The Effect of Trade Openness on Environmental Quality in Southern African Customs Union (SACU) Countries: The CS-ARDL Approach

Abstract

The Southern African Customs Union (SACU), as a bloc, is compelled to commit to trade in environmentally friendly goods. This study investigated the short-run and long-run relationships between trade openness and environmental quality in the SACU. The Cross-Sectional Autoregressive Distributed Lag (CS-ARDL) approach was applied to the data from 1985 to 2023. The results show that the estimated coefficients of trade openness positively and significantly contribute to carbon emissions in the short run and the long run. The results demonstrate that the gains-from-trade hypothesis does not hold in the SACU. Also, the results indicate that foreign direct investment inflow does not significantly contribute to CO₂ emissions; therefore, the pollution haven hypothesis does not hold. The Dumitrescu–Hurlin Granger non-causality test was employed, and the results show that there is bidirectional causality between CO₂ emissions and trade openness, CO₂ emissions and economic growth, and CO₂ emissions and population growth and no directional causality between foreign direct investment and CO₂ emissions. This study recommends that SACU countries should encourage the trade of eco-friendly goods, which is likely to lessen environmental consequences.

1. Introduction

Climate change, particularly the accelerating greenhouse effect, is a global environmental concern that is a result of both human-related and natural causes. Therefore, the need to protect the environment continues to be a world priority (J. Wang et al., 2024; World Bank, 2025). As such, the effect of trade openness on the environment is a significant issue in international trade policies and agreements among trading blocs. The empirical and theoretical literature on the trade–environment nexus is riddled with mixed results and suggestions.

J. Sun (2024) noted that the interconnectedness and interdependence of global economies continue to increase, and international trade has increased to a new level. There is a continued shift in the production of goods to China, which are then exported to emerging and developing African economies. Notably, J. Sun (2024) further suggested that a large number of Chinese industrial firms also move their operations to African countries, among which are Southern African Customs Union (SACU) members.

Founded in 1910, the SACU is one of the oldest customs unions in the world and has evolved from a colonial-era setup to a more democratic entity (Gibb, 2006). The customs union was at first controlled by South Africa, but it experienced significant changes with new accords in 1969 and 2002 that addressed imbalances in development and decision-making among member nations. The 2002 agreement established shared governance frameworks, including a Council of Ministers and a permanent Secretariat, as a shift from South Africa’s sole decision-making authority (Ngalawa, 2013). The union’s basic goal is to make trade free and easy for its members and enhance trade openness. In its strategic plan for 2022–2027, the SACU admits that it is concerned with environmental quality and emphasises moving away from environmentally damaging economic activities. The plan also suggests that significant advances should occur in new industrial production methods, commodity demand, trade, and investment movements.

Also, the SACU countries face a range of environmental pollution concerns, such as air and water pollution, waste management issues, and threats to wildlife and the extinction of species. In 2023, the union’s economy grew by 5.4 percent in comparison to 6.4 percent in the year 2022 (Southern African Customs Union, 2025). This increase indicates not only the region’s resiliency but also its prospects for long-term and sustainable growth. The mining and quarrying and manufacturing sectors are among the major drivers of economic growth in the SACU economies. However, this economic progress is closely linked to the extraction of natural resources beyond regeneration capability and potentially increases carbon emissions and deteriorates environmental quality (T. Mosikari, 2024). Therefore, many of the SACU economies grow at a cost. According to the Department of Mineral Resources and Energy (2023) in South Africa, 80% of the primary energy supply comes from non-renewable energy, and coal is dominant. In light of the energy security advantage from coal, Gava et al. (2025) noted that it appears deceptive to assume that non-renewable energy can easily be substituted by other main energy sources without lowering economic growth. However, pollution of the environment due to mining activities has become a source of sorrow. The Department of Forestry, Fisheries and the Environment (2018) noted that even though there is precise execution of well-defined rehabilitation procedures, degradation from pre-mining to post-mining land capability is unavoidable. Also, Greenpeace (2024) stated that in South Africa, energy generation alone accounts for 23% of the fine particulate matter in the air, with particles less than 2.5 micrometres in diameter (PM2.5) being a high burden. At the national level, South Africa and Eswatini, both SACU members, have significant coal-related PM2.5 emissions.

Figure 1 below is a basic line graph plotted as a panel of combined cross-sections, which shows trends in pollution of the environment proxied by CO₂ (carbon emission) in metric tonnes per capita for each SACU country member. Figure 1 shows that, on average, CO₂ emissions remained stable in Lesotho from 1985 to 2023, while in Eswatini, there has been a slight decline in 1993, 2011, and 2012 to just below 1, followed by an increase. CO₂ emissions in South Africa and Botswana are above those of all other SACU members, with South Africa above the rest of the countries between 1985 and 2023. Both Botswana and South Africa showed a considerable increase in carbon emissions per capita in 2014.

Figure 2 shows a basic line graph plotted as a panel of combined cross-sections to show trends in trade openness proxied by the composite trade index for each SACU country member. In South Africa, CO₂ emissions in metric tonnes per capita continued to rise until 2008. Despite some fluctuation, trade openness generally increased from 1994 to 2008 before declining in 2009. After 2009, the composite index increased steadily in South Africa until 2020. Furthermore, the composite trade index for other SACU members also suggests similar patterns, although it is below that for South Africa. The link between trade openness and carbon emissions could explain the same pattern in Figure 1 and Figure 2. It can be seen from the graphs that the composite index and CO₂ emissions decreased between 2007 and 2009. This could possibly be due to a drop in trade around the same period owing to the financial crisis, during which natural resource exports declined, as did the CO₂ emissions throughout the economies of SACU countries.

However, what remains to be answered is whether trade openness is beneficial or detrimental to environmental quality, particularly carbon emissions, in the SACU bloc. There is a lack of studies focusing on the short-run and long-run relationships between trade openness and environmental quality in the SACU bloc in Africa. There have been mixed findings from several studies that focused on selected Sub-Saharan Africa (SSA) countries (see Wicaksana & Karsinah, 2022; Oumarou & Nourou, 2024; Andriamahery et al., 2022; Acheampong et al., 2019) and Africa (see Ibrahim et al., 2021a; Twerefou et al., 2019; Mignamissi et al., 2024), without including all SACU member countries, as well as studies that investigated individual countries within the SACU, such as South Africa, as reported by Kohler (2013), Shahbaz et al. (2013), and Hasson and Masih (2017), who found that trade openness improves environmental quality, suggesting that trade openness contributes to lowering CO₂ emissions, but Dingiswayo et al. (2023) demonstrated that trade openness significantly increases CO₂ emissions. L. Wang and Ibrahim (2024), Udeagha and Ngepah (2021), and Ngepah and Udeagha (2022) found that trade contributes to rising emissions over time, and their results supported the existence of the Environmental Kuznets Curve (EKC).

According to J. Wang et al. (2024), Namibia, Botswana, and Lesotho are among the top ten African countries in import-based carbon emissions per gross domestic product, while South Africa is dominant in export-based carbon emissions per gross domestic product. This highlights the fact that carbon is embodied in traded items crossing borders. This ignites the complex debate on the relationship between trade openness and environmental quality in SACU. Van Tran (2020) and Shahbaz et al. (2017) suggested that trade openness is often linked to an increase in greenhouse gas emissions, especially in economies that are in developing regions characterised by weak ecological and environmental regulations. However, Hakimi and Hamdi (2020) noted that the effects are heterogeneous across regions and are influenced by institutional frameworks, income levels, and the structure of the economies.

In the SACU as a region, notable studies include those of T. Mosikari (2024) and Biyase et al. (2024), but none of the studies focus on trade openness. T. Mosikari (2024) examined the heterogeneous impact of industrialisation on environmental quality, while Biyase et al. (2024) looked at the relationship between remittance and CO₂ emissions. Since the SACU as a region shares trade policies and emphasises the need to move away from environmentally damaging economic activities, this study was also motivated by this argument.

The main objective of the investigation in the present study is to examine the short-run and long-run relationships between trade openness and environmental quality in the SACU in the context of the gains-from-trade hypothesis. Furthermore, the direction of causality between environmental equality and trade openness was also examined. The contribution of this study to the literature is that (1) it is the first to investigate the short-run and long-run effects of trade openness in the SACU; (2) the current study uses the comprehensive composite trade index, the dependence ratio, and the ratio of exports to real gross domestic product as proxies for trade openness to enhance the robustness of this estimation technique; (3) the study used a sample of data (1985 to 2023) that enhances the policy debate since there have been major developments in international trade arguments, growing decarbonisation calls, and shifts in environmental policy across Africa.

The Cross-Sectional Autoregressive Distributed Lag (CS-ARDL) approach was employed to examine the short-run and long-run relationships between trade openness and environmental quality. Furthermore, Granger non-causality tests were performed, especially for the Cross-Sectional Autoregressive Distributed Lag (CS-ARDL). This was performed to improve the predictive power and the robustness of the findings, since the study relates to energy and environmental policy. This will enhance the shift in trade policies and environmental policy frameworks in the region.

The findings of this study reveal that the estimated coefficients of trade openness (both proxies) positively and significantly (at the 1% level of significance) lead to carbon emissions in the short run and the long run. This implies that the gains-from-trade hypothesis does not hold in the SACU in either the short run or the long run. Trade openness significantly increases carbon emissions in the long run by a greater coefficient magnitude compared to the short run. Therefore, this can imply that in the SACU region, trade openness is a long-term contributor to poor environmental quality since it leads to an increase in carbon emissions by a higher magnitude. The Dumitrescu–Hurlin Granger non-causality test was employed, and the results show that there is bidirectional causality between CO₂ emissions and trade openness, CO₂ emissions and economic growth, and CO₂ emissions and population growth; unidirectional causality between CO₂ emissions and the exports-to-real-gross-domestic-product ratio; and no directional causality between foreign direct investment and CO₂ emissions. The empirical findings are robust, owing to the estimation technique employed and the proxies of trade openness used.

The outline of this study is as follows: Section 2 presents the literature review. Section 3 describes the research design. Section 4 discusses empirical analysis. Section 5 provides the conclusion and policy implications.

2. Literature Review

This section focuses exclusively on the empirical relationship between trade openness and environmental quality. However, the empirical literature on the association of foreign direct investment, population growth, and economic growth with environmental quality will be briefly discussed.

Turning to the empirical literature on the relationship between trade openness and environmental quality, the findings show complex and often contradictory evidence. The empirical literature is grouped as follows: firstly, country-specific studies conducted in South Africa, Namibia, Botswana, Eswatini and Lesotho; secondly, regional studies in SACU countries, Sub-Saharan Africa (SSA), and Africa; and thirdly, some international studies that examined the relationship between environmental quality and trade openness in countries in the European Union (EU), the Association of Southeast Asian Nations (ASEAN), and the North American Free Trade Agreement (NAFTA)/US–Mexico–Canada Agreement (USMCA). Therefore, the literature on the relationship between trade openness and environmental quality is divided into three groups.

The first group comprises country-specific empirical studies that explain the link between trade openness and environmental quality in South Africa, Namibia, Botswana, Eswatini, and Lesotho. In South Africa, the findings on the relationship between trade openness and environmental quality are mixed, revealing positive, negative, and insignificant effects. Kohler (2013), Shahbaz et al. (2013), and Hasson and Masih (2017) established that trade openness increases environmental quality, suggesting that trade openness contributes to reducing CO₂ emissions due to increased access to cleaner technologies and more efficient production methods. In contrast, a growing branch of the literature demonstrates that trade openness increases environmental deterioration. Dingiswayo et al. (2023) demonstrate that trade openness significantly increases CO₂ emissions, affirming the pollution haven hypothesis. L. Wang and Ibrahim (2024), Udeagha and Ngepah (2021), and Ngepah and Udeagha (2022) similarly demonstrated that trade liberalisation contributes to increasing CO₂ emissions in the long run, and they further affirmed the existence of the Environmental Kuznets Curve (EKC).

In Eswatini, Sacolo et al. (2018) found that between 1968 and 2015, even though the country improved its institutional structures, exports continued to be less than imports (trade balance deficit). Furthermore, Phiri (2019) investigated the effect of changes in economic activities on the EKC for a period spanning from 1970 to 2014 and found that economic growth significantly increases greenhouse gas emissions in the expansion phase of the business cycle. Still, when economic activities decline, the impact is insignificant. However, both studies, Sacolo et al. (2018) and Phiri (2019), fail to report on trade openness and environmental quality. Furthermore, empirical evidence on the environmental impact of trade openness in Botswana remains notably limited. Existing country-specific studies by Malefane (2020) and Chiwira et al. (2024) indicate that trade openness, particularly through an increase in exports, significantly increases economic growth. In Lesotho, Malefane and Odhiambo (2019) showed that trade openness has no significant impact on economic growth in either the short run or the long run; however, this was in contrast with the long-run findings by Makhetha and Rantaoleng (2017) that trade openness significantly reduces economic growth, but in short run, the results are similar to those of Malefane and Odhiambo (2019). Also, Sanusi and Dickason-Koekemoer (2024) found no long-run link between economic growth, trade openness, and financial development. In Namibia, Sunde et al. (2023) looked at the effects of exports, imports, and trade openness on economic growth and found that imports negatively affect economic growth, but exports and trade openness are positively linked to economic growth. The results were statistically significant. Also, taking into account the non-linear link between the variables, T. J. Mosikari and Eita (2020) examined the relationship between the main export sectors and economic growth in Namibia. The results reveal that growth in exports increases economic growth, while the decline in exports lowers economic growth.

Most available studies, which are country-specific, focus on South Africa, even though some SACU members are also exposed to climate change risks. However, in other countries, like Botswana, Lesotho, Eswatini, and Namibia, studies omit environmental variables, despite the resource-dependent nature of SACU economies, which rely heavily on the mining, agriculture, and manufacturing sectors’ exports and imports that are often pollution-intensive.

The second group comprises panel studies, which investigated the link between trade openness and environmental quality in the SACU, SSA, and Africa. In the SACU, T. Mosikari (2024) employed the quantile technique to examine the heterogeneous impact of industrialisation on environmental quality, and the results revealed that when industrialisation grew in the 4th to 6th quantiles, so did environmental degradation, but in the 7th to 8th quantiles, there was a negative relationship between industrialisation and environmental degradation. Biyase et al. (2024) looked at the relationship between remittance and CO₂ emissions and found that a modified EKC hypothesis holds in the SACU. However, none of the studies attempted to incorporate trade openness into their empirical models.

In Africa, Twerefou et al. (2019) analysed the relationship between trade liberalisation and environmental quality for a panel of 30 African countries using the generalised method of moments (GMM) estimation technique. The results revealed a positive relationship between trade openness and CO₂ emissions because of the composition effect, since factor endowment provided a comparative advantage. However, the original results suggest that trade openness enhances environmental quality. However, Van Tran (2020) looked at how trade openness influences the environment for a panel of 66 less developed countries, of which the bulk were African countries. Using two-step generalised method of moments (GMM) estimators, which included a finite-sample correction, the results revealed that trade openness increases emissions due to industrial growth and regulatory weaknesses, and this suggests the existence of the pollution haven hypothesis. Ibrahim et al. (2021a) explored various channels to determine whether trade influences environmental pollution (CO₂ emissions) in a panel of 47 African countries, and the study employed the dynamic GMM estimation technique. The results showed a statistically significant positive relationship between trade and CO₂ emissions. Also, the results failed to substantiate the claim of a turning point between the variables.

Also, Mignamissi et al. (2024) examined the link between CO₂ emissions and trade openness using the Two-Stage Least Squares technique. Despite the elasticity coefficients differing depending on the proxies used for trade openness, the overall results validated the PHH.

In Sub-Saharan Africa (SSA), H. Sun et al. (2020) empirically analysed the link between CO₂ emissions, energy consumption, trade openness, and economic growth. The results showed a cointegrating link amongst variables. Furthermore, the study found that in the long run, trade decreases environmental pollution. Wicaksana and Karsinah (2022) explored the impact of the effect of trade openness, energy, technology, and population on the environmental performance index. Using the panel data regression techniques, the results suggest a positive and significant impact of trade openness, energy, and population on the environmental performance index. In the study by Andriamahery et al. (2022), who used the panel GMM technique, the results revealed that trade has a positive relationship with nitrous oxide, agricultural methane, and CO₂ emissions for the selected countries in SSA and its income groups.

