Deusches Insiu für Wirschafsforschung www.diw.de Discussion Papers 940 Guglielmo Maria Caporale Chrisophe Raul Rober Sova Anamaria Sova Financial Developmen and Economic Growh: Evidence from Ten New EU Members Berlin, Ocober 2009
Opinions expressed in his paper are hose of he auhor and do no necessarily reflec views of he insiue. IMPRESSUM DIW Berlin, 2009 DIW Berlin German Insiue for Economic Research Mohrensr. 58 10117 Berlin Tel. +49 (30) 897 89-0 Fax +49 (30) 897 89-200 hp://www.diw.de ISSN prin ediion 1433-0210 ISSN elecronic ediion 1619-4535 Available for free downloading from he DIW Berlin websie. Discussion Papers of DIW Berlin are indexed in RePEc and SSRN. Papers can be downloaded free of charge from he following websies: hp://www.diw.de/english/producs/publicaions/discussion_papers/27539.hml hp://ideas.repec.org/s/diw/diwwpp.hml hp://papers.ssrn.com/sol3/jeljour_resuls.cfm?form_name=journalbrowse&journal_id=1079991
Financial Developmen and Economic Growh: Evidence from Ten New EU Members Guglielmo Maria Caporale Brunel Universiy (London), CESifo and DIW Berlin Chrisophe Raul LEO, Universiy of Orleans and IZA Rober Sova CES, Sorbonne Universiy and ASE Anamaria Sova CES, Sorbonne Universiy and EBRC Ocober 2009 Absrac This paper reviews he main feaures of he banking and financial secor in en new EU members, and hen examines he relaionship beween financial developmen and economic growh in hese counries by esimaing a dynamic panel model over he period 1994-2007. The evidence suggess ha he sock and credi markes are sill underdeveloped in hese economies, and ha heir conribuion o economic growh is limied owing o a lack of financial deph. By conras, a more efficien banking secor is found o have acceleraed growh. Furhermore, Granger causaliy es indicae ha causaliy runs from financial developmen o economic growh, bu no in he opposie direcion. Keywords: Financial Developmen, Economic Growh, Causaliy Tess, Transiion Economies JEL classificaion codes: E44, E58, F36, P26 Corresponding auhor: Professor Guglielmo Maria Caporale, Research Professor a DIW. Cenre for Empirical Finance, Brunel Universiy, Wes London, UB8 3PH, UK. Tel.: +44 (0)1895 266713. Fax: +44 (0)1895 269770. Email: Guglielmo-Maria.Caporale@brunel.ac.uk
1. Inroducion The relaionship beween financial developmen and economic growh has been exensively analysed in he lieraure. Mos empirical sudies conclude ha he former, ogeher wih a more efficien banking sysem, acceleraes he laer (Levine, 1997, 2005; Wachel, 2001). Levine (2005) suggess ha financial insiuions and markes can foser economic growh hrough several channels, i.e. by (i) easing he exchange of goods and services hrough he provision of paymen services, (ii) mobilising and pooling savings from a large number of invesors, (iii) acquiring and processing informaion abou enerprises and possible invesmen projecs, hus allocaing savings o heir mos producive use, (iv) monioring invesmen and carrying ou corporae governance, and (v) diversifying, increasing liquidiy and reducing ineremporal risk. Each of hese funcions can influence saving and invesmen decisions and hence economic growh. Since many marke fricions exis and laws, regulaions, and policies differ markedly across economies and over ime, improvemens along any single dimension may have differen implicaions for resource allocaion and welfare depending on oher fricions in he economy. The reform of he financial secor in Cenral and Easern Europe (CEE) sared from he banking secor. Is ransformaion has been one of he mos imporan aspecs of he ransiion process from a cenrally planned o a marke economy. Iniially a heavily regulaed indusry, he banking sysem has been rapidly urned ino one of he mos dynamic secors of he economy. The process sared in he early 1990s when foreign banks began invesing in he region. From 2004, hese have been holding majoriy shares in all CEE counries. Their enry ino he marke has resuled in considerable benefis for he secor and he economy in general, bu hey have had o face various challenges deriving mosly from he underdevelopmen of key insiuional suppor for banking growh. Alhough accession o he European Union (EU) has helped he reform process in he CEE counries, real convergence in erms of real GDP per capia remains a challenge. The presen sudy invesigaes wheher financial developmen can be insrumenal in reducing he gap visà-vis he oher EU members. Specifically, afer reviewing he main feaures of he banking and financial secors in hese counries, i examines he empirical linkages beween financial developmen and economic growh by esimaing a Barro ype growh regression augmened 1
wih he inclusion of financial variables using panel daa for en ransiion counries over period 1994-2007. As financial developmen varies considerably across hese counries, we spli hem ino hree more homogenous groups: Cenral and Easern European counries (CEE-5), Balic counries (B-3) and Souheasern European counries (SEE-2). We also consider he deerminans of cred given is imporance for financing invesmen projecs and is impac on economic growh. We analyse hese issues by employing he sysem GMM mehod o conrol for endogeneiy and measuremen errors and obain unbiased, consisen and efficien esimaes. Finally, Granger causaliy ess are carried ou. The layou of he paper is he following. Secion 2 provides a brief review of he lieraure on he relaionship beween finance and growh. Secion 3 analyses he evoluion of he financial and banking secor in en ransiion economies. Secion 4 discusses he daa and he economeric approach, as well as he panel evidence on he nexus beween financial developmen and economic growh. Secion 5 carries ou bi-direcional causaliy ess beween financial developmen/efficiency of he banking sysem and economic growh. Secion 6 offers some concluding remarks. 2. Lieraure Review The relaionship beween financial developmen and economic growh is a conroversial issue. Some auhors consider finance an imporan elemen of growh (Schumpeer, 1934; Goldsmih, 1969; McKinnon, 1973; Shaw, 1973; King and Levine (1993), whils for ohers i is only a minor growh facor (Robinson, 1952; Lucas, 1988). Schumpeer (1934) sees he banking secor as an engine of economic growh hrough is funding of producive invesmen. On he conrary, Lucas (1988) argues ha he role of finance has been oversressed. Greenwood and Jovanovic (1990) model he dynamic ineracions beween finance and growh and emphasise he wo-way causaliy beween hem. Financial inermediaries produce beer informaion and improve resource allocaion. An expanded sysem of financial inermediaion is able o allocae more capial o efficien invesmens and hus o foser economic growh. Bencivenga and Smih (1991) highligh he fac ha, by eliminaing liquidiy risk, banks can raise economic growh. Financial inermediaries boos produciviy, capial accumulaion and growh by improving corporae governance. 2
Exising sudies ypically focus on variables capuring he size, aciviy or efficiency of specific financial insiuions or markes. Early conribuions used aggregae daa on banks for a large number of developed and developing counries including he raio o GDP of moneary variables (M2 or M3), or financial deph indicaors (credi o he privae secor). Laer sudies on he link beween financial developmen and economic growh have added indicaors of he size and liquidiy of sock markes, bu hese are available for fewer counries and shorer ime periods. The same applies o indicaors of he efficiency and compeiiveness of financial insiuions. Single-counry sudies allow researchers o use more exensive micro-based daa and/or analyse specific policy measures or reforms. Goldsmih s paper (1969) was he firs o show empirically he exisence of a posiive relaionship beween financial developmen and GDP per capia. King and Levine (1993) used mosly moneary indicaors and measures of he size and relaive imporance of banking insiuions and also found a posiive and significan relaionship beween several financial developmen indicaors and GDP per capia growh. Levine and Zervos (1996) included measures of sock marke developmen and found a posiive parial correlaion beween boh sock marke and banking developmen and GDP per capia growh. More precisely, hey repored a posiive and significan link beween liquidiy of sock markes and economic growh, bu no robus relaionship beween he size of sock markes and economic growh. Levine e al. (2000) found ha he developmen of financial inermediaion affecs growh posiively, and ha cross-counries differences in legal and accouning sysem largely accoun for differen degrees of financial developmen. More recenly, some auhors have suggesed ha here is a posiive relaionship beween financial deepening and per capia income in he ransiion economies (Éger e al., 2007; Backé e al., 2007). A posiive effec of financial developmen on economic growh hrough is sources (capial accumulaion and produciviy), and even on income inequaliy and povery, has also been repored (de Haas, 2001; Levine, 2005). Only a few sudies have focused on he ransiion economies from Cenral and Easern Europe (Bonin and Wachel 2003, Bonin e al., 2005; Hermes and Lensink, 2000; Berglöf and Bolon, 2002; Haas, 2001; Fink e al., 2005, 2008; Kenourgios and Samias (2007), mosly finding a posiive relaionship beween several financial indicaors and economic growh. Hermes and Lensink (2000) provide an overview of he main relevan issues, in 3
paricular he role of sock markes in he process of financial inermediaion (wih an emphasis on he imporance of regulaion in hese markes), and he role of deposi insurance o improve sabiliy of he banking secor. Berglöf and Bolon (2002) find ha he link beween financial developmen and economic growh does no appear o be very srong during he firs decade of ransiion, a leas when one looks a he raio of domesic credi o GDP. Kenourgios and Samias (2007) examined he long-run relaionship beween finance and economic growh for Poland and concluded ha credi o he privae secor has been one of he main driving forces of long-run growh. Hagmayr e al. (2007) invesigaed he finance-growh nexus in four emerging economies of Souheasern Europe for he period 1995-2005 and found a posiive and significan effec of bond markes and he capial sock on growh. Fink e al. (2005), using a sample of 33 counries (11 ransiion economies and 22 marke economies), found ha financial developmen has posiive growh effecs in he shor run raher ha in he long run. Fink e al. (2008) invesigaed he impac of he cred bond and sock segmens in nine EU-accession counries over he early ransiion years (1996 2000) and compared hese o maure marke economies and o counries a an inermediae sage. They found ha he ransmission mechanisms differ, and ha financial marke segmens wih links o he public secor (bu no o sock markes) conribued o sabiliy and growh in he ransiion economies. Winkler (2009) reviews he process of rapid financial deepening and he associaed vulnerabiliy and risks for he Souheasern European counries. He argues ha he sraegy of pursuing financial developmen hrough he enry of foreign banks does no guaranee financial sabiliy. Finally, a srong consensus has emerged in he las decade ha well-funcioning financial inermediaries have a significan impac on economic growh (Bonin and Wachel, 2003). 3. The Banking and Financial Secor in he Transiion Economies In he cenrally planned economies, money played only a limied role as a medium of exchange. In he banking secor, he cenral bank combined he sandard funcions of moneary auhoriies wih some of hose of a commercial bank. Besides, in mos economies here were banks specialising in differen secors, namely expor rade operaions, financing of long-erm invesmen, and he agriculure and food indusry. A he ime, here was only a sae savings bank collecing available resources and household deposis. Thus, banking 4
aciviies were characerised by segmenaion along funcional lines. The ransacions wihin he sae secor, including hose beween sae-owned producion enerprises, involved no moneary paymen while households used cash for ransacions. The firs sep in he ransiion process for he financial secor was he developmen of markeoriened financial insiuions, banks being he mos visible and ofen he dominan ones. The ransiion o a marke economy sared in he CEE counries in 1991 wih reforms of he banking secor. In all ransiion counries, he firs sep was he aboliion of he mono-bank sysem. New banking legislaion was inroduced allowing privae banks o develop and foreign financial insiuions o ener he domesic banking secor. Banks were allowed o operae as universal rade banks, whils he new Cenral Bank remained in charge of moneary policy, including exchange rae policy, and monioring of he newly creaed banking secor. The new sysem was very similar o ha already exising in EU. Thus, mos ransiion counries experienced a rapid expansion of he banking secor due o he enry of new (foreign) banks and he decline in sae ownership. The ransiion generaed macroeconomic urbulence and made any new bank lending exremely risky. During he 1990s, he increase in socks of non-performing loans led o banking crises in many ransiion counries. The sock of bad loans evolved parly as a resul of he gradual recogniion of he qualiy of exising relaionships in sae-owned banks (he sock issue), and parly because of coninuing bad lending pracices (he flow problem) (Bonin and Wachel, 2003). The privaisaion of he sae-owned banks and he paricipaion of foreign sraegic invesors in banking represened effecive ways o solve hese problems. Thus, progress in he banking secor in CEE counries has led o a smaller amoun of nonperforming loans. Foreign banks have played an imporan role in he developmen of he financial sysem of he CEE counries by increasing credi availabiliy, echnology ransfers and compeiion. They have been more innovaive in erms of he number and range of new producs offered, some of hem already available in he foreign banks home markes. Besides, hey have helped consolidae he CEE s banking sysems, producing waves of mergers and acquisiions ha have decreased he number of banks. The majoriy of banks in he newly privaised banking secor are in fac foreign owned. 5
Financial indicaors of he developmen of he banking secor in several ransiion economies are shown in Table 1. Table 1: Main financial indicaors of banking secor developmen Year Counry Number of oal banks Number of foreign owned banks Asse share of sae owned banks (%) Asse share of foreign owned banks (%) 1996 2007 1996 2007 1996 2007 1996 2007 Bulgaria 49 29 3 21 82.2 21.0 29.3 82.3 Czech.Rep 53 37 3 15 69.9 2.4 19.0 84.8 Esonia 15 15 4 13 6.6 0.0 1.6 98.7 Hungary 42 40 26 27 15.3 3.7 46.2 64.2 Lavia 34 25 18 14 6.9 4.2 51.5 63.8 Lihuania 12 14 3 6 54.0 0.0 28 91.7 Poland 81 64 28 54 51.6 19.5 16 75.5 Romania 31 31 10 26 80.9 5.7 10.7 87.3 Slovakia 29 26 14 15 54.2 1.0 12.7 99 Slovenia 36 27 4 11 40.7 14.4 5.3 28.8 Source : EBRD As can be seen, he majoriy of banks have been privaised and foreign banks hold he larges share of asses. This has increased sharply in he pas decade in all ransiion counries, while he level of sae ownership has fallen below 20 % in each counry. Thus, he influence of he sae-owned banks has declined subsanially. In 2007, no sae-owned bank exised any longer in Esonia and Lihuania. The enry of foreign banks ino he local marke had a posiive influence by increasing compeiion and efficiency of he banking sysem, encouraging beer regulaion of he financial secor in he form of banking supervision, and enhancing access o inernaional capial. In addiion, he higher efficiency of foreign banks has simulaed economic growh, and he paricipaion of foreign sraegic invesors in banking is an effecive way o avoid bad loans. Almos all ransiion counries have experienced a decline in he number of banks. For example, in Bulgaria his has fallen from 49 in 1996 o 29 in 2007. Many smaller banks 6
became insolven owing o sricer regulaions for banking supervision. An excepion is Lihuania, where he number of banks increased from 12 in 1996 o 14 in 2007. 3.1 Liquid Liabiliies The raio of liquid liabiliies o GDP is an indicaor of he size of he financial secor. The highes moneisaion raios are found in Slovenia (74.4% in 2007). Romania has recorded a decline in his raio (from 46% in 1991 o 36% in 2007) and has now he lowes one. Generally, he raio of broad money o GDP is a leas 60% in high-income counries wih developed banking secors. Thus, he banking secors in he ransiion economies canno be considered o be highly developed wih a few excepions. 7
3.2 Privae secor lending growh Mos ransiion counries have recorded high privae secor lending growh in recen years. This expansion of credi has been a feaure of he ransiion counries, foreign banks being he main source of credi for he privae secor (see Table 2). Table 2. The evoluion of he raio of privae secor credi o GDP (in percen) Year 2000 2001 2002 2003 2004 2005 2006 2007 Counry Bulgaria 12.5 14.8 19.4 26.7 35.2 42.9 47.1 66.8 Czech.Rep 44.0 33.0 29.4 30.7 31.6 35.8 40.0 41.0 Esonia 23.3 24.3 26.0 30.7 39.7 57.0 78.2 89.3 Hungary 29.9 30.9 33.6 41.0 44.6 49.8 54.1 59.2 Lavia 21.5 26.3 29.5 40.2 50.8 68.2 87.5 93.9 Lihuania 11.3 13.5 16.2 22.9 28.8 41.3 50.6 61.2 Poland 26.9 28.0 28.2 29.2 27.5 29.2 33.4 35.2 Romania 7.2 8.7 10.1 13.7 15.7 20.0 26.1 32.9 Slovakia 43.7 33.0 30.8 31.6 30.1 34.7 38.6 42.3 Slovenia 36.7 38.8 38.6 41.3 48.1 56.4 65.9 79.0 Source: EBRD Empirical sudies sugges a posiive relaionship beween credi o he privae secor and per capia income in he ransiion economies (Coarelli e al., 2005). However, he banking sysem in he CEE counries appears o be more and more dependen on he aciviies of foreign banks. These, mainly from he EU counries, conrol he majoriy of asses and capial flows in he financial markes. Their enry has indeed boosed economic growh, enhanced compeiion and conribued o arac foreign direc invesmen. However, he lack of effecive ani-rus legislaion and mergers and acquisiions can lead o excessive concenraion, while ani-compeiive pracices and abuse of dominan posiion may also occur. In mos CEE counries he financial archiecure has converged owards a bank-based sysem wih subsanial foreign ownership. 8
3.3 Household lending growh Anoher feaure of he ransiion economies was he rapid growh of consumer credi resuling from an increase of public confidence in he banking secor as well as in per capia income. Currenly, he main business in he banking secor is indeed consumer credi (including credi cards and morgage loans). Is growh also reflecs he anicipaion of higher fuure income and consumpion smoohing. However, his conribues o widening curren accoun deficis hrough increased demand for impored consumer goods and currency appreciaion. One of he reasons for he boom in consumer lending is he relaive unaraciveness of wholesale lending owing o insiuional weaknesses, above all he poor funcioning of he legal sysem. Table 3 gives some informaion abou he evoluion of household lending growh. Table 3 Evoluion of credi o households in percen of GDP Year 2000 2001 2002 2003 2004 2005 2006 2007 Counry Bulgaria 2.1 2.8 3.7 7.1 10.0 14.4 16.6 23.0 Czech.Rep 5.6 5.9 7.3 9.1 11.2 13.8 16.5 20.0 Esonia 7.1 8.4 10.6 14.3 19.7 28.1 38.2 43.3 Hungary 3.2 4.7 7.4 10.9 12.8 15.6 18.5 21.7 Lavia 3.3 4.6 7.3 11.6 17.6 26.8 38.0 42.7 Lihuania 1.3 1.5 2.4 4.2 7.1 12.0 17.9 24.4 Poland 7.5 8.7 9.4 10.3 10.6 12.4 15.6 20.0 Romania 1.2 1.7 1.9 3.8 4.8 7.2 11.2 17.7 Slovakia 4.7 5.1 5.5 7.0 8.6 11.2 13.1 16.3 Slovenia 11.3 10.9 10.5 10.8 12.2 14.8 17.0 19.2 Source: EBRD Widening curren accoun imbalances are a concern for policy-makers, and measures migh be necessary o slow down he growh in credi o households and o allocae more resources o producive invesmens. A he same ime, he financial infrasrucure should be improved as crediors need proecion hrough he enforcemen of bankrupcy and insolvency legislaion meeing inernaional sandards. In addiion, improving corporae governance and 9
providing beer credi informaion migh help banks channel resources owards he producive corporae secor. 3.4 Sock marke capialisaion The marke capialisaion raio measures he size of he sock marke and is equal o he value of lised domesic shares divided by GDP. Sock marke capialisaion in he ransiion counries grew due o he privaisaion process. However, he developmen of he sock marke was affeced by he economic and financial crisis ha he ransiion economies have experienced. A he end of 2007, hese counries sill displayed differen levels of sock marke developmen, is capialisaion ranging from 8.6 % o 57.2 % in he counries covered in his sudy, being a is lowes in Slovakia and a is highes in Slovenia (see Table 4). Table 4 Evoluion of sock marke capialisaion in percen of GDP Counry Year 2000 2001 2002 2003 2004 2005 2006 2007 Bulgaria 4.8 3.7 4.2 7.9 10.4 19.7 31.1 51.3 Czech.Rep 18.9 14.1 19.4 17.6 24.5 31.6 31.6 37.4 Esonia 31.5 24.1 29.9 38.4 47.1 25.2 34.6 26.9 Hungary 25.1 18.7 17.2 18.3 25 31.6 33.8 32.4 Lavia 7.3 8.4 7.3 9.5 11.5 16.5 12.9 10.8 Lihuania 13.9 9.9 9.3 16.9 26.1 31.7 32.6 24.7 Poland 17.4 13.2 13.6 16.5 23 31.1 40.9 44.1 Romania 3.4 5.8 10.1 9.2 13.9 22.2 24.4 27.3 Slovakia 6.3 7.4 6.8 7.4 9.4 9.4 8.8 8.6 Slovenia 16.8 16.8 24.1 22.5 26.2 22 37.2 57.2 Source: EBRD Despie an upward rend, he figures sill remain below he corresponding ones for he EU developed economies. Capial marke developmen is complicaed by he need o suppor he developmen of insiuional infrasrucure and regulaory mechanisms. Overall, here has been significan progress in he banking secor, as also indicaed by he EBRD index of banking secor reform (see Table A2 in he Appendix). 10
4. Financial Developmen and Economic Growh: Empirical Analysis In his secion, we analyse he linkages beween financial developmen/efficiency and economic growh using panel daa for en ransiion counries during he period 1994-2007. Firs, we esimae he impac of financial indicaors over he whole sample. Second, we spli he daa ino subpanels corresponding o hree more homogenous groups of counries and compare he resuls. 4.1 The Model To sudy he relaionship beween finance and growh we esimae an augmened Barrogrowh regression including financial developmen variables which akes he following form: [ FINANCE] + γ i[ CONDITIONINGSET ε i GROWTH α i + β i ] +, = (1) or g y y 1 = α i + β i f + γ ic + μi + ε = (2) where y is real GDP per capia, g is growh rae, f an indicaor of financial developmen, C a se of condiioning variables, μ i and ε error erms, i (where i = 1,2,.N) he observaional uni (counry), and (where =1,2,,T) he ime period, while ε is a whie noise error wih zero mean, and μ a counry-specific componen of he error erm ha does no necessarily have a zero mean. The parameer α i is he counry-specific inercep which may vary across counries. One imporan issue concerning he link beween financial secor developmen and growh is he difficuly o idenify proxies for measuring hem. Beck e al. (2000, 2008) discuss differen indicaors of financial developmen capuring he size, aciviy and efficiency of he financial secor, insiuions or markes. In our analysis, we consider several indicaors, namely: he raio of credi o he privae secor o GDP as a measure of financial deph; indicaors of he size of sock markes as sock marke capialisaion (as a percenage of GDP); moneisaion variables such as he raio of broad money o GDP as a measure of he size of he financial secor; indicaors of he efficiency and compeiiveness of he financial sysem such as he margin beween lending and deposi ineres raes and he EBRD ransiion index of financial insiuional developmen. Deails are provided below. 11
Aciviy of he financial secor: - The raio of credi o he privae secor o GDP (DCPS), which is he value of loans made by banks o privae enerprises and households divided by GDP, is used as a measure of financial deph and banking developmen. This indicaor isolaes credi issued by banks, as opposed o credi issued by he cenral bank, and credi o enerprises, as opposed o credi issued o governmens (Levine and Zervos, 1996). Size of he financial secor - The sock marke capialisaion o GDP raio (STMC), which is an indicaor of he size of he financial secor given by he marke value of lised shares divided by GDP. Alhough large markes do no necessarily funcion effecively and axes may disor incenives o lis on he exchange, he marke capialisaion raio is frequenly used as an indicaor of marke developmen. - Liquid liabiliies o GDP raio (LLG), which equals liquid liabiliies of he financial sysem divided by GDP. I is used as a measure of "financial deph" and hus of he overall size of he financial inermediaion secor (King and Levine,1993a). Efficiency of he financial secor - The ineres rae margin (INT), which measures he difference beween deposi and lending raes in he banking marke is used o measure he efficiency of he secor. Levine (1997) suggesed several possible indicaors for economic growh: real per capia GDP growh, average per capia capial sock growh and produciviy growh. Here we use real per capia GDP growh. Oher variables influencing economic growh were inroduced in our model, including per capia income, average educaion, poliical and sabiliy indicaors as well as indicaors reflecing rade, fiscal and moneary policy such as governmen consumpion or rade openness and inflaion. 12
