The Endogeneity of the Money Supply and Expectations | RDP 7704: Money and Money–Income: An Essay on the “Transmission Mechanism”

RDP 7704: Money and Money–Income: An Essay on the “Transmission Mechanism” V. The Endogeneity of the Money Supply and Expectations

D.E.W. Laidler

August 1977

The standard IS-LM mcdel treats the money supply as independent of fiscal policy and it deals with a closed economy. These properties do not immediately imply that within that model the money supply must be treated as an exogenous variable, but they do limit the cases in which it can become endogenous to those where the monetary authorities set a target for the nominal interest rate and let the money supply adjust to any level necessary to achieve that target. It is certainly true that interest rate levels have frequently been pegged by central banks, not least as I have already noted, during the 1941–51 period in the United States, but the adoption of such a policy regime is in fact far from being the only possible source of endogeneity in the money supply.^[35] The way in which the government decides to meet its own budget constraint, and the exchange rate regime it adopts, both have potentially important implications for the way in which the money supply will interact with other economic variables, and hence for the way in which we should interpret evidence about that interaction.

It was Carl Christ's (1969) paper that first drew widespread attention to the government's budget constraint. Government expenditure not financed by taxes or by borrowing from the public must be financed by borrowing from the banking system. This latter method of finance, the only one Christ analyzed, necessarily involves money creation and, will alter the structure of portfolios in the private sector. In a series of papers (e.g., 1976), Brunner and Meltzer have greatly extended Christ's analysis, analyzing both money and bond financed fiscal policy in a model in which government debt, like money, is an imperfect substitute for private capital. Their work stressed the fact that policy induced is curve shifts must also lead to LM curve shifts except in the special case in which changes in government expenditure are matched by equal changes in taxes, even if we assume that government interest bearing debt is not to any degree net wealth. Moreover, because anything but complete tax financing of government expenditure implies the creation of money and/or government debt, these shifts will persist for as long as any deficit arising from a fiscal policy change also persists (unless money and debt are created in just the right proportions to keep the LM curve stationary).

These considerations have important implications for macroeconomics in general. They underlie the recent contributions of Blinder and Solow (1973) and Tobin and Bultor (1976), to the “crowding out” debate. They argue that provided government bonds are net wealth to a degree sufficient to ensure that a bond financed deficit sets in motion an expansion of output towards a stable equilibrium (or of the price level if output is held constant at a capacity level), the equilibrium in question will occur at a higher value of output than if the deficit were money financed. This result stems from two properties of their model. First, equilibrium occurs when the budget is brought into balance by an expansion of income that increases tax revenues. Second, the government's cash outlays must necessarily be higher when its debt has been increased by the issues of interest-bearing bonds than when it has been increased by the issue of non-interest-bearing money. The behavior of the private sector is important here only to the extent that it guarantees that the stable equilibrium about which inferences are drawn exists. Otherwise the characteristics of that equilibrium depend completely on the arithmetic properties of the government budget constraint. Thus, we need pass no judgement on its importance for macroeconomics in general to conclude, that the “crowding out” debate is of only peripheral interest in the context of the problems under discussion in this paper. However, none of this means that no noteworthy implications for the way in which we discuss the transmission mechanism and evidence about it show from other aspects of the literature on the government's budget constraint.

First we can hardly ignore the fact that if the monetary authorities of a country are expected to accommodate that country's treasury-as until recently they have been to a greater extent in Britain and Canada than in the United States-then expansionary fiscal policy will be accompanied by ex-pansionary meontary policy. In such circumstances, it is bound to be difficult for “reduced form” studies that rely on correlations between money and autonomous expenditure on the one hand, and money income on the other, to distinguish between the direct influence or fiscal changes, and of tho indirect influence that they have as a result of being financed by money creation. This is not to mention the problems that arise from government expenditure, as well as taxes, being variables that themselves depend in part upon real income and prices. The endogeneity of the deficit combined with the existence of the government budget constraint implies the possibility of causation running from money income to money as well vice versa. This would cause no problem if increases in money income could be relied upon to have a negative effect on the rate of money creation, since then, the two way causation would be easy to disentangle, but there can be no a priori presumption that this will be the case.

