You are on page 1of 27



Financial Development and Economic Growth: The Role of Stock Markets

Utilizing time series methods and data (rorn five developed economies, we examine the relationship between stock market development and economic growth, controlling for the effects of the hanking system and stock market volatility. Our results support the view thai, although both hanks and stock markets may be able to promote economic growth, the effects of the former are more powcrlul. They also stiggcst that the contribution of stock markets on economic growth may have been exaggerated by studies that utilize cross-country growth regressions. GROWiNG IMPORTANCE of stock markets around the

world has recently opened a new avenue of research into the relationship between financial devcloptnent and economic growth, which focuses on the effects of stock market development. In this context, various stock market development indicators have heen found to explain part of the variation of growth rates across countries, in some cases over and above the effects of" the banking system (Atje and Jovanovic 1993; Levine and Zervos 1998). Since these results have been obtained from crosscountry growth regressions, they ean at best provide oniy a broad-brush picture of the relationship between stock markets and growth, the details of which may reasonably be expeeted to vary considerably across countries, depending on institutional characteristics and circumstances. Furthermore, given ihe widespread skepticism concerning the robustness of econometric results derived from cross-country growth regressions, these results must be viewed with some caution (Levine and Renelt
The auihors are gralcfiLl to two anonymous referees Tur construclive coirmients. They also (hank the purlieipants of the 1997 Development Hconomics Study Group annual conference (University of Birmingham) and the 1999 Royal Kconomic Society (RES) annual conference (University uf NoUingham) for useful comments. They acknowledge tinancial support from the ESRC (Grant No. R000236463). I. World stock market capitali/alion grew from $2 trillion in 19^2 toS4.7 trillion in 1986. $10 trillion in 1993 and $15.2 Irillion in I99(i, implying an average annual growth rate of I.*! percent: emerging markel eapilali/ation urcw from less than 4 to 13 percent of total world capitalization (Dcmirgu^-Kunt and Levine 1996: Singh i997).

PHIIJP ARI-.STIS is il professor of economics at Soiijli Bank University. London. E-mail:

p.arestis^^ PANICOS O. DKMI'.TRIADHS is a professor of economics at Leicester Uni-

versity. KUL B, LL:INI[;L is a reader In economics at Brunei University. Journal of Money, Credit, and Bcmking. Vol. 33, No. I (February 2(K)I) Copyright 2(X)I by Tbe Obio State University

P l i l l . l P X R h S I I S . P A N I C O S O . D h . M H r H I A D K S . A . M ) k l I. B . H I N I K I .

. 17

1992; Arestis and Dcmctriadcs 1997; Luintcl and Khan 1999)." In the spec-ilic context of the cross-country relationship between stock market development and growth, for example, tbc presence of endogeneity has been sbown to considerably weaken the estimated effect of stoek market Indieators (Harris 1997). Tbere are, therefore, important eeonometric advantages in examining tbe role of stock markets in the relationship between financial development and growtb using time series methods. Besides being better able to address issues of causality and endogeneity, they are also less likely to suffer from otber limitations of eross-eountry growth regressions. Even setting aside econometric issues., time series methods can provide useful insigbts into differences of tbis relationship aeross eountries and may illuminate important details tbat are hidden in averaged-out results. This paper utilizes time series methods to reexamine the relationship between eeononiie growth and stoek market development, while controlling tor the effeets of the commercial banking seetor and stoek market volatility. Inevitably, a time series analysis has its ov/n limitations. Specifically, the need to obtain a long time series of stoek market development indicators narrows down the focus of tbe empirical analysis to five developed eeonomies, namely. Germany, tbe United States. Japan, the United Kingdom, and France. While the absenee oi' Ics.s-deveioped economies from our sample means that no direct inferences ean be made about tbe contribution of stock markets at early stages of economic development, our findings ncvertbeless bave implications for tbe debate on bank-based versus eapital-market-bascd linancial systems (see, for example, Rajan and Zingalcs 1993 and 1996; iloriucbi and Okazaki 1994: Hdwards and Fiscber 1994; Corbett and Jenkinson 1994). Thus, our results eould be indirectly valuable for less-developed economies, in tbat tbey may inform poiiey decisions relating to the adoption or otherwise of speeilic types of financial system. The rest of tbe paper is slruetured as follows. In seelion I we provide a discussion ofthe role of stock markets and banks in tbe proeess of economic growtb and summarise tbe existing empirieal literature. In section 2 we ontline our data and eeonometric metbodology. In section ^ we present our findings and discuss their implieations for the debate on finaneial systems. Finally in seetion 4 we provide a summary and some eoncluding remarks.


Positive Effects of Stock Markets Recent theoretieal eontributions suggest tbat stock markets may promote long-run growth. Stock markets eneourage specialization as well as acquisition and dissemi2. Quati (1993) eniphasi/.es Ihc nonc\isli?iiL-e (tfbalimccJ growth paiiis. Casclli. t',sc|iiivcl. ami I clorl (1996) anci Levhie and Rcnell(l992) focus on oDiillcJ vaiiatile hiasor niissix-cification. Hvan.s ( 1995 land Pesaran and Smith ( I 9 9 5 | dwell on the helerogcncil\ iiC slope codlicicnis across counirics. while prohlems of causality and endogeneity arc explored hy Demetriades and llitsseiii 11996) and Harris (19971. 3. The view that time series studies ot economic growth olTer iinporiani advantages over eross-eounIiy growth regres.sions is gaining aeceptance. .See. for example, .lones i 199,'i). Evans ( I997|. Koehcrlaknta and Yi (1997), and Klcnow and Rodrisjue/.-Clare (1997).



nation of information (Diamond 1984; Greenwood and Jovanovic 1990; Williamson 1986) and may reduce the cost of mobilizing savings, thereby facilitating investment (Greenwood and Smith 1997). Well-developed stock markets may enhance corporate control by mitigating tbe principal-agent problem through aligning the interests of managers and owners, in which case managers would strive to maximize firm value (Diamond and Verrccchia 1982; Jensen and Murphy 1990). Stock Market Liquidity Levine (1991) and Bencivcnga, Smith, and Starr (1996) suggest thai stock markets make financial assets traded in them less risky because they allow savers to buy and sell quickly and cheaply wben they wish to alter tbeir portfolios. Companies at the same time enjoy easy access to capital through equity issues. Less-risky assets and easy access to capital markets improve the allocation of capital, an important channel of economic growth. More savings and investment thereby may also ensue, further enhancing long-term economic growth. It is conceded, though, that increased liquidity can also influence growth negatively (sec, for example, Levine 1997). There are three channels through which this may take place (Demirgu^-Kunt and Levine 1996). The first is tbat greater stock market liquidity, by increasing the returns to investment, may reduce savings rates. The second is that, given the ambiguous effect of uncertainty on savings, greater stock market liquidity might in fact reduce savings rates through its negative impact on uncertainty since less uncertainty may decrease the demand for precautionary savings. The tbird channel operates through the euphoria and myopia that may be encouraged by highly liquid stock markets. Dissatisfied participants find it easy to sell quickly which can lead to disincentives to exert corporate control, thus affecting adversely corporate governance and hurting economic growth in the process (sec, however. Jensen and Murphy 1990), TheRoleofVolaliliiy
Another important characteristic of stock markets is that of price volatility, as this may undermine the ability of stock markets to promote an efficient allocation of investment. The undesirable side effects of volatility were recognized early on in tbe literature, notably by Keynes (1936), who was particularly shaip-tongued: As the organisation of investment markets improves, the risk tif ihc predominance of speculation docs . , . increase . . . Speculators may do no harm as hubbies on a steady stream of enterprise ,., a serious situation can develop ., , when enterprise becomes ihe bubble on a whirlpool of speculation. When the capital developtnent of a country becomes li by-product of the activities of a casino, the job is likely to be ill-done. , , . It is usually agreed that casinos should, in the public interest, be inaccessible and expensive. And perhaps the same is true of Stock Exchanges, (pp. 158-59) More recent literature is less conclusive on this issue. Wbile a certain degree of price volatility in the stock market is clearly desirable, since it may reflect the effects of new information flows in an efficient stock market, some evidence suggests ihat the observed levels of volatility may be "excessive." This may reflect independence of stock-market-a,sset values from underlying fundamentals (Sbiiler 1981 and 1989),

P i n i IP.AHF.STIS. I ' A N K ' D S O . l l l M H T K I A I M . i S , A M ) K I L B. l . t l N T E t .

