Interest rate setting by universal banks and the monetary policy transmission mechanism in the euro area pot

Interest rate setting by universal banks and the monetary policy transmission mechanism in the euro areaGabe de Bondt, Benoît Mojon and Natacha Valla* 6 November 2002 Preliminary Draft A

Interest rate setting by universal banks and the monetary policy transmission mechanism in the euro area

Gabe de Bondt, Benoît Mojon and Natacha Valla*

6 November 2002

Preliminary Draft Abstract

This paper empirically analyses the pass-through of changes in market interest rates to retail bankinterest rates in euro area countries The results confirm earlier findings that retail bank rates adjustsluggishly to market rates Differences in the degree of this pass-through persisted after theintroduction of the euro, both across the five segments of the retail bank markets analysed and acrossthe ten euro area countries considered First, the sluggishness is not generally due to an ability of

“universal” banks to fund loans by deposits rather than securities Second, estimation results of linearand state dependent error correction models show that retail bank interest rates adjust to changes infunding (lending rates) or opportunity costs (deposit rates), which are approximated by a weightedaverage of short and long-term market interest rates Furthermore, it is found that the introduction ofthe euro has speed up the adjustment of retail bank interest rates to market interest rates One possiblefactor behind this evolution is in the evolution could be related to competitive forces in the differentsegments of the retail bank market

Keywords: retail bank interest rates; market interest rates; euro area countries

JEL classification: E43; G21

European Central Bank, Kaiserstrasse 29, D-60311, Frankfurt am Main, GERMANY E-mail addresses:

gabe.de_bondt@ecb.int, benoit.mojon@ecb.int, and natacha.valla@ecb.int We thank Jesper Berg, Francesco Drudi, Michael Ehrmann and Oreste Tristani for their reflexions on a previous draft, Hans-Joachim Klockers for his comments and Rasmus Pilegaard for excellent data assistance All views expressed are those of the authors alone and do not necessarily reflect those of the ECB or the Eurosystem.

1 Introduction

The level of interest rates is one of the main determinants of savings and investment decisions Whenmaking these decisions, euro area households and firms are mainly confronted with retail bank interestrates In 2000, the amounts outstanding of loans to non-financial corporations of the euro area wereseven times as large as debt securities Moreover, traditionally deposits are larger than money mutualfunds (Agresti and Claessens, 2002; ECB’s Report on Financial Structures, 2002) It is also clear thatthe response of bank retail rates to changes in the interest rates on the refinancing operationscontrolled by the central bank is a major link in the transmission of the ECB monetary policy

A growing literature has shown the sluggishness of retail banks interest rates in the euro area Asshown in Table 1 (reproduced from De Bondt 2002), the complete pass-through from changes in themoney market rates to retail bank rates takes at least several months1 These results are usually based

on reduced form regressions of retail bank rates on the money market rate While this modellingapproach provides a good summary evaluation of this sluggishness, it falls short of explaining itsdeterminants To remedy this gap, this paper estimates semi-structural equations of interest rate setting

by euro area banks, which we see as being in their majority “universal” 2 Our study covers fivecategories of euro area retail bank rates, four loans (short-term and long-term loans to firms,mortgages and consumer credit), and one time deposits interest rates We use time series of retail ratesand market rates from ten of the twelve euro area countries as well as for the euro-area aggregate

“Universal” banks, which we define by opposition to “specialised” banks, should in principle enjoyeconomies of scope In particular, the rates on loans granted by universal institutions may depend onthe cost of raising deposits rather than issuing securities Such a deposit-based funding of loanactivities could imply that retail bank rates remain little responsive to market conditions once depositrates are accounted for On the contrary, specialised banks without branches collecting deposits wouldset their retail loan rates on the basis of their market-based funding

Within our framework, this ability of continental European banks to avoid market funding usingdeposits instead is one possible explanation for the retail rates sluggishness

“narrow” definition of universal banking and refer to the set of “traditional” retail banking activities.

Another explanation for the sluggishness of retail bank rates is that funding costs are not entirelyindexed to the money market rate First, banks may try to limit interest rate risk on long-term loans byincreasing the maturity of the funding of such loans Second, in the presence of adjustment/menucosts, uncertainty about the persistence of changes in money market rates may induce banks to define

a target retail rate as a function of long-term rates, as a smooth indicator of future changes in moneymarket rates

Our initial tests reject the idea that retail bank rates on deposits have, conditionally on the level ofmarket interest rates, a significant influence on retail bank rates on loans We then show that thedynamics of each retail bank interest rate can be specified separately within an error correction model(ECM) In the long run, banks set their retail prices in line with their marginal costs, i.e the fundingcosts of loans and the opportunity costs of deposits Both the funding and opportunity costs aremodeled as a weighted average of the three-month money market rate (MMR) and the 10-yeargovernment bond yield (BR) This way, the marginal cost of retail bank instruments are moreaccurately captured than in previous studies, which typically examined only the short-term marketinterest rate Nevertheless, it can not be ruled-out that our empirical findings are distorted by maturitymismatches or yield curve effects, since it is unclear whether our freely estimated weighted average ofthe short and long-term interest rates have a comparable maturity with the underlying retail bankinstrument.3 In the short run, changes in retail bank rates depend on changes in the MMR and in the

BR and on the deviation from the long-run equilibrium relationship between the retail bank interestrate and short- and long-term interest rates

The stability of the baseline linear ECM before and after the introduction of the euro is then tested and

we assess whether more general state dependent models are preferable to the linear specification Abreak in January 1999 is found in the estimated linear ECM in about half of the cases One possibleexplanation for this break may be associated to the evolution of the competitive forces in the differentsegments of the retail bank market since the introduction of the euro For instance, the time depositand mortgage markets have seen some new entrants, in particular internet banks In addition, for non-financial corporations, non-bank sources of finance, in particular debt securities, have increased inrecent years (De Bondt, 2002c)

This paper has three key contributions First, it shows that the sluggishness in the response of retailrates to market rates is not due to the possibility of banks to fund their loans through the issuance ofdeposits Second, it shows that retail bank interest rates in euro area countries adjust to changes inmarginal funding or opportunity costs, approximated by a weighted average of short and long-term

3 At the euro area level De Bondt (2002a) examines the adjustment of retail bank interest rates to market interest rates with a comparable maturity.

interest rates The weight of market rates in the equilibrium target of retail bank rates, which has beenlargely ignored in the literature on the pass-through, is found to be an important factor behind thesluggishness of the response of bank rates to market rates Third, we find that in a large number ofnational retail markets, the introduction of the euro has affected bank pricing.

The paper is structured as follows Section 2 presents available evidence on interest rates pass-throughprocess in individual euro area countries The theoretical model of bank pricing is presented in section

3 Section 4 provides an overview of the data Section 5 discusses the empirical results and Section 6concludes

2 The pass-through to retail bank interest rates: literature review(s)

While recent studies on the retail bank interest rate pass-through stress the sluggishness in theadjustment of retail bank rates, they usually do not provide explanations for it Against thisbackground, we discuss how the complementarities across bank’s activities can explain the lack ofresponsiveness of bank rates to market conditions

2.1 An incomplete interest rate pass through in individual euro area countries

Table 1 summarises the main findings of interest rate pass-through studies performed for individualeuro area countries All studies show cross-country differences in the interest rate pass-through,although no clear pattern in those differences seems to emerge

Studies from the mid-1990s broadly show that changes in official and/or money market rates are notfully reflected in short-term bank lending rates to enterprises after three months, but that the pass-through is higher in the long term (BIS, 1994, Cottarelli and Kourelis, 1994, and Borio and Fritz1995) Recent cross-country studies by Kleimeier and Sander (2000 and 2002), Donnay and Degryse(2001), Toolsema et al (2001), and Heinemann and Schüller (2002) confirm this finding Hofmann(2000) and Mojon (2000) also find short-term sluggishness in short-term bank lending rates toenterprises, but assume a priori a complete long-term pass-through

As regards long-term bank lending rates to enterprises and households, all studies, except BIS (1994),typically show that the pass-through tends to be less complete than for short-term bank lending rates toenterprises This finding may be driven by the fact that the funding costs are approximated by moneymarket interest rates, which may not always be the most appropriate marginal funding costs, inparticular for long-term loans to enterprises and mortgages

Furthermore, changes in and convergence of financial structures among euro area countries is apotential determinant of the interest rate pass-through For instance, Mojon (2000) concludes that

deregulation has significantly affected the interest rate pass-through process for deposits, but not forloans.

