The overshooting of the broad money target in recent years begs the question as to whether there is a serious risk of a pickup in inflation down the road. This chapter tries to shed some light on this issue by providing an empirical analysis of different monetary indicators. It finds that both money growth and a monetary conditions index (MCI), constructed as a weighted average of real short-term interest rates and the real effective exchange rate, have in the past provided useful early warning signals for inflation. Nevertheless, neither indicator would have predicted developments in inflation around the turn of the decade very well—perhaps not a surprising result, given the unprecedented nature of the demand shock to the west German economy associated with unification. During 1992-93, when money growth increased significantly, the MCI indicated that monetary conditions remained fairly tight. Since there were good reasons to suspect that the monetary data were distorted by special factors, the behavior of the MCI at this time suggests that the policy of cutting official interest rates during 1993 and the first half of 1994, even in the face of further surges in monetary growth, was appropriate. The validity of this policy is also borne out by the subsequent slowdown in monetary growth in the second half of 1994 and the favorable evolution of wage and price inflation.

## Background

The conduct of monetary policy in Germany is predicated on the view that there is a long-run relationship between money growth and inflation. An explanation for this relationship, which underpins the calculation of the Bundesbank’s annual monetary target, begins with the Quantity Theory identity:

where *M* is the money stock (M3 in the target definition), *V* is money velocity, *P* is the average price level, and *Y* is real income. On the assumption that velocity follows a predictable path and that real income growth is constrained to the potential growth of supply in the long run, it is then possible to translate a specific inflation goal (price stability) into an annual target for monetary growth.^{1} By adhering to the target, monetary conditions would tighten (in the sense that the quantity of money would be lower than normally warranted by the nominal value of transactions in the economy) if actual inflation were higher than the objective or actual output growth were above potential. In this way, while the objective of monetary policy would be price stability, the framework would in principle allow monetary conditions to vary in a countercyclical manner.

From an empirical perspective, this framework assumes that causality runs from money to inflation—although the route of causality may be indirect, through other variables such as incomes, exchange rates, or asset prices—and that velocity movements are predictable (money demand is stable).^{2} Studies tend to confirm the existence of a positive causal link between developments in money and future prices for Germany. In addition, there is evidence—although the consensus is not unchallenged in the literature—that money demand was stable, at least in the period up to unification in mid-1990.^{3} However, the continuation of this stability in recent years is more contentious, with some evidence that, at the very least, there has been a once-off break around unification in the hitherto stable path of velocity. See Box 1, as well as Chart 5-1 and Table 5-1.

**Cointegration Tests: Money, Income, and Interest Rates ^{1}**

^{1}The variables are as follows: *M*, broad money (M3); *P*, the GDP deflator; *Y*, real GDP; *V*, money velocity; *TT*, a time trend; and *R*, three-month money market rates. Variables other than interest rates are in logarithms and, for the period up to unification, refer to west Germany only. Data on velocity (*V*) refer to united Germany, with estimates before unification based on west German data. See Appendix for splicing assumptions. *DF* stands for the Dickey-Fuller statistic and *ADF* its augmented version with four lags. A single asterisk denotes significance (likely cointegration) at the 10 percent level and a double asterisk denotes significance at the 5 percent level.

**Cointegration Tests: Money, Income, and Interest Rates ^{1}**

Test Statistics | ||||
---|---|---|---|---|

DF | ADF | |||

Pre-unification data (1970:1-1989:4) | ||||

Model | ||||

M–P = 1.69Y – 0.32R – 8.07 | 5.32** | 3.02 | ||

M–P = 1.06Y – 0.16R + 0.013TT – 30.92 | 4.30** | 3.38* | ||

V = -0.015TT + 0.14R + 30.00 | 3.97* | 3.35* | ||

V = -0.015TT + 30.42 | 3.75** | 3.67** | ||

Full data set (1970:1-1994:4) | ||||

V = -0.012TT + 25.32 | 2.22 | 3.09* | ||

V = -0.012TT + 0.47R + 25.05 | 2.62 | 2.93 | ||

V = -0.015TT + 30.50 +.073DUM | 3.68* | 4.22** | ||

V = -0.015TT + 0.20R + 29.90 +.066DUM | 3.89* | 3.84** |

^{1}The variables are as follows: *M*, broad money (M3); *P*, the GDP deflator; *Y*, real GDP; *V*, money velocity; *TT*, a time trend; and *R*, three-month money market rates. Variables other than interest rates are in logarithms and, for the period up to unification, refer to west Germany only. Data on velocity (*V*) refer to united Germany, with estimates before unification based on west German data. See Appendix for splicing assumptions. *DF* stands for the Dickey-Fuller statistic and *ADF* its augmented version with four lags. A single asterisk denotes significance (likely cointegration) at the 10 percent level and a double asterisk denotes significance at the 5 percent level.