: I'l

even though the debale on the presence of CACCSS volatility in slock returns is far from settled. If preseni. excessive volalilily is likely to result in an ineftieient alloeation of resources, upward pressures on interest rates in view of the higher uncertainty, harnpering both the volume and the productivity of investtiient atid, therefore, reducing growth (Federer 1993: DeLong et al. 1989). Furlhennore. excessive slock trading may very well induee "noise" into the market lo the tietritneni of efficient resouree alloeation (f:)eLong et al. 1989).


fhe relationship between st(K-k markets and growth may also be inllueneed hy the link between stoek markets and linaneial ititermediaries. which is not unatnbiguous. Stock tnarkets and banks are elearly subslilule sources for corporate ftnance since when a firm issues new equity its borrowing needs kvm the banking system deeline. Assuming that banks and linatieial intermediaries are in a better position than stoek markets to address agency problems (for example. Diamond 1984; Stiglilz 1985). it is then possible that sioek market developtnent may hamper economic growth if it happens al the expense of banking sysletii developt)ient. Similar views are expressed by tbe lilerature on capital-market-based linancial sysletns that predicls a very weak relationship belween stoek markets and growth since corporate investtnent is tiot linaneed through issues of equity (Mayer 1988; see, also. Fry 1997). Corbcll and Jenkinson (1994) wben discussing the contribulion of stock market To corporate invesltnenl financing, suggest thai il was negative in the United Kitigdotn and only stnall positive overall in the United Stales during the I97()s and 198()s. Akyuz (1993) and vSingh (1997) argue Ihat unfavorable economic shocks prt)duce tnacroeconomic instability Ihrough the interactions between stock markets and foreign exchange markets, which affect economic growth adversely. On the other hand, at the aggregate level inereased stoek market capilali/ation may be accompanied by an iticrease in the volutne of bank bustness. if tiot an inerease in new lending, as financial intermediaries may provide cotnplemenlary services to issuers of new equity sueh as underwriting. Thus, it is likely that at the aggregate level the developtnent of the stock markei goes hand in hand with the developtnent of the hanking system. Empirical Evidemc Existing evidetice points lo slock market development taking place in tandem with other aspects of linancial deve!<ipnient. Lising data for forty-four induslrial and developing countries for the period 1986-1993. Demirgui^-Kunt atid Lcvitie (19%) conclude that countries with well-devekiped slock tnarkels also have well-developed banks atid nonbank linancial inlertnediaries, while cottntries wiih weak stock markets tend to have weak banks and linancial ititertnediaries. Demirgui^-Kunt and Maksimovie (19%). hi their investigation of the effect of stoek tnarkel development on llrtiis' financitm choices in thirty industrial and developitig econotnies from 1980 to

2(1 : MONt-:Y, rROOl!' .\SD BANKING

1991, find that initial itTiprovcmcnts in the functioning of a developing stock market produce higher debt/equity ratios for large firms, with small firms not being significantly affected, hi already developed stock markets, further development leads to substitution of equity for debt financing, especially for long-term debt. Boyd and Smith (1996) argue that as an economy develops the ratio of debt to equity tends to increase, with the iwo soutxes of finance being complementary. Levitie and Zcrvos (1998), utilizing cross-country regressions for a number of countries covering the period 1976-1993, detnonstrate that various measures of equity trtarket activity are positively correlated with treasures of real activity and that the associatioti is parlieularly strotig for developing countries. Conditiotiing on a number of variables, including indicators of banking development, they conclude that stock markets provide different financial services frotn banks. They argue that stock markets may enhance growth through liquidity, which tnakes investment less, thereby enabling companies to etijoy permanent access to capital through liquid equity issues. Atje and .lovanovic (1993), usitig a similar approach, also find a significant correlation between economic growth and the value of stock market trading relative to GDP for forty countries over the period 1980-88. However. Harris (1997) shows Ihat this relationship is at best weak. Reestimating the same model for forty-nine countries over the period 1980-91, but using current investment rather than lagged, and utilizing two-stage least squares, he detnonstrates that in the case of the full sample (which includes both developed and developing countries), and of the subsample of developing countries, the stock market variable does not offer much incremental explanatory power. In the subsample of developed countries, although the level of stock market activity has some explanatory power, its statistical significance is weak. The volatility characteristic of stock tiiarkets has been investigated from the point of view of its relationship to the size of stock tnarkets and capital control liberalization. Dcmirgu^-Kutit and Levine (1996) find that in a sample of forty-four developed and emerging markets frotn 1986 lo 1993, large markets tend to be less volatile. Also, that internationally integrated markets tend to be less volatile. Levine and Zervos (1995) explore the effect of liberalizing capital controls in sixteen countries which reduced substatitially barriers to international capital and dividend flows in the 1980s. They eoticlude that stock market volatility increases significantly immediately following capital control liberalization in approximately half of the countries considered and does not decrease significantly in any of them. This result tnay be combined with Ihat of Demirgui^-Kunt and Levine (19%) to suggest thaL in the long run, stock return volatility is lower in countries with more open capital markets. However, tnore recently Levitic and Zervos (1998) have examined volatility in a relationship that coticentrates on stock market liquidity and econotiiie performance. They conclude that in a cross-section approach, this link is not statistically robust (and it is unexpectedly positive in tnost cases). We investigate this link further in this paper and are able to improve substatitially on this result and show that the link between volatilUy and growth is significantly strong and negative in our time series analysis as shown below.

I ' H I I . I P A R K S r i S , I ' A N I C O S O . I M ' M L I RI X D H S , A N D K l i L B l . t i | N I K I .

: 21


We arc motivated by two primary objectives: First, to explore the long-run relationship between stock market volatility, stock market development, hanking system development and the level of output. In so doing, the magnitude of the estimated long-run output elasticities wilh respect to the treasures of banking system development and the stoek market development is likely to shed light ()n the relative importance of the two components of the linancial system for output growth, Seeond. to investigate the causal flows in this relationship, that is, hetween output and banking system developtnent on one hand and output and stock market development on the other, Dala and Mea.sitrcment We employ quarterly data on output and indicators of hanking system development, stoek tnarket developmenl and stoek market volatility for Germany during 1973:1-1997:4, the United States for I972:2-I998:L Japan for 1974:2-1998:1, the United Kingdom for 1968:2-1997:4, and France tor 1974:1-1998:1 ."^Our variables are measured as follows. Output is measured by the logarithm of real C3DP (LY); stock market development hy the logarithm of the stoek market capitalization ratio (LMC), defitied as the ratio of stock market value to GDP; hanking system development hy the logarithm of the ratio ofdomestic bank credit to notiiinal GDP (LBY); stock market volatility (SMV) is measured by an eight-quarter moving standard deviation of the end-of-quarter change of stock market prices.' There are,, other possible indicators of financial development. As far as banking sector development is concerned a deposil-hased measure eould also be used. Earlier work by Arestis and Demetriades (1996), however, suggests that in the case of developed economies credit-based indicalors are more likely to exhibit a stable long-run relationship with output than deposit-based ones. As lar as stock market deveiopinenl is concerned, cross-section sludies have found liquidity-hased measures, sueh as the ratio of the value of traded shares to GDP to be more closely linked with economic growth than market capitalization indicators. Nevertheless, in a time series context the capitalization indicator has a number of advantages over transaetion-hased measures. Firstly, it is a stock variable rather than a Ilow variable that makes comparison with the bank development-based indicator, which is also a stock, more mcanitigful. Secondly, primarily for the same teasoii it is tiiore likely to have time series properties that make it suitable for coinlegration analysis. Furthermore, the discussion of section I suggests that there are also conceptual reasons as to why stock market capitalization may be more closely linked t<i economic growth than transactions-based measures. Despite this we tesi the sensitivity of our results

4. AH dala series were extracted from lhc online informalion service naiastrcani InlenHUioiml Slock market variables are end-nl-quaHer price indices and market values. 5. We tjrsi ealculaled the logarithmic lirst dilferencus of Ihe eiid-ol-qiiarler slock market price index We then computed a moving eighi-quarter siandard devi;ttion as a measure markcl volatihlv.



using alternative measures of stock market development; however, this is only possible for the United Kingdom and the United States since data on these variables are not available for the other countries in our sample for a sufficiently long period. Methods We carry out our empirieal investigation in a vector autoregression (VAR) framework. Reeent literature (for example, Gonzalo 1994; Hargreaves 1994; Haug 1996) suggests that for sample sizes of around one hundred data points, the maximum likelihood approach of Johansen (1988) performs in general better than a range of other estimators of long-run relationships (cointegrating veetors). Further, Toda and Phillips (1993). Hall and Wickens (1993), and Hall and Milne (1994) show it to be an effieient method of testing causality. We therefore follow this method to identify the number of cointegrating veetors amongst the variables specified in the VAR and then examine the direction of causality. The Jobansen (1988) method is based on a vector error correction (VECM) representation of a VAR(/.') mode! whicb can be written as

wbere -v is an nxl vector of the first order integrated |that is, 1(1 )J variables, ri,r2,...,r,are/ix/7 matrices of unknown parameters, D is a set of 1(0) determimistic variables such as constant, trend, and dummies, and ii is a vector of normally and independently distributed errors witb zero mean and eonstant variance. The steadystate (equilibrium) properties of equation (1) are characterized by the rank of n , a square matrix of size n. In our case n = 4. The existenee of a cointegrating vector implies that 11 is rank deficient. Johansen (1988) derives tbe maximal eigenvalue and trace statistic for testing the rank of n . Appropriate eritieal values are tabulated in Osterwald-Lenum (1992). If 1 Iis of rank r (0<r<n) then it can be decomposed into two matrices a {nxr) and p (nx/-)sueh tbat

II = ap'.


Tbe rows of p are interpreted as the distinct cointegrating vectors whereby ^'x form stationary proeesses. The as are tbe error correction eoeffieients that indicate the speeds of adjustment toward equilibrium. Substituting (2) into (1) we get A.,=r,Av,_i+r,Av,_,+ + r^, ,AK,_^,,,+a(^'x,_^,)+^D,+u, . (3)

Tbis is a basic specification for the test of long-run causality. A test of zero restrictions on the as is a test of weak exogcneity when the parameters of interest are long run (Johansen and Juselius 1992). Hall and Wickens (1993) and Hall and Milne (1994) interpret weak exogeneity in a eointegrated system as a notion of long-run eausaliiy. We employ weak exogeneity tests to examine the issue of long-run eausal-

P A \ N 1 C ( ) S ( ) . D L M L ' I KIADLiS, A N D KUl^li


ity between tbe variables in tbe system. The null of a = 0 can be tested by tbe standard likelihood ratio test. A number of issues arc important in the estimation and interpretation of coinlegrating vectors. First, in view of the various (financial) policy changes that have taken place during the sample period, it is plausible to allow for the possibility of structural break in the cointcgraling relationships. We address this issue directly by testing the null of parameter and rank constancy in tbe cointcgriiting relationships following Quintos (I99.'i) and Hanson and Johanscn (1993, 1998). The hypothesis of interest here is that II and the rank of IT, pdl), orthc number of cointegrating relationships remains stable overtime. The null of parameter and rank constancy can be stated as H,;. p(II,) = r, or pdl,) = p(n) for all ( . (4)

The alternative hypothesis that allows for both the parameters and number o! cointegrating ranks to change is /(; p(ll,) ^ p(ll), for all or some of tbe / . (5)

Quintos (1995, p. 412) provides a likelihood ratio (LR) statistic which lests the null of no structural break under a single break date. Implementation of Quintos's test requires splitting the sample at the break date; estimating separate models for the preand postbreak dates; and testing whether subsampic eigenvalues arc significantly different from those of the full satnple. In view of the fairly small sample we bave we do not follow this approach. Instead, we implement the rank stability tests in a recursive framework as suggested by llansen and Johanscn (1993, 1998).'' The relevant LR test can be shown as

where X and >^, are the full and rcsursive sample estimates of the eigenvalues of matrix II; subscript; indicates tbe starting date of recursion such that T^ = 1\ + I, r, + 2,...,r. Thus, our approach essentially involves estimating the cointegrating vector(s) using full sample and tben testing whether the full sample results (tbat is, cointegrating parameters and ranks) remain stable when the model is estimated over tbe recursive subsample. The recursive LR test is x\2) distributed. Second, it is shown that cointegrating relationships are sensitive to the treatment of deterministic terms in the cointegrating space (Baillie and Bollcrslev 1994; Diebold, Gardeazabal. and Yilmaz 1994). To resolve this. Johansen (1992) suggests
6, It should be iioled tliat Quiiilus's tcsl is biiscd on Ilaiisen and Jnhanscn (19^)3), O n e of the advantages u\ recursive lests of siruclural hreak is ihat w e d o nol requite KI ideiilily break dale endogeiidusly which is important in view of our sacnplc s i / e .



identifying appropriate deterministic terms in the cointegrating space through rank tests following the so-ealled Pantula principle. Crowder and Hoffman (1996) and Luintel and Paudyal (1998), among others, implement cointegration tests along these lines and we follow this approach. Third, it is well known that the results are sensitive to lag length selection in the VAR. Generally lag lengths are specified following some information eriteria (for example, Akaike l973;Schwar7 1978). However, Hall (1989) and Johansen (1992) suggest that the lag length should be specified such that the VAR residuals are empirically Gaussian. Cheung and Lai (1993) show that the lag length selection based on information eriteria may not be adequate when errors contain moving average terms. We specify lag lengths as tbe minitnum length for which there is no significant autocorrelation in Ihe estimated VAR residuals. Finally, the identifieation and interpretation of the eointegrating vectors, p, as long-run economic relationships is another pertinent issue. Given r cointegrating vectors, Johansen (1991) suggests identi licalion through the lesl of r'jusl-idennfyin^ restrictions. Pesaran and Shin (1994), however, show that Johansen's identification scheme is deficient in that the maximized log-likelihood values associated with any set oi just-identifying restrictions would be identical, which makes it impossible to disfinguisb between the competing set of restrictions. Instead, they suggest identification of long-run economic relationship through fhe tests of over-identifying restrictions. This requires testing for r'+k (where /: > 1) restrictions in a cointegrating space spanned by r stationary relationships which gives k over-identifying restrictions. It should be noted that when a unique cointcgraling vector is found then the problem of interpretation does not arise, and one may simply test for just identifying restrictions in the form of a normalization restriction.


We begin by carrying out unit root tests whicb suggest that tbe all variables are 1(1).^ We then perform cointegration analysis for each of the W\Q countries, the results of which arc reported in Tables 1-5. In order to allow for any deterministic seasonality, centered quarterly dummies are ineluded in unrestricted form throughout the estimation. Part (a) of each table contains the results frotn recursive estitnation. In view of the sample .size, ihe starting year for the recursive estimation is chosen as 1990(4) for all countries except for the United Kingdotn, in which case because of the longer sample we begin recursive estimation at 1987(4). The last column of each table reports the LR tests under the null that the full sample eointegrating rank is stable over the recursive subsample. The rcjeetion of the null indieates structural breaks in the cointegrating rank that fortns the basis for the introduction of appropriate shift dummies. Once the structural break is identilied, we then reestimate the cointegration rank
7. Uiiil rool tcsis tor all variahtcs (which are not reported here) can be otitmned from ihc authors upon request.

i ' l i n . l P A R E S T I S , I ' A N I C O S O . DI-.Mt-l R I A D I ^ S . - \ N I ) K t H . Ii. I . i : i N r i . l ,

: 25

for the full satnple by allowing for structural shifts through shift durntnies.^ Sinee trace statistics and maxitnal eigenvalue statistics provide qualitatively similar results, only the former are reported for the sake of brevity, hi the cvetit ol' multiple cointegrating vcetors, identifteation of each vector (o an econoinically interprelable relationship is achieved through tests of over-identifying restrictions (see Pesaran aticl Shin 1994). Each vector is normalized on the variahle for which we could lind evidence of error correction (that is, negative and sigtiilicatil loading iactors, a). These results are reported in part (b) of eaeh table. Finally iti part (c) we report the resttlts of weak cxogeneity tests, which arc expected to shed light on the pattertis of longrun causality in each system. The results on Germany, reported in Table I. show evidence of a break iu theeointegrating rank during 1991-92, which coincides with the period of ihe Gertnan reunification. Once this structural hrcak is taken into account through the introduction of an intercept dutnmy for the period of I99I( I)-1992(4) in the coinlegrating spaee, we continue to tind evidence of a single cointegrating vector. Tests of normality and serial correlation suggest that the VAR residuals arc empirically Gaussian.'' The cointegrating vector is normalized on output, given the correctly signed and strongly sigtiificant error correction term (a). The cointegrating vector for (his country shows a positive relationship between the level of real GDP atid banking systetn development, as well as a positive stock tnarket capitalization elTecl. It also shows that stock market volatility, a variable treated weakly exogenous to the systetn, has a positive but insignificanl effect.'" Flanking system development is endogeneous to the output vcetor whereas stock market capitalization is tiot. Hence, in Germatiy, there is bidirectional causality between banking system development and the level of output while stock market capitalization is weakly exogeneous to the output vector in the long run. St<ick tnarket eapitalization, however. alTccts Gf:)P through the positive and significant cointegrating paratncter. The coeflicients on LBY and LMC arc significant at the I percent level, with the former heing tnore than three tirnes larger than the latter. These results are clearly not surprisitig given the close relationship of the banking system with industry in Germany and the tekttively tninur role played hy the stock tnarket there (see, for example. Arestis and [)ctnetriades 1996). In the case of the United States (see Tahle 2) the picture is rather dilTerent in view of the endogeneity of stock market capitalization and the weak cxogeneity of real GDP, once the structural shift in 1990:1-1994:4 is accounted fur. This period eoiticidcs with a downturn in [he U.S. economy, a suhstaiuial fail in hotid market yields and a serious nutnber of defaults of "jutik" bonds. There is otily one eoititegratitig vector for this country, which is normalized on LMC. According to this vector, LMC
K. W e tested for both slope a n d iiitcrcep: J u m n i R ' s but tiiL- Inruicr w e n : iri.sigiii(jc;int ii) all ciiscs, y. In all cases recursive e s t i m a t i o n is c o n d u c t e d wiihoui ihe iiitroduciion (il any diiniiiiy variable. H o w e v e r , l e w d u m m i e s were intRiduced in the olhL-r csliniatioiis in capture bliiis iu ihc data, Ir'xehision oT these d u m m i e s d o e s not e h a n y c the results (|ualilatively, except lor failure of ihe n o m i a l i : \ lest. T h e s e are not reported hut ean be obtaiued from Ihe aulhors. 10, T h e likelihood ratio tests eiiuld uot rejeci the weak exojieiieitv of SMV. T h e test staiisiic is distributed as ehi-sc|uare( I) w h i c h gives a / ' - v a l u e o r o . l . ^ r i . In Tables I-7."the cocfficieuts tor S M V . unlike ilic o t h e r coeftieieuts, are not elasticiiics.


Tra.o Stali>.lk- II,,: rank =1' ,,-0 p---\ >, 2.71 3.05 2.24 1.92 2.07 1.92 1.64 1.71 0.369 0.404 0.365 0.320 0.309 0-296 0.272 0.273

Ki^envakiL-s A.. >.. Rant, stability tL-;l

1990(4) 1991(4) 1992(4) 1993(41 1994(4) 1995(4) 1996(4) 1997(4)

48-18'" 52-89' ' 49.16' 45.19' 46.46' 46.80" 46.03" 47.90"

17.31 16.10 15.06 14.68 15.79 16.22 17.12 I7.5K

0-196 0.168 0.157 0.149 0.152 0.152 0.156 0.154

0.039 0.042 0-029 0.024 0.025 0.022 0.018 0.018

0.54 6.54 5.98 ' 0-81 0.34 0.26 1.42 Full sample


53.67"' N(ii[iiuh/.L:d on LY Const.n,

18.03 Dummy: I W K I ) IW2(4) LMf

2.59 1 BY SMv'

Coefficient /3-value

5.893 0.0002

0.0633 0.0623

0.1316 0.0079

0.4405 0.0000

1.239 0.1243

/'VLiliiL's arc tlij[ iif thf likelihood ralio Icsts iimlci lhc null (hjt Uie paramwcr i s / o r ( VLvliiraulouirralalionlcsts: r"(4S,lSN) = 1 I ' l ^ i |(l 21 IO| Vcclor iKiriiialily IL-SI: i-lii-sqiiare (hi: AXKl |0.()652]

I r : WRAK ExoGtiNtiiiY Tiisrs LY \MV LBY

Loading (a) /)-value

-0.1450 0.000

0-2550 0.3678

0.2148 0.0{K)

N()ii,s:/j-valiicsiirc Ih.ilof ilic likclihiKid liitio tL-sl-i under I hi,' null Lhal [lif loadiii|! hit Ior ih zero. * " , +" and * indit:ali: suiiiMnial .ignilR-jnui; al 1 (X-rCL'iil, ^ pOKTiil. and W piTconi. rL->|iL\-lively. ^

TABLf- 2
Fl.NANci.M. !,n)'Ml'M 2A: W I ) |-,c(i\<i.\iii ( l i - ; n w i i i i s-Dii- l ; \ r i i - i i S i M i ^ Vi'C'i'Oks ( I . M , 4 } K t( TKsiM', [ i s ' r i M . M K A UI C'oi\ii'i;R,\ii\"(i

Silllipk'. I'IT;,:,

1 p (1


r 1 4.^.23 43.2K 42.28 ' 41.21 42.89 42.73 43.76 42.24 "


>., 0.400 0(17 0.278 0.277 0.267 0.260 0.251 0.261

A ,

'.. 0 0 0 (1 0 0 0 0. 176 162

R.Lni,-.[.ihililv k'M

1990(4) 1991|4) 1992(4) 1993(4) 1994(4) 1995(4) 1996(4) 1998( 1)

78.0 71.29 67.04 67.14 69.00 69.23' 70..VJ 7! .58

20.08 I''.7O 19.00 19.42 19.41 19 91 19,62

0 ^98 0 ^8^


0 2'^7 0 "'4'' 0.249 O.\^^ 0 ''"'8 0,208

1 ^7 135
1 "^2


12.01 4.61 3.8'' "^.39 2.31

1.4S Iiil1-s;imple


2t!: H.S'IIMAIM)

c;K\n\(] V i - c i o'i; ,\rii',R ,'\i.i owiNd 1 O .Si R

ll.n-i- SlilllMU p 0
|[,, 1.111k /I

(I .x<;^^-1)

Norrii^ili/.-ihiii i.MI

.MM, -0.1939 1,1

1 \i^

CoefficiciU /)-\llllK'
P >.;ilin.-s a i c IIKLI n l Ilu- l i k . ' \\\h. .lLllll^.^lrlcl;llHllllLM^



3.208 0.000

0.447 0.779

6.331 0.181

I[IIIC[ IIU- n u l l III.LI llk> |>,ii^ii

Ic: Wi.AK h


l.iiudiiig (a)

0.0013 0.617

0.109 0.0001

0.0(13 0.269

Nulls./>-\.iUiL'-,ii 111. JElll M)pi'll-ITL[. [..p,\


is positively related to real GDP and to banking sector development (LBY) and negatively related to slock market volatility. However, only the GDP coefficient is significant in this vector. The weak exogeneity of real GDP and banking system development suggests that capital market development, which in the long run is influenced by these two weakly exogenous variables, has no long-run causal influence on either of these variables. This result shows that there is clear evidence to suggest that in the U.S. financial development does not cause real GDP in the long run. Consequently, these results are in sharp contrast to the ones obtained for Germany and may, to some extent, reflect the international character and the nature of the banking system in the United States which is a capital market-based system (as opposed to the bank-based system in Germany). The results for Japan are reported in Table 3. The recursive estimates show two cointegrating vectors throughout 1990-1995 and one cointegrating vector after 1996. Such a shift in the number of significant eigenvalues is indicative of a structural break in the long-run relationship. We therefore treat the data points up to 1995:4 as the full sample and compute LR tests of rank stability for the other years reported in the table. Tests reveal that in Japan structural breaks in the cointegrating ranks appear during 1991-1992 and 1996-1998. The former period coincides with a steep decline in bank proflts, reflecting worsening in the scale of nonperforming loans, due to a sharp fall in asset values, especially in the property market, and a tight monetary policy. The structural break during the period 1996:1-1998:1 is clearly not surprising given that this period coincides with one of the worst economic crises in Japan's postwar history. Importantly, the Japanese financial system has been at the center of these problems, with many financial institutions becoming insolvent. Once this break is taken into account through the introduction of intercept dummies for 1996:1-1998:1 two coinlegrating vectors emerge. These vectors are respectively normalized on real GDP and banking system development, which show clear evidence of error correction. The over-identifying restrictions that are accepted by the data include two normalization restrictions, linear homogeneity of LBY and a positive coefficient on LMC in the first vector and exclusion of LY from the second vector. The first vector shows that both the stock market and the banking system development indicators influence output positively; however, the influence of the former is about one-sixth of the latter. Stock market volatility, on the other hand, exerts a negative and significant influence on the development of the real economy. The second vector is essentially a positive and significant relationship between the banking system development and stock market capitalization. Once again, stock market volatility enters with a negative and significant coefficient, suggesting that inereased stock market volatihtya weakly exogenous variable to tbe system"impacts negatively on the development of the banking system. The estimated banking development vector displays a downward shift during 1996:1-1998:4, reflecting a significant autonomous shift of 1.13 percent. This effect is not significant in the output vector.
II. Weak exogeneity tesl of SMV from the system asstimcs a/rvalue of 0,972 which is chi-squarc(2) distributed.

KlNANtlAL Di:vi"l OPMLNT ANI) R c O N d M [<' (.JRdWlH [N J A I ' A N 3 A : R E C T RSiVI: E S T l M A r i d N O l - C o i N l F X i R A l l S d V l - . t K I K S ( l , M i - 5 ) T.I, f Stall^lle H,,: ratik-" /' 0




i-m,n. 1990(4) 1441(4) 1992(4) 1993(4) 1944(41 1445(4) 1996(4) 1998(1)

'-, 0.333 O..125 0.356 O..^21 0.282 (1.234 0.2.1(1 0.230


'., 0.167 0.135 0.160 0.1M 0.112 0.1 14 0.114 0.11S


68.46' 64.46 ' 68.56 65.73 61.42

5 1.80 53.40'

41.27"" 41.54 35.53 35.12 33.97 .36.04 ' 27.99 28.32

18.91 16.56 15.35 14.58 12.38 13.34 13.71 14.71

0.2K4 (1.247 (1.236 (1.224 (1.224 (1.23(1 0.145 (1.1.32

3.11 6.47 6.77 4.72 2.61 # 22.63 21.37

3B: QiiN'iFxi

i V i x r o R . s Arrr.R A L L O W I N G I U R S I R I ( I T R A L B R I ^ A K ( L A G ^-5 iank--r ;.- 1 "

; ' -

80..,Vector 1
Normalized on l.V C'on>[ant

23.01 "

1 MC


Cocl'licJcnl /J-value Veetor 2 Nmiiali/fdmiLlty Coettieient /)-value

Niin^s' I Treated /i-v:iliies aiv tliul o

12.92 0.000



1.01)0 0.000

0.130 0.008

-1.336 0.000

,.>..,....,., , . m
0.5434 0.0002 0.1130 0.0381

1 ,M(



1.755 (1.0(1(1

: likdilioi>d EiiluilL-> i.'ivsliirtions:i.hi-M wsts I-(4.^.lf)l) -= elii-M|iJaio (()) S 72

: the mill tli;il Ilk-


Loading (a) ot vector 1 /'-value Loading (a) of vcctur 2 /'-value

-0.1471 O.(KH)O 0.0763 O.OiKlO

0.1661 t).OII<) -0.1410 0.01 II

(i.yoii (1.0000 0.4(105 0.0002

NOTI s: ;j-valjos are that ol the likelihoixl lali,. Icsls Lindci tlie iiiill thai [he li>adip - - ' . "* :irld * indn'iitc stiiti-.lii::il Miiililkanti- ;it I (lerecnt. ."i |>eri.eiit. ;i[id HI jM-'ie

30 :


Finally, the weak exogeneity tests suggest a feedback relationship between real GDP and both parts of the tinaneial system sinee all three variables are endogenous to the system. With perhaps the exception of the negative and signiftcant influctice of stock market volatility, these results should not be surprising in view of the relative importanee of the banking sector in Japan (Corbett and Jenkinson 1994; Arestis and Demetriades 1996).'The results pertaining to the United Kingdom, presented in Table 4, display significant differetiees with those of the other countries, reflecting perhaps the uniqueness of its linancial system. To start with, we tind evidence of a structural shift in the estitnated relationships during 1987:1-1991:4. As 1987 was the year of one of the most important deregulations of the U.K. linancial system in recent historythe Big Bangthis is once again not a surprising result. There were also statistical redefinitions in the mid-1980s, pertaining to the inclusion of Building Societies in the banking system statistics. Once these structural shifts are taken into account, we find evidence of two cointegrating vectors. They are normalized on stoek market capitalization and banking development. The data-identifying restrietions are the following five: exelusion of real GDP from the tirst eointegrating vector, exclusion of LMC and linear homogeneity hctween LBY and LY in the second vector, and two normalization restrictions. The first vector is a simple and straightforward positive relationship between stock market capitalization and banking sector developtnent: the two parts of the financial systetn exhibit a stable long-run positive association^subject to a shift iti the 1987-91 period. It also shows that stock market volatility impacts negatively on stock market capitalization. The second vector suggests that banking sector developtnent is explained by real GDP growth, while stock market volatility exerts a significant negative influence. The weak exogeneity tests show that real GDP is weakly exogeneous with respect to the LBY vector, and marginally so in the case of the LMC. Thus, the LMC vector appears to eause LY, although marginally, in the long run. The long-run banking development vector has no relationship with the level of output and stock market development. In the long run, causality runs from LBY to LMC and from LMC to GDP. There is no direct long-run causal relationship between banking system developmctit and real GDP for this country. In conclusion, the evidence on the Utiited Kingdom suggests that in the long run causality flows from banking systetn development to stock tnarket development. It is also evident, however, that the fiow of eausality from financial systetn development to real GDP is, at best, weak. On the other hand, banking system development and stock market development are both negatively affected by stock market volatility. This evidence could also be interpreted as suggesting that the U.K. financial system is not a strong promoter of domestic economic growth, which to sotne extent reflects its weak links with industry, in that it is a typical capital market-based system, and its international character. Turning finally to France, we find evidence of instability in the cointegrating rank
12. Results iueluding shift dumuiies for the period m'JI(l)-l992(4) are tiualilaiively similar. Hence they arc not reported for the sake iif brevity but are available on request.

I'lNA.NCIAl DlA'l.l,nP\U;NT AND HCd'>;()\1I<' (iROWIll IN 1 IKrii-iJ KiN<-,i)O.\i III. 4A: RixiiRsivi', J-STIMAl"i()\ nt- ('(ll^;ii-<',RAnN(; Viicroi'iS (L..\(i 5)
S.inipkI't7(,(h la'SlaliMK- !!: Enilk p
p:. 1 p: 1

'-' 0.234 0 "'^l 0.158 0.156 1).! 54 0.153 0.143 0.121

1). ! 1 6 0.108 0.094


R: u,ks,ah,hl,lw

1987(4) 1988(4) 1989(4) 1990(4) 1991(4) 1992(4) 1993(4) 1994(4) 1995(4) 1996(4) 1997(4)

67.40" 59.47 49.66 51.45 5 1.85 53.32" 54.65 53.K9 55.52 56.60 56.75

35.85 36.67 25.44 26.1 1 25.62 25.93 26.39 24.79 25.13 25.16 24.34

14.78 15.94 10.44 10.62 9.6S 9.52 10.54 10.96 11.46 12.01 12.59

0.329 0.240 0.243 0.243 0.241 0.242 0.240 0.238 0.240 0.240 0.240

0.I4S 0.153 0.090


''0.171' ' 11.099 S.5.t2 7.062''


0.071 0.065 0.067 0.069 (1.069 0.068 0.067

4.017 4.429 2.610 "'.019 0.969 1'uii-sample

4H: HsiiMAii.n Coixn-x IkAIINfi Vl.( lOKs

Ai'ii'K Ai.i.owiNc; i-ok S i R i r r i ' R A i . B K I - A K ( L A I ; - ^ 5 )

rrn.L-Slali.[K 1 1,,
,'.:- 1


1'' -

105.50 " Vector 1

47.7'' ..,,nu,,>;,m,,.,."II (41


CoefiieienI /i-v;tkte Vector 2

Nii[[iialL/alon [.in

I).79I 0.0tH)4

0.490 0.026

l).499 0.0527

- 33.57 0.000



CociVicienl /I-value



0.00 II

1.000 0.000

-31.080 0.000

N i r i i s; //-vahii.". aiv t l U l l l l lIlL-,i

ItsiMindci IliL-niill [h;ail ll- [i.ii-anu'[i.-i IS /L-ro

Vector jiiliiL-iH-EL'laiL i i i i k ' s i . - !"(; VrtHii iiiiri[i;Llit> iLL-(Xi: 1 1 i>.V- 1 LVi;] |(,
' ''""""'"

4c: Wi",\K Hxoc;i.M.nT Ti',s"is


i.llY 0.081 0.000 -0.082 0.0001



Loading (a) of M'clor 1 /'-\aluc Limdjiij; (a.) ol vcclor 2


0.0012 0.601 1

(1.0 l.M) 0.0531 0.006 0.3247

-0.014 0,000 0.003 0.000

Miiii>:;i-uiliii.-saK-ili.L[iil ilio liki-lilnunl r.iin'IL-M. umlor [ho luill Hint llu- k ' - . - * und iiidii-;LlL'sMli'.ln.Ml Mj>rn(k,in,c ;il I |H-in-nl. . i i r m - n l . and M) S



during 1990:1-1993:4. This period follows the completion of fmancial reforms, wbich commenced in 1984 witb the comprehensive reform of French banking legislation, the official policy of deregulation and the harmonization of institutional arrangements wilhin the European Union. By 1992 capital account liberalization had been completed. Also, 1992-93 saw tbe FRM turmoil with the French franc coming under strong speculative attack. Furthermore, given the important deregulation ofthe Freneh financial system that took place during the mid-1980s, we allowed for additional structural shifts by incorporating an intercept dummy for the period of 1985:1-198^:4 in our estimation.'-'Thus, for France, we allow for two sets of shift dummies, for 1985:1-1985:4 and 1990:1-1993:4. Cointegration results for France, wbich take into account the above structural shifts, are reported in Table 5b. The traee statistic shows evidence of two eointegratitig vectors. On the basis of the sign and significance of the associated loading factors we normalize them on real GDP and stock market capitalization respectively. The identifying restrictions that are data aeceptable comprise exclusions of LMC and the 1985 dummy from the first vector, exclusion of LY from the second vector, and the two normalization restrictions. The first vector shows a positive impact of banking system dcvelopmenl and capital market on real GDP and a negative and significant effect of stock market volatility. The second vector, portrays a positive relationship between tbe banking system and stock market capitalization while stock market volatility is found lo have a negative influence on LMC. This veetor appears to have been affected by lhe financial reforms of the mid-1980s, which seem to have boosted stock market capitalization. In this sense this provides support to the argument put forward by Arestis and Demelriades (1996) that the French tinancial sysicm shifted toward becoming a capilal-market-based one. following those reforms (see, also, Bcrtero 1994). Altbough both the stock market capitalization and banking system development appear with positive and signiticant cointegrating parameters in the output vector, it is iniportant to note that the magnitude of the latter's eoefficient is almost seven times that ofthe former. Tbus, the effect of banking development on real GDP appears to be much stronger than that of the stock market. The weak exogeneity tests reveal that no variable is weakly exogenous with respect to any ofthe vectors, suggesting an abundance of feedback relationships between the variables in the system. In the long run there is feedback between LY and LMC as all loadings are significant. Both real outpul and stock tnarket development vectors enter into the banking system developtnent suggesting that in the long run both LY and LMC cause LBY. In conclusion, the results on France suggest that while the French financial system has directly contributed to long-term output growth, the role of the banking system has been much more substantial than that ofthe stock market; this was in spite ofthe growing importance of the latter as a result of the financial reforms of the 1980s. Furthermore, even though the long-run dcvelopmenl of the stock tnarket has gone
13. in view of Ihc sample si/e we did nol curnpute recursive lests of stability for 1985:1-1985:4 period.

