I did a GMM system estiamtion as well as a Fama MacBeth (with the add in). I analyse the classical asset pricing theory having 25 portfolios with with 600 monthly obserations each, which I regress on three independent variables and a constant.
Now, I would like to test for autocorrelation, heteroscedasticity and endogeneity in the data to see which procedure is more appropriate. The results are very different as different variables are significant.
autocorrelation: durbin watson test given by standard output: 2.158
Question: Are the residals from the regression to estimate beta safed in resid (is this the last regression in the programm). Then I would use them to regress the residuals and the sqaured residuals on the independent variables to test for heteroscedasticity and endogeneity.
Used: HAC standard errors and covariance (Bartlett kernel, Newey West fixed bandwidth= 3000
GMM (time series (HAC))
Hansen J test rejects the model (.12 with 600 oservations). I used a barlett kernel bandwidth fixed nw, which should already be consistent in the presence of both heteroskedasticity and autocorrelation.
Question: What does it indicate that the results vastly different in terms of coefficients and t-tests, considering that both models should be consistent in the presence of both heteroskedasticity and autocorrelation.
Thanks for the help in advance.
For econometric discussions not necessarily related to EViews.
1 post • Page 1 of 1
Who is online
Users browsing this forum: No registered users and 4 guests