Heteroscedastic regression in robust econometrics

We study the effect of heteroscedastic errors on different robust regression methods. Firstly we derive asymptotic heteroscedasticity tests for the least weighted squares regression, which is one of robust regression methods with a high breakdown point. These tests have the same form as standard heteroscedasticity tests for the least squares context. Secondly the idea of downweighting less reliable observations is used to define a robustification of the instrumental variables estimator, which is a popular method in econometrics. We derive asymptotic tests of heteroscedasticity of the errors. Finally it is examined how to use the heteroscedasticity of an uknown form to obtain a more efficient robust estimator in the linear regression model.