2.2 Bank studies on the interest rate pass-through

The industrial organisation literature typically examines the link between bank interest rate marginsand the market structure of the banking system using bank data (Hannan and Berger, 1991, Neumarkand Sharpe, 1992, Angbazo, 1997, Hannan, 1997, Wong, 1997, and Corvoisier and Gropp, 2001) Themain lesson of these banking structure studies is that the pricing behaviour of banks may depend onthe degree of competition and contestability in the different segments of the retail bank market Forinstance, Corvoisier and Gropp (2001) conclude that for demand deposits and loans increasing bankconcentration in individual euro area countries during the years 1993–1999 may have resulted in lesscompetitive pricing by banks, whereas for savings and time deposits the opposite seem to be the case

2.3 Loans pricing by universal banks

The vast empirical literature that has tested the existence of economies of scope is largelyinconclusive However, the complementarities across bank activities can explain the sluggishness ofretail bank rates

One factor of retail bank rate sluggishness that has not received much attention is the ability of banks

to exploit the complementarity of their activity In a universal banking environment, the way banks setretail interest rates pertains to the general - and overall inconclusive - debate on complementarities andscope in banking A widespread belief suggests that banks tend to expand the scope (and possiblyscale) of their activities because of the allegedly increased competition in traditional bankingactivities4 Unfortunately, the economic benefits from expanding scope seem are not overwhelminglyechoed in the data (Berger et al 1993) The enormous literature estimating costs and productionfunctions of banks remains relatively inconclusive on the issue (see the survey in Clark (1988) andAltunbas (2001) and Bikker (2001)) Regressing three different measures of bank profits on bank-

4 Because they need to improve their cost efficiencies to operate more effectively, scale expansion would be justified In parallel, by squeezing margins, tougher competition would force banks to seek profits outside familiar territories This argument, however, is not strongly supported empirically In addition, cost efficiency arguments suppose a focus on

“core” competencies They are therefore difficult to reconcile with scope expansion (Hamel and Prahalad (1990)) A more appealing argument relates to the strategic benefits that may arise from increasing scope (and size) Milbourn, Boot and Thakor (1999) suggest that banks may enlarge the scope of their activities because it enhances the reputation of their management and/or increases the wealth of shareholders In addition, strategic benefits may also arise if (i) current operations are sufficiently profitable to finance the fixed costs associated with scope expansion, (ii) uncertainty about the core competencies required to successfully run new operations is sufficiently high, and (iii) expected competition in the prospective market is low enough so as to make the effort worthwile By stressing the role of informational uncertainty and learning for a wider scope to be optimal, this argument suggests that unless very specific conditions are met, it does not pay, on average, to be diversified See also Berger and Humphrey (2000) and, for European studies, Altunbas (2001) and Bikker (2001).

specific and country specific variables, Steinherr (1994) singles out the significantly larger influence

of costs, competition and regulation on the profits of universal banks relative to those of specialisedinstitutions However, those results do not say much on the disaggregation of those three factors byactivity

Overall, estimating complementarities and economies of scope is subject to a range of practicalproblems Among the issues identified in Berger, Hunter and Timme (1993), two are of direct interest

to us First, data on specialized banks are scarce This issue is particularly fierce when using a narrowdefinition of universal banking as we do here In particular, banks in the euro area tend to produce theentire array of retail banking outputs.5 Second, the data used to evaluate economies of scalecorrespond to a point which is far from the efficiency frontier In that sense, scope economies may bemistaken for X-efficiencies.6

However, for our purpose, the prevalence of universal banks in the euro area may affect thetransmission of market interest rates to retail banking conditions even if economies of scope are notlarge In particular, the multi-business nature of universal banks may help them to manage interest raterisk and dampen the effect of fluctuations in market conditions To that respect, specialized institutionsmay either add a portfolio of liquid securities and/or recourse to money market instruments and centralbank liquidity Universal banks have another option As they handle both loans and deposits, they may

in principle shelter lending activities from market conditions not only by spreading risk across loanmarkets segments, but also by “using deposits” as an input to producing loans By doing so, banksshould buffer the impact of market conditions on retail loans rates, and create a causal link fromdeposit rates to lending rates

