Log in Page Discussion History Go to the site toolbox


From BluWiki

Seth Greiner 624 summary notes – 23 October 2013

Gerber, A.S., and Green, D.P. (2000). The Effects of Canvassing, Telephone Calls, and Direct Mail on Voter Turnout: A Field Experiment. American Political Science Review 94:3, 653-663


Modern campaigns have less face-to-face contact, more phone, mailers etc. being used with a drop in the related numbers of adults who work for a political party along with a decline in civic organization participation – has contributed to lower voter turnout numbers since the 1960’s (pg. 653)

Hypothesis: personal canvasing is more effective that other methods for spurring voter turnout (pg. 653)

Previous studies have relied on survey data, non-experimental, to see relationships between voter turnout and political contact. Drawback = contact may not predict turnout (partisan direction). Also, control variable issues with regression analysis and covariate analysis (in previous studies) (pg. 653)

ANES (American National Election Studies) – no measure difference between phone contact and canvasing, same measure used (pg. 653)

Experimental Studies – have been rare, small sample sizes, no reliable inferences with lots of uncertainties present. These are significant limitations on previous research (pg. 654)

Methods (pg. 665): • Location – New Haven, Conn. September 1998 • Sample size of 29,000 registered voters (controlled for student/temporary population) • Design was to measure the effect of personal canvasing, telephone calls, and direct mail appeals on voter turnout • Random division of sample into control vs. experiment groups (2x2x4 design) • Treatment and Control overlap for the three experiments o 10,000 no intervention (control) o 7,300 1+ mailing, nothing else o 2,600 only personal face-to-face contact o 1,000 only telephone o 7,500 (remaining sample) 2 or more treatments • Personal canvasing designed to be uncorrelated with telephone and mail experiments for separate analysis (for future studies) • Random assignment to mail/telephone treatments, made calls more frequent for those who received mail because the two are correlated so must use multivariate analysis • Canvasing treatment = 5,800 persons • Canvasing control = 23,600 persons • Direct mail treatment = 14,700 persons • Direct mail control = 14,600 persons • Effectiveness of randomization checked with 1996 voter turnout data

Canvasing (pgs. 665,656): • Saturday and Sunday, the 4 weeks preceding the election, contact randomly selected registered voters • Questioners matched with ethnicity of neighborhood, questioners were majority graduate students, in pairs. • Specific addresses targeted, not whole streets, therefore were only able to cover 28% or 1,600 of 5,800 in treatment group. Size: 29 wards, average 28%, 9% standard deviation • External Validity: nonpartisan appeals to 1) civic duty, 2) close election, 3) neighborhood solidarity. Standardized script used for each appeal.

Direct Mail (pg. 656) • Intended to measure turnout effect of both number of mailings and message • First effect: treatment divided by three, sent one, two, or three mailings respectively o Subgroup = 4,900 persons (x3 groups) o Interval: group one: 15,13,8 group two: 13,8 group 3: 8 days before election o Subgroups (x3), one for each type of message matching canvasing and mail message o Nine postcards used to present same message (layout, color, style same for all)

Telephone calls (pg. 656): • Done the three days before the election • Sunday evening, Monday and Tuesday all day • 30 second scripted calls • Scripts mirrored those of the canvasing (civic duty and close election) • Neighborhood solidarity dropped due to implausibility from using out of state phone bank • Survey Sample Inc. used to conduct, cross-checked phone numbers, still many wrong numbers • Treatment group: 2,100 of 6,700 (32%) completion rate

Results (pgs. 657-661) • Tables show basic findings on personal canvassing • Regression analysis to confirm results with more statistical precision o All treatments taken into account o Covariates (past data) can be used

Conclusions (pgs. 661,662) • Canvasing for greater influence on voter turnout • Phone banks least effective influence on voter turnout • Hypothesis = “on the road to confirmation” o Falling rates of voter turnout reflect a decline in face-to-face political activity • Unanswered questions: o Partisan mobilization, this experiment was nonpartisan, similar pattern? o small but discernible effect of direct mail may offset some of the decline in personal mobilization (face-to-face) • Further research: o For generalizability of results o Other settings, other types of elections o Very little known about personal contact mechanisms • Importance of work/experiment: o Clues on why voter turnout has declined ♣ Some individuals won’t vote without face-to-face contact, personal encouragement o Focus on costs as well as personal connection for voting process • Final question: o “The question is whether the long-term decay of civic and political organizations has reached such a point that our society no longer has the infrastructure to conduct face-to-face canvassing on a large scale” (pg. 662)

Site Toolbox:

Personal tools
GNU Free Documentation License 1.2
This page was last modified on 22 October 2013, at 01:25.
Disclaimers - About BluWiki