How Large and Long-lasting Are the Persuasive Effects of Televised Campaign Ads? Results from a Randomized Field Experiment
American Political Science Review, 2011 · paper · Political science
Reports a randomized field experiment using about $2 million of television and radio ads in a 2006 governor's race, with daily tracking polls. Ads had strong effects on voting preferences, but the effects faded within a week or two.
Checked: PDF from gwern.net, fetched 5 Oct 2026. Page numbers are those of that PDF.
6 findings
Substance Footnote 9 gives Baselice as "Perry ahead with 41%... Strayhorn with 24%" (a 17-point lead) but says "Perry has a 17–percentage point lead over Strayhorn in our poll and a 19-point lead in the Baselice poll". The two leads appear swapped, or a percentage is misprinted.
- p. 7 footnote 9
put Perry ahead with 41%, the Democratic candidate with 14%, Strayhorn with 24%
- p. 7 footnote 9
point lead over Strayhorn in our poll and a 19-point lead in the Baselice poll
- p. 8 Table 3 panel (a), January 12-18 column
32.8%
- p. 7 footnote 9
ahead with 41%, the Democratic candidate with 14%, Strayhorn with 24%
Only the authors can say what is right here.
Judge's note: The p. 7 footnote and the p. 8 Table 3 images were checked. 41-24=17, which contradicts the 19 stated for Baselice in the same sentence. The authors' unweighted Table 3 lead is 18.7 for Jan 12-18 (close to 19). Their weighted figure is not shown, so only the authors can say which number is wrong.
Second check: another model agreed
Report a problem with this finding
Reference or label "For 0 > ρ > 1, media effects decay geometrically in time": no ρ satisfies this range; it should be 0 < ρ < 1.
- p. 9 right column, paragraph after Equation (2)
For 0 > ρ > 1, media effects decay geometrically in time
- p. 9 text after Equation (2)
For 0 > ρ >
Judge's note: The C012-a crop shows both inequality signs printed as '>'. The example that follows (ρ = 0.5) uses 0 < ρ < 1. The meaning is clear from context, but the printed condition is impossible.
Second check: another model agreed
Report a problem with this finding
Reference or label "In column (1), for example, ... raises Perry's vote share by 4.73 percentage points per 1,000 GRPs (SE = 1.42); a week later ... −0.17 (SE = 1.42)": these are the column (2) values of Table 5. Column (1) shows 4.72 (1.41) and has no lag term.
- p. 10 Results, right column, last lines
In column (1), for example, we see that the current week’s advertising raises Perry’s vote share by 4.73 percentage points per 1,000 GRPs (SE = 1.42)
- p. 12 Table 5, row TV (No Lag), columns (1) and (2)
4.72 4.73
- p. 12 text below Figure 1
have receded to −0.17 percentage points (SE = 1.42)
- p. 12 text below Figure 1
later, the effects of these ads have receded to −0.17 percentage points (SE = 1.42)
Judge's note: p. 10 text and the p. 12 Table 5 image: col (1) 4.72 (1.41), no 1-week lag; col (2) 4.73 (1.42) with lag −0.17 (1.42). The Figure 1 footnote also cites column (2). Wrong column pointer.
Second check: another model agreed
Report a problem with this finding
Reference or label Huber and Arceneaux (2007) is given as "American Journal of Political Science 24: 957–77"; for a 2007 AJPS article the volume would be 51.
- p. 16 References, Huber and Arceneaux 2007
Political Science 24: 957–77.
- p. 16 References, Huber and Arceneaux 2007
Political Science 24: 957–77
- p. 16 References, Panagopoulos and Green 2008
American Journal of Political Science 52 (1): 156–
- p. 16 Reference list
American Journal of Political Science 24: 957–77.
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: AJPS volume 51 is 2007 and volume 24 is 1980. Internal evidence on p. 16 points the same way: the other AJPS entries follow volume = year - 1956 (1998→42, 2000→44, 2008→52; Brader 2005→49 on p. 15). The in-text citation on p. 2 also misspells the name as 'Arcenaux' (not part of this cluster).
