Not everyone accepts the Rational Expectations Hypothesis, but most now agree that it is worthy of attention, at least as a tractable approximation in an imperfectly understood world. An attempt to justify the hypothesis would take us too far afield; instead, this essay examines its implications for monetary policy in a closed economy, although these arguments are easily extended to an open economy. I first consider how monetary shocks might affect output in the presence of rational expectations. Traditional theories of the transmission mechanism are undermined by rational expectations, but there are difficulties with alternative accounts. I also discuss the important implications of rational expectations for the definition of monetary policy and its relationship to fiscal policy. I conclude by examining whether monetary policy is capable of stabilizing output and more importantly whether such stabilization is desirable.


Does money affect output?

How might monetary shocks affect output? The neo-classical economists had a simple answer: it did not (at least in the long run). The money supply affected only the general price level: it had no effect on relative prices or any real variable. This is too simple, however, for we do observe a relationship between movements in the money supply and output.

One influential account of a transmission mechanism was advanced by Phelps and Friedman, involving a Phillips curve relationship between wage levels and employment with exogenously given or backward-looking price expectations. In the equilibrium version of this story, increases in the money supply drive up prices while wages move relatively slowly, held in place because price expectations adjust only sluggishly. Output therefore increases in response to falling real wages and rising profitability.

This account hinges on a labour supply curve in which there is a sizeable response of labour supply to current expected real wages. There is an information lag for workers with respect to the general price level; they perceive their own nominal wages accurately, but they cannot sample all the prices of the many goods they will be buying in a typical year. Workers therefore form an expectation of the price level based partly on the prices they do sample and partly on out-of-date general information about the economy, including the money supply. Price expectations will therefore incorporate the effect of lagged money supply movements on prices; but current money supply movements that could not be predicted will be incorporated into workers expectations only in so far as they are reflected in individual prices observed by the workers. These will reflect not only the money supply but also individual market influences and economy-wide real shocks; so when the money supply rises unexpectedly and drives up the general price level, price expectations will only partially reflect the actual increase in the price level. Real wages are actually lower than workers believe, and the workers are in a sense 'fooled' into supplying more labour. The demand for labour depends upon the maximizing decisions of firms and therefore on actual prices and real wages. Firms hire more workers because the real wage has fallen (although workers do not know this yet); output therefore rises. A money-supply shock therefore raises prices and output in a very familiar way, but only if the shock is unexpected. Expected movements in the money supply have no effect on output because workers fully incorporate this information into their price expectations and labour market behaviour.

This basic story can be enriched by allowing people to predict movements in the general price level from the prices they observe. They will base such inferences on the past correlation between these individual prices and unforeseen movements in the general price level. If the economy is one where monetary shocks dominate all others (e.g., one of high and variable inflation), then this correlation will be high. In this case, people will interpret any local price movements as a signal that the general price level has changed; their price expectations will change rapidly, and they will not change their labour market behaviour. Output therefore will be unaffected by monetary shocks. At the opposite extreme, when there is monetary stability, expectations concerning the general price level will respond only very slowly to changes in individual prices.

The development of this equilibrium account of the transmission mechanism is mainly due to Barro, Lucas, and Sargent and Wallace. It is often called the 'islands' story because individuals are consigned to local markets and cannot obtain up-to-date information on the rest of the economy. The main difficulty with such insularity is its reliance on a rather implausible information deficiency. Aggregate information on the price level and the money supply is routinely available with a very short delay, at least in developed economies. It is hard to believe that errors in forecasting prices or the money supply are therefore serious enough to mislead workers into significant changes in their labour supply. The story seems most plausible in developing countries where aggregate information is subject to long delays or is unreliable. Indeed Lucas, in a 1973 study, found cross-country support for his islands theory from a sample containing many developing countries.