The question as to whether deposits are inputs or outputs for universal banks has been raised in theintermediation literature On the one hand, they have been considered as inputs for the production ofloans, i.e as a source of liquidity, which is then redistributed under various maturities to agents inneed of financing Alternatively, we may argue that deposits are also an output, as retail banks areservice producers to depositors Without clinching the matter, Sealey and Lindley (1977) viewdeposits as an intermediate input which is produced by banks (they offer means of payments and aremuneration to depositors), and then used in the production of loans Hughes and Mester (1993a, b)estimate a variable cost function with a fixed level of deposits Their estimate of this function’sderivative with respect to deposits being negative, they conclude that deposits are inputs The “usercost methodology” proposed by Hancock (1991) is also insightful Hancock regresses bank profits on

the real balances of all items of its balance sheet without earmarking loans and deposits as being exante outputs or inputs The input/output distinction emerges endogenously from the sign of theregression coefficients Positive estimates correspond to outputs, while negative estimates correspond

to inputs She finds that loans and deposits are outputs, while cash (time deposits and borrowedmoney) is an input

As a result, retail interest rates in loans markets may depend on retail pricing for deposits, aphenomenon less likely to emerge when banks are specialised Hence it appears that the dominance ofuniversal banking in the euro area could be a factor for the sluggishness of retail bank rates

3 A model of bank pricing

From a static point of view, the links between market and retail interest rates has an immediateinterpretation in terms of bank profit maximizing and pricing An equilibrium relationship betweenretail and market rates may be obtained in a simple static Monti-Klein bank where banks hold moneymarket instruments and longer term assets.7 A “universal” bank with some monopoly power choosesthe volumes of loans L and deposits D that maximise its profits given by

[1] π=r l L+r s M+r b B−r d D−C(D,L)

C(.) is a well behaved cost function M is the bank’s net position on the interbank market and B its net

longer term assets holdings (“bonds”) which, given mandatory reserve requirements a, satisfy the

balance sheet condition:

[2] (1- a)D – L = M+B

s

r and rb are the interest rates prevailing in the money and bonds markets They are taken as given by

the bank Money market instruments and bonds are held with proportions k and (1-k), that is

[3] M = k [(1- a) D + L]

[4] B=(1-k)[(1- a) D + L]

Rewriting profits accordingly, the first order conditions for the supply of loans and the demand for

deposits yield a pricing rule for each market i (i = deposits, various loans), for given inverse demand

and supply functions rl(L ) and rd( D ):

7

More elaborate profit-maximizing oligopoly models have exploited large detailed datasets, see Steinherr and Huveneers (1994).

11][

')1([

]

5

i i

b s

Overall, the resulting pricing equation [5] highlights the key role played by long term rates when thelonger term assets held by banks are taken into account when they maximise profits [5] states thatretail rates are set as a weighted average of market interest rates, corrected for market structure andbanking costs

The interdependence between the deposit and lending activities in which a universal bank is involvedappears via the cost structure of banking activities If the markets for deposits or loans are segmented,the properties of the cost function determine whether developments in one market have an impact onretail rates in other markets In particular, the cost function is defined on both deposits and loans andshould not be separable in those arguments for a bank engaging simultaneously in various lending anddeposit markets In particular, the equilibrium volume of deposits chosen by the bank will depend onthe rate prevailing in that market, so that the marginal cost of loans in [5] can be explicitly written as

of quantities and cost parameters As a consequence, we concentrate on a linear specification includingboth market rates but leaving deposit rates out of the information set.9 This is coherent with the fact

that a baseline Monti-Klein bank considers retail interest rates to be independent from each otheracross markets.

