Assumptions Built into the Proof of Media Bias

I appreciate Tim Groseclose’s reply below, but I fear his response only makes me more skeptical of the reasoning he uses. I’m going to hide most of my post “below the fold” because I don’t want to steal our guest’s thunder.  (It’s his guest-blogging stint, not mine!)  But for those interested, I’ll explain a bit below why I don’t think the argument adds up.

In the post below, Professor Groseclose suggests that journalists are about 2/3 interested in persuading people and 1/3 interested in informing people. The reasoning appears to be as follows:

1) Assuming for the sake of argument that all of Groseclove’s analysis is correct, unbiased journalists just trying to inform should have a slant quotient of 50;
2) Unbiased journalists would report as a centrist politician would report, and therefore would also have an sq of 50;
3) The pq of a representative in a particular district he found that happens to have a republican/democratic split similar to a particular group of reporters is 100.
4) If reporters were trying to persuade people, they would have the same sq as a class than the sq of the one representative he found — that is, 100.
5) The average slant quotient calculated by his study is 65.
6) Because 65 is about 1/3 the way between 50 and 100, it should be that journalists are about 1/3 interested in persuading people (the goal of politicians) and 2/3 interested in informing people (the idealized goal of journalists).

Assuming I have described Groseclose’s reasoning accurately, I find it quite puzzling. Each step contains a series of assumptions built into it that are unproven at best and seem made up out of thin air at worst. For example, I don’t know why we would expect a group of people that has the same Democratic/Republican ratio as the district that a particular elected official represents to share any particular characteristic with that elected official. Similarly, I don’t know why we would expect the politics of a small group of reporters (washington correspondents) to match the politics of a large group (reporters as a whole). Further, I don’t know why would expect a reporter’s level of interest between two goals to have a linear relationship between their idealized sq and pq scores. And finally, the assumption at step 1 seems to make the argument circular: The study is correct if you assume the study is correct. But what if you don’t think the study is measuring anything real at all?

Just to be extra clear, I don’t doubt that media bias exists. I see it all the time. I also think it’s plausible that the end result is right that journalists are mostly focused on informing but also somewhat focused on persuading. But there’s a critical difference between explaining data in a rigorous way and coming up with a long series of seemingly-arbitrary steps that end up with a plausible-sounding result.  After spending some time studying it, my fear is that Groseclose’s reasoning seems more like the latter than the former.