EP011.tex What Is Wrong with the Eurosystem’s Money-Growth Indicator, and What Should the Eurosystem Do about It?¤ Lars E.O. Svensson Institute for International Economic Studies, Stockholm University; CEPR and NBER; Homepage: www.iies.su.se/leosven November, 2000 The primary objective of Eurosystem monetary policy is to maintain price stability, de…ned as an annual increase in the HICP below two percent.1 Because of the lags in the e¤ects of monetary-policy actions on aggregate demand and in‡ation, monetary-policy actions cannot a¤ect current in‡ation and output, nor in‡ation or output in the near future. A rough benchmark is that monetary policy a¤ects output in about a year and in‡ation in about two years. Therefore, Eurosystem monetary policy has to be guided by in‡ation forecasts about two years ahead. As discussed further in Svensson [11], for successful policy, the Eurosystem must construct conditional in‡ation forecasts. These forecasts should depend on all relevant information, in- cluding the Eurosystem’s view of the transmission mechanism for monetary policy, its view of the current economic and monetary situation within and outside the Euro area, information about current and future …scal policy, private-sector in‡ation expectations, etc. In particular, the forecasts should be contingent on alternative paths for the monetary-policy instrument rate, that is, the interest rate on the main re…nancing operations. This way, the Eurosystem can select an instrument-rate path, for which the conditional in‡ation forecast about two years ahead is in line with the de…nition of price stability, and then set the instrument rate accordingly. ¤Brie…ng paper for the Committee on Economic and Monetary A¤airs (ECON) of the European Parliament for the quarterly dialogue with the President of the European Central Bank. I thank Stefan Gerlach for comments and Annika Andreasson for secretarial and editorial assistance. Expressed views and any errors are solely my own responsibility. 1 As many commentators have suggested, it would be better to formulate this de…nition in an unambiguous and symmetric way as a point in‡ation target, say 1.5%, possibly with a tolerance interval, §1%. Since forecasts are crucial in a forward-looking monetary policy, transparency requires that the Eurosystem’s forecasts are published (including the assumptions and reasoning used in constructing the forecasts) and made available for external scrutiny. This makes it easier for the general public to hold the Eurosystem accountable for its conduct of policy, which in turn provides stronger incentives for the Eurosystem to conduct good policy. The world is waiting for this publication to occur, presumably this fall. State-of-the-art forecasting for monetary policy, including assessments of the uncertainty of the forecasts, is presented, for instance, in the regular In‡ation Reports by the Bank of England and by Sveriges Riksbank and in the regular Monetary Policy Statement by the Reserve Bank of New Zealand.2 In spite of the example set by these central banks, the Eurosystem has selected a contro- versial two-pillar approach in its monetary-policy strategy. The …rst pillar refers to monetary developments and emphasizes a money-growth indicator, the growth of M3 relative to a reference value (currently 4.5%/year). The second pillar is a “broadly-based assessment of the outlook for price developments and the risks to price stability in the euro area... using a wide range of economic and …nancial variables as indicators for future price developments.” As many commentators have pointed out, there is neither theoretical nor empirical support for this separation of monetary developments, in general, and the money-growth indicator, in particular, from other indicators. The only reasonable approach is to use all relevant informa- tion, including international developments, output-gap estimates, cost and wage developments, private-sector in‡ation expectations, monetary developments, etc., in the construction of an in‡ation forecast. That is, an obvious improvement in Eurosystem strategy is to combine the …rst pillar with the second pillar, and rely on one pillar only. Any information in the mon- etary aggregates that is deemed to have implications for future in‡ation should be combined with other information in order to construct the in‡ation forecasts. This is, for instance, the approach taken by the Swiss National Bank, after its announcement in December 1999 that it will abandon monetary targeting and instead rely on an in‡ation forecast in monetary-policy decisions. The separation of monetary developments into the …rst pillar appears even stranger when combined with theoretical and empirical insights about the insigni…cance of, or even lack of, any information in money growth for future in‡ation at horizons relevant for monetary policy 2 Many other central banks publish regular forecasts; recently the Swiss National Bank and Bank of Japan have started to do so. 2 (about 1–3 years). Signi…cantly, almost two and a half years after its creation, the ECB has not yet provided any theoretical model or any empirical results supporting