P&C the Jan/Feb 2025 issue

Examining Loss Estimates

Insurers and reinsurers are grappling with changing loss conditions that make initial estimates seem out of touch after a natural catastrophe.
By Chris Schneidmiller Posted on January 14, 2025

Insurers and reinsurers are wrestling with sharp increases over time in losses following natural disasters. In one instance, loss estimates rose from $2.2 billion to nearly $6 billion after hailstorms in Northern Italy.

Multiple factors are driving the phenomenon, industry sources say, including climate change, outdated information that leads to underestimation of loss exposure, and inflation.

Addressing the dilemma will require up-to-date property valuation data, transparency between insurers and their reinsurers, and improved modeling, among other measures, according to industry experts.

Swiss Re highlighted the event in a June 2024 report as a signature example of loss creep, in which loss figures escalate drastically over time following a disaster (also known as loss development).

It was not an isolated case, Swiss Re said in its report; rather, loss creep is an “industry-wide phenomenon” that in Europe has occurred after 2021 flooding in Germany, hail in France in 2022, and the Turkey-Syria earthquakes in 2023, the reinsurer said.

“When it comes to extreme weather events as we have seen in Italy, loss creep fundamentally boils down to massive underestimation of loss exposure, risk values and inflationary impact. That was exacerbated by insufficient availability of loss adjusters, experts, builders and building materials as well as a lack of claims handling in these times of high demand,” Rita Müller, Swiss Re head of facultative casualty underwriting, and Balz Grollimund, head of catastrophe perils, wrote in their paper, “Tackling the toxic loss creep in European NatCat events.”

The Climate Change Factor

It is widely accepted in the industry that losses may increase following an event as new claims come in and additional damages are uncovered.

“The fundamental premise behind loss development is the reality that it takes time after an event to fully ascertain both the number of claims and the loss amounts on those claims as new information comes to light while the queue of claims is being worked through by the insurance company,” says Alex Korb, a consultant and former head of ICEYE’s parametric insurance practice.

However, Swiss Re and others say there is more to consider when looking at this trend. Loss creep tends to be found in events that have complicating factors that are difficult to model, says David Flandro, head of industry analysis and strategic advisory for Howden Re.

One of those factors is climate change. For example, while wildfires have long been a threat to many regions around the world, research has found that climate change is exacerbating the risk by increasing temperatures and drought conditions. These factors, alongside others such as fire seasons that can extend by months, must now be incorporated into models for losses.

Flandro cites several massive wildfires from recent years, including the November 2018 Camp Fire in California, the most destructive in state history. Modelers initially estimated losses in the single-digit billions—AIR Worldwide, for one, projected $6 billion to $9 billion in December of that year, according to reporting at the time—but losses rose into the teens of billions, Flandro says. Losses from other wildfires have followed the same path, he adds.

(At the time this article was posted, a series of wildfires in the Los Angeles region had killed at least 24 people, damaged over 12,000 structures, including residences, and forced widespread evacuations in some areas. The private AccuWeather service as of Jan. 13 estimated total losses at $250 billion, The Associated Press reported. Insured losses have been projected at up to $20 billion, according to Reuters.)

When it comes to extreme weather events as we have seen in Italy, loss creep fundamentally boils down to massive underestimation of loss exposure, risk values and inflationary impact. That was exacerbated by insufficient availability of loss adjusters, experts, builders and building materials as well as a lack of claims handling in these times of high demand.”
Rita Müller, head of facultative casualty underwriting, and Balz Grollimund, head of catastrophe perils, Swiss Re

The effects of climate change are being seen in other perils as well. In 2024, modeler Verisk predicted $151 billion in insured average annual loss (AAL) from natural catastrophes over the long term. Verisk modeling showed severe storms as the top contributor to AAL, at 32%, followed in order by tropical cyclones (24%), crop loss (21%), earthquakes (10%), flooding (7%), and wildfire (6%). Apart from earthquakes, all of these perils are exacerbated by climate change.

“Currently, climate change accounts for approximately 1% of the annual increase in losses, with exposure growth and inflation being the main contributors,” according to the company’s 2024 Global Modeled Catastrophe Losses report, issued in September. “Nonetheless, its influence is expected to become more significant over the next few decades. The insurance industry needs to be proactive and utilize advanced, forward-looking models to better estimate risk and guide internal decision-making.”

