27 February 2007

The Rational Expectations Hypothesis (REH)

The Rational Expectations Hypothesis (REH) emerged in the early 1960's as a set of theories about the financial markets. At that time, the prevailing economic theory was Keynesianism, which had evolved massively since 1936 without much supervision from its namesake. The crucial departure of Keynesianism can be summarized as "imperfect markets," which meant,
  • prices may "never" adjust to reflect the intersection of supply & demand;
  • money is not neutral (i.e., disinflation or deflation has a serious impact on economic conditions; the available money supply, accompanied by a distinct market for cash balances, has an impact on real output);
  • factor markets, such as for labor, land, and capital, may sometimes not clear;
  • state intervention may sometimes cause more good than harm.
The last proposition is a bit difficult for some people to understand. Keynes had sought to establish a macroeconomic role for the state precisely so it could avoid a microeconomic (i.e., socialistic) one, and he attempted to establish clear, immutable boundaries for the state under conditions of economic emergency.

In the period 1946-1980, economists influenced by Keynes' system (and more decisively, by his apostles) devised a framework of economic research and policy methods that bypassed any assumption of an equilibrium. There was understood to exist a microeconomic realm, which included transactions among many buyers and sellers—and which was subject to familiar Neoclassical laws of economics—and there was the macroeconomic realm of inflation, interest rates, and factor employment, where a different set of laws applied. The chief divergence in this realm was that the market never completed its process of adjustment; it could easily stop "adjusting" to full-employment equilibrium without ever getting close.

The key objection to this was always in the financial markets, i.e., the markets for bonds, commercial paper, foreign exchange, or money equivalents. Specifically, Keynesianism models developed for government agencies tended to rely on assumptions of how the money markets would respond to government actions. Everyone reading this is no doubt accustomed to news reports of the Fed changing interest rates; but it's long been noted that the Fed has limited effect in that regard. Basically, if the administration is running a "structural deficit" and the national debt is expected to remain on a permanent upward trajectory, then a reduction in the federal funds rate (the shortest-term and lowest interest rate of all) will very likely lead to an increase in longer-term rates. This was interpreted as evidence that investors make purposeful, and therefore, rational, decisions.

From there, the REH evolved to encompass a greater scope of economic behavior. After all, money markets merely respond to professionally anticipated demand for money or credit. The real sector of the economy, the part that actually does stuff, also makes plans based on inflationary expectations. At this point, the departure from pre-1979 macroeconomics becomes a little more vivid. Whereas, Keynesian economics uses simple charts to map the entire economy (IS-LM; AS-AD), REH models abandoned such visually appealing charts. Instead of graphs linking liquidity preference to interest rates and real output (the LM curve), post-Keynesian analysis relies on modeling the utility-maximizing behavior of a representative individual through time, given certain limiting assumptions: savings occurs at the expense of consumption; labor occurs at the expense of leisure; taxation influences economic activity, and most importantly, strategies to save and work are guided by expectations of the future.

The behavior of an agent in response to economic conditions is represented as a mathematical equation, nearly always the Ramsey-Cass-Koopmans Model. The RCK model is pretty flexible, and any student can edit its objective or constraint functions to reflect new hypotheses. The purpose of doing so is to create a precise map of what happens when something happens to the prevailing conditions: a change in interest rates, a scheme of fiscal stimulus (deficit spending), or central bank interventions. The RCK model is the backbone of dynamic general equilibrium (DGE) schools of economic thought, which incorporate varying degrees of REH.

However, it is not correct to claim that all this was alien to Keynesian economics. The RCK model clearly evolved in the employ of David Cass (1965) and Tjalling Koopmans (1965), both of whom were addressing questions of long-term economic growth. Neither were concerned with fiscal or monetary policy; however, it is true that by assuming economic growth was, in effect, a function of capital accumulation, both were certainly reverting to a world in which aggregate demand was not an issue. Keynes himself devoted a chapter of the General Theory to the role of expectation in business cycles. But he did not consider expectations to be so sensitive as to respond reliably to plausible state policies, and so his exposition deals exclusively with the role of the business cycle. Not surprisingly, whereas REH tended to think expectations were something business planners used to combat the state (the "policy ineffectiveness propositions"), Keynes saw expectations as undermining the neoclassical assumptions of a self-correcting economy. Hence, Keynes' use of [adaptive] expectations tended to support the idea of fiscal and monetary policy, in contrast to the REH's school's position that [rational] expectations make both pointless.