**Cointegration Tests: Money, Income, and Interest Rates ^{1}**

Test Statistics | ||||
---|---|---|---|---|

DF | ADF | |||

Pre-unification data (1970:1-1989:4) | ||||

Model | ||||

M–P = 1.69Y – 0.32R – 8.07 | 5.32** | 3.02 | ||

M–P = 1.06Y – 0.16R + 0.013TT – 30.92 | 4.30** | 3.38* | ||

V = -0.015TT + 0.14R + 30.00 | 3.97* | 3.35* | ||

V = -0.015TT + 30.42 | 3.75** | 3.67** | ||

Full data set (1970:1-1994:4) | ||||

V = -0.012TT + 25.32 | 2.22 | 3.09* | ||

V = -0.012TT + 0.47R + 25.05 | 2.62 | 2.93 | ||

V = -0.015TT + 30.50 +.073DUM | 3.68* | 4.22** | ||

V = -0.015TT + 0.20R + 29.90 +.066DUM | 3.89* | 3.84** |

^{1}The variables are as follows: *M*, broad money (M3); *P*, the GDP deflator; *Y*, real GDP; *V*, money velocity; *TT*, a time trend; and *R*, three-month money market rates. Variables other than interest rates are in logarithms and, for the period up to unification, refer to west Germany only. Data on velocity (*V*) refer to united Germany, with estimates before unification based on west German data. See Appendix for splicing assumptions. *DF* stands for the Dickey-Fuller statistic and *ADF* its augmented version with four lags. A single asterisk denotes significance (likely cointegration) at the 10 percent level and a double asterisk denotes significance at the 5 percent level.

More generally, even if money demand is stable, practical use of a target framework requires that the starting point for basing the money target can be assessed properly. In effect, a judgment has to be made as to what extent the target growth rate of money should allow for any earlier monetary overhang. As a case in point, the added measurement uncertainties caused by unification make it unclear whether the ensuing increase in monetary growth reflects over-loose monetary conditions or a return of velocity to trend.

**Stability of the Demand for M3**

The stability of the demand for M3 can be tested using the two-stage estimation procedure of Engle and Granger (1987). This procedure tests first for the existence of a stable, long-run statistical relationship among money, income, and interest rates. Such tests confirm that money was cointegrated with money and income in the pre-unification era. Moreover, adding a time trend to the relationship suggests that the income elasticity of real money demand was unity or, put another way, that money velocity fluctuated about a stable trend (Table 5-1). The trend was downward at an estimated rate of about 1 1/2 percent a year, or a little more than the amount the Bundesbank usually factors into its calculation of the monetary target range. Either short- or long-term interest rates, with correctly signed coefficients, can also be legitimately included in the cointegrating relationship.

If the sample period is extended beyond unification, the cointegration results underpinning the existence of a long-run stability in money demand break down unless a sizable one-off increase in velocity is allowed for. The shift in velocity, which is estimated to have occurred in 1990, could reflect a combination of unusual portfolio shifts around the time of unification as well as data measurement errors. The size and permanence of the velocity shift is sensitive to data splicing assumptions and the specification of dummy variables in the regression analysis. For example, analysis in the 1994 *OECD Economic Survey* for Germany assumes that a shift in money demand occurs after 1990, and, partly as a result, the analysis comes to more sanguine conclusions about the underlying stability of money demand. However, developments in velocity suggest that unusual portfolio shifts probably began ahead of unification, which occurred in the middle of 1990.

Nevertheless, it should be stressed that it is too early to conclude whether there has been a permanent break in trend velocity or whether the rapid monetary growth through early 1994 was in the process of reversing an earlier portfolio shift (Chart 5-1, top panel). This issue has, of course, important implications for the interpretation of postunification monetary growth. If the velocity shift was permanent, velocity would have been well below trend and a sizable monetary overhang would have existed by early 1994. But if a reversal of an earlier portfolio shift was taking place (or a new portfolio shift was occurring), the rapid monetary growth through the early part of 1994 would probably pose little concern for medium-term inflation.

In the second stage of the analysis, an error correction equation for money demand is estimated. This equation exhibits parameter instability in the postunification period, even if a shift in trend velocity is factored into the long-run properties of the equation (see below). The equation begins to break down in the second half of 1992 (Chart 5-1, lower panel). This could, in part, reflect the increased variance of monetary growth owing to the large bouts of currency intervention during the two ERM crises. The equation breaks down completely in the first quarter of 1994 as it fails to predict the surge in M3 growth—although special factors, notably changes in the taxation of interest income, may again provide part of the explanation.

Money Demand Equation

where *V* = *PY/M* and *V** = -0.015*TT* + 0.200*R* + 29.897 + 0.066*DUM*.

The variables are *M* for money stock (M3); *P* for the GDP deflator, *Y* for real GDP; *V* for velocity; *V* for three-month money market rates; *TT* for a time trend; and *DUM* for a shift dummy taking the value 1 from 1990:1 onward. The symbol Δ represents a fourth difference (Δ*X* – *X* – *X*-4). The reported test statistics are *DH* for Durbin’s *H*-statistic; *AUTO* for a test for serial correlation of up to lag five; and *FORE* (17) for an out-of-sample parameter stability test for the period 1990:1-1994:1.

In summary, it is difficult to avoid the conclusion that money demand has exhibited instability since unification. However, the nature of the instability and the implications for the interpretation of monetary growth are unlikely to become clear for some time.

