EU-wide money and cross-border holdings

EU-Wide Money and Cross-Border Holdings By

Carlo Monticeili

C o n t e n t s : I. Introduction. - II. Traditional and Extended Money Measures. - III. Cointegration Analysis. - IV. Dynamic Money Demand Equations. - V. Conclu- sions. - Appendix.

I. Introduction

T he purpose of this paper is to provide new empirical evidence on the economic properties of EU-wide monetary aggregates focusing on one aspect so far unexplored: the role of cross-

border holdings (CBHs). The analysis of money demand functions defined for a group of EU countries as a whole has recently attracted widespread attention. 1 The interest in this approach and the significance of the encouraging results stem from two arguments.

First, the estimation of area-wide equations may be regarded as an effective device to overcome the problems encountered in the specification of stable and well-performing money demand functions for several EU countries (see the surveys by Boughton 1991 and by Mon- ticelli and Strauss-Kahn 1992). As single-country equations typically omit foreign variables (partly as a consequence of the high degree of correlation between domestic and foreign variables), cross-country spiUover effects can be effectively captured by aggregating money demand equations across countries which have strong economic and financial interlinkages. Indeed, the increasing degree of economic and financial integration within the EU has been associated with an intensification of international portfolio diversification and with growing currency substitution, calling for an explicit examination of the role of foreign variables on the residents' demand for monetary assets. The specification bias affecting traditional single-country equations is re-

Remark: The opinions expressed in the paper are those of the author and should not be interpreted as reflecting the views of the Banca d'Italia. Comments by an anonymous referee, M. Artis, P. Masson, and L. Papi, and the competent research assistance provided by L. Schoett Jensen are gratefully acknowledged. i For a survey of the research following the path-breaking work of Bekx and Tullio (1989) and Kremers and Lane (1990), see van Riet (1993) and Monticelli and Papi (1996).

216 Weltwirtschaftliches Archiv 1996, Vol. 132(2)

duced by the estimation of area-wide equations. However, an aggregation bias may be introduced to the extent that there are different money demand relationships in the countries in the group (Pesaran et al. 1989; Kremers and Lane 1992a).

Secondly, the analysis of the EU-wide demand for money has important implications for the conduct of monetary policy. In Stage III of the Economic and Monetary Union, monetary policy will be set on the basis of area-wide monetary relationships and thus the analysis of money demand equations for a group of countries as a whole can provide a useful background to assist policy implementation in the early days of EMU. 2 During the transition to the Economic and Monetary Union, a well-performing money demand at the area level can provide a valuable tool for the coordination of national monetary policies. Moreover, the progress of economic and financial integration, the increase of international portfolio diversification, and the intensification of currency substitution might increase the volatility of national money demands, while making the area-wide money demand significantly more stable than in any individual country. If this were the case, it would be to the benefit of all the countries in the area to implement monetary policy with reference to area-wide monetary relationships. This arrangement would also imply a mutually satisfactory solution to the issue of the symmetry of the ERM (Monticelli and Papi 1996: Chapter 2).

CBHs considered in this paper consist of the monetary assets of the private non-financial sector held with financial intermediaries located in countries other than those in which the holders are resident. Even though European CBHs have been growing rapidly in recent years, their effects on monetary analysis have been explored only by few studies which have obtained promising results with reference to the properties of national monetary aggregates (Angeloni et al. 1991; Giucca and Levy 1992). Furthermore, the inclusion of (some) CBHs in an EU-wide broad monetary aggregate seems warranted on the grounds of logical consistency: for example, French residents' Deutsche mark deposits in German banks should be part of the measure of the European money stock since they are held in Europe by European residents and are denominated in a European currency. Yet, the em-

2 Two caveats are, however, in order. First, the move to a single monetary policy will be such an important structural change that it might have devastating consequences on the stability of money demand equations. Secondly, the uncertainty surrounding which countries will participate in EMU from the outset implies that studies of area-wide economic relationships should consider different groups of countries.

Monticelli: EU-Wide Money 217

pirical analysis of money demand for groups of EU countries has so far neglected CHBs, since it has relied on area-wide measures of money obtained from the sum of national aggregates, thereby under- estimating the EU money stock.

The paper is organized as follows. Section II considers the preliminary issues relating to the definition of area-wide monetary aggregates which include CBHs and presents six different 'extended' measures of broad money. The economic properties of these aggregates are then explored, using as a benchmark the results from the analysis of the traditional aggregate obtained from the sum of national definitions of broad money. Section III is devoted to the cointegration analysis to assess the stability of the money demand equations; Section IV to the estimation of error-correction models in order to investigate the shorter-run dynamics of monetary relationships. Section V concludes.

II. Traditional and Extended Money Measures

The definition of measures of the money stock requires the application of at least one, but possibly more (including all), of the following three criteria which underlie the definition of the monetary aggregate for a particular geographical area, typically (but not necessarily) a nation:

Residence of the holder (R): the aggregate should include the monetary assets held by residents in the area (irrespective of the currency of denomination and of the location of the issuer with whom the assets are held). Location of the issuer (L): the aggregate should include the monetary assets issued by the intermediaries located in the area (irrespective of the currency of denomination and of the residence of their holders). Currency of denomination (C): the aggregate should include the monetary assets denominated in the currency(ies) of the area (irrespective of the location of the issuers and of the residence of the holders).

The measures of money which originate from the implementation of the three criteria (or a combination thereof) will be different, unless the extent of international financial integration is so low that residents in a particular area keep all their monetary assets with intermediaries located in that area and hold only assets denominated in domestic currency (Goodhart 1990).

Currently, in most EU countries monetary aggregates are defined by applying more than one criterion, typically (but not always) the

218 Weltwirtschaftl iches Archly 1996, Vol. 132 (2)

criteria of residence and location. The preference granted in practice to these criteria, however, has rarely been examined by econometric analysis and is mainly the result of convenience in data collection.

