Which model to use: panel data, Poisson regression or Neg binomial regression (count data)

[Harvey Jackson]

Hi everyone,

I am examining the effect of 5 macroeconomic variables on the # of deals (count variable) for 4 countries and period of 20 years. I am doubting which model I have to use. Panel data regression, Poisson regression or negative binomial regression?

Can someone explain to me which model is best and why?

Thank you!

[Maxence Morlet]

Try the user-written ppmlhdfe command

Other than that, you may want to run a two-way fixed effects regression as baseline, and then give the user-written xtscc command a go.