Financial Development and Economic Growth in Sub-Saharan Africa Revisited: Disentangling the Role of Banks and Stock Markets

Abstract

The objective of this paper is to contribute to the analysis of the relationship between banks and stock markets in sub-Saharan African countries and their impacts on economic growth. While the literature on this issue is abundant, our article focuses on disentangling the effects of banks and stock markets on economic growth. Our approach is based on two indicators: the “activity–structure” variable, which measures the importance of stock markets in relation to banks, and the “activity–finance” variable, which takes into account the simultaneous development of banks and stock markets. Our empirical strategy is based on the estimation of a dynamic fixed effect model. The results indicate that the development of banks has a negative and significant impact, while the development of stock markets seems to have a positive influence on economic growth. It also shows that government authorities need to focus on developing stock markets at the expense of banks to promote economic growth.

1. Introduction

The world’s economies are usually hit by economic and financial crises. In 2007, the world witnessed a crisis that finally reached its peak on 15 September 2008 with the bankruptcy of the Lehman Brothers bank. This crisis was the most violent since that of 1929. It has shown the limits of the role of the financial system and its integration into the economy (Jacquet & Pollin, 2012). It should be noted that until the 1980s, the links between real and financial economics were analyzed relatively little in the empirical literature. Indeed, growth theories left no room for the role of the financial sector.

The pioneering work of Bagehot (1873) and Schumpeter (1934) provided early analyses of the relationship between financial development and economic growth. These two authors postulated that finance plays a key role in the process of economic development. Bagehot (1873) emphasized in his book “A Description of the Money Market” the links between financial development and economic growth. Indeed, he postulates that financial development promotes growth in that it mobilizes savings to finance various long-term investments. He then highlights the particular strength of the English financial market in terms of the relative ease of mobilizing savings to finance various long-term investment projects. This easy access for entrepreneurs is said to have contributed to the rapid industrialization of England. Schumpeter (1934) argued that the proper functioning of the financial system through the granting of bank credits was essential for economic growth since these improved productivity and strongly encouraged technological innovation.

Since the work of the two authors, many studies have attempted to examine the relationship between finance and economic development. These studies are not unanimous on the role of the financial sector in economic growth. Goldsmith (1969), McKinnon (1973) and Shaw (1973) confirmed the importance of the financial sector in economic development. Goldsmith (1969), who worked on data from 35 countries over the period 1860–1963, assumes that the size of the financial intermediation sector has a positive influence on the level of economic activity. McKinnon (1973) and Shaw (1973), continuing the investigations of Goldsmith (1969), showed the negative impact of financial repression (capped interest rates, financial protectionism, etc.), which would harm investment and, moreover, capital accumulation. More recently, on the other hand, for Robinson (1952), causality is reversed. In other words, it is rather financial development that is the result of economic growth. Other economists leave aside the contribution of the financial system to economic growth. This is the case of the economists of the classical school, who emphasize the quantity of capital, labor, or productivity and innovation to explain economic growth. For example, Lucas’s (1988) theory of economic development focuses on physical and human capital and leaves aside the dimension of financial development. The theoretical foundations and empirical studies on the relationship between financial and economic development do not lead to a uniform conclusion. There are also many variations across countries in the development of markets and financial intermediaries. As a result, financial systems rely on intermediaries and markets in varying ways.

In this respect, several questions, therefore, deserve to be asked. What are the links between banks and stock markets in sub-Saharan African countries? Is there a duality or complementarity between the banking system and the stock markets? What is the impact of the development of banks and stock markets on economic development?

Many studies have focused on the link between financial development and economic development (Kpodar, 2005; Mlambo, 2024; Yahyaoui & Rahmani, 2010). They suggest that financial development is necessary to achieve a more efficient allocation of resources and, consequently, to promote growth. However, the question of complementarity or substitutability between financial development and development has received less attention. The contribution of this article lies in examining the relationship of complementarity or substitutability between banks and stock markets in promoting growth.

The objective of this research is twofold. On the one hand, it examines the complementarity between banks and stock markets in sub-Saharan African countries, and on the other hand, it analyses the impact of these links on economic growth. The results of the article will go beyond the conclusions of Sène and Thiam (2018) to consider whether policies that promote financial development should be prioritized by the authorities and whether the focus should be on banks or stock exchanges.

In sub-Saharan African countries, financial development has not necessarily contributed to economic growth for a number of reasons, including the nature and orientation of credit distribution, bank concentration, and high interest rates. Added to this is the weakness of institutional, accounting, regulatory, and prudential systems. This justifies our choice of sub-Saharan African countries.

Our article is structured in four sections. The first is devoted to the theoretical and empirical review. The second part sets out the methodology adopted to measure the nature of the relationship between financial and economic development. The third focuses on the stylized facts relating to financial development. The results are presented and analyzed in the fourth section. The final section concludes and makes recommendations for economic policy.

2. Literature Review

This section presents the theoretical foundations and empirical studies on the links between finance and growth on the one hand and the separate effects of banks and financial markets on growth on the other.

Several studies on the nature of the relationship between finance and growth have shown the causal effect of financial development on growth (Bagehot, 1873; Goldsmith, 1969; McKinnon, 1973; Shaw, 1973).

Bagehot (1873) emphasized the mobilization of savings to finance various long-term investment projects. For the author, it was the particular strength of the English financial market that contributed to the rapid industrialization of England. Schumpeter (1934) evokes the decisive role of banks in economic growth. Indeed, he emphasizes the role of bankers, who, by targeting and financing entrepreneurs, encourage innovation and capital accumulation. Therefore, financial development would promote economic development through an increase in the investment rate and an efficient allocation of capital in the various most productive investment projects.

Other authors, such as Goldsmith (1969), McKinnon (1973), and Shaw (1973), followed the same logic as Schumpeter and provided other theoretical arguments. Goldsmith (1969) was interested in the causal impact of financial intermediaries on economic growth. Indeed, using data from 35 countries over a time horizon of 103 years (1860–1963), Goldsmith (1969) shows a positive correlation between the size of financial intermediation and economic growth. Subsequently, Goldsmith’s investigations were extended by Shaw (1973) and McKinnon (1973). The latter examined the effects of government intervention on financial development. They argue that the restrictions imposed by the authorities on the banking system are detrimental to investment and thus to capital accumulation and consequently reduce economic growth. These different authors did not use advanced econometric methods, nor did they even use causality analysis.

Several more recent studies have confirmed the significant impact of financial development on economic growth and use fairly advanced techniques (causality, taking into account simultaneity, etc.).

Kpodar (2005) attempted to study the link between financial development and economic growth in developing countries (DCs), taking into account the specific nature of sub-Saharan African countries. He used the method of generalized moments (GMM) in a dynamic panel on a sample of 64 developing countries (including 25 in sub-Saharan Africa) over the period 1968–1997. The results of the econometric estimations confirmed the hypothesis that the expansion of the financial system stimulates economic growth. In fact, a 1% increase in credit to the private sector as a proportion of GDP or in the ratio of the M3 aggregate would lead to a 0.33 percentage point increase in the growth rate of GDP per capita, compared with a 0.46 percentage point increase in the case of a 1% increase in the share of commercial bank assets in GDP. By adding interaction terms between the Africa dummy variable and each of the financial development indicators, he obtains that the impact is not homogeneous between the two groups of countries. He concludes that the impact of financial development is negligible for African countries, unlike other developing countries where the impact is positive and significant.