It is natural to think of an increase in money income leading to a fall in the budget deficit and hence in the rate of monetary expansion, because one normally thinks of an increase in income involving an increase in tax recepts and a fall in government expenditures. That is how “built in stabilizers” are expected to work. However, for some econcmies at least, it is crucial to consider the division of changes in money income between real output and the price level before accepting such a conclusion. For example, in Britain, until recently, government expenditure was planned in volume terms: nominal expenditures therefore automatically rose with the general price level. Moreover, government employees' salaries, and many transfer payments, not least those whose volume is inversely related to the level of real economic activity, for example, unemployment benefits, have increasingly come to be indexed in recent years, both in Britain and Canada, either formally or informally. On the revenue side, in Canada, income taxes have recently been formally, albeit partially, indexed. In Britain, marginal tax rates are constant across a wide band of income and the influence of inflation on affective real tax rates, that in the United States can perhaps be relied upon to produce a falling deficit in the face of inflation, is not so sharply present there.

The point of all this is that, in situations where the major source of an increase in money income is rising prices rather than rising real output, it is conceivable that, in countries such as Britain and Canada at least, the government's nominal deficit will increase rather than decrease. This is especially likely to be the case if we have a state of affairs in which rising money income is the net result of rising prices and falling output. If the nominal deficit is permitted to feed monetary expansion, this in its turn, will cause further increases in money income-in what could prove to be an explosive inflationary spiral. Thus there exists the possibility of the interaction between money and money income over time involving positive effects running in both directions. It would take empirical work, of a type not yet carried out, to test the validity of this conjecture. However, if it is valid then it would be difficult indeed to disentangle the complex causative patterns involved from the results generated by reduced form equations of the^[36] type used in the studies mentioned earlier, or indeed from studies that employ Sims' methods. The extent to which problems of this sort have influenced the outcome of work of this type for Britain and Canada would be well worth looking into.

The second but crucial, implication of the existence of a government budget constraint for the transmission mechanism arises when we analyze it in conjunction with the rational expectations hypothesis. Earlier we looked at the consequences of agents using direct observation of the rate of monetary expansion to form expectations about the behavior of the price level. However, if the money supply itself is an endogenous variable whose time path depends upon the size of the government's deficit, and the manner in which it is financed, rational agents could use information about the deficit, among other variables, to form expectations about the time path of the money supply in order to generate, in turn, expectations of inflation. We can thus conceive of the actual rate of monetary expansion as the sum of two components, one expected and the other unexpected. Expected variations in the monetary expansion rate should lead directly, via a rational expectations mechanism, to variations in the inflation rate, but unexpected variations should have their effects transmitted through a mechanism involving portfolio disequilibrium leading on to output and eventually to inflation rate fluctuations. Empirical work by Barro (1977) for the United States, and by Wogin (1976) and Saidi and Barro (1976) for Canada has put this proposition to the test. They all attempt to divide the monetary expansion rate up between a forecast component and residuals from that forecast, and then show that unemployment and output fluctuations correlate only with the residuals so derived. However they do not go on to show that forecast changes in the monetary expansion rate directly lead into changes in the inflation rate, and so their evidence stops short of providing a complete set of tests of the above propositions.

The work of Feige and Pearce (1976a) might be thought to have produced results that are inconsistent with these applications of the rational expectations notion. They apply optimal time series forecasting techniques to U.S. data on inflation for the period 1953–71, and then show that information on the behavior of the money supply does not permit any improvement to be made in inflation forecasts derived by these methods. However, the feedback rules used by the monetary authorities to decide upon their policy actions may bo implicit in the lag structure which Feige and Pearce's forecasting technique applies to the past behavior of the inflation rate. Thus their results are not in conflict with those of Barro, but do show that his is not a strong test of the rational expectations hypothesis. Sargent (1975) has stressed the need to test the rational expectations hypothesis against data sets within which different policy regimes have been in force in order to overcome the problems posed by considerations such as these.

