Irrational Economist_ Making Decisions in a Dangerous World - Erwann Michel-Kerjan [13]
In explaining these concepts, we cite the example of a mythical and obsessive character called Klaus who every day charts his commuting times to and from work (using the subway) so that, over a certain period, he accumulates a statistical representation of his travel times in the form of what is essentially a normal distribution—as shown in Figure 3.1. Indeed, assuming that nothing systematic disturbs the statistical pattern observed, Klaus can use the characteristics of the normal distribution to calculate the probability that his arrival time on any given day will fall within specified limits based on past history. Moreover, he can actually validate his model’s predictions every day when he goes to work.
By “subway uncertainty,” then, we mean a source of uncertainty whereby the statistical properties are well known and future “surprises” fall within well-specified limits. This kind of uncertainty—or approximations thereof—can be well handled within our decision-theoretic and forecasting models.
By “coconut uncertainty” we mean something quite different. Imagine you are sitting under a palm tree on a South Seas island and a coconut happens to fall on your head, causing considerable distress. Now, there are many disasters that you can imagine in life, but being hit on the head by a coconut when you are on vacation is probably not one of them (until you read this story, at least). In other words, by realizations of coconut uncertainty we mean events that you probably never even imagined could occur—there are so many different ones and you don’t know which particular one will happen. Indeed, you might not even have a good handle on the class of events that could be described as coconuts.1
Interestingly, although there is a history of coconuts in some domains, people are still surprised by their occurrence. Daily returns on the stock market provide a case in point. Figure 3.2 shows the distribution of daily returns of the Dow Jones Industrial Average (DJIA) for the period from January 1, 1900, to December 31, 2007. As you can see, many observations lie outside the plus-or-minus-three-standard-deviations limits, and this graph does not even contain the observations for the wild market movements that occurred in 2008 and 2009. Parenthetically, if you look at the same data as a time series, there are periods—mainly at “crisis” times (e.g., October 1987 or 2001)—during which there seems to be dependence in the size of fluctuations, rather like the pre- and after-shocks that accompany earthquakes. In principle, in considering the stock market and other financial data one could always model some coconuts; but for some reason, many practitioners fail to do this and inevitably suffer the consequences—as evidenced by the worldwide financial crisis of 2008.
This observation is not limited to the financial markets, however. My point is that for future-choice decisions many variables are inherently unknowable. We cannot characterize the uncertainty because, if we are honest, we cannot even specify the events that might or might not occur. Consider, for example, how few people at the beginning of the 1980s foresaw the widespread use of personal computers and the development of the Internet. If you had been given correct forecasts for these developments at the time, would you have believed them? My contention is that you probably would not have known how to evaluate the forecasts.
What developments will occur over the next twenty-five years? I believe that we are all quite blind with respect to the future—and, moreover, that if we simply extrapolate past trends, huge errors will ensue. The path of social and economic development follows an evolutionary trail and, as is well known, although evolution provides a good story for explaining the past, it makes no predictions.
FIGURE 3.2 Dow Jones Industrial Average from January 1, 1900, to December 31, 2007
Source: Makridakis, Hograrth, and Gaba (2009)
COMMIT ONLY AS FAR AS YOU CAN PREDICT
It would be wonderful to come up with a “good” or “amended”