When considering area-wide measures of money, i.e. when the money stocks of different countries are aggregated, the issue of the consistency between national and area-wide measures arises. The straightforward adding up of national aggregates across countries yields a consistent area-wide measure only if one, and only one, of the three criteria is applied in each and every country. If different criteria are, jointly or singly, applied in the countries in question, the sum of national aggregates will be inconsistent and will involve omissions and/or duplications. Conversely, if any combination of the criteria is applied in each and every country, the sum of national aggregates will systematically underestimate the area-wide measure consistently defined on the basis of the same criteria with reference to the whole area. The intuition underlying this result is simple: foreign, but 'within-the- area', variables become 'domestic' with reference to the area as a whole. Thus, for example, Deutsche marks in France and French francs in Germany are foreign currencies, which, however, have to be considered as domestic currency with reference to any area bloc including both France and Germany.

Table 1 illustrates the relationship between the application of the three criteria at the national and at the EU level. Thus, for example, if national aggregates are set out according to the national residence of the holder and the national location of the issuer (as is the case with traditional broad monetary aggregates in EU countries), 3 their sum will be equal to column EH. In contrast, the EU measure resulting from the application at the EU level of the same criteria includes column EW in addition (i.e., EU residents' CBHs held within the EU).

Table 1 is also useful to illustrate the following definitions of the EU-wide monetary aggregates analyzed in the paper, where super- scripts indicate the criteria applied at the EU level to identify the CBHs included in the extended definition, with the exception of T which stands for traditional:

M r -- Sum of traditional national broad monetary aggregates (column EH);

M n = M r extended to include EU residents' CBHs irrespective of their location and currency denomination (columns EH + EW + EO);

a The UK is an exception: M4 includes only assets denominated in pounds.

Monticelli: EU-Wide Money

Table 1 - Definition of EU-Wide Aggregates

219

NO EO EW EH NW

EU currencies 1 2 3 4 5

Non-EU currencies

6 7 8

9 I0 11 12

EU currency of the country where the bank is located

Other EU currencies

Note: NO represents Non-EU residents' monetary holdings, denominated in EU currencies, held at intermediaries located Outside the EU.

EO represents EU residents' cross-border holdings Outside the EU. EW represents ELI residents' cross-border holdings Within the EU. EH represents ELI residents' holdings with intermediaries at Home, i.e.

located in their country of residence. NW represents Non-EU residents' cross-border holdings Within the EU.

M r = M r extended to include CBH s held within the E U irrespective o f the residence o f the holder and the currency denomina t ion (columns E H + EW + NW);

M R, L = M r extended to include EU residents' CBHs held within the E U irrespective o f their currency denomina t ion (columns EH + EW);

M R. c = M r extended to include E U residents ' CBHs denomina ted in EU currencies irrespective o f their locat ion (column E H + cells 2, 3, 6);

M L'c = M r extended to include CBHs held with banks located within the EU and denomina ted in EU currencies irrespective o f the residence o f the holders (column EH + cells 3, 5, 6, 8);

M R, L, C = M r extended to include EU residents ' CBHs denomina ted in EU currencies and held with intermediaries located in the E U (column E H + cells 3, 6).

This list omits M c, the aggregate extended to include CBHs denomina ted in EU currencies irrespective o f the locat ion o f the intermediaries and o f the residence o f the holders. The omission is moti- vated by the belief that n o n - E U residents ' CBHs held outside the EU - such as the deposits denomina ted in deutsche marks which US

220 Weltwirtschaftl iches Archiv 1996, Vol. 132(2)

residents hold in Hong Kong (represented by cell 1 in Table 1) - cannot realistically be expected to have a link with output and price developments within the Community. 4

Before moving to the empirical analysis of the role of CBHs in EU-wide monetary relationships, it has to be stressed that the factors underlying the importance of extended aggregates (independently of the definition) relative to traditional money measures are distinct from those which support the validity and usefulness of EU-wide monetary aggregation. The relevance of CBHs derives from the delocalization of deposits, that is from the decoupling of the country of residence of the holder from the country where the intermediary issu- ing the liability is located. The economic rationale for the definition of area-wide aggregates is instead provided by currency substitution, international portfolio diversification, and, more in general, by the spillover effects across integrated economies which render the demand for money sensitive to changes in foreign, but 'within-the-area', variables. These factors, however, need not lead to significant changes in the location of monetary assets even when they exert an important influence on the choice between monetary and non-monetary assets and on decisions regarding the currency composition of assets.

Large shifts in the currency denomination of monetary holdings, for example due to currency substitution, may not imply fluctuations in the size of CBHs, although they will in any case have a significant impact on the stability of the demand for money and on the control- lability of the traditional monetary aggregates. When shifts in currency denomination take place, money demand is influenced by variables which are traditionally excluded from the specification and which are typically outside the control of the 'domestic' central bank. For the sake of analytical simplicity, macroeconomic models usually employed to explore the implications of currency substitution (see for example the classic paper by Girton and Roper 1981) implicitly identify the criterion of the location of the issuer with the criterion of currency denomination, and therefore currency substitution automat- ically implies deposit delocalization. In contrast, the empirical coun- terpart to the theoretical notion of money is sensitive to the choice of the criteria underlying its measurement. Whereas the implications for the conduct of monetary policy derived from theoretical models hold

4 As regards the criterion of currency of denomination, it should be noted that its application would imply the exclusion of foreign-currency deposits in non-EU currencies held at home (cell 11) from the aggregates M a'c, M L'c, M a'~'c. This, however, was not possible because of statistical limitations.


irrespective of whether currency substitution indeed coincides with shifts of funds abroad - the consequences for the aggregates extended to include CBHs will instead be different.