4 Data

National retail bank markets provide independent observations to investigate bank pricing Indeed,these markets remain segmented in spite of the institutional changes and the market consolidation thatfinancial intermediaries went through in the last two decades In contrast to the increasing integration

of securities markets and wholesale finance services across countries (Pagano et al 2002), theconsolidation of the banking sector took mostly place within national borders This nationalsegmentation explains, among other factors, why the pass-through of changes in market interest rates

to retail bank interest rates is different across euro area countries (see Table 1)10

The analysis is carried out on 46 retail interest rate series for all euro area member states exceptLuxembourg and Greece, the euro area and the associated MMR and BR All series have a monthlyfrequency - except for France, where the model is estimated on quarterly data - and are available fromthe ECB national retail interest rate database.11 They correspond to five main financial instruments thatreflect different segments of the banking sector They include interest rates on short (10 series) andlong-term (6 series) loans to firms, mortgages to households (10 series), consumer credit (7 series) andtime deposits (9 series) One should note that from January 1999, the MMR is the EURIBOR for allcountries

Each of those five categories may differ across countries by their main characteristics, namely habitat,maturity, the average size of each transaction, and risk The rates on short-term loans are reported,when specified, for maturities ranging from up to three months (Spain) to up to 18 months (Italy).Long-term loans to enterprises refer to investment credit of over one year Consumer loans includeoverdrafts (e.g Ireland), but usually correspond to a weighted-average of short-term credit lines,personal loans, and longer-term installment credit Housing and mortgage loans typically have alonger maturity, specified over 18 months (Italy) to three (Spain) to five years (Germany, Portugal).Finally, we restrict our analysis of deposits to the interest rate on time deposits, which are the onlydeposit rates that are available for a large enough number of countries

We estimate the models on sample periods that excludes the turbulent years of the early 1990s, wheninterest rates have been quite volatile, as they had to respond to a number of shocks, notably exchange

10 It is worth stressing that, while the levels of retail bank rates, and their spread with respect to market rates are not directly comparable, the pass-through from market rates to retail bank rates is more comparable.

11

The series we use and a detailed description of their characteristics is available at www.ecb.int

rate crises This clearly appears in Charts 1 and 2 For most countries, the ERM crises of the earlynineties led either to out-liers (Ireland, Belgium, France, Italy) or to periods of high volatility ofmarket rates (Spain, Portugal, Finland) We trust that such turbulence can never take place in EMU.Hence, for these countries, the sample period of estimation starts in 1994:4 For Germany and theNetherlands the estimation sample starts in 1991:1 In Austria, which also was part of the core ERM,the estimation starts in 1995:7 because retail bank rates are not available before.12 Nevertheless, it cannot be excluded that our results are distorted by a change in yield curve effects.

Two final observations on the data are worth noting First, Charts 1 to 4 indicate a clear downwardtrend in all the market and retail bank interest rates in the period prior to EMU In addition to theupturn, which took place after April 1999, the other main episode of rising interest rates corresponds

to the winter 1994 crash on bond markets which was triggered by the February 1994 increase in theFed funds rate Second, the well-known hierarchy in the mark-ups across retail bank markets, i.e.largest on consumer credit, lowest on mortgages with loans to firms in between, is widely observedacross countries

5 Empirical estimations

5.1 Do lending rates depend on market rates?

The first step in our empirical analysis is to check whether banks insulate lending rates from marketconditions thanks to the funds collected as deposits We look at this issue by investigating the extent towhich deposit rates have a predictive power with respect to lending rates For that purpose, weimplement Fisher tests on the coefficients of market rates and deposit rates in equations of lendingrates such as:

++

+

i

t n

i

i t i n

i

i t i i

t i n

i

i t i n

222

12

21

β α

where x, mmr, br, d1, d2 are retail loan rate, the short-term and long-term market rates, interest rates

on deposits (time deposits, savings accounts or current accounts depending on the country), while liisalternatively the interest rate on loans to firms (short and long), consumption credit and mortgages

We test which interest rate Granger causes each of our four lending rates

Results shown in Table 2 reveal that deposit rates are not relevant for interest rates on loans in a clearmajority of cases (25 out of 32 across the four types of loans) The 7 observations where the lags of atleast one deposit rate are relevant are concentrated in Austria (all markets but mortgages) and to someextent in Belgium In terms of markets, cases are concentrated in loans to firms, both short and long

12

In the case of France where our data is quarterly, however, the full sample is extended to 1991:1 – 2001:4 (1998:1 onwards for the EMU-sample).

