Page **1** of **1**

### Autocorrelation and heteroskedasticity in Panel Data

Posted: **Sun Mar 18, 2018 11:22 pm**

by **NimUnited**

I am using Eviews 9, and I am having some troubles with panel data. I have a panel data set of 400 Companies with observations from 2007-2016. It is almost balanced, but not totally, i.e. I have some missing observations now and then.

I have 2 questions:

1. How do I test for autocorrelation and heteroskedasticity in my dataset? I can't seem to find the same tests as with cross-sectional.

2. If autocorrelation and heteroskedasticity exists, how do make my standarderrors robust to it?

I have been told, that Eviews can't handle the above questions, and that I should use Stata, but I think that sounds a bit weird? However I do not seem to find any guides on how to deal with it on Eviews online-forum.

When I do my panel regressions I can choose White Period standarderrors, but I am not sure that they do the job I want them to do in this particular example? (question 2), and it does not say anything about whether autocorrelation and heteroskedasticity actually exists (question 1).

I hope someone can help me.

Best regards.

### Re: Autocorrelation and heteroskedasticity in Panel Data

Posted: **Tue Mar 20, 2018 11:12 am**

by **NimUnited**

Do anyone has an answer for this? I truly hope to get some help.

Best regards

### Re: Autocorrelation and heteroskedasticity in Panel Data

Posted: **Thu Apr 19, 2018 9:45 am**

by **sarchi**

Hi,

That is my question too. Can anybody help us?

I use EViews 10, but in this version ,is there any possibility for white test. I want to solve Hetroscedastisty Problem with white, but Cent find it.

Is it enough , if we choose white cross-section In Equastion Estimation, in Panel Option ? or should I choose fix Effect in Effect spesification too ?

Im looking for any answer please!!!

Thank you

### Re: Autocorrelation and heteroskedasticity in Panel Data

Posted: **Thu Apr 19, 2018 9:54 am**

by **EViews Gareth**

The standard introductory approaches to autocorrelation and heteroskedasticity don't apply in panel data. Thus there is no "White test" or simple "White standard errors". The idea of heteroskedaticity is more complicated in a panel setting - is the covariance between errors within cross-sections only, within period only, or a mixture?

Thus the options for testing and adjusting covariances are different. You have to figure out what you want to correct for, then choose the appropriate covariance option.

### Re: Autocorrelation and heteroskedasticity in Panel Data

Posted: **Thu Apr 19, 2018 10:25 am**

by **sarchi**

Thanks for your answer,

I have workfile mentioned below. Durbin-watson stat is near zero. That means there is Heteroscedastisity here. What can I do to handle Heteroscedastisity? is it enough to Choos "white cross- section" in "Coef covariance method " ?

Thank you so much

### Re: Autocorrelation and heteroskedasticity in Panel Data

Posted: **Thu Apr 19, 2018 11:06 am**

by **sarchi**

white diagonal means correct both cross-sections and within period ?

### Re: Autocorrelation and heteroskedasticity in Panel Data

Posted: **Thu Apr 19, 2018 11:53 am**

by **EViews Gareth**

### Re: Autocorrelation and heteroskedasticity in Panel Data

Posted: **Thu Apr 19, 2018 12:46 pm**

by **sarchi**

thank you very much. that was very helpful.

but now i have another question.

"Ignoring cross-sectional dependence in estimation can have serious consequences, with unaccounted for residual dependence resulting in estimator efficiency loss and invalid test statistics."

what should I do, wenn I have cross-section Dependance in my Panel data ?

in EViews Help just says how can we test it. But did not say how can we solve it.

Thanks a lot

### Re: Autocorrelation and heteroskedasticity in Panel Data

Posted: **Fri Apr 20, 2018 12:39 pm**

by **NimUnited**

Hi Gareth,

We have clustering in both cross-section and in time. We use yearly time-dummies to fix the time-clustering. Would you recommend White Period to fix cross-section clustering?

Thanks in advance.

### Re: Autocorrelation and heteroskedasticity in Panel Data

Posted: **Fri Apr 20, 2018 12:45 pm**

by **EViews Gareth**

I do not give econometric advice