Abstract
This paper develops an econometric technique to test for political bias in news reports that controls for the underlying character of the news reported. Because of the changing availability of the number of newspapers in Nexis/Lexis, two sets of time are examined: from January 1991 to May 2004 and from January 1985 to May 2004. Our results suggest that American newspapers tend to give more positive coverage to the same economic news when Democrats are in the White House than when Republicans are; a similar though smaller effect is found for Democratic control of Congress. Our results reject the claim that “reader diversity is a powerful force toward accuracy.” When all types of news are pooled into a single analysis, our results are significant. However, the results vary greatly depending upon which types of economic data are being reported. When newspapers are examined individually the only support that Republicans appear to obtain is from the president’s home state newspapers during his term. This is true for the Houston Chronicle under both Bushes and the Los Angeles Times during the Reagan administration. Contrary to rational expectations, media coverage affects people’s perceptions of the economy.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
For a more systematic examination of positive and negative news coverage from late March through early June 2004 regarding that year’s Presidential campaign, see: Project for Excellence in Journalism (2004).
Two papers directly test for how the ownership of television and radio stations across countries varies with the returns to controlling information (Lott 1990, 1999). Another recent paper discusses how media coverage affects asset prices (Dyck and Zingales 2003). A final paper by Leeson (2008) shows that press freedom is associated with more informed citizens.
Other surveys of reporters “deny the existence of this bias, claiming that their decisions are premised solely on professional norms” (Patterson and Donsbach 1996, p. 466).
See also Opensecrets.org. http://www.worldnetdaily.com/news/article.asp?ARTICLE_ID=40862.
Patterson and Donsbach’s (1996, p. 466) survey of journalists lead them to conclude that “there is… a perceptual gap between journalists’ self-image and their actions, and it leads them to reject any suggestion that they are politically biased.”
Krupnikov et al. (2006) show that Democrats become more supportive of eliminating the estate tax as they get more information about the issue.
Gentzkow and Shapiro (2010) address Groseclose and Mylo’s reliance on 200 think tanks by looking to see if there is a general bias in phrases used in news stories (e.g., the rates at which “death tax” or “estate tax” is used) to rank newspapers by whether they are liberal or conservative. However, this still (does) may not solve the problem of identifying which newspapers are politically biased since the underlying news story is still unknown.
In addition, during both the end of the Clinton administration and the beginning of the Bush administration only Clinton era government officials were interviewed for prospective on government regulations.
Groeling and Kernell control for an even year dummy for elections in 1990, 1992, and 1994, but do not distinguish the particular emphasis on presidential approval ratings during a presidential election year.
We believe that Puglisi has misinterpreted his results (2004 version of his 2011 paper, p. 29) and that they are possibly even more interesting than he believes. Immediately prior to an election, the emphasis on Democratic issues is 24.3 % higher when there is a Democratic incumbent and an almost identical 26.3 % higher for a Republican incumbent. While it would be interesting to see how this pattern varies across individual administrations, the impression is that there is a consistent Democratic bias no matter which party currently controls the presidency. One possible explanation for the time variation in his results is that if there is diminishing marginal returns to transfers, those who are winning transfers at the end of a Democratic administration will fight less hard to keep getting additional transfers and those who expect to win transfers from Democrats will fight more to win if a Republican administration is currently in office (Jung et al. 1994).
Another paper by DellaVigna and Gentzkow (2010)) shows that Republicans gained 0.4–0.7 percentage points in towns where the Fox News Channel was carried before the 2000 election, although the study suffers from a potential endogeniety problem in terms of the cities that had access to the Fox News Channel first.
While the first version of Larcinese et al. (2011)) was written in October 2006, the first version of our paper was written in August 2004.
Ansolabehere et al. (2006) use the same methodology of relying on newspaper endorsements to measure bias.
A similar approach on a different media issue was also used in John R. Lott, Jr., “Media Bias: Political correctness invades sports coverage,” Washington Times, October 10, 2003, p. A21.
The Appendix is available at SSRN (http://ssrn.com/abstract=2319005).
We started by assigning all research assistants to look at the same five hundred headlines, but did not find any disagreements. We then assigned two research assistants to code the same headlines when we originally did our sample back to 1991, but we didn’t continue doing that for all the newspapers when we expanded the sample back to 1985 because there were again no disagreements in how headlines were being classified (the two newspapers that we didn’t do this for were the Washington Post and the Chicago Tribune). This double-checking was done for all those involved in the coding and for some part of all the different indexes. Neither Hassett nor Lott classified the headlines themselves and the coders were not told why we were classifying the headlines. Our reliance on headlines versus looking at the content of was the reason we believe that there were so little disagreement on classifications.
