The Resource Paradox: Coal Export Dependency, Environmental Degradation, and Growth Constraints in Jambi Province, Indonesia

The global reliance on coal as a primary energy source raises urgent concerns regarding long-term economic sustainability and environmental integrity. By 2024, global coal demand is projected to reach 8.7 billion tons and remain elevated through 2027 [1]. While this surge brings short-term economic benefits for exporting countries through revenues and sectoral growth, it also entails long-term structural risks. Coal price volatility illustrates this vulnerability, as fluctuations in coal market prices have been shown to impact broader economic and financial conditions in energy markets [2].

In resource-rich developing economies, prolonged dependence on fossil fuel exports often leads to structural stagnation, institutional fragility, and environmental degradation—commonly referred to as the resource curse. Many countries fail to transform resource wealth into broad-based growth due to rent-seeking, weak institutions, and limited diversification [3-7]. This vulnerability is reinforced by global trade dynamics: the Prebisch–Singer hypothesis posits a long-term decline in terms of trade for primary commodities, which increases risks for economies reliant on raw material exports [8, 9]. In Indonesia, mining-based provinces have been particularly vulnerable, with limited diversification, as illustrated by evidence from East Kalimantan, where mining dominance has not translated into broad-based income gains [10].

Coal exploitation also generates ecological and social pressures. The Environmental Kuznets Curve (EKC) hypothesis posits a non-linear relationship between economic growth and environmental quality: degradation worsens during early growth stages but may improve after reaching a certain income threshold, supported by stronger regulation and cleaner technologies [11, 12]. Yet, evidence from China, India, and South Asia shows that the turning point is often absent or delayed, particularly in coal-dependent regions with weak governance [13, 14]. In Indonesia, Prasetyanto and Sari [15] found that growth and primary energy consumption drive environmental degradation, although an EKC pattern may emerge if regulatory enforcement strengthens. Beyond environmental concerns, coal mining leads to land degradation, pollution, and increased health risks from particulate matter and heavy metals [16, 17]. Moreover, its economic benefits are unevenly distributed, with profits concentrated among elites while local communities experience widening socio-economic disparities [18, 19].

Several countries have addressed the structural challenges of resource dependence through institutional innovation. Norway, for example, redirected rents into a sovereign wealth fund to finance green transitions [20]. By contrast, most Indonesian provinces remain in the early stages of transition, and coal-producing regions, such as Jambi, have yet to reach the EKC turning point [21].

In Jambi Province, coal has increasingly dominated the export structure [22]. As shown in Figure 1, this growing dependence heightens vulnerability to global shocks and ecological stress while constraining economic diversification. At the same time, national policies—such as the government’s downstreaming agenda. At the same time, national policy directions related to energy transition and development priorities [23], the 2025 amendment of the mining law to broaden access for small enterprises and religious groups [24], and Indonesia’s pledge to achieve carbon neutrality before 2050 with a 15-year fossil fuel phase-out plan [25]—underscore the urgency of reducing coal dependence.

Figure 1. Volume of coal exports and total exports of Jambi province, 2009–2023

Despite extensive research on the resource curse and EKC, several gaps remain. First, most studies focus on national or macroeconomic effects, with limited attention to regional coal dependency. Second, prior work often uses aggregate production or exports, overlooking how concentration in a single commodity destabilizes economic resilience and ecological sustainability [26-29]. Third, many empirical studies employ linear models that fail to capture structural breaks triggered by policy reforms or global shocks [7, 28].

This study differs from previous provincial research in Indonesia in two key ways. First, it introduces the Coal Export Concentration Ratio (CECR) to quantify coal’s dominance in Jambi’s export profile. Earlier studies in East Kalimantan used coal activity as a proxy for emissions but ignored export dependence as a structural vulnerability [30]. Research in South Kalimantan analyzed the economic and environmental effects of coal mining but did not measure how export concentration amplifies external shocks [31]. Other Indonesia-focused and Sumatra-relevant analyses examined economic–environmental linkages without incorporating export concentration as a determinant [32].

Second, this study extends the export concentration literature [33-35], which typically uses concentration ratios (CR) and the Herfindahl–Hirschman index (HHI) at the national level, to a subnational context. Export concentration increases vulnerability to external shocks and growth volatility [36-38]; yet, no prior research has systematically linked provincial coal export concentration with both growth and environmental quality in Indonesia. By combining ARDL estimation with Bai–Perron (BP) multiple structural break tests, this study also addresses methodological gaps by capturing policy- and market-driven regime shifts often overlooked in linear models. This framing aligns with global evidence that energy series—coal in particular—exhibit structural breaks that affect their long-run properties [39] and with findings on the dynamic relationships between international energy price trends and coal prices in China [40].