the money-growth indicator. Theoretical criticism of the money-growth indicator, for instance, by Svensson, using a traditional aggregate supply/aggregate demand model in [8] and the so-called P ¤ model in [9], shows that nominal money growth is not the best predictor of future in‡ation. Instead, a policy aimed at stabilizing nominal money growth leads to a policy reaction function that di¤ers signi…cantly from the one corresponding to e¢cient stabilization of in‡ation around an in‡ation target. Empirical research reaches the same conclusions. Estrella and Mishkin [2] and Stock and Watson [7] …nd that money growth contains little or no information on future in‡ation. In empirical models based on U.S. data, Rudebusch and Svensson [6] …nd that monetary targeting, in the sense of stabilizing nominal money growth, would cause considerable instability in in‡ation and the output gap, both in a traditional aggregate demand/aggregate supply framework and in a P ¤ model (in spite of the P ¤ model often being used by advocates of money-growth targeting). Thus, the results of Rudebusch and Svensson support the theoretical criticism of the money- growth indicator in Svensson [8] and [9]. For Euro-area data, in the …rst part of a recent ECB Working Paper, Trecroci and Vega [12] demonstrate that the hypothesis that money growth does not Granger-cause in‡ation cannot be rejected. This jargon means that there is no evidence that current money growth helps to predict future in‡ation in the Euro area (more precisely, there is no information in money growth that is not already available in other indicators).3 Somewhat surprisingly, preliminary results by Gerlach and Svensson [3] (con…rmed by Tre- croci and Vega [12] in the second part of their paper) for Euro-area data indicate that the level of the real money gap (the gap between the current real M3 stock and the real M3 stock con- sistent with a long-run equilibrium) helps to predict future in‡ation. But the real money gap is di¤erent from nominal money growth. The distinction is similar to the distinction between the output gap (the gap between real GDP and real potential GDP) and nominal output growth. Gerlach and Svensson …nd that the money-growth indicator does not contain any information about future in‡ation that is not already contained in the real money gap and the output gap (they also conduct Granger-causality tests with the same conclusion and hence support the con- clusion of the more extensive Granger-causality tests of Trecroci and Vega [12]). Thus, although 3 Even if the Eurosystem would eventually be able to publish results indicating that the money-growth indicator contains information about future in‡ation at relevant horizons, it is worth noting that it is not enough that the money-growth indicator helps in constructing unconditional forecasts. What is need is a structural forecasting model that allows the Eurosystem to construct conditional forecasts, forecasts that are conditional on alternative monetary-policy actions. 3 the empirical results of Gerlach and Svensson support the P ¤ model, they also support the theoretical criticism of Svensson [9] that the P ¤ model does not provide any support for the money-growth indicator. Furthermore, although the results of Gerlach and Svensson suggest that the real money gap, rather than the nominal money-growth indicator, may be one of the useful indicators for the Eurosystem, they do not support the idea that monetary developments should be detached from other indicators into a separate pillar rather than integrated with the general construction of forecasts. Similarly, empirical results by Meltzer [4] and Nelson [5] suggesting an e¤ect of the real money stock on aggregate demand propose that such an e¤ect is taken into account when constructing output forecasts; they do not imply any support for the money-growth indicator as a predictor of future in‡ation. With regard to the real money gap, the P ¤ model that underlies the analysis is highly controversial and so far lacks so-called microfoundations (that is, it has not been rationalized by sound microeconomic principles). Furthermore, the real money gap shares with the output gap the dependence on notoriously uncertain estimates of potential output. The real money gap, in addition, requires a stable long-run demand for money, in particular a stable income elasticity of money. There is some evidence in favor of a stable long-run money demand for Euro-area data, for instance, in Coenen and Vega [1] and Gerlach and Svensson [3], although practical experience from several countries shows that shifts in the long-run demand for money are quite common and may occur at any time. The real money gap is hence a fragile construction, even more so than the output gap. Counter to some conventional wisdom, as discussed in Rudebusch and Svensson [6], a stable money-demand function does not imply that monetary targeting is advisable or that the money- growth indicator is good predictor of future in‡ation. Indeed, a stable money-demand function does not necessarily imply that even the contemporaneous correlation between nominal money growth and in‡ation is very