The Effects of Canvassing, Telephone Calls, and Direct Mail on Voter Turnout: A Field Experiment
American Political Science Review, 2000 · paper · Political science
A field experiment with voters in New Haven comparing nonpartisan get-out-the-vote efforts before the 1998 election. Personal canvassing raised turnout substantially, direct mail slightly, and telephone calls not at all.
Checked: PDF from web.archive.org, fetched 5 Oct 2026. Page numbers are those of that PDF.
0 findings
None that passed a second check.
8 minor (typos and the like; most checked by the judge only)
Wording The text says of Miller, Bositis, and Baer (1981) that "the total sample size was only 215", but Table 1 lists 79, 80 and 81 under "Number of Subjects (Including Control Group)" for that study's canvass, mail and telephone rows; neither separate nor shared control groups give 215.
- p. 2 Right column, first paragraph
was only 215
- p. 2 Table 1, Miller, Bositis, and Baer (1981) rows, Number of Subjects (Including Control Group)
79
- p. 2 Table 1, Miller, Bositis, and Baer (1981) rows
79 Canvass ... 80 Mail ... 81 Telephone
- p. 2 text, right column
the total sample size was only 215
Only the authors can say what is right here.
Judge's note: Read Table 1 and the right column on the page 2 image. The three rows sum to 240. If one control group of size c is counted in each row, 240 - 2c = 215 gives c = 12.5, not a whole number. The paper gives no further explanation of the column. The original report is not in the files, so only the authors can say which figure is meant; the counts may be on a basis the table does not describe.
Second check: another model agreed
Report a problem with this finding
Wording Footnote 4: "Even if the true effects of leafleting and mailers were equivalent to three mailings, the overall bias to our estimate of the effect of personal canvassing would be less than 1 percentage point"; with the paper's own figures the bound holds for the turnout differential but not for the contact-rate-adjusted estimate.
- p. 4 Footnote 4 (continued from previous page)
be less than 1 percentage point
- p. 3 Footnote 4
719 were randomly
- p. 4 Footnote 4 (continued)
(683
Only the authors can say what is right here.
Judge's note: Read footnote 4 on pages 3-4 and Table 5 on page 7. My reworking: 719 + 683 = 1,402 of the 5,794 in the treatment group (24.2%) got a mailer or leaflet. Three mailings are worth 3 x .0063 = 1.89 points (Table 5, column 2), so treatment-group turnout would rise by about 0.46 points. The paper's estimator (equation 6) divides the differential by the contact rate of 27.87%, giving about 1.6 points; with the column 1 coefficient (.0058) about 1.5 points, with the probit figure of 2.5 points about 2.2. The stated bound is met only if it refers to the unadjusted differential. The authors may have used a different basis (for example, not all 683 actually received a leaflet), so this is a query rather than a certain error.
Second check: another model agreed
Report a problem with this finding
Wording Cost calculation: "sending three mailings to each household (containing an average of 1.5 voters)"; the sample is "29,380 individuals (22,077 households)", an average of 1.33 voters per household.
- p. 9 Regression Results, last paragraph (top of left column)
taining an average of 1.5 voters
- p. 3 Experimental Design, first paragraph
29,380 individuals (22,077 house-
- p. 9 Discussion, cost paragraph
containing an average of 1.5 voters
- p. 3 Experimental Design
29,380 individuals (22,077 house- holds)
Only the authors can say what is right here.
Judge's note: Read page 3 and page 9 images. 29,380/22,077 = 1.33 per household. The sample has 14,774 one-voter and 7,303 two-voter households, so 1.5 is the average household size seen by an individual (half the individuals live in two-voter households), not the average per household that the per-household mailing cost calls for; it may also be meant as a round figure. With 1.33 the mail figure would be about $45 rather than $40. The paper does not say which is intended.
Second check: another model agreed
Report a problem with this finding
Wording Table 4, civic duty: "Turnout Differential (2.43%)/Contact Rate (26.90%) = 9.1%" (repeated in the text as "the 9.1% associated with the civic duty appeal"); the printed terms give 9.0%.
