Regionalisation versus globalisation in
European ®nancial market integration:
Evidence from co-integration analyses
q
Stefanie Kleimeier
a,c,*, Harald Sander
b,c aMaastricht University, FdEWB, LIFE, P.O. Box 616, 6200 MD Maastricht, The Netherlands bUniversity of Applied Sciences Cologne, Claudinsstr. 1, 50678 Cologne, Germany cMaastricht School of Management, P.O. Box 1203, 6201 BE Maastricht, The Netherlands
Abstract
Motivated by recent regulatory changes, this study investigates the degree of inte-gration in retail lending in six core European Union (EU) countries using co-inteinte-gration methodology which allows to investigate the presence and eects of structural breaks. While in the pre-break period we could detect integration to a limited degree, the evi-dence for integration weakened in the post-1992 period. This could however re¯ect a convergence process, particularly with respect to spreads. This result is clearly a re-gional, not a global phenomenon. As European lending rates are not yet fully inte-grated, the still segmented ®nancial markets pose a challenge for a uni®ed monetary policy.Ó2000 Elsevier Science B.V. All rights reserved.
JEL classi®cation:F36; G28; G15
Keywords:Co-integration; Economic integration; European Union; Financial markets; Interest rates
www.elsevier.com/locate/econbase
q
We appreciate the comments oered by seminar participants at the VII Financial Conference 1998 at the University of Rome ``Tor Vergata'', the 1998 Multinational Finance Society meeting, the 1997 European Financial Management Association meeting, the 1997 Financial Management Association meeting, and the National University of Singapore.
*Corresponding author. Tel.: +31-43-3883733; fax: +31-43-3258530. E-mail address:s.kleimeier@ber®n.unimaas.nl (S. Kleimeier).
0378-4266/00/$ - see front matterÓ2000 Elsevier Science B.V. All rights reserved.
1. Introduction
This study provides evidence on the degree of European UnionÕs (EU)
in-tegration in international ®nancial markets, both on a global as well as regional scale. Motivated by recent regulatory eorts within the EU to bring a re-gionally integrated ®nancial market into eect, we investigate the progress and state of integration in the retail lending market of EU member countries, both from an intra-regional as well as an inter-regional perspective. The study
ad-dresses two issues. First, has the EUÕs single market policy with respect to retail
banking been eective in producing convergence in lending rates and second, is the degree of regional ®nancial market integration dierent from the degree of global ®nancial market integration?
In 1988, the Commission of the European Communities commissioned a study, now widely known as the Cecchini report (Commission of the European Communities, 1988), which derived quantitative estimates of the bene®ts of ®nancial market integration, i.e. the gains in consumer surplus resulting from price reductions for ®nancial services. This study relies on the application of the law of one price to ®nancial markets, predicting that post-integration prices will fall to a level equal to the prices of the country with the lowest pre-inte-gration prices. As Gardener and Teppett (1995) report, these price reductions range between 21% and 4% for the dierent member countries with an average reduction of 11% and are expected to result in a total gain in consumer surplus of 21.6 billion ECU for the 8 member states: Belgium, Germany, Spain, France, Italy, Luxembourg, The Netherlands, and the UK. In order to realise these bene®ts, the second banking directive (2BD) was implemented on 1 January 1993 for the member countries of the EU with the intent to provide an appropriate regulatory environment for the single European banking market. Next to the politically promoted process of European ®nancial market in-tegration, global ®nancial markets are equally subject to increasing trends towards liberalisation and deregulation. Moreover, for the ®rst time within the framework of multilateral trade talks the Uruguay Round of the GATT re-sulted in a ``General Agreement on Trade in Services'' (GATS) as part of the ®nal agreement. However, the progress in putting trade in ®nancial services on
a Ômost favoured nation principleÕ basis has only very recently resulted in an
agreement. This contrasts sharply with the far-reaching integration approach of the EU, which ± by means of adopting the mutual recognition rule ± has created conditions for intensi®ed competition among dierent European reg-ulatory regimes.
global one? To answer these questions, the study is structured as follows. Section 2 reviews the expected bene®ts from ®nancial market integration as suggested by the Cecchini report, the state of the regulatory implementation of the 2BD, and the actual trends in the banking markets of the member states. In Section 3, hypotheses and methodological foundations of the study are laid out. In particular, a co-integration approach and the corresponding error correction model (ECM) will be adopted in order to analyse integration and convergence in retail credit markets as well as the eectiveness of the market arbitrage process. Section 4 reports the results of the European and the global analyses. Finally, Section 5 contains the conclusions and highlights some im-plications of our results for the relative eectiveness of a uni®ed European monetary policy within the European Monetary Union.
2. European ®nancial market integration
2.1. Potential bene®ts from ®nancial market integration
Perfectly integrated markets are by de®nition obeying the ``law of one price''. As identical assets are selling at one single price, one can expect the usual ``gains from trade'' that should materialise as an increase in consumer rents when markets are being integrated. In the Cecchini study such gains in allocative eciency, which lead to a higher industry output at lower prices, have been advocated as one major source of potential bene®ts from ®nancial integration within Europe. A second source of potential bene®ts is derived from deviations from the perfect market assumption, speci®cally by advocating economies of scale and scope and an increase in competitive pressure. If in-ternal economies of scale are important in the ®nancial sector, restricting op-erations within national boundaries implies foregone cost and price reductions. Similarly, scale eects would allow only for a limited number of domestic suppliers and thus generate an oligopolistic market structure that may lead to mark-up pricing strategies that generate super-normal pro®ts.
banks.1No consistent evidence for economies of scope could be found in the US research. A study speci®cally looking at the European market by Steinherr and Gilibert (1989) ®nds no signi®cant relationship between size and pro®t-ability for most member countries and thus, similarly to the US studies, in-dicating no economies of scale in the banking sector. Therefore, signi®cant economies of scale and the associated bene®ts according to Cecchini are doubtful.
Next to these shortcomings, the Cecchini study is subject to criticism from other sides as well. For example, it does not consider the dynamic growth ef-fects of ®nancial integration. Cecchini bene®ts ± if they materialise ± will have a long-run impact on savings and thus investments and growth (see e.g. Baldwin, 1989). Moreover, the Cecchini study does not address the issue of trade di-version. Regional ®nancial market integration has the potential to discriminate against non-members of the regional club. If least-cost suppliers are located outside the regional integration area at least for some members of the club regional ®nancial market integration may be worse than none. Therefore, we investigate integration among the EU countries as well as integration of the EU
vis-a-vis the US and Japan.
A replication of the microeconomic methodology of the Cecchini report is provided by Gardener and Teppett (1995) and reveals many of the short-comings of Cecchini's analysis, i.e. with respect to the forecasted price reduc-tion process. The authors conclude that ``the economic gain estimates are highly suspect; they can only be indicative at best and are practically irrelevant at worst''. The microeconomic bene®ts of ®nancial integration are therefore far from clear and call for an empirical examination of the actual degree of integration.
2.2. Trends in national regulation and bank behaviour
The 2BD as the key legislation towards the single European banking market was adopted by the European Commission in 1988. It was preceded by the ®rst banking directive of 1977, which allowed branch banking under the host country rule. In additional to these two general banking directives, eight ad-ditional directives have been designed between 1986 and 1992 focusing on bank supervision, capital requirements, solvency standard, money laundering, credit risk, annual report requirements and consolidation. Relying on the principles
of harmonisation, mutual recognition, and home country control and super-vision, the developers of the 2BD expected a deepening ®nancial integration within the banking markets in the EU. Formulated in form of a directive, the implementation of this legislation into national law is required by the member states. All other single market legislations are also formulated in the form of directives. Thus, European ®nancial market integration is dependent not only on the eectiveness of the banking directives but also on the state of imple-mentation by the member countries. Zimmerman (1995) has reviewed the state of implementation of all single market directives across member countries and showed that between the end of 1991 and April 1994 the percentage of all directives which had been adopted by member countries rose steadily from 58% to 89%. For the 2BD, 11 of the 12 EU member states had adopted the legis-lation by April 1994, the only exception being Spain, which Zimmerman in-terpreted as a ``serious setback to the completion of the single banking market''. Overall, however, Zimmerman considered the legal framework for ®nancial market integration to be in place. This opinion of advanced legal integration is also shared by other researchers such as Bredemeier (1995) who views the EU legal system as ``far advanced''.
