Search This Blog

Saturday, July 14, 2007

The 400 Year Storm

Traveling across the Twin Span bridge from New Orleans to the City of Slidell, I always found it odd and a bit frightening to see on the shoreline rows of piers abutting houses on stilts extending into the horizon. A good hurricane would reduce it all to kindling, but that was not in cards it seemed to those who lived on the shore. After all, the last major storm to strike this region was Hurricane Betsy in 1965, and the next 400 year storm was many years away. That was before August 30, 2005, when that perfect storm arrived, 360 years early.

So Katrina destroyed the coast, and by merely happening, changed the odds. Certainly for the insurance companies it did, as rates rose to match a more dangerous world. In actuality however, no one knows the true likelihood of when the next big storm will hit, and in lieu of time consuming explanations of events, which only with great labor can even attempt to reveal reliable probabilities (think of the global warming controversy), a better bet is to infer probabilities from a mere sampling of events. A quick rule of thumb way of assessing the probability of some event is to just extrapolate from a simple pattern in the past. This 'heuristic' method is called induction. The problem however comes when we assume that the success of our heuristics is due to some set of natural laws that we cannot explain. Often this may conflict with the 'true' odds as revealed through nature. Thus, although forty years without a storm does not predict next year's weather any better than the first year after the storm, we would be more comfortable with our beach house investment with the former case than the latter.

This thinking becomes even more illogical if we preselect the sample that makes a set of startling predictions post-hoc or after the fact. For example, the frequency of sun spots and the size of the hem of women's dresses can be shown to have a positive correlation with stock prices. Similarly, one may choose to build a vacation home in the Florida coastal town of Apalachicola. After all, it's never been hit by a hurricane, yet.

So why do we have a predilection for choosing positive post hoc predictions? It's not because logic demands it, because such predictions are at root illogical. A better answer is that is affective, or in other words is an intrinsically pleasurable thing. Positive post-hoc predictions are 'inspiring' if we perceive them, and 'motivating' if they engender our intention to act to like them. Thus, it is inspiring to see the exploits of a war hero, and motivating if we promptly enlist to follow his footsteps. (Of course, we will likely get killed in the process, rather than make a killing, as the next example also attests.)

The conflation of logic with affect takes perhaps its most ironic turn when we look to investor strategy. Consider the op-ed section of the Wall Street Journal, filled with articles brimming with academic acumen that invariably extol the fact that you can't beat the market, as your bright insight into a stock is already factored into the cost of the stock. The recommendation: just put your cash into a stock index fund and sit back and think of other things, like fishing. Of course, if all investors followed that wisdom, there would be little need for a Wall Street Journal. Hence the journal becomes a selection of Dr. Seuss tales for mature children, selected post-hoc because of their happy endings. That means of course an avalanche of stories about this or that investor who made the right call and invested in gold, hedge funds, pharmaceutical stocks, or sub-prime mortgages (wait, scratch that last one). Inspiring stuff to be sure, and motivating too! Just call your broker and get in on the action! Ultimately, whether investing success is dumb luck or skill is often decided because it's simply more fun to impute skill to dumb behavior, even if we know better.

Postscript: Such dumb behavior has been confirmed experimentally (as if we still need confirmation that people are naturally inclined to do dumb things-- and we do!). Consider the following experiment that demonstrates once and for all that a subscription to Field and Stream magazine is preferable to the Wall Street Journal.

SooChow Gambling Task
Researchers in Taiwan developed a fascinating gambling task (Soochow gambling task) with counterintuitive results. Even when experimental subjects know that they have a choice between a positive and a negative expected value bet (+$250 versus -$250), they still choose the negative option in large numbers (about half the time). When not told the odds, they do even worse. They lose all the money they are given at the beginning of the experiment, and still continue to lose. So what could persuade them to lose over and over again when they have a clear choice for how to make money? The negative expected value gamble usually results in a small win, but every 5 trials a big loss happens, and on average the subjects lose $250 per trial. The reverse happens in the positive expected value gamble - a series of small losses and one big gain. Even though they lose money, those 4 small gains are so valuable to people that they persist in choosing to play from that deck of cards. It seems to support Nassim Taleb's strategy of investing for catastrophe payouts.

No comments: