Economic theory, similar to all other fields of knowledge, has been debated by several researchers. Thus, throughout the history of the economic literature, we can observe controversies over the importance of the financial system.

1) The Theory of Economic Evolution

In his theory of economic evolution, Schumpeter (1912) argued that banks offer credit to entrepreneurs to finance economic activities that promote economic growth. For him, the banking sector must be able to select and finance the most creative investment projects of private enterprises, which are a source of economic growth. Economic growth is achieved if the banker provides capital and credit to the entrepreneur; thus, credit and banking occupy a primordial position in economic activity.

2) The Theory of Financial Liberalization

McKinnon (1973) and Shaw (1973) developed the theory of financial liberalization in the early 1970s. They presented financial system liberalization as an effective and simple technique for activating economic growth in developing countries. Their hypotheses were based on the negative effects of financial repression on savings and investments in developing countries. According to them, only banks play a key role in financing investments by private companies. Furthermore, they question the role of the informal sector in financing economic activities. They added that low interest rates lead to low levels of investment as bank deposits decrease. This resolution affects the quality of investment, as banks that pay their deposits at low rates and lend them at lower lending rates increase risk and choose less risky projects instead of riskier but more profitable ones.

However, beyond the ambitious objectives provided by Schumpeter (1912) and McKinnon (1973) and Shaw (1973) , several criticisms are worth analyzing.

1) The Theory of Asymmetric Information

Neo-Keynesians such as Stiglitz and Weiss (1981) asserted that asymmetric information is an obstacle in the banking sector that fundamentally impedes efficiency. They challenge the theory of economic evolution and financial liberalization by determining that asymmetric information in the credit market leads to credit rationing, which can justify the financial repression policy. Stiglitz and Weiss (1981) added that it is difficult for the policy of financial liberalization to act through a better allocation of resources and the orientation of savings toward productive investments since borrowers who agree to pay higher interest rates are partly those with the riskiest projects in the context of informational asymmetry.

2) Informal Market Theory

According to Taylor (1983) and Wijnbergen (1983) , the informal sector is important for improving economies. For them, this market is characterized by an equilibrium rate that equalizes credit supply and demand, similar to the formal market. They argued that increasing bank lending rates can reduce the demand for money balances, which, in turn, reduces the supply of loanable money in the sector. In this sense, the growth rate in the informal sector will lead to an acceleration in the cost of capital and a reduction in investment.

3) Effective Demand Theory

According to Burckett and Dutt (1991) , in line with Keynesian concepts, investment is determined by the anticipated effective demand and not by the sum of bank deposits (savings). They argued that the interest rate strongly influences the level of investment by private companies. Thus, an increase in bank interest rates can reduce investment and activity levels by exerting downward pressure on household consumption.

Orji (2012) examines the determinants of bank savings in Nigeria as well as the impact of bank savings and bank credits on economic growth in Nigeria between 1970 and 2006 using two impact models: the distributed lag error correction model (DL-ECM) and the distributed model. The empirical results showed that bank credits have a positive influence on economic growth.

Dembele and Machrafi (2021) studied the effect of the banking sector on economic growth in Côte d’Ivoire during 1990-2019, using a VAR model. The results indicate a positive relationship between the banking sector and economic growth in Côte d’Ivoire.

Vasconcelos et al. (2021) tested the hypothesis that bank credit is necessary for economic growth, depending on the country’s level of economic and financial development. To do this, the Granger causality methodology is used for panel data, with data from 106 countries for the period between 1970 and 2016. The main empirical results indicate that, in general, credit causes economic growth and vice versa.

Chuba and Hitlar (2022) examined the effect of commercial bank credit allocated to the agricultural sector on economic growth in Nigeria. Their survey results revealed that commercial bank credit allocated to the agricultural sector has a significant positive effect on economic growth in Nigeria.

Furthermore, Ekamena Ntsama et al. (2022) worked on the effect of bank credit and savings on growth in Cameroon using the ordinary least squares (OLS) method from 1980 to 2019. They observed that bank credit to private companies and savings had a positive and significant impact on Cameroon’s economic growth during the study period.

Koivu (2002) analyses the link between finance and growth using a fixed-effects panel model and unbalanced panel data from 25 transition countries over the period 1993-2000. The results indicate that an increase in the amount of bank credit does not appear to accelerate economic growth. In some cases, credit growth may have led to a decline in growth rates. According to these results, the interest rate differential is significantly and negatively related to economic growth.

