st: When to use Poisson or Negative Binomial

["Querze, Alana Renee" <arq@ku.edu>]

I am trying to figure out how to justify the use of either

xtpqml dvar ivar, fe

or

xtnbreg dvar ivar, fe

where my data is overdispersed. 

Cameron and Trivedi write that:

"the negative binomial has the attraction that, unlike Poisson, the estimator is designed to explicitly handle overdispersion... this may lead to improved efficiency in estimation and a default estimate of the VCE that should be much closer to the cluster-robust estimate of the VCE... at the same time, the Poisson panel estimators rely on weaker distributional assumptions˗˗essentially, correct specification of the mean˗˗and it may be more robust to use the Poisson panel estimators with cluster-robust standard errors" (Cameron and Trivedi, 2010, p.641).

But I don't know whether it is better to sacrifice robustness or efficiency.

Anyone know how to justify the use of one over the other? (BTW I have just under 400 districts and 52 months in my panel data).

Best,
Alana

*
*   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/