Parametric insurance products are increasingly being used as an alternative to traditional insurance to address the “protection gap” and increase the speed of insurance payments, particularly in the face of natural catastrophes. Advances in modelling of parametric insurance triggers have the potential to decrease the basis risk of parametric products, and make them more attractive to both insurers and their insureds.
The term ‘basis risk’ can be used to describe imperfect hedging; when the amount by which the exposure in a futures contract differs from the projected value. In the catastrophe bond market, basis risk is the risk that a cat bond may not be triggered, or inadequately triggered, even where the sponsor has suffered a loss.
In the insurance context, basis risk is usually used to describe the amount by which the loss to be indemnified under an insurance contract differs from the pre-modelled loss anticipated when the policy was underwritten.
When insurance cover is indemnity-based, recovery will be determined by reference to the actual loss; in theory there should be very little basis risk. It may be claims handlers, loss adjusters or lawyers who ultimately determine what is covered according to the application of contractual terms to the facts on the ground. Discrepancies between insurer and insured in anticipated recoveries will usually lie in divergent understandings of what was in fact covered according to the policy’s terms and conditions.
With a parametric insurance product, the trigger for payment is a set of pre-defined objective parameters, such as the force of a hurricane or the magnitude of an earthquake. The arbiter for insurance coverage effectively becomes the parametric trigger itself: if the parameters are met, payment is made.
One of the benefits of parametric insurance products is the potential speed by which payments are released when the defined parameters are met. Disputes about coverage are also much reduced.
However, the potential for basis risk remains the major and oft-cited downside to parametric insurance products and for insurance linked securities with a parametric trigger. One example of imperfect parametrics came in the experience of the African Risk Capacity (ARC), a specialised agency of the African Union, in Malawi.
The ARC uses a bespoke modelling tool called Africa RiskView to interpret a range of weather data – including rainfall estimates, satellite-based rainfall information and information on population vulnerability – in order to estimate the size of drought-affected populations. Payments from the ARC risk pooling insurance fund are triggered based on the data – releasing key funding when populations are calculated to be at risk of suffering from droughts.
The basis risk issue was brought into sharp relief when, in 2016, the Africa RiskView model predicted that 21,000 people would be affected by drought in Malawi, whereas in reality it was closer to 6.5 million.
Although the model may have been sound, the input data was not. ARC’s investigations into the issue revealed that farmers had switched to a different crop with a shorter growing cycle. With this more accurate data, the model provided a more accurate estimate and triggered payout to Malawi from the ARC fund.
This shows the potential magnitude of the basis risk problem in parametric insurance products and the importance of accurate and up-to-date input data for modelling.
Basis risk may be minimised by more detailed and accurate modelling and availability of better data. The good news is that models for parametric products are becoming increasingly sophisticated and innovative.
For example, AIG’s Compass Re II 2015-1 has a parametric trigger designed to respond to latitude and longitude of a storm at landfall, calculated by a weighting formula. Another example is the Kenya Livestock Insurance Program, which delivers parametrically-triggered insurance payments directly to pastoralists via phone networks, using satellite images to monitor grazing conditions.
The cost of developing better parametric models is also likely to decrease as enabling data and technology are increasingly shared as an open resource. For example, the Global Flood Monitoring System provides real-time satellite data and hydrological runoff analysis as an online resource, while the OpenQuake Platform allows modelling analysts to share data and tools for earthquake risk.
Insurers have long been in the business of attempting to predict the future through the art and science of risk modelling. Improvements in satellite technology, the widespread availability of key data and open resource sharing are now bringing risk modelling to the fore.
As with other technological disruption, it seems likely that, with an ever-growing range of applications and users and an increased rate of knowledge sharing, the accuracy of risk modelling will increase and costs decrease.
The more accurate modelling for parametric triggers can become, the lower the basis risk and the more attractive parametric insurance products will be in the long term. This will help insurers, who wish to more accurately price their products and reserve against possible future loss; and insureds, who want to be more predictably recompensed.
This article was originaly published in Commercial Risk on 20th October 2017.