Uneasy Money

After writing my previous post, I reread Robert Lucas’s classic article “Econometric Policy Evaluation: A Critique,” surely one of the most influential economics articles of the last half century. While the main point of the article was not entirely original, as Lucas himself acknowledged in the article, so powerful was his explanation of the point that it soon came to be known simply as the Lucas Critique. The Lucas Critique says that if a certain relationship between two economic variables has been estimated econometrically, policy makers, in formulating a policy for the future, cannot rely on that relationship to persist once a policy aiming to exploit the relationship is adopted. The motivation for the Lucas Critique was the Friedman-Phelps argument that a policy of inflation would fail to reduce the unemployment rate in the long run, because workers would eventually adjust their expectations of inflation, thereby draining inflation of any stimulative effect. By restating the Friedman-Phelps argument as the application of a more general principle, Lucas reinforced and solidified the natural-rate hypothesis, thereby establishing a key principle of modern macroeconomics.

In my previous post I argued that microeconomic relationships, e.g., demand curves and marginal rates of substitution, are, as a matter of pure theory, not independent of the state of the macroeconomy. In an interdependent economy all variables are mutually determined, so there is no warrant for saying that microrelationships are logically prior to, or even independent of, macrorelationships. If so, then the idea of microfoundations for macroeconomics is misleading, because all economic relationships are mutually interdependent; some relationships are not more basic or more fundamental than others. The kernel of truth in the idea of microfoundations is that there are certain basic principles or axioms of behavior that we don’t think an economic model should contradict, e.g., arbitrage opportunities should not be left unexploited – people should not pass up obvious opportunities, such as mutually beneficial offers of exchange, to increase their wealth or otherwise improve their state of well-being.

So I was curious to how see whether Lucas, while addressing the issue of how price expectations affected output and employment, recognized the possibility that a microeconomic relationship could be dependent on the state of the macroeconomy. For my purposes, the relevant passage occurs in section 5.3 (subtitled “Phillips Curves”) of the paper. After working out the basic theory earlier in the page, Lucas, in section 5, provided three examples of how econometric estimates of macroeconomic relationships would mislead policy makers if the effect of expectations on those relationships were not taken into account. The first two subsections treated consumption expenditures and the investment tax credit. The passage that I want to focus on consists of the first two paragraphs of subsection 5.3 (which I now quote verbatim except for minor changes in Lucas’s notation).

A third example is suggested by the recent controversy over the Phelps-Friedman hypothesis that permanent changes in the inflation rate will not alter the average rate of unemployment. Most of the major econometric models have been used in simulation experiments to test this proposition; the results are uniformly negative. Since expectations are involved in an essential way in labor and product market supply behavior, one would presumed, on the basis of the considerations raised in section 4, that these tests are beside the point. This presumption is correct, as the following example illustrates.

It will be helpful to utilize a simple, parametric model which captures the main features of the expectational view of aggregate supply – rational agents, cleared markets, incomplete information. We imagine suppliers of goods to be distributed over N distinct markets i, I = 1, . . ., N. To avoid index number problems, suppose that the same (except for location) good is traded in each market, and let y_it be the log of quantity supplied in market i in period t. Assume, further, that the supply y_it is composed of two factors

y_it = Py_it + Cy_it,

where Py_it denotes normal or permanent supply, and Cy_it cyclical or transitory supply (both again in logs). We take Py_it to be unresponsive to all but permanent relative price changes or, since the latter have been defined away by assuming a single good, simply unresponsive to price changes. Transitory supply Cy_it varies with perceived changes in the relative price of goods in i:

Cy_it = β(p_it – Ep_it),

where p_it is the log of the actual price in i at time t, and Ep_it is the log of the general (geometric average) price level in the economy as a whole, as perceived in market i.

Let’s take a moment to ponder the meaning of Lucas’s simplifying assumption that there is just one good. Relative prices (except for spatial differences in an otherwise identical good) are fixed by assumption; a disequilibrium (or suboptimal outcome) can arise only because of misperceptions of the aggregate price level. So, by explicit assumption, Lucas rules out the possibility that any microeconomic relationship depends on macroeconomic conditions. Note also that Lucas does not provide an account of the process by which market prices are established at each location, nothing being said about demand conditions. For example, if suppliers at location i perceive a price (transitorily) above the equilibrium price, and respond by (mistakenly) increasing output, thereby increasing their earnings, do those suppliers increase their demand to consume output? Suppose suppliers live and purchase at locations other than where they are supplying product, so that a supplier at location i purchases at location j, where i does not equal j. If a supplier at location i perceives an increase in price at location i, will his demand to purchase the good at location j increase as well? Will the increase in demand at location j cause an increase in the price at location j? What if there is a one-period lag between supplier receipts and their consumption demands? Lucas provides no insight into these possible ambiguities in his model.

Stated more generally, the problem with Lucas’s example is that it seems to be designed to exclude a priori the possibility of every type of disequilibrium but one, a disequilibrium corresponding to a single type of informational imperfection. Reasoning on the basis of that narrow premise, Lucas shows that, under a given expectation of the future price level, an econometrician would find a positive correlation between the price level and output — a negatively sloped Phillips Curve. Yet, under the same assumptions, Lucas also shows that an anticipated policy to raise the rate of inflation would fail to raise output (or, by implication, increase employment). But, given his very narrow underlying assumptions, it seems plausible to doubt the robustness of Lucas’s conclusion. Proving the validity of a proposition requires more than constructing an example in which the proposition is shown to be valid. That would be like trying to prove that the sides of every triangle are equal in length by constructing a triangle whose angles are all equal to 60 degrees, and then claiming that, because the sides of that triangle are equal in length, the sides of all triangles are equal in length.

Perhaps a better model than the one Lucas posited would have been one in which the amount supplied in each market was positively correlated with the amount supplied in every other market, inasmuch as an increase (decrease) in the amount supplied in one market will tend to increase (decrease) demand in other markets. In that case, I conjecture, deviations from permanent supply would tend to be cumulative (though not necessarily permanent), implying a more complex propagation mechanism than Lucas’s simple model does. Nor is it obvious to me how the equilibrium of such a model would compare to the equilibrium in the Lucas model. It does not seem inconceivable that a model could be constructed in which equilibrium output depended on the average price level. But this is just conjecture on my part, because I haven’t tried to write out and solve such a model. Perhaps an interested reader out there will try to work it out and report back to us on the results.

PS:  Congratulations to Scott Sumner on his excellent op-ed on nominal GDP level targeting in today’s Financial Times.