Fuel-mining exports and growth in a developing state: the case of the UAE

This study examines the causal effects of traditional UAE exports on economic growth over the period 1981-2012, using a neoclassical production function augmented with fuel-mining exports and imports of goods and services. To investigate the existence of a long-run relationship between fuel-mining exports and economic growth, the study applies the Johansen cointegration test, while the direction of the short-run causality is examined by applying the Granger causality test in a vector error correction model framework. In addition, a modified Wald test in an augmented vector autoregressive model, developed by Toda and Yamamoto (1995), is used to investigate the existence of a long-run causality between the variables. The cointegration analysis confirms the existence of a long-run relationship between the variables, while fuel-mining exports are found to have a negative impact on economic growth. Moreover, the study finds that fuel-mining exports do not cause economic growth in the short-run or the long-run.


INTRODUCTİON
Many studies have investigated the relationship between exports and economic growth, noting the positive effect of exports on growth. However, some studies have shown that aggregate measures may mask the different causal effects of subcategories of exports. Total exports may have a positive effect on economic growth, while disaggregated exports have an inverse effect, or vice-versa. This inverse effect often arises in countries in which primary exports comprise a large proportion of total exports.
The UAE is a country that relies heavily on oil exports and has experienced a high rate of economic growth, but are primary exports the cause of this growth? There is no published evidence on the causal relationship between fuel-mining exports and economic growth for the UAE. This study, in investigating this relationship, is intended to contribute to the discourse on future policies for sustaining economic growth in the UAE and other oil-producing countries.
This may lead to biased conclusions, as causality tests are considered to be sensitive to omitted variables. Moreover, most recent studies have examined the existence of a long-run causality between exports and economic growth based on error correction models (ECMs), but in the case of multivariate ECMs, the long-run causal effect of each variable on the dependent variable cannot be identified. To overcome these issues, this study uses an augmented production function, including variables omitted in previous studies, and performs the Toda-Yamamoto long-run causality test.
The empirical analysis confirms the existence of a long-run relationship between the variables, while fuel-mining exports are found to have a negative impact on economic growth. Moreover, the study finds that fuel-mining exports do not cause economic growth in the short-run or the long-run.
The remaining sections of this paper are organized as follows: Section 2 presents the literature on the relationship between exports and economic growth, while the chosen methodology and data sources are presented in Section 3. Section 4 reports and interprets the empirical results, while Section 5 presents the conclusions and policy implications of this research.

