Title: Does Media Concentration Lead to Biased Coverage? Evidence from Movie Reviews
1Does Media Concentration Lead to Biased Coverage?
Evidence from Movie Reviews
- Stefano DellaVigna, UC Berkeley and NBER
- Alec Kennedy, San Francisco Fed
- Moscow Media Conference
- October 29, 2011
- Preliminary, comments most welcome
2Introduction
- Dec. 13, 2007 News Corp. acquires from Bancroft
family Dow Jones Company ? Wall Street Journal - Unlike Bancroft family, Murdoch's holdings
include - Cable channels (i.e., Fox Sports and Fox News)
- Satellite television (Sky group)
- Movie distributor (20th Century Fox)
-
- Media conglomerate implies conflict of interest ?
Coverage of such businesses (or their
competitors) in the Wall Street Journal may be
biased.
3Introduction
- Wall Street Journal situation hardly unique
- Media conglomerates
- Comcast owns NBC
- Hearst Corporation owns ESPN and print outlets
- Time Warner owns AOL and Time magazine
- ? Pervasive conflict of interest
- But should media consolidation lead to
distortion? - Cost Loss of reputation if bias is revealed
- Benefit Can persuade audience (if audience naïve
about bias) - Generally, persuasion rates of 5-10 from the
media (DellaVigna-Gentzkow, ARE 2010)
4Introduction
- This paper Focus on two conglomerates
- News Corp.
- Time-Warner
- Measure how media outlets in these groups review
movies distributed by an affiliate - 20th Century Fox
- Warner Bros.
- Identification of bias transparent
(diff-in-diff) - Compare review of 20th Century Fox movie by WSJ
and NYT - Further control group review of Paramount movie
by WSJ and NYT
5Introduction
- Why movie reviews?
- Frequent (500,000 reviews in data set)
- Easily quantifiable (coded on 0-100 scale by
metacritic.com) - Large industry (60bn annual revenue)
- Some evidence that movie reviews influence movie
attendance (Reinstein and Snyder (2005)? benefits
to the distributor from increased ticket sales
6Data
- First data set of reviews from metacritic.com,
scored 0-100
7Data
- Second data set of reviews from
rottentomatoes.com, scored on a 0-1 freshness
scale
8Data
- Merged data set contains 548,764 reviews from 336
media from 1985 to July 2011
9Data
- Studios which distribute movies in the sample
Major, Independent, and Other
10Average Bias News Corp.
11Average Bias Time Warner
12Average Bias Statistical Test
- Is the bias for News Corp. statistically
significant? - Is it robust to introducing controls?
- OLS regression
- captures the effect of conflict of
interest in Fox - captures the effect of conflict of
interest in Time Warner - Sample of 473,727 reviews (qualitative reviews in
RT have no 0-100 score) - Standard errors clustered at the movie level
13Average Bias Statistical Test
14Average Bias Statistical Test
15Average Bias Statistical Test
- In favorite specification, conflict of interest
increases rating for News Corp. by 2.6 points out
of 100 - Estimate of bias increases with extra controls ?
Unobservables unlikely to bias coefficient upward
(Altonji, Elder, and Taber, 2005) - Magnitude of bias
- Equivalent to 1 extra star every ten reviews
- Small but still economically significant impact
- Reinstein and Snyder (2005) estimate 25 higher
revenue for two thumbs up by Roger Ebert - Are all media conglomerates the same?
