This article compares various measures of on-the-job training, from a
new source that matches establishments and workers, allowing us to com
pare the responses of employers and employees to identical training qu
estions. Establishments report 25% more hours of training than do work
ers, although workers and establishments report similar incidence rate
s of training. Both establishment and worker measures agree that there
is much more informal training than formal training. Further, informa
l training is measured about as accurately as formal training. Finally
, we show that measurement error reduces substantially the observed ef
fect of training, in particular the effect of training on productivity
growth.