LİTERATURE REVİEW
A number of studies have described the positive impact of export expansion on economic growth, through enhanced economies of scale, adoption of advanced technology and greater capacity utilization (Feder, 1982;Lucas, 1988;Al-Yousif, 1997;Vohra, 2001;Abou-Stait, 2005). In particular, export growth increases investment in those sectors in which a country has a comparative advantage, fostering the adoption of advanced technologies, increasing national production and raising the rate of economic growth. Moreover, an increase in exports increases the inflow of foreign exchange, allowing the expansion of imports of services and capital goods, which are essential to improving productivity and economic growth (McKinnon, 1964;Chenery and Strout, 1966;Gylfason, 1998;Shirazi and Manap, 2004). This is referred to as the export-led growth (ELG) hypothesis. In contrast, some studies provide evidence that the causality runs from growth to exports, referred to as the growth-led exports (GLE) hypothesis, or conclude that there is a bi-directional causal relationship (ELG-GLE) between exports and economic growth (Panas and Vamvoukas, 2002;Love and Chandra, 2005;Elbeydi et al., 2010;Kalaitzi and Cleeve, 2017).
At the same time, other studies have found that exports have a negative effect on economic growth (Myrdal, 1957;Meier, 1970;and Lee and Huang, 2002;Kim and Lin, 2009). This negative effect may arise from the fact that in some countries primary exports constitute a large share of total exports. Primary exports may be subject to large price fluctuations (Myrdal, 1957), and seldom offer the knowledge spillovers and other externalities of manufactured good exports (Herzer et al., 2006). In general, as Sachs and Warner (1995) note, primary exports comprising a large share of total exports are associated with lower economic growth.
These findings have led a number of economists to investigate the impact of export categories on economic growth, as aggregate measures may mask the different causal effects of subcategories of exports. In particular, studies by Fosu (1990), Ghatak et al. (1997), Siliverstovs and Herzer (2007), Hosseini and Tang (2014) and Kalaitzi and Cleeve (2017) examine the effect of export composition on economic growth, and find that not all exports affect economic growth equally.
In particular, the study by Fosu (1990) argues that the heterogeneity of exports plays an important role in explaining the economic growth differentials among countries. The study uses crosssectional data for sixty-four developing countries over the period 1960-1980, to examine the effect of manufacturing and primary exports. Fosu (1990) uses ordinary least squares based on a production function augmented with non-manufacturing and manufacturing exports growth. The results indicate that nonmanufacturing exports have a negligible impact on economic growth; in contrast, the manufacturing export sector has a positive and significant effect. Ghatak et al. (1997) examine the ELG hypothesis for Malaysia over the period . Their study focuses on the effects of aggregate and disaggregated exports on real GDP and nonexport real GDP, using cointegration and causality testing. Two production functions are used, where, in addition to human and physical capital, manufactured exports, fuel primary exports and non-fuel primary exports are included as inputs. Real gross domestic investment as a percentage of real GDP is used as a proxy for physical capital, while the enrolment ratio in primary and secondary school is used as a proxy for human capital. This study provides evidence that non-fuel primary exports have a significant negative causal effect on economic growth. Siliverstovs and Herzer (2007) examine the ELG hypothesis using annual time series data for Chile, over the period 1960-2001. This study uses Johansen's cointegration methodology to investigate the impact of manufactured and mining exports. Siliverstovs and Herzer (2007) estimate an augmented neoclassical production function, including manufactured exports, mining exports and imports of capital goods. As for the human capital and physical capital variables, the study uses total working population and accumulated capital expenditure. The results indicate that there is a uni-directional Granger causality from manufactured exports to economic growth, while a bi-directional causality exists between mining exports and real non-export GDP. Hosseini and Tang (2014) examine the causal effects of oil and nonoil exports in Iran, over the period 1970-2008, using a production function augmented with oil exports, non-oil exports and imports of goods and services. This study applies Johansen's cointegration test and the Granger causality test to examine the existence of a long-run relationship and the direction of the short-run causality respectively. The cointegration results indicate that a long-run relationship exists between the variables, while oil exports and imports are found to have a negative effect on economic growth. The Granger causality results show evidence of a uni-directional short-run causality from oil exports and non-oil exports to economic growth, indicating that the ELG hypothesis is valid in the case of Iran.
International Journal of Energy Economics and Policy | Vol 10 • Issue 4 • 2020 302 Kalaitzi and Cleeve (2017) investigate the causality between primary exports, manufactured exports and economic growth in the UAE over the period 1981-2012, using a Cobb-Douglas production function augmented with primary exports, manufactured exports and imports of goods and services. This study employs the Johansen cointegration test, while the Granger causality test in a VECM framework and the Toda-Yamamoto causality test are performed to find the short-run and long-run causality respectively. The cointegration results show that there is a long-run relationship between the variables, while manufactured exports contribute more to economic growth than primary exports. This study also provides evidence to support a bi-directional short-run causality between manufactured exports and economic growth, while no causality exists between primary exports and economic growth. In the long-run, economic growth causes manufactured exports in the case of UAE.
The present study extends the work of Kalaitzi and Cleeve (2017), by testing the validity of the ELG hypothesis, focusing on the causality between fuel-mining exports and economic growth in the UAE for the same period.

RESEARCH METHODOLOGY
The relationship between fuel-mining exports and economic growth, assuming that aggregate production can be expressed as a function of physical capital, human capital, fuel-mining exports and imports of goods and services, is as follows: Where Y t denotes the aggregate production of the UAE economy at time t, A t is total factor productivity, and K t and HC t represent physical capital and human capital respectively. The constants α and β are between zero and one, measuring the impact of physical capital and human capital on national income. As mentioned above, in order to test the relationship between fuelmining exports and economic growth, it is assumed that total factor productivity can be expressed as a function of fuel-mining exports, FX t , imports of goods and services, IMP t , and other exogenous factors C t : Combining equations (1) and (2), the following equation is obtained: α, β, γ and δ represent the elasticities of production with respect to the inputs of production: K t , HC t , FX t , and IMP t . After taking the natural logs of both sides of equation (3), we obtain the following: where c is the intercept, the coefficients α, β, γ and δ are constant elasticities and ε t is the error term, which reflects the influence of other factors not included in the model.
The study uses annual time series for the UAE over the period 1981-2012, obtained from national and international sources. Specifically, gross domestic product (Y) is obtained from the World Bank, population (HC) is taken from the National Bureau of Statistics of the UAE, while fuel-mining exports (FX) come from the World Trade Organization. The data series for imports of goods and services (IMP) and gross fixed capital formation (K) are taken from the IMF, the UAE National Bureau of Statistics and the World Bank. All the variables are expressed in logarithmic form and real terms, using the GDP deflator obtained from the World Bank. Table 1 provides a brief summary of the descriptive statistics for all variables used in this study.
In order to investigate the existence of a causal relationship between fuel-mining exports and economic growth, the study applies the following tests: (a) unit root tests, in order to assess the stationary properties of the variables; (b) a cointegration test to confirm the existence of a long-run relationship between the variables; (c) the multivariate Granger causality test to find the direction of any short-run causality; and (d) a modified Wald test (MWALD) in an augmented vector autoregressive model, to investigate the existence of long-run causality between exports and economic growth.