Okelele et al. (2022) investigated the effect of international trade (proxied by trade openness and foreign direct investment flows) on environmental quality (proxied by ecological footprint) in 23 SSA countries and concluded that ecological footprint in per capita terms is negatively correlated with trade openness but positively correlated with foreign direct investment inflows. Oumarou and Nourou (2024) also found that in a panel of 38 SSA countries, trade openness increases CO₂ emissions. Acheampong et al. (2019) used data from 1980 to 2015 for 46 Sub-Saharan African countries to study the effect of globalisation (foreign direct investment and trade openness) and renewable energy on CO₂ emissions. Using the first-generation panel data techniques, namely, fixed and random effects, the study found that renewable energy and foreign direct investment contribute to lowering carbon emissions, whereas trade openness harms the environment. Furthermore, by making use of time-series data spanning from 1975 to 2020, Ewane and Ewane (2023) explored the effects of trade openness and foreign direct investment (FDI) on environmental degradation in selected SSA countries in the context of the EKC; one of the key findings suggests that trade openness and FDI contribute to lowering CO₂ emissions in the short run while increasing it in the long run.

Besides the aforementioned classification of the above empirical research, results on the link between trade openness and environmental quality around the globe as a result of different trade agreements based on trade blocs, such the European Union (EU), the Association of Southeast Asian Nations (ASEAN), and the North American Free Trade Agreement (NAFTA)/US–Mexico–Canada Agreement (USMCA), are also mixed, revealing positive, negative, and insignificant effects. For example, in the EU, Tachie et al. (2020) obtained results that supported the presence of the EKC, pollution halo, and PHH. The Dumetriscu–Hurlin Granger causality test results suggest that unidirectional Granger causality exists, from trade openness to CO₂ emissions. Using the fully modified ordinary least squares, Al-Mulali et al. (2015) revealed that GDP growth, urbanisation, and financial development increase CO₂ emissions in the long run, while trade openness reduces them. Le et al. (2016) found that openness appears to lead to environmental degradation for the global sample. Moreover, Destek et al. (2018) found that trade openness decreases environmental degradation in EU countries, but Nam and Ryu (2024) indicated that trade initially negatively impacts the environment through greater CO₂ emissions but eventually contributes to a reduction in CO₂ emissions beyond a certain threshold.

Ling et al. (2020) noted that trade openness and carbon dioxide emissions do have a positive relationship among ASEAN-5 countries. The results reveal that only foreign direct investment is related to trade openness in the long run, while all the other variables, namely, carbon dioxide emissions, economic growth, and energy consumption, show both short-run and long-run relationships with trade openness in ASEAN-5 countries. However, A’yun and Khasanah (2022) found that economic growth lowers environmental quality, while trade openness (exports and FDI) leads to growth in CO₂ emissions in ASEAN countries. Hu et al. (2021) found that trade openness increases CO₂ emissions, but foreign direct investment, gross domestic product, and patents lower CO₂ emissions in ASEAN countries. Mahrinasari et al. (2019) found that trade liberalisation has a significant positive impact on carbon dioxide emissions in ASEAN countries. This was further supported by Burki and Tahir (2022), who revealed that trade openness and financial development increase environmental degradation in ASEAN countries. In the context of USMCA countries, Gómez and Rodríguez (2020) proxied environmental quality by ecological footprint to explore the presence of the Environmental Kuznets Curve in the USMCA countries. Using the method of moments quantile regression technique, the results suggest that trade openness has no significant effect on environmental quality in the USMCA countries, and only renewable energy was found to be a variable that significantly lowered environmental degradation. However, the results from the study by Miranda et al. (2020) indicated that the EKC hypothesis holds in both Mexico and the United States of America (USA) but does not hold in Canada.

Foreign direct investment inflows also continue to be an influential factor, especially for the destination countries. FDI inflow may have either benefits or detrimental effects on the environment of the receiving country. The study by Huay et al. (2022) suggests that polluting firms from developed nations reallocate their production units to less developed nations via direct investment. Ali et al. (2022) found that FDI inflows negatively influence the environmental quality via the PHH. On the contrary, the pollution halo hypothesis suggests that foreign investments can improve environmental quality in the receiving countries, since foreign firms can bring clean and efficient technical knowledge. Therefore, domestic industries may improve overall ecological quality, as suggested by the results of studies by C. Sun et al. (2017) and Ahmad et al. (2021). Several authors have explored the link between foreign direct investment inflow and environmental quality in the context of the PHH, such as Tsoy and Heshmati (2023), Qamruzzaman (2023), and Musah et al. (2022), and yielded inconclusive and mixed results. Therefore, it is argued that foreign direct investment (FDI) inflow into the SACU economies may be of great importance, especially in the era of green and sustainable growth and clean energy consumption calls.

Changes in environmental quality can depend on population growth patterns. According to Tal (2025), growth in population leads to irreversible effects on the environment. Martínez-Zarzoso and Maruotti (2011) pointed out that, in some scenarios, population density leads to better environmental outcomes. This is due to the role of economies of scale in energy use and more efficient infrastructure. However, in the study by Rehman et al. (2021), the results indicate that population growth generally contributes to higher ${\mathrm{CO}}_2$ emissions. Also, Dimnwobi et al. (2021) investigated the nexus between population changes and environmental degradation in five selected countries in Africa. The results indicate that population growth increases environmental decay.

Economic growth contributes to environmental quality, and the link is embodied in the Environmental Kuznets Curve (EKC) hypothesis. The study by Hunjra et al. (2024) found a positive relationship between economic growth and ${\mathrm{CO}}_2$ emissions in the early stages, but at a later stage, the relationship became negative. Phiri (2019) found that economic growth significantly increases greenhouse gas emissions in the expansion phase of the business cycle. Olaoye (2024) investigated the relationship between ${\mathrm{CO}}_2$emissions and economic growth in African countries, and the results indicate that ${\mathrm{CO}}_2$emissions increase economic growth.

Throughout the literature, studies which focused on the SACU region as a bloc in Africa have not investigated the short-run and long-run relationships between trade openness and environmental quality. Apart from mixed findings, several studies focus on selected Sub-Saharan Africa (SSA) and Africa without including all SACU member countries or investigating individual countries within the SACU. In the SACU as a region, notable studies include T. Mosikari (2024) and Biyase et al. (2024), but none of the studies focus on trade openness. T. Mosikari (2024) examined the heterogeneous impact of industrialisation on environmental quality, while Biyase et al. (2024) looked at the relationship between remittance and CO₂ emissions. Furthermore, the study used the traditional trade openness measures (exports-to-real-gross-domestic-product ratio, dependency ratio) and the composite trade index, which is a comprehensive measure for evaluating whether trade openness influences environmental quality. Therefore, the existing literature still lacks evidence on the relationship between trade openness and environmental quality based on the Cross-Sectional Autoregressive Distributed Lag (CS-ADRL), and as suggested by Bergougui and Zambrano-Monserrate (2025), Granger non-causality tests were employed to improve the predictive power, especially since the study relates to energy and environmental policy, which was ignored by various studies.

3. Research Design

3.1. Theoretical Analysis

The interplay between trade openness and environmental quality, specifically emission levels, continues to be widely studied using two dominant theoretical strands: neoclassical trade and environmental economists’ perspective (gains-from-trade hypothesis) and the ecological school of thought perspective (the race-to-the-bottom-effect hypothesis). Furthermore, there exist arguments in the theoretical literature regarding the direct and indirect impacts of trade openness and other variables on environmental quality. Two important hypotheses are the Environmental Kuznets Curve (EKC) hypothesis and the pollution haven hypothesis (PHH), which offer insights into how variables influence environmental quality. The theoretical explanations of the gains-from-trade hypothesis, the race-to-the-bottom-effect hypothesis, the pollution haven hypothesis (PHH), and the Environmental Kuznets Curve (EKC) hypothesis will be discussed.

Using the gains-from-trade hypothesis, the proponents of the neoclassical trade theory and environmental economists suggest that trade openness enhances and magnifies environmental quality (Muradian & Martinez-Alier, 2001). This is based on the suggestion that trade openness has a long-run growth-promoting effect, which allows for the conservation of the environment (the environment is viewed as a normal good). Muradian and Martinez-Alier (2001) further noted that this key proposition largely relies on the Heckscher–Ohlin–Samuelson theory, which suggests that if a country enjoys a comparative advantage for a certain good, then the production of that good should increase. With trade openness, more efficient goods that are environmentally friendly will be allowed to move, and clean production techniques may be transferred.

Contrary to the neoclassical trade and environmental economists’ view is the ecological school’s suggestion that free trade may lead to a race to the bottom, formalized as the race-to-the-bottom-effect hypothesis. That is, faced with competitive pressure and the need to specialise, trading countries may be pushed to maximise their profits at the cost of the environment (Gerritse, 2021).

