The Effect of Banking Industry Development on Economic Growth: An Empirical Study in Jordan

Abstract

This study aims to investigate whether economic growth is elevated by banking industry development in Jordan. The study adopts time-series econometric methodologies, which comprise the bounds testing approach within the autoregressive distributed lag (ARDL) and the conditional causality analysis. Consistent with the assumptions of the adopted methodology, the study utilized annual time-series data for a relatively long period of thirty-nine years, between 1980 and 2018. The empirical results show that Jordan's economic growth is strongly responsive in respect to any changes in banking industry development. Also, the results reveal the harmful impact of rising lending interest rate; as this rate increases, economic growth will decrease. The findings are in line with the conceptual arguments of the supply-leading hypothesis, which confirmed that banking development is considered as one of the main pillars that have stimulating effects on economic growth. The evidence of the current study may provide important implications for policymakers and bankers. Those professionals should work to maintain a stable regulatory system that enhances the banking system function in activating economic growth. Also, a considerable focus should be placed on designing a steady interest rate policy to avoid the inherently undesirable impacts of high-interest rates on the Jordanian economy.

1. Introduction

It is confirmed in the economic literature that banking sectors play an intermediation role by facilitating money movements between both fund-suppliers and fund- demanders. This can be through attracting savings and then channeling these savings as loans to finance different economic activities, which eventually encourage economic growth (EG). It is important to assert that banking institutions should direct the funds to the appropriate economic activities that can elevate EG rapidly; this, in turn, reflects how important these institutions in the EG processes. However, bankers should pay considerable attention to maintain a solvent banking sector to ensure its ability in providing the necessary fund needs, which sustain economic development and growth. A developed banking system is a major requirement for an economy to grow hurriedly. Therefore, the banking system should be stable and sound to absorb all of the possible shocks; regardless of internal or external shocks. Therefore, a well-developed banking sector can play an effective role, which, surely, elevates EG.

The linkage between EG and banking industry development has attracted the attention of economists and researchers, especially after the growth of the endogeneity theories. The development of the banking industry can be described by the ability of the banking sector to providing broad financial services and products that satisfy the rabid expansions of economic activities. Today, banking development is an important factor causing EG, while the direction of the causal impact between these two variables is one of the debated issues in the literature. Some support the supply-leading hypothesis that argues that banking development is the key supporter of the EG, whereas others confirmed the demand-leading hypothesis that posits the EG is a source of banking development. Historically, Schumpeter (1961) mentioned that the functional role of banking sectors, as financial intermediaries, is one of the most important roles that cause to accelerating EG. This can be through the main functions of this sector that include providing funds, managing risks, evaluating and selecting projects, and facilitating transactions among different parties. In the same manner, Patrick (1966) confirmed the role of the banking sector to be one of the most important elevators to EG, which asserts the importance of the banking sector especially in the early stages of EG. As a consequence, the direction of the causal impact becomes reversed in the advanced stages of EG; economic expansion creates a demand on the financial services and products leading to a developed banking sector. The former view is strongly supported by the evidence from a broad panel of countries (among others, Robinson, 1952; King & Levine, 1993; Levin, 2005).

EG has been considered a crucial objective of countries throughout the world over the past number of decades and up to date. However, the development of this process of growth needs certain prerequisites to satisfy demand. Satisfying this demand may occur by providing sufficient finance for investment and production activities. These requirements appear heavily in developing countries where numerous economic problems, such as high inflation and interest rates, high unemployment rate, very low living and wellbeing standards are prevalent (Pieroni, 2009). According to these problems, developing countries constantly seek to maintain their EG to increase their national income, creating more job opportunities, thus improving living standards. The banking industry is currently playing a vital contribution to accelerating the process of economic development. Moreover, the banking sector leads to enhance rates of EG in both developing and developed economies (Rodrik, 2000). The theory of EG postulates that banking sectors are useful instruments for improving the production capacity of the economies to be more productive; assuming that the banking sector comprises an important internal source for any country, especially the countries where still in the first stages of economic expansion (Schumpeter, 1961; Shokr & Al-Gasaymeh, 2018). The need to stimulate and manage EG is a global issue of significant importance in developed, emerging, and under-developed countries. While this need is not new, given the continued population growth, changing demographics, and widening income disparities in under-developed countries. Tekin (2012) asserted that the most successful economies have been sufficiently sophisticated to develop their financial systems to sustain EG.

