In this paper we use the overtaking method to calculate marginal rates
of return to on-the-job training. As we have no prior information on
who invests in OJT and who does not, we use a switching regression mod
el with unknown regimes to obtain parameter estimates of the wage prof
iles. The estimation results show that lower intercepts are associated
with steeper wage profiles. This confirms the prediction that those w
ho invest in OJT have lower initial wages but higher wage growth. We f
urther find that the marginal rate of return to OJT is high, and incre
ases with education.