econstor Make Your Publications Visible.

Transcription

1 econstor Make Your Publcatons Vsble. A Servce of Wrtschaft Centre zbwebnz-informatonszentrum Economcs Buch, Clauda M. Workng Paper Captal Market Integraton n Euroland - The Role of Banks Kel Workng Paper, No. 93 Provded n Cooperaton wth: Kel Insttute for the World Economy (IfW) Suggested Ctaton: Buch, Clauda M. (1999) : Captal Market Integraton n Euroland - The Role of Banks, Kel Workng Paper, No. 93 Ths Verson s avalable at: Standard-Nutzungsbedngungen: e okumente auf EconStor dürfen zu egenen wssenschaftlchen Zwecken und zum Prvatgebrauch gespechert und kopert werden. Se dürfen de okumente ncht für öffentlche oder kommerzelle Zwecke vervelfältgen, öffentlch ausstellen, öffentlch zugänglch machen, vertreben oder anderwetg nutzen. Sofern de Verfasser de okumente unter Open-Content-zenzen (nsbesondere CC-zenzen) zur Verfügung gestellt haben sollten, gelten abwechend von desen Nutzungsbedngungen de n der dort genannten zenz gewährten Nutzungsrechte. Terms of use: ocuments n EconStor may be saved and coped for your personal and scholarly purposes. You are not to copy documents for publc or commercal purposes, to exhbt the documents publcly, to make them publcly avalable on the nternet, or to dstrbute or otherwse use the documents n publc. If the documents have been made avalable under an Open Content cence (especally Creatve Commons cences), you may exercse further usage rghts as specfed n the ndcated lcence.

2 Kel Insttute of World Economcs üsternbrooker Weg Kel Kel Workng Paper No. 93 Captal Market Integraton n Euroland The Role of Banks by Clauda M. Buch June 1999 The author herself, not the Kel Insttute of World Economcs, s solely responsble for the contents and dstrbuton of each Kel Workng Paper. Snce the seres nvolves manuscrpts n a prelmnary form, nterested readers are requested to drect crtcsms and suggestons drectly to the author and to clear any quotaton wth her.

3 Abstract * The ntroducton of the euro marks a mlestone n the process of European fnancal market ntegraton. Ths paper analyzes the mplcatons of the euro for cross-border bankng actvtes. A portfolo model s used whch captures the role of banks as provders of nformatonal and of rsk-dversfcaton servces. By elmnatng exchange rate rsks, the euro enhances the ncentves of banks to expand wthn Euroland. Yet, whle the currency bas n bank portfolos wll be elmnated, the home bas wll reman. It s also argued that postve dversfcaton effects may outwegh possble negatve effects on the rsk takng of banks. (98 words) Keywords: JE-classfcaton: captal moblty, European fnancal ntegraton, bankng, asymmetrc nformaton, portfolo choce 8, F36, G1 Malng address: The Kel Insttute of World Economcs üsternbrooker Weg Kel Phone: * Fax: * E-Mal: cbuch@fw.un-kel.de * Part of ths paper has been wrtten whle the author has been on a research vst at the Unversty of Mchgan, Ann Arbor, n September The hosptalty of the Unversty of Mchgan s gratefully acknowledged. The author would lke to thank Stefan M. Golder, Ralph P. Henrch, and Jörn Klenert for most helpful comments on an earler draft. Remanng errors and naccuraces are solely n my own responsblty.

4 3 Contents 1 Motvaton... 4 Stylzed Facts A Portfolo Model of Cross-Border Bankng Fnancal Market Integraton and Captal Moblty The Bankng Sector Impact of the Euro Portfolo ecsons and the Euro Home Bas versus Currency Bas Montorng Actvtes Balance of Payments Effects Market Integraton and Rsk Takng Summary References... 3

5 4 1 Motvaton The ntroducton of the euro at the begnnng of 1999 has been the sngle most mportant change affectng European fnancal markets for the years to come. Fnancal markets n Europe have already undergone profound changes n the past two decades. Captal controls wthn Europe and vs-àvs the rest of the world have been lfted; the Second Bankng rectve has leveled the playng feld for banks n Europe. The advent of the euro completes these processes and, at the same tme, may serve as a catalyst of future nsttutonal change wthn Europe s fnancal markets. The euro s lkely to affect the way n whch fnancal markets operate and to mpact upon captal moblty. 1 The magntude of these effects, n turn, wll have mportant mplcatons for other polcy areas such as the effectveness of fscal polcy and the conduct of monetary polcy (see e.g. ornbusch et al. 1998). Ths paper provdes a framework n whch the mplcatons of the euro on captal moblty wthn Europe can be analyzed. Costs of cross-border transactons and asymmetres n nformaton between domestc and foregn nvestors are ntroduced nto a standard mean-varance framework n whch banks can hold both assets and labltes at home and abroad. Although the focus of the analyss s on the portfolo choces of commercal banks, t s not lmted to ths class of nvestors. The mpact of nformatonal asymmetres on cross-border captal flows has been shown already before. Montgomery (1990), for nstance, consders a two-country model, n whch one ntermedary s present n each country. Intermedares have pad a sunk cost whch gves them access to funds below the rsk-free rate. All nternatonal captal flows are effected through these ntermedares whch compete across borders by grantng loans to each other but not to foregn resdents. Returns on domestc loans nter ala depend on the montorng effort exerted by domestc ntermedares. ue to asymmetres n nformaton, ntermedares cannot observe each other s montorng efforts. Hence, compared to a full-nformaton framework, under- 1 See Begg et al. (1999), or ermne and Hllon (1999).