In line with the previous discussion, this study aims to explore the role played by the banking industry development in elevating EG of Jordan between 1980 and 2018. Jordan’s financial system is characterized as a bank-based system; the banking sector is considered the prime sector in the financial system that plays a dynamic role in the economy (Almahadin et al., 2020). Therefore, the banking sector is deemed the solidest aspect of the financial services industry. The financial services industry is considered the most established and vigorous sectors in Jordan, remaining resilient in the face of significant external shocks and retaining its role as the key stimulator of economic growth. For that the Central Bank of Jordan (CBJ) maintains a pro-growth monetary stance, following on from growth in deposits and profits at commercial banks (CBJ, 2018). Also, this study is motivated by the lack of empirical studies that investigated the role of banking development in elevating EG of Jordan. The empirical outcomes of the present study are presumed to provide interested parties with informative content that will be useful in decision-making processes and policy designing.

The remaining of the study is organized as follows: the literature review is presented in section 2; the model specification is presented in section 3, followed by econometric methodology in section 4; the empirical results are reported in section 5, and section 6 delivers the conclusions and policy implications.

2. Literature Review

In past decades, many researchers and analysts have discussed the debated relationship between finance and EG (among others, Baghehot, 1873; Schumpter, 1961; Robinson, 1952; Hicks, 1969; Miller, 1998). The debate was whether financial development is the source of EG, or the EG is the key factor that enhances financial development. For example, Robinson (1952) suggests that EG creates a demand for financial services and products in which, as a result, accelerates financial development. On the other side, Lucas, (1988) indicates that, when finance is excessive, it is somehow complicated in explaining growth. Moreover, growth may be affected by financial development when combining savings, high diversifications, risk management, and when goods and services are easily exchanged (Levine, 2005; Bong & Premaratne, 2019; Camba & Camba, 2020; Kumar & Paramanik, 2020; Ngoc, 2020).

In the context of developing countries, several studies have discussed the indispensable role of banking sectors in developing and promoting EG in different countries and using different methodologies (Demetriades & Hussein, 1996; Rousseau & Watchel, 1998; Kar & Pentecost, 2000; Christopoulos & Tsionas, 2004; Bolbol et al., 2005; Abu- Bader & Abu-Qarn, 2008; Abusharbeh, 2017). Demetriades and Hussein (1996) suggest a bidirectional causal relation accurses between money supply and Gross Domestic Products for developing countries. Rousseau and Watchel (1998) suggest that in developing countries, finance has a supportive impact on EG; the functional role of banking sectors elevates economic to grow. For Turkey, Kar and Pentecost (2000) point out that the direction of the causality depends heavily on the used indicators especially; the impact causality analysis might be varying across different used gauges of both EG and financial development. While Christopoulos and Tsionas (2004) suggest that in the long run there are high frequently factors influencing the finance, which lead to EG. Recently, Abusharbeh (2017) indicates that banking sectors are playing a momentous role in EG progress by enhancing economic productivity. For Egypt, Bolbol et al. (2005) point out that the country’s total factor productivity can be enhanced by diversifying the financial system and by reforming the stock market. For MENA region, Abu-Bader and Abu-Qarn (2008) found a bidirectional causality to suggest finance leads growth and the vice versa is correct as well. They assert that financial sector supports EG through efficient investment allocation. The authors also suggest that reforms programs need to be maintained to motivate savings, investments, and, eventually, long-run EG.

