MacD~ffle (1995) although the grouping of practices was derived based on
theoretical rationales rather than factor analysis. What these studies have in
common is that they computed measures of “systems” of HRM practices by
grouping practices together and then taking the average or sum of those practices.
These scales or indexes were then used to estimate the effect of the HR
system on performance.
Researchers in this area have begun making distinctions between and
ac~owled~ng the differing assumptions of scales and indexes. As DeVellis
(1991, p. 9) pointed out, using scales assumes the responses to items are
“caused by an underlying construct.” An index, however, is made up of items
that “determine the level of a construct.” The importance of this distinction is
apparent when one investigates how these combinations of practices, either
scales or indexes, are to be used. If one assumes that the HRM practices are
caused by an underlying construct and are equivalent measures of this construct,
they should have relatively high intercorrelations. In this case, Cronbath’s
alpha is an appropriate indicator of the reliability of such a measure. On
the other hand, if one assumes that these practices determine the level of a
construct (e.g., work force skills), then they should be combined in a linear
fashion, weighted by their influence on the construct and they may not be
highly correlated. In a very real sense, the researcher is creating a regression
equation and computing a predicted level of the outcome construct. Under
these circumstances, it is not assumed or expected that the practices should
have high intercorrelations and Cronbach’s alpha is not appropriate.
Huselid (1995) provides an example of how factor analysis has been used in
SHRM research. In that study, Huselid (1995) conducted an exploratory factor
analysis of HRM practices that resulted in the extraction of two factors. For
example, one of the factors was labeled employee motivation. One could argue
that this factor should be interpreted as an employer’s concern for employee
motivation and it is this concern that led to the adoption of the practices
loading on this factor. In fact, that the practices load together on this factor,
suggests that organizations adopt these practices simultaneously Huselid
(1995, p. 6471, however, stated that that these practices are “designed to recognize
and reinforce desired employee behaviors.” This seems to be an argument
more in line with treating this group of practices as an index rather than a
scale. Youndt et al. (1996), however, explicitly combined measures of practices
into indexes based on prior theory that indicated how the practices influenced
firm outcomes. This is apparent in their labeling of one of the indexes, “humancapital-enhancing.”
They stated that “such an additive approach to combining
HR practices into an index suggests that firms can improve performance either
by increasing the number of practices they employ within the system or by
using the practices in an HR system in a more comprehensive and widespread
approach” (p. 849). What they really argued here is that HRM practices can
influence firm performance by increasing the human capital of the firm. Thus,
the index they created could be interpreted as a predicted level of human
capital in the firm. A measure of internal consistency, like that reported in the