German monetary targeting: a retrospective view.

Linda S. Kole and Ellen E. Meade, of the Board's Division of International Finance, prepared this article. Matthew Field provided research assistance.

Nineteen ninety-five marks the twentieth year of German monetary targeting and the eighth year that the Deutsche Bundesbank has set an objective for the monetary aggregate M3.(1) Among the central banks of the major industrial countries, only the Bundesbank continues to use a monetary aggregate as its guidepost for monetary policy. In the other Group of Seven (G-7) countries - Canada, France, Italy, Japan, the United Kingdom, and the United States - the statistical relation between prices and the supply of money has weakened and become unstable over the past decade and a half, making money an unreliable guide for policy. As a result, monetary aggregates have been relegated to a smaller role in monetary policy decisions in these countries.

The erratic growth of German M3 in recent years has raised questions about the usefulness of the Bundesbank' s target as well. Although the German central bank appears to be committed to M3, the aggregate's recent volatility, particularly since the political and economic unification of East and West Germany in 1990, has called into question the stability of the statistical relation between German money and prices.

The primary goal of German monetary policy, as set forth in the Bundesbank Act of 1957, is to safeguard the currency, which in practice has been interpreted to mean keeping price inflation at or below 2 percent per year. The Bundesbank believes that targeting M3 is the most reliable means of achieving this goal. At its meeting on December 22, 1994, the Bundesbank Council (the central bank's policymaking body) set the monetary target for 1995 and described it as follows:

In 1995 the Deutsche Bundesbank will conduct its monetary policy in such a way as to ensure that inflation continues to decline and, at the same time, that the monetary conditions for sustained economic growth remain in place. To this end, the Bank regards it as appropriate for the money stock M3 to expand by 4 percent to 6 percent between the fourth quarter of 1994 and the fourth quarter of 1995 ....

The Bundesbank is thus abiding by its tried and tested strategy of monetary targeting, and underlines that, despite the disruptive influences of the past few years, it still regards the money stock M3 as the key reference variable for its monetary policy.(2)

The study reported in this article explored the question of whether M3 targeting remains a useful approach to achieving low inflation in Germany by investigating the stability of German money demand in recent years. The findings indicate that at the time of unification, when residents of eastern Germany initially acquired deutsche marks, there was a change in the short-run dynamics of money demand as well as a shift in the level of M3 demanded. Over the longer run, however, unification apparently has not altered the demand for money in a fundamental way: The long-run, or equilibrium, relationship among German real M3 and its key determinants - real economic activity (GDP), interest rates, and real net financial wealth - seems to have remained largely unchanged over the past two and one-half decades despite unification and other major events. That said, the period since unification is short relative to the entire period studied, and more time must pass before it can be determined conclusively whether money demand has remained stable.

The article begins with a brief history of monetary targeting in the G-7 countries and then summarizes in some detail the German experience with monetary targeting. Next, recent studies of the stability of German money demand in the postunification period are reviewed briefly and the results of the current investigation are discussed. Finally, some conclusions are drawn from the analysis. Details of the empirical investigation and estimation results and a description of the data used are given in the appendixes.

The Rise and Fall of Monetary

Targeting in the G-7 Countries

Since the end of World War II, monetary policy in the G-7 countries has been devoted to the pursuit of two broad goals - economic growth and price stability.(3) Progress toward achieving these goals has been measured with intermediate targets - objectives for financial variables such as monetary aggregates or interest rates.(4) A good intermediate target variable is one that is readily observable, can be controlled by monetary authorities, and has a predictable link to one or more policy goals. Such ideal variables are uncommon, however; some meet one or two of the criteria but not all three. For example, interest rates are readily observable, but monetary authorities can influence them only indirectly, as private sector decisions ultimately determine their behavior.

In the absence of ideal intermediate target variables, monetary authorities have also formulated operating targets for variables that they can control more directly and that are closely related to their intermediate targets. In the United States, one such operating target has been the federal funds rate (the rate charged by a depository institution on an overnight sale of federal funds to another depository institution). Since the early 1970s the Federal Reserve has often used as its operating target a narrow range for the federal funds rate.

Often, the type of variable a central bank chooses as its operating target coincides with the variable chosen as an intermediate target. For example, if a central bank is aiming at a particular level of three-month or one-year interest rates as its intermediate target, it would likely use a shorter-term rate (such as the federal funds rate) as its operating target. If, on the other hand, a central bank has chosen a particular monetary aggregate as its intermediate target, it would more likely adopt a closely related operating target such as the amount of reserves in the banking system.

The choice between using the level of interest rates or a monetary aggregate as an intermediate target depends crucially on the relation of these variables to ultimate monetary policy goals. Economic theory suggests that if the nonfinancial side of the economy is stable, then interest rate targets are the most effective way to achieve economic stability, whereas if the financial side of the economy is more stable, targeting the stock of money is better. The oil shocks of the 1970s destabilized the real economy in all the G-7 nations, perhaps increasing the attractiveness of targeting monetary growth. In addition, the growth of the money stock was considered a good alternative to short-term rates as an intermediate target because it was published weekly and was thought to have a stable long-run relation to inflation. Eventually, however, the process of financial innovation and deregulation in some G-7 countries made the supply of money less controllable by monetary authorities, rendering it less useful as an intermediate target.