The Environmental Kuznets Curve (EKC) hypothesis, which originated from the seminal work of Grossman and Krueger (1991), claims that the link between environmental degradation and the growth of an economy resembles an inverted U-shaped graph. Based on this claim, pollution levels tend to rise with income at the early stages of economic growth because of growth in industrialisation and structural changes in the economy. However, once the region passes a certain level of income, economic growth enhances environmental quality. This is because the community will adopt environmentally friendly technologies, and more rules to address environmental degradation will be enforced. Shafik (1994) and Soytas et al. (2007) extended this framework by incorporating energy use and trade openness variables. Ang (2007) empirically validated the EKC hypothesis using carbon emission data in both short-run and long-run contexts. Cole (2004) advanced the debate by interrogating the robustness of the EKC across different pollutants and countries, finding that the relationship holds more consistently for local air pollutants than for global pollutants like CO₂ emissions.

Furthermore, intricately linked to the EKC is the pollution haven hypothesis (PHH). The pollution haven hypothesis (PHH) asserts that trade can lead to environmental degradation in countries with weak environmental regulations, as pollution-intensive industries relocate from stricter to laxer authorities. This hypothesis is especially pertinent for developing and middle-income countries like those in the SACU. Copeland and Taylor (2004) modelled this link, suggesting that disparities in environmental policy rigour affect comparative advantages in trade. Cole (2004) concurred with the empirical existence of the PHH and indicated that trade-induced pollution depends not only on regulatory frameworks but also on capital intensity and industrial composition. Kearsley and Riddel (2010) refined the PHH by accounting for heterogeneity across sectors and countries, noting that pollution havens are not inevitable but conditional on trade patterns, factor endowments, and regulatory enforcement.

In the context of SACU countries, the theoretical relevance of the gains-from-trade hypothesis is based on the fact that the hypothesis is interpreted based on the direct effects of trade on environmental equality in open economies. Also, the theoretical relevance of the PHH is that the inclusion of foreign direct investment inflow in the model is based on the notion that polluting firms from developed nations reallocate their production units to less developed nations via direct investment, making it central to the trade–environment nexus. However, despite the inclusion of RGDP in the model, as well as the acknowledgement that the EKC provides a foundation for hypothesising dynamic environmental responses to growth, the evidence of the EKC is not investigated. The study aims to look at the short-run and long-run effects of trade openness on environmental quality in SACU countries.

3.2. Model Design

In this study, we started with the formulation of the model motivated by Oumarou and Nourou (2024) and H. Sun et al. (2020) to establish the relationship between trade openness (proxied by the CTI, which is composite trade intensity, explained in Equation (2), and the dependency ratio (DRA), elaborated in Equation (3)) and environmental quality (proxied by CO₂ emissions) in SACU countries from 1985 to 2023. Furthermore, the widely used trade-to-GDP ratio was used as a proxy for trade openness (X/RGDP) to check the robustness of the results. The factors that influence environmental quality (CO₂ emissions) were decomposed based on the Stochastic Impacts by Regression on Population, Affluence, and Technology (STIRPAT) framework by Richard & Kaya in 1989, as refined by York et al. (2003). The STIRPAT framework allows for the inclusion of additional explanatory variables in the standard trade and environment relationship.

$${{\mathrm{CO}}_2}_{\mathrm{it}}={\mathsf{\alpha }}_{\mathrm{it}}+{\mathsf{\gamma }\mathrm{TO}}_{\mathrm{it}}+{\mathsf{\theta }\mathrm{X}}_{\mathrm{it}}+{\mathsf{ϵ}}_{\mathrm{it}}$$

(1)

where ${\mathrm{CO}}_2$ is the carbon emissions, $\mathrm{TO}$ is trade openness, and $\mathrm{X}$ is the $\mathrm{K}\times 1$vector of control variables ($\mathrm{K}=3)$. The control variables include $\mathrm{RGDP},\mathrm{PPNGR},$and $\mathrm{FDI}.$In addition, $\mathrm{t}$ is the time, i denotes the country, and ${\mathsf{ϵ}}_{\mathrm{it}}$ is the error term.

${\mathrm{CO}}_2$ = Environmental quality proxy: Carbon dioxide (${\mathrm{CO}}_2$) emissions (in metric tons) per capita (see Pham & Nguyen, 2024). We used CO₂ emissions as the basic proxy for environmental quality for three key reasons. First, CO₂ emissions have been extensively used as a basic indicator of environmental quality, as noted by Yang et al. (2023) and Le et al. (2016); therefore, its use enables us to draw comparisons with previous research. Second, historical data indicate a significant surge in CO₂ emissions, making it one of the most concerning contributors to air pollution and the exacerbation of global warming. Thirdly, the CO₂ emissions variable, which was used as a proxy for environmental quality in this study, lumps together emissions from various industries and sectors, such as industrial combustion, the power industry, waste, agriculture, and industrial processes, for each and every country.

PPNGR = Population growth (annual %): The population between the ages of 15 and 64. The population is based on the de facto definition of population, which counts all residents regardless of legal status or citizenship.

FDI = Foreign direct investment, net inflows (% of RGDP): The series indicates net inflows (new investment inflows minus disinvestments) in the reporting economy from foreign investors and is divided by the country’s GDP, according to the World Bank.

TO = Trade openness is represented by (TO), which shows the extent to which a particular country is integrated into the global economy and is measured by the composite trade intensity (CTI) (see Squalli & Wilson, 2011 and Ibrahim et al., 2021b). The dependency ratio (DRA) and trade-to-GDP ratio (X/RGDP) are also included to ensure the robustness of the results.

RGDP = Economic growth (GDP at constant prices, 2015 in USD).

Therefore, the calculations of the CTI and DPR are as follows:

$${\mathrm{CTI}}_{\mathrm{i}}={\displaystyle \frac{{\left(\mathrm{EXP}+\mathrm{IMP}\right)}_{\mathrm{i}}}{\frac{1}{\mathrm{n}}{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{n}}{\left(\mathrm{EXP}+\mathrm{IMP}\right)}_{\mathrm{j}}}}\times {\displaystyle \frac{{\left(\mathrm{EXP}+\mathrm{IMP}\right)}_{\mathrm{i}}}{{\mathrm{RGDP}}_{\mathrm{i}}}}$$

(2)

$${\mathrm{DRA}}_{\mathrm{i}}={\displaystyle \frac{{\left(\mathrm{EXP}+\mathrm{IMP}\right)}_{\mathrm{i}}}{{\mathrm{RGDP}}_{\mathrm{i}}}}$$

(3)

where $\mathrm{i}$ is a country subscript; ${\mathrm{CTI}}_{\mathrm{i}}$ is country $\mathrm{i}$’s composite trade share, with the adjustment of a country’s trade share to the gross domestic product at constant prices (RGDP) relative to the world at large; ${\left(\mathrm{EXP}+\mathrm{IMP}\right)}_{\mathrm{i}}$ is country i’s aggregated exports and imports; n denotes the total number of countries; ${\displaystyle \frac{1}{\mathrm{n}}}$ is the average value of the world trade share of all the countries included; $\mathrm{j}$ denotes the whole world; and ${\left(\mathrm{EXP}+\mathrm{IMP}\right)}_{\mathrm{j}}$ denotes the sum of exports and imports in the whole world. ${\mathrm{RGDP}}_{\mathrm{i}}$ shows country i’s gross domestic product at constant prices (RGDP).

Table 1. Definitions of the variables.

Variable	Definition of Variable and Measurement	Source
RGDP	Economic growth (GDP (constant 2015 USD))	World Bank Development Indicators
${\mathrm{CO}}_2$	$\mathrm{Environmental}\mathrm{quality}:\mathrm{measured}\mathrm{by}\mathrm{per}\mathrm{capita}\mathrm{carbon}\mathrm{dioxide}({\mathrm{CO}}_2$) emissions (in metric tons)	World Bank Development Indicators
CTI	Composite trade intensity	World Bank Development Indicators
DRA	Dependency ratio	World Bank Development Indicators
PPNGR	Population growth (annual %)	World Bank Development Indicators
FDI	Foreign direct investment, net inflows (% of GDP)	World Bank Development Indicators
IMP	Exports of goods and services (constant 2015 USD)	World Bank Development Indicators
EXP	Imports of goods and services (constant 2015 USD)	World Bank Development Indicators

below reports a summary of the description of the variables.

Also, prior economic theory and explanations suggest that trade openness (TO) has an expected positive or negative relationship with environmental equality. TO has a positive sign (${\displaystyle \frac{{\mathrm{CO}}_2}{\mathrm{TO}}}>0)$ or negative sign (${\displaystyle \frac{{\mathrm{CO}}_2}{\mathrm{TO}}}<0)$. Also, economic growth (measured by RGDP) may have a negative sign (${\displaystyle \frac{{\mathrm{CO}}_2}{\mathrm{RGDP}}}<0$) or positive sign $({\displaystyle \frac{{\mathrm{CO}}_2}{\mathrm{RGDP}}}>0)$. Economic growth (RGDP) is included because several studies have suggested that ${\mathrm{CO}}_2$ emissions increase, especially in developing economies where growth is driven by energy-intensive sectors, but others found the opposite. For example, Hunjra et al. (2024) suggest that economic growth typically increases ${\mathrm{CO}}_2$ emissions in the early stages but may eventually lead to environmental improvements at higher levels of income, as seen in the Environmental Kuznets Curve (EKC) hypothesis. Phiri (2019) found that economic growth significantly increases greenhouse gas emissions in the expansion phase of the business cycle. Given that growth in many SACU countries is often linked to the agriculture and mining sectors, its inclusion ensures that the impact of trade openness on environmental quality is not confused with emissions induced by economic growth.

Furthermore, population growth is anticipated to positively affect environmental quality (${\displaystyle \frac{{\mathrm{CO}}_2}{\mathrm{PPNGR}}}>0).$Also, population growth is included to account for the rising demand for energy and natural resources in SACU economies, which often leads to increased ${\mathrm{CO}}_2$emissions. A study by Rehman et al. (2021) has shown that larger population growth generally contributes to higher ${\mathrm{CO}}_2$ emissions. However, as Martínez-Zarzoso and Maruotti (2011) pointed out, in certain contexts, population density can lead to better environmental outcomes due to the role of economies of scale in energy use and more efficient infrastructure. Thus, including population growth helps to capture these potentially contradictory effects from theoretical explanations.

Foreign direct investment inflow is anticipated to positively influence ${\mathrm{CO}}_2$ emissions (${\displaystyle \frac{{\mathrm{CO}}_2}{\mathrm{FDI}}} >0)$. Foreign direct investment inflow (FDI) is included to account for the dual role that FDI can play in environmental quality. FDI inflows to SACU countries are fundamentally driven by the availability of rare mineral resources. Consequently, international companies will move their production activities to SACU countries. Furthermore, however, if the countries have inadequate environmental policies, it eventually leads to environmental degradation (${\displaystyle \frac{{\mathrm{CO}}_2}{\mathrm{FDI}}} >0$). On the other hand, FDI inflow can enhance the adoption of cleaner energy use technologies in the region through changes in production methods, resulting in eco-friendly activities (T. Mosikari & Xaba, 2025). This justifies the inclusion of the variable in the model.

To investigate the short-run and long-run elasticities of the relationship between trade openness and environmental quality, including the corresponding control variables in SACU countries, Equation (1) is estimated using the Cross-Sectional-Autoregressive-Distributed Lag (CS-ARDL) approach, which Chudik and Pesaran (2015) expressed in general form (see Equation (4) below) as

$$∆y_{it}={\mu }_i+{\varnothing }_i\left(y_{it-1}-{\beta }_i{x}_{it-1}-{\theta }_{1i}{\overline{y}}_{it-1}-{\theta }_{2i}{\overline{x}}_{it-1}\right)+{\displaystyle \sum }_{j=1}^{p-1}{\rho }_{ij} ∆y_{i, t-j}+ {\displaystyle \sum }_{j=0}^{q-1}{\gamma }_{ij}∆x_{i,t-j}+{\tau }_{1i}∆{\overline{y}}_t+{\tau }_{2i}∆{\overline{x}}_t+{ϵ}_{it}$$

(4)

In Equation (4) above, $y_{it}$ denotes ${\mathrm{CO}}_2$, which is the response variable; ${\mu }_i$ is a constant term; and ${\beta }_i$ represents the slope coefficients of the explanatory variables as well as the lagged response variable. Also, $x_{it}$ denotes a vector of explanatory variables (PPNGR, FDI, RGDP, TO (CTI, DRA, or X/RGDP)). The symbol ${\varnothing }_i$ denotes the error correction term (ECM). The ECM indicates the speed of adjustment of the response variable to its long-run equilibrium after a shock. Lastly, ${\overline{y}}_{t-1}$ and ${\overline{x}}_{t-1}$ serve as proxies for unobserved components in the long run, while $∆{\overline{y}}_t$ and $∆{\overline{x}}_t$ provide proxies for unobserved components in the short run in Equation (4), which assists in capturing the persistent effect of unobserved components on the outcome variables over time, and ${ϵ}_{it}$ is the error term.

3.3. Variables and Data

The data for this study was collected from the World Bank Development Indicators (WDI) and covers the period 1985 to 2023 for SACU countries (South Africa, Botswana, Namibia, Lesotho, and Eswatini). The definitions and the measurements of variables used in the study are explained in Table 1 above.

4. Empirical Analysis

4.1. Cross-Sectional Dependence Test

De Hoyos and Sarafidis (2006) suggest that before conducting panel unit root testing, the first step should be to check cross-sectional dependency. This is an important procedure in panel data analysis because it informs the type of the panel unit root test to be employed in a study. Furthermore, in the presence of cross-sectional dependence, all first-generation panel unit root tests, namely, the Levin and Chu test (LLC, 2002), Im–Pesaran–Shin (IPS, 2003), and Fisher-Augmented Dickey–Fuller (Fisher-ADF) test, become unsuitable for checking the stationarity of the variables. Herwartz and Siedenburg (2008) indicated that the first-generation panel unit tests assume cross-sectional independence, and in the presence of cross-sectional dependence, the tests will provide misleading results. Therefore, second-generation panel unit root tests, such as the Cross-Sectionally Augmented IPS (CIPS) test and Cross-Sectionally Augmented ADF (CADF) by Pesaran (2007), are suitable and more robust in a scenario where the assumption of cross-sectional independence is violated.

According to Pesaran (2020), in panel data techniques, the error terms are assumed to be cross-sectionally independent, but this is correct only with fairly large cross-sectional units (N). If the number of cross-sectional units (N) is insufficient, namely less than ten, and the time dimension(T) is fairly large, Pesaran (2020) indicated that, to model the cross-correlations of the errors, a seemingly unrelated regression equation (SURE) setup is suitable. The Breusch and Pagan (1980) Lagrange multiplier (LM) test, which is based on the expected value of the squared pair-wise correlation of the residuals, follows the seemingly unrelated regression equation (SURE) framework. It is important to note that the LM test is a suitable test for cross-sectional dependence if T is greater than N; however, the LM test is not desirable in the case of T being less than N (Pesaran, 2020).

In the present study, T is 39 and N is 5; therefore, according to Pesaran (2020), the Breusch–Pagan LM test is adopted.

Table 2. Summary results for cross-sectional dependence.

Variables	Breusch–Pagan LM	Pesaran Scaled LM	Pesaran
${\mathrm{CO}}_2$	53.50297 ***	9.727559 ***	2.025752 **
RGDP	92.56337 ***	18.46173 ***	2.194617 **
PPNGR	138.9527 ***	28.83470 ***	11.23921 ***
CTI	28.00046 ***	4.025025 ***	−0.632443
FDI	11.30038	0.290774	−0.149889
DRA	27.05682 ***	3.814021 ***	0.426581
X/RGDP	148.3399 ***	30.93374 ***	−0.131084 **

Note: *** 1 percent level of significance, ** 5 percent level of significance.

indicates the summary results of tests for cross-sectional dependence. For the Breusch and Pagan (1980) Lagrange multiplier (LM), the null hypothesis is that residuals are cross-sectionally independent. The results from Table 2 below indicate that the Breusch–Pagan LM test rejected the null hypothesis at a 1% level of significance and a 5 percent level of significance; therefore, we concluded that there is cross-sectional dependence.

