EViews.com

Hi

I am running an OLS regression model and according to the white test there exists heteroscedasticity in my regression model. I have estimated the equation with the White option available. However when I re test the model for heteroscedasticity, according to the white test there still exists heteroscedasticity. I understand that the White adjustment makes my model more conservative by increasing standard errors, but is it enough to 'tackle' the heteroscedasticity problem or am I required to do further corrections/remedies? If yes what can be done?

I have logged data that can be logged in my model (except rates) but it hasn't helped. I tried lagging the data but they are insignificant so it's pointless lagging. Any suggestions?

I have read that it may be a sympton of other mis specifications in my model, such as non normality. However it's not possible to solve the non normality problem (and it apparently can be neglected)... So I am wondering if I can get away with just a white test?

Thanks!

What you do just corrects the standard errors (not the parameters nor the residuals) of estimation taking into account the heteroscedasticity in the data. If you are dealing with time series data and heteroscedasticity appears to be an important issue, then you can try fitting GARCH models for the "variance part".

You can find such information in a typical econometrics textbook or in any reliable online source on the subject. And you should really go over model building and specification issues before going any further.