Trigger Trouble

Is there an insurance trigger politics?

Recent news reports caught my eye for renewed attention to an important and technical contractual mechanism: insurance triggers.

Writing in the Wall Street Journal, reporter Telis Demos, does a fine job covering constructions of trigger events. The story is about the Los Angeles fire- or fireS- with the distinction being important for how losses are distributed within the industry and potential consequences for the future cost of cover in fire prone states.

Demos notes a the 9/11 attack on the Twin Towers as a similar situation:

How disasters are designated has been contentious in the past, including after the Sept. 11 attacks. The developer that controlled the World Trade Center at the time fought with insurers for years over whether it could receive coverage for two occurrences rather than one after planes hit each Twin Tower.

Who decides how many fire “events” occurred in LA this past month? Demos posits,

Verisk PCS, a provider of catastrophe loss information that is often used as an industry benchmark, has designated two separate events, the Palisades fire and the Eaton fire. The California Department of Forestry and Fire Protection, known as Cal Fire, is also tracking them as two incidents.

I will guess that the CalFire process for incident management was not set up to act as official mechanism to decide about who bears the brunt of billions of insured losses.

In Bloomberg, an investment manager explains that the triggers are set to favor investors. Explains one investment manager, “Investors have been able to obtain more robust structures, making cat bonds safer.”

The journalist explains:

Over the past five years, however, trigger events have become less common as fund managers use increasingly advanced models to set the terms of the bonds they hold. At the same time, investors have been pushing for higher interest payments in return for taking on the risk posed by higher inflation, rising property values and climate change.

In niche internet circles, commentators noted the story supports past speculations that the triggers are biased to the interests of investors and this is particularly problematic for developing countries.

There is an established literature in the social science of insurance. Because insurance is forefront public concern it is relevant to note that it is also a subject of critical study beyond the business and finance disciplines.

Classic texts of political science emphasize that media, publics, and political elites shape- or construct-the meaning of facts, actions, and events. The benefit of examining events as constructs is the opportunity to understand when and why controversy arises around triggers and how to make them more robust.

Significant advances in the study of risk and decision making were made on the grounds that risk is a social construct. Esteemed psychology researcher Paul Slovic quipped that “Danger is real, but risk is socially constructed.” Understanding risk as a construct helps explain differences in the perceived value and uptake of insurance cover for high loss, low probability events, for example.

Here are four cases of insurance trigger problems stemming from the different ways events are constructed in society versus in contracts: Hurricane Sandy, Pandemic, Food Insecurity, and Cyber Attacks.

I suggest considering challenges in the design of insurance triggers using the two continuous dimensions: Clear Criteria and Recognized Authority. The industry and publics are using different criteria and sources of authority to realize that a particular event occurred¹.

1. Was Superstorm Sandy a hurricane landfall?

Windstorm insurance policies commonly contain two different deductibles. One deductible applies to wind losses generally. The second deductible applies to hurricanes specifically. The so-called ‘hurricane deductible’ is much larger than the general wind loss deductible and presents additional cost burden to homeowners.

Hurricanes may become extratropical storms but the societal experience of either may be the same.

To make a long story way shorter…

Because of Sandy’s extratropical status at landfall and the lack of official hurricane watches and warnings the storm did not serve as a triggering event for the hurricane deductible in states north of North Carolina. Billions of dollars that would have been shouldered by property owners had the hurricane deductible been triggered were instead shouldered by industry.

The high stakes attached to the decision about Sandy’s categorization at landfall was felt by both weather agencies and policymakers of affected states.

A couple of days after Sandy’s landfall the National Weather Service announced an assessment committee to review its performance in the storm (something that is usually done well after hurricane season ends). When the assessment was announced policymakers put pressure on the weather agencies to maintain the original decision of extratropical status.

The agencies terminated their assessment maintaining the storm’s extratropical status.

2. When has a pandemic occurred?

The financial industry plays a role in providing humanitarian and development aid that mobilizes resources to limit spread and treat the ill to reduce deaths. In 2017, World Bank developed the Pandemic Emergency Financing Facility (PEF) bond after West African nations struggled to fund response to the 2013-2016 Ebola outbreak.

Critics of the PEF argued that the way the triggering event was characterized did not align with the way people experience public health crises. By characterizing the triggering event through number of deaths and rate of spread, the bonds provide funding for response to an already dire situation.

The World Health Organization (WHO) declared a global pandemic from COVID-19 on March 11 but even by the end of the month the PEF still had not triggered.

AIR Worldwide, a private catastrophe modeling firm, served as the PEF’s calculating agent. Public pressure mounted for AIR to justify its position that the triggering event had not occurred despite the 100,000 deaths from COVID-19 and a WHO pandemic declaration.

In March, AIR took over two weeks to determine that the final criterion of the triggering event- exponential growth rate in the countries poor enough to be eligible for funding- had not yet been met. The company held this position until April 17, when it reported that the final criterion had been met.

The PEF was set to expire in July 2020 at which point it would provide investors with a 13% return. The timeline created speculation of internal pressure to delay triggering coverage which would present a loss to investors.

3. Is drought a sufficient proxy for food insecurity?

Malawi purchased a parametric drought insurance policy from the African Risk Capacity Insurance Company (ARC) for the 2015/2016 agricultural season. The policy defined the triggering event as 1.39 million people effected by drought and estimated this to have a probability of occurring, on average, once in five years.

Drought ensued and in April 2016 the Malawian president declared a state of disaster. The nation’s maize production had decreased by 12% and there was not enough food to feed the 2.8 million Malawians, about 20% of the population, that were food insecure.