The monetary framework assumes nothing specific about the transmission channels from money to prices. In principle, several channels are possible. From the domestic side, the effect might be through interest rates and their impact on output, as in the conventional IS-LM framework. Where a broad money aggregate is considered, more direct wealth effects might also come into play and, in some cases, the impact might be felt first in asset markets.^{4} From the external side, there might be a link between monetary growth and the exchange rate, which has a direct impact on domestic prices via import costs and prices and an indirect effect via changes in trade volumes.

Thus, an alternative to basing the monetary framework on a particular money aggregate would be to gauge monetary conditions from developments in other financial indicators such as interest rates, exchange rates, and asset prices, which may play more direct roles in determining output and inflation. If money demand and other behavioral relationships in the economy are stable, such indicators would convey information about monetary conditions similar to that provided by the monetary aggregates. It is conceivable that if money demand were behaving less predictably, alternative financial indicators could provide more reliable information.

In a number of countries, including Canada. New Zealand, and Sweden, where developments in domestic monetary aggregates have been difficult to interpret, increasing reliance has been placed on financial indicator variables. In the case of Canada, an MCI consisting of a weighted average of interest rates and the effective exchange rate has proved to be a helpful indicator of inflationary pressures.^{5} The weighting in the index reflects the relative partial impacts on inflation or output of each financial indicator over a period of about two years. The rationale for combining the two elements rests upon the fact that the domestic indicator of monetary conditions (interest rates) and the external indicator (exchange rates) can, in practice, convey independent information about monetary conditions. For example, an increase in interest rates that was accompanied by exchange rate appreciation might have a greater effect on output and prices than an interest rate rise alone.

In what follows, a monetary indicator is constructed for Germany. This index is compared with M3 as a predictor of future inflation, and the analysis is used to provide an assessment of recent monetary conditions.

## A First Look at the Data

By international standards, German inflation has been low in the past few decades. Annual increases in the GDP deflator (one of the broadest measures of inflation) reached 8 percent in the early 1970s but then declined to about 4 percent in the second half of that decade (Chart 5-2). After picking up briefly at the turn of the decade, inflation slowed again and for most of the 1980s was contained in a 1-3 percent band. More recently, inflation in west Germany showed a persistent rise from its low point in 1988 of 1 1/2 percent to a peak of 4 1/2 percent in 1992—a period that included unification in 1990. Inflation has since abated: by end-1993, it had eased to about 3 percent and fell further during 1994.

**Inflation, Money, Output Gap, and Import Prices**

(Four-quarter percentage changes)

Source: Author’s estimates.^{1}West Germany before mid-1990. Data spliced to remove discontinuity.

^{2}Actual minus potential real GDP as a percent of potential.

**Inflation, Money, Output Gap, and Import Prices**

(Four-quarter percentage changes)

Source: Author’s estimates.^{1}West Germany before mid-1990. Data spliced to remove discontinuity.

^{2}Actual minus potential real GDP as a percent of potential.

**Inflation, Money, Output Gap, and Import Prices**

(Four-quarter percentage changes)

Source: Author’s estimates.^{1}West Germany before mid-1990. Data spliced to remove discontinuity.

^{2}Actual minus potential real GDP as a percent of potential.

From inspection of Chart 5-2, it is clear that several of the turning points in inflation were preceded by changes in the growth of broad money (M3) in the same direction. The lag appears to be a couple of years. Thus, the slowdowns in inflation in the mid-1970s and the early 1980s were both preceded by sharp decelerations in money growth, while the pickup in inflation in the late 1980s was preceded by a pickup in money growth in the mid-1980s. However, the most recent rise in inflation is less obviously related to earlier money growth developments—although this period is more difficult to interpret because of the disruptions to key data stemming from unification. Indeed, money growth was on a *declining* path from 1987 to around 1991.

Two other potential factors behind historical movements in inflation are also included in Chart 5-2. The first, denoted the output gap, measures the percentage deviation of real GDP in west Germany from its trend level.^{6} These deviations appear broadly correlated with inflation. In particular, the climb down from the peak of inflation in 1980 to the low levels in the rest of that decade corresponds to a period in which output generally persisted at a level below normal capacity. The more recent increase in inflation took place against a shift to above-normal capacity utilization—particularly as the west German economy boomed in the immediate postunification period. As the recession hit in 1992-93, inflation slowed.

The second factor, changes in import prices, also appears to bear some loose correlation with the historical movements in inflation—although the visual evidence is less strong for west Germany than it is for other countries. In particular, it can be seen that inflation picked up temporarily around the times of the oil price hikes of 1973-74 and 1979-80, while the fall in inflation to its low of about 1 percent in 1987 appears to have been helped by the collapse in oil prices in the middle of the 1980s.

## Determinants of Inflation: Evidence from VARs

This section utilizes vector autoregressive (VAR) models to evaluate the predictive roles of various monetary indicators in the inflation process. VAR models are based on regressions of each variable on lagged values of itself and the other variables in the system. They have the advantage of permitting an evaluation of the properties of the data without imposing prior restrictions suggested by economic theory or by policy reaction rules. However, the analysis is subject to practical constraints, such as limits on degrees of freedom and problems of multi-collinearity in the data, and the results can be quite sensitive to changes in model specification. A key metric of the importance of the variables is provided by a decomposition of the VAR system’s forecast error variance for inflation.^{7} Further information is provided by examining the response of each variable to random perturbations in each equation (the “impulse responses”).

