All of these studies have used the interaction approach with regression analysis. In
a study predicting turnover, Shaw et al. ( 1998 ) found that the selection ratio
interacts with valid selection techniques. The lowest discharge rate was found
whenWrms use valid selection techniques and have a low selection ratio, an eVect
that was stronger than the individual main eVects. When the selection ratio was
high, however, valid selection techniques were associated with high discharge rates.
Also, the eVect of the selection ratio (low or high) on quit rates was relatively stable
when the use of selection procedures was low. TheseWndings clearly show that
there are ‘powerful connections’ and ‘deadly combinations’ within the selection
activity area.
MoreWt-supportingWndings on intra-HRM activity areaWt come from com-
pensation research. Studies at the individual level of analysis by PfeVer and
colleagues (e.g. PfeVer and Davis-Blake 1992 ; PfeVer and Langton 1993 ) found
that certain combinations of speciWc compensation activities (e.g. pay dispersion
and pay level, pay dispersion and pay structure knowledge) interact with each other
and inXuence satisfaction as well as individual and collective performance. Certain
combinations yield ‘powerful connections’ while others are ‘deadly combinations’
and result in less satisWed and productive employees.
At the organizational level, Shaw and Gupta ( 2001 ) found that pay system
communication interacts with performance-based pay and seniority-based pay
in predicting quit rates. Pay system communication was critical when pay disper-
sion was high and signiWcantly reduced turnover. A combination of low pay
system communication, however, showed diVerent eVects. More recently, Brown
et al. ( 2003 ) investigated the interaction between pay level and pay dispersion. As
predicted, the interaction of both sets of compensation practices predicted work-
force productivity and organizational performance. SpeciWcally, the authors
observed an inverted U-shaped relationship between pay level andWnancial per-
formance that varied with the degree of pay dispersion. For example, low degrees
of pay dispersion and low pay levels as well as high pay levels and high degrees of
pay dispersion were ‘powerful connections’ and resulted in equally high levels
of organizational performance. The opposite arrangements were ‘deadly combin-
ations.’ It should be noted, however, that Brown and colleagues treated pay dis-
persion as a ‘within-job category’ variable (horizontal pay dispersion) but
operationalized it across the organizational hierarchy (vertical pay dispersion).
This may have led to spurious outcomes and misleading conclusions (see PfeVer
and Langton 1993 ).
As with inter-HRM activity areaWt, a serious shortcoming in the research on
intra-HRM activity areaWt is that nearly all studies have focused on only two levels
of abstraction (e.g. policy and practice levels). Research at the process level is
desirable since it is at this level where employees interact with HRM activities
and experience their direct impact. It seems thus likely that the synergistic eVects
found at higher levels of abstraction are even more dramatic.
hrmsystemsandtheproblemofinternalfit 399