Khalouki (2016) set out to test whether economic growth is influenced by financial and banking development. Unlike Kpodar (2005), Khalouki used time series data for the case of Morocco. To test this relationship, he measures economic growth by GDP growth, capital accumulation, and total factor productivity (TFP). As indicators of financial development, it uses bank loans granted to the private sector (banking development) and two indicators of the development of financial intermediation (money supply in the M2 sense and loans granted by all financial institutions (including banks) to the private sector). Cointegration tests revealed a long-term relationship between economic development (GDP growth and total factor productivity (TFP)), banking development, and the development of financial intermediation. Econometric estimates reveal the existence of both short-run and long-run relationships between banking development, stock market development, and economic development.

In contrast to previous studies, some have focused on the role of a sound institutional framework in the relationship between financial development and economic growth.

This is the case of Yahyaoui and Rahmani (2010), who attempt to demonstrate the importance of a sound institutional framework in the relationship between financial development and economic growth in the case of 22 developing countries over the period 1990–2006. They adopt a theoretical Solow growth model augmented by human capital and examine the relationship between financial development, institutions, and economic growth. To estimate the model, they adopt different methods: traditional estimators (fixed or random effects), which can yield biased estimates in the presence of unit roots in the series, and estimators based on panel integration-cointegration analysis (fully modified least squares (FMOLS) and dynamic least squares (DOLS). All the techniques used (OLS, FMOLS, and DOLS) confirm that good governance provides a favorable environment for financial development and, consequently, economic growth.

Like Yahyaoui and Rahmani (2010), Kuipou-Toukam et al. (2015) examined the effect of financial development on growth through institutions, but in the Franc zone. Using the Generalized Moments method in a dynamic panel (GMM), they show that the direct effect of governance on growth in the Franc zone is insignificant. However, analysis by sub-region reveals that it has a positive and significant impact in the West African Economic and Monetary Union (WAEMU), while it has a positive and insignificant effect in the Central African Economic and Monetary Community (CAEMC) zone. They also find that financial development, as measured by private sector credit and the liquidity ratio, has a negative impact on economic growth in the CAEMC and WAEMU zones. The results of their analysis also show that credit from the banking sector has a positive impact on certain specifications. They also find that the combined effect of financial development and government remains insignificant in the CAEMC and WAEMU zones. Exceptionally, in the WAEMU zone, the combined variable of governance and credit from the banking sector has a negative and significant impact on growth. They, therefore, conclude that the institutional framework does not contribute to improving the impact of financial development on economic growth in the CAEMC and WAEMU zones.

The particularity Sène and Thiam’s (2018) study focuses on the separate effects of banks and financial markets on economic growth. Indeed, the author attempts to analyze, on the one hand, the relationship between banks and financial markets in sub-Saharan African countries and, on the other hand, their relative contributions to growth. Therefore, he used two indicators: the ‘activity–structure’ variable, which measures the importance of the equity market in relation to banks, and the ‘activity–finance’ variable, which takes into account the development of both banks and financial markets. The analysis covers seven sub-Saharan African countries (Côte d’Ivoire, Botswana, Ghana, Kenya, Mauritius, Nigeria and South Africa) and the period 1990 to 2011. Granger causality tests and extensions show that heterogeneous causality is rejected. The mean group (MG) estimator is therefore used to assess the long-term coefficients of the relationship. The results of the various estimations reveal the dominant impact of banks over financial markets on growth. The study also concludes that the two sources of financing for the economy are complementary in promoting economic growth.

However, other studies contradict the active role of finance in economic development. For these authors, financial development is the result of economic growth and not the opposite.

For these economists, the expansion of economic activity would push the real sector to demand funds from financial institutions to meet the increase in productivity. Consequently, growth in economic activity leads these institutions to intermediate and create money. In this vein, we find Robinson (1952), Kuznets (1955), Stern (1989), Singh (1997), and Beck et al. (2000). For Lucas (1988), the role of finance in growth is exaggerated. More recently, certain authors have arrived at this relationship between economic growth and financial development. These include Odhiambo (2009), who studied the causality between finance and growth in South Africa over the period 1960–2006. In the case of African and other low-income countries, Bist (2018) supports the hypothesis that growth promotes financial development.

Finally, Pinshi and Kabeya (2020) also tested the causal relationship between financial development and economic growth in the Democratic Republic of Congo (DRC) using data from 2004 to 2019. Using the Granger causality test, these authors show a unidirectional relationship between economic growth, measured by real gross domestic product (GDP), and financial development (measured by the ratio of bank credit to the private sector to GDP, the ratio of money supply to GDP and capital flows in the financial system). They conclude that any policy aimed at promoting economic growth is crucial to improving financial development in the DRC. More recent studies in sub-Saharan African countries include Mengesha and Berde (2023) in Ethiopia, which found a unidirectional causal link from growth to financial development, and Odhiambo and Nyasha (2022) in Ouganda, for whom the results are more mixed depending on the proxy used to measure financial development.

Aka (2010) sought to identify the nature of the relationship between financial development and economic growth in WAEMU countries over the period 1961–2005. His synthetic indicator is constructed from four indicators: monetary aggregate M2 over GDP, monetary aggregate M3 over GDP, domestic credit granted to the private sector divided by GDP, and domestic credit granted by banks over GDP. To analyze these links, he adopted a time series model for each country, performing cointegration and causality tests in the Granger sense. The results indicate a stable long-term relationship between financial development and economic growth in the WAEMU countries. Regarding the direction of causality, it runs from finance to economic growth for three countries (Guinea-Bissau, Niger, and Senegal) and is bidirectional for the other five countries. It concludes that any measure aimed at promoting financial development will stimulate economic growth. At the sectoral level, the results show cases of one-way causality, from finance to the sectors of activity, and cases of two-way causality. At the sectoral level, the results reveal cases of one-way causality, from finance to the sectors of activity, and cases of two-way causality. However, there are more cases of non-cointegration and non-causality between financial development and the agricultural sector. This could suggest that the former does not drain sufficient funds into the latter to enable its development.

Mlambo (2024) examines the relationship between financial development and economic growth in low-income nations in the SADC region. Using quantitative analysis and panel data, the results showed that there is a positive relationship between financial development and economic growth. The relationship was also found to be causal: financial development is not only a result of economic growth; it also influences growth. This calls on the governments in the countries under investigation to create environments that foster financial development.

The literature review shows that the nature of the relationship between finance and growth is not obvious and varies from one country to another or from one group of homogeneous countries to another.

3. Methodological Approach

This section describes the theoretical growth model used and the empirical model adopted to analyze the effects of financial development on economic growth.

3.1. Theoretical Model

Our objective in this section is to study the effects of financial development on economic growth. Therefore, we adopted the model of growth augmented by human capital presented by Mankiw et al. (1992).

We consider the following Cobb–Douglas production function:

$$Y\left(t\right)=K{\left(t\right)}^{\alpha }H{\left(t\right)}^{\beta }{\left(A\left(t\right)L\left(t\right)\right)}^{1-\alpha -\beta }$$

(1)

where Y is output, K is the capital factor, L is labor, A is the level of technology, and H is the stock of human capital.

The model assumes that output can be either consumed or invested and that capital depreciates at a constant rate (δ). The model also assumes that labor supply (L) grows exogenously at rate n.

$$L\left(t\right)=L\left(0\right)e^{nt}$$

(2)

The factor reflecting the technological level and efficiency of the economy is defined by

$$A\left(t\right)=A\left(0\right)e^{gt+W\left(t\right)\theta }$$

(3)

where g is the rate of technical progress assumed to be constant, W is a set of financial development factors and institutional policies and other factors affecting the level of technology and efficiency of the economy, and θ is a vector of coefficients relating these variables to the level of technology.

Considering $s_k$, the proportion of the proceeds invested in physical capital, and $s_h$, which is invested in human capital, is the variables describing the dynamics of economic development are determined by the following relationships:

$$\dot{k}\left(t\right)=s_{k}y\left(t\right)-(n+g+\delta )k\left(t\right)$$

(4)

$$\dot{k}\left(t\right)=s_{h}y\left(t\right)-\left(n+g+\delta \right)h\left(t\right)$$

(5)

where $y=Y/AL$, $y=K/AL$ et $h=H/AL$ are the quantities per effective unit of work.

The model also assumes that returns to scale are decreasing for all capital ($\alpha +\beta <1$).

The capital intensities (physical and human capital) associated with the long-run dynamic equilibrium are determined by solving the system formed by the equations: $\dot{k}\left(t\right)=0$ and $\dot{h}\left(t\right)=0$.

Solving the system implies that the economy converges to a steady state defined by

$$k^{\mathrm{*}}={\left({\displaystyle \frac{s_k^{1-\beta }{s}_h^{\beta }}{n+g+\gamma }}\right)}^{{\displaystyle \frac{1}{1-\alpha -\beta }}}$$

(6)

$$h^{\mathrm{*}}={\left({\displaystyle \frac{s_k^{\alpha }{s}_h^{1-\alpha }}{n+g+\gamma }}\right)}^{{\displaystyle \frac{1}{1-\alpha -\beta }}}$$

(7)

By substituting the equilibrium values into the production function and making a few transformations, we obtain the log value of the per capita income:

$$\ln \left[{\displaystyle \frac{Y\left(t\right)}{L\left(t\right)}}\right]=lnA\left(t\right)-{\displaystyle \frac{\alpha +\beta }{1-\alpha -\beta }}\ln \left(n+g+\gamma \right)+{\displaystyle \frac{\alpha }{1-\alpha -\beta }}\mathrm{l}\mathrm{n}(s_k)+{\displaystyle \frac{\beta }{1-\alpha -\beta }}\ln \left(s_h\right)$$

(8)

The equation translates the impact of demographic growth and the accumulation of physical and human capital on per capita income. Like Solow’s original model, the model predicts that high population growth reduces per capita income, while investment in physical and human capital has a positive impact on per capita income.

Solow’s (1956) model predicts that differences in real per capita income tend to narrow between countries. Solow (1956) found that poorer countries grew faster than richer ones. For him, countries converge towards the same income in the long term in the presence of diminishing returns and perfect technology diffusion. On the other hand, endogenous growth models characterized by the hypothesis of non-decreasing returns to factors of production imply that countries that invest more will grow indefinitely.

The equation translates the impact of demographic growth and the accumulation of physical and human capital on per capita income. Like Solow’s original model, the model predicts that high population growth reduces per capita income, while investment in physical and human capital has a positive impact on per capita income.

Mankiw et al. (1992) also analysed the speed of convergence to the steady state. Assuming $y^{*}$ income per worker in the stationary state and $y\left(t\right)$ the income at time t, the speed of convergence is given by

$${\displaystyle \frac{dln\left(y\left(t\right)\right)}{dt}}=-\lambda [\ln \left(y\left(t\right)\right)-\ln \left(y^{\mathrm{*}}\left(t\right)\right)]$$

(9)

With $\lambda =(1-\alpha -\beta )(\delta +n+g)$.

This Equation (1) is a differential equation in ln(y) whose solution is given by

$$\ln \left(y\left(t\right)\right)=\left(1-e^{-\lambda t}\right)\ln \left(y^{\mathrm{*}}\right)+e^{-\lambda t}\ln \left(y\left(0\right)\right)$$

(10)

To obtain the income growth for a period, we use the following expression:

$$\ln \left(y\left(t\right)\right)=\left(1-e^{-\lambda }\right)\ln \left(y^{\mathrm{*}}\right)+e^{-\lambda }\ln \left(y\left(t-1\right)\right)$$

(11)

By substracting $\ln \left(y\left(t-1\right)\right)$ from both parts of the equation, we obtain the expression for the growth of income per capita:cap:

$$\ln \left(y\left(t\right)\right)-\ln \left(y\left(t-1\right)\right)=\left(1-e^{-\lambda }\right)\ln \left(y^{\mathrm{*}}\right)-(1-e^{-\lambda })\ln \left(y\left(t-1\right)\right)$$

(12)

Substituting $\ln \left(y^{*}\right)$, we have

$$\begin{array}{l}\mathsf{\Delta }\ln \left(y\left(t\right)\right)=-\left(1-e^{-\lambda }\right)\ln \left(y\left(t-1\right)\right)+\left(1-e^{-\lambda }\right)[\ln \left(A\left(t\right)\right)-{\displaystyle \frac{\alpha +\beta }{1-\alpha -\beta }}\ln \left(n+g+\gamma \right)+\\ {\displaystyle \frac{\alpha }{1-\alpha -\beta }}\ln \left(s_k\right)+{\displaystyle \frac{\beta }{1-\alpha -\beta }}\ln \left(s_h\right)]\end{array}$$

(13)

Using the equation for $lnA\left(t\right)$, we have

$$\ln \left(A\left(t\right)\right)=lnA\left(0\right)+gt+\theta W\left(t\right)$$

(14)

By replacing $lnA\left(t\right)$ by its expression in ($\mathsf{\Delta }\ln \left(y\left(t\right)\right)$), we obtain

$$\begin{array}{l}\mathsf{\Delta }\ln \left(y\left(t\right)\right)=\varphi \ln \left(y\left(t-1\right)\right)-\varphi \ln \left(A\left(0\right)\right)-\varphi gt-\varphi \theta W\left(t\right)+{\displaystyle \frac{\varphi \left(\alpha +\beta \right)}{1-\alpha -\beta }}\ln \left(n+g+\gamma \right)-\\ {\displaystyle \frac{\varphi \alpha }{1-\alpha -\beta }}\ln \left(s_k\right)-{\displaystyle \frac{\varphi \beta }{1-\alpha -\beta }}\ln \left(s_h\right)\end{array}$$

(15)

where $\varphi =-(1-e^{-\lambda })$ et $\lambda =(1-\alpha -\beta )(\delta +n+g)$.

Simplifying this equation gives the following equation:

$$\begin{array}{l}\mathsf{\Delta }\ln \left(y\left(t\right)\right)={\eta }_0+{\eta }_1\ln \left(y\left(t-1\right)\right)+{\eta }_{2}t+{\eta }_{3}W\left(t\right)+{\eta }_4\ln \left(s_k\right)+{\eta }_5\ln \left(s_h\right)+\\ {\eta }_6\ln \left(n+g+\gamma \right)\end{array}$$

(16)

With

$$\begin{gathered} {\eta }_0=-\varphi \ln \left(A\left(0\right)\right) \\ {\eta }_1=\varphi \\ {\eta }_2=-\varphi g \\ {\eta }_3=-\varphi \theta \\ {\eta }_4=-{\displaystyle \frac{\varphi \alpha }{1-\alpha -\beta }} \\ {\eta }_5=-{\displaystyle \frac{\varphi \beta }{1-\alpha -\beta }} \\ {\eta }_6={\displaystyle \frac{\varphi \left(\alpha +\beta \right)}{1-\alpha -\beta }} \end{gathered}$$

3.2. Empirical Model and Estimation Strategy

3.2.1. Empirical Model

Based on Solow’s theoretical model augmented by human capital presented in the previous section and the various empirical works cited above, our model is specified in two ways: firstly, to analyze the separate effects of banks and stock markets on growth, and secondly, to assess the complementarity or substitutability between banks and stock markets in explaining growth.

First model: separate effects of banks and financial markets on growth.

$$\begin{array}{l}\mathsf{\Delta }\mathit{ln}\left(y_t\right)={\beta }_0+{\beta }_1\ln \left(y_{t-1}\right)+{\beta }_2\mathit{ln}\left(bank\right)+{\beta }_3\mathit{ln}\left(stock\text{\_}exc\right)+{\beta }_4\mathit{ln}\left(inv\right)+{\beta }_5\mathit{ln}\left(educ\right)+\\ {\beta }_{6}pop+{\beta }_7\mathit{ln}\left(trade\right)+{\beta }_8\mathit{ln}\left(dep\right)+{\beta }_{9}inflation+{\beta }_{10}institution+\epsilon \end{array}$$

(17)

where the following apply:

✓

$y_t:$ per capita income at time $t$;

✓

$y_{t-1}:$ per capita income in the previous period;

✓

$bank:$ financial development indicator for banks;

✓

$stock\text{\_}exc:$ financial development indicator for stock markets;

✓

$inv:$ ratio of private investment to GDP;

✓

$educ:$ level of human capital;

✓

$trade:$ of trade openness measured by (exports + imports) to GDP;

✓

$dep:$ public spending as a percentage of GDP;

✓

$inflation:$ inflation rate;

✓

$pop:$ population growth rate;

✓

Institution: indicator measuring the quality of institutions or good governance includes two variables stabilite_pol (political stability) and efficacite_gov (government effectiveness).

This first model measures the separate impact of banks and financial markets on economic growth.

Second model: banks and stock markets: complementarity or substitutability

$$\begin{array}{l}\mathsf{\Delta }\mathit{ln}\left(y_t\right)={\beta }_0+{\beta }_1\ln \left(y_{t-1}\right)+{\beta }_2\mathit{ln}\left(struc\right)+{\beta }_3\mathit{ln}\left(fin\right)+{\beta }_4\mathit{ln}\left(inv\right)+{\beta }_5\mathit{ln}\left(educ\right)+\\ {\beta }_{6}pop+{\beta }_7\mathit{ln}\left(trade\right)+{\beta }_8\mathit{ln}\left(dep\right)+{\beta }_{9}inflation+{\beta }_{10}institution+\epsilon \end{array}$$

(18)

where the following applies:

✓

$struc:$ represents the financial structure variable “activity–structure”: this is the ratio of market capitalization to total bank assets;

✓

$fin:$ “activity–finance”, market capitalization plus total bank assets.

This second model makes it possible to study the complementarity or substitutability between banks and financial markets in explaining the growth of countries.

The activity–structure indicator provides insight into the relative structure of the financial market by assessing the weight of the stock market in relation to the banking sector. This measure thus enables an evaluation of the relative importance of market-based financing compared to intermediary-based financing, offering a structural perspective on the financial configuration. Conversely, the second indicator termed the activity–finance indicator, corresponds to the sum of stock market capitalization and total banking assets. It provides an aggregate measure of the overall size of the financial sector within the economy, capturing the intensity of financial development regardless of the financing channel. The simultaneous inclusion of these two variables in the econometric regression allows for a distinction between the specific effects of financial structure, on the one hand, and the overall magnitude of financial activity, on the other, on economic growth.

3.2.2. Estimation Strategy

Panel data modeling is used to estimate these empirical models. Panel data econometrics involves carrying out an analysis using both time-series and cross-sectional data. It has advantages over cross-sectional and time-series approaches. While time series data allows dynamic behavior to be analyzed, it does not allow heterogeneity between individuals to be taken into account. Conversely, cross-sectional data takes heterogeneity into account but does not allow dynamic behavior to be analyzed. Panel data, on the other hand, makes it possible to analyze both aspects: heterogeneity between individuals and dynamic behavior.

In addition, by using panel data, the researcher works with more observations, leading to convergent estimators and asymptotic properties. The large number of observations means that estimators are more accurate, and the bias and variance of estimators tend toward zero.

In addition, this approach also makes it possible to control for unobserved heterogeneity—a potential source of endogeneity discussed in the literature.

Because of all its advantages, like Sène and Thiam (2018), Kuipou-Toukam et al. (2015), Aka (2010), and Yahyaoui and Rahmani (2010), panel data estimation is preferred to determine the nature of the links between finance and economic growth.

✓

Indicators for measuring financial development.

The choice of indicators for measuring financial development plays a crucial role. To study the relationship between finance and economic development, it is crucial to discuss the choice of indicators for measuring financial development. In most studies, indicators of size and activity are used.

✓

Indicators for measuring bank size and activity.

○

Size measures based on bank assets.

Like Sène and Thiam (2018), we use two approaches to measure bank size: a relative and an absolute approach. The relative measure is defined as the assets of deposit banks relative to the total financial assets (i.e., the combined assets of deposit banks and the central bank), while the absolute measure corresponds to the assets of deposit banks expressed as a share of GDP.

To measure financial development according to the relative approach, deposit bank assets are used as a ratio of total financial assets (banks + central bank). This variable has been used in several studies (King & Levine, 1993; Beck et al., 2000). According to the absolute approach, two indicators are used: deposit bank assets relative to GDP and banking system assets (banks + central bank) relative to GDP. This approach was used by Kpodar (2005) when he used the share of commercial bank assets in GDP to measure financial development.

○

Measuring activity on the basis of loans.

Bank activity is assessed on the basis of bank loans. Studies are mainly based on two indicators measuring banking activity. These are private lending by deposit banks divided by GDP and private lending by the banking system (banks + central bank) divided by GDP.

✓

Indicators for measuring the development of financial markets.

Several indicators are used to measure the development of financial markets. The indicators used in this study relate to the size, activity, and efficiency of the equity market:

➢

Measure of equity market size: market capitalization as a percentage of GDP. It represents the value of shares listed on the market in relation to GDP.

➢

Measure of equity market activity: total value traded of the equity market over GDP.

➢

Measure of equity market efficiency: this is the equity market turnover ratio. It is calculated by dividing the total value of shares traded by market capitalization. It is used to assess market activity relative to market size. For example, a small but active market will have a very high turnover.

✓

Indicators for measuring the structure of financial systems.

One of the aims of this study is to analyze the separate effects of banks and stock markets on growth in Sub-Saharan African countries. It also aims to verify the existence of a complementary relationship between the two sources of financing for these economies. To answer this question, we introduce into the model, like Sène and Thiam (2018) and Beck (2003), two indicators of summary measures of financial development. The first indicator measures the importance of the equity market in relation to banks, while the second takes into account both the development of banks and financial markets. These two indicators will help determine whether the authorities in these countries should develop stock markets at the expense of banks.

3.3. Data

The sample covers the period 1995 to 2015 for the following countries: Côte d’Ivoire, Botswana, Ghana, Kenya, Mauritius, Nigeria, and South Africa. This choice is justified by the fact that certain stock market data for the chosen period are not available for all countries. The data were collected from the World Bank (Global Financial Development Database, World Development Indicators), the International Monetary Fund (IMF, International Financial Statistics), and the National Institute of Statistics for these countries.

4. Stylized Facts

This section is devoted to analyzing the characteristics of the two sources of financing economic activity (banks and financial markets).

4.1. Analysis of Financial Markets

Two indicators are used to compare the financial development of the countries in our study: market capitalization and turnover ratio.

Defined as the value of listed shares in relation to GDP, market capitalization is obtained by multiplying the number of shares making up a company’s capital by their stock market price.

Table 1. Market capitalization.

Country	1995–1999	2000–2004	2005–2009	2010–2015	1995–2015
South Africa	153.6	146.8	220.0	232.3	190.2
Botswana	10.3	23.8	35.5	29.4	25.0
Côte d’Ivoire	7.4	9.1	24.7	31.0	18.7
Ghana	19.6	6.9	6.6	7.9	10.1
Kenya	17.1	14.4	33.5	28.1	23.5
Mauritius	34.4	26.4	58.0	67.9	47.7
Nigeria	17.2	8.5	15.7	11.5	13.1
Together	37.1	33.7	56.3	58.3	46.9

Source: authors’ calculations from the Global Financial Development Database (World Bank).

shows the average market capitalizations of the various countries over the period 1995–2015. Over the entire period, South Africa occupies a prominent position in terms of market capitalization. On average, it has the highest market capitalization (190.2%). Mauritius came second (47.7%). This significant difference reveals the gap in financial market development between these Sub-Saharan African countries. Ghana has the lowest average (10.1%).

South Africa, Côte d’Ivoire, and Mauritius also showed an upward trend over the period, while Ghana saw a huge drop, from an average capitalization of 19.6% between 1995 and 1999 to an average of 7.9% between 2010 and 2015.

The turnover ratio is a measure of equity market efficiency calculated by dividing the total value of shares traded by market capitalization. It enables us to assess the activity of financial markets in relation to their size. For example, a small but active market will have a very high turnover.

Table 2. Turnover ratios.

Country	1995–1999	2000–2004	2005–2009	2010–2015	1995–2015
South Africa	20.1	25.5	30.4	28.1	26.1
Botswana	9.5	4.2	2.5	3.1	4.7
Côte d’Ivoire	14.3	2.2	2.9	3.1	5.5
Ghana	2.3	6.0	7.9	4.5	5.1
Kenya	4.0	4.7	8.2	7.0	6.0
Mauritius	6.1	6.4	5.7	5.1	5.8
Nigeria	3.2	9.9	17.6	9.0	9.9
Together	8.5	8.4	10.7	8.6	9.0

Source: authors’ calculations from the Global Financial Development Database (World Bank).

below shows the averages of these ratios over the study period. It shows that South Africa has the most efficient financial markets, with an average of 26.1 over the whole period. Nigeria ranks second in terms of financial market efficiency (9.9). By contrast, Botswana (4.7), Ghana (5.1), and Côte d’Ivoire (5.8) have the least efficient financial markets.

4.2. Analysis of the Banking System

The financial systems of African countries are characterized by the preponderance of banks over financial markets. The analysis of the banking systems of the countries in the sample is based on two indicators: assets of deposit banks as a percentage of GDP and private loans of deposit banks as a percentage of GDP.

✓

Deposit bank assets as a percentage of GDP.

Table 3. Deposit bank assets as a percentage of GDP.

Country	1995–1999	2000–2004	2005–2009	2010–2015	1995–2015
South Africa	63.5	67.6	76.5	77.5	71.6
Botswana	11.6	18.6	23.3	30.8	21.6
Côte d’Ivoire	22.1	18.1	18.4	23.3	20.6
Ghana	7.2	14.7	19.3	24.6	16.8
Kenya	25.8	31.3	31.8	40.9	32.9
Mauritius	64.6	76.9	94.6	112.4	88.3
Nigeria	8.5	11.4	17.4	19.6	14.5
Together	29.1	34.1	40.2	47.0	38.0

Source: authors’ calculations from the Global Financial Development Database.

shows average deposit bank assets as a percentage of GDP over the period 1995–2015. It shows that, on average, over the whole period, the highest averages are recorded for the Republic of Mauritius (88.3%) and South Africa (71.6%). The other countries in the sample have low averages (less than 35.0% for all countries).

✓

Private loans from deposit banks as a percentage of GDP.

Private credit from deposit banks represents the financial resources provided to the private sector by deposit banks in the form of loans, purchases of non-equity securities, trade credits, and other claims. The availability of credit to the private sector attests to the development of the banking and financial sector in each country.

Table 4. Private loans as a percentage of GDP.

Country	1995–1999	2000–2004	2005–2009	2010–2015	1995–2015
South Africa	58.7	61.6	69.4	66.5	64.2
Botswana	10.8	16.9	21.8	28.6	20.0
Côte d’Ivoire	16.2	13.8	14.3	17.2	15.5
Ghana	4.2	6.9	11.6	14.7	9.6
Kenya	19.3	22.4	22.7	28.7	23.5
Mauritius	47.4	58.2	71.3	93.0	68.7
Nigeria	6.9	8.0	13.0	13.3	10.5
Together	23.4	26.8	32.0	37.4	30.3

Source: authors’ calculations from the Global Financial Development Database (World Bank).

shows the average private credit as a percentage of GDP over the period 1995–2015. It shows that, on average, the Republic of Mauritius (68.7%) and South Africa (64.2%) offer the highest private credit from deposit banks as a percentage of GDP, while Nigeria (10.5%) and Ghana (9.6%) have the lowest ratios.

4.3. Main Economic Indicators

The real GDP growth rate measures the evolution of real GDP, which is an economic indicator measuring the wealth produced in a country over a given period.

Over the study period, Ghana (5.7%) and Nigeria (5.7%) recorded the highest average growth rates (

Table 5. Real GDP growth rate.

Country	1995–1999	2000–2004	2005–2009	2010–2015	1995–2015
South Africa	2.6	3.6	3.6	2.3	3.0
Botswana	6.3	3.1	4.0	5.5	4.7
Côte d’Ivoire	5.0	−0.7	2.2	5.8	3.2
Ghana	4.4	4.6	6.1	7.3	5.7
Kenya	2.9	2.6	4.6	6.0	4.1
Mauritius	4.8	4.7	4.2	3.8	4.3
Nigeria	2.0	8.6	6.8	5.5	5.7
Together	4.0	3.8	4.5	5.2	4.4

Source: authors’ calculations from the World Development Indicators (World Bank).

). South Africa, with its relatively more developed financial system, had the lowest average growth rate (3.0%).

In most empirical studies, GDP per capita is typically employed to compare GDP across countries with varying population sizes.

Table 6. Average real GDP growth rates per capita.

Country	1995–1999	2000–2004	2005–2009	2010–2015	1995–2015
South Africa	0.8	2.3	2.2	0.8	1.5
Botswana	3.8	1.2	1.9	4.0	2.8
Côte d’Ivoire	1.8	−3.0	−0.1	3.3	0.6
Ghana	1.8	2.1	3.5	4.8	3.1
Kenya	0.0	−0.2	1.7	3.3	1.3
Mauritius	3.7	3.9	3.8	3.6	3.7
Nigeria	−0.5	5.9	4.0	2.7	3.0
Together	1.6	1.7	2.4	3.2	2.3

Source: authors’ calculations from the World Development Indicators (World Bank).

shows the average real GDP per capita growth rates for the countries in the sample. When population dynamics are taken into account, Mauritius has the highest average growth rate (3.7%). This is followed by Ghana (3.1%) and Nigeria (3.0%). Côte d’Ivoire recorded the lowest average growth rate over the entire period (0.6%).

5. Analysis and Interpretation of Results

The econometric estimation process begins with a study of the different series. This involves testing the unit root hypothesis and any cointegration and causality between the series.

5.1. Unit Root Test

The usual panel data techniques require the assumption that the various variables are stationary. By combining the individual dimension with the time dimension, stationarity problems may arise, as in the case of time series. Failure to take this assumption into account would bias the estimates.

There is a vast amount of literature on unit root tests. The first (first-generation) tests take into account individual heterogeneity and the shapes of individual dynamics (i.e., parameter heterogeneity). In this line of tests, we find the tests of Levin and Lin (1992), Im et al. (1997), and Maddala and Wu (1999). The second generation of tests challenges the hypothesis of independence between individuals. These are the tests of Bai and Ng (2001), Moon and Perron (2004), and Pesaran (2007).

The individual dependence test makes it possible to discriminate between the first and second-generation tests.

5.1.1. Individual Dependency Test

To test the inter-individual dependence of each variable, the Pesaran (2007) test is used.

Table 7. Correlation coefficients—inter-individual dependence test.

Variable	CD-Test	p-Value	Mean ρ	Mean abs (ρ)	Conclusion
L_gdp_cap_2010	12.065	0.000	0.57	0.76	Dependence
L_investment	−1.336	0.182	−0.06	0.43	Independence
L1_exp_gov	−0.493	0.622	−0.02	0.31	Independence
L_nb_educ_expected	14.240	0.000	0.68	0.68	Dependence
L_nb_average_educ	19.256	0.000	0.92	0.92	Dependence
Inflation	6.602	0.000	0.31	0.37	Dependence
L_trade	0.766	0.444	0.04	0.26	Independence
efficacite_gov	−1.342	0.179	−0.06	0.34	Independence
stabilite_pol	−0.569	0.569	−0.03	0.46	Independence
L_M2GDP	9.744	0.000	0.46	0.47	Dependence
L_asset_banq_gdp	13.375	0.000	0.64	0.65	Dependence
L_credit_prive_banq	13.519	0.000	0.64	0.64	Dependence
L_credit_prive_all_banq	14.567	0.000	0.69	0.69	Dependence
L_stock_exc	5.672	0.000	0.27	0.49	Dependence
L_turnover	−1.009	0.313	−0.05	0.42	Independence
L1_struc	2.697	0.007	0.13	0.35	Dependence
L_fin	14.859	0.000	0.71	0.71	Dependence

Source: authors’ calculations.

shows the results of this test based on the null hypothesis of cross-sectional independence. The null hypothesis assumes cross-sectional independence. The test is based on Pesaran’s CD statistic, which is the average of the correlation coefficients between the different countries taken in pairs for each time period. Under the null hypothesis, this statistic is asymptotically distributed according to a centered reduced normal distribution ℕ(0, 1). The table below also shows the average correlation coefficients $Moy.\rho$ and that of the absolute value of the correlation coefficients by $Moy.\left|\rho \right|$.

The results indicate the existence of a strong cross-sectional dependence for the seven (07) countries. In fact, of the eighteen variables, only six showed inter-individual independence.

5.1.2. First and Second Generation Unit Root Test

Given that the results of the individual dependence test reveal strong individual dependence in the data, it is essential to use second-generation unit root tests. In this study, the Pesaran (2007) test is used. In addition to taking inter-individual dependence into account, this test assumes parameter heterogeneity. The null hypothesis of this test is the presence of a unit root in each individual time series, while the alternative hypothesis is that some of the individual series are stationary.

Table 8. Pesaran tests.

Variables	Dependence	In Levels		In First Differences		Conclusion
		CIPS	5%	CIPS	5%
L_gdp_cap_2010	Dependence	−1.765	−2.86	−3.074	−2.34	I(1)
L_investment	Independence	−2.863	−2.86	-	-	I(0)
L1_exp_gov	Independence	−2.000	−2.86	−3.945	−2.34	I(1)
L_nb_educ_expected	Dependence	−2.058	−2.86	−3.502	−2.34	I(1)
L_nb_average_educ	Dependence	−2.332	−2.86	−3.825	−2.34	I(1)
Inflation	Dependence	−3.947	−2.86	-	-	I(0)
L_trade	Independence	−2.389	−2.86	−4.065	−2.34	I(1)
efficacite_gov	Independence	−3.067	−2.86	-	-	I(0)
stabilite_pol	Independence	−3.348	−2.86	-	-	I(0)
L_M2GDP	Dependence	−1.564	−2.86	−3.613	−2.34	I(1)
L_asset_banq_gdp	Dependence	−1.671	−2.86	−2.791	−2.34	I(1)
L_credit_prive_banq	Dependence	−1.589	−2.86	−2.817	−2.34	I(1)
L_credit_prive_all_banq	Dependence	−1.677	−2.86	−2.748	−2.34	I(1)
L_stock_exc	Dependence	−2.682	−2.86	−3.583	−2.34	I(1)
L_turnover	Independence	−3.197	−2.86	-	-	I(0)
L1_struc	Dependence	−2.557	−2.86	−3.743	−2.34	I(1)
L_fin	Dependence	−2.634	−2.86	−4.052	−2.34	I(1)

Source: authors’ calculations.

presents the CIPS statistic and the critical values (at a threshold of 5%) of the Pesaran (2007) test.

The results of the Pesaran (2007) test indicate that all the variables respecting the hypothesis of individual dependence are integrated of order 1 except inflation, which is stationary in level.

5.1.3. First-Generation Test

The first category of unit root tests is based on the assumption of cross-sectional independence of observations. In this study, three first-generation tests are performed: Levin-Lin and Chu (LLC); d’Im, Pesaran and Shin (IPS); and Maddala and Wu.