- No evidence of bias for Time Warner, can reject
bias of 0.9 points
16Average Bias Robustness
- Evidence of bias using RT 0-1 freshness indicator
17Calculation of benefits
- Back-of-the-envelope calculation of estimated
benefits from distortion of movie review - Case of New York Post (NYP)
- Suppose that NYP gives one extra star per Fox
movie - 500,000 average readers
- Persuades an extra 1 of readers to watch movie
- Ticket sales increase by 5,000, or
85,00040,000 - Studio receives about half of increased sales,
plus another half from higher rights from DVD
licensing ? 40,000 - About 400 20th Century Fox movies since 1985
- ? Potential benefits to NewsCorp. from one extra
star per Fox review by NYP over 25 years
16,ooo,ooo
18Average Bias Explanations
- Three main explanations for the results
- (E) Explicit editorial policy conveyed to
journalists - (J) Bias by a journalist ultimately due to the
conflict of interest, but lacking editorial
pressure - (T) Correlation in taste between the media
reviewer (or the media audience) and the
affiliated studio - To separate explanations, we present evidence
on - Clustering of bias within a conglomerate (E and
T, maybe J) - Editorial policies (E only)
- Selective bias by type of movie (E or J, not T)
- Omission of reviews (E or J)
19Clustering of Bias By Media
- News Corp. media Statistical evidence for NYP,
2-3 point bias for all 6 media - No evidence of bias for any of the Time Warner
media
20Clustering of Bias By Journalist
- Most media have a small number of journalists
reviewing movies - Chicago Sun-Times, News of the World, Wall Street
Journal and CNN.com have essentially only one
21Clustering of Bias By Journalist
- News Corp Statistical evidence of bias for 3 out
of 4 NYP journalists, and one of TV Guide
journalists - Time Warner No evidence of positive bias
- Significant Clustering Pattern suggestive of
editorial bias, but could also be correlated
tastes, or similar journalists
22Editorial Policies
- Two tests of editorial policies
- Personnel Policy Change of reviewers at change
of ownership No evidence - Bias in Assignment Assign affiliated movies to
more generous reviewers - Estimate average reviewer generosity in score
- Reviewers differ significantly in their
generosity - Are movies by affiliate studios more likely to be
assigned to more generous reviewer? - No evidence assignment quasi-random
- Bias at Newscorp. is unlikely of editorial origin
- Could be correlation in journalist bias or in
tastes
23Editorial Policies
24Selective Bias
- If bias is due to conflict of interest, should be
optimal - Bias when marginal return (persuasion) is highest
- Maximize impact on revenue
- Proposition 2. Small or no bias for very low
quality movies - If bias is due to correlated tastes, no such
prediction - Assumption for this test
- Movies with negative reviews by others are
unlikely to benefit from a lone positive review - Movies with other positive reviews more likely to
benefit from more positive review
25Selective Bias News Corp.
26Selective Bias News Corp.
27Selective Bias New York Post
28Selective Bias Time Warner
29Selective Bias Statistical Evidence
- Evidence of selective bias for NYPost and
qualitatively for WSJ Bias due to conflict of
interest - No evidence for other media
30Bias by Omission
- If bias is due to conflict of interest, and
audience not fully rational, bias by omission - Review high-quality affiliated movies
- Omit review of low-quality affiliated movies
- Do not expect this pattern if correlated tastes
- Rare setting to separate bias by omission and by
commission - Outlets differ substantially in probability of
review ? Use matching procedure - For each media, find ten matching media in terms
of average probability of review of a movie - Plot local polynomial regression of dummy for
review on average review score
31Bias by Omission New York Post
32Bias by Omission Time
33Bias by Omission Entertainment Weekly
34Bias by Omission Statistical Test
- Newscorp. No systematic evidence of omission
bias - Time Warner Evidence for 2 outlets
- Bias by omission and by commission substitutes,
not complement
35Bias in Movie Aggregator
- So far focus on most obvious conflict, for
reviewers - Conflict of interest hardly stops there
Rottentomatoes.com, also independent when
launched in 1998, was acquired by IGN
Entertainment in June 2004, and IGN was purchased
by News Corp. in September 2005. IGN, and hence
RottenTomatoes, was then sold in January of 2010
by Newscorp. - Conflict of interest for RT more positive
reviews of the 20th Century Fox movies in
2006-2009 - Control for rating of same review in MC
36Bias in Movie Aggregator
- Remarkably, no evidence of bias, even for
qualitative reviews, where bias is easier to hide - Can reject small bias
37Bias in Movie Aggregator
- Local polynomial regression of fresh indicator
on average movie score no evidence of bias
38Bias in Movie Aggregator
- Event study comparing residual freshness (after
controlling for score) for FOX and non-FOX movies
39Summary of Results
- We documented the extent of bias due to conflict
of interest for two media conglomerates - Average bias
- 2.6 points bias out of 100 for Newscorp. outlets
- No bias for Time Warner outlets, can reject even
small bias - Bias is clustered within a media conglomerate
- No evidence of editorial policy to assign movies
- Selective Bias Evidence for New York Post, not
for other media - Omission Bias Evidence for two Time Warner
outlets - Interpretation
- Best fits with bias due to conflict of interest
for journalists, with clustering of such bias
40Conclusion
- Overall, remarkably little evidence of distortion
from conflict of interest - No distortion in review for Time Warner
- No distortion for Rottentomatoes
- No distortion in editorial assignment
- However, bias still does occur
- Small, but significant, bias for Newscorp.
Outlets - Some omission bias for Time Warner outlets
- Suggests that transparency and emphasis on
reputation (for example because of competition)
critical to keep media honest - Paper allowed us to decompose potential media
bias in novel ways - Relates to Conflict of interest in other
settings - Significantly less distortion than for analysts
- Less distortion than for advertising
41(No Transcript)
42Average Bias
43Average Bias
44Selective Bias II
- Different types of movies can have different
returns to positive reviews - Snyder and Reinstein Larger effect of movie
reviews for independent movies - However Independent movies also have smaller
revenue, so bias may be less worthwhile - Examine the effect of conflict of interest
separately
45Selective Bias Indy movies
- Significant bias for News Corp. only for
mainstream movies