There is an alternative, disequilibrium version of the Phillips curve developed by Stanley Fischer and by Phelps and Taylor in the 1970s. They assumed that workers and firms sign contracts extending over several periods, in which labour is supplied in whatever quantity firms should require, at a nominal wage fixed in the contract. In this model output also responds to unanticipated shocks in money and prices. An unanticipated rise in the price level lowers the real wages paid by firms. Firms can obtain all the labour they want at a nominal wage which is fixed by contract and therefore at a lower real wage; firms therefore expand employment and output. Output is therefore affected by any money supply movements which were not predicted at the date of the longest-running contract.
The difficulty with this account lies in its assumption that contracts fix nominal and not real wages. This assumption must ultimately rest on the existence of transactions costs in changing nominal wages; without such costs, contracts signed between firms and risk-averse workers would index nominal wages to the price level, so that real wages were the object of the contract. Why should this indexation not occur in contracts? The experience of European countries such as Italy and Belgium, where formal indexation close to 100% is practised, suggests that real wages can be made largely impervious to monetary shocks. If this is so in countries where there is a statutory indexation mechanism, why should it be any less feasible in other countries?

The difficulties with both versions of the Phillips curve have led to a search for alternative 'nominal rigidities' which might explain the observed relationship between money and output. A natural place to look for such rigidities is in asset holdings. Non-indexed government bonds are one example of assets denominated in nominal terms, which are in effect contracts whose terms were set at various dates in the past, rather like overlapping wage contracts. Holdings of such assets would affect aggregate demand and hence output, provided these assets are regarded by the private sector as additions to its net wealth, because of incomplete discounting by the private sector of its future tax liabilities.

One may enrich this model, in which bonds appear as net wealth, by appealing to a possible confusion, not between local and economy-wide price shocks, as in the island story, but between temporary and permanent monetary shocks. Suppose that monetary and real shocks each have both a transitory and a permanent component, and that these components cannot be distinguished in the current period. This emphasis on the confusion between transitory and permanent components can be found in the 1980 work of Brunner, Cukierman and Meltzer, although they use it rather differently. This approach suggests that money does have an effect on output, one which is similar to that described in recent work by Lucas. The greater is the variability of the permanent relative to the transitory component of money, the smaller is the effect of a monetary shock on output. Real interest rates vary directly with output in this approach, but these results depend on the private sector treating its bond holdings as net wealth.

This approach places confusion firmly where it belongs, namely in our knowledge about the future. Individuals know today everything there is to know about shocks today; what they cannot know is which of these shocks will endure into the future. The basis for this confusion is a lack of information, but it is not the sort of information that could readily be provided by more surveys; it is not factual information, rather it is interpretive information. Suppose there is an unexpected rise in the money supply. That is a fact, to be established by a statistical recording process. But what one really needs to know is what that fact means for the future: have the authorities decided to raise the level of the money supply permanently, or the growth rate permanently, or is this merely a temporary disturbance? How can this be discovered? One may ask civil servants or Ministers, and discover their intention; that in turn may be conditional on how backbenchers or the government's supporters in the country react to events. One may ask a sample of these their opinions. And so on. There is no fact to be discovered at this time, because no one knows what will be.

Some researchers have argued that explanations of a transmission mechanism linking the money supply to output are wide of the mark: observed correlations between money and output reflect the influence of output on money! Long and Plosser have argued in their theory of the 'real business cycle' that correlations between output and money reflect causation running from output to the money supply. The banking system, it is argued, accommodates the higher real balance requirements resulting from a higher level of output. Money shocks have no causal influence on output in this theory. If correct, this theory would devastate macroeconomic explanations of business cycle history. For example, it would suggest that the Great Depression happened because of some unknown shock to tastes and technology. It is too early to pass judgement on this explanation, but it is worth noting that Kam-Fai Chan and CEPR Research Fellow Michael Beenstock have found evidence for the United Kingdom that an increase in economic activity will tend to expand the money supply through the operation of the money multiplier (CEPR Discussion Paper No. 106, 'The Determinants of the Money Multiplier in the United Kingdom').

This discussion has shown that, under rational expectations, there are many mechanisms by which monetary shocks may affect output. All these mechanisms are implausible in some respects, however. Information delays are relevant mainly in developing countries; contracts are drafted in nominal terms only when the transactions costs of indexation are high; money may only have a significant role as wealth when alternative assets are restricted; and there may be no money-denominated bonds. Institutional details of the monetary framework will therefore be important in assessing through which channel, if any, monetary shocks work. This judgement may lead us to accept a real business cycle theory for certain situations, like that of a high-inflation country with a weak government incapable of enforcing a currency monopoly. It also suggests that the monetary explanation of the Great Depression due to Friedman and Schwartz can stand up if we can show that the monetary framework of that time made one at least of these channels operative - surely not a difficult task.

