Reexamining the Returns to Training: Functional Form, Magnitude, and Interpretation

We investigate the functional form for formal training in a wage equation and derive estimates of its rate of return. The cube root fits best in our two data sets. We show that if wages are not adjusted continuously, estimating the return to training requires one lag and one lead of training. Using the cube root and a semi-nonparametric estimator, estimated returns are 150–180 percent. Adjusting for heterogeneity in wage growth, promotions, and direct costs reduces the return to 40–50 percent. We find evidence of heterogeneity in returns. Our estimates can thus be regarded as the return to training for the trained, but cannot be extrapolated to the untrained.