Its sort of hard to tell from your paper. My best reading of your data is that you have observations on 175 countries over 11 years from 2000 through 2010 (p.2). So its a pretty good panel - way more than enough to run a two-way fixed-effects model which would, I think, take care any selection bias concerns in at least as convincing manner as your Heckman stuff. But I'm staring at your main regression specification (equation 1, p.3) and I see no mention of country fixed-effects which is odd since you do specifically mention the year fixed effects. Then in the "SI" appendix the "outcome equation" specified in your Heckman procedure does show a country fixed effect (mu_i) term but there is no discussion of it and that equation omits the year dummies (but maybe there is a reason for that?)

Anyway, if your results are not coming out of two-way FE model, I loved to see how well they hold up when you run that way. Thanks.

Joseph Guse
Department of Economics
Washington & Lee University
gusej@wlu.edu