Monday, March 9, 2015

The Economics of Media Bias

Here are four basic questions about media bias:
"First, is media news reporting actually slanted? ...
Second, if reporting is biased, what is the reason? Is such bias driven by the
supply-side, as when reporting reflects the prejudices of an outlet’s owners or journalists? ...
Third, what is the effect of media competition on accuracy and bias?  ...
Finally, does media reporting actually matter for individual understanding and
action? Does it affect knowledge? Does it influence participation in the political
process? Does it influence how people vote?"
The questions are posed by Andrei Shleifer in his paper on "Matthew Gentzkow, Winner of the 2014
Clark Medal," in the Winter 2015 issue of the Journal of Economic Perspectives. As background, the Clark medal is given by the American Economic Association each year "to that American economist under the age of forty who is judged to have made the most significant contribution to economic thought and knowledge." Shleifer is describing the academic work for which Gentzkow won the award last year. Shleifer argues: "In a very short decade, economic research has obtained fairly clear answers to at least some of these questions."

(Full disclosure: My paid job has been Managing Editor of the JEP since the first issue in 1987. All papers in JEP from the first issue to the most recent are freely available on-line courtesy of the American Economic Association. Shleifer was Editor of JEP, and thus my boss, from 2003-2008.)

On the first question of the existence of media bias, how does one go beyond anecdotes about how different newspapers or TV channels covered certain stories to come up with a defensible quantitative way of detecting media bias? The modern approach has been to use text analysis. For example, have a computer search a dataset of all speeches given in Congress during the year 2005. Have the computer search for phrases that are much more commonly used by Republicans or by Democrats. For example, in 2005 Democrats were much more likely to refer to the "war in Iraq" while Republicans were more likely to refer to the "war on terror." Now do a search on the text used by media outlets, and see if they are more likely to be using Republican phrases, Democratic phrases, or an even mixture of the two.

Gentzkow didn't invent this approach to meausuring media bias. For earlier work on the subject in the research literature, a starting point would be the article by Tim Groseclose and Jeffrey Milyo, “A Measure of Media Bias," in the Quarterly Journal of Economics in 2005 (120:4, pp. 1191–1237). But in work with co-author Jesse Shapiro, Gentzkow applied the approach to newspapers across the US and was thus able to provide hard evidence that many newspapers indeed exhibit partisan bias in how they report the news.

Does the bias of newspapers reflect their owners, or their customers? Here's how Shleifer describes it:
"Gentzkow and Shapiro then collected data on the use of these highly diagnostic phrases in US daily newspapers and used these data to place news outlets on the ideological spectrum comparable to members of Congress. In addition to this large methodological advance in how to measure partisan newspaper slant, the paper used detailed information on newspaper circulation and voting patterns across space to estimate a model of the demand for slant and to show that—consistent with the theory—consumers gravitate to like-minded sources, giving the newspapers an incentive to tailor their content to their readers. They also show that newspapers respond to that incentive and that variation in reader ideology explains a large portion of the variation in slant across US daily newspapers. ... [A]fter controlling for a newspaper’s audience, the identity of its owner does not affect its slant. Two newspapers with the same owner look no more similar in their slant than newspapers with different owners. Ownership regulation in the US and elsewhere is based on the premise a news outlet’s owner determines how it spins the news. Gentzkow and Shapiro produced the first large-scale test of this hypothesis, which showed that, contrary to the conventional wisdom and regulatory stance, demand is much more influential in shaping content than supply as proxied by ownership.
Does more competition in the media tend to increase or diminish this bias? This question is tough  to answer, but in a different paper by Gentzkow and Shapiro, they look at a closely related topic of how people of different political beliefs use the Internet. Specifically, do people tend to cluster at the websites that that match their ideology, or do they surf around? Shleifer describes the result:
One might worry that the increase in choice among news suppliers as a result of the Internet would allow news consumers to self-segregate, reading only news that confirms their preconceptions. Gentzkow and Shapiro test this claim using data from a panel of Internet users for which they have a survey-based measure of political ideology and tracking data on online news consumption. They find that ideological segregation is surprisingly low online. The average conservative’s news outlet on the Internet is about as conservative as; the average liberal’s is as liberal as Strikingly, the Internet is less ideologically segregated than US residential geography: two people using the same news website are less likely to have an ideology in common than two people living in the same zip code.
Finally, is it the case that people just choose the media outlets that reflect their bias, in which case the media bias doesn't affect their opinions or their voting patterns? Or is there reason to believe that the extent of media bias does affect opinions and voting patterns?

In one study, Gentzkow looked at historical data on how television coverage spread across the United States, and what changes in voting patterns followed. As Shleifer writes; "He estimates a huge negative effect: the availability of television accounts for between one-quarter and one-half of the total decline in voter turnout since the 1950s. Matt argues that a principal reason for this is substitution in media consumption away from newspapers, which provide more political
coverage and thus stimulate more interest in voting."

In a different study, Gentzkow and co-authors look at the patterns of newspapers being born and dying from 1869 to 2004, and compare this with voting patterns. Shleifer writes:
They find that newspapers have a large effect in raising voter turnout, especially in the period before the introduction of broadcast media. However, the political affiliation of entering newspapers does not affect the partisan composition of an area’s vote. The latter result contrasts with another important finding, by DellaVigna and Kaplan (2007), that the entry of Fox News does sway some voters toward voting Republican. An interpretation consistent with these findings is that newspapers motivate but don’t persuade, while television does the opposite.
Research on media bias and its political effects is certainly not settled, but for what it's worth, I'd sum up the existing evidence in this way. There's lots of political bias in the media, mainly because media outlets are trying to attract customers with similar bias. But in the world of the Internet, at least, people of all beliefs do surf readily between news websites with different kind of bias. The growth of television to some extent displaced the role of newspapers and lowered the extent of voting. For the future, a central question is whether a population that gets its news from a mixture of websites and social media becomes better-informed or more willing to vote, or whether it becomes a population that instead becomes expert at selfiesm, cat videos, World of Goo, Candy Crush, Angry Birds, and the celebrity-du-jour.