6 5 nvestment occurs because the ntermedary whch has access to a greater (exogenous) supply of funds s less wllng to lend cross-border. As a result, the country wth lower ntal savngs s confned to lower nvestment, whch can explan the emprcal observaton made by Feldsten and Horoka (1980) and confrmed by many subsequent studes that domestc savngs and nvestment are hghly correlated. Gordon and Bovenberg (1996) lkewse show the mpact of asymmetres n nformaton for the effcency of the nternatonal allocaton of captal. They assume that foregners can ether make greenfeld nvestments n the domestc economy or purchase shares n exstng domestc frms from resdents. Foregners dffer from resdents n that they cannot observe the (stochastc) component of project returns when bddng for shares and that they are less effcent than domestc owners n runnng frms themselves. These asymmetres n nformaton and sklls mples that resdents can overcharge foregners when sellng shares and that greenfeld nvestment occurs despte the lower productvty of frms run by foregners. Although Gordon and Bovenberg look manly at foregn drect nvestment decsons, ther results could easly be re-nterpreted n terms of other forms of captal flows. In a smlar ven, Gehrg (1993) argues that asymmetres n nformaton can be one explanaton for the home bas typcally observed n nternatonal asset portfolos (Tesar and Werner 199). In hs model, nvestors receve nose sgnals about returns on assets at home and abroad, and the average precson s hgher for domestc than for foregn sgnals. The followng analyss wll extend these deas n four regards. Frst, rather than explctly modelng the prncpal-agent relatonshp between domestc and foregn banks as n Montgomery (1990), cross-border actvtes of banks wll be analyzed n a portfolo framework. The advantage of ths approach s that t allows for greater flexblty n modelng asset and lablty choces of banks. Second, exchange rate effects wll be taken nto account explctly. Thrd, n addton to the nformatonal role of banks, ther rsk poolng functons wll be consdered. Fourth, the model wll be used to derve mplcatons of the euro for cross-border bankng actvtes, captal moblty, and bankng rsks. The baselne portfolo model s presented n Secton 3. The model wll be used to show the mpact of exchange rate rsks and of costs of obtanng nformaton on the cross-border actvtes of banks n Secton 4. Secton 5 concludes and summarzes the man fndngs. We start wth a bref

7 6 summary of stylzed facts on the cross-border actvtes of commercal banks. Stylzed Facts espte the wdely dscussed globalzaton of fnancal markets, foregn assets and labltes of commercal banks account for less than 0 percent of the balance sheet total n most ndustralzed countres (Graph 1). Exceptons are countres that host fnancal centers such as the Unted Kngdom where foregn busness consttutes almost two-thrds of all actvtes. For the EU countres, French commercal banks also have a relatvely large exposure towards foregn countres whereas the balance sheet shares for German or Italan banks are n a range of 15-0 percent. The Unted States are at the lower end of the spectrum as foregn actvtes of commercal banks account for less than 10 percent of the balance sheet total. Judged on the bass of total foregn actvtes, the EU s Sngle Market Program of 199 seems to have enforced an already exstng trend for an expanson of foregn assets n countres such as France, Italy, or Germany. Snce 1995, a smlar trend could be observed for the Unted States. As regards the mportance of foregn labltes, n contrast, only German and French banks have ncreased foregn actvtes after 199 whle Italan banks have reduced ther relance on foregn funds. For all countres, the fgures presented n Graph 1 nclude not only loans and deposts granted to and rased from abroad, but also securtzed assets and labltes. Hence, the data gve the upper bound for the share of crossborder lendng and borrowng n the retal market. For the countres of the European Unon (EU), calculatons of the Bank for Internatonal Settlements show a share of foregn lendng n total lendng to non-banks of less than 10 percent (Table 1).

8 Graph 1 Foregn Assets and abltes (n % of End-Year Balance Sheet Total) a) Assets FRA GER ITA US b) abltes FRA GER ITA US Source: IMF (1999), own calculatons.

9 Table 1 Cross-Border Actvtes of European Banks Cross-border loans to non-banks / domestc credt (%) Cross-border labltes to non-banks / domestc money (%) Austra Belgum France Germany Italy Netherlands Span Swtzerland Unted Kngdom Source: Whte (1998: 5). Graph Foregn Assets and abltes of German Banks (n % of Total) a) EU-countresa Clams abltes

10 9 b) Unted States Clams abltes a) After 1994: Includng Austra, Fnland, and Sweden. Source: eutsche Bundesbank (1999), own calculatons. For Germany, Graph shows the share of cross-border clams and labltes of banks dsaggregated by countres. Snce the late 1970s, clams and labltes of German banks towards EU countres have expanded rapdly. By the end of 1998, labltes vs-à-vs EU countres accounted for about 70 percent of the total; clams for about 60 percent. Hence, a strong regonal bas n German banks foregn actvtes s thus vsble. Actvtes n the US are much less mportant but stll account for 6-10 percent of the total. Interestngly, t does not seem as f the Sngle Market Program of 199 has had an mpact on the overall trend to expand European operatons. Fnally, t has been argued that fnancal lberalzaton and ntegraton of fnancal markets could lower the screenng actvtes of commercal banks (Azenman 1998). Ths mght show up n an ncreased need to provson for loans losses. Table presents selected data from the ncome statements of commercal banks. It shows that real returns on equty (ROE) for commercal banks n Europe have on average been lower durng the perod between 1980 and 1995 than for banks n the Unted States. Over tme, the proftablty of banks has developed qute dfferently. After a substantal declne n the ROE between 1986 and 1990, US banks have mproved ther perform-

11 10 ance n the subsequent fve-year perod. Manly, ths mprovement was acheved through an ncrease n the proft margn,.e. through cuttng costs. Provson expenses declned for US commercal banks n the early 1990s as compared to the late 1980s but tended to rse for the European banks. Ths evdence could be nterpreted n two ways. Ether, external condtons could be the same for the two bankng systems but US banks have superor rsk management systems. Alternatvely, banks n Europe may be under greater compettve pressure and would thus have faced dfferent external condtons than banks n the US. Table Income Statement Analyss for Commercal Banks Germany ROE Real ROE Provsons / gross ncome France ROE Real ROE Provsons / gross ncome Italy ROE Real ROE Provsons / gross ncome Span ROE Real ROE Provsons / gross ncome Unted States ROE Real ROE Provsons / gross ncome ROE = return on equty (net after tax ncome / equty). 1 France , Italy and UK All banks. Source: OEC (1997) In the followng, we wll present a formal model whch helps to determne the factors whch decde over foregn actvtes of commercal banks. In addton, lnks between market ntegraton and rsk takng of commercal banks wll be analyzed.