Indeed, the related literature has gauged EG by various measures (i.e., real per capita growth, total productivity growth, and growth in capital accumulation). Also, the banking development has been expressed by various indicators in which the most popular are: total bank credits, central bank assets, and credit to private sector all are computed as a percent of GDP (King & Levine, 1993). A strand of literature has focuses on the banking development-growth nexus as the main focus of the investigation. For instance, King and Levine (1993) illustrate that the banking sector considers among the most important sectors that have a desirable impact on the EG. The authors provide their evidence from a wide range of countries and by employing different indicators for both banking development and the EG. This view is supported by many studies (Ghali, 1999; La Porta et al. 2002; Al-Aawad & Harb, 2005; Kar et al., 2011; Trabelsi, 2012; Abdelhafidh, 2013). For example, Ghali (1999) has found a unique causal impact that runs from the finance development toward EG of Tunisia. His investigation has been done by employing a couple of proxies to financial development (i.e., total bank claims on the private sector as a portion of nominal GDP and total bank deposit as a portion of GDP). However, some literature provides evidence of a strong long-term impact and weak short-term impact from financial development to EG (Al-Aawad & Harb, 2005; Kar et al., 2011; Ben Salem & Trabelsi, 2012; Abdelhafidh, 2013).

Using panel data analysis of the Generalized Method of Moments (GMM), the findings of Levine et al. (2000) support the elevating impact of the banking sector on EG in a broad panel of countries. In a similar approach, Beck et al. (2000) reveal that financial development leads to growth, which is achieved through economic productivity, not through capital accumulation. Rousseau and Watchel (2000) indicate that the banking and market developments are exogenous mechanisms that have significant contributions in growth process. For the Mediterranean economies, Boulila and Trabelsi (2004) provide a strong support for the demand-leading hypothesis; as the economy grow, the size of the demand on the financial services increases leading to develop financial systems. Rioja and Valev (2004a, b) confirmed that the EG is affected by the financial development more strongly in high-income economies when compared to low-income economies.

Based on the reviewed literature, one can conclude that the banking sector is important and indispensable for EG, especially, in emerging economies in which banking development contributes significantly to EG. Based on these findings, the question raised here is whether this conclusion applies in the context of emerging countries, within the context of the lack of research in relation to the investigation of whether banking industry development elevates economic growth in Jordan as an example of an emerging country. While the banking system of Jordan is anticipated to elevate EG by facilitating sufficient credit to different economic activities. Therefore, this study attempts to investigate whether or not banking development is making a significant influence on the economy. Due to the lack of studies on this regard for Jordan banking sectors, this study will fill the gap in the literature to examine the impact of banking finance on EG in Jordan-as an emerging country.

3. Model Specification

In line with the theoretical and empirical arguments, the functional model below is suggested to explore the role that banking industry development played in elevating Jordan’s economic growth utilizing a time-series data during the period of 1980–2018:

\(\mathrm{GDP}=f(\mathrm{BID}, R, \mathrm{BD})\)       (1)

where GDP is the economic growth indicator measured by GDP per capita; BID indicates to banking industry development as proxied by the domestic credits facilitated by the banking sector to the private sector as a percentage of GDP; R is the lending interest rate that considers the major determinants of economic growth with a significant impact on the credit providing function, also used to capture the impact of domestic monetary policy; and BD is the ratio of bank deposits to GDP that are used to control the size of the banking sector. The raw data of the considered variables have been downloaded from the Word Bank (2019).

To capture growth impacts in the long-term economic period, the functional specification of Equation (1) is expressed in logarithmic form:

\(\operatorname{LnGDP}_{t}=\beta_{0}+\beta_{1} \operatorname{LnBID}_{t}+\beta_{2} R_{t}+\beta_{3} \operatorname{LnBD}_{t}+\varepsilon_{t}\)       (2)

where at period t, Ln GDP is the natural log of GDP per capita; Ln BID is the natural log of the percentage of private credit; R is the lending interest rate; Ln BD is the natural log of the banking deposits ratio; ε is the error disturbance; β₀ represents a constant or intercept term; and β₁, β₂, and β₃ are parameters to be estimated with expected signs to be positive, either positive or negative, and positive, respectively.

4. Econometric Methodology

The primary phase in the empirical analysis is the unit root testing using augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) tests to examine the stationarity of the considered variables. By doing so, identifying the integration order of the considered variables provides significant insights that will be valuable in testing the relationship between the variables within the appropriate econometric techniques.

As a subsequent phase, the existence of the long-run level relationship should be examined between each pair of variables. For this, the study employs the bounds testing approach within the ARDL framework of Pesaran et al. (2001). This technique is applicable regardless of the integration order of the explanatory variables: it can be either a first-order, I(0); second-order, I(1); or mixed-orders of integrations. The Wald F-statistics are used to test the null hypothesis of no long-run level relationship between the variables. Rejecting the stated null hypothesis provides evidence for existing the long-term level relationship among the suggested variables. Compared to other traditional approaches, this approach is more superior by having a set of advantages: The integration order of the variables can be mixed; it is suitable in case of small sample size; selecting the optimal lags is reachable by allowing to use different lag for each variable. By adopting this approach, the following bounds model is estimated using the ordinary least squares method (OLS):

\(\begin{aligned} \Delta \operatorname{In} \mathrm{GDP}_{\mathrm{t}} &=a_{0}+\sum_{i=1}^{p} b_{i} \Delta \ln \mathrm{GDP}_{r-i}+\sum_{i=0}^{q} c_{i} \Delta \mathrm{BID}_{t-i} \\ &+\sum_{i=0}^{q} d_{i} \Delta R_{t-i}+\sum_{i=0}^{q} e_{i} \Delta \operatorname{In} \mathrm{BD}_{t-i}+\sigma_{1} \operatorname{In} \mathrm{GDP}_{t-1} \\ &+\sigma_{2} \operatorname{In} \mathrm{BID}_{t-1}+\sigma_{3} R_{t-1}+\sigma_{4} \operatorname{In} \mathrm{BD}_{r-1}+\varepsilon_{t} \end{aligned}\)       (3)

where Δ symbolizes a change in the GDP, BID, R, and BD variables; a₀ is the intercept; b_i, c_i, d_i, and e_i are the short-term parameters to be estimated; σ₁, σ₂, σ₃, and σ₄ are the long-term parameters of the first lag of the explanatories at their levels; and ε_t is the disturbance error-term of the econometric function.

Once the null-hypothesis of the above bounds test is rejected; the existence of a level long-term relationship between the variables will be, formally, confirmed. Then, the following dynamic long-term ARDL model of Pesaran and Shin (1995, 1999) can be applied to extract the estimated long-term parameters:

\(\begin{aligned} \operatorname{LnGDP} &=\vartheta_{0}+\sum_{i=1}^{p} \vartheta_{1} \Delta \mathrm{GDP}_{t-j}+\sum_{i=0}^{4} \vartheta_{2} \Delta \mathrm{BID}_{t-i} \\ &+\sum_{i=0}^{q} \vartheta_{3} \Delta R_{t-i}+\sum_{i=0}^{4} \vartheta_{4} \Delta \ln \mathrm{BD}_{t-i}+\varepsilon_{t} \end{aligned}\)       (4)

where all of the variables are well-defined in previous section; and ϑ₁, ϑ₂, ϑ₃, and ϑ₄ are the long-term dynamic parameters to be estimated by the ARDL mechanism.

Subsequently, the short-term parameters will be estimated by utilizing the conditional error correction model (ECM). By doing so, the error correction term (ECT, ε_t–1) will also be estimated through the following mechanism:

\(\begin{aligned} \Delta \operatorname{InGDP} &=\beta_{0}+\sum_{j=1}^{p} \beta_{1} \Delta \operatorname{InGDP}_{t-j}+\sum_{j=0}^{q} \beta_{2} \Delta \operatorname{InBID}_{t-j} \\ &+\sum_{j=0}^{4} \beta_{3} \Delta \operatorname{In} R_{t-j}+\sum_{j=0}^{q} \beta_{4} \Delta \operatorname{InBD}_{t-j} \\ &+\beta_{5} \varepsilon_{t-1}+u_{t} \end{aligned}\)       (5)

where all of the variables are well-defined in previous section; β₀ is the equation constant term; β₁, β₂, β₃, and β₄ are the short-term parameters to be estimated by the ECM; β₅ is the estimated coefficient of one lagged period of the ECT, with the expected negative and statistically significant value, thus proving how quickly the disequilibrium between the short-run and long-run values of the explained variable is eliminated in each specific period (Gujarati, 2003); and u_t is the random error term.

As the final step of the adopted econometric methodology, the conditional Granger causality analysis should be conducted to determine the directions of the causal impacts between the considered variables. This analysis applies within the mechanism of the VECM, allowing for the inclusion of an ECT to capture the short-run deviations of series from their long-run equilibrium path (Narayan & Smyth, 2004). The conditional Granger causality test is conducted as follows:

\(\begin{aligned} \left[\begin{array}{c} \Delta \ln \mathrm{GDP}_{t} \\ \Delta \ln \mathrm{BID}_{r} \\ \Delta R_{t} \\ \Delta \ln \mathrm{BD}_{t} \end{array}\right] &=\left[\begin{array}{l} \mu_{1} \\ \mu_{2} \\ \mu_{3} \\ \mu_{4} \end{array}\right]+\left[\begin{array}{llll} \varrho_{1,1} & \varrho_{12,1} & \varrho_{13,1} & \varrho_{14,1} \\ \varrho_{21,1} & \varrho_{22,1} & \varrho_{23,1} & \varrho_{24,1} \\ \varrho_{3,1} & \varrho_{32,1} & \varrho_{33,1} & \varrho_{34,1} \\ \varrho_{4,1} & \varrho_{42,1} & \varrho_{43,1} & \varrho_{44,1} \end{array}\right]\left[\begin{array}{c} \Delta \ln \mathrm{GDP}_{t-1} \\ \Delta \ln \mathrm{BID}_{t-1} \\ \Delta R_{t-1} \\ \Delta \ln \mathrm{BD}_{t-1} \end{array}\right] \\ &+\cdots+\left[\begin{array}{llll} \varrho_{11, i} & \varrho_{12, i} & \varrho_{13, i} & \varrho_{14, i} \\ \varrho_{21, i} & \varrho_{22,1} & \varrho_{23,1} & \varrho_{24, i} \\ \varrho_{31, i} & \varrho_{32, i} & \partial_{33, i} & \varrho_{34, i} \\ \varrho_{41, \tilde{i}} & \varrho_{42,1} & \varrho_{43, j} & \varrho_{4, i} \end{array}\right]\left[\begin{array}{c} \Delta \ln \mathrm{GDP}_{t-i} \\ \Delta \ln \mathrm{BID}_{t-i} \\ \Delta R_{t-i} \\ \Delta \ln \mathrm{BD}_{t-i} \end{array}\right] \end{aligned}\\ +\left[\begin{array}{l} \emptyset_{1} \\ \emptyset_{2} \\ \emptyset_{3} \\ \emptyset_{4} \end{array}\right] \mathrm{ECT}_{t-1}+\left[\begin{array}{l} \varepsilon_{1, r} \\ \varepsilon_{2, r} \\ \varepsilon_{3, t} \\ \varepsilon_{4, r} \end{array}\right]\)       (6)

where some of the variables present here have been defined in the previous section, Δ indicates the difference operator, μ₁ to μ₄ are the constant terms, ϱ’s are the causal parameters to be estimated, Ø₁ to Ø₄ are the parameters of the lagged ECT (ECT_t-1) derived from the estimated long-run model, ∂’s are the causal parameters to be estimated, and ɛ₁ to ɛ₄ are random- error terms which presumed to be serially independent.

5. Empirical Results

5.1. Unit Root Testing

Table 1 presents the test statistic values of the ADF and PP unit root tests for all of the considerable variables. The statistic values of these tests indicate that GDP is not stationary at the level but stationary at the first difference. According to statistic values, the banking industry development indicator (BID) was found stationary at level based on the ADF test at a marginally significant level of (0.10) but had a non-stationary series based on the PP test, whereas it presented a stationary series at the first difference in both tests. The lending interest rate series (R) is non-stationary at level in both ADF and PP tests but is stationary at the first difference. The BD variable observed a non-stationary series based on the ADF and PP tests at their levels, but it is stationary at the first difference. To conclude, the empirical outcomes of the unit root tests provide evidence of a mixed order of integrated; the dependent variable of GDP is a first-order stationary series, whereas the explanatory variables are mixed-order stationary series. The unit root tests affirm that none of the variables are stationary at the second order; this is compatible with the ARDL model assumptions. Thus, this is sufficient justification to investigate the existence of a long-run level relationship using the ARDL bounds testing framework.

Table 1: Series Stationary Testing with ADF and PP Tests

ADF is the Augmented Dickey-Fuller and PP is the Phillips-Perron tests of unit root, respectively. τ_T, τ_µ, and τ represent the models with a drift and trend, with a drift and without trend, and without a drift and trend, respectively. ***, ** and * denote the rejection of the null hypothesis of the individual series has a unit root (non-stationary) at the 0.01 0.05 and 0.10 levels, respectively. I(0) denotes that the individual series is integrated at the level. I(1) denotes that the individual series is integrated at the first order.

5.2. Bounds Testing

The F-statistics of the bounds testing approach are reported in Table 2. The empirical outcomes reveal that the null hypothesis of no level long-term relationship between the variables is rejected under all scenarios of the bounds test (F_iii, F_iv, and F_v). In other words, the calculated F-statistics of the bounds test exceed the upper bounds’ critical values. Thus, the alternative hypothesis that the level long-term relationship between the variables exists will be accepted and adopted by the study. This, in turn, provides a strong justification to proceed with the next steps of the utilized methodology, as explained in the econometric methodology section. Estimating the ARDL long-term model, the ECM, and the conditional Granger causality tests will comprise the coherent analytical framework for this study.

Table 2: Bounds F- and t-statistics for the Existence of a Levels Relationship

Notes: F_iii, F_iv, and F_v denote the F-statistics of the bounds model with unrestricted intercepts and without deterministic trends, whit both unrestricted intercepts & restricted deterministic trends, and with unrestricted intercepts & without deterministic trends, respectively within a corresponding P-values. t_iii and t_v are the t-ratio used to test if α₀ = 0 of the bounds model without & with deterministic trends, respectively. *** and ** denote that the calculated F-statistic is exceeds the UCB value at 0.01 and 0.05 levels of significance, respectively which means that the null hypothesis of no level long-term relationship between the variables is rejected.

5.3. Level Long-Term Relationship

The estimated coefficients of the long-term ARDL model are reported in Table 3. The coefficient of the BID is positive, with a value of 21.87 percent, and statistically significant, at a 0.05 level of significance. This indicates that banking industry development positively elevates economic growth, which is compatible of the results of previous studies (for instance Rousseau & Watchel, 1998; Kar & Pentecost, 2000; Levine, 2005). Thus, the empirical outcome supports the supply-leading hypothesis and is consistent with the findings of King and Levine (1993) and Levine et al. (2000). In contrast, the estimated coefficient of the interest rate on credit facilities (lending interest rate, R) is negative and significant, with a value of –5.24 percent. This, in turn, indicates the harmful role that interest rates play in economic growth. Thus, we can conclude that as the lending interest rates increase, economic growth will decrease, which is consistent with empirical evidence from Abusharbeh (2017), Almahadin (2019), and Almahadin and Tuna (2019).

Table 3: The Estimated Coefficients of the Long-Term ARDL Model

Notes: ***, ** and * indicate the significance level of 0.01, 0.05 and 0.10, respectively.

Regarding the effect of banking sector size, as measured by the ratio of banking sector deposits to GDP, the estimated coefficient is reported to be positive and significant, with a value of 21.68 percent. This result is in line with the role of banking industry development, indicating that the size of the banking sector has a positive relationship with economic growth. In other words, banking sector size is an important factor that enhances the role of the banking industry in accelerating economic growth (King & Levine, 1993).

5.4. Short-Term Relationship

The estimations of the ECM in the short-term relationship are presented in Table 4. As shown, the estimated coefficient of banking industry development is recorded as positive and significant, with a value of 24.18 percent. This value indicates the positive short-term relationship between banking development and economic growth in Jordan. This relationship is consistent with the empirical results of (Ghali, 1999; Abdelhafidh, 2013; Almahadin, 2019). In opposition to this, the estimated short-term coefficients of the lending interest rate provide evidence of negative impact at the first and second lags. Thus, one can conclude that increased lending interest rates negatively affect economic growth. Regarding the relationship between banking sector size and economic growth, the reported results show the coefficient to be positive and significant, with an estimated value of 23.53 percent. The empirical results of the ECM of the short-term relationship are, thus, strongly compatible with the results of the long-term model.

Table 4: The Error Correction Estimations of the Short-Term Relationship

Diagnostics: R-squared (0.8034), Adjusted R-squared (0.7525), F-statistic (15.7713), (F-statistic probability = 0.0000), and Durbin- Watson statistic (1.71488). Notes: ECT is the estimated coefficient of the error correction term in the error correction model, ***, ** and * indicate the significance level of 0.01, 0.05 and 0.10, respectively.

Concerning the estimated coefficient of the ECT (ɛ_t–1), a negative and highly significant value of 85 percent is observed. The negative and significant ECT value indicates a long-term relationship between the dependent and independent variables. In other words, the value of the ECT indicates that the convergence of the economic growth process will reach the long-term equilibrium path with a speed of adjustment of approximately 85 percent through the channels of the explanatory variables. Finally, the bottom panel of Table 4 presents a set of diagnostic tests of the estimated ECM. As clearly shown, the value of R-squared is around 80 percent, and the F-statistic value is 15.77 and highly significant, thus supporting the fitness of the model. At the same time, the Durbin Watson statistic is around 1.7, which is close to the rule of thumb of 2, indicating that there is no autocorrelation problem with the model.

5.6. Causality Investigation

The empirical outcomes of the conditional Granger causality test are reported in Table 5. Among the main results, the banking industry development gauge is found to result in economic growth. This result, in turn, supports the theoretical arguments of the supply-leading hypothesis and is strongly consistent with the empirical results of the level long-term estimations. Economically, one can conclude that banking development is an essential and critical source of economic growth in Jordan. Thus, the intermediary role of the banking sector as a fund facilitator leads to elevated economic growth (Ben Salem & Trabelsi, 2012). Moreover, the causality test outcomes provide empirical evidence of an existing bidirectional causal relationship between banking industry development and bank size; as measured by banking sector deposits, which is in line with the empirical outcomes of Almahadin and Tuna (2019). At the same time, the bidirectional causal relationship is observed between bank deposits and lending interest rates. In contrast, the unique direction of the causal associations tends to pass from lending interest rate to banking industry development and from economic growth to banking sector size. Regarding the causality in the short term, as can be shown in the t-statistics of the ECM_t–1; presented in the last column of Table 5. The empirical outcomes indicate that three causal relationships being with banking industry development, bank size, and lending interest rate and move toward all other variables.

Table 5: The Conditional Granger Causality Test

Notes: H₀: Variable X does not Granger Cause Variable Y. P-values are given in parentheses. ***, ** and * indicate the significance level of 0.01, 0.05 and 0.10, respectively.

5.7. Diagnostic and Stability Tests

Table 6 reports the most recommended diagnostic tests to ensure the adequacy of the estimated models. As shown, the statistic value of the JB test provides evidence for the normal distribution of model residuals. To test for serial correlation in the residuals of the estimated models, the Breusch–Godfrey LM test was used, and the statistic value indicates that the residuals are not serially correlated. In addition, the Breusch–Pagan–Godfrey test, the Harvey test, the Glejser test, the ARCH test, and the White test are employed to test for heteroscedasticity problems in the residuals of the estimated models. The statistical values provide evidence of inexistent heteroscedasticity in the residuals of the estimated models; the statistic values of all of the tests are insignificant. Finally, the Ramsey regression equation specification error test (RESET) is utilized to ensure that substantial nonlinear combinations exist among the independent variables of the linear model. Based on the reported test statistic of the RESET, the estimated model did not suffer from any misspecification problems, meaning that the estimated model is well-specified.

Table 6: Diagnostics Tests for the Estimated ARDL Models

The stability of the estimated long-term coefficients is tested by using the cumulative sum (CUSUM) and the cumulative sum of squares (CUSUMSQ) tests of Brown et al. (1975). The outcomes of these tests are presented in Figure 1. The plots of both the CUSUM and CUSUMSQ tests did not cross the critical boundary of a 0.05 level of significance. Thus, one can conclude that the estimated long-term coefficients were stable over the period of investigation. In other words, the estimated models do not suffer from any structural or systematic instability over the sample period of the current study.

Figure 1: CUSUM and CUSUMSQ Stability Tests

6. Conclusion and Policy Implications

This study aimed to investigate whether banking industry development elevated economic growth in Jordan between 1980 and 2018. For this purpose, the study adopted one of the most acceptable financial time series econometric methodologies, namely, bounds testing within the ARDL framework. When using this sophisticated methodology, the possibility of finding a level long-term relationship between the considered variables should first be found before proceeding to estimate the long-term model. Moreover, the ECM is also estimated to assess the short-term relationships in addition to estimating the speed of the adjustment coefficient. In order to ensure more robust results, the study analyzed the causal effects among the variables using the conditional Granger causality within the VECM.

The empirical outcomes of the unit root tests affirmed that the dependent variable is stationary at the first order and that none of the variables found the stationary condition at the second order. The bounds test revealed that a level long-term relationship exists between GDP and the considered explanatory variables (i.e., BID, R, and BD). This, in turn, provided sufficient justifications to proceed with estimating the long-term ARDL model. The estimated long-term parameter of the banking industry development indicator was found to positively affect economic growth with the highest coefficient value compared to other estimations in the model. From an economic perspective, this result provides significant support for the supply-leading hypothesis, which asserts that banking development is one of the main elements that determines economic growth. Thus, the empirical outcomes indicate that Jordan’s economic growth is elastic to changes in banking industry development. In contrast, the lending interest rate proxy has a negative impact on economic growth. The negative impact of this variable reveals the harmful and undesirable role that increased lending interest rates play in economic growth. As a result, monetary policymakers should expend considerable effort in maintaining steady and effective interest rate policies that lead to more investment in both domestic and international fields. The estimations of the ECM provide a comprehensive picture of the short-term relationship, and the results are highly consistent with the results of the long-term model. The estimated ECT reveals that the convergence process of economic growth will reach a long-term equilibrium path with a relatively high speed of adjustment through the channels of the explanatory variables. This provides supplementary evidence for the existence of long-term feedback between the dependent and determinant variables of the functional model.

Among the major results of the causality analysis, a unique direction causal effect was found to pass from banking industry development to economic growth. This result also supports the theoretical arguments of the supply-leading hypothesis, as in the empirical results of the level long-term model. Economically, one can conclude that banking development is an essential and critical source of accelerated economic growth in Jordan. Thus, the intermediary role of the banking sector as a fund facilitator leads to elevated economic growth. In terms of the diagnostic and stability tests, the estimated models were found to be robust and stable over the sample period without any observable problems. Finally, the empirical evidence of the current study may provide important policy implications for both policymakers and bankers. They should work seriously to maintain a stable regulatory system that enhances the role of the banking system in stimulating economic growth. At the same time, considerable attention should be paid to designing a steady policy of interest rates to avoid the inherently undesirable impact of high interest rates on the economy.