The Fox News Effect: Media Bias and Voting

• p. 2 Introduction

  available in about 35 percent of households, the impact of Fox News is estimated to be 0.15 to 0.2 percentage points

• p. 2 Introduction, paragraph 4

  News is estimated to be 0.15 to 0.2 percentage points, 200,000

• p. 25 Section III.B, Magnitudes

  0.34 (0.0055) 0.0019

Only the authors can say what is right here.

Judge's note: p. 2 gives the effect as 0.4 to 0.7 points and the share as about 35 percent. On p. 20 the benchmark estimates are 0.0042 and 0.0069. Section III.B (p. 25) computes only the midpoint, 0.34*0.0055=0.0019. Scaling gives a lower end of about 0.15, which matches, but the upper end comes to about 0.24 (0.23 using 34%), not 0.2. The authors may have rounded loosely, so this is put to them as a query.

Second check: another model agreed

• p. 2 Introduction

  an audience of 17.3 percent of the U. S. population by June 2000

• p. 7 Section II, Fox News Audience

  105,201 respondents to the August 2000–March 2001 survey. The recall audience for Fox News is 17.3 percent

Only the authors can say what is right here.

Judge's note: p. 2 dates the figure to June 2000. p. 7 and Table I (column 1, note: 'August 2000–March 2001') show 17.3% (0.173) as the recall audience from the survey run August 2000–March 2001. No other Scarborough wave dated June 2000 is mentioned. The only 'June 2000' elsewhere is footnote 2, about satellite subscribers. The authors should say which date applies.

Second check: another model agreed

• p. 2 Introduction, paragraph 3

  the availability of Fox News is uncorrelated with town-level demographic controls

• p. 18 Table III, column (6), F-test row

  F = 1.57**

Judge's note: The Table III image shows F = 1.11 in col (4), 1.28 in col (5), 1.57** in col (6) and 1.31 in col (7). The text on pp. 18–19 discusses columns (6)–(7) only for the 1988–1992 trend and concludes that towns 'are no different in the demographics'. So the claim holds for the benchmark columns (4)–(5) on the full sample, but the text does not mention the significant result in col (6). This is an unqualified statement, not a wrong number.

Second check: another model agreed

• p. 9 Table I, No. observations row

  N = 105,201

• p. 7 Text, paragraph on Table I

  105,201 respondents to the August 2000–March 2001 survey

• p. 2 Introduction, paragraph 2

  audience of 17.3 percent of the U. S. population by June 2000

• p. 7 Section II, Fox News Audience

  105,201 respondents to the August 2000–March 2001 survey. The recall audience for Fox News is 17.3 percent

Only the authors can say what is right here.

Judge's note: The Table I (continued) image confirms the N row. The notes describe columns (2) and (3) as the regular and nonregular audience of the whole sample, and the text gives no reason for missing recall answers. The matched columns do sum (403+10,985=11,388). The gap of 10 is small and may be missing responses, so this is a query to the authors. The June 2000 dating point that some reviewers raised here is handled under C003.

Second check: another model agreed

• p. 6 Section II

  the towns reached by Fox News in 2000 were more than twice as large as the remaining towns

• p. 14 Summary Statistics

  are 25 percent larger

• p. 12 Table II, Population, columns (2)–(3)

  11,516

Only the authors can say what is right here.

Judge's note: The Table II image (cols 2–3) gives a ratio of 1.26, which matches '25 percent larger'. The p. 6 claim is used to explain why 20% of towns hold 34.3% of the population, and that would need a ratio of about 2.1. The two statements may use different town sets or weighting, but the paper does not say so. The authors should clarify.

Second check: another model agreed

• p. 36 Section IV.A, Robustness

  (t = 0.583 and d = 0.583 * (1 − 0.538))

• p. 12 Table II, Log turnout in 1996, column (1)

  −0.583

• p. 36 Robustness

  t 0.583

Only the authors can say what is right here.

Judge's note: Table II (unweighted; its note says 'Observations unweighted') shows all-towns log turnout −0.583 in 1996 and −0.522 in 2000, and Rep. share 0.538 in 2000. By Jensen's inequality, mean turnout ≥ exp(mean log turnout) = exp(−0.522) ≈ 0.593 > 0.583, if the 2000 sample and definition are the same. The value 0.583 matches the magnitude of the 1996 mean log. A different sample or turnout definition could explain it, and the paper does not say, so this is a query for the authors.

Second check: another model agreed

• p. 13 Table II notes

  Potential subscribers is defined as the total voting-age population of the towns reached by a cable provider.