F i N A N r i A i , Di-:\'i.i.(ii'Mi'M 5A: .\M)EcoNdMif CiH(m [ i M N I ' K A N C L 4) Ri-(TR,si\-i. i{sTi\i.\iii)N (II' C o i M i < i R , M ! \ ( ; V L C I D K S ( L . M i

S-,,n k1'17,1(1)

1 nnY

Stiiti-tK H,,: i.iiik /' 1

,- -0

p- 1


>' 0.204 0,190 0,181) 0.178 0,16.^ 0.167 0,154 0.149

'0,1 27 0,1 31 0,1 ,t3 0,1 32 l),l 15 0,1 18



1990(4) 1992(4) 1993(4) 1994(4) 1995(4) 1996(4) 1998(1)

55. W

24,72 26,25 27,2(1 2,"^ .41 27,31 26,89 27,53

56,22 5H.42 60,21 60,43

U),20 11.18 1 [ ,.^^8 10,26 11,28 1 1,53 11.89

0,333 0,322 0,316 0,295 0,307 0,298 0.288

5,40" 4.90 4,97 4.98 2,08 1.79 17 0,42 ]5 Inill-satuple

OUMi .(;UATiN(iVl,(n()K S ;' "


MTIJi A M . O V,-|N.; [''OK SCKt ( 11 K.\l 11,, i.nk ,,

, BKI,\K (1 ..\(. 1 4 )

r :



Vector 1
N"o[jii;Lh/,Mliiii 1.1

0,0561 0,0017




Cocl'licient /I-value Vector 2



3,35S 0.0023


nn 5.164 0,000


Cdcnicicnl /)-value
.N"(ii]. p v,iliiL-.;irv[ha(ot ill. TL^l ot miT u]onlil>m_E EV-lELt

0,738 0,007(1


thi-si(ii,Liv( I

5r: WLAK


l.oaiijng ( a ) of \x'ctor 1 /I-value Loading ( a ) iil' veelor 2 />-value

-0.0160 0.0059 0.002

1,0730 0,0031 0,057 0,002,S

I),O79I) 11,0003 0,0060 0.0039

0.1040 0.0000 -0,0006 0,IH)(i8



hand in hand with the development of the banking systetn, the rt>rmer appears to have been rnuch more of a follower in the process of economic development, responding posilively to both output growth and banking system development. On the other hand, stock market volatility seems to have been detrimental to both long-term output growth and stock tnarket development. Aliernative Measures of Stock Markei Develupment: Sensitivity Checks In order to check tbe sensitivity of our results vis-a-vis the alternative measures of stock market development, we repeated the empirical analy.sis using two further measures, namely, the logarithms of the ratio of lotal stock markei transactions to GDP (TRY) atid the ratio of total stock market transactions to market valuation (TRMV) for the United States and the United Kingdom. The stock market transaction data for the United Kingdom are available for a period of 1976: i-1998:1 and for the US 1973:1-1998:1. In view of the shorter samples for the United Kingdom, this sensitivity analysis should be taken as indicative only. For other sample countries lack of data on slock market transactions, or the insufficient length of this time scries, prevented us from carrying out sensitivity analysis. The results.obtained from alternative treasures of financial development are reported in Tables 6 and 7. Tbey are, in qualitative tenns. broadly similar to those obtained using the stock market capitalizatioti variable. Therefore, in the interests of brevity we report only a summary of these findings. Speciftcally, for the United States we continue to tind one cointegrating vector, whichever of the above measures is used. In either case the cointegrating vector is nortnali/.ed on the stock market variable in view of the correct sign and signilicance of the corresponding loading factor. The normalized vector depicts a positive association between stock market development, banking system development, atid real GDP. Real GDP and banking system development appear weakly exogenous when using the first variable but become endogenous when using the second. Stock market volatility has negative effects in both vectors, albeit significant only in the TRMV vector. Further, weak exogeneity tests reveal qualitatively sitnilar causal patterns to those reported in Table 2c with respect to TRY vector whereas LY atid LBY become endogenous with respect to TRMV vector. In the case of the United Kingdom, the results reported in Tabie 7 still show two cointegrating vectors with the first alternative measure of stock market development. TRY The iirst cointegrating vector, normalized on TRY, is very similar to the one normalized on LMC (sec Table 4b) and shows a positive association between stock market dcveioptnent and banking sector development; stock market volatility exerts a significant negative effect on TRY. The second vector, normalized on LBY, now shows an insignificant real output effect but the volatility effect is negative and much stronger. Weak exogeneity tests that there exists a feedback effect between stock market development and banking systctTi development; however, the causal flows from financial variables to output are more significant now. With the second alternative measure (TRMV). we find only one cointegrating vector, which is normalized on LBY and shows a positive and significant relationship between l RV r.nH

TABLE 6 F I N A N C I A L D [ : V I : I , ( ) | ' \ I I - . M A N D He O N O M I C ( j R o w d i IN riii. I I M I C D .SiAii.s T K A N S A C T I O N S - B A S I - : ! ) M i . A s u R i s III' .SiiHK M A R K L T D L M . I . O I ' V K M S . ' W I I ' M : l ' ) 7 3 : i -

A . R A I I O O F S T O C K M.M<1SIJ TkANSAdiONS C l H N I " i i ; K A I I N < ] V l . C I O R .WVV.R


Al.l.OWINCI f O k S I RI C l I R M. B R I . \ K

]r.\i.\' Sl.ili>[iL I I , , liir

31.32 CoiiUcgiatirig Vector Ndiniali/cdorTkV c,,,.u,n Diiiiiiin l'(.S7(h- l'WI(,|) ,,,n 1 V .S\l\ '

t'oefiicieiil /'-value


-0.128 0.532

i .464 (1.235

5.2')') 0.003

1.775 0.641

Noil s: I. TieiitL-d : f wi;;ikly V.Tt^lr^lUHNll^vl;l1n>lUL.^^ Vi-tHirjuiiiiulKy U-M i,-lll-x

:si |0

Wi-.AK hx(K;i:\t:n Y 1 I , S I S 1 V lin 1,1! Y

l.nadiug P'\'ii\uC


0.0024 0,5881 (I.OIHI

0,005 0.3026

H. RAMOOI- S K K K MARKI;I '["i<A\,SAr IIONS r n S l I K K MARKI',1 VA: H S I I M A I I , ! ) C d l N l l i G K A T I N i : ; V l C I ' O l ; .M H-.K .1^il,I.O\MN{} [ U S i R r c ' i f i R .K; SAMI'I,!.: 1973:1-I99X: 1

36.80 C ouite^ratini:[ Vector N"i.n.i,ili/L-duirn ( M V 1 UIISI.LIU






Ciiefticietit /'-\alue


0,181 0.426

.i,372 0.044

2.527 (1.009

7.144 0.054

12'^ \i r niii-|ii;Lli[y l e s l t h i M t l i ^ i K - I U ) K : i 4

t 1 Y TkMV I.liV

Loading (a) /'-\alue

0,0068 (1.0440

-0,0815 0.078 (1.0642

' ' ". " .md iiidKiilc ,kill^ll^:;^l 'i^rnlKanti- .il *


r-2 18.75

Diimiiij: l9K7(h-l99l(4)

Vector 1
Noriiiuh/cd on 1 KY Constant

Coefticient /7-value Vector 2

Norniali/ediin LBY Coefficient / > value

-0.932 0.762

0.553 0.020

0.630 0.000

-34.74 0.006


Diiiiimy: 1987U) IWK4)



-0.697 0.006

2.122 0.003
I lliLit lhc pLiraiiictcr i^

-8.506 0.120

131.40 0.0054

NiJ-iT'S /.-valJ^:^ arc thai of lhc likchlmod rLilio Icsts iiiidi'i lhc; Vector iuilocoriclalmnK-sK FiSG.IK.) - ^m^\l)Af,2] Vector mil malily Idsc ehi-sijiiaii.-(S): II 8.W |O.I5f>| W E A K HXOCJF.NFITY T E S I S

Loading (a) of vector 1 /*-value Loading (a) of vcclor 2 /;-vaiuc

0.0069 0.0045 -0.002 0.005

-0.154 0.0004 0.030 0.0002

0.1096 0.000 -0.0363 0.000

0.1340 0-0004 -0.0065 0.0006


76..57 Comtcgraling Vector

Normiilimi on I.HY Ciin>.l,iiii

Duniiiiy: 19870 )-lWl(4)