The papers cited above are not the only ones that contain evidence that is relevant to the rational expectations hypothesis. Application of the notion to the analysis of the behavior of open economies under alternative exchange rate regimes leads to a set of predictions, which are supported by mor and stronger, empirical evidence than that which we have just considered. How the openness of an economy impinges upon the conduct of monetary policy and its transmission mechanism depends upon the exchange rate regime in force. Consider first a fixed rate. Its maintenance must involve a commitment by a country's monetary authorities to buy and sell their own currency at a fixed price in terms of others. Thus, unless sterilization is feasible, they must surrender control over the quantity of money, exactly as they would were they to peg the price of bonds instead of the price of foreign exchange. The traditional view of the operation of a fixed exchange rate has always recognized the balance of payments as a source of monetary expansion or con-traction unless reserve flows are sterilized; but the period of time over which sterilization operations can be expected to be successful is widely agreed to have diminished markedly with the growth of international capital mobility in the 1960s.^[37] Hence, we here neglect sterilization operations as being a short-term complication that does not alter the essence of the analysis.

The traditional view of the balance of payments mechanism under fixed exchange rates would lead us to expect that, if the inflation rate in the world economy were to accelerate, and if the domestic authorities did not simultaneously undertake an expansionary policy, the home country's balance of payments would become increasingly favorable. Its rate of monetary expansion would increase as a result, and ultimately a readjustment of its domestic price level and inflation rate to values compatible with balance of payments equilibrium would take place. In this traditional view, the purely domestic aspects of the chain of causation would be no different from those through which a change in the monetary expansion rate would operate in a closed economy, although it is worth noting that the same characteristics of international capital markets that render sterilization operations less viable also put severe limits on the extent to which domestic interest rates can deviate from those ruling in the world economy.

The rational expectations notion underlines this traditional view of the transmission mechanism in an open economy. If an increase in the world inflation rate is going to lead to an increase in the monetary expansion rate via the consequences of a balance of payments surplus, then rational agents would expect this to affect the time path of domestic prices. Hence domestic inflation expectation will be directly influenced by the time path of world prices. Domestic interest rates, even on assets not directly tradeable on world markets, would then rise when the world inflation rate increased. The effect of inflation expectations on price setting behavior would also result in there being a direct causative link between the behavior of world and domestic prices. If rational expectations work in this way, then monetary expansion, either coming through the balance of payments (or generated by a change in the rate of domestic credit expansion), must be regarded as acccmmodatino rather than causing any change in the time path of domestic money income the results from a change in the world inflation rate. Moreover, it is quite coceivable that money income changes, brought about by changes in the world inflation rate, might lead rather than lag behind accompanying variations in the quantity of money.^[38]

The process I have just sketched out could give all the appearances of “reverse causation” between money and money income in data generated by an open economy operating a fixed exchange rate. The key element in it is the role played by the world price level, by way of its effect on inflation expectations in determining domestic price setting behavior. Cross and Laidler (1976) in a study of 19 countries for the years 1952–1970 found that the behavior of the world price level did seem to influence inflation expectations in all of them (such influences being at a minimum in the case of the United States). Parkin, Sumner and Ward (1976), in a study of wage price behavior in the United Kingdom, generated a similar finding for that country as did Spinelli (1976) in a study of Italy for the period 1954–1973.^[39]

Exchange rate changes under a fixed rate regime also have, for given behavior of the domestic credit expansion rate, predictable consequences for the balance of payments and hence the money supply. It is therefore notewort that Laidler (1972) found that, over the period 1919–1970, the qualitative nature of price and output interaction for Britain could be predicted with an expectations augmented Phillips curve that utilized error learning, except for the years following exchange rate changes. Moreover, Carlson and Parkin (1976) derived an estimate of the expected inflation rate for Britain directly from survey data, and found that it increased markedly and otherwise unexplicably after the November 1967 devaluation. These results, dealing as they do with the behavior of expectations in the wake of abrupt policy changes, provide evidence in favor of what I have termed the “loose” version of the rational expectations hypothesis. Taken in conjunction with the more general arguments advanced earlier about the generation of inflation expectations in a fixed exchange rate open economy, they also go a long way towards explaining why British data in particular produce such ambiguous results about the direction of causation between money and money income.^[40]

