10
A test to determine whether or not the tenure effect was linear was also reported in Baker et al. (1994b;
3rd last line of p.935,.
11
These coefficients are also highly non-linear, just as Baker et al.(1994b) argued.
24
workers with the same number of years tenure, but who had entered the firm at different times, earned different wages. Again the cohort dummies’ exclusion F test value was 15.2, which also provides evidence to support the existence of the cohort effect.
Now we can apply the same interpretation above to investigate all of the six type-pay matches. First of all, we find that cohort dummies are significant determinants for all types of workers and pay variables. The respective F statistics for the six type-pay matches (b against a) are 24.35, 3.87, 9.62, 2.41, 21.38 and 9.04. All of these passed the critical F exclusion test value at the 1 per cent level.12 The large cohort dummy exclusion F test values for specification (d) and (e) also verify the existence, importance and non-linearity of the cohort effect. Baker et al. (1994b) were concerned that the cohort effect might simply be due to changes in the composition of entrants, thus they ran the regression on entry wages only with the human capital and year dummies. In their test for whether the year dummies represented a set of meaningful dependant variables they concluded that this was not the case, since they found a 22.97 F-statistic.We ran the same specification, and found that the respective F statistics for the six type-pay matches were 17.83, 11.25, 12.21, 26.41, 4.90 and 14.61, which supports the Baker et al. (1994b) findings.
Another interesting issue is the determination of whether a limited number of tenure variables is sufficient to pick up all of the tenure dummy effects. We can see that for staff workers, linear tenure, or the functional form which contains its square, cubic and quartic terms, provides a sufficient approximation of the tenure effect. The same conclusion can be drawn for the salaries of both technicians and salespeople, but not for their bonuses. This finding contradicts the observations of Baker et al. (1994b), that linear tenure is a good proxy for the tenure effect. Finally, by comparing the size and significance of the coefficients, we can see that there is a weaker cohort effect amongst salespeople that amongst other workers. This is consistent with the evidence provided earlier that salespeople have higher rates of demotion, and that most of their compensation comes from bonuses, which ties them more directly to their market value. Overall, we have confirmed that there is strong evidence supporting the existence of the cohort effect, similar to many of the findings in the prior literature, and that this effect is not driven by the composition of the entrants.
12
The 1 per cent critical value of F (10, ∞ )=2.32.
25
7. CONCLUSIONS
This study has presented an analysis of the internal labor economics of a Taiwanese auto dealer, Company X, comprising of three different sub-internal labor markets, salespeople, technicians and staff workers. We have offered evidence on ILMs from the perspective of another culture, and have therefore attempted to render these stylized facts more reliable and universal. Our empirical results are similar to those of Baker et al. (1994a, b), providing mixed evidence on Doeringer and Piore (1971).
Using job title, authority level and the structural hierarchy of the company to identify the various levels, we find significant differences in the employment transition paths across the different types of workers. We argue that the effects of jobs and levels, on both the salary and the bonus equations, are positive and smaller under a fixed effects model than under an OLS (combined) model; however, when adding in individual fixed effects, the reduction is greater in the bonus equations than in the salary equations. With changes in economic conditions, greater variations occur in bonuses than in salaries, and higher ranking employees feel the effects of these variations more than lower level staff workers and salespeople. Wage variations do exist within and between levels, and they are greater for bonuses than for salaries
.
Furthermore, the variations for both salaries and bonuses, defined by the coeffficient variations, are greater in the years when demand is high than in years of low demand.Entry and exit is observed at all levels, but this is more likely to occur in the lower levels of the hierarchy. We have also verified the existence of the cohort effect, and find that it is not driven by the composition of the entrants.Our evidence shows that the external and internal markets are both working, and that ILMs can only partially shield workers. Furthermore, a very interesting finding is that although this is a Taiwanese company, heavily influenced by Japanese culture, most of the findings corraborte the earlier literature, in terms of the broad patterns found in US and European firms, which suggests that ILMs work in a more general setting.
The dataset has demonstrated a number of interesting comparisons with the stylized facts of ILMs. Through the concepts of the internal hierarchy, levels, ports of entry and exit, and the relationship between salaries, bonuses, jobs and levels, we have verified the existence of sets of rules and procedures that effectively define the internal labor markets. It is hoped that the findings of this study will provide some contribution to the ongoing quest of many economists for the opening up of ‘the black box of the internal economics of the firm’.