RDP 7704: Money and Money–Income: An Essay on the “Transmission Mechanism” IV. The Role of Inflation Expectations

The discussion of the preceding section held expectations about prices constant, and kept them in the background. We are new in a position to look more closely at the role such expectations play in the transmission mechanism. Most recent work, because it has been done against a background of severe inflation, takes it for granted that future prices are expected to differ from current prices. Even the substitution of a constant expected rate of inflation for the constant expected price level implicit in IS-LM analysis somewhat complicates the analysis of the channels of causation that run between money and money income.

Equation (11a) tells us that, ignoring real growth, if output is held at its “natural” level, prices, in sence money income, will increase at the same case as the price level is expected to increase. The existence of a stable aggregate demand for money function implies that this can happen only if the money supply also expands at the expected rate of inflation in order to validate the price increase in question. If inflation expectations remain constant over time, then so will the actual inflation rate so long as the percentage rate of change of the money supply also remains constant at the appropriate value. Money income and the money supply thus grow at the same rate, but there is no sense in which one can be said to be “causing” the behavior of the other.^[24]

What really complicates matters, however, is not the existence of inflation expectations nor, as is well-known, that they influence nominal interest rates as well as price setting behavior; it is that they themselves change endogenously over time. To begin with, the presence of endogenous inflationary expectations makes it inappropriate to follow IS-LM analysis in treating an increase in the money supply as the typical expansionary policy and a decrease as the typical contractionary one. Once we permit inflationary expectations to enter as proximate determinants of the behavior of prices, so that it is possible for any ongoing rate of inflation to be validated by the appropriate monetary expansion rate, it is more helpful to think of a rate of monetary expansion in excess of the expected inflation rate as expansionary, and a rate of expansion below the expected inflation rate as contractionary. Thus in dealing with the “transmission mechanism” we should, as Friedman argued as long ago as 1958 (though not on the grounds advanced here), be concerned with a series of events set in motion by a change in the rate of monacary expansion rather than in the level of money supply.^[25]

Let us trace out the consequences of a particular expandionary policy, namely an increase in the monotary expencetion race occurring which the economy is initially in a full equilibrium situation. To do so, we must extend the transmission mechanism sketched out earlier to allow for the influence of endogenous inflation expectations on both price setting behavior and the time path of interest rates, but we must also say something about how such endogenous expectations are formed. Let us, for the moment, adopt a postulate, whose first order adaptive expectations special case is commonly found in the literature (but which is by no means uncontroversial because of that), that agents form expectations of inflation by observing the time path of the actual inflation rate, and extrapolating from it in such a way as to ensure that, if a constant inflation rate persists over time, the expected inflation rate will eventually come into equality with it.

An increase in the rate of expansion of the money supply to a pace faster than that necessary to validate an ongoing anticipated inflation will first lead to a buildup of real money balances, whose implicit own rate of return will therefore begin to fall relative to that on other assets. As a consequence, a process of substitution into all other assets and into current consumption will be set in motion, driving down interest rates, both observable and unobservable. The ensuing increase in current production will set in motion a multiplier process. Instead of having described the whole transmission process, as they would have done in the context of the IS-LM model, the last two sentences have sketched only the first step in a more complex pattern of events.

Along with the increase in output just postulacec goes a tendency for firms to increase their prices, and for money wages to rise, to levels in excess of the values these variables were initially expected to take. Given that there initially exists an expected rate of inflation, this involves an acceleration of the actual inflation rate relative to that expected rate. Is the actual inflation rate influences the expected rate, the latter must also begin to rise. In its turn, an increase in the expected rate of inflation has two interrelated effects on variables involved in the transmission mechanism. It puts upward pressure on the rates of interest that assets denominated in nominal terms bear, and in increasing the opportunity cost of holding money, accentuates the very portfolio disequilibrium which sets going the first stage of the transmission mechanism, and which accelerating inflation begins to offset.^[26] It is impossible to follow subsequent steps in this ongoing dynamic process with verbal argument. However, since both the FMP and RDX2 models contain elements of the transmission process we are here describing, as do the small-scale models of Laidler (1973) and Stein (1975), and since all four models are dynamically stable, the assumption that this process is stable is defensible. Let us then describe the new equilibrium to which the economy will eventually move; we can derive some information about the path whereby it is reached by considering its properties.

Because, as equation (11a) tells us, the expected and actual inflation rates will differ so long as output is not at its “natural” level, the new equilibrium, like the initial one, will see the economy operating at such a level of real output. The expected rate of inflation will be higher in this new equilibrium, and so the quantity of real balances held by the public will be smaller. If money is “superneutral” so that the “natural” output level is independent of the inflation rate, and of any past history of disequllibrum in the economy (both of these being dubious assumptions supported by no empirical evidence of which I am aware, and the former being contradected by a good deal of theoretical argument^[27]), then we would also expect to fund real rates of inetrest offered to their initial levels, with normal some interest: formation of inflation expectations challenges this assumption, and does so in such a way as to make the transmission mechanism even more awkward to get to grips with than the above analysis would suggest.^[30] This approach argues that economic agents act “as if” they form their expectations about the inflation rate by using the forecast that would be yielded by a correct model of the economy in which they are operating, and “as if” they expected every other agent in the economy to form his or her expectations in the same way.