In he esimaion we used real GDP per capia wih a one-year lag as iniial income per capia o conrol for he seady-sae convergence prediced by he neoclassical growh model. For human capial, we inroduced a proxy for educaional aainmen, more precisely he secondary school enrollmen raio whose expeced influence on growh is posiive hrough is effec on produciviy. Inernaional rade openness is proxied by an inernaional rade policy variable, i.e. he rade o GDP raio, wih an expeced posiive coefficien. Higher openness enhances growh hrough higher compeiion and echnological progress (see Winer, 2004). Inflaion measures he degree of uncerainy abou he fuure marke environmen, firms becoming more relucan o make long-run commimens in he presence of higher price variabiliy; he expeced sign of his variable is herefore negaive. 1 The esimaed model, which includes a proxy for financial developmen, is he following: RGDPC + β DCPS 7 = α i + β1rgdpc 1 + β STMC 8 + β LLG 9 + β INV 2 + β RI 10 + β TOP 3 + β INT 11 + β INFL 4 + u i + ε + β GVE 5 + β HC 6 + (3) where: RGDPC = real per capia GDP growh; RGDPC = iniial income per capia; INV = invesmen/gdp (percenage); TOP = rade/gdp (percenage); INFL = inflaion, average consumer prices; GVE = governmen expendiure/gdp; HC = secondary school enrollmen raio; DCPS = domesic credi o he privae secor (as a percenage of GDP); STMC = sock marke capialisaion (as a percenage of GDP); LLG = liquid liabiliies (as a percenage of GDP); RI = Reform index of financial insiuional developmen (which is he average of he EBRD s indices of banking secor reform and of reform of non-bank financial insiuions); INT = ineres rae margin. 4.2 Daa Our panel consiss of daa for en ransiion counries from Cenral and Easern Europe over he period 1994-2007. The daa are annual and he counries included in he sample are: 1 Oher sudies on he finance-growh nexus for he ransiion economies including inflaion as a condiioning variable are Rousseau and Wachel, 2002; Gillman and Harris, 2004. 13
Bulgaria, Czech Republic, Esonia, Hungary, Lavia, Lihuania, Poland, Romania, Slovakia and Slovenia. We also carry ou he analysis for hree more homogeneous sub-groupings: (a) he Balic counries (B-3): Esonia, Lavia and Lihuania; (b) he CEE-5: he Czech Republic, Hungary, Poland, Slovakia and Slovenia; (c) Souheasern Europe (SEE-2): Bulgaria and Romania. The daa were obained from he EBRD daabase and he Inernaional Moneary Fund (IFS). For more deails on daa sources and definiions, see he Appendix. 4.3 Mehodology The mos common mehods for invesigaing he finance-growh nexus are cross-counry regressions and panel daa echniques. Noe ha he esimaes of β i (financial developmen indicaors) can be biased for a variey of reasons, among hem measuremen error, reverse causaion and omied variable bias. Therefore, a suiable esimaion mehod should be used in order o obain unbiased, consisen and efficien esimaes of his coefficien. To deal wih hese biases, researchers have uilised dynamic panel regressions wih lagged values of he explanaory endogenous variables as insrumens (see Beck e al., 2000; Rioja and Valev, 2004). Such mehods have several advanages over cross-secional insrumenal variable regressions. In paricular, hey conrol for endogeneiy and measuremen error no only of he financial developmen variables, bu also of oher explanaory variables. Noe also ha, in he case of cross-secion regressions, he lagged dependen variable is correlaed wih he error erm if i is no insrumened (see Beck, 2008). The dynamic panel regression akes he following form: 1 2 g α i + βf + γ 1C + γ 2C + δy 1 + μi + λ + ε = (4) where C 1 represens a se of exogenous explanaory variables, C 2 a se of endogenous explanaory variables, and λ a vecor of ime dummies. In our analysis, we employ he sysem GMM esimaor developed by Arellano and Bover (1995), which combines a regression in differences wih one in levels. Blundell and Bond (1998) presen Mone Carlo evidence ha he inclusion of he level regression in he 14
esimaion reduces he poenial bias in finie samples and he asympoic inaccuracy associaed wih he difference esimaor. The consisency of he GMM esimaor depends on he validiy of he insrumens used in he model as well as he assumpion ha he error erm does no exhibi serial correlaion. In our case, he insrumens are chosen from he lagged endogenous and explanaory variables. In order o es he validiy of he seleced insrumens, we perform he Sargan es of overidenifying resricions proposed by Arellano and Bond (1991). In addiion, we also check for he presence of any residual auocorrelaion. Finally, we perform saionariy ess belonging o he firs- (Levin-Lin-Chu, 2002) and second-generaion uni roo es (Pesaran, 2007), (see he Appendix for deails). The resuls sugges ha all series are saionary (see Table A5 in he Appendix), and consequenly no co-inegraion analysis is necessary. Therefore we proceed direcly o he GMM esimaion. 15
4.4 The esimaion resuls The dynamic panel regressions were run boh for he en ransiion economies as a whole and he hree subgroupings menioned before. The esimaion resuls are presened in Tables 5 and 6. Table 5: The financial developmen and economic growh nexus: dynamic panel regression (1) (2) Variables RGDPC RGDPC L.RGDPC 0.229 0.201 (3.40)*** (4.62)*** INV 0.292 0.342 (4.50)*** (5.50)*** TOP 0.015 0.011 (2.21)** (2.33)** INFL -0.008-0.006 (3.59)*** (4.01)*** GVE -0.057-0.066 (2.56)** (5.66)*** HC 0.018 0.020 DCPS (3.61)*** (3.61)*** 0.007 STMC (0.23) 0.004 LLG (2.95)*** 0.013 RI (2.42)** 0.493 INT (1.82)* -0.027 (5.64)*** Consan 0.070-0.059 (2.84)*** (0.58) Observaions 140 140 Arellano-Bond AR(2) -0.17 0.15 Prob > z (0.867) (0.878) Sargan es chi2 27.45 30.94 Prob > chi2 (0.237) (0.156) Absolue value of z saisics in parenheses * significan a 10%; ** significan a 5%; *** significan a 1% 16
The firs regression represens a sandard growh equaion wih he GDP per capia growh rae as an endogenous variable. The resuls sugges ha capial accumulaion, i.e. invesmen, is he mos relevan deerminan of he growh process. As expeced, human capial and rade openness have a posiive and significan impac on economic growh, he former hrough improved produciviy, and he laer (resuling from he signing of regional agreemens) hrough higher compeiion and echnological progress. To analyse he link beween financial secor developmen and economic growh we added o he sandard growh regression (1) hree financial indicaors, i.e. he raio o GDP of privae cred liquid liabiliies and sock marke capialisaion respecively. We find ha credi o he privae secor has a posiive bu insignifican effec on economic growh, possibly as a resul of he numerous banking crises caused by he large proporion of non-performing loans (and hus unsusainable credi growh) a he beginning of he ransiion process in he counries of Cenral and Easern Europe (Tang e al., 2000). However, credi graned o privae companies is essenial for financing invesmen projecs, which in urn affec posiively long-run growh. Furher, he sock marke capialisaion o GDP raio has a posiive bu minor effec on economic growh. Despie an upward rend for his indicaor in he CEE counries during he period being invesigaed, heir sock markes sill have a small size, and i is herefore very imporan o arac foreign invesors. The raio of liquid liabiliies as a proporion of real GDP has a posiive and significan coefficien, consisenly wih he idea ha money supply helps growh by faciliaing economic aciviy. As he size of he financial secor by iself migh no be sufficien o esimae he role of financial developmen in he growh process, we added o he model wo indicaors of financial efficiency: he ineres margin raes beween he lending and deposi as a measure of efficiency in he banking secor, and he EBRD index of insiuional developmen which measures he progress in reforming he financial secor. The former variable measures ransacion coss wihin he secor bu may also reflec an improvemen in he qualiy of borrowers in he economy. If he margin declines due o a decrease in ransacion coss, he share of saving going o invesmen increases and economic growh acceleraes. Boh hese variables appear o be highly significan (see column (3) of Table 5). The margin beween 17
