Unlocking re/insurance capacity: Strengthening models and data-driven underwriting resilience

The re/insurance industry has always been at the forefront of natural catastrophe risk assessment. But a series of sobering setbacks from 2011-2020, where actual losses far exceeded the expected modelled losses – such as the 2011 Thailand floods,¹ and typhoons Jebi (2018) and Hagibis (2019) in Japan² – highlight the urgency of fine-tuning our risk assessment methods, so they remain fit for purpose at a time when both the frequency and impact of catastrophes are on the rise.  

This is especially apparent in the wake of this ‘lost decade' where events continue to take us by surprise. In 2021, Malaysia recorded its highest-ever insured flood losses of around USD 600 million, while Australia announced record-breaking insured losses of USD 4.3 billion ³ from the 2022 Eastern Australia floods alone. Just this year, New Zealand experienced two unprecedented floods with combined insured losses of about USD 2.3 billion. 

The factors behind these unexpected losses, especially those associated with weather-related events need to be better understood. With Asia Pacific home to many vulnerable metropolises, it is vital to ask: How can the insurance industry build capacity to withstand shocks from the next big surprise event? And what lessons can we glean from recent unexpected occurrences?

Though there is no silver bullet, we believe the agile adaptation of models that integrate high-quality datasets throughout the insurance value chain is key. It enables us to assess risks and likely losses with more adequacy, thereby enhancing overall underwriting resilience.

For re/insurers to develop a sustainable offering that ensures the right pricing in a timely manner, continuous assessment of losses and potential downside is crucial. Despite improvements in nat cat risk modelling in Asia Pacific over the past decade, progress continues to lag in a rapidly changing landscape.

Catastrophe models – cornerstones of nat cat underwriting – are necessary to ensure that the risks from floods and other secondary perils are adequately measured, though they are not consistently available. For example, while Australia and New Zealand generally maintain high quality datasets, there are only a few prominent vendor models for flood peril in Australia, a country that recently experienced its highest flood insured losses in history.

When models are not available or not used along the insurance value chain despite being available, risks tend to be overlooked or calculated using highly subjective experience-based methods, resulting in a lack of consistency and rigour.

Consequently, these methods are prone to material underestimation of losses. In Malaysia, for example, limited loss experience prior to 2021, which does not reflect the loss potential from the country’s rapid urbanisation, would have indicated expected flood-related losses in the order of USD 50-100 million (Graph 2), a far cry from the USD 600 million recorded in that year.

Data quality is also crucial to achieving meaningful modelled results, and is another area where Asia Pacific still needs to catch up.

One significant challenge for less mature markets is that data flowing through the insurance value chain often comes in aggregated forms and sometimes, with vague data definition. In the Chinese market, for example, for the larger portion of portfolios, only province-level aggregated exposure values are shared in the insurance value chain, and it is sometimes ambiguous whether these represent replacement, gross, or net insured values. The lack of basic details such as insurance conditions or line of business value splits limit our ability to fully leverage this data in catastrophe models – which in the end, are only as good as the data fed into them.

Barriers aside, any effort to put catastrophe models in place can encourage the systematic collection of quality exposure data and instil discipline and consistency in risk assessment. Insurers can also turn to tools such as Swiss Re’s CatNet® to assess and visualise natural hazard exposure for an individual location and portfolio. A flow of quality modelling data into the insurance value chain creates positive feedback loops, driving continuous improvements in models and data over time.

Taking a 360-degree view of risk will further minimise avoidable surprises with adequate risk benchmarks. While it is prudent to constantly review nat cat tools and underwriting assumptions in the context of the present environment to identify potential blind spots, it is equally crucial not to overlook valuable lessons from the region’s past.

The arrival of typhoon Jebi (2018) and typhoon Hagibis (2019) reminded re/insurers of the vulnerability of Japan’s urban areas to severe typhoon wind and flood risks after more than 25 relatively uneventful years. Similarly, China, with its extensive coastline and concentrated exposure in the prosperous Pearl River Delta and Shanghai regions, faces significant risk of severe losses from direct typhoon landfall. While these areas have been free of direct impact in recent memory, historical events such as typhoon Hope (1979) in the Pearl River Delta and typhoon Gloria (1949) in the Shanghai region, highlight the likelihood and potential impact.

Given the unprecedented changes in the physical, socio-economic and insurance landscapes in China as shown in Graph 3, estimating losses for events that occur today carries a degree of uncertainty. But surprises for the insurance industry with multi-billion-dollar losses are certain if we do not seek to inform our risk views with all the knowledge at our disposal.

We can do this by analysing and factoring in the impact of short to long-term macro trends. These include how urbanisation has influenced exposure concentration and modified flood risk vulnerability in Jakarta over the past decades; how flood insurance practices have evolved in Australia; and the role of climate change and variability in extreme rainfall in India.

Without insights from the past filtered through current realities like the region’s rapid growth, what has been observed in complex urban areas during severe natural catastrophe events show the likelihood of massive losses from unexpected outcomes such as storm surges, infrastructure failure leading to massive flooding, catastrophic fires, and blackouts.

As nat cat risks become increasingly dynamic, particularly in fast-growing markets exposed to multiple weather perils, we must enhance our response agility. Embedding faster feedback loops where critical learnings are quickly incorporated into nat cat tools – even if that requires significant changes in risk views – will contribute to a better functioning market. Japan offers a recent example, where typhoon-induced flood risk, which was largely ignored in pricing, was quickly accounted for in underwriting following the typhoons in 2018 and 2019.

Insurers, brokers and reinsurers must work hand-in-hand to advance best practices and sustainability of nat cat capacity. Collecting and sharing detailed exposure and results data from widely available models across all APAC, more importantly for flood peril, can help ensure these data and risks become an integral part of the re/insurance value chain.

Where models are lacking or being updated, and loss experience become important risk benchmarks, it is essential to pay particular attention to rapid urbanisation across APAC as historical event loss experience ages very quickly and might be full of gaps that relate to new, untested megapolises.

All of this will not only ensure that we retain the capacity but also unlock billions of dollars of more market capacity at competitive terms and strengthen nat cat resilience for both industry and society.

¹ https://www.swissre.com/institute/research/sigma-research/Economic-Insights/the-costliest-flood-thailand-flood.html
² https://www.spglobal.com/marketintelligence/en/news-insights/trending/lFwqDt32qXnFb56-A3uOZQ2
³ https://www.reinsurancene.ws/australia-flood-loss-up-29-to-6-3bn-in-fresh-perils-update/