hiring individuals who perform suYciently well so as to add unique value to the
system, but who are not necessarily the top applicants, thereby reducing selection
costs.
Another interesting pattern in the current literature is the diVerential growth of
team composition and ‘compilation’ models. Composition models seek to explain
the aggregation of lower-level behaviors that are similar while compilation models
seek to explain group-level phenomena that result from specialized individual
behaviors (Kozlowski and Klein 2000 ). As Chan’s ( 1998 ) typology of aggregation
models suggests, compilation models require complex theoretical justiWcations for
aggregation rather than simple demonstrations of agreement or similarity, as do
most composition models. The consequence of this distinction is that we still lack
standard methods for combining performance behaviors when each person in a
team or organization has a specialized function. The problem becomes further
complicated when team membership is continually shifting and the selection of
a new member depends on the current team composition/compilation.
Regardless of whether aggregation occurs through composition or compilation,
it is always important to establish that individual-level predictors of performance
will also predict higher-level outcomes, over which the individual has some
control. Ployhart and Schneider ( 2002 a) elaborated on the classic validity model
depicted by Binning and Barrett ( 1989 ) to show how the validity of individual-level
predictors translates into validity at multiple, higher levels.
Despite the progress made by multilevel theorists and the evident costs of failing
to consider relevant issues, it appears that future researchers will encounter diY-
culties in deWning and measuring tasks and KSAs that have meaning across multiple
levels of organizational structures (Schmitt 2002 ). The necessary but daunting task
of weighting predictors across levels will also be an obstacle to the development of
appropriate selection systems (Ployhart and Schneider 2002 b). Establishing validity
at multiple levels also carries with it practical concerns about how toWnd suY-
ciently large sample sizes that contain variance (e.g. at the plant level).
- 5 Deviance and Counterproductivity
Given the tradition of pathology and dysfunction in psychology, it is somewhat
surprising that critical examinations of organizational deviance and counterpro-
ductive work behaviors began just over a decade ago. The ‘dark side behaviors,’ as
GriYn and O’Leary-Kelly ( 2004 ) label them, range from the mildly annoying to the
criminal, but often pose serious threats to an organization’s resources and prod-
uctivity, social system, or public image even when the base rate is low (Harris and
Ogbonna 2002 ). As such, organizations should be concerned not only with select-
ing productive employees, but also with selecting out employees who will harm the
organization.
selection decision-making 311