Worth mentioning that the Freakonomics guys (Donohue and Levitt) work tying abortion law to crime reduction was debunked. There was a bug in the regression analysis code, "coding errors" (as in typos while copying data tables from print sources), and flat out statistical errors (most egregious was their use of total arrests instead of arrests per capita).
More interesting is that reduction in childhood exposure to lead is the most significant factor. Levitt counters the debunking of his theory two paragraphs down from your link:
I know, Levitt and Donohue made the corrections, and then made even more "corrections", until they got back their original results (or as close as they could get).
That's how it often goes when researches are accused of scientific misconduct (not to accuse LD, just noting). One famous example was Potti & Nevins at Duke. There were several back-and-forth "Comments" published to counter the questions Baggerly & Coombes were raising. PN repeatedly addressed the issues, each time releasing another "corrected" supplementary data set. Eventually the journals stopped publishing the BC critiques, PN seemingly convinced the reviewers that BC were just being pedantic about inconsequential statistical subtleties. It only came to a head because Duke was proceeding with human clinical trials based on that research.
To this day Potti still claims that he can address all the issues BN raised and correct them, and that his research was valid.
http://en.wikipedia.org/wiki/The_Impact_of_Legalized_Abortio...