I lL Cointegration Analysis

The empirical analysis includes all the countries in the EU except Greece, Luxembourg, Portugal (which were excluded because of limitations in the available data), and Austria, Finland, and Sweden (which joined the EU after the end of the sample considered in the analysis). CBHs held in these countries, and most notably in Luxem- bourg, are however included in the extended aggregates. Given the relatively small size of the countries excluded, the group underlying the analysis can be regarded as an accurate proxy for the whole EU.

Turning to the aggregation method to calculate area-wide series, for variables expressed in the same units, such as rates and indices, the use of a weighted averaging scheme is standard, with the shares of the real income variables in the aggregate employed as weights. The choice of the method to convert quantity variables (money and income) into a common currency is more controversial, as three alternative rates have been used in the literature: (i) current exchange rates; (ii) fixed base-period exchange rates; and (ii i) purchasing power parity (PPP) rates, s However, the theoretical underpinnings of each method have not been explored to date 6 and hence not even tentative conclusions have yet been reached on the issue.

In this paper, preference is given to the use of current exchange rates, selecting the ECU as the numeraire. 7 This method is intuitively appealing and provides a consistent market evaluation of the stocks of financial assets across countries and currencies. Furthermore, it is in line with the notion that money balances (be they transactions or precautionary balances) are held because of their (current and future) effective purchasing power, which, in terms of foreign goods and assets, is appropriately measured by market exchange rates.

For example, method (i) has been used in Gray et al. (1976) and in Gurney and In't Veld (1991); method (ii) in McKinnon (1982) and in Bekx and Tullio (1989); method (iii) in Kremers and Lane (1990). 6 For example, in the exchange between Arnold (1992) and Kremers and Lane (1992 a), the way in which method (ii 0 should be applied is discussed in detail without clarifying why this particular conversion method should be used. v In order to assess the robustness of the findings, some of the empirical exercises have been replicated employing quantity variables converted at PPP rates; analogous results were obtained.

222 Wel twi r t scha f t l i ches Arch iv 1996, Vol. 132(2)

The Appendix contains a description of the national variables selected for aggregation. National broad monetary aggregates form the basis of the traditional, 'non-extended', area-wide measure of money M r . Broad aggregates, rather than narrow ones, are chosen since they are currently used as targets or indicators in several EU countries and hence they may be expected to have a greater bearing on the implementation of monetary policy, either at the national or at the area-wide level. Moreover, this selection is in line with the main thrust of the study: it would be contradictory to compare 'very broad' money measures - i.e., definitions extended to include CBHs - with a narrow aggregate. The income variable Y is either real GDP or GNP, depending on availability, while the price variable is the corresponding deflator, P. The interest rate employed in the analysis, r, s is the yield on government bonds, taken as a proxy for the return on the whole range of assets alternative to money.

The sample spans from the first quarter of 1979 (inception of the European Monetary System) to the first quarter of 1992 (latest period for which detailed data on CBHs are available to the author).

Once the area-wide variables are defined, the first step in the empirical analysis is to ascertain the order of integration of the series and to assess whether they exhibit time trends. Stationarity tests, not reported for the sake of brevity, were performed on the logarithm of the EU-wide variables 9 (denoted by lower cases of the notation presented above), exploring the possibility that the series might have up to two unit roots and time trends up to second order. The results quite clearly support the view that all the series, both in nominal and in real terms, are non-stationary and exhibit a single unit root, i.e. differenc- ing only once is sufficient to obtain stationary variables. As is often the case, less clear-cut results are obtained regarding the order of the deterministic components in the specification, particularly for money and income. Replicating the analysis for different subsamples does not help to obtain a more definite picture, as the trend properties of the variables seem to vary over the sample. The results are mixed and difficult to interpret in a consistent manner, but tend to suggest that

s In preliminary estimates, a short-term interest rate (three-month interbank rate) was also considered. However, when both the short- and the long-term interest rates are included in the specification, severe multicollinearity problems arise. The long-term rate has been preferred throughout the analysis since it is more in line with the standard theory underpinning the specification of the long-run equation (1). 9 The logarithmic transformation is not applied to EU-wide interest rates since the long-run equations are specified in semi-logarithmic form.


real money and output could be integrated of order one around a linear trend starting only in the mid-1980s. Therefore, cointegration analysis is conducted conditionally on different assumptions regarding the presence of a time trend (either spanning the entire sample or starting in the mid-1980s) in the specification of the cointegrating vector.

The stability of the demand for EU-wide money is explored on the basis of a simple, traditional specification for the long-run equation:

(m-p )~=ao+~l yt + ct2 re, cq >0 , ~2 < 0 . (1)

In line with the mixed findings on the trend characteristics of the variables, and as a test of the robustness of the results, a linear trend from 1986.1 to 1989.4 (denoted by segt) lo is also added to the specification, either on its own or together with a time trend spanning the entire sample.

Table 2 presents the results of the long-run equations for the different definitions of EU-wide money.

Inspection of the table may begin with the first row - which refers to m r, the traditional definition of broad money excluding CBHs - since these results can be considered as a benchmark to assess the performance of the extended aggregates. Irrespective of the insertion of time trends in the specification, test results do not reject the hypothesis that the simple trivariate model of money demand for m r cointe- grates. This finding confirms the key result of a growing body of literature surveyed in Monticelli and Papi (1996): the degree of economic and financial integration within the EU is already sufficient for an area-wide demand for broad money to be stable. Area-wide monetary aggregates thus stand as a valuable tool to analyze and interpret monetary developments within the Community.

The income elasticity in the model with no time trends is high, well above the results typically obtained from national equations. When the linear trends are added to the specification, the income elasticity is significantly reduced, reaching values in line with the estimates exceeding unity which are found for Italy (e.g., Muscatelli and Papi 1990), Belgium (e.g., Jeanfils 1992), Germany (e.g., Schmid and Herr- man 1991) and France (e.g., de Bandt 1991).