4.2. Panel Unit Root Tests

The second-generation panel unit root test, namely, the cross-sectionally augmented Im–Pesaran–Shin (CIPS) panel unit root test proposed by Pesaran (2007), and the cross-sectionally Augmented Dickey–Fuller (CADF) panel unit root tests proposed by Pesaran (2003) were used due to the presence of cross-sectional dependence in the variables used in this study. Cerasa (2008) first specified a dynamic linear heterogeneous panel data model, as presented in Equation (5) below:

$$x_{it}=(1-{\varnothing }_i){\mu }_i+{\varnothing }_i{x}_{it-1}+u_{it}$$

(5)

From Equation (5) above, $u_{it}$ is assumed to have a single common factor structure ($u_{it}={\gamma }_i{f}_t+{\epsilon }_{it}$), in which ${\epsilon }_{it}$ is an idiosyncratic component, ${\gamma }_i$ is an individual factor loading, and $f_t$ is the unobserved common effect. ${\epsilon }_{it}$, ${\gamma }_i$, and $f_t$ are all independent and identically distributed, with each component having an expected value of zero and a constant variance. Equation (5) then, can then be rewritten as

$$∆x_{it}= {\propto }_{i }+{\beta }_i{x}_{it-1}+{\epsilon }_{it}$$

(6)

Given Equation (6), Pesaran (2007) then proxies $f_t$ with the cross-section average of $x_{it}$. Pesaran (2003) augmented the standard Dickey–Fuller (DF) regressions with the cross-section average of lagged individual series in levels and first differences. Hurlin and Mignon (2006) also mentioned that in the absence of a serial correlation of residuals, Equation (5) below should be employed for regression for the ith cross-section (country).

$$∆x_{it}= {\propto }_{i }+b_i{x}_{it-1}+{\rho }_i{\overline{x}}_{t-1}+{\tau }_i∆{\overline{x}}_t+{\epsilon }_{it}$$

(7)

From Equation (5), ${\overline{x}}_t=\left({\displaystyle \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$N$}\right.}\right){{\displaystyle \sum }}_{i=1}^N{x}_{it}$, which is the cross-sectional mean of $x_{it}$ and its lagged values ($x_{t-1}, x_{t-2, }{x}_{t-3} . . .)$. Also, the testing of the null hypothesis of a unit root with respect to unit $i$ will then be based on $t_i\left(N,T\right),$ which is the t-statistic of the ordinary least squares (OLS) estimate of ${\beta }_i$ in the CADF regression Equation (5) above. According to Cerasa (2008), for the test of the null hypothesis $H_0=0$ for all $i$ against the heterogeneous alternative hypothesis $H_1:{\beta }_1<0, . . . {\beta }_{N_0}<0, N_0 \le N,$ the entire panel data set is represented by the average of the individual CADF statistic given in Equation (8) below:

$$CIPS\left(N,T\right)=\left({\displaystyle \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$N$}\right.}\right){\displaystyle \sum }_{i=1}^N{t}_i\left(N,T\right)$$

(8)

The results of the CADF and CIPS tests are shown in

Table 3. Summary results for CIPS and CADF panel unit root tests.

Variables	CIPS I (0)	CIPS I (1)	CADF I (0)	CADF I (1)
${\mathrm{CO}}_2$	−2.308 ***	-	−1.911 **	-
RGDP	−2.417 **	-	−2.555 ***	-
PPNGR	−2.863 ***	-	−1.086	−5.094 ***
CTI	−1.041	−4.939 ***	1.361	−2.173 **
FDI	−3.641 ***	-	−1.388 *	-
X/RGDP	−2.614 ***	-	−1.685 **
DRA	−1.490	−5.102 ***	0.294	−2.592 ***

Note: *** Significance at 1% level. ** Significance at 5% level. * Significance at 10% level.

below. Prior to conducting CADF and CIPS tests and running the CS-ARDL models (1, 2, and 3), ${\mathrm{CO}}_2$, CTI, DRA, RGDP, and X/RGDP, but not FDI and PPNGR, were log transformed. For the CADF and CIPS tests, the deterministic choice used is the intercept, and the maximum number of lags is two. The results indicate a mixed order of integration for both CIPS and CADF. The CIPS test shows that DRA and CTI are non-stationary at the 1% level with an integration order of 1, while ${\mathrm{CO}}_2,$FDI, PPNGR, X/RGDP, and RGDP are stationary variables, and the order of integration is 0 (${\mathrm{CO}}_2,$ FDI, PPNGR, and X/RGDP are statistically significant at 1% while RGDP is statistically significant at 5%). The CADF test shows that RGDP (statistically significant at 1%), X/RGDP, ${\mathrm{CO}}_2$ (statistically significant at 5%), and FDI (statistically significant at 10%) are stationary (the order of integration is 0), whereas PPNGR, DRA, and CTI are non-stationary and become stationary after differencing the variable once (integrated of order 1).

4.3. CS-ARDL Approach

After establishing that there is a mixed order of integration, we then estimated the long-run and short-run elasticities of the relationship between environmental quality (${\mathrm{CO}}_2$) and the independent variables by employing the CS-ARDL approach.

Table 4. Summary results for CS-ARDL model 1, with CTI as proxy for trade openness.

Dependent Variable: CO2
Long Run			Short Run
Variable	Slope Coefficient	Standard Errors	Variable	Slope Coefficient	Standard Errors
${\mathrm{CO}}_2\left(-1\right)$	-	-	$∆{\mathrm{CO}}_2\left(-1\right)$	0.5750071 ***	0.0651804
RGDP	1.882709	1.249833	$∆$RGDP	0.7232274	0.5909511
PPNGR	−0.1355439 **	0.0534207	$∆$PPNGR	−0.0627808 **	0.0299405
CTI	0.4249157 *	0.2240551	$∆$CTI	0.1475776 **	0.0602636
FDI	−0.0061098	0.0124836	$∆$FDI	−0.0046199	0.0056475
-	-	-	ECT (−1)	−0.4249929 ***	0.0651804

Note: *** Significance at 1% level. ** Significance at 10% level. * Significance at 10% level.

reports the results of CS-ARDL estimation of model (1), where the CTI was used as a proxy for trade openness. The results reveal that trade openness (CTI) and PPNGR have a statistically significant relationship with ${\mathrm{CO}}_2$ emissions in both the short run and the long run. Trade openness (CTI) is statistically significant at 5% in the short run and statistically significant at 10% in the long run. While PPNGR has a long-run and short-run statistically significant relationship with ${\mathrm{CO}}_2$ emissions at a 5% level of significance, the lag of ${\mathrm{CO}}_2$ (${\mathrm{CO}}_2 \left(-1\right))$ is statistically significant in the short run, and FDI is statistically insignificant in both the short run and long run. The positive values in the short and long run (CS ARDL) of the coefficient of RGDP and trade openness (CTI) show that the increase in these variables also increases carbon emissions in the SACU countries. However, negative values of the coefficients of FDI and PPNGR in the short and long run (CS-ARDL) suggest that an increase in these variables will result in a decrease in ${\mathrm{CO}}_2$ emissions.

Furthermore, the second CS-ARDL model (model 2) with DRA as a proxy for trade openness was estimated, and the results are presented in

Table 5. CS-ARDL model 2 (with DRA as proxy for trade openness) results.

Dependent Variable: CO2
Long Run			Short Run
Variable	Slope Coefficient	Standard Errors	Variable	Slope Coefficient	Standard Errors
${\mathrm{CO}}_2\left(-1\right)$	-	-	$∆{\mathrm{CO}}_2\left(-1\right)$	0.5895166 ***	0.0751611
RGDP	2.017252	1.229326	$∆$RGDP	0.7570938	0.5497164
PPNGR	−0.1395158 ***	0.0509793	$∆$PPNGR	−0.0645091 **	0.0278822
DRA	0.5051549 **	0.2270173	$∆\mathrm{DRA}$	0.1988687 **	0.1043904
FDI	0.0006616	0.0154135	$∆$FDI	−0.0034782	0.0060871
-	-	-	ECT (−1)	−0.4104834 ***	0.0751611

Note: *** Significance at 1% level. ** Significance at 10% level.

below. The results show that trade openness (DRA) and PPNGR have a statistically significant relationship with ${\mathrm{CO}}_2$ emissions in both the short run and the long run, and the lag of ${\mathrm{CO}}_2$ (${\mathrm{CO}}_2 \left(-1\right))$ is statistically significant in the short run, but FDI and RGDP are statistically insignificant in the short run and the long run. The positive values of the coefficients of RGDP and trade openness (DRA) in the short and long run (CS-ARDL) show that the increase in these variables increases ${\mathrm{CO}}_2$ emissions in the SACU countries. Lastly, negative values of the coefficients for FDI and PPNGR in the short run and PPNGR in the long run (CS-ARDL) suggest that if there is an increase in these variables, there will be a decrease in ${\mathrm{CO}}_2$ emissions, which implies an improvement in environmental quality.

The third CS-ARDL model (model 3) with X/RGDP as a proxy for trade openness was estimated, and the results are presented in

Table 6. CS-ARDL model 3 (with X/RGDP as proxy for trade openness) results.

Dependent Variable: CO2
Long Run			Short Run
Variable	Slope Coefficient	Standard Errors	Variable	Slope Coefficient	Standard Errors
${\mathrm{CO}}_2\left(-1\right)$	-	-	$∆{\mathrm{CO}}_2\left(-1\right)$	0.6177472 ***	0.0919526
RGDP	1.577557	1.02025	$∆$RGDP	0.4495487	0.5211319
PPNGR	−0.302566 ***	0.0553669	$∆$PPNGR	−0.0988562 ***	0.0222368
X/RGDP	0.0210102 *	0.0127453	$∆\mathrm{X}/\mathrm{RGDP}$	0.0051716	0.0034995
FDI	0.0066822	0.0132771	$∆$FDI	−0.0020812	0.0020100
-	-	-	ECT (−1)	−0.3822528 ***	0.0919526

Note: *** Significance at 1% level, * Significance at 10% level.

below. The results show that trade openness (X/RGDP) and PPNGR have a statistically significant relationship with ${\mathrm{CO}}_2$ emissions in the long run (X/RGDP at 10% while PPNGR is at 1%). PPNGR further shows a 1% statistically significant relationship in the short run, while X/RGDP shows an insignificant relationship with ${\mathrm{CO}}_2$ emissions; the lag of ${\mathrm{CO}}_2$ (${\mathrm{CO}}_2 \left(-1\right))$ is statistically significant in the short run at the 1% level of significance, but FDI and RGDP are statistically insignificant in the short run and the long run. The positive values of the coefficients of RGDP and trade openness (X/RGDP) in the short and long run (CS-ARDL) show that an increase in these variables increases ${\mathrm{CO}}_2$ emissions in SACU countries. The negative values of the coefficients of FDI and PPNGR in the short run and PPNGR in the long run (CS-ARDL) suggest that if there is an increase in these variables, there will be a decrease in ${\mathrm{CO}}_2$ emissions, which implies an improvement in environmental quality.

4.4. Panel Granger Causality Test

Given the fact that this study attempts to give an account of the effect of trade openness on environmental policy, Granger non-causality tests are employed, as suggested by Bergougui and Zambrano-Monserrate (2025), to enhance the predictive power and the robustness of the results for policy analysis. The Dumitrescu and Hurlin (2012) Granger non-causality test, which is a better type of causality test, was used. The test enjoys advantages such as allowing all the coefficients to differ across cross-sections (heterogeneity across cross-sections) and the fact that it can be used when cross-sectional dependency is detected. Furthermore, Dumitrescu and Hurlin (2012) Granger non-causality can be employed in both cases where T is greater than N and those where T is less than N and in unbalanced and heterogeneous panels. In the present study, T is 39 and N is 5, and there is cross-sectional dependence. Therefore, given the context, heterogeneity may either emanate from the regression model or be based on the causal link from CTI, RGDP, PPNGR, and FDI to ${\mathrm{CO}}_2$, from X/RGDP to ${\mathrm{CO}}_2$ and from DRA to ${\mathrm{CO}}_2$. The results are presented in

Table 7. Dumitrescu–Hurlin Granger non-causality results.

Causality	Zbar-Stat	p-Value	Conclusion
${\mathrm{CO}}_2\to \mathrm{CTI}$	2.5825 ***	0.0098	Bidirectional causality between carbon emissions and trade openness
$\mathrm{CTI}\to {\mathrm{CO}}_2$	1.9129 *	0.0558
${\mathrm{CO}}_2\to \mathrm{RGDP}$	2.1452 **	0.0319	Bidirectional causality between carbon emissions and economic growth
$\mathrm{RGDP}\to {\mathrm{CO}}_2$	2.3173 **	0.0205
${\mathrm{CO}}_2\to \mathrm{PPNGR}$	4.6423 ***	0.0000	Bidirectional causality between carbon emissions and population growth
$\mathrm{PPNGR}\to {\mathrm{CO}}_2$	6.3258 ***	0.0000
${\mathrm{CO}}_2\to \mathrm{FDI}$	−0.8965	0.3700	No directional causality between foreign direct investment and carbon emissions
$\mathrm{FDI}\to {\mathrm{CO}}_2$	−0.2487	0.8036
${\mathrm{CO}}_2\to \mathrm{DRA}$	3.0623 ***	0.0022	Bidirectional causality between carbon emissions and dependency ratio
$\mathrm{DRA}\to {\mathrm{CO}}_2$	1.8463 *	0.0649
${\mathrm{CO}}_2\to \mathrm{X}/\mathrm{RGDP}$	6.9322 ***	0.0000	Unidirectional causality between exports-to-real-gross-domestic-product ratio and carbon emissions
$\mathrm{X}/\mathrm{RGDP}\to {\mathrm{CO}}_2$	1.4888	0.1365

Note: *** Significance at 1% level. ** Significance at 5% level. * Significance at 10% level.

below, which shows that there is bidirectional causality between ${\mathrm{CO}}_2$ emissions and trade openness (CTI as well as DRA), unidirectional causality between X/RGDP and ${\mathrm{CO}}_2$ emissions (from ${\mathrm{CO}}_2$ emissions to X/RGDP), bidirectional causality between ${\mathrm{CO}}_2$ emissions and economic growth, bidirectional causality between ${\mathrm{CO}}_2$ emissions and population growth, and no directional causality between foreign direct investment inflow and ${\mathrm{CO}}_2$ emissions.

4.5. Results and Discussion

Table 4 shows that trade openness (CTI) positively and significantly contributes to carbon emissions in the SACU both in the short run (statistically significant at the 10% level) and in the long run (statistically significant at the 5% level). In the baseline model (1), the coefficient of CTI is 0.1475776 in the short run, while in the long run, it is 0.4249157. This shows that the CTI significantly increases carbon emissions in the long run as well as in the short run. The contribution (in terms of the magnitude of the coefficient) of the CTI to carbon emissions is higher in the long run, though weakly statistically significant. Also, as shown in Table 5, with the proxy of trade openness being the dependence ratio (DRA), DRA positively and significantly contributes to carbon emissions in the SACU, both in the short run and in the long run, which is statistically significant at the 5% level of significance. The coefficient of DRA is 0.1988687 in the short run, while in the long run, it is 0.5051549. This shows that DRA significantly increases carbon emissions in the long run by a greater coefficient magnitude compared to the short run. Therefore, for both models, this implies that in the SACU region, trade openness is a long-term contributor to poor environmental quality since it leads to an increase in carbon emissions by a higher magnitude. Also from Table 6, the results show that trade openness (X/RGDP) positively and significantly affects carbon emissions (statistically significant at the 10% level).

Regardless of the differences in the coefficients in results from models 1 and 2 in the short run and long run, the results imply that the gains-from-trade hypothesis, which asserts that trade openness can lead to improvement in environmental quality, does not hold in SACU countries. Also, for model 3, the results suggest that the gains-from-trade hypothesis does not hold. This shows that trade is dominated by the importation and exportation of goods, which are unfriendly to the environment. The empirical findings are in line with studies by Wicaksana and Karsinah (2022), Oumarou and Nourou (2024), Mignamissi et al. (2024), and Andriamahery et al. (2022), who found that trade openness contributes to rising ${\mathrm{CO}}_2$ emissions in Sub-Saharan Africa. However, this was contrary to the results obtained by H. Sun et al. (2020), who found that trade openness has a negative impact on carbon emissions in the long run; that is, it decreases environmental pollution in Sub-Saharan Africa. Pham and Nguyen (2024) found no evidence of a statistically significant effect of trade openness on environmental pollution in developing countries.

Also, as shown in Table 4, Table 5 and Table 6, RGDP positively but insignificantly increases ${\mathrm{CO}}_2$ emissions in the SACU, both in the short run and the long run. In Table 4, with the base model 1, where trade openness is proxied by the CTI, the coefficient of RGDP is 0.7232274 in the short run, and in the long run, it is 1.882709. In Table 5 the coefficient of RGDP is 0.7570938 in the short run, and in the long run, it is 2.017252. Therefore, for both models, it can also be concluded that in the SACU region, economic growth does not significantly contribute to poor environmental quality. The findings disagree with Osadume and University (2021), who found that economic growth (RGDP) positively and significantly contributes to carbon emissions in developing countries. Also, Osobajo et al. (2020) found that a significant long-run relationship exists between economic growth and ${\mathrm{CO}}_2$ emissions. In African countries, Boamah et al. (2023) found that per capita GDP increases ${\mathrm{CO}}_2$ emission levels, while trade openness causes a reduction. However, Aye and Edoja (2017) suggested that in developing countries, the effect of economic growth on carbon emissions depends on the business cycle regimes of the countries being investigated, and, in their study, they found that during times of poor growth, a negative effect on ${\mathrm{CO}}_2$ emissions exists, while a positive effect exists during high levels of economic growth.

