Why the Rational Expectations Hypothesis Does Not Characterize Rational Expectations in Real-World Markets
Characterizing rational expectations in real-world markets requires acknowledging that we cannot have complete probabilistic knowledge of the future
In Economics, the Rational Expectations Hypothesis (REH) is widely considered the way to represent the expectations of rational individuals in real-world markets theoretically.
A vast empirical literature, however, has empirically rejected REH’s key predictions.
Relying on survey forecasts of macroeconomic variables, such as those from the Survey of Professional Forecasters, individuals’ forecast errors are found to be systematic. In particular, their forecast errors are biased (not zero, on average) and correlated with information available when the forecasts were made. This has been confirmed for different macroeconomic variables, countries, and time periods.
These findings reject REH’s core prediction that if individuals are rational, their forecast errors should be unpredictable: they should be unbiased (zero, on average) and uncorrelated with the information available when the forecasts were made.
The empirical rejection of REH has led economists to question whether individuals in real-world markets form their expectations rationally.
Providing different answers to this question, several theoretical approaches have been proposed in the literature to account for REH’s empirical rejection.
One approach, limited-information REH, maintains the assumption that individuals are rational, but argues that the predictability of their forecast errors arises because they only have access to, and base their expectations upon, limited or noisy information instead of full information.
Another approach, relying on behavioral economics, argues that the predictability arises because psychological factors, such as Kahneman and Tversky’s representativeness heuristic, distort individuals’ expectations relative to the rational expectation. Thus, the influence of psychological factors is interpreted as individuals being irrational.
In our recent papers, Roman Frydman and I offer a different explanation:
REH does not represent the expectations of rational individuals in real-world markets because it rests on the premise that they can have complete probabilistic knowledge of the future. This premise—not the assumptions that individuals are rational or don’t have access to full information—is the root cause of REH’s empirical rejection.
Model-Consistent Expectations are the Only Way to Acknowledge that Individuals are Rational
Most economic decisions are characterized by uncertainty because the consequences of individuals’ decisions occur in the future. Thus, modeling economic decision-making requires a representation of their expectations of the future outcomes.
Most of today’s economic models rely on John Muth’s (1961) Rational Expectations Hypothesis (REH) to represent these expectations.1
Muth argued that economists should acknowledge that individuals are rational by representing their expectations as consistent with the predictions of their model. REH does this by representing their expectations with a model’s conditional expectation of future outcomes. This implies that their expectations only differ from the model’s actual outcomes by a random error term, so that their forecast errors should be unpredictable.
Muth warned that REH’s representation of expectations should be viewed as a bold abstraction, not as a literal representation of how individuals actually form these expectations.
His key argument, however, was that if individuals are rational, their expectations should correspond to how the economy works. An economist’s model formalizes a hypothesis about how the economy works by specifying a stochastic process that characterizes how outcomes unfold over time. If the model’s hypothesis about how the economy works is correct and individuals are rational, their expectations should be consistent with this process.
Indeed, a representation of expectations that deviates from this process would be inconsistent with the model’s hypothesis about how the economy works. Thus, it would be inconsistent with the assumption that individuals are rational.
To summarize, the only way an economist can acknowledge that individuals are rational is by representing their expectations as model-consistent.
Model-Consistency is Not Sufficient to Characterize Real-World Rational Expectations
Although representing expectations as model-consistent is the only way an economist can acknowledge that individuals are rational, it is not sufficient to characterize the expectations of rational individuals in the real world.
This distinction is crucial.
The reason is that model-consistent expectations are necessarily contingent on the model’s assumptions about how the economy works. Thus, model-consistent expectations are only rational given the model’s assumptions.
If these assumptions are wrong—meaning that the economist’s model abstracts from elements that are important for real-world outcomes—then the model does not represent a correct hypothesis about how the economy works. Consequently, its model-consistent expectations will not correspond to the expectations of rational individuals in real-world markets.
In other words, a model-consistent representation of expectations only characterizes the expectations of rational individuals in real-world markets if the model’s assumptions about how the economy unfolds over time are valid.
This brings me to the root of Roman Frydman’s and my argument: REH does not characterize the expectations of rational individuals in real-world markets because it rests on models that assume the future is a probabilistic replica of the past.
Such a model of how the economy works assumes that individuals can have complete probabilistic knowledge of the future.
Characterizing Rational Individuals’ Expectations Requires Acknowledging Knightian Uncertainty
In real-world markets, complete probabilistic knowledge of the future is inherently impossible—even when individuals are completely rational and have access to full information.
REH models abstract from this inherent imperfection of our knowledge of the future.
In contrast, Roman Frydman and I argue that to characterize how the economy works, we must acknowledge that its structure changes in nonrepetitive ways that cannot be perfectly foreseen, even in probabilistic terms.
As I wrote about last week, such change—which we call unforeseeable change—gives rise to Knightian uncertainty: the uncertainty about future economic outcomes cannot be reduced to a probability distribution a priori.
By representing individuals’ expectations as model-consistent in a model open to unforeseeable change and the Knightian uncertainty arising from such change, we can acknowledge that they cannot have complete probabilistic knowledge of the future. Instead, their knowledge of the future is inherently imperfect.
This implies that even rational individuals, who base their expectations on full information, form expectations that deviate from actual subsequent outcomes. As a result, their forecast errors can be systematic, but the bias of these forecast errors and their correlation with ex ante information will change over time. This is consistent with the empirical findings relying on survey-based forecasts of macroeconomic variables.
The hypothesis we explore is that opening economic models to unforeseeable change and Knightian uncertainty enables us to provide a model-consistent theoretical representation of the expectations of rational individuals in real-world markets.
In future posts, I will explain our theoretical approach to doing, how it changes our understanding of rational expectations, and how it enables us to account for the key features of survey-based forecasts.
This post is based on Roman Frydman’s and my recent working paper, titled “Rational Expectations of Inflation Undergoing Unforeseeable Change”.
In the paper, you can read a more detailed version of the argument in this post. Moreover, you can read the full details of how we propose opening economic models to unforeseeable change and representing rational expectations in the presence of the Knightian uncertainty arising from such change.
Muth, John (1961), “The Rational Expectations Hypothesis and the Theory of Price Movements,” Econometrica, 29, 3.