Imperfect Knowledge Economics. Exchange Rates and Risk by Roman Frydman and Michael D. Goldberg is one of the most interesting and topical books in economics to appear in recent years. Before you order a copy, be warned that from chapter 3 onwards it quickly gets technical. It is basically a graduate textbook.

Frydman and Goldberg argue that the Rational Expectations Hypothesis that underlies many contemporary economic models is seriously flawed. There is, of course, nothing new about criticism of the Rational Expectations Hypothesis, and in the first two chapters Frydman and Goldberg call upon a neglected tradition in economics that goes back to Hayek and Keynes. But many economists stick to it for lack of a better alternative and because assuming economic agents are rational is convenient from a modeling point of view.

In a world of imperfect knowledge macroeconomic models that are based on models of individual behavior may lead to internal inconsistencies. Aggregate outcomes are based on individual decisions, but individual decisions depend on forecasts of aggregate outcomes. The Rational Expectations Hypothesis avoids any inconsistencies that may result from this feedback loop by assuming that economic agents have perfect foresight and, in the long run and on average, do not make systematic errors when predicting the future.

Unfortunately, models based on Rational Expectations cannot account for the data they are designed to explain. This is perhaps most evident in financial markets. Currency markets and stock markets for instance, are far more volatile than the theory would predict and exhibit features that are at odds with the Rational Expectations and the Efficient Market Hypothesis. Frydman and Goldberg name the long-term deviations from parity in foreign exchange markets, which in the Rational Expectations framework can only be explained by assuming systematic misperceptions on the part of market participants.

In recent years, faced with the failure to explain the empirical data with conventional economic models, there has been a turn in economics to psychology and neuroscience as a possible foundation of individual economic behavior. But as Frydman and Goldberg argue, insofar as these models are fully determined, they cannot form the microfoundation of an aggregate, macroeconomic model, since, by design, they entail the above mentioned inconsistency between the individual and aggregate level.

What’s more, it may be true that people value gains and losses differently, that they are biased towards recent events, mimic the actions of a larger group, don’t understand probability theory, are overconfident, overreact and often base their decisions on known but irrelevant facts. But even if we were to take all of this into account in a psychologically realistic model, the model may still produce inaccurate forecasts, because individuals may change their behavior in non-predictable, but rational ways in response to novel situations that arise over the course of time.

Frydman and Goldberg observe that in behavioural economics deviations from the correct answer or the optimal response are regarded as forms of irrational bias. But what may appear as irrational with hindsight may be rational in the face of uncertainty and imperfect knowledge. As Tim Harford argues extensively in his forthcoming book The Logic of Life, economics actually shows that people are more rational than they appear to be. This is also the point made by Frydman and Goldberg. The Rational Expectations Hypothesis is wrong for a precise reason: we live in a world of imperfect knowledge.

If we look at the labour market, a sudden change in U.S. immigration law may deteriorate the job prospects of Indian engineering students. The easing of visa procedures for Chinese wishing to work abroad may increase the number of highly qualified job seekers in the U.K. A change in products and attitudes may mean that financial firms suddenly start hiring mathematics and physics PhD’s instead of MBA’s.

One of the essential points made by Frydman and Goldberg is that we cannot assign probabilities to the above scenarios, since they represent a fundamental change in the socio-economic landscape.

As Frydman and Goldberg point out, many economic models incorporate change as exogenous shocks, of which a change in immigration law may be an example. But endogenous change in the form of innovation is at the heart of much economic activity. A theory which treats all forms of change as exogenous is therefore unrealistic.

The alternative representation proposed by Frydman and Goldberg, and I am oversimplifying enormously here, is to model (consumer) demand not with one but a set of two or more equations for different price levels. If a price rises above or falls below a certain level consumers may switch to a different forecasting regime. Investors may for example change their overall outlook from bearish to bullish above a certain level and vice versa. This change need not be symmetric.

As a result of this formulation the models proposed by Frydman and Goldberg don’t yield sharp predictions, but only qualitative, directional predictions. Still, their model of the FX market fares better than Rational Expectations models.

Imperfect Knowledge Economics is an interesting albeit challenging read. Its perspective on economic theory is fresh and novel, although it remains to be seen whether the approach can be fruitfully extended to realms of economics other than the modelling of exchange rates. I must admit that I’m now stuck somewhere in the middle, because it is pretty hard-going and I have only browsed through the remaining chapters.

The first chapter of Imperfect Knowledge Economics. Exchange Rates and Risk.

Macroeconomics for a Modern Economy. The Nobel prize lecture (2006) by Edmund Phelps.