Applied Epistemology in Actuarial Practice

Much of actuarial practice can be viewed as making various types of forecasts, whether it is expected costs of future claims for individual risks, aggregate portfolio outcomes, economic or natural disaster scenarios and, more recently, customer behaviour. In this paper we examine the status of these forecasts from the point of view of applied epistemology – under what circumstances are our beliefs about the future justified? We make distinctions between three types of forecasts:

• What is likely to happen in the absence of any explicit intervention
• What is the expected effect of an intervention
• What is the best course of action (choice of intervention).
• And three types of systems with respect to which forecasts are made:
• Large ensembles of independent and homogenous risks
• A collection of Individually unique risks or events where it is possible for a forecaster to establish a track record, such as regular weather forecasts or outcomes of sporting events
• A unique system or event, e.g. long term forecasts of national mortality or GDP.

The proposed framework is illustrated with examples from fields including weather forecasting, quality control, betting markets and quantitative social science and is meant to help actuaries to choose appropriate modelling and verification strategies and better appreciate their limitations.

Find the Q&A here: Q&A on 'Understanding Risks in Practice'