An adequate legal environment by itself might, however, not be sucient for ®nancial market integration. Zimmerman consents that barriers to integration can exist even after full legal implementation. Mitchell (1991) similarly sees constraints for ®nancial integration which are independent of regulatory issues, for example cultural dierences in consumer behaviour including preferences for types of credit and investment, or for investment horizons which dier among countries.
Additional insights into the practical aspects of ®nancial market integration can be found when reviewing bank activities in anticipation of legislative
harmonisation.2 The ®rst banking directive did not lead to any signi®cant
amount of cross-border banking since national regulatory dierences remained intact (e.g. Heremans, 1993; van Cayseele, 1992; Zimmerman, 1995). Accord-ing to Heremans (1993), the European bankAccord-ing market prior to the 2BD was characterised by natural entry barriers such as transaction cost and arti®cial entry barriers caused by diering regulatory systems. Despite the reduction of natural barriers through advances in information technology, which bene®ted mostly the wholesale markets, arti®cial barriers in retail markets were still in place and prevented ®nancial integration. Between the adoption of the 2BD in 1988 and its implementation in 1993, however, European banks seemed to prepare for the single market as an increased number of mergers and
acquisitions, cross-border alliances, or co-operative agreements has shown.3 The mid-1990s recession in Europe slowed this trend down with most banks adopting more cautious strategies which focus on selected ®nancial centres or
activities rather than Europe-wide full service networks.4 With the entry
barriers mentioned earlier and these institutional aspects in mind, the extent of ®nancial market integration in the EU after the 2BD appeared less obvious than regulation alone might suggest. Only recently and coinciding with the arrival of the EMU, intensi®ed international consolidation of ®nancial services could be witnessed inside and outside the EMU-11 countries.
3. Hypotheses and methodology
Our study will not re-examine the Cecchini analysis but attempt to answer the following two questions in the context of the EU: First, are European ®-nancial markets already integrated or at least in a process towards integration?
In other words, has the EUÕs single market policy with respect to retail banking
already been eective in producing convergence in lending rates? Secondly, if ®nancial market integration is actually occurring, is it an exclusively regional phenomenon restricted to the EU member countries, or is it a global one? More speci®cally, is there a dierence in the degree of ®nancial market integration on an intra- and inter-regional level?
3.1. Financial market integration as co-integration
The Cecchini study advances the hypothesis of price equalisation for ®-nancial assets within Europe as one characteristic of completely integrated markets. This hypothesis is known as interest rate parity theory. It is well es-tablished that under perfect capital mobility the covered interest parity typi-cally holds while it is more dicult to establish the empirical validity of the uncovered interest parity (UIP) due to exchange rate volatility or exchange rate expectations. It is however noteworthy to recall that more recently ± and somewhat surprisingly contrasting with the before-mentioned research results ± interest rates in all three major ®nancial centres of the Triad tend to move more closely together as reported in the OECD Economic Outlook (1996). While the UIP mechanism is typically thought of as a relatively fast working arbitrage mechanism, the role of time-dependent risk-premiums suggests that UIP may hold only as a long-run relation. In the short-run deviations from this parity
3For a list of international alliances or acquisitions of European banks in the 1980s and a table regarding the presence of EU banks in other member countries see Canal (1993, ch. 10).
induce adjustment towards long-run UIP equilibrium (Johansen and Juselius, 1992). Such a setting where time series of individual variables ``can wander extensively and yet some pairs of series may be expected to move so they do not drift too far apart'' (Engle and Granger, 1987) is best being studied in the context of a co-integration analysis. The underlying idea of co-integration is that non-stationary time series (such as interest rates) can move apart in the short-run, but will be brought back by market forces to an equilibrium relation
in the long run.5
Since interest rate parity is suggested as parity for interest rates on such assets like government bonds which are at least close if not perfect substitutes, in retail banking it seems even more dicult to argue that credit portfolios are perfect substitutes across countries. Rather, heterogeneity is caused by risk dierences, cultural in¯uences in bank±client relationship, diering monetary policy conditions, country-speci®c strategic bank behaviour in order to cope with informational imperfections (moral hazard, incentive eects etc. as anal-ysed, e.g. by Stiglitz and Weiss (1981)), to name just a few. Consequently, one cannot expect the law of one price to hold in the credit market. Retail interest rates may not as easily equalise as suggested by the Cecchini study. However, as integration will take place pricing in retail banking will be more closely tied together within an increasingly integrating region. Thus, we can employ the concept of co-integration to analyse the state of market integration in retail banking when lending rates are tied together by a long-term equilibrium re-lationship that does not demand strict price equalisation.
Short-run deviations from this long-run equilibrium can be corrected by three mechanisms. First, one can think of an international arbitrage process where banks increasingly shift their lending activities to countries where lending rates are the highest while consumers borrow in low interest countries. The second mechanism is an extension of the UIP. When money market rates equalise by means of an international arbitrage process such changes will have an impact on lending rates via domestic competition that ties lending and borrowing rates together. Finally, international competition, or the threat of it (as suggested by the theory of contestable markets) will help to harmonise the pricing behaviour of banks and thus lead to a co-integration of retail prices.
In this study, we are focussing on nominal rather than real interest rates because nominal rates re¯ect more appropriately the arbitrage process in ®-nancial markets. Borrowers compare nominal lending rates of banks from
5
A non-stationary variablextis said to be integrated of orderd(denotedI(d) withd> 0) if an
dierent countries adjusted for exchange rate expectations. Banks that do not cross borders physically via direct investment are equally interested in nominal rates adjusted for exchange rate expectation. Banks that do cross borders will have to deal with local money market rates and lending rates that are subject to the same price de¯ator when expressed in real terms. As Lemmen (1998) in accordance with Marston (1995) clearly states: ``Real interest dierentials are not comparable across countries because they are denominated in their own national currency. Consequently, they do not oer pro®t opportunities for an individual investor or borrower since no single agent compares real interest rates across countries''. Lemmen (1998) also shows that the real interest parity (RIP) requires (1) a zero country premium, as measured by the validity of the covered interest parity, (2) a zero exchange rate risk premium, and (3) a zero deviation from the ex ante relative purchasing power parity (PPP). In the context of the EU therefore, incomplete goods market integration and subse-quent PPP violations could lead to a rejection of the ®nancial market
inte-gration hypothesis if real interest rates were employed.6On the other hand, in
situations where PPP holds at least approximately, using real rates to test ®-nancial integration would be appropriate. Moreover, Marston (1995) argues that contrary to common belief average real interest rates are actually similar across countries. This fact should not be surprising when recognising that very long-run average deviations from PPP are small even if in the short-run RIP fails to hold. When the real cost of borrowing is ultimately closely related to the returns on investments, a comparable set of investment opportunities in
dierent countries could produce the long-run result even without ®nancial
arbitrage taking place.7 Thus, testing real interest rate integration could be
both too restrictive and potentially misleading. We therefore decide in the context of this study to focus on nominal interest rates as the indicator of integration in European retail ®nancial markets.
Our study will test the ®nancial market integration hypothesis by per-forming co-integration tests for both lending rates and spreads using monthly interest rates starting in 1985. For lending rates the respective national prime rates are chosen and for calculating spreads we chose money market interest rates as a proxy for deposit rates as advocated by Steinherr and Huveneers (1992). The tests are performed ®rst for those European ``core countries'' that have also been included in the original Cecchini study: France, Germany, UK,
6
Empirical evidence (e.g., Froot and Rogo, 1995; Froot et al., 1995; Wei and Parsley, 1995) reveals that PPP deviations are persistent as they have a half-life of about four years, and readjustment is predominantly brought about by nominal exchange rate moves. Engel, (1995, 1996) furthermore argues that even this evidence might overstate PPP validity.
Netherlands, Belgium, and Italy.8Later the analysis is extended in order to compare regional and global ®nancial market integration by including the US and Japan. Co-integration analysis is performed for the various countries
under scrutiny vis-a-vis the weighted average of the remaining EU countries
and Germany, respectively.9 The national commercial bank prime lending
rates are obtained from Datastream and are recorded as o January 1985.