lending and deposi ineres raes is negaively correlaed wih economic growh, consisenly wih heory (see Harrison e al., 1999). This means ha a shrinking ineres margin rae can increase economic growh. In all ransiion counries from Cenral and Easern Europe efficiency increased over ime bu reached differen levels (see Appendix), depending on he privaisaion mehods and he influence of more efficien foreign banks (see Maousek and Tac 2005; Bonin e al., 2005). The oher financial efficiency indicaor, i.e. he EBRD index, has a posiive effec, implying ha reforms in he banking and financial secor such as marke regulaion and monioring, increase economic growh. The resuls for he hree subgroups are repored in Table 6. The privae credi o GDP raio is found o have a posiive bu insignifican effec in all hree groups. As for sock marke capialisaion, his has a posiive, small effec in he case of he CEE-5 counries, and a sill posiive bu insignifican one in he SEE-2 and B-3 counries. In he former group he sock marke expanded more rapidly due o early privaisaion and he enry of foreign invesors, bu i is sill relaively underdeveloped. 18
Table 6: The financial secor and economic growh nexus in he ree subgroups: dynamic panel regression Subgroup CEE-5 B-3 SEE-2 (1) (2) (3) Variables RGDPC RGDPC RGDPC L1.RGDPC 0.236 0.045-0.083 (2.69)*** (0.33) (0.65) INV 0.181 0.032 0.089 (5.85)*** (1.70)* (6.99)*** TOP 0.025 0.221 0.023 (3.31)*** (3.96)*** (0.47) INFL -0.004-0.003-0.016 (1.84)* (1.67)* (2.70)*** GVE -0.023-0.034-0.237 (1.86)* (0.68) (3.30)*** HC 0.022 0.142 0.078 (2.42)** (2.97)*** (1.74)* DCPS 0.042 0.014 0.058 (1.70) (0.79) (1.05) STMC 0.010 0.015 0.002 (2.61)** (0.68) (1.31) LLG 0.008 0.006 0.002 (2.10)** (2.44)** (1.81)* RI 1.046 0.634 0.311 (4.74)*** (2.62)** (2.17)** INT -0.031-0.011-0.067 (2.85)** (2.33)** (4.89)*** Consan 0.098-0.252 0.267 (2.31)** (1.20) (1.50) Observaions 70 42 28 Arellano-Bond AR(2) -0.57 0.15-1.30 Prob > z (0.570) (0.878) (0.193) Sargan es chi2 10.45 30.94 7.65 Prob > chi2 (0.235) (0.156) (0.364) Absolue value of z saisics in parenheses * significan a 10%; ** significan a 5%; *** significan a 1% 19
The index of financial insiuional developmen also has a posiive effec in all hree groups, especially so in he CEE-5, followed by he B-3 and he SEE-2, reforms of he financial sysem being more advanced in he wo former groups. Moneisaion is also significanly and posiively correlaed wih real per capia GDP growh in all hree cases. In mos high-income counries wih developed banking secors, he raio of broad money o GDP is a leas 60 percen (Bonin and Wachel, 2003). In he ransiion counries, he highes moneisaion raio in 2007 is found in Slovenia (75.4), and he lowes in Romania (36.6). The degree of moneisaion can be seen as an indicaor of macroeconomic sabiliy, which represens an incenive for foreign invesors. The efficiency of he banking secor has an imporan role in economic growh. This indicaor is negaively correlaed wih economic growh in all cases. Achieving higher efficiency remains a challenge for hese hree groups of counries. The CEE-5 have recorded an increase of his indicaor due o he early privaisaion of he banking secor and he enry of foreign banks. The SEE-2 counries insead have sared privaisaion laer and seen high ineres rae margins during he ransiion period (for example, 20.8 in Romania in 2000 in comparison wih 7.2 in Poland and 2.1 in Hungary). Overall, underdevelopmen of he sock and credi markes, and herefore lack of financial deph, remains one of he main feaures of hese counries compared wih he oher EU counries (see Coricelli and Masen, 2004). 4.5 The role of credi in he economy and is deerminans Lending o he privae secor is one of he main driving forces of economic growh. Thus, increasing he supply of loans is a key challenge for he CEE counries. Alhough credi markes are sill underdeveloped, in recen years in mos of hese counries he credi o GDP raio has risen. A he end of 2007, hese counries displayed a heerogeneous privae secor credi o GDP raio ranging from 33% o 94%, he lowes increase being recorded in Romania and he highes in Lavia. This credi expansion has been largely he resul of increased morgage loans o households. Rapid credi growh parly reflecs he very low iniial level of inermediaion and he convergence owards he levels of he developed EU counries, bu he figures sill remain below hose for he euro area (Éger e al., 2007). Some sudies have addressed he quesion wheher lending growh has become excessive in he CEE counries (see Boissay e al., 2007; Brzoza-Brzezina 2005; Backé e al, 2007). Given he imporance of credi for economic growh, nex we invesigae economerically is deerminans. 20
Specifically, we expand he model proposed by Éger e al. (2007) by adding hree new variables, namely: non-performing loans (as a percenage of oal loans), asse share of foreign-owned banks (in per cen) and domesic credi o households (as a percenage of GDP): DCPS = f( GDPC, BCPS, INFL, INT, HCR, LR, IBR, NPL, PCFB) (5) where DCPS is raio of privae secor credi o GDP, and he explanaory variables include: GDP per capia a purchasing power pariy (GDPC), bank credi o he public secor as a percenage of GDP (BCPS), producer price inflaion (INFL), he margin beween lending and deposi ineres raes (INT), domesic credi o households as a percenage of GDP (HCR), nominal ineres raes (lending raes) (LR), an index of banking reform (IBR), nonperforming loans (as a percenage of oal loans) (NPL), asse share of foreign-owned banks (in percenage) (PCFB). The empirical specificaion is he following: DCPS + β IBR 7 = α i + β1dcps 1 + β NPL 8 + β PCFB 9 + β GDPC 2 + β INFL + β BCPS + u 10 3 i + ε + β INT 4 + β HCR 5 + β LR 6 + (6) Again, he model is esimaed firs for he whole panel and hen for he subgroups using he sysem GMM mehod, and he sample period is he same as before. The esimaion resuls are repored in Table 7. 21
Table 7: The deerminans of credi o he privae secor: dynamic panel regression ZONE TOTAL CEE-5 B-3 SEE-2 (1) (2) (3) (4) Variables DCPS DCPS DCPS DCPS L.DCPS 0.730 0.728 0.737 0.563 (14.17)*** (16.31)*** (9.29)*** (3.13)*** GDPC 0.187 0.114 0.216 0.079 (2.13)** (1.86)* (1.96)* (2.10)** INFL -0.084-0.018-0.028-0.119 (3.29)*** (1.93)* (1.76)* (2.15)** INT -0.023-0.053-0.034-0.293 (1.74)* (1.66)* (1.88)* (2.45)*** HCR 0.129 0.029 0.167 0.274 (4.07)*** (3.33)*** (3.83)*** (2.47)** LR -0.108-0.057-0.098-0.172 (1.94)* (1.86)* (1.70)* (2.57)*** IBR 0.717 0.781 0.953 0.526 (3.04)*** (1.88)* (3.44)*** (1.76)* NPL -0.046-0.139-0.034-0.121 (2.07)** (4.55)*** (2.11)** (1.77)* PCFB 0.041 0.033 0.028 0.073 (2.25)* (2.44)* (3.45)*** (1.51) BCPS -0.160-0.121-0.093-0.143 (2.32)** (1.92)* (1.84)* (2.25)** Consan -0.045 0.589 1.667 0.234 (0.14) (1.62) (3.64)*** (0.06) Observaions 140 70 42 28 Number of counry 10 5 3 2 Arellano-Bond AR(2) -1.28 1.25-0.62-1.21 Prob > z (0.199) (0.212) (0.535) (0.227) Sargan es chi2 23.67 23.88 16.79 16.51 Prob > chi2 (0.699) (0.123) (0.819) (0.790) Absolue value of z saisics in parenheses * significan a 10%; ** significan a 5%; *** significan a 1% We noe ha GDP per capia has a posiive effec on privae cred increasing financial deph. Higher disposable income, as well as low foreign ineres raes, made i easier for households o finance heir expendiure and service heir deb. Privae credi growh has been largely he resul of more loans o households, primarily morgage-based housing loans (see he Appendix). The lending rae is negaively linked o privae credi. Thus, a decrease in lending raes, i.e. in he cos of borrowing, leads o financial deepening. Inflaion also has a negaive effec, 22