### Pre-unification Data

The analysis of VAR models was initially restricted to data from the pre-unification period in order to avoid issues of data splicing. In explaining inflation in the VAR models, the following variables were included: broad money growth and the spread between ten-year government bond yields and three-month money market rates (which capture domestic monetary conditions);^{8} the output gap, defined as the ratio of real GDP to its trend value; and import prices, which incorporate both a potential endogenous outlet for the external transmission of monetary policy (the exchange rate) and an exogenous cost push element (world commodity prices). Import prices, the GDP deflator, and money were expressed as the four-quarter change in the logarithm of each series—the inflation rate over the course of four quarters—in order to turn these variables into stationary time series.^{9} On the basis of likelihood ratio tests for higher-order terms, a maximum lag length of five was selected for the VAR models.

The analysis found evidence that all the chosen variables provide some information about inflation developments (Table 5-2). In general, the results suggest that money is a fairly robust indicator of future inflation, explaining anywhere between 27 percent and 48 percent of the error variance of inflation after five years. Nevertheless, the other variables can explain a large proportion of the remaining variance and generally provide useful independent information on future inflation.

**Contribution to Forecast Error Variance of Inflation: Estimation Sample, 1971:3-1989:4**

(In percent of total)

^{1}The orthoganalization process needed to decompose the error variance is not independent of the ordering of the variables (Sims, 1980). In model 6a, the ordering runs money growth, spread, import price inflation, gap, price inflation. In model 6b, the order of the money and spread variables are reversed. In the other models, changes in ordering had little effect on the results.

**Contribution to Forecast Error Variance of Inflation: Estimation Sample, 1971:3-1989:4**

(In percent of total)

Money Growth | Yield Spread | Import Price Inflation | Output Gap | Lagged Inflation | ||
---|---|---|---|---|---|---|

Model 1 | ||||||

After two years | 5.8 | … | … | … | 94.2 | |

After five years | 36.8 | … | … | … | 63.2 | |

Model 2 | ||||||

After two years | 13.1 | … | … | 31.7 | 55.2 | |

After five years | 48.0 | … | … | 19.8 | 32.2 | |

Model 3 | ||||||

After two years | 20.2 | … | 7.8 | 36.5 | 35.8 | |

After five years | 46.0 | … | 14.4 | 22.2 | 17.5 | |

Model 4 | ||||||

After two years | … | 12.7 | 6.7 | 35.5 | 45.1 | |

After five years | … | 13.5 | 5.0 | 41.9 | 39.5 | |

Model 5 | ||||||

After two years | 37.2 | 17.4 | 2.9 | … | 42.6 | |

After five years | 43.0 | 22.2 | 7.9 | … | 26.9 | |

Model 6a^{1} | ||||||

After two years | 26.6 | 14.5 | 4.4 | 21.7 | 32.7 | |

After five years | 34.5 | 25.9 | 5.5 | 15.9 | 18.2 | |

Model 6b1 | ||||||

After two years | 31.5 | 9.7 | 4.4 | 21.7 | 32.7 | |

After five years | 27.0 | 33.4 | 5.5 | 15.9 | 18.2 |

^{1}The orthoganalization process needed to decompose the error variance is not independent of the ordering of the variables (Sims, 1980). In model 6a, the ordering runs money growth, spread, import price inflation, gap, price inflation. In model 6b, the order of the money and spread variables are reversed. In the other models, changes in ordering had little effect on the results.

**Contribution to Forecast Error Variance of Inflation: Estimation Sample, 1971:3-1989:4**

(In percent of total)

Money Growth | Yield Spread | Import Price Inflation | Output Gap | Lagged Inflation | ||
---|---|---|---|---|---|---|

Model 1 | ||||||

After two years | 5.8 | … | … | … | 94.2 | |

After five years | 36.8 | … | … | … | 63.2 | |

Model 2 | ||||||

After two years | 13.1 | … | … | 31.7 | 55.2 | |

After five years | 48.0 | … | … | 19.8 | 32.2 | |

Model 3 | ||||||

After two years | 20.2 | … | 7.8 | 36.5 | 35.8 | |

After five years | 46.0 | … | 14.4 | 22.2 | 17.5 | |

Model 4 | ||||||

After two years | … | 12.7 | 6.7 | 35.5 | 45.1 | |

After five years | … | 13.5 | 5.0 | 41.9 | 39.5 | |

Model 5 | ||||||

After two years | 37.2 | 17.4 | 2.9 | … | 42.6 | |

After five years | 43.0 | 22.2 | 7.9 | … | 26.9 | |

Model 6a^{1} | ||||||

After two years | 26.6 | 14.5 | 4.4 | 21.7 | 32.7 | |

After five years | 34.5 | 25.9 | 5.5 | 15.9 | 18.2 | |

Model 6b1 | ||||||

After two years | 31.5 | 9.7 | 4.4 | 21.7 | 32.7 | |

After five years | 27.0 | 33.4 | 5.5 | 15.9 | 18.2 |

^{1}The orthoganalization process needed to decompose the error variance is not independent of the ordering of the variables (Sims, 1980). In model 6a, the ordering runs money growth, spread, import price inflation, gap, price inflation. In model 6b, the order of the money and spread variables are reversed. In the other models, changes in ordering had little effect on the results.

