1The African continent is endowed with large mineral reserves. According to the United Nations, Africa counts with 30% of the world’s mineral reserves.[1] From the global energy reserves, Africa shares 3.6% of coal reserves, 7.5% of natural gas reserves, and 7.6 oil reserves (United Nations Environment Programme [UNEP], 2017). In 2010 Mozambique stood out with its potential in mineral resources, including natural gas, coal, heavy sands, and precious stones, among others (Centro de Integridade Pública [CIP], 2016). The primary focus in Mozambique's mining sector is on coal, with estimated reserves of 38.4 billion tons ( Extractive Industries Transparency Initiative [EITI], 2020). In recent years, there has been a notable contribution of the various natural resources to Mozambique’s economic growth, through the production and export of these resources. For the case of coal, it is major contribution started in 2012, with the opening of the Moatize coal mine, one of the world’s biggest coal mines by reserve, currently operated by Vulcan. Data from Ministry of Mineral Resources and Energy of Mozambique (MIREME) showed that coal export increased from 2.26 tons in 2012 to 13.26 tons in 2018. According to the National Institute of Statistics (INE) data, the coal exports value was 596.58 million USD in 2012 and 1.48 billion USD in 2020. The World Bank (WB) Coal Rents Ranking, ranked Mozambique as the second country with the largest contributions from the coal sector to GDP. Since 2014, Mozambique coal rents grew 26.6 points year to year (WB, 2019). [2]
- 3 Based on IMF yearly exchange rate
2The contributions of the coal sector to the economy come from foreign megaprojects in the mining sector. Among the multinationals that produce coal in Mozambique, especially metallurgical and thermal coal, are the Brazilian Vale Mozambique, Indian Coal Venture Limited (ICVL), and Jindal Steel Power Limited (JSPL), both from India. These companies produce and sell coal in the foreign market, particularly China and India. They respond to the growing coal demand in the metallurgical and electrical industry, which corresponds to one use of metallurgical and thermal coal, respectively. These companies have largely contributed to the government budget, through the payment of coal exploitation taxes. Data from the Extractive Industry Transparency Initiative (EITI) showed that between 2010 and 2018, the government of Mozambique received from Vale Mozambique, approximately 349.34 million USD; 9.05 million USD from ICVL, and 3.16 million USD from JSPL. [3] Since 1999, megaprojects exports in Mozambique have been contributing to GDP growth by about 3.2% (Saxegaard, 2008).
3Although the above growing scenario of the coal trade in Mozambique, there is a concern regarding the vulnerability of the country in facing a resource curse phenomenon. Natural resource curse is a phenomenon that refers to slow or negative growth in the resource-rich countries (Sachs and Warner,1995). According to the authors when natural resources abundance provides a positive impact on economic growth in the short-term but in the long-term the result may be different, causing a lower Gross National Product (GNP) than the initial boom of natural resources, i.e., the country is likely to experience the natural resource curse or Dutch disease.
4Green and Otto (2014) analyzed the context of resource abundance in Mozambique and concluded to be highly vulnerable to a resource curse caused by the negative impact that natural resource wealth. In the same perspective, Colom-Jaén and Bidaurratzaga-Aurre (2015) emphasized the point of whether foreign energy companies and government policies will benefit both the energy sectors and the population, or will bring a resource curse. According to Orre and Rønning (2017) the resource curse has already sparked in the country and the phenomenon may get worst with the start of Liquefied Natural Gas (LNG) projects. Resenfeld (2012) outlined three channels through which the coal industry might affect the Mozambican economy negatively, namely: Dutch disease, exchange rate volatility, corruption, and waste.
5The above-mentioned studies made an overview of the resource curse based on the dynamic context of the natural resource abundance and exploitation. However, no estimation was made on how the abundance and dependence of mineral resources affect Mozambican growth and human development. Therefore, this study fills this gap and makes contributions to the literature by applying econometric methods. This will allow short and long-run estimations providing a clear conclusion of the resource curse phenomenon.
6Some authors like Murshed (2004) have argued that it is not a set of natural resource endowments that intrinsically lead to growth but the type of natural resource. Therefore, the study is based on a specific natural resource sector, the coal sector. The main aim of this study is to measure the coal abundance and dependence on Mozambique’s economic growth and human development, which will help to answer the flowing research question: are coal resources a blessing or a curse?