high. Intuitively, one can understand why from the following identity, In‡ation ´ Nominal money growth¡Real money growth, or more formally, ¢pt ´ ¢mt ¡¢(mt ¡ pt); where pt is the log price level and mt is the log M3 stock in quarter t, so ¢pt ´ pt ¡ pt¡1 is the rate of in‡ation, ¢mt is the rate of nominal M3 growth, and ¢(mt ¡ pt) is the rate of growth 4 and the real M3 stock. Thus, real money growth draws a wedge between in‡ation and nominal money growth. If it were the case that real money growth was very stable, there would be a close relation between nominal money growth and in‡ation. However, in practice, real money growth is far from stable. Empirical money-demand functions (for instance, in Coenen and Vega [1], Gerlach and Svensson [3] or Rudebusch and Svensson [6]) are variants of the so-called error-correction form ¢(mt+1 ¡ pt+1) = ¡·m[(mt ¡ pt)¡ (·0 + ·yyt ¡ ·iit)] + "t+1; (0.1) where real money growth, ¢(mt+1 ¡ pt+1); is negatively proportional to the deviation between current real money, mt ¡ pt, and the long-run real money demand, ·0 + ·yyt ¡ ·iit; where yt is real GDP, it is the opportunity cost of money, and "t+1 is a shock corresponding to an unexplained residual (·0 is a constant and the coe¢cients ·y, ·m and ·m are positive). Even if the long-run money demand is stable (in the sense of ·0, ·y and ·i being constant over time) and there are no or only small shocks "t+1 to real money growth, unavoidable variability in real GDP and the opportunity cost of money will cause real money to deviate from the long-run real money demand, which will cause variability in real money growth. This variability in real money growth will reduce the correlation between in‡ation and nominal money growth. What should the Eurosystem do in this situation, now that the theoretical and empirical evidence against its money-growth indicator is accumulating? (The ECB has demonstrating commendable openness and scienti…c integrity by publishing some of this evidence itself.) First, the Eurosystem should reduce its emphasis on nominal money growth and the money-growth indicator, and instead refer more to levels of real money and credit aggregates, perhaps start reporting the real money gap. In particular, it should always report its view on the impact (in- cluding the magnitude and time perspective) on in‡ation of monetary developments. (Tellingly, there is currently no attempt to report any likely consequences on future in‡ation of monetary developments.) Second, the Eurosystem should consolidate the two pillars into one, and use all its information and resources to produce the best forecasts. Central banks traditionally avoid admitting mistakes, perhaps believing that such admissions would negatively a¤ect their credibility. I believe such beliefs are irrational. Correcting mistakes and demonstrate learning and improvement would most likely improve credibility. A majority of experts watching the Eurosystem would certainly welcome such a development. Delaying 5 obvious improvements of the monetary strategy may not only deteriorate the quality of policy but also adversely a¤ect the credibility of the institution. References [1] Coenen, Günter and Juan-Luis Vega (1999), ”The Demand for M3 in the Euro Area,” Working Paper No. 6, European Central Bank. [2] Estrella, Arturo, and Frederic S. Mishkin (1997), “Is There a Role for Monetary Aggregates in the Conduct of Monetary Policy,” Journal of Monetary Economics 40, 279–304. [3] Gerlach, Stefan, and Lars E.O. Svensson (1999), “Money and In‡ation in the Euro Area: A Case for Monetary Indicators?” Working Paper (www.iies.su.se/leosven). [4] Meltzer, Allan H. (1999), “The Transmission Process,” in Deutsche Bundesbank, The Mon- etary Transmission Process: Recent Developments and Lessons for Europe, MacMillan, London, forthcoming. [5] Nelson, Edward (2000), “Direct E¤ects of Base Money on Aggregate Demand: Theory and Evidence,” Working Paper No. 122, Bank of England. [6] Rudebusch, Glenn, and Lars E.O. Svensson (2000), “Eurosystem Monetary Targeting: Lessons from U.S. Data,” Working Paper (www.iies.su.se/leosven). [7] Stock, James H., and Mark W. Watson (1999), “Forecasting In‡ation,” Journal of Monetary Economics 44, 293–335. [8] Svensson, Lars E.O. (1999), “Monetary Policy Issues for the Eurosystem,” Carnegie- Rochester Series on Public Policy 51-1, 79–136. [9] Svensson, Lars E.O. (2000a), “Does the P ¤ Model Provide Any Rationale for Monetary Targeting?” German Economic Review 1, 69–81. [10] Svensson, Lars E.O. (2000b), “Monetary Policy and the Current Economic and Monetary Situation,” Brie…ng Paper for the Committee on Economic and Monetary A¤airs (ECON) of the European Parliament, June 2000 (www.iies.su.se/leosven). [11] Svensson, Lars E.O. (2000c), “Forward-Looking Monetary Policy, Leading Indicators, and the Riksbank’s In‡ation Report vs. the ECB’s Monthly Bulletin,” Brie…ng Paper for the 6 Committee on Economic and Monetary A¤airs (ECON) of the European Parliament, September 2000 (www.iies.su.se/leosven). [12] Trecroci, Carmine, and Juan Luis Vega (2000), “The Information Content of M3 for Future In‡ation,” Working Paper No. 33, European Central Bank. 7