Seemingly unrelated and fixed-effect panel regressions: collinearity and singular dispersion

Abstract: The paper discusses identification conditions, representations and relations of generalized least squares estimators of regression parameters in multivariate linear regression models such as seemingly unrelated and fixed effect panel models. Results are presented on identification for unrestricted dispersion structure and general heteroskedasticity and cross-equation dependence, considering explicit and implicit restrictions, singularity of the dispersion and multicollinearity in the design matrix.