-effectiveness) of mobilization because it fails to account for the effect of the intervention on people who interact with the contact voter in the treatment group. Nickerson measures interpersonal influence by analyzing two placebo-controlled experiments conducted in Denver and Minneapolis. The treatments involved face-to-face canvassing of one person within two member households and a parallel recycling canvassing effort that served as a baseline. Turnout among uncontacted members of households was treated as contagion.

Electoral Context: Both experiments were conducted in the context of the 2002 Congressional Primary elections.

Subject Population: The sample consisted of two-member households in Minneapolis, Minnesota and Denver, Colorado. Minneapolis and Denver are large cities with majority white populations. Neighborhoods with a high density of two-voter households were targeted in order to facilitate efficient door knocking campaigns. These neighborhoods exhibited a higher ratio of home ownership than the national average.

Randomization Procedure: Households were randomly assigned to three conditions. Subjects would a) receive a GOTV appeal; b) receive encouragement to recycle (placebo); or c) receive no contact from the campaign.

Treatment: Prior to the 2002 Congressional Primaries, households in Denver and Minneapolis with two registered persons were culled from the voter rolls and randomly assigned to three conditions: a) receive a GOTV appeal; b) receive encouragement to recycle; c) receive no contact from the campaign. Each appeal was delivered door to door the weekend prior to the Tuesday primary by paid workers. Canvassers were trained and instructed to provide the correct appeal at the correct house (voting or recycling); give the pitch to whoever answered the door; and ask the name of the individual and record the person directly contacted. 486 households received the GOTV treatment and 470 received the recycling treatment.

Findings: The GOTV campaign effect sizes among treated individuals for Denver and Minneapolis were 8.6 percentage points (standard error: 4.2 percentage points) and 10.9 percentage points (standard error: 4.1 percentage points) respectively. The mobilization effects among untreated members of the contacted households are estimated to be 5.5 percentage points (standard error: 4.1 percentage points) in Denver and 6.4 percentage points (standard error: 4.1 percentage points) in Minneapolis. No direct or secondary effect is statistically significant, but pooled together, the estimated secondary effect is estimated to be 6.0 percentage points which is significant at the .05 level (standard error: 2.9 percentage points). Nickerson finds that the placebo (recycling) group voted at roughly the same rate as the control group (38.3% in Denver and 17.2% in Minneapolis.) Four two-stage least squares models are introduced that control for progressively more demographic and geographic attributes as well as past voting behavior. In all four models, contagion effects are substantively and statistically significant. The base model estimates a contagion effect of 6.1 percentage points (standard error: 2.3 percentage points). Controlling for voter history, the contagion effect is estimated to be 5.6 percentage points (standard error: 2.2 percentage points). Controlling for demographic variables, the contagion effect is estimated to be 5.9 percentage points (standard error: 2.2 percentage points); and, finally, controlling for geography, the estimated contagion effect is 6.8 percentage points (standard error: 2.5 percentage points). It appears as though mobilization exhibits the positive externality of mobilizing household members other than those who were directly treated by the campaigns.