Uncertainty is not a new topic to economics, but you wouldn't know it from the state of the discipline. Frank Knight wrote the most comprehensive analysis of uncertainty back in 1921. Most certainly it was the basis for uncertainty in Keynes' General Theory fifteen years later. Yet economics not only has not adopted uncertainty, it believes the opposite: people have perfect foresight. Sadly, without perfect foresight, equilibrium and thus most of economic theory would have to be thrown out. This is why Knight, despite his comprehensive understanding of the subject, still accepts this assumption, and also why Keynes discussion of the subject (Chapter 12) was a "digression" (his words) rather than the core of the book.
Let's take a quick tour of the certainty/uncertainty spectrum according to Knight.
We start out in Certaintyland. In certainty land things happen with a probability of 100%. Choice in this land is reduced to cost-benefit analysis. Since all things are certain, we simply weigh costs versus benefits when making decisions and choose the path with the best payoff. Very few choices in real life are made in Certaintyland, but some general broad choices fall into this category.
A personal example is my choice to drop out of grad school in 1996. The math was hard and the theory was ridiculous, but I suppose I could have stuck it out and passed my comps if I cared to. However, the starting salary for economics professors at the time was $40,000; and I already had a Masters Degree. Since the best I could do was known with certainty, it was easy to do the cost benefit analysis and determine it would be better to look elsewhere for future work.
Moving out of Certaintyland, we enter the realm of Probabilistic Certainty. This is also known as "Cards and Dice" Land, because conditions are much like many games of cards and dice. In the realm of Probabilistic Certainty, we know the set of possible outcomes and we know the probability of all individual outcomes with certainty. A single die has six possible outcomes and each outcome has a probability of 1/6. Risk enters into the PC realm, but only in an orderly fashion. We know the distributions of risk and we can write off a string of rolling five consecutive fours as a one in 7776 event.
The realm of Probabilistic Uncertainty is unfortunately where many experimental economists go, thinking it is Uncertaintyland. They design experiments where probabilities are known and payoffs can be calculated. Though uncertainty involuntary enters their experiments by design, they generally write it off as some decision pathology of the test subjects.
For example, Kahneman and Tversky in their work on Prospect Theory discuss the Pseudocertainty Effect, but the games that are played are all defined probability outcomes where an expected value can be calculated. The experiments "prove" that humans have a thinking disorder which craves certainty and thus shuts down logical thought. However, the same outcome would be achieved if the test subjects simply didn't believe in the probabilities given to them by the experiment - preferring to trust their intuition that probability under uncertainty is undefined. We'll return to this concept when we get to Uncertaintyland.
Leaving the PC realm, we enter the tiny, yet quite important, municipality of Probabilistic Uncertainty. This is a land of horse races and sports books, where the set of possible outcomes is known, but the probability of an individual outcome is not. In a horse race, there may be ten horses running. One of them will certainly win, but the odds of each horse winning are determined by the betting, and not the true qualities of the participants. Risk distribution in probabilistic uncertainty can assume any form, yet will tend to resemble a normal distribution with fat tails. Betting on horses will tend to keep the odds close to the true odds, because many bettors weigh indicators of success in the betting process. However, enough freakish events will occur (the 2-5 favorite injures a leg out of the gate) to make the normal distribution seem implausible. In this municipality, choices and risk are as much an analysis of the way people make choices and risk as about the actual risk. For this reason, the stock market and investment markets are located in this realm. Often, success can be reached by analyzing other bettors/investors behavior and looking for a pattern of mistakes: undervaluation or overvaluation of certain characteristics that takes place, due to the municipality's border with Uncertaintyland.
Once we leave the Probabilistic Uncertainty municipality, we enter the wilds of Uncertaintyland. In Uncertaintyland, the set of possible events is unknown and possibly infinite. Probabilities are undefined, because the denominator of the probability function is not known. Things happen in Uncertaintyland that nobody can imagine would happen. In addition, even things that can be forecast to happen with some reasonable expectation lose their meaning. A "bad" event may occur and not be bad. A "good" event may occur and not be good. This is because in Uncertaintyland events have more than one dimension. An outcome may have a magnitude or a duration, or it may be contingent on other uncertain events to determine its payoff.
Think about the possibility of losing your job. What do you think that probability is right now? I, as a Federal Government employee, may determine that probability to be nearly zero, but it is not. I could accidentally do something that is a dismissable offense. There are also things like shutdowns and sequesters that might not have been conceivable even a few years ago. One might not lose their job, but face a pay cut (dimension) or even get promoted to a higher paying position, but under a terrible manager that results in quitting the job soon after (good event being bad). The dimension effects of uncertainty mean that even logical, rational choices may not result in success, and the true meaning of the "odds" of any event reduce to 1 or 0. Things will either happen or not, and until the outcome occurs, the payoffs cannot be calculated.
That's enough for me today. I'll discuss the ramifications of uncertainty on choices in a following post.
No comments:
Post a Comment