Is export-led growth hypothesis still valid for sub-Saharan African countries? New evidence from panel data analysis

Purpose –This study examines the causal relationship between exports and economic growth in sub-Saharan African (SSA) countries during the period 1980 to 2017. The study also examines whether the causality between these two macroeconomic variables depends on the countries’ stage of development as proxied by their per capita income. Design/methodology/approach – The study uses a panel cointegration test and panel Granger-causality model to examine the link between exports and growth. The study also incorporates external debt as an intermittent variable in a bivariate setting between exports and economic growth, thereby creating a dynamic multivariate panel Granger-causality model. Findings – Although the study found the existence of a long-run relationship between exports and economic growth, the study failed to find any export-led growth response in both low-income and middle-income countries. Instead, the study found evidence of a bidirectional causality and a neutrality response in middleincome and low-income countries, respectively. The study, therefore, concludes that the benefits of an exportled growth hypothesis may have been oversold, and that the strategymay not be desirable to some low-income developing countries. Practical implications – These findings have important policy implications as they indicate that the causality between exports and economic growth in SSA countries varies with the countries’ stage of development. Consistent with the contemporary literature, the study cautions low-income SSA countries against over-relying on an export-led growth strategy to achieve a sustained growth path as no causality between exports and economic growth has been found to exist in those countries. Instead, such countries should consider pursuing new growth strategies by building the domestic demand side of their economies alongside their export promotion strategies in order to expand the real sector of their economies. For middleincome countries, the study recommends that both export promotion strategies and pro-growth policies should be intensified as economic growth and exports have been found to reinforce each other in those countries. Originality/value – Unlike the previous studies, the current study disaggregated the full sample of SSA countries into two subsets – one comprising of low-income countries and the other consisting of middle-income countries. In addition, the study uses a multivariate Granger-causality model in order to address the emissionof-variable bias. To our knowledge, thismay be the first study of its kind in recent years to examine in detail the causal relationship between exports and economic growth in SSA countries using an ECM-based multivariate panel Granger-causality model.


