Underwriting: The next generation
Driven primarily by the pace of technological progress, underwriting is changing rapidly. With new insights from an ability to track perils in real time, we are able to change the way we model and understand risk. This will allow new means of risk assessment and underwriting, augmenting our traditional process of using past data. These shifts will see the nature of insurance products begin to change from ex ante compensation packages to anticipative risk management services. Swiss Re Group Chief Underwriting Officer Edi Schmid considers the evolution of his profession; and how it will alter our approach to risk.
Article information and share options
Once upon a time, in the distant past before even I had underwritten my first insurance contract, insurers underwrote life insurance with a single piece of data – age. Edward Rees Mores, the founder of the Society for Equitable Assurances on Lives and Survivorship in 1762, regulated premiums to their risk pool by age alone, regardless of any other factors.
Those same life tables, relying solely on age, were the basis of underwriting until the 1940s, when gender was introduced as an underwriting criteria. Women tended to live longer than men, and this was reflected in premiums. In the 1970s, following the Federal Trade Commission's report to Congress on the danger of smoking in 1967, insurers differentiated between smokers and non-smokers. The emergence of HIV/AIDS in the 1980s saw blood tests introduced into underwriting. Those blood tests could yield considerably more data than just the presence of HIV at incremental cost, a quality, which saw them become increasingly standard. Thus, the portfolio of underwriting criteria I was familiar with when I started my career – including weight, blood pressure, heart rate, family history, cholesterol, foreign travel, hazardous activities – came into being.
The point of this brief history lesson is that underwriting changes. Moreover, the speed with which it changes has been increasing as we have learned more about risk and been able to use that knowledge to improve our capital allocation and risk selection.
The pace of change is ratcheting up again, and with one common driving force: technology. Even when I began as an underwriter – which does not seem that long ago, especially after writing about insurance 250 years past – it was unimaginable that most of our insured portfolio would be walking around with GPS enabled devices, tracking their steps and activities. Some even tracking heart rate, hydration and other vital factors. I could not foresee cars able to monitor driver performance constantly. I had no idea we would have drones capable of flying over fields, able to interpret the health and yield of crops; nor did I think we could not only follow a container across the world, always updated on its temperature and humidity. And I did not anticipate we would be able to capture all this detail remotely in real time, then incorporate and interpret it through sophisticated data modelling.
New underwriting approaches
As digital transforms all our lives, so it is transforming our underwriting. We may not be talking revolution – don't throw out those life tables just yet – but we are looking at an accelerated pace of evolution, changing how insurers accept and manage risk.
- Forward-looking underwriting – Underwriting has been built on data, and data records what happened in the past. One can however, take a portfolio of data, model and analyse it on a forward-looking basis. Swiss Re has undertaken this with reference to Canada wildfire, with risk prediction based on data points including water vapour, vegetation cover, surface temperature and lightening index. This modelling was enabled by innovative deep learning techniques. Such techniques are invaluable in a new line, such as cyber risk, where we have relatively little in the way of historical data to help us make decisions. Swiss Re is also using forward-looking approaches across its casualty lines.
Distribution of wildfire hazard with respect to the expected intensity of fires, six months ahead
Note: Modelling using data as of November 2015. Green represents areas where no fires are predicted in six months’ time. Yellow represents areas where small fires are likely and orange represents areas where large fires are likely in six months’ time. Black dots are actual fires in April/May 2016. Source: Swiss Re --> see Swiss Re Institute Expertise Publication "Wildfire in Canada: fostering resilience through advances in modelling".
- Predictive modelling – Sometimes proxy data can provide a powerful indicator as to underlying risk conditions. Swiss Re has been working with life insurance partners for many years on leveraging data such as using bank data as a proxy to the health of applicants, thus reducing the need for traditional underwriting tests, and allowing for pre-approved products.
- Dynamic modelling - Underwriting has traditionally been a risk snap shot in time. However, risks change. For many years, insurers have been capturing fitness data among their insured populations and using it as reward points or premium discounts. Next generation technology will go beyond fitness to health trackers, providing underwriters with a dynamic view into past health status, and the ability to monitor current health. Another line changing with the dynamic use of data is motor, with telematics giving underwriters an individual and updated overview of the insured party's driving style and risk profile.
