using fixed-effects or lagged dependent variable in this case?

[Alp Marta]

I use panel data for 10 years to estimate the effect of the corporate tax rate on the hourly median wage of a company. When I use fixed effects, the coefficient is - 0.41. When I use lagged dependent variable, the coefficient for tax rate becomes - 3.8%! I suspect that maybe previous wage influences mostly the change of a company's hourly wage and not tax). I have the following inquiries:
1. Is it correct to say in case of fixed-effect model: 1% increase in corporate tax rate negatively influences hourly wage and decreases its growth by 41% (I use ln of wage)?
2. How do I know which of these models give me the better coefficient for tax?

Thanks

[Carlo Lazzaro]

Marta (I do hope I'm not mistaking your surname for your given name):
welcome to this forum-
1) since all models are approximation of reality, the best approach is to give a fair and true view of the data generating process you're investigating. "Smart" coefficients and their statistical significance should be angles of your research.
2) you should not suspect, but refer to the literature in corporate finance (that I left > 30 years ago) and see if a lagged dependent variable is actually used. However, be informed that a lagged regeressand means leaving the realm of static panel regression (and, as such, -xtreg- is no more an option);
3) as per FAQ, please share what you typed nad what Stata gave you back and/or post an example/excerpt of your dataset via -dataex-. Thanks.

[Maxence Morlet]

Carlo is absolutely right, check out GMM estimation, and Kripfganz's 2017 xtdpdgmm command