Monetary Policy Uncovered: Theory and Practice

Transcription

1 International Review of Applied Economics, Vol. 18, No. 1, 25 41, January 2004 Monetary Policy Uncovered: Theory and Practice GIUSEPPE FONTANA* & ALFONSO PALACIO-VERA** *Department of Economics, University of Leeds, UK, **Departamento de Economía Aplicada III, Facultad de Ciencias Económicas y Empresariales, Universidad Complutense de Madrid, Madrid, Spain CIRA sgm / International Print Research 2004 Taylor January AlfonsoPalacio-Vera, Departamento , & 2004 Article Francis Review de Economía Online Ltdof Applied A plicada E conomics III,Facultad de Ciencias Económicas y Empresariales,Universidad Complute nse de Madrid,Madrid,2 8223,Spain.ap v@ccee.ucm.es. ABSTRACT This paper discusses the current new consensus view on monetary policy and the theoretical framework on which that practical view relies, namely, the targetsand-instrument approach. We argue that in the modern world of financial innovation and liability management central banks cannot choose between an interest rate-targeting policy and a money-targeting policy. A money-targeting regime is not desirable, if not unfeasible. In addition, in the context of Poole s approach to the instrument problem, the implementation of a money-targeting regime would raise the expected value of the loss function of the central bank and would thus shift the balance in favour of an interest-rate targeting regime. KEY WORDS: Monetary policy; Poole s approach; Post Keynesian economics; endogenous money Central bankers and even some monetary economists talk knowledgeably of using interest rates to control inflation, but I know of no evidence from even one economy linking these variables in a useful way, let alone evidence as sharp as that displayed in figure 1 [showing a simple correlation between inflation and money growth]. The kind of monetary neutrality shown in this figure needs to be a central feature of any monetary or macroeconomic theory that claims empirical seriousness (Robert E. Lucas, Nobel Lecture, 1996). Introduction In a recent issue of the Journal of Economic Perspectives, De Long (2000, pp ) has claimed that the influence of monetarism over current macroeconomics is profound and widespread. In particular, he argues that the emergence of monetary policy as one of the most critical government responsibilities is the result of the triumph of classic monetarism as developed, for example, by Friedman & Correspondence Address: Alfonso Palacio-Vera, Departamento de Economía Aplicada III, Facultad de Ciencias Economicas y Empresariales, Universidad Complutense de Madrid, Madrid, 28223, Spain. apv@ccee. ucym.es ISSN print; ISSN online/04/ Taylor & Francis Ltd. DOI: /

2 26 G. Fontana & A. Palacio-Vera Schwartz (1963). A similar sort of evocation of monetarism is also at the core of a recent paper by Laidler on the transmission mechanism (Laidler, 2002). Some policy-makers would find comfort in these assessments. They would see them as providing further evidence to the view that modern monetary policy is monetarist in nature despite some degree of play-acting, even deception, by modern central bankers. For instance, Pianalto, first vice president of the Federal Reserve Bank of Cleveland has argued that central bankers have finally abandoned the impulse to control fluctuations in national output and movements in the rate of unemployment. Economic policymakers, monetary and otherwise, are replacing their business cycle focus with a longer-term perspective (Pianalto, 2001, p. 3). Thus, according to Pianalto, current central banking practice would seem consistent with Milton Friedman s prescription that monetary authorities should avoid fine-tuning policies (Friedman, 1982). However, critics of monetarism would disagree with these conclusions. They would, for instance, recall another of Milton Friedman s prescriptions, namely the dismissal of interest rates as a policy instrument. Experience has demonstrated that it is simply not feasible for the monetary authority to use interest rates as either a target or as an effective instrument. Hence, there is now wide agreement that the appropriate, short-run tactics are to express a target in terms of monetary aggregates, and to use control of the base, or components of the base, as an instrument to achieve the target (Friedman, 1982, p. 101). Those critics would thus argue that the adoption of an interest rate policy by modern monetary authorities marks, if any, the collapse rather than the triumph of the monetarist position on policy. 1 They would also contend that the current practice of central banking is at least in part consistent with Keynes s theory and policy prescriptions. For instance, Dalziel (2002) has shown that the cornerstone of monetarism, the quantity theory of money, is no longer used by central banks. He maintains that, today, monetary policy is aimed to maintain aggregate demand growth compatible with supply-side capacity growth, a framework that he traces back to Chapter 21 of Keynes s (1936) General Theory. Other critics, such as Sawyer, have emphasised that modern monetary policy is supposed to operate through interest rates, which influence aggregate demand and thereby inflation. However, according to Sawyer, this is the way Keynes (1930) used the fundamental equations in the Treatise on Money for explaining the movement of prices (Sawyer, 2002; also Fontana, 2003). Does current monetary policy then mark the triumph of Friedman and monetarism, or is it rather the triumph of Keynes and his policy prescriptions? This paper invites a non-dualistic response to the question, and in this way aims to uncover the tension between the theory and the practice of modern central banking. The structure of the paper is as follows. The second section reviews the current practice of central banks. The third section discusses the theoretical framework on which that practice relies, namely, the targets-and-instrument approach and its core element, Poole s approach to the instrument problem. The fourth section focuses on the theoretical and practical problems faced by money-targeting regimes, and offers a Post Keynesian critique of them. The fifth section then shows that the implementation of a money-targeting regime would inevitably raise the relative variance of shocks to the monetary sector and, through changes in the co-variances of shocks to the commodity and monetary sectors, it would thus shift the balance in favour of an interest rate-targeting regime. The final section concludes.

3 Central Banking in Practice: the New Consensus View Monetary Policy Uncovered 27 The main tenet of the so-called new consensus view in macroeconomics is that monetary policy is the major direct determinant of inflation (Bernanke et al., 1999, p. 3). According to a prominent monetary economist, this means that if one were to discuss monetary policy with a representative group of central bankers they would undoubtedly start to discuss their policies in terms of the following expression (Laidler, 2002, p. 3): dp dm dv dy + dt dt dt dt (1a) The logarithmic growth rate of prices P that is, the inflation rate is equal to the logarithmic rate of growth of some representative monetary aggregate M, plus the logarithmic rate of change in the velocity of circulation of money V, minus the logarithmic growth rate of real income Y. According to Laidler (2002), those central bankers would also argue that in the long run changes in real income and the velocity of circulation of money are exogenous and, therefore, outside their control so that expression (1a) can be written in the following way: dp dm dt dt (1b) For what matters, expression (1b) is perfectly compatible with standard monetarist propositions. Causality is read from right to left, the independent and causal variable being the monetary aggregate M and the dependent variable being the price level P. Thus, any change in the money supply that monetary authorities succeed in bringing about manifests itself in the long run in higher prices but not in higher real output. 2 But does it really matter that expression (1b) could be made consistent with a monetarist approach to monetary policy? 3 What do central bankers really do in the day-to-day setting of monetary policy? Again, according to Laidler, the same central bankers would surely agree that in the short run they use roughly a set of three equations namely, an expectations-augmented Phillips curve, an IS curve, and a Fisher equation: ( d P P d dt dt gy Y = P e = ( ) Y Y h r, X r = P i d dt e ( (2) (3) (4) where (Y Y*) measures the output gap (for the problematic nature of the computation of potential output, see Dalziel, 2002), (dp/dt) e is the expected value of the inflation rate, X is a vector of variables that shift the IS curve, and r and i are the real and nominal rates of interest, respectively. However, as Laidler promptly recognises, there is an unresolved tension over the way the new consensus literature approaches the setting of monetary policy.

4 28 G. Fontana & A. Palacio-Vera Expression (1b) seems to make money all important for inflation in the long run while equations (2 4) treat it as irrelevant in the short run. From a monetarist perspective, equation (3) is a Trojan horse in the citadel of the quantity theory of money. What a true monetarist would like to see in its place is the following equation namely, a positively sloped LM curve: M P = miy (, ) (5) where (M/P) indicates the equilibrium real money balances of our economy and M is assumed to be exogenously determined. Of course, there is a good, though often neglected, reason for using an IS curve as in equation (3) rather than an LM curve. It is related to the so-called instrument problem, a major component of the targets-and-instrument approach that is currently used by most central bankers (Blinder, 1998). The debate around the instrument problem began with Poole s seminal work (Poole, 1970). In Poole s approach, the analysis was conducted in the context of a stochastic IS-LM framework where the money supply was assumed to be controllable by the monetary authorities. The crux of the matter was to determine the conditions under which an interest rate r policy was preferred to a monetary aggregate M policy. The main result of the debate was that, in theory, the latter policy tends to represent an optimal choice of monetary policy when ceteris paribus the variance of shocks to the commodity market is larger than the variance of shocks to the monetary sector. However, in practice, monetary targets were often missed and, in addition, at least in the US during the mid-to-late 1980s, fluctuations in money growth ceased to anticipate fluctuations in either output or prices. Central banks were thus led to interpret the empirical evidence as suggesting that the theoretical conditions for an interest rate policy had actually arisen. In a recent study using the VAR methodology, Friedman & Kuttner (1996, pp ) provide support for this view and, in addition, they show that in the US economy the variance of shocks to the monetary sector rose substantially relative to the variance of shocks to the real sector in the mid-to-late 1980s. Thus, the controversy between advocates of r targets and advocates of M targets was settled in practice. As argued by Blinder (1997, p. 7), these [theoretical] conditions then arose in practice, and one central bank after another abandoned M targets in favour of r targets. This is all well known, and the controversy over the optimal choice of monetary policy instruments is now history. However, there is a problem with this argument and the often-quoted successful interaction between theory and practice of central banking. Consistency with the targets-and-instrument approach requires that, were the relative size of the variance of shocks to the commodity market to be larger than the variance of shocks to the monetary sector, it might again be appropriate to use monetary aggregates as intermediate targets. This possibility may justify, among other things, keeping the base money-multiplier approach as the essential tool for explaining the money supply process in standard textbooks. Indeed, this seems to be the view of the president of the Federal Reserve Bank of New York, who is not short of praise for using monetary aggregates as a targeting framework for monetary policy (see also Volcker, 2002). According to him, it is only the momentary absence of a stable and predictable relationship between the money target and nominal income that prevents the adoption of that policy framework in modern times (McDonough, 1997, p. 4; also Taylor, 1995, p. 12, n. 1).

5 Monetary Policy Uncovered 29 However, critics like Moore (1988) and Wray (1990) as well as leading monetary practitioners like Goodhart (2002) have talked openly about the myth of the money multiplier. In particular, these authors have denounced the ex-post tautological nature of the multiplier and thus its uselessness for forecasting purposes. The money multiplier seems to ignore both modern liability management practices implemented by banks and the effects of the loan demand by the non-bank private sector. As shown in the next sections, these critics thus maintain that a moneytargeting strategy is neither feasible nor desirable. Before moving to the problems of the current use of the targets-and-instrument approach, it is worthwhile noting that there is much more than a numerical substitution in having an IS curve rather than a positively sloped LM curve in the set of equations (2) (4). For the day-to-day setting of monetary policy, central bankers seem to have lost any confidence in the ability of targeting monetary aggregates in order to deliver price stability. 4 It is the central bank s key interest rate that is now seen as the policy instrument for achieving the desired inflation rate through its effects on aggregate demand (Bank of England, 1999). Recent work by Arestis & Sawyer (2002) on current monetary policy in the United Kingdom confirms this view. They argue that interest rate-targeting policy has little to do with monetary aggregates. In the basic macroeconomic models of the Treasury and the Bank of England, the supply of money is not even mentioned. More importantly, the demand for money is either viewed as unstable or is treated as a residual. Similarly, in the macroeconomic model of the US economy used by the Federal Reserve, shifts in monetary policy are fully captured by innovations to the federal funds rate, with no role for monetary aggregates (Federal Reserve Board, 1996). Thus, what Laidler (2002) defines as an unresolved tension in the modern setting of monetary policy is simply evidence of the large bridge that has, for long time now, separated theory from practice. As Wray (1998, p. 98) explains, the central bank never has controlled, nor could it ever control, the quantity of money; neither can it control the quantity of reserves in a discretionary manner. For the current setting of monetary policy, the monetarist interpretation of expression (1b) is then simply a relic of what theorists suggested that central banks should do, and what these same central banks could not do. Therefore, a more general or encompassing interpretation of expression (1a) would be that, in the long run and in the absence of significant changes in the velocity of circulation of money, the money supply will move in line with nominal income. Similarly, the implied direction of causation, if any, in expression (1b) would then be from changes in nominal income to changes in the stock of money. 5 Conventional Theories of Central Banking: the Instrument Problem Framework The instrument problem of monetary policy arises because of the need to specify how central banks implement open market operations. In particular, the instrument problem consists of the choice of a variable that will be set directly by central bankers via the purchase or sale of securities in financial markets. A central bank may buy or sell a certain amount of securities, thereby providing or withdrawing the equivalent amount of bank reserves. Alternatively, a central bank may purchase or sell whatever amount of securities other market participants want to exchange at a specified price. In this case, central banks would let the market determine the quantity of bank reserves to be held at that price. Beginning with Poole

6 30 G. Fontana & A. Palacio-Vera (1970), the literature regarding the choice of monetary policy instruments has laid down the conditions under which a price variable, for instance an interest rate, was to be preferred to a quantity variable like borrowed (BOR) or non-borrowed (NBOR) bank reserves. In Poole s seminal contribution (Poole, 1970) a stochastic IS LM framework is formulated where the commodity and monetary sectors are subjected to exogenous random shocks. The monetary policy strategy that minimises a loss function, presented as the quadratic deviation of current Y from desired level of output Y, will be the preferable one: ( ) L= E Y Y 2 (6) If σ m is the standard deviation of disturbances to the monetary sector, σ u the standard deviation of disturbances to the commodity sector, and β 1 the income elasticity of the demand for money, a sufficient condition for a money-targeting regime to be superior to an interest rate-targeting regime is (σ m / σ u < β 1 (Poole, 1970, p. 206). In the context of the stochastic IS-LM framework used by Poole, targeting money supply damps the impact on income of disturbances to aggregate demand (i.e. to the IS function), whereas targeting the interest rate damps the impact of disturbances to the demand for money (i.e. to the LM function). In general, the choice of the monetary policy instrument depends on the values of σ m and σ u, the structural parameters of the model determining the slopes of the IS and LM functions, and the covariance of disturbances to the commodity and monetary sectors. Given the values of the behavioural parameters of the model and the covariance of shocks, a larger variance of disturbances to the commodity sector makes the money stock more likely to be the preferable instrument, and vice versa. As a result, the policy choice is inherently empirical. In recent years more complicated models have brought in other factors, yet both the logic and essential results of Poole s approach have remained unchallenged (Friedman, 1990, pp ). Notwithstanding the merits of Poole s approach, it suffers from several problems (see Palley, 1996a, p. 593), not least that it ignores the interdependence between the commodity market (IS function) and the money market (LM function). 6 The focus of this paper is on only two issues. First, Poole s analysis implicitly assumes that central banks can have full control of the money supply, if they choose to. However, most of the money used by the public, either as a means of payment or as a liquid store of value, represents the liabilities of private depository institutions. Central banks have some influence over the lending activity of these institutions, but they cannot fix the stock of money in a country. The money supply is not an exogenously set policy variable. Rather, it is the result of the income and portfolio decisions of the bank and non-bank private sector (Wray, 1998; Fontana, 2000). This is what Post Keynesians mean when they say that the money supply is endogenously determined as a residual of the economic process (Palley, 2002). Indeed, the increased ability of banks to avert central bank-imposed reserve constraints has led to the view that, under the current institutional framework, central banks can only affect money supply growth indirectly that is, through variations in short-term nominal interest rates. 7 For instance, Goodhart (1995, p. 252) suggests that, with a few possible exceptions, no central bank has ever run a true system of monetary base control and that instead, central banks sought to

