Using worker-firm matched data, we study the incidence and the effects of f
irm-provided training in France. When we use the same kinds of data and mod
els as the existing ones, the estimated impact of training on wages is as h
igh in France as in other countries. When we build on the uniqueness of our
French data and control for the selectivity of firms' training practices,
the estimated impact of training falls close to zero. The wage differential
s observed across trained and untrained workers reflect that individuals wh
o are the most likely to be placed in training programs are also those with
the highest unmeasured abilities, (C) 2000 Elsevier Science B.V. All right
s reserved. JEL classification: J24; J31.