The perils of "unknown unknowns"

Robert Shiller has long advocated the use of structured insurance and hedging instruments to mitigate long-term risks from climate change and natural disasters. He writes in a new Times article,

We need to worry about the potential for greater-than-expected disasters, especially those that concentrate their fury on specific places or circumstances, many of which we cannot now predict. That’s why global warming needs to be addressed by the private institutions of risk management, such as insurance and securitization. They have deep experience in smoothing out disasters’ effects by sharing them among large numbers of people. The people or entities that are hit hardest are helped by those less badly damaged.

But these institutions need ways to deal with such grand-scale issues. Governments should recognize that by giving these businesses a profit incentive to prepare for these unevenly distributed disasters. After all, fire insurance does no good unless you buy it before the house burns down. And you have to diversify your portfolio before the stock market crashes.

And about the prevailing state of the market for hedging against such disasters, he writes,

They tend to focus on relatively short-term risks, and don’t hedge against the increasing cost of disasters over distant future years. Yet if the problems of global warming become more serious, they will very likely be long-lasting, raising some complex, tough-to-quantify issues. Some kinds of crises, like hurricanes, may remain intermittent, but their tendency toward severity may build in a slow, hard-to-predict process and in complex geographical patterns.

I think Prof Shiller himself displays a market cognitive bias here - the assumption that markets can provide the risk mitigation solution for such complex risks. Consider this data snippet,

In the 1960s, 1970s, and 1980s, catastrophe loss costs in the US property casualty insurance industry were about 1% of written premiums. That jumped to approximately 3.5% in the 1990s and in the 2000-2010 period, and doubled to 7.2% from 2010-2013. This was due in large part to catastrophic weather from 2008-2012, including inland wind (tornadoes), hail, and coastal wind (hurricanes), which had become more difficult to predict and model.

I think it is impossible, even for sophisticated insurers and reinsurers, to model such Knightian uncertainties. This is all the more so with assuming long-term risks. Climate trajectories, especially at local levels, are subject to mutations (abrupt changes) which could render previous assumptions completely irrelevant. Once the risk materializes, its holders are left with no options, howsoever much they have hedged for it.

Since the holders of such insurance are citizens, governments would then be forced to step in.  The too-big-to-fail associated with large insurers (who are most likely to be counter-parties to such risks) would anyways make bailouts inevitable. Insurers and reinsurers who have internalized this belief would therefore be less than rigorous with their under-writing standards. This moral hazard at the insurer's end is mirrored at the buyer's end. The availability of insurance whose price is presumably less than the true cost of the risk assumed would encourage people to build on vulnerable areas and so on. The systemic incentive distortions would be considerable.

This is not to suggest that markets have no role to play in mitigating such risks. Far from it. Insurance products can be used to mitigate several risks, especially shorter-term natural disasters. And there is an active market in such products.

But when faced with "unknown unknowns", there are limits to how much smart we (or us collectively, in the form of markets) can be. In some ways, the belief that such uncertainties can be hedged is similar to the classic belief in "free lunch". In fact, the insurance industry itself, in the words of a leading insurance industry think tank Geneva Association, has described such risks as "uninsurable".

Such uncertainties are best hedged by plain-simple regulations - identification of vulnerable areas, with specific risk probabilities, and rules that limit construction and economic activities. The costs of any disasters that go beyond these (ie one which hit even ex-ante less vulnerable areas) are best covered through some form of direct public payouts (even with its moral hazard) than through the market mechanism.

Some times the transaction costs and incentive distortions associated with risk diversification become too high that it more than offsets the benefits from diversification. And long-term climate risk mitigation may be one such case.