There has recently been revived interest in the role of money as a buffer stock; money balances, it is argued, are highly liquid assets that act as a 'shock absorber' and allow agents to postpone costly adjustments to holdings of other financial assets or to economic variables such as employment, investment, and output. Interest in the buffer stock approach has been prompted by the difficulties experienced in the estimation of stable demand for money functions, in understanding the 'long and variable lags' of monetary policy, and in the phenomena of interest rate and exchange rate overshooting under monetary targets.

The Centre held a one-day workshop on June 13 to assess whether buffer stock approach to monetary theory could resolve these difficulties and illuminate macroeconomic theory more generally. The workshop was opened by Keith Cuthbertson and Mark Taylor (Newcastle University), with a paper entitled 'Buffer Stock Money: An Appraisal'. They outlined the four main approaches to buffer stock money. The first can be found in work by Mervyn Lewis and CEPR Research Fellow Michael Artis, who attempted to remedy the instability of estimated money demand functions by advocating that the demand for money function should be inverted before estimation. If, for example, the price level is present in the demand for money function and this variable adjusts only slowly to its long-run equilibrium value, then the demand function should be estimated with the price level as the dependent variable. The second approach takes the Artis and Lewis analysis somewhat further, suggesting that the economy is characterized by pervasive disequilibria, and in particular an excess demand for money. This gives rise to a system of equations in which the variables that determine money demand are themselves affected by the excess demand for money.

The third approach described by Cuthbertson and Taylor was the 'shock-absorber' model, which originated with Carr and Derby. This model supposes that the economy is in equilibrium but is affected by random shocks, both real and monetary. Individuals' money balances respond to these shocks initially, and this gives rise to unanticipated levels of money holdings: it is these unanticipated holdings that constitute the buffer stock. Suppose the economy experiences a shock, i.e. an unanticipated change in the money supply. If expectations are formed rationally, anticipated changes in the money supply are immediately reflected in price expectations, and if prices are flexible, then real money balances are unchanged. Unanticipated changes in the money supply, however, will be absorbed in the buffer stock and the effect on prices will only be felt in later periods. Carr and Derby supported their theory with empirical estimates, using US data, of a demand for money function which included 'unanticipated money', i.e. the residuals from an equation 'explaining' the money stock. An immediate problem with their work, pointed out by McKinnon and Millbourne, is that the money stock is assumed to be exogenous, and the evidence does not seem to support this assumption. Other tests also suggest the falsity of the shock absorber hypothesis, and Cuthbertson had found that UK data confirmed this. The tests also involve the assumption that expectations are formed rationally, and the data rejected this hypothesis. If expectations are modelled as a learning process however, the Carr and Derby hypothesis is not rejected by the data.

In the fourth approach, characterized by Cuthbertson and Taylor as a 'forward-looking buffer stock model', individuals are assumed to (choose a series of desired) money holdings over future periods so as to minimize the costs associated with being out of equilibrium, making allowance for the costs of changing money holdings. This gives rise to a complicated estimation problem: the resulting demand equation relates the change in money demand to the unexpected changes in income, prices and interest rates, but the underlying theory imposes complex restrictions on the models coefficients. Cuthbertson and Taylor had estimated the forward-looking model, and argued that the results were reasonable and confirmed the role of money as a buffer stock. For example, the effect on money holdings of an unanticipated change in real income was three times that of an anticipated change: money holdings appeared to absorb income fluctuations. One drawback of this approach is that the buffer stock money holdings are the residual in the demand equation and are not modelled explicitly. Cuthbertson and Taylor then discussed an alternative model that used Kalman Filtering to model the buffer stock explicitly; estimation of this model had yielded similar results. The analysis suggested that models that allowed for forward-looking behaviour generally produced more satisfactory results, Cuthbertson and Taylor concluded.

There was some discussion of the technical issues raised by the 'forward-looking' approach. Mike Wickens (University of Southampton and CEPR) suggested an alternative method of modelling rational expectations; Jane Edwards (Queen Mary College, London) pointed out that the estimated standard errors were biased. Michael Artis (University of Manchester and CEPR) observed that Cuthbertson and Taylor were taking the money stock to be demand-determined, and pointed out that this was a controversial assumption. Peter Spencer (HM Treasury) argued that the significant costs of adjustment came not from changing money holdings but from changing the holdings of other assets, so that buffering could affect a wide range of assets.

In 'Buffer Stock Money and Money Demand Equations', James Davidson (LSE) discussed the problems of formalizing the buffer stock approach at a microeconomic level. Davidson hypothesized following Akerlof's earlier work that individuals have a range of acceptable money holdings. In addition to the upper and lower thresholds which define this range, there is a target point within the range to which the individual tries to return if money holdings cross one of the thresholds. Changes in money holdings which result from payments and receipts are termed 'autonomous', these are distinguished from 'induced' changes, designed to restore money holdings to the acceptable band. In such a framework, changes in the money supply will have no effect unless they cause the money holdings of some individuals move outside their thresholds. Even then it is only the induced transactions between the non-bank private sector and other sectors that will have real effects: other forms of transaction will merely lead to an increased velocity of circulation of money. Therefore there are two sources of lagged adjustment: that arising from the buffer stock itself, and that arising from the fact that some of the excess money is passed to other agents in the non-bank sector.