7The main aim of this study is to measure the coal abundance and dependence on Mozambique’s economic growth and human development, which will help to answer the question: are coal resources a blessing or a curse?
8The paper is structured as follows: Section 2 presents theoretical and empirical evidence from the literature on natural resource abundance and dependence regarding the effects on economic growth. Section 3 describes the data and empirical method employed in the study. The empirical estimations and discussion of the results are presented in section 4. Finally, section 5 points out the study's main findings and recommendations.
9There has been a vast empirical discussion about the effect of natural resources on economic growth. However, there is no consensus on the effect of natural resource endowments on economic growth. Havranek et al., (2016) based on a meta-analysis found that 40% of empirical papers reported a negative effect of natural resources on economic growth, 20% found a positive effect and 40% found no effect. One study that is considered to be one of the preeminent studies addressing the relation between natural resources and economic growth is Sachs and Warner (1995). The authors found that in the long-term, economic growth in resource-rich countries grows more slowly than in resource-scarce countries. The negative correlation between resource abundance and GDP growth remains one of the most robust findings in empirical studies, and it is known as the resource curse hypothesis (Bulte et al., 2005). Du et al. (2020) found a resource curse in China, one of the largest producers and consumers of coal in the world. They found that the negative impact of resource dependence in high-quality economic development was due to the crowding-out effect of innovation investment and talents.
10Nhabinde and Heshmati (2020) analyzed the direct and indirect impacts of extractive industries on economic growth. They made their analysis in 11 founding countries of the Southern African Development Community (SADC). They found evidence of a resource curse given that extractive industries have direct negative impacts on the countries’ economic growth. The authors found that the validation of the resource curse hypothesis was high in Malawi and Mozambique, followed by Lesotho, Tanzania, Zimbabwe, and Zambia, respectively. Also, based on cross-country data, Saeed (2021) found robust evidence of a negative association between per capita growth and natural resource dependence. The author found this presence of a natural resource curse to persist in a sub-sample of 14 oil-rich Middle East and North Africa (MENA) countries. Pérez and Claveria (2019) explained the negative association between resource rents and human development in 10 African countries (including Mozambique), because of a high level of corruption that may impede economic development.
11Lederman and Maloney (2007) used different econometric techniques with panel data and found natural resource exports to have a positive rather than a negative effect on subsequent economic growth. Nasir and Redmond (2020) used Fixed Effects (FE), Random Effects (RE), Panel Fully Modified Least Squares (FMOLS), and Panel Dynamic Least Squares (DOLS) to investigate the impacts of natural resource abundance on economic growth and a human development index. Their panel study of 30 countries revealed that natural resource abundance has a positive impact on economic growth, while a negative effect was found on human development. Aljarallah and Angus (2020) used ARDL modeling to understand the dilemma of natural resource abundance. They found that resource rents increase per capita GDP merely in the short run. In the long run, they found that resource rents deteriorate productivity, human capital, and institutional quality. Using an ARDL approach, Kwakwa et al. (2021) found natural resource rents to significantly and positively affect GDP growth. Using the same model, Satti et al. (2014) found evidence of the resource curse hypothesis, since natural resource abundance hinders economic growth.
12Havranek et al. (2016) explained the conflicting evidence in understanding the effect of natural resources on growth as being associated with distinguishing growth between different natural resources also between different measures of natural resources wealth. Some studies use resource abundance measures, such as resources per capita (e.g. Brunnschweiler and Bulte, 2008; Lashitew and Werker, 2020) and others based on resource dependence measures, such as the contribution of resources to exports or GDP (e.g. Ding and Field, 2005; Boschini et al., 2013)
13Bulte et al., (2005) stated that the use of primary exports to GDP or total exports is an imperfect proxy in investigating the resource curse as primary export shares include non-resource output, such as agricultural production. To avoid this, the study focuses only on mineral exports and coal endowment. Thus, the study considers coal resource abundance by using Sachs and Vial's (2001) alternative measure of resource abundance, i.e., the share of natural resource exports in total mineral exports and natural resource exports as a share of GDP by Sachs and Warner (1997).
