correlations are found between the adoption of a certain mix of practices and
performance we do not know why this occurs. We have no evidence on the nature
of any intermediary processes that need to occur to produce such relationships.
Not for nothing is this referred to as the ‘HR black box.’ Boselie et al. ( 2005 : 77 )
conWrm ‘the impression that the ‘‘linking mechanism’’ between HRM and per-
formance and themediatingeVects of key variables are largely disregarded. Indeed,
while we found plenty of acknowledgement of the existence of the ‘‘black box’’
and some speculation on its possible contents, few studies tried to look inside’
(emphasis in the original).
Looking inside the black box requires specifying an HR causal chain. At the
centre of the chain are employee attitudes and behavior and it is this which raises
the most vital question in the HR–performance debate. If ‘the distinctive feature of
HRM is its assumption that improved performance is achieved through people in
the organisation’ (Guest 1997 : 269 ), why is it that so few researchers actually study
the people: the employees and their attitudes and behavior? While Delery and
Shaw ( 2001 : 190 ) argue that ‘HRM practices and job design have the most sign-
iWcant direct inXuence on the skills, motivation and empowerment of the work-
force,’ they go on to say that ‘measuring the most important aspects of workforce
characteristics may, however, be beyond our capabilities.’ They do not say why.
Only three out of twenty-Wve studies examined by Wall and Wood ( 2005 ) and
eleven of the 104 reviewed by Boselie et al. ( 2005 ) used employee survey data.
Edwards and Wright ( 2001 : 570 ) correctly assert that ‘it remains rare for studies to
assess links in the chain, with eVects on employee commitment being a particularly
rarely studied issue.’ This absence is hard to understand, let alone justify, given the
tradition of employee-centered research in organizational behavior and industrial
relations. It stems, almost certainly, from the use of multiple-Wrm datasets where
single management respondents are only able to indicate the intended practices
and their coverage. They cannot reliably report on employee perceptions of these
practices as they experience them. Thus, the usual steps in research, of theory
determining the research questions and thence the choice of method, have
been reversed. Methodological considerations have determined what questions
can be asked while factors beyond the reach of the chosen method, however
important, have been ignored. As Guest ( 2001 : 1095 ) candidly put it, ‘almost
inevitably, both for the sake of brevity and to increase the chances of publication,
many published papers tend to play down a number of methodological and
analytical concerns.’
Wright and Nishii ( 2004 ) address this issue by proposing an elaborated model of
the HR causal chain which is divided intoWve steps, moving from intended, to
actual, to perceived HR practices, followed by employee reactions, and then
performance. This model provides an excellent basis for understanding the links
between HRM and performance, but it needs further development. First, we
examine the subsets of the casual chain in more detail. We then go on to look at
540 john purcell and nicholas kinnie