Unlike previous provincial studies in Indonesia that rely on export values or export shares, this study introduces the CECR as a structural indicator of commodity dependence. CECR captures the degree of export concentration and long-run vulnerability more accurately than conventional export metrics, thereby offering a more precise framework for understanding the structural exposure of resource-dependent regions such as Jambi.

Against this backdrop, the study investigates the long-term relationship between coal export concentration, environmental quality, and economic growth in Jambi Province. Using a non-linear empirical approach with ARDL and BP structural break tests, this study aims to provide new insights into the interplay among export dynamics, ecological stress, and sustainable regional growth. In doing so, it contributes to theoretical debates on resource dependence and the EKC, while offering empirical evidence to inform Indonesia’s policy reforms and its broader post-coal transition.

3.1 Bai-Perron (BP) test

To account for potential regime shifts in the time-series data, the BP multiple structural break test was applied to the two independent variables, CECR and EQI. Structural breaks indicate statistically significant changes in the underlying data-generating process—typically caused by regulatory reforms, market volatility, or environmental shocks—and may affect model stability and the validity of long-run relationships in the ARDL framework.

The BP test identified two significant structural breaks in the CECR series, occurring in 2015 and 2022, with F-statistics of 12.18 and 13.85, respectively. Both values exceed the 5% critical thresholds (8.58 and 10.13), providing robust evidence of structural change. For EQI, one significant breakpoint was detected in 2018 (F = 68.65), indicating a substantial shift in environmental quality conditions, whereas the second-break test was not significant (F = 5.00 < 10.13) (refer to Table 2).

Table 2. BP structural break test results

These breakpoints closely correspond to major policy and market developments. The 2015 breakpoint in CECR aligns with Indonesia’s large-scale revocation of mining permits and the tightening of district-level coal export regulations. The 2022 breakpoint reflects a surge in global coal demand triggered by geopolitical tensions—particularly the Russia–Ukraine conflict—and the sharp rise in global coal prices. Meanwhile, the 2018 structural shift in EQI is consistent with the environmental impacts of large-scale forest fires and the introduction of stricter environmental controls that altered the province’s ecological conditions.

By identifying these breakpoints, the analysis incorporates key non-linear dynamics into the subsequent ARDL modelling. This step enhances the robustness and reliability of the estimation by ensuring that long-run relationships are evaluated within an empirically stable structural environment.

3.2 Unit root test

Before estimating the ARDL model, unit root tests were conducted to determine the integration order of each variable. The Augmented Dickey–Fuller (ADF) test was applied in EViews using a specification with an intercept and no deterministic trend. Lag length was selected automatically using the Schwarz Information Criterion (SIC), with a maximum of 1 lag, given the characteristics of the small annual dataset. If the null hypothesis of a unit root could not be rejected at the level form, the test was repeated at the first-difference form.

As a confirmatory step, the Phillips–Perron (PP) test was also employed because it is more robust to heteroskedasticity and residual autocorrelation. The PP test results were used to ensure consistency with the ADF findings.

Based on the ADF and PP test results presented in Table 3, all variables in the study—economic growth (EG), coal export concentration (CECR), and the EQI—become stationary after first differencing. Accordingly, the three variables are classified as I(1) processes, and none are integrated of order two. This condition is essential, as it ensures that the ARDL model can be validly applied and allows for the examination of potential long-run relationships through the bounds testing approach.

Table 3. Augmented Dickey–Fuller and PP unit root test results

3.3 Bound cointegration test

In estimating the ARDL model, the maximum lag length for both the dependent and independent variables was restricted to two lags to maintain model parsimony, given the relatively small annual sample size. The optimal lag combination was selected automatically using the Akaike Information Criterion (AIC), yielding ARDL(1, 0, 2) as the best-fitting specification. Based on this model, the Bounds Cointegration Test was conducted to assess the existence of long-run relationships among the variables.

Table 4. Bound cointegration test results

Table 4 presents the Bounds test results. The F-statistic of 13.580 is far above the upper bound critical values (I(1)) at the 10%, 5%, and 1% significance levels. Considering that the number of explanatory variables is two (k = 2) and the sample size is relatively small, the interpretation relies on finite-sample critical values (n ≈ 30), which similarly indicate that the F-statistic lies well above the upper bound for cointegration.

Based on these results, the null hypothesis of no long-run relationship can be rejected at all significance levels. Thus, there is strong evidence of cointegration among coal export concentration, environmental quality, and economic growth in Jambi Province.

Overall, the Bounds test findings indicate that the ARDL model used is sufficiently capable of capturing the long-run linkages among the variables. This supports the model’s validity for further analysis of short-run dynamics and long-run equilibrium relationships in the province.

3.4 Autoregressive distributed lag model estimation

Given the limited number of observations for the 2009–2023 period, the ARDL model was estimated using a maximum of two lags. Model selection was performed automatically using the AIC, which identified the ARDL (1, 0, 2) as the optimal specification. The complete estimation results are presented in Table 5.