Roots in the 1960s

Throughout the 1950s and most of the 1960s, the industrial world experienced rapid growth with low inflation. Under the Bretton Woods system of fixed exchange rates, instituted in the mid-1940s, the central banks of the major industrial countries were limited in their ability to pursue independent monetary policies. Occasionally, they were obliged to intervene in the foreign exchange market to maintain exchange rate parities, directly affecting their reserves and the supply of money.(5) When exchange markets were tranquil, central banks attempted to influence liquidity in the banking system; some regulated interest rates, some imposed controls on the growth of particular credit aggregates such as bank loans, and some did both.

The practice of targeting rates of growth of key monetary aggregates had its roots in the late 1960s when monetary authorities in several major industrial countries reduced their emphasis on short-term interest rates and began to pay more attention to monetary aggregates.(6) Around that time, economic conditions raised awareness of the influence of the money stock on economic activity over the short to medium run and on prices over the longer run. The U.S. credit crunches in 1966 and 1969 and the German recession in 1966-67, for example, were seen as having been caused in part by monetary contractions that, at least in retrospect, seemed excessive. And the U.S. fiscal expansion associated with the Vietnam war, along with an increase in the pace of money growth in 1967 and 1968, clearly precipitated a substantial rise in U.S. inflation in the late 1960s, which was transmitted to the other G-7 countries through the fixed exchange rate system. From 1970 to 1973, frequent exchange market interventions during the terminal stages of the Bretton Woods system led to a worldwide expansion in liquidity and a rise in inflation in all the G-7 countries in 1972-73 (chart 1).

[GRAPHICAL DATA OMITTED]

Announcement of Targets in the 1970s

The orientation of monetary policy in the G-7 countries changed significantly during the 1970s. The breakdown of the Bretton Woods system gave monetary authorities greater latitude to pursue independent policies. In the early 1970s, rising inflation pressures, together with the move to flexible exchange rates, led some G-7 central banks to set objectives for the pace of monetary expansion. For instance, the Federal Reserve adopted monetary targets in 1970 with the intention of using them to gradually reduce inflation. As a result, the Federal Open Market Committee (FOMC) began to set targets for M1 and M2 growth over the period between its bimonthly meetings.(7) These objectives were initially used to guide internal policy discussions and were not released publicly.

The rise in inflation in the early 1970s was exacerbated by the outbreak of the Arab-Israeli war in autumn 1973 and the OPEC oil embargo, which caused the international price of oil to quadruple in January 1974. The negative supply shock associated with the rise of oil prices in 1973-74 contributed to recessions in the G-7 nations, which were worsened by the monetary contraction that accompanied it. High levels of inflation, interest rates, and unemployment prevailed in the G-7 countries by the mid-1970s and led to growing public concern about monetary policy.

It was in this atmosphere of stagflation that monetary authorities took a new approach to implementing monetary policy: publicly announcing targets for the growth of the supply of money. Monetary authorities in most G-7 countries adopted explicit targets for the growth of one or more money or credit aggregates in the mid-1970s, and during the second half of the 1970s these targets became a major focus of monetary policy in Canada, Germany, the United Kingdom, and the United States. By the end of the decade, all the G-7 countries except Japan and Italy had official objectives for the rate of monetary expansion.(8) The experience of the G-7 countries with announced monetary targets or (in the case of Japan) projections is summarized in table 1.(9)

[TABULAR DATA OMITTED]

Experience with Targets

In the early years of monetary targeting - the second half of the 1970s - the targets often were exceeded; only Canada was able to meet its targets consistently. At the same time, monetary authorities were trying to reduce the variability of interest rates and exchange rates, actions that were sometimes at odds with meeting the monetary targets. Nevertheless; rates of monetary growth eventually slowed somewhat, and target ranges were reduced. Then global oil prices more than doubled during the course of 1979, contributing to a worldwide rise in inflation that peaked near the turn of the decade.

The appointment of Paul Volcker as Chairman of the Federal Reserve Board in August 1979 marked a turning point in the implementation of monetary targeting. In October 1979, the Federal Reserve announced a major change in its conduct of monetary policy - adoption of an operating target for nonborrowed reserves to improve its control over M1. Interest rates shot up, and M1 growth declined from an annual rate of 8 1/2 percent over the first three quarters of 1979 to 5 1/2 percent during the four quarters of 1981. The change in operating procedure was largely successful at reducing inflation, which peaked at around 13 percent in early 1980 and fell below 5 percent during 1982. The shift toward tighter monetary policy and lower rates of inflation eventually occurred in all G-7 countries over the course of the 1980s and generally has continued in the 1990s.

Decreased Emphasis in the 1980s

Financial liberalization and innovation over the 1970s and 1980s changed the relationship between money stocks and other economic variables in most G-7 countries. Structural changes in these countries' banking systems altered the role of banks in the monetary transmission mechanism, and technological changes, particularly the advance of computer technology, fundamentally revolutionized financial markets and increased the speed with which financial transactions could occur.

In many countries, previously stable money demand relationships broke down during the 1980s.(10) Financial innovation and deregulation altered the elasticity of money demand with respect to income and to interest rates, making monetary aggregates less reliable as intermediate targets. For example, by late 1981 the velocity of M1 in the United States had begun to deviate significantly from its trend and the linkages between money, income, interest rates, and prices had become less stable and predictable. Since the summer of 1982 the monetary aggregates have been deemphasized in the United States, although the Federal Reserve continues to announce annual growth ranges.

In some countries, monetary aggregates have been deemphasized through a change in targeting procedure. For example, as the demand for the broad U.K. aggregate M3[pounds] became less stable, the British government shifted to the much narrower aggregate M0. Canada abandoned monetary targeting altogether in 1982. By the end of the 1980s, the problems with monetary targeting were widely recognized and most G-7 nations had ceased to rely on monetary targets as a consistent guide for policy.

During the 1980s, monetary policy in some countries became increasingly oriented toward external objectives, in particular, toward maintaining a competitive or stable value of the nation's currency. Countries that participated in the exchange rate mechanism (ERM) of the European Monetary System progressively directed their monetary policy actions toward maintaining exchange rate parities. At times, monetary policy in the United Kingdom and Canada was also aimed at exchange rate objectives.

In the second half of the 1980s and in the early 1990s, some G-7 central banks adopted official targets for the rate of inflation, making the ultimate goal of monetary policy a target variable. The Bank of Canada and, more recently, the Bank of England have taken this approach.

History of Monetary Targeting

in Germany

The Deutsche Bundesbank has used monetary targeting to guide policy for twenty years.(11) The Bundesbank Council usually publicizes its target range for monetary growth during the subsequent year in December and reviews the target in July. In setting the target, the Council takes into account three factors: an estimate of the growth of potential output for Germany, a maximum tolerable rate of price inflation, and an adjustment for the downward trend in velocity.(12) Recent targets have been formulated assuming growth of potential output of 2 1/2 percent or 2 3/4 percent, inflation of 2 percent, and a decline in M3 velocity of 1 percent.(13)

The Early Years

In December 1974, the Bundesbank became the first central bank to announce a monetary target.(14) The target was 8 percent growth of the stock of central bank money between December 1974 and December 1975.(15) Germany's experience with monetary targeting since that first target was announced is summarized in table 2. For the next three years, the target was left at 8 percent growth of the stock of central bank money, but the measure was changed to annual average growth. The target was missed in all four early years, making the link between the stock of central bank money and monetary policy actions unclear.

In 1979, the target was changed in two ways. First, the measure was changed once again, from annual average growth of central bank money to growth from fourth quarter to fourth quarter. Second, the target was specified as a range of acceptable outcomes rather than a single rate. The latter change was intended to permit some scope for policy maneuvering, understanding that the prediction of money growth over a short period is inherently imprecise. Following these changes, until 1986, the growth of the stock of central bank money was consistently within its target range.(16)

In 1986, the growth of central bank money was more than 2 percentage points above the upper limit of its target range. Even so, the Bundesbank did not raise official interest rates during the year. A number of factors likely contributed to the Bundesbank's reluctance to raise interest rates. The deutsche mark increased in value 35 percent against the U.S. dollar over the year, making German exports less competitive in international markets and, along with a decline in world oil prices, putting considerable downward pressure on domestic prices. As a result, consumer prices in West Germany (as measured on an annual average basis) declined. Falling prices, combined with a flat yield curve, encouraged the holding of short-term assets included in the stock of central bank money, boosting the growth of the aggregate. Raising interest rates under such circumstances would have risked further appreciation of the mark and the consequent negative effects on growth and prices. In 1987, the growth of the central bank money stock again substantially exceeded the upper limit of its target range, owing primarily to very rapid expansion of bank notes and coins.

The Switch to M3

The wayward behavior of central bank money in 1986 and 1987 led the Bundesbank to switch its focus to the monetary aggregate M3 and to adopt it as a formal target in 1988. At the time, the Bundesbank cited the rapid growth of currency and its relatively heavy weight in the central bank money stock, along with the similarity between the behavior of the central bank money stock and M3, as the reasons for its move to *e broader aggregate.(17) M3 growth slightly exceeded the upper limit of its target range in 1988, as the lagged effects of interest rate cuts executed in 1987 continued to boost money demand and also because more deutsche mark notes were being demanded abroad. Also, in anticipation of a withholding tax on capital income announced in the second half of 1988 and scheduled to take effect at the beginning of 1989, many German investors had begun to liquidate their longer-term deposits not included in the M3 aggregate and to move into cash.(18)

For 1989, the Bundesbank announced a target for M3 growth of "about 5 percent," the single figure intended to imply somewhat more flexibility than that in earlier years because the figure was at the upper end of the range that would have been set on the basis of the standard criteria (that is, 3 percent to 5 percent). M3 growth slowed considerably in 1989 in response to the Bundesbank's tightening of monetary conditions and also as a result of the rescinding of the withholding tax on capital income at midyear.(19) M3 expanded 4.6 percent between the fourth quarter of 1988 and the fourth quarter of 1989, and thus the target was achieved.

Unification

For 1990, the Bundesbank once again specified a target range, to reflect the uncertainties associated with the monetary and economic union of West and East Germany and to reaffirm to the public that unification had not changed the strategy or ultimate goal of monetary policy. The readoption of a target range may also have been motivated by a desire to harmonize monetary targets across European countries in the first stage of European monetary union, which began July 1, 1990. German monetary and economic unification in mid-1990, in which residents of the former East Germany traded in their ostmarks for deutsche marks, led to an increase in the level of M3 on the order of 16 percent (the shift is apparent in table 3 but has been adjusted for in the figures shown in table 2). Still, M3 expanded within its target range for the year.

[TABULAR DATA OMITTED]

In 1990, the contribution of eastern Germany to total German output was roughly 7 1/2 percent, substantially smaller than the 16 percent share of M3 held by eastern residents.(20) Sluggish growth of M3 through mid-l991 was thought to reflect portfolio shifts by eastern German residents out of the shorter-term assets included in M3 and into longerterm assets. When M3 growth picked up toward the end of 1991, the Bundesbank cited as factors underlying its acceleration the increased use of deutsche marks in the countries of eastern Europe and the republics of the former Soviet Union and the brisk pace of credit expansion (about one-fourth to onethird of which was thought to represent subsidized lending in eastern Germany) (table 3).

ERM Turmoil and the Withholding Tax

Despite the broad changes in the German economy at unification, monetary targeting was largely on track until 1992, when various "special factors"

resulted in growth of 9.5 percent (fourth quarter over fourth quarter), exceeding the upper limit of the Bundesbank's target range by 4 percentage points. One element of M3's brisk expansion in 1992 was the continued rapid growth of currency (table 3): Continued high demand in eastern European countries and the former Soviet republics boosted currency growth throughout the year. Other, more important special factors were the September 1992 crisis in the ERM and the massive deutsche mark intervention associated with it,(21) as well as the reintroduction of a withholding tax on interest income, effective January 1, 1993, which initially induced Germans to trade securities for cash.(22) Another special factor in 1992 was the fall of long-term interest rates to levels below shortterm rates, which led to shifts from longer-term assets excluded from German M3 to time deposits having maturity of less than four years (which are included in M3).

[TABULAR DATA OMITTED]

The growth of M3 slowed significantly in early 1993 as the shifts into currency associated with the imposition of the withholding tax began to unwind (as currency moved out of Germany), but the aggregate began to grow rapidly again in early summer, and the growth rate ended the year about 1 percentage point above the 6 1/2 percent upper limit of the target range. The particular special factors that influenced the behavior of M3 in 1993 are less clear than those present in 1992. The Bundesbank again cited the greater use of the mark as a parallel currency outside Germany, pointed to the effects of the massive intervention in foreign exchange markets during another episode of ERM turmoil in late July, and mentioned the continued inversion of the yield curve.

Though M3 growth exceeded its target in 1993, it was the excessive growth during the first half of 1994 that caused intensified concern about the ultimate viability of M3 targeting for Germany (chart 2). The Bundesbank attributed M3's surge to a number of special factors, both economic and statistical. One was the anticipation of a change in the law regarding the withholding tax on interest income that was to take effect on January 1, 1994, which induced a reflow of investment funds from Luxembourg to Germany in December 1993.(23) At the end of December 1993 the level of M3 was already above the average level for the fourth quarter (the base period for the target) by some DM 64 billion (or 3.6 percent), adding a "monetary overhang" to the aggregate that was carried over into 1994. Another special factor cited by the Bundesbank was the expiration of tax incentives for mortgage borrowing, which boosted M3 around the turn of the year.

A steep rise in long-term interest rates in early 1994 also contributed to the rapid growth of M3 early in the year. Uncertainty about whether these movements were likely to be temporary or permanent led to a buildup of liquidity and slowed the flow of the repatriated investment funds into longer-term assets not included in M3. Eventually, the speculative inflows into time deposits were reversed, and by the end of the year M3 had fallen back into its target range.

The Introduction of Money Market Funds

Another factor influencing M3 growth over 1994 was legislation effective August 1, 1994, that encouraged the development of money market funds in Germany. Although money market funds had already been permitted in principle, they had been allowed to invest no more than 49 percent of total assets in short-term paper or bank deposits, a restriction that effectively prevented their development; the new legislation allowed the funds to invest exclusively in short-term instruments.(24) From August through December 1994, DM 46 billion flowed into money market funds, helping the Bundesbank meet its monetary target last year.(25) Had M3 been redefined to include these funds, growth would have been 6.3 percent, slightly above the upper limit of the target range but still a considerable slowdown from the pace in 1993.(26)

In December 1994, the Bundesbank reaffirrned that money market funds would not be included in the M3 aggregate but said that they would be included in the next broadest monetary aggregate, extended M3. The Bundesbank also announced its intention to monitor extended M3 and the development of money market funds.

Developments in 1995

The contraction of M3 through mid-1995 has once again raised questions about the appropriateness of monetary targeting as a guide for Bundesbank policy. Despite a leveling off of funds flowing into money market accounts, M3 has remained weak, with growth mainly in longer-term assets not included in the monetary aggregate ("monetary capital formation," in Bundesbank parlance). In July, M3 growth returned to negative territory (with June posting the only positive figure so far this year), and the level of M3 remained well below that in the fourth quarter of last year. In its midyear review of the monetary target, the Bundesbank reaffirmed the 4 percent to 6 percent target range, stating: "[T]he judgment that the long-term demand for money has remained stable, and that the current weakness of monetary growth is largely a reaction to previous dislocations, warrants neither an abandonment nor a revision of this year's monetary target. However, since the monetary target is being considerably undershot at present, it seems rather unlikely that the target will be met at the end of the year." On August 24, the Bundesbank reduced official interest rates, stating that "[t]he main reason for this interest rate cut was the persistently weak trend in the money stock."

Stability of German Money Demand

The viability of M3 as an intermediate target for monetary policy rests on the predictability of the macroeconomic consequences of a change in the supply of M3. If a stable money demand function exists in Germany, then a policy-driven change in M3 should have predictable influences on GDP, interest rates, and, ultimately, prices.

A number of recent studies have examined the stability of German money demand since the unification of West and East in 1990. Though the studies have differed in some ways, including the number of data observations since unification, the number and definition of explanatory variables, and the seasonal adjustment of the data, they have generally taken the same approach: specification and estimation of an equation that explains the demand for German M3 in terms of income and a measure of the opportunity cost of holding M3. Some studies analyzed the demand for nominal M3, whereas others focused on real M3 (the nominal stock of M3 deflated by some measure of prices).

The common formulation of the money demand function, used in most previous studies and in the work reported here, is a dynamic error-correction representation in which the long-run, or equilibrium, relationship between M3 and its determinants is embedded in an equation that captures short-run variation and dynamics.(27) Stability of M3 since German unification is essentially a question about the stability of the long-run relationship within the dynamic error-correction representation. This section focuses on the estimated long-run relationship; details of model testing and empirical results are given in appendix A.

The general approach in previous work, and the approach taken in the current study, is as follows: The M3 equation is initially estimated over the pre-unification period and statistical tests are performed to ensure that the estimated coefficients are stable. Then the equation's estimation range is extended to include data since 1990, with one or more dummy variables included to account for shifts due to unification. (An impulse dummy is used to detect shifts in a single quarter, whereas a shift dummy is used to detect a permanent change. A statistically significant impulse dummy would not be evidence of a break in the money demand relation, but a statistically significant shift dummy would be.) Stability is examined by comparing the pre-unification results with results for the extended period.

In general, the results of earlier studies have been mixed (table 4). Von Hagen and Hansen and Kim rejected the stability of M3 demand, whereas Issing and Todter and the Bundesbank found little evidence that the long-run relation between money demand and its key determinants changed after German unification. Gerlach, using a money demand relation estimated through 1990, forecast German M3 through 1992 and, based on the accuracy of the forecast, concluded that the demand for M3 had remained stable. The OECD concluded that its findings lend tentative support to stability despite a small but statistically significant estimate for the shift dummy variable.

[TABULAR DATA OMITTED]

As in some earlier studies, our equations for estimating the long-run money demand relationship explained real M3 in terms of real GDP and an interest rate to proxy for the opportunity cost of holding money (table 5). In contrast to most earlier studies, however, we included real net financial wealth in the equation and estimated it using data that had not been adjusted for seasonal variation. We constructed a measure of real net financial wealth by extrapolating biannual data on flows and stocks of financial assets held by households. The wealth variable was not significant, perhaps indicating statistical problems with our measurement of real wealth. However, inclusion of real net wealth substantially reduced the estimated responsiveness of money demand to real income, producing a more plausible result (most earlier studies reported an income elasticity well in excess of unity, a finding that does not accord with economic theory). The inclusion of wealth also reduced the estimated semi-elasticity with respect to the opportunity cost variable, which some earlier studies of German money demand found to be relatively high.

5.  Equations for estimating equilibrium money  demand
used current study

For period before German unification:

  Real M3 = 0.58 x real income
          - 1.76 x opportunity cost
          + 0.41 x real net walth
          + constant

For entire period:
  Real M3 = 0.75 x real income
          - 1.83 x opportunity cost
          + 0.34 x real net wealth
          + constant

The use of data that were unadjusted for seasonal variation allowed us to estimate the seasonal effects directly and to test whether the seasonal pattern shifted at unification. It also lengthened the postunification sample period by two observations.

Although our approach differed somewhat from those of earlier studies, our findings are similar to the earlier results regarding the effects of German unification on the long-run demand for real M3. That is, the equilibrium money demand function appears to have been largely stable since unification despite some changes in short-run dynamics and a level shift in 1990 (table 6). Our long-run estimates do, however, indicate some increase in the sensitivity of real money demand to real income and some decline with respect to changes in real net wealth. The shift in the income and wealth elasticities could be signaling a shift in the long-run relationship. However, at the time of unification, residents of eastern Germany were holding most of their financial wealth in the form of M3, and some additional diversification out of M3 assets is likely. Thus, it is probably still too early in post-unification period to ascertain conclusively whether the long-run demand for German M3 has remained stable.

[TABULAR DATA OMITTED]

To assess the accuracy of our model, we examined the prediction errors for the level of real M3 (chart 3).(28) Despite the apparent stability of long-run money demand, prediction errors and the variability of those errors have become noticeably larger since unification. Recently, large prediction errors have tended to coincide with periods when special factors, such as tax or regulatory changes, were present.

Conclusions

Empirical evidence from the study reported here as 11 as from earlier studies indicates that the demand for German M3 has remained largely stable over the past two and a half decades despite the unification of East and West Germany in 1990. Thus, the Bundesbank's continued use of M3 as its primary intermediate target in pursuit of price stability - while other central banks have de-emphasized monetary aggregates as a guide to policy - seems justified.

With the many changes worldwide in financial markets and economic relations over the past two decades, the stability of the long-run relation of the stock of money to income and prices in Germany is perhaps surprising. The relative stability may be due in part to the historical pattern of financial change in Germany. In contrast to that in other G-7 countries, domestic financial liberalization was well advanced in Germany by 1970 and has progressed relatively slowly since then. Foreign exchange controls were removed in the 1950s, and controls on interest rates were abolished in 1967 - in line with a longstanding tradition of liberal banking and capital market policies.