Ojeaga et al. (2013) examined the effect of bank lending on growth in Nigeria using data from 1989 to 2012. The estimation method used in the study was the quantile regression estimation method. Their results show that commercial bank lending has a negative effect on growth, whereas institutions do not sufficiently protect their customers from the negative effects that often arise during bank liquidation.

Ishioro (2017) analyzed the effects of banking sector reforms on economic growth in Nigeria using time series data from 1970 to 2013 and an ARDL approach. The study reveals that credit from banking sector to the private sector has been negative and statistically insignificant in Nigeria’s economic growth.

Tchouassi and Tomo (2022) study the effects on economic growth of institutional reforms of financial development in the Economic and Monetary Community of Central Africa. The results obtained using the method of generalised moments on a balanced panel show that credit to the private sector does not contribute to economic growth.

Okonkwo et al. (2022) examine the contributions of the banking sector to the growth of the Nigerian economy from 1986 to 2020 using ordinary least squares regression and Granger causality test. The results of the study revealed that bank credit has a significant negative effect on economic growth.

As far as economic growth is concerned, Figure 1 shows an upward trend. Between 2005 and 2008, the CEMAC zone experienced moderate economic growth, rising from 3.2% to 4.5%. The increase in commodity prices, particularly oil prices, has contributed to the improvement in the level of GDP growth, with the price of a barrel of oil rising from $54.4 ($) per barrel to $97.6 per barrel during this period ( UNCTAD, 2022 ).

In 2009, the growth rate decreased by 2.5 percentage points year-on-year to 2%. This contraction in the growth rate was exacerbated by the global economic recession linked to the subprime crisis (2007-2008) and the sharp fall in oil prices—35.7 percentage points year-on-year at the end of December 2009, to $61.9 per barrel ( UNCTAD, 2022 ).

During the 2009-2011 period, economic growth accelerated to 5.5% in 2011, largely due to higher oil prices to $110.9 per barrel in 2011. This increase in oil prices was caused by political instability in Arab countries (Arab Spring) and the growing global demand for oil, especially in China.

Between 2011 and 2020, economic growth in the CEMAC zone was weak and unstable, falling from 4% in 2012 to −2.1% in 2020. Growth has been held back due to two shocks. The first was the sharp drop in oil prices from $112 per barrel in 2012 to $42.3 per barrel in 2020 ( UNCTAD, 2022 ) and the second was the economic recession in 2020 due to the COVID-19 pandemic. During this period, GDP growth was negative, i.e., −0.2% in 2016 and −2.1% in 2020. A moderate recovery in economic growth was observed in 2021 with an estimated growth rate of 1.3%. This improvement in economic growth is mainly due to the post-COVID-19 response measures under IMF and World Bank-supported programs.

Figure 1 shows that GDP growth is more volatile than bank credit growth. The dynamics of bank credit and GDP growth diverge, which means that demand for bank credit comes more from the non-oil sector and not from the oil sector, which is the engine of growth in the CEMAC. As shown in the data in Figure 1 , while bank credit provided to the private sector by banks held up well during the 2014 commodity crisis, GDP growth had slowed ( Figure 1 ), showing the little impact of bank credit in its ability to boost GDP growth in the CEMAC. Similarly, in 2016, at the height of the commodity crisis, bank credit growth reached a historic level of 14.3% while GDP growth recorded a negative rate of −0.2%. These developments showed that the increase in bank credit could have a negative effect on GDP growth in the CEMAC zone.

The period between 2005 and 2016 saw a strong development of bank credit to the private sector. This significant increase in bank credit was supported by an increase in credit demand from companies and households. Credit momentum slowed between 2018 and 2021 ( Figure 1 ). This slowdown can be explained by the collapse of oil prices, leading to a significant increase in unpaid and bad loans, and also by the COVID-19 crisis, which has slowed down economic activity. In response to the COVID-19 pandemic, bank credit has benefited the community, social and personal services to the tune of 25% as shown in Figure 2 .

Overall, in 2021, bank credit in the CEMAC benefited the non-oil sector more ( Figure 2 ). Bank credit allocated to manufacturing industries was limited to 12%, the lowest rate recorded compared to the transport and telecommunications sector, which captured 25%, or the trade (wholesale and retail) and catering-hotel sector, which accounted for 24% of bank credit distributed to the private sector. Another interpretation is that the extractive industries sector (oil, gas, minerals and metals) on which the CEMAC economy is based is by its nature financed more by extractive FDI from multinationals than by national bank credit. This explains the low level of bank credit enjoyed by manufacturing industries, which are mainly located in the extractive sector of CEMAC countries.

Our study used the Solow (1956) model developed by Pagano (1993) . Capital and labor are remunerated according to marginal productivity. To analyze the effects of bank credit on growth, we followed Ekamena Ntsama et al. (2022) approach. We assumed a Cobb-Douglas production function whose output at time t is given by:

$Y_t=A_t{K}_t^{\alpha }{L}_t^{\beta }$ (1)

with $0<\alpha <1$ et $0<\beta <1$.

Y: product, K: capital, L: labor, and A: level of technology.

In this sense, the productivity of Factor A complements the other factors of production and is presented by the following equation:

$A_t=A_0{F}_t^{e}C{P}_t^z{\left(F\ast CP\right)}_t^{\partial }{e}^{\left(g_t+{\mu }_t\right)}$ (2)

The initial equation becomes:

$Y_t=A_0{F}_t^{e}C{P}_t^z{\left(F\ast CP\right)}_t^{\partial }{e}^{\left(g_t+{\mu }_t\right)}{K}_t^{\alpha }{L}_t^{\beta }$ (3)

Regarding the assumption of convergence in the Solow (1956) method, after developing and introducing the logarithm, we obtained the following final equation:

$\Delta Y_t={\beta }_0+\left(\alpha -1\right)Y_{t-1}+\beta L_t+\alpha K_t+e{F}_t+zC{P}_t+\partial {\left(F\ast CP\right)}_t+g_t+{\mu }_t$ (4)

with CP private sector credit and F financial development, Equation (4) links our two study variables.

Considering the different variables retained, the estimation model is given by

$GD{P}_{it}={\delta }_0+{\delta }_{1}C{P}_{it}+{\delta }_{2}CP{S}_{it}+{\delta }_{3}FD{I}_{it}+{\delta }_{4}IMPOR{T}_{it}+{\delta }_{5}CP{I}_{it}+{\epsilon }_{it}$ (5)

where ${\delta }_i$ are parameters to be estimated; i and t represent the country and year indices, respectively; and ${\epsilon }_{it}$ is the random term. We focused on the main variables that serve as the basis for our estimates.

The data used in this study come from the online databases of the World Bank (WDI) and the United Nations Conference on Trade and Development (UNCTAD). Our sample is made up of the six CEMAC member countries. The data are observed on an annual basis, with a frequency ranging from 2005 to 2021.

GDP growth represents the growth rate of gross domestic product as an annual percentage. This indicator assesses a country’s level of economic activity. Its expected sign is positive ( Dembele & Machrafi, 2021 ).

CP represents the rate of credit granted to the private sector by banks and is an indicator that measures the ratio between the amount of credit granted to private companies by banks and GDP. Its expected sign is positive ( Olowofeso et al., 2015 ).

CPS refers to the number of cell phone subscribers. It is an essential variable of the digital economy. Its expected sign is positive ( Gbame & Silue, 2022 ).

FDI captures the importance of foreign financial flows in relation to production. Its expected sign is positive ( Ekamena Ntsama et al., 2022 ).

Inflation rate (CPI) measured by the relative change in consumer price index, corresponds to a general increase in the prices of goods and services in an economy. Its expected sign is negative ( Noula et al., 2016 ).

IMPORT represents the value of all goods and other market services received from the rest of the world. Its expected sign is positive ( Karaca, 2023 ).

This section presents a descriptive statistical analysis of the variables to be estimated. The results are presented in Table 1 .

Source: Author based on WDI (2023) data.

Table 1 shows that the average GDP of all CEMAC countries is 2.53%. Additionally, the standard deviation between countries is equal to 1.15%, whereas for each country, it is 6.63% of an overall standard deviation of 6.72%. Thus, economic growth in the CEMAC region averaged approximately 2.53%. This average was far from 100%. Hence, on average, CEMAC countries demonstrate weak economic growth—the minimum level of the GDP growth rate is −36.39%, the maximum level is approximately 17.79%, and the dispersion is strong around the average value. The importance of the temporal dimension (17 years per country) over individual dimensions (6 countries) explains this result.

We used an appropriate estimation method to achieve our specific objective. However, to determine this model and the types of data to be applied, it is necessary to conduct economic and econometric tests.

Pesaran and Yamagata (2008) proposed the ∆ test as an improved version of Swamy’s slope homogeneity test. Table 2 presents the results of homogeneity tests.

Source: Author based on WDI (2023) data.

The null hypothesis for the slope heterogeneity test is that the slope coefficients are homogeneous. Therefore, we opted to study panel data with a model for estimation purposes.

$GD{P}_{it}={\delta }_0+{\delta }_{1}C{P}_{it}+{\delta }_{2}CP{S}_{it}+{\delta }_{3}FD{I}_{it}+{\delta }_{4}IMPOR{T}_{it}+{\delta }_{5}CP{I}_{it}+{\epsilon }_{it}$

This section examines whether the variables used in the model are stationary. In other words, we need to determine whether there is at least a cointegrating or long-term relationship among the variables. This relationship is verified using several econometric tests. Therefore, we retain three unit root tests in this study: the Levin, Lin, and Chu (LLC) test (2002) ; Im, Pesaran, and Shin W-stat (IPS) test (2003) ; and Hadri’s LM test (2000) . The findings are presented in Table 3 .

The significance thresholds were set to 1% (***), 5% (**), and 10% (*). Author based on WDI (2023) data.

H0: The panel is non-stationary.

H1: The panel is stationary

The acceptance rule consists of rejecting H0 if the p-value is insufficient at the chosen risk of error threshold (5% in this case).

In fact, the elaboration of these results allows us to conclude that the LLC, IPS, and Hadri tests, notably those of the first and second generations, indicate that all six (6) panel variables are non-stationary in level. In other words, all variables in the model are integrated in order (1), that is, the first difference, which leads us to analyze whether there is at least one cointegration relationship among the selected variables.

To examine the existence of a long-term relationship between growth and credit, as well as other control variables (number of CPS, foreign direct investment, imports, and inflation), we use Pedroni (1999 , 2004) cointegration tests based on stationarity tests on estimated residuals. The test results are listed in Table 4 :

Source: Author based on WDI (2023) data. The significance thresholds were set to 1% (***), 5% (**), and 10% (*).

Pedroni (1999 , 2004) cointegration test results indicate that the set of statistics for a between-panel (panel: rho, PP, and ADF) and a within-panel (group: rho, PP, and ADF) that are significant are greater than those that are nonsignificant. In fact, all of these tests validate the existence of a cointegrating or long-term relationship.

Subsequently, having noted the existence of a cointegrating relationship based on Pedroni (1999) test, it is now crucial to apply an efficient model to estimate systems of cointegrated variables in panel data. To this end, we can distinguish several mechanisms: the GMM model; the error-correction estimators of Pesaran, Shin, and Smith (1999) ; mean group; pooled mean group; static fixed effect; dynamic fixed effect; the fully modified OLS model suggested by Phillips and Hansen (1990) and applied by Pedroni; the dynamic OLS method by Saikkonen (1991) , Stock and Watson (1993) and developed by Kao and Chiang (2001) .

Among the abovementioned models, we retain the GMM dynamic panel model because it allows for estimation when explanatory variables are assumed to be weakly exogenous. Thus, they can be influenced by present or past realizations of the variable to be explained, although they are uncorrelated with future realizations of the error term. Therefore, it is an exceptional model that has attracted the attention of macroeconomists for several years ( Kpodar, 2005 ).

The GMM model is an efficient estimation technique for nonstationary panel data. Although several models exist, it has been widely applied in econometrics. A method is dynamic when it includes one or more lags of the dependent variable as explanatory variables.

Thus, by integrating the lagged dependent variable, Equation (6) becomes

$Y_{i,t}-Y_{i,t-1}=\left(\alpha -1\right)Y_{i,t-1}+{\beta }^{\prime }{X}_{i,t}+v_i+{\epsilon }_{it}$ (6)

where $Y_{i,t}$ represents the GDP rate, X represents the model’s explanatory variables, ${\beta }^{\prime }$ represents the model’s parameters, v the time-specific effect, $\epsilon$ the error term, i the country index, and t the time index.

Equation (7), which is a growth equation following Arellano and Bond (1991) , Arellano and Bover (1995) , and Blundell and Bond (1998) studies, corresponds to in our work and can be formulated as follows:

$\begin{array}{c}GD{P}_{i,t}=\alpha GD{P}_{i,t-1}+{\delta }_{1}GD{P}_{i,t}+{\delta }_{2}C{P}_{i,t}+{\delta }_{3}CP{S}_{i,t}+{\delta }_{4}FD{I}_{i,t}\\ +{\delta }_{5}IMPO{R}_{i,t}+{\delta }_{6}CP{I}_{i,t}+v_t+{\epsilon }_{it}\end{array}$ (7)

Therefore, the presence of the lagged explained variable in the above method hampers the use of standard econometric techniques such as OLS because they generate biased results.

It is also noteworthy that there are two variants of dynamic panel GMM estimators: the first-difference GMM estimator was developed by Arellano and Bond (1991) , and the second system GMM estimator was developed by Arellano and Bover (1995) and Blundell and Bond (1998) . The latter provides solutions for the simultaneity bias, reverse causality, and omitted variables. The consideration of these two estimators assumes quasi-stationarity of the variables in the level equation and the absence of autocorrelation in the residuals.

Each estimator exhibits unique features. Considering the particularity of the model, the results are presented in Table 5 .

The results of the panel data estimation using the GMM in difference and in system with long-run coefficients on the effects of bank credit on economic growth in CEMAC countries are presented in Table 6 . Similarly, considering the results of the two GMM estimates ( Table 5 and Table 6 ), the difference estimate yields better results than the system estimate. Therefore, in our study, we applied the GMM in difference.

The results obtained in Table 5 indicate that according to the Wald chi² test, in addition to the number of instruments being lower than the number of observations, we can conclude that the method is appropriate. The Sargan and Hansen tests did not reject the hypothesis of acceptance of the instruments used, because the probabilities assigned to them were 0.743 and 1, respectively. Similarly, according to the Arellano-Bond tests, the presence of the AR effect (1) is shown at

Source: Author based on WDI (2023) data. The significance thresholds were set to 1% (***), 5% (**), and 10% (*).

Source: Author based on WDI (2023) data. The significance thresholds were set to 1% (***), 5% (**), and 10% (*).

the 10% threshold; however, the probability associated with it is 0.073, and the existence of the AR effect (2) is rejected, that is, an allied probability is 0.562. In other words, according to the Arellano-Bond tests, we can conclude that there is no second-order autocorrelation, but rather a first-order autocorrelation, which validates the obtained results that can be interpreted.

The results of the estimate contained in Table 6 show that a 1% increase in bank credit to the private sector leads to a decrease in GDP of 2.24%. However, bank credit to the private sector has a negative and significant effect on economic growth in the CEMAC at the 1% threshold. Similar results are found by many previous studies, including ( Ishioro, 2017 ; Tchouassi & Tomo, 2022 ; Okonkwo et al., 2022 ). These results can be explained in the context of the CEMAC by the fact that the extractive industries sector (oil, gas, minerals and metals) on which economic growth in the CEMAC is based is financed more by foreign direct investment (FDI) from the multinationals that dominate this sector than by financing by bank credit from the host country. Another explanation is the rigidity of banking regulations in the CEMAC, the narrowness of the financial sector, the low level of financial inclusion, and poor management of bank credit. The predominance of short-term credit in the CEMAC hinders long-term investment projects and benefits traders who use them to finance imports of common consumer products into the CEMAC. More specifically, imports have a positive and significant effect on economic growth in the CEMAC at the 5% threshold ( Table 6 ). Thus, an increase in imports of 1 point, all other things being equal, improves the economic growth of the CEMAC by 0.15%. Similar results are found by many studies ( Karaca, 2023 ; Esposito & Hassan, 2003 ; Jinjarak, 2007 ). However, the coefficient associated with the inflation rate variable is negative and significant at the 1% threshold, suggesting that inflation has a negative effect on economic growth in the CEMAC. An increase of 1 point in the inflation rate translates into a decrease in the level of economic growth in the CEMAC of 1.06%. This result is in line with those found by Noula et al. (2016) ; Ekamena Ntsama et al. (2022) and Nkwenka Nyanda (2021) . In the CEMAC as elsewhere, the countries’ import activities are carried out through trade credits. Inflation and imports are influenced by the CEMAC government to varying degrees of complexity. The BEAC (2021a) , which defines and conducts CEMAC monetary policy, notes that imports have maintained an upward trend in the general price level due to a structural deficit in agricultural production in the CEMAC but also due to supply difficulties for importers. Thus, inflation in the CEMAC could have an external origin or trade with the outside world could be the transmission channel of inflation in the CEMAC, thus depressing economic activity ( Nkwenka Nyanda, 2021 ). Future research will try to address this concern.