Introduction
The relationship between exports and economic growth has attracted numerous studies in recent decades. The thrust of the debate has been whether exports drive economic growth or Exports and economic growth in SSA whether it is the growth of the real sector that drives exports. While the former view is referred to as the export-led growth (ELG) hypothesis, the latter is popularly known as the growth-led export (GLE) hypothesis. According to the ELG hypothesis, real GDP growth does not only rely on the increase in the amounts of labour and capital, but also on the growth of exports through a multiplier effect. This makes export one of the engines of economic growth. Moreover, an increase in exports as a result of export-oriented policies can also indirectly stimulate economic growth through the efficient allocation of resources, greater capacity utilisation and exploitation of economies of scale (Awokuse, 2003). Apart from stimulating technological enhancement due to foreign market competition, exports also play a critical role in enabling investment and technological transfer, which accelerates the process of globalisation (see Keesing, 1967;Bhagwati and Srinivasan, 1975;Dervis, 1979) [1]. An increase in exports also provides foreign exchange, which can be used for importing capital goods and intermediate goods, thereby leading to higher capital formation, which, in turn, leads to higher economic growth (McKinnon, 1964;Balassa, 1978;Buffie, 1992). Indeed, the remarkable performance by a number of Asian countries can be attributed to the beneficial effects of exports on economic growth (see Salim and Hossain, 2011;Awokuse and Christopoulos, 2009;Lee and Huang, 2002;El-Sakka and Al-Mutairi, 2000) [2]. Although exports can significantly contribute to economic growth, some studies have argued that there is a danger in over-relying on exports to boost economic growth, especially in developing countries. This is mainly because the market for the exports of developing countries is limited by the capacity of industrialised countries. Hence, stagnation in demand in developed countries may lead to overinvestment and excess capacity in developing countries (see Blecker, 2002;Felipe, 2003). Moreover, some recent studies have argued that the benefits of an exportled growth hypothesis may have been oversold, and that the strategy may not be desirable to some low-income developing countries; hence, a new development paradigm is needed. According to Pillay (2011), there is a need for a shift towards a domestic demand-led growth strategy, while maintaining exports as countries still need exports to pay for their imported inputs and some finished goods that cannot be produced locally (see Pillay, 2011, p. 9).
As opposed to the ELG hypothesis, the GLE hypothesis postulates that an increase in economic growth could also lead to an increase in exports through a realisation of economies of scale and a reduction in the cost of production (see Bahmani-Oskooee, 2009). Previous studies have also argued that an increase in GDP is likely to lead to a corresponding increase in trade, unless an anti-bias trade is created by the growth-induced supply and the corresponding demand (Bhagwati, 1988). The GLE hypothesis has also been supported by the neoclassical trade theory. According to the neoclassical trade theory, economic growth, through its effects on the supply of the economy (factor endowments), may create more demand for exports within a country, thereby affording a country a strong export production base (Mahadevan, 2007).
Although a number of studies have been conducted on the relationship between exports and economic growth, especially since the 1960s, the majority of these studies have mainly been conducted on Asia and Latin America, thereby leaving many SSA countries with little or no coverage at all (see, for example, Ahmad et al., 2018;Ali and Li, 2018;Shakeel and Ahmed, 2020;Dinç and G€ okmen, 2019;Kalaitzi and Chamberlain, 2020, among others). Even where such studies have been conducted, the findings on the causal relationship between exports and economic growth remains mixed at best and controversial at worst. In addition, some of these previous studies have fundamental methodological weaknesses. It is against this background that the current study aims to examine the causal relationship between exports and economic growth in sub-Saharan African countries using a panel Granger causal model. In order to address the omission-of-variable bias, which has been reported in some of the previous studies, the current study uses a multivariate panel Granger-causality model to examine this linkage. In order to examine whether the causality between exports and economic growth depends on the countries' stage of development as proxied by their per capita income, the study disaggregated the full sample of SSA countries into two subsetsone comprising of low-income countries and the other consisting of middle-income countries.
To our knowledge, the studies that are closest to the current research are based on the work done by Ee (2016) and Ahmad and Kwan (1991). However, the current study differs fundamentally from these two studies in various ways. For example, Ee (2016) used fully modified OLS (FMOLS) and dynamic ordinary least square (DOLS) to test the export-led growth hypothesis, while the current study uses an ECM-based multivariate panel Grangercausality model to examine the short-run and long-run causality between exports and economic growth. In addition, in the current study, two panels of SSA countries are used, namely low-income and middle-income panels. Ahmad and Kwan (1991), on the other hand, used a bivariate Granger-causality model, while the current study uses a multivariate ECMbased Granger-causality model, which reduces the omission-of-variable bias and captures the short-run and long-run causal dynamics.
The rest of the paper is structured as follows. Section 2 reviews some of the empirical literature on the relationship between exports and economic growth in developing and developed countries. Section 3 deals with the methodology, empirical analysis and discussion of the results. Section 4 concludes the study.

Literature review
Previous studies on the relationship between exports and economic growth vary significantly between those that are in favour of the export-led growth (ELG) strategy and those that are in favour of growth-led export (GLE) strategy. Theoretically, the export-led growth (ELG) strategy hinges on whether a country should focus on export promotion or import substitution. In the main, the proponents of export-led growth theory support export promotion policy instead of import substitution policy. According to a comprehensive study by World Bank (1987), export-promotion strategy is the best strategy for less developing countries (LDCs) that intends to industrialise and transform their economies into more developed economies (see Tang et al., 2015). This view argues that growth could be achieved better through ELG strategies. A case in point is the growth rate that has been achieved by the Asian economies, such as Hong Kong, Singapore, South Korea, Taiwan, Malaysia and Thailand that were found to have been supported by the export promotions strategies. Over a period of 30 years, these countries were found to have doubled their standards of living every ten years (see Giles and Williams, 2000). According to the proponents of ELG theory, export growth leads to an increase in the demand for the country's output, which leads to an increase in real output. An increase in a country's exports may inter alia lead to an increase in the specialisation of export goods, which may, in turn, boost the country's productivity level and eventually leads to output growth (see Giles and Williams, 2000). In addition, the outward-oriented trade policy resulting from the ELG strategy may also give access to advanced technologies, learning by doing gains and better management practices, which may lead to further efficiency gains (see Giles and Williams, 2000;Hart, 1983;Ben-David and Loewy, 1998). Apart from the ELG, recent studies have shown that there is also a potential for growth-led export (GLE). Bhagwati (1988), for example, argues that an increase in GDP generally leads to a corresponding expansion of trade, unless the pattern of growth-induced supply and corresponding demand creates an anti-trade bias. Neoclassical trade theory also stresses the causality that runs from home-factor endowments and productivity to the supply of exports (see Findlay, 1984).
On the empirical front, there are a number of studies that have been conducted to examine the causal relationship between exports and economic growth in both developed and developing countries. However, the findings of such studies remain at best inconclusive and often contradictory. Broadly speaking, previous studies on this subject can be divided into Lanka; El-Sakka and Al-Mutairi (2000) for Kuwait, Libya, Qatar, Sudan and Tunis; Shirazi and Manap (2005) for Sri Lanka and India; Tang (2006) for China; Tang (2006) for China; Shirazi and Manap (2005) for Sri Lanka and India; more recently, Kalaitzi and Chamberlain (2020) for the case of the United Arab Emirates in the long run. Table 1 gives a summary of previous empirical findings on the causal relationship between exports and economic growth in both developed and developing countries, based on these four groups of studies.