1o The period spanned by segt was chosen on the basis both of the hints provided by the stationarity tests and of the results in Monticelli and Strauss-Kahn (1993). Perron (1989; 1990) proposes parametric tests to identify the presence of breaks in the trend characteristics of economic series. This type of analysis, however, is beyond the scope of this paper.

2 2 4 W e l t w i r t s c h a f t l i c h e s A r c h i v 1996, Vol. 132(2)

T a b l e 2 - Cointegrat ion Analysis , 1 9 7 9 . 1 - 1 9 9 2 . 1

Monetary aggregates

m r - p m s - - p m L --p mR,L_p mR, C--p m L,c-p mS,L,C_p

m I" _p m R - - p

m L - p m~,L_p m~,C_p m L,c_p mR,L,C_p

m r - p m R - - p

m L - p mR, L_p mR,C--p mL, C--p m~.L.C_p

Coefficients on Test statistics

y r DF PP

Constant

1.65 -0 .61 -4 .14"** -4 .60*** 1.99 -0 .57 -2 .20 - 2.99 1.86 -0 .59 - 1.50 - 1.28 1.90 -0 .58 -2 .28 - 3 . 1 6 " 1.93 -0 .54 - 2.55 - 3.73"* 1.92 - 0.63 - 2,47 - 3.46"* 1.89 - 0.59 - 2.47 - 3.61 **

Constant, segt

1.38 - 0.66 - 3.56" - 3.98 ** 2.10 -0 .55 -2 .31 -3 .28 2.32 -0 .50 - 1.52 - 1.61 1.87 -0 .59 -2 .25 - 3.08 1.88 -0 .54 -2 .48 -3 .64* 1.89 - 0.63 - 2.43 - 3.37 1.82 - 0 . 6 0 -2 .39 -3 .71 *

Constant, segt, t

1.14 -0 .45 -4.31 ** -4 .84*** 1.40 -0 .29 -3 .61 * -4 .42** 0.84 -0 .02 -3 .54 -3 .19 1.21 -0 .35 -3 .97** -4 .43** 1.32 -0 .35 - 3.77" -4 .58"* 1.27 -0 .41 -4 ,02"* -4 .68"** 1.24 -0 .40 -3 .95* -4 .59**

Note: *** significant at the 1%, ** at the 5%, and * at the 10% level. Segt denotes a linear trend spanning the period 1986.1 -1989.4, while t denotes a time trend over the whole sample. DF and PP denote the Dickey-Fuller cointegration test (Dickey and Fuller 1981) and the test statistic proposed by Phillips (1987) and by Phillips and Perron (1988), respectively. The PP test was performed in a specification with four lags. Critical values are from McKinnon (1991).

T h e s e n s i t i v i t y o f t h e i n c o m e e l a s t i c i t y t o t h e n o n - s t o c h a s t i c c o m -

p o n e n t o f t h e s p e c i f i c a t i o n is o p e n t o t w o c o m p l e m e n t a r y i n t e r p r e t a -

t i o n s . F r o m a s t a t i s t i c a l p o i n t o f v i e w , i t s t r e s ses t h e i m p o r t a n c e f o r

c o i n t e g r a t i o n a n a l y s i s o f a t h o r o u g h e x p l o r a t i o n o f t h e t r e n d c h a r a c -

t e r i s t i c s o f t h e v a r i a b l e s , c a l l i n g f o r a c l o s e e x a m i n a t i o n o f deterministic t r e n d s s i n c e t h e y m a y h a v e a p r o f o u n d e f f e c t o n t h e s t a t i s t i c a l

MonticeUi: EU-Wide Money 225

inference about s t o c h a s t i c trends (see, e.g., Perron 1989 and Sims et al. 1990). From an economic point of view, time trends may pick up the effects of other determinants of money demand which are missing in the simple trivariate model. Trends are certainly ad hoc regressors to capture the effects of these developments, and ideally economic variables should be used instead such as proxies for the intensity of financial innovation and measures of real and financial wealth, which for example is an essential ingredient of the demand for broad money in the UK (Hall et al. 1990). Unfortunately, data limitations for several EU countries do not allow any improvement in this direction in the specification of the area-wide model. However, it is reassuring that the time trends, besides their statistical justification, may be given a reasonable economic interpretation in terms of the structural developments in the financial field in the second half of the 1980s, when financial innovation and deregulation raised the liquidity of and in- creased the returns on the interest-bearing assets included in broad aggregates, leading to a steady decline in the EU-wide velocity of money.

Moving to the definitions of money extended for CBHs, the second and the third rows of Table 2 show that the demand for aggregates which include 'broad' measures of CBHs (i.e., measures based on the application of one defining criterion only, see the description of m R and m L above) fails to cointegrate, particularly in the absence of the linear trends. This finding is not surprising when it is recalled that these monetary aggregates include some assets which could hardly be expected to have a direct link with spending decisions within the EU. In fact, m R contains EU residents' CBHs located outside Europe, possibly in off-shore financial centres, and denominated in non-EU currencies. Delocalization of deposits outside the EU pre- sumably involves higher transaction costs and is more closely connected with the avoidance of domestic regulations than delocalization within the Community, while the denomination in non-EU currencies exposes the holder to a particularly intense exchange rate risk. These considerations suggest that the absence of a stable money demand for this aggregate is plausible from an economic viewpoint and indicate that only an explicit analysis of portfolio allocation can provide insights on the factors underlying the demand for CBHs (see Alworth and An- dresen 1992). Similar reasons account for the lack of cointegration for m L. The latter aggregate includes non-EU residents' CBHs kept with intermediaries located in European financial centres - such as London or Luxembourg - which can be expected to attract funds from outside

226 Weltwirtschaftliches Archiv 1996, Vol. 132(2)

the Community more in relation to the financial services they provide than as a result of the developments in EU-wide aggregate income.