The frequency of severe convective storms makes them particularly dangerous—of the 30 events that cost up to $5 billion in 2023, 21 were storms of this type, Swiss Re said in a separate March 2024 report on natural catastrophes.

While 85% of the $64 billion in worldwide insured losses from severe convective storms in 2020 came out of the United States, “SCS-related insured losses were fastest-growing in Europe, exceeding USD 5 billion in each of the last three years,” according to that report. “Hail risk in particular is increasing, mainly in Germany, Italy and France.”

First-generation catastrophe models are not accurately estimating insured losses from severe convective storms, Artemis quoted modeler Karen Clark & Company as saying.

“Because the industry for years has been underestimating the losses from the frequency (aka secondary) perils, most notably SCS, it appears to many (re)insurers as if there is a significant increasing trend,” according to the company. “Not only have the models used by many (re)insurers underestimated the risk, sources of market losses have also underestimated the total industry losses from these events.”

Outdated Data

Multiple factors complicate loss projections for wildfires and other natural hazards, Flandro cautions, including:

  • Climate change, particularly related to less-modeled risks such as wildfire, hail, and derechos. While hurricane development in the Atlantic has been modeled for decades, they can still surprise “when they’re not ‘plain vanilla,’” he says.
  • Lack of risk modeling for certain regions. He specifically calls out the limited amount of modeling in northern Alberta and Saskatchewan, Canada, prior to the Fort McMurray fire from 2016 to 2017, the most expensive in the nation’s history. While the area seemed a “peripheral” risk, the boreal forests, growing population near the oil sands operations, and insufficient forest management instead “was (we can now all tell with hindsight) a disaster waiting to happen.”
  • Mismodeling of increasing urbanization, land-use changes, infrastructure vulnerability, and other factors, such as climate change driven rainfall amounts and intensities in Europe.

“The root cause for underestimating the cost of extreme weather events comes down to a lack of data about the up-to-date exposure and the current risk values,” Müller and Grollimund wrote in their paper. “Frequently, many of the multiple data re/insurers use to assess risk are missing or outdated. And as an industry, we consistently give too much weight to previous events when modelling losses.”

Müller and Grollimund argued that failing to manage the danger promises underpricing losses and reduction of trust in the insurance sector. The industry has recognized loss creep for years, but it has not taken steps to address it, they said.

I think that the better data that we’re able to get, the more granular data…if you’re able to get way more details on the type of construction, the location, the year built, all of these other factors are really important in terms of trying to reduce the uncertainty anytime you are running some of these events in real time, trying to get a quick estimation of what an event loss may be.”
Steve Bowen, chief science officer, Gallagher Re

Pointing to the hailstorms in Italy, the Swiss Re authors noted that insurers were broadly ignorant that many homeowners had added solar panels to their roofs with help of government incentives. They also failed to account for cost differentials in repairs to roofs with and without damaged solar panels. Reinsurers, in turn, receive little to no information from insurance companies regarding hail exposure in Italy, Grollimund said in the paper. “For some, we don’t receive any information at all. For others, we just receive very broad information that covers a large area.”

Inflation is another contributor over the past few years, solidifying the case against modeling that relies on outdated information on repair expenses, Swiss Re said. In the paper, Müller directly called out insurers that applied a reserving model from 2014 incident hail events in France ($1.4 billion in losses) to setting preliminary reserve estimates for similar severe convective storms eight years later (ultimately $5.5 billion in losses): “You can’t do that.”

Initial cost analyses from catastrophe modeling companies are generally updated several times over a period of weeks as the amount of area affected becomes clearer and to incorporate expenses from inflation and other factors, Gallagher Re said in its report on insured losses in first-quarter 2024. A series of large events in a short period will limit available supplies and labor for repairs or replacement, says Steve Bowen, chief science officer for Gallagher Re, further driving up costs over an extended period.

“Current market conditions and impact of catastrophe events can have an impact on relevant material and labor market pricing, through disruption of market equilibrium for major cat events, and/or with inflationary conditions expected to continue during reconstruction after catastrophe events,” says Cagdas Kafali, senior vice president of research and modeling for Verisk Extreme Event Solutions.

