Title: A Jacobian approach for the incidental parameter problem
Authors: Guangjie Li - Cardiff University (United Kingdom) [presenting]
Abstract: The information orthogonal reparameterization method for the incidental parameter problem is studied and extended. We found that the properties of the estimators for the common parameters depend on the function on the right-hand side of the differential equation used to find the reparameterization. When the function is not linear in the incidental parameters, the Jacobian from the original incidental parameters to the new incidental parameters can be used as a bias-reducing rather than a bias-removing prior as in the static panel logit and probit models. When the function is linear, it is possible to obtain consistent estimators of the common parameters as in the panel linear autoregressive models with strictly exogenous regressors and predetermined regressors. When the regressors are strictly exogenous, though the information orthogonal reparameterization does not exist, the Jacobian can still be found. When the regressors are predetermined, though the Jacobian cannot be found, a related moment condition can be used to obtain the consistent estimators.