- Parametric modelling – Emerging from agriculture, this form of underwriting takes an easily measurable variable (say rainfall) and correlates it against a close dependency (crop yields) to reduce underwriting costs. Other examples include wind speed as a proxy for power output of windfarms, sun exposure for solar plants, and a number of natural catastrophe lines, such as wind speed in the case of storms. Swiss Re even has a flight delay product requiring no damage assessment or claims process. The qualitative approach is critical here, as the underwriter is effectively providing an automated service, managing both demand and supply in the transaction.
- Synthetic modelling – Natural catastrophe modelling originally took historical losses and built factors such as wind speed and landscape profiles into predictive models. The layers of analytic data have grown considerably. Satellite images in combination with weather, traffic, building and other ground sensors can be combined and overlaid to provide a more nuanced and accurate picture of risk.
- Combined modelling – Structural models are being trained to combine cost and damage indices, an approach being trialled in mortality and epidemiology.
- Although advances in the gathering and use of data will augment underwriting, they will not, at least in the foreseeable future, replace the fulcrum of the process, namely the underwriter. The human factor in risk assessment remains crucial to the process, particularly in underwriting complex insurance policies.
Insurance as a service
It is not just modelling and underwriting that stands on a precipice of change. Our products have evolved very little since the 1760s. We still provide a compensation promise ex ante for a stated and categorised risk. It is worth adding that traditional insurance is seen largely as a necessary evil rather than valued product. Even though I do not think that the polls are always statistically reliable, it still pains me to see insurance agents regularly appear among lists for 'the most hated profession', alongside politicians, estate agents, car salesmen and (more recently) bankers.
This might change in coming years as insurers are beginning to create products that are more than just post-damage payments. In the life and health space, many insurers are adding services to their main products, including free or subsidised gym membership, and other benefits, such as regular health checks. It is easy to imagine such services developing over coming years, particularly as health monitoring devices become more sophisticated. As the main payee in most established health systems, insurers need to develop their product, so that it is a central part of the digital health value chain.
From fitness to health: Wearables in the insurance value chain
Swiss Re Institute recently hosted a conference on wearables to discuss the latest developments and solutions. We caught up with Kelvyn Young, Head of L&H Partnerships at Swiss Re, to get his view on how insurers are using wearables data and the importance of health ecosystems --> see the interview with Kelvyn Young, Head of L&H Partnerships at Swiss Re.
This brings us a step closer to insurers providing active risk management solutions for their clients. By monitoring relevant data flows, insurers may be able to help prevent, or at least mitigate damages prior to them occurring. We could advise immediate hospitalisation, for example, if we saw abnormalities suggesting a possible stroke. Smart homes will provide us with real time data that will aid risk management, such as networked burglar or fire alarms. We could help small farmers identify crop disease early and react. We could offer proactive natural catastrophe consulting to local governments, using our models to help them prepare for disasters.
Key to these product improvements will be the ability of insurers to participate in the digital value chain. Where risk is involved and data is collected, we need to be in those data flows. It is vital to our relevance in a digital world. It will further aid insurers in underwriting certain risks that are considered currently uninsurable. In some lines of business we have been relatively successful in accessing data; in others we are some way behind. We need to build partnerships around where the data flows. This is one of the most significant roles of Swiss Re Institute, helping Swiss Re secure data partners, curating and analysing that data.
A last factor we should keep in mind – underwriting is only one part of the value added by insurers. The use of real time data affects not only how we assess and underwrite risk, it affects how we distribute our product, how we administer our product, and how our customers can make claims against our products. While this article is focused primarily on underwriting, the impact of data flows will be felt across the insurance value chain.
We have been on a journey from some of the first hand-written life tables to predictive risk analysis using artificial intelligence. It is a journey I have a vested interest in, and to some degree reflects my own career. The underwriting profession I entered will be very different to the one I (eventually) leave.
The worst mistake we can make is to resist change. The digital age has left us with the corpses of many companies that responded too little too late to tectonic changes. It is why I endorse the 'next generation underwriting' initiative at Swiss Re.
This article ignites our initiative to explore the modern ways of underwriting together with our primary insurance network worldwide. The initiative runs in 2019 and 2020, involving different lines of business, world regions, and communities. Please connect with us if you want to become part of this project.