ABI ORIC RESEARCH PAPER 2009 |
ANALYSING OPERATIONAL LOSSES IN INSURANCE |
Evidence on the need for scaling from the ORIC database |
EXECUTIVE SUMMARY
Historical loss data are essential for the effective measurement and management of operational risks. Operational losses in insurance do not happen very often, though, and take time to crystallise. Internal losses are thus unlikely to provide a full picture of the spectrum of risks faced by the firm. They are a biased sample of the universe of potential losses because they reflect idiosyncratic features such as the firm’s business, culture and control environment. Information on losses experienced by other firms can fill important gaps in this knowledge, but to deliver meaningful results it is vital these loss events are “comparable” to the losses the firm might experience.
This research studies robust methodologies for scaling the size and number of external losses to make them equivalent to a firm’s internal loss events. Adjusting for potential scaling biases is important when external and internal losses are merged for operational risk management and economic capital calculations.
We use operational loss event in the ORIC database to provide real-world applications of the methodologies discussed. It is the first time we have used our data in this way. We set this against data on the size of the insurer where the loss occurred and additional scaling factors controlling for business lines and loss event types. The purpose of our research is not to provide final answers, but to illustrate our empirical approach and uncover early trends in operational loss data from insurance business.
Among the main conclusions of our empirical analysis are the following:
The size of the insurer is strongly associated with the severity and number of its operational losses. More specifically, loss amounts are positively correlated with the number of full-time employees whereas the number of loss events in a given quarter is more sensitive to premium income.
Increasing the number of full-time employees by 1% results in an increase of about 0.8% in the predicted loss amount, holding all other variables constant. The standardised loss amount per event, which according to our estimates is around £27,000, can increase to nearly £300,000 in firms with high number of full-time employees.
For a standard deviation rise in premium income, roughly £3.7bn, the projected number of operational losses per quarter increases by 24 per cent, holding other variables constant.
Insurers at the lower end of the distribution of premium income are predicted to experience more than nine losses only 4 out of 100 times, whereas insurers in the upper end of the distribution would experience more than nine losses 12 out of 100 times.
The business functions Customer Service/Policy Administration and Claims are strongly associated with smaller losses, other things being equal. Business functions can often predict the variability in observed operational losses better than business lines.
While the goodness-of-fit of our scaling models is often higher than the results reported by similar studies in the banking sector, a great deal of the variability in observed losses remains unexplained by our models.
For scaling the frequency of operational losses, the negative binomial regression model was preferred to the commonly used Poisson regression model.
To ensure the robustness of these findings, we fitted additional scaling models using different sample sizes and econometric techniques (e.g. quantile regression). The final results did not change much.
To download the full paper please click here.
Operational Risk Consortium Limited is a company registered in England and Wales with Company Number 05510364.
Its Registered Office is 51 Gresham Street, London EC2V 7HQ. VAT Registration No. 882 3901 09