This article was downloaded by: [Florida Atlantic University]On: 24 September 2013, At: 05:18Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number:1072954 Registered office: Mortimer House, 37-41 Mortimer Street,London W1T 3JH, UK

The Journal of DevelopmentStudiesPublication details, including instructions forauthors and subscription information:http://www.tandfonline.com/loi/fjds20

Financial developmentand economic growth:Additional evidenceRati Ram aa Department of Economics, Illinois StateUniversity, Campus Box 4200, Normal, IL,61790–4200, USAPublished online: 23 Nov 2007.

To cite this article: Rati Ram (1999) Financial development and economicgrowth: Additional evidence, The Journal of Development Studies, 35:4,164-174, DOI: 10.1080/00220389908422585

To link to this article: http://dx.doi.org/10.1080/00220389908422585

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of allthe information (the “Content”) contained in the publications on ourplatform. However, Taylor & Francis, our agents, and our licensorsmake no representations or warranties whatsoever as to the accuracy,completeness, or suitability for any purpose of the Content. Anyopinions and views expressed in this publication are the opinions andviews of the authors, and are not the views of or endorsed by Taylor& Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information.Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities

whatsoever or howsoever caused arising directly or indirectly inconnection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private studypurposes. Any substantial or systematic reproduction, redistribution,reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of accessand use can be found at http://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

Flor

ida

Atla

ntic

Uni

vers

ity]

at 0

5:18

24

Sept

embe

r 20

13

Notes

Financial Development and EconomicGrowth: Additional Evidence

RATI RAM

This note suggests that, contrary to the conclusions reached inseveral recent studies, the empirical evidence does not support theview that financial development promotes economic growth. It isfirst noted that the predominant pattern in the data for 95individual countries is that of a negligible or weakly negativecovariation between financial development and growth of realGDP per capita. Second, the individual-country correlationalpicture is a sharp contrast to the correlations based on cross-country data that have been used in most research on the subject.Third, individual-country estimates of a basic multiple-regressiongrowth model also do not indicate a positive association betweenfinancial development and growth. Fourth, in cross-country dataand models of the kind that have been used in most studies, whenthe regression structure is permitted to vary across threesubgroups, a huge parametric heterogeneity is observed, and theoverall indication is that of a negligible or negative associationbetween financial development and growth.

INTRODUCTION

Besides the earlier research by Goldsmith [1969], McKinnon [1973] andother scholars, several recent studies have indicated that the empiricalevidence suggests a substantial positive effect of financial development oneconomic growth.1 For example, King and Levine [1993: 734-35]concluded that 'indicators of the level of financial development ... are

Rati Ram, Department of Economics, Illinois State University, Campus Box 4200, Normal, IL61790-4200, USA. Maryam Arabshahi provided conscientious research assistance. Twoanonymous referees and the Editors of this journal gave helpful suggestions on an earlier version.The usual disclaimer applies.

The Journal of Development Studies, Vol.35, No.4, April 1999, pp.164-174PUBLISHED BY FRANK CASS, LONDON

Dow

nloa

ded

by [

Flor

ida

Atla

ntic

Uni

vers

ity]

at 0

5:18

24

Sept

embe

r 20

13

FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 165

strongly and robustly correlated with growth ... and ... the predeterminedcomponents of these ... indicators significantly predict subsequent values ofthe growth indicators'. Similarly, Odedokun [1996: 119, 135] summarisedhis work by saying that financial intermediation promotes economic growthin most countries. A little more recently, based on a fairly extensive reviewof the literature, Levine [1997: 688-89] stated that 'the preponderance oftheoretical reasoning and empirical evidence suggests a positive, first-orderrelationship between financial development and economic growth' and thata 'growing body of work would push even most skeptics toward the beliefthat the development of financial markets and institutions is a critical andinextricable part of the growth process'.2

This note suggests that the empirical results reported in such studies areat best uncertain and ambiguous, and that the predominance of evidenceindicates the lack of a significant positive association between financialdevelopment and economic growth. Four points are noted to support thisview. First, the predominant pattern in the data for 95 individual countriesis that of a negligible or weakly negative covariation between financialdevelopment and growth. Second, the individual-country correlationalpattern is a sharp contrast to the large, positive and significant correlationbetween the same variables in intercountry data on which most research inthe area has relied. Third, multiple regression estimates of simple growthequations from individual-country data indicate the same pattern as thebivariate correlations. Fourth, in multiple regression estimates from thetypical cross-country averaged data, when the parametric structure ispermitted to vary across three subgroups, a huge structural heterogeneity isobserved, and the indication again is that of a negligible or weakly negativeparameter for the financial-development variable in most cases.

INDIVIDUAL-COUNTRY EVIDENCE

While it is common to consider cross-country regression estimates to judgethe growth effects of such factors, the implications obviously relate to eachcountry, and it is important to study the individual-country evidence at leastat a simple level. Since the paucity of observations for most countries makesit difficult to estimate a structural model for each country, a good startingpoint is to look at the covariation between financial development andeconomic growth in each country. Out of several indicators of financialdevelopment, DEPTH, which is the ratio of 'liquid liabilities' to GDP,seems most appropriate since it has been used widely as a prime indicatorof financial development and since data for it are relatively more plentiful.Therefore, the coefficient of correlation between DEPTH and growth of realGDP per capita is calculated from annual observations for each country. To

Dow

nloa

ded

by [

Flor

ida

Atla

ntic

Uni

vers

ity]

at 0

5:18

24

Sept

embe

r 20

13

166 THE JOURNAL OF DEVELOPMENT STUDIES

keep the results comparable with the previous research, the period 1960-1989 is studied, and all data are taken from International FinancialStatistics Yearbook 1990. Every country for which at least ten observationsover the period are available has been included. The resulting sampleconsists of 95 countries. In almost all countries, there are substantialvariations in the values of the variables over the sample period.

Table 1 contains the basic information for each country. Out of the entireset of 95 countries, the correlations are positive in 39 cases, of which nineshow statistical significance at the conventional five per cent level.3 Thecorrelations are negative in the remaining 56 cases, and 16 of these aresignificant at the five per cent level.4 The mean of the 95 correlationcoefficients is —0.06. The preponderance of evidence thus suggests anegligible or weakly negative association between economic growth and aprime proxy for financial development.5

Two additional points about Table 1 might be mentioned. First, as thenote in that Table explains, DEPTH is used with a one-year lag, and is thustreated as the "prior" variable in a temporal sense. Second, to see how farthe correlations might have been affected by cyclical factors, five-yearaverages of DEPTH and growth of GDP per capita are examined for 39countries which had at least 25 data points. The averages are taken for theperiods 1960-64, 1964-68, 1968-72, 1972-76, 1976-80, 1980-84 and1984-88.6 Despite the limited number of observations in the averaged data,correlations based on five-year averages of the variables show a patternwhich is remarkably similar to that indicated by the annual data. Correlationbetween the 39 coefficients of correlation based on annual and averageddata is 0.92, and the means of the correlation coefficients in the two sets are0.01 and -0.04 respectively. Moreover, the inferences yielded by the twosets of correlations are virtually identical in almost all cases.7

The individual-country pattern in Table 1 presents a sharp contrast to thecross-country correlation between the same variables. Taking the averagevalue of DEPTH for each country over the period 1960-89 and the averageannual rate of growth of GDP per capita during the same period, thecorrelation between the two variables across the 95 countries is noted to be0.33, which is a sizeable positive number, is statistically significant at almostany conventional level, and thus presents a stark contrast from the mean of-0.06 for the individual-country correlations.8 The cross-country correlationin this sample of 95 countries is fairly similar to that reported by King andLevine [1993: 723] for their sample of about 80 countries. Given thepredominant individual-country scenario, it seems likely that the sizable,positive and highly significant cross-country correlation is an artifact andprobably reflects something other than the 'effect' of DEPTH on growth.

While the limited number of observations for individual countries does

Dow

nloa

ded

by [

Flor

ida

Atla

ntic

Uni

vers

ity]

at 0

5:18

24

Sept

embe

r 20

13

FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 167

TABLE 1COEFFICIENT OF CORRELATION BETWEEN RATIO OF LIQUID LIABILITIES

TO GDP AND RATE OF GROWTH OF REAL GDP PER CAPITA:INDIVIDUAL-COUNTRY DATA

1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

15.

16.

Country(period)

Algeria(1966-82)

Australia(1960-89)

Austria(1960-89)

Bangladesh(1974-88)

Barbados(1966-86)

Belgium(1960-88)

Benin(1970-85)

Bolivia(1964-88)

Botswana(1976-89)

B. Faso(1972-85)

Burundi(1970-88)

Cameroon(1969-85)

Canada(1966-89)

Chile(1972-85)

Colombia(1960-84)

Costa Rica(1968-89)

Com(p-value)

-.002(.996)

.471(.011)

-.375(.049)

-.293(.330)

.564(.012)

.632(.000)

.152(.574)

-.283(.191)

-.039(.905)

-.023(.940)

-.019(.940)

.691(.003)

-.296(.180)

-.190(.554)

-.460(.047)

-.588(.006)

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

Country(period)

Cyprus(1963-88)

Denmark(1960-87)

Dom. Rep.(1960-88)

Ecuador(1960-88)

El Salvador(1960-89)

Fiji(1966-88)

Finland(1960-87)

France(1960-89)

Gambia(1970-86)

Germany(1960-89)

Ghana(1960-88)

Greece(1960-87)

Grenada(1975-87)

Guatemala(1960-89)

Guyana(1967-87)

Haiti(1960-87)

Corr.(p-value)

.094(.661)

.158(.441)

-.079(.695)

.375(.054)

-.379(.046)

-.164(.467)

.141(.492)

-.692(.000)

.664(.005)

-.092(.641)

-.000(.998)

-.569(.002)

.069(.831)

-.444(.018)

-.305(.205)

-.107(.604)

33.

34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

Country(period)

Honduras(1960-89)

Iceland(1960-88)

India(1960-88)

Indonesia(1965-88)

Iran(1962-85)

Ireland(1971-88)

Israel(1976-88)

Italy(1960-88)

Jamaica(1970-88)

Japan(1970-88)

Jordan(1969-88)

Kenya(1973-88)

Korea (So.)(1960-89)

Kuwait(1976-88)

Liberia(1974-86)

Libya(1960-80)

Corr.(p-value)

-.190(.332)

.224(.261)

.258(.193)

-.163(.469)

-.078(.745)

-.199(.460)

.368(.239)

.098(.628)

.161(.537)

.073(.782)

.079(.748)

-.573(.032)

.296(.127)

.350(.291)

-.010(.976)

-.385(.103)

contd.

Dow

nloa

ded

by [

Flor

ida

Atla

ntic

Uni

vers

ity]

at 0

5:18

24

Sept

embe

r 20

13

168 THE JOURNAL OF DEVELOPMENT STUDIES

TABLE 1 (cont.)

49.

50.

51.

52.

53.

54.

55.

56.

57.

58.

59.

60.

61.

62.

63.

64.

Country(period)

Madagascar(1962-85)

Malawi(1964-87)

Malaysia(1970-88)

Malta(1960-88)

Mauritius(1960-87)

Morocco(1969-88)

Myanmar(1960-88)

Nepal(1960-88)

Netherlands(1960-89)

New Zealand(1965-86)

Nicaragua(1960-83)

Niger(1962-81)

Nigeria(1960-88)

Norway(1960-89)

Oman(1972-87)

Pakistan(1960-89)

Com(p-value)

-.435(.105)

-.200(.372)

-.331(.211)

.452(.018)

.556(.003)

.042(.870)

-.373(.056)

.337(.086)

-.295(.127)

.084(.724)

-.230(.303)

.275(.269)

-.228(.253)

-.139(.479)

-.447(.109)

.297(.125)

65.

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

76.

77.

78.

79.

80.

Country(period)

Papua N.G.(1973-88)

Paraguay(1960-88)

Peru(1960-88)

Philippines(1960-88)

Portugal(1960-87)

Rwanda(1964-87)

Saudi Arabia(1968-87)

Senegal(1967-86)

S. Leone(1964-87)

Singapore(1968-89)

So. Africa(1960-88)

Spain(1960-89)

Sri Lanka(1960-88)

Sweden(1964-85)

Switzerland(1960-88)

Syria(1963-87)

Corr.(p-value)

.013(.964)

.241(.226)

.130(.519)

-.439(.022)

-.451(.021)

-.390(.072)

-.634(.004)

.325(.256)

-.339(.114)

-.216(.359)

.293(.139)

-.560(.002)

-.026(.898)

.444(.044)

.134(.504)

-.355(.096)

81.

82.

83.

84.

85.

86.

87.

88.

89.

90.

91.

92.

93.

94.

95.

Country(period)

Tanzania(1966-88)

Thailand(1961-88)

Togo(1970-86)

T. & Tobago(1963-85)

Tunisia(1960-86)

Turkey(1970-88)

Uganda(1970-86)

U.A. Emir.(1973-88)

U.K.(1960-88)

U.S.A.(1960-89)

Uruguay(1968-88)

Venezuela(1963-88)

Yemen(1973-87)

Zaire(1963-87)

Zambia(1965-87)

Corr.(p-value)

-.720(.000)

.120(.560)

-.100(.711)

-.597(.003)

-.159(.449)

.410(.102)

.168(.533)

-.243(.402)

.229(.251)

.516(.005)

-.368(.111)

-.444(.030)

-.729(.005)

-.328(.117)

-.301(.210)

cont.

Dow

nloa

ded

by [

Flor

ida

Atla

ntic

Uni

vers

ity]

at 0

5:18

24

Sept

embe

r 20

13

FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 169

TABLE 1 (cont.)

Note: The correlation coefficients are computed by using Statistical Analysis System (SAS). Thep-values are also taken from the SAS printout, and both have been rounded off to the thirddecimal place. The underlying data are obtained from International Financial Statistics Yearbook1990 (IFS). The ratio of liquid liabilities to GDP is calculated by following the procedureindicated by King and Levine [1993: 720] and Levine [1997]. For example, the number for 1961is the ratio of the mean value of liquid liabilities (line 551 of IFS) for 1960 and 1961 to GDP (line99b of IFS) for 1961. This ratio is correlated with growth of GDP per capita from 1961 to 1962,the expression for which is ln(RGDP/POP)1962-ln(RGDP/POP)196I, where RGDP denotes realGDP (line 99b.r of IFS), POP stands for population (line 99z of IFS), the subscript indicates theyear, and In stands for natural logarithm. In cases where adequate data on liquid liabilities werenot available, line 351 of IFS (money plus quasi-money) has been used, as done by King andLevine [1993]. Out of the 95 countries listed, correlation is positive in 39 cases, of which nineshow statistical significance at the 5% level. The remaining 56 correlations are negative, and 16of these are significant at the 5% level. The simple mean of the 95 correlation coefficients is-0.064.

not permit estimation of a full-scale structural growth equation for eachcountry, it is possible to get some flavor for the DEPTH parameter fromindividual-country regressions of simple growth models. For that purpose,a slight modification of the growth model specified in equation (4) ofOdedokun [1996: 123] appears to be a reasonable candidate. Thespecification may be written as

GY = ao + ajCGL) + a2(I/Y) + a3(GX) + a4(DEPTH) (1)

where GY, GL and GX denote annual rates of growth of real GDP,population (as a proxy for the labor force) and exports respectively, DEPTHis (as already noted) the ratio of current-price liquid-liabilities to GDP, andI/Y is the ratio of gross domestic investment to GDP. Except for the additionof DEPTH, this is a fairly standard model for growth of GDP, and is thesame as that used by Odedokun [1996] after replacing by DEPTH hisfinancial-development variable which is likely to be subject to a severeendogeneity problem.9 To keep the project tractable, out of the 71 lessdeveloped countries (LDCs) studied by Odedokun [1996], a random sampleof eight countries is used by taking every 10th country starting with thefirst.10 Table 2 contains abridged estimates of equation (1) for each of theseeight countries and reports the DEPTH parameter along with thecorresponding t-statistic. It is evident that the overall picture here is verysimilar to that in Table 1, and most estimates of the DEPTH parameter arenegative and statistically insignificant at the conventional levels."

Dow

nloa

ded

by [

Flor

ida

Atla

ntic

Uni

vers

ity]

at 0

5:18

24

Sept

embe

r 20

13

170 THE JOURNAL OF DEVELOPMENT STUDIES

TABLE 2ESTIMATES OF DEPTH COEFFICIENT FROM ANNUAL DATA FOR SELECTEDCOUNTRIES ON THE BASIS OF MULTIPLE REGRESSION OF GDP-GROWTH

SPECIFIED IN EQUATION (1)

Country (period) Coefficient of DEPTH(t-statistic)

Algeria (1966-1982) -0.181 (-0.63)Greece (1960-1987) -0.148 (-3.88)India (1960-1986) -O.070 (-0.24)Jamaica (1970-1988) +0.069 (+0.50)Morocco (1969-1988) -0.331 (-1.85)Portugal (1960-1987) -0.098 (-2.25)Thailand (1962-1988) -O.024 (-0.46)Zambia (1965-1984) -0.128 (-0.67)

Note: This is a random 10% sample from the 71 countries listed by .Odedokun [1996: 136-45].All data are based on International Financial Statistics Yearbook 1990. Ordinary least-squares(OLS) estimates of equation (1) of the text are reported since the first-order autoregressiveparameter is not statistically significant at any meaningful level for any country except India.Moreover, the feasible generalised least-squares estimates based on a first-order autoregressiveerror indicate the same picture as the OLS estimates for all countries, including India. Parameterestimates for other regressors are not reported so as to focus on the DEPTH variable, and areavailable from the author.

THE C R O S S - C O U N T R Y S C E N A R I O : AN EXPLORATION OF

PARAMETRIC H E T E R O G E N E I T Y

The extensive individual-country evidence in Tables 1 and 2 seems toprovide a strong refutation of the view that there is a positive effect offinancial development on growth or that there is a positive associationbetween the two variables. However, most empirical research on the topicis based on cross-country data that contain averaged observations forsizable periods, and strong conclusions are drawn from such cross-sectionestimates about the favorable effect of financial development on growth.Moreover, following the tradition in the many recent empirical studies ofeconomic growth, such constant-parameter cross-section estimates are usedto make predictions about the growth rate if there is a change in thefinancial-development variable in a country or a subset of the countries, andthere is not even a mention of the obvious question about the structuralhomogeneity across the sample countries. For instance, Levine [1997: 706]observed that the coefficient of 0.024 in his Table 2 implies that a countrythat 'increased DEPTH from the mean of the slowest growing quartile ofcountries (0.2) to the mean of the fastest growing quartile of countries (0.6)would have increased its per capita growth rate by almost one percent peryear'. However, before an inference of that kind can be reasonably drawn,

Dow

nloa

ded

by [

Flor

ida

Atla

ntic

Uni

vers

ity]

at 0

5:18

24

Sept

embe

r 20

13

FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 171

it seems appropriate to ask whether the coefficient of DEPTH is the samefor the fastest-growing and the slowest-growing quartiles. If, as one mayreasonably believe and as Table 1 indicates rather dramatically, theassociation between DEPTH and growth shows considerable variationacross countries, the implications could be quite different.

Although structural homogeneity is assumed in almost all empiricalgrowth research based on cross-section observations, its validity can beexplored in several ways. This is illustrated by taking the dataset compiledby Levine and Renelt [1992], adding the DEPTH variable, and estimatingthe multiple regression models used by King and Levine [1993: 727] andLevine [1997: 706] for the entire sample and for three subsamples of high-growth, mid-growth and low-growth countries.12 Although they used severalindicators of financial development, since all of them yielded very similarresults, the illustration is done with DEPTH which, as noted before, hasbeen widely used as a prime proxy for financial development. The five otherexplanatory variables are the same as those noted by King and Levine[1993: 727] and Levine [1997: 706].

Table 3 reports abridged estimates of the King-Levine [1993] andLevine [1997] regressions and contains the DEPTH coefficient and the

TABLE 3ESTIMATES OF THE COEFFICIENT OF DEPTH IN CROSS-COUNTRYREGRESSIONS FOR ENTIRE SAMPLE AND THREE SUBSAMPLES:

AVERAGED DATA FOR 1960-89

Country group

Full sample

Low-growth subsample

Mid-growth subsample

High-growth subsample

DEPTH coefficient (t-statistic)

6.027* (3.38)

-0.003 (-0.11)

-0.008 (-1.70)

0.018* (2.20)

R2

0.41

0.44

0.35

0.35

N

83

25

29

29

Note: The basic dataset compiled by Levine and Renelt [1992] is used. Since they do not provideinformation on DEPTH, the average value of the variable for 1960-89 has been computed fromthe data underlying Table 1. The resulting sample consists of 83 countries. As explained in note12, the low-growth group consists of countries whose annual growth rate of GDP per capitaduring 1960-89 (as reported by Levine and Renelt) is at or below 1.27%; the growth rate for themid-growth group lies between 1.28% and 2.79%, and the high-growth group includes countrieswhose growth rate equals or exceeds 2.80%. Note that other five regressors in the models are thesame as in King and Levine [1993] and Levine [1997], namely, log of initial income, log of initialsecondary school enrolment rate, ratio of government consumption expenditure to GDP, inflationrate, and ratio of exports plus imports to GDP. Full regression estimates are available from theauthor. An asterisk (*) indicates significance at least at the 5% level.

Dow

nloa

ded

by [

Flor

ida

Atla

ntic

Uni

vers

ity]

at 0

5:18

24

Sept

embe

r 20

13

172 THE JOURNAL OF DEVELOPMENT STUDIES

related t-statistic for each group. It reveals vividly the huge parametricheterogeneity across the sample countries and the illegitimacy of statementsbeing made about the subgroups on the basis of the full-sample estimates.In particular, there is nothing to suggest that the low-growth or the medium-growth countries are likely to experience a higher growth rate if DEPTH isincreased. It is only for the high-growth group that one might be able to saythat there is a positive association between financial development andgrowth.13 Thus even for cross-country data, the limited parametric flexibilitypermitted in Table 3 leads to a set of estimates that show lack of significantpositive association between financial development and economic growthin most cases, and the evidence broadly supports the individual-countryscenario reflected in Table 1 and Table 2.

It is also possible to conduct formal tests of structural homogeneity tosee if the 'pooling' of all countries implied by the regression estimatesbased on the entire sample is legitimate. The simple two-group test ofstructural homogeneity described by Maddala [1992: 171] for the low-growth and the high-growth subgroups yields an F-statistic of 15.94 whilethe critical one per cent F(7, 40) is 3.12. It is, therefore, obvious that thefull-sample estimates are an illegitimate basis for proper conclusions aboutthe parameters, especially if these are related to subsamples based ongrowth rates. The application of the full-sample estimates to makestatements about an individual country appears even more illegitimate.14

CONCLUDING REMARKS

The main theme of this note is to suggest that the preponderance ofempirical evidence does not encourage one to share the view that financialdevelopment is observed to have a positive effect on economic growth. Fourpoints are made to support this conclusion. First, data for 95 individualcountries suggest that the predominant correlation between financialdevelopment and economic growth is negligible or weakly negative.Second, the average individual-country correlation presents a sharp contrastto the cross-country correlation between the same variables, and indicatesthat the cross-country estimates, which have been used in most studies,might be spurious. Third, individual-country estimates of a basic multiple-regression growth model also indicate a picture that is consistent with thebivariate correlations. Fourth, therefore, positive parameter estimates for thefinancial-development variables reported in cross-country studies are not areliable basis for parametric inference, especially when statements are madeabout individual countries or sample subgroups. This point is furtherillustrated by reporting multiple regression estimates of the kind used byKing and Levine [1993] and Levine [1997] for the entire cross-country

Dow

nloa

ded

by [

Flor

ida

Atla

ntic

Uni

vers

ity]

at 0

5:18

24

Sept

embe

r 20

13

FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 173

sample and for three subsamples. The estimates reveal a dramatic structuralheterogeneity across the subsamples and support the individual-countryscenario of there being no significant association between financialdevelopment and economic growth in most cases. Moreover, a formal two-group test strongly rejects the null hypothesis of structural homogeneity andshows that it is illegitimate to consider constant-parameter estimates fromthe full sample.

Following the foregoing points, it is suggested that the future researchon this important topic may include a greater focus on individual-countrystudies, and an exploration of parametric heterogeneity across the samplesubgroups in cross-country data, especially if a quantitative interpretation isplaced on the parameter estimates.15 It is perhaps obvious that thesesuggestions apply not merely to the research on the nexus between financialdevelopment and economic growth, but also to most of the numerous otherrecent empirical studies of economic growth.

final version received May 1998

NOTES

1. A comprehensive discussion of the many dimensions of the subject is provided by Fry[1995].

2. Lucas [1988: 6], however, had noted 'In general, I believe that the importance of financialmatters is very badly over-stressed in popular and even much professional discussion (ofeconomic development)'.

3. The correlations are computed by using Statistical Analysis System (SAS), and the p-valuesare taken from the SAS printout.

4. Despite the dramatic variation in the correlations across the sample countries, it is difficultto discern any clear patterns in terms of country characteristics.

5. Since most correlations lack statistical significance, there should be no question of a'spurious' correlation in these time-series observations.

6. For DEPTH, a simple average for each five-year period is taken. Rate of growth of GDP percapita for each period is estimated through an exponential regression on 'time'. Additionaldetails may be requested from the author.

7. The 39 countries and the correlation coefficients from averaged data for these are: Australia(0.853), Austria (-0.743), Belgium (0.812), Denmark (0.419), Dominican Republic (-0.389),Ecuador (0.395), El Salvador (-0.785), Finland (-0.068), France (-0.877), Germany(-0.411), Ghana (-0.071), Greece (-0.865), Guatemala (-0.636), Haiti (-0.344), Honduras(-0.302), Iceland (0.411), India (0.726), Italy (-0.247), (South) Korea (0.837), Malta (0.433),Mauritius (0.423), Myanmar (-0.729), Nepal (0.657), Netherlands (-0.754), Nigeria(-0.406), Norway (-0.174), Pakistan (0.556), Paraguay (0.250), Peru (0.032), Philippines(-0.747), Portugal (-0.779), South Africa (0.773), Spain (-0.732), Sri Lanka (0.105),Switzerland (-0.070), Thailand (-0.026), Tunisia (-0.504), UK (0.872), and USA (0.430).SAS is used for all means, correlations, and regressions.

8. The annual rate of growth of real GDP per capita is estimated from a regression of thelogarithm of real GDP per capita on 'time'.

9. Since the rate of growth of population is on the right-hand side, this is almost equivalent toestimating an equation for growth of real GDP per capita.

Dow

nloa

ded

by [

Flor

ida

Atla

ntic

Uni

vers

ity]

at 0

5:18

24

Sept

embe

r 20

13

174 THE JOURNAL OF DEVELOPMENT STUDIES

10. Since information for Congo was incomplete, that country was replaced by India.11. Complete estimates are available from the author.12. Of the 95 countries in Table 1, Burkina Faso, Grenada, Guyana, Kuwait, Libya, Myanmar,

Nepal, Oman, Singapore, Uganda, United Arab Emirates and Yemen are not included incross-country regressions because of complete information not being available in the Levineand Renelt [1992] dataset. The subsamples are defined so as to generate an almost equalnumber of observations in the three subgroups. The low-growth group consists of countrieswhose annual growth during 1960-89 (as reported by King and Levine) is at or below 1.27per cent; the growth rate for the mid-growth group lies between 1.28 per cent and 2.79 percent; and the high-growth group includes countries whose growth rate equals or exceeds 2.80per cent.

13. Some other issues, particularly that of the potential endogeneity of the DEPTH variable,discussed by Levine [1997] and other scholars, are overlooked here so as to focus on theimportant aspect concerning structural homogeneity.

14. One example of such statements is the observation by Levine [1997: 70S] about Bolivia,which country belongs to the low-growth sub-group.

15. As noted in the text, ratio of liquid-liabilities to GDP appears to have been a prime proxy forfinancial development and has been used extensively. Along with the two aspects suggestedin the text, consideration of other measures of financial development is also a good area forfuture research.

REFERENCES

Fry, M.J., 1995, Money, Interest, and Banking in Economic Development, Second Edition,Baltimore, MD: Johns Hopkins University Press.

Goldsmith, R.W., 1969, Financial Structure and Development, New Haven, CT: Yale UniversityPress.

International Monetary Fund, 1990, International Financial Statistics Yearbook 1990,Washington, DC.

King, R.G. and R. Levine, 1993, 'Finance and Growth: Schumpeter Might Be Right', QuarterlyJournal of Economics, Vol.108, No.3, pp.717-37.

Levine, R., 1997, 'Financial Development and Economic Growth: Views and Agenda', Journalof Economic Literature, Vol.35, No.2, pp.688-726.

Levine, R. and D. Renelt, 1992, 'A Sensitivity Analysis of Cross-Country Growth Regressions',American Economic Review, Vol.82, No.4, pp.942-63.

Lucas, R.E., Jr., 1988, 'On the Mechanics of Economic Development', Journal of MonetaryEconomics, Vol.22, No.1, pp.3-42.

Maddala, G.S., 1992, Introduction to Econometrics, Second Edition, New York, NY: MacmillanPublishing Company.

McKinnon, R.I., 1973, Money and Capital in Economic Development, Washington, DC:Brookings Institution.

Odedokun, M.O., 1996, 'Alternative Econometric Approaches for Analysing the Role of theFinancial Sector in Economic Growth: Time-Series Evidence from LDCs', Journal ofDevelopment Economics, Vol.50, No.1, pp. 119-46.

Dow

nloa

ded

by [

Flor

ida

Atla

ntic

Uni

vers

ity]

at 0

5:18

24

Sept

embe

r 20

13