The Impact of Banks and Stock Market Development on Economic Growth in South Africa: An ARDL-Bounds Testing Approach

This paper examines the impact of both bank- and market-based financial development on economic growth in South Africa during the period from 1980 to 2012. Unlike some previous studies, the current study employs means-removed average to construct both bank- and market-based financial development indices. The study uses the newly developed autoregressive distributed lag (ARDL) bounds testing approach to examine this linkage. The empirical results of this study show that there is a positive relationship between bank-based financial development and economic growth in South Africa. The results, however, fail to find any relationship between market-based financial development and economic growth in South Africa. The results apply irrespective of whether the regression analysis is conducted in the short run or in the long run. These results imply that it is bank-based financial development rather than market-based financial development that plays a pivotal role in propelling South Africa’s real sector.

This paper examines the impact of both bank-and market-based financial development on economic growth in South Africa during the period from 1980 to 2012. Unlike some previous studies, the current study employs means-removed average to construct both bank-and market-based financial development indices. The study uses the newly developed autoregressive distributed lag (ARDL) bounds testing approach to examine this linkage. The empirical results of this study show that there is a positive relationship between bank-based financial development and economic growth in South Africa. The results, however, fail to find any relationship between market-based financial development and economic growth in South Africa. The results apply irrespective of whether the regression analysis is conducted in the short run or in the long run. These results imply that it is bank-based financial development rather than market-based financial development that plays a pivotal role in propelling South Africa's real sector.

Introduction
The relationships among banking sector development, stock market development and economic growth have generated a considerable amount of debate for many years among development economists -but with little consensus. Although a growing body of work reflects the close relationship between financial development and economic growth (Gelb, 1989;King and Levine, 1993a;1993b;Roubini & Sala-i-Martin, 1992), alternative views nevertheless exist. Studies that support a positive relationship between financial development and economic growth include those of Schumpeter (2012), Goldsmith (1969), McKinnon (1973), Shaw (1973, King and Levine (1993a) and Odedokun (1996), among others. Studies that support a negative relationship include Van Wijnbergen (1983) and Buffie (1984). Apart from these two groups, there are studies that find either no association or a negligible relationship between financial development and economic growth. These include Robinson (1952), Lucas (1988) and Stern (1989).
Previous studies on this subject, however, suffer from four major limitations. First, the majority of previous studies relied on bank-based proxies of financial development, giving market-based proxies little attention. Where the latter studies have been undertaken, the empirical findings have been inconclusive (Akinlo & Akinlo, 2009;Levine & Zervos, 1996;Ujunwa & Salami, 2010), and the evidence suggests that the outcome is both country-and proxy-dependent. Second, the bulk of previous studies have been based mainly on the causal relationship between financial development and economic growth. Very few studies have examined in detail the relative impact of both bank-and market-based financial development on economic growth. Third, the majority of previous studies have mainly used either the residual-based cointegration test associated with Engle and Granger (1987) or the maximum likelihood test based on Johansen (1988) and Johansen and Juselius (1990).
Yet, it is now well known that these cointegration techniques may not be appropriate when the sample size is too small (see Odhiambo, 2009). Fourth, some of the previous studies were over-reliant on crosssectional data, which may not have satisfactorily addressed country-specific issues (Ghirmay, 2004).
It is against this backdrop that the current study attempts to examine the relative impact of bank-and market-based financial development on economic growth in South Africa using the newly developed autoregressive distributed lag (ARDL) bounds testing approach. To incorporate various proxies of bank-and market-based financial development in the empirical analysis, the current study employs the method of means-removed average to construct both bank-and market-based financial development indices.
The rest of this paper is organized as follows. Section 2 provides an overview of financial sector reforms in South Africa. Section 3 highlights the theoretical and empirical linkages between bank-and marketbased financial development and economic growth.
Section 4 presents the literature review, while section 5 addresses the empirical model specification, estimation technique, and empirical analysis of the regression results. Section 6 concludes.

Financial Sector Reforms in South Africa
The South African financial system is by far the largest, most developed, and most sophisticated in Africa, and it compares well with the financial systems of the developed world (Bank of    institutions -more than it is driven by financial markets -such as stock and bond markets -that economy's financial system is generally referred to as a bankbased financial system. If securities markets share center stage with banks in driving economic growth via savings mobilization and allocation, corporate control, and risk management, that economy's financial system is generally referred to as a market-based financial system.
Most economists still believe that a bank-based financial system is better than a market-based system.
In particular, it is argued that economic growth could be encouraged more in a bank-based system because it can induce longer-term investment in the real sector, whereas investment in a market-based system is too sensitive to stock market prices with short-term investment (Hoshi, Kashyap, & Scharfstein, 1990). The bank-based system can encourage productive investment because it is less affected by unstable financial markets. Even in recessions, the close relationship between banks and businesses can allow firms to continue investment without pushing them into bankruptcy (Odhiambo, 2011). Moreover, it is argued that expensive government policies can be carried out more easily in a bank-based system because it provides governments with more measures with which to intervene in the financial sector (such as credit policy and interest rate regulation) than a market-based system (Pollin, 1995).
However, the bank-based financial system is not without its own disadvantages. According to Odhiambo (2011), a bank-based system is vulnerable to problems, such as inefficient capital allocation, an intimate relationship between banks and firms and a higher debt ratio. Banks may not be effective gatherers or processors of information in new, uncertain situations involving innovative products and processes (Allen & Gale, 1999). This situation has prompted some to lend support to a market-based financial system, arguing that markets provide a richer set of risk-management tools that permit greater customization of risk-ameliorating instruments.
According to Levine (2004), while bank-based systems may provide inexpensive, basic risk-management services for standardized situations, market-based systems provide greater flexibility through tailor-made products. Thus, as economies mature and require a richer set of risk-management tools and vehicles for raising capital, they may benefit from a concomitant legal and regulatory environment that supports the evolution of market-based activities; otherwise, overall growth may be retarded (Levine, 2004).

Literature Review
The relationship between financial development and economic growth has received considerable attention in the empirical growth literature, and it is now widely acknowledged that bank-based financial development is positively associated with economic growth. Using the ratio of bank credit to private sector as a measure of bank-based financial development, De Gregorio and Guidotti (1995) examined the empirical relationship between economic growth and financial development in a large cross-country sample. Overall, they found that bank-based financial development is positively related to economic growth. However, its impact tends to vary across countries, and in panel data for Latin America, the relationship was found to be negative. Odedokun (1996) analyzed the effects of financial development on economic growth in 71 less developed countries (LDCs) using annual data over varying periods from the 1960s through to the 1980s. He found that financial development promotes economic growth in approximately 85% of these countries. However, a negative association between financial development and economic growth was revealed in at least 15% of the 71 countries he investigated.

Cointegration -ARDL Bounds Testing Procedure
This study utilizes the newly proposed ARDL bounds testing approach originally introduced by Pesaran and Shin (1999) and later extended by Pesaran, Shin and Smith (2001) to examine the cointegration relationship between bank-and market-based financial development and economic growth. This test has numerous advantages over previous cointegration tests, such as the residual-based technique by Engle and Granger (1987) and Full Maximum Likelihood (FML) test based on Johansen (1988; and on Johansen and Juselius (1990). First, unlike other conventional cointegration techniques, the ARDL bounds testing approach does not impose the restrictive assumption that all the variables under study must be integrated of the same order. In other words, the ARDL approach can be applied to test the existence of a relationship between variables regardless of whether the underlying regressors are integrated of order one [I(1)], order zero [I(0)], or fractionally integrated.
Second, while conventional cointegration methods estimate the long-run relationship within the context of a system of equations, the ARDL method employs only a single reduced form equation (Pesaran & Shin, 1999). Third, the ARDL technique generally provides unbiased estimates of the long-run model and valid t statistics -even when some of the regressors are endogenous (Odhiambo, 2008;Odhiambo, 2011). Fourth, while other cointegration techniques are sensitive to the size of the sample, the ARDL test is suitable even when the sample size is small. Thus, the ARDL test has superior small sample properties compared to the Johansen and Juselius (1990) cointegration test (Pesaran & Shin, 1999). Consequently, the approach is considered very suitable for analyzing the underlying relationship and it has been increasingly used in empirical research in recent years.
The empirical model used in this study to test the impact of financial development, both bank-and market-based, on economic growth is based on Ram (1999), Christopoulos and Tsionas (2004);Majid (2008); and Kargbo and Adamu (2009). The ARDL representation of the model is expressed as follows: 1 1

THE IMPACT OF BANKS & STOCK MARKET DEVELOPMENT …
(1) where GRO is the growth rate of real gross domestic product, a proxy for economic growth; BFD is an index of bank-based financial development, which is a means removed average of M2, M3 and credit provided to the private sector by financial intermediaries, a proxy for bank-based financial development (see also, Demirguc-Kunt and ; MFD is an index of market-based financial development, which is a means removed average of stock market capitalization, stock market traded value and stock market turnover, a proxy for market-based financial development (see also Demirguc-Kunt and Levine, 1996) of the investment in gross domestic product; SAV is a share of savings in gross domestic product; TOP is trade openness; α 0 is a constant; α 1 -α 6 and σ 1 -σ 6 are regression coefficients; ∆ is the difference operator; n is the lag length; and ɛ t is the white noise error term. The error correction model is specified as follows: (2)

Data Sources
This study utilized annual time-series data covering the period from 1980 to 2012. The annual data used in the study were obtained from the World Bank

Stationarity Tests
Before any analysis was made, the variables were first tested for stationarity using the Dickey-Fuller generalized least squares (DF-GLS) and Phillips-Perron (PP) tests. To attend to possible structural breaks within the dataset, the Perron 1997 unit root test (PPU-Root) was utilized as the third unit root testing method.
The DF-GLS lag length was selected automatically by SIC, the PP truncation lag was selected automatically on the Newey-West bandwidth, and the PPU-Root break years were also automatically selected; these dates ranged from 1987 to 2006, depending on the variable (see Appendix 1). The results of DF-GLS, PP and PPU-Root stationarity tests for the variables are presented in Table 1.
The results reported in Tables 1 show that after dif-ferencing the variables once, all the variables were confirmed to be stationary. Although the ARDL test does not require the pretesting of variables, the unit root test provides guidance as to whether ARDL is applicable because it is only applicable to the analysis of variables that are integrated of order zero [I(0)] or one [I(1)]. In this instance, the variables are conclusively stationary after being differenced once; hence, the ARDL bounds testing procedure can be satisfactorily performed.

Cointegration and ARDL-ECM Model
In this section, the long-run relationship between the variables in the general model is examined using the ARDL bounds testing procedure. The first step is to obtain the order of lags on the first differenced variables in equations (1) by using the Akaike Information Criterion and the Schwartz Bayesian Criterion. This is followed by the application of a bounds F-test to equation (1) to establish a long-run relationship between the variables under study. The results of the bounds Ftest are reported in Table 2.
The results of the F-test suggest that there exists a long-run relationship between GRO, BFD, MFD, INV, SAV and TOP. Following the estimation of the ARDL model and the use of AIC or SIC for optimal lag-length selection, the SIC-based ARDL (1,1,0,1,0,1) model was selected because it is more parsimonious than the AIC-based model. The long-run results of the selected model are reported in Table 3 Panel A and the shortrun results are reported in Table 3 Panel B.
The results displayed in Table 4 show that the model passes all the diagnostic tests performed for serial correlation, functional form, normality and heteroscedasticity. Figures 3 and 4 show plots of the cumulative sum of recursive residuals (CUSUM) and cumulative sum of squares of recursive residuals (CUSUMQ). The reported CUSUM and CUSUMQ are within the boundaries, showing that the model is stable and confirms the stability of the long-run coefficients of the regressors.

Conclusion
In this paper, we have examined the relative impact  Engle and Granger (1987) and the maximum likelihood test based on Johansen (1988) and Johansen and Juselius (1990). Yet, it is now known that these techniques may not be appropriate when the sample size is too small. Unlike the majority of previous studies, the current study uses the newly developed ARDL bounds testing approach to examine this linkage. In addition, the study employs the method of means removed averages to construct both bank-and marketbased financial development indices. The empirical results of this study show that there is a positive relationship between bank-based financial development and economic growth in South Africa. The results apply irrespective of whether the regression analysis is conducted over the short or long run. However, the results further reveal that there is no relationship between market-based financial development and eco-The Impact of Banks and Stock Market Development on Economic Growth in South Africa: an ARDL-bounds Testing Approach nomic growth in South Africa irrespective of whether the regression analysis is conducted over the short or long run. These results imply that it is bank-based financial development rather than market-based financial development that plays a pivotal role in propelling South Africa's real sector, both in the short run and in the long run.