Why is xtnbreg yielding different results than nbreg with id / year dummies?

I am using a data set of industry-year observations to calculate the effect of an IV [lobbying dollars spent by a particular industry in a particular year (continuous variable)] on a count DV [how many mergers for that industry-year were investigated, controlling for the count of how many were initiated]. By the way, I use the count DV instead of a percentage of success DV because otherwise too many observations are zero out of zero, and get dropped.

My question involves the built-in function of the xt panel data. I thought that

xtset industryid year

created a panel data set so that a panel version of the command:

xtnbreg countdv continuousiv

was equivalent to using dummies on a non-xt version of nbreg as in:

nbreg countdv continuousiv i.industryid i.year


But the results are quite different. What is the difference between using industry / year dummies and declaring panel data to use those commands?

Thanks so much!