EVENTS

Date:	Monday, September 13, 2004
Time:	1:00 PM -- 2:30 PM
Location:	Wohlstetter Conference Center, Twelfth Floor, AEI 1150 Seventeenth Street, N.W., Washington, D.C. 20036

September 2004

Media Bias

Members of both political parties frequently allege the existence of political bias in the media, yet there have been few systematic studies of such bias to date, and none that examine whether the media treat Democratic and Republican presidents differently. AEI scholars John R. Lott Jr. and Kevin A. Hassett have developed a simple test for political bias by comparing, in 389 newspapers, how economic reports on unemployment, GDP, durable goods, and retail sales have been covered under different presidential administrations from Ronald Reagan to the present. Following Lott and Hassett's presentation of their findings, Stephen Hess of the Brookings Institution and George Washington University, Jack Shafer of Slate, and George Washington University economist Donald O. Parsons responded.

John R. Lott Jr.
AEI

Many across the political spectrum believe that the media is biased. Where this consensus breaks down is in the inclination of that bias: of those who believe the media is biased, they usually consider the media to be biased against them. Indeed, according to surveys, many reporters, as well as consumers, believe that the media is biased. There is a burgeoning empirical literature suggesting such bias. The effects of biased media coverage are revealed in discrepancies between public perceptions, shaped by news coverage, and the actual state of affairs--in this case, the economy. 

According to surveys of journalists, daily newspapers are typically viewed as the best source of economic news, outstripping television and magazines. The availability of concrete economic data, as well as the apparent confidence in such reporting, makes daily newspaper coverage of economic news an appropriate test subject for political bias in news coverage. Furthermore, the Nexis database provides a ready source of newspaper stories and headlines to examine. Headlines were tested as they can be more definitively classified than whole news stories, which may contain subtle biases that are not so easily gauged or classified. For the purpose of this study, headlines for news stories covering reports for unemployment, GDP, durable goods, and retail sales were classified.

We found that the across all newspapers, headlines provided less positive coverage during Republican administrations than during Democratic administrations. Using simple averages, only the variables of GDP and unemployment varied significantly across administrations, with Republicans receiving fewer positive headlines. Regression analysis revealed more precisely variation in positive news coverage between Republican and Democratic administrations, with Republicans again receiving less positive coverage by between 9.6 and 14.7 percent for all newspapers. The study also provided information on coverage for major newspapers, wires, and for each variable. Ultimately, the significance of such bias lies in its effect on perception. With this in mind, a fifteen-percentage-point partisan gap produces a 3- to 4-percentage-point change in people's perception on the economy

Stephen Hess
Brookings Institute and George Washington University

The study, as it is based on a data set of classified headlines, would be more aptly regarded as a study on news headlines, rather than news coverage generally. Classifying headlines to assess news coverage can be problematic because headlines can be written out of ignorance, for the sake of promotion or hype, or to foreshadow the underlying news story.  Indeed, headlines can be inaccurate but are not necessarily biased.

It is also possible that using Nexis to search headlines might be problematic, as only twelve news bureaus cover major economic releases from the Federal Reserve or other such agencies. Perhaps it would both be more manageable and effective to classify the lead paragraphs from the twelve news bureaus, rather than the headlines from all major newspapers.

It is also important to remember that newspapers are complex institutions wherein many factors other than bias, including limited time for headline editors, can effect the creation of headlines.

Donald O. Parsons
George Washington University

Media bias is present and can often be regarded as clear and predictable. One need only look at coverage by Al-Jazeera on U.S. policy in the Middle East as an example. Such bias in the reporting of economic data, however, is harder to imagine, given the clear-cut nature of the hard numbers involved.

With regard to the methodology of the study, blind categorizations such as "good," or "bad," as used in psychology studies, may have been a better approach to classifying the headlines. As it is, the study uses large standard errors and imprecise estimates. Typically, one could improve the estimates by expanding the data set. Here, the study is limited by Nexis, which limits the data to the time parameters of the Nexis database. The study addressed this by using stacked variables to compensate. This approach however, only makes sense if the variables being tested are within the same parameters. In this case, each variable may be more or less subject to bias. Perhaps the study could observe the difference between politically sensitive variables such as unemployment and less political variables such as retail sales, using a difference-in-difference framework.

After evaluating the study, it is safe to make the cautious conclusion that Republicans are not favored in the media, but to go beyond this finding would require further examination.

Jack Shafer
Slate

The study does not provide a convincing argument of media bias, which would at least require expanding the data set used. The study might be bolstered if other forms of media beyond newspapers were included, such as television coverage. Furthermore, it would serve the study to extend the data set to include the first Reagan administration as well as the Carter administration. While this process would be more difficult and time consuming, it would bolster the data set.

Headlines should not be interchanged with coverage. According to surveyed reporters, headlines are typically written by copywriters, who may not write headlines that are necessarily reflective of the content of the piece and are more removed from the reporting process. Furthermore, it is estimated that readers spend about one second on headlines before moving to the underlying story, and as such the significance of headlines may be overstated.

Evaluating economic news may also be problematic because of the nature of the economy. According to Gregg Easterbrook's piece, "The Sky is Always Falling," in The New Republic all economic news is bad. Even when it is good or positive, it can always be a mixed bag. The economy is one of tradeoffs: for example, when the economy grows, more jobs are created, yet inflation and interest-rate worries rise.

Kevin A. Hassett
AEI

The comments and participation of the discussants is greatly appreciated and welcomed. The idea to examine the reporting of the twelve bureaus reporting on economic reports is interesting and could perhaps be incorporated, potentially using random sampling.

While much attention has been given to the usage of headlines in this study, the poll results noted in the paper indicate that headlines are in fact significant. To verify the findings in the study, perhaps an outside firm could be commissioned. The internal process and the involvement of copy editors in creating headlines is also worth examining.  The findings in the study, however, that noted a bias towards Reagan by the Los Angeles Times is telling, as its owner was known to be a Reagan supporter.

AEI staff assistant Gordon Gray prepared this summary.