SIGMA Actuarial Consulting Group’s Jason Luckett and Rich Moncher discuss property analytics for the captive insurance market
The property insurance market is evolving, and analytics can support effective decision making in a rapidly changing environment. Property risks are now frequently being integrated into the captive space to stabilise or reduce costs and customise coverage. Captives often start out with low property coverage limits that grow over time, as surplus becomes adequate to support higher amounts of coverage.
Laying the foundation Looking back on experience
Property risk can be challenging to analyse when compared to casualty coverages such as workers compensation, general liability, and automobile liability due to the significant volatility in year-to-year loss experience, as well as the evolving landscape of natural disasters and weather-related catastrophes.
Larger companies may have credible property loss experience for some perils (e.g., non-weather water losses, crime, etc.) which can be analysed relative to historical exposures (such as total insured values) and used to forecast future expected losses.
- Total Insured Value (TIV) — the total value of the insured’s combined property, business income, and any other items (equipment, inventory, etc.) covered by a policy.
It typically represents the maximum payout an insurer will make to the insured in the event of a total loss. However, some property exposures can be subject to catastrophes at a scale that has not been captured in historical data due to a loss of that magnitude not yet occurring or historical records not going back far enough in time. Furthermore, the risk a property faces can change over time, with a recent example being the rising frequency of wildfires in the Western United States.
For these reasons, even when there is historical loss data for a property-related risk which enables the use of standard actuarial and analytical methods, it is important to supplement this data with resources that consider information outside the scope of history.
Expanding the view Using external data in property analytics
There are numerous resources available to use as benchmarks for property analytics.
Within the US, nationally managed databases, such as the National Centers for Environmental Information (NCEI) and the US Geological Survey (USGS), contain information on the frequency and severity of natural disasters, extreme weather events, and the property damage caused by them.
This data includes locations, dates, storm types, and more, which can provide key insights on the historical frequency and estimated losses caused by storms across the US.
Other increasingly utilised resources for property analytics are catastrophe models which utilise industry-wide insurance data and tens of thousands of iterations of simulated losses for a given exposure to determine average annual losses and probable maximum losses.
- Catastrophe models (CAT Models) — a model of low-frequency events with potential high-severity or widespread effects.
- Average Annual Loss (AAL) — the expected (mean) annual loss for an exposure.
- Probable Maximum Loss (PML) — an estimate of the largest loss that a risk is reasonably likely to suffer.
It is important to note that the probable maximum loss is not the maximum possible loss.
CAT Models include four components: the event module, hazard module, vulnerability module, and financial module.
These models simulate a wide range of potential disasters (event module), translate those simulated event scenarios into physical impacts (hazard module), evaluate the extent of damage to insured properties based on hazard intensity and specific property characteristics (vulnerability module), and compute the financial implications resulting from the simulated events while accounting for property damage, policy terms, and reinsurance contracts (financial module).
An AAL provides information related to the expected loss for an exposure year, which can be useful in consideration of loss accruals.
A PML, on the other hand, enables an understanding of the cost for a reasonably likely worst-case scenario, which can be useful in considering solvency requirements and determination of appropriate loss limits.
Reading the room Keeping an eye on market dynamics
Ideally, property risk analytics will merge (weight) the unique historical experience with benchmark information. However, an additional consideration that may come into play is how the market and/or a captive regulator views the risk.
If an AAL for an exposure is US$10,000 and that exposure represents a potential loss layer of US$100 million, would the market write that risk for only US$10,000 or would a regulator approve of a captive writing that risk for only US$10,000?
This scenario seems unlikely, highlighting another key challenge with property analytics: aligning quantitative data (historical loss experience, AAL, PML) with qualitative considerations (market pricing, minimum funding amounts).
When computing property analytics, market premium quotes can be used as a reasonableness check on results obtained from the use of historical loss experience and/or catastrophe modeling.
The intent is to ensure that the loss estimates line up with the information that the market is providing. For situations in which loss estimates diverge from market pricing, it is important to revisit the underlying assumptions of those estimates to ensure that they are sound and align with their intended purpose.
An additional consideration is a reinsurance concept known as rate on line (ROL), which is utilised to compare exposures to premiums for the purposes of assessing risk transfer.
- ROL — the ratio of the reinsurance premium to the reinsurance limit for an exposure.
In the scenario above, the ratio of premium (US$10,000, assuming premium has been set equal to expected losses) to the limit (US$100 million) of coverage results in a ROL of 0.01 per cent, which is unlikely to be acceptable to a regulator.
While there is no specific textbook minimum ROL, in our experience a rule-of-thumb would be to target 10 per cent or higher.
This could vary by type of risk/exposure and would be an important item to discuss with regulators.
The inverse of the ROL is known as the payback period, which is used as a measure of how long it may take for a reinsurer to recover the cost of a loss.
- Payback period — the number of years it would take for the reinsurance premium to account for a full limit loss.
In this scenario, the payback period would be 10,000 years, in other words, it would take 10,000 premium payments with no associated losses to fund one full limit loss.
When evaluating the payback period, it is important to consider the return period, which presents as a counter to the payback period and provides a competing view from a loss frequency perspective.
- Return period — the estimated number of years between loss occurrences of a given magnitude.
If there is a less than 10,000-year return period, then there is a potentially likely scenario in which an insured is faced with multiple occurrences before enough premium has been collected to cover the expected cost of a single occurrence.
This scenario may provide some clarity as to why a regulator could be reluctant to sign off on a policy providing US$100 million of coverage for US$10,000 of premium.
Seeing the full picture A unified property analytics approach for captives
As we have discussed, there are many important factors when computing property analytics.
It is ideal to consider historical loss experience for an exposure, but that experience does not always guarantee that you are seeing the full scope of potential losses and may need to be supplemented.
Benchmark information such as nationally managed databases and catastrophe models can provide reasonable projections of expected average losses and potential worse case scenarios, but these estimates may not align with the marketplace and may require some adjustment to account for market considerations.
Market information such as premium quotes and ROLs can provide a starting point that regulators may be comfortable with, but they may lack accuracy relative to actual experience.
The most successful property analytics blend data from all these sources together and, in doing so, will take advantage of the full spectrum of information available within the marketplace.