It is worth noting that the expectations mechanism sketched above is not the only route whereby changes in world prices, or in the exchange rate, could impinge directly upon domestic prices without the intervention of changes in the money stock. The “mark-up” pricing hypothesis has long been an important component of models of the inflationary process, and still plays a crucial role in determining prices in large macro models. In open economies, the role of import prices as a component of production costs has long attracted attention as a means whereby inflationary impulses could be directly imported from abroad. For example, Dicks-Mireaux (1961) and Lipsey and Parkin (1969) both found that such a variable could be given an important role in an aggregate price determination equation for Britain, while import prices contribute significantly to the proximate determination of the price level in both the LBS model of Britain, and the RDX2 model of Canada. I am aware of only one test (contained in Laidler (1976b)) that has attempted tc discriminate between an expectations and an “import cost push” mechanism as the principal means by which world price level fluctuations are transmitted into open economies under fixed exchange rates. The results of that test favored the expectations mechanism, but, given that there has been only one test of this question, any conclusion based on its ouccome should be regarded as extremely tentative.^[41]

Either of the above mechanisms could in principle lead to an appearance of “reverse causation” between money and money income in an open economy, but rising prices accommodated by monetary expansion that might ben generated by either mechanism would exist only at the level of the individual economy. The foregoing analysis might help us to interpret evidence for individual open economies but, because it treats the behavior of the price level in the rest of the world as exogenous, it cannot be looked upon as providing a complete account of the relationship between the behavior of money and money income. A full treatment of the problems with which this paper is trying to deal, for a fixed exchange rate open economy, would have to include an account of what it is that determines the behavior of money income at the level of the “closed” world economy. It would treat the mechanisms described above as elements in the process whereby variations in economy-wide income impinge upon various regio of the economy.^[42]

The contrast between a fixed and flexible exchange rate regime is not necessarily a sharp one, particularly as far as the matters being discussed in this paper are concerned. The adoption of a flexible exchange rate does give the monetary authorities, in principle, the ability to do what they will with the domestic money supply. This does not mean, though, that the transmission mechanism for monetary policy will be just as it would be in a closed economy even if we ignore questions concerning expectations. For example, in the LBS model (as in RDX2) import costs play an important role in determining prices. Under flexible rates, the LBS model has expansionary monetary policy driving down the exchange rate. The subsequent rise in the domestic price of imports plays a key role in driving up domestic prices. (See Ball and Burns (1976).) Such a chain of causation obviously could not be incorporated in a model of a closed economy.

If we look to an expectations mechanism, rather than to an import cost mechanism, to link domestic prices to those in the rest of the world under fixed exchange rates, then the adoption of perfect exchange rate flexibility could break this link completely. However, it would do so only if all agents ignored the behavior of world prices, and of the exchange rate, in forming their expectations, if all agents concentrated solely on the behavior of domestic variables. The truth or falsity of such a proposition is obviously an empirical question. Rational agents might look only to domestic variables in forming expectations under a regime in which the monetary authorities were willing to ignore completely the behavior of the exchange rate in designing their policies, but they might equally look at the behavior of foreign prices and the exchange rate. In any event such a policy stance is an extreme one. When they abandon a fixed rate, monetary authorities frequently entertain the notion that there is a desirable value, or range of values, for the exchange rate.