Davidson used this microeconomic framework to derive a money demand function, in which demand depends on lagged money holdings, desired long-run money holdings, and autonomous transactions outside the non-bank sector. This equation bore some resemblance to a conventional demand for money equation, but its coefficients were subject to a number of testable restrictions. Davidson noted that, in particular, if the money supply is assumed to be exogenous, then the money demand equation he had derived reduced to a money supply equation! Those who estimated traditional demand for money equations ought to have some faith in the endogeneity of money, he observed. Davidson also noted that the speed of response to a change in the money supply will depend on how many induced transactions occur within the non-bank private sector. If credit is rationed, Davidson argued, this proportion will be quite high and therefore monetary shocks will have a delayed impact on the real sector of the economy. The fact that individuals have a very limited portfolio of assets suggests that this proportion might be quite high. Davidson concluded by discussing some of the modelling problems which arose in this approach.

David Laidler (University of Western Ontario) argued that buffer stock analysis should be viewed from the perspective of a more general model of non-clearing markets. If markets cleared, buffer stock money had no real role to play. If markets were in disequilibrium, however, the buffer stock approach could be important. Laidler also stressed the distinction between the individual and the market experiment.

Charles Goodhart (LSE) argued that buffer stock analysis essentially involved responses to supply shocks. While agreeing with Cuthbertson and Taylor about the need to allow for forward- looking behaviour, he doubted whether such an analysis could explain, for example, the recent high rate of growth of M3. This must, he argued, involve a shift in demand, and Goodhart believed that this shift was related to the process of financial innovation. There was then the problem of distinguishing between responses to supply shocks and underlying demand changes. Traditional aggregate econometrics could not deal with this problem, he argued, although micro data might be of some use. In other words, the models available at the moment do not seem good enough to predict the impact of changes in the underlying financial structure and thus they are less than ideal for policy purposes. Laidler responded to Goodhart's comments by arguing that this meant that we really do not have good models, since the hallmark of a good model is its ability to forecast. Cuthbertson speculated that allowing certain parameters in these models to change may go some way to alleviating the problem.

David Begg (Bank of England and CEPR) then outlined some of the conclusions of recent work by Gregor Smith on 'Stochastic Inventory Theory and Money Demand'. Smith treats the holding of money and other assets as an inventory problem, and supposes that the cost of changing asset holdings is independent of the size of the transactions involved. Asset demands under such circumstances have thresholds similar to those discussed by Davidson. The value of the threshold is determined by forward- looking variables, and the speed of adjustment of asset holdings will depend on the probability of passing one of the thresholds. The degree to which buffer stocks of assets are held will depend on how predictable is the path of interest rates: if they are difficult to predict, adjustment will be more rapid and buffer stocks will not be held. Begg also suggested a possible solution to the apparent paradox that money demand responded more to long- term than to short-term interest rates: on the basis of Smith's analysis, the long-term rates may be acting as a measure of the expected value of the stream of future short-term rates.

Charles Goodhart raised the question of the policy implications of the buffer stock analysis. Over the past decade monetary policy had raised the real interest rate. Did the buffer stock analysis suggest that this was a desirable policy? Laidler suggested that high interest rates may in some part have been due to high real rates in the United States, although Peter Spencer disagreed, believing that the increases had begun earlier. Spencer speculated that some of the effect of high real rates is to choke off consumer expenditure, and Palle Andersen (Bank for International Settlements) agreed that this seemed to have been the case in Denmark. David Currie (Queen Mary College, London, and CEPR) believed that the evidence was mixed on this point. Andrew Bain (Midland Bank) drew attention to behaviour in the corporate sector, where financial transactions were much more frequent. The real lag in the impact of policy, Bain argued, occurred because moving between financial assets was very much cheaper than moving between real and financial assets. Mervyn Lewis (Nottingham University) argued that a purely monetary focus might be inappropriate, since we were moving from a purely monetary to a 'liquidity' economy.

David Currie argued that investigation at a more disaggregate level would be useful, and he wondered if disequilibrium econometrics had supplied any evidence. Some encouraging preliminary work was now available. Laidler agreed that the focus should be on disequilibrium and argued that one of the main attractions of buffer stock money was that it provided some form of rationale for the maintenance of disequilibrium, since buffer stocks of money allowed individuals to postpone the costly process of adjustment. Goodhart commented that the increasing instability in money demand may lead model builders to abandon money demand equations. Simon Wren-Lewis (National Institute of Economic and Social Research) suggested that, particularly for the corporate sector, the distinction between long- and short- term borrowing was very blurred, which only added to the problems of modelling. Spencer argued that the process of financial innovation was closely related to the fact that real interest rates were so high. This, he believed, was the consequence of the earlier M3 overhang and therefore may be only temporary.

The discussion revealed that the participants had different ideas about the importance and even the definition of buffer stocks. What emerged was an agreement that a full understanding of these issues was essential for a proper analysis of policy options.