14For the coal dependence measure, coal rents were used as a percentage of GDP (e.g., Lashitew and Werker, 2020 and Aljarallah, 2020). The study looks at the effect of both coal abundance and dependence on GDP per capita and the Human Development Index (HDI). The study also includes regressors such as coal taxes, inward Foreign Direct Investment (FDI), and Corruption Perception Index (CPI) in interaction with coal resource abundance.
15Corruption was pointed out by Resenfeld (2012) to be a source of the resource curse in the coal industry in Mozambique. Considering the natural resource abundance context in Mozambique, Green and Otto (2014) have identified six risk factors for conflict, namely: its history of violence and political party dynamics; geographic distribution; poverty and inequality; expectations; corruption; and the cost of extraction. They found that a high level of corruption in the government does not bode well for an influx of natural resource rents. Busse and Gröning (2011) found that exports of natural resources lead to an increase in corruption. The study takes corruption in interaction with coal abundance, as a share of total mineral exports. With this interaction, a negative effect on economic growth is expected. Kwakwa et al. (2021) and Alpha and Ding (2016), found the interaction of corruption and natural resources export ends up in a decline in economic growth.
16Pineda and Rodríguez (2010), and Lashitew and Werker (2020) found evidence of changes in HDI as positively and significantly correlated with natural resource abundance. From the FDI-led growth hypothesis, many studies have found a positive relationship between the FDI inflows and the GDP per capita (see Hakizimana, 2015; Thanh et al., 2019). Studying the effect of natural resource abundance on economic growth, Amini (2018) found no significant effect. However, the investments in this sector increase, on average, the growth rate of real per capita GDP. Resource abundance shifts the reliance of the government from tax incomes to resource rents (Veisi, 2017). Also, natural resources are the main source of revenue in resource-rich countries (Kwakwa et al., 2021). Quixina and Almeida (2014) found that mining revenues Granger causes economic growth. Thus, it is expected that coal taxes have a positive effect on GDP growth.
17The study takes coal resource abundance and dependence to investigate their effect on economic growth (equation 1) and development (equation 2) in Mozambique. From the discussion in Section 2, different proxy measures of coal resource abundance and dependence were considered (Table 1). The study uses quarterly data from 2010Q1 to 2019Q4 got from different sources, as presented in Table 1. Therefore, the study relies on the empirical models’ equations, as presented below:
Table 1. Variables definition and source
18The study used the Autoregressive Distributed Lag (ARDL) approach developed by Pesaran et al. (2001), to find the long-run relationship between the explanatory variables and the economic growth and development. From equations (1) and (2), the ARDL models are specified:
19Compared to other co-integration techniques which require all the variables to be integrated in the same order, the ARDL bound test approach provides very efficient and consistent test results whether the variables are I(0) or I(1) and deals well with small and large sample sizes (Pesaran et al., 2001). The order of integration of the variables was first checked by performing unit root tests of the augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) tests.
20After estimating the equations (3) and (4), was performed the F-test by applying the coefficient diagnostic Wald restriction test on the long-term variable parameters. Following Pesaran et al. (2001) was tested the following hypothesis:
21The decision regarding the null hypothesis is made under the following conditions: if the computed F-statistic is greater than the upper critical bound value, then Ho is rejected meaning that the variables are co-integrated, otherwise, we fail to reject the null hypothesis and conclude that there is no long-run relationship among the variables.
22Once it has confirmed the long-run relationship between coal resource abundance and dependence on economic growth and development in Mozambique, the long-run equilibrium relationship between the variables was estimated. To do so, ARDL error correction model was estimated following the equation bellow:
23The study used Akaike’s Information Criterion (AIC), for the selection of lags. For model diagnostic were used different tests such as the cumulative sum of recursive residuals (CUSUM) and the cumulative sum of squares of the recursive residuals (CUSUM) tests to test for the stability of the model; Breusch–Godfrey test for residual serial correlation, Jarque–Bera for residual normal distribution, and ARCH test for heteroscedasticity.