Fiscal constraints on monetary policy

Rational expectations also has important implications for the definition of monetary policy and its relationship to fiscal policy. The literature on the 'government budget constraint' drew attention to the instability which could arise if monetary and fiscal policy were 'inconsistent'. Such an inconsistency would arise if a fiscal deficit were permanently bond-financed. Because expectations are forward-looking, models which incorporate this hypothesis have in principle to be solved into the indefinite future in order to arrive at a solution for the current period: this implies that a model must have a reasonable long-run equilibrium if it is to be useful. Under rational expectations, therefore, the long-run instability threatened by inconsistent policies becomes an immediate one: it is impossible even to define an equilibrium path. Thus fiscal deficit policy implies bounds on feasible monetary policy. Furthermore, the future paths of monetary policy implied by fiscal policy have effects on current output and inflation.

This has important and rather surprising implications. Suppose we make the very mild assumption that in any steady state, the ratio of real government debt to GDP should approach a constant, thus ruling out 'Ponzi financing' of government. The government budget constraint now implies that the growth rate of money in a steady-state equilibrium depends upon the ratio of GDP to the steady-state deficit inclusive of real debt interest (sometimes called the 'inflation-adjusted real deficit') minus an allowance for growth.

This condition can be stated in another way: money growth must equal the ratio of the Public Sector Borrowing Requirement to GDP, multiplied by the 'velocity' of 'outside money'. Though exactly equivalent, one statement uses the inflation-adjusted deficit while the other uses the unadjusted deficit. When the inflation-adjusted deficit is used to assess whether fiscal policy would be consistent with a particular counter-inflationary monetary policy, great care must be taken to allow for the effects of falling inflation and interest rates on the value of outstanding bonds. This adjustment can be very large indeed when a large proportion of these bonds are non-indexed and of long maturity, and it may be that large cuts in the government deficit excluding interest are necessary for consistency. The assessment is easier when the unadjusted deficit is used, since debt interest will not change in nominal terms except for short- maturity stocks which are rolled over before inflation comes down. The debate on the Thatcher government's Medium Term Financial Strategy has not always taken into consideration this point. Willem Buiter and Marcus Miller, for example, have used the inflation-adjusted deficit to argue that fiscal policies under the Medium Term Financial Strategy were 'unnecessarily' restrictive. Their method of adjusting for the effects of current inflation on the current market value of debt, however, have led Buiter and Miller astray.

Consideration of a steady state in which the ratio of debt to output is constant also highlights the essential interdependence of monetary and fiscal policy, a point made most forcefully by Sargent and Wallace in a 1981 paper. For a given time path of money supply growth, larger deficits along the path require either higher inflation or a reduced deficit at the end of the path. Conversely, for a given time path of fiscal deficits, lower monetary growth along the path implies higher debt and so higher inflation at the end. Sargent and Wallace took this argument one step further. If expectations were forward-looking, then at given levels of government expenditure and tax rates, lower money growth today could produce higher inflation today as well as in the long run! This surprising result has been shown to hold in more general models. The importance of the general result is clear, and further generalizations in more dynamic models are of great interest.

Monetary policy as an output stabilizer

Monetary and fiscal policy are therefore interdependent, and it is difficult to analyse the stabilizing role of monetary policy in isolation. One way of avoiding this complex interdependence is to think of monetary policy as 'independent' in the short to medium run, but constrained by or constraining the fiscal deficit in the long run. This procedure also has the merit that monetary stabilization policy - to which we turn next - can be thought about separately from fiscal policy or long-run monetary growth.

I take as axiomatic the role of monetary policy in controlling inflation via the systematic long-run control of money supply growth. The role of monetary policy in stabilizing output fluctuations demands more careful examination. Before the advent of the rational expectations hypothesis, no one doubted that in principle monetary policy could and should stabilize output, given slowly moving price expectations. The debate resolved around three assumptions: that the central bank has up-to-date information at least as good as that available to the private sector, that the central bank has a good model of the economy with which to forecast the effects of policy, and that it is efficient in implementing required policy. Those opposed to an 'activist' monetary policy argued that all three assumptions were invalid and concluded that activist policy would be at least as likely to increase as to dampen fluctuations.

