Improved Heteroscedasticity-Consistent Ridge Estimators for Linear Regression with Multicollinearity

This paper formalizes ridge estimators that specifically address the simultaneous presence of multicollinearity and heteroscedasticity in the data. We propose improved ridge estimators that can also tackle the issue of heteroscedasticity and show that our suggestion improves the existing estimators in terms of smaller mean square error. The improvement is more significant with highly collinear and heteroscedastic data. The proposed estimators are very easy to implement and provide us with a simple alternative to handle both of these issues simultaneously. Finally, we have compared our proposed estimator with the existing ones using simulations and have illustrated their use with application to real data.