The evidence supporting cointegration is instead stronger for the other extended aggregates, which include the narrower measures of CBHs based on the joint application of two (or three) defining criteria. Test statistics have, however, lower significance than for the traditional definition of broad money - particularly when no trend is inserted in the specification - suggesting that, even though cointegration holds, the EU-wide link between aggregates extended for CBHs, output and the interest rate may be less strong and dependable than for the money measure which excludes CBHs.

Irrespective of the presence of time trends in the specification, the income elasticity of the demand for extended aggregates is systematically higher than for the traditional definition of broad money. This finding hints at the fact that portfolio considerations play a more important role than transaction motives in the demand for CBHs, no matter which definition is considered. The pattern of interest rate semi-elasticities is instead quite sensitive to the presence of trends, with values that tend to be notably lower for extended aggregates that include 'broad' measures of CBHs (m R and m L) than for traditional aggregates in the specification including both linear trends.

Both the long-run coefficients and the values of cointegration tests are, however, quite similar for the extended aggregates defined by applying the residence criterion jointly with (one of) the other two criteria (i.e., mR'r; mR'C; and mR'r"c). This finding reflects the fact that EU residents have increasingly tended to place the CBHs they keep within the Community in EU currencies. Therefore, combinations of these criteria yield measures of CBHs which have approxi- mately the same size and which possess similar economic properties. In contrast, the results obtained for m ~' c are markedly different from those for m r, providing evidence to support the view that non-EU residents' CBHs denominated in non-EU currencies, although kept with financial intermediaries located in the EU (these CBHs account for a large part of the difference between m L and mt"c), are not linked with spending decisions within the Community.

IV. Dynamic Money Demand Equations

The estimation of dynamic models of money demand comple- ments the empirical analysis of the long-run equations in two respects. First, the evaluation of the performance of dynamic equations is a


crucial test to ascertain whether monetary relationships are sufficiently stable and predictable to provide useful guidance in the conduct of monetary policy. Secondly, from a statistical point of view, the estimation of error-correction models provides a test on the robustness of the conclusions from cointegration analysis. Not only does the Granger Representation Theorem - which states that the cointegrated series can be represented by error-correction models (Granger 1983; Engle and Granger 1987) - establish the fundamental rationale to cross-check the empirical results from the two approaches, but more recently Kremers et al. (1992) have argued that the standard t-ratio for the coefficient on the error-correction regressor in the dynamic equation is a more powerful test for cointegration than other more commonly applied methods, such as the Dickey-Fuller test.

The estimation of the error-correction models follows both the single-stage approach - which leaves the long-run coefficients of money demand unconstrained - and the two-stage approach, which employs the lagged cointegrating residual to represent the feedback towards the long-run equilibrium and thus imposes the long-run coefficients obtained from the static regression. However, the two proce- dures yield virtually identical results and the appropriate F-test (denoted 'Equality test' in Table 3) never rejects the joint hypothesis that the estimates of the long-run coefficients of the money demand obtained from the single-stage unconstrained equation are equal to coefficients obtained from the static cointegrating equation (1). Hence, for the sake of brevity, results from the single-stage approach are not reported.

The specification search used Hendry's general-to-specific modelling strategy to arrive at the equations presented in Table 3, which typically display a fairly simple dynamics. Since the contemporary change in EU-wide income figures among the regressors, the test put forward by Hausman (1978) was carried out to ascertain the presence of a statistically significant simultaneity bias. The test statistics (denoted 'Hausman test' in Table 3) are well below the 5 per cent critical values, thus failing to reject the null hypothesis that the estimates are equal to those obtained from instrumental variable estimation: if not absent, the simultaneity bias is negligible.

The dynamic equation for the traditional definition of broad money is reported below together with the adjusted R 2, the estimated standard error of the equation and some misspecification tests (for a description, see Hendry 1989). ECMT_, denotes the lagged residual from the cointegrating regression which includes both time trends.

228 Weltwirtsehaftliehes Archiv 1996, Vol. 132(2)

( ~ -

o~

I e~

o~

~s

~s

~s

I

,z :

~...., ~ ,',,,', ~ ~ ,~- oo ,,'-, ,~ ._ = .~ ~ . v

�9 ~. ~ ~ ~ o ~ ,~. ".~ o .r

~ x o o o o o o ~ ~ ~ ~ : ~ o ~

- = - I ~ ~ o ~ ,

0 ( ' ~ 1 O ~ 0 { " 4 ~ ( ' ~ 1 ~ ~ ~/~

I . . . . . . . ~ -,.-, ~ ~ . o I

m ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ i

i i ~ ~ ~ - - ~ ~ I ~ ' ~ ~ - ~ r - - ~ - - ~ . o ~ . ~ I

i ~ ~ . ~ ~ i I ~ ~ " ~ " ~ o ' ~ , - < ~ I i o~H o~ ~,.H ~ . ~ ~ i

~ l , I I I I I , ~ . ' ~ ~ ~ I l ---" ~ ~ o ~ o ,~ ~, . ,u i I ~ . ~ o. , . . , . - ' . ~ i

I

I ~ =~ ~,~,~ I I ,= ~ ' ~ ~ " ~ i

-~."1~. ~ ...~ = ~ o ~ . ~ I

~ l o , ' ~ o ~ o ~ 0 ~ ~ > ~ i I I I ~ ~ 0 ,'~ i

I ~ ' = ~ . o o - , I l ~ ~ .~. ~ '= ~ I

i - ~ o ~ " , ~ ~='= o ~ ~ _~ I

~ i ~ , ~ ~ ~ ~ ~= ~ 0 . ~ ' ~ I n i = ~ 0 ~ O:._~ ~ I I ~ = ~ '

