Thailand Law Journal 2009 Spring Issue 1 Volume 12

The Wald test in the table shows that the interactions between educational attainment and training are statistically significant. Evaluated at the sample mean years of education, the estimated percentage increase in (monthly) wages following a one-unit increase in the hours of monthly training ranges from 0.0016 to 0.0025 for OJT and from 0 to 0.0024 for OFFJT. The p-values of the Wald tests, reported within brackets below the estimated effects of training on log wages, show that these effects are precisely estimated only in the case of OJT. We conclude that the marginal returns to training eW/eSOJT and eW/eSOFFJT--also evaluated at sample mean education--are increasing in training investment.

Are these returns low? To answer this question, suppose that an additional year of education represents about 1,000 hours of instruction. [FN19] This is equivalent to about 83.3 hours per month. If we increase the monthly hours of OJT and OFFJT by a similar amount, and use the results in the second column of Table 9, monthly earnings in the full sample increase by 18.3% (83.3 x 0.0022) and 7.5%, respectively. According to Psacharopoulos (1994), the private return to an additional year of education in Thailand is close to 10%, about half the size of the returns to an additional year of OJT and about the size of the returns to OFFJT. Our estimates of the returns to training are lower than those obtained for the United States by Frazis and Lowenstein (2005), who found that 60 hours of training increased wages by 3-4%, but not far from the rates of return reported for the United States by Carneiro and Heckman, which were in the range between 16% and 26% (2003). [FN20]

As for the training incidence equations, we checked to see whether our results are driven by the measurement error associated with retrospective data by replicating the fixed-effects estimates in a subsample consisting of the last two years in the sample. We found that in the shorter sample the marginal returns to OJT and OFFJT are equal to 0.0041 (p-value of the Wald test: 0.003) and 0.0021 (p-value of the test: 0.64), respectively.

Discussion

In summary, we find that the relationship between education and training is positive, as expected, for OFFJT and negative for OJT. How do we explain these results? As anticipated in the introduction, a useful starting point is the definition of equilibrium investment in training as the level of investment that equalizes the marginal costs and marginal benefits of training. Our estimates of the Mincerian earnings function suggest that the marginal returns to both types of training-- evaluated at the sample mean value of education--are increasing in the level of training. Therefore, a stable equilibrium in the training market requires that (unobserved) marginal costs also be increasing with training investment, and at a faster pace than marginal benefits.

Suppose now that educational attainment increases marginally over the sample mean. Since the marginal benefits of training are higher for the better educated, a marginal increase in education shifts the marginal benefits schedule upward. For given marginal costs, the new training equilibrium occurs at a higher training investment, which is consistent with the results for OFFJT but inconsistent with the results for OJT. For this latter type of training, our empirical results show that the new equilibrium level of investment should be lower than the original level, not higher. An equilibrium with lower training intensity (incidence) can be obtained in this framework only if the marginal costs of training also increase when educational attainment is marginally increased above the mean.

The natural reason marginal training costs increase for the better educated is that employees with higher education are more productive and thus have higher opportunity costs for the time spent receiving training. An alternative, but we believe less compelling, possible reason is that the better educated, who have higher cognitive skills, are either less willing or less able than other workers to master the practical skills associated with OJT, contrary to the view suggested by Thurow and Rosen.

Conclusions

Our analysis of data from a survey of Thai employees has shown that the relationship between education and training depends on the type of training. In particular, we found that OJT is a substitute for and OFFJT a complement to higher education. The negative relationship between educational attainment and OJT suggests that the firms in the sample were compensating for the relative scarcity of human capital among their less educated employees by investing more in on-the-job training for those employees. This compensation might have been necessary given the rapid growth of manufacturing in Thailand during the past twenty years, which has greatly outpaced the growth of educational attainment.

When individuals with higher education are systematically more likely to receive employer-provided training, the skill gap between them and less educated individuals increases, with important consequences for the earnings distribution. Widening differences in human capital can be a problem in a developing country such as Thailand, with its relatively low educational attainment and its dispersed distribution of income. Our findings suggest that OJT in Thailand partially offsets the existing differences in educational attainment.

We have explained the negative relationship between education and OJT by arguing that the marginal costs of this type of training are higher for the better educated, who are more productive than other workers and therefore have higher opportunity costs for the time spent receiving training.

It should be borne in mind that this study is limited to training in a handful of business sectors in Thai manufacturing. Whether the results apply to the Thai economy as a whole is an open question that can only be addressed using data with more generality.

[FN19]. The mean number of hours of instruction received per year by pupils older than 14 in the United States was 1,032 in 1991 (OECD 1992).

[FN20]. We do not include in the computation of the returns to OJT and OFFJT the return to tenure and labor market experience, because upward-sloping earnings profiles are not necessarily generated by investments in human capital.

This article is published with the kind permission of Kenn Ariga and Giorgio Brunello. The article originally appeared in Volume 59, Issue 4, July 2006, of the Industrial and Labor Relations Review. Copyright Cornell University.