maria petrova new economic school esnie, 4 th june 2010
TRANSCRIPT
Maria PetrovaNew Economic School
ESNIE, 4th June 2010
OutlineTwo big questions:1) Is media important?
Any independent effect on people’s behavior? Different counterfactuals:
Different media… …or no media at all
If media is important, what it is important for?
2) If media is important, what drives media content?
Media effects
Traditional studiesPeople come to be interested in media effects
during and after WWIIBut: early studies did not find any effects
Self-selection to media consumption is the main problem
“Minimal effects” paradigmMedia reinforce existing beliefs and
predispositionsLazarsfeld, Berelson, and Gaudet 1948;
Berelson, B. R., P. F. Lazarsfeld, and W. N. McPhee 1954; Klapper 1960
Does media matter? ExamplesFox News and 2000 Bush-Gore electionsRussian public television and elections of
Unity and Vladimir Putin in 1999-2000German radio in 1930s and Nazi supportRTLM radio in Rwanda in 1994 and Tutsi
genocideTelevision and Mexican pivotal 2000
elections
MethodologyThe main problem is self-selection
People choose media which reflect their preferences and prior beliefs
As a result, effects are either too small, if a study controls for individual pre-existing preferences, or too large, if a study does not do it
Need some exogenous variation to identify the effect
Field experiments (e.g. Gerber, Karlan, and Bergan 2009, free 10-week subscription to Washington Post or Washington Times)
Methodology - 2Another potential solution: use geography…
Ground conductivity, proportion of woodland (e.g. Strömberg 2004)
Irregular Terrain Model (ITM) and mountains (e.g. Olken 2009)
ITM and idiosyncrasy of Soviet times resource allocation (e.g. Enikolopov et al. 2010)
Elevation (our work in progress in Croatia)
Methodology - 3… or other source of idiosyncratic variation
Cable industry variables (DellaVigna and Kaplan 2007)
Different overlap between media markets and Congressional districts (Strömberg and Snyder 2009)
Variation in coverage due to Olympic Games or other exogenous events (Einesee and Strömberg 2007)
Different distance to the nearest newspaper publishing information about school grants in Uganda (Reinikka and Svennson 2004)
QuestionsWhat are the outcomes?
CorruptionVoting behaviorSocial capitalViolenceConsumption, investment…
Is the presence of media important?i.e. compare situations with and without media
Is media content important?i.e. compare situations with and without a
particularly biased media outlet
Comparison of effectsPersuasion rates (DellaVigna Kaplan 2007, DellaVigna
Gentzkow forthcoming)How many people are convinced by a media outlet to
change their behavior?
Here i{T,C}, ei is the share of those receiving the message,
yi is the share of those who adopt behavior of interest,
y0 is share of those who would adopt it without the message
01
1*100
yee
yyf
CT
CT
Voting studiesEffect of Fox News on Bush vote share 11.6% (DellaVigna and Kaplan 2007)Effect of NTV, the national channel independent
from the government, onvote share of opposition parties in Russia: 7.7% vote share of pro-government party in Russia: 66%
(Enikolopov, Petrova, Zhuravskaya 2010)Effect of Washington Post free subscription on
Democratic vote share19.5% (Gerber, Karlan, Bergan 2009)Effect of unexpected Democratic endorsements
on Gore vote share 6.5% (Knight and Chiang 2009)
Turnout studiesEffects of different GOTV technologies (Gerber and Green
2000): 15.6% persuasion rate for door-to-door canvassing1.0% persuasion rate of 1-3 mailing cards
Effect of the introduction of television in 50s and 60s (Gentzkow 2006):4.4% persuasion rate of exposure to television
Effect of newspaper entry in 19th and 20th century U.S. counties (Gentzkow and Shapiro 2009)12.9% persuasion rate of access to some local newspaper
Effect of local news in Spanish on turnout of Spanish population7.6 % effect on individual turnout in midterm elections, 3.5%
effect in presidential elections
Mechanism of persuasionRational choice models, with media consumers
not fully discounting information from biased sources (e.g. Genzkow and Shapiro 2006; Petrova 2008; Gehlbach and Sonin 2009)
Behavioral models, with consumersunderestimating the biases in media content (e.g.,
Cain et al. 2005; Eyster and Rabin 2009), thinking categorically (Mullainathan 2001;
Mullainathan, Schwarzstein, and Shleifer 2008), or double counting repeated information(DeMarzo,
Vayanos, and Zwiebel 2003)More detailed survey in DellaVigna and
Gentzkow (forthcoming)
Outside United States…Convincing evidence of media effects is scarceEffect of West German television on anti-communism
feelings in Eastern Germany (Hainmueller and Kern 2009)
Effect of RTLM radio on killing of Tutsi in Rwanda (Yanagizawa 2010)
Some studies which do not use field or natural experiments
Lawson and McCann (2005): media effects in Mexican 2000 elections
Ladd and Lenz (2009): Effect of unexpected change in endorsement pattern of Sun and Daily Star on vote for Labour in Britain
Reinikka and Svensson (2004): Effect of newspaper coverage of funding arrangements on school enrollment in Uganda
OverallHot topic for empirical research; many gaps in
the literatureMedia effects in countries other than U.S.
