Sunday, January 25, 2009
Posted by Jeff Lipshaw
Although I sometimes wonder if all the various behavioral psychology theories ultimately cancel each other out (sort of like Karl Llewellyn's famous table of contradictory construction axioms), Gretchen Morgenson's New York Times column on credit default swaps got me to thinking this morning about optimism bias. This is the documented tendency of human beings to over-estimate the likelihood of positive future events and under-estimate the negative. I suspect optimism bias is, for example, a personality staple for all entrepreneurs. The other tendency (does it cancel out optimism? I'm not sure) comes from prospect theory, which says people evaluate prospective outcomes based on the change from a reference point, not on the absolute expected utility of the outcome. The idea here is that people evaluate the risk of positive and negative outcomes based on how things will change from the present reference point. What is important to understand here is that the weighting of positive and negative outcomes (i.e. risk assessment) may change based on the size of the outcome. I may be perfectly happy to risk $100 for an evening of blackjack at the casino (the risk of my losing $100 is far greater than chance I will win $100), whereas I may weigh outcomes quite differently if you ask me to take all my savings and invest it in one very risky start-up company.
I decided a couple days ago that I wanted to lose ten pounds. These are "vanity" pounds, so it's not easy to lose them. Weight loss is actually a pretty digital affair: you will lose roughly one pound for every net 3,500 calorie deficit. So if you burn on average 2,300 calories a day, and you restrict yourself to 1,800 calories of intake a day, you will lose about one pound a week (7 x 500 calorie deficit per day). Consistent with the behavioral observations, there is optimism bias. When faced with the decision how much ice cream to cram into the dish, I under-estimate the future cost. The solution is regulation. Diet experts tell you to keep a log of what you eat, and count the calories (or the Weight Watchers points). But the optimism bias is so strong, even then I'm inclined to understate in the records how much I'm really consuming, either by using the most optimistic data, or by understating how much I'm really consuming. Example: unadulterated air-popped popcorn has 31 calories per cup. Orville Redenbacher's "Smart-Pop" microwave stuff, which says it is 94% fat-free (note that it still contains oil and butter), claims 15 calories per cup on the box. I'm sorry, but that cannot be, and while I am not accusing Orville of lying, somebody is trying to appeal to my optimism bias!
What Ms. Morgenson observes this morning seems reasonable: there is a place in the world for credit default swaps, just like there's a place in the world for any kind of insurance, whether it be fire insurance or currency hedging. It does, however, seem plausible to me, for reasons explained by optimism bias and prospect theory, that anybody making lots of small bets will under-estimate the risk of a negative outcome (paying on the swap or finding more fat around one's middle) against the present utility (getting the premium or eating the Ben & Jerry's Yes Pecan). Back in the 1980s, I was up to my eyeballs as a working lawyer involved in insurance company insolvency, and a fellow by the name of Charles McAlear in Michigan wrote a little book called The Foundering Ark about the state of the property and casualty insurance industry. What he wondered was why an insurance company given an "A" rating by A.M. Best (the equivalent of a credit agency rating) could be in receivership a year later. The answer had to do with the relationship of minimum capital requirements and loss reserving. In a nutshell, loss reserves (which in the insurance industry constitute costs or liabilities) are merely estimates by human beings. If you regularly under-estimate your reserves, say, by 20%, it's entirely possible that you have real world liabilities or expenses that would wipe out what you have otherwise stated on your books as your net worth. McAlear's point was that the industry (a highly regulated one, by the way) was fraught with systemic under-reserving of losses.
Moreover, each instance of underwriting is little compared to the whole business, and thus more like tossing $100 at the roulette wheel than staking your entire fortune on your cousin's idea to start an online dog-washing service. That's the prospect theory part of it.
Why, then, is it not a surprise when Ms. Morgenson observes that "[s]ellers of C.D.S.'s spent years raking in premiums while underestimating or simply ignoring the possibilities of rising defaults. Regulators let the market grow unchecked. . . . In the end, far too much of this insurance was written at way too cheap a cost"? I'm not into the blame game, so I will leave for others demonizing boards and management who oversaw companies who did it, but there are indeed regulatory systems in place for industries like banking and insurance in which there is, or can be, a systemic "over-optimistic, small bet" bias. Shareholders like to eat their ice cream now. Managers will respond to that, even if they aren't the demons they can be made out to be.
A little Coaseian economics about regulatory solutions below the fold.
This is how we dealt with the cost of casualty insurance failure, at least when I was practicing it some twenty years ago. There was no federal bankruptcy (as is the case today) for insurance companies. Failed companies when into receivership in their domiciliary state. Each state had a "guaranty association" chartered pursuant to statute. Most all of the them worked on what was known as the "post-insolvency assessment model." I'll use Michigan as an example. Let's say the Failsafe Insurance failed, and went into receivership in New York. The Michigan guaranty association statute governed losses related to Michigan policyholders of the failed company. The GA employs a group of adjusters who step in and calculate the expected losses for Failsafe's Michigan policies, and assess the rest of the insurance companies registered in the state pro-rata in accordance with the amount of premium they have written in the most recent year. (Michigan was interesting because it excluded GA coverage for policy holders with a large net worth, on the rationale either than they could bear the loss themselves, or should have been sophisticated enough to do a better job choosing their insurers).
Now, all the other insurance companies were entitled to count these assessments when setting their own future premium rates for insurance. In other words, the state didn't bail out either the insurance company or the policy holders. The good news was that the initial charge was placed on the lowest cost avoiders - the insurance companies themselves - who might or might not have to absorb the cost or pass it on to customers depending on competitive state of the business. The bad news was that if the costs couldn't be passed on, and the companies wanted to keep their profitability at the same level, this might well spur additional under-reserving and more systemic insolvency. But overall, while running in cycles, the system seemed to hold up, probably because the total failure rate and the ensuing assessments were still a very small component of the industry's costs.
The difference, however, was that there was solvency regulation. It didn't stop insurance companies from going belly up, but it did have to have a salutary effect. I suppose this is all another way of saying that the problem here is the migration of a market system under relatively benign regulation to one entirely unregulated. But doesn't an enforced system of counting, like my little calorie-counting notebook, force us to consider the biases at work?
[Clipart - (c) 2009 www.clipart.com]