like/dislike, similarity of outlook, or
in-group/out-group effects, because this
information is not routinely recorded, and
often it is not readily accessible.
This means that many sources of bias
render performance appraisal ratings less
accurate, but in ways which neither employer
nor employee can easily identify. In technical
terms, the performance ratings are contaminated, and the contamination is invisible.
This in turn has further implications for
research on personnel selection. Between
60 per cent and 80 per cent of researches on
selection use supervisor rating as their criterion of “true performance”. Bias in ratings,
from the sources described above, means that
rating may often be a fairly poor criterion of
true performance, which in turn implies that
much research may actually underestimate
the true validity of selection methods.