Whereas financial change in the form of financial liberalization came early in Germany, financial change in the form of the development of new financial instruments - a source of instability in money demand in many other major industrial countries - has been relatively slow to evolve. For example, the German short-term money market is quite limited; instruments similar to U.S. three-month Treasury bills do not exist, and the public has little access to interbank money markets.

Several factors underlie the slower pace of financial innovation in Germany compared with that in other G-7 countries. In several G-7 countries, controls on exchange rates and interest rates in the 1970s and early 1980s prompted the creation of new financial instruments designed to circumvent these controls. In Germany, the early removal of such controls, combined with the relatively low rate of inflation, may have reduced the need for financial innovation. Germany' s proximity to Luxembourg, a financial center that offers a full array of financial instruments and that has often served as an investment outlet for German asset holders, may also have played a role.

The relative stability of German money demand resulting from the slower pace of financial change over the past few decades has been a key ingredient in the Bundesbank's success in using monetary targeting to achieve low inflation. Another key ingredient has been the central bank's flexibility and credibility. Whenever the monetary target has become less useful because of exchange-rate or other considerations, the Bundesbank has allowed M3 to miss its target temporarily and has suffered little loss of credibility. Another important factor has been the popular consensus against inflation, which is probably stronger in Germany than in most other countries.

The future for monetary targeting in Germany is not clear. Further financial changes undoubtedly will come, raising the potential for instability in the relation between money and prices. As Europe moves toward establishing a European Central Bank, issues regarding the best intermediate targets and operating procedures are being actively discussed at the European Monetary Institute.(29) The Bundesbank's experience with monetary targeting undoubtedly will influence the future course of European monetary policy.

Appendix A:

Details of Estimation and Results

This appendix discusses our general empirical approach. The estimated long-run (or equilibrium) money demand relationship was presented earlier and is not repeated. Here we lay out the general specification in which the long-run relationship is embedded and discuss the estimates for the short-run dynamics of the money demand function.

We began our analysis with a very general specification of a dynamic error-correction formulation of the demand for M3 balances:

(1) A(L)[delta](m - p) = [alpha.sub.0] + [alpha.sub.1](L)[delta]y + [alpa.sub.2](L)[delta]p

+ [alpha.sub.3](L)[delta]i + [alpha.sub.4]y-1

+ [alpha.sub.5i]-1 + [alpha.sub.6](w - p)-1

+ [alpha.sub.7(m - p)-1 + [alpha.sub.8]SQ1

+ [alpha.sub.9]SQ2 + [alpha.sub.10]SQ3,

where A(L) and [alpha](L) are polynomials in the lag operator, L; [delta] denotes the difference operator; (m - p) is real M3 balances; y is real GDP; p is the GDP deflator; i is the interest rate on public bonds (a proxy for the opportunity cost of holding M3 balances); (w - p) is real net financial wealth of households; and SQ1, SQ2, and SQ3 are seasonal dummies that take the value of 1 in the quarter indicated and 0 elsewhere. The lower case variables (other than i) are natural logarithms.

Our initial specification included four lags of the differenced variables (the variables preceded by [alpha](L)[delta] in eq. 1). The equation was subsequently pared down because the estimated parameters proved to be statistically insignificant. (At each stage of the reduction process, we also used the Schwarz information criterion to judge the information content of each equation and checked for serial correlation of the residuals.) The parameters of the dynamic money demand equation estimated over the pre-unification period are shown in table 6.

We were unable to find many explanatory variables in differenced form that entered the equation describing the change in real money balances. In particular, we never found that the addition of the growth rate of real GDP or its four lagged values improved the fit of the equations.

The estimated coefficient on the contemporaneous change in the yield on public bonds was positive and significant at the 5 percent level, but it was smaller in magnitude than the coefficient on the lagged change in interest rates, which was negative, as expected. In the short run, the change in interest rates may proxy for changes in short-term or own interest rates (that is, the rate paid on assets included in the M3 aggregate), which would tend to boost the demand for real M3 balances.

The parameters of our relationship proved to be sensitive to the particular interest rate used, as other authors have noted. The own rate was consistently insignificant in estimation for a number of different definitions. We constructed average own rates and marginal own rates from available data on the yields on savings deposits and time deposits (the interest-bearing components of German M3). However, the data did not allow any of these measures of the own rate to be constructed without some approximation, and the process of approximation may have introduced enough measurement error to make the variable insignificant in our regressions.(30)

The estimated coefficient on [delta]p (differenced prices) is significantly negative, indicating that, within a quarter, the demand for M3 is not homogeneous with respect to prices: A change of 1 percentage point in the GDP deflator erodes real balances by almost 3/4 percentage point. To some extent, this result may be related to the different periodicities of money and prices, as M3 data are end-of-quarter figures whereas the GDP deflator is a quarterly average.(31)

We performed several diagnostic procedures to test for the presence of autocorrelation, autoregressive conditional heteroskedasticity (ARCH), and normality of the residuals and found no evidence of any problems.(32) However, we did find evidence of some conventional heteroskedasticity, owing in part to the inclusion of dummy variables to detect shifts at unification.