Other democratic countriesAuthoritarian states and dictatorshipHistorical studies
Media effects in other circumstancesCivil wars International conflictsRegime changes
Media effects on other types of behaviorConsumption, investment Protest behaviorFirm’s strategy
Example: media effects in Russia in 1999
Motivationbased on Enikolopov, Petrova, Zhurvaskaya (2010)
Two storiesIn August 1999, Putin, with popularity rating
between 1 to 2%, was appointed as prime minister. Eight months later, he was elected president by getting 52.9% of the vote.
In December 1999 Parliamentary elections pro-government party, Unity, which did not even exist two months prior to the election, scored the second with 23.8% of the total vote
Mass media seem to play important roleThe goal of the paper is to establish causal
effect of media on voting behavior
Identification - 1Compute predicted availability of NTV, the only
national channel independent from the government in 1999Use data on location and power of transmittersUse ITM model (Hufford 2002, Olken 2008) which
takes into account geographic obstacles to signal propagation to predict signal strength
Use signal strength and survey data to predict NTV availability
Compare subregions with different availability of NTV with region fixed effects included
Use signal strength as instrument for NTV exposure in individual-level analysis
Identification - 2Look at the determinants of availability of NTV
transmittersSystem of transmitters was inherited from Soviet
educational channelTransmitters were more likely to be located in large
industrial towns, so 3 basic variables explain their location
Availability of transmitters not correlated with pre-existing political preferences after these 3 variables are taken into account
Placebo experimentIn 1995 NTV was not able to use this national
system of transmittersCheck if NTV had any effect on voting in 1995
ResultsEffects of predicted NTV availability
+6.3% on vote for opposition parties, positively covered by NTV
- 8.9% on vote for pro-government party, criticised by NTV
-3.8% on turnoutOn individual level, significant effect of NTV even
controlling for voting intentions 1 month before elections
Persuasion rates7.7% for positive message65.6% for negative message
Media bias
All the news that fits to printWhat drives media slant?Can we talk about media bias?Who are the actors?