Second check: another model agreed
Report a problem with this finding
Reference or label Hovland, Lumsdaine and Sheffield (1949), Experiments in Mass Communication, is listed with the publisher "Ann Arbor, MI: Association for Consumer Research".
- p. 16 References, Hovland et al. 1949
Ann Arbor, MI: Association for Consumer Research.
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: the book was published by Princeton University Press, and the Association for Consumer Research was founded in 1969. The p. 16 entry reads as quoted.
Second check: another model agreed
Report a problem with this finding
Wording The Discussion says "the volume, timing, and location of TV and radio ads were determined randomly", but the design section says "The GRP weight of the radio ads was not varied randomly" and that stations and programs were not randomized.
- p. 14 Discussion, first paragraph (continued from p. 13)
the volume, timing, and location of TV and radio ads were determined randomly
- p. 5 Experimental Design
The GRP weight of the radio ads was not varied randomly.
- p. 5 Experimental Design
The experiment did not randomize the stations and programs on which the ads were placed.
Judge's note: Page 5: radio was randomized only in timing, and the radio purchaser had discretion over placement. Houston and Dallas–Fort Worth were excluded from randomization. The p. 14 sentence overstates the design for radio volume and location.
Second check: another model agreed
Report a problem with this finding
20 minor (typos and the like; most checked by the judge only)
Reference or label Ashworth and Clinton is dated "2006" in QJPS volume 2 (pp. 27–41), while Vavreck "2007" is also in QJPS volume 2; the year of one entry is probably wrong.
- p. 15 References, Ashworth and Clinton
Ashworth, Scott, and Joshua D. Clinton. 2006.
- p. 16 References, Vavreck 2007
ical Science 2: 325–43.
- p. 2 right column, third paragraph
(Ashworth and Clinton 2006;
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: the Quarterly Journal of Political Science began with volume 1 in 2006, so volume 2 is 2007. Inside the paper, one volume carries two different years, and the in-text citation on p. 2 also says 2006. The reference list was read on pp. 15-16.
Second check: another model agreed
Report a problem with this finding
Wording "Perry had from 1985 to 1991 served two terms as the Commissioner of Agriculture and in the Texas House of Representatives" puts both offices inside 1985-1991.
- p. 3 Description of the Experimental Setting, first paragraph
from 1985 to 1991 served two terms as the Commis-
- p. 3 Description of experimental setting
Perry had from 1985 to 1991 served two terms as the Commis- sioner of Agriculture and in the Texas House of Repre- sentatives.
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: Perry was in the Texas House 1985-1991 and Commissioner of Agriculture 1991-1999. The p. 3 image matches the quote. As written, the dates seem to cover both offices; the sentence is ambiguous at best.
Second check: another model agreed
Report a problem with this finding
Wording The maximum TV dose (1,000 GRPs) is said to have "boosted Perry's relative standing by approximately six percentage points", but the estimate just quoted is 5.27 per 1,000 GRPs (Table 4 col. 1), and p. 13 uses "5 percentage points".
- p. 10 Results, left column
boosted Perry’s relative standing by approximately six percentage points
- p. 10 Results, left column
tising are 5.27 (SE = 1.50) and 4.25 (SE = 5.91)
- p. 13 right column, cost paragraph
To shift opinion 5 percentage points
- p. 13 last paragraph before Discussion
To shift opinion 5 percentage points among
Only the authors can say what is right here.
Judge's note: Page 10: the sentence follows the 5.27 (SE = 1.50) estimate per 1,000 GRPs, with a maximum of 1,000 weekly GRPs, which gives about 5.3 points. The fixed-effects estimates in Table 4 run from 5.02 to 5.44. Only the 2-DMA stratum estimate in footnote 15 (6.21) or the Table 5 PDL estimates exceed 6, and the text does not point to either. 'Relative standing' is undefined, so this goes to the authors as a query.
Second check: another model agreed
Report a problem with this finding
Wording "Table 5 shows how the estimated effects of TV and radio change when dynamic parameters are added", but Table 5 reports television only (title: "Dynamic Effects of Television Advertising...").
