Description.Three members of Yale University's Department of Economics have recently published a working paper that addresses an important gap in our understanding of whether, how, and to what degree media influences public opinion, political attitudes, and voting behavior. One of the most important features of this paper is the question that it does not address- whether or not media bias exists. The reason why they don't is that it is already settled:

There is substantial evidence that media sources have identifiable political slants, but there has been relatively little study of the effects of media bias on the views and behavior of media consumers, or of the effect of mere exposure to news (irrespective of the slant, for example Gentzkow (2006)).

Thus, taking bias as a given, the authors focus on how that bias affects on media consumers, particularly the shaping of their political opinions and their voting behavior. To underscore how little understood is the effect of bias on opinion and behavior, the authors describe several possible models and mechanisms:

While it is possible that news with a distinctive political perspective will move readers’ opinions, it is also possible that readers will compensate for the leaning of the source and thus not shift their opinions towards that of the source. This could be because readers incorporate their perception of the bias when they update their beliefs, or simply because the sources are deemed not credible. It is even possible that readers will be inspired to contest the biased source, and thereby overcompensate for any bias. Further, it may be that media bias may have a real effect on readers or viewers, but that the effects of bias are much smaller than the effect of information provision or the heightened salience given to the issues or events selected for coverage. In other words, whether a particular topic is covered may be more important in shaping opinions than the slant of the media source.

Measures.The study employs the Groseclose and Milyo measure of media bias, one based on who papers cite as authorities in their stories:

The conventional wisdom regarding the political slant of these papers’ news coverage is supported by a recent study. Groseclose and Milyo (2005) propose an innovative measure locating different media outlets on the right-left political spectrum based on the similarity of the experts used by the media outlet and those cited by conservative and liberal members of Congress. They find that the Washington Times is by far the most conservative of the six papers they assess, on average citing a mix of think tanks and other research organizations similar to those cited by members of Congress with an Americans for Democratic Action (ADA) score of 35 (lower scores signify more conservative voting on House and Senate roll call votes). In contrast, the Washington Post is much more liberal; the Post cites experts similar to those cited by members of Congress with an ADA score of 66.

Method.The study made use of the unique demographics of the north eastern Virginia and the occasion of the Virginia Gubernatorial election to conduct a "natural experiment" to test their hypotheses:

The presence of a liberal and conservative paper serving the same region creates an outstanding opportunity to study the effect of media bias in a naturalistic setting within a single population. Approximately one month prior to the Virginia Gubernatorial election in November 2005 we administered a short survey to a random selection of households in Prince William county, a northeastern Virginia county. We used the survey to establish whether the household already subscribed to the Post and/or the Times and to obtain some pre-treatment information about demographics and political attitudes. From the set of households reporting that they received neither the Post nor the Times, we randomly assigned households to get subscriptions to either the Post or the Times, or to a control group which was not sent either paper.

Results. Despite a limited intervention- study subject received the papers for only a few weeks, effects were discernible and consistent with some of their predictions:

There was evidence that getting the Post increased the probability a subject supported the Democratic candidate for Governor in Virginia, and weak but reasonably consistent evidence that getting either paper shifted subjects away from the President and Republican party. Looking at the newspaper coverage, there was clear evidence of a slant in the news consistent with the conventional wisdom about the papers. However, the month prior to the post-election survey was a difficult period for President Bush, one in which his overall approval rating fell by approximately 4 percentage points nationwide. It appears that heightened exposure to both papers’ news coverage, despite opposing ideological slants, moved public opinion away from Republicans.

Conclusion.The authors propose two explanations of their findings:

First, media bias may alter behavior and opinions. The Post, for the most part, moved individuals left whereas the Times did not move individuals to the right (and in fact, in many cases, the statistically insignificant point estimate indicates a slight shift to the left). Hence, the mechanism is not as simple as newspapers moving individuals in the direction of their reporting bias. If individuals behave as Bayesians, then one must know their perception of the bias, not just the actual bias. Hence, in this setting, if individuals perceived the Times to be conservative, then biased news reporting may not influence public opinion at all. On the other hand, if individuals perceive the Post as more neutral, they may be more likely to be influenced by the reporting. (Emphasis added)

Second, the shift leftward may simply be a result of a reduction in the administration’s standing due to unfavorable news coverage during the four weeks in which we sent households the newspapers. The experiment coincided with a difficult political period for the Bush administration, and the exposure to newspapers made the treatment groups (both the Post and the Times) more aware of current events. From October 15 to November 8 national presidential approval rates fell by 4 percent. Examination of the national news coverage in the Times and Post showed that while the Times selected more favorable stories than the Post, and the Times headlines were more
sympathetic to the administration, both papers carried many stories about the Iraq war, political scandals, and the failed attempt to place Harriet Miers on the Supreme Court. The evidence suggests that those exposed to this stream of news, even when the news was given a relatively pro-administration slant, held the President and his party in lower regard than those who were less likely to read the news.

Commentary

As to be expected, the authors warn against over-generalizing from their own findings. In doing so they make eviden thow complex is the relationship between media bias and media consumer behavior:

As in all empirical work, experimental or not, there is still the important question of generalizing from our particular findings. Any broad inferences from this study about the effects of media bias on political decisions should recognize that the results may depend on several specific features of our experiment, such as the political context, choice of subjects, intensity of treatment, length of the study, timing of the study, and choice of media outlets and type.

That more research in this area should be undertaken is a given. My preference would be to see more attention given over to the question of the perception of bias. To my mind this one of the most important insights of the study, i.e. the possiblity that not recognizing bias makes media consumers more susceptible to it. If so, it might be that case that just like in other games, it helps to have scorecard.

See also: The Belmont Club: Page 53