• Choose a category
• All
• Classic-Fiction

Shleifer, Andrei, and Robert W. Vishny (1997). “The Limits of Arbitrage.” Journal of Finance 52, no. 1: 33-55.

3

Subways, Coconuts, and Foggy Minefields

An Approach to Studying Future-Choice Decisions

ROBIN M. HOGARTH

“Future-choice” decisions are decisions that have three important sources of complexity.

First, actions taken today can have unknown consequences at future horizons that are difficult to specify.

Second, decisions imply complex inter-temporal tradeoffs.

And, third, it is problematic to specify relevant states of the world, let alone to assess their probabilities.

Consider, for example, making decisions today that will affect the layout of a city. How far into the future should such decisions look? What volumes of traffic are likely to develop on alternative highways? What unknown technological advances could change costs and benefits? Will public preferences remain stable? What are the possibilities of different natural or social disasters? And so on. The list of possibilities and complications is endless. Imagine further a 23-year-old who is planning a career and ways of acquiring capital across her life. How can she evaluate different career paths? How can she know today what expertise will be in demand in ten or twenty years? How can she predict changes in her personal situation as well as her health?

Today, the standard economic model for dealing with these situations is the discounted utility model. This has the advantage of reducing the net benefits of each alternative stream of actions to a single number so that “rational” comparisons can be made. However, achieving these single numbers implies a series of heroic assumptions that exceed most mortals’ capabilities.

In thinking about these issues, I first want to salute the considerable advances made in our understanding due to decision theory. In so doing, however, we need to keep clearly in mind just what decision analysis can and cannot take into account.

Second, I will go on to distinguish two types of uncertainty that characterize decisions labeled “subway” and “coconut” uncertainty, respectively.

Third, the presence of coconut uncertainty essentially implies the breakdown of decision-theoretic and forecasting models and demands a new approach for future-choice decisions. Whereas I cannot offer a new approach yet—or an intellectual breakthrough—I can suggest a heuristic principle and metaphor that may help us deal with some of the issues. Perhaps by considering their advantages and disadvantages, we might illuminate some of the problems of future-choice decisions and, at the least, set an agenda for future research.

THE FAILURE OF DECISION THEORY

Like many graduate students of my generation, I was totally seduced by statistical decision theory. But statistical decision theory never really became the universal tool that many imagined it would.

Nor do I believe that decision theory has been handicapped by the fact that many of its axioms are violated by people’s choices. Indeed, it is precisely “violations” of this sort that illustrate why a good theory is necessary.

My belief is that statistical decision theory fails in many important problems—particularly future-choice problems—because humans are incapable of characterizing the uncertainties of the world in which they operate. Decision theory applies only to well-defined “small worlds” (Savage, 1954).

In addition, although there can be no doubt that the ability to predict many phenomena has been an important element of human development, when it comes to predicting events in the socioeconomic domain the human predictive track record is, frankly, dismal.

TWO TYPES OF UNCERTAINTY

I think that what really hampers the use of formal analytical techniques for future-choice decisions is that uncertainty can take two forms. In a new book (Makridakis, Hogarth, and Gaba, 2009), my co-authors and I refer to these as “subway uncertainty” and “coconut uncertainty,” respectively.

FIGURE 3.1 Klaus’s Travel Time to Work (in minutes) Source: Makridakis,