Future uncertainty and present indeterminacy

"Arguing in My Spare Time" 2.02
by Arnold Kling

April 7, 1999

May not be redistributed commercially without the author's permission.

When I was studying economics, a fad called "rational expectations" was sweeping through the profession. It was an interesting concept, worthy of study, but I am afraid that it threw us off track.

The fad had two things going for it. One was that it involved some mathematical complexity. Like the Unix operating system, this gave the initiated a sense of superiority. Another point in its favor was consistency with the traditional economic paradigm.

The concept of rational expectations applies to the way in which individuals make forecasts about future economic variables. For example, any firm that plans a long-term capital investment must at least implicitly forecast market conditions that will affect the future returns on that investment.

Economists believe that irrational or inefficient behavior will be weeded out by the economic system. The argument for rational expectations runs like this: if people make systematic forecasting errors, they will start to adjust their forecasting processes. As the processes are corrected, forecasts will improve and mistakes will be reduced to small, random errors.

This theory is very plausible for repetitive short-term forecasting. If people are making forecasts each month for what is going to take place next month, then it is plausible that they will get better at it over time.

However, the argument is less compelling for long-term expectations. If I make an error in my forecast for my income ten years from now, it could take me many years to discover the mistake, and it may be impossible for me to figure out which of many possible corrections are needed to fix my forecasting process.

The theory of rational expectations came to include a doctrine that long-term expectations are uniquely determined. That is, only one long-term outcome can be forecasted rationally. The virtue of this doctrine is that it matches the number of equations to the number of unknowns, resulting in determinate solutions within economic models. But in fact it has no compelling basis as a description of the world.

Consider instead an alternative in which many long-term outcomes are plausible. Perhaps corporate profits will be 40 percent higher ten years from now than they are today, after adjusting for inflation. But perhaps profits will be only 15 percent higher. If most people believe the higher figure, they will behave differently than if they believe the lower figure.

If people believe that corporate profits are going to grow rapidly, investment will be high; conversely, if they are pessimistic about profits, investment will be low. Either way, it may be a long time before enough information accumulates to cause people to change their expectations.

If we take it as given that the long-term future is so uncertain that a wide range of forecasts is plausible, then we introduce an element of indeterminacy to economic outcomes. Economic performance will depend on the degree of optimism and pessimism that happens to prevail at any one time about inflation, profits, and other variables.

According to Skidelsky, among others, this was part of Keynes' view of the economy. Keynes focused on the possibility of people becoming pessimistic about the long-term future. In this case, he argued, they would attempt to save for the future in the form of liquid financial instruments rather than in the form of risky capital investments. This would reduce the demand for output and cause an economic slump.

For the past five years, a Keynesian slump has been taking place in Japan. Meanwhile, in the U.S., optimistic expectations are the rule. While the emerging markets of Asia and Latin America have been abandoned by many investors, the emerging market of the Internet is very popular. Such is the state of long-term expectations today.

In short, the real world is characterized by:

1. Enormous uncertainty about the long-term future

2. Very slow rates of information gain about the long-term future

3. Wide variations in short-term outcomes depending on the state of expectations for the long-term future

Thus, reality is nearly the opposite of the neat, tidy scheme of rational expectations. The rational expectations approach is to start with a given long-term outcome and work backward to derive the unique state of expectations that should hold today in order to reach that outcome. Instead, one must acknowledge that a variety of states of expectations are possible today, and different states of expectations will lead to different paths and scenarios going forward.