7 Monetary Policy Uncovered 31 control the level of interest rates in order to affect the rate of monetary expansion indirectly. An important implication of this argument is that control of the rate of expansion of the money supply ultimately rests on the ability of central banks to affect bank lending to the non-bank private sector. This is a proposition that has long been advocated by Post Keynesians like Kaldor & Trevithick (1981), Moore (1988), Niggle (1990, 1991), Arestis & Howells (1992), Dow (1993) and Chick (1995). Within a Post Keynesian framework, the key question is how variations in interest rates impinge on the decisions to deficit-spend by the non-bank private sector and the willingness to lend by the banking system. In this respect, Arestis & Howells (1992, p. 149) argue that it is clear that the UK authorities now see interest rate changes as having an impact upon aggregate demand through a number of diverse channels. These include the cost of borrowing, income and wealth effects, and the exchange rate. In summary, the type of institutional arrangements characteristic of modern economies with developed financial markets make it very unlikely that both the theoretical and practical conditions necessary to run a money-targeting regime will be fulfilled. In an extension of Poole s analysis by Modigliani et al. (1970), the money supply process incorporates two types of disturbances. First, there are the standard shocks affecting the aggregate demand function and the money demand function. Second, and importantly, the model also includes shocks affecting the relationship between high-powered money and the money supply. The difference with Poole s analysis is that targeting interest rates now damps the impact on income of monetary disturbances that could consist of variations in the money multiplier as well as shocks to the money demand function. On the other hand, targeting highpowered money now damps to a lesser extent than in Poole s model the impact on income of shocks to the IS function. This is because aggregate demand shocks may be accompanied, for instance, by variations in the money multiplier, thus leading to changes of the same sign in the LM function. We will return to this point below. Secondly, another feature of Poole s approach is that the standard deviation of shocks to the monetary sector σ m and the covariances of shocks to the commodity and monetary sectors are assumed to be exogenous and, therefore, independent of the monetary policy regime implemented. But these parameters, and this is our main contention, are likely to be a function of the monetary policy regime pursued by a central bank. In particular, it will be argued below that the implementation of a money-targeting regime will raise σ m and vary the covariance terms in such a way as to shift the balance in favour of an interest rate-targeting regime. As a result, the choice of a money-targeting regime is likely to be self-defeating. Would then its implementation raise σ m and vary the value and sign of the covariance terms as much as to make an interest rate-targeting regime preferable? The experience with money-targeting regimes is not helpful here because, in practice, no central bank has ever succeeded in determining money growth with precision over a reliable period of time (Bernanke et al., 1999, p. 304). This fact raises serious doubts about the possibility of a central bank ever being able to implement successfully a money-targeting regime. 8 For instance, Friedman and Kuttner (1996, pp ) present evidence showing that the Federal Reserve did for a while genuinely use money growth targets to conduct monetary policy in the sense that it varied either the federal funds rate or NBOR in response to observed fluctuations of either M1 or M2 that departed from the corresponding target. However, around the mid-1980s, the Federal Reserve started to ignore monetary aggregates although legislation calling for their use remained in force for much longer. As

8 32 G. Fontana & A. Palacio-Vera argued in Bernanke (1996), even those countries traditionally associated with money-targeting regimes, such as Germany and Switzerland, often missed the announced monetary targets. In addition, the announced monetary targets were adjusted from year to year to reflect economic conditions and competing objectives of the monetary authorities. Monetary Aggregates Targeting: a Post Keynesian Perspective According to Post Keynesians, money-targeting regimes are undesirable, if not infeasible. For a money-targeting regime to be viable, at least two conditions must be met: (a) central banks need to have full control of the money supply and (b) the relationship between the money supply and nominal income needs to be stable. In turn, if the first condition is to hold, two additional sub-conditions must be fulfilled. First, central banks need to have full control of the level of bank reserves or so-called monetary base (a1). Secondly, a stable relationship between the monetary base and the money supply has to hold once the money-targeting regime is implemented, i.e. the money-multiplier needs to be stable (a2). The discussion below intends to show that, in practice, most (if not all) of these conditions do not hold in economies with modern financial systems. Do Central Banks Have Full Control of the Money Supply? The first type of problem for a money-targeting policy is related to the effectiveness of quantity control of bank reserves. Central banks may not indeed have full control of bank reserves (condition (a1)). Cramp (1970) and Kaldor (1985, p.10) were among the first modern economists to argue that central banks cannot close the discount window, since the maintenance of the solvency of the banking system is their most important function. More recently, several economists have highlighted several institutional features of the US and UK monetary policy frameworks that make attempts at controlling bank reserves ineffective, at least in the short run. For example, Goodhart (1994, p.1425) has argued that: If the CB [central bank] tried to run a system of monetary base control, it would fail. Much, perhaps most, of the time it would still be accommodating the day-to-day demand of the banking system for reserves at a penal interest rate of its own choice, whenever its M o [monetary base] target was below the system s demand for reserves. Otherwise when its target was above the system s demand, overnight rates would fall to near zero. Similarly, talking of the experience of the United States, Moore (1988, p. 122) has argued that, by keeping an unsatisfied demand of NBOR, the Fed controls the amount of discount-window borrowing and, indirectly, it sets the federal funds rate. As the demand for BOR rises, the marginal effective total cost (discount rate plus frown costs) of obtaining reserves rises above the discount rate. As a result, the federal funds rate rises pari passu above the discount rate. Increases in NBOR relative to total reserves (TR), with TR = BOR + NBOR, operate in the opposite direction, reducing BOR and, therefore, short-term interest rates. Regarding the newly created institution of the European Central Bank (ECB), its deposit and lending facilities set a lower and upper bound to overnight rates in the euro-zone. To the extent that, in principle, there is no explicit limit other than the assets used

