Table 8. Tobit Estimates of Training Intensity, Production Workers Only,
1998-2001. (dependent variables: OJT and OFFJT intensity; bootstrapped standard errors [50 replications] in parentheses)
-------------------------------------------------------------------------------
Years of Education Treated as Testing for Exogeneity of
Exogenous Education
----------------------------- ------------------------------
Variable OJT OFFJT OJT OFFJT
-------------------------------------------------------------------------------
Sex 0.590 0.208 0.509 -0.055
(.531) (.205) (1.125) (.405)
Age 0.483 0.625** 0.313 0.371
(.664) (.313) (1.163) (.407)
Age2 -0.018 -0.019** -0.013 -0.012
(.020) (.009) (.034) (.013)
Age3 * 10 0.002 0.001** 0.001 0.001
(.000) (.000) (.002) (.001)
Tenure in 1998 -0.012 0.017 -0.022 0.014
(.096) (.028) (.097) (.024)
Previous -0.089 -0.007 -0.078 -0.012
Experience in
1998
(.067) (.023) (.096) (.021)
Education -0.618*** 0.094*** -0.627 0.272
(.090) (.031) (.507) (.211)
First Step -- -- -0.011 -0.181
Residuals
(.526) (.212)
Observations 3,466 3,508 3,432 3,472
-------------------------------------------------------------------------------
Notes: See notes to Table 5.
Our estimates use four years of retrospective data. A drawback of these data is that they introduce measurement error, as respondents may suffer from recollection problems, which are expected to increase with the period of time between the training spell and the interview and with the detail of the training questions (see Leuven 2005). To minimize the influence of these measurement errors and check the robustness of our results, we replicate Tables 5 and 6 in a subsample that retains only the observations for the last year (2001), and find that the sign of the relationship between years of education and type of training is confirmed. In a further effort to reduce measurement error, we re-define training incidence by allowing it to take a value of one if any training occurred during the period 1998-2001 and zero otherwise, and we re-define training intensity as average hours of training over the sample period. Again, our statistical results are qualitatively unchanged. [FN17] |