If he actually implemented the protocol he states he is implementing on that data, he should not get the results he does. He claims to be doing perfectly ordinary within estimator on fixed effects panel data modeling. That shouldn't turn a figure like the one you have into a figure like the one he gets.

Unless of course he tortured his model in a way that he should know (because it would trigger more than one of the several misspecification tests built into all modern econometric software packages) is wrong.

Good catch.

- Jake

Friends come and go. Enemies accumulate.