9 Monetary Policy Uncovered 33 by banks as collateral to the amount individual banks can borrow (lend) through the lending (deposit) facilities, the ECB cannot closely control the amount of total aggregate reserves. Thus, more generally, it is fair to claim that central banks only set the supply price of reserves. As a result, the reserve supply function is horizontal in the market period, at an interest rate exogenously administered by central banks. Secondly, the money-multiplier is not stable (condition (a2)). The explanation for the instability and, hence, for the uncertainty attached to any forecast of money multipliers, were a money-targeting regime ever be implemented, relies on the idea that the money supply in modern monetary production economies is creditdriven and demand-determined. Credit-money is created when loans are granted by the banking system to the non-bank private sector, and is extinguished when loans are repaid, so that the level and the rate of expansion of the money supply are ultimately a decision of the private sector of the economy. The rate of monetary expansion is thus determined by the level of planned aggregate deficit-spending net of loans repayment. Thus, for instance, if a central bank increases NBOR through open market operations, banks may voluntarily opt not to increase lending if they do not identify any profitable outlet, and, instead, increase their level of excess reserves (ER). According to several authors, this was the behaviour of US banks in the aftermath of the Great Depression. Similarly, non-bank private sector units that is, firms and households can easily dispose of any unwanted money balances through loans repayment (the reflux mechanism). In these cases, an increase in NBOR will be offset by a rise in the aggregate reserve-deposit ratio and, therefore, a decrease in the money multiplier. Thus, any attempt by central banks to impose a high rate (relative to the rate desired by the private sector) of monetary expansion through, for instance, an increase in NBOR will be ineffective. This rate is ultimately determined by lending and/or net deficit-spending decisions of the private sector. Likewise, attempts by central banks to restrain the level of bank reserves below either the level of required reserves (RR), where TR = BOR + NBOR = RR + ER, or TR will encourage banks to look for reserves elsewhere. For instance, banks may borrow from the discount window or on the inter-bank market. In any case, interbank market interest rates will tend to rise and banks will only be able to obtain additional reserves at a higher cost. As a result, banks will be encouraged to implement portfolio adjustments on the liability side of their balance sheets by searching for alternative sources of funds with lower reserve requirements. These adjustments are commonly known as liability management practices (Earley & Evans, 1982; Evans, 1984; Goodhart, 1984, Ch. 3; Podolski, 1986; De Cecco, 1987). As these funds are obtained, banks transform their balance sheets, allowing them to increase their loan/reserve and deposit/reserve ratios (Pollin, 1991; Palley, 1996a, 1996b, 1998; Moore, 1998). As profit maximisers, banks continually seek to raise their deposit/reserve ratio by discovering new financial products and procedures, thus making money multipliers and money velocity especially for narrow definitions of money exhibit an upward trend. 9 However, with rising interest rates and increasing returns on non-reserve assets, banks will have an additional incentive to minimise average holdings of excess reserves (Davidson, 1990b, pp ). Furthermore, as interest rates rise (fall), banks feel more (less) pressure to search for alternative sources of funds with lower reserve requirements such that, other things being the same, the rate of financial innovation will tend to speed up (slow down) (Palacio-Vera, 2001). As a result of these two processes,

10 34 G. Fontana & A. Palacio-Vera money multipliers (as well as the velocity of circulation of narrow money) will also exhibit a pro-cyclical pattern. In this way, financial innovation loosens the connection between reserves and bank lending (Earley & Evans, 1982; Evans, 1984, p. 444). Since central banks cannot know in advance either the rate or the pattern of financial innovation, they will not be able to predict accurately the rate of growth of reserves necessary to obtain a given growth rate of the monetary aggregate selected as an intermediate target variable (Pierce, 1984). For instance, in the context of a stochastic IS-LM model, financial innovations will shift both the slope and the position of the LM curve in an uncertain and unpredictable way. Notwithstanding, it may well be the case that under a sufficiently elastic supply of reserves, money multipliers remain stable. This is the case when the central bank fully satisfies the demand for reserves of banks at a constant short-term interest rate. Thus, the issue of the stability of money multipliers is independent of the direction of causality between reserves and deposits. Is the Velocity of Circulation of Money Stable? The idea of a stable relationship between the money supply and nominal income (condition (b)) has not enjoyed much empirical support in recent times. Beginning in the 1970s, the instabilities showing up in money demand functions in the United States, in the United Kingdom, and in many other countries, led academics and policy makers alike to conclude that a money-targeting regime was not a viable option. For instance, in an influential study, Estrella & Mishkin (1997) conclude that in the case of the United States and Germany the empirical relationships involving monetary aggregates, nominal income, and inflation are not sufficiently strong and stable to support a straightforward role for monetary aggregates in the conduct of monetary policy. Similar results are presented by Friedman & Kuttner (1992, 1996). A common explanation for the abandonment of money-targeting regimes is thus the notion that the relationship between the money supply and nominal income has proven to be unstable. In more formal terms, the unstable nature of the relationship between monetary aggregates and nominal income shows up in the empirical observation that the various monetary aggregates are not co-integrated with nominal income that is, velocity is non-stationary (Goodhart, 1989, p. 316). 10 In turn, the break-up of this relationship is attributed to a combination of deregulation of financial markets, improvements in transactions technology, and financial innovation (Hester, 1981; Minsky, 1982, Ch. 7; Goodhart, 1984, Ch. 3, 1986, 1989; Podolski, 1986; De Cecco, 1987; Arestis & Howells, 1992; Clarida et al., 1999, p. 1685). It is now commonplace to conclude that changes in financial markets in the last three decades have made the growth rate of the money supply much less controllable and predictable. For instance, financial institutions have substantially increased their capacity to circumvent restrictive policies of the central bank by operating in wholesale markets. In particular, liability management practices have allowed banks to capture non-deposit funds in wholesale markets by paying market interest rates. In turn, as argued before, the ways these non-deposit funds alter the structure of the liability side of the balance sheets of banks is uncertain and unpredictable (Pierce, 1984; Podolski, 1986, p. 164). In short, financial innovation and liability management practices have made both the relationship between reserves (or monetary base) and deposits (or money supply) and the relationship between deposits and nominal GDP an unstable one. Under such circumstances,