12 11 3 A Portfolo Model of Cross-Border Bankng Ths secton presents a smple portfolo model of cross-border bankng n whch domestc and foregn banks compete both for loans and deposts at home and abroad. The model allows us to gauge the mpact of a reducton n exchange rate rsk as well as nformatonal asymmetres on the behavor of commercal banks (and thus on captal flows). The man focus of the analyss are retal bankng actvtes of banks but an extenson to the wholesale market and/or nvestment bankng would be straghtforward. 3.1 Fnancal Market Integraton and Captal Moblty Before startng wth the more formal analyss, t s useful to clarfy some concepts related to the ntegraton of nternatonal fnancal markets whch can essentally take three forms: Cross-border captal flows: If domestc savngs are exported to fnance nvestment abroad, a cross-border captal flow s regstered n the fnancal account of the balance of payments. Cross border captal flows have an mpact on the domestc bankng system because they affect the supply and demand functons for loans and deposts. Increased captal flows and easer access to foregn markets mply that households and frms get access to a wder choce of fnancal assets and may thus react more quckly to changes n domestc nterest rates. Trade n fnancal servces: If nformaton and transacton costs were neglgble, fnancal ntermedares would not exst. In realty, however, only a fracton of fnancal contracts s concluded drectly between the ultmate supplers and users of funds. Ths holds n partcular n an nternatonal context where savers and nvestors are locatonally separated. Hence, fnancal ntermedares are nvolved. These ntermedares earn nterest rate spreads and other fees on cross-border captal flows whch are regstered n the balance of servces of the current account. Trade n fnancal servces can but need not be lnked to cross-border captal flows. It s concevable that a captal flow s regstered between country A and B but that ths deal s arranged by a fnancal ntermedary located n country C. Even though a foregn ntermedary may not be present physcally n the home country, ts presence abroad would thus have a compettve mpact on domestc banks.

13 1 Market presence of foregn banks: In some market segments, physcal presence n the market s needed to servce domestc clents. Hghly nformaton-senstve relatonshp loans, for example, are rarely arranged through off-shore ntermedares but rather through ntermedares whch hold close, personal contacts to ther clents. A smple applcaton of the tradtonal foregn trade theory to the bankng ndustry, whch would mply that prces for fnancal servces can be equalzed ether through trade n bankng servces or through foregn drect nvestment n the fnancal sector, s thus not possble. Instead of beng substtutes, trade n fnancal servces and foregn drect nvestment (FI) n bankng must be vewed as complementary to the extent that the provson of fnancal servces requres the physcal presence n the market (Walter 1988). An alternatve vew would be that f foregn banks are present n a gven country,.e. f they have ncurred the sunk cost assocated wth market entry, they are lkely to focus both the asset and lablty actvtes on the market of that country. Crossborder captal flows and/or fnancal servces may vansh. Ultmately, the queston whether FI n bankng and cross-border captal flows are complements or substtutes s thus an emprcal one. These consderatons mply that the degree of cross-border competton n bankng s closely related to the degree of captal moblty but that the two are not necessarly the same. Rather, market ntegraton s a broader concept than captal moblty. Hence, an assessment of the degree of market ntegraton must take nto account all three channels descrbed above. ookng at nterest party condtons or net captal flows alone would gve an ncomplete pcture of the degree of ntegraton. In Europe, market ntegraton has taken all three forms. Cross-border captal flows have become fully lberalzed wth the successve abolton of captal controls n the early 1990s. Trade n fnancal servces has been deregulated by applyng the home country prncple, whch s enshrned n the Second Bankng rectve, to the fnancal servces ndustry. In addton, low entry barrers for outsde fnancal nsttutons make European fnancal markets hghly contestable. Stll, t s commonly asserted that the full compettve mpact of the creaton of a sngle market for captal lags behnd expectatons as market shares of foregn banks are low and as neffcences n some segments of European bankng preval (Prat and Schnas 1997, McCauley and Whte 1997).

14 13 In the followng, we present a smple model of cross-border bankng whch shows that even though captal flows have n prncple been lberalzed, the presence of transactons costs nduces a home bas n banks portfolos. The baselne model assumes that trade n fnancal servces (nterest payments) and captal flows (changes n loans and deposts abroad) can move freely between two countres. Clents abroad can be servced from the domestc bank s homebase, but we assume hgher transacton costs of crossborder lendng and borrowng whch captures the fact that costs are lower f customers are close to the banks. 3. The Bankng Sector Two countres wth a fxed number of banks (n and n* where = 1,..., n, and j = 1,..., n*) operatng n each are consdered. Each bank gves out loans () and rases deposts () on ts home market as well as on the foregn market. Yet, they mantan a presence only n ther domestc market,.e. there s no FI n bankng. In addton to deposts and loans, banks can nvest nto a rskless securty but cannot borrow at the rskless rate. Arbtrage between the home and the foregn market s exerted through banks only. Ths s equvalent to assumng that households and frms face hgher transacton costs than banks. Ths assumpton on market access squares wth the observaton that, despte the creaton of a Sngle Market for captal n Europe, retal markets reman largely segmented (European Commsson 1997: 4). Because borrowers and lenders do not nteract drectly, depost and lendng rates are not dentcal, the spread between the two reflectng the costs of fnancal ntermedaton. These costs are motvated by the presence of asymmetres n nformaton and by a superor allocaton of rsks, whch make tradng through ntermedares less costly than drect trades (Allen and Santomero 1997). In prncple, there are four dfferent ways n whch captal can flow nternatonally n order to arbtrage between markets. omestc banks can rase deposts at home or abroad and nvest nto foregn and domestc loans. The same optons are avalable for foregn banks. To analyze the resultng port- In order to focus on exchange rate effects, markets for the rskless assets are assumed to be segmented.

