Skip to content

Whitepaper: A Recipe for (Avoiding) Disaster

Establishing Best Practices for the Assessment of Physical Peril Risk Models

Climate-related financial risks are a real and present danger, with impacts already reverberating throughout the United States’ (U.S.) financial and housing markets. Under the Biden Administration, several federal agencies have made significant progress toward establishing regulations and oversight mechanisms that address this issue. However, much of the focus has been placed on transition risk, which has led to a lack of focus on physical risk.

Currently, financial and housing markets are flooded with third-party service providers claiming to provide reliable physical risk assessments. The majority of these service providers have neither the breadth nor the quality of data necessary for reliable and actionable analyses. Confusion persists for both regulators and their covered financial institutions. As such, there is a need to establish a clear set of best practices for assessing physical climate-related financial risks.


In order for these best practices – and potential future regulations – to be effective, the following questions must be addressed:

  • What attributes are necessary for a peril risk model to be considered reliable and actionable?
  • How should peril risk models and climate models be used in tandem, for which policy issues, and in what manner?
  • How can regulators and companies contextualize the results of their physical risk analyses in financial terms?

Accordingly, this whitepaper – the first in a three-part series – addresses the first question above by outlining eight best practices focused on specific techniques and capabilities that improve reliability in a peril risk model’s outputs and conclusions.

Model inputs

Natural hazards are highly gradient perils, meaning their impacts can vary over incredibly short distances, making them wide-reaching yet still acutely felt.

1. All inputs — and outputs — must be granular to the individual property-level.

To reliably assess peril risk, one must be able to identify the property itself and the specific structure(s) on the property that require separate assessments. This identification requires location data that can reliably assess the geographical boundaries of a property and its structure(s), along with rich data sources that describe the attributes of the parcel and the structure itself. If the underlying property location data is not reliable, financial damage assessments will not reflect the true risk to the property.

Climate-related parcel data can be broken down into four distinct categories:

  • property location
  • property elevation
  • structure location
  • structure elevation

2. Detailed valuation data is necessary to develop a true understanding of risk.

A model must be developed with a detailed knowledge of building characteristics, and it must have the ability to reliably calculate reconstruction cost value (RCV). The more specific information about the home supplied, the more reliable the RCV calculation will be. This includes:

  • Square footage
  • Year built
  • Architectural style
  • Number of stories
  • Foundation and roof
  • Kitchen and bathrooms
  • Garages

3. All model inputs must be updated at the highest frequency possible.

Depending on the specific data set, the location from where it is being retrieved, and the ability to transmit it in a timely manner, update intervals can vary from minutes to years. For example, tax roll data is typically updated by county governments on an annual basis. However, some counties may be more frequent.

Model structure

With the expanding catastrophe risk management needs of the insurance, housing and financial services industries, regulators and their covered financial institutions require confidence and transparency in model results to manage the potential financial effects of natural hazards. To gain this confidence, it will be necessary to revisit existing peril risk management and loss adjustment strategies by improving overall understanding of peril risk models themselves.

4. The model must be capable of assessing multiple perils collectively.

It is critical that any peril risk model used is capable of reliably assessing all natural hazards that could potentially affect a property as well as understanding the correlation between those individual risks. As such, integrated peril risk assessments are needed to represent the total hazard risk for any location across the U.S.

Peril risk models must incorporate the following, at a minimum, as these perils have a significant effect on properties in the U.S.:

  • Earthquake (Ground-shaking, Fire Following Earthquakes, Tsunami)
  • Hurricane/Tropical Storm Surge
  • Hurricane/Tropical Storm Wind
  • Inland Flood
  • Severe Convective Storm
  • Wildfire
  • Winter Storm (Ice, Snow, Straight-Line Wind)

5. The model should be capable of aggregating risk at the portfolio level.

It is important for regulators and their covered financial institutions to use models that are capable of aggregating individual properties’ composite risks at the portfolio level without loss of reliability.

The combination of risks must consider the geographic distribution and correlation of risks of each property to the whole, while also considering the geographic extent of the disaster itself.

6. The model should use the most comprehensive stochastic event set available.

Peril risk models that can simulate thousands of years are able to capture a broader range of potential disasters. Reliance on the observational record alone underestimates risk, while the use of stochastic event sets (i.e., a random event set) can reveal risks and allow for insights that can only be attained by looking at greater expanses of history.

7. The model should be routinely back-tested against natural hazard events.

The validation process is necessary to ensure model credibility and should be performed on each of the individual components essential to the model’s performance — including hazard risk data, vulnerability calculations, and financial information. Additionally, the full model should be validated against historical losses through a method known as back-testing.

8. The model must be externally validated by credible third parties via proper channels.

The credibility of a third-party evaluator for peril risk model is built through a variety of factors, including but not limited to:

  • Independence
  • Expertise
  • History
  • Published methodology

Read the full paper at

Recent Stories

These Democrats have called on Biden to quit the race

Gaffe track — Congressional Hits and Misses

Trump’s presidential office hours were the shortest since FDR, Biden’s not far behind him

Biden admits other Democrats could beat Trump, but sends potential rivals a message

Photos of the week ending July 12, 2024

At high-stakes news conference, Biden calls Harris ‘Vice President Trump’