Because REH assumptions could be adjusted within the DGE schema, economists have been free to tweak them to fit the available data. However, there have been grave shortcomings in the REH results. First, the REH predictions are especially resistant to falsification. The problem is that the algorithm for selecting coefficients on any RCK equation are determined from the historical data one is trying to fit. In my opinion, this is mere data mining. It's an adaptation of the statistical procedure called regression analysis, except that the procedure is also supposed to identify "technology shocks," or disruptive events that cause spikes in the steady growth of the economy.

Let's consider these technology shocks: a contentious matter in the early days of REH, skeptics asked for examples of them. REH proponents appealed to the principle of positive economics, arguing that the point was that their model explained what actually occurred, so technology shocks were merely stylized facts. But surely there would have been a body of literature on these all-important nodes in history, particularly as REH models tended toward a standard set of assumptions. Since no such equilibrium has occurred, each version of REH produces a different set of technology shocks. One could reasonably have expected the luminaries of the field to furnish a list of the greatest shocks, and some discussion of how they affected existing technology. Instead, the official story remains the same: it's a stylized fact, don't look at it.

Stylized facts are an understandable shorthand, but when they are constantly edited so the theory fits the observations, they don't converge to a stable version of reality. Instead, they lurch about every time the data set is increased, as with the passage of time. For this reason, I would have to say that I don't expect REH to perform better in a Bretton Woods-style crisis than Keynesianism did.

Structural Deficit: given a fixed tax code, fixed entitlement policies, and a roughly stable rate of direct government purchases of goods & services (G), a recession will cause tax revenues to fall and government expenditures to rise. This creates a deficit (or reduces the surplus, if there was one). Conversely, if the economy is in recovery, tax revenues rise and expenditures fall, causing the budget deficit to shrink or even move to a surplus.

An incoming administration has no control over the business cycle, and has little (if any) control over the duration of its tenure. Rhetoric aside, changes to the tax code or favorable trade agreements have effects over only long periods of time. So an administration that comes to office at the beginning of a recession may defend its deficit spending by pointing out that the state is structurally in balance; i.e., the inevitable recovery will bring with it a return to fiscal balance. The point is not trivial; if the deficit is structural, then money markets will have to assume that the state's growing debts will lead both to a higher demand for credit, and higher tax rates further off in the future.

Those interested in the so-called "Ricardian Equivalence Hypothesis" of Robert Barro, please note that one study of the historical evidence established a split between the structural and non-structural elements of deficits: non-structural deficits don't stimulate increased saving.

Stylized facts: a presentation of an empirical finding, usually in statistical format. An example is the result of a regression analysis, which might correlate many variables with GDP growth. One of the variables could be a dummy variable for "Protestant majority," which might have a positive coefficient; the stylized fact here is, "there's some favorable impact on economic growth from having a Protestant majority." But it's not a universally valid generalization; it depends on what explanatory variables one used, and what the data set was. Unless the data set is relevant to the conclusions one hopes to draw, it's not useful information.

When a regression analysis is done well, it will produce a set of stylized facts that, while individually inconsequential, can be used as a group to produce valid inferences. One of the arguments used by REH theorists is that, sure, their assumptions may be individually absurd, but as a rich system of equations, they are powerful analytical tools that produce useful results. Hence, a point-by-point rebuttal of REH assumptions is an exercise in futility.

BOOKS: Roger Guesnerie, Assessing Rational Expectations, MIT Press (2001); Stephen M. Sheffrin, Rational Expectations, 2nd Edition, Cambridge (1996); Preston Miller (editor), The Rational Expectations Revolution, MIT Press (1996);

Labels: , , ,


Post a Comment

<< Home