Door-to-Door GOTV Canvassing — New Haven
Yale University (Green & Gerber) · New Haven, CT, USA · 2000
Summary
The New Haven canvassing experiment established one of the most durable findings in political science: personal contact substantially increases voter turnout, while impersonal contact (mail, phone) has little effect. The mechanism is human social connection — canvassers are neighbors and community members, not call-center operators. The size of the effect (5–9 pp) is large relative to most civic interventions and has been replicated across party contexts, candidate types, and countries. The experiment also established a key null: direct mail, despite its ubiquity in political campaigns, has no reliable effect on turnout. Money spent on mail could produce 6x more turnout if spent on canvassers.
Research question
"Does in-person door-to-door canvassing increase voter turnout compared to direct mail or phone contact?"
Methodology
Intervention
Registered nonvoters and sporadic voters randomly assigned to: (1) personal visit from trained canvasser with civic duty message, (2) direct mail with same message, (3) phone call with same message, (4) no contact. Canvassers visited multiple times if initial contact failed.
Assignment
Randomized controlled trial (individual)
Sample size
29,380 registered voters in New Haven; multiple replications totaling 100,000+
Primary outcome
Voter turnout in November 2000 general election (verified through official records)
Effect estimate
Personal canvassing: +6 to +9 pp; direct mail: +0 to +1 pp (not significant); phone calls: +3 pp. Effect of canvassing has been replicated in 50+ subsequent studies with consistent range of 5–10 pp.
Decision
Canvassing became standard in campaign strategy for competitive races; nonpartisan organizations (Obama campaigns, state Secretaries of State) adopted structured canvassing programs; finding replicated in US, UK, Canada, and multiple European countries
Result
Positive
Personal canvassing: +6 to +9 pp; direct mail: +0 to +1 pp (not significant); phone calls: +3 pp. Effect of canvassing has been replicated in 50+ subsequent studies with consistent range of 5–10 pp.
Evidence strength
Strong
Randomized trial, replicated across multiple sites or studies.
Replication status
Replicated
Institution
Yale University (Green & Gerber)
Location
New Haven, CT, USA
Year
2000
Policy area
Voter Engagement
Mechanism
Community engagement