Money market rates are collected from the IMFÕs International Financial
Statistics (IFS) also starting in January 1985.10Spreads are calculated in two
ways. First, nominal spreads are calculated by subtracting the money market rate from the lending rate. Secondly, relative spreads are calculated by dividing the lending rate by the money market rate.
The co-integration analysis follows closely the approach promoted by Engle and Granger (1987) and proceeds in three steps. First the time series must proven to be unit roots. Only then the co-integration vector can be estimated. Finally, once co-integration has been established, the corresponding ECM will
be estimated. In order to establish whether interest rates or spreads areI(1) two
test statistics, a t-statistic and an F-statistic, will be employed based on
re-gressions on levels as well as ®rst dierences of the underlying series. Both include next to lagged observations of the interest rate in question also a trend variable:
Dytab ytÿ1cDytÿ1d Tet; 1
D2ytabDytÿ1cD2ytÿ1d Tet: 2
The null hypothesis states that the series follows random walks. For the
t-statistic, this corresponds to a null hypothesis of H0:b0 and for theF
-sta-tistic to a null hypothesis of H0:bd 0. We fail to reject the null hypothesis
of a random walk if the calculatedtorFvalues are smaller in absolute terms
than the critical values.
8With the exception of Luxembourg and Spain which are included in the Cecchini study but will not be part of the analysis in this study. For Luxembourg, no prime lending rates are available on Datastream, whereas for Spain the IFS money market rates were characterised by too many missing values.
9
The weighted average is calculated using GDP weights. For detailed information see OECD Economic Outlook, December 1996, Table 8. For the core EU countries used in this study the following GDP weights ± as percentages of the total GDP of all OECD countries ± are provided: Germany 8.34%, France 6.37%, Italy 5.97%, UK 5.54%, Belgium 1.06%, Netherlands 1.52%. For the weighted averages, the sum of the remaining country weights is re-based to 100%.
10
Once theI(1) characteristic has been established, co-integration testing can commence starting with Durbin±Watson (DW) statistics obtained from the
co-integration regression using the rate or spreadytfor the individual country as
the dependent variable and the weighted average rate or spread for the
re-maining EU countriesxt as the independent variable:
ytab xtut: 3
The null hypothesis of no co-integration can be rejected when the calculated DW values resulting from the regression of Eq. (3) are larger than the critical values. As Engle and Granger point out, the DW test can be used as a good but only approximate indicator for co-integration and should be followed by a more speci®c testing procedure such as the Dickey±Fuller (DF) and augmented Dickey±Fuller (ADF) tests. The DF test is based on the residuals of the co-integration regression
Dut ÿ/utÿ1et; 4
where thet-statistic for the estimated coecient / provides an indication
re-garding the co-integration of the two series. In particular, the null hypothesis
of no co-integration can be rejected when the t-statistic is larger in absolute
value than the critical value. Furthermore, the augmented DF test is obtained from the regression
Dut ÿ/utÿ1RibDutÿiet; 5
where the null hypothesis of no co-integration can be rejected when the
t-statistic for the estimated coecient / is larger in absolute value than the
critical value.
Finally once co-integration is established, the corresponding ECM will be
estimated in order to investigate the speed of adjustment.11To ®nd the correct
speci®cation of the ECM, the procedure suggested by Engle and Granger (1987) is followed. First, an unrestricted vector autoregression (UVAR) is es-timated based on the regression
Dytd0d1ytÿ1d2xtÿ1RidyiDytÿiRidxiDxtÿiet: 6
From this regression, the signi®cant lagged ®rst dierences of the exogenous and endogenous variables are identi®ed and included in the ®nal ECM in
combination with any error correction terms (ECT) obtained from the estimated errors that were found signi®cant in the co-integration regression
Dytu0u1ECTtÿ1RiuyiDytÿiRiuxiDxtÿiet: 7
Thet-statistic of the estimated coecientu1of the ECT measures the speed of
adjustment. For example an estimatedu1 of )0.2 indicates that if there is a
shock to the variableyt which raises its value relative to the equilibrium
rela-tionship to the co-integrated series xt, then one ®fth of the divergence is
eliminated in the following period.
3.2. Co-integration, structural breaks, and convergence
If the single market project is eective in tying interest rates more closely together this implies a tendency for interest dierentials between countries to decline. Contrary to widely held beliefs, the existence of non-stationary re-sidual, i.e. non-co-integration, could re¯ect the dynamics of a convergence process, as it has recently been argued in the context of interest parity analyses by Caporale et al. (1996). Co-integration is a powerful test of the arbitrage process before and after convergence has occurred, but not during the
con-vergence process itself.12 Therefore, we investigate ®rst whether or not
structural breaks are present in the data. If yes, the sample will be divided into a pre- and post-break period that will both be tested for co-integration. Note that the absence of co-integration can well be consistent with the presence of a convergence process.
Based on the perception that the 2BD could have had a signi®cant impact
on European ®nancial markets by in¯uencing banksÕ interest rate policies as
well as the competitive conditions in general, interest rates series might very well exhibit structural breaks. Thus, we employ test procedures for unit root testing and as well as co-integration which consider the existence of such breaks. As the 2BD was adopted in 1988 for implementation by each individual member country in their respective national law by 1993, the exact timing of such a structural break for each member country is dicult to specify with any degree of exactness. In the presence of structural breaks at an a-priory un-known point in time, the above two unit root tests have very little power. Better speci®ed are the following test statistics proposed by Banerjee et al. (1992).
12
Based on Eqs. (1) and (2), recursive minimumt-statistics can be calculated in
order to test the null-hypothesis ofb0.13Furthermore, sequential unit root
tests that distinguish between a shift in the mean or the trend of the series can be calculated based on the following regressions for levels and ®rst dierences:
Dytab ytÿ1cDytÿ1d TfDet; 8
D2ytabDytÿ1cD2ytÿ1d TfDet; 9
whereDindicates a dummy variable. For the mean-shift tests,Dis coded as 1 if
t>kand 0 otherwise. For the trend-shift tests,Dis coded astif t>kand 0
otherwise. Here, both minimum stests regarding H0: b0 and maximum F
tests regarding H0: bd 0 can be calculated and compared to the critical
values.14
As for the unit root characteristic of the interest rate series, the co-inte-gration relationship might also be characterised by structural breaks. To test for structural breaks in the co-integration relationship, we proceed as follows: First, the co-integration regression of Eq. (3) is performed for the full sample ranging from January 1985 to January 1997 and DW, DF, ADF, and ECM are reported. However, in the presence of a structural break, these tests have low power, i.e. the rejection frequency of the ADF test is clearly reduced (e.g. Gregory et al., 1996). In a second step, the co-integration vector is tested for
structural breaks such that H0:at1at2 andbt1bt2 with sub-samplest1 1
tokandt2k1 toT. Ifk, the time of the break is known, the two samplest1
andt2are clearly identi®ed and a Chow test can be conducted. In our case, the
exact timing of the break ± if indeed there is any ± is not known. Thus, rather
than using a standard Chow test, a supremumF(supF) test is calculated. This
test was ®rst proposed by Quandt (1960) and has recently been the focus of various studies (e.g. Andrews, 1993; Diebold and Chen, 1996; Hansen,
1992).15After determining the break, the total sample will be split in two
sub-13This unit root test is based on a series of sub-samples which span about one quarter of the total sample and comprise data fromt  1 tok,t  2 tok+ 1, untilt  ntoTwithkequal to months 37 for our study. The recursive minsis found as the smallestt(k/T) over all sub-samples. To reject the null-hypothesis, the calculated mins-value which has to be smaller than the critical value.
14
In particular, a series of regressions is run on the full sample so that dierent k can be chosen in accordance with Banerjee et al.Õs suggestion to movekthrough the mid-70% of the total sample. Thus, for our sample,kranges from month 22 to 124. Similar to the recursive test, a minimums
value is calculated as the smallestt(k/T) over all sub-samples. 15
periods for which the complete co-integration analysis will be conducted. In this way we hope to shed some light on the changes in the European ®nancial market integration process over the past decade.
