Monday, June 22, 2009

Levitt's Abortion Result

Steve Levitt's most famous empirical finding in his best-selling Freakonomics is that abortion cuts crime. The basic idea is that abortions cut unwanted births, and because unwanted kids would have less nurturing, they would be more prone to crime. Hence, abortion lowers crime. Looking at the data between states with interaction terms, using 1973 as the cut-off date for legal abortion in the US, Levitt and Donohue argue their analysis supports this hypothesis. Many have found data contrary to this, and note that those who abort appear have socio-economic characteristics suggesting they would be better than average parents.

In a new NBER paper, Theodore Joyce notes:
Economists, on the other hand, have corrected mistakes in the original analyses, added new data, offered alternative tests and tried to replicate the association in other countries. Donohue and Levitt have responded to each challenge with more data and additional regressions. Making sense of the dueling econometrics has proven difficult for even the most seasoned empiricists.

Empirical debates can be very tricky. But the bottom line is that a positive, true result becomes more clear the more data and people look at it. If the issue becomes less clear, this favors the null that there is no effect. With GMM, instrumental variables, and interaction terms, you can torture the data to confess to anything, but when others try to replicate your results without the same bias such confessions are rarely observed.

No comments: