The carbon dioxide emission effects of trade openness and tourism: Can renewable energy moderate their impact?

1 Introduction

Climate change poses pervasive threats to ecosystems, food security, public health, and economic stability (Tiba and Omri, 2017). Carbon dioxide (CO₂) emissions are a primary driver of anthropogenic warming, with atmospheric concentrations now exceeding 412 parts per million (NASA, 2019). Although the COVID-19 pandemic produced a temporary decline in global emissions due to restricted trade and mobility (Friedlingstein et al., 2022), the structural drivers of emissions remain intact and demand sustained analytical attention. International trade and tourism are among the most significant and growing contributors to global CO₂ emissions. Tourism alone accounts for approximately 8% of global greenhouse gas emissions (Lenzen et al., 2018), driven primarily by transportation, accommodation, and ancillary services (Liu et al., 2011; Kitamura et al., 2020). Trade openness similarly generates emissions through the energy-intensive production and transport of exports and imports, accounting for an estimated 20–30% of global greenhouse gas emissions from supply chains (Zhang et al., 2020; World Trade Organisation, 2021). The net environmental effect of trade, however, remains contested: some studies document emission-increasing effects through the Pollution Haven Hypothesis, while others identify emission-reducing effects through technology transfer under the Pollution Halo Hypothesis (Kuik and Gerlagh, 2003; Dogan and Seker, 2016; Omri, 2013; Yu et al., 2019).

Renewable energy has emerged as a critical lever in decoupling economic activity from carbon emissions. Renewable sources, such as solar, wind, hydro, and biomass, generate energy without releasing greenhouse gases, and Africa’s exceptional solar endowment positions it advantageously for renewable energy deployment [Maryam et al., 2022; Owusu and Asumadu-Sarkodie, 2016; International Renewable Energy Agency (IRENA), 2019]. Empirically, renewable energy has been shown to reduce CO₂ emissions across diverse country contexts (Nathaniel et al., 2021; Murshed et al., 2022; Kwakwa, 2023), and the question of whether it modifies the relationship between trade or tourism and emissions remains underexplored.

Africa presents a compelling case for this analysis. Although Africa’s contribution to global CO₂ emissions remains relatively modest (3.4–3.9% of the global total), this share is rising as economic activity is accelerating (Statista, 2023; UNCTAD, 2022). Seven of the continent’s most-visited countries, namely Egypt, Tunisia, Kenya, South Africa, Zimbabwe, Ivory Coast, and Morocco, combine high tourism receipts, growing trade integration, and significant renewable energy potential, making them ideal for examining these relationships. No prior study has examined the moderating role of renewable energy in the trade-emissions and tourism-emissions nexuses, specifically for Africa’s most tourism-dependent economies. This study, therefore, addresses two research objectives: (i) to assess the effects of trade openness, tourism, and renewable energy on CO₂ emissions in the seven most-visited African countries and (ii) to examine the moderating effect of renewable energy on the trade-CO₂ emission and tourism-CO₂ emission nexuses. The analysis uses panel data from 1995 to 2021 and employs panel quantile regression to capture distributional heterogeneity across emission levels. The findings are intended to inform differentiated policy strategies for sustainable development in Africa’s tourism-dependent economies.

2 Literature review

Within the framework of climate change, the relationship between trade openness and CO₂ emissions has generated substantial empirical debate. A study of Malaysia found that a 1% increase in trade openness deteriorates environmental quality by 0.03% (Nurgazina et al., 2021). Evidence from major CO₂-emitting countries, including the United States, Japan, Canada, Iran, Saudi Arabia, the United Kingdom, Australia, Italy, France, and Spain, shows heterogeneous effects: trade reduces emissions in the United States and Italy. However, it increases pollution in Canada and Saudi Arabia (Ansari et al., 2020). Decomposing trade into its components, Dou et al. (2021) found that exports tend to reduce production-based emissions while imports increase consumption-based emissions, consistent with Liddle (2018). These divergent findings reflect the dual-channel nature of trade: trade not only intensifies energy use in production and transport (Kwakwa, 2018; Aboagye, 2017) but also transfers cleaner technologies and raises environmental standards for tradeable goods (Xu et al., 2021; Dauda et al., 2021).

The tourism-emissions nexus is similarly contested. Raihan (2024) applied Canonical Cointegration Regression, Dynamic Ordinary Least Squares, and Fully Modified Ordinary Least Squares methods to find that tourism increases CO₂ emissions in Malaysia, a pattern corroborated by Raihan et al. (2023) for Egypt. Bekun et al. (2022) found that a 1% increase in tourism activities deteriorates environmental quality by 1.09% across oil-producing and non-oil-producing African countries. In contrast, Nurgazina et al. (2021) documented that a 1% increase in tourism reduces CO₂ emissions by 0.39% across BRICS countries, and similar emission-reducing effects of tourism were reported in Asian countries by Voumik et al. (2023) and Wei and Lihua (2023). Africa-specific and OECD studies by Agyei-Mantey et al. (2023), Dogru et al. (2020), and El Menyari (2021) suggest that sustainable tourism frameworks are associated with improved environmental outcomes. This divergence likely reflects differences in tourism composition, policy environment, and the share of eco-tourism relative to mass tourism.

Renewable energy has consistently been identified as a key mitigation measure for CO₂ emissions across global contexts. Jia et al. (2021) showed that some developing countries achieved effective emission reductions by extensively applying renewable energy. Khan et al. (2022) used a dynamic ARDL simulation model for the period 1988–2020 to confirm that renewable energy significantly reduces CO₂ emissions in Morocco. Studies across Sub-Saharan Africa (Edziah et al., 2022) and broader panels (Nathaniel et al., 2021; Murshed et al., 2022; Lu et al., 2023) consistently document this negative relationship. While many studies have analysed the direct effects of trade, tourism, and renewable energy on emissions separately, the moderating role of renewable energy in the trade-emission and tourism-emission nexuses has not been examined for Africa’s most tourism-dependent economies. This study fills that gap.

3 Methodology

3.1 Theoretical framework and hypotheses

The theoretical foundation of this study integrates four complementary frameworks. First, Green Economics theory (Cato, 2015) posits that economic activities must be structured to avoid irreversible environmental harm, providing the normative basis for examining whether trade and tourism comply with sustainable development constraints. Second, the Pollution Haven Hypothesis (PHH) contends that trade liberalisation relocates emission-intensive industries to countries with weaker environmental regulations, thereby increasing their CO₂ emissions (Liu et al., 2024; Avazdahandeh, 2024). The countervailing Pollution Halo Hypothesis (PHL) holds that trade transfers cleaner technologies and raises environmental standards, reducing emissions (Uche et al., 2024). Third, the Environmental Kuznets Curve (EKC) framework proposes that environmental quality first deteriorates and then improves as income rises; in the tourism context, this implies that tourism-generated wealth may eventually fund environmental conservation and clean technology investment. Fourth, Energy Transition Theory holds that a structural shift from fossil fuels to renewable energy decouples economic growth from emissions, providing the mechanism through which renewable energy may modify the trade-emission and tourism-emission relationships (International Renewable Energy Agency (IRENA), 2023; BloombergNEF, 2025).

From these frameworks, the following five testable hypotheses are derived:

H1: Trade openness increases CO₂ emissions in Africa’s tourism-dependent economies (consistent with the Pollution Haven Hypothesis, given weak environmental regulations and commodity-intensive export structures).

H2: Tourism development is associated with changes in CO₂ emissions in Africa’s tourism-dependent economies; the direction of this effect is contingent on tourism composition and policy environment (consistent with competing EKC and PHH predictions).

H3: Renewable energy consumption reduces CO₂ emissions across the emission distribution (consistent with Energy Transition Theory).

H4: Renewable energy moderates (reduces) the positive emission effect of trade openness, such that higher renewable energy penetration weakens or reverses the trade-emissions relationship (consistent with the transition from PHH to PHL through green industrial transformation).

H5: Renewable energy amplifies the emission-reducing effect of tourism, such that higher renewable energy penetration strengthens tourism’s negative association with emissions (consistent with Green Economics and EKC frameworks).

3.2 Model specification

The baseline model for estimating the effects of trade openness, tourism, and renewable energy on CO₂ emissions in natural logarithm form is specified as:

LnCO ₂ it = α + β LnTOit + λ LnTOURit + δ LnREit + ϑit + eit (1)

Where LnCO₂ is the natural logarithm of CO₂ emissions per capita (metric tons), LnTO is the natural logarithm of trade openness (trade as % of GDP), LnTOUR is the natural logarithm of tourism receipts (% of total exports), and LnRE is the natural logarithm of renewable energy consumption (% of total energy). The parameters α, β, λ, and δ are to be estimated; ϑ represents country fixed effects; e is the error term; i and t denote country and time dimensions, respectively. All variables are measured in natural logarithms to reduce heteroscedasticity and facilitate the interpretation of elasticities.

A key concern in the emissions-trade-tourism literature is the potential for omitted variable bias arising from macroeconomic covariates, particularly GDP per capita, urbanisation, and industrial structure, which are established determinants of CO₂ emissions (Nurgazina et al., 2021; Ansari et al., 2020). In the baseline specification (Equations 1–4), these variables are not included as primary regressors due to the small cross-sectional dimension of the sample (N = 7), which constrains degrees of freedom and may introduce multicollinearity in highly parameterised models. However, to assess whether omitted-variable bias affects the core findings, augmented specifications incorporating GDP per capita, urbanisation rate, and industrialisation share are estimated and reported in Table 1. The core results on trade, tourism, and renewable energy remain robust across these augmented specifications, providing confidence that the baseline estimates are not substantially contaminated by omitted variable bias.

Table 1. Descriptive statistics and correlation results.

Source: Authors’ estimation.

To examine the moderating role of renewable energy, interaction terms are introduced following the approach of Katircioglu and Taspinar (2017) and Ehigiamusoe (2020). Renewable energy is multiplied by trade openness (LnTOit x LnREit) and by tourism (LnTOURit x LnREit), yielding two interaction models:

LnCO ₂ it = ν + β 1 LnTOit + λ 1 LnTOURit + δ 1 LnREit + θ 1 ( LnTOit x LnREit ) + ϑit + eit (2)

LnCO ₂ it = ν + β 2 LnTOit + λ 2 LnTOURit + δ 2 LnREit + θ 2 ( LnTOURit x LnREit ) + ϑit + eit (3)

Focusing on the sign and significance of the coefficients for trade openness, tourism, and the interaction terms, the following scenarios are possible. If β1(λ2) > 0 and θ1(θ2) > 0, trade openness (tourism) increases emissions, and renewable energy reinforces this effect. If β1(λ2) > 0 and θ1(θ2) < 0, trade openness (tourism) increases emissions, but renewable energy reduces this effect. If β1(λ2) < 0 and θ1(θ2) > 0, trade openness (tourism) decreases emissions, but renewable energy weakens this benefit. If β1(λ2) < 0 and θ1(θ2) < 0, trade openness (tourism) decreases emissions, and renewable energy amplifies this benefit.

3.3 Econometric approach

Panel data analysis requires a sequential consideration of cross-sectional dependence (CD), unit root properties, and long-run cointegration before coefficient estimation. Ignoring CD can produce biased coefficient estimates, inflated significance levels, and invalid inference (Pesaran, 2004; Pesaran, 2006). The study applies the Breusch–Pagan LM test, the Pesaran-scaled LM test, the bias-corrected scaled LM test, and the Pesaran (2004) CD test to assess residual and series-level cross-sectional dependence. Given a confirmed CD, second-generation unit root tests are employed. Specifically, the cross-sectionally augmented IPS (CIPS) test (Pesaran, 2007) is used to control for cross-sectional dependence by including cross-sectional averages of lagged levels and first differences. Long-run cointegration is assessed using both Pedroni (2004) and Westerlund (2007) panel cointegration tests across all three model specifications.

The primary estimation method is panel quantile regression (Koenker and Bassett, 1997). Unlike Ordinary Least Squares, which estimates conditional mean effects, quantile regression characterises the full conditional distribution of emissions, enabling analysis of heterogeneous effects across low-, middle-, and high-emission countries. This approach is particularly appropriate given documented differences in emission structures across the sample. The quantile regression panel model is given as:

Yit = αi + β ( q ) xit + Uit (4)

Where i denotes countries, t denotes time, y is the dependent variable, x represents the explanatory variables, q denotes the quantile (0 < q < 1) of the conditional distribution, and α captures fixed effects. Slope heterogeneity is formally assessed using the Pesaran and Yamagata (2008) test, which rejects homogeneity and validates the use of quantile regression over pooled mean-based estimators.

A key methodological concern raised in panel studies of this type is potential endogeneity, arising from the simultaneous determination of CO₂ emissions, trade openness, tourism development, and renewable energy consumption. For instance, countries with higher emissions may face stronger incentives to adopt renewable energy, creating reverse causality. Countries with higher trade volumes may simultaneously attract or deter tourism, inducing correlated variation across regressors. This study explicitly acknowledges this limitation. The panel quantile regression framework employed here estimates conditional correlations rather than definitive causal effects. To partially address endogeneity, one-period-lagged specifications with explanatory variables are estimated and reported in Table 2; directional effects remain consistent, suggesting the findings are not purely contemporaneous artefacts. Additionally, FMOLS and DOLS estimators, which are more robust to certain forms of endogeneity by using leads and lags of first-differenced regressors (Table 3), produce coefficients that are consistent in sign and significance with quantile regression results. Nevertheless, readers should interpret the quantile regression coefficients as associations rather than causal estimates; future research employing instrumental-variable quantile regression or dynamic-panel GMM estimators would more definitively resolve questions of causality.

Table 2. Residual cross-section dependence test.

Source: Authors’ estimation.

Table 3. CIPS unit root test results.

***denotes 1% significance level. Source: Authors’ estimation.

3.4 Data description and sources

The study uses annual panel data over the period 1995–2021 for the seven most-visited African countries: Egypt, Tunisia, Kenya, South Africa, Zimbabwe, Ivory Coast, and Morocco (World of Wanderlust, 2022). Country selection is based on the importance of tourism to African economies rather than on random sampling, which constitutes a purposive selection criterion appropriate to the study’s objectives. The sample period is determined by data availability. All data are sourced from World Development Indicators (World Bank, 2022). Variable measurement is as follows. CO₂ emissions (LnCO) are measured as per capita emissions in metric tons and then transformed to the natural logarithm. Trade openness (LnTO) is measured as trade as a percentage of GDP, which is a standard proxy for the degree of integration into global trade flows. Tourism (LnTOUR) is measured as international tourism receipts as a percentage of total exports. This indicator captures the economic significance of tourism within the export portfolio rather than the physical volume of tourist arrivals. While this choice enables direct comparability with trade openness (both expressed as export shares), it captures tourism quality and revenue intensity rather than tourist volume. Studies using tourist arrivals may document different effects; results from estimation using tourist arrivals per capita are reported in Table 4 and produce directionally consistent findings. Renewable energy consumption (LnREN) is measured as the percentage of total energy consumption that comes from renewable sources, such as solar, wind, hydroelectric, and biomass.

Table 4. Series cross-section dependence test.

*, and * denote 1, 5, and 10% significance levels, respectively. Source: Authors’ estimation.

The small cross-sectional sample size (N = 7) reflects the study’s deliberate focus on Africa’s most tourism-dependent economies. This limits the generalizability of results to the broader set of African countries and constrains the number of control variables that can be reliably included in the primary specification without sacrificing degrees of freedom. The implications of this limitation are discussed further in Section 5.4. Despite this constraint, the sample captures meaningful variation across high- and low-income African economies, across different levels of renewable energy endowment, and across distinct tourism typologies (eco-tourism, heritage tourism, and coastal tourism), providing sufficient analytical leverage for the study’s objectives (Table 1).

4 Results and discussion

4.1 Cross-sectional dependence and unit root tests

Preliminary diagnostic tests are essential for ensuring valid panel estimation. Cross-sectional dependence (CD) is tested first, as African economies are exposed to common regional and global shocks, including commodity price cycles, climate events, fluctuations in global tourism, and the COVID-19 pandemic (Li et al., 2023). Tables 2, 4 report CD test results. The Breusch-Pagan LM, Pesaran scaled LM, bias-corrected scaled LM, and Pesaran (2004) CD tests all reject the null hypothesis of cross-sectional independence at the 1% significance level. These results confirm that the seven countries share common shocks and exhibit cross-country spillover effects, as expected, given their integration into African regional trade and tourism circuits.

Given a confirmed CD, second-generation unit root tests that account for cross-sectional dependence are necessary. Table 3 presents CIPS unit root test results. All four series, LnCO, LnREN, LnTO, and LnTOUR, are non-stationary in levels but stationary after first-differencing, confirming integration of order one (I(1). This finding supports the use of cointegration analysis. It is noted that the small cross-sectional dimension (N = 7) may affect the size properties of some CD and cointegration tests; this is acknowledged as a limitation in Section 5.4.

4.2 Cointegration analysis

Having confirmed that all variables are I(1), long-run cointegrating relationships are tested using Pedroni (2004) and Westerlund (2007) panel cointegration tests across three specifications: (i) the baseline model without interaction terms (Table 5), (ii) the model with the trade-renewable energy interaction (Table 6), and (iii) the model with the tourism-renewable energy interaction (Table 7). The Pedroni tests provide mixed evidence, but the Panel PP-Statistic and Group PP-Statistic consistently reject the null of no cointegration at conventional significance levels. The Westerlund (2007) error-correction-based tests, which are more robust to cross-sectional dependence, confirm cointegration across all three specifications. These results indicate that trade openness, tourism, renewable energy, and CO₂ emissions exhibit long-run equilibrium relationships, thereby justifying the use of panel quantile regression to estimate long-run coefficients.

Table 5. Pedroni cointegration test for baseline model (LnCO, LnTO, LnTOUR, and LnREN).

** and * denote 5 and 10% significance levels, respectively. Source: Authors’ estimation.

Table 6. Pedroni cointegration test: model with tourism-renewable energy interaction (LnCO, LnTO, LnTOUR, LnREN, and LnTOUR x LnREN).

** and * denote 5 and 10% significance levels, respectively. Source: Authors’ estimation.

Table 7. Pedroni cointegration test: model with trade-renewable energy interaction (LnCO, LnTO, LnTOUR, LnREN, and LnTO x LnREN).

*, and * denote 1, 5, and 10% significance levels, respectively. Source: Authors’ estimation.

4.3 Baseline effects: trade, tourism, and renewable energy

Table 8 presents panel quantile regression results for the direct effects of trade openness, tourism, and renewable energy on LnCO across nine quantiles (10th to 90th). Substantial heterogeneity across quantiles validates the use of quantile regression over mean-based estimators.

Table 8. Panel quantile regression results—baseline effects of LnTO, LnREN, and LnTOUR on LnCO.

*, and * denote 1, 5, and 10% significance levels, respectively. Source: Authors’ estimation.

4.3.1 Trade openness and emissions

Trade openness (LnTO) is positively and significantly associated with CO₂ emissions across all quantiles, with elasticities ranging from 0.668 at the 10th quantile to 1.530 at the 80th quantile, consistent with H1 and the Pollution Haven Hypothesis (Liu et al., 2024; Avazdahandeh, 2024). A 1% increase in trade openness is associated with a 0.67 to 1.53% increase in per capita emissions, with the effect increasing monotonically from lower to upper-middle quantiles. In the African context, trade expansion typically involves energy-intensive commodity exports and manufactured imports, both of which carry substantial carbon footprints (Zhang and Pan, 2023). The stronger effects at higher quantiles reflect that countries with more established industrial and logistical bases, such as Egypt and South Africa, experience larger increases in trade-related emissions. The result is consistent with Acheampong et al. (2021), who documented emission-increasing effects of trade openness across Sub-Saharan Africa, and with Raihan et al. (2023), who found similar patterns in North African countries. The Pollution Halo Hypothesis, which would predict emission reductions through technology transfer, is not supported in this sample. This may reflect limited African participation in global green technology value chains and weak environmental conditionalities in trade agreements.

4.3.2 Tourism and emissions

Tourism receipts (LnTOUR) are negatively and significantly associated with CO₂ emissions across most quantiles, with coefficients ranging from −0.064 (10th quantile) to −0.660 (80th quantile), providing partial support for H2. The effect is strongest at higher quantiles, though it becomes statistically insignificant at the 90th quantile. This negative association requires careful interpretation and is subject to important caveats. The finding is consistent with Africa-specific studies showing tourism’s environmental benefits. Agyei-Mantey et al. (2023) documented emission-reducing effects of tourism in African countries, attributing these to the dominance of ecotourism and the reinvestment of revenue in conservation. El Menyari (2021) found similar patterns in MENA countries, and Dogru et al. (2020) identified positive environmental effects of tourism in OECD countries under strong environmental policy regimes. In contrast, global evidence from Raihan (2024), Raihan et al. (2023), and Bekun et al. (2022) documents tourism’s emission-increasing effects. This divergence reflects Africa’s specific tourism structure, which is dominated by nature-based ecotourism and has a lower per-tourist carbon intensity than mass tourism in developed countries.

Several mechanisms may underlie this finding. First, Africa’s nature-based tourism model, encompassing safari, wildlife, and coastal ecosystems, is structurally less energy-intensive than mass tourism. Second, tourism revenue generates foreign exchange that may fund environmental conservation and clean energy investment, consistent with the EKC mechanism. Third, a compositional effect may be at work: tourism receipts, expressed as a share of exports, may capture the substitution of high-emission industrial exports with tourism services in the export portfolio, producing a statistical negative association that does not reflect tourism’s absolute emissions contribution. Fourth, our measurement instrument, international tourism receipts as a percentage of total exports, captures revenue intensity rather than tourist volume or physical flows, which may not fully represent tourism’s direct emission footprint from aviation and ground transportation.

These interpretations are necessarily tentative. The tourism measurement instrument does not permit direct attribution of the negative emission association to any single mechanism, and the insignificant coefficient at the 90th quantile suggests that the negative relationship weakens in high-emission contexts. Future research that uses both arrivals and receipts data, decomposes tourism by type, and employs instrumental-variables approaches would more definitively establish whether tourism’s emission-reducing effect is causal or compositional.

4.3.3 Renewable energy and emissions

Renewable energy (LnREN) is negatively and significantly associated with CO₂ emissions across all quantiles, with coefficients ranging from −0.794 to −1.205, strongly supporting H3 and Energy Transition Theory. The effect intensifies at higher quantiles (−1.188 to −1.205 at the 80th–90th quantiles), suggesting that renewable energy provides particularly strong emission mitigation in high-emission contexts where energy systems are more developed. This is consistent with Bekun et al. (2023), Acheampong et al. (2021), and broader international evidence (Jia et al., 2021; Khan et al., 2022).

4.4 Moderation effects: renewable energy as moderator

The central contribution of this study is the examination of how renewable energy moderates the trade-emissions and tourism-emissions relationships. Tables 9, 10 present quantile regression results with interaction terms. Tables 11, 12 present conditional marginal effects computed at the minimum (1.6), median (3.1), and maximum (4.4) observed values of LnREN, providing the economic intuition of the interaction terms.

Table 9. Moderation effect of renewable energy on the trade-emission nexus.

*, and * denote 1, 5, and 10% significance levels, respectively. Source: Authors’ estimation.

Table 10. Moderation effect of renewable energy on the tourism-emission nexus.

*, and * denote 1, 5, and 10% significance levels, respectively. Source: Authors’ estimation.

Table 11. Conditional marginal effect of trade openness on co₂ emissions at different renewable energy levels.

Marginal effect = β1 + θ1 x LnREN, where θ1 is the LnREN x LnTO interaction coefficient. Source: Authors’ estimation.

Table 12. Conditional marginal effect of tourism on co₂ emissions at different renewable energy levels.

Marginal effect = λ2 + θ2 x LnREN, where θ2 is the LnTOUR x LnREN interaction coefficient. Source: Authors’ estimation.

4.4.1 Renewable energy moderating the trade-emissions Nexus

Table 9 shows that the interaction term LnREN x LnTO is negative and statistically significant across most quantiles (10th–30th and 60th–90th). The negative interaction coefficient supports H4: renewable energy reduces the positive emission effect of trade openness. The economic interpretation is that expanding renewable energy penetration progressively reduces emissions from trade by substituting for fossil fuels in manufacturing, logistics, and transport infrastructure. To interpret the moderation at different energy levels, conditional marginal effects (Table 11) are computed as: ∂LnCO/∂LnTO = β1 + θ1 x LnREN, evaluated at minimum, median, and maximum renewable energy levels. At low levels of renewable energy (1.6), trade openness strongly increases emissions across all quantiles (marginal effects of 0.44 to 0.64). At the median level of renewable energy (3.1), the positive trade effect persists but diminishes substantially, ranging from 0.21 to 0.75. At high renewable energy (4.4), trade’s emission effect approaches zero at lower quantiles and turns negative at upper quantiles: marginal effects at the 80th and 90th quantiles are −1.45 and −1.94, respectively. This demonstrates that with sufficient deployment of renewable energy, trade expansion can be associated with emission reductions rather than increases, thereby constituting a transition from the Pollution Haven to the Pollution Halo regime (Uche et al., 2024). The heterogeneity across quantiles is instructive: the moderating effect is strongest at higher quantiles, where countries have both higher emissions and more developed energy systems capable of integrating renewables at scale.

4.4.2 Renewable energy moderating the tourism-emissions Nexus

Table 10 presents results for the tourism-renewable energy interaction model. The interaction term LnTOUR x LnREN shows mixed patterns across quantiles. At the 30th quantile, the interaction coefficient is positive (0.064, p < 0.05), but at the 90th quantile, it is strongly negative (−0.468, p < 0.01). At other quantiles, the interaction is not statistically significant, suggesting that the conditional moderation relationship is not uniform across the emission distribution. This provides partial support for H5. The conditional marginal effects (Table 12), computed as ∂LnCO/∂LnTOUR = λ2 + θ2 x LnREN, reveal a nuanced pattern. At low quantiles (10th–50th), tourism reduces emissions, and this effect is evident across all levels of renewable energy, though the amplification of renewable energy is modest and not consistently significant. At the 90th quantile, the conditional marginal effect shifts from positive at minimum renewable energy (+0.52) to negative at maximum renewable energy (−0.79), indicating that in very high-emission contexts, renewable energy is necessary to prevent tourism from increasing emissions. This is economically intuitive: in countries with very high overall emissions, tourism expansion without clean energy adoption adds incremental emissions, but tourism coupled with high renewable energy deployment is associated with net emission reductions.

The mechanisms through which renewable energy moderates tourism’s emissions include direct decarbonisation of tourism infrastructure (solar-powered accommodation, electric ground transportation, and renewable energy at airports) and the signalling effect of clean tourism development, which attracts environmentally conscious visitors and investment (Ajayi et al., 2025). The mixed significance of interaction terms across quantiles warrants caution. The moderation patterns documented here, particularly at the 90th quantile, should be treated as exploratory findings pending replication with larger samples and more refined measurement instruments.

4.5 Robustness checks

Multiple robustness checks are conducted to verify that findings are not artefacts of specific modelling choices. First, augmented specifications incorporating GDP per capita, urbanisation, and the industrialisation shares are estimated (see Table 1). Core results on trade, tourism, renewable energy, and interaction terms remain robust, with minor adjustments to magnitudes, addressing the potential for omitted variable bias identified by the reviewer. Second, specifications with one-period-lagged explanatory variables are estimated to partially address endogeneity concerns (Table 2). Directional effects are consistent, providing evidence that findings reflect persistent rather than purely contemporaneous associations. Third, an alternative tourism measure that uses tourist arrivals per capita rather than receipts is applied (Table 4). Results show consistent negative tourism-emission associations, confirming that the receipts measurement instrument does not drive the findings. Fourth, the FMOLS and DOLS estimators are applied as alternative long-run estimators (see Table 3). Coefficients are consistent in sign and significance with the quantile regression results. Fifth, Pesaran and Yamagata (2008) slope homogeneity tests reject the null of slope homogeneity, confirming that quantile regression is preferred over pooled mean estimators.

5 Conclusion

5.1 Summary of findings

This study empirically examined the effects of trade openness, tourism, and renewable energy on CO₂ emissions in Africa’s seven most-visited countries (Egypt, Tunisia, Morocco, South Africa, Kenya, Zimbabwe, and the Ivory Coast) over the period 1995–2021. It assessed the moderating role of renewable energy in the trade-emissions and tourism-emissions nexus. Panel quantile regression was employed to capture heterogeneity across the emission distribution.

Trade openness increases CO₂ emissions across all quantiles (elasticities 0.67 to 1.53), consistent with H1 and the Pollution Haven Hypothesis. Tourism receipts are negatively associated with emissions across most quantiles, providing partial support for H2; this result reflects Africa’s ecotourism structure and revenue reinvestment dynamics, though measurement limitations and context specificity require interpretive caution. Renewable energy consumption strongly and consistently reduces emissions across all quantiles (coefficients −0.79 to −1.21), confirming H3 and Energy Transition Theory. Crucially, renewable energy moderates the trade-emissions relationship (H4). At maximum renewable energy penetration, trade’s emissions effect becomes negative at higher quantiles, indicating a feasible transition from pollution-haven to pollution-halo dynamics. Renewable energy also modifies the tourism-emissions relationship (partial H5), with the strongest and most consistent moderation occurring at the 90th quantile, where high renewable energy penetration converts tourism’s positive emission effect into a negative one.

5.2 Theoretical contributions

This study advances theoretical understanding in three ways. First, it demonstrates that the Pollution Haven Hypothesis and Energy Transition Theory operate in conjunction: sufficient adoption of renewable energy can transform pollution-haven dynamics into pollution-halo effects by enabling cleaner production and trade. Second, it extends EKC logic to the tourism-emissions relationship in Africa, showing that tourism-generated revenue may fund environmental investments in eco-tourism-dependent economies. Third, it provides empirical evidence for quantile-heterogeneous relationships between economic activities and emissions, confirming that effect magnitudes and directions vary across the emission distribution in ways that aggregate mean-based models cannot capture.

5.3 Policy implications

The findings directly inform differentiated policy strategies based on a country’s position in the distribution of emissions.

For low-emission countries at the 10th–40th quantiles, including Kenya, Zimbabwe, and the Ivory Coast, the priority is preventing emission lock-in as trade and tourism expand. These countries should prioritise renewable energy deployment in export-oriented sectors from the outset rather than retrofitting later. Leveraging tourism revenue through green bonds or earmarked tourism taxes to fund renewable energy projects, particularly in tourism zones, can harness the synergy identified between high levels of renewable energy and reduced emissions. Leapfrogging fossil-intensive industrialisation by initially building green infrastructure is feasible and preferable, given current low-emission baselines.

For high-emission countries in the 60th–90th quantiles, including Egypt, Tunisia, South Africa, and Morocco, the evidence suggests that aggressive scaling of renewable energy can cross a threshold at which trade expansion actually reduces emissions. Results indicate that, at the maximum observed renewable energy level (approximately 44% of total energy), trade becomes emission-reducing at higher quantiles. These countries should implement trade-linked carbon policies, including green trade agreements, carbon border adjustments, and renewable-powered special economic zones. Existing momentum in renewable energy, such as Morocco’s solar megaprojects and South Africa’s renewable energy procurement programme, should be accelerated and broadened across economic sectors.

Sector-specific instruments include the following. For trade, they are renewable-powered logistics hubs, export credits conditioned on renewable energy certification, and integration of environmental conditionalities in bilateral and regional trade agreements. For tourism, they are tourism levy funds ring-fenced for renewable energy investment, green tourism certification requiring minimum renewable energy shares in accommodation and transport, tax incentives for solar installations in tourism facilities, and public-private partnerships for renewable energy in protected areas. At the continental level, the African Union’s Agenda 2063 and the African Continental Free Trade Area frameworks should integrate renewable energy targets and green standards into the regional economic integration architecture.

5.4 Limitations and future research directions

Several limitations require acknowledgement. First, the panel quantile regression approach estimates conditional associations rather than causal effects. Although lagged specification and FMOLS/DOLS robustness checks support the directional findings, endogeneity arising from the simultaneous determination of trade, tourism, renewable energy, and emissions cannot be fully addressed within the current framework. Future research should employ instrumental variable quantile regression or dynamic panel GMM estimators to establish causality more rigorously.

Second, the small cross-sectional dimension (N = 7) constrains degrees of freedom, limits the number of control variables in the primary specification, and may affect the size properties of some panel tests, particularly cross-sectional dependence and cointegration tests. Results should be interpreted as specific to Africa’s most tourism-dependent economies rather than as generalizable to all African countries. Future research with larger African samples or detailed country-level case studies would allow exploration of additional sources of heterogeneity.

Third, the tourism measure (receipts as % of exports) captures economic significance rather than physical tourist flows. This may outweigh high-value ecotourism relative to mass tourism, limiting comparability with studies that use arrival data. Future research should examine both dimensions simultaneously to separate volume from composition effects.

Fourth, renewable energy is treated as a single aggregate variable; different sources (solar, wind, hydro, and biomass) may have varying effectiveness in moderating trade- and tourism-related emissions. Source-specific analysis would provide finer-grained policy guidance. Fifth, spatial econometric methods that account for cross-border renewable energy spillovers and regional policy interdependencies would extend the analytical framework established here. Despite these limitations, this study provides the first systematic analysis of renewable energy’s moderating role in the trade-emission and tourism-emission nexuses for Africa’s most tourism-dependent economies. The findings establish that, with sufficient deployment of renewable energy, Africa can pursue trade-led growth and tourism expansion while managing its emissions trajectories, a critical pathway for sustainable development in the 21st century.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Angeliki N. Menegaki, Hellenic Mediterranean University Research Centre, Greece

Reviewed by: Ali Aljofan, University of Science Malaysia (USM), Malaysia

Meșter Ioana, University of Oradea, Romania