15 14 folo choces of banks, we assume that all contracts are denomnated n local currency. When calculatng returns on actvtes abroad, exchange rate rsks have thus to be taken nto account. Furthermore, we consder only one perod. At the begnnng of the perod, the bank chooses ts optmal portfolo structure. Hereby, t must observe ts balance sheet restrcton whch s gven by bank s loans on the domestc and on the foregn market and the rskless asset ( R ): (1) W + + * = + * + R where W = ntal wealth, () = domestc deposts (loans), and *(*) = foregn deposts (loans) n domestc currency terms. At the end of the perod, returns are realzed. The expected proft of a representatve domestc bank s thus gven by: * * * * * * () E[ Π ] = ( r c ) + ( r c + e& ) + r R ( r + c ) ( r + c e& ) K ( µ ) t,, F,, where &e = expected rate of change n the exchange rate (prce of foregn currency n domestc currency terms), r, r = expected nterest rates on loans and deposts, r F = nterest rate on the rsk-free asset, c = varable costs of makng loans and rasng deposts 3 and K ( µ ) = montorng costs wth K '( µ ) > 0, K ''( µ ) < 0. Snce we assume that banks bear the exchange rate rsk, a deprecaton of the domestc currency ( &e > 0 ) rases both the return on loans abroad and the costs of deposts abroad. Exchange rate changes are stochastc wth a standard devaton σ e > 0, and are taken as exogenous by the banks. A smlar proft functon can be derved for the foregn bank. Upon substtutng the balance sheet restrcton (1) nto (), one obtans: ( ) [ ] ( ) ( * * ) * Π = + ( ), +, + & +, * * * ( r + c, e& rf ) K( µ ) E r W r c r r c e r r c r F F F F Rasng deposts and grantng loans s costly for banks because t requres, for nstance, the mantenance of a branch network. Varable costs are assumed to be hgher n an nternatonal context than domestcally as these comprse the costs of cross-border fnancal transactons ( c < c* and,, j, * j, c, < c*, ). The reverse relatonshp holds for foregn banks: c > c and 3 Note that these varable costs add to the nterest cost of deposts whle they lower the nterest rate earned on loans.

16 15 c j, > c* j,. omestc (foregn) banks are assumed to have a comparatve advantage n the provson of domestc (foregn) fnancal servces,.e. c < c and c* > c*. A smlar condton apples to the loan market., j,, j, In addton to the expected profts of ther actvtes, banks also care about the rsk of ther portfolo: (3) σ ( ) 4 Π = x, m σ m + x, x, COV m= 1 m= 1 n= 1 m n 4 4 m n mn x x where x denote portfolo shares wth x = x x =, and COV = co- * * varances of returns., 1,, 3, 4 The objectve functon of the representatve bank s ncreasng n expected profts and decreasng n the varance of the portfolo: 4 (4) U U E( Π ) σ ( Π ) U U [, ], E( Π ) σ ( Π ) = > 0 < 0 Ths rsk averson of banks could be endogenzed by assumng that banks face a postve probablty of nsolvency, and that nsolvences are costly. Baltensperger and Mlde (1987), for example, argue that n the case of bankruptcy banks have to cover costs of reorganzaton and admnstraton. The same qualtatve results are obtaned f banks have to meet an equty requrement (Helblng 199). If, due to an unexpectedly low return on assets, ths equty requrement s volated, banks are charged wth a penalty propor- 4 Ize and evy-yeyat (1998) use a smlar mean-varance approach to determne the mpact of macroeconomc rsks on the degree of dollarzaton. In contrast to our approach whch focuses on the portfolo choces of commercal banks, they model drectly the behavor of households and frms whle assumng banks as relatvely passve ntermedares between the two groups of market partcpants. For earler applcatons of portfolo models to the management of country specfc rsk and to the assessment of the foregn exchange rsk ncurred by US banks see Walter (1981) and Grammatkos et al. (1986).

17 16 tonal to the amount by whch equty falls short of the threshold. 5 Internatonally accepted bankng standards requre banks to hold equty to cover open foregn exchange postons as, n 1995, the Basle Commttee on Bankng Supervson has ntroduced a specal captal charges applyng to banks foregn currency rsks (BIS 1996). Before analyzng optmal portfolo choces of banks, t s useful to dstngush the rsks that banks are exposed to. On a general level, nterest rate and exchange rate rsks can be dstngushed. Interest rate rsks arse because the return on lendng actvtes s assumed to be stochastc. For smplcty, we gnore uncertanty about the magntude of depost rates. Banks are assumed to be able to reduce ther exposure to lendng rsks by nvestng nto screenng actvtes whch allow them to better classfy prospectve borrowers. Followng Baltensperger and Mlde (1987: 169n), we assume that screenng of loan applcants helps banks to reduce the standard devaton of returns from lendng although not allowng them to fully elmnate lendng * * * rsks: σ ', ( µ ), σ ', ( µ ) < 0 and σ '', ( µ ), σ '', ( µ ) > 0 where σ, ( σ, ) = standard devaton of loan returns for domestc bank when lendng to domestc * (foregn) clents, σ j, ( σ j, ) = standard devaton of loan returns for foregn bank when lendng to domestc (foregn) clents, and µ ( µ j ) = montorng actvtes of domestc (foregn) banks. omestc banks are assumed to have a comparatve advantage n classfyng domestc borrowers (and vce versa): * (5) σ ' ( µ ) σ ' ( µ ) > and σ ', ( µ ) σ ' j, ( µ j ),, >. Whereas the volatlty of domestc returns depends on characterstcs of the borrower populaton only, foregn actvtes also expose the bank to an exchange rate rsk. The standard devatons of foregn lendng and depost rates are thus gven by: (6) * * ( r ) 3 = [, + e + COVe ] σ σ σ σ σ * ( r ) = σ e 5 The key assumpton s that banks hold equty by the amount needed to cover the expected nsolvency cost. Hence, n the case of nsolvency, equty s zero. In the present settng, ths specal role of equty has not been taken nto account explctly. 1