24The process above allowed us to verify the existence of a causality effect between coal resource abundance and dependence on economic growth and development, though it does not show the direction of causality. Thus, to find the direction of the short-run, as well as long-run causality, was considered vector error correction model (VECM) which extends the error correction term. According to Granger (1986), VECM is more appropriate to examine the causality between the series if the variables are integrated at order one, I(1). Therefore, following Engel and Granger (1987) was apply the VECM Granger causality test as presented in the equations below:
25This study uses quarterly data covering the period from 2010Q1 to 2019Q4 and Table 1 presents summary statistics of the variables. In the period of analysis, the coal resource dependence was on average 2.01% while the coal resource abundance as a share of coal exports in total mineral exports was 13.84%. The coal resource abundance as a share of coal exports in per capita GDP was 1.86%. Still, in this period, the PGDP and HDI were on average 4.68% and 0.43%, respectively.
26From the standard deviation statistics, we can see how the values of the variables are close to the mean. In this perspective, with the high standard deviation, the values of CRAExp and the interaction Corr*CRAExp were found to be far from the mean, i.e., they spread out over a wider range of the normal distribution plot. From the values of Skewness, I found that the distribution of those variables is concentrated to the right, i.e., are right-skewed. The negative values of skewness were observed for the dependent variables and the independent variables of coal taxes and inward foreign direct investment. Given that the value of kurtosis is greater than 3 for most of the variables (except for HDI, CRD, and LnCoalTax) it said their distribution has heavier tails than a normal distribution. In this sense, extreme values can influence the estimation in the series.
Table 2: Descriptive statistics
27Table 3 reveals the correlation matrix between the variables. The results show a negative correlation between coal dependence, coal abundance, and corruption in interaction with coal abundance regarding the per capita GDP. The outcomes also reveal a positive correlation between tax paid by the coal firms and inward FDI in the mining and quarrying sector regarding the per capita GDP. Considering the negative correlation between coal resource dependence and abundance on economic growth can indicate presence of resource curse which is consistent with Bulte et al. (2005) findings. From the magnitude of the correlation coefficients, which vary from 0.3 to 0.5, we can say that the explanatory variables have a low correlation with per capita GDP.
Table 3. Correlation results among variables
28Table 3 also presents the relationship between human development measured by HDI and coal resource dependence and abundance. Contrary to the first model, in the second model, the results can show an absence of resource curse when looking at the human development perspective. The results reveal a positive and significant correlation (varying from moderate to strong correlation) between coal resource dependence and abundance and human development.
29By testing the stationarity of the series, we also get to know the order of integration of the variables, which is an important step before applying a co-integration technique. For this purpose, I applied Augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) tests, in Table 4. For both dependent variables, the null hypotheses of the presence of unit root in the series were rejected after applying the first difference in the series. This implies that both dependent variables are integrated of order one, i.e. I(1). From the independent variables, the results show that only coal resource abundance (measured coal exports share of GDP) and coal taxes were stationary at levels, which means to be integrated of order zero I(0). To confirm the existence of a long-run relationship among the variables was performed the ARDL bound test. This test provides very efficient and consistent results whether the variables are I(0) or I(1) (Pesaran et al., 2001). The results a presented in Table 5.
Table 4. Unit root test results
30The results in Table 5 show that, in both models, the null hypothesis of no cointegration is rejected. In model one, there is a long-run relationship between the explanatory variables and economic growth as the computed F-statistic exceeds the upper bounds critical values at a 10% significance level. These results imply the existence of a weaker relationship between the explanatory variables and economic growth. Contrary, in the second model the F-statistic exceeds the upper bounds critical values at 1%, 5%, and 10% significance levels. It can be said that there is a strong long-run relationship between per capita GDP, coal resource dependence and abundance, and economic development.
Table 5. ARDL bounds test for cointegration
31In the first model, the study looks at the effect of coal dependence and abundance, and other related variables on economic growth. Tables 6 and 7 present the estimated output for both long and short-run analyses, respectively. The results show that coal resource dependence significantly affects negatively economic growth in both the long and short-run (although positive on its first lag). A one percent increase in coal resource dependence leads to a decrease in the economic growth of 83.6% in the long run and 34.0% in the short-run, ceteris paribus. In the short-run, the results reveal that coal resource abundance has a positive effect on Mozambique’s economic growth. This, therefore, aligns with Aljarallah and Angus (2020) study and Sachs and Warner (1995) classical study. Both studies found natural resource abundance to have a positive effect on growth in the short-run, but in the long run, it turns into a natural resources curse. Recently, Nhabinde and Heshmati (2020) found the resources curse in extractive industries in Africa, being high in Malawi and Mozambique, followed by Lesotho, Tanzania, Zimbabwe, and Zambia, respectively.
32The results reported in Table 6 also revealed the significance of other coal-related variables. As expected, the inward foreign direct investment in the mining sector and also the taxes paid by the coal companies have a positive and significant effect on economic growth in the long run. These findings are aligned with the so-called FDI-led growth hypothesis and are also consistent with Hakizimana (2015) and Thanh et al. (2019) findings. Similarly, Amini (2018) found that investments in the natural resource sector increase, on average, the growth rate of real per capita GDP. This positive contribution of both FDI and coal taxes is reinforced by the World Bank 2019 data, which places Mozambique in the second position among the countries with the largest contributions from the coal sector to GDP.
33The results also showed that when corruption interacts with coal resource abundance, it negatively affects economic growth in the short-run. This implies that coal resource abundance can be a fertile environment for corruption to increase which compromises efforts for economic growth. This is consistent with the findings of Resenfeld (2012) who pointed out corruption as a source of the resource curse in the coal industry in Mozambique. Alpha and Ding (2016), and Kwakwa et al. (2021) found that an interaction between natural resources export and corruption leads to a decline in economic growth. The coefficient of the error correction term ECT (-1) is negative and highly significant, which reinforces the existence of a long-run relationship between economic growth and coal resource dependence, coal resource abundance, coal taxes, inward foreign direct investment, and corruption.
34Considering the estimated output, it can be said that Mozambique is prone to face the resource curse phenomenon in the coal sector. In this situation, the Mozambican government cannot benefit fully from coal resource wealth, and may not respond effectively to public needs. Various authors (such as Sachs and Warner, 1999; Asanuma, 2008 and NRGI, 2015) argue that the resource curse can cause public expenditures to increase political pressures that may jeopardize the fiscal and balance-of-payments sustainability; price volatility of the commodities; inefficient spending and borrowing; limited government capture of benefits; weaker institutional development, social and environmental problems and the so-called Dutch disease phenomenon (in which the revenues from natural resource exports damage a nation's productive economic sectors by causing an increase in the real exchange rate and wage).
Table 6: Long run ARDL Error Correction estimates
35From the development perspective, measured by the Human Development Index, the estimation shows a great significance of the coal resources variables in Mozambican development. Despite being highly significant, the results showed an unexpected sign for some variables. In both the short and long run, FDI and Coal Taxes (although not significant) negatively affect the HDI. Hilmawan and Amalia (2019) found coal and mineral revenues to positively affect HDI. Contrarily, coal resource dependence showed an opposite sign in both the short-run and long-run. The results show that a percent increase in coal dependence improves the HDI by 0.01 and 0.08 in the short and long run, respectively. Also, coal resource abundance, taken as a share of coal exports in total mineral exports, was found to have a significant and positive effect on HDI. These results corroborate with those found by Pineda and Rodríguez (2010) and Lashitew and Werker (2020). In this sense, aligned with the authors’ findings, coal resources may be a blessing rather than a curse for human development, primarily through their effects on improving the standards of living in poor communities, mainly those where coal resource is exploited. Pérez and Claveria (2019) found a negative association between resource rents and human development in 10 African countries (including Mozambique). The differences in outputs between Pérez and Claveria study and the present study can be because both studies analyzed different commodities. Pérez and Claveria took a whole set of mineral resources and the present study focus only on coal commodity, following Murshed (2004) observation (see Section 1).
36The interaction of corruption and coal abundance, as expected, reduces the human development index. Looking to the main channels of extractive activities and human development mentioned by Greener and Dodd (2015) this interaction can affect the Government revenue, employment, procurement, infrastructure, social investment, and investor returns, finally, this can take to a reduction of the poor communities’ wellbeing. The significance of the error correction term ETC (-1), which is also negative, confirms the existence of the long-run convergency between human development and coal resource variables.
Table 7: Short-run ARDL Error Correction estimates
37From Table 8, the Breusch–Godfrey results reveal the absence of first-order residuals for serial autocorrelation as the probabilities are great than 5% in both models. Similarly, the heteroscedasticity test results show that the residuals are homoscedastic. In the Jarque–Bera test for residual normality, the null hypothesis is that the skewness and excess kurtosis of the variables are zero. From the results, it is concluded that the residuals are normally distributed in model two at a 5% significance level, however, at the same significance level, the null hypothesis was not rejected in model 1.