- p. 6 Table 4, Implied Effects, Civic duty row
Turnout Differential (2.43%)/Contact Rate (26.90%) = 9.1%
- p. 6 Right column, paragraph beginning 'Table 4 suggests'
slightly better than the 9.1% associated with the civic duty appeal
- p. 6 Results text below Table 4
better than the 9.1% associated with the civic duty appeal
- p. 6 Text under Table 4
9.1% associated with the civic duty appeal
Only the authors can say what is right here.
Judge's note: Read Table 4 and the text on the page 6 image. 2.43/26.90 = 9.03. The contact rate is exactly 534/1,985 = 26.90%, so 9.1 (at least 9.05) would need a differential of at least 2.4346, a sliver of the range that rounds to 2.43; I checked that no whole-number vote counts for 1,985 treated and 23,586 control subjects give a differential in that sliver. The other rows reproduce (1.48/29.03 = 5.10, 3.36/27.75 = 12.11, Table 3 2.43/27.87 = 8.72), and the three differentials weight-average to the overall 2.43. Either the 2.43% or the 9.1% is off in the last digit; the difference is 0.1 point and changes nothing of substance.
Second check: another model agreed
Report a problem with this finding
Typo Footnote 2 cites "Kraut and McConohay 1973"; the reference list has "Kraut, Robert E., and John B. McConahay. 1973."
- p. 2 Footnote 2
and McConohay 1973
- p. 11 References, Kraut entry
Kraut, Robert E., and John B. McConahay. 1973.
- p. 2 Footnote 2
Kraut and McConohay 1973
Judge's note: Both spellings are visible in the page images (footnote 2 on page 2, reference list on page 11), so this is not an OCR slip. The two spellings are inconsistent within the paper; the reference-list form is presumably the intended one.
Second check: none (judge only)
Report a problem with this finding
Typo "none of the experiment relied on self-reports."
- p. 9 Conclusion, top of right column
none of the experiment relied on self-reports.
Judge's note: The wording is as quoted on the page 9 image (top of right column). The singular is awkward and "experiments" may be intended, since the paper elsewhere speaks of its experiments, but the conclusion also says "our experiment", so the singular can be read as deliberate (no part of the experiment).
Second check: none (judge only)
Report a problem with this finding
Typo Reference list prints "Putnam, Robert C. 2000. Bowling Alone: The Collapse and Renewal of American Community. New York: Simon and Schuster."; the author of Bowling Alone is Robert D. Putnam (and the book's subtitle reads "Collapse and Revival").
- p. 11 References, Putnam entry
Putnam, Robert C. 2000. Bowling Alone
- p. 11 References, Putnam entry
Putnam, Robert C. 2000. Bowling Alone: The Collapse and Renewal of American Community.
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: the author of Bowling Alone is Robert D. Putnam and the published subtitle is "The Collapse and Revival of American Community". Viewed the page 11 image: the entry clearly reads "Putnam, Robert C. 2000. Bowling Alone: The Collapse and Renewal of American Community", so this is not an OCR artefact. The in-text citations give only "Putnam (2000)" and so do not settle the initial. Cosmetic bibliographic slip.
Second check: none (judge only)
Report a problem with this finding
Typo Reference list prints "Sayre, Wallace, and Herbert Kaufman. 1960. Governing New York City: Politics in the Metropolis. Philadelphia: Russell Sage."; the Russell Sage Foundation is a New York publisher.
- p. 11 References, Sayre and Kaufman entry
Philadelphia: Russell Sage.
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: the Russell Sage Foundation is based in New York and the 1960 edition of Governing New York City was published there, not in Philadelphia. Viewed the page 11 image: the entry reads "Philadelphia: Russell Sage." as quoted, not an OCR artefact. The place of publication cannot be verified from the files; a cosmetic bibliographic point for the authors to check.
Second check: none (judge only)