Expert Witness Journal Issue 66 April 2026 - Flipbook - Page 70
It can be seen from the horizontal axis and the
blue line that results outside the range 16 – 38 are
unlikely. However, many people would assume that
reality is the red line, ie, that large (or small) 昀椀gures
are realistic, unless other factors are substantially in
play.
formal thresholds have been passed.
If we now assume an even distribution of merit over
race (so, in the above example, any talent band
replicates the total population), we can isolate the
e昀昀ect of Factor Q. Awarding a notional number of
points to each individual (say 1 – 1,000) allows us to
rank the population, and assume every nth individual
is in our category (for our BAME example, we can
use 25% as an approximation for 27%, which is one
in four). So if, for example, we have recruitment of
45%, what is the e昀昀ect of Factor Q?
One interesting factor is that the smaller the
representation, the smaller the range of likely
outcomes, which will make it harder to conform to
“Mans昀椀eld Rule” advances.
2. Selection Mechanisms and Recruitment
Outcomes – Factor Q
An obvious way to do this is to apply a multiplier to
the marks obtained by each BAME individual, then
see what multiplier delivers the required result. For
the above example, the boost (Factor Q) required is
13%, as the illustration below shows:
Factor Q represents the in昀氀uence of attributes not
captured by academic scores — such as interview
performance, perceived potential or organisational
昀椀t — which may alter the ranking of candidates once
3. The Stubborn States Framework
This can be approached in two ways:
– Why Workforce Composition Changes Slowly
a) Single group – a Recruitment Flow Model.
Assume our feeder group is stable and consider the
behaviour of our main group. For example, consider
progress of the partnership just looking at principal
promotions, and
These programmes examine targets set for the total
workforce (eg, raise female representation from 35%
to 50%), or a subset (eg, the partnership). Here we
set assumptions for recruitment and exit/promotion
rates for each group and feeder group. Calculations
can be deterministic (ie, one set of assumptions),
or stochastic (running on many assumption setts,
possibly up to a million runs), and seeing what is
required in order to achieve those targets.
b) Multistate – a Promotion Flow Model. This
considers the changeability of all support groups.
This could be appropriate if assessing the e昀昀ect
of graduate recruitment mix in changing the
partnership.
The diagram on the next page gives a sample of type
a) run results.
For any group, the ONLY sources of change are
volumes and shares of recruits and movers – and the
constitution of feeder groups is essential in assessing
the feasibility of published or even internal targets.
EXPERT WITNESS JOURNAL
68
APRIL 2026