• p. 45 Appendix II notes

  The number of potential subscribers is the total population covered by a cable system.

Judge's note: Both note images were read (pp. 13 and 45). They give two different definitions of the same control variable (voting-age population vs total population), and the paper does not say which one was used.

Second check: another model agreed

• p. 29 Table VII, column (6), Republican vote share in 1996 presidential elections

  (0.1111)***

• p. 29 Table VII, Republican vote share in 1996 presidential elections, column (6)

  0.8289 (0.1111)***

Only the authors can say what is right here.

Judge's note: The upright Table VII image shows 0.1111. Col. (6) is the col. (4) specification plus two interactions. Its other SEs are close to col. (4) (Fox 0.0035 vs 0.0026; NY race 0.0067 vs 0.0067), so a tenfold SE on an unchanged coefficient looks like a digit slip. The stars hold either way. Only the authors can confirm.

Second check: another model agreed

• p. 29 Table VII, column (3)

  −0.0207 (0.0087)*

Only the authors can say what is right here.

Judge's note: The upright Table VII image confirms both cells. Both coefficients come from the same regression, with the same clustered SEs and degrees of freedom, so the larger ratio should not get fewer stars. The note defines * as 10 percent and ** as 5 percent. The authors should check which star is right.

Second check: another model agreed

• p. 36 Section IV.A, first lines

  0.560 * (1 − 0.453) = 0.306

• p. 38 Table IX, Fox News Study rows, column (7)

  0.556

Only the authors can say what is right here.

Judge's note: I looked at p.36 (the 0.560*(1−0.453)=0.306 line), p.35 (t = 0.560), p.38 (Table IX Fox rows: 0.556, 0.560, 0.563) and Table II on p.12, where the unweighted 2000 Republican share is 0.538, matching the robustness text (1−0.538). The value 0.556 appears nowhere else in the text or tables as a vote share. The Fox News treatment values equal 0.556 plus the vote effects 0.0042/0.0069, so 0.556 is being used as the baseline. Since neither 0.453 nor 0.538 equals 0.556, only the authors can say what 0.556 is. The text also says 'Democratic two-party vote share' while writing (1 − 0.453), which adds to the ambiguity.

Second check: another model agreed

• p. 39 Table IX notes

  Columns (7) and (8) report the share of Republican voters in the control and treatment group.

Judge's note: I checked the Table IX rows on pp.38–39. Column (3) states the variable for each row, so readers can recover the meaning, but the note's general description is wrong for most rows.

Second check: another model agreed

• p. 39 Table IX, Gentzkow and Shapiro Al Jazeera row

  0.215 0.133 1.000 0.105

Judge's note: I confirmed the values on the p.39 image. The CNN (0.083) and Kull et al. (0.141) rows reproduce exactly with the formula. The Al Jazeera row matches only as an absolute value: 0.082/0.785 = 0.1045, printed 0.105. The text on p.40 ('persuasion rates between 0.08 and 0.10') treats the figures as magnitudes, so the sign convention is implied rather than wrong, but neither the table nor the note states it.

Second check: another model agreed

• p. 1 Introduction

  De Marzo, Vayanos, and Zwiebel

• p. 4 Introduction

  DeMarzo, Vayanos, and Zwiebel 2003

• p. 1 Introduction, first paragraph

  De Marzo, Vayanos, and Zwiebel 2003

• p. 47 References

  DeMarzo, Peter, Dimitri Vayanos

Judge's note: The page-1 image clearly shows 'De Marzo' with a space. The page-4 image shows 'DeMarzo, Vayanos, and Zwiebel 2003', and the reference list on p. 47 has 'DeMarzo, Peter, Dimitri Vayanos, and Jeffrey Zwiebel'. This is a cosmetic inconsistency.

Second check: none (judge only)

• p. 4 Introduction

  Kull, Ramsey, and Lewis 2003

• p. 48 References

  Kull, Steven, Clay Ramsay, and Evan Lewis

• p. 4 Introduction

  Kull, Ramsey, and

• p. 48 Reference list

  Kull, Steven, Clay Ramsay

Judge's note: The page-4 image shows 'Ramsey' and the reference entry on p. 48 shows 'Ramsay'. This is a cosmetic misspelling.

Second check: none (judge only)

• p. 5 Section II, Fox News History and Content

  ABC, CBS, and NCB—and CNN

• p. 5 Fox News History and Content

  ABC, CBS, and NCB

• p. 5

  the major networks—ABC, CBS, and NCB—and CNN

Judge's note: The page-5 image shows 'NCB'. A few lines later the same paragraph has '(ABC, CBS, CNN, and NBC)'.

