81 Chancery Lane
0207 112 8493
View map


1 Lochrin Square
92 Fountainbridge
0207 112 8493
View map


24A Baggot Street Upper
D04 V970
0207 112 8493
View map

Registered Office

7 Bell Yard,
0207 112 8493
View map

Send us a message

Contact Us

Actuarial Modelling and Diversity & Inclusion

Modelling is one of the core practices of actuarial science, but why has it become so widespread in financial organisations?

In this article we look at applying the scientific method to evaluate the successfulness of actuarial modelling. We will go on to look at how diversity and inclusion affects actuarial modelling and whether this is for the better.

Applying the scientific method to evaluate actuarial modelling

The last few centuries of civilised society have been characterised by the rapid expansion of science, technology and their philosophical underpinnings at breakneck speeds.  This period has seen heavy debate over what science is, where its value lies and the constitution of its methodology.  A full discussion of this history is beyond the scope of this article.  We will use the formulation of the scientific method as brought forward by Karl Popper of the early 20th century. His hypothetico-deductive method is outlined in the following steps:

  1. Identify the problem area
  2. Define the problem statement
  3. Develop a testable and falsifiable hypothesis
  4. Determine appropriate measures
  5. Collect relevant data
  6. Analyse the data
  7. Interpret the data within the context of the hypothesis

Some might attempt to apply these steps to the domain of actuarial modelling, a cornerstone of actuarial science, in a similar way:

  1. Problem area: Actuarial Modelling
  2. Problem statement: To evaluate the success of actuarial modelling.
  3. Hypothesis: If actuarial modelling is successful, its most established model results will have approximate accuracy.
  4. Measures: Back test well-established actuarial model forecasts.
  5. Data collection: While beyond this article’s scope, there are many sources available for established actuarial model forecasts for things like mortality rates or interest rates spanning over decades.
  6. Data analysis: Upon analysis, many forecasts over the long term are far from real world observations. Two illustrative examples are[1]:
    1. Mortality rate projections, a classic actuarial application, have often deviated by more than 50% over several decades.
    2. Some well-known actuarial projections and probability distributions of interest rates from the 1980s and 1990s estimated the low interest rate environment seen in the years leading up to 2022 as close to a 1 in 10,000 event.
  7. Interpret data: Even the most established actuarial models lack approximate accuracy over longer time horizons, undermining the hypothesis and indicating a need for revaluation.

If actuarial predictions diverge so incredibly from observed reality, why then should the lay person have any faith in our work?  Why should they not come to the conclusion that all the efforts of actuarial science is a waste of time and money?

What went wrong here?

As has been alluded to, and is often the case in practice, the issue here lies with the hypothesis.  It should not be difficult to see why the hypothesis is flawed in the example above for those familiar with actuarial science and similar disciplines.

Formulating useful hypotheses and effective measures remain the crux of applying the scientific method.  This process demands a deep understanding of the problem domain, coupled with creativity. Unfortunately, even for those possessing these attributes, there’s no well-defined, evidence-based methodology to follow.  Application of the scientific method is an art and each new problem its own frontier.

The success and usefulness of actuarial modelling doesn’t solely hinge on the precision of its predictions – though it is often surprisingly precise over shorter periods.  Rather, its value emerges as an interpretive tool for expert decision making.

Consider the analogy of a child learning to safely cross roads, which serves to underscore this point:

A modeller must begin with a good understanding of the problem area.  The process of identifying the influencing factors needed to design a useful model backed by relevant data produces further insights leading to a deeper understanding.  The model and its results allow a decision making process which would not be nearly as well informed without them.  Even so, this approach acknowledges the eventual inevitability of the unexpected.

Humility in the approach to modelling is reflected in common actuarial practices such as clearly stated assumptions and expert judgements, independent audit and review, maintaining additional capital for extreme scenarios and regular adjustments and developments to integrate the latest knowledge and information.

Undoubtedly, certain prerequisites underpin meaningful contributions to actuarial modelling.  Numerical and analytical skills, problem area knowledge, and effective communication are typical examples.

The importance of these prerequisites can be understood when comparing the impact of an experienced doctor’s diagnosis and advice on treatment for an illness to that of an unqualified friend.  Despite the possibility of the doctor being mistaken a wise person would generally take their advice, even if in conflict with the friend, unless there were more compelling reasons not to do so.

How can diversity and inclusion help?

With prerequisites such as those above in mind, the importance of having diversity of thought and a variety of perspectives for those involved in actuarial modelling and beyond is being increasingly recognised.

Numerous factors, some yet unknown, influence phenomena being modelled by actuaries.  The ever-changing world demands constant adaptability – today’s material factors may look quite different to tomorrow’s.  While we don’t know where the next innovative and useful insight will come from, it can only increase the likelihood of its occurrence if there are competent people with diverse perspectives included and empowered to influence actuarial modelling and business decision making.

There are a variety of admirable motivating factors for an organisation to have a diverse and inclusive workforce, with strategies in place to prevent any form of discrimination.  Improving the robustness of actuarial modelling is probably not one of the first factors which come to mind, but it is encouraging to know that it is among the likely positive impacts of an active D&I strategy.

APR has a clearly defined and embedded D&I policy and strategy in place which is backed by an open and inclusive culture throughout.  These translate into a wide range of day to day actions.  There are small but consistent actions such as anonymising online tests during recruitment processes and reviewing the wording of written material with D&I considerations in mind.  To larger initiatives such as working with organisations who specialise in advocating these values, minimum D&I training requirements for all staff (which vary depending on their role) and sponsorship of these initiatives by APR leadership.

It would be bold to claim that having a diverse demographic, as measured by proportions of people with different protected characteristics, will guarantee less groupthink, more challenge and creativity in general and for actuarial modelling in particular.  The Prudential Regulation Authority has highlighted the importance of more difficult to measure factors like diversity of experience and diversity of thought in addition to demographic diversity in a recent consultation paper[2].

It is impossible to predict exactly when or where the next significant breakthrough will arise.  However, nurturing an environment and culture which maximises the probability of such things occurring is necessary for the continued success of a business.  A highly skilled and diverse group of people working on similar problems will certainly contribute to this.




Mujtaba Syed

April 2024