Unit Root Tests
This study uses the augmented Dickey-Fuller (ADF) test (1979), the Phillips-Perron (PP) test (1988) and the unit root test with a structural break proposed by Saikkonen and Lutkepohl (SL) (2002). In each test, the null hypothesis, H o , is that a unit root exists, while the alternative hypothesis, H A, is that the time series is stationary. In the case where the series is found to be nonstationary, the first difference is used. In particular, the ADF test is based on the following equations: α 0 and α 2 represent the deterministic elements, equation (5) is a random walk with intercept and time trend; equation (6) is a random walk with intercept only; while equation (7) is a random walk (Gujarati, 2003). The random errors, ε t , are assumed to be identically distributed, with zero mean and variance σ 2 {ε t ~ ii(0,σ 2 ) for t = 1, 2,…}, and not correlated. The PP unit root test, which proposes a semi-parametric correction of serial correlation and time-dependent heteroskedasticity (Enders, 1995), involves the following equations: T is the number of observations and the error term μ t is such that E(μ t ) = 0, but is not necessarily serially uncorrelated. If there are structural breaks in the data, the ADF and PP test statistics are biased toward the non-rejection of a unit root. The SL test is based on the following equations, allowing the inclusion of a shift dummy variable.
μ 0 is the constant term, μ 1 and δ are the coefficients of the trend term and the shift dummy variable respectively, while u t is the error term. d 1t is a shift dummy variable with break date T break : d 1t = 0, for t<T break and d 1t = 1, for t>T break .

Cointegration Test
To test the existence of a long-run relationship between the variables, this study performs the Johansen cointegration test (Johansen, 1988), which is based on the following restricted vector autoregressive model (VECM): and D t is a vector of the deterministic variables 1 . If the coefficient matrix Π has rank r<n, but is not equal to zero, the variables are cointegrated and r is the number of cointegrating vectors. The number of cointegrating vectors can be determined by using the likelihood ratio (LR) trace test statistic suggested by Johansen (1988). The LR trace statistic, which is adjusted for small sample size, as proposed by Reinsel and Ahn (1992) 2 , is given by where T is the sample size and λ is the eigenvalue. The trace test is a test of the null hypothesis of at most r cointegrating vectors against the alternative hypothesis of n cointegrating vectors.

Short-run Granger Causality Test
The Granger causality test is used to examine the direction of the causality between fuel-mining exports and economic growth.
1 The Pantula Principle (Pantula, 1989) is used to determine the appropriate specification of the deterministic term in the VECM. 2 Trace statistics are adjusted by using the correction factor (T-n*p)/T proposed by Reinsel and Ahn (1992), where T is the sample size, and n and p are the number of variables and the optimal lag length respectively.
If the variables are cointegrated, the causality can be tested by estimating the following VECM model: LY t represents economic growth, while LK t , LHC t , LFX t and LIMP t represent the independent variables of equation (1). Δ is the difference operator, β ij , γ ij , δ ij , ζ ij , θ ij , μ ij and λ ij are the regression coefficients and ECT t-1 is the error correction term derived from the cointegration equation.
Once the models have been estimated, the following diagnostic tests are conducted in order to determine whether the models are well-specified and stable: (a) the Jarque-Bera normality test; (b) the Portmanteau test; (c) the Breusch-Godfrey LM test; (d) the White heteroskedasticity test; (e) the multivariate ARCH test; and (f) the AR roots stability test. In addition, the parameter constancy of the ECM estimates are assessed by applying the cumulative sum of recursive residuals (CUSUM) and the CUSUM of squares (CUSUMQ) tests proposed by Brown et al. (1975). The CUSUM and CUSUMQ tests detect systematic and haphazard changes in the parameters respectively. In particular, the CUSUM test is based on the statistic: 1 / , 1, s is the standard deviation of the recursive residuals (w t ), defined as: The numerator 1 − − t tt y xb is the forecast error, b t-1 is the estimated coefficient vector up to period t-1 and ´t x is the row vector of observations on the regressors in period t. The X t-1 denotes the (t-1) × k matrix of the regressors from period 1 to period t-1. If the b vector changes, W t will tend to diverge from the zero mean value line; if the b vector remains constant, E(W t )=0. The test shows parameter instability if the cumulative sum of the recursive residuals lies outside the area between the two 5% significance lines, the distance between which increases with t.
The CUSUM of squares test uses the square recursive residuals, w t 2 , and is based on the plot of the statistic: where t=k+1,…., T. The expected value of S t , under the null hypothesis of the b t 's being constant, is E(S t )=(t-k)/(T-k), which changes from zero at t=k to unity at t=T. In this test, the S t are plotted together with the 5% significance lines. Movements outside the 5% significance lines indicate instability in the equation during the period examined. After assessing the stability of the model parameters, the causality from fuel-mining exports to economic

Long-run Granger Causality Test
This paper applies the modified version of the Granger causality test (MWALD) proposed by Toda and Yamamoto (1995). This test is based on the following model: p is the optimal lag length, selected by minimising the value of the Schwartz information criterion (SIC), while dmax is the maximum order of integration of the variables in the model. The selected lag length (p) is augmented by the maximum order of integration (dmax) and the chi-square test is applied to the first p VAR coefficients.

Unit Root Tests
Before testing for the causality between fuel-mining exports and economic growth, the order of integration of the variables is examined. Tables 2 and 3 present the results of the ADF, PP and SL tests for the logarithmic level and first difference of the time series respectively.
The results of the unit root tests at the log level indicate that the null hypothesis of non-stationarity cannot be rejected for any of the variables at the 5% significance level. In contrast, after taking the first difference of LY, LK, LFX and LIMP, the null hypothesis of the presence of a unit root can be rejected at the 1% level, while the firstdifferenced series of LHC is found to be stationary at 5%. Hence, the time series for the period 1981-2012 are integrated of order one I(1).

Cointegration Test
The Johansen cointegration test is conducted in order to investigate the existence of a long-run relationship between LY, LK, LHC, LFX and LIMP. Table 4 shows that the null hypothesis of no cointegration is rejected at the 5% level, indicating the existence of one cointegrating vector. This suggests that real GDP, real gross fixed capital formation, population, real fuel-mining exports and real imports are cointegrated and follow a common long run path.
The cointegrating vector is estimated after normalizing on LY.
The following long-run relationship is obtained, with the absolute t-statistics reported in the parentheses: percent increase in imports can lead to an increase in real GDP of 0.322%. These results suggest that fuel-mining exports do not enhance economic growth in the long-run.

Granger Causality in VECM
The short-run causality results reported in Table 5 show that the null hypothesis of non-causality from fuel-mining exports to economic growth can be rejected at the 10% significance level, indicating some support for the ELG hypothesis in the short-run. However, the null hypothesis of non-causality from economic growth to fuelmining exports cannot be rejected at any conventional significance level. In other words, the GLE is not valid in the case of fuel-mining exports. As for the other causal relationships, a uni-directional Note: * , ** , ***Indicate significance at the 10%, 5% and 1% levels respectively (df in parentheses). The diagnostic tests for the VECM model show that serial correlation is not present, while the residuals are multivariate normal and homoscedastic. In addition the stability of the VECM is confirmed based on calculations of the inverse roots of the characteristic AR polynomial.
Note: Arrows indicate the direction of Granger causality between the variables  Note: ** , ***Denote the rejection of the null hypothesis of a unit root at 5% and 1% respectively. Numbers in [ ] corresponding to the ADF and SL test statistics are the optimal lags, chosen based on the SIC. Bandwidth in { } (Newey-West automatic) using the Bartlett kernel estimation method. Critical values for SL test are tabulated in Lanne et al. (2002). The maximum lag length for the ADF test is found by rounding up P max =[12* (T/100) ¼ ]=[12* (32/100) ¼ ]≈ 9 (Schwert, 1989). For the ADF and PP tests, all the time series are tested for the unit root including intercept and trend (a), intercept only (b) and no constant or trend (c). The letters in brackets indicate the selected model following Dolado et al. (1990). The years in the   Lanne et al. (2002). The maximum lag length for the ADF test is found by rounding up P max =[12* (T/100) ¼ ]= [12* (32/100) ¼ ]≈ 9 (Schwert, 1989). For the ADF and PP tests, all time series are tested for the unit root including intercept and trend (a), intercept only (b) and no constant or trend (c). The letters in brackets indicate the selected model following Dolado et al. (1990). The years in the table refer to the shift dummy variable d 1t with break date T break : d 1t =0, for t<T break and d 1t =1, for t>T break in equations 10 and 11 causality runs from physical capital to fuel and mining exports at a 5% level of significance, while fuel-mining exports Granger cause imports at 1%. These results show that investments cause the expansion of fuel-mining exports, allowing the expansion of imports of services and capital goods, which are essential to improving productivity and economic growth (Gylfason, 1998;McKinnon, 1964;Chenery and Strout, 1966).
In addition, a Chi-square test is performed to investigate the joint significance of the explanatory variables. The results indicate that the variables ΔLY t , ΔLK t , ΔLHC t and ΔLFX t jointly cause imports, ΔLIMP t, at the 5% significance level. Table 6 summarizes the short-run Granger causality results.
The structural stability of the parameters of equation (14) is tested by applying the cumulative sum of recursive residuals (CUSUM) and the CUSUM of squares (CUSUMQ). The CUSUM plots ( Figure 1) show that there is no movement outside the 5% critical lines of parameter stability. Therefore, the model for economic growth is stable, even during the oil crises of 1986 and 2000. The CUSUM plots (Figure 2) for the estimated ECM for fuel-mining exports (equation 17) show that there is no movement outside the 5% critical lines. The model for fuel-mining exports is thus also stable and, again, even during the oil crises of 1986 and 2000.

