Abstract
I derive conditions under which a class of fixed-effects estimators consistently estimates the population-averaged slope coefficients in panel data models with individual-specific slopes, where the slopes are allowed to be correlated with the covariates. In addition to including the usual fixed-effects estimator, the results apply to estimators that eliminate individual-specific trends. I apply the results, and propose alternative estimators, to estimation of average treatment in a class of nonlinear unobserved-effects models.
This content is only available as a PDF.
© 2005 President and Fellows of Harvard College and the Massachusetts Institute of Technology
2005
You do not currently have access to this content.