Expert Witness Journal Issue 66 April 2026 - Flipbook - Page 69
Statistical Evidence in Recruitment
Discrimination Cases
by Peter Crowley, Actuary & Owner of Windsor Actuarial Consultants Ltd
Employment Tribunal cases frequently involve
arguments about recruitment or promotion
statistics. claimant may argue that an observed
outcome - for example, a particular demographic
pattern among successful candidates is unlikely
to have arisen without discriminatory decisionmaking. Conversely, respondents may argue that
the outcome simply re昀氀ects the composition of the
applicant pool or normal statistical variation.
There are four useful tools in this area:
1. Recruitment Probability Curve
If a particular group represents approximately 27%
of the relevant recruitment population (for example,
among graduates achieving 昀椀rst or upper-second
degrees), recruitment outcomes might normally be
expected to cluster around that proportion, subject
to variation arising from the selection process and
other factors.
Determining which interpretation is more plausible
requires quantitative analysis. By modelling the
relationship between the applicant population, the
selection process and the recruitment outcome, it
is possible to estimate the probability of observing
particular results under di昀昀erent assumptions.
Such analysis cannot determine the legal outcome
of a case, but it can assist tribunals and legal
representatives in interpreting recruitment statistics
more rigorously. The analytical framework described
here has been implemented in the Diversity Maths
Modelling Toolkit.
EXPERT WITNESS JOURNAL
Probability modelling allows the expected range
of outcomes to be calculated. Assuming that
recruitment is totally anonymised for factors other
than degree, which avoids any unconscious bias, the
likelihood of various results is shown below:
NB, In practice, recruitment outcomes may di昀昀er from the
underlying population proportion for a variety of reasons,
including selection criteria, interview performance, or
organisational priorities.
67
APRIL 2026
