such as a situational interview. The level at which one finally decides to take
measures has important implications that have only briefly been discussed in
the literature. Each of these measures is assessing a different staffing-related
construct. The construct ultimately measured should be chosen based on the
chosen theoretical rationale. In the literature, however, these measures are
often all described as measures of HRM practices.
Most SHRM research is focused on predicting organization- or facility-level
outcomes. Given this focus, I believe that it is important to focus on measuring
higher level systems constructs, in addition to measuring individual HRM
practices. This is particularly true given the complications in using combinations
of individual HRM practices to construct HRM system measures. Snell
and Dean (1992) provide an example of one method of assessing a higher level
HRM system construct, that of selective staffing. Instead of asking respondents
whether their organizations used cognitive ability tests, or situational
interviews and whether they had a low selection ratio, Snell and Dean (1992)
simply asked respondents to report on issues such as “How extensive is the
employee selection process for a job in this unit?” This selective staffing measure
is broader in scope than measures of individual HRM practices. This may
allow better measurement of the underlying construct than those obtained by
combining individual staffing practices, particularly unit weighted additive
combinations. Two organizations using quite different selection practices that
in each case lead to selection of highly skilled employees may receive very
different scores on a selective staffing composite created by combining individual
practices. Using the scale developed by Snell and Dean (19921, however,
they may score equally on selective staffing. I see this problem as similar to
that encountered in multi-level research when a researcher attempts to combine
individual-level measures to assess a group-level phenomenon.
This approach, however, is not without problems. One criticism of the approach
taken by Snell and Dean (1992) is that it is too subjective a measure of
staffing practices. It is clearly somewhat more subjective than the approach
used to measure HRM practices by Ichniowski et al. (1997). Before abandoning
the idea of measuring individual HRM practices, however, it is important that
researchers investigate the construct validity of these higher level systems
measures. Delery, Gupta, and Shaw (19971, for instance, presented evidence
that a measure of selective staffing adapted ,from Snell and Dean (1992) was
strongly correlated with both the number of valid selection devices employed
by an organization and their selection ratio. Firms scoring higher on selective
staffing used a greater number of valid selection devices and had lower selection
ratios. This provides initial evidence that researchers may be able to
measure higher level system characteristics directly, rather than inferring
them from individual HRM practices, some of which may be substitutes for one
another or act synergistically.
Although direct measurement of higher-level system characteristics may
provide better measures of HRM systems, researchers are still left with the
question of which level to measure. The example above is focused on sta%ng