PPML regression

[Nathalie Kool]

Dear Statalists,


For my masters thesis I’m running a ppml regression (dependent variable = total size of deals) and the independent variables consist of a series of bilateral variables, investor country variables and target country variables - all in log except dummy variables and the dependent one.

I’m quite surprised with the results (the coefficients are somewhat high, same for std errors and my RESET test failed). I then tried a poisson regression and the results seem a bit more coherent but still not what I expected.

I. PPML regression

II. Poisson regression

Could someone give me some guidance abt which model to use and how can I improve it?

Thanks a lot!

Nathalie

[Joao Santos Silva]

Dear Nathalie,

PPML and Poisson should give you the same results in this example. The reason why you are not getting the results is that you are not including the same regressors in the two regressions.

Best wishes,

Joao

[Nathalie Kool]

Ok thanks for your reply.

But expect this, do you think that my model is still valid?

Thanks;

Nathalie

[Joao Santos Silva]

Sorry, I cannot comment on that.

Best of luck,

Joao