Dear statalisters

This is my first post, but I have followed the forum for years so I hope that I follow the FAQ.

I am currently estimating a model for 47 US states for the period 1950 to 2016. The dependent variable are fiscal variables like general expenditures and taxes for the individual states.

As most people in the litterature use a fixed effects estimator, sometimes with year dummies I am doing the same. As I have found that the year dummies are significant I did a two-way fixed effects estimator, utilizing the user-written program reghdfe by Correia (2017).

My models essentially looks like this

reghdfe y x1 + additional controls, absorb(state year) cluster(state)

I cluster the standard errors because of serial correlation in the error term.

My question however is, if it might be better to utilize a model as xtpcse that use panel corrected standard errors as proposed by Beck and Katz (1995)? If this might be better, how should I then take care of the state and year fixed effects. I fear that by including both year and state dummies that I might be exhausting the degrees of freedom of the model.

I know that cross sectional dependence might also be a problem, something that is not taken care of in my model, and is rarely questioned in the litterature. That is primarily why I havent focused on it either.

Any help is highly appreciated.