The fact that an employer and employee incur a loss when a trained wor
ker changes jobs provides an incentive for on-the-jab training to be s
electively provided to workers who are less likely to change jobs. Con
sequently, if there is belated information about employees' future mob
ility, it may be optimal to delay training, even if doing so means for
going the returns to training during the early part of the employment
relationship. The training literature, however, assumes that training
is concentrated at the beginning of the employment relationship. The a
uthors of this paper examine the relationship between tenure and the p
robability of ever having received training using data from the Curren
t Population Survey and the National Longitudinal Surveys of Youth. Th
eir findings indicate that delayed formal training is the norm rather
than the exception.