Thursday, May 3, 2007

The interpretation of dummies

What is the simple mistake that several great economists made? Halvorsen and Raymond point out that Robert Lucas and Zvi Griliches -among others- misinterpreted the coefficients of dummies in semilogarithmic regressions.
Halvorsen, Robert & Palmquist, Raymond, 1980."The Interpretation of Dummy Variables in Semilogarithmic Equations,"American Economic Review, American Economic Association, vol. 70(3), pages 474-75, June.
The interpretation of a dummy coefficient b is not a change of b*100 in the dependent variable. The correct interpretation is (exp(X)-1)*100. For small values of b there is no problem, but the difference can be quite large.
Mea culpa: Yes, I think that I've made this mistake before. But at least I am in good company.

No comments: