You are on page 1of 14

Research in International Business and Finance xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Research in International Business and Finance
journal homepage: www.elsevier.com/locate/ribaf

Stock markets, banks, and economic growth: Evidence from more
homogeneous panels☆

Tolina Fufaa, Jaebeom Kimb,
a
Department of Economics, Northwest Missouri State University, Maryville, MO 64468, USA
b
Department of Economics, Oklahoma State University, Stillwater, OK 74078, USA

AR TI CLE I NF O AB S T R A CT

JEL classification: The present paper investigates whether the link between stock markets, banks, and economic
G10 growth becomes more evident as more homogeneous groups of countries are considered. The
G21 dynamic panel generalized method of moment (GMM) estimator with Windmeijer (2005) cor-
O16 rection is employed using data of European and non-European high-income countries as well as
O40
upper and lower middle-income countries averaged over five and three years. Our results indicate
Keywords: that the link between financial development and economic growth depends on the stages of
Economic growth economic growth of the countries. As more homogeneous economies are involved in a panel, a
Stock market development
more economically stylized link is uncovered.
Financial development
Homogeneous panel
GMM

1. Introduction

The present paper investigates whether there exists a link between financial development and economic growth by considering
the financial sectors across stock markets and banks as more homogeneous groups of countries are involved. Since the pioneering
works of Schumpeter (1934) and, more recently, Goldsmith (1969), McKinnon (1973), and Shaw (1973), there has been an intense
debate regarding the link between financial sector development and economic growth. Some early studies, including Robinson
(1952) and Lucas (1988), contend that the financial sector develops merely in response to economic growth or has no significant
contribution to growth. However, a body of recent literature, including King and Levine (1993), Levine (1997), Bekaert et al. (2005)
and Bertocco (2008), emphasizes that a well-functioning financial system improves the allocation of resources and hence promotes
economic growth by mitigating the effects of information asymmetry and transaction costs.1
The theory presents somewhat conflicting explanations about whether stock markets and banks have independent roles in eco-
nomic growth and whether the two have any comparative importance in economic activity. On the one hand, Holmstrom and Tirole
(1993), Boyd and Smith (1998), and Allen and Gale (1999), among others, argue that well-functioning stock markets are better at
reducing information and transition costs and thus fostering economic growth. On the other hand, a number of studies, including
Boot and Thakor (1997) and Coval and Thakor (2005), argue that banks are relatively better at reducing market frictions associated


We thank Lee Adkins, Wenyi Shen, David Carter, Dongwook Lee, Donggyu Sul and seminar participants at the MEA meeting, SEA meeting, World Finance
Conference at New York, KEA-KAEA International Economics Conference, AMES at Kyoto, Japan, CAFM international conference for helpful comments and discus-
sions. All remaining errors are our own.

Corresponding author.
E-mail address: jb.kim@okstate.edu (J. Kim).
1
See also Diamond (1984), Stiglitz (1985), Greenwood and Jovanovic (1990), Bencivenga and Smith (1991), Bhide (1993), Bencivenga et al. (1995), Allen and Gale
(1997), and Khan (2001). For a thorough review of the literature refer to Levine (1997, 2005) and Beck (2013).

http://dx.doi.org/10.1016/j.ribaf.2017.07.120
Received 30 January 2017; Accepted 5 July 2017
0275-5319/ © 2017 Elsevier B.V. All rights reserved.

Please cite this article as: Fufa, T., Research in International Business and Finance (2017),
http://dx.doi.org/10.1016/j.ribaf.2017.07.120

As can be seen in many recent empirical studies. The issue of fundamental and empirical importance in this study is the extent to which the role of financial intermediaries and markets in economic growth across countries can be measured in terms of a degree of homogeneity dealing with the endogeneity problem. find evidence of differential effects of financial depth on growth. see Zingales (2015). and Demirguc- Kunt et al. Kim Research in International Business and Finance xxx (xxxx) xxx–xxx with the mobilization and allocation of resources towards more productive activities. on the other hand. Banks. existing panel studies fail to settle two important issues: heterogeneity of the countries in a panel. however. nurture long-term relationships with investors and hence provide a stable source of finance for achieving long-term economic growth and industrialization. comparisons are made between the dynamic panel system and difference GMM estimators. we employ a dynamic panel generalized method of moments (GMM) estimator by Blundell and Bond (1998) with Windmeijer (2005) correction to the data from 64 countries. serious income inequality. they did not employ any corrections of downward bias for small sample. the results from panel regressions based on pooled heterogeneous cross-country observations may have limited policy relevance. However. including both stock markets and banks. 4 According to Zingales (2015). see Deidda and Fattouh (2002). with data averaged over three and five years to abstract from business cycle relationships in European and non-European HICs as well as the upper and lower MICs. upper middle-income countries (MICs). and Arcand et al.5 Second. and Song and Thakor (2010). Still others. following the World Bank's income classification. In general.and the lower-MICs. unlike bank credit. including Deidda and Fattouh (2002). it is significant to note that the more homogeneous the economies involved in a panel. our empirical results are consistent with the views that the financial system provides important services for economic growth. and Arcand et al. Bank credit is a strong determinant of economic growth for both the upper. 2 . and high regions). and the impact of banks and markets on economic growth is assessed in one pass. see Boot and Thakor (1997). Finally. to control for the heterogeneous characteristics of countries in a panel. to understand the relationship between the financial system and economic growth more com- prehensively. intermediate. three different variables are employed to measure both stock markets and banks. emphasize that the focus should be on creating well-functioning banks and markets rather than on making a choice between the two as they are not only competing but also complementary sources of financing. (2013). and the endogeneity problem with the inclusion of a lagged dependent variable as a regressor. In sharp contrast to the existing empirical findings. Further. For European HICs. (2012). the greater the opportunity to 2 For details. implying that the link between financial development and economic growth depends on the stages of economic growth of the countries considered in the study. Singh and Weisse (1998) show that stock markets are unlikely to spur long-term economic growth in developing countries as they encourage short-term profits and also require sophis- ticated monitoring systems to function effectively. including Levine (1997). and Song and Thakor (2010). For details. including Beck and Levine (2004). and different levels of economic growth of countries in a group. developing and developed countries are usually pooled together under a strong assumption of homogeneity. Fufa.3 The challenge is even more vivid if we have to use different indicators of financial development as there is no single best indicator of financial development for a country. the same cannot be said of their effect on the growth of the HICs. 3 For details. this may not be very informative as it is likely that financial intermediaries and markets have quite different impacts on growth depending on the phases of countries’ economic development. there is little evidence that the existence or the size of an equity market matters for growth. in order to consider the highly integrated financial systems that have close relationships with one another. non-European HICs. and lower MICs. The literature also stresses that the relative merits of banks and stock markets evolve over time and vary at different stages of economic growth of countries. For instance. Though few studies. (2012). Rioja and Valev (2004). More specifically. omitted variable bias. They divided a group of countries in three regions according to the level of financial development (low. Deidda and Fattouh (2008). 5 Rioja and Valev (2004) did not include stock markets in their study. Because of the strong spillovers and externalities in financial systems. First. The main question in this study is whether the link between financial development and economic growth becomes more evident as more homogeneous groups of countries are included after considering the endogeneity issue. stock market liquidity is not a strong determinant of growth. In particular. a group of 64 countries is divided into four different subgroups: European high-income countries (HICs). however. Rioja and Valev (2004). and that stock markets and banks play different roles. the services provided by financial markets become relatively more important and countries become more market-based and imply that the services provided by banks make a significant contribution to the process of economic growth in the early stages of development. we can at least assess whether the role of financial development. and high correlation and strong integration of financial intermediaries and markets between cross-sectional units.4 differs based on the various stages of economic growth of the countries after controlling for simultaneity bias. We find that bank credit is not robust and stock market liquidity is only significant in the case of non-European HICs. Thus. this paper finds that the effects of stock markets and banking system development on economic growth differ for the various income groups of the economies. stock market liquidity exerts a robust influence on the economic growth of the upper-MICs only. Allen and Gale (2000). Boyd and Smith (1998). The answer to the question especially depends on how persistent financial systems are and how big economic heterogeneity is among the countries. This study is distinct from the existing literature in that the dynamic panel system GMM with Windmeijer (2005) correction not only utilizes information more efficiently from financial systems and economies but also provides more economically reasonable estimates as more homogeneous countries become involved.T. For the sensitivity analysis. these studies consistently indicate the difficulty of setting threshold levels of financial development indicators while analyzing the non-monotone effects of financial development on growth. we present a GMM with more homogeneous groups to utilize reasonable information from a variance–covariance matrix to account for cross-equation correlations among the cross-sec- tional units in each group.2 These studies show that as economies grow. The regression results show that while bank credits and stock market liquidity have a positive and robust impact on the economic growth of the MICs. to estimate the link and test the impact of the development of stock markets and the banking system on economic growth in terms of a degree of homogeneity. J. By doing this.

2004–2008. 6 According to the World Bank.746 or more.045 but less than $12. However.125. 3 . and market capitalization. Traded value equals the value of all domestic shares traded in the stock market divided by GDP. 1994–1998. the HICs have the largest and most active stock markets while the lower MICs have the least active markets. traded value. Data permitting.2. In general.002% in Swaziland in 2004–2008 to a maximum of 621% in Hong Kong in 2009–2012. are sourced from the electronic version of the World Bank's World Development Indicators. we use bank credit.746. relative to the total sample. to GDP. private credit. Fufa. Private credit equals credit issued by deposit banks and other financial institutions. with the exception of schooling from the Barro and Lee (2010) data base. Turnover ratio equals the value of the traded shares in the domestic stock market divided by the total value of listed shares. and facilitation of exchange of goods and services.and middle-income economies are sometimes referred to as developing economies. The lower and upper MICs are separated at a GNI per capita of $4. pooling and mobilization of savings. As a result. J. high-income (31) and middle-income (33). with the minimum credit of about 5% issued by deposit banks in Ghana in 1989–1993 and the maximum amount of about 224% extended by banks in Iceland in 2004–2008.1. and liquid liabilities. there is no single best indicator that shows the extent to which banks and stock markets provide these services across a number of countries. and it measures the size of the market relative to the economy. excluding central banks. there is a huge cross-country variation in these measures. it could also show over-lending. For banking system development. (2012). The theory predicts that more liquid/active markets facilitate efficient allocation of resources and foster growth. Liquid liabilities range from a minimum of 12% in Peru in 1989–1993 to a maximum of 354% in Luxembourg in 2004–2008. taking into account the time dimension of the data. with the average being 38% for the full sample (62% in the HICs and 15% in the MICs). However. and their average is divided by GDP deflated by annual CPI. 2.7 We source the financial system indicators from the World Bank's “Financial Structure and Development Data Base. while the HICs have a GNI per capita of $12. and it measures the liquidity of the stock market relative to its size. three-year averaged data are also used in this study. and the MICs are composed of upper and lower MICs. 9 For details. following the World Bank's income classification system. It excludes credits by other financial institutions and to the government and public enterprises. and IMF's International Financial Statistics Yearbook for banking system measures. The HICs include both European and non-European countries. Preliminary analysis Table 1 presents the summary statistics of the indicators. and 2009–2012. or the ratio of broad money. respectively). see Arcand et al. For instance. the financial system indicators are updated using the original sources: Standard & Poor's Emerging Stock Markets/ Global Stock Markets Factbook for stock market measures. These two measures were almost identical until the late 1990s. monitoring of investments. Traded value also varies from a minimum of 0. Well-functioning financial systems mitigate market frictions and spur economic growth through the provision of information about investment opportunities. with the average being 72% (95% in the HICs and 50% in the MICs). but they started diverging at the beginning of the new millennium. Bank credit equals total credit extended by deposit money banks to the private sector as a share of GDP. we rely on multiple standard measures of the size and activity of banks and stock markets. Table 1 also shows the variation in banking system measures across the sample. which could deter growth. for example. averages about 66% of GDP (94% in the HICs and about 40% in the MICs). M3. diversification and sharing of risks. the MICs have a gross national income (GNI) per capita of more than $1. As we can see. where Swaziland had the minimum value of 0. Data The study employs a panel of 64 countries for which we have complete stock market and banking system data for the period 1989–2012. Data and preliminary analysis 2. 2. Liquid li- abilities are calculated as the ratio of liquid liabilities10 of the financial system to GDP. All the financial system indicators are deflated by end-of-period prices. The average size of the stock market is about 58% of GDP (79% in the HICs and 38% in the MICs). However. and Pakistan had the maximum value of 377% in 1999–2003. Private credit also averages about 71% of GDP (101% and 43% in the HICs and MICs. to the private sector divided by GDP.02% in 2004–2008.T. Market capitalization equals the total value of listed shares in the stock market divided by GDP. 1999–2003. The countries are divided into two groups.6 It is a common practice in panel studies of the finance and growth relationship to rely on a less frequent data set to minimize the effects of business cycle fluctuations masking the true long-run relationship between growth and financial development. It measures the overall size of the financial intermediary sector and is commonly used in the literature under the assumption that the size of the sector is directly correlated with the financial services it renders. Bank credit. the data are averaged over five-year non-overlapping periods: 1989–1993.9 Higher levels of each could represent higher degrees of financial services to the private sector and thus greater financial intermediary development. we use three measures: turnover ratio. 8 Where available. Note that the low. and it measures how active the stock market is relative to the size of the economy. 7 For sensitivity analysis. 10 This includes currency plus demand and interest-bearing liabilities of banks and other financial institutions. To measure stock market development.”8 whereas the data for all the control variables. and also to allow easy comparison with the existing literature. Kim Research in International Business and Finance xxx (xxxx) xxx–xxx observe the link in terms of more economically reasonable and stylized evidence. it does not necessarily measure the degree to which financial institutions overcome the adverse effects of information asymmetry and transaction costs to provide effective service. turnover ratio averages about 53% (72% in the HICs and 34% in the MICs).

62 Inflation 49.01 2282.94 2.67 4. dev.72 Inflation 27.39 1.76 0.92 −0.89 Turnover ratio 52.59 366.50 85529.80 Schooling 9.90 15.97 11.11 the ratio of total final government consumption expenditure.04 25.67 Initial income 28469.61 18. We use a Wald test to examine if the indicators enter our regression jointly significantly. 3.03 5.73 Traded value 61.05 81.97 Liquid liabilities 50.02 47.30 29.68 −8.43 0.82 −2.82 52. while traded value and market capitalization are positively and significantly correlated with growth in the case of the HICs.65 6.12 The correlations in Tables 2 and 3 show that for the MICs.58 239. Kim Research in International Business and Finance xxx (xxxx) xxx–xxx Table 1 Summary statistics.65 0.63 As is standard in the literature on the finance-growth nexus.8 Schooling 7.05 4.18 5. and trade openness.002 124.52 1.40 −8. size 13. Blundell and Bond (1998).01 Liquid liabilities 71.44 69. J.75 47. Arellano and Bover (1995).10 51.85 Traded value 15. as measured by the government size.002 621. which is the sum of imports and exports as a fraction of GDP. we include the following control variables.33 64.68 0.37 5.20 325.97 13.38 10.69 0.033 236.14 366. size 16.76 Bank credit 93.40 Openness 84.21 30.23 621. implying that both bank and stock market measures may offer more conservative estimates of the role of banks and stock markets in economic growth than alternative approaches. which have been shown empirically to have robust growth effects: real GDP per capita. growth is positively and significantly correlated with all financial measures except market capitalization.12 14490.94 353.20 4631.23 85529.72 Inflation 4. 13 At the same time.69 5.89 Turnover ratio 71. the banking system and stock market measures are generally significantly correlated to each other for both sets of countries (especially in the case of the MICs).94 169.07 0. Min.59 1.38 40.47 2.67 1.91 16424.62 55.26 51.14 The cross-country growth regression can be written as follows: 11 Both variables are commonly referred to as a simple control set.59 79.67 Initial income 15228.84 224.77 Openness 97.92 7.84 147. refer to Hansen (1982).38 10. form our policy control set.17 42. All countries (64) Economic growth 2.28 Private credit 43.02 1.02 376.69 Govt.87 425.84 224. capitalization 79.23 8312.37 Initial income 2715. together with the simple control set.08 30. capitalization 58.26 11.63 Middle-income countries (33) Economic growth 2.98 High-income countries (31) Economic growth 1.80 62.87 206.80 0.04 2.10 478.23 18.31 1916.88 173.98 −5.57 −2.82 17.49 28. Arellano and Bond (1991).15 4.00 49. Variable Mean Std.22 60.033 478.84 4.13 The summary statistics for individual countries are given in Table 4.01 Private credit 101.40 Openness 72.84 150. On the other hand.38 33.58 224.97 2282. capitalization 38.53 15.39 Bank credit 66.78 15. to control for human capital accumulation. to capture the tendency for growth rates to converge across countries and over time.85 Traded value 37. 12 The last three variables for macroeconomic stability are policy related and.25 Govt.68 0.33 Mkt. credit measures are negatively and significantly correlated with growth.62 53.99 191.08 Turnover ratio 34. and Roodman (2009). the CPI-based inflation rate.33 Mkt.02 30. Econometric methodology We examine the link between economic growth and both stock markets and banking system development using the dynamic panel GMM estimators.77 62.03 425. 14 For a detailed description of the various GMM estimators.76 Schooling 6. average years of schooling in the population 25 years of age or older.25 Govt.01 Private credit 71.49 30.56 Mkt.33 4. 4 . size 19.48 224. Max.T.39 Bank credit 40.01 Liquid liabilities 94.021 376. Fufa.00 353.31 11.31 2.91 4.37 13.

061 −0.745 (0.000) 0.154 0. 0.079 (0.947 (0.015 0.206 (0.444 (0.222 (0.000) 0.006 −0.246 (0. 0. J.313) 0.000) 1.00 5 InitialIncome −0.300 −0.082 (0.043) 1.484 1.148) 0.592 (0.262 (0.113 (0.000) 1.006) 0.00 TurnoverRatio 0.002) 0. Fufa.042) −0.238 (0.144 (0.576 (0.000) 0.373) (0.004) 0.002) 1.000) 1. growth Bank credit Private Credit Liquid liab.Size −0.334 (0.000) 0.169 (0.300) 0.012 (0. cap.000) 0.238 0.284 (0.00 (0.411) (0.090 (0.144 (0.223 0.745 (0.195 1.123 (0.00 (0.571 (0.080 0.034) (0.148) (0.Cap.000) Govt. Econ.000) 0.244 (0.049) (0. Kim Table 2 Correlations.000) 0.000) 0.012) Note: Significance levels are reported in parenthesis.249) 0.00 BankCredit 0.071 −0.000) 0.532 (0.027) (0.436) (0.495) (0.113 (0.511 (0.03 (0.706) −0.000) 0.083 (0. Initial Schooling Govt.000) 0.002) 0.064) −0.881) −0.234 (0.119 0.153) 0.075 (0.122 (0.122) 0.156 (0. Research in International Business and Finance xxx (xxxx) xxx–xxx .125 0.381 (0.845 (0.248 (0.066 1.114 0.110 (0.215 (0.001) 0. T.160) 0.00 PrivateCredit 0.065) 0.000) 1.00 TradedValue 0.291) 0.134) (0.627 (0.008) 0.159 (0.000) 1.173 0.064 −0. middle-income countries.035 (0.296 (0.006) 0.000) (0.031) −0.001) 1.167 0.00 LiquidLiab.300 (0.00 (0.00 (0.000) −0.121) 0.398) Openness 0.846) Schooling 0.003) 0.213 (0.00 Mkt.684 (0.002) Inflation −0.136 (0.307) (0.005) (0.659) −0.085) 0.00 (0.347) 0. Turnover ratio Traded value Mkt. size Inflation Openness income EconGrowth 1.000) 0.442 (0.054 −0.112) (0.936) (0.

474 (0.881 (0.119 0.218 −0.007) LiquidLiab.199 (0. Kim Table 3 Correlations.000) 0.615 (0.274) 0.099 (0.000) 0.660) TurnoverRatio 0.072 −0.004) 0.00 (0. growth Bank credit Private credit Liquid liab.099 0. cap.071 1.307 (0.065 1.287 (0. T.450 (0.310 (0.386) 6 TradedValue 0.000) 0.831) (0.000) 1.000) 0.440 0.157 0.180 (0.479 (0.114 −0.011) (0.00 (0.00 (0.052) (0.089 (0.088 (0.387 (0.204 −0.00 (0.269 −0.159) (0.00 (0.000) 0.000) 0. size Inflation Openness EconGrowth 1.386 (0.224) 0.215 −0.Size −0.421) Note: Significance levels are reported in parenthesis.177 −0.367 −0.586) 0.140 (0.177 (0.176) −0.00 (0. 0.000) 0.279) 0.000) Inflation −0.000) 1.012) PrivateCredit −0.292 −0.023) 0.013) 0.007) (0.496 (0.353 (0. Econ.00 BankCredit −0.Cap.941) (0.00 InitialIncome −0.010) 0.000) 0.076) (0.000) 0.00 (0.000) 1.00 Mkt.241 0.000) 1.031 (0.000) (0.701) 1.141) Govt.201 1.000) 0.000) 0.615 (0.144 −0.160 (0. Research in International Business and Finance xxx (xxxx) xxx–xxx .000) 0.000) (0.216) (0.001) 1.000) −0.540 (0.000) 0.006 −0.101 −0.149 (0.626 (0.183 (0.219) Openness 0.026) (0. high-income countries. −0.017 −0.456 (0.154 −0.409 −0. Initial income Schooling Govt.755 (0.036 0.234 (0.274 (0.026) 0.583 (0.391) 0.000) 0.00 (0. Turnover ratio Traded value Mkt.218 0.375) (0.082) 0.000) −0.109 (0.065) 0.206 (0.182 −0.153 (0.023) 0. J.023) (0.00 (0.208 (0.044 (0.01) 0.356 (0.533 (0.070 (0.007) (0.047) 1.000) (0.000) 1.057) 0. Fufa.561 (0.002) Schooling −0.056) (0.001) (0.

806 21.487 Norway 1.410 0.748 33.066 Singapore 3.827 Hong Kong 2.302 22.260 159.299 95.964 14.684 20.913 41.353 125.316 72.352 0.072 5.139 62.183 78.014 30.320 2.406 125. Country Econ.294 19.224 3.988 33.694 161.451 15.646 19.750 2.698 54.271 39.688 3. 7 .558 25.777 60.680 31.014 149.890 32.858 32.573 147.663 13.811 93.465 98.570 78.855 24.712 3.645 38.866 77.998 74.547 Trinadad & Tob.980 Spain 1.075 0.753 31.850 5.961 Switzerland 0.096 31.968 18.656 28.877 1.703 Ecuador* 1.166 101.378 83.876 44.367 54.515 19.406 22.372 12.452 Netherlands 1.317 27.407 Hungary 0.640 131.962 65.670 58.509 1.456 17.451 8.084 29.999 32.117 110.556 63.867 93.808 125.291 23.262 57.606 2.795 Botswana* 2.813 64.939 81.508 105.328 1.563 57.314 51.904 109.687 Tunisia* 2.913 88.945 79.396 27.950 29.563 Thailand* 3.619 25.194 78.988 105.746 44.416 Sweden 1.112 161.081 Philippines* 1.454 45.264 18.074 10.238 1.453 24.770 81.937 Israel 1.061 41.952 2.940 193.109 India* 4.903 100.210 51.354 68.611 17.595 41.763 159.525 40.718 38.718 23.066 190.430 169.948 107.946 Japan 1.901 42.560 60.997 53.377 74.988 92.589 Mauritius* 3.798 33.559 China* 8.574 85.630 114.882 5.145 47.726 7.899 Morocco* 2.852 152.295 42.133 25.551 143.845 159.066 Barbados 0.203 84.329 62.060 91.158 20.658 93.281 89.942 Cote d’Ivoire* −0.556 54.005 24.090 53.575 50.846 59.391 12.841 Kenya** 0.838 16.065 20.872 12.614 84.852 22.746 39.195 Portugal 1.977 22.463 2.367 75.754 27.571 Denmark 0.228 89.778 85.137 63.910 18.823 28.867 122. country average.558 83.689 5.386 Nigeria* 2.418 20.028 51.471 48.317 Luxembourg 2.087 38.832 Colombia* 1.957 73.753 28.063 108.527 39.357 80.814 0.558 77.804 56.514 45.043 18.684 Jamaica* 0.557 59.786 66.422 118.664 56.066 105.232 52.376 57.457 109.879 23.146 Uruguay* 2.533 75.526 48. 2.977 40.428 24.556 82.271 63.044 95.655 31.034 79.778 41.455 22.263 113.113 65.036 59.739 Poland 2.310 16.571 324.758 27.368 32.321 20. Kim Research in International Business and Finance xxx (xxxx) xxx–xxx Table 4 Bank and stock market measures and economic growth.067 25.534 Sri Lanka* 4.107 72.078 139.417 104.437 31.246 15. J.340 22.031 0.548 138.723 11.992 33.226 98.272 8.967 10.249 39.007 32.421 27.701 23.374 35.027 115. Fufa.278 New Zealand 1.670 79.350 76.175 61.952 54.864 31.T.370 Swaziland* 1. The values are based on 5-year average data.102 13.776 13.322 92.322 11.100 37.724 18.459 116.645 183.053 64.434 106.173 Australia 1.556 33.477 17.015 105.380 35.622 62.287 18.633 43.783 Italy 0.12 47.778 125.286 0.821 0.729 0.667 14.085 202.159 Jordan* 1.394 114.839 Panama* 3.556 119.586 40.390 138.245 90.647 90.322 70.570 52.272 12.510 Bolivia* 1.800 35.132 96.214 30.235 127.550 Russia* 0.167 42.546 62.738 45.590 96.852 25.945 52.541 54.705 Austria 1.503 110.616 7.297 France 1.065 166.411 65.112 118.945 110.586 10.302 39.433 3.172 Iceland 1.683 46.564 23.051 Iran* 3.286 Peru* 2.620 105.033 Finland 1.549 123.843 Note: * represents MIC and the rest HIC.21 3.793 Indonesia* 3.946 2.675 61.395 Saudi Arabia 1.561 80.999 24.104 Ghana* 2.369 23.759 22.355 15.880 71.534 77.518 72.296 320.718 60.303 92.725 Belgium 1.987 119.282 20.227 20.162 58. 1989–2012.945 67.980 16.604 67.891 22.001 68.155 1.333 61.012 44.012 75.576 24.550 2.584 16.510 5.856 112.605 Greece 0.724 56. growth Bank credit Private credit Liquid liabilities Turnover ratio Traded value Market capitalization Argentina* 2.490 65.554 5.424 230.545 33.169 120.330 UK 1.402 Egypt* 2.752 155.450 98.035 115.832 Korea 4.347 6.291 60.320 159.785 18.201 0.730 52.508 263.271 98.283 56.524 18.412 99.116 Turkey* 2.446 99.330 33.121 Malaysia* 3.391 25.447 72.674 90.034 39.950 Brazil* 1.647 105.966 43.326 110.270 150.707 28.166 45.293 31.504 18.897 34.616 34.777 154.201 18.162 Pakistan* 1.639 92.601 23.917 0.468 0.411 26.287 40.107 72.286 1.682 48.141 128.885 75.258 Germany 1.028 65.608 39.886 149.630 63.790 Canada 1.124 127.187 60.291 Mexico* 1.022 96.910 27.545 82.235 32.422 50.020 41.407 South Africa* 0.923 9.504 143.638 95.242 77.916 United States 1.334 44.744 39.106 1.709 42.097 17.072 61.

These are valid instruments under the following additional assumption: although a correlation may exist between the levels of the explanatory variables and country fixed effects in (2). t − 1)] = 0 forl ≥ 2. and εi. To increase efficiency.16 We also use two specification tests. and this adversely affects the small sample and asymptotic properties of the difference GMM.t − ei. is to capture time-specific effects as well as unobserved cross-sectional dependence across national outputs that might be correlated through common economic shocks. t − 1) + (εi. μi represents time invariant country-specific effects. This method. t − l − Xi. We can reduce this instrument count problem by either restricting the instruments to certain lags instead of all available lags or by collapsing the instrument matrix. t − εi.t−l(ei. t )] = 0 forl = 1 (7) E [(Xi. t − εi. t − 1) (3) Although the fixed effects are expunged. t + p μi ] = E [yi. known as the Windmeijer correction.t is a set of explanatory variables. measures of stock markets and bank development. which was proposed by Arellano and Bover (1995). (1988) and Arellano and Bond (1991) propose the first-difference transform of Eq. Blundell and Bond develop a system GMM. (2) contains the lagged dependent variables correlated with the unobserved country fixed effects. t − l − 1)(μi + εi. rather than separate moments ∑t. t − l (εi. t = 3. uses the following moment conditions. and (8) to generate consistent and efficient estimates. This dynamic panel estimator. Fufa. t − l (εi. (7). Kim Research in International Business and Finance xxx (xxxx) xxx–xxx yi. t (2) where yi. t − 1 = (α − 1) yi. (5). (4). we replace the moment conditions of the standard difference GMM (4) and (5) with (9) and (10). In our estimation. including average years of schooling. This assumption results in the following stationarity properties: E [yi. as no lags are actually dropped. t − l − 1)(μi + εi. government consumption expenditure.T. 16 See Roodman (2009) for details. (4) and (5)) but we only want the estimator to minimize the magnitude of the empirical moments ∑tyi. (2) as follows: yi. t − 1 + β′Xi.15 The inclusion of time dummies. and Xi. significantly minimizes the potential biases that arise due to over-identification problems and boosts the efficiency of our estimates without losing information. In particular. inflation rate. T (5) However. t − Xi. …. the lagged per capita GDP as well as any of the control variables in X are still potentially endogenous. t − εi. t + q μi ] for all p and q (6) The additional moment conditions for the regression in levels are: E [(yi. The consistency of the dynamic panel GMM results rests on the validity of the instruments and the assumption that the error terms do not exhibit serial correlation. t − yi. t + p μi ] = E [Xi. t − 1)] = 0 forl ≥ 2. J. with the two equations being distinctly instrumented. uses the moment conditions in Eqs. t + q μi ] andE [Xi. The new moment conditions state the same orthogonality assumption between the lagged levels and differenced error term (namely. τt. t )] = 0 forl = 1 (8) The dynamic panel GMM-sometimes referred to as system GMM-. T (4) E [Xi.t − ei. t − 1)] = 0 for each l ≥ 2 (10) In a dynamic panel GMM. To overcome the issue of endogeneity. commonly referred to as difference GMM. to augment the difference estimator by simultaneously estimating in differences and levels. t = 3. this term is removed by using cross-sectionally demeaned data for all variables. t (1) Note that we can rewrite Eq. (1) as: yi. giving rise to bias and inconsistency of estimators. t − 1 = α (yi. E [yi. 8 . μi.lyi. t + τt + μi + εi. t − εi. t − 1)] = 0 for each l ≥ 2 (9) E [Xi. ….t−1) for each l. the instrument count may become equal to or larger than the number of cross-sectional units. t − 1 + β′Xi. While the instruments for the equation in differences are the same as above. t + τt + μi + εi.t−l(ei. t − l − yi. and they may fail to remove the endogenous components of the variables and result in biased parameter estimates towards those from non-instrumenting estimators. Holtz- Eakin et al. The addition of regression in levels also allows us to examine the cross-country re- lationship between our variables of interest. The latter can be formally expressed as: E [yi. thus over- fitting the instrumented variables. t − 1 − yi.t is the logarithm of real per capita GDP in country i at time t. Beware that Eq. τt captures time-specific effects. To remove this dynamic panel bias. the in- struments for the equation in levels are the lagged differences of the explanatory variables.t−1) for each l and t. Blundell and Bond (1998) demonstrate that when explanatory variables are persistent over time. t − l (εi. the untransformed lagged levels of these variables are weak instruments for transformed variables. t − 2 ) + β′ (Xi. t − l (εi. and trade openness. thus. the estimator can suffer from a potential instrument proliferation. t − εi. Arellano and Bond (1991) use the lagged levels of the explanatory variables as instruments.t denotes the idiosyncratic shocks. there is no correlation between the differences of these variables and the country-specific effect. t − yi. The first relates to instruments and includes a Hansen-J test of joint validity of the instruments and the difference-in-Hansen tests of exogeneity of instrument subsets (the null 15 This is a modified version of Barro's (1991) growth regression model. t = αyi.

In order to attenuate this problem and obtain a more policy relevant outcome. respectively.20 A panel regression that does not address the heterogeneity in the cross-country units may conceal the likely differential growth effects of financial development arising. Both GMM estimators show that while bank credit is statistically not significant at the 5% level. MIC (which takes a value of 1 and 0 respectively if the country is a middle-income and high-income country. respectively) with bank and stock market measures. but they have no discernible effect on that of the HICs. the interaction terms between the indicator variable and bank and stock market measures support our findings. Our specification tests also confirm that we cannot reject the null hypotheses of no second-order serial correlation in the differenced error term. 20 In the case of the HICs. the stock market turnover ratio is significant in all regressions for both the three-year and five-year averaged data. All results are qualitatively similar and are available upon request from the authors. the Wald test for joint significance shows that.17 4. although only stock market measures enter the regressions with statistically significant coefficients. Bank credit. we run regressions on the full sample. It is interesting to see that though the overall banking system measures are statistically insignificant for the HICs.19 As can be seen. they are significant for the European HICs. although they are marginally significant for the non-European HICs. It may also be that the stock markets may suffer from lack of transparency – limited firm-specific information. by hindering effective exercise of property rights and corporate governance by investors. the differenced error term is likely first-order serially correlated. 9 . The results also indicate that stock market liquidity is a major determinant of economic growth for both groups of countries. among other factors.S. In the case of the two sub-groups of the HICs. The above regression results indicate that after controlling for liquidity of stock markets. both GMM estimators show that bank credit and turnover ratio enter all the regressions with positive and statistically significant coefficients. 21 For details. from differences in the stages of economic development of the countries. For the MICs. we further sort the countries into upper-and lower-MICs and European- and non-European HICs and run separate regressions for each of these relatively more homogeneous sub-groups. the credits extended to the private sector strongly boost the economic growth of the MICs. both bank and stock market measures enter the regressions jointly significantly. The results imply that the size of credit extended to the private sector is a strong determinant of economic growth for both the upper. Fufa. the insignificant coefficient estimates of stock market measures perhaps imply that their already established stock markets may not yet have reached some minimum size or activity level beyond which they can robustly influence economic growth. particularly those with long-term maturity. These results are robust to the inclusion of more control variables. stands for countries with the most developed stock markets such as U. However. exerts a robust influence on the economic growth of upper-MICs only.18 Each of the two reported regressions includes bank credit and turnover ratio. the banking system may also have continued to be the major source of financial services in these countries. 19 In addition to sorting out our sample countries into HICs and MICs and running separate regressions. including Rose (2014).A. we employ private credit and traded value as well as liquid liabilities and market capitalization as indicators of banking system and stock markets. Some studies. respectively. For the HICs. indicate that banks and bond markets.and three-year average data for the MICs and HICs.and three-year averaged data. the coefficient estimate of bank credit is insignificant. see Jin and Myers (2006). is not robust for non-Europeans. The findings that the banking system measures are statistically significant unlike the stock markets for the lower-MICs lends credence to the argument put forward by Singh and Weisse (1998). 18 In addition to bank credit and turnover ratio. in general. the estimation results are qualitatively very similar in both the difference and the system GMM estimators for five. in particular – which. They contend that developing countries should focus on making the banking sector the primary source of their financial services rather than diverting scarce resources to the promotion of 17 By construction. are more effective at influencing economic growth under inflation-targeting regimes. the turnover ratio enters all regressions for the full sample with a statistically significant coefficient. This implies that only stock market liquidity exhibits a meaningful growth effect for all countries in our sample. The regression also controls for government consumption. the lower-MICs in particular. where we interact the group dummy variable. However. We expect that the more homogeneous the countries. trade openness. Stock market liquidity. is somewhat intriguing as stock markets are regarded as having a comparative advantage over banks in raising funds for the innovative and high-tech investments that characterize these countries. stock market liquidity is statistically not significant for the European-HICs but it is statistically sig- nificant in all the regressions for the non-European HICs. only bank credit enters the lower-MICs growth regression with significant coefficients. Although we implicitly assume that the coefficients of the control variables are the same for all countries.. The second test examines the hypothesis that the error term is not second-order serially correlated. and Japan.21 By establishing and nurturing a long-run personal relationship with investors. Canada. and controls for lagged income and average years of schooling. in the case of the MICs. the more representative the coefficient estimates of each country in the group. Kim Research in International Business and Finance xxx (xxxx) xxx–xxx hypothesis being that the lagged differences of the explanatory variables are uncorrelated with the residuals). separate regressions for the two income groups – the MICs and HICs – show somewhat interesting outcomes.and lower-MICs. however. and the overall validity of our instrument sets and subsets. although it enters the European-HICs regressions with significant coefficients.T. supporting the inclusion of both measures in our regressions to examine if each exerts an independent effect on growth. It appears that the overall impact of financial development on economic growth is the same whether we use a three-year or five-year average data set. Empirical results Tables 5 and 6 show the empirical results from the dynamic panel GMM for the full sample as well as the five. As can be seen. and inflation. on the other hand. The finding that stock market measures are statistically not significant for the European HICs. The empirical results from both the dynamic panel difference and the system GMMs in Tables 7–10 show that while the coefficient estimates of both bank credit and turnover ratio are statistically significant for the upper-MICs. J. The fact that the banking system explains economic growth better than the stock market for the European HICs may be attributed to the strong common monetary policy adopted in the region. leads to poor performance of stock markets.

186 0.159) −0.054** 0.237 0.235) (0.234 0.346 0.004 (0.140) Schooling 0.162** 0.143) 0.449 Diff in Hansen(c) 0. respectively.101 (0.329 0.264 −0.332 0.048) (0.237 0. with robust standard errors in parentheses.110** 0.145 0.037) 0.070** 0. t − 1 and schooling.315) −0.286 (0.174 0.047) (0.039 (0. while (2) controls for those in (1) and government size.047** (0.276) (0.596) (0.140) (0.027 0. * and ** indicate significance at the 10% and 5% level.065) 0.164 0.231) −0.402) 0.524** −0.425) −0. T.480 0.167** (0.040) TurnoverRatio 0.034 −0.155 0.071) (0. and trade openness.179) −0.328** −0.105) Observations 189 189 189 189 91 91 91 91 98 98 98 98 Wald test(a) 0.003 0.168) −0.071 0.042) (0.099) 0.562** −0.054 (0.130 0. inflation.057) (0.126 (0.054) −0.179) 0.094) (0.026) (0.404 Note: All variables are logged and cross-sectionally demeaned.056) (0.343 (0.040 0.138) (0.075) (0.021) (0. P-values of post-estimation tests are reported.135 0.233** 0.217 (0. −0.050 (0.454** (0.174 0. all regressions incorporate Windmeijer (2005) correction. (c) The null hypothesis is that the instruments used are not correlated with the residuals.469 (0.105 0.184) (0.038 (0.448 0.066) TradeOpenness 0.020) (0.260* −0.174) −0.281 0.377) 0.238 0.015) 0.007 0.041) (0.001 0.154) (0.283) (0.068) (0.350) −0.420* −0.082 −0.122) (0.000 AR(2) test(b) 0.163 (0.015 (0.050** 0.042 −0.150 0.037 0.019 −0.533** −0.570 (0.297 −0.135* 0.049) 0.079 (0. Fufa.200) (0.149) (0.209) (0.185 0.222) (0.147) (0.079) (0.124) (0.110 0.065 −0.008 0.203) (0.077 −0.069** (0.004 (0. Variable Full sample (64) HICs (31) MICs (33) Difference GMM System GMM Difference GMM System GMM Difference GMM System GMM (1) (2) (1) (2) (1) (2) (1) (2) (1) (2) (1) (2) BankCredit −0.078** 0.123) (0.118 0.900 0.044* 0.150** 0. Kim Table 5 Dynamic panel GMM regression results (5-year average data).099 −0.055 (0.057** (0.277 0.310) (0. (1) controls for income per capita.264* −0.061) 0.043 (0.124 0.015 0.123** 0.326 0.083** (0.510** (0.021) (0.004 0.061 −0.124 (0.000 0.032 0.227) (0.468) 0.572 (0.197 0.014 0. Research in International Business and Finance xxx (xxxx) xxx–xxx .101) (0.072 −0. (b) The null hypothesis is that the errors in the first-difference regression exhibit no second-order serial correlation.022) (0.323 (0.281 Hansen J(c) 0.220 0. implying test of joint significance of banking and stock market measures.092 −0.022) yt−1 −0.025) 0. (a) The null hypothesis is that banking and stock market measures enter the regression with zero coefficients.129) Inflation 0.208) (0.004 0.046 0.130) (0.417) 10 Gov’tSize.261 0. J.030 0.

030) 0.051** 0.064 0.204) (0.121 0.271* 0.230* 0.086) (0. Fufa. Kim Table 6 Dynamic panel GMM regression results (3-year average data).041** 0.785 0.027** 0.379** 0.125) (0. T. respectively.166 0.163) (0.052 0.252** −0. Research in International Business and Finance xxx (xxxx) xxx–xxx .161** −0.251** −0.030 (0.041 (0.150** 0.144 −0.021) (0.134) (0.036 (0.044 (0.101) (0.034** 0.085 −0.098 −0.233 0.037** 0.046 −0.156** (0.000 0.076) (0. with robust standard errors in parentheses.019 0.035) (0. while (2) controls for those in (1) and government size.088 −0.133 (0.057) TradeOpenness 0.436 0.192) (0.087) (0.008 (0.080) (0.166** 0.196** 0. (a) The null hypothesis is that banking and stock market measures enter the regression with zero coefficients.025 (0.173 0.034) (0.015) (0.017) (0.011) (0. Variable Full sample (64) HICs (31) MICs (33) Difference GMM System GMM Difference GMM System GMM Difference GMM System GMM (1) (2) (1) (2) (1) (2) (1) (2) (1) (2) (1) (2) BankCredit 0. implying test of joint significance of banking and stock market measures. inflation.732 0.034 (0.039** 0.085) (0.017 0.175** −0.075) Observations 377 377 377 377 182 182 182 182 195 195 195 195 Wald test(a) 0.130 −0. (b) The null hypothesis is that the errors in the first-difference regression exhibit no second-order serial correlation.059 (0.086) (0.321 Hansen J(c) 0.083) (0.041 −0.105** 0.060) (0.024** 0.000 0.011) (0.016) (0.899** −0.123 −0.295 0.075) Inflation 0.000 0.058) 0.212** −0.019 0.101) (0.108) (0.073) (0.022 0.139) (0.028) (0.037 (0.782** −0.020) yt−1 −0.719 0.313** 0.039) (0.040) TurnoverRatio 0.470 0.038) −0.006 0.083** 0.120* −0.172 0.062 0.089) 0.059** 0.058** (0.026 0.040** 0.130* (0. all regressions incorporate Windmeijer (2005) correction.068) 0.059 −0.012) (0.119) (0.175 0.087) 0.105) 11 Gov’tSize.043) (0.059) (0.108 −0.107 −0.098) (0.039) (0.094 0.017 0.148* −0. −0.029) (0.057) (0.065** 0.000 0.012) (0.098) Schooling 0.024 0.724 Note: All variables are logged and cross-sectionally demeaned.024) (0.176 0.100** 0.085) (0.024) (0. P-values of post-estimation tests are reported.082** (0.775 Diff in Hansen(c) 0.474 0.009) (0.010) (0. and trade openness.143** −0.028) −0.341 0.039** 0.000 AR(2) test(b) 0.057) (0.283 0.212) (0.124 (0.005 0.107 0.094** −0.140) (0. J.304** −0.027) 0.265 0.077) (0.140 (0.100) (0. (c) The null hypothesis is that the instruments used are not correlated with the residuals.148) (0. t − 1 and schooling.198* −0.058 0.161 0.029** 0. (1) controls for income per capia. * and ** indicate significance at the 10% and 5% level.056) (0.023) 0.059 0.152 0.185** −0.051 −0.351) (0.241 0.

019) (0.241 0.025) 0.004 (0. respectively.151) (0.044 (0.907** −0.146** 0.044) 0.857** −0.453 0.052* (0.592 0.982 0. sophisticated financial institutions like stock markets.034) 0.072* (0.335) Observations 54 54 35 35 57 57 33 33 Wald test(a) 0.008 0. Our finding regarding the lack of robust growth effects of financial sectors.130 0.138) −0.017 0.131 2.021) 0.427) −0.170 (0.069** 0.213** 0.402 0.203 (0.449) (0.212 (0.046 (0.039) 0.408) 0.169) (0. while (2) controls for those in (1) and government size.033) 0.108 (0.099) (0. * and ** indicate significance at the 10% and 5% level.251) TradeOpenness 0. −0.249 (0.112) TurnoverRatio 0.021 (0.059) 0.390) Inflation −0.693) 0. which was characterized by 22 It is a period analyzed by most of the early studies that established a positive relationship between financial development and economic growth.437) 0.020 0.046* (0.779 0. especially for the HICs.116** 0. (c) The null hypothesis is that the instruments used are not correlated with the residuals.258** −0. Kim Research in International Business and Finance xxx (xxxx) xxx–xxx Table 7 Dynamic panel difference GMM results (5-year average data).162) 0.107 0.332 −0.078) (0.450) Gov’tSize.144) (0.702 0.192 (0.066 0.781** −0.418) (0.120) 0.051* (0. (1) controls for income per capia.973** (0.370** −0.274) (0.300 (0.194) (0.337 (0.851 0. 12 .514* (0.262) 0.331 0.463) 0.437 0.026 0.775** −0.804 Note: All variables are logged and cross-sectionally demeaned.414 0.068 (0. Fufa.572) −0.291) (0. all regressions incorporate Windmeijer (2005) correction.038 0. −0.927 Diff in Hansen(c) 0.127 −0.111) TurnoverRatio 0.072 (0.657) 1.310 0.071 (0.990 (1.023 AR(2) test(b) 0.188) 0.T.001 0.631 (0.029) yt−1 −0.049) yt−1 −0.154** (0.297 (0.186** 0.087) (0.257 (0. respectively.068) (0.171) 0. and trade openness.387 −0.135 0.283 (0.372) Schooling 0.225 0.353 0.292) 0.755** (0. (1) controls for income per capia.517) (0. Variable European HICs (18) Non-European HICs (12) Upper-MICs (20) Lower-MICs (11) (1) (2) (1) (2) (1) (2) (1) (2) BankCredit 0. is in line with the findings of some recent cross-country and panel studies: although financial deepening has had a positive and significant effect on growth between 1960–1989.027 (0.403) 0.009 0.328** (0.235) 1.214 (0.053) −0.626) 1.063) (0. with robust standard errors in parentheses. Table 8 Dynamic panel system GMM results (5-year average data).161** (0. inflation. t − 1 and schooling.544** 0.040) 0. which not only rely on more advanced monitoring and infrastructure but also encourage short-term profits.295 0.153 (0.410) 0.104) −0. J.133 (0.004 AR(2) test(b) 0.007 0.318 (0.340** −0.569) Inflation −0.178** 0.073 (0.032) 0.459 0. and trade openness. with robust standard errors in parentheses.094) (0.392) −0.260) −0.006 0.230) (0.494) (1.651 Hansen J(c) 0.073** (0. (b) The null hypothesis is that the errors in the first-difference regression exhibit no second-order serial correlation.869 (0.680 (1.400 0.418 0.060) 0.426 0.217) −0.254 (0.107* (0.323 (0.183) 0. while (2) controls for those in (1) and government size. inflation.032) 0.845 0.064 (0.101) (0.887 0.301** (0.004 (0.174) TradeOpenness 0.946** −0.045 (0.355 0.528 (0. P-values of post-estimation tests are reported.321 0.277 0.180 (2.176) Schooling −0.230 0.135 −0.438 0.390 Note: All variables are logged and cross-sectionally demeaned.081) 0.303) (0.077* (0.450) (0.184 0.192 0.299) Observations 54 54 35 35 57 57 33 33 Wald test(a) 0.132 0.337) 0.046 0. (a) The null hypothesis is that banking and stock market measures enter the regression with zero coefficients.073** 0.015 (0.616** (0.387) −0.359 −0.275 0.091) 0.22 its effect vanishes during the later period of 1990–2004.000 0.115) (0.155) (0.381 0.266) Gov’tSize.252 −0. (c) The null hypothesis is that the instruments used are not correlated with the residuals. (b) The null hypothesis is that the errors in the first-difference regression exhibit no second-order serial correlation.068) 0. (a) The null hypothesis is that banking and stock market measures enter the regression with zero coefficients. all regressions incorporate Windmeijer (2005) correction.896 0.019) −0.163 −0.068) 0.003 0.056* (0. Variable European HICs (18) Non-European HICs (12) Upper-MICs (20) Lower-MICs (11) (1) (2) (1) (2) (1) (2) (1) (2) BankCredit 0.120** (0. t − 1 and schooling.388 0.547 0.069) (0.030) 0.164) (0.076 (1. * and ** indicate significance at the 10% and 5% level.356 0.195) 0.413 (0. P-values of post-estimation tests are reported. implying test of joint significance of banking and stock market measures.310) (0.262** (0.331) (0.024 −0.058 Hansen J(c) 0.125) (0.121) 0.252) (0.824 0. implying test of joint significance of banking and stock market measures.327** (0.042 (0.116 (0.752 0.

054 (0.283) 0. 2012).724) TradeOpenness 0.321* (0.230) −0.124 0.101 (0.558** −0.324 0.056 (0.032) 0.663 0.838 (1.156 (0.999 0.991 0.220) (0.045 0.128** (0.048 0. t − 1 and schooling. all regressions incorporate Windmeijer (2005) correction. Conclusion The main question in this study is whether the link between financial development and economic growth becomes more evident 13 .132 (0.087) 0.626 (0. (c) The null hypothesis is that the instruments used are not correlated with the residuals.837 0.124) (0.110) (0.216** (0.175) (0.999 Diff in Hansen(c) 0. −0.159) Gov’tSize.148 (0.056** 0.113) 0. * and ** indicate significance at the 10% and 5% level.029) 0.058) −0.282) 0.523 0.046** (0.211) (0.678 0. all regressions incorporate Windmeijer (2005) correction.110** (0.391) Inflation −1.217 −0.021) (0.031) yt−1 −0. respectively.934 0.081 (0. P-values of post-estimation tests are reported.823 0.040) (0.030) (0.154 (0.242) Schooling −0.192) (0.t-1 and schooling.183) (0. (b) The null hypothesis is that the errors in the first- difference regression exhibit no second-order serial correlation.461) 0.102) (0. (b) The null hypothesis is that the errors in the first-difference regression exhibit no second-order serial correlation.155) (0.673) 0.002 (0.023) yt−1 −0.764** −0.011 0.100** 0.163) Observations 108 108 70 70 112 112 66 66 Wald test(a) 0.104** 0.017 0.630 (0.126) −0.T.166 0.998 (0.599 0.298 0.177 (0. (1) controls for income per capia.999 0.111 (0.045 0.044 (0.120) 0.183** (0.652) −0.418** −0.823 0.909** −0.113 (0.658 (0.199) −0.965 0.025 (0.049) (0.276) 0.136 0.503) −0.474 0.033 0.999 0.442** −0.999 Note: All variables are logged and cross-sectionally demeaned. inflation.293) (0.158 (0. the prevalence of excess bank credit – credit boom – in many developed countries is dampening economic growth (Arcand et al.148 (0.149 (0.221 0.015) 0. with robust standard errors in parentheses.041) TurnoverRatio 0.493) (0.366) 0. and trade openness.120* (0.005 AR(2)Test(b) 0.430) Gov’tSize.519 −0.810 0.201 0.036) TurnoverRatio 0.018) 0.063) 0.148) Inflation −0. 2011).076* 0. * and ** indicate significance at the 10% and 5% level. Kim Research in International Business and Finance xxx (xxxx) xxx–xxx Table 9 Dynamic panel difference GMM results (3-year average data). P-values of post-estimation tests are reported.073 (0.025) 0.049** (0.067) (0.. J.642) −0.140) 0. Variable European HICs (18) Non-European HICs (12) Upper-MICs (20) Lower-MICs (11) (1) (2) (1) (2) (1) (2) (1) (2) BankCredit 0. respectively.006 0. and trade openness. recurrent financial crises (Rousseau and Wachtel.206) (0.098) Observations 108 108 70 70 113 113 66 66 Wald test(a) 0.046* (0.994 0.266 0.007 (0.693 0.050** (0. implying test of joint significance of banking and stock market measures.280 (0.248 (0.830) −0.071) 0. −0.115** 0.346) (0. although the correlation between the two is well-established (Favara.297** −0.177) 0.455 0. 5.095** (0.261** (0.006 (0.070) TradeOpenness 0.215) 0.468 0. (a) The null hypothesis is that banking and stock market measures enter the regression with zero coefficients.202) 0.207 −0.177) 0.028 (0.021 (0.014 (0.002 0.039 (0. (c) The null hypothesis is that the instruments used are not correlated with the residuals.004 0.437 Hansen J(c) 0. Variable European HICs (18) Non-European HICs (12) Upper-MICs (20) Lower-MICs (11) (1) (2) (1) (2) (1) (2) (1) (2) BankCredit 0.087) Schooling 0.097 −0.001 0.891** −0.075 AR(2) test(b) 0.788 0.316) 0.266 (0.097) −0.303) (0.299) −0.039) (0.145 −0.274 0.017) (0.006 0. implying test of joint significance of banking and stock market measures.999 Note: All variables are logged and cross-sectionally demeaned.091** (0.186 (0.151) (0.133) (0.239) (0.004 (0.271) 1.057** 0.324** −0.954 0. (a) The null hypothesis is that banking and stock market measures enter the regression with zero coefficients.058 −0. with robust standard errors in parentheses.017 (0.026 −0.515 0.522 0.342 0.019) 0.090) (0.251* (0.022) 0.561 Hansen J(c) 0.193) (0. while (2) controls for those in (1) and government size.092** (0.749 0.076 −0.769* −0.013) 0.999 0.077 (0.040) 0.133) (0. while (2) controls for those in (1) and government size.085** 0. Table 10 Dynamic panel system GMM results (3-year average data).068 0.234) (0.075) 0.355** −0.032 (0.076) 0.147 0.202) (0.081 (0.004 0.123) (0.179 0. and there is no clear evidence that finance spurs growth.489) (0.147 0. (1) controls for income per capia.033** (0.143 (0. 2003).176) 1.001 0. inflation.018) 0.133 (0.034) 0.135 0. Fufa.133* 0.031) (0.999 0.259 −0.

1999. Polit. Rev. Holtz-Eakin... J.. Gale. European I Department. 104. (Eds. 10 (3). 1969. Beck. Economic growth in a cross section of countries. Hansen.. D.D. Windmeijer. Stock markets.. Int. Diamond. A.. Bank. R.The bond market: An inflation-targeter's best friend. 423–442. 693–733. U. Roodman. Money & Capital in Economic Development. Fattouh. 2010. Econ.. J. but stock market liquidity has a robust impact on the growth of the upper-MICs alone. Dev. Oxf. J. 6–36. CT. D. Valev. 1990. A. M. G. B. Finance 5 (1/2). New-York. Berkes..A. D. J. Finance 32 (6). 1995. H. Bertocco. J. 58. Financ. 1076–1107. G. Econ. Amsterdam. 98 (5). T. 2008. 339–345. pp. Thakor. Econ. A. Transactions costs. S. R. C.. R. L. 2014. 2000. Financ. Durlauf. Theory 67. J. 3–55. V. Financ. Financial intermediation and endogenous growth. Cambridge. Bank. J.. Favara. 2003. F. Capital. R. 17 (2).. Too much finance?. Arellano. 1984. 1988.. 1029–1054. 2012. L. 393–414. S. intermediaries. 26 (4). J. Financial intermediation and delegated monitoring.W. Dev. D.L. 135–158. Finance and growth: Schumpeter might be right... O. Demirguc-Kunt. World Dev. Levine. J. in order to understand the relationship between the financial system and economic growth more comprehensively. 1973. Econ. Lit. 1934. J.. A. J. R. J.D.G. References Allen.D. R. 257–292. The evolution of debt and equity markets in economic development. 1998. 22. Polit. The empirical results show that while the size of credit extended to the private sector and stock market liquidity exert strong positive impacts on the economic growth of the MICs. 25–51. Singh.. bank credit is not robust and stock market liquidity is only marginally significant. Market liquidity and performance monitoring.. Stud.A. Stud. Econ. Does financial liberalization spur growth? J.. 1995.. Does one size fit all? A reexamination of the finance and growth relationship.R. J. D. J. Inq.MICs..A. stock market liquidity is not a strong determinant of growth. 120 (547). A finite sample correction for the variance of linear efficient two-step GMM estimators. P. 87. 865–934. 277–297. Financ. and endogenous growth. 523–546. Initial conditions and moment restrictions in dynamic panel data models. B. 49 (1). Bencivenga.. Schumpeter. Banks. 1991.. Thakor. Financial development and economic growth: views and agenda. Rosen. J. 607–622. World Bank Econ..M. and the Business Cycle... Levine. Deidda. In the case of European HICs. Stat. Bond. 1371–1395.and lower. Financial markets. 115–143. Fattouh. Financ. J. our empirical study. Econometrica 50 (4). 35 (2). Financ. 2006.... Harvey. MA.. R. Money. B. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Brookings Institution. Stud. 535–569. 1997. Econ. 678–709. An empirical reassessment of the relationship between finance and growth. Levine.E. 1982. J. however. 8. Credit Bank. 1993.. Edward S. F.. F. Shaw. Interest. J.R. J. Washington. Econometrica 56 (6). Monet. Comparing Financial Systems. 58 (2). Bencivenga. B. IMF Working Paper.. 1950–2010. Financ. Estimating vector autoregressions with panel data. Beck. Macroecon. Econ. Presidential address: does finance benefit society? J. What is happening to the impact of financial deepening on economic growth? Econ. 17. Wachtel. Bank. finds that the growth effects of banks and stock markets differ for the various groups of economies... Rose. only stock market liquidity positively and strongly influences the economic growth of the HICs. Jovanovic. Smith. 1997. R. Kim Research in International Business and Finance xxx (xxxx) xxx–xxx as more homogeneous groups of countries are involved after dealing with simultaneity bias. 2013. P. MA. 1991. MIT Press.. 413–433. J. Jin. Econ. A. Econ. Econ.D. Rev. unlike bank credit. Financial development and economic growth.P. 133–152. Stiglitz.. B. 184–198. Credit. The generalization of the general theory. J.R. Financial system architecture and the co-evolution of banks and capital markets. 2004. W.. so that more economically reasonable and stylized evidence is uncovered regarding the link. E. 03/123. Elsevier... portfolio capital flows and long-term economic growth: micro and macroeconomic perspectives. 519–560.. 126. and the en- dogeneity issue. D. Tirole... 717–738. R2 around the world: new theory and new tests. 2005. Polit. Bekaert. Robinson.. F. Econ. 1973.. 68. it is vital to include more homogeneous economies in the panel. R.. The Theory of Economic Development: An Inquiry into Profits. Econom. Allen. Econ. 476–490. which uses more homogeneous income groups. Q. 67–142. Rev. Econom. Yale University Press.N.. Econ. T. B. Song. In general. Arcand. Feyen. Deidda. F. World Bank Group 27 (3). E. Thakor. L. Econ..V. omitted variable bias. DC. B. Lucas. 75.. 105 (3). Levine. technological choice. Starr. Non-linearity between finance and growth.L.. King. 2005. 1985. Emerging stock markets. 1993. Arellano. 195–209. Khan. Finance 28. Econ.. Econ. 2013. 34. Rev. Finance. Lett. implying that the link between financial development and economic growth seems to depend on the stages of the countries’ economic development. 68–89. J.. financial markets and growth. 51 (3). Financial Structure and Development.. Handbook of Economic Growth. 108 (3). S. G.. Levine.. Rousseau. Boot. Greenwood. Diversity of opinion and financing of new technologies. 2005. 2001. A.W. A. 77. 1161–1175. J.V.. C. and growth: panel evidence. 2010. 74. Weisse. Newey. J. N. Panizza. 1021–1055.. McKinnon. L. 1998. J. Econ. M.. 429–447. B. 2011. 276–288.T.. 1327–1363. F. Macmillan.. London. 2005. Myers.. Thus. J. Financial system architecture. 1988. and intertemporal smoothing..H.. Econ. Goldsmith. growth and the distribution of income. Financial development. 74 (2). 153–177. Finance and development: is Schumpeter's analysis still relevant? J. 31–51.. Econ. In sharp contrast to the existing empirical findings.. Coval. Allen. R.. 407–443. R.. A. Intermed. Rev. Econ. Bover. 2015. Finance and growth: theory and evidence. pp. Large sample properties of generalized method of moments estimators. Bull. Dyn. Stud. It is also found that bank credit spurs growth for both the upper. R. Blundell.W. Lee. Fufa. 79. Lundblad. 1998. 14 . V. P. J. The Rate of Interest and Other Essays. Bond. banks... 688–726. Holmstrom.W.. Financial Deepening in Economic Development.. 2004. The evolving importance of banks and securities markets. 29–52. Another look at the instrumental-variable estimation of error-components models. we employ the dynamic panel GMM estimator with relatively more homogeneous groups of countries. 1952. 101 (4). 2002. 49–77.. 5 (3). J. Account. Boyd. 1997.. Gale. 2008. Finance 70 (4). The bond market: An inflation-targeter's best friend. Smith. J... Credit markets and the control of capital. S. Smith. J.E. Econom. R. Intermed. On the mechanics of economic development. A new data set of educational attainment in the world. Barro... J.. Gale. Econ. Q. J. For this purpose. Theory 12. New Haven. 1993. our empirical results are consistent with the views that the financial system provides important services for economic growth and that stock markets and banks play different roles. Oxford University Press. In: Aghion. 3–42. L. 2009.K. Cambridge. Rioja. Zingales. 71 (1).C. growth and fragility: the role of government. A note on the theme of too many instruments.. For the non-European HICs.. Bhide.. A. 1991. The hidden costs of stock market liquidity. Barro.)... Harvard University Press.. Financial intermediation as a beliefs-bridge between optimists and pessimists.