The Economist Magazine on Steve Levitt and Abortion
But what is most amazing is that despite all the statistical significance being eliminated from their panel data set being eliminated when the results are done Donohue and Levitt's way with the the data they want to use, John Donohue now says what they wanted to do:
There was no warning that these were the wrong tests at all in either of their previous papers or in Levitt's book. Levitt now refers to the “collage of evidence," but the panel data was the only test that really amounted to anything worthwhile. The rest of the data was merely cross sectional or time series.