The LLC test assumes the homogeneity of the autoregressive root under the alternative hypothesis. However, if the null hypothesis of the presence of a unit root is rejected, it is unlikely that the hypothesis of an autoregressive root identical to all the individuals in the panel will be accepted. The IPS test, therefore, remedies this limitation by taking into account not only heterogeneity in the autoregressive root but also heterogeneity in the presence of a unit root in the panel.

In contrast, Maddala and Wu (1999) used a non-parametric Fisher test. It is based on a combination of the significance levels (p-values) of N-independent individual unit root tests. Like the test proposed by IPS, this test does not rely on the restrictive hypothesis of the LLC test, according to which the autoregressive coefficient is identical for all individuals.

The results of the three tests are summarised in

Table 9. Independence test.

Variable	LLC		IPS		Fisher		Conclusion
	Adjusted t *	p-Value	Statistic	p-Value	Statistic	p-Value
In levels
L_investment	−3.9230	0.0000	−1.6546	0.0490	1.8464	0.9676	Stationary
L1_exp_gov	18.0628	1.0000	1.0632	0.8562	1.4201	0.9222	Non-Stationary
L_trade	18.6025	1.0000	−0.4937	0.3108	0.3230	0.6266	Non-Stationary
efficacite_gov	−0.5096	0.3052	1.1995	0.8848	0.8394	0.7994	Non-Stationary
stabilite_pol	−2.7302	0.0032	−0.0226	0.4910	1.3928	0.9182	Non-Stationary
L_turnover	−11.7957	0.0000	−5.3889	0.0000	−3.4602	0.0003	Stationary
In differences
L1_exp_gov	6.0758	1.0000	−6.7010	0.0000	−10.1285	0.0000	Stationary
L_trade	8.1942	1.0000	−7.8656	0.0000	−11.3055	0.0000	Stationary
efficacite_gov	3.1778	0.9993	−8.1638	0.0000	−12.7227	0.0000	Stationary
stabilite_pol	−1.0847	0.1390	−5.9640	0.0000	−10.1658	0.0000	Stationary

Source: authors’ calculations. * indicate that statistically significant at 10%.

. Of the six variables respecting inter-individual independence, only two are stationary in level.

5.1.4. Granger Causality Test

We now perform Granger causality tests to highlight the bidirectional relationships between certain financial development variables and growth (

Table 10. Causality test.

Sense of Causality	W-Bar	Z-Bar	Z-Bar Tilde
asset to GDP	4.3616	0.3383	−0.4777
GDP to asset	4.5915	0.5533	−0.3890
credit to GDP	3.3460	−0.6117	−0.8695
GDP to credit	5.9169	1.7931	0.1222
stock_exc to GDP	11.7767	7.2745 **	2.3827 **
GDP to stock_exc	7.8496	3.6009	0.8678
Exchange val to GDP	2.6627	−1.2509	−1.1331
GDP to exchange val	6.7514	2.5737	0.4442
turnover to GDP	3.3734	−0.5861	−0.8589
GDP to turnover	5.2859	1.2029	−0.1212
structure to GDP	17.3760	12.5121 **	4.5427 **
GDP to structure	6.8680	2.6828	0.4891
finance to GDP	6.3333	2.1826	0.2829
GDP to finance	6.1702	2.0301	0.2200

Source: authors’ calculations. *, **, and *** indicate that the estimated coefficients are statistically significant at 10%, 5%, and 1% levels, respectively.

). The results indicate only two causal relationships, the first running from market capitalization to gross domestic product and the second from structure to gross domestic product.

6. Cointegration Test

In the presence of non-stationary variables, it is advisable to study possible cointegration relationships between these variables. Indeed, when applying the usual estimation methods, two main problems could arise: the problem of spurious regressions and the problem of the invalidity of certain asymptotic laws.

In this study, we perform the Pedroni (1999, 2004) cointegration test. Pedroni proposes several tests to verify the null hypothesis of the absence of cointegrating relationships for all the individuals in the panel by allowing heterogeneity in the parameters for each of them. He develops two categories of tests. The first, consisting of four tests, assumes coefficient homogeneity, while the second category of tests (group tests), consisting of three tests, assumes heterogeneity.

Table 11. Cointegration tests.

Market capitalization and assets of deposit banks as % of GDP
Test Stats.	Panel	Group
V	1.484
Rho	3.165	3.916
T	−3.421	−6.579
Adf	7.497	10.59
Turnover and assets of deposit banks as % of GDP
Test Stats.	Panel	Group
v	1.234
rho	2.75	3.542
t	−2.267	−3.624
adf	2.936	3.473
Total stock market value and assets of deposit banks as % of GDP
Test Stats.	Panel	Group
v	1.487
rho	3.172	3.988
t	−2.719	−5.222
adf	6.397	8.008
Stock market capitalization and private credit of deposit banks as a % of GDP
Test Stats.	Panel	Group
v	1.586
rho	3.392	4.217
t	−2.319	−4.196
adf	6.577	8.464
Structure and financial activity
Test Stats.	Panel	Group
v	1.257
rho	3.219	4.033
t	−2.72	−4.879
adf	1.814	4.187

Source: authors’ calculations.

presents the results of the seven different Pedroni (1999, 2004) cointegration tests. A linear trend was added for these tests.

It should be remembered that under the null hypothesis of non-cointegration, all these statistics are distributed according to the normal distribution N (0, 1). For the five models, most of the statistics confirm the presence of cointegration relationships. It is, therefore, important to take these cointegration relationships into account in order to obtain good results.

7. Presentation of Econometric Estimation Results

To take into account the cointegration relationships between the variables, we have three models: the PMG (pooled mean group) model, the MG (mean group) model, and the DFE (dynamic fixed effect) model.

The PMG estimator proposed by Pesaran et al. (1999) is constructed under the assumption of heterogeneity of short-term coefficients and homogeneity of long-term slope coefficients. However, the MG (mean group) estimator estimates the mean group model where the model coefficients are calculated from the means of the unconstrained and totally heterogeneous model (in the short and long term). The DFE (dynamic fixed effect) estimator, on the other hand, estimates the dynamic fixed-effect model where all the parameters, with the exception of the constant, are constrained to be identical for all the individuals in the panel. The Hausman test is used to select the appropriate estimator. The DFE estimator has been validated by the various Hausman tests.

The results of the econometric estimates are summarised in the

Table 12. Econometric estimates of the models.

