This variable is also not statistically different

This variable is also not statistically different
from zero, with the exception of the benefits
administration equation. Again, therefore,
strategy—or at least these measures—seems
not related to HRM expenditures across
firms.
The third strategy-related question con-
cerns the major performance criterion on
which the HRM function is evaluated. This
variable may also reflect the employee rela-
tions philosophy of the firm and its top
management. If the performance goal is
cost-containment, for example, we hypothe-
size that less is spent on per capita HRM
whereas if it is employee morale and satisfac-
tion, then a larger amount is spent. To test
this, we enter a dummy variable for those
respondents who chose the criterion goal
“employee morale and satisfaction.” 29 Our
hypothesis is broadly supported. The perfor-
mance variable has statistical significance of
varying degrees in six regressions, including
the Total Expenditure equation. Interest-
ingly, the coefficient is negative and signifi-
cant in the Compensation sub-function
regression. This result may be a statistical
quirk; alternatively, it is congruent with the
prediction of satisfaction/hygiene theory
that firms find non-wage measures to be a
more effective means to boost morale and
satisfaction.
The next variable drawn from the BNA
survey is the percentage of employees in the
reporting unit represented by a union. It is
not statistically different from zero in all ten
regressions, indicating that in this data set
per capita expenditure on HRM does not
29
The question in the BNA survey allows respondents to
circle two of the five possible answers; thus using a set of
(say) four categorical dummy variables is precluded
since some observations take more than one value. The
results for all three strategy variables are problematic to
the extent that the firm’s HRM expenditure level influ-
enced the survey respondent’s choice in answering the
firm’s strategic involvement level.
show discernible variation with respect to
union status. 30 One could well expect that
for the Employee Relations and Safety and
Health sub-functions this variable would be
positive, which it is in both cases, but not sig-
nificantly so.
The survey also categorized firms into
different broad industries/sectors: manu-
facturing, non-manufacturing or service,
government, and non-profit. We treated
manufacturing as the excluded category and
created dummy variables for the other three.
In the aggregate HRM equation, per capita
HRM expenditures are lower in services (rel-
ative to manufacturing) and, using a ten-
percent significance test, also lower in the
government sector but not in the non-profit
sector. At the sub-function level, per capita
expenditure on Compensation and Benefits
Administration is lower in Service sector
firms and, for Compensation, also lower in
non-profit firms (p  .10). Recruitment ex-
penditure per employee is higher in non-
profit firms (p  .10), however.