4. European intra-regional and inter-regional ®nancial integration
4.1. Trends in interest rates and spreads
In order to analyse the degree of European and global ®nancial market integration, our analyses will be performed for six core EU countries, Japan, and USA. Table 1 presents the prime lending and money market rates as well as the nominal and relative spreads on which these analyses are based. De-scriptive statistics are calculated for the full sample period ranging from January 1985 to January 1997, for the pre-integration period of January 1985 to December 1990, and the post-integration period of September 1993 to
January 1997 separately.16
Focusing on the EU countries reveals dierent levels of average national money market rates ranging from a low Germany of 5.89% to a high in Italy of 11.66% during the full sample period. For prime lending rates a similar situ-ation presents itself with a range including German and Dutch average rates of about 8.2% at the low end of the scale and Italian average rates of 12.86% at the high end of the scale.
Comparing the pre- and post-integration period interest rates can provide some initial insights into the claims of the Cecchini study that ®nancial market integration should provide consumer surpluses. Thus, if Cecchini is correct, one expects to see a reduction in nominal interest rates to the level of the country with the lowest pre-integration interest rate. For prime as well as money market rates, a general reduction in interest rates can be observed be-tween the pre- and post-integration period in Table 1. This however seems to be in line with a global reduction in interest rates as the rates for the USA and Japan show. In the post-integration period, national interest rates appear closer together, one reason for which could be the overall lower level of interest rates. The data do, however, not reveal equal national interest rates as money market rates still dier by as much as 4.97 percentage points and prime rates by as much as 4.28 percentage points between the extremes of Italy and the Netherlands.
16
Table 1
Statistics on money market rates, commercial bank prime lending rates, and spreadsa
Country Belgium France Germany Italy Nether-lands
UK USA Japan
Money market rates 1:1985±12:1996
Mean 6.78 8.01 5.89 11.66 6.19 9.51 6.17 4.09 S.D. 2.03 2.10 2.07 2.19 2.00 3.32 1.93 2.23 1:1985±12:1990
Mean 6.82 8.70 5.33 12.69 6.18 11.81 7.76 5.07 S.D. 1.49 1.14 1.58 1.63 1.37 2.16 1.03 1.52 9:1993±1:1997
Mean 4.94 5.38 4.56 9.17 4.20 5.54 4.81 1.39 S.D. 1.68 1.37 1.13 1.05 1.17 0.70 1.00 0.88
Prime rates 1:1985±12:1996
Mean 10.74 9.37 8.16 12.86 8.17 10.76 8.47 4.34 S.D. 2.07 1.26 1.96 2.06 2.16 3.20 1.51 1.82 1:1985±12:1990
Mean 10.99 10.11 7.58 14.03 8.09 12.95 9.45 4.72 S.D. 1.74 0.71 1.57 1.54 1.50 2.21 1.11 1.37 9:1993±1:1997
Mean 8.79 7.65 7.10 10.42 6.14 7.01 7.86 2.42 S.D. 1.64 0.64 1.38 0.95 1.38 0.53 1.05 0.71
Nominal spreads 1:1985±12:1996
Mean 3.96 1.36 2.28 1.21 1.99 1.25 2.31 0.25 S.D. 0.63 1.13 0.59 0.94 0.38 0.69 0.70 0.73 1:1985±12:1990
Mean 4.17 1.41 2.25 1.34 1.92 1.14 1.68 )0.35
S.D. 0.71 0.68 0.65 0.71 0.42 0.87 0.43 0.48 9:1993±1:1997
Mean 3.86 2.28 2.54 1.25 1.94 1.47 2.95 1.02 S.D. 0.39 0.82 0.43 1.03 0.32 0.50 0.16 0.21
Relative spreads 1:1985±12:1996
Mean 1.65 1.23 1.45 1.11 1.35 1.16 1.45 1.39 S.D. 0.26 0.25 0.21 0.09 0.13 0.12 0.27 0.83 1:1985±12:1990
Mean 1.64 1.17 1.47 1.11 1.33 1.10 1.22 0.94 S.D. 0.17 0.10 0.21 0.06 0.11 0.09 0.06 0.08 9:1993±1:1997
An analysis of spreads shows that in all countries except Belgium and Italy nominal spreads have decreased between the pre- and post-integration period which in itself does not provide support for consumer bene®ts from ®nancial market integration since at the same time nominal interest rates are falling. Focusing on relative spreads, however, reveals that spreads have actually in-creased rather than dein-creased over time. In the pre-integration period, relative spreads range from a high of 1.64 in Belgium to a low of 1.11 in Italy and 1.10 in the UK whereas in the post-integration period spreads have increased to a range from 1.87 in Belgium and 1.15 in Italy. Part or all of this increase in relative spreads can be explained by the reduction in interest rates. If prime rates are seen as a function of money market rates plus a mark-up, then relative spreads de®ned as prime rates divided by money market rates will increase as the money market rate decreases. We can, however, observe that in the post-integration period relative spreads seem to move closer together towards the German level in all studied EU countries except Belgium, thus providing some preliminary indication for convergence. Overall, the descriptive statistics pre-sented here do not provide any support for the consumer bene®ts advocated by the Cecchini study but ± at best ± give a small indication for closer alignment of spreads across countries. These observations are also illustrated by the graphs provided in Appendix A.
4.2. Unit roots in interest rates and spreads
As a ®rst step in the co-integration analysis, we now have to establish that the prospective candidate time series for co-integration are integrated of order
1, denoted asI(1), i.e. the time series are random walks. Therefore we test for
unit roots in both, levels and ®rst dierences. As shown by the level regressions in Table 2, for all interest rates we clearly fail to reject the hypotheses that the time series are not random walks regardless whether we employ unit-root tests that allow for the presence of a structural break of unknown timing or not. For
individual countryÕs series, only twice does a test-statistics indicate rejection of
the null hypothesis.17Since we are furthermore able to reject the random walk
hypotheses for the ®rst dierences, we conclude that the interest rates are
in-deedI(1) and thus suitable for co-integration analyses.
Applying the unit roots tests to the spreads, here calculated as the prime lending rate divided by the money market rate, Table 2 gives the interesting
result that we can in general establishI(1) for all countries. Only for Italian,
Table 2 Unit root testsa
Country Levels First dierences
Recursive Mean shift Trend shift Recursive Mean shift Trend shift
) ) ) ) ) ) ) )
aThis table contains the results of unit root tests for the dierent national interest rate series as well as for several average series. EU6 for example is the
weighted average series including all six EU member countryÕs series, whereas EUB, EUF, EUG, EUI, EUN, and EUUK represent the weighted
average series excluding Belgium, France, Germany, Italy, the Netherlands, and the UK, respectively.t(b) andFgive unit-root test-statistics which
ignore the possible presence of a structural break. The remaining test statistics allow for the presence of a structural break of unknown timing. The
critical values for 100 observations are as follows:)3.46 (1%),)2.88 (5%),)2.57 (10%) for thet(b) test;)8.73 (1%), 6.49 (5%), 5.47 (10%) for theFtest;
)4.62 (2.5%),)2.88 (5%),)2.57 (10%) for the recursive minstest;)5.07 (2.5%),)4.80 (5%),)4.54 (10%) for the mean-shift minstest; 20.83 (2.5%),
18.62 (5%), 16.20 (10%) for the mean-shift maxFtest;)4.76 (2.5%),)4.48 (5%),)4.20 (10%) for the trend-shift minstest; and 16.30)4.80 (5%), 13.64
(10%) for the trend-shift maxFtest.
Dutch, and British spreads do some selected test-statistics indicate I(0). 18 Earlier statistical analyses not reported here has revealed that for the 1985±
1991 period the result ofI(0) is con®rmed for Italy, the Netherlands, and the
UK and marginally also for Belgium and Japan. In the post-1992 period
spreads can easily be classi®ed asI(1) for all countries.