11 Monetary Policy Uncovered 35 any attempt to run a money-targeting regime is bound to lead to severe short-run volatility in interest rates and, perhaps, even in money growth, as the monetarist experiment in the United States revealed (Friedman, 1984, 1988, p. 55). Money-Targeting, Liability Management, and the Instrument Problem A closely related explanation for the loss of stability of the money income relationship is Goodhart s Law (Goodhart, 1984, p. 96). According to this, any observed statistical regularity tends to collapse once pressure is placed upon it for control purposes. In the field of monetary policy, Goodhart s Law implies that the adoption of a money-targeting regime that exploits the (previously) observed stability of the relationship between nominal income and a specific monetary aggregate will lead to increasing instability and unpredictability in that relationship. In the context of the discussion on the instrument problem presented above, Goodhart s Law and the ability of banks to avert central bank-imposed reserve constraints by adopting liability management practices means that, were a central bank ever to adopt a money-targeting regime σ m would tend to rise. 11 Similarly, the covariance of shocks would vary in a way as to shift the balance of preference towards an interest rate-targeting regime. One reason is that, as experience has revealed, the implementation of a money-targeting regime substantially increases volatility in short-term interest rates (Friedman, 1988, p. 55). This volatility raises the degree of uncertainty about the expected value of the marginal lending costs, thereby encouraging banks to hedge against it. Since loan rates are determined prior to the determination of the lending costs, Cukierman (1991) argues that unanticipated credit or money demand shocks after banks have entered into loan commitments create a negative correlation between competitive deposit rates and bank profits (Cukierman, as quoted by Goodfriend, 1991, p. 14). As a result, the policy of smoothing short-term interest rates protects banks against the risk of widespread insolvency. In the absence of interest rate smoothing as would be the case under a money-targeting regime resorting to liability management practices allows banks to reduce this uncertainty but, as shown above, this in turn increases instability in the relation between reserves and the money supply on one hand, and between money and nominal income on the other hand. According to Goodhart (1995, p. 259), the likely outcome of short-term volatility in interest rates would be the development of a mechanism for shifting financing flows between banks and non-bank intermediaries. In turn, such developments would facilitate stabilisation of money growth rates at the price of inducing greater variability in the relationship between money and nominal income. Another reason why σ m would tend to rise and the covariance of the different types of shocks to change dramatically is that, under a money-targeting regime, if the income-velocity of money and loan demand are not sufficiently interest-elastic, implementation of a money-targeting regime will not only cause short-run volatility in interest rates but may also lead to higher real interest rates. This will occur if, for instance, the central bank attempts to impose on the private sector a rate of monetary growth that is lower than its desired rate. Interestingly, Mishkin (1998, pp ) argues that, in reality, the Fed monetarist experiment in the period was a smokescreen that was used by Paul Volcker to divert attention away from interest rates. In this way, the setting of targets for the growth rates of monetary aggregates allowed the Fed to set interest rates high enough to bring inflation down. Similarly, Goodhart (1989, p. 296; 2002, pp ) argues that the policies

12 36 G. Fontana & A. Palacio-Vera adopted in the early 1980s allowed the authorities freedom to raise interest rates to levels that could subdue inflation. However, several authors have argued that high real interest rates stimulate financial innovation. In turn, a higher rate of financial innovation reflects a higher σ m. For instance, Hester (1981, p. 183) argues that the one clear lesson from recent history is that financial institutions innovate whenever customer relationships are jeopardised by slow monetary growth. 12 Likewise, Minsky (1982, pp ) argues that high interest rates are likely to stimulate institutional changes that lead to an increasing lending ability of the banking system. Porter et al. (1979, p. 217) c also show that high market interest rates in , 1978 and early 1979 increased the incentive for managers to implement new cash-management techniques. In turn, the implementation of these techniques led to a reduction in their average holdings of bank deposits beyond what could be expected from higher interest rates according to any standard model of the demand for money by firms. All the arguments provided above can be formally presented by resort to an extension of Poole s original model developed by Modigliani et al. (1970) and intended to allow for some degree of endogeneity of the money supply. The model is made up of the following three equations: y= 1 r+ u m= y r+ 1 2 m= h+ r+ q 1 2 (7) (8) (9) where y is real output, m is the money supply, h is high-powered money, and u, v and q are random terms with zero-mean which capture shocks to the commodities markets, the demand for money and the relation between high-powered money and the money supply, respectively. Following the discussion above, it will be assumed that q and v capture respectively shocks to the money multiplier and the demand for money brought about by the implementation of liability management practices by banks. Therefore, equation (7) is an IS function, equation (8) is an LM function, and equation (9) is the relationship between high-powered money and the money supply. In principle, no assumption is made about the policy instrument of the central bank. It could be either r or h. Thus, the model is compatible with both an exogenous or an endogenous money supply approach. In the former case, h and the money supply will be the operating and intermediate-target variable of the central bank, respectively. 13 In the latter case, the interest rate will be the operating target, and the total amount of deposits will be determined by the demand for loans of the non-bank private sector. Since there is no credit market in the IS-LM model, equation (8) would then measure the residual of the loan supply process in terms of real money balances willingly held by the public. As shown in Friedman (1990, pp ), the minimum expected loss caused by output volatility under an interest rate-targeting regime is: ( ) = E y 2 2 u r (10) where σ 2 u is the variance of shocks to the commodities sector.