Coeflicicni /7-value

-8.607 0.001

0.452 0.003

0.612 0.0004

1.598 0.258

- 15.85 0-000


i>rrclaiuiii [cMs: l^Xd.l Id) = 1.072 |0..t ^iljlv ItM: fhi-square (,S|: 9,920 10.2711


Loading (a)
and " indnatc slatl.slic.ll

0.009 0.038

-0.006 0.939
nd 10 |icrccill, i

-0.072 0.001

-0.012 0-0004

K t - . S r i S , I ' A N U O S O , D h M I . I R I A D t i . S . A . \ i ) K U i . l i I.L IN I L L


TRMV. The real GDP effeet is insigniticant and stoek market volatility is signitieantly negative. Weak exogeneity tests indieate catisality from LBY to LY and no association between LBY and TRMV. Impliealions for the Debate on Financial Systems Hven though our results vary aeross countries, they aceord reasonahly well with widely accepted views regarding the comparative ability of variotis types of financial system to stimulate investment and growth. Specilieally, our lindings are broadly consistent with the view that bank-based financial systems are more eapable of promoting long-run growth than Anglo-Saxon type systems beeause they are hetter able to address ageney problems and sbort-termism (for example. Slight/ 19X5: see also Singb 1997). Speeitieally. both (i) the positive inlluence of the hanking system on real GDP in Germany, Japan, and Franee and (ii) the ahsence or weakness of a po.sitive causal link from hnancial development to real GDP iii the United Kingdom and the United States, aceord well with this view. What is also interesting in the cases of Germany and Japan is that both banking system and stoek market development seem to have played a positive role in promoting long-run growth, even though in quantitative terms the contribution ofthe stoek market was stibstaiitially smaller.


Our empirical analysis shows that while stoek markets may be able to contribute to long-term output growth, their inlluence is, at best, a small fraetion of that of lhe banking system. Speeitiealiy, both stoek markets and banks seem to have made important eontributions to output growth in France. Germany and Japan, even though the latter"s eontrihution has ranged from about one-seventh to around one-third ot the latter. Finally, the link between tinaneial development and growth in the United Kingdt>m and the United States was found to be statistiealiy weak and, if anything, to run from growth to linaneiai development. Thus, otir lindings are consistent with the view that bank-based linaneial systems may be more able to promote long-term growth than capital-market-based ones. Our findings also suggest that stock market volatility had negative real effeets in Japan attd Franee. In the case ofthe United Kingdom stock itiai-ket volatility seems to have exerted negative effects both on financial developmeni and output. Finally, the effects of stock market volatility in Germany were found to be insignificant. While in principle the presence of volatility in stoek priees may refleet eftieieiit functioning of stock markets, our tindings do not support this hypothesis. Furthermore, they are consistent with the findings of Ai/enman and Marion (1996), who found that other measures of volatilityliscal, monetary and externalalso have negative real efiects. This, of course, may suggest that volatility of any kind rellects general econotiiie uncertainty and is, therefore, negatively correlated to real economic activity. Clearly further research is needed before more delniitive eonelusions ean be drawn on this issue, espeeially on the ehannels through whieh stoek market volatility may affeet eeonomic activity.



The rich diversity in our results complements the findings obtained from crosscountry growth regressions concerning the re]ation,ship betvv'een stock markets and growth. It also confirms the view that cross-country growth regressions at best only provide a broad- brush picture of the relationship between financial development and growth, whieh misses out many important details. There are good theoretical reasons why the reiation.ship between the financial system and growth may vary substantially across countries. This paper's findings suggest that these reasons are empirically sound. Thus, the broad-brush conclusion that stock market development helps promote economic growth must now be viewed with some caution.

LITERATURE CITED Ai/enman, Joshua, und Nancy P. Marion, "Voliililily and ihe Investment Response," National Bureau of Eeonomic Research, Working paper 5841. November 1996. Akaikc, Hirottikii, "Information Theory and an Extension of the Maximum Likelihood Principle," In Second International Symposium on Information Theory, edited by B, Petrov and F. Csaki, pp. 267-81. Budapest: Akademiai Kiado. 1973. Akyu/, Yiltnaz, "Financial Liberalization: The Key Issties," In Finance ami ihe Real Economy, edited by Yilnia/, Akyuz and Gunther Held. Santiago: United Nations University. 1993, Arestis, Philip, and Panicos O, Demetriades, "Finance and Growth: Institutional Consideraiions and Causality." Paper presented at ihe Royal Economic Society Annual Conference, University of Wales Swansea. 1-4 April, !996, _____ . "Financial Development and Economic Growth: Assessing the Evidence," Economic Journal 107 (May 1997), 783-99. Atje, Raymond, and Boyan Jovanovic, "Stocks Markets and Development," European Economic Review 31 (Apn\ 1993), 632-40. Baiilie. Riehard T, and Tim Bollerslev. "Cointegration, Ftactional Cointegration, and Exchange Rate Dynamics." Journal of Finance 49 (June 1994), 737-45. Bencivenga, Valerie R., Bruce D. Smith, and Ross M. Starr, "Equity Markets, Transactions Costs, and Capital Accumulation: An Illtistration," World Bank Fconomic Review 10 (May 1996), 241-65. Bertero, Elisabetta. "The Banking System, Financial Markets, and Capital Structure: Some New Evidence from France." Oxford Review of Fconomic Policy 10 (Winter 1994), 68-78. Boyd, John H., and Bruce D, Smith, "The Coevolution of the Real and Financial Sectors in the Growth Process." World Bank Economic Review 10 (May 1996), 371 -96, Caselli. F., G. Esquivel, and F, Lefort. "Reopening the Convergence Debate: A New Look at Cross-Country Grt>wth Empirics," Journal of Economic Growth I (1996), 363-90.
14, One of the refcrtjcs suggests tliat the comparison ot" our results with tliosc of cross-sectional studies is sotiiewhat unfair, in that our study is not rciilly ahoiil the lung-run effects of financial intermediation on L^cotiouiic growth in the spirit of cross-section studies thai include both developed and developing countries. Whilst we accept the premise of the referee's argtuiienl, we would make two comments. The first is ihat mixing developed atid developing countries may not he so fruitful in view of" the institutional differences hcfween them. The second comment is that a cnjnparison hetween time scries results and tht.ise from cross-section studies is slill useful per se, although the qualification by the anonymous referee should be borne in mind when attempting such comparisons.

' l l l l . l l ' A R F i S r i S . I'.XNICOSO. Dh.MinKI.AIM'.S. A M ) Kt.iL H. I.CINTEL

Cheung, Yin-Wong, and Kcin-S. I.M. "Finite Sample Si/es of Johanseti's likelihood Ratio Tests Tor Cointcsiration." Oxjoril BiiUelin of F.coucmics and Suitisiics 55 ( 1993). 313-28. Ct)rbett. Jentiy. andTitn Jenkinson. The Financing ol Industry. I97()-I9S9: An International Comparison." Discussion Paper no. 948, Lond()n: Centre for Heonomie Poliey Research, 1994. Crowder. William J.. and Dennis L. HolTman. "The Long-Run Relationship hetween Nominal Interest and Inllation: The Fisher Hquation Revisited." ./('((//K;/(;/M(;/(c\', Credit, and Banking 2S (February 19%), IO."^-!X. Davi.s. E. Phihp. Pension Funds. Rclin'mciH-hwoine Sccuiin. and Capikil Markets: An Inlcriialional Perspcdive. Oxford: ClaieiKk)n Press, 199.'i. DeLong. J.Bradtbrd. Andrei Sehleiler, Lawrenec H. .Suminers. and Robert .i. Waldtnann. "The Si/.e atid incidence of ihe Losses from Noise Trading." Jounmi of Finance 44 (Septcinber 1989). 6S I-96. Demetriades, Panicos O.. and Khaled Hussein. "Does Finaiiciul Development Cause Bconomie Growth'.'" Journal of Di'vetopnien! Fconomics 51 (December 1996), 387-41 L Dcmeiriades, PanieosO., and Kill B. Luiniel. "I'inancial Development. Ecciiiomic Growth and Bitnking Sector Controls: Lividcnce from India." Fcimomic Journal 106 (March 1996), 359-74. Demirgui;-KutU. Aslie, and Ross Levine. "St()ck Markets, Corporate Finance, and Econotnic Growth; An Ovet-view." World Hank Fconomic Review 10 (May 1996). 223-39. Demirgu^-Kunt, Aslie. and Vojislav Maksimcn ic. "Stock Market Development and Finaneing Choices of I'irnis." World Bank Fconomic Review 10 (May 1996), 341-69. Dianu)iid, Douglas W. "Financial Intertncdiation and i)elegated Monitoring." Review of Fvonomic Studies 5\ (.luly 1984), 393-414. Diamond, Douglas W., and R. (i. Verreeehia. "Optimal Managerial Contracts and Uquilibriutn Seeurity Vx\cc'^:' Journal of Finance 31 (May 1982). 27.S-S7. Diebold, FYancis X., Javier Gardca/abal, and Kamil Yilma/. "Cointegralion and F^xehange Rate Dynamies." 7f);j;7!(;/(^/7-((;/i( ( 49 (.lunc 1994), 727- 35. Edwards, Jeremy, and Klaus Fischer. Banks. Finance, and Investmen! in Germany. C'anibridgc; C;iinhridge University Press. 1994. Hvans, Paul. "How to Estimate Growth Fquations Consisicntly." Mimeo. Ohio State Uitiversity; presented at the 7th Worid Congress oi the HcoiH>nietrie Society. Tokyo. 1995. . "Government C<insumption and Cirowth." Fconomic Inquiry'}<5 (April 1997), 209 17. Federet, J. Peter. "J\\c Impact of Uncertainty on AggicgLite Investment Spending." Journal of Money. Credit, and Hankini- 25 (February' 1993). 30-45. Fry. Maxwell J. "In Defenee of Financial Libcralisanon." Fconomic Journal 107 (May 1997). 754-70. Cu)ii/.ah), Jesus. "Five Alternative Methods of Estimating Long-Run F-qiiilibriutii Kelaliotihs.'' Journal of Economeirics fyi) (January-February 1994), 203 33. GreenuDod. Jeremy, and Boyan Jovanovic. "Finaneial f>evc!opment. CJrowth, aiid the Distribution of Income." y(;/-/!(;/fVAW/niv;//iVr-zK^jm 98 (October 1990). 1076 -I 107. Greenwood, Jeremy, and Bruce Smith. "Financial Markets in Developmenl and the Development of Financial Markets." Journal of Fconomic Dynamics and Control 21 (January 1997), 145-82. Hall. Stephen G. "Maximum Likelilumd Estimation of ('ointegrating Veckirs; An Example of Johansen's Procedure." Oxford Bullelin of Fconomics and Statistics 51 (March 1989), 2L1-I8.

