"Accepting that there is suhstantial real rigidity does not require adopting the view that many types of shocks are important to fluctuations. In the daylight-saving-time example, for instance, although there appears to be considerable real rigidity in individuals preferences about their schedules, we do not observe sharp short-run variations in economy-wide real schedules arising from sources other than changes in the time standard. Finally, an intermediate possibilit) is that when there are large real rigidities, many kinds of shocks, both real and nominal, are important to fluctuations (see, for example, Greenwald and Stiglitz, 1988).

Equation (6.99) shows that when the reaction function slopes up, there is a "multiplier" that magnifies the effect of the shift of the reaction function at a given level of y, dyf/Sx. In terms of the diagram, the impact on the equilibrium level of is much larger than the upward shift of the reaction function. The closer the slope is to 1, the larger the multiplier is.

In a situation like this, any factor that affects the reaction fimction has a large impact on overall economic activity. In the terminology of Summers (1988), the equilibrium is fragile. Thus it is possible that there is substantial real rigidity but that fluctuations are driven by real rather than nominal shocks. When there is substantial real rigidity, technology shocks, credit-market disruptions, changes in government spending and tax rates, shifts in uncertainty about future pohcies, and other real disturbances can aU be important sources of output movements. Since, as we have seen, there is unlikely to be substantial real rigidity in a Walrasian model, we refer to theories of fluctuations based on real rigidities and real disturbances as real non-Walrasian theories. Just as there are many candidate real rigidities, there are many possible theories of this type.*"

This discussion suggests that whether there are multiple flexible-price equilibria or merely a unique but fragile equUibrium is not crucial to fluctuations. Suppose first that (as we have been assuming throughout this section) there are no barriers to nominal adjustment. If there are multiple equUib-ria, fluctuations can occur without any disturbances at all as the economy moves among the different equilibria. With a unique but fragile equilibrium, on the other hand, fluctuations can occur in response to small disturbances as the equilibrium is greatly affected by the shocks.

The situation is similar with small barriers to price adjustment. Strong real rigidity (plus appropriate insensitivity of the profit function) causes firms incentives to adjust their prices in response to a nominal disturbance to be small; whether the real rigidity is strong enough to create multiple equilibria when prices are flexible is not important.

6.15 Limitations

Real-business-cycle research has developed precisely specified models that take strong stands concerning the sources of shocks and how they are transmitted to the aggregate economy. As a result, it is often easy to confront the

models with the data and to identify ways in which they are unsuccessful. Keynesian theory, in contrast, encompasses a wide range of models, most of which are intended to address specific issues rather than to approximate the behavior of the economy as a whole. In addition, Keynesian accounts of fluctuations usually ascribe important roles to many different kinds of shocks and many different types of market imperfections.

These features of the Keynesian approach form the basis for the major criticism that can be made against it: Keynesian models are so vague and so flexible that they are almost impossible to refute. Like Ptolemaic astronomers with their epicycles to explain every new observation, Keynesian macroeconomists can modify their theories and postulate unobserved shocks to fit the data in almost any situation.

It is easy to find examples of the flexibihty of Keynesian analysis, involving issues ranging from the basic assumptions of the models to the specifics of individual episodes. Shortly after the publication of the General Theory, Dunlop (1938) and Tarshis (1939) provided strong evidence against its prediction of a countercychcal real wage. Rather than abandoning his theory, Keynes (1939) merely argued that its description of price-setting behavior should be changed. To give another example, the Keynesian response to the breakdown of the output-inflation relationship in the late 1960s and early 1970s was simply to modify the models to include supply shocks and core inflation. Similarly, confronted with clear evidence that the microeconomics of nominal adjustment differ greatly from what one would expect if the only barriers to adjustment are small fixed costs of changing prices, new Keynesians did not discard their theories; instead they argued that the actual barriers to nominal flexibility are a complicated combination of adjustment costs and other factors (D. Romer, 1993a), or that menu costs are just a metaphor that is no more intended to describe reality than is the Walrasian auctioneer of competitive models (Ball and Mankiw, 1994). And so on.

The same flexibility characterizes not just Keynesian models, but Keynesian accounts of specific episodes. The models allow for disturbances in essentially every sector of the economy-money supply, money demand, fiscal pohcy, consumption, investment, price-setting, wage-setting, and international trade-and thus are consistent with almost any combination of movements in the different variables. For example, conventional Keynesian accounts attribute the 1981-82 U.S. recession to tight monetary policy. The fact that most measures of money growth did not decline sharply is not viewed as an important problem for this view, but is accounted for by postulating a shift in money demand that was only partly acconmiodated by the Federal Reserve. Similarly, conventional Keynesian accounts attribute a large part of the 1990-91 U.S. recession to an unexplained fall in "consumer confidence."

Of course, it is possible that the economy is complicated, that there are many types of shocks, and that the modifications of Keynesian models

reflect gradual progress in our understanding of the economy. But a theory that is so flexible that it cannot be contradicted by any set of observations is devoid of content. Thus if Keynesian theory is to be useful, there must be some questions about which it delivers clear predictions.

One view is that the central issue on which Keynesian theory must stand or fall is the real effects of nominal disturbances. A central element of all Keynesian models is that nominal prices or wages do not adjust immediately. As a result, the models predict that independent monetary distiu--bances affect real activity. If this prediction is contradicted by the data, it appears that the models would have to be abandoned rather than modified, and that the study of fluctuations would have to pursue the real-business-cycle models of Chapter 4 or the real non-Walrasian theories of the previous section. If this view of the defining element of Keynesian theory is right, evaluating and extending the evidence described in Section 5.6 concerning the effects of monetary shocks is critical to business-cycle research.

Problems

6.1. Consider the problem facing an individual in the Lucas model when P, IP is unknown. The individual chooses L, to maximize the expectation of tJ,; U, continues to be given by equation (6.3).

(a) Find the first-order condition for I,, and rearrange it to obtain an expression for Li in terms of E[PilP]. Take logs of this expression to obtain an expression for ,.

(b) How does the amount of labor the individual supplies if he or she follows the certainty-equivalence rule in (6.17) compare with the optimal amount derived in part (a)? (Hint: how does E{\n(Pi IP)] compare with \n(E{Pi /P])?i

(c) Suppose that (as in the Lucas model) 1 ( ,- IP) = E\ln(Pi IP) \ Pi] + , where Ui is normal with a mean of zero and a variance that is independent of P,. Show that this implies that ln{£[(P, IP) Pi]} = E\MPi IP) \Pi] + C, where is a constant whose value is independent of P,. (Hint: note that Pt/P = exp{E[ln(PiIP) I Pi]} exp{u,}, and show that this implies that the £i that maximizes expected utility differs from the certainty-equivalence rule in (6.17) only by a constant.)

6.2. (This follows Dixit and Stiglitz, 1977.) Suppose that the consumption index C, in equation (6.2) is Q = [J,lo2/C""4/]"-", where Qj is the individuals consumption of good J and Zj is the taste shock for good J. Suppose the individual has amount / to spend on goods. Thus the budget constraint is SjKoPjCijdJ = Yi.

(a) Find the first-order condition for the problem of maximizing C, subject to the budget constraint. Solve for Qj in terms of Zj, Pj, and the Lagrange multiplier on the budget constraint.

(b) Use the budget constraint to find Cy in terms of Zj, Pj, Yi, and the Zs and Ps.