18 17 * wth COVe = COV ( r, e) = ρ eσ σ e = covarances of foregn lendng rate and exchange rate changes (ρ e = coeffcent of correlaton). The standard devaton of domestc currency returns of foregn lendng s therefore below the sum of lendng and exchange rate rsk (Elton and Gruber 1995: 66). If correlatons between foregn nterest rates and exchange rate changes are suffcently small n absolute terms, the rsk of foregn lendng ncreases f exchange rate volatlty goes up: σ σ 3 ε σ ε + ρ εσ = > 0. σ 3 The bank s optmal demand for asset m s gven by maxmzng (4) wth respect to loans and deposts. The frst order condtons are thus gven by: (4 ) U x ( Π ) ( Π ) U E σ U = + E x σ x ( Π ) ( Π ), and, by denotng the degree of the bank s relatve rsk averson wth (7) λ U = 1 E ( Π ) ( ) σ Π, U optmal portfolo shares can be obtaned from: 1 (8) $x = λ V r and V 1 s the nverse of the varance-covarance matrx of excess returns r. We assume that there s no uncertanty wth regard to the magntude of varable costs, and that V s dstrbuted normally. The vector of excess returns s gven by: r r r = r r,, *, *, r c, rf r c, + rf = r* c* + e& r F, r* c* e& + rf, Exchange rate changes are assumed to be relatvely small,.e. r, r > 0; r, r < 0., *,, *,

19 18 Thus knowng the bank s relatve rsk averson, the expected excess returns, and the covarances between rsky assets, ts optmal demand for each of the assets n terms of mean-varance-effcency can be determned. Under the assumpton that excess returns on loans (deposts) are postve (negatve) and that all elements n the varance-covarance matrx are postve, one obtans negatve portfolo shares for deposts and postve portfolo shares for loans. An ncrease n the excess return of an ndvdual securty ncreases the share of ths securty n the portfolo (and reduces the absolute value f the securty s a lablty). An ncrease n the varance of a securty reduces ts portfolo share. These results are hardly surprsng and follow the standard lterature (Frexas and Rochet 1998, Hart and Jaffee 1974). Yet, the frst mportant result from equaton (8) s that unless ther vectors of excess returns are dentcal, domestc and foregn banks wll hold dfferent portfolos. For all practcal purposes, ths wll be the case. Ths result wll also hold for dfferent types of domestc banks to the extent that they have dfferent cost structures. Hence, the separaton theorem, whch says that all banks should hold the same co-lnear portfolo rrespectve of ther degree of rsk averson (Hart and Jaffee 1974), does apply only wthn subgroups of homogeneous banks but not between them. 6 Under certan parameter constellatons, some assets may not even be traded (Stulz 1981). 4 Impact of the Euro 4.1 Portfolo ecsons and the Euro The above framework can be used to analyze the reacton of banks to changng market opportuntes such as the ntroducton of the euro. For ths purpose, note that the frst order condtons for domestc and foregn loans are gven by: (9a) U = + + = 0 E U [ ] ( ) U [ ] [ * * r c r σ COV COV ], F Π σ Π 6 Ths result s dentcal to that of Stulz (1981) who assumes that domestc nvestors have to pay a tax proportonal to ther holdngs of foregn assets. In our framework, ths tax corresponds to the varable costs of cross-border transactons. kewse, Gehrg (1993) concludes that a market portfolo ceases to exst when asymmetres n nformaton are allowed for.

20 (9b) U * 19 = + + = 0 E U [ ] ( * * ) * U r c r [ ] [ * * σ COV COV ], F Π σ Π The most mportant change that the euro precptates s that t elmnates exchange rate rsks n Europe. The response of domestc loans to a declne n exchange rate rsk, n turn, s gven by σ $ U' σ U'' e e =. BecauseU ''< 0 holds n the optmum, the sgn of the numerator of ths term on the RHS determnes the sgn of the HS: (10a) U' σ e (10b) U' σ ε U U σ = * ρ σ' * ρ σ' * * σ + σ σ σ e > 0 < < 0 [ σ 1 COV13 COV14 ] σ > 0 > 0 > = < x1 0 U U σ = * * σ σ ' * ρ σ ' * σ COV * COV σ + σ σ { σ > 0 < 0 e > < 0 < 0 > 0 > 0 where the prme denotes the frst dervatve wth respect to the standard devaton of the exchange rate. Wthout addtonal assumptons, the effects of a change n exchange rate rsk on the demand for loans are undetermned. Assumng that the ndrect effects stemmng from the elmnaton of the exchange rate rsk on the second dervatve of the utlty functon are small, only the sgn of the frst terms on the RHS n (10a) and (10b) matters. If the bank holds no foregn deposts ( * = 0 ) and f loan returns are postvely correlated ( ρ 13 > 0 ), both domestc and foregn lendng rse f exchange rate rsks fall. The effect on foregn lendng s larger because t works drectly va the reducton of the exchange rate rsk whereas the effect on domestc lendng arses only va the correlaton of loans returns. Notce that the reverse result may be obtaned f foregn deposts are large and f loan and depost returns are postvely correlated ( ρ 14 > 0). In ths case, exchange rate rsks n lendng decsons would serve as a hedge aganst depost rate rsks. Wth the elmnaton of exchange rate rsks, the need to hedge exchange rate rsks would dmnsh, however, and demand for loans would be affected negatvely.

