Newspaper Endorsements, Candidate Quality, and the Influence of Biased Media

• In Event: Voting, Political News and the Campaign

Abstract

In this paper, I study the effect of partisan local media on election outcomes using a newly collected set of over 16,000 local newspaper endorsements of candidates for president, Congress, statewide offices, state legislature, and local government. A newspaper’s pattern of partisan endorsements reflects the paper’s partisan bias, while the pattern of election-specific endorsements across multiple newspapers provides an estimate of candidate quality. Using this large set of endorsements, I simultaneously estimate newspaper bias for hundreds of newspapers and candidate quality for candidates in elections between 1960-2004. I show that this endorsement-based candidate quality measure is correlated with preexisting measures of candidate quality, such as incumbency, years of experience, prior political experience, and (lack of) scandals. Additionally, I show that a one standard deviation increase in the endorsement-based measure of candidate quality is associated with a 2 percentage point increase in candidate vote share, further validating it as a measure of candidate quality.

I next examine the effect of local media bias on county-level vote shares for statewide offices. I merge endorsements and estimates of newspapers’ partisan bias to county-level newspaper circulation data to get localized measures of partisan news and circulation-weighted shares of Democratic endorsements for each county. In each county I calculate the overall partisan lean of the local newspapers as well as their specific endorsements for candidates in statewide elections. I use within-state variation in these measures across counties to identify the influence of local news bias and endorsements on election outcomes while controlling for election- and candidate-specific factors. I also create a “credibility-weighted” measure of endorsements that takes into account newspapers’ partisan lean. A key aspect of the credibility-weighted endorsement measure is that the most credible endorsements come from the most biased papers -- specifically, the most credible endorsement is one where a highly partisan paper endorses a candidate who is of their typically-less-preferred political party. The credibility-weighted endorsement measure adjusts for the fact that local newspaper endorsements and newspaper partisan bias may be endogenous to local voting patterns, by discounting “expected” endorsements.

I estimate the effects of local news’ partisan bias, endorsement shares, and credibility-weighted endorsements on Democratic two-party vote shares and Democratic over-performance at the county level. Local news bias is significantly correlated with two-party vote shares, but the effect disappears when controlling for the explicit political endorsements of the local papers. I find that going from 0% to 100% of local papers endorsing Democrats increases Democratic vote shares in a county by 3.3 percentage points, and that going from the most credible Republican endorsement to the most credible Democratic endorsement increases Democratic vote shares by about 3 percentage points.

My results are consistent with a model where local news coverage affects voter information and, for some voters, their vote choice in elections. My study also highlights the fact that even biased news can be valuable to voters -- namely, that the most credible and influential endorsements come from the most biased local newspapers, and that these highly credible endorsements can effectively help voters identify bad candidates of their typically-preferred partisan lean. While biased news may make it harder to differentiate between good and bad candidates in general, the tradeoff is that biased news may help enable the electorate to weed out extremely bad candidates more easily.

Author

• Kevin DeLuca, Harvard University