Gary Chamberlain, Distinguished Fellow 2015

Gary Chamberlain has made early and influential contributions to economics in diverse fields. Along with Zvi Griliches, he studied returns to education using data on siblings. His work on this and other models with unobservable effects provided key insights for linear and nonlinear models with latent variables, which have been widely applied. He has continued his interest in the economics of labor and education. His inaugural paper in the Proceedings of the National Academy of Sciences on the occasion of his election to the Academy in 2011 estimates the effects of teachers and schools on later college attendance and test scores of students.

Using an original approach based on multinomial approximations, he provided the first treatment of efficiency bounds in a variety of moment-based semi-parametric models. In linear regression, for example, his work can be thought of as examining how much improvement we can expect in estimating a linear regression if we model the distributions of errors, instead of proceeding as if the standard normal assumption were appropriate. Economists are used to assuming that regression model errors have mean zero and finite variance, and it is well known that in that case, in large samples, the distribution theory based on assuming normal errors is approximately correct, even if errors are not normal. One might think that by using a non-normal model that includes normality as a special case, one could get an improvement in efficiency when the model is correct, while losing nothing, asymptotically, because the normal model is a special case. Chamberlain’s work showed in this context that any such method either fails to retain the normal model’s robustness to misspecification, or is no more efficient asymptotically than the normal model.

He has continued his interest in econometric methods that are useful in applications. In 2007 and 2009 Econometrica papers he used a decision-theoretic approach to derive original, elegant and practical new approaches to instrumental variables and to panel data models.

In two 1983 Econometrica papers, one joint with Michael Rothschild, he was an early contributor to factor models of arbitrage-free asset pricing.

He is the Louis Berkman Professor of Economics at Harvard University and received his PhD from Harvard in 1975. From 1979 to 1987 he was on the faculty of the University of Wisconsin in Madison.