On its own, money growth is capable of explaining some 37 percent of the error variance of inflation after five years (Table 5-2, Model 1). This simple model also confirms that causality runs from money to prices (in the sense that lagged values of money are significant determinants of inflation). The peak impact of a shock to money growth occurs after two to three years—a result in tune with the earlier visual impression in Chart 5-2.

The proportion of the error variance explained by money rises to nearly one half (Models 2 and 3) if the two nonmonetary variables—the output gap and import price inflation—are included. At the same time, the nonmonetary variables considerably reduce the forecast variance attributed in the Table to lagged inflation.^{10} The importance of the output gap appears to be particularly pronounced in the short run, and after two years it explains about one third of the forecast error variance of inflation.

Analysis also confirms that the yield spread variable contains independent information that augments the overall explanatory power of the VAR models.^{11} At the same time, there is some evidence that the yield spread is a useful substitute for the output gap, which would be consistent with the results in Hu (1993) that suggest yield spreads are useful predictors of output developments. For example, in the most general model (Model 6 in Table 5-2), the inclusion of the yield spread renders the output gap variable jointly insignificant for the VAR as a whole, even though lags of the output gap are jointly quite significant in the inflation equation. In addition, there is a degree of symmetry between the results for Models 3 and 5 in Table 5-2 in which the output gap and yield spread variables appear, respectively, on their own. In general, however, the yield spread does not appear by itself to have been a good substitute for money as an indicator for future inflation (compare Models 3 and 4).

Additional information in interpreting the results is provided by an examination of the impulse responses of the VAR systems to different shocks. In the most general model, where all four monetary and nonmonetary variables are included along with inflation, a shock to money growth has a fairly immediate, but temporary, impact on the output gap and a more drawn out impact on price inflation. The latter effect gradually builds up to a peak toward the end of two years (Chart 5-3). By contrast, a positive shock to the yield spread, which might be interpreted as a lowering of short-term interest rates, has a delayed impact on output, which begins to rise significantly after a period of 18 months to 2 years.^{12} While inflation reacts perversely in the short term, it is significantly higher in the medium term (Chart 5-3, middle panel). Finally, a shock to import price inflation (which could reflect unanticipated currency depreciation) appears to have a rather small impact on inflation, largely because it leads to a contraction of monetary growth—perhaps reflecting the typical policy response in the past—and the opening up of an output gap.

**Impulse Responses: Estimated 1970-89**

(In percent)

Source: Author’s estimates.**Impulse Responses: Estimated 1970-89**

(In percent)

Source: Author’s estimates.**Impulse Responses: Estimated 1970-89**

(In percent)

Source: Author’s estimates.### Extending the Estimation After Unification

Estimates of the VAR models incorporating data from the postunification period (1990-93) indicate a smaller role for financial variables—both money and yield spreads—in explaining inflation and a significantly greater role for the output gap (Table 5-3).^{13} This would be in keeping with the earlier observation that the most recent inflationary episode was not preceded by a significant upsurge in monetary growth. The result is particularly strong if united German estimates of real GDP and the GDP deflator are used: the higher inflation rate in the unified economy, against a background of much weaker output growth, was even more at odds with the relatively benign earlier expansion of the money slock. Of course, the fall in output in east Germany and rise in price level after unification was more of a one-off adjustment, suggesting that underlying economic developments may, for the time being, be more reasonably measured by west German indicators.

**Contribution to Forecast Error Variance of Inflation: Estimation Sample, 1971:3-1993:4 ^{1}**

(In percent of total)

^{1}Data for inflation and the output gap refer to west Germany only.

^{2}The orthoganalization process needed to decompose the error variance is not independent of the ordering of the variables (Sims. 1980). In model 6a, the ordering runs money growth, spread, import price inflation, gap price inflation. In model 6b, the order of the money and spread variables are reversed. In the other models, changes in ordering had little effect on the results.

**Contribution to Forecast Error Variance of Inflation: Estimation Sample, 1971:3-1993:4 ^{1}**

(In percent of total)

Money Growth | Yield Spread | Import Price Inflation | Output Gap | Lagged Inflation | ||
---|---|---|---|---|---|---|