leading o macroeconomics insabiliy. Credi is insead posiively affeced by he asse share of foreign-owned banks. These have become increasingly imporan for he expansion of domesic credi in hese counries. Moreover, in he CEE counries he financial secors are dominaed by privae banks where foreign banks (mainly from he EU) hold he larges share of asses. As expeced, non-performing loans have a negaive effec, as heir growh leads o banking crises and herefore slower credi growh. By conras, he index of banking reform has a posiive effec, confirming ha reforms o he banking sysem simulae credi growh and he developmen of credi markes. Credi o he public secor has a negaive effec. The margin beween lending and deposi ineres raes also has a negaive effec, a more efficien banking secor leading o financial deepening. Heerogeneiy in credi dynamics can have various causes, such as a differen degree of economic developmen and of financial inermediaion, and differen insiuional and regulaory frameworks. The facors ha are normally found o simulae credi growh in he ransiion counries, such as an increase in income or a decrease in lending raes, inflaion and non-performing loans, coninue o play an imporan role in he case of he CEE counries. Progress in heir economic and moneary inegraion can accelerae credi growh, wih benefis in erms of financial and economic developmen, bu also wih poenial risks: a credi boom can have negaive repercussions such as sizeable exernal imbalances, for insance consumpion and invesmen booms leading o economic overheaing and banking and currency crises. 5. Financial developmen and economic growh: he causal linkages In his secion we invesigae causaliy beween financial developmen and economic growh in he en new EU members included in our panel using Granger-ype causaliy ess. 23
5.1 Granger causaliy es As menioned above, our series are saionary and herefore i is legiimae o perform sandard Granger Causaliy ess. Consider wo saionary variables X and Y observed over T periods and N unis. Le x, (y ) denoe he variable X (Y) associaed wih uni i = 1,2... N and = 1,2,... T. We es he hypohesis of no causaliy using he following linear models: J J j j y = α i + δ i y 1 + β i x j + ε j= 1 j= 1 (7) J J j j x = α i + δ i x 1 + β i y j + ε j= 1 j= 1 (8) wih * J N and ε i.i.d. 0, ) ( σ ε,i 5. 2 Resuls We invesigae causaliy linkages in boh direcions by esimaing equaions (7) and (8) o es for causaliy in boh direcions for he following pairs of variables in urn: (i) economic growh (RGDPC) and financial developmen (proxied by domesic credi o he privae secor DCPS); (ii) economic growh (RGDPC) and banking efficiency (INT), and finally (iii) economic growh (RGDPC) and sock marke capialisaion (STMC). The Granger causaliy es was originally designed for ime series (Granger, 1969). However, i has recenly been exended o panels (see Granger and Lin, 1995, Granger, 2003). The esimaion is carried ou here using he sysem GMM mehod developed by Arellano and Bover (1995) and Blundell and Bond (1998) which was designed o overcome some of he limiaions of he difference GMM. We perform he Sargan/Hansen es for he validiy of he addiional momen resricions required by he sysem GMM esimaor. In order o avoid model misspecificaion hree condiions should be saisfied: a significan AR(1) serial correlaion, lack of AR(2) serial correlaion and a high Sargan es saisic (Arellano and Bond, 1991). 24
In he AR(2) model (J=2) described in Eq. (7, 8) he join null β 1 =β 2 =0 is inerpreed as a panel daa es for Granger causaliy and is disribued as a χ 2 wih wo degrees of freedom (see Casu and Girardone, 2009). A p-value < 0.10 implies a rejecion a he 10% significance level of he null hypohesis of no causaliy. To esablish if here is a long-run linkage beween x and y, we es he resricion β 1 +β 2 =0, where he null is no long-run effec. The sign of he causal relaionship is given by T=β 1 +β 2. Table 8: Type of causal relaionship and inerpreaion Equaion T=β 1 +β 2 Type of Inerpreaion causal relaionship Eq. 7 >0 posiive An increase of x implies an increase of y and viceversa Eq. 8 >0 posiive An increase of y implies an increase of x and viceversa Eq. 7 <0 negaive An increase of x implies an decrease of y and viceversa Eq. 8 <0 negaive An increase of y implies an decrease of x and viceversa A posiive (negaive) T implies ha he causal relaionship beween pas x and presen y (eq. 7) or beween pas y and presen x (eq. 8) is also posiive (negaive). The resuls of he Granger Causaliy es are repored in Tables 9a and 9b. 25
Table 9a: Granger Causaliy es beween domesic privae credi secor and economic growh Variables RGDPC Variables DCPS L1.RGDPC 0.670 L1.DCPS 1.022 (3.49)*** (2.31)** L2.RGDPC -0.147 L2.DCPS -0.431 (0.69) (2.23)* L1.DCPS 0.010 L1.RGDPC 9.520 (0.50) (1.82) L2.DCPS 0.027 L2.RGDPC -0.478 (1.28) (0.23) Consan -0.010 Consan 0.345 (0.47) (0.66) L1.DCPS + L2. DCPS 0.037 L1.RGDPC + 8.042 L2.RGDPC Granger causaliy p- (0.017) Granger causaliy p-value (0.120) value Arellano-Bond es for 0.127 Arellano-Bond es for 0.211 AR(1) p-value AR(1) p-value Arellano-Bond es for 0.412 Arellano-Bond es for 0.155 AR(2) p-value AR(2) p-value Sargan es of overid. 0.752 Sargan es of overid. 0.998 resricions: p-value resricions: p-value Tes (β 1 + β 2 )=0 Prob>F 0.325 Tes (β 1 + β 2 )=0 0.206 Observaions 140 Observaions 140 Number of counry 10 Number of counry 10 Absolue value of saisics in parenheses * significan a 10%; ** significan a 5%; *** significan a 1% One can see ha he relaionship beween privae credi and economic growh is posiive bu he esimaed coefficiens are no saisically significan, leading o he conclusion ha here are no causal linkages beween hese wo variables. Also, here is no Granger causaliy in eiher direcion beween economic growh and financial deph (DCPS). 26
Table 9b: Granger Causaliy es beween ineres rae margin and economic growh Variables RGDPC Variables INT L.RGDPC 0.646 L1.INT -0.010 (8.73)*** (0.10) L2.RGDPC -0.146 L2.INT 0.274 (3.09)** (3.21)** L1.INT 0.001 L.RGDPC 7.157 (1.96)* (1.96)* L2.INT -0.003 L2.RGDPC -17.938 (2.17)** (4.29)*** Consan 0.019 Consan 1.382 (4.90)*** (3.99)*** L1.INT+ L2.INT -0.002 L.RGDPC + L2.RGDPC -10.781 Granger causaliy p-value (0.035) Granger causaliy p-value (0.13) Arellano-Bond es for 0.151 Arellano-Bond es for 0.250 AR(1) p-value AR(1) p-value Arellano-Bond es for 0.658 Arellano-Bond es for 0.579 AR(2) AR(2 p-value) Sargan es of overid. 0.852 Sargan es of overid. 0.901 resricions: p-value resricions: p-value Tes (β 1 + β 2 )=0 0.300 Tes (β 1 + β 2 )=0 0.101 Observaions 140 Observaions 140 Number of counry 10 Number of counry 10 Absolue value of saisics in parenheses * significan a 10%; ** significan a 5%; *** significan a 1% Furher, causaliy runs from banking efficiency (INT) o economic growh bu no in he opposie direcion, i.e. he ineres rae margin Granger-causes economic growh. This linkage is negaive and significan. Again, here is no evidence of long-run effecs of causaliy from INT o RGDPC (Prob>F = 0.300, implying ha H 0 : no long-run effec is no rejeced). 27
Table 9c: Granger Causaliy es beween sock marke capialisaion and economic growh Variable RGDPC Variable STMC L1.RGDPC 0.652 L1.STMC 0.373 (8.33)*** (2.14)* L2.RGDPC -0.092 L2.STMC 0.133 (1.37) (1.35) L1.STMC 0.007 L1.RGDPC -1.206 (3.44)*** (0.24) L2.STMC -0.005 L2.RGDPC 5.439 (2.64)*** (2.64)** Consan 0.012 Consan 0.494 (5.54)*** (1.94)* L1.STMC + L2.STMC 0.002 L1.RGDPC + L2.RGDPC 4.233 Granger causaliy p-value 0.002 Granger causaliy p-value 0.176 Arellano-Bond es for AR(1) 0.13 Arellano-Bond es for AR(1) 0.151 p-value p-value Arellano-Bond es for AR(2) 0.349 Arellano-Bond es for AR(2) 0.421 p-value p-value Sargan es of overid. 0.625 Sargan es of overid. 0.836 resricions: p-value resricions: p-value Tes (β 1 + β 2 )=0 Prob>F 0.350 Tes (β 1 + β 2 )=0 Prob>F 0.451 Observaions 132 Observaions 132 Number of counry 10 Number of counry 10 Absolue value of saisics in parenheses * significan a 10%; ** significan a 5%; *** significan a 1% Finally, Granger causaliy runs from sock marke capialisaion (STMC) o economic growh (RGDPC) bu no in he opposie direcion. There is also no evidence of long-run effecs (Prob>F = 0.350, implying ha H 0 : no long-run effec is no rejeced). 6. Conclusions In his paper we have reviewed he main feaures of he banking and financial secor in en new EU members, and hen invesigaed he relaionship beween financial developmen and economic growh in hese economies by esimaing a dynamic panel daa model over he period 1994-2007. To summarise, financial deph is found o be lacking in all en counries, and herefore he conribuion of he relaively underdeveloped credi and sock markes o growh has been raher limied, wih only a minor posiive effec of some indicaors of 28
