Can “traditional” actuarial methods be sexy?

There’s plenty of discussion going on amongst the profession these days about machine learning, artificial intelligence and applying these methods and techniques for predictive analytics. It’s clearly an area of growth for data-driven professionals such as actuaries, and of course we all stand to benefit from adding these emerging capabilities into our technical toolkit.

I do sometimes feel though that implicit in this excitement is an “un-sexiness” associated with traditional actuarial methods. More importantly, I think that this runs the risk of missing opportunities for actuaries to be relevant in other ways. If I step back and consider what traditional actuarial methods are trying to do, I would say that broadly speaking they focus upon:

• Managing volatility – The methods are trying to take inherently volatile outcomes and analyse them in a way that allows for patterns to be identified that allows an outcome to be reliably projected in order to support management and strategic decision making
• Observation – The methods are constructed such that the key known and observable drivers of volatility are accounted for in the modelling, without a reliance upon highly granular data
• Cutting through the noise – The primary objective is to form a reliable view at an aggregate level (be it product, portfolio or claim segment), rather than an individual outcome. The need for and scope to apply judgment is explicit in the processes, recognising the inherent data limitations that need to be worked around.

These methods were developed to estimate a “present value” figure that would be expected to cover the anticipated costs of an uncertain set of cashflows expected to occur sometime in the future. They are intended to provide a stable basis by which to put a value on these uncertain outcomes. Specifically the methods have been developed with the following in mind:

• Costs and outcomes to emerge slowly over time
• A time bias in outcomes (e.g. smaller claims are settled first)
• High variability at the individual level
• Relatively limited “explanatory” data available for predictive modelling
• A need to provide a robust, evidence-based method for valuing assets and/or liabilities to management, Directors, Regulators and share markets to ensure decision makers adequately informed and confident in their decisions.

I think this helps to shed some light on why we’ve seen a divergence in techniques in the general insurance industry between personal lines pricing and other insurance actuarial work. As the availability of detailed risk data has increased, the industry has moved toward highly sophisticated predictive modelling, and away from aggregate techniques. The lack of access to such rich data in other insurance areas (commercial lines, life, etc.) perhaps explains why more traditional actuarial methods still play a large in those industry segments.

Going back to those attributes I set out above, I’d argue that insurance isn’t the only business that exhibits these patterns. Indeed, we have seen some high profile work done by actuaries in Australia and New Zealand valuing Government Welfare liabilities. Overseas, there are examples of actuaries working with Oil & Gas valuations, helping companies understand the value relating to assets where future cashflows can be highly volatile. In these examples, I think actuaries can add as much by helping management understand not just the likely outcome, but also by helping quantify risks and assist the companies devise risk mitigation strategies.

A personal example that comes to mind is of an instance where I built state-transition models with movements between each state modelled using the equivalent of a claim finalisation triangle. What was it for? It was for a large scale construction project; the purpose was to assist the company understand its resourcing needs over the coming years. Each phase of the construction process had different skillsets required to support it, and therefore the company needed to have a handle on how many “active” construction sites would be in any given phase of construction, so that staff resourcing could be accordingly managed in advance. Reliance on external parties, such as councils, sub-contractors, materials suppliers, engineers etc. meant that progress rates were subject to variability. In other words, the perfect set of circumstances to tackle with a finalisation triangle! By taking this tried and true approach to projecting outcomes, we were able to help the company understand:

• Expected timeframes for completion of the overall project
• The required resourcing needed to support the project in all future months
• The range of variability in that resourcing demand, which in turn informed strategies to manage the right level of flexibility in staffing levels
• They key risks to meeting targets (e.g. project phases that had a higher propensity for substantive delays)

As a result, this particular business has successfully navigated a stressful period of high growth, and subsequent slowdown, very well. Quality and timeframe KPIs were delivered very well, despite a previously unseen level of demand and growth in the market. The subsequent slowdown has seen a controlled and profitable reduction in the size of the organisation. If you think about boom and bust cycles the construction industry goes through, this is actually a rather impressive feat. By making long-term decisions based on an informed projection, rather than based simply on demand today, the business has been managed effectively and profitably. Things haven’t been so easy for those players not making a more forward looking assessment.

Now, for me this opportunity to do an interesting project and work with a different industry altogether actually came out of my work with an insurer who happened to be working closely with the construction company. The issue came up in discussion and after some creative thinking was applied, we were soon working on trying to build a solution to the problem. In this case, simply by thinking carefully about the broader issues the insurer faced, we were able to identify an opportunity to help with something a bit left-field.

This is just one example, and the welfare valuation work being done by actuaries in Australia and New Zealand is a big endorsement of the profession’s ability to make a broader contribution than insurance.

There are plenty of opportunities in the business world to help companies grapple with issues of uncertainty in a systematic way that links cause and effect, and allows management to understand the business implications of the uncertainty. We just need to open our minds up to these potential applications.

By all means, I encourage actuaries everywhere to increase their understanding of emerging modelling and analytical techniques. This is an important part of making sure our technical work remains relevant. However, don’t lose sight of what it is that the profession offers – the ability to manage uncertainty through evidence-based decision making. There are plenty of business problems out there, where decisions have to be made an environment where causal data is extremely limited (so your GBM, neural network, etc. won’t help you), and outcomes are driven by multiple underlying real-world dynamics. Our training gives us the ability to tackle these problems systematically and to help stakeholders devise management strategies. So let’s make sure that while we pursue greater technical knowledge, we also expand our horizons and help other sectors exploit the value that actuaries can readily add.