Alex Tabarrok scooped you all on this back in September:

It seems to me that using the median congressman to measure media political bias is a little like trying to determine the bias of a scale by comparing it with a collection of scales that have been selected because they look good in people’s bathrooms.

Well, as I point out at DPS, even if you take 50 as the “objective” midpoint (as opposed to using the median congressman—a position significantly to the right of 50), the only media outlet that is right-wing is FNC Special Report. The other media outlets cluster just right of the Democratic median, around an ADA score of 60.

Like I also mentioned at DPS, I’d rather they used other ideology measures than ADA scores… Groseclose has invested a lot of effort into normalizing them over time, but I think they’re inferior to techniques that include all votes like the NOMINATE and C-J-R item-response models. The method in the paper wouldn’t be hard to adapt to use those scoring methods (the hard part of the project is compiling the interest group citation rates); in fact, I’d be surprised if he hasn’t done it already for the APSA version of the paper. And some bootstrapped errors would be nice, too.

That reminds me: the one thing I hate about ADA scores is that the scale is reversed from every other liberal-conservative measure on the planet (where “low” is most liberal and “high” is most conservative). When you put it on a horizontal scale, the “right” of the ADA scale is the ideological left, and vice-versa.

The normalization over time of ideology measures might have some interesting results. I’d hypothesize that the major media outlets and the median congressman have drifted apart over the last 30 years. Have the media drifted leftward or congress rightward? And what does this tell us about “media bias”?

The evidence from roll calls, etc., suggests that Congress hasn’t so much drifted left or right as drifted apart—polarization has increased as the parties have become more distinct and the Dixiecrats have died off (or, to a lesser extent, changed parties, retired, or lost elections).

Trying to track the media over time would probably be a great, long-term project, although I suspect the RAs would rebel at the prospect of doing a content analysis of every news article in the New York Times (and the other outlets) since, say, 1960. Of course, there are automated methods to speed this stuff up with machine-readable information, assuming the archives are available in electronic format.

For those with an interest in this topic, Poole and Rosenthal’s book, Congress: A Political-Economic History of Roll-Call Voting is an excellent place to start. There’s an article in this month’s American Political Science Review by Josh Clinton, Simon Jackman, and Doug Rivers that extends that research line (and makes it more econometrically sound).