By Dr Marcus Schmalbach
“We could carry a local epidemic. A global event like a pandemic is simply not insurable.”
Michael Diekmann, chairman of the supervisory board, Allianz SE
There have been many discussions in trade journals in recent weeks and months on the subject of “pandemic/COVID-19/coronavirus.” We are in the midst of a world economic crisis triggered in a relatively unknown region in China. Who could have imagined such a scenario beyond the screenwriters of Hollywood? Unfortunately, such incidents are “the new normal.” We are confronted with impacts that nobody would have guessed, but the consequences are felt globally. In this context, one very often reads it is a “systemic risk.”
What is that? Systemic risk is a risk that can impair the functioning or continued existence of an entire economic system. While the specific risk or individual risk only affects certain system participants in isolation without endangering the system as a whole, the systemic risk spreads to other economic entities or systems, spilling over with “contagion,” and, in essence, a domino effect. In the insurance industry, one would speak of an accumulation loss – something that every insurance company in the world fears and that can lead to a financial collapse.
So, let us turn our attention to the crucial question of this article: Is insurance the answer to systemic risks? The answer is no. Why?
Law of Large Numbers – The Principal of Insurance
A small appendix and the description of the principles of “insurance” are warranted. Insurance refers to the basic principle of “collective risk assumption” (often referred to as the insurance principle or equivalence principle), whereby many people pay a sum of money (an insurance premium) into the insurer’s “capital collection center” to receive future compensation from this capital collection center when a corresponding loss, the insured event, occurs. Since the insured event will only occur with a small number of insured persons, the assets of the collection agency are sufficient if the premiums are paid. As a prerequisite for the insurance model, the extent of the losses can be statistically estimated, and the contribution required from each member of the collective can, therefore, be determined using actuarial methods.
By definition, insurance is based on two pillars: the “collective” pillar and the “assessability of risks” pillar. Is this possible in the case of systemic risk? No. The extent can be modelled with certainty and derived from previous catastrophes, but a collective approach is useless because many or most of all of the parties are affected and the insurance premium would, therefore, have to be 100 percent of the loss amount. The premium would have to be even higher because the insurer has additional distribution costs, administrative costs, costs for “claims settlement,” etc.
As an interim conclusion, the insurance approach is not the right one. Are there alternatives, or does the US government have to actually ratify PRIA, CRIA next time after TRIA? While these structures could provide an economic backstop, there are indeed other alternatives to be considered.
Traditional insurance is based on the principle of indemnification: a demonstrable loss against a definite asset. Take home insurance, for example. A house will be in a particular location and of a known size and built from specific materials, which makes its asset value fairly easy to establish. If the house burns down, a loss adjuster can estimate the damage, and this can be used as the measure for the claim payment. With parametric insurance, the payout is not linked to identified damage but instead to an index or set of parameters that gauge the severity of the event.
A loss adjuster will ask many questions in the claims process, such as the following:
- What caused the damage?
- When did it happen?
- What items were damaged, and can the insured prove a loss?
Parametric insurance does not require any questions like this. The simple fact that an index reached a specified level is sufficient to trigger the claims payment. Examples of parameters that can be used as triggers are rainfall volume or seismic intensity. So, for flood insurance, if the rainfall volume in a particular area exceeds a defined, measurable amount, a payout will be made without having to demonstrate that any flood damage has occurred. Likewise, in an earthquake scenario, if the seismic intensity exceeds, say, a 7 on the Richter scale, a parametric insurance contract will pay out even if there is no loss to compensate.
A claim is the “moment of consummation” in an insurance relationship. After all, that is the real product that is being sold—a contractual promise to pay. Parametric insurance has the ability to improve this relationship by avoiding arguments about causality and valuation and delivering a speedy payment. Knowledge, trust, and price were identified earlier as three reasons why customers may not be buying traditional insurance. Parametric insurance can deliver improvements in all three of the following:
- Knowledge. Parametric insurance is more transparent as it is based on a single identified numerical value equally understood by both parties.
- Trust. There’s no tricky “small print” or obfuscation around exclusions, causes, or damage. Payouts are streamlined and much faster.
- Price. It significantly reduces underwriting and claims settlement costs; these savings can be passed on to customers in the form of lower prices.
In addition to these, there are other benefits to parametric cover. There is a greater time flexibility as the contracts can be tailored for specific scenarios and do not have to be renewed annually. Typically, a parametric contract is multiyear, from 3 to 5 years’ duration. The normal insurance annual cycle requires exposures and asset values to be changed every year based on accountants’ reports and the like, whereas parametric insurance has no such limitations because it is not linked to underlying assets. Contracts can also be shorter than 1 year, for example, just covering the Christmas shopping season or summer holiday periods.
