This is the final blog in a series of four blogs examining three potential “protection gaps” and the importance of “protection gap analytics”. To read the first blog post in this series, click here.
We are not going to be able to take effective action to reduce any of these three protection gaps unless we can first learn how to consistently measure the difference between insured and total loss. Such measurement means we can know the current situation as well as set appropriate targets and monitor progress in reducing the gap. It can also help to focus investment and action.
At present, the only form of measurement is to acknowledge the difference between insured loss and the estimated total economic loss once the claims have settled, one or two years after a significant disaster.
In the same way that probabilistic catastrophe risk models were developed to enable insurers and reinsurers to look beyond the latest event loss, so the same models are now required to monitor the protection gap. This is the focus of “protection gap analytics”.
The calculation of total loss (or “economic” loss) can itself be something of a mystery as no agency gathers together all the costs in the way that insurance loss is the sum of all the claims paid. In the U.S., for example, the National Oceanographic and Atmospheric Agency (NOAA) calculates economic loss from applying a formula to the total insured loss.
What should be included within protection gap analytics and how does one go about generating the relevant information?
Starting with the “high-risk protection gaps” found in high income countries, in most situations we are already likely to have the probabilistic catastrophe loss models for that region and peril. Into the specific peril model, we can enter data on the total industry insured exposures as well as the total economic, or “insurable” exposures. Such data will need to be as detailed as possible. For residential exposures we will need the location and values, information on building type, construction materials, and age of the building, as well as the value of what is contained within the property. In addition for commercial policies, we will need information covering income and profits to explore the impact of business interruption.
In an emerging market, there is less likely to be a preexisting probabilistic catastrophe loss model. In which case it may be necessary to select one or a small number of scenario events, or even a well-resolved historical event. In an emerging economy there may also be challenges around gathering data, including exposure data on the building stock, as well as knowledge of what is insured.
For intangible risks in commercial policies, these are early days for having fully probabilistic modeling capability as well as complete industry wide exposure data. We should start with some scenario events, and then attempt to evaluate how much insurance is currently in place to pick up the losses, as well as what would be the total economic consequence of a particular incident.
Separating the “Penetration Gap” from the “Coverage Gap”.
For each peril and region we can split the total protection gap into two components: the “penetration gap” and the “coverage gap”.
The “penetration gap”, at its simplest, could be the proportion of people without insurance for that peril. More completely it could be defined as the proportion of that peril’s risk that is not redressed by insurance, assuming that all policyholders were fully covered for their potential losses. The difference between the proportion of properties insured, compared with the proportion of the risk insured, would reveal whether insurance take-up rates are higher among those with the highest value properties or with the highest levels of risk of monetary loss. From the perspective of the insurer this gets termed “adverse selection”, although for the consumer this is informed insurance purchase.
In identifying the penetration gap for hurricane, we would need to consider individually all the different perils that combine to generate the loss: wave damage to offshore platforms, coastal storm surge, direct wind damage and inland rainfall-related flooding.
The peril specific “coverage gap” is the proportion of the risk that is not covered by that policy. The coverage gap includes two components — the impact of exclusions and what part of loss repayment is not provided because of the financial terms of the coverage.
Examples of exclusions could be an earthquake policy not covering “appurtenant structures”,external to the building, such as swimming pools and patios. Peril-specific exclusions, like excluding flood in a wind policy, would be picked up separately when measuring the penetration gap for that additional peril.
The financial terms of the policy could include the size of a deductible, or the loss threshold provided by a limit to the coverage, as well as any co-insurance. There will also be a gap introduced when the insured values are lower than the cost of reconstruction for the property.
Overall the “coverage gap” is determined by how much of the risk the insurer wants to cover in the structure of a policy. Meanwhile the “penetration gap” reflects the attractiveness of the policy to the homeowner, or whether insurance is mandated. Clearly the two gaps will be interdependent. The more that exclusions and financial terms restrict the coverage, the less likely that consumers will purchase the policy.
When concerned with impacts arising from separate perils, such as hurricane wind and storm surge, then we first consider the coverage and penetration gaps per peril and then combine these to identify the total hurricane protection gap.
The Protection Gap EP
In any region, the protection gap — measured as the percentage of the economic loss not covered by insurance — is likely to vary according to the severity and geography of the catastrophe. This variation could be expressed as a protection gap exceedance probability “EP” curve, revealing how the percentage of uninsured risk varies by return period.
The protection gap could also be output as a map, indicating how the percentage gap varies across a region. For example, with hurricane, one might find a larger protection gap where the principal agent of loss is coastal or inland flooding.
According to the use of policy deductibles, protection gaps should reduce for more extreme events where the insurer pays a higher proportion of each claim. For more frequent events, the insured can expect to pay a greater proportion of the loss.
However, the application of limits on total loss per policy will have the opposite effect, widening the protection gap for the most extreme losses. An insurer may be less interested in helping expand coverage and shrink the protection gap unless they can be confident that appropriately priced reinsurance will be available.
There should be standardization in measuring and reporting the protection gap. This should include analyzing insured and economic loss using a probabilistic catastrophe model, separating the effects of the coverage gap from the penetration gap and identifying how the gap varies with annual probability and location.
With access to the appropriate demographic data we could dig deeper into those factors that determine insurance purchase. For example, how does the purchase of insurance vary according to income level or age of the homeowner? How does this proportion vary between occupiers and landlords, by whether the homeowner has a mortgage or the value of the property? For SMEs this could also include what proportion purchase business interruption insurance, and how this relates to the value of the business.
Such demographic analysis will be important when it comes to designing the actions to be taken to reduce the protection gap.
Planning Action on the Protection Gap
A full analysis of the protection gap and its components should be required for all pre-existing catastrophe risk insurance schemes, to provide a baseline understanding and to rank how different schemes compare to one another. In introducing a new or modified insurance product there should be a standard procedure to demonstrate what this achieves in terms of reducing the protection gap.
Having begun by separating the coverage gap from the penetration gap, an insurer could explore how insurance take-up expands as a result of reducing the gap in coverage within the terms of the policy, as by lowering deductibles.
The reality of the “protection gap” is more complex than the simple slogan “to close the protection gap” implies. We should avoid setting generic targets across a country for “closing the gap” but rather make these goals specific to the sectors most in need of assistance, in particular the poor. The focus on economic loss inevitably skews the analysis to wealthier sectors. Helping those on middle incomes acquire insurance may even exacerbate the depth of inequality in a society. The InsuResilience initiative has set itself the goal of “providing access to insurance” for up to 400 million additional people by 2020. Perhaps it would be better to measure new insurance schemes in terms of how much of the total risk is now being covered by insurance as well as how the protections benefit the poor.