For all newspapers, 23 % of the sample is at zero and 7 % is at one and the distribution between these values is uniform. For the top ten newspapers, 34 % of the sample is at zero and 19 % is at one and the distribution between these values is convex from above. While STATA’s Tobit estimates don’t allow us to adjust for possible serial correlation of the error term, the presence of a lagged dependent variable is intended to minimize the importance of that factor. As a check, the appendix reports regressions that make those adjustments (using Prais–Winsten regressions with Cochrane–Orcutt transformations) and they produce very similar results to those reported here.
The fewer observations for the top ten papers occurs almost exclusively because retail sales numbers don’t get much news coverage. Even if the top ten papers don’t cover a retail sales number release, at least some of the 389 newspapers in our sample will mention the announcement. Only five of the remaining missing observations are due to durable good sales, and the reasons are the same. Rerunning the first four specifications in Table 5 so that the sample is limited to what was available for the top 10 newspapers has almost no effect on the results. For example, the Republican dummy coefficient for the first specification is −0.143 (t = 3.75). The results shown later in Table 7 for durable goods and Table 9 for retail sales are also unaffected. For example, the Republican dummy coefficient in the first specification of Table 7 becomes −0.203 (t = −2.09) and for the first specification of Table 9 becomes −0.048 (t = −0.62). The alternative of reporting all the regressions for the 389 with these observations deleted didn’t seem like the best approach both as it would throw out data and it didn’t alter the results we reported.
Summarizing comments by Alan Blinder, the New York Times wrote: “if anything, current economic coverage favored Mr. Bush by letting the administration get away with blaming 9/11 for the economy's poor performance” (Porter 2004). While a dummy for news coverage after September is positive, it is never statistically significant. Including it implies that news coverage for all newspapers exhibits an even larger partisan bias against Republicans (−0.178 (t-statistic = −3.8)), but the bias against Republicans in the top ten newspapers is smaller (−0.139 (t-statistic = −1.68)).
We reran all the regressions shown in Tables 5 through 11 using the Gallup survey data on economic conditions. There were two ways of controlling for this variable: including separate variables for the percent who thought that economic conditions were excellent, good, and fair or those who thought that economics were either excellent or good. These variables were almost never statistically significant at least at the 10 % level for a two-tailed t test, and did not really alter the results. Despite the large reduction in sample size, the estimates for all newspapers were slightly weakened and those for the top ten newspapers were slightly strengthened. The eight results for the Republican presidential variable that correspond to those reported in Table 5 are as follows (absolute t-statistics in parentheses): −0.128 (2.61), −0.111 (2.28), −0.105 (2.08), −0.086 (1.71), −0.280 (2.93), −0.220 (2.25), −0.207 (2.09), and −0.141 (1.40). For Table 10 that examines unemployment, the results for columns 1, 2, 5 and 6 are as follows: −0.092 (1.2), −0.86 (1.11), −0.342 (2.59), and −0.320 (2.41).
The time series of daily returns on all three indices, S&P 500, Value-weighted return and Dow Jones Industrial Average were obtained from Wharton Research Data Services (WRDS) from February 1, 1985 to July 30, 2004. While coefficients for all changes in market return are positive, the Dow Jones Industrial Average is the only one that is statistically significant. The coefficient for the DJIA for one day for all newspapers is 0.04 (t-statistic = 2.86) and for the top 10 newspapers the coefficient is 0.13 (t-statistic = 4.48).
Chris Carroll made another related suggestion. Replacing the simple one-month lag for the percent of the headlines that are positive with the percentage for the last six months, resulted in the Republican coefficient for specification 1 equaling −0.14 (t-statistic = −3.4) and specification 5 equaling −0.16 (t-statistic = −1.75). In both cases the new lagged variable was extremely small and had a t-statistic less than 0.5.
We tried this even though a plot of the percentage of positive news stories for either all newspapers or the top ten papers on the corresponding number of stories did not indicate any real heteroskedasticity; weighting the Tobits by the number of stories did not appreciably alter the results.
Because of Chris Carroll’s concern that there might be “an underlying 'momentum' term plus a higher-frequency 'noise' term,” we included a series of monthly dummy variables that cover the previous 48 months. The first dummy is for one month lagged, the second for two months lagged, and so on up to 48 months.