I . . . . . . . ~ - ~ ~ , ~ 1 I ~ . ~ . . . . . . ~ . . . . . ~ _ . . ~ 1 I o , ~ o ~ o M o ~ o ~ o ( ~ o ~ .~ .~, = ' = - ~ ~ ~ ' 1 I ~ o ~ , ~ 1 m ~ ~ O * - . O ~ ~ I

~.~ ~ ~ ~ i I ~ . ~ . ~ . 0 . m r - : . ~ . ~ m ~ ~ . ( ' - I ~ r ~ . I ' v e = 0 ~- "~ i i o ~ o ~ o ~ o ~ o ~ o , ~ , ~ = ~ ~ ~ ~ 1

~ , I ~ I I ~ I I ( i I I I I I " ~ ~ i I .=-=, ~ ~ : ~ I ' ~ ~ ,..:.~0 u I

i)) i i '


The restrictions imposed on the coefficients were tested and could not be rejected at the 10 per cent level.

A (m r - p), = 0.01 + 0.45 A (m r - P)t- 1 + 0.21 A (m r - p),_ 2

(0.61) (4.40) (2.22)

+ 0.27 A (m r - p),_ 4 + 0.54 Ay t - 0.27 Ayt_ 1

(2.52) (5.34) (-3.25)

-0 .27 Ayt-4-0.45 Art-1 +0.35 Art_ 4 (-3.25) (-3.46) (2.67)

-0 .52 ECM~'_ 1

(-5.51)

T = 1979.2 - 1992.1

R 2 =0.708

S.E. per cent =0.451

AR (4)~ = 1.72

NORM (2) =0.68

HET (10, 35) = 2.06

ARCH (4, 38)=0.43

CHOW(17, 29)=0.55

RESET (3, 43)= 1.97

The statistical performance of the above equation is very satisfactory: it passes the diagnostic tests, tracks real money growth well, and the equation standard error compares favourably with those of national money demand functions (see, e.g., Fase and Winder 1993) and of the previous area-wide studies for groups of EU countries (see the survey by van Riet 1993 and by Monticelli and Papi 1996). Results from Chow tests and from recursive estimation over the successive samples obtained by adding one observation from 1979.2 - 1988.1 (omitted here for the sake of brevity) point to the stability of the equation.

The pattern of dynamic responses to changes in real income and in the rate of interest is plausible from an economic point of view. The error-correction term is strongly significant and above the 1% critical value in McKinnon's (1991) tables (which Kremers et al. 1992 conjec- ture to be the appropriate ones), providing further support for the hypothesis of cointegration. The value of the coefficient is relatively high for an equation for broad money, implying that more 50 per cent of the correction of shocks from equilibrium takes place within one quarter. If the traditional argument by Goodfriend (1985) that the misspecification of money demand equations typically results in very

230 W e l t w i r t s c h a f t l i c h e s A r c h i v 1996, Vol. 132(2)

low estimates of the adjustment coefficients is accepted, 11 this finding is yet another indication of the good economic and statistical properties of the money demand function presented above.

The rows after the first row in Table 3 report the results of the dynamic money demand functions for the monetary aggregates which include CBHs. For all aggregates, including m R and m L, the statistical performance of the equation is quite good and the error-correction term has high t-statistics. The latter, however, are in all cases smaller than the corresponding statistic in the dynamic equation for traditional money; in particular, the t-statistic decreases sharply for m R, m L

and m L" c This evidence is in lirte with the conclusions of cointegration analysis and suggests that, even if cointegration holds also for m a and m L, it is more tenuous than for all the other aggregates.

The dynamic money demands for the extended aggregate possess good statistical properties by the usual standard - especially in the case of m R' L and m R, L, c _ pointing to the fact that aggregates including CBHs can provide useful information about economic developments at the EU level. Nevertheless, they never outperform the traditional definition of money, which consequently remains the preferred monetary aggregate. The difference in performance is, however, very small, especially in terms of the standard error of the estimated equation, one of the key parameters in assessing the relevance of demand functions for the conduct of monetary policy. This suggests that extended aggregates may well become an increasingly important tool for monetary analysis at the EU level if the progress of financial and economic integration is associated with significant additional delocalization of deposits.

The estimates of the coefficient on the error-correction term are systematically lower in the equations for the extended aggregate than for m r , although they remain quite high by usual standards for broad aggregates. This result is consistent with the reasonable expectation that adjustment costs are higher for assets held abroad than for assets held at home. The pattern of dynamic responses is very similar to that for traditional money, with possibly a simpler structure as typically fewer lags of the changes in money and income enter the short-run equations for the aggregates including CBHs.

t 1 Kremers and Lane (1992 b) have shown that this argument also applies for national money demand equations when the misspecification arises from the omission of regres~ sots vehic~ can capture eurr~'acy subs~itutior~ arid other internatiox~a! sp~l~ver effects.


V. Conclusions

The econometric evidence presented in this paper confirms, for an extended sample and new data, that the degree of economic and financial integration within the EU is already sufficient for EU-wide monetary aggregates to be useful for the analysis of economic developments within the Community and for the coordination of national monetary policies. Cointegration analysis and error-correction modelling show that the EU-wide demand for money is stable and predictable, with a statistical performance at least as good as the best-performing national equations.

