Interactions using the PPML command for the purpose of estimating a gravity model

[Chinmay Sharma]

Dear Statalist Users,


I am working with a model of bilateral trade, for which I have information on geographic characteristics of country pairs. Given that there are a host of problems using OLS type estimators (despite knowing the scedastic function), or even NLS estimators (that suffer from inefficiency), I have decided to make use of the Poisson Pseudo ML estimator, which is the state of the art in estimating gravity models nowadays.

Although in the original paper by Silva and Tenreyro, the authors recommend using the poisson command, I feel that it has now become customary to use instead the ppml command, as it presents some added benefits. There is one problem with the PPML command, however: my model contains a lot of interaction terms. However, when I type for istance

ppml y (x1)##(c.x2 i.x3)

I receive an error message "interactions not allowed". Short of writing out the interaction terms by brute force, is there any other succinct way of doing this?

Many Thanks,
CS

[Joao Santos Silva]

Dear Chinmay Sharma,

I an not sure if it works, but you can try using the xi: prefix. Alternatively, you can use the xi command to generate the variables and then run ppml.

Best wishes,

Joao

[Chinmay Sharma]

Thanks a lot, Professor. I really appreciate it.