Model 1 | ||||||

After two years | 4.3 | … | … | … | 95.7 | |

After five years | 28.6 | … | … | … | 71.4 | |

Model 2 | ||||||

After two years | 8.8 | … | … | 36.0 | 55.2 | |

After five years | 29.2 | … | … | 29.1 | 41.7 | |

Model 3 | ||||||

After two years | 17.8 | … | 7.5 | 41.2 | 33.5 | |

After five years | 42.3 | … | 7.4 | 29.9 | 20.5 | |

Model 4 | ||||||

After two years | … | 11.0 | 10.8 | 36.0 | 42.2 | |

After five years | … | 9.3 | 13.0 | 36.2 | 41.5 | |

Model 5 | ||||||

After two years | 25.7 | 16.6 | 6.6 | … | 51.1 | |

After five years | 39.8 | 17.1 | 5.6 | … | 37.6 | |

Model 6a^{2} | ||||||

After two years | 22.7 | 8.9 | 5.6 | 32.6 | 30.2 | |

After five years | 32.6 | 15.2 | 4.3 | 26.0 | 22.0 | |

Model 6b^{2} | ||||||

After two years | 24.7 | 6.8 | 5.6 | 32.6 | 30.2 | |

After five years | 29.4 | 18.4 | 4.2 | 26.0 | 22.0 |

^{1}Data for inflation and the output gap refer to west Germany only.

^{2}The orthoganalization process needed to decompose the error variance is not independent of the ordering of the variables (Sims. 1980). In model 6a, the ordering runs money growth, spread, import price inflation, gap price inflation. In model 6b, the order of the money and spread variables are reversed. In the other models, changes in ordering had little effect on the results.

**Contribution to Forecast Error Variance of Inflation: Estimation Sample, 1971:3-1993:4 ^{1}**

(In percent of total)

Money Growth | Yield Spread | Import Price Inflation | Output Gap | Lagged Inflation | ||
---|---|---|---|---|---|---|

Model 1 | ||||||

After two years | 4.3 | … | … | … | 95.7 | |

After five years | 28.6 | … | … | … | 71.4 | |

Model 2 | ||||||

After two years | 8.8 | … | … | 36.0 | 55.2 | |

After five years | 29.2 | … | … | 29.1 | 41.7 | |

Model 3 | ||||||

After two years | 17.8 | … | 7.5 | 41.2 | 33.5 | |

After five years | 42.3 | … | 7.4 | 29.9 | 20.5 | |

Model 4 | ||||||

After two years | … | 11.0 | 10.8 | 36.0 | 42.2 | |

After five years | … | 9.3 | 13.0 | 36.2 | 41.5 | |

Model 5 | ||||||

After two years | 25.7 | 16.6 | 6.6 | … | 51.1 | |

After five years | 39.8 | 17.1 | 5.6 | … | 37.6 | |

Model 6a^{2} | ||||||

After two years | 22.7 | 8.9 | 5.6 | 32.6 | 30.2 | |

After five years | 32.6 | 15.2 | 4.3 | 26.0 | 22.0 | |

Model 6b^{2} | ||||||

After two years | 24.7 | 6.8 | 5.6 | 32.6 | 30.2 | |

After five years | 29.4 | 18.4 | 4.2 | 26.0 | 22.0 |

^{1}Data for inflation and the output gap refer to west Germany only.

^{2}The orthoganalization process needed to decompose the error variance is not independent of the ordering of the variables (Sims. 1980). In model 6a, the ordering runs money growth, spread, import price inflation, gap price inflation. In model 6b, the order of the money and spread variables are reversed. In the other models, changes in ordering had little effect on the results.

## Monetary Conditions Index

The previous section suggests that other financial variables, in principle, could augment or even substitute for money as an indicator of future inflation. In this section, a monetary conditions index, like the one used as an operational target for monetary policy in Canada, is constructed and its ability as an inflation indicator is compared to that of M3.

### Constructing the MCI

The MCI described in Freedman (1994) is constructed as a weighted average of changes in short-term interest rates and the effective exchange rate. The two components can be expressed in either real or nominal terms, with the choice of weights based on the relative impact of each component on either real demand or prices over a six- to eight-quarter horizon. In principle, Freedman indicates a preference for a weighting of real rates based on their individual effects on real demand because this largely determines the output gap and hence inflationary pressures. In Canada’s case, empirical evidence suggests a three-to-one weighting; a 1 percentage point increase in real interest rates is assumed to have the same impact on demand as a 3 percent appreciation of the real effective exchange rate. In practice, Freedman points out, the nominal-and real-based MCIs are highly collinear, and so the nominal and real MCIs provide comparable information.

A similar index was constructed for Germany. Based on simulations of a simplified model of output and inflation, a three-to-one weighting also appeared broadly appropriate—although the results were not oversensitive to different weighting choices.^{14} The precise formula was as follows:

where *RR* represents real three-month money market rates, *RER* the real effective exchange rate, and the subscript 85 refers to the level in the first quarter of 1985. It should be noted that the choice of the latter period as a benchmark is arbitrary so that no significance can be attributed to the level of the MCI. However, an increase in its value should generally reflect a loosening of monetary conditions, and vice versa for a decrease.