Media consumersMedia outletsPotentially, some other group interested in
media content Government
Special interest groups Advertisers Journalists
Media biasDiscretion which media outlets have over
contentChoice of topicsChoice of expertsEditorialsEndorsementsReporting/non-reporting valuable information
(e.g. the state of the world)Assume that some unbiased point is defined…
Bias can be viewed as deviation from this unbiased point
Some measuresExperts cited by different members of
Congress (Groseclose and Milyo 2005, Gasper 2009)
Language used by different members of Congress (Gentzkow and Shapiro 2010)
Support of Supreme Court decisions by editorials (Ho and Quinn 2008)
Recommendations of mutual funds with and without advertising (Reuter and Zitzewitz 2006)
Empirical evidenceGentzkow and Shapiro (2010):
Use a measure of media bias based on phrases used by Democrats and Republicans in Congress, e.g. “war on terror” vs “war in Iraq” or “tax cut” vs “tax relief”
Evidence for demand-side effect, 20% media bias is explained by political preferences of local population
No evidence of economically significant supply-side effects
Reuter and Zitzewitz (2005): Financial recommendations biased to mutual funds
advertised in newspapers No bias for Wall Street Journal
Supply-side explanationsGovernment
Besley and Prat 2006, Gehlbach and Sonin 2009, Qian and Yanagizawa 2009
Journalists Baron 2004, Puglisi 2006
Special interest groups Herman and Chomsky 1988, Dyck et al. 2008, Alston et
al. 2010, Petrova 2008Advertisers
Baldasty 1992, Ellman and Germano 2009Political parties
Kaplan 2003, Petrova 2010
ExamplesCBS and Abu Graib story (2004)
CBS received informationchairman of the Joint Chiefs of Staff personally called
CBS anchor asking to suppress photos and videosSecret-police chief Montesinos and TV channels in
Peru (2000)Bribes to TV owners were substantially higher than
bribes to judges or members of parliamentsSmall cable channel finally exposed bribe videos
which facilitated the change of regimeGeneral Motors and LA Times (2005)
GM cancelled its advertising contract with LA Times after a negative story about its products was published
Demand-side explanationsSome media consumers want media slanted in
particular directionCoverage consistent with their priors
Confirmatory bias (Rabin and Schrag 1999)Models of media markets with consumers’ preferences
for bias (Mullainathan and Shleifer 2005) or reputation Gentzkow and Shapiro 2006)
Models with exogenous changes of control in media markets (Durante Knight 2009)
Lippmann (1922) “a newspaper can flout an advertiser, it can attack a powerful banking or traction interest, but if it alienates the buying public, it loses one indispensable asset of its existence”
Effect of competitionSupply-side driven bias: competition helps
Increases incentive of media outlets to please the audience
Increases payment necessary for influencing media (e.g. Besley and Prat 2009)
Demand-side driven bias: competition does not necessarily reveals truthIncreases incentives of media outlets to stick to
people’s priors (Gentzkow Shapiro 2006)Allows consumers to self-segregate more
effectively (Mullainathan Shleifer 2005)
Effect of advertisingDoes increasing reliance on advertising revenues
reduces supply-side media bias?Theoretically, two answers:
Yes (Besley Prat 2006, Gentkow et al. 2006, Gabszewicz et al. (2001, 2002)
It depends (Gehlbach and Sonin 2009, Petrova 2008)Anecdotal evidence suggest advertising stimulated
development of independent press in the U.S. Baldasty, 2002, Smythe 2002, Starr 2004
Empirically , true for American newspapers of 1880s Petrova 2010
OverallWhat determines media bias?
Interesting question, theoretically and empirically
Potential directions for future research:How supply-side and demand-side interact in a
two-sided market framework?Media competition and interest group
competition with heterogenous consumersMore empirical evidence for supply-side effects More empirical evidence for countries other
than United States
Example: how advertising revenues created an independent press
Argumentbased on Petrova (2010)
Higher advertising revenues create incentive for media to become independent
Theoretical conditions for advertising effect:Besley and Prat (2006): independent media is always
preferred by the audienceGehlbach and Sonin (2008): truth is preferred and
government does not want to interveneChen and Riordan (2007): market expansion
increases varietyGabszewicz et al. (2001): pooling equilibrium
becomes possibleGentzkow et al. (2006): size of bias does not matter
Necessary to look for empirical evidence
Empirical evidence Data from unique dataset on independent and
partisan newspapers in U.S. ,1881-1886Newspapers had political affiliation: Democratic,
Independent, or RepublicanPolitical affiliation implied control by political parties
through printing contractsthrough access to political informationthrough convincing constituency to subscribe
Downside of party control: inability to choose editorial policy
Data on advertising rates and location of these newspapers
MethodologyMain hypothesis: higher profitability of
advertising makes newspapers more likely to be independent
Fixed effects estimationcounty or newspaper fixed effects
Analysis of entryIV results:
local laws and ordinancesoutdoor advertising: strict regulation in some
places, after Niagara Falls storyhandbill and newspaper distribution
ResultsPositive and significant effect of local
profitability of advertising on newspapers’ independence2-3% higher probability of being independent
after 1 standard deviation changeChanges to independent affiliation more likelyEntries of independent newspapers more likely
Implies within-county growth of advertising rates explains 32% of the growth of independent newspapers from 1881 to 1886
Thank you!