STATA does not provide clustering or robust standard errors with the Tobit option. .
The negative binomial estimates provide another way of dealing with truncation at zero. For negative binomial incident rate ratios with clustering by President and robust standard errors, the Republican coefficient for corresponding to specification 1 equaled 0.72 (t-statistic = −5.97) and specification 5 equaling 0.78 (t-statistic = −5.65). Not including clustering by President changes the t-statistics for specification 1 to −18.3 and for specification 5 to −6.99.
When the number of positive articles is zero, the natural log is taken of 0.1.
We also tried including the percentage increase in new jobs and the change in that number along with the GDP growth rates in these GDP regressions, but that did not alter the results. The coefficient on specification 1 increases from −0.163 to −0.187 with an absolute t-statistic of 1.79 and for specification 5 it shrinks from −0.24 to −0.23 with an absolute t-statistic of 1.19. We also redid the regressions in Table 5 the same way and again there was little change in the estimates.
We redid specifications 1 and 5 in Tables 7 through 11 and obtained estimates similar to those reported in the text. For all newspapers, the coefficient on Republican for durable goods equaled −0.20 (t-statistic = 1.99); for GDP, −0.183 (t-statistic = 2.11); for retail sales, −0.04 (t-statistic = 0.65); and for unemployment, −0.06 (t-statistic = 1.20). In none of the cases was the recession variable statistically significant. In the one case it came closest to being statistically significant the coefficient for GDP was 0.26 (t-statistic = 1.44). For the top 10 newspapers, the coefficient on Republican for durable goods equaled −0.24 (t-statistic = 1.82); for GDP, −0.28 (t-statistic = 1.67); for retail sales, 0.13 (t-statistic = 0.70); and for unemployment, −0.16 (t-statistic = 1.90). The recession variable was statistically significant only in the unemployment regression.
In a simple single-variable regression, measurement errors will bias down coefficient estimates. For multiple variable regression estimates, measurement errors can alter coefficient signs (Leamer 1978). The true coefficient values can be determined by estimating the direct and reverse regressions for those variables estimated with error (Koopmans 1937; Klepper and Leamer 1984). Koopmans showed that if the K reverse regressions and the direct regression all yield estimates of the same sign on every variable, then the maximum and minimum values for a coefficient from this set of K+1 estimates are consistent estimates of the endpoints of an interval that contains the true coefficients. Measurement error is not a problem for our regressions. That is indeed the case for specification 1 in Table 7 for durable goods. We follow Klepper and Leamer’s lead in introducing prior beliefs to bound the measurement error for GDP. They show that as long as R*2, the R2 obtained with this model assuming no measurement error, is less than the maximum value of R*2 consistent with all the normalized regressions lying in the same orthant, the estimates are bounded. We used Klepper and Leamer’s approach and found that as long as the expected R2 in specification 1 in Table 8 if there was no measurement error was less than 0.91, the measurement error would be bounded. To expect such a simple model to explain over 91 % of the variation in positive headlines seems doubtful as long as the reader agrees with that the measurement error issue is not a problem for that regression either. The estimates for retail sales were not bounded and only the estimates for the top ten newspapers for Unemployment in specification 5 were bounded.
The Associated Press provides 15.3 % of all headlines in our sample. Stories were classified as AP when either the reporter credited was the AP reporter or the story was simply identified as being from the AP. We did not do a similar breakdown for the other news services. AP and other wire service stories that appear in many papers are often not put into Nexis because they figure that there will be at least a few other papers that will put them up and they figure that it adds little to have a large number of papers all report the same thing. As we show later, the headlines on the AP stories that are run in newspapers are very similar to the headlines that are on the AP wire service version.
It was not possible to use ordered logits on our earlier regressions because there were too many categories. Even here simple logit regressions are not appropriate because newspapers occasionally run multiple stories on the economic numbers when they are released. Using Tobits or ordered logits for these individual newspapers has little effect on which newspapers have a statistically significant partisan gap. The one exception is for the New York Times, which goes from being statistically significant at the 13 % level in the equivalent of specification 4 using a Tobit to being significant at the 0.08 level with an ordered logit. Using these different estimation procedures did have an effect on the size of the estimated partisan bias.
We also did this using the percentage of the headlines that were negative and obtained similar results. The omitted categories were for neutral or mixed headlines.
Even the New York Times recently acknowledged a bias on social issues, though there was no discussion of its coverage of economic topics. See Okrent (2004).