- p. 10 right column, paragraph introducing Table 5
Table 5 shows how the estimated effects of TV and radio change when dynamic parameters are added
- p. 12 Table 5 title
Dynamic Effects of Television Advertising on Voter Preference, Measured Weekly
Judge's note: Table 5 on p. 12 has no radio rows, and its note says radio was omitted. Footnote 16, attached to the same paragraph, says radio was excluded, so a reader is told within a few lines. The opening sentence still describes radio estimates that are not in the table.
Second check: another model agreed
Report a problem with this finding
Typo DMA is expanded as "designated market areas, hereafter “DMAs”" on p. 2 and as "the DMA, or designated media market" on p. 5.
- p. 2 left column, second paragraph
(designated market areas, hereafter “DMAs”)
- p. 5 Experimental Design, first paragraph
tion is the DMA, or designated media market
Judge's note: Both crops checked. The two expansions differ. The second reads as a gloss, so this is a minor consistency point.
Second check: none (judge only)
Report a problem with this finding
Typo In-text citation "Huber and Arcenaux 2007" (p. 2) misspells Arceneaux, which is the spelling in the reference list ("Huber, Gregory, and Kevin Arceneaux. 2007.") and on p. 3.
- p. 2 right column, third paragraph
and Arcenaux 2007; Krasno and Green 2008)
- p. 2 Previous Literature, paragraph on natural experiments
and Arcenaux 2007; Krasno and Green 2008
- p. 16 References
Huber, Gregory, and Kevin Arceneaux. 2007.
- p. 2 Previous literature, natural experiments sentence
Huber and Arcenaux 2007
Judge's note: p. 2 crop shows 'Arcenaux'. The reference list on p. 16 and the citation on p. 3 have 'Arceneaux'.
Second check: none (judge only)
Report a problem with this finding
Typo The senator's surname is printed "Hutchinson" ("U.S. Senator Kay Bailey Hutchinson", "Hutchinson decided not to challenge Perry", "opponent than Hutchinson"). The correct spelling is Hutchison.
- p. 3 right column, last paragraph
U.S. Senator Kay Bailey Hutchinson
- p. 3 right column, last paragraph
Hutchinson decided not to challenge Perry
- p. 4 left column, first paragraph
opponent than Hutchinson
- p. 4 next paragraph
a less formidable opponent than Hutchinson
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: the senator's name is Kay Bailey Hutchison. pp. 3-4 confirm the spelling 'Hutchinson' is used in all three places.
Second check: none (judge only)
Report a problem with this finding
Typo "kept the Perry camp on edge though the end of 2005": 'though' should be 'through'.
- p. 4 left column, first lines
though the end of 2005
- p. 4 first lines, left column
kept the Perry camp on edge though the end of 2005
Judge's note: The p. 4 image, first line of the left column, shows 'though'.
Second check: none (judge only)
Report a problem with this finding
Typo "the graph makes clear that the Strayhorn ads bear little relationship..." refers to the right panel of Table 2, which is a table of GRP figures, not a graph.
- p. 5 right column, paragraph beginning 'As we noted earlier'
graph makes clear that the Strayhorn ads bear little
Judge's note: p. 5 text and the p. 7 Table 2 image: Table 2 is a numeric table and there is no graph of the schedules. Loose wording.
Second check: none (judge only)
Report a problem with this finding
Typo "the two campaigns competing communications" lacks the possessive apostrophe (campaigns').
- p. 5 right column, end of the paragraph on Strayhorn ads
entiating the effects of the two campaigns competing
Judge's note: The p. 5 image, end of the paragraph in the right column, shows no apostrophe.
Second check: none (judge only)
Report a problem with this finding
Typo Table 1, 1,000 GRPs column: "60 Minutes6-7p" has no space before the time slot.
- p. 6 Table 1, 1,000 GRPs column
60 Minutes6-7p
Judge's note: The p. 6 image shows the words run together; the entry above reads 'Dateline 6-7p'.
Second check: none (judge only)
Report a problem with this finding
Typo Table 1, 1,000 GRPs column lists "Sunday Today" twice ("Sunday Today / Sunday AM News / Sunday Today / Meet the Press").