Empirical analysis
3.1 Model specificationa trivariate Granger-causality model This study uses panel data and a trivariate Granger-causality model to examine the causal relationship between exports and economic growth in SSA countries. The use of this technique is deemed most suitable in this study because of the various advantages it renders. Firstly, a panel data technique has the ability to test more complicated behavioural models than a single cross-sectional or time-series data technique (see Hsiao, 2003). Secondly, panel data contains more degrees of freedom and more sample variability than cross-sectional or time-series data (Hsiao et al., 1995). Thirdly, panel data analysis generates more accurate predictions for individual outcomes by pooling the data rather than generating predictions of individual outcomes using the data on the individual in question (Hsiao et al., 1989(Hsiao et al., , 1993 [3]. The Granger causality model adopted in this study is expressed as follows (see Odhiambo, 2015):  : 1960India: 1960Bangladesh: 1973Sri Lanka: 1960Nepal: 1975 Exports →Y (Pakistan)

Data
The data used in this study cover the period 1980 to 2017. The studied countries were divided into two panels where data were availablelow-income panel and middle-income panel. The data were sourced from the World Bank's World Development Indicators. Although a number of proxies could be used to measure economic growth, in this study, real GDP per capita was used to measure the growth of the real sector. The advantage of using real GDP per capita is that it takes into consideration the effect of a population on economic growth. Some of the studies that have used this proxy include those of Shan et al. (2001), Thangavelu and James (2004), Rousseau and Vuthipadadorn (2005), Cooray (2010), Demirguc-Kunt et al. (2011), Odhiambo (2014, 2021, to mention a few. The exports variable is measured by the value of the exports of goods and services, while external debt, which has been used as an intermittent variable between exports and economic growth, is measured by the value of the external debt as a percentage of GNI.

The panel unit root test
In order to identify the order of integration of the variables used in the study, three panel unit root tests are employed: (1) Levin et al. (2002); (2) Table 2 for both low-income and middle-income countries.
The results of panel unit root tests reported in Table 2 show that the variables are consistently stationary in first difference.

The panel cointegration test
Having confirmed the order of integration of the variables used in this study, the next step is to examine the long-run relationship among these variables. For this purpose, two panel cointegration tests are employed in order to ensure the veracity of the findings. These are: (1) the Pedroni (2004) residual cointegration test; and (2) the Kao (1999) residual cointegration test. The cointegration results are reported in Table 3.
Overall, the results of the two panel cointegration tests reported in Table 3 reveal that the variables in the two models (1-2) are cointegrated; hence, the Granger-causality test could be performed.

Trivariate Granger-causality results
In this section, a dynamic multivariate panel Granger-causality model is employed to examine the causal relationship between exports, debt and economic growth in both low-LLC income and middle-income countries. The short-run causality is given by the F-statistics, which is expected to be statistically significant (see Asongu, 2014;Odhiambo, 2015). The long-run causality, on the other hand, is based on the coefficient of the error-correction term (ECT), which is expected to be negative and also statistically significant (see Odhiambo, 2021;Asongu et al., 2016). Table 4 presents the Granger-causality results for both low-income and middle-income countries. Based on the findings reported in Panel A, it is clear that exports do not Granger-cause economic growth in low-income countries. This applies irrespective of whether the causality is estimated in the short run or in the long run. The short-run causality has been rejected by the corresponding F-statistic in the growth equation, which has been found to be statistically significant. Likewise, the long-run causality has been rejected by the coefficient of the error correction term in the economic growth in low-income countries' panel, which has also been found to be statistically insignificant. The same findings apply to the reverse causality from economic growth to exports. This can be confirmed by the corresponding F-statistic in the export's equation, which has been found to be statistically insignificant. This finding, therefore, shows that there is no causal relationship between exports and economic growth in either direction in low-income countries. This finding, though contrary to some of the previous studies, is consistent with previous studies, such as Ahmad and Kwan (1991) for the case of 47 African Countries, Ahmed et al. (2000) for the case of Bangladesh, Pakistan and Sri Lanka and Shirazi and Manap (2005) for Sri Lanka and India, among others.
In middle-income countries (Panel B), the results show that there is bidirectional causal relationship between exports and economic growth. This applies irrespective of whether the causality is conducted in the short run or in the long run. The causal flow from exports to economic growth has been confirmed by the coefficient of the ECM term and the corresponding F-statistic in Panel B, which have been found to be both statistically significant. Likewise, the reverse causal flow from economic growth to exports has been confirmed by the coefficient of the ECT and the corresponding F-statistic in the export's equation, which have been found to be both statistically significant. Overall, the results of both low-income and middle-income countries show that the export-led growth paradigm, which gained prominence in the 1970s, may no longer be relevant to the countries under study. Other results show that for panel A, there is a long-run and short-run unidirectional causal flow from economic growth to debt in low-income countries. This is confirmed by the coefficient of the error correction term and the corresponding F-statistic in the debt equation, which have been found to be statistically significant. The results also show that for low-    Table 4. Granger-causality results for all models income countries, there is a unidirectional causal flow from debt to exports both in the short run and in the long run. This finding is confirmed by the coefficient of the ECM and the corresponding F-statistic in the export's equation, which have been found to be both statistically significant. In Panel B, the results show that there is a short-run unidirectional causal flow from exports to debt. This has been confirmed by the corresponding F-statistic in the debt equation, which has been found to be statistically significant. However, no causality was found to exist between economic growth and debt in either direction. This applies irrespective of whether the causality was estimated in the short run or in the long run.

Conclusion
In this study, the dynamic causal relationship between exports and economic growth has been examined. The study was motivated by the current debate on the export-led growth versus growth-led export nexus. Unlike in some previous African studies, in the current study, SSA countries are divided into two groups, namely low-income and middle-income countries. In addition, external debt has been used as an intermittent variable in a bivariate setting between exports and economic growth, leading to a multivariate panel Grangercausality model. Using an ECM-based panel Granger-causality model, the study found that there is a long-run relationship between exports and economic growth in both groups of countries. However, the causality between these two variables varies significantly between low-income and middle-income countries. Specifically, the study found a short-run and longrun bidirectional causality between exports and economic growth to prevail in middle-income countries. However, in low-income countries, no causality was found to exist between these two variables in either direction. This applies irrespective of whether the causality was estimated in the short run or in the long run. These findings have important policy implications as they indicate that the causality between exports and economic growth in SSA countries varies with the countries' stage of development. The study, therefore, concludes that the argument that exports always Granger-cause economic growth may have been oversold to many SSA countries. This finding is not surprising given the nature and the composition of the exports of many SSA countries. Indeed, the exports of many SSA countries, especially low-income countries, are dominated by primary products, whose prices are relatively low when compared to those of manufactured goods. Moreover, given the fact that industrialisation in some SSA countries has been relatively slow, some SSA countries have been forced to continue importing some consumer goods that could be produced locally, thereby leading to widening current account deficits. Consistent with the contemporary literature, the study cautions low-income SSA countries against over-relying on an export-led growth strategy to achieve a sustained growth path as no causality between exports and economic growth has been found to exist in those countries. Instead, such countries should consider pursuing new growth strategies by building the domestic demand side of their economies alongside their export promotion strategies in order to expand the real sector of their economies. For middle-income countries, the results show that the expansion of exports through various exports promotion strategies has been an integral component of their economic growth path. Consequently, the study recommends that both export promotion strategies and pro-growth policies should be intensified as economic growth and exports have been found to reinforce each other in those countries.