For the twenty-two regressions reported in Table 12 none of the Republican coefficients are affected by the inclusion of the recession variable; the recession dummy is statistically significant in just four of those estimates.
Including the dummy variable for recessions did not alter these general results, though this was the one set of regressions where the recession variable was itself consistently statistically significant.
These estimates were also redone using a dummy variable for coverage after September 11th, but there was little change in the partisan gap, with the gap still being statistically significant for all newspapers other than the Los Angeles Times. The dummy for September 11th was only close to being statistically significant for the Chicago Tribune. Even for the Chicago Tribune, the estimates implied that the Republicans got an even smaller percentage of positive headlines than would have normally been expected in the brief but short economic downturn after the attack (the coefficient was −0.58 with a t-statistic = −1.53).
Earlier evidence from 1985 also shows that newspaper readers are relatively conservative. A Los Angeles Times survey of 6,000 newspaper readers found that 24 % of newspaper readers described themselves as liberal, 33 % as middle-of-the-road, and 29 % as conservative (Shaw 1985).
Following Puglisi (2004), we interacted the dummy for a Republican president with the election year. Puglisi says that when there is a Republican president the press becomes more favorable towards Democrats. While the results are mixed, none of them seem to support his hypothesis. For all newspapers, the news coverage actually appears to become more favorable towards Republicans. For the top 10 papers there is no statistically significant effect. Since Puglisi examined the New York Times, we re-examined that newspaper and again found no change during election years. We also tried limiting the election variables to just the three months studied by Puglisi (August, September and October). The interaction for a Republican president with those three months until the election is now consistently negative (consistent with his theory), but while the coefficients for these interactions are relatively large, none are close to being statistically significant.
One of the authors came across an example of this during the spring of 1994 at a lunch with some members of the Philadelphia Inquirer’s editorial board. The board briefly discussed endorsing a Republican candidate but decided against it because Democrats controlled the governorship, the state legislature, and the city government of Philadelphia and there was the concern that they would not have the same access to news from the Democrats if they endorsed the Republican.
Along a similar line, Tierney (2004) recently surveyed journalists at the 2004 Democratic National Convention and found that while ones based in Washington favored Kerry by at 12 to 1 margin over Bush, those outside the Beltway favored Kerry by a 3 to 1 margin. He wanted to see whether reporters’ support for Kerry might be explained by self-interest instead of ideology or whether covering Kerry might be better for their careers (“mainly because any change in administration means lots of news”). If anything he found that they supported Kerry despite its impact on their careers, with a 77–67 majority believing that covering Bush would be more likely to produce a “front page [story] and get on the air.” The test that we conduct differs in that we see how the partisan gap varies as the cost of helping a particular party changes. While very interesting, the cross-sectional survey data studied by Tierney do not examine these marginal changes.
Even after including the variables for media coverage and the economy, including the dummy variable for whether there is a Republican president finds that respondents during Republican administrations were 10.5–11.4 percentage points less likely to believe that the economy was getting better (the t-statistics were about 3). This was true both for all newspapers and the top ten in specifications 4, 5, 9, and 10. This implies that even after accounting for newspaper coverage and the actual economic numbers, respondents still had a relatively negative impression of how the economy was doing during Republican administrations. Including a dummy variable for recessions (while usually statistically significant) has almost no effect on the Republican or press coverage variables. Whether this additional deficit is due to Republicans being unable to explain their accomplishments or due to other types of news coverage is not clear.
These estimates are available at SSRN (http://ssrn.com/abstract=2319005).
References
Alterman, E. (2004). What liberal media? The truth about bias and the news. New York: Basic Books. (Reprint edition March 2, 2004).
Andrews, E. L. (2004). “Economy remained strong in 4th quarter, U.S. reports,” New York Times, Saturday, January 31, 2004, 1.
Ansolabehere, S., Lessem, R., & Snyder, J. M, Jr. (2006). The orientation of newspaper endorsements in U.S. elections, 1940–2002. Quarterly Journal of Political Science, 1(4), 393–404.
Associated Press/Ipsos-Public Affairs survey (2004). 1101 Connecticut Avenue NW, Suite 200, Washington, DC 20036, June 7–9, 2004.
Baron, D. P. (2006). Persistent media bias. Journal of Public Economics, 90, 1–36.
Berry, J. M., & Allen, M. (2003). “U.S. economic growth surges; output rises at highest rate since 1984, but jobs still decline,” Washington Post, October 31, 2003, p. A1.