21 0 An extenson of the above framework would be to allow for short-sale constrants of banks to be bndng. We have so far assumed that excess returns on loans are postve such that banks offer a postve amount of loans n equlbrum. It s concevable, however, that the costs of offerng loans to foregn customers are prohbtvely expensve and that the net yeld from gong abroad turns negatve. In ths case, a short-sale constrant would become bndng and $ * = 0. Hence, even though the elmnaton of exchange rate rsks may tend to ncrease the ncentves to go abroad, ths parameter change mght not be suffcent to nduce actvtes n the foregn market. 7 An assessment of the mpact of the euro on the market opportuntes of banks gong beyond the exchange rate effect s to a large extent speculatve. One effect could be that the operatng costs of holdng foregn assets and labltes may fall as the costs of cross-border transactons declne, thus ncreasng the net return from gong abroad. Ths would ncrease banks ncentves to offer both deposts and loans abroad. In the medum- to long-run, however, the expansonary effect of the euro s not that clear-cut. If the ntroducton of a common currency fosters the ntegraton of both real and fnancal markets, t may actually rase the correlaton between rates of return on assets n the euro-zone. Hgher covarances of domestc and foregn assets would then lower the ncentves to expand across borders as European securtes would provde a poorer hedge aganst dosyncratc (country-specfc) rsks. An alternatve scenaro would be that ndustres become more concentrated across Europe whch would mply a decrease n return correlatons across countres. Cross-border bankng actvtes would thus become relatvely more attractve. Notce, however, that banks may also reap the benefts of dversfcaton by nvestng nto foregn securtes. Even f correlatons between returns fall, cross-border bankng actvtes may not ncrease f, at the same tme, securtzed assets become more readly avalable. Overall, we have thus two effects (reduced rsks and hgher net returns) whch would cause an expanson of actvtes wthn Euroland and one effect (hgher correlatons) whch may work nto the opposte drecton. The nterestng pont s that f the euro precptates a greater convergence of markets n Europe, t 7 A smlar reasonng would apply to foregn drect nvestments of banks. See Buch and app (1998) for detals.

22 1 may actually provde ncentves for banks to expand out- rather than nsde Europe. 4. Home Bas versus Currency Bas Emprcal evdence shows that nternatonal nvestment portfolos exhbt a strong bas towards assets ssued by home-country borrowers n the currency of the home country. In order to solate the effect of the euro on portfolo decsons, t s mportant to dscern whether ths asset allocaton s the result of a home or of a currency bas of nvestment portfolos (Buch and app 1998). Whle the home bas towards assets ssued by home-country borrowers wll reman even after the euro has been ntroduced, the currency bas wthn Euroland wll dsappear. Wthn the framework of ths model, the home bas n nvestment portfolos s captured through the costs of cross-border transactons whch can be defned n a broad sense as comprsng nformaton costs, dfferences n nsttutons, and more techncal transportaton costs. The currency bas, n contrast, s reflected by the fact that foregn transactons expose domestc banks to addtonal (exchange rate) rsks, n partcular f labltes are denomnated n the home currency. To see how the home and the currency bas are affected by the euro, a slghtly modfed model whch comprses three regons (domestc economy, Euroland E, the rest of the world R) nstead of two could be consdered. Costs of nternatonal transactons would exceed those n Euroland E R ( c < c < c ) because of a greater nsttutonal proxmty of European,,, markets, whch comes, not least, as a result of the ntegraton process. Obvously, the man mplcaton of the euro s that t elmnates the currency bas wthn Euroland. In terms of exchange rate rsks, Euroland assets become perfect substtutes for domestc assets. Ths should promote the expanson of banks wthn Euroland. At the same tme, the potental for dversfcaton wthn Euroland s reduced. Ths potental countervalng effect would nduce banks to expand outsde rather than nsde Europe. In addton, the home bas wthn Euroland remans to the extent that transacton costs, nsttutonal structures, and asymmetres n nformaton adjust only gradually.

23 4.3 Montorng Actvtes Apart from choosng optmal amounts of loans and deposts, the bank has to decde on the optmal amount of montorng whch can be derved from: (11) U µ [ Π ] [ Π ] [ Π ] [ Π ] U E U σ = + = 0 E µ σ µ Ths equaton essentally mples that the margnal costs of montorng must equal the margnal revenue n terms of a reducton n lendng rsk: (1) [ ] U * * * * * ( µ ) = ( ) ( ) + + * K' ' ' σ σ σ ρ σ ρ σ σ σ ρ σ ρ σ where the prme denotes the frst dervatve of the standard devaton wth respect to montorng,.e. σ' 1, σ ' 3 < 0. Hence, the response of the optmal amount of montorng wth respect to a change n exchange rate rsk s gven by: (13) µ $ U' σ = σ U'' e e U σ σ { ρ σ ρ σ * * * σ * * = ' σ σ σ { ' ρ σ U'' e e + σ e σ < < e > 0 13 Π < 0 > 0 < 0 > 0 < 0 where σ 3 µ σ e = 0 has been assumed. Wth U''< 0, montorng actvtes and exchange rate rsks are postvely related f the term n squared brackets s negatve. Agan, no defnte statement s possble because there are two postve and two negatve terms. Assumng that cross-border competton s confned to the loan market ( * = 0 ), a declne n exchange rate rsk reduces screenng actvtes. The ntuton behnd ths result s that f exchange rate rsks declne, the same level of rsk can be obtaned at a lower cost.

24 4.4 Balance of Payments Effects 3 The above results can be used to assess the mpact of a reducton (or elmnaton) of exchange rate rsk on the balance of payments. The current account balance (CUR) s gven by the trade balance (TB) plus the balance of servces (SB),.e. net nterest payments on cross-border captal flows: (14) CUR = TB( e) + SB where TB > 0 (e = exchange rate). The captal account s gven by the change n net foregn labltes of domestc resdents: (15) CAP = NF = F FA where CAP > 0 represent a net captal nflow, (N)F = (net) foregn labltes, FA = foregn assets of the domestc economy, and = change n stocks durng the perod. Note that banks affect cross-border captal flows through ther headquarters. Hence, f a foregn bank rases deposts on the domestc market, these deposts are placed n the foregn bank. Net foregn assets of the domestc economy are then gven by the loans granted by domestc banks abroad plus deposts rased by foregn banks on the home market. Net foregn labltes equal loans granted by foregn banks on the domestc market plus deposts rased by domestc banks abroad. Under the assumpton that all domestc (foregn) banks are dentcal, the captal account of the balance of payments thus reads: [ ] * * * * (15 ) CAP = NF = NFA = ( n + n j ) ( n j + n ) Net nterest recepts n the servce account are: (16) SB = n( * * * * ) + n* ( j j ) Assumng flexble exchange rates, central bank s net foregn assets do not change. The balance of payments s thus defned as: (17) BOP CUR + CAP = 0 The adjustment n the balance of payments nduced by an elmnaton of exchange rate rsks s thus gven by