Table 5 shows that the CECR variable has a positive and significant effect at p = 0.0010. This indicates that increases in coal export concentration provide a direct short-run stimulus to economic growth. Meanwhile, the second lag of EQI has a negative, significant coefficient (p = 0.0217), suggesting that declines in environmental quality exert a detrimental impact that materializes with a two-year lag. The negative coefficient on EG(-1) reflects an adjustment mechanism toward long-run equilibrium, although it is not statistically significant at conventional levels.

The existence of long-run relationships among the variables was examined using the Bounds Cointegration Test, presented earlier in Table 4. The F-statistic of 13.580 exceeds all upper-bound critical values (I(1)) at the 10%, 5%, and 1% significance levels, including the finite-sample category. This allows rejection of the null hypothesis of no cointegration, indicating that a stable long-run relationship binds economic growth, coal export concentration, and environmental quality. Accordingly, the use of the ARDL model to investigate both short-run dynamics and long-run equilibrium relationships is fully justified.

Table 5. ARDL estimation results

Long-run estimates obtained through the level equation are shown in Table 6. The results indicate that environmental quality, represented by the coefficient on EQI(-1), has a negative and significant long-run effect on economic growth (p = 0.0017). This implies that deteriorating environmental conditions are associated with slower long-run economic growth. Conversely, there is no empirical evidence supporting a long-run effect of CECR. In the EViews output, CECR appears in the long-run form as Z = Z(-1) + D(Z), so its coefficient cannot be interpreted as a long-run parameter. Therefore, environmental quality stands as the only significant long-run determinant of economic growth, while coal export dominance exerts influence only in the short run.

Table 6. Long-term ARDL estimation results

3.5 Error-correction model estimation

The short-run dynamics are presented in Table 7. The ECM estimation results indicate that none of the differenced variables—whether changes in CECR, changes in EQI, or their respective lags—are statistically significant at conventional levels. This condition is common in annual data with small sample sizes, where short-run variation is relatively limited and typically influenced more by random fluctuations than by substantial structural changes.

Table 7. Short-term ARDL-ECM estimation

Table 7 shows that the ECT(-1) coefficient is negative, as predicted by cointegration theory. Its sign indicates the presence of an adjustment mechanism toward the long-run equilibrium when deviations occur in the previous period. However, the p-value of 0.2126 shows that this adjustment process is not statistically significant. In other words, although the correction toward long-run equilibrium moves in the theoretically expected direction, its speed is not strong enough to be confirmed econometrically.

The two structural dummy variables—DUM_CECR15 and DUM_CECR22—are also not significant. This suggests that the regime shifts identified by the BP test in 2015 and 2022 did not generate measurable short-run effects on economic growth. The regulatory changes or coal market shocks in those years appear to have had a stronger influence on long-run dynamics, as indicated by the ARDL results, rather than on short-run fluctuations.

Overall, the ECM results indicate that the effects of coal export dependence and environmental quality on Jambi’s economic growth operate primarily through long-run mechanisms. Short-run effects are not statistically significant, suggesting that annual growth dynamics are shaped more by structural factors reflected in the long-run relationship than by year-to-year volatility.

3.6 Parameter stability testing

To ensure that the parameters in the estimated ARDL model remain stable throughout the observation period, two key diagnostic tests were conducted: the Cumulative Sum (CUSUM) and the Cumulative Sum of Squares (CUSUMSQ) tests. These tests are part of the Recursive Estimates procedure and are widely recommended in the econometric literature for detecting potential parameter instability [41]. The results are illustrated in Figure 2.

Figure 2. The plot of cumulative sum (CUSUM) and cumulative sum of squares (CUSUMSQ)

The first panel (CUSUM) displays the dynamics of the accumulated recursive residuals. The CUSUM line trends downward from 2019 and remains within the 5 percent significance bounds across the entire sample period. This pattern indicates the absence of major structural shocks or abrupt shifts in model parameters. In other words, the estimated ARDL coefficients are stable over time.

The second panel (CUSUMSQ) reflects the stability of the residual variance. Although an upward trend appears after 2020, the CUSUMSQ line stays within the 5 percent critical bounds, suggesting that the increase is not large enough to signal variance instability. Thus, the observed fluctuations in variance remain within an acceptable range.

Overall, both tests show that the CUSUM and CUSUMSQ lines remain within the 5 percent critical bounds throughout the estimation period. No patterns emerge suggesting parameter instability, and the ARDL model is confirmed to be stable with respect to both coefficient behavior and residual variance.

Given this evidence, the ARDL model used in the analysis is reliable for examining both the short-run dynamics and the long-run relationships among coal export dependence (CECR), environmental quality (EQI), and economic growth (EG) in Jambi Province. These findings are also consistent with the econometric literature, which emphasizes parameter stability as a fundamental requirement for valid time-series inference.