Generally speaking, a linear combination of twoI(1) series with unit
coe-cients, such as spreads, could be expected to beI(0), i.e. innovations will have
only temporary eects on the value of the spreads and both interest rates could not drift too far apart, thus ¯uctuating around a ``normal level'' of a long-run equilibrium. It appears, however, that the sluggishness of the prime rates to adjust to changing condition in the money market might have been contributing
to an I(1)-ness of the spreads in most countries under scrutiny. Technically
speaking, co-integration between money market rate and prime lending rate is not too strong. An inspection of the co-integration relationship between both rates (shown in Table 7 in Appendix A) reveals that in those countries where co-integration between both rates is strong and the co-co-integration parameter takes
a value close to 1, we had more problems in establishingI(1). The reader may
also inspect the ®gures of the spread time series in Appendix A, which clearly
show this sluggish adjustment in spreads as well as the ``I(1)-ness'' of the series.
Without driving the point here too far, one might imagine a kind of ``institu-tional convergence'' in the recent years within Europe in terms of retail price
setting and may think of usingI(1)-ness as a rough indicator for price stickiness
in lending and the relative eciency of competition. All this is, however, beyond the scope of our paper and opens another agenda for future research.
4.3. Testing for co-integration in ®nancial markets
If prime rates and spreads are I(1), are they then at least somehow tied
together, so that they cannot drift too far apart? In other words are they co-integrated, and if so, is there any evidence in favour of a regime shift that is consistent with the implementation of the 2BD? We start the analysis by in-vestigating the possibility of structural breaks that we do expect to occur somewhere in the ®rst half of 1990s. On top of the potential impact of the 2BD, the presence of such a break in that period is the more likely since major events like the German uni®cation and the EMS crisis had a strong impact on Eu-ropean ®nancial markets, exchange rates, and exchange rate expectations.
Table 3 presents the results of the structural break test for the co-integration relationship of spreads, prime rates and money market rates. For all series a
18For Italian spreads thet(b) test and the recursive min
structural break could be detected during the sample period. Overall, it appears that the breaks in the money market rates occurred the earliest during the late 1980s and beginning 1990s. Due to the sluggishness of prime rate adjustments it is not surprising that prime rate series show structural breaks shortly after-wards but still in the early 1990s. Spreads, however, exhibit breaks only in the years between 1993 and 1995 with the only exception of Germany, which probably could be considered to be the market leader. Overall, the timing of the breaks leads us to the conclusion that the European ®nancial markets underwent major changes at the beginning of the 1990s. This pattern is con-sistent with the adoption and implementation of the 2BD in 1988 and 1993, respectively. Thus, it appears justi®ed to analyse the European ®nancial market in two distinct time periods: The ®rst period ranges from January 1985 until December 1990 and re¯ects the pre-integration stage of the European ®nancial Table 3
Structural breaks in the co-integration relationship between national and European interest rates (January 1985±January 1997)
Country Series Co-integration vector supF Breakpointa
a bb
Belgium Spreads )0.47 1.75 29.45 March 1995
Prime rates )0.45 1.12 39.44 May 1991
Money market rates )0.59 0.88 28.66 December 1990
France Spreads )1.18 1.96 76.05 June 1994
Prime rates 3.10 0.62 126.53 August 1988 Money market rates )0.12 0.97 76.03 December 1992
Germany Spreads )0.13 1.34 65.80 May 1988
Prime rates 3.41 0.44 163.23 January 1990 Money market rates 0.66 0.56 182.67 May 1990 Italy Spreads 0.73 0.30 24.72 April 1994
Prime rates 5.98 0.74 76.34 August 1988 Money market rates 6.03 0.76 75.36 October 1986 Netherlands Spreads 0.48 0.72 28.28 February 1995
Prime rates )2.61 1.07 110.26 February 1991
Money market rates )1.45 0.91 74.10 November 1990
UK Spreads 0.85 0.25 59.39 February 1993 Prime rates )3.23 1.42 150.07 August 1991
Money market rates 0.43 1.13 191.26 August 1991 Japan Spreads )4.23 4.61 459.91 March 1995
Prime rates )5.44 0.98 32.17 March 1990
Money market rates )4.23 1.00 77.62 December 1991
USA Spreads 1.73 0.23 228.54 August 1991 Prime rates 5.71 0.28 111.46 November 1991 Money market rates 4.21 0.24 196.03 January 1991 a
Based on the critical values reported by Hansen (1992) all of the above listed structural breaks are signi®cant.
b
markets. The second period ranges from September 1993 until January 1997 and re¯ects the post-integration integration stage of the European ®nancial markets. The period between December 1990 and September 1993 is excluded to provide suciently separated time periods and to eliminate the eects of the European currency crisis in the summer of 1993.
Starting now with the co-integration analyses by looking at prime rates, we
could not detect co-integration of prime rates of individual countries vis-a-vis
the rest of the EU in the full sample judged by both, the DW statistics and the DF statistics as presented in Table 4. This remains true when we divide the
sample in pre- and post-integration periods.19 Reasons for this lack in
integration could be: (1) a lack of integration in lending, (2) a lack of co-integration in the money market, i.e. diverging money market rates, or (3) a Table 4
Co-integration between national and EU average money market rates and prime ratesa
Country Belgium France Germany Italy Nether-lands
UK USA Japan
Money market rates 01:1985±12:1996
DW 0.424 0.376 0.029 0.156 0.185 0.052 0.023 0.093 DF )4.143 )3.775 )1.361 )2.863 )3.237 )1.356 )1.192 )1.929
01:1985±12:1990
DW 1.178 0.352 0.149 0.193 0.589 0.280 0.206 0.172 DF )5.606 )2.898 )1.753 )2.528 )3.946 )2.606 )1.488 )1.620
09:1993±01:1997
DW 0.118 0.683 0.052 0.114 0.103 0.407 0.055 0.109 DF )3.082 )2.862 )1.904 )0.659 )1.355 )2.121 )0.960 )1.046
Prime rates 01:1985±12:1996
DW 0.182 0.129 0.022 0.071 0.065 0.065 0.031 0.109 DF )2.874 )3.000 )0.641 )2.929 )2.225 )1.813 )1.388 )1.973
01:1985±12:1990
DW 0.419 0.163 0.080 0.062 0.225 0.303 0.090 0.216 DF )3.964 )3.556 )0.191 )2.561 )2.363 )4.081 )1.420 )0.723 09:1993±01:1997
DW 0.155 0.260 0.042 0.132 0.134 0.126 0.061 0.117 DF )1.341 )1.972 )0.618 )1.175 )1.009 )1.120 )0.776 )1.044
a
For EU countries, the national interest rate series are regressed against the EU average excluding the country under investigation. For the USA and Japan, the US or Japanese rates are regressed against the full EU average. DW gives the Durbin±Watson statistics. The critical values for re-jecting the null hypothesis of non-co-integration are 0.511 (1%), 0.386 (5%) and 0.322 (10%) for 100 observation. DF gives thet-statistics for/in the simple DF regression. The critical values for 100 observations are 4.07 (1%), 3.37 (5%), and 3.03 (10%).
lack of co-integration between money market and lending rates. As far as the last point is concerned, our analysis of co-integration of lending and money market rates in Table 7 in Appendix A has revealed that in most cases we cannot reject the co-integration hypothesis, however, the relation is less than perfect and unitary co-integration vectors are not the rule. While this point may partly be blamed for the ®nding, it is unlikely that it bears the full re-sponsibility for it.
We therefore examine co-integration in the money market. Again, there appears to be a lack of co-integration among European countries with a few marginal exceptions as Table 4 shows. It should, however, be recalled that the
countriesÕ money market rates have been regressed against an EU average.
