Reinvestigating the environmental Kuznets curve (EKC) of carbon emissions and ecological footprint in 147 countries: a matter of trade protectionism

The study addresses how trade protectionism (proxied inversely by trade openness) affects the relationship between economic growth and environmental degradation within the Environmental Kuznets Curve (EKC) framework. Motivated by rising concerns over climate change, global trade disruptions (financial crisis, COVID-19, geopolitical conflicts), and debates over the environmental consequences of trade, the paper seeks to: (1) test whether the EKC hypothesis holds in a large global sample, (2) assess the role of trade protection measures in the growth–environment nexus and identify potential nonlinear (threshold) effects, and (3) evaluate heterogeneity of these effects across income groups. The context underscores competing theories (pollution haven vs. pollution halo) and the need for a comprehensive, cross-country, multi-indicator assessment (CO2 emissions and ecological footprint) using methods capable of capturing nonlinearity. The study aims to inform policy on balancing economic development and environmental sustainability under differing income levels and trade regimes.

The literature on EKC finds mixed evidence: while many studies confirm an inverted U-shaped relationship between income and various environmental indicators, others document N-shaped or more complex dynamics. Two strands are highlighted: country/region-specific versus multi-country analyses, revealing that results vary by context and methodology. The paper also reviews nonlinear modeling in energy/environment economics, emphasizing panel threshold regression as a tool to capture regime-dependent relationships. Competing views on trade’s environmental impact are summarized via Grossman and Krueger’s scale, composition, and technique effects, alongside the pollution haven and pollution halo hypotheses. Empirical studies show both adverse and beneficial environmental impacts of trade, FDI, and globalization, implying that net effects are context- and regime-dependent. The authors position their contribution as re-examining the EKC through the lens of trade protection using a nonlinear threshold panel model, employing two environmental indicators (CO2 and ecological footprint), and comparing effects across four World Bank income groups.

Data: Panel of 147 countries, 1995–2018. Environmental indicators: per capita CO2 emissions (CO2) and per capita ecological footprint (EF). Core explanatory variable: GDP per capita (PGDP, constant 2015 USD). Threshold variable: trade (TRA), defined as merchandise trade (exports + imports) as a share of GDP. Controls: industrial structure (IS, industry value added share of GDP), foreign investment (FI, net FDI inflows share of GDP), globalization index (GI, KOF index capturing economic, social, and political dimensions). Data sources: World Bank, Global Footprint Network, KOF Swiss Economic Institute. Variables transformed to natural logs where appropriate: LnCO2, LnEF, LnPGDP, LnIS, LnGI, LnTRA; FI kept in levels. Models: (1) Baseline EKC-type panel models with LnCO2 or LnEF as dependent variables, including LnPGDP and (implicitly) its squared term via nonlinear threshold specification; (2) Hansen (1999) panel threshold regression using LnTRA as the threshold variable to allow the effect of LnPGDP on environmental outcomes to vary by trade regimes. Both single and double-threshold models estimated. Preliminary tests: panel unit root tests (LLC, IPS) indicate I(1) processes; Pedroni residual cointegration tests confirm long-run relationships among variables for both CO2 and EF specifications across global sample and income groups. Threshold significance assessed via bootstrap F-tests; threshold confidence intervals via likelihood ratio statistics. Estimation conducted with fixed effects and OLS within each regime, reporting regime-specific elasticities of environmental indicators with respect to income, conditional on trade openness regimes.

