Dear all,

I am running a PPML regression for a trade gravity model using bilateral trade flows and my stata code is as follows:

ppml Tradeflow i.year i.pair i.Exporter i.Importer BorderYear1 BorderYear2 BorderYear3 ,nocons cluster (dist)

Now if I want to see whether the border measure has increased between the first and last year then, do I test whether their difference (i.e. BorderYear3 - BorderYear1) is significant or whether their ratio (i.e. BorderYear3 / BorderYear1) is significant? I am confused because from my understanding, ppml estimates the exponentiated value of the coefficients so wouldn’t it be correct to check the statistical difference in the ratios instead of the differences?

Thanks in advance