Rational expectations has widened the scope of this debate. Individuals in the private sector incorporate knowledge of the central bank's reactions into their expectations, if these are indeed formed rationally. Under certain conditions this can neutralize the effects of monetary policy on output. In general, rational expectations complicate the economy's responses to stabilization policy, and this raises questions about the desirability of activist policy.
An economic theory must upset strongly held policy convictions in order to be noticed and to acquire a following quickly - Keynes and Friedman understood this, and the policy-ineffectiveness proposition advanced by Sargent and Wallace in 1975 proved the point once more. Sargent and Wallace dealt a strong blow to activist policy prescriptions by merely adding the assumption of rational expectations to a standard Keynesian model with an expectations-augmented Phillips curve. They demonstrated that a monetary authority possessing no better information than its private citizens could not have any influence on the behaviour of output by implementing any planned policy action. Only the authority's monetary 'errors', i.e. the deviations from its monetary plan, could affect output.

This proposition had two effects on the economics profession. It created an antipathy to the hypothesis of rational expectations among those committed to monetary stabilization policy. It also generated a voluminous literature searching for ways to generate policy effectiveness. Yet, as Lucas and Sargent later emphasized, the proposition was never intended as more than a cautionary tale, designed to illustrate the general point that monetary policies may have effects very different from those supposed to occur under naive expectations hypotheses. The intended moral was that 'expectations must be allowed for carefully.'

David Peel and I have surveyed the ways in which policy effectiveness may be reestablished in rational expectations models. If one retains the assumption that the government and the private sector have access to the same information, then any nominal rigidity will do the trick. All that is required is that some nominally denominated asset or contract has a maturity longer than the information lags on nominal variables. The intuition is obvious. Suppose the economy experiences a shock at time t, and its effects on prices are perceived in period t+1. Then provided there are some contracts or assets whose value in period t+2 is affected by the price level in period t+2, the central bank can, by changing the money supply and so the price level, offset the effects of the shock during period t+2. Phelps and Taylor and Fischer showed this for nominal wage and price contracts, while my work with Peel shows that this holds for wealth effects of nominal government liabilities. We have to appeal to these very rigidities in order to argue that money has any effect on output at all, and the case for an activist monetary stabilization policy rests on the same arguments.

We have written as if there were no possibility for the money supply to react immediately to the current information available to the central bank. Yet such reactions are what most central bankers themselves think of as monetary policy; reserve injections in response to rises in interest rates, exchange rates and so on. Fifteen years ago William Poole analysed the monetary authority's choice between a policy of fixing interest rates and one of fixing the money supply. Interest rate policies, he argued, would stabilize output in response to monetary shocks but destabilize it in response to real shocks, while policies to govern the money supply would do the opposite. This analysis suggests that the central bank should employ a combination of the two policies which reflected the relative likelihood of monetary versus real shocks to the economy.

While Poole's approach remains valid for the models with nominal rigidities, it is less convincing when information on interest rates is generally available and expectations are taken to be rational. People can infer the current money-supply response from interest rate changes, and can then incorporate this information into their expectations, so that the policy response has no effect on output. This argument against Poole's approach would be inapplicable only if the central bank reacts to information which is available only at an individual market level, but this is hard to argue in the case of variables such as interest rates and exchange rates upon which central bank policies are typically based.
In spite of these difficulties the balance of the argument seems to confirm the effectiveness of monetary policy even in the most 'classical' models. But this does not establish its desirability. Three main problems have been identified: the 'Lucas critique', whereby the system's response can change as policies change; the time-inconsistency or credibility problem raised by Kydland and Prescott, whereby announced policies may not be followed through; and the appropriateness of using social welfare functions based on output gaps and inflation to design or assess stabilization policies.

Lucas's critique is very convincing, given the complexities introduced when people second-guess government intentions and attempt to interpret economic signals. It essentially returns us to Milton Friedman's original argument: that policy-makers do not have a good enough model of the economy to be certain that their policies will improve matters. The Lucas critique suggests that models can never be good enough: even if adequate within the sample period over which they are estimated, their structures will shift outside the estimation period as private sector behaviour changes in response to new policies which are introduced. This also constitutes a powerful argument against basing monetary policy on models which rely on nominal rigidities, for the structure of these models above all will shift with changes in the policy framework.

