Using a wide array of examples from the literature and from original estima
tes, the author examines the pitfalls that make good empirical research in
labor economics at least as much craft as statistical technique. Among the
subjects discussed are the appropriateness and cleanliness of data; problem
s of extreme observations; the validity of attempts to produce exogeneity u
sing instrumental variables, "natural experiments," and structural models;
and the treatment of selectivity and unobservable individual effects. The a
uthor stresses the importance of testing empirical results to ensure that t
hey make sense, and of presenting them clearly and economically.