25 (18) CAP σ ( e) & 4 * * * ( ' ') ( j ' j ') = n + n where the prme denotes the frst dervatve wth respect to exchange rate rsk. Hence, the captal account effect s undetermned a pror. It depends on the relatve degree of rsk averson of domestc and foregn banks, on ther cost structures, and on the sze of domestc and foregn markets (n and n * ). Moreover, the mpact of an elmnaton of the exchange rate rsk on captal moblty depends on the way the latter s defned. Measurng captal moblty n terms of ether the volume of gross captal flows or the degree of competton n bankng, captal moblty ncreases. Measurng t n terms of net captal flows, captal moblty may ncrease or decrease, dependng on the strength of the exchange rate effect on the varous captal account tems. 5 Market Integraton and Rsk Takng As European fnancal markets are becomng ncreasngly ntegrated, the queston arses whether ntegraton has an effect on the propensty of banks to take rsks. As the prevous dscusson has shown, a declne n exchange rate rsks may reduce montorng actvtes of banks. Ths secton dscusses recent contrbutons whch suggest that montorng and market ntegraton may n fact be nversely related. Hence, ntegraton would ncrease rsk takng and may undermne the stablty of the bankng system. These papers are revewed n the present secton, and the condtons under whch the negatve welfare effects of ntegraton are obtaned are analyzed. Azenman (1998) argues that the ntegraton of fnancal markets mght lead to a declne n economc welfare as t reduces the ncentves of banks to screen borrowers. 8 In what follows, we extend hs model to take the effects of fnancal market ntegraton on portfolo rsks and on the foregn bankng market nto account. A smple asymmetrc nformaton framework s used. Frms can nvest an amount nto a project whch s successful wth probablty p and yelds a 8 A related argument has been made by Gehrg (1998).

26 5 return X and whch s a falure and yelds a zero return wth probablty 1-p. Assumng that projects wth a low probablty of success yeld a lower return f successful as compared to projects wth a hgh probablty of success, X can be wrtten as a functon of p: X = X ( p) wth X '( p) < 0. The probablty of success and the return n the good state of the world are thus nversely related. Informaton about the actual probablty of success s prvate nformaton to entrepreneurs. Outsde nvestors such as banks can observe only the expected value of the pay-off. Consder the frm s profts under self-fnancng as a benchmark. Normalzng the return on alternatve nvestment opportuntes to zero, the frm s proft s gven by: (19) Π F px ( p) =. Assumng that the entrepreneur can choose among projects whch dffer n ther probablty of success, the frst order condton for a proft maxmum s: (19 ) p Π F = X + px ' = 0. Hence, the optmal probablty of success under self-fnancng s gven by: (0) p * s = X X' > 0 ( p) If the frm has no funds to fnance nvestment from nternal sources, t has the opton to obtan a bank loan at a real nterest rate r. Assumng lmted lablty of frms, entrepreneurs servce ther loans only n the good state of the world. Hence, net profts under bank fnance are gven by: [ r ] (1) Π F p X ( p) = In ths case, frms chose projects wth a probablty of success: () p * B r = X' X( p) ( p) whch s lower than that under self-fnance: (3) p r X X r ps = + = < 0 X' X' X' * * B ( p)

27 6 Ths s a standard result of the asymmetrc nformaton lterature: lmted lablty mples that the entrepreneur has a preference for rsky projects, and overnvestment occurs. As before, the bank can nvest nto a screenng technology. In contrast to the prevous analyss, we assume that screenng (µ) not only reduces the varance of project returns but also ncreases the probablty of success,.e. σ = σ ( µ ) and p = p( µ ) wth σ '< 0 and p'> Moreover, we now assume that screenng takes place on a project-byproject bass such that total screenng costs are obtaned by multplyng varable screenng costs k ( µ ) by the number of loans granted. The effcency of banks s captured through a shft parameter η,.e. the hgher η, the less effcent the bank s n usng the screenng technology. We assume that there are n dentcal banks present n the domestc market. Industry supply of loans and ndustry demand for deposts are thus gven by: = n and =. Under autarky, profts of a representatve bank are: n (4) ( µ ) η ( µ ) Π B = p r k r where r = domestc depost rate and = s the balance sheet restrcton. In contrast to Azenman, we assume that banks not only care about expected profts but also about the rskness of ther actvtes: σ = σ. Hence, the bank s utlty s gven by: U = U ( Π B,σ B ) The bank has two choce parameters. Assumng mperfectly compettve markets, t optmzes on the scale of ts actvtes () by takng the responses of the other compettors n the market as gven (Cournot-competton). We thus depart from the analyss of Azenman who assumes perfectly compettve markets whch, n the presence of restrctons on the free flow of captal, seems an unrealstc assumpton. In addton, the bank chooses the optmal amount of screenng. The frst condton for a proft maxmum s thus gven by: B (5) U U r σ pr p η r r k U B = + + Π B σ B = 0 9 Note that these condtons do not hold smultaneously n the general case but rather depend on the strength of adjustment of p and X wth respect to µ.

28 7 where = = 1 d j j = 0, d j = 0, j whch can be transformed nto: (5 ) U U 1 = r p 1 + r 1 + Π B nε r ε (, ) n (, r ) 1 U σ B ηk + = 0 σ B r r where ε(, r ) = < 0 s the elastcty of demand for domestc loans and r r ε(, r ) = > 0 s the elastcty of supply for domestc deposts. In the optmum, > 0 nε must hold. Equaton (5 ) can be used to derve the response of the optmal volume of lendng ( $ ) to changes n lendng and depost rates as well as to changes n the effcency of screenng: 10 (6a) U 1 p 1+ $ U nε, r r Π B = = r U U ( ) > 0 (6b) U + $ U nε, r Π B 1 1 = = r U U ( r ) < 0 (6c) U k $ U η Π = = η U U B < 0 because U <0 n the proft maxmum. Hence, the scale of actvtes ncreases n the lendng rate and declnes n the depost rate and n the degree of neffcency of the bankng system. In addton to choosng the scale of ts actvtes, the representatve bank maxmzes profts by choosng screenng accordng to: 10 In addton, ndrect effects whch result from the second dervatve of the utlty functon wth respect to profts, e.g. U Π B, would have to be taken nto account. If we assume that the drect effects always domnate these ndrect effects, we Π B r can drop the latter n order to smplfy the exposton.

