market and vertical hierarchy variables are used as alternate measures of the
same construct. The overall model is always statistically significant, and
joint-F tests indicate that the HRM practices jointly explain a significant
amount of the variance in perceived organizational performance. These re-
sults show that five of the six HRM practice coefficients in each equation are
positive and three (those on training, incentive compensation, and vertical
hierarchy) are statistically significant. The HRM practice coefficients are
smaller in models 2 and 3 than in model 1, suggesting that the results obtained
in analyses focusing on individual HRM practices overstate the effects. In-
deed, the decentralized decision making and grievance procedure coefficients
become insignificant when entered with the other HRM practice variables.
Table 4 presents results for the perceived market performance dependent
variable in a format identical to that used in Table 3. The Table 4 findings
are similar to those described above. For example, the coefficients on five of
the seven HRM practices are positive and significant when the variables are
entered individually into the regression (model 6). At least two of the HRM
practice coefficients retain their statistical significance in models 7 and 8.
In addition, the coefficients on the HRM practice variables are smaller when
the variables are entered together than when they are entered individually.
In general, the results in Tables 3 and 4 generally suggest that HRM practices
are positively associated with perceptions of performance, consistent with
Hypothesis 1.
Hypothesis 2 concerns the potential for complementarities among HRM
practices. The first variable we constructed to evaluate this hypothesis indicated the number of HRM practices for which a firm was above the sample