13 Monetary Policy Uncovered 37 Similarly, the minimum expected loss caused by output volatility under a money-targeting regime is: 2 ( ) = h E y 2 2 ( 2 + 2) u v q 2 2 1( 2 + 2) uv + 2 1( 2 + 2) 2 ( ) 2 uq 2 1 vq where σ 2 u is the variance of shocks to the demand for money, σ2 q is the variance of shocks to the relation linking h to the money supply, and σ 2 uv, σ2 uq, and σ2 vq are the corresponding co-variance terms. There are two considerations to make. First, the adoption of liability management practices by banks or the adoption of cash-management techniques by firms cannot be fully captured by increases in σ q and σ v, respectively. In particular, the behaviour of banks and firms will lead, along the lines of the discussion above, to structural change in equations (8) and (9). However, it may be argued that, for policy purposes the consequences of structural change are akin to a rise in σ q and σ v, i.e. the forecasting ability of the central bank is hindered. Secondly, the discussion above has not made a distinction between v and q. We do so below. Now, a rise in aggregate demand by the private sector (u>0) stimulates the demand for loans and this, in turn, leads to a rise in the demand for reserves. Under a moneytargeting regime, and in a static context, the central bank would keep h unchanged at its target value. As a result, interest rates in the money markets would rise. As interest rates increase, banks have incentives to restructure the liability side of their balance sheets in order to maintain or enhance their lending capacity. In other words, banks encourage their customers to move from high-reserves liabilities into low-reserves liabilities. In turn, the transformation of the liability side of the balance sheets of banks would cause a fall in RR. Thus, for a broad measure of money, q>0 in equation (9) so that σ uq should generally be positive in equation (11). For a narrow measure of money, however, the restructuring of the liability side of the balance sheets of banks does not have such a clear-cut effect upon the corresponding money multiplier. As the public shifts away from high-reserves deposits, the narrow measure of money contracts but, as banks reduce RR and subsequently increase their lending, new deposits are created such that the narrow measure of money expands. The net effect is uncertain. Regarding the sign of σ uv, the expansion in lending brought about by the fall in RR will lead to an expansion in the supply of both narrow and broad money. Since there has been a previous shift away from high-reserves into low-reserves bank liabilities, the demand for narrow money relative to its supply has, in principle, fallen so that v<0, but the demand for broad money relative to its supply would have changed in an uncertain way. As a result, for the broad money case the values of σ uv and σ qv in equation (11) are uncertain. In the narrow money case, it is quite likely that v<0 so that σ uv should be negative under normal circumstances. However, the value of σ qv is uncertain. Therefore, under a money-targeting regime σ 2 q and σ2 v will tend to go up, thereby raising the value of the minimum expected loss in equation (11). In addition, and in the broad money case, the changes occurred in the covariance terms may raise that value even further since σ uq >0 while the signs of σ qv and σ uv are uncertain. Similarly, in the narrow money case, it is likely that σ uv <0, whereas the signs of σ uq and σ qv are uncertain. In short, the implementation of a money-targeting regime tends to increase the value of the minimum expected loss in equation (11) relative to the value of (10) thereby shifting the balance in favour of an interest rate-targeting regime. (11)

