Here’s an interesting and timely new paper by Brown economists Brian Knight and Chun-Fang Chiang on the impact of newspaper endorsements on voter behavior.

They analyze reader opinions before and after their local newspaper endorsed Bush or Gore in 2000. Their major findings:

— Newspaper endorsements do have an impact. It’s not huge, but it is detectable.

— Endorsements have the biggest impact when they are unexpected given the perceived bias of the newspaper. For example, they determine that in 2000, there was only a 10 percent chance The New York Times would endorse Bush, and only a 17 percent chance The Dallas Morning News (my old employer) would endorse Gore. So the NYT endorsing Gore and the DMN endorsing Bush each had only minimal impact; each was reliably set on their ideological ways long before 2000. Knight and Chiang: “These results suggest that voters are sophisticated and attempt to filter out any bias in media coverage of politics.”

By their measure, the most convincing endorsements came from The Denver Post (which went Gore when they estimate Bush would have been a more likely pick) and the Chicago Sun-Times (which did the opposite). Each swung about three percent of voters.

I have plenty of concerns with economists’ methods for determining “bias” in newspapers (some of them summarized here; not differentiating between the editorial pages and the news section is a big one), but I did find it interesting how Knight and Chiang estimated two key numbers: the percentage of each newspaper’s readers who supported Gore vs. the estimated chance that the newspaper would endorse Gore. Here they are:

Note that in 12 of the 16 cases, the newspaper’s readers were more pro-Gore than the newspaper was.