Tuesday, September 21, 2004

Update on newspaper bias paper  

When I posted before on the Lott and Hassett paper I drew a large number of comments from Tim Lambert, who has now posted a blog that claims to refute LH. I would encourage people, before drawing conclusions, to try to read the original LH paper, which Prof. Lambert did not link. Did they run many regressions? Sure they did. Did they cherry pick the results? They report that the different specifications did not matter; readers on Lambert's site have gotten the data from LH and are checking. This is as it should be. I'm not convinced by the evidence Lambert has provided so far, but the issue should remain open until the data are further explored. I will agree with him that the estimation technique LH use is open to criticism, but they included a second. Note to people writing working papers -- even if you think the results are invalid, I like to see your ordinary least squares regressions anyways. We often know how the results are biased, and seeing how the bias is resolved helps me to see whether your story is correct.

Again, let me be clear -- if the LH results are correct, all they really mean to me is that coverage of economic events under Clinton differed from that under Bush pere et fil. Calling it bias is conjectural. Worth noting -- does it differ between the two Bushes? I don't think LH checked.