Costs Creeping Up

Bowen notes that in some instances the damage might be less widespread than initially estimated, driving early loss figures lower over time. Still, he acknowledges that “there are more instances these days, especially when we’re talking about thunderstorm losses, where loss costs tend to creep higher than they do go down.”

There is a long list of other contributing factors as well, according to industry representatives, from demand surge that drives up prices over time to litigation to involvement of third-party assessors.

Internationally, areas with a less robust insurance market could be at greater risk of loss creep, Bowen says. Those regions might simply not have the comprehensive data needed for well-calibrated models that are applied to develop preliminary loss estimates, he adds.

Assessing the various forms of damage is another challenge. While structural damage is perhaps the No. 1 issue following a natural catastrophe, these events can also damage or destroy separate property such as vehicles, says Robb Lanham, chief sales officer for Hub Private Client.

Continued underestimation of losses would mean the industry is underpricing its product, Müller and Grollimund wrote. It can also undermine consumers’ trust in the industry, they added. “As we expect more extreme weather events in the future, systemic loss creep is not acceptable and not sustainable for our industry.”

The “Long-Tail” Effect

The potential damage from loss creep extends into the insurance-linked securities market, Artemis has noted in reporting on the risk. In September 2024 it cited a report from Switzerland-based Icosa Investments AG that three catastrophe bonds had partly or completely defaulted based on increasing losses dating as far back as September 2022. The development “presents an opportunity to explain a crucial concept—loss creep—and why investors should be mindful of this risk before allocating to cat bond funds,” Icosa said at the time.

“ILS fund managers have already experienced how prolonged loss creep can hurt them when capital gets trapped for the duration of the development,” according to an earlier Artemis article on the Swiss Re paper. “The uncertainty this creates and the drag on portfolio performance can be a significant issue and inaccurate approaches to initial reserving can be a real issue here, especially where an investment assumed likely to be safe, ends up facing losses thanks to significant creep.”

Howden Re’s Flandro concurs that under-reserving is a major threat to the industry but on a wider scope. It is essentially the “long-tail” version of loss creep and the biggest historical cause of reinsurer failure, he says. Historically, the threat has involved lines of business such as directors and officers liability, commercial auto, or workers compensations, according to Flandro.

“I think one of the last truly existential moments that the sector faced was in the early 2000s when we had the liability crisis and that was driven by asbestos and mesothelioma losses, which really did surprise a lot of people—especially when initial loss picks for those years are considered,” he adds. “And you want to talk about loss creep, I mean, with hindsight, in real terms the sector lost hundreds of billions on that. And that was 25 years ago.”

In terms of large-scale industry risks, adverse loss development must also stand beside massive fast-moving balance sheet impairments such as those during the 2008 financial crisis and exceedingly rare but ruinous mega-catastrophes such as the 1906 San Francisco earthquake or the 1926 Miami hurricane, according to Flandro.

The Fix

To maintain an accurate assessment of losses, insurers must have accurate, updated data and transparency throughout the value chain, Müller and Grollimund wrote. That begins with ensuring that loss exposure data does not become outdated “at the point of primary underwriting.”

“Valuations need to represent true replacement costs and new investments must be captured in a timely fashion,” according to the Swiss Re experts. “With a clearer view of the true costs, re/insurers can more accurately price the risks. And as a consequence, initial reserve setting after an event can be based on realistic exposure data rather than outdated historic events.”

They also called to steer preliminary reserving models away from overreliance on historic data and slow loss adjustment. The models must instead consider actual loss exposure, encompassing current inflation levels.

Underwriters should be skeptical of vendor models rather than accepting them at face value, industry sources say.

“I think that the better data that we’re able to get, the more granular data…if you’re able to get way more details on the type of construction, the location, the year built, all of these other factors are really important in terms of trying to reduce the uncertainty anytime you are running some of these events in real time, trying to get a quick estimation of what an event loss may be,” Bowen says.

Chris Schneidmiller Senior Editor, Leader's Edge Read More

More in P&C

Houses of Worship Face Extreme Weather
P&C Houses of Worship Face Extreme Weather
Insurer offers insights into the unique challenges of covering religious buildin...
P&C Risky Business
RiskScan 2024 identifies what buyers and sellers in the insurance industry are w...
Sponsored By Munich Re
When Disaster Knocks
P&C When Disaster Knocks
A continuous series of natural catastrophes around the United States might final...