Toda-Yamamoto Granger Causality Test
The maximum order of integration of the variables is dmax = 1, while the optimal lag length is two. Therefore the selected lag length (p = 2) is augmented by the maximum order of integration (dmax = 1) and Wald tests are applied to the first p VAR coefficients. The results are presented in Table 7.
The results of the Wald tests show that there is no evidence to support the ELG hypothesis in the long-run, as the null hypothesis that LFX does not Granger cause LY cannot be rejected at any conventional significance level. According to Siliverstovs and Herzer (2007), such results show evidence of the productivity-limiting effects of fuel-mining exports. In addition, the absence of any direct long-run causality from economic growth to fuel-mining exports indicates that the GLE is not valid.
In contrast, a long-run bi-directional causality exists between imports and economic growth. In particular, the null hypothesis that LIMP does not Granger cause LY can be rejected at the 10% level, while LY Granger causes LIMP at the 1%. This suggests that economic growth can increase the country's capacity to import essential materials for domestic production, improving existing technology and leading to further economic growth. Moreover, imports are also affected directly by physical capital at the 1% level.
At the same time, a long-run causality runs from physical capital and imports to fuel-mining exports at 1% and 5% significance respectively, suggesting that investments in advanced technology and imports contribute to the expansion of fuel-mining exports.
Therefore, economic growth indirectly causes the expansion of fuel-mining exports. In addition, the results show that LY, LK, LHC and LIMP jointly Granger cause LFX in the long-run at the 1% level, while all variables in the model jointly cause LIMP at 1%. Figure 3 summarizes the long-run causal relationships among the variables in the model.

CONCLUSION
The cointegration analysis of the model confirms the existence of a long-run relationship between GDP, physical capital, human capital, fuel-mining exports and imports, while fuel-mining exports are found to have a negative impact on economic growth. These findings are consistent with previous studies, which argued that this category of exports does not enhance economic growth in the long-run (Myrdal, 1957;Sachs and Warner, 1995;Herzer et al., 2006;Hosseini and Tang, 2014;Kalaitzi and Cleeve, 2017). Moreover, the study finds that fuel-mining exports do not cause economic growth in the short-run or the long-run.
UAE fuel-mining exports have been decreasing steadily since 1981, with fuel mining exports as a proportion of total merchandise exports decreasing from 84% in 1981 to 31% in 2014. In other words, the UAE is moving away from fuel-mining as its primary Source: Created by the authors for the purpose of this study  Note: * , ** and *** indicate significance at the 10%, 5% and 1% levels respectively. The diagnostic tests for the select VAR(p) model prior to the application of the Toda-Yamamoto procedure show that serial correlation is not present, while the residuals are multivariate normal and homoscedastic.
source of export revenue. Further reductions in the relative importance of fuel-mining exports should enhance the rate of economic growth, with greater emphasis on imports and physical capital accumulation, directly or indirectly increasing growth in the long-run.