Nevertheless, all studies using cluster analysis support Arthur’s ( 1992 )Wnding that
diVerent organizations have distinct, theoretically internally consistent HRM
activities, supporting indirectly the notion of internalWt. With a similar theoretical
approach, applying pattern analysis, Verburg et al. ( 2004 ) concluded thatWt per se
was not associated with superior performance. Instead, certain types of HRM
systems were related to better organizational performance, aWnding aligned with
previous research (e.g. Arthur 1992 ; Becker and Huselid 1998 ).
Several empirical studies have used factor analytic techniques and scales or
indices as measures of HRM systems (e.g. De Saa ́-Pe ́rez and Garcı ́a-Falco ́n 2002 ;
Huselid 1995 ; MacDuYe 1995 ; Vandenberg et al. 1999 ; Zacharatos et al. 2005 ),
possibly due to the familiarity and acceptance of these techniques in the social
sciences. While the studies reviewed by Delery ( 1998 ) only employed exploratory
factor analysis, more recent studies have used a conWrmatory approach. The paper
by Zacharatos et al. ( 2005 ) is most notable. In their Wrst study, the authors
measured ten high-performance HRM practices with several items on Likert-type
scales. Fit measures supported their prediction and all parameter estimates were
signiWcant. In a second study, using almost identical scales and items, Zacharatos
et al. ( 2005 ) further showed that their high-performance HRM system had con-
vergent and discriminant validity when aVective commitment was included in
a two-factor model; both factors were positively correlated as theory would predict
(Arthur 1992 ; Meyer and Smith 2000 ).
Taken together, the results of all the above cited studies clearly demonstrate that
certain HRM activitiesWt with each other and form a coherent ‘bundle.’ However,
all of the studies only investigatedWt at the policy and practice levels. Some
studies even combine the policy and practice levels, which is possibly a serious
shortcoming. Further, although most studies conclude that their results show
coherent HRM systems, they really show only that organizations have internally
aligned HRM policies and/or practices. No study actually looks at theWtofan
entire HRM system, theWt between diVerent HRM activity areas across several
levels of abstraction, or even theWt of a complete HRM architecture. Nevertheless,
all of these studies provide some support for the notion of inter-HRM activity
areaWt.
Another shortcoming of these studies is that they do not necessarily account for
synergistic eVects, one of the hallmarks of internalWt. Hence, while the studies
show support for the presence of ‘bundles’ of HRM activities, they do not prove the
existence of ‘powerful connections’ or ‘deadly combinations.’ The interaction
approach using regression analysis allows for a better test of synergistic eVects,
although it only allows for testingWt between a limited number of practices.
Adopting this approach, Delery et al. ( 1997 ) used broadly measured HRM practices
(e.g. selective staYng, performance-based pay, and participative decision-making)
and performed multiple two- and three-way interactions that unmistakably sup-
port the notion ofWtandsynergies.
hrm systems and the problem of internal fit 397