However, the event did not trigger the insurance coverage because ARC’s calculating agent, its risk model, Africa RiskView, estimated that only about 21,000 people were at risk in the central portion of the country rather than the millions affected in its southern region.

ARC said that the model did not identify the food insecurity being experienced because of incorrect assumptions about the type of crop farmers were planting.

ARC then changed its model assumptions to produce “a reasonable representation of the situation on the ground” and triggered the bond. By the time ARC announced a payout of USD $8.1 million in November 2016 and finally delivered in 2017, critics argued that it was too little, too late- people had already starved.

What seemed to stand out to critics most urgently however was that the way the model provided judgement about food insecurity did not, and likely cannot, account for shifting social dynamics that increase a population’s vulnerability to extreme weather. Because of poor harvest in prior seasons many had tenuous food stores making them highly vulnerable to even minor fluctuations in weather.

3. When is a cyber attack an act of war?

A leading problems in expanding a catastrophic cyber insurance market is developing a shared understanding of the cyberattacks that have transpired. Insurers have tried to manage this issue through the use of exclusions, definitions of terrorism, and war clauses. Conflicts arise around the interpretation of these boundaries because each event has unique characteristics and interpretation carries high stakes of financial and diplomatic relevance.

In 2017, the global cyberattack, NotPetya, caused $3 billion or more in insured losses. While some insurers paid out on the claims they receives others did not citing the policies’ war clause to deny the claims.

Merck, a pharmaceutical company, filed suit against their insurers for the denied claims. The courts decided that the war clause was not intended for cyberattacks but for physical acts of warfare. The cases settled. News reports indicate that ambiguities of the war clause in the case of cyberattacks persist.

Similarly, using network software supplied by the company SolarWinds, hackers conducted an enormous cyberattack in 2020. The SolarWinds attack affected critical and high power government institutions and businesses.

Attribution of the SolarWinds attack was a point of political controversy.

US intelligence authorities attributed the SolarWinds attack to the Russian Foreign Intelligence Service. The Russian Embassy in Washington denied responsibility. President Trump posted to Twitter that the SolarWinds attack was overblown and suggested the source was China.

The US Senate Intelligence Committee discussed uncertainty in the information underpinning decisions of attribution. Clearly, Congress did not make a war declaration.

These examples are situations where an insurance trigger became contested and entwined with societal constructions of the event. A critical view situates the cases as a spectrum of events in which the industry and public are using different criteria and sources of authority for enactment of an event.

In the below 2x2 matrix the X-axis is a continuum of the extent to which industry and society recognize the same decision making authority. The right side describes situations where there is an institutionalized and authoritative mechanism to pronounce an event’s enactment. The left side are situations where public and private market interests look to different authorities to determine when an event has occurred.

The Y-axis represents a continuum of the extent to which an event has shared agreement over clear criteria. The bottom means a lack of agreement on criteria. The top means clear and widely accepted criteria.

A quick analysis

• Sandy is situated in the upper right hand corner of the matrix. In the United States, tropical meteorologists at the NHC have the authority to make decisions about tropical cyclone weather characterizations and there are clear technical criteria for doing so. Still, the process requires expert judgement which became subject to political pressure because of the financial stakes and proximity to land.

• War is situated in the bottom right quadrant of the matrix. Declarations of war are made by Congress but beyond their declaration there are not objective measures for distinguishing war from ongoing conflicts. This makes it difficult to delimit cyberattacks from war before such divisions are declared by Congress which are in turn dependent on public sentiment.

• The COVID-19 pandemic case is situated in the lower left hand side of the matrix. Public sentiment turned to the authority of the WHO for declaration of a pandemic and to local health care workers for indication of a crises. In contrast, industry turned to a private model vendor. Even still, there were no well-established criteria for the threshold of pandemic. As well, WHO declarations have some degree of attachment to international relations.

• Realizing a state of food insecurity requiring humanitarian aid is tied to social and political dimensions that are not predicted by drought alone. Political actors and humanitarian groups played an important role in identifying the food insecurity emergency. However, ARC relied on their risk model. Nonetheless, there was a shared sense that an emergency had manifested that the model failed to capture. This is less than an ideal sense of a shared acceptance of clear criteria.

What to do?

Triggers are more robust the better their characterizations mirror the decisions made by society’s authoritative voices on the matter. A designated authority can be created where there is no existing formal institution or the recognized institution is not well equipped to withstand pressure created by the significant financial dynamics. For instance, it is likely unwise to use CalFire as a delimiter of events when their priorities are incident command and public safety. An independent body created among industry interest but maintains transparency may be well suited to make needed decisions on the characterizations of cyber attacks when they occur.

Reducing the range of judgement required is also useful. After the Sandy debacle hurricane deductibles turned to the use of a ‘named storm’ rather than a forecast watch or warning. The move does not eliminate the use of expert judgement but limits it because the point at which storms are named is less ambiguous then deciding the point in space and time when a hurricane becomes extratropical.

Risk models are a good source of criteria that are (in theory) resistant to external pressure. But their proprietary aspects prevent transparency which means the public will not likely look to a modeling firm to tell them what sort of event they’ve experienced. Also a model creates the optics of a deeply bankrupt morality when used to disperse aid to people facing starvation and disease.

1. This post is an excerpt from a paper I was working on and I cut references for ease of posting here. I’m not sure I’ll finish the paper- maybe, if I get some good comments/critiques!