- Descriptive statistics (global): average per capita CO2 emissions 1.8225 metric tons; average per capita ecological footprint 2.5226 gha; average per capita GDP about $4,619.31 (2015 USD). High-income countries record the largest per capita CO2 and EF; low-income the smallest; high-income also have higher trade shares on average. - Stationarity and cointegration: LLC and IPS tests indicate variables are first-difference stationary; Pedroni tests largely confirm cointegration between environmental indicators and covariates in global and income-group panels. - Global threshold tests: Significant double-threshold effects when LnCO2 is dependent (10% level); thresholds at LnTRA λ1=4.188 (≈65.9% of GDP) and λ2=5.019 (≈150.6%). For LnEF, double threshold significant at 10%; thresholds at λ1=2.979 (≈19.7%), λ2=5.540 (≈255.6%). - Global threshold regressions: For CO2, the elasticity of CO2 w.r.t. income declines as trade openness moves to higher regimes: 0.3845 (q≤4.188), 0.3766 (4.188<q≤5.019), 0.3613 (q>5.019). For EF, coefficients follow a U-shape: 0.2818 (q≤2.979), 0.2727 (2.979<q≤5.540), 0.2915 (q>5.540), with the second threshold regime sparsely populated. Industrial structure raises both CO2 and EF; FI is globally insignificant; GI increases CO2 and EF. - High-income group: CO2 model shows double threshold (λ1=4.075≈59.2%, λ2=4.925≈137.7%), with income elasticities decreasing across regimes: 0.0609 → 0.0545 → 0.0362. EF shows a single threshold (λ≈4.931≈138.9%), with coefficients 0.2754 → 0.2645 as trade rises. Trade openness mitigates the growth–emissions link in high-income economies. - Upper-middle-income group: CO2 has thresholds at λ1=3.675 (≈39.4%), λ2=4.022 (≈55.9%); coefficients 0.4941 → 0.5109 → 0.5041 (inverted-U across trade regimes). EF shows a single threshold at λ=3.706 (≈40.7%); coefficients 0.3252 → 0.3315, indicating trade openness slightly amplifies EF pressures. - Lower-middle-income group: CO2 thresholds λ1=3.267 (≈26.2%), λ2=4.714 (≈111.0%); coefficients increase with trade openness: 0.7340 → 0.7502 → 0.7766. EF thresholds λ1=3.496 (≈33.1%), λ2=4.957 (≈142.0%); coefficients 0.2579 → 0.2687 → 0.2979, all significant at 1%. Trade openness strengthens the growth–environment pressure link, consistent with pollution haven dynamics. - Low-income group: CO2 coefficients across regimes approximately 0.6716 → 0.6459 → 0.6619; EF approximately 0.1732 → 0.1898 → 0.1794. Thresholds reported at λ1=3.142 (≈23.1%), λ2=3.811 (≈45.1%) for CO2; λ1=3.562 (≈35.3%), λ2=3.595 (≈36.4%) for EF. Effects generally indicate that increased trade openness does not yield the same mitigation as in high-income economies. - EKC validation: Across income strata, the impact of economic growth on environmental pressures tends to rise from low- to middle-income and then fall in high-income groups, consistent with the EKC. - Mechanisms: Industrial structure intensifies environmental pressure; globalization tends to raise CO2 and EF (global panel), while FI effects vary by group.

The findings address the research questions by confirming an EKC-type relationship and demonstrating that trade openness conditions the growth–environment nexus through nonlinear threshold effects. Globally, higher trade openness attenuates the elasticity of CO2 with respect to income, suggesting that openness fosters technological diffusion, stricter standards, and efficiency gains that can mitigate emissions intensity. However, for ecological footprint, the relationship is more nuanced, showing regime-dependent U-shaped changes with sparse data in the highest-trade regime. Heterogeneity across income groups is central: high-income countries benefit environmentally from greater openness, while upper- and lower-middle-income (and in some cases low-income) countries experience stronger positive links between income and environmental pressures as openness rises. These patterns support pollution haven dynamics, where stringent regulations in advanced economies and global value chains reallocate pollution-intensive activities toward developing economies. The results emphasize that the environmental implications of trade are not uniform; they depend on countries’ income levels, industrial composition, regulatory capacities, and positions in global supply chains. Tailored policies, including environmental standards embedded in trade agreements, technology transfer, and structural transformation toward cleaner sectors, are needed to align trade and growth with environmental goals across development stages.

This study contributes by re-examining the EKC through a nonlinear threshold framework that treats trade openness as the conditioning variable, applied to two environmental indicators (CO2 and ecological footprint) for 147 countries and four income groups (1995–2018). The main contributions are: (i) empirical support for the EKC across income groups, with the growth–environment elasticity rising from low to middle income and falling for high income; (ii) identification of significant trade-related thresholds that alter the growth–environment elasticity, with openness generally mitigating CO2 pressures globally and in high-income countries but amplifying pressures in many developing economies; and (iii) evidence of heterogeneous mechanisms, highlighting roles for industrial structure, globalization, and investment. Policy-wise, the results argue for environmentally friendly trade frameworks, stronger international cooperation, technology transfer, and stricter environmental governance, especially in developing economies to avoid pollution haven outcomes. Future research should expand datasets temporally and cross-nationally, explore alternative specifications and variables, and employ advanced econometric techniques to test additional nonlinearities and mechanisms.

The analysis is limited by data availability to 147 countries over 1995–2018, potentially constraining generalizability and the precision of threshold estimates, particularly in sparsely populated high-trade regimes. The methodology, while robust (panel cointegration and Hansen threshold models), could be complemented by alternative identification strategies and more advanced econometric techniques to address endogeneity and dynamic effects. Variable selection focuses on a core set (PGDP, trade, IS, FI, GI); additional factors (energy mix, policy indices, innovation measures, institutional quality) could refine mechanisms. Future work should extend the time span and country coverage, enrich covariates, test alternative functional forms and thresholds, and probe the substantive policy implications and causal channels in greater depth.