4(1 :


Hall, Stephen G.. and Alistair Milne. "The RelcvanceofP-Star Analysis to UK Monetary Policy." Economic Journal 104 (May 1994), 597-604. Hall. Stephen, and Mike Wickcns. "Causality in Integrated Systems." London School Discussion Papers 27-93 (November 1993). Hansen, Hcnrik, and Sorcn Johansen. "Recursive Estimation in Cointegrated VAR Models." Preprint I. Insuutte of Mathematical Statistics, University of Copenhagen, 1993. . "Some Tests of Parameter Constancy in Cointegrated VAR Models." Institute ofMalhemalical Statistics. University of Copenhagen, 1998. Hansen, Henrik. and Katarina Juselius. Manual of Cointegration Analysis of Time Series CATS in RATS. Institute of Economics: University of Copenhagen, 1994. Hargreaves. Colin. "A Review of Methods of Estimating Cointegrating Relationships." In Non.statiomirv Time Series Analysis and Cointegration, edited by Colin Hargreaves, pp. 87-131. New York: Oxford University Press, 1994. Harris, Richard D.F. '"Stock Markets and Development: A Reassessment." European Economic Review 4\ (1997), 139-46. HiiLig Alfred A. "Tests for Cointegration: A Montecarlo Comparision." Journal of Econometricsl\ (March 1996), 89-115. Horiuchi, Akiyoshi, and Ryoko Okazaki. "Capital Markets and the Banking Sector: Efficiency of Japanese Banks in Reducing Agency Costs." In Japan. Europe and International Financial Markets: Analytical and Empirical Perspectives, edited by Ryuzo Sato, Richard M. Levich, and Rama V Ramachandran, pp. 197-219. Cambridge: Cambridge University Press, 1994. Jensen, Michael C , and Kevin J. Murphy. "Performance Pay and Top Managemeni Incentives." Journal of Political Economy 98 (April 1990), 225-64. Johansen, Soren. "Statistical Analysis of Co-Integrating Vectors." Journal of Economic Dynamics and Control 12 (June-September 1988). 231-254. ^___ . "Estimation and Hypothesis Testing of Cointegrating Vectors in Gaussian Vector Autorcgressive Models." cY/(f>mffma 59 (November 1991), 1551-80. . "Determination of Cointergation Rank in the Presence of a Linear Trend." Oxford Bulletin of Economics and Statistics 54 (August 1992), 383-97. Johansen, Soren, and Katarina Juselius. "Some Structural Hypotheses in a Multivariate Cointegration Analysis of Purchasing Power Parity and Uncovered Interest Parity for the UK." Journal of Econometrics 5^ (My-Scptcmber 1992), 211-44. Jones. Charles I. "Time Series Tests of Endogenous Growth Models." Quarterly Journal of Eeonomies 1 10 (May 1995), 495-525. Klenow, Peter J., and Andres Rodriguez-Clare. "Economic Growth: A Review Essay." Journal of Monetary Economics 40 (December 1997), 597-618. Kocherlakota, Narayana R., and Kei-Mu Yi. "Is There Endogenous Long-Run Growth? Eviilenee from the United States and the United Kingdom." Journal of Money, Credit, and Banking 29 (May 1997). 235-60. Keynes, John Maynard. The General Theory of Employment. Interest, and Money. London: Macmillan. 1936. Levine, Ross. "Stock Markets, Growth, and Tax Policy." Journal of Finance 46 (September 1991), 1445-65. "Financial Developmeni and Economic Growth: Views and Agenda." Journal of Economic Literature 35 (1997), 688-726. Levine, Ross, and David Renelt. "A Sensitivity Analysis of Cross-Country Growth Regressions." American Economic ReviewH2 (September 1992). 942-63.

Plill.ll>.'\HKSIIS. P A . M f O S O

niiMHII^IADI.S. A N D K I L Ii l.lilNTI'l


Levine, Ross, and Sara Zcrvos, "Capiliil Control l.ibcrali/alion Lind Slock Markcl Devckipmenl." Policy Research Working Paper 1622. Washinglon D.C: World Bank. 1995. . "Stock Markels, Banks, and Economic Growth.'" .American luonomic Review 88 (June 1998), 5.37-5K. Luintcl. Kul B., and Mosahid Khan. "A Quantitative Reassessment ot Financc-Cirowth Nexus: Evidence from Muilivariaic VAR." Journal of Development Economics 60 (1999), 381-405. Luintel, Kul B.. and Krishna Paudyal, "Cnmmoii Suxhastic Trends bclwccn Forward and Spot Exchange Rales." Journal of International Money and Finance 17 11998). 279 97, Mayer, Colin. "New Issues in Corporate Miuince."" European Famomic Review 32 (June 1988). 1 167-88. Oslerwald-Lenum. Michael. "A Note with Quartiles ol ihe Asymptotic Distribution of the ML Cointegration Rank Test Statistics." O.xfoni Bulletin of Fconomics and Statistic s 54 (August 1992), 461-72. Pesaran, M. Hashcin, and Yongchcol Shin. '"Long-Run Structural Modelling," University of Camhridge: Department of Applied Economics. Working Paper no. 9419. September 1994, Pesaran, M, Hashcm, and R. Smith, "Fslimating Long-Run Relationships frtim Dynamic Hclerogeneous Panels." Journal ofEconomclrics 68 (1995). 79 113. Quah, Danny. "Empirical Cross-Section Dynamics in EcoiK)mic Cli*t)wth." European Economic Review 37 (April 1993), 426-34, Quintos H. Carnicla. "Sustainability oi' the Dclieit Process with Siructural Shifts." Journal of Business and Economic Statistics 13 (1995), 409-16. Rajan, Raghuram G.. and Luigi Zingalcs. "What Do We Know aboul Capital Structure? SIMIIC Evidence From International Data." Journal of Finance 50 (1995). 1421-60. Rajan, Raghuratn G., and Luigi Zingalcs. ""Financiai Dependence and (Jrowlh." Mimeo. University of Chicago, 1996, Schwar/, Gideon. "Estimatinij the [Jimension of a Model." Annals of Statistics 6 (1978), 461-64. Shillcr, Robert J, "Do Stock Prices Move Too Much to Be Justilicd by Subsequent Changes in Dividends?"/i/jzcr/rc/n F.conomic Review 71 (1981). 421 36. . Market W)/f(r///n\ Cambridge. Mass.: MIT Press, 1989. Singh. Ajit. "Stock Markets, Financial Liberalisation and Economic Dc\clopnicnt." Economic Journal t()7 (May 1997). 771-82. Stiglit/, Joseph E, "Credit Markets and theConlrol ot Capital." Journal of Monev, Credil. and Banking- 17 (February 1985), 133-52, Toda, Hiro Y., and Peter C.B- Phillips. "Vector Auiorcyrcssions and C'ausalit>." Fcononietrica 61 (November 1993). 1367-9.3. Williamson, Stephen D. "Costly Monitoring. Financial Intermediation, and Equilihrium Credit Rationing." Journal of Monetary Economies 18 (Scptcnihcr I9S6), 159 79,