Birz, G., & Lott, J. R., Jr. (2011). The effect of macroeconomic news on stock returns: new evidence from newspaper coverage. Journal of Banking & Finance, 35, 2791–2800. co-authored with Gene Birz.
Bovitz, G. L., Druckman, J. N., & Lupia, A. (2002). When can a news organization lead public opinion? Ideology versus market forces in decisions to make news. Public Choice, 113, 127–155.
Chan, J., & Stone, D. F. (2013). Media proliferation and partisan selective exposure. Public Choice, 156(3–4), 467–490.
Chiang, C., & Knight, B. (2011). Media bias and influence: Evidence from newspaper endorsements. Review of Economic Studies, 78(3), 795–820.
Coulter, A. (2002). Slander: Liberal lies about the American right. New York: Crown Publishers.
Croteau, D. (1998). Examining the ‘liberal media’ claim. New York: Fairness & Accuracy in Reporting. (June).
DellaVigna, S., & Gentzkow, M. (2010). Persuasion: Empirical evidence. Annual Review of Economics, 2(1), 643–669.
DellaVigna, S., & Kaplan, E. (2007). The Fox News effect: media bias and voting. Quarterly Journal of Economics, 122(3), 1187–1234.
Demsetz, H., & Lehn, K. (1985). The structure of corporate ownership: causes and consequences. Journal of Political Economy, 93(6), 313–346.
Doms, M., & Morin, N. (2004). Consumer sentiment, the economy, and the news media. San Francisco: Federal Reserve Bank of San Francisco. (July).
Duff, C. (1996). “Factory orders post a 1.3 % jump for December,” Wall Street Journal, Friday, February 16, 1996, A2.
Dyck, A., & Zingales, L. (2003). The media and asset prices. University of Chicago Business School Working Paper, August 1.
Fields, G. (2004). “Unemployment rate rises,” Miami Herald, Friday, April 2, 2004, p. 1.
Gentzkow, M., Petek, N., Shapiro, J. M. & Sinkinson, M. (2012). Do newspapers serve the state?: Incumbent party influence on the US press, 1869–1928. NBER Working Paper 18164.
Gentzkow, M., & Shapiro, J. M. (2010). What drives media slant? Evidence from U.S. daily newspapers. Econometrica, 78(1), 35–71.
Gerber, A., Karlan, D., & Bergan, D. (2009). Does the media matter?: A field experiment measuring the effect of newspapers on voting behavior and political opinions. American Economic Journal, 1(2), 35–52.
Goldberg, B. (2001). Bias: A CBS insider exposes how the media distort the news. Washington, DC: Regnery Publishing.
Groeling, T., & Kernell, S. (1998). Is network news coverage of the president biased? Journal of Politics, 60(4), 1063–1087.
Groseclose, T., & Milyo, J. (2005). A measure of media bias. Quarterly Journal of Economics, 120(4), 1191–1237.
Hagenbaugh, B. (2003). 7.2% GDP growth fastest in 19 years, USA Today, October 31, 2003, A1.
Hamilton, J. T. (2004). All the news that’s fit to sell: how the market transforms information into news. Princeton: Princeton University Press.
Ivanovich, D. (2003). “U.S. economy shows biggest rise in years; officials hope improvement spurs new jobs. Houston Chronicle, 31(2003), A1.
Jung, G., Kenny, L. W., & Lott, J. R, Jr. (1994). An explanation for why senators from the same state vote differently so frequently. Journal of Public Economics, 54(1), 65–96.
Klepper, S., & Leamer, E. E. (1984). Consistent sets of estimators for regression with error is all variables. Econometrica, 52(1), 162–183.
Kohut, A. (2008). “The web: alarming, appealing and a challenge to journalistic values,” Project for Excellence in Journalism, The PEW Research Center, March 17 (http://www.stateofthemedia.org/files/2011/01/Journalist-report-2008.pdf).
Koopmans, T. C. (1937). Linear regression analysis of economic time series. Amsterdam: Econometric Institute Itaarlem—de Erwin F., Bohn NV.
Krupnikov, Y., Lupia, A., & Prior, M. (2006). Public ignorance and estate tax repeal: the effect of partisan differences and survey incentives. National Tax Journal, 59(3), 425–437.
Larcinese, V., Puglisi, R., & Snyder, J. M., Jr. (2011). Partisan bias in economic news: evidence on the agenda-setting behavior of U.S. newspapers. Journal of Public Economics, 95(9), 1178–1189.
Leamer, E. (1978). Specification searches: ad hoc inference with non experimental data. New York: Wiley.