Table 6 also presents the estimation results over the extended sample period. Initially, we included several dummy variables to capture possible effects of changes in definitions of data and in economic behavior due to unification. When estimated through the fourth quarter of 1994, the results gave strong support for impulse dummies in the second and third quarters of 1990 (coinciding with breaks in the M3 and GDP series respectively), but a permanent shift dummy variable in mid-1990 was not statistically significant. The coefficients indicate a rise in real M3 balances of about 14 percent in the second quarter of 1990, followed by a further 3 1/4 percent boost in the third quarter. In addition, there is some evidence that the seasonal pattern of money demand changed at unification, particularly in the first quarter.

Appendix B: Data

All observations used in the study were quarterly log levels of not seasonally adjusted data.

Money. End-of-period levels of M3. Data for West Germany only through 1990:Q1; unified Germany thereafter. (The targeted aggregate M3 is defined on the basis of seasonally adjusted, period average data.) Source: Bundesbank.

Real GDP. Data for West Germany only through 1990:Q2; unified Germany thereafter. Sources: Bundesbank (West Germany and western states) and Federal Statistical Office and the Deutsches Institut fur Wirtschaftsforschung (eastern states).

GDP Deflator. Constructed from real and nominal GDP. Data for West Germany only through 1990:Q2; unified Germany thereafter.

Interest Rate. Average yield in the secondary market on all public sector bonds. Source: Bundesbank.

Wealth. Constructed and interpolated from biannual flows of net financial wealth of households. Data for West Germany only through 1990:Q2; unified Germany thereafter. Source: Bundesbank. (1.) German M3 consists of notes and coins in circulation plus German residents' holdings of sight deposits at banks in Germany, time deposits having original maturities of one month to less than four years held at banks in Germany, and savings deposits having a period of notice of three months or less. (2.) Deutsche Bundesbank, Monthly Report (January 1995), p. 23. (3.) Technically, the term "G-7" to refer to the seven largest industrial countries was not used before the mid-1980s. Other major industrial countries, notably Switzerland, also used annual monetary targets in the 1970s and 1980s. (4.) For a discussion of intermediate targets and the goals of U.S. monetary policy, see R. Glenn Hubbard, Money, the Financial System, and the Economy (Reading, Mass.: Addison-Wesley, 1994), chap. 21. (5.) Under the Bretton Woods system, the U.S. dollar was the nth currency, in terms of which the (n - 1) other currencies were defined. In theory, the United States was responsible for fixing the price of gold at $35 per ounce. In practice, however, the United States intervened only rarely in the market for gold or foreign exchange, letting the central banks of other countries intervene to fix their exchange rates against the dollar. (6.) For a discussion of the economic events leading up to the implementation of monetary targets, as well as early experience with targets, see Organisation for Economic Co-operation and Development, Monetary Targets and Inflation Control (Paris: OECD, 1979). Also see Brian Griffiths and G.E. Wood, eds., Monetary targets (New York: St. Martin's Press, 1981), introduction and chap. 1. (7.) See Ann-Marie Meulendyke, US. Monetary Policy and Financial Markets (New York: Federal Reserve Bank of New York, 1989), pp. 38 47, for a discussion of U.S. monetary policy in the 1970s and early 1980s. (8.) From 1974 to 1980, Italy did, however, set targets for total domestic credit expansion. (9.) See Karen H. Johnson, "Foreign Experience with Targets for Money Growth," Federal Reserve Bulletin, vol. 69 (October 1983), pp. 746 54, for a comprehensive review of early experience with monetary targets abroad. Also see Paul Meek, ed., Central Bank Views on Monetary Targeting (Federal Reserve Bank of New York, 1983) and Michele U. Fratianni and Dominick Salvatore, eds., Monetary Policy in Developed Economies (New York: Greenwood Press, 1993). (10.) One study that considered the stability of money demand in several countries was Ray C. Fair, "International Evidence on the Demand for Money," Review of Economics and Statistics, vol. 69 (1987), pp. 473-80. Fair concluded that the demand for money had become unstable in most major countries, the only exception being West Germany. (11.) This section relies heavily on material published by the Deutsche Bundesbank in its Monthly Report (various issues, 1975-95). Also see Norbert Kloten, "The Control of Monetary Aggregates in the Federal Republic of Germany under Changing Conditions, "in S.F. Frowen and D. Kath, eds., Monetary Policy and Financial Innovations in Five Industrial Countries: The UK. the USA, West Germany, France, and Japan (New York: St. Martin's Press, 1992), pp. 32-58. (12.) Velocity is the rate at which the stock of money turns over in an economy and defines exactly the relation between the nominal money supply and nominal income. The velocity of German M3 has trended down over time. (13.) The Bundesbank's figure for "acceptable" inflation - typically 2 percent - is based on a goal of stable prices taking into account potential errors in measurement. (14.) Bundesbank Council member Otmar Issing provided an interesting discussion of Germany's experience with monetary targeting in a paper delivered at the conference on "Monetary Policy in an Integrated World Economy," Kiel, Germany, June 1995. (15.) Central bank money is defined as total currency in circulation plus (1) sight deposits multiplied by a factor of 0.166, (2) time deposits having maturity of less than four years multiplied by a factor of 0.124, and (3) savings deposits having maturity of less than four years multiplied by a factor of 0.081. (The factors represent the reserve ratios effective in January 1974.) (16.) Technically, the target was undershot slightly in 1980 and 1981. However, if actual growth rates are rounded up to the nearest half percentage point, as was the Bundesbank's practice, the targets were met. See Deutsche Bundesbank, The Monetary Policy of the Bundesbank (Frankfurt: Deutsche Bundesbank, March 1994), p. 118. (17.) The Bundesbank stated: "[I]n the past two years the central bank money stock has tended to exaggerate monetary growth because of the large share of currency in it (and the exceptionally fast growth of this component). Over the long term, the central bank money stock and the money stock M3 have moved largely parallel to one another. They are therefore of comparable value as guidelines for a monetary policy oriented to non-inflationary economic growth over the medium term" (Monthly Report of the Deutsche Bundesbank, February 1988, p. 9). (18.) The withholding tax (known as the "minor investment income tax") of 10 percent was withheld at the source. (19.) The withholding tax was abolished on July 1. Because the tax had applied to foreign as well as domestic investors, foreign asset holders had fled German capital markets - an unintended consequence of the tax. (20.) That eastern Germans held 16 percent of M3 balances while producing only 7 1/2 percent of all German output reflects the exchange rate used to convert ostmarks to deutsche marks at the time of unification and the greater fraction of wealth held as money balances (due to the underdevelopment of financial markets in the east). After unification, economic activity in the eastern states fell dramatically, bottoming out in 1991, when output was only 7 percent of all German GDP. Output in the eastern states has recovered steadily since and at the end of 1994 accounted for 9 percent of all German real GDP.

