Testing for Heteroscedasticity in Panel Data

[MattvanPallen]

Hi Again!

Hope this is the last time I'm forced to bother you, as the sas help doc is for me

I wanted to test for heteroscedasticity in my panel data sample and eventually correct it.

My model:

FE=FELO FEUP FE_L FELOWER FEUPER;

FELO - dummy (1 if below FE_L Q1)

FEUP - dummy (1 if above FE_L Q1)

FE_L - laged FE

FELOWER = FELO*FE_L

FEUPER =FEUP*FE_L

On SAS Help I can only find the example with "incomes" which is not as complicated as my...

I would really appreciate your time and concern.

[ets_kps · Accepted solution]

Hello and thanks for the question. 

There are two parts to your question.  If you want to conduct a Breusch-Pagan test which is requested as an option on the MODEL statement in PROC PANEL. 

SAS/ETS(R) 13.1 User's Guide

The second, and frankly what I would do, is to just assume you have heteroskedasticity and autocorrelation and correct for it using the HAC=NEWEYWEST option on the MODEL statement.SAS/ETS(R) 13.1 User's Guide

If you wanted to employ simply Heteroskedastic-corrected se's you can do that here.

SAS/ETS(R) 13.1 User's Guide

Good luck and let us know if you need anything else. -Ken

[rohitg]

Hi,

I was trying to follow your steps using SAS 9.4, but it is not working: I need to test if there is any hetroscedasticity and also how I can correct it.

proc panel data=mydata;

model y=x1 x2 x3 x4/RANTWO HAC=NEWEYWEST;

id firm time;run;

Please help!

..Rohit

[SteveDenham]

In 9.4 (SAS/ETS 14.1), the specification is now:

HAC(neweywest94)

or

NEWEYWEST (parameters inserted for gamma, rate and constant, if you want something other than the default)

Hope this helps.

Steve Denham

[rohitg]

Hey, Thanks for your reply! I have access to 9.4 (SAS/ETS 13.2).

To correct  heteroscedasticity, I have succesfully tried and used:

proc panel data=mydata;

model y=x1 x2 x3 x4/RANTWO HAC(BANDWIDTH=NEWEYWEST94);

id firm time;run;

Could you please suggest some way through which I can check for hetroscedsaticty in my panel data before looking for any correction ?

Thanks.

Rohit