Enhancing Accuracy in Insurance Forecasting with Behavioural and Decision Science

Whether predicting long-term mortality improvements in an emerging markets economy or the number of traffic fatalities amid a shift to autonomous driving, accurate forecasting is the bread and butter of any re/insurance organisation.

Supported by the digitalisation of insurance operations and evolving data science approaches, predictions for costing and reserving are increasingly driven by sophisticated quantitative forecasting models. However, in contexts of high uncertainty and complexity, research has shown human expert judgement can still outperform quantitative forecasting models, or offer a valuable complementary perspective. Behavioural economics studies suggest, however, that the quality of expert judgement can be undermined by cognitive biases and inconsistencies (noise).

While several methodologies, including 'estimate-talk-estimate' approaches (using the widely recognised Delphi approach), have been tested for their contributions to enhancing forecasting accuracy, applications to insurance problems have been more limited to date.

Here, we present our findings on utilising underwriting and reserving performance data, online experiments, and sentiment analysis to assess the prevalence of cognitive biases and noise in key costing and reserving activities. Informed by these diagnostics, we assessed the value of structured approaches to expert judgement, specifically their ability to better navigate cognitive biases and minimise inconsistency, finding that improvements of between 5% and 40% in forecasting accuracy are possible using these techniques. Lastly, we evaluated strategies for achieving greater scalability of best-practice approaches, including through integration into work process, forecasting competitions and other learning and development initiatives.

In conclusion, our research suggests valuable improvements in forecasting accuracy are possible through best practice approaches for enhancing expert judgement, helping to improve loss ratio performance and reserving adequacy.