Leeson, P. T. (2008). Media freedom, political knowledge, and participation. Journal of Economic Perspectives, 22(2), 155–169.
Leeson, P. T., & Coyne, C. J. (2005). Manipulating the media. Institutions and Economic Development, 1(2), 67–92.
Lott, J. R., Jr. (1990). An explanation for public provision of schooling: the importance of indoctrination. Journal of Law and Economics, 33(1), 199–231.
Lott, J. R., Jr. (1999). Public schooling, indoctrination, and totalitarianism. Journal of Political Economy, 107(S6), S127–S157.
Lott, J. R., Jr. (2003). The bias against guns. Washington, DC: Regnery Publishing, Inc.
Lott, M. (2007). Do media outlets really not provide both sides in news coverage: Re-evaluating Groseclose and Milyo. Working paper.
Media Research Center (1995). MediaWatch: June 1995. 9(6).
Media Research Center. (2004) “The liberal media: every poll shows journalists are more liberal than the American public—and the public knows it. June 30.
Miller, J. P. (2004). “GDP disappoints,” Chicago Tribune, Saturday, January 31, 2004, 1.
Mullainathan, S., & Shleifer, A. (2005). The market for news. American Economic Review, 95(1), 1031–1053.
Neikirk, W. (2003). “GDP aside, jobless rate is number that matters,” Chicago Tribune, October 31, 2003, 1.
Okrent, D. (2004). “Is the New York Times a liberal newspaper?” New York Times, Sunday, July 25, 2004.
Patterson, T. E., & Donsbach, W. (1996). News decisions: Journalists as partisan actors. Political Communication, 13, 453–468.
Pew Research Center for the People and the Press. (2004a). Survey of journalists. March 10–April 20, 2004.
Porter, E. (2004). “Do newspapers make good news look bad?” New York Times, September 12. D3.
Project for Excellence in Journalism (2004). Character and the campaign: what are the master narratives about the candidates in 2004 and how is the public reacting to them? Project for Excellence in Journalism, Pew Research Center and University of Missouri journalism school.
Puglisi, R. (2011). Being the New York Times: the political behaviour of a newspaper. The B.E. Journal of Economic Analysis & Policy, 11(1), 20.
Puglisi, R. & Snyder, J. M. Jr. (2011). The balanced U.S. press. NBER Working Paper 17263.
Rasmussen Reports (2008). Belief growing that reporters are trying to help Obama win. Rasmussen Reports, July 21, 2008.
The PEW Research Center (2004b). “Bottom-line pressures now hurting coverage, say journalists,” Project for Excellence in Journalism, May 23, 2004 (http://www.people-press.org/2004/05/23/bottom-line-pressures-now-hurting-coverage-say-journalists/).
San Jose Mercury News. (1996). “Factories break losing streak,” Mercury News Staff writers, the Associated Press, Bloomberg New and Reuters Friday, February 16, C1.
Shapiro, R. (2012) “Money, politics and the press: media political donations to Democrats,” The Huffington Post, September 6, 2012 (http://www.huffingtonpost.com/2012/09/06/media-political-donations-democrats_n_1855502.html).
Shaw, D. (1985). “Public and the press—two viewpoints,” Los Angeles Times, August 11, A1.
Smith, T. (1988). The vanishing economy. Arlington: The Media Institute.
Tierney, J. (2004). Finding biases on the bus. New York Times, Sunday, August 1, 16.
Van Riper, T. (2004). “Jobs, jobs everywhere,” Daily News (New York), Friday, April 2, 2004, 17.
Wire Reports (1996). “Clinton calls for debate on economic growth, latest round of indicators show marked slowdown in recent months,” Dallas Morning News, Friday, February 16, D1.
Acknowledgments
We would like to thank John Cogan, James Hamilton, Joe Harrington, Larry Kenny, Mark Klamer, Sam Peltzman, Florenz Plassmann, Alex Tabarrok and the participants in seminars at the University of Texas, University of British Columbia, Simon Fraser University, University of Canterbury (NZ), and the University of Maryland for their comments. Brian Blase, Jill Mitchell, Gordon Gray, Refael Lav, Aarif Morbi, Neil Dutta, Robert Harrison, and Robert Jeffrey for their helpful research assistance.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Lott, J.R., Hassett, K.A. Is newspaper coverage of economic events politically biased?. Public Choice 160, 65–108 (2014). https://doi.org/10.1007/s11127-014-0171-5
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11127-014-0171-5