LGDP Per Capita V	Model 1: Bank (Asset) and Stock Exchange (cap_bours)	Model 2: Bank (Asset) and Stock Exchange (Turnover)	Model 3: Bank (Asset) and Stock Exchange (Valeur Echange)	Model 4: Bank (Credit) and Stock Exchange (cap_bours)	Model 5: Bank (Finan) and Stock Exchange (struc)
Ec
L_nb_educ_expected	1.8511 ***	1.9735 ***	1.7285 ***	1.8662 ***	1.9239 ***
L_trade	0.3107	0.3270	0.2433	0.2750	0.3293
L1_exp_gov	0.3007 ***	0.2333 *	0.2601 ***	0.2989 ***	0.3014 ***
efficacite_gov	0.3512 ***	0.3567 ***	0.3197 ***	0.3242 ***	0.3505 **
stabilite_pol	0.1918	0.1786	0.1864	0.1888	0.1885
Stock_exc	0.0111	-	0.1015	−0.0051	0.8666
Bank	0.0151	−0.0094	0.0114	0.0600	−0.8680 *
SR
Ec	−0.1032 ***	−0.0926 ***	−0.1034 ***	−0.1077 ***	−0.1014 ***
L_investment	0.0166	0.0130	0.0112	0.0190	0.0173
Inflation	−0.0006	−0.0005	−0.0006	−0.0004	−0.0006
L_nb_educ_expected	−0.1081 **	−0.1443 **	−0.1594 ***	−0.0956 ***	−0.1161 ***
L_trade	0.0049	0.0050	0.0062	0.0003	0.0034
L1_exp_gov	−0.0119	−0.0104	−0.0139	−0.0132	−0.0124
efficacite_gov	0.0001	0.0042	0.0076	−0.0082	−0.0005
stabilite_pol	0.0144	0.0116	0.0072	0.0167	0.0145
Stock_exc	−0.0018	0.0074	0.0199 **	−0.0001	0.0879 *
Bank	−0.0897 ***	−0.0837 **	−0.1026 **	−0.0948 **	−0.0880 **
_cons	0.1511	0.1259	0.2263	0.1508	0.1306

Source: authors’ calculations. *, **, and *** indicate that the estimated coefficients are statistically significant at 10%, 5%, and 1% levels, respectively.

. The first four models measure the separate impacts of banks and stock markets on economic growth. The last model verifies whether or not the two sources of financing economic activity are complementary in promoting economic growth.

The results of the econometric estimations show that the error correction terms are negative and significantly different from 0. We can, therefore, interpret the results.

The results show that the development of banks has a negative and significant effect on economic growth in the short term. In fact, a 1% increase in bank assets as a percentage of GDP would lead to a fall in growth of between 0.09% and 0.10%. Model 4 also shows that a 1% increase in private credit from deposit banks as a percentage of GDP would lead to a 0.09% fall. The development of stock markets seems to have a positive effect on economic growth. Similarly, a 1% increase in the value traded on financial markets as a percentage of GDP would imply an increase in short-term growth of 0.02%. On the other hand, increases in market capitalization and turnover have no significant impact on economic growth.

The expected level of education has a negative impact in the short term and a positive impact in the long term. In the short term, a 1% increase in the expected level of education would lead to a fall in economic growth of between 0.10% and 0.16%. The results also indicate that government spending as a percentage of GDP has a positive impact on long-term economic growth. In fact, a 1% increase in government spending as a percentage of GDP would lead to a long-term increase in growth of between 0.23 and 0.30% in the different specifications. The government efficiency variable measures the impact of institutional efficiency on economic development. The results indicate that government efficiency has a positive effect on long-term economic growth.

To verify the relationship of complementarity or substitutability between banks and stock markets, this study uses two indicators. The first indicator is the financial structure variable defined by the ratio of market capitalisation to bank assets. The second measures financial activity as the sum of market capitalisation and total bank assets. The results show that the structure variable has a positive impact, while the financial activity variable has a negative impact on economic growth. This suggests that government authorities should focus on developing stock markets rather than banks. While banks play a crucial role in financing segments of the private sector that lack access to capital markets, stock markets offer long-term funding and promote transparency. The stronger effect of stock markets on economic growth observed in our analysis may reflect their superior ability to allocate capital toward high-return, productivity-enhancing investments, thereby fostering more dynamic and innovation-driven growth.

8. Macroeconomic Developments and Caveats

Certain international macroeconomic developments may introduce endogeneity bias into our study, particularly through the potential omission of variables that could simultaneously influence both financial development and economic growth in sub-Saharan Africa during the period under review (1995–2015). Specifically, three major global phenomena were not explicitly accounted for in the empirical model.

First, the 2008 global financial crisis—triggered by the collapse of the subprime mortgage market—deeply disrupted international capital flows and financing conditions, which may have affected African economies through both trade and financial channels. Cortes et al. (2022) find that during the subprime crisis, unconventional monetary policies in the United States contributed to capital outflows from emerging markets, thereby exacerbating financial stress through a negative contagion effect linked to the global flight to safety.

Second, the emergence of the BRICS (Brazil, Russia, India, China, and South Africa) as a geopolitical and economic bloc had an indirect but notable influence on capital flows into sub-Saharan Africa (SSA). The dynamism of BRICS countries encouraged global investors to seek opportunities in other high-potential emerging markets, including several SSA economies. Deych (2015) shows that Nigeria, Kenya, and Ghana—three countries included in our sample—were among the main recipients of foreign direct investment (FDI) from BRICS countries, owing to their economic potential and natural resource endowments.

Third, the growing importance of responsible investment flows, particularly Environmental, Social, and Governance (ESG)-oriented investments, may have altered the nature and structure of international capital allocations. Chen et al. (2023) highlight that ESG assets have expanded significantly, now representing a substantial share of global asset management. This shift has implications for banking practices and financial markets, as well as their impact on economic growth. Banks are increasingly incorporating ESG criteria into their lending policies. Traditionally, stock markets have focused primarily on financial performance; however, the integration of ESG factors introduces new dimensions of value, namely sustainability, equity, and transparency.

Other alternative and expanding forms of financing in Africa may also be reshaping traditional financial markets. Islamic finance, for instance, is gaining traction in several countries, with a growing number of banks offering Sharia-compliant financial products, some of which are now publicly traded on stock markets. Similarly, the rise of fintech and mobile money services is prompting banks to undergo digital transformation or to engage in partnerships with fintech firms.

Future research could further investigate these dynamics, assessing how they influence the relationship between financial development and economic growth in sub-Saharan Africa.

9. Conclusions

The 2007/2008 crisis, with the collapse of Lehman Brothers, showed the limits of the role of the financial system and its integration into the economy. A number of economists have examined the relationship between financial development and economic growth. Theoretical and empirical reviews find mixed results, indicating that the nature of this relationship is complex and not uniform and varies from one country to another. This study, therefore, seeks to analyze the relationship between banks and financial markets and their impact on economic development in sub-Saharan Africa. The contribution of this paper lies in examining the relationship of complementarity or substitutability between banks and stock markets in promoting growth. Two indicators are used: the ‘activity–structure’ variable, which measures the importance of financial markets relative to banks, and the ‘activity–finance’ variable, which takes into account the simultaneous development of banks and stock markets. Based on the Hausman tests, we used the DFE (dynamic fixed effect) models instead of the PMG (pooled mean group) and MG (mean group) models. The results indicate that the development of banks has a negative and significant impact, whereas the development of stock markets seems to have a positive influence on economic growth. The stronger effect observed for the stock market in our analysis may stem from its greater capacity to allocate resources efficiently to productive investments. It also emerges that government authorities should focus on the development of financial markets rather than banks in order to promote economic growth. However, banks and stock markets perform complementary roles within the financial system. Stock markets are generally better suited to large firms capable of meeting high governance standards, whereas banks remain the primary source of funding for small and medium-sized enterprises. A key policy challenge is to rethink the operational model of banks to enhance their contribution to productive investment. In the African context, this may require improved risk assessment mechanisms and the development of innovative financial instruments tailored to local firms.