Since exchange rate expectations are militating against interest parity, the ®nding is plausible. As a crosscheck we regress the money market rate against the German money market rates the results of which are presented in Table 5. Here we ®nd co-integration between German money market rates and Bel-gium, Dutch, and British money market rates in both sub-periods, with a
Table 5
Bi-lateral co-integration between national and German interest ratesa
Country Belgium France Italy Netherlands UK
Money market rates 01:1985±12:1996
DW 0.432 0.169 0.108 0.217 0.037
DF )4.556 )2.567 )2.353 )2.691 )1.256
01:1985±12:1990
DW 0.671 0.189 0.186 0.702 0.758
DF )4.178 )2.592 )2.548 )3.899 )3.838
09:1993±01:1997
DW 0.523 0.275 0.115 0.528 0.518
DF )4.027 )1.634 )0.272 )2.503 )2.409
Prime rates 01:1985±12:1996
DW 0.146 0.030 0.039 0.178 0.036
DF )3.283 )1.665 )2.390 )2.602 )1.041
01:1985±12:1990
DW 0.163 0.091 0.048 0.425 0.376
DF )2.674 )3.037 )2.516 )3.153 )3.244
09:1993±01:1997
DW 0.403 0.194 0.152 0.492 0.149
DF )2.750 )1.219 )1.288 )2.268 )1.212
Table 6
Co-integration between national and European spreadsa
Country Period Co-integration vector DW ADF (k) u1
a
(t-statistics)
b
(t-statistics)
(t-statistics)
Belgium 01:1985±01:1997 0.47 1.75 0.949 )6.59 (0) )0.23
()2.85) (12.82) )3.00 (1,2,4) ()2.98)
01:1985±12:1990 0.52 0.94 1.739 )7.35 (0) )0.6
(1.43) (3.07) )3.48 (1,2) ()4.05)
09:1993±01:1997 )1.31 2.38 0.369 )2.44 (0) 0.05
()2.70) (6.56) )0.58
France 01:1985±01:1997 )1.18 1.96 0.369 )3.43 (0) )0.11
()6.01) (12.30) )2.87 (3) ()2.98)
01:1985±12:1990 0.39 0.65 0.382 )2.68 (0) )0.19
(1.95) (3.98) )1.97 (3) ()3.13)
09:1993±01:1997 )2.74 3.18 0.571 )2.64 (0) )0.14
()3.88) (6.02) ()1.83)
Germany 01:1985±01:1997 )0.12 1.34 0.347 )3.48 (0) )0.09
()0.80) (10.04) )2.53 (2) ()2.12)
01:1985±12:1990 )1.93 2.96 0.600 )3.64 (0) )0.04
()4.43) (7.81) ()0.55)
09:1993±01:1997 0.61 0.75 0.511 )2.39 (0) )0.17
(2.38) (3.79) )1.63
Italy 01:1985±01:1997 0.73 0.3 0.465 )4.01 (0) )0.22
(11.58) (6.16) ()3.91)
01:1985±12:1990 0.73 0.31 1.116 )5.02 (0) )0.58
(7.31) (3.79) )4.83(1) ()5.27)
09:1993±01:1997 0.26 0.6 0.230 )1.28 (0) )0.05
(1.04) (3.46) )0.77
Netherlands 01:1985±01:1997 0.48 0.72 1.101 )7.36 (0) )0.2
(4.86) (9.04) )2.59 (1)4) ()2.23)
01:1985±12:1990 1.01 0.26 1.540 )6.57 (0) )0.37
) )
(4.07) (1.60) ()2.16)
UK 01:1985±01:1997 0.85 0.25 0.646 )5.61 (0) )0.14
(8.85) (3.21) )2.34 (1)3) ()2.08)
01:1985±12:1990 0.91 0.15 1.181 )6.11 (0) )0.5
(5.45) (1.14) ()4.70)
09:1993±01:1997 1.92 )0.46 1.177 )4.11 (0) )0.55
(10.59) ()3.55) ()3.56)
Japan 01:1985±01:1997 )4.24 4.61 0.085 )1.09 (0) 0
()6.46) (8.61) )1.06 (1,2) ()0.24)
01:1985±12:1990 0.77 0.14 0.712 )3.88 (0) )0.37
(5.55) (1.25) ()3.91)
09:1993±01:1997 )11.85 10.81 0.372 )1.91 (0) )0.04
()6.87) (8.25) )2.81 (3) ()0.76)
USA 01:1985±01:1997 1.73 )0.23 0.054 )1.40 (0) )0.03
(6.63) ()1.07) )0.26 (1,4) ()1.32)
01:1985±12:1990 1.06 0.13 0.941 )4.39 (0) )0.46
(9.71) (1.52) ()4.19)
09:1993±01:1997 2.77 )0.87 0.166 )1.81 (0) )0.11
)7.1 ()2.93) ()1.97)
a
For EU countries, the national spreads are regressed against the EU average spread excluding the country under investigation. For the USA and Japan, the US or Japanese spreads are regressed against the full EU average spread. The co-integration vector displays the coecient estimates with
their respectivet-statistics in parentheses. DW gives the Durbin±Watson statistics. The critical values for rejecting the null hypothesis of
non-coin-tegration are 0.511 (1%), 0.386 (5%) and 0.322 (10%) for 100 observation. ADF (k) gives thetstatistics for/in the (augmented) DF regression.k, given
in brackets, gives the lags used in augmented DF regression, so thatk0 represents the simple DF regression. The critical values for 100 observations
are 4.07 (1%), 3.37 (5%), and 3.03 (10%) for the DF regression, and 3.77 (1%), 3.17 (5%), and 2.84 (10%) for the augmented DF regression. The error
correction termu1is given in the last column for which alsot-values are reported in brackets.
stronger relationship being present in the pre-integration period. This fact that can be attributed to the Maastricht process and the close tying of exchange rates among the core EU countries. This view is particularly con®rmed by the lower test statistics for France and Italy which indicate no co-integration be-tween these two countries and Germany. As suggested by the money market study by Caporale et al. (1996), we interpret this lack of co-integration as the outcome of a convergence process. The tying together of money market rates by seemingly irrevocable exchange rates translates to some extent into lending rate co-integration. As can be seen again in Table 4, in the case of Belgium where money market rates are integrated in the pre-integration period, prime rates are also integrated in the pre-integration period. In the case of the Netherlands, however, money market integration did not result in lending market integration. As a tentative conclusion we take from these results that convergence in money market rates is a necessary, but by no means a sucient condition for lending rate co-integration.
Convergence in retail banking as envisaged in the Cecchini report is promising lowest possible retail banking prices. Since lending rates are related to the general interest level prevailing in a country it is useful to examine to what extent spreads between money market and lending rate are co-integrated. A co-integration of spreads would essentially re¯ect arbitrage forces that tie together the markets towards a long-run equilibrium relationship. Such forces could be money market integration, convergence in retail pricing behaviour, and arbitrage processes in the lending market, be it suppliers or consumers crossing borders. As in all co-integration analyses, however, the driving forces behind this process are not individually investigated. The existence of such an equilibrating process itself is the subject of interest here.
We generally found spreads to be co-integrated in the European sample as can be seen from the results regarding the whole sample period reported in Table 6. If we, however, divide the sample into pre- and post-integration pe-riods we observe a number of structural changes. Generally, in the ®rst sub-period the evidence is overall in favour of the co-integration hypothesis, the only exception being France, which shows only marginal evidence for in-tegration. In the post-integration period we found that the evidence for co-integration is weakening. Looking at the ADF test-statistics, co-co-integration could only be established for the UK in the post-integration period. One has, however, to exercise caution since we had problems to establish that the spreads in Italy, the UK and possibly the Netherlands are random walks
in the pre-period. As expected by the results of the supF test for structural
at a slower pace.20 The ECT for Germany is negative but not signi®cantly dierent from zero. This implies that German spreads are not aected by movements of spreads in the rest of the EU which might in addition to the
evidence in Tables 5 and 8 underline GermanyÕs leading role in setting interest
rates in the EU.
4.4. Co-integration and convergence
The latter result of no co-integration in the post-period is, however, not contradicting the hypothesis of a closer market integration. In fact, when convergence occurs this implies a tendency for spread dierentials between countries to decline. In line with the co-integration and convergence inter-pretation suggested by Caporale et al. (1996), we interpret our results as indicating a loosely connected retail banking market before 1993. This has been particularly true for France, Italy, and the UK, as it is re¯ected in the lack of co-integration as in the case of France or in the low parameter for the ECT as in the case of Italy and the UK. The time pro®le of the residuals from the co-integration regression also reveals that in the post-1993 period convergence occurred, as it was already visible in the descriptive statistics. The average level of spreads to which European economies are converging to seems to be around 1.5 (at the current interest rate level). This implies for quite some economies increasing rather than falling spreads. One should, however, recall from our earlier remark that even in a fully integrated retail banking market spreads need not equalise fully. National dierences in spreads make sense in the presence of diering loan portfolios. In fact, in the presence of uni®ed monetary conditions they are the only means banks have to adjust for dierent portfolio risks. Convergence demands only that spreads must ®nd their new equilibrium level given the same money market rates. Once convergence has occurred, changes in money market rates should be absorbed by lending rates in a similar way, thus allowing the spreads to return to their equilibrium level with the same speed of adjustment every-where. During this adjustment process co-integration should be observable in integrated lending markets.