- p. 6 Table 1, 1,000 GRPs column
Sunday Today
Only the authors can say what is right here.
Judge's note: The p. 6 image confirms the repeat. Two airings of the same programme are possible, but nothing distinguishes them; possibly a duplicated entry.
Second check: none (judge only)
Report a problem with this finding
Typo "with larger values of ρ imply greater persistence" is ungrammatical; it should read 'implying' or 'larger values of ρ imply'.
- p. 9 right column, paragraph after Equation (2)
values of ρ imply greater persistence
Judge's note: Checked in the C012-a crop and on p. 9.
Second check: none (judge only)
Report a problem with this finding
Typo Footnote 14 calls β0, β1, ..., βK "the K lag coefficients" (there are K+1 of them), and switches from the defined index k to an undefined j in "equals zero at j = K + 1".
- p. 9 footnote 14
the K lag coefficients β0, β1, β2, . . . , βK
- p. 9 footnote 14
polynomial equals zero at j = K + 1
- p. 9 footnote 14
this polynomial equals zero at j = K + 1
- p. 9 Footnote 14
equals zero at j = K + 1
Judge's note: C014-b crop checked. The index is defined as k = 0,1,...K, and j is never defined. The meaning can be inferred. A notation slip.
Second check: none (judge only)
Report a problem with this finding
Typo Table 4 note: "in a given DMA and in given week" is missing an article ("in a given week").
- p. 11 Table 4 note
in a given DMA and in given week
Judge's note: Read in the upright image of the Table 4 note on p. 11.
Second check: none (judge only)
Report a problem with this finding
Typo The text says "Figure 1 depicts the regression estimates in columns (1) and (2)" of Table 5; the figure footnote says "Specification is the same as column (2) of Table 5".
- p. 12 text below Figure 1
Figure 1 depicts the regression estimates in columns (1) and (2)
- p. 12 Figure 1 footnote
Specification is the same as column (2) of Table 5
Judge's note: Page 12: the right panel (lagged GRPs) can only come from column (2). The left panel's slope is almost the same under either column (4.72 vs 4.73). This is a small labelling inconsistency with no effect on the data.
Second check: none (judge only)
Report a problem with this finding
Typo "Columns (1) and (2) interactions with a 1-week lag, and columns (3) and (4) interactions with both 1-week and 2-week lags." has no verb (e.g. "include").
- p. 13 right column, first paragraph
Columns (1) and (2) interactions with a 1-week lag, and columns (3) and (4) interactions with both 1-week and 2-week lags.
Judge's note: Read on p. 13, right column.
Second check: none (judge only)
Report a problem with this finding
Typo The Ashworth and Clinton reference reads "Quarterly Journal of Political 2: 27–41"; the word "Science" is missing from the journal title.
- p. 15 References, Ashworth and Clinton 2006
Quarterly Journal of Political 2: 27–
- p. 16 References, Vavreck 2007
Quarterly Journal of Polit- ical Science 2: 325–43.
Judge's note: The crop of the p. 15 reference list shows 'Political' followed directly by the volume, so this is not a line-break artefact. The Vavreck 2007 entry on p. 16 gives the full title.
Second check: none (judge only)
Report a problem with this finding
Typo There is a stray period after the journal title in the Haugtvedt et al. entry: "Journal of Consumer Research. 21: 176–89."
- p. 16 References
Journal of Consumer Research. 21: 176–89.
Judge's note: Page 16: no other journal entry has punctuation between the journal and the volume.
Second check: none (judge only)
Report a problem with this finding
Typo Missing punctuation: "University of Michigan Press 17–78." (Zaller 1996) and "Electoral Advantage” Working Paper" (Hill et al. 2010).
- p. 16 References, Zaller 1996
Ann Arbor: University of Michigan Press 17–78.
- p. 16 References, Hill et al. 2010
Campaign Advertising and Electoral Advantage” Working Paper
Judge's note: Both are visible on p. 16. Neither entry has the separator found elsewhere in the list.
Second check: none (judge only)