By contrast to this very occasional relevance of deposit rates, the coefficients for short and/or longterm market rates contain quasi systematically valid information for lending rates On the basis ofthose properties, we vastly reject the buffer role of deposits for the pricing of loans and consider thatbanks in continental Europe follow market conditions in spite of being universal We therefore focus

go ahead with a specification of loans pricing based on market rates only

5.2

Presentation of the empirical model

After finding that lending rates usually do not depend on deposit rates, we present the error correctionform of the model presented in section 3 All specifications proposed below introduce, in addition tothe MMR, the long-term market interest rate as an explanatory variable for the behavior of retailinterest rates This approach has two merits First, the long-term interest rate inherently influences thefunding cost of banks As suggested in section 2, the latter depends on the structure of the bank’sassets, which include short and long term market instruments as well as loans.13 Moreover, we expectthat banks consider that the funding costs of a loan depend on its maturity For example, mortgages arefunded with the issuance of long-term bonds while short-term loans to firms are funded with theissuance of certificates of deposits

In practice, the funding cost of loans would depend on both short-term and long-term market interestrates, with a stronger impact of the latter for long-term loans The second merit is that long-termmarket rates contain information about market expectations of the level of future short-term interestrates

We choose to provide an empirical characterization of those issues within an ECM that relates eachretail interest rate to the short and long-term market rates This approach is a simple and intuitive way

to relate the adjustment of retail bank interest rates to changes in the funding or opportunity costs inthe banking sector In particular, the ECM approach gives a view on long-term relationships betweenretail bank prices and market interest rates, on the adjustment dynamics of the former, and finally has asay over their stochastic properties and the equilibrium conditions between them

The advantage of our approach over an analysis of cointegrating relationships between retail andmarket rates following Johansen (1988) is its very intuitive interpretability as a marginal cost pricemodel of bank pricing Under imperfect competition, intermediaries impose a mark-up over expectedrefinancing conditions when setting their retail interest rates (Rousseas, 1985, and De Bondt, 2002a)

13

We note that the latter also depends on the structure of the banks’ liability, which comprises short and long-term market instruments as well as deposits However, the interest rate on deposits is a mark down over the opportunity cost of raising alternative liabilities, i.e issuing debt securities at market rates Overall, the reasoning developed in section 3 may be applied to net holdings of money market instruments and long-term bonds, assuming them to be perfect substitutes on the asset and the liability side of the bank balance sheet.

This mark-up is related to the multiplicative term [1− 1]−1

i

maturities of the supplied financial instruments, those conditions are reflected by both short and term market interest rates In addition to be intuitive, the specification is appropriate to discriminatebetween the short-run dynamics (first difference terms) and the adjustment towards the long-runequilibrium relation (in level terms) This can be done here without reducing the specification unless

long-we want to impose specific testable restrictions.14 Overall, we do distinguish the long-term relationshipbetween retail and market rates (i.e the drift of retail rates along a linear combination of market rates)from the short-run dynamics In the short-run, changes in market rates are transmitted to changes inretail bank rates through the lagged differenced terms in [1] The long-run rigidities that relate retailrates to market rates are reflected by the correction term towards equilibrium For this interpretation,

we rely on the fact that market interest rates are (weakly) exogenous to retail bank interest rates andmay themselves depart from their long-run equilibrium value

Baseline specification

Our baseline specification is a symmetric linear error correction model (ECM) relating each retailinterest rate on loans as shown in [1] to both long-term bond rates and short-term money marketinterest rates

j

j t j j

j t j j

j t j t

1 2

1 0

For time deposits, we have

[6’’] ect t =(AAi t−1+BBl t−1−CC)−r t−1

14 If we were to give a structural economic interpretation to unit roots that may not be statistically rejected, a fully-fledged cointegration analysis of the system formed by all rates would have been the way to go More than one cointegration relationship is needed to give a role to expectations, structures and policy regimes as determinants of the interest rate pass-through Since any linear combination of cointegration relations would also be stationary, cointegrating vectors could not be given a direct interpretation as meaningful economic relations, and identifying restrictions have to be imposed To our view, imposing such restrictions would be too strong.