4.5. Intra-regional versus inter-regional integration
Finally, we contrast the results obtained for Europe with an inter-regional perspective. Speci®cally, we perform the empirical tests we employed for Europe in a global context, i.e. we investigate the ®nancial integration of our
EU6 sample as a whole vis-a-vis the US and Japan, respectively. In this way we crosscheck to what extent the results obtained for European ®nancial in-tegration are a consequence of regional inin-tegration policies or merely a result of the globalisation of ®nancial markets. This seems the more important since inter-regional interest rates have been converging recently, as investigated by the OECD (1996). In fact, however, neither European money market rates nor lending rates are co-integrated with these rates in the USA or Japan, respec-tively. The same is true for spreads, suggesting that integration is predomi-nantly a regional phenomenon. The result of these tests can also shed some light on the relative importance of a multilateral agreement on trade in ®-nancial services under the roof of the World Trading Organisation (see Schott, 1996): When least-cost suppliers are located extra-regionally, the European bank customers could possibly obtain additional gains when preferences given now exclusively to European banks will ± in one way or the other ± be ex-tended to non-European banks, thus intensifying competition in European banking.
5. European ®nancial market integration, the EMU, and the world trading system: A conclusion
With the launch of the single European Monetary Union (EMU) on 1 January 1999 and, consequently, the introduction of a single European mon-etary policy the potential bene®ts from ®nancial market integration go far beyond ``Cecchini gains''. With only one pan-European short-run interest rate the issue arises how fast changes in money market rates will be passed on to-wards borrowers. If credit costs are an important element in the monetary
transmission process,21then convergence in lending rate behaviour will play a
vital role in unifying the impact of a single monetary policy. An increasing
number of recent studies22have all pointed out that an interest rate shock may
have a dierent impact on dierent EMU member countries, suggesting that a single currency and monetary policy cannot suit every member in the same way. For example, Ramaswany and Sloek (1997) found that in Austria, Bel-gium, Finland, Germany, the Netherlands, and the UK a contractionary monetary shock aects output almost twice as much, but takes twice as long to occur than in a second group of countries comprising Denmark, France, Italy,
21
For a discussion see the various contributions to the ÔSymposium on the Monetary Transmission ProcessÕin the Journal of Economic Perspectives, Vol. 9, No. 4, Fall 1995.
Portugal, Spain, and Sweden. It is also noteworthy that this grouping of countries is orthogonal to the traditional grouping of countries into core and non-core EMU countries according to optimum currency area exercises in the
spirit of MundellÕs (1961) seminal analysis, with e.g. France now grouped
to-gether with southern European countries. Other studies come to similar con-clusions, although the country list diers somewhat from the one given above. The crucial dierence between these studies and the Mundell optimum cur-rency area analyses is that the latter suggests that under asymmetric shocks one country may need a dierent dose of monetary policy than another, while here it is suggested that the same dose of that policy could have dierent eects in dierent member countries.
While it is clear that some of the dierences in the impact of monetary policy on output are rooted in goods market behaviour ± in particular how demand shock are transformed into price and output eects in the short run and long-run ± the workings of the ®nancial markets may well play an important role. Although it is controversial to what extent the cost of borrowing is a major transmission channel of monetary policy, a mounting body of research shows that lending rates adjust with a dierent speed to changes in ocial interest
rate.23 While the dierences in stickiness of lending rates are therefore well
reported for the past, there is a lack of investigations that take into account the structural changes in competition in retail banking that are supposed to have occurred recently and may have been intensi®ed in the presence of the Euro. Failure to take the impact of these changes into account may lead to wrong policy conclusions. Although we recognize that the database is not yet large enough to make secure inference from recent statistical evidence, our study provides initial evidence on the convergence process in retail banking. Full integration in retail banking does, however, not mean equalization of lending rates since the loan portfolios in dierent countries will dier in risk. In the presence of a uni®ed monetary policy, interest spreads must ®nd their new equilibrium level. Once this level has been found, lending rates in Europe should move in tandem. Therefore, the results of our study can be interpreted in the context of a ``®fth'' or ``forgotten'' convergence criteria in the Maastricht treaty: To what extent has ®nancial market integration already laid the foun-dation for a smooth functioning of a monetary policy under the condition of a single European monetary policy?
23
Regarding this and the other questions that our study attempts to answer, our results indicate that ®nancial integration seems to change the pattern of European banking. While some degree of co-integration has been there before the 1992 project has commenced, the arbitrage process was quite limited in its eectiveness. Thereafter, the evidence provided here seems to be consistent with the view that the structure of European banking is now changing rapidly and convergence is occurring. It should, however, be noted that during the process when convergence occurs, co-integration tests are of limited use. They are most informative after convergence has happened, so that a repetition of the exercise at a later time would be most useful.
Having said that convergence is occurring, three channels of arbitrage have been identi®ed: Convergence of money market rate, lending arbitrage across borders, and ``institutional convergence'' of pricing behaviour. The ®rst mechanism is best being promoted by irrevocably ®xed exchange rate or a single European Currency. It is in this respect, that the Euro could contribute to a ``completion of the single European market'' also in retail banking. Likewise, elimination of exchange rate risks makes cross-border lending and
borrowing more attractive, especially from the consumerÕs side, although
consumers are less likely to shop around for credits than for goods. Given structural dierences in lending activities it would be too farfetched to expect equal lending rates throughout Europe. It would equally be too farfetched to expect that retail banking prices will converge to the lowest level within the EU. Convergence of spreads has not necessarily been going into that direction.
Integrating ®nancial markets is in itself not a guarantee that a perfectly competitive market will be established. The fact that spreads are often non-stationary suggests that market imperfections, in particular informational imperfection, continue to prevail. Market integration alone cannot assure lowest possible prices and the expected subsequent dynamic economic growth eects. If the European ®nancial marketplace will succeed in deepening ®-nancial integration within Europe, the bene®ts may be tremendous, but the do not come automatically and continue to require regulatory eorts. Moreover, since regional integration scheme are potentially discriminatory, inter-regional integration is rarely visible, and market imperfection are intrinsic in ®nancial markets, extending integration eorts into a multilateral setting under the
umbrella of the World Trade OrganisationsÕtalks on trade in ®nancial services
could result in intensi®ed competition and thus additional ''bene®ts beyond Cecchini''.
