AIR: Tackling the cluster
COVID-19 and evolving technology are both providing risk modellers with a challenge.
By Marc Jones, Associate Editor
As the COVID-19 pandemic continues to roil the world, the long-term implications of the coronavirus remain to be seen. However, risk-modelling companies are taking the opportunity to learn the lessons of just what a global pandemic can look like.
According to Bill Churney, President of AIR Worldwide, for models to provide the latest view of a risk, they need to evolve by incorporating the latest science. This remains a crucial component to developing cutting-edge risk models, a point on which he says AIR will not compromise.
“We continue to see more instances where risk is correlated across lines of business,” he explains. “For example, we now have a better understanding of the ramifications of global pandemics across multiple lines of business as a result of COVID-19. To fully capture all the potential impacts, risk models should reconsider the perils they simulate as events that could potentially impact all lines of business.”
He adds that models also need to better understand clustering: events occur in clusters that impact a region in a short amount of time, or they cluster within a season and impact one geography multiple times. AIR is already tackling clustering across seasons, such as what the company is doing in its tropical cyclone, flood, and wildfire models. AIR knows the risk in a given season is dependent on what has occurred in prior years. For example, the company is simulating weather leading up to droughts in a multi-year fashion. This is a work in progress, but an area where AIR sees that there is significant room for further improvement.
Machine learning is an area that AIR has been investigating and leveraging for a few years. Churney says that at a macro level, it remains a small data vs. a big data problem i.e., in general, AIR is not overwhelmed with data, rather it is always looking for more data to inform its models because catastrophes are inherently rare. However, AIR is currently employing machine learning and artificial intelligence techniques in certain areas. For example, to enhance commercial exposure data, the company is developing and training algorithms to identify high-rise buildings from satellite and aerial imagery, based on shadows, patterns, and their shapes. This is a more efficient way to accurately identify these critical structures, especially in areas of the world where detailed exposure data isn’t readily available.
Evolving technology
In addition the evolution of technology is changing the world of insurance, as insurers start to see the possibilities that that evolution provides.
According to Churney AIR is seeing the need for digital transformation in the insurance industry – the need to digitise workflows and processes. “We’re seeing digital transformation occur around how clients use extreme event models and integrate them into their associated workflows,” he says. “That’s leading to a lot of efficiencies, savings, and more resilient operations that companies can and should be taking advantage of. One example is companies moving more of their core operations to the cloud, including modelling. Advantages of this include an environment optimized for cat modelling that’s fully virtualized and is accessible anywhere. Frankly, it is amazing to see how fast the insurance industry has moved to embrace the cloud in the past few years.”
Churney says that AIR is also seeing the industry focus on automating internal workflows by leveraging application programming interfaces (APIs) to connect their own systems with platforms like AIR’s. This enables clients to transform a modelling process that might have involved dozens of steps each involving a person to monitor the process into an automated flow requiring one click of a mouse that occurs in a fraction of the time. What this means is more time spent on understanding the model results vs just obtaining them. Companies can make better business decisions in less time and with less cost. Churney thinks that there will be a growing gap between insurers who are embracing such technology and those who are not.
Looking ahead, Churney thinks that the future of risk modelling is in modelling interconnections, adding that this becomes clear when you look at new areas of risk, like supply chain, liability, and cyber – it’s all about interconnections.
“Applying this type of thinking will lead us to the future of natural catastrophe risk,” says Churney. “We are working on the first truly global atmospheric peril model that will connect weather systems around the globe and capture the way a hurricane in Florida could affect storm systems in Europe, or how monsoons might impact typhoon formation.”
Climate risk is another important area of focus for AIR. According to Churney a crucial factor in understanding and managing climate risk is assessing how the risk could change over time and under different climate scenarios. Assessing and mitigating the current and near-future (10-15 years from now) view of risk is the highest priority for the insurance industry. AIR has committed significant resources to study how climate change is impacting extreme weather events now, and the future impacts of climate change on specific weather phenomena.
Beyond accurately underwriting and pricing the risk of climate change, Churney concludes that developing a holistic approach to managing this risk is essential for re/insurers. “The risk of changing climate has extended beyond property and agriculture,” he says. “Today, for example, we know climate change is impacting the growth and spread of pandemics. We have also seen the impact of sea-level rise and extreme events on global supply chains. The liability risk of climate change is rising as well, as more and more lawsuits are filed against corporations and municipalities worldwide. Therefore, new analytical tools that can estimate these interconnected risks are poised to play a critical role in helping many different types of entities assess their risk going forward.”
