related practices) remains fairly small (4% from table 1).
Given the previous results, I selected the three employee involvement practices mainly used and
computed a set of dummy variables describing the eight possible combinations of one, two and three of
the practices including non adoption. The bottom half of table 2 shows the frequencies for the different
possible bundles. One can see that for some of the bundles like teamwork only (T) or teamwork and
job rotation (TJ) the frequencies are very low especially when the sample is divided into manufacturing
and non manufacturing. Given that the analysis hereafter will use each sample separately, a broader
definition of bundles will be more appropriate.
A broader definition of bundles can be used by using a criteria based on the number of practices
adopted. The full system is defined as a system where the three practices are adopted, the partial
system is defined as having one or two of the practices adopted. The last system is the null system
where none of the employee involvement practices are adopted. The frequencies associated with these
systems are shown at the bottom of the table. One can see that 24% of the firms adopted the full
system in 1993 and 31% in 1996. The percentage of firms under the null system did not change over
the two years while the proportion of firms under the partial system decreased from 69% in 1994 to
63% in 1996. The remainder of the analysis will focus on these systems, controlling for the other types
of workplace practices variables. Note that the variables indicating adoption of a practice or system of
practices have been created using a broad definition of adoption. If at least 1% of the workers in the firm
operate under the given practice, the establishment is considered to have adopted the practice. Since
the sample includes large and medium sized establishments, 1% may represent a reasonable minimum 6
Appendix B shows the means of all the variables used in the analysis hereafter by type of system
used (as defined above) for both years. Comparing the columns of the tables associated with the full
system to those with the null system for both years allows one to emphasize possible differences in the
average profiles of adopters and non adopters. For both years, firms that adopted the full system have
on average more educated managers and front-line workers. In terms of average wage dispersion, the
ratio of average wages for managers and production workers is greater under the full system relative
to the null system suggesting greater wage dispersion in establishments that adopted the full system.
Comparing average wages of managers and front-line workers separately, one can see that production
workers tend to have slightly higher wages in establishments that did not adopt a system of employee
involvement practices.
In summary, from the descriptive analysis, a comparison of non adopters to all-practices adopters
shows substantial differences. In particular, wage dispersion is higher in firms adopting all of the
6Appendix A.1 discusses this point and shows the average percentage of (non-managerial and non-supervisory) workers
under the given bundles of practices defined at the 1% level.
11