Appendix A
Co-integration between national prime and national money market ratesa
Country Period Co-integration vector DW ADF (k) u1
a
(t-statistics)
b
(t-statistics)
(t-statistics)
Belgium 01:1985±01:1997 4.14 0.97 1.196 )7.94 (0) )0.24
(22.69) (37.69) )6.00 (1,4) ()3.61)
01:1985±12:1990 3.71 1.07 1.645 )7.03 (0) )0.28
(9.43) (18.94) ()3.89)
09:1993±01:1997 4.11 0.95 1.693 )5.66 (0) )0.07
(21.76) (26.13) )2.38
France 01:1985±01:1997 5.08 0.54 0.347 )4.02 (0) )0.08
(26.64) (23.26) )3.16 (2) ()2.86)
01:1985±12:1990 5.66 0.51 0.314 )2.96 (0) )0.27
(15.31) (12.14) )2.19 (3) ()6.19)
09:1993±01:1997 5.35 0.43 0.876 )3.20 (0) )0.18
(31.79) (14.12) ()1.73)
Germany 01:1985±01:1997 2.80 0.91 0.325 )3.34 (0) )0.26
(19.77) )40.07 ()6.83)
01:1985±12:1990 2.73 0.91 0.315 )2.46 (0) )0.27
(10.23) (18.92) ()5.40)
09:1993±01:1997 1.73 1.18 0.479 )2.06 (0) )0.21
(6.86) (21.84) )2.77 (1,4) ()2.19)
Italy 01:1985±01:1997 2.95 0.85 0.551 )4.88 (0) )0.14
(7.37) (25.23) )4.54 (4) ()3.29)
01:1985±12:1990 3.24 0.85 0.931 )4.44 (0) )0.16
(5.17) (17.33) ()3.39)
09:1993±01:1997 6.50 0.43 0.116 )1.11 (0) )0.14
Table 7 (Continued)
Country Period Co-integration vector DW ADF (k) u1
a
(t-statistics)
b
(t-statistics)
(t-statistics)
Netherlands 01:1985±01:1997 1.61 1.06 1.208 )7.88 (0) )0.21
(16.52) (70.68) )5.85 (1) ()3.13)
01:1985±12:1990 1.60 1.05 1.407 )6.07 (0) )0.11
(6.93) (28.84) ()1.01)
09:1993±01:1997 1.30 1.15 0.742 )2.97 (0) )0.09
(8.18) (31.62) ()0.50)
UK 01:1985±01:1997 1.81 0.94 1.069 )8.28 (0) )0.44
(10.72) (56.12) ()4.12)
01:1985±12:1990 1.83 0.94 1.045 )5.93 (0) )0.33
(3.21) (19.87) ()4.70)
09:1993±01:1997 4.08 0.53 0.623 )2.72 (0) )0.66
(8.42) (6.12) )3.20 (1) ()9.44)
Japan 01:1985±01:1997 1.16 0.78 0.278 )3.33 (0) )0.08
(12.39) (38.82) )2.04 (1,2) ()2.60)
01:1985±12:1990 0.37 0.86 0.597 )3.37 (0) )0.03
(2.06) (25.53) ()0.40)
09:1993±01:1997 1.30 0.8 1.572 )5.27 (0) )1.19
(35.97) (36.27) )5.51 (3) ()14.95)
USA 01:1985±01:1997 3.91 0.74 0.324 )3.64 (0) )0.08
(28.63) (34.97) )2.73 (2) ()2.29)
01:1985±12:1990 1.72 1.00 0.861 )4.36 (0) )0.32
(4.44) (20.12) )2.00 (2) ()3.90)
09:1993±01:1997 2.77 1.04 1.699 )5.37 (0) )1.18
(22.75) (42.67) ()10.84)
aFor each country, the national prime rate is regressed against the national money market rate. For further comments and explanations see Table 6.
Bilateral co-integration between national and German spreadsa
Country Period Co-integration vector DW ADF (k) u1
a
(t-statistics)
b
(t-statistics)
(t-statistics)
Belgium 01:1985±01:1997 0.62 0.71 0.721 )5.27 (0) )0.21
(5.04) (8.51) )2.09 (1,2,4) ()3.22)
01:1985±12:1990 1.37 0.18 1.617 )6.94 (0) )0.85
(9.72) (1.90) )4.65 (1) ()7.30)
09:1993±01:1997 )0.35 1.40 0.418 )1.55 (0) 0
()0.88) (5.63) ()0.00)
France 01:1985±01:1997 0.18 0.72 0.257 )2.22 (0) )0.09
(1.63) (9.36) )1.04 (2) ()2.59)
01:1985±12:1990 0.80 0.26 0.394 )2.72 (0) )0.28
(11.28) (5.41) )3.67 (1,4) ()3.90)
09:1993±01:1997 )0.52 1.28 0.431 )1.48 (0) )0.16
()1.31) (5.07) ()1.96)
Italy 01:1985±01:1997 0.84 0.19 0.461 )3.78 (0) )0.21
(17.06) (5.50) ()3.90)
01:1985±12:1990 0.92 0.13 1.106 )4.92 (0) )0.69
(20.72) (4.27) )4.76 (1) ()5.26)
09:1993±01:1997 0.67 0.30 0.206 )0.834 (0) )0.06
(3.07) (2.16) ()0.97)
Netherlands 1:1985±01:1997 0.88 0.33 0.840 )6.05 (0) )0.18
(13.27) (7.29) )1.91 (1-4) ()2.44)
01:1985±12:1990 1.16 0.11 1.500 )6.43 (0) )0.39
(12.80) (1.86) )1.60 (1-4) ()2.27)
09:1993±01:1997 0.82 0.42 0.762 )2.91 (0) )0.19
(4.44) (3.64) ()1.66)
UK 01:1985±01:1997 0.95 0.14 0.632 )5.59 (0) )0.14
(13.86) (3.04) )2.29 (1)3) ()2.18)
01:1985±12:1990 0.93 0.11 1.227 )6.33 (0) )0.65
(13.54) (2.47) ()5.98)
09:1993±01:1997 1.7 )0.27 1.031 )3.65 (0) )0.29
(7.95) ()1.97) ()1.71)
a
For each country, the national spread is regressed against the German spread. For further comments and explanations see Table 6.
References
Andrews, D.W.K, 1993. Tests for parameter instability and structural change with an unknown change point. Econometrica 61, 821±856.
Baldwin, R.E., 1989. The growth eects of 1992 economic policy. A European Forum 9, 248±281. Banerjee, A., Lumsdaine, R.L., Stock, J.H., 1992. Recursive and sequential tests of the unit-root and trend-break hypotheses: Theory and international evidence. Journal of Business & Economic Statistics 10, 271±287.
Barran, F., Coudert, V., Mojon, B., 1996. The transmission of monetary policy in the European countries. CEPII Working Paper 96-03.
Bredemeier, S., 1995. Integration within the banking sector, In: Lang, F.P., Ohr, R. (Eds.), International Economic Integration. Physica Verlag, Heidelberg, pp. 159±180.
Britton, E., Whitley, J., 1997. Comparing the monetary transmission mechanism in France, Germany, and the UK: Some issues and results. Bank of England 37(2), May.
Canal, J., 1993. Competitive Strategies in European Banking. Oxford University Press, Oxford. Caporale, G.M., Kalyvitis, S., Pittis, N., 1996. Interest rates convergence, capital controls, risk
premia, and foreign exchange market eciency in the EMS. Journal of Macroeconomics 18, 693±714.
Clark, J.A., 1988. Economics of scale and scope at depository institutions: A review of the literature. Federal Reserve Bank of Kansas City Economic Review, September±October, pp. 16±33.
Commission of the European Communities, 1988. European economy: The Economics of 1992, No. 35, March, Commission of the European Communities, Brussels.
Cotarelli, C., Ferri, G., Generale, A., 1995. Bank lending rates and ®nancial structure in Italy: A case study. IMF Working Paper 95/38.
Cotarelli, C., Kourelis, A., 1994. Financial structure, bank lending rates, and the transmission mechanism of monetary policy. IMF Sta Papers 41, No. 4.
Dermine, J. (Ed.), 1993. European Banking in the 1990s. Blackwell Publishers, Oxford.
Diebold, F.X., Chen, C., 1996. Testing structural stability with endogenous breakpoint ± a size comparison of analytic and bootstrap procedures. Journal of Econometrics 70, 221±241. Dornbusch, R., Favero, C., Giavazzi, F., 1998. A red letter day? CEPR Discussion Paper No.1804. Eades, M., 1994. Can there ever be a truly commercial bank? Banking World 13±18.
Engel, C.M., 1995. Accounting for U.S. real exchange rate changes. NBER Working Paper No. 5394.
Engel, C.M., 1996. Long-run PPP may not hold after all. NBER Working Paper No. 5646. Engle, R.F., Granger, C.W.J., 1987. Co-integration and error correction: Representation
estimation and testing. Econometrica 55, 251±276.
Froot, K.A., Rogo, K.S., 1995. Perspectives on PPP and long-run real exchange rates. NBER Reprint No. 2049.
Froot, K.A., Kim, M., Rogo, K.S., 1995. The law of one price over 700 years. NBER Working Paper No. 5132.
Gardener, E.P.M., Teppett, J.L., 1995. A select replication of the Cecchini microeconomic methodology on the EFTA ®nancial services sectors: A note and critique. The Service Industries Journal 15, 74±89.
Gerlach, St., Smeets, F., 1995. The monetary transmission mechanism: Evidence from the G-7 countries. CEPR Discussion Paper 1219.
Gregory, A.W., Nason, J.M., Watt, D.G., 1996. Testing for structural breaks in cointegrated relationships. Journal of Econometrics 71, 321±341.