Up to unification, the MCI had a strong positive correlation with inflation two to three years ahead. The correlation coefficient is almost as high in magnitude as that between inflation and lagged money growth (Table 5-4).^{15} However, for both money and the MCI, the correlation is very weak in the postunification period, suggesting that neither indicator would have predicted developments in inflation around the turn of the decade very well (Chart 5-4). This is perhaps not surprising given the unprecedented nature of the demand shock to the west German economy brought on by unification. The MCI, for example, was signaling (largely on the basis of the strong real appreciation of the deutsche mark in the second half of the 1980s) that monetary conditions had been tightening for some time before unification.^{16} However, it was not until after 1992 that inflation stabilized and then began to head downward. But it was also the case that, as discussed above, the rise in inflation in the late 1980s and early 1990s was not preceded by a pickup in monetary growth. Furthermore, the pickup in monetary growth from the end of 1991 onward would, on the basis of past experience, have been suggesting that inflation should perhaps have begun to accelerate in 1993 and early 1994. The opposite occurred.

**Inflation Correlation Coefficients**

**Inflation Correlation Coefficients**

1973-89 | 1973-93 | |||
---|---|---|---|---|

MCI backward-looking | ||||

price expectations | ||||

Lagged 8 | 0.49 | 0.35 | ||

Lagged 10 | 0.63 | 0.45 | ||

Lagged 12 | 0.66 | 0.49 | ||

MCI forward-looking | ||||

price expectations | ||||

Lagged 8 | 0.53 | 0.39 | ||

Lagged 10 | 0.65 | 0.49 | ||

Lagged 12 | 0.63 | 0.50 | ||

M3 growth | ||||

Lagged 8 | 0.74 | 0.64 | ||

Lagged 10 | 0.69 | 0.62 | ||

Lagged 12 | 0.64 | 0.59 |

**Inflation Correlation Coefficients**

1973-89 | 1973-93 | |||
---|---|---|---|---|

MCI backward-looking | ||||

price expectations | ||||

Lagged 8 | 0.49 | 0.35 | ||

Lagged 10 | 0.63 | 0.45 | ||

Lagged 12 | 0.66 | 0.49 | ||

MCI forward-looking | ||||

price expectations | ||||

Lagged 8 | 0.53 | 0.39 | ||

Lagged 10 | 0.65 | 0.49 | ||

Lagged 12 | 0.63 | 0.50 | ||

M3 growth | ||||

Lagged 8 | 0.74 | 0.64 | ||

Lagged 10 | 0.69 | 0.62 | ||

Lagged 12 | 0.64 | 0.59 |

**Monetary Conditions**

(Four-quarter percent change)

Source: Author’s estimates^{1}West Germany.

^{2}In percent. A decline reflects a tightening of monetary conditions.

^{3}Spliced data. See Appendix for details.

**Monetary Conditions**

(Four-quarter percent change)

Source: Author’s estimates^{1}West Germany.

^{2}In percent. A decline reflects a tightening of monetary conditions.

^{3}Spliced data. See Appendix for details.

**Monetary Conditions**

(Four-quarter percent change)

Source: Author’s estimates^{1}West Germany.

^{2}In percent. A decline reflects a tightening of monetary conditions.

^{3}Spliced data. See Appendix for details.

### Implications for Recent Monetary Policy

The MCI and M3 growth provide different evaluations of German monetary conditions in the past two to three years. The early revival of money growth, dating from the end of 1991, would have suggested a significant and ongoing loosening of monetary conditions. By contrast, the MCI would have suggested that monetary conditions continued to tighten until 1993. Even in 1993, the MCI only pointed to a modest easing of monetary conditions because the effects on real interest rates of declines in short-term interest rates were offset to some extent by falling inflation, while at the same time some further appreciation of the exchange rate took place. This interpretation of monetary conditions would support the decision to cut official short-term interest rates significantly during 1993—and to continue doing so in the first half of 1994 despite the substantial overshooting of the monetary target. The decision is further validated by the sharp slowdown in M3 growth in the second half of 1994, which helped to reduce inconsistencies between the two indicators of monetary conditions.

## Conclusions

On average in the past, a monetary conditions index based on interest and exchange rates would probably have provided a useful early warning of future inflation. However, as was the case with M3, the MCI would not have been a particularly reliable predictor of inflation in the early postunification period. More recently, as M3 growth continued to exceed its target, the MCI signaled tight monetary conditions, at least until 1993, when they only began to case gradually. This tends to support the view that concerns about a revival in inflation in the next few years are not warranted—a view corroborated by the subsequent slowdown in monetary growth in the second half of 1994.

## Appendix: Data Splicing Assumptions

Time-series analysis of the postunification period requires the splicing of key data series in order to remove discontinuities associated with the enlargement of the economic area. For nominal and real GDP data, quarterly estimates for united Germany were available from DIW from the beginning of 1990. These were spliced with earlier estimates for west Germany on the assumption that the growth rate in the first quarter of 1990 was the same in both the old and new Lander. As a result, and given the initial sharp fall in output in east Germany, real GDP growth in the unified economy is estimated to have been considerably weaker in 1990-91 than in the western Lander alone (Chart 5-A1), while inflation would have been somewhat higher because of price adjustment in the east. Thereafter, the revival of the east German economy would have had a small offsetting effect on the recession in the west.