14 38 G. Fontana & A. Palacio-Vera Conclusions This paper has shown that there is a not fully resolved tension in the modern setting of monetary policy. In the long run, money is made all-important for inflation, but in the short run it is treated as irrelevant. However, this tension is quickly resolved when a distinction is made between what some academics have suggested that central banks should do and what these same central banks felt right to do. In the first part of the paper we have analysed the theoretical foundations for the current practice of central banking, namely, the targets-and-instrument approach whose core element is Poole s approach to the instrument problem (Poole, 1970). This approach provides a criterion for choosing between a moneytargeting regime and an interest rate-targeting regime. We have then argued that the endogenous money hypothesis, i.e. the proposition that the level and rate of expansion of the money stock is ultimately determined by the demand for loans of the private sector (the demand for money of the latter being also relevant notwithstanding), precludes the implementation of a money-targeting regime. In addition, we have discussed several arguments whereby the implementation of a money-targeting regime will tend to raise the relative variance of shocks to the monetary sector and vary the covariances between shocks to the commodity and monetary sectors in such a way as to shift the balance in favour of an interest rate-targeting regime. The main result of our analysis is thus that the implementation of a money-targeting regime is undesirable on the basis of Poole s approach if not unfeasible. Acknowledgements The authors are grateful to the editor of this journal and an anonymous referee for stimulating comments. We are also grateful for comments and suggestions from Philip Arestis, David Laidler, Tom Palley, Randall Wray and participants at the First ADEK International Conference, held at the University of Bourgogne, Dijon, France, in November A previous version of the paper was written when Giuseppe Fontana was visiting research scholar at both C-FEPS and Economics Department, University of Missouri Kansas City, Kansas City, USA. He would like to express appreciation to members of those institutions for providing a stimulating and pleasant working environment and to David Foster for proofreading the text. Notes 1. Indeed, as a referee has pointed out, current central banking practice can be characterized as hyper fine-tuning policy, with central banks potentially changing short-term nominal interest rates monthly (or even more frequently) in pursuit of a medium-term inflation target. 2. This equation is a long-run relation as seen by monetarists. As a referee has noted, this equation does not say anything about real aggregate demand. Presumably, the equation only says that, in the long run, aggregate demand passively adjusts to aggregate supply, the latter being exogenously given. As a result, a higher rate of growth of money supply can only, as far as monetarists are concerned, lead to a higher rate of inflation in the long run. 3. Of course, in posing this question, this paper breaks away from Lucas s policy recommendation as expressed in his Nobel Lecture and quoted at the beginning of this paper (Lucas, 1996). 4. For instance, the ECB (European Central Bank, 2001, p. 47) remarks that although most empirical studies for the euro area support the view that there is a stable (long-run) money demand relationship linking M3 to the price level and other macroeconomic variables the reference value [for M3] is not a monetary target. The ECB does not attempt to keep M3 growth at the reference value at any particular point in time by manipulating interest rates.

15 Monetary Policy Uncovered In an empirical study for the period , Hoover (1991) shows that, for the US economy, the balance of evidence supports the view that money does not cause prices, and that prices do cause money. 6. For an earlier example of this critique, see Davidson (1990a), and further developments by Rochon (1999, pp ) and Palley (2002, pp ). In the context of Poole s framework, this interdependency could, in principle, be accounted for by the covariance of shocks to the commodities and monetary sectors as we show above. However, Poole assumes this covariance to be random and exogenous. 7. In this respect, Goodfriend (1991, p. 8) comments: except for the period from 1934 to the end of the 1940s when short-term interest rates were near zero or pegged, the Fed has always employed either a direct or an indirect Federal funds rate policy instrument. 8. Indeed, as argued in Goodhart (1984), both the low short-run interest elasticity and instability of loan demand make it extremely difficult for a central bank to fine-tune aggregate demand growth so as to hit annual money growth targets. 9. However, this upward trend is not a universal prediction. As shown in Pollin and Schaberg (1998), the spectacular increase in non-gdp transactions (financial and real state market trading) in the US economy since the early 1980s explains, at least partially, the so-called velocity puzzle, i.e. the fact that M1 velocity stopped rising in the early 1980s and has been falling since then. 10. According to Howells & Hussein (1999) the lack of co-integration of money and nominal income is also due to the growth of the money stock being influenced by non-gdp transactions. The latter have grown more rapidly than income in recent years. 11. As a referee has noted, not only may σ m tend to rise, but whatever monetary aggregate is used for policy control purposes, it will lose its previous meaning and, in effect, there should now be some other monetary aggregate which is being targeted. 12. Hester (1981) identifies two waves of financial innovation in the US economy, respectively and , which, according to him, were induced by restrictive monetary policies and high interest rates. 13. If the central bank instrument is assumed to be high-powered money, we then have that, in practice, the central bank varies the level of high-powered money to achieve a certain rate of interest which, in turn, determines the level of aggregate demand and, ultimately, (through equation (8) and together with the level of real income) the amount of money demanded by the public. Indeed, if we substitute equation (8) into (9) for the money supply, the resulting expression provides the level of high-powered money required (for a given level of real income) to set a certain interest rate, the money supply thus being a residual with no casual significance. Therefore, in this first case, the central bank would only use equations (8) and (9) for computation purposes. In the case where the interest rate is explicitly assumed to be the instrument, the central bank does not pay any attention to equations (8) and (9) and simply injects or withdraws high-powered money in order to set the interest rate which, in turn, affects the level of real aggregate demand. References Arestis, P. & Howells, P. (1992) Institutional developments and the effectiveness of monetary policy, Journal of Economic Issues, 26(1), pp Arestis, P. & Sawyer, M. (2002) The Bank of England macroeconomic model: its nature and implications, Journal of Post Keynesian Economics, Summer, 24(4), pp Bank of England (1999) The transmission mechanism of monetary policy, Bank of England Quarterly Bulletin, 39(2), pp Bernanke, B. S. (1996) What does the Bundesbank target?, National Bureau of Economic Research, Working Paper No. 5764, Cambridge, MA. Bernanke, B. S., Laubach, T., Mishkin, F. S. & Posen, A. S. (1999) Inflation Targeting: Lessons from the International Experience (Princeton: Princeton University Press). Blinder, A. S. (1997) What central bankers could learn from academics and vice versa, Journal of Economic Perspectives, 11(2), pp Blinder, A. S. (1998) Central Banking in Theory and Practice (Cambridge, MA: MIT Press). Chick, V. (1995 [orig. 1986]) The evolution of the banking system and the theory of saving, investment and interest, in: M. Musella & C. Panico (Eds) The Money Supply in the Economic Process: A Post Keynesian Perspective, pp (Cheltenham and Brookfield: Edward Elgar). Clarida, R., Galí, J. & Gertler, M. (1999) The science of monetary policy: a New Keynesian perspective, Journal of Economic Literature, 37(4), pp