Parametric cover is based on inclusion rather than exclusion. A traditional insurance wording starts with a base premise and then carves parts out through detailed exclusions, deductibles, and limits/sublimits. The parametric approach remains at a high level. All that is required to be demonstrated is simply that the event happened, not what caused it nor what harm resulted.
Traditional insurance is well-suited to high-frequency, low-severity events aimed at households and small businesses. A multitude of small-scale losses is easier to model and manage due to the richness of historical data and the fact that the law of large numbers will enable accurate macrolevel predictions. Parametric covers in the past have been focused on low-frequency, high-severity events. It was initially developed in the form of catastrophe bonds to provide extra reinsurance capital for major disasters. Natural catastrophes, in particular, are relatively easy to predict and determine their probability of occurrence. But pandemics? Terrorist attacks? Global cyber-attacks? How do you design an insurance coverage concept for a completely unknown event? And how do you price it?
How Parametric Solution Works
The parametric approach removes the need for human judgment, investigation, and debate in the claims process and replaces it with an index-based trigger. While this gives many advantages in speed, validation, and transparency, it also has a potential disadvantage: what if the index does not properly match the risk being covered? This gap between customer expectations and the eventual outcome is known as basis risk. It is an imperfect correlation between the risk and the index. If the parametric product is poorly designed, the trigger level in the contract and the damage suffered by the client will be discrepant. There are two types of basis risk:
- Adverse basis risk. Damage occurs, but index is not triggered.
- Perverse basis risk. A “false positive payout”—index is triggered but no damage.
If there is too much perverse basis risk, then the product should not really be considered insurance or risk transfer and acts more like a derivative. Strictly speaking, insurance should not offer any upside reward. This balance of risk and reward is critical. A key principle of insurance is that insurers should offer premiums that are proportional to a customer’s risk. High premiums will attract only the riskiest customers, leading to greater payouts, thus reinforcing a vicious spiral leading to market collapse. This effect is known as adverse selection by insurers.
A major consideration in balancing these risks is granularity. If the product is pitched at a macro level, such as blanket countrywide coverage, then mismatches at a local level are very likely. Conversely, reducing scale to only encompass small localities is costly, cumbersome, and hampered by scarce and patchy data. So, finding the correct level of granularity at which to construct the models is critical to success.
Index Independence & Moral Hazard
Another key consideration is the independence of the index trigger. The index must be detached from any potential influence by either the insurer or the insured. There are two reasons for this. First, it eliminates any subjectivity over the payouts and also removes the risk of moral hazard. This is related to the concept of adverse selection, but where the latter addresses the type of product, moral hazard is concerned with actions. If the insured, through its actions, can manipulate the index so that it rises above the trigger level, that would constitute moral hazard.
Indexes based on climate or geological data are safe from moral hazard as customers are unlikely to be able to make the wind blow harder, the rain last longer, or an earthquake more violent. They could, however, tamper with the local measuring equipment, so it is important that the reporting body is a trusted, independent entity. Other measurement factors are less robust. Corporate financial figures like revenues and profits, though independently audited, can be quite subjective and rely to a great extent on the honesty of the corporate in question. The huge accounting discrepancies that bankrupted Enron and Carillion are salutary lessons in this regard. Likewise, cost-based measures, where the spending is under the control of the insured, are rife with moral hazard. Why not spend as much as you can, if the insurance company is going to pay for it?
A second consideration to take into account is the risk of index failure. What would happen if the index could not be calculated on a particular day? Maybe the extreme weather has damaged the sensors so no readings can be taken, or a cyber-system failure creates a gap in the data record. What interpolation method is specified to patch the void? Are there any backup providers? These types of eventualities need to be appraised and mitigated when defining the parametric trigger.
Conclusion & Outlook
In summary, it can be said that the insurance approach is not suitable for systemic risks. And, although the statements of the insurance experts caused a lot of resentment among the population, I completely agree with Michael Diekmann that a pandemic is not insurable. However, this does not mean that it is not transferable! Are people facing global catastrophes and systemic risks the need to insure them slowly ending traditional insurance companies?
Tatjana Winter, an expert on digital transformation, evaluates this thesis as follows.
“The insurance industry is transforming very slowly, and the focus has so far been almost exclusively on distribution. Other industries are significantly further along in this process. However, parametric solutions can speed up the digital transformation very much, because things have to be developed from scratch and can be rolled out digitally from the beginning.”
The future of risk transfer- especially for systemic and catastrophe losses – will be parametrically driven, and indemnity insurances will become obsolete. It will be exciting to see whether traditional insurers drive this development or whether they will also be obsolete in the end and tech companies – in cooperation with the capital market – bring the solutions to the market faster and more convincingly.