Time inconsistency is a general problem with any policy where the policy-maker is not bound by some external force to stick to his announced policies. The assumption of rational expectations only adds force to the problem. It has played a prominent role in recent discussions of international economic policy coordination. (See the article by David Currie and Marcus Miller in Bulletin No. 13 for current CEPR research in this area, as well as Discussion Papers No. 94 and 102.) The essence of time consistency is that the policy-maker induces the private sector to behave in a certain way, based on what the private sector expects the government to do given announced policy. Once the policy-maker has induced this behaviour, the situation is now a fresh one in which it will generally pay him to renege (provided this is a one-off episode).

The best macroeconomic example is the announcement of a tough anti-inflation policy, which causes people to expect low inflation in the future and so to sign moderate wage contracts today. Once the private sector is committed to moderate wage growth, the government will be tempted to expand the money supply to achieve higher output growth as well as not quite so low inflation. Unfortunately this tactic cannot be repeated indefinitely: people come to realize in advance that the government will renege, and so they expect high money growth and high inflation. The government then finds that it has to deliver exactly this amount of money growth and that attempts to increase output will result only in higher inflation.

Though it may seem attractive to allow the central bank to exercise discretion in its attempts to stabilize the economy, the potential benefits of stabilization may be lost through the failure to carry out announced policies. Recent research by Paul Levine and Programme Co-Director David Currie (reported in Discussion Papers No. 94 and 102) and by David Backus and CEPR Research Fellow John Driffill has examined whether a government's concern for its reputation would be enough to limit the temptation to renege on its policy announcements. We do not yet have firm answers because it is not easy to model convincingly the behaviour of political decision-makers or political parties, although I have attempted to do so in Discussion Paper No. 79. I would conjecture that reputation, though important, is not enough to limit the abuse of discretion.

The implications for stabilization policy are clear. Stabilization policy is in practice always discretionary; there are no cases known to me where a pre-committed stabilization policy rule has been adopted. Therefore the scope for backsliding or 'U-turns' is greater when stabilization responses are permissible. It is minimized when rules, which in practice should be simple and fixed rules, are enforced by some higher authority such as a constitutional court.

The social welfare functions used for assessing optimal stabilization policy are usually simple functions of what appears to concern policy-makers, i.e. output or unemployment and inflation. But is this how policy-makers should act? In a world of rational expectations individuals with the same access to macroeconomic information as the government may be presumed to have reacted to this information in an optimal fashion already. If shocks occur, individuals should react. If, for example, oil prices change then certain industries should contract and others expand: the respective rates of change may not exactly coincide, so that a recession (or boom) may occur. This represents an optimal reaction. To smother it would lower welfare.

This argument has been made powerfully by CEPR Research Fellow Michael Beenstock and also by Thomas Sargent. The argument must be qualified when there are distortions at a microeconomic level, the most obvious of which are taxes and unemployment benefits that increase the desired level of unemployment in the economy. When such distortions are present, I would argue, welfare will be raised by reducing unemployment below its 'natural' (equilibrium) rate. This would appear to justify stabilization policy and, in principle at least, the presence of unemployment (or output) in the social welfare function. Nevertheless it does not follow that stabilization is preferable to a policy which removes the distortions. This will only be so if there is some strong political objection which makes their removal impossible.

I would conclude from these arguments that rational expectations has weakened but not destroyed the case for monetary stabilization policy. The conditions for successful policy are difficult to achieve, and the onus of proof has been shifted onto those who wish to pursue it despite all the difficulties described above.

Rational expectations has changed fundamentally the way economists think about monetary policy. The channels through which policy operates, the constraints upon policy, and its potential to stabilize the economy all look different in the light of rational expectations. But very few of the insights provided by the hypothesis can be regarded as at all certain. If anything, money and the institutions that accompany it - the nominal denomination of contracts and assets - are more of a mystery than ever before. How can we explain why rational people commit themselves to arrangements whereby purely nominal shocks will affect the real quantities they buy and sell? Until we can convincingly unravel that mystery, we will not be able to claim a firm foundation for a theory of monetary policy.

Patrick Minford is Edward Gonner Professor of Applied Economics at the University of Liverpool, a Research Fellow in the Centre's programme in International Macroeconomics, and a member of the Economic Policy Panel. The arguments contained in this article are developed at greater length in CEPR Discussion Paper No. 104 entitled 'Rational Expectations and Monetary Policy'. Further details of the research described in this article can be found in Discussion Papers No. 11, 63, 78, 79, 94, 102 and 106.