The main novelty of the paper is the analysis of the role of cross- border holdings (CBHs) for monetary analysis at the EU level, which had previously been considered only with reference to single countries. Several aggregates extended to include different measures of CBHs are put forward as a result of the application of different combinations of the criteria which underpin the definition of the money stock. Ex- tended aggregates are shown to have very different economic properties according to the definition. 'Very broad' measures of EU money (which contain EU residents' CBHs kept outside the EU and denominated in non-EU currencies as well as non-EU residents' CBHs kept within the EU but denominated in non-EU currencies) are poorly linked with EU-wide income, since they fail to pass cointegration tests. In contrast, extended definitions 'focused on the EU' (which hinge on the inclusion of CBHs denominated in EU currencies and/or kept within the EU) have a satisfactory performance, already sufficiently good to play an auxiliary role in EU-wide monetary analysis. How- ever, no extended aggregate yet outperforms the EU-wide measure of money obtained from adding traditional national definitions: the traditional definition of broad money remains the preferable aggregate for the analysis at the EU level, even though CBHs are likely to become increasingly important in size and economic significance.

Appendix

D a t a S o u r c e s a n d D e f i n i t i o n s

Broad monetary aggregates: All aggregates, published by national central banks, are quarterly averages of monthly, seasonally adjusted figures. Belgium: M3H; Denmark: M3H (= M2 national less Treasury Bills); Germany: M3 (to avoid a break in the series, for the most recent

232 Weltwir t schaf t l iches Archiv 1996, Vol. 132(2)

observations the rate of change in M3 for all Germany following unification was applied to the series for M3 for West Germany prior to unification); Spain: M3H; France: M3H; Ireland: M3; Italy: M3H (= M2 national); the Netherlands: M3H (corrected for items in transit and breaks); United Kingdom: M3H.

Real income: Real income is measured by real GDP and obtained from the BIS data bank for all countries except Denmark, Spain and the Netherlands (estimates from national central banks) and for Bel- gium and Ireland (OECD Economic Outlook; yearly data interpo- lated), where GNP is used.

GNP/GNP implicit price deflators: All deflators derive from the same source as the corresponding seasonally adjusted income variables.

Long-term interest rate: Yield on ten-year government bonds; for some countries the yield on bonds with shorter maturities had to be employed as ten-year ones were not available during the whole sample period.

Exchange rates: Exchange rates are quarterly averages of the spot rates implied by the official definition of the ECU.

Cross-border holdings (CBHs): CBH series are elaborated on the basis of elementary data from the BIS banking statistics and include assets with initial maturities below (or equal to) one year. Measures of CBHs obtained from the joint application of the criteria R and C do not include CBHs denominated in Danish kroner, Greek drach- mas, Spanish pesetas or Portuguese escudos, as reporting banks do not single out these currencies in their statistical returns. For measures obtained from criteria L and C, the underestimation is however smaller, because Danish and Spanish banks do report non-residents' CBHs denominated in kroner and pesetas, respectively.

References

Alworth, J. S., and S. Andresen (1992). The Determinants of Cross-Border Non-Bank Deposits and the Competitiveness of Financial Market Centres. Money Affairs 5 (2): 105-133.

Angeloni, I., C. Cottarelli, and A. Levy (1991). Cross-Border Deposits and Monetary Aggregates in the Transition to EMU. IMF Working Paper 91/114.

Arnold, I.J.M. (1992). The Derivation of the Liquidity Ratio in the EMS: Comment on Kremers and Lane. IMF Staff Papers 39(1): 195-202.


Artis, M.J., R.C. Bladen-Hovell, and Wei Bin Zhang (1993). A European Money Demand Function. In P. Masson and M.P. Taylor (eds.), Policy Issues in the Opera- tion of Currency Unions. Cambridge: Cambridge University Press.

Bekx, P., and G. Tullio (1989). A Note on the European Monetary System, and the Determination of the DM-Dollar Exchange Rate. Cahiers Economiques de Bruxelles 123 (3): 329-343.

Boughton, J. M. (1991). Long-Run Money Demand in Large Industrial Countries. IMF Staff Papers 38 (1): 1 - 32.

De Bandt, O. (1991). Politique Mon6taire et Agr6gats de Monnaie en France. Bulletin Trimestriel de la Banque de France 78 (1): 79-93.

Dickey, D., and W Fuller (1981). Likelihood Ratio Statistics for Autoregressive Time Series with a Unit Root. Econometrica 49(4): 1057-1072.

Engle R.F., and C. W J. Granger (1987). Cointegration and Error Correction: Repre- sentation, Estimation and Testing. Econometrica 55(2): 251-276.

Fase, M.M.G., and C. C.A. Winder (1993). The Demand for Money in the Nether- lands and the Other EC Countries. De Economist 141 (4):471-496.

Girton, L., and D. Roper (1981). Theory and Implications of Currency Substitution. Journal of Money, Credit, and Banking 13(1): 12-30.

Giucca, P., and A. Levy (1992). Monetary Aggregates and Monetary Policy Coordina- tion on the Way to Economic and Monetary Union: the Role of Cross-Border Deposits. Temi di Discussione del Servizio Studi 162, Banca d'Italia.

Goodfriend, M. (1985). Reinterpreting Money Demand Regressions. In K. Brunner and A.H. Meltzer (eds.), Understanding Monetary Regimes. Carnegie-Rochester Conference Series 22. Amsterdam: North Holland.

Goodhart, C. (1990). International Financial Linkages. LSE Financial Market Group, Special Papers, 21.

Granger, C. W. J. (1983). Cointegrated Variables and Error-Correcting Models. Univer- sity of California Working Papers 83-13. San Diego.

- - (1988). Some Recent Developments in a Concept of Causality. Journal of Economet- rics 39(1/2): 199-211.

Gray, M.R., R. Ward, and G. Zis (1976). The World Demand for Money Function: Some Preliminary Results. In M. Parkin and G. Zis (eds.), Inflation in the World Economy. Manchester: Manchester University Press.

Gurney, A., and J.W in't Veld (1991). Monetary Control in a European Monetary Union. Unpublished paper. National Institute of Economic and Social Research, London.

Hall, S.G., B.S.G. Henry, and J.B. Wilcox (1990). The Long-Run Determination of UK Monetary Aggregates. In B. S. G. Henry and K.D. Patterson (eds.), Economic Modelling at the Bank of England. London: Chapman and Hall.