financial developmen. This migh be a consequence of he large sock of non-performing loans and he banking crises experienced by hese economies a he beginning of he ransiion period. In general, he CEE-5 have more developed financial secors han he B-3 and SEE-2 counries. By conras, he implemenaion of reforms, he enry of foreign banks and he privaisaion of sae-owned banks have reduced ransacion coss and increased credi availabiliy. This has improved he efficiency of he banking secor (Fries e al., 2006), which has played an imporan role as an engine of growh. Beer regulaion and supervision was parly moivaed by he European inegraion process and he need o adop EU sandards. Thus, many of he banking secor weaknesses radiionally characerising emerging markes have gradually been eliminaed. Given he prospec of EU accession, foreign banks, mainly from he euro area, seized he opporuniy and esablished subsidiaries in all CEE counries, seeing hem as an exension of he common European marke and becoming dominan players in heir banking secors. However, he massive presence of foreign banks has also increased conagion risks, and he consolidaion process (wih he majoriy of banks being foreign owned) could limi compeiion. Thus, a financial crisis produced in he maure markes of he euro area could also reach he CEE counries. A sraegy of financial developmen based on foreign enry from he anchor currency area is no guaranee for a smooh process of finance and growh, an example being he curren crisis which sared in he maure economies in he summer of 2007 and caused a sudden sop of capial flows o Souheasern Europe (Winkler, 2009). Granger causaliy es sugges ha causaliy runs from financial developmen, measured as credi o he privae secor and he ineres rae margin, o economic growh, bu no in he opposie direcion. Credi o he privae secor has risen rapidly in hese counries in recen years bu a a differen rae, he lending boom being paricularly srong in he segmen of loans o households, primarily morgage-based housing loans. The heerogeneiy in credi dynamics can have various causes, such as a differen degree of economic or financial inermediaion developmen, and differen insiuional and regulaory frameworks. Our analysis of he deerminans of credi o he privae secor highlighs differen facors ha simulae credi growh in he ransiion counries, such as an increase in income or a decrease in lending raes, inflaion and non-performing loans, and he implemenaion of reforms in he 29
banking secor. Furher, he high growh of credi o households can affec negaively he curren accoun, which migh be a serious problem for he ransiion economies. Overall, he underdevelopmen of sock and credi markes, wih he consequen lack of financial deph, remains one of he main feaures of hese economies. However, elemens of marke-oriened inermediaion are now he rule raher han he excepion hroughou hem (Bonin and Wachel, 2003), and appropriae policies can reduce financial secor insabiliy ha could impair growh (Kraf, 2005).The adopion of he euro could have a furher posiive impac on financial developmen and economic growh in hese counries, bu his issue is beyond he scope of he presen paper. 30
References Arellano M., Bond S. (1991), Some Tess of Specificaion for Panel Daa: Mone Carlo Evidence and an Applicaion o Employmen Equaions, Review of Economic Sudies, 58(2), 277-297. Arellano M., Bover O. (1995), Anoher Look a he insrumenal variable esimaion of error-componens models, Journal of Economerics, 68(1), 29-51. Backé P., Éger B., Walko Z. (2007), Credi Growh in Cenral and Easern Europe Revisied, Focus on European Inegraion 2, 69-77. Beck T., Levine R., Loayza N. (2000), Finance and he Sources of Growh, Journal of Financial Economics, 58(1-2), 261-300. Beck T., Feijen E., Ize A., Moizeszowicz F. (2008), Benchmarking financial developmen, Policy Research Working Paper Series 4638, The World Bank. Beck T., Demirgüç-Kun A., Levine R. (2000), A New Daabase on he Srucure and Developmen of he Financial Secor, World Bank Economic Review 14 (3), 597-605. Beck T. (2008), The Economerics of Finance and Growh, Policy Research Working Paper Series 4608, The World Bank. Bencivenga V.R., Smih B.D. (1991), Financial Inermediaion and Endogenous Growh, Review of Economic Sudies, 58(2), 195-209. Berglöf E., Bolon P. (2002), The Grea Divide and Beyond: Financial Archiecure in Transiion, The Journal of Economic Perspecives, 16 (1), 77-100. Blundell R., Bond S. (1998), Iniial Condiions and Momen Resricions in Dynamic Panel Daa Models, Journal of Economerics, 87(1), 115-143. Boissay F., Calvo-Gonzalez O., Kozluk T. (2007), Using Fundamenals o Idenify Episodes of Excessive Credi Growh in Cenral and Easern Europe, in Enoch C. and I. Öker-Robe (eds.). Rapid Credi Growh in Cenral and Easern Europe: Endless Boom or Early Warning? 47 66. Bonin J., Wachel P. (2003), Financial Secor Developmen in Transiion Economies: Lessons from he Firs Decade, Financial Markes, Insiuions and Insrumens 12 (1), 1-66. Bonin J., Hasan I., Wachel P. (2005), Privaisaion Maers: Bank Performance in Transiion Counries, Journal of Banking and Finance 29, 2153-78. Brzoza-Brzezina M. (2005), Lending booms in he new EU Member Saes: will euro adopion maer, ECB Working Paper No. 543. 31
Casu B., Girardone C. (2009), Tesing he relaionship beween compeiion and efficiency in banking: A panel daa analysis, Economics Leers 105 (1), 134 137. Coricelli F., Masen I. (2004). "Growh and Volailiy in Transiion Counries: The Role of Cred" presened a he IMF Conference in honor of Guillermo A. Calvo, Washingon, DC, April 15-16, 2004, Inernaional Moneary Fund. Coarelli C., Dell Ariccia G., Vladkova-Hollar I. (2005), Early Birds, Lae Risers and Sleeping Beauies: Bank Credi Growh o he Privae Secor in Cenral and Easern Europe and in he Balkans, Journal of Banking and Finance, 29(1), 83 104. De Haas R.T.A (2001), Financial developmen and economic growh in ransiion economies A survey of he heoreical and empirical lieraure Research Series Supervision 35, Neherlands Cenral Bank. Éger B., Backé P., Zumer T. (2007), Privae-Secor Credi in Cenral and Easern Europe: New (Over) Shooing Sars?, Comparaive Economic Sudies 49 (2), 201-231. Fink G., Haiss P., Manler H.C. (2005), The Finance-Growh-Nexus: Marke Economies vs. Transiion Counries, Europainsiu Working Paper 64. Fink G., Haiss P., and Vuksic G. (2008), Conribuion of Financial Marke Segmens a Differen Sages of Developmen: Transiion, Cohesion and Maure Economies Compared Journal of Financial Sabiliy, forhcoming. Fries S., Neven D.J., Seabrigh P., Taci A. (2006), Marke Enry, Privaizaion and Bank Performance in Transiion, Economics of Transiion, 14(4), 579-610. Gillman M., Harris M.N. (2004), Inflaion, Financial Developmen and Growh in Transiion Counries, Monash Economerics and Business Saisics Working Papers 23. Goldsmih R.W. (1969), Financial Srucure and Developmen, New Haven, CT, Yale Universiy Press. Granger C.W.J. (1969), Invesigaing Causal Relaions by Economeric Models and Cross- Specral Mehods, Economerica, 37(3), 424-38. Granger C.W.J., Lin J. L. (1995), Causaliy in he long run, Economeric Theory, 11 (3), 530-36. Granger C.W.J. (2003), Some Aspecs of Causal Relaionships, Journal of Economerics, 112 (1), 69-71. Greenwood J., Jovanovic B. (1990), Financial Developmen, Growh, and he Disribuion of Income, Journal of Poliical Economy, Universiy of Chicago Press, 98(5), 1076-1107. 32
Hagmayr B., Haiss P.R., Sumegi K. (2007), Financial Secor Developmen and Economic Growh - Evidence for Souheasern Europe EuropaInsiu Working Paper. Harrison P., Sussman O., Zeira J. (1999), Finance and growh: Theory and new evidence, Finance and Economics Discussion Series 35, The Federal Reserve Board. Hermes N., Lensink R. (2000), Financial sysem developmen in ransiion economies Journal of Banking and Finance, 24 (4), 507-524. Im K., Pesaran M.H., Shin Y. (2003), Tesing for uni roos in heerogeneous panels, Journal of Economerics 115 (1), 53-74. Kenourgios D., Samias A. (2007). Financial Developmen and Economic Growh in a Transiion Economy: Evidence for Poland Journal of Financial Decision Making, 3, (1), 35-48. King R.G., Levine R. (1993a), Finance, Enrepreneurship, and Growh: Theory and evidence, Journal of Moneary Economics, 32(3), 513-42. King R.G., Levine R (1993b), Finance and Growh: Schumpeer migh be righ, Quarerly Journal of Economics, 108 (3), 717-37. Kraf E., (2005), EU Accession - Financial Secor Opporuniies and Challenges for Souheas Europe, Springer Berlin Heidelberg, 61-91. Levine R. (1997), Financial developmen and economic growh: Views and agenda, Journal of Economic Lieraure, 35(2), 688-726. Levin A., Lin. C.F., Chu C.S., (2002), Uni roo ess in panel daa: Asympoic and finie sample properies, Journal of Economerics, 108(1), 1 24. Levine R., Zervos S. (1996), Sock Marke Developmen and Long-Run Growh, World Bank Economic Review, 10(2), 323-339. Levine R., Loayza N., Beck T. (2000), Financial inermediaion and growh: Causaliy and causes, Journal of Moneary Economics, 46(1), 31-77. Levine R. (2005), Finance and Growh: Theory and Evidence, Handbook of Economic Growh, in: Aghion P. and S. Durlauf (ed.), vol 1, 865-934. Lucas R.E. (1988), On he mechanics of economic developmen, Journal of Moneary Economics, 22(1), 3-42. Maousek R., Taci A. (2005), Efficiency in banking: Empirical Evidence from he Czech Republic, Economic Change and Resrucuring, 37(3), 225-244. 33
