This week I have been running regressions on the determinants of industrial location in 1920 Brazil. The first results are quite promising.
Before starting any an applied paper, however, I think it is healthy to read Peter Kennedy's "Oh no! I got the wrong sign! What should I do? and The Ten Commandments of Applied Econometrics. If I had learned his lessons when I was younger, maybe I wouldn't have sinned as much as I did. I strongly recommend grad students to read the paper before applying the last rocket-science technique as well.
Andrew Gelman provides further comments on Kennedy's papers.