A thorough-going application of this approach raises a number of difficulties. It ignores the fact that the gathering and processing of information is costly, so that many agents might not find it worthwhile to compute the (statistically speaking) optimal forecast of the inflation rate. As Benjamin Friedman (1975) has argued, it ignores the fact that agents are not automatically endowed with knowledge of the economy's structure, so that learning about it must be an ongoing process.^[31] And it ignores the fact, noted by Laidler (1976), Poole (1976), and Modigliani (1977) and recently subjected to formal analysis by Fischer (1977) and Phelps and Taylor (1977), that, if they are bound by long-term contracts, agents will be unable to act upon new information however it might affect their view of the future. It is one thing to expect the inflation rate to behave in a particular way, and quite another to act upon chat ex-pectation and hence anticipate the behavior of the inflation rate. Moreover, under rational anticipations, any change in the monetary expansion rate, not accompanied by an appropriate step change in the level of the money supply would lead to an instantaneously explosive inflation or deflation, unless the money market were cleared by an instantaneous, and unforeseen, step change in the price level before these anticipation became effective.^[32] However, arguments such as these dispose only of a particularly extreme form of the normal e-pectations notion, not of the idea in general.

A looser version of the same hypothesis would recognize that agents knowledge of the way in which the economy works is imperfect, that data on the behavior of particular variables are expensive to generate and process, and that changed expectations do not immediately lead to changes in activities. It would nevertheless insist that, for some agents at least, it is possible to use extraneous information on the behavior of such variables as the money supply, and others with which we shall deal in more detail below, in order to generate a more accurate forecast of the behavior of the inflation rate than could be had simply by extrapolating from past data on that variable, to do so at a cost which makes the exercise worthwhile, and then to act upon that forecast.^[33]

If some agents were to use data on the time path of the money supply in forming expectations of inflation, and then act upon these expectations, their behavior would “short circuit” the transmission mechanism that I have described above. The effect would be to make the expected rate of inflation that underlies price setting activities, and the determination of nominal interest rates, depend directly upon the behavior of the money supply. Thus if specialist dealers in security markets were to note that the rate of monetary expansion had increased, and to change the prices of securities to reflect changes in expectations of inflation, they could do so before any discrepancy between ex ante supply and demand in such markets appeared. More important, if the rate of monetary expansion increased, and this very fact led some firms to expect that there would be an increase in the inflation rate, they would be-gin to increase the prices of their output at a more rapid rate, without any intervening chain of asset disequilibrium or output change being necessary to prompt such behavior. If all agents acted in this way, and expected all other agents to do the same, and if we accept the somewhat artificial assumptions necessary to rule out instantaneous and explosive inflation, then the only effects of any change in the behavior of the money supply would be on prices. The transmission mechanism operating through portfolio disequilibrium and output changes would never be called into play and monetary policy would have no effect on any real variables. We have already seen that there are a number of reasons for not taking the extreme form of rational expectations that underlies these propositions too seriously. That in no way detracts from the importance of the hypothesis for the analysis of the transmission mechanism; because we have also seen that, even if only some agents act upon “rational” expectations, their activities imply the existence of yet another channel whereby monetary changes affect money income and expenditure, one which operates directly through expectations and their influence on price setting behavior.

Now, the arguments presented in this section of the paper have important, and, to a degree, destructive, implications for much of the evidence on the transmission mechanisms of monetary policy sketched out in Section II. If the division of changes in money income between the price level and output must be regarded as an integral part of that transmission mechanism rather than as a matter that can be analyzed separately from it, then we must pay particular attention to the way in which endogrenous, and variable, inflationary expectations impinge upon behavior when we study that transmission mechanism. This is of crucial importance in assessing how much attention we should pay to evidence on the influence of market interest rates on various categories of expenditure. As far as the demand for durable goods is ccncerned-both consumer and producer durables-it is real rather than nominal rates of interest that matter, or nominal rates taken in conjunction with expected inflation rates, and yet, all too often it is nominal interest rates alone that have been used in empirical work.^[34] Of course, if the expected inflation rates is more or less constant, variations in nominal interest rates will reflect variations in real rates, and little if any harm is done by using the former. It is, perhaps, not without significance that work utilising United States data drawn from periods starting after the end of the Korean War and terminating during the 1960s-such as, for example, Hamburger's study of the demand for consumer durables–has produced evidence showing an influence of nominal interest rates on expenditure, for this was a period of unprecedented price stability. Moreover, the studies of Tanner (1960) and Moroney and Mason (1971), based on an IS-LM framework cast in nominal terms, seemed to produce satisfactory enough results, and used data from these years. This too is significant, because the fore-going discussion would imply that the IS-LM model is more likely to be viable at times and places where the expected rate of inflation is approximately zero, and fluctuates little, than at others.

Whenever or wherever there is any reason to suppose that variations in the actual inflation rate might have been reflected in variations in the expected inflation rate, the relevance of the IS-LM framework becomes suspect. The same circumstances render any correlation, or lack thereof, between nominal interest rates and expenditures of any type irrelevant to deciding upon the existence, or otherwise, of a well-determined linkage between monetary policy and nominal income. After all, it is real, and not nominal, interest rates that should influence expenditure decisions. Moreover, the “rational expectations” hypothesis makes it possible to picture circumstances in which variations in the quantity of money can directly affect money income-specifically prices-without generating any evidence that expenditure decisions are senstivie to variations in interest rates or to any other relative price fluctuations. Such linkages will come into play and be observate only when the consequences of monetary changes are not fully anticipated by the agents generating the data for us to study.

This means that the crude correlations between money and money income which we discussed earlier are more important than one might initially have supposed, or than many might think desirable, as pieces of evidence about the transmission mechanism. That such correlations vary in strength between time periods and across countries may no doubt be explained in part by the considerations that earlier led us to the conclusion that the division of money income fluctuations between output and price level changes is of vital importance. However, such an explanation still leaves questions about the direction of causation between the variables to be discussed further. Results on this matter are, as we have seen, much more clear-cut for the United States than for other economies. In the following section of this paper I shall consider how the potentially endogenous nature of the money supply impinges upon the transmission mechanism and hence upon the way in which we might interpret these results.