Hausman, J. A. (1978). Specification Tests in Econometrics. Econometrica 46(6): 1251 - 1272.

Hendry, D. E (1989). PC-GIVE: An Interactive Econometric Modelling System. Insti- tute of Economics and Statistics, University of Oxford.

234 Weltwir tschaf t l iches Archiv 1996, Vol. 132(2)

Jeanfils, P. (1992). La Demande de Monnaie. Banque Nationale de Belgique, mimeo.

Kremers, J. J. M., and T.D. Lane (1990). Economic and Monetary Integration and the Aggregate Demand for Money in the EMS. IMF Staff Papers 37(4): 777-805.

- - (1992 a). The Derivation of the Liquidity Ratio in the EMS: Reply to Arnold. IMF Staff Papers 39 (1): 203 - 207.

- - (1992 b). The Implications of Cross-Border Monetary Aggregation. IMF Working Paper 92/71.

Kremers, J. J. M., N. R. Ericsson, and J. J. Dolado (1992). The Power of Cointegration Tests. Oxford Bulletin of Economics and Statistics 54(3): 325-347.

Lane, T.D., and S.S. Poloz (1992). Currency Substitution and Cross-Border Monetary Aggregation: Evidence from the G-7. IMF Working Paper 92/81.

McKinnon, R.I. (1982). Currency Substitution and Instability in the World Dollar Standard. American Economic Review 72(3): 320-333.

Monticelli, C., and L. Papi (1996). European Integration, Monetary Co-Ordination and the Demand for Money, Oxford: Oxford University Press.

Monticelli, C., and M.-O. Strauss-Kahn (1992). Broad Monetary Aggregates and Na- tional Money Demand in Europe. Economies et Soci~t~s 26(9/10): 147-197.

- - (1993). European Integration and the Demand for Broad Money. The Manchester School of Economic and Social Studies 61 (4): 345-366.

Muscatelli, V.A., and L. Papi (1990). Cointegration, Financial Innovation and Mod- elling the Demand for Money in Italy. The Manchester School of Economic and Social Studies 58 (3): 242-259.

Perron, P. (1989). The Great Crash, the Oil Price Shock and the Unit Root Hypothesis. Ecanometrica 57 (6): 1361 - 1401.

- - (1990). Further Evidence on Breaking Trend Functions in Macroeconomic Vari- ables. Princeton University Working Paper 350.

Pesaran, H. M., R. G. Pierse, and M. S. Kumar (1989). Econometric Analysis of Aggre- gation in the Context of Linear Prediction Models. Econometrica 57(4): 861-888.

Phillips, P. C. B. (1987). Time Series Regression with a Unit Root. Econometrica 55 (2): 277-301.

Phillips, P. C. B., and P. Perron (1988). Testing for a Unit Root in Time Series Regressions. Biometrika 75 (2): 335-346.

Riet, A.G. van (1992). European Integration and the Demand for Money in the EU. De Nederlandsche Bank Quarterly Bulletin 1992/3: 33-43.

- - (1993). Studies of EU Money Demand: Survey and Assessment. De Nederlandsche Bank Quarterly Bulletin 1992/4: 63-75.

Schmid, P., and H. Herrman (1991). Federal Republic of Germany: Old and New Problems Posed by a Target-Guided Monetary Policy. In Bank for International Settlements (ed.), The Orientation of Monetary Policy and the Monetary Policy Decision-Making Process. Basle.

Schmidt, P., and P. C. B. Phillips (1992). LM Tests for a Unit Root in the Presence of Deterministic Trends. Oxford Bulletin of Economics and Statistics 54 (3): 257-287.


Sims, C.A., J.H. Stock, and M.W Watson (1990). Inference in Linear Time Series Models with Some Unit Roots. Econometrica 58 (1): 113-144.

Stock, J.H., and M.W Watson (1987). Testing for Common Trends. Journal of the American Statistical Association 83 (400): 1094-1107.

A b s t r a c t: EU-Wide Money and Cross-Border Holdings. - The paper explores the economic properties of several measures of EU-wide money that include different definitions of cross-border holdings (CBHs). 'Very broad' aggregates are poorly linked with EU-wide income and price developments; in contrast, the demand for aggregates which is 'focused on the European Union' (which hinge on the inclusion of CBHs denominated in EU currencies and/or kept within the EU) is shown to be stable and predictable_ Although aggregates extended to include CBHs do not yet outperform the traditional measure of broad money, they are likely to become an increasingly important tool for monetary analysis at the EU level. JEL no. E41, F36

Z u s a m m e n fa s s u n g : Europaweite Geldmengen und grenziiberschreitende geldnahe Forderungen. - Der Verfasser untersucht die 6konomischen Eigenschaften einiger Marie fiir europaweite Geldmengenaggregate, welche grenziiberschreitende For- derungen unterschiedlich welter Definitionen einschlieflen. Sehr umfassende Aggregate sind nur schwach mit tier EU-weiten Einkommens- und Preisentwicklung verbunden. Dagegen erweist sich die Geldnachfrage im Fail yon Aggregaten, die auf die Europa- ische Union konzentriert sind (wobei es auf grenziiberschreitende Forderungen an- kommt, die in EU-W~ihrungen denominiert sind und/oder in der EU gehalten werden), als stabil und vorhersehbar. Obwohl Aggregate, die so erweitert sind, dab sic die grenziiberschreitenden geld- nahen Forderungen einschliel3en, bisher nicht mehr leisten als das traditionelle MaB fiir die weite Geldmenge, dfirften sie vermutlich ein immer wichtigeres Werkzeug fiir monetfire Analysen auf EU-Ebene werden.

Documents

EU-wide money and cross-border holdings