Table 8: Diagnostic tests
38Further, the study assessed the stability of the model the stability basing on the cumulative sum (CUSUM) and cumulative sum of squares (CUSUMSQ) tests. From Figure 1 and Figure 2, it is observed that the parameters are stable, given that the CUSUM and CUSUMSQ lie within the two 5% critical bounds in both models.
Figure 1. The plot of CUSUM and CUSUMQ for coefficient stability of ECM model 1.
Figure 2. The plot of CUSUM and CUSUMQ for coefficient stability of ECM model 2.
39In this study, the Granger causality test shows the direction of causality among the coal resource variables and economic growth and human development. Table 9 presents both short and long-run Granger causality results. The results reveal the existence of a short-run unidirectional causality moving from PGDP to both types of coal abundance (coal exports share of GDP) and the interaction of corruption and coal abundance. Mohamed (2020) found a short-run unidirectional Granger causality relationship running from resource rents to PGDP growth. At a 10% significant level, coal taxes Granger causes PGDP. This means that coal taxes have a certain predictive effect on economic growth in the short run. Still, in the short run, the results show that coal dependence and coal abundance have a predictive effect on coal taxes at 5% and 1% significance levels, respectively. The short-run results also revealed a bi-directional Granger causality between PGDP and coal abundance, between the corruption and both measures of coal abundance. Alpha and Ding (2016) found a unidirectional Granger causality effect moving from natural resource abundance to corruption and also found a bi-directional causality effect between natural resource abundance and economic growth.
Table 9: Granger short-run and long-run causality test
40The long-run causality among the variables was confirmed by the significance of the estimated one period lagged error correction term (ECTt-1). The long-term causality results show that the bidirectional causality relationship only exists between PGDP and coal dependence, and also between PGDP and the interaction of corruption and coal abundance. In the long-run coal resources variables and HDI Granger causes PGDP. Shahbaz et al. (2017), Ben-Salha et al. (2018), and Mohamed (2020) found evidence of bi-directional causality between natural resources and economic growth in the long run.
41The study examined the existence of the resource course phenomena in the Mozambican coal sector. To do so, the study measured the effect of coal abundance and dependence on Mozambique’s economic growth and human development. The results revealed that coal resource dependence significantly affects negatively economic growth in both the short and long term. The results also revealed that coal resource abundance has a positive effect on Mozambique’s economic growth in short run, however, in the long-run, coal resource abundance yields a negative effect on Mozambique’s economic growth. From these results, the study concluded that the Mozambican coal sector is at risk of facing the resource curse. From the human development perspective, the results showed to exist a positive effect of coal abundance and coal dependence. This means that coal megaprojects contribute to relief the standards of life of the population. The study also found inward FDI in the mining sector and coal taxes to have a positive and significant effect on economic growth. In the long-run, when corruption interacts with coal abundance, the economic benefits are reduced. Granger causality test revealed that coal abundance, coal dependency, and other coal-related variables have a long-run predictive effect on economic growth. The study found bi-directional Granger causality between PGDP and coal abundance, also between the corruption and both measures of coal abundance. The study recommends similar analyzes to be carried out with other sectors, such as hydrocarbons, considering the large investments in LNG and oil in the Rovuma Basin. Given the limitations of the data, it is recommended future research with recent data on coal exports.
42Based on the results of the empirical analysis, the policy recommendation of the study is that the government should take advantage of the coal resource endowments and make a capital investment. Such investment should be aimed at coal manufacture, considering its vast use in thermal and metallurgical industries. This would reduce the strict dependence on the external market. Diversification should be also a key focus of the Mozambican economic policies. Diversification of over one key determinant of economic growth, such as agriculture and the manufacturing industry, could help control the negative effects of resource (in the case of coal) abundance and dependence. Also, revenues from natural resources should enhance the living standards of the communities. This can be done by investing in different dimensions of the human. The results showed the negative effect of the interaction between corruption and coal abundance on economic growth. In this sense, more work needs to be done to improve institutional quality and, therefore, control corruption.