st: Heteroskedasticity Test in Panel Data - Panel and Non-Panel Test

[Chris Parker <cparker.phd2007@london.edu>]

Hi Statalisters,

I have daily panel data covering many crops in many markets for a
period of several months. In this panel I want to know if a disruption
had an impact on prices in the markets. I run a fixed effects
regression of the form

. xtreg price during after, fe

where during and after are binary variables indicated whether the time
period is before or after the disruption. Before the disruption is
used as the base case. My hypothesis is that prices will be higher
during the disruption and return to normal levels after the ban which
I can test using the output from the regression. However, I am
concerned about two sources of heteroskedasticity in the error
variances. The first is along the group (crop-market) variable which I
have tested using xttest3 (see Baum (2001)) and I can reject the null
of homoskedastic error variances.

The second is along the disruption period. We suspect that the
disruption will not only impact price levels but price variation
within the disruption periods. I know I can do this type of test using
robvar for Levene (1960) or Brown and Forsythe (1974) style tests. Now
comes my two questions:

1) Are the robvar tests appropriate for panel data when the group cuts
across panels ie along the time dimension?

2) If I run robvar it will not account for the panel-level
heteroskedasticity I have observed. Is there a way I can test for
differences in error variances at the disruption level after
accounting for differences at the panel level?

And in line with the Statalist guidelines here are my full citations:

Baum, C.F. (2001), "Residual diagnostic for cross-section time series
regression models," Stata Journal 1: 101-104

Brown, M. and Forsythe, A. (1974), "Robust test for the equality of
variances," Journal of the American Statistical Association 69:
364-367

Levene, H. (1960), "Robust tests for equality of variances," In
Contributions to Probability and Statistics, ed. I. Olkin, 278-292.
Palo Alto, CA: Standford University Press

Thanks for your consideration,

Chris Parker

________________________________

PhD Candidate | Management Science & Operations
London Business School | Regent's Park | London NW1 4SA | United Kingdom
Direct line +44 (0)20 7000 8816 | Email cparker.phd2007@london.edu
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/