McKinnon R.I. (1973), Money and Capial in Economic Developmen, Washingon D.C., The Brookings Insiuion. Pesaran H.M. (2007), A simple panel uni roo es in he presence of cross-secion dependence, Journal of Applied Economerics 22(2), 265-312. Rioja F., Valev N. (2004), Does one size fi all? A reexaminaion of he finance and growh relaionship, Journal of Developmen Economics 74(2), 429-47. Robinson J. (1952), The generalizaion of he general heory, in The Rae of Ineres and Oher Essays, London: Macmillan, 69-142. Rousseau P.L., Wachel P. (2002), Inflaion hresholds and he finance-growh nexus, Journal of Inernaional Money and Finance, 21(6), 777-793. Schumpeer J.A. (1934), The Theory of Economic Developmen, Cambridge, MA, Harvard Universiy Press. Shaw E.S. (1973), Financial Deepening in Economic Developmen, New York: Oxford Universiy Press. Tang H., Zoli E., Klychnikova I. (2000), Banking Crises in Transiion Counries: Fiscal Coss and Relaed Issues, World Bank Policy Research Working Paper 2484. Wachel P. (2001), Growh and Finance: Wha do we know and how do we know i?, Inernaional Finance, 4(3), 335-362. Winer L.A. (2004), Trade Liberalizaion and Economic Performance: An Overview, The Economic Journal, 114, 4-21 Winkler. A., (2009), Souheasern Europe: Financial Deepening, Foreign Banks and Sudden Sops in Capial Flows Focus on European Economic Inegraion 1, 84-97. 34
APPENDIX A Table A1: Lis of variables VARIABLE (series) Source CODE NOM BCPS Bank credi o he public secor as a percenage of IFS daabase GDP DCPS Domesic credi o privae secor (in per cen of GDP) EBRD daabase GDPC GDP per capia (in PPP) EBRD daabase GVE General governmen expendiure o GDP EBRD daabase HC Secondary school enrollmen raio UNESCO daabase HCR Domesic credi o households (in per cen of GDP) EBRD daabase IBR EBRD index of banking secor reform EBRD daabase INFL Inflaion, average consumer prices IMF daabase INV Invesmen/GDP (in per cen) EBRD daabase INT Ineres margin raes beween lending and deposi (in per cen) Auhors calculaion using EBRD daabase LLG Liquid Liabiliies (in per cen of GDP) EBRD daabase LR Lending rae (average) EBRD daabase NPL Non-performing loans (in per cen of oal loans) EBRD daabase PCFB Asse share of foreign-owned banks (in per cen) Auhors calculaion using EBRD daabase RGDPC Real GDP per capia growh Auhors calculaion using EBRD daabase RI Reform index of financial insiuional developmen Auhors calculaion using EBRD daabase STMC Sock marke capialisaion (in per cen of GDP) EBRD daabase TOP Trade openness o GDP EBRD daabase 35
Table A2. EBRD indicaors of reform Indicaor EBRD index of banking secor reform EBRD index of reform of nonbank financial insiuions 1996 2007 1996 2007 Year Counry Bulgaria 2.0 3.7 2.0 2.7 Czech.Rep. 4.3 4.3 2.7 3.0 Esonia 4.0 4.3 2.0 3.7 Hungary 4.3 4.3 3.0 3.3 Lavia 4.0 4.3 2.0 3.0 Lihuania 4.0 4.3 2.0 3.3 Poland 4.3 4.3 2.7 3.3 Romania 3.0 4.3 1.0 2.7 Slovakia 4.3 4.3 3.0 3.3 Slovenia 4.3 4.3 2.0 2.7 Source EBRD Table A3 : Ineres rae margin (%) Year 2000 2001 2002 2003 2004 2005 2006 2007 Counry Bulgaria 8.4 8.2 6.6 5.9 5.8 4.9 4.9 6.3 Czech.Rep 3.8 6.1 4.8 4.8 4.8 4.5 4.3 4.6 Esonia 2.1 5.6 2.9 2.7 4.1 6.2 3.6 4.1 Hungary 2.9 2.6 2.3 2.5 1.9 2.2 1.8 2 Lavia 7.7 5.5 2.3 2.4 4 2.7 3.7 4.8 Lihuania 9.7 7.4 5.8 4.8 5.4 5.3 5.2 5.3 Poland 7.2 8.8 7.4 6.7 7.4 4.2 4.1 4.5 Romania 20.8 19.5 16.2 14.4 14.1 13.2 9.2 6.7 Slovakia 4.5 5 3.6 3.2 5 4.3 4.1 4.2 Slovenia 5.7 5.3 5 4.8 4.9 4.6 4.6 2.3 Source: EBRD 36
Table A4: Morgage lending (as a percenage of GDP) Year 2000 2001 2002 2003 2004 2005 2006 2007 Counry Bulgaria - - - 1.2 2.7 4.8 7.2 10.4 Czech.Rep 2.0 2.4 3.0 4.2 5.9 7.7 10.0 12.5 Esonia 2.3 3.5 5.5 9.5 14.6 22.6 33.0 37.7 Hungary 1.1 1.7 4.1 8.0 9.5 11.5 13.9 16.4 Lavia 1.6 2.4 4.1 7.6 12.4 19.5 28.9 33.7 Lihuania - - 1.9 3.4 5.5 9.0 12.6 17.2 Poland - - 2.4 3.4 3.8 5.0 7.2 9.9 Romania - - - 0.3 0.5 0.6 0.9 1.4 Slovakia 0.1 0.4 1.0 2.2 2.9 3.6 4.1 4.5 Slovenia 1.7 1.8 2.0 2.3 2.8 4.2 4.5 6.2 Table A5: Levin-Lin-Chu saionariy es Coefficien -value -sar P > Level Series BCPS -0.54670-7.485-4.53383 0.0000 I(0) DCPS -0.24590-5.356-2.25883 0.0119 I(0) GDPC -0.15040-5.358-3.71969 0.0001 I(0) GVE -0.52960-7.119-4.00032 0.0000 I(0) HC -0.25120-6.727-4.99531 0.0000 I(0) HCR -0.14939-3.812-1.71353 0.0433 I(0) IBR -0.47511-7.459-4.42017 0.0000 I(0) INFL -0.46330-6.384-2.61235 0.0045 I(0) INT -0.63380-8.358-5.17992 0.0000 I(0) INV -0.19084-3.633-1.26133 0.0136 I(0) LLG -0.19990-5.282-3.15713 0.0008 I(0) LR -0.65490-8.049-4.40804 0.0000 I(0) NPL -0.21493-3.994-1.29016 0.0985 I(0) PCFB -0.55450-9.596-7.97387 0.0000 I(0) RGDPC -0.58719-8.584-5.15507 0.0000 I(0) RI -0.43460-8.835-5.78175 0.0000 I(0) STMC -0.89160 15.682-13.67317 0.0000 I(0) TOP -0.28015-8.240-6.20488 0.0000 I(0) 37
The Levin-Lin-Chu (LLC) es The LLC es is based on esimaing he following equaion: Δ i = 1,2, N, = 1,2, T y = α i + δ i + θ + ρi y 1 + ς This model allows for wo way fixed effecs (α and θ) and uni specific ime rends. Because he coefficien of he lagged dependen variable is resriced o be homogeneous across all unis of he panel, he uni specific fixed effecs are an imporan source of heerogeneiy. The es involves he null hypohesis H 0 : ρ i = 0 for all i agains he alernaive H A : ρ i = ρ < 0, wih auxiliary assumpions abou he coefficiens of he deerminisic componens also being required under he null. The LLC es assumes ha he individual processes are cross secionally independen. Given his assumpion, condiions (and correcion facors) are derived under which he pooled OLS esimae of ρ will have a sandard normal disribuion under he null hypohesis. Levin e al. (2002) analyse he asympoic disribuion of his pooled panel esimae of ρ under differen assumpions on he exisence of fixed effecs and homogeneous ime rends. This es can be viewed as a pooled Dickey Fuller (or ADF) es, poenially wih differing lag lenghs across he unis of he panel. The Pesaran (2007) es The Pesaran (2007) es is based on esimaing he following equaion: Δy ρi = α i + ρy 1 + β jδy j + ε j= 1 i = 1,2, N, = 1,2, T I is essenially a -es for uni roos in heerogeneous panels wih cross-secional dependence. Similarly o he Im, Pesaran and Shin (IPS, 2003) es, i is based on he mean of individual DF (or ADF) -saisics of each uni in he panel. The null hypohesis is ha all series are non-saionary. To eliminae cross-secional dependence, he sandard DF (or ADF) regressions are augmened wih he cross-secional averages of lagged levels and firsdifferences of he individual series (CADF saisics). This avoids size disorions, especially in he case of models wih residual serial correlaions and linear rends. When T is fixed, in order o ensure ha he CADF saisics do no depend on he nuisance parameers he effec of he iniial cross-secional mean mus also be eliminaed; his can be achieved by applying 38
he es o he deviaions of he variable from he cross-secional mean. Lags of he dependen variable can be inroduced o conrol for serial correlaion in he errors. The order of augmenaion can be esimaed using model selecion crieria such as Akaike or Schwarz applied as usual o he underlying ime series specificaion. The exac criical values of he -bar saisic are given by Pesaran (2007). The Z[-bar] saisic is disribued sandard normal under he null hypohesis of nonsaionariy. Pesaran (2007) suggess ha a generalisaion of he es o unbalanced panels can be made sraighforwardly as IPS (2003) show. In he case of unbalanced panels only sandardised Z[bar] saisics can be compued. 39