Rational Expectations

The Idea

How do people form expectations about the future? Before the 1960s, most macroeconomic models assumed some form of adaptive expectations: people look at what happened recently and extrapolate. If inflation was 3 percent last year, they expect it to be roughly 3 percent next year, adjusting slowly as new data arrives. This approach is simple and intuitive, but it has a troubling implication: people are systematically wrong. If inflation is rising, adaptive expectations always lag behind, and policy-makers can exploit this lag to stimulate the economy by generating surprise inflation.

John Muth proposed an alternative in a 1961 paper on agricultural markets. He argued that expectations, since they are informed predictions of future events, should be “essentially the same as the predictions of the relevant economic theory.” People do not have to be omniscient or infallible. They simply use all available information, including knowledge of how the economy works and how policy-makers behave, to form the best possible forecast. Their predictions will be wrong sometimes, but the errors will be random and unpredictable, not systematic.

Muth’s idea sat relatively quietly for a decade before Robert Lucas brought it to macroeconomics with explosive force.

Lucas and the New Classical Revolution

In a series of papers in the early 1970s, Lucas applied rational expectations to the core questions of macroeconomics: inflation, unemployment, and the effectiveness of government policy. The results upended the Keynesian consensus.

Consider the Phillips curve, the empirical relationship between inflation and unemployment that had guided policy since the 1960s. Keynesian policy-makers believed they could permanently reduce unemployment by accepting a bit more inflation. Lucas argued that this trade-off is illusory once expectations adjust. If the government repeatedly uses inflationary policy to boost employment, people will come to expect inflation and incorporate it into their wage demands and pricing decisions. The stimulus evaporates, and the economy ends up with higher inflation but no reduction in unemployment, exactly what happened during the stagflation of the 1970s.

Rational expectations did not merely predict that adaptive expectations were too slow. It predicted that any systematic, predictable policy would be fully anticipated by the public and therefore neutralized. Only unanticipated policy, true surprises, could have real effects. This is the policy ineffectiveness proposition, associated with Thomas Sargent and Neil Wallace: if the central bank follows a known rule (raise the money supply by 5 percent per year), rational agents will price it in completely, and the rule will affect only the price level, not output or employment.

The Lucas Critique

Perhaps Lucas’s most enduring contribution is methodological rather than substantive. In a 1976 paper, he argued that the econometric models used to evaluate policy are fundamentally unreliable because the parameters of those models are not structural constants. They are functions of policy itself. When policy changes, behavior changes, and the estimated relationships break down.

For example, suppose a statistical model estimated during a period of stable monetary policy finds a robust correlation between money growth and output growth. A policy-maker uses this model to predict the effect of doubling the money supply. But the act of doubling the money supply changes the environment in which agents operate. Their expectations shift, their behavior adapts, and the historical correlation no longer holds.

The Lucas critique demands that economic models be built on “deep parameters,” preferences and technologies that remain stable across policy regimes. This insistence reshaped macroeconomic modeling, leading to the dominance of dynamic stochastic general equilibrium (DSGE) models in which agents optimize intertemporally and expectations are formed rationally.

Why RE Disciplines Models

Whatever one thinks of rational expectations as a description of reality, it serves as a powerful modeling discipline. It forces economists to think about how agents will respond to policy changes, rather than treating behavior as mechanical and fixed. It eliminates “free lunches” in which policy-makers can permanently exploit a static relationship. And it ensures that models are internally consistent: the expectations embedded in agents’ decisions must be compatible with the outcomes the model generates.

In this sense, rational expectations is less an empirical claim about how people think and more a methodological standard for how models should be built. A model that assumes agents are systematically fooled by predictable policies is, by construction, a model in which agents fail to use information that is freely available. Such a model may be useful for some purposes, but it has an uncomfortable internal contradiction.

Empirical Challenges

The rational expectations hypothesis faces serious empirical difficulties. Survey data on inflation expectations consistently show that people’s forecasts are biased, sluggish, and heterogeneous in ways that rational expectations does not predict. Professional forecasters do better, but even they exhibit systematic patterns of over- and under-reaction.

Financial markets, often cited as the best candidates for rational expectations (since professional traders have strong incentives to use information efficiently), are plagued by anomalies: momentum effects, excess volatility, bubbles, and crashes that are difficult to reconcile with the hypothesis. The efficient market hypothesis, a close cousin of rational expectations in finance, has been under sustained empirical attack since the 1980s.

Behavioral economics has documented a long list of cognitive biases, anchoring, overconfidence, recency bias, narrative thinking, that cause systematic departures from the rational expectations ideal. These are not random errors that wash out in the aggregate; they are predictable patterns that persist across individuals and contexts.

Diagnostic Expectations and Other Modifications

In response to these challenges, economists have developed several modifications. One of the most promising is diagnostic expectations, proposed by Pedro Bordalo, Nicola Gennaioli, and Andrei Shleifer. Under diagnostic expectations, agents overweight information that is representative of the current state of the world. If the economy is booming, people exaggerate the probability that the boom will continue. If it is contracting, they exaggerate the probability of further decline.

This approach preserves the discipline of forward-looking expectations while introducing the kind of systematic bias that survey data and market behavior suggest is real. It can generate boom-bust cycles, excess volatility, and other phenomena that rational expectations struggles with.

Other modifications include learning models (agents start with imperfect knowledge and update over time), sticky information models (not everyone updates their information set at the same time), and noisy information models (agents receive signals contaminated by noise). Each relaxes the rational expectations assumption in a different direction while retaining some of its analytical structure.

When RE Is Useful vs. Misleading

Rational expectations is most useful in settings where agents have strong incentives to get things right, where information is readily available, and where the environment is relatively stable. Professional traders pricing well-understood financial instruments come close to the ideal. Central bankers communicating policy intentions in a transparent framework can reasonably assume that markets will process the information efficiently.

It is most misleading in settings characterized by deep uncertainty, novelty, and complexity: financial crises, technological revolutions, pandemics. In these environments, agents lack a reliable model of the world to form expectations against. They resort to heuristics, narratives, and social imitation, precisely the behaviors that rational expectations assumes away.

The legacy of rational expectations is therefore double-edged. It raised the bar for macroeconomic modeling and destroyed the naive belief that policy-makers could permanently exploit a fixed Phillips curve trade-off. But it also encouraged a generation of economists to build models in which crises, bubbles, and systematic errors are impossible by assumption, leaving them poorly prepared for the events that mattered most.