Applying xt commands for skewed dependent variable, 2 period panel

[Charles Bondi]

I am developing a causal model using 2 period panel data with a strong suspicion of endogeneity. xtreg, fe seems a natural choice but for the skewed dependent variable. When checking "methods and formulas" for xtreg,fe in xt.pdf p. 421 in the Stata 14 version), I see that

"xtreg, fe produces estimates by running OLS on a model with the dependent variable

(yit 􀀀 yi + y) " (in words y sub it minus panel mean + grand total mean). This latter expression has no skew at all according to summarize, detail. So is it correct to assume no skew so xtreg is potentially an appropriate algorithm, or should one assume skewness as in the dependent variable of interest to me as a researcher, and therefore look for a different xt algorithm?

[Carlo Lazzaro]

Charles:
welcome to this forum.
If you suspect endogeneity in your panel data regression model, you should consider -xtivreg-.

[Charles Bondi]

Thank you Carlo. xtivreg is on the back-burner, as choosing ivs is a separate challenge and it is on the list of things to do. In any case, there's the skewness in the dependent variable of interest to address.

[Carlo Lazzaro]

Charles:
What if you log the regressand?

[Charles Bondi]

The skew is much diminished though by going back to the formula I refer to in my first post, it disappears.