29 (7) [ η ] 8 U U U σ p r k B = ' ' + = 0 µ Π σ µ B where the prme denotes the frst dervatve wth respect to µ. Equaton (7) can be used to derve the response of banks optmal screenng actvtes $µ to changes n effcency and n nterest rates: B (7a) (7b) µ $ r µ $ r U $ p ' + U µ r Π B r = = U U µµ µµ U U µ r Π = = U B ( p' r ηk ') > 0 $ r p' r p' $ η ' r $ k + > 0 U < µµ µµ (7c) µ $ η U U µη Π = = U µµ B $ η ' ' ' $ r k + p r + p η ' η k < 0 U where U µµ s the second dervatve of the utlty functon wth respect to montorng and U µµ < 0 n the optmum. A pror, the sgns of the expressons are undetermned. In (7a), screenng always ncreases n the lendng rate f the frst, drect effect domnates the second, ndrect effect. Ths s because the term n round brackets may become negatve. Montorng declnes f the depost rate declnes f the absolute value of the second and the thrd term n the brackets exceed the frst term n (7b) because of (6b). Fnally, the mpact of an ncrease n the neffcency of banks (hgher η) on screenng s negatve f the drect effect domnates. These latter two results conform to Azenman who fnds that lower depost rates and less effcent banks ncrease the rskness of projects. Under autarky, smlar condtons can be derved for the bankng system n the foregn country. Now, consder what happens n a two-country-model f the countres moves from autarky to an ntegrated captal market. Azenman argues that fnancal lberalzaton (.e. the abolton of captal controls) has two effects. On the one hand, fnancal lberalzaton ncreases the level of rsk takng f, pror to the openng of the captal account, domestc depost rates were below the nternatonal nterest rate level. On the other hand, f-

30 9 nancal lberalzaton by ncreasng the effcency of fnancal ntermedaton ncreases the amount of montorng. These results are manly due to the fact that the analyss has been restrcted to a sngle country. Hence, potental welfare mplcatons for the foregn country va changes n nterest rates there have not be consdered. Takng these nto account, however, the followng welfare mplcatons arse: If countres are symmetrc, the only effect of ntegraton s that t ncreases the degree of competton n domestc fnancal markets as the number of compettors ncrease from n and n*, respectvely, to n + n*. endng rates fall and depost rates ncrease. ower lendng rates, n turn, affect negatvely banks propensty to screen borrowers. Hgher depost rates, n contrast, would rase screenng actvtes. If countres are asymmetrc, ntegraton not only affects the number of compettors but also the relatve supply of funds n each country. If, under autarky, savngs are relatvely scarce n the domestc economy whle they are relatvely abundant n the foregn economy, domestc nterest rates exceed foregn nterest rates. After fnancal lberalzaton, foregn captal thus flows nto the domestc economy, and nterest rates converge to a common world level. 11 Foregn lendng rates ncrease and domestc lendng rates declne. Ths has postve effects on the screenng actvtes of foregn banks and negatve effects on the screenng actvtes of domestc banks. The net effect s undetermned and the depends on the relatvely sze of the economes. As n the Azenman-model, ncreased effcency of fnancal ntermedaton ncreased montorng. Fnally, the possblty to lend abroad allows hedgng of dosyncratc rsks f the correlaton of loan returns s below one. Ceters parbus (.e. at the same level of actvtes), portfolo rsks are thus lower than under autarky. In summary then, fnancal market ntegraton has postve welfare mplcatons because t enhances the effcency of fnancal ntermedaton and because t gves banks better dversfcaton opportuntes. These effects are 11 For smplcty, ths argument gnores the spread between depost and lendng rates.

31 30 mtgated by the negatve effects that changes n nterest rates have on the propensty of banks to montor ther clents. The net effect depends on the relatve sze of markets, on the compettve structure of markets, and on relatve demand and supply condtons. Overall, the precedng dscusson mght therefore provde a ratonale for the need to strengthen bankng supervson when allowng a greater ntegraton of fnancal markets. 6 Summary espte the substantal efforts that have been made to level the playng feld for banks n Europe and to abolsh barrers to the free flow of captal, crossborder bankng actvtes n Europe reman modest. The purpose of ths paper has been to provde a framework for analyzng the lnks between bankng actvtes and captal moblty whch can explan ths dchotomy. Its man argument has been that due to asymmetres n nformaton and other costs of cross-border bankng actvtes, commercal banks asset portfolos can be expected to exhbt a relatvely strong home bas. These results have been derved n a standard mean-varance framework. Although the focus has been on cross-border portfolo choces of banks, an extenson to other market partcpants would be straghtforward. The nsghts of the paper have furthermore been used to derve mplcatons for Euroland. It has been argued that the euro affects captal flows because t elmnates exchange rate rsks wthn Euroland. The persstence of transacton costs of cross-border fnancal flows and of asymmetres n nformaton, however, lmts the mpact of the euro. Three results are noteworthy: Frst, the ntroducton of the euro wll stmulate captal flows wthn Europe as t elmnates exchange rate rsks and thus ncreases the ncentves of (rsk-averse) nvestors to go abroad. Instead of assumng rsk averson of banks, the lack of rsk neutralty on whch these results are based could be endogenzed by assumng that banks have to meet captal-adequacy requrements. The effect of the euro s strengthened f other costs of crossborder transactons declne as well. Gross captal flows and competton n fnancal markets unequvocally ncrease. The mpact on net captal flows s