Though the Nobel Committee has yet again inexcusably overlooked the matchless contributions to economics of Armen Alchian, its selection of Thomas Sargent and Christopher Sims to receive this year’s Nobel Prize in economics was, by any objective standard, an outstanding choice. No one can dispute that Sargent and Sims are truly deserving of the honor bestowed on them.
In explaining the selection, the Nobel Committee focused on the contributions of Sargent and Sims in developing new econometric techniques by which to analyze macroeconomic time series data, techniques now essential to empirical macroeconomics. The motivation for developing these techniques in Sargent’s case was to test empirically assumptions about expectation formation. Using these techniques, Sargent was able to provide empirical support for a vertical long-run Phillips curve, a key implication of the rational expectations hypothesis, perhaps the most important empirical result in macroeconomics of the last several decades. The Nobel Committee cited Sargent for his contributions to the empirical testing of rational expectations rather than for the development of the rational expectations hypothesis itself, presumably because Robert Lucas had already received the Nobel Prize for developing the rational expectations hypothesis. But Sargent could easily (and justly) have been chosen to receive the prize in 1995 along with Lucas. It gets complicated.
At any rate, because Sargent is a key figure in the development of rational expectations, his selection as winner of the Nobel Prize provides an occasion for some reflections on rational expectations and the place of rational expectations in economic theory.
Rational expectations emerged as an empirical hypothesis in the course of the debates in the 1960s about the Phillips Curve and whether a stable trade-off exists between inflation and unemployment that, as Samuelson and Solow suggested in a famous paper, could be viewed as a menu by policy makers. Friedman and Phelps independently refuted that interpretation of Phillips’s empirical result, a less original refutation, by the way, than is generally supposed, Mises, Hayek, Haberler, Alchian and Kessel, Irving Fisher and David Hume, among others, having already long since showed that inflation would have no stimulative effect on output and employment once it became expected. The original step taken by Lucas, building on John Muth’s seminal paper formalizing the concept of rational expectations, was to argue that even a policy of accelerating inflation designed to stay a step ahead of the public’s expectations of inflation could not work, because the public would catch on to the implicit policy rule, thereby frustrating its implementation.
This was an important advance both at the conceptual and practical levels because it helped clarify how to think about the role of expectations in economic models and because it exposed clearly constraints on economic policy-making not previously recognized (though as mentioned above many economists had for a long time been generally aware of the issue and had argued that policy makers had to take it into account). But from an empirical (i.e., testable) hypothesis about expectations formation, rational expectations (along with its cousin the efficient markets hypothesis) fairly quickly evolved into an axiomatic (and hence irrefutable) principle, more or less on the same level in economic theory as the rationality (wealth- and utility-maximization) postulate. As a result, the substantive (i.e. empirical) content of macroeconomic theories was increasingly dictated by the adoption of a methodological principle, having only limited and ambiguous empirical support.
New Keynesian theorists have tried to strike a balance between the Lucasian rational expectations paradigm and a desire to rationalize a role for counter-cyclical stabilization policy by adopting the dynamic stochastic general equilibrium (DSGE) paradigm while positing a variety of informational imperfections causing departures from the optimality results associated with DSGE models combined with rational expectations. This was in fact the basis of Lucas’s own early approach positing the inability of agents to distinguish between relative and absolute price changes. All in all, I don’t think that this has been a good bargain for the New Keynesians or for the profession at large.
The problem, it seems to me, is that elevating rational expectations into a methodological principle converts a property that one would expect to obtain only in a full general equilibrium into a necessary property of any acceptable economic model. It is legitimate to test the coherence of a model by asking whether the model’s results are consistent with the assumption of rational expectations. We want models that have that fixed point property. But that is not the same as saying that every model of the real world must exhibit rational expectations under all circumstances. To some extent, Sargent’s own work, and that of his students, on learning is a recognition that it does not make sense to impose rational expectations as a universal property of economic modeling.
I am pressed for time and there is much more to be said on the subject. Perhaps I will have a further post on the subject next week or the week after. This will be my last post for the week, and I will not be replying to comments either, so don’t be alarmed or upset at my silence for the next several days. I hope to return to the blogosphere on Sunday.