Second check: none (judge only)

• p. 7 Section II, Fox News Audience

  sorting to Republicans into the Fox News audience

Judge's note: The page-7 image shows 'sorting to Republicans into'. This is a grammatical slip.

Second check: none (judge only)

• p. 17 Section III.A, text after equation (1)

  Xk,2000–1990, the set of changes in controls

• p. 15 Equation (1)

  Γ00–90Xk,00–90

Judge's note: I checked the page 15 image (eq. 1: Γ00–90 X_{k,00–90}) and the page 17 crop (text: X_{k,2000–1990}). Both clearly mean the same vector; the short form is used again on p. 1206. This is a cosmetic notation inconsistency only.

Second check: none (judge only)

• p. 20 Section III.B

  The observation are weighted by the votes cast in 1996

Judge's note: The p. 20 crop shows the text as quoted.

Second check: none (judge only)

• p. 23 Table V notes

  The observations are weighted total votes cast in the 1996 presidential election

Judge's note: The upright Table V image shows the note as quoted. The other table notes (e.g., Tables III and VII) say 'weighted by total votes cast'.

Second check: none (judge only)

• p. 23 Table V, row label and notes

  Unweighted, turnout > 2000

• p. 24 Section III.B, Robustness

  at least 2,000 votes

Judge's note: I checked the Table V image and the p. 24 crop. A strict versus non-strict threshold (and 'people' versus 'votes') differs only at the boundary. It is a trivial consistency point.

Second check: none (judge only)

• p. 26 Section III.C

  in rural towns and in Republican Districts

Judge's note: The page 26 image shows 'Republican Districts' capitalised one line after the lower-case 'Republican districts'. This is a cosmetic style point.

Second check: none (judge only)

• p. 34 Table VIII notes

  is computed at the ratio

Judge's note: The upright Table VIII note image shows both wordings.

Second check: none (judge only)

• p. 36 Section IV.A, paragraph on diary measure

  are 3.42 times larger if eF − eN is measured

• p. 35 Section IV.A, Persuasion Rates

  35.3/10.3 = 3.43

• p. 35 Persuasion Rates

  35.3/10.3 3.43

• p. 36 Persuasion Rates

  3.42 times larger

Judge's note: The page images confirm both quotations. A search of the text finds eF − eN only on p.36. 35.3/10.3 = 3.427, so 3.43 and 3.42 are two roundings of the same ratio; the two estimate pairs give 3.43 and 3.42. The inconsistent rounding is trivial. The undefined symbol is a real, minor notation slip; the evident fix is eT − eC.

Second check: none (judge only)

• p. 36 Section IV.A, Persuasion Rates

  if eF − eN is measured using the diary data

• p. 36 Persuasion Rates paragraph on the diary measure

  measured using the diary data

• p. 36 same paragraph

  eF eN

Judge's note: The page 36 image shows eF − eN. A grep finds no other occurrence or definition. Expression (6), equation (7) text and p. 33 all use eT − eC.

Second check: none (judge only)

• p. 39 Table IX notes

  We obtain the baseline share of vectors

• p. 39 Table IX notes

  baseline share of vectors t{C}

• p. 39 Table IX notes

  baseline share of vectors

• p. 39

  We obtain the baseline share of vectors t{C} from Table B1.1

Judge's note: The p.39 upright image shows 'vectors'. From context the intended word is plainly 'voters'.

Second check: none (judge only)

• p. 40 Laboratory Experiments paragraph

  Ansolabehere and Iyengar (1995]

• p. 40

  Ansolabehere and Iyengar (1995] expose experimental subjects

• p. 40

  re and Iyengar (1995]

Judge's note: The zoomed crop of p.40 shows '(1995]' in the printed text, so this is not an extraction artefact.

Second check: code check

• p. 46 Appendix III, dependent-variable header

  Presidential republication vote share change between 2000 and 1996

• p. 46 Appendix III table header

  Presidential republication vote share change

Judge's note: The p.46 upright image shows 'republication'. The table notes say 'two-party Republican vote share', so the intended word is clear.

Second check: none (judge only)

• p. 47 References

  How Attack Ads Shrinks and Polarize the Electorate

Judge's note: The p.47 reference list shows 'Shrinks and Polarize'. The verb disagreement inside the title is evident without outside knowledge of the book's actual title.

Second check: none (judge only)