If they do treat the exchange rate as a policy target, they must also stand ready to make the behavior of the domestic money supply compatible with the maintenance of whatever range of values they decide to aim for. Such a policy regime would differ only in degree from one of rigidly fixed rates. The domestic money supply would still be open to influence from the rest of the world in a systematic way, so that events in the outside world would be relevant to expectations about the time path of domestic prices. Thus, even with some degree of exchange rate flexibility, the traditional channels whereby monetary changes influence money income may be bypassed just as, it has been argued, they are bypassed under fixed rates.^[43] Arguments such as these might well explain why difficulties of interpreting money-money income correlations similar to those encountered with British data also arise in studies of Canada, despite that economy having operated nominally flexible exchange rates for a good part of the postwar period. This conjecture receives support from the work of Caves and Feige (1976) who, using a Sims-type test show that, for Canada, causation has run primarily from the exchange rate to money, rather than vice versa. This result implies that, under flexible exchange rates, monetary policy has been geared to maintaining the exchange rate at a particular value rather than to achieving domestic targets. However, things are not always so. Howson's (1976) conclusions about post-1931 Britain noted earlier, suggest that, in that episode, monetary policy was geared to achieving domestic targets and hence could be though of as “causing” the behavior of the exchange rate.

The implications of the foregoing analysis for the transmission mechanism between money and money income are easy enough to draw, but are of profound importance for the way in which we view macroeconomics. We have seen that there is evidence that agents do use information on the behavior of variables that influence the behavior of the money supply to form expectations about inflation. However, forecast changes in the money supply impinge directly upon money income through the effect of expectations on prices. Only changes in the rate of monetary expansion that are not forecast have their effects transmitted to money income through the traditional channels of portfolio disequilibrium and output changes. Thus, the extent to which a particular change in the monetary expansion rate is divided up between a forecast component and an unforeseen one will affect the way its effects are transmitted. This division of course depends upon the way in which agents themselves form their expectations, and we have empirical evidence to the effect that this in turn depends upon such things as the way in which the authorities finance their deficits and conduct themselves in the foreign exchange market. In short, the nature of the transmission mechanism for monetary policy itself depends upon the manner in which monetary policy is carried out and the way in which it interacts with fiscal and exchange rate policy.^[44] Far from regarding the structure of the economy as something which may be taken as given for purposes of analyzing alternative policies, we must recognize that it varies systematically with the conduct of policy. If the structure of the economy through which policy effects are transmitted does vary with the goals of policy, and the means adopted to achieve them, then the very notion of a unique “transmission mechanism” for monetary policy is a chimera. It is small wonder that we have had so little success in tracking it down.

Footnotes

It is worth noting that this essentially “Wicksellian” type of monetary policy is incompatible with the existence of rational expectation. So long as there is a gap between the natural and money rates of interest, there is a potential for inflation to go on without limit. Hence, with rational expectations, the price level would instantaneously explode. Note that Wicksell himself recognized this problem. Sea Wicksell (1898), p. 97. indebted to Malcolm Gray for discussion of this matter. [35]

This is a point of which Kaldor (1970) made much. However, the experience in Britain over the years since 1973 has involved large budget deficits combined with relative monetary stringency and does provide a potentially crucial experiemtn on this question. [36]

For a discussion of these issues see Bell (1974). [37]

The properties of the model in which this occurs have been investigated in Laidler (1975), Ch. 9 and Laidler (1976). [38]

Note that there now exists a series of working papers building on the work of Brunner (1974) dealing with the United States (Dutton (1977)), Germany (Neumann (1977)), Sweden (Myhryman (1977)), France (Fourcans (1977)), The Netherlands (Korteweg (1977)), Italy (Fratianni (1977)) that test, among other things, for the direct influence of world inflation on domestic inflation. All find a role for it to play, though, because with the exception of Korteweg's study, import prices, rather than world price in general , are used in this work, its results are not directly comparable to these undar discussion here. See also p. 40 below. [39]

The argument of this paragraph is not intended to imply that either this author or his colleagues knew that they were providing evidence on the rational expectations hypothesis in the work cited. We were aware that an open economy made a difference to the manner in which expectations were formed, and that the nature of the exchange rate regime would also make a difference. However we did not at the time recognize that this could be regarded as a special case of the more general phenomenon that is now known as rational expectations. [40]