included. Specification II shows the results when specification I is augmented by technology controls and
specification III adds workplace organization controls. In specification IV, the employee involvement
system dummies are interacted with the variable indicating whether production workers received training
about problem-solving meetings and teamwork during the last three years. Specification V adds a
dummy for the use of profit sharing or stock options. The results for each specification are summarized
into two columns showing the estimated coefficients and the value of the χ2 statistic from a test of
equality between the estimated log wage ratios. Finally, each table is further divided into two panels
each showing the results from the regressions applied to the samples of manufacturing (upper panel)
and non manufacturing firms (lower panel).
The first two columns of table 5 show the estimated managerial wage premia associated to each
system (null, partial and full) with controls for workforce characteristics. Notice that compared to the
raw averages presented in appendix B, the inclusion of workforce characteristics reduces substantially
the estimated gap with an estimated 33% managerial premium in the manufacturing sector among
establishment that did not adopt employee involvement practices and 34% in the non manufacturing
sector. The estimated premia varies across systems and specifications and remain strongly significant.
Note that it is not always the case on the non manufacturing sector but this may be explained by the
much smaller sample size of non manufacturing firms.
To better illustrate the differences in the wage premia across systems, figures 1 and 2 below graph
the estimated managerial wage premia by type of employee involvement system adopted for the year
1996 for the manufacturing and non manufacturing sectors respectively. From figure 1 describing the
manufacturing sector case, one can see that the systems of partial use of the involvement practices (when
only one or two of the three practices is used) and the full systems are both associated with greater
wage dispersion relative to the null system (none of the practices adopted) The additional controls do
not change these results.
Going back to table 5 for the manufacturing sector, one can see that the differences in the premia
are significant between the null system and the partial system as well as the full system. Indeed, for the
difference between the null system and the partial system, the value of the statistics in specifications
IV or V (which controls for stock option and profit sharing) are equal to 6.44 and 6.59 which implies
that equality in the estimated wage premia between establishments with no system and those with a
partial system is rejected at the 99% confidence level.
Figure 2 illustrates the non manufacturing case. The results are similar to the manufacturing sector
in that adoption of an employee involvement system (whether full or partial) is associated with greater
wage dispersion. The statistics of the tests of equality between wage premia across systems in the second
panel of table 5 show that there are significant differences between the managerial wage premia under
14