**Inflation and Real GDP**

(Four-quarter percent change)

Source: Author’s estimates.**Inflation and Real GDP**

(Four-quarter percent change)

Source: Author’s estimates.**Inflation and Real GDP**

(Four-quarter percent change)

Source: Author’s estimates.Money stock data for the unified economy have been published on a consistent basis since the beginning of 1991, although a separate east-west monetary survey also exists for the second half of 1990. On the basis of the overlapping observations, the unified German money stock was estimated for the last two quarters of 1990 and spliced with earlier data for west Germany on the assumption that money growth was the same in the eastern and west-em Lander in the third quarter of 1990.^{17}

The data indicate that money velocity in the united economy during the second half of 1990 was below that in the western Lander. This might reflect, in part, the lack of nonmonetary financial investment opportunities in the former German Democratic Republic (Chart 5-A2). According to the overlapping data for the second half of 1990, the velocity of money in the united economy was as much as 6 percent lower than in the western Lander alone. However, on the basis of the splicing assumptions, the discrepancy would have been only 3 percent prior to monetary union in mid-1990.

## References

Cassard, Marcel, Timothy Lane, and Paul Masson, “ERM Money Supplies and the Transition to EMU,”

*IMF Working Paper No. 94/1*(Washington: International Monetary Fund, 1994).Davis, E.P., and S.G.B. Henry, “The Use of Financial Spreads as Indicator Variables: Evidence for the United Kingdom, and Germany.”

Vol. 41 (September 1994). pp. 517 –25.*Staff Papers*, International Monetary Fund,Deutsche Bundesbank,

(January 1994).*Monthly Report*Engle Robert F. and C.W.J. Granger, “Co-Integration and Error Correction: Representation, Estimation, and Testing,”

Vol. 55 (March 1987), pp. 251 –76.*Econometrica*,Freedman, Charles, “The Use of Indicators and of the Monetary Conditions Index in Canada,”

*in*(Washington: International Monetary Fund, 1994).*Frameworks for Monetary Stability: Policy Issues and Country Experiences*Hodrick, Robert, and Edward C. Prescott, “Post-War U.S. Business Cycles: An Empirical Investigation” (unpublished; Pittsburgh: Carnegie-Mellon University, 1980).

Hu, Zuliu, “The Yield Curve and Real Activity,”

Vol. 40 (December 1993), pp. 781 –806.*Staff Papers*, International Monetary Fund,Schinasi, Gary, “Asset Prices, Monetary Policy, and the Business Cycle,”

*IMF Paper on Policy Analysis and Assessment No. 94/6*(Washington: International Monetary Fund, 1994).Sims, Christopher A., “Macroeconomics and Reality.”

Vol. 48 (January 1980), pp. 1 –48.*Econometrica*,Todd, Richard M., “Vector Autoregression Evidence on Monetarism: Another Look at the Robustness Debate.”

(Spring 1991), pp. 19 –37.*Federal Reserve Bank of Minneapolis Quarterly Review*von Hagen, Jürgen, “Monetary Union, Money Demand, and Money Supply: A Review of German Monetary Union.”

Vol. 37 (1993).*European Economic Review*,

^{}1

In 1994, for example, the target was based on an objective of 2 percent inflation, a decline in trend velocity of 1 percent a year, and potential growth of 2 1/2 percent a year, giving an indicative monetary growth target of *5 1/2* percent. The actual target was expressed as a range (4-6 percent) with adjustments added for target overshooting in previous years. See Deutsche Bundesbank. *Monthly Report* (January 1994), pp. 17-21.

^{}2

Note that equation (1) always holds true by identity and can be consistent with an entirely different causal ordering: inflation could originate from a completely separate process and money adjust subsequently to accommodate a higher price level.

^{}4

Schinasi (1994), for example, notes that this phenomenon played an important role in the latest business cycle in a number of countries.

^{}6

The trend is defined by means of a Hodrick-Prescott filter of the actual GDP series (see Hodrick and Prescott, 1980). Alternative estimation methods, based on production function analysis, produced similar estimates.

^{}8

Initial estimates also examined the predictive value of equity prices. However, little significance could he found for this variable, and so the results are not reported here.

^{}9

Because statistical tests suggested that the levels of all variables, including interest raies, were integrated of order unity, there would be a risk of spurious regression results if the data were not differenced.

^{}10

Block exclusion tests—analogous to F-tests in a single regression but extended to the VAR system—confirm that the out put gap and import price inflation provide additional explanatory power for inflation over and above that provided by money growth alone

^{}12

This interpretation of shocks to the yield curve would be consistent with historical experience: most of the variance in the spread variable can be accounted for by variance in short-term rates as opposed to variance in bond yields.

^{}13

See the Appendix at the end of this chapter for data splicing assumptions. A dummy variable was included to allow for a potential break in behavior after the end of 1989 or to compensate for inappropriate splicing assumptions.

^{}14

As in the VAR simulations, interest rates influence inflation in the model through their impact on the output gap: exchange rate changes have a more direct impact on inflation.

^{}15

As in the case of Canada, the properties of the MCI were similar regardless of whelher it was expressed in nominal or real terms. Nor does it make much difference (as reported in Table 5-4) whether real interest rates are defined using forward or backward measures of inflation expectations.

^{}16

The real exchange rale measure is based on relative normalized unit labor costs in manufacturing. As pointed out in Chapter II, this indicator may have exaggerated the implied loss of competitiveness.