shown in appendix Table B2.2 with a higher proportion of blue-collar moving from the low
to the middle level and a higher proportion of white-collars moving from the middle to upper
level, the different impacts of measured and unmeasured skills at different levels of the job hier-
archy suggest that measured skills would be more important in the assignment of blue-collared
workers whereas unmeasured ability would be more important for white-collared workers.
The second panel of Table 4 show the estimation results when learning about unobserved
innate ability is introduced. One can see that assuming that mobility is generated by learning
about unobserved ability changes substantially the preceding results. Overall, the coefficients
are less precisely estimated with most of the standard errors doubling in magnitude. None of
the tests of equality in the slope coefficients reject the null implying no evidence of comparative
advantage.
These results cast some doubt on the ability of the learning hypothesis to be supported by
the data. Note that with the introduction of learning, the pure rank effects cease to be signif-
icant. Taken in isolation this result would suggest that mobility generated by learning about
unobserved ability explains all of the rank effects in the wage dynamics. On the other hand, it
is difficult to reconcile with the absence of evidence on the workers’ comparative advantage in
a given rank. A better explanation for this result would be that mobility of German workers
across ranks is not important enough to identify any differential rank effects, either pure rank
effects or differential skills and ability effects across ranks. This result is consistent with Bauer
and Haisken-Denew (2001) who use the same data to analyze the covariance structure of wages
resulting from learning about workers’ unobserved ability and do not find evidence of learning
effects in the estimated covariance structure.
Concerning the estimation of the human capital ratio, in both specifications, the results
correspond to the estimation of a ratio defined as a constant b0 . Estimations based on the
functional form given in (14), either using experience or tenure with the firm, all lead to an
estimated ratio close to 1 with the b0 estimate varying between 0.99 and 1.01 and the b1
estimate of .001 suggesting the ratio is independent of experience or tenure. 27 For that reason
I re-estimated the model focusing on the estimation of the constant term in (14) and present
the results associated with a simpler functional form defined as the constant b0 .
From the two panels of Table 4, one can see that the ratio is significantly different from zero
but not significantly different from unity. As mentioned previously, a ratio of unity implies
that the function of accumulation of human capital (proxied by years of experience) is constant
27Results available upon request.
21