Dear Statalist,

I have a panel of firms where I estimated the effect of the introduction of a patent law reform on the stock of patents possessed by the firm. The specification is

yit = beta treated x treatment + firm fe + year fe + e

where treated is a dummy identifying the treated firms and treatment is another dummy identifying the period over which the treatment (the reform) was active.
I got positive and significant results, but then I wanted to transform the dependent variable in logarithm log(1+y) in order to have an easier interpretation of the coefficient. Results are still significant but the sign is now negative.
Might it depend from the fact that 40% of values of the dependent variable are zeros? I thought I could try with a tobit, but then I don't know how to deal with fixed effects. Is it a good idea or anybody has other suggestions for how treating this problem of log transformation?

Thank you for your time.

Luca Gallorini