White's heteroscedasticity test and Panel

[FOTIS]

I have a panel dataset and I want to estimate White's test for heteroscedasticity.

I run the regression and I then go to View-Residuals but there is only one option namely Histogram - Normality Test

The only way to get White's test in the options is by entering the sample using the unstructured dataset option (i.e. not panel).

Is that correct or is there another way to do it with panel?

[EViews Gareth]

We do not offer White's Heteroskedasticity test for panels.

[FOTIS]

Thanks Gareth. May I ask the following then?

When I run the test with OLS using the unstructured dataset (i.e. not specified as panel), it indicates the existence of heteroscedasticity.

Does it indicate the existence of heteroscedasticity in the panel as well? Or is it irrelevant?

[startz]

Thanks Gareth. May I ask the following then?

When I run the test with OLS using the unstructured dataset (i.e. not specified as panel), it indicates the existence of heteroscedasticity.

Does it indicate the existence of heteroscedasticity in the panel as well? Or is it irrelevant?

If you are running OLS with no fixed effects, that is ignoring the panel structure, then the unstructured test is valid. I don't think it is otherwise.

[paneltester]

Thanks Gareth. May I ask the following then?

When I run the test with OLS using the unstructured dataset (i.e. not specified as panel), it indicates the existence of heteroscedasticity.

Does it indicate the existence of heteroscedasticity in the panel as well? Or is it irrelevant?

If you are running OLS with no fixed effects, that is ignoring the panel structure, then the unstructured test is valid. I don't think it is otherwise.

Hi Eviews experts, following on from the above, I have a related question that I can't find an answer to in any of my econometrics texts.

I too, have panel data and want to test for heteroskedasticity. I'm using Eviews 5.1. My panel data has 17 cross sections which I have been representing by 17 dummy variables (and thus no constant) rather than the built-in fixed effects option.

In seeking to test for heteroskedasticity, it seems the only way to use the Eviews built-in tests is to, as suggested above, run an OLS model on the unstructured dataset (ie. as stacked data rather than as a pooled time-series cross section model) and apply the Eviews White's Test that way. Does your advice above imply that I should exclude the set of 17 dummy variables in the OLS model being used for the White's Test? Thanks

[assaf]

Please advice if I found an structural break point , do you recommend to separate the regression over two periods ? Does it apply on fixed effect regressions ? I am applying the regression to obtain the residuals not to estimate the coefficients and their significance(although it is important to do so and I will of course ).

[alfred002]

I don’t think that EVIEW offer White's Heteroskedasticity test for panels. The only way I could do the White's Heteroskedasticity test was by using the unstructured dataset option so, it does not concern about the panel. I hope it will be solved in the coming versions of EVIEW.