(21.) During the ERM crisis in September 1992, intervention by the Bundesbank totaled more than DM 92 billion ($63 billion) as authorities attempted to keep ERM currencies from rising above their upper target limit relative to the deutsche mark. The Bundesbank did not sterilize this intervention immediately; in its view, the available monetary instruments (weekly repurchase operations) were not sufficient to withdraw such a large volume of liquidity from the markets at one time. Thus, the sterilization took place gradually, causing M3 to swell during the interim period. (Had the Bundesbank chosen to withdraw such a volume of liquidity rapidly, the resulting spike in interest rates would have jeopardized an already weak domestic economy and risked further destabilization of the ERM.) The Bundesbank attempted to lessen the difficulties associated with sterilizing the large inflow into M3 by immediately shortening the maturity of repurchase contracts (from predominantly four-week to two-week contracts) and, beginning in early 1993, by issuing short-term paper (known as Bulis) to nonbanks. (22.) In response to a ruling by the Federal Constitutional Court in mid-1991 that declared the prevailing treatment of interest income to be unconstitutional, the withholding tax on interest income was reinstated. However, the withholding tax instituted in 1993 differed in several ways from the one effective in 1989: First, the interest income exemption was raised by a factor of ten, to DM 6,000 per person; second, the rate of taxation was increased from 10 percent to 30 percent (35 percent for over-the-counter transactions); third, foreign holders of German assets were effectively exempt from the tax. Because the new withholding tax applied only to interest on investments with banks located in Germany, residents seeking to avoid the tax traded securities for currency in the late months of 1992. Eventually the cash was transported to Luxembourg (and other locations outside Germany) and reinvested, often in branches of German banks (which were not legally required to withhold tax on interest income). These actions caused a swelling of the currency component of M3 toward the end of 1992 and a subsequent reversal early in 1993 as the funds were reinvested outside Germany. (23.) Effective January 1, 1994, the withholding tax was extended to income accruing to investors in Germany from cumulative foreign investment funds. (24.) As financial markets in other major industrial countries were introducing money market funds, the Bundesbank had resisted their development because it feared that a proliferation of short-term instruments would lead to a loss of monetary control. In September 1994, the Bundesbank stopped issuing Bulis (see note 21), thus reducing the variety of short-term instruments available to money market funds. (25.) More than half the inflow into money market funds - some DM 27 billion - occurred in December and was related to a change in the taxation of assets effective at the end of the year. The change increased the tax rate on deposits while leaving the tax rate on equities (including money market funds) unchanged. (26.) The calculation assumes that domestic money market funds are included in the M3 aggregate. Foreign money market funds (for instance, funds offered by German banks located in Luxembourg) have risen as well, but the M3 aggregate by definition is restricted to holdings of assets extended by banks located in Germany. (27.) See D.F. Hendry and N.R. Ericsson, "Modeling the Demand for Narrow Money in the United Kingdom and the United States," European Economic Review, vol. 35 (1991), pp. 833-86, and J.M. Kremers, N.R. Ericsson, and J.J. Dolado, "The Power of Cointegration Tests," Oxford Bulletin of Economics and Statistics, vol. 54 (1992), pp. 325-48, for discussions of this approach. (28.) As the dynamic error-correction representation is based on changes in holdings of real M3, we cumulated the predicted changes from the initial level to obtain constructed levels. (29.) Issues of currency substitution and stability of European money demand in Europe are discussed in Deutsche Bundesbank, "Demand for Money and Currency Substitution in Europe," Monthly Report (January 1995), pp. 33 49. (30.) In his estimated equations for the pre-unification period, Von Hagen omitted the interest rates on savings deposits and time deposits included in M3 on the grounds that the rate on savings deposits did not vary enough during the sample period and the rate on time deposits was not available for a long enough period of time. See Jurgen Von Hagen, "Monetary Union, Money Demand, and Money Supply: A Review of the German Monetary Union," European Economic Review, vol. 37 (1993), pp. 803-36. (31.) We used end-of-quarter M3 data to reduce the possibility of any within-quarter feedback from money to prices. (32.) ARCH is a special type of heteroskedasticity in which the variance of the residuals exhibits an autoregressive pattern.