Also, for other variables, as shown in Table 4, Table 5 and Table 6, FDI does not have a significant influence on carbon emissions in the long run or the short run in the SACU. This was, however, contrary to the findings by Boamah et al. (2023), who revealed that foreign direct investment inflows contribute to the rising and falling levels of ${\mathrm{CO}}_2$ emissions in African countries, implying that the pollution haven and halo hypotheses do hold. Also, Kivyiro and Arminen (2014) and Amoah et al. (2023) found that FDI inflows appear to increase ${\mathrm{CO}}_2$ emissions in Sub-Saharan Africa. PPNGR contributes negatively and significantly (5% level of significance in Table 4) to carbon emissions in the short run and long run. In Table 5, where the DRA is used as a proxy for trade openness, PPNGR is statistically significant at 1% in the long run and 5% in the short run. Also, Table 6 shows that PPNGR is statistically significant at 1%. This is contrary to the findings from Aye and Edoja (2017), who revealed a positive relationship between population and ${\mathrm{CO}}_2$ emissions in the context of developing countries, which include African countries.

As shown in Table 4, the coefficient of ECT (−1) is negative and statistically significant at the 1% level of significance. This suggests that the convergent values are negative and significant, implying that there is a long-run relationship (cointegration) between ${\mathrm{CO}}_2$ emissions and the regressors (FDI, PPNGR, CTI, and RGDP). ECT (−1) further showed the speed of adjustment with which any shock in the short run is restored to equilibrium in the long run. The value of −0.4249929 indicates that the disequilibrium in the short run is adjusted for in the long run at a speed of 42.50%. Also, as shown in Table 5 and Table 6, the coefficient of ECT (−1) is negative and statistically significant, as expected. This means that the convergent values are negative and significant. As illustrated in Table 5, this indicates that there is cointegration between ${\mathrm{CO}}_2$ emissions and the regressors (FDI, PPNGR, DRA, and RGDP). The value of −0.4104834 indicates that the disequilibrium in the short run is adjusted for in the long run at a speed of 41.05%. In Table 6, the results imply that there is a long-run relationship (cointegration) between ${\mathrm{CO}}_2$ emissions and the regressors (FDI, PPNGR, X/RGDP, and RGDP). ECT (−1) further showed the speed of adjustment with which any shock in the short run is restored to equilibrium in the long run. The value of −0.3822528 indicates that the disequilibrium in the short run is adjusted for in the long run at a speed of 38.23%.

As suggested by Bergougui and Zambrano-Monserrate (2025), Dumitrescu–Hurlin Granger non-causality tests were conducted to improve the predictive power and the robustness of the findings, especially since the study relates to energy and environmental policy. The results from Table 7 show that bidirectional causality exists between ${\mathrm{CO}}_2$ emissions and trade openness (CTI as well as DRA), results that are similar to previous studies in Saharan African countries, such as Yılmaz (2023) and Zhou et al. (2025) on emerging economies. Furthermore, unidirectional causality between ${\mathrm{CO}}_2$ emissions and the exports-to-real-gross-domestic-product ratio and carbon emissions exists, from ${\mathrm{CO}}_2$ emissions to the exports-to-real-gross-domestic-product ratio. This can be attributed to the fact that SACU countries may be using high-emission energy-intensive processes to produce goods that are export-oriented. Also the results revealed bidirectional causality between carbon emissions and economic growth, as well as bidirectional causality between carbon emissions and population growth, which is in line with a study by Osobajo et al. (2020), and no directional causality between foreign direct investment and carbon emissions, which is contrary to Kivyiro and Arminen (2014), who found one-way causality from FDI to ${\mathrm{CO}}_2$ emissions in Saharan African countries.

5. Conclusions and Policy Recommendations

The primary objective of this study was to investigate the short-run and long-run relationships between trade openness and ${\mathrm{CO}}_2$ emissions, along with other key regressors, including population growth, foreign direct investment inflow, and economic growth, in five SACU member countries from 1985 to 2023. The results of CS-ARDL estimation reveal that trade openness (CTI and DRA) has a positive and statistically significant relationship with ${\mathrm{CO}}_2$ emissions in both the long run and the short run. When trade openness is proxied by the ratio of exports to real gross domestic product (X/RGDP), the results show a statistically significant long-run relationship. FDI and RGDP are statistically insignificant in both the long run and short run, while PPNGR is statistically significant in the long run and short run. The key findings that can be drawn from this study are as follows:

• The positive and significant long-run and short-run links between trade openness and carbon emissions in the SACU region invalidate the gains-from-trade hypothesis by providing evidence of the important role played by trade openness in the region.
• Foreign direct investment inflow in the SACU region does not significantly influence carbon emissions in either the short run or long run, and this invalidates the pollution haven hypotheses, while population growth has a significant influence in the short run and long run.
• There are positive but insignificant short-run and long-run relationships between economic growth and environmental pollution in the SACU region.
• The Dumitrescu–Hurlin Granger non-causality test shows two-way causality between carbon emissions and trade openness, bidirectional causality between carbon emissions and economic growth, two-way causality between carbon emissions and population growth, unidirectional causality between the exports-to-real-gross-domestic-product ratio and carbon emissions, and no directional causality between foreign direct investment and carbon emissions.

The invalidation of the gains-from-trade hypothesis in the short and long run, which asserts that trade openness can lead to better environmental equality in the SACU region, has policy implications. The main objective of the SACU is to enhance free trade among its member countries, but in its strategic plan for 2022–2027, the SACU admits that it is deeply concerned about environmental quality and stresses the need to move away from ecologically destructive economic activity to eco-friendly economic activities. The strategic plan suggests that significant advances should be made in new industrial production techniques, commodity demand, and trade, along with changes in investments. The findings of this study suggest the need to readjust trade policies and regulations across the SACU. The strategic plan, which aims to intensify the industrial base of the bloc and stimulate sectoral complementarities in production and encourage industries to diversify across the bloc, should be implemented with great emphasis on the environmental implications. Furthermore, the SACU’s strategic plan for 2022–2027 outlines the integrated approaches and practical initiatives to encourage industrialisation, while at the same time positioning the bloc to fully capitalise on opportunities afforded by the African Continental Free Trade Area (AfCFTA), which would enhance trade integration and openness among several countries in Africa. However, as shown by the findings, trade openness is one of the main factors that increase CO₂ emissions; SACU countries need to intensify the trade of environmentally friendly goods.

Also, this two-way causality between trade openness and carbon emissions further enhances the complexity of the relationship. However, it may imply that in the SACU region, with growth in the rate at which the environment is being polluted, owing to the mining and agriculture sectors’ activities, countries may impose restrictions on trade in goods that are unfriendly to the environment. This encourages companies to embrace green technology and trade in environmentally friendly products. However, companies may be reluctant, as in most SACU countries, the main source of electricity is coal; therefore, companies may be reluctant to adopt renewable energy sources. Consequently, SACU countries may actively engage in efforts to combat climate change and embrace trade in environmentally friendly products.

The results show positive and insignificant short-run and long-run relationships between economic growth and environmental pollution in the SACU region. This implies that growth in the region is not linked to detrimental effects on the environment. However, this is not surprising because the majority of SACU countries have poor adoption strategies for industrialisation while acknowledging the complex nature of the link between economic growth and carbon emissions in the region. Furthermore, the finding of two-way causality between economic growth and carbon emissions implies that, in the SACU, there is a need for a balanced and holistic approach that acknowledges the extent to which economic growth and environmental factors are related. The study further revealed that foreign direct investment inflow in the SACU region does not strongly and significantly influence carbon emissions in the long run or short run. This implies that FDI perhaps does not significantly influence carbon emissions, because the SACU region’s environmental policies may be yielding improved carbon emission outcomes.

Also, the results show that population growth negatively and significantly influences carbon emissions in both the long run and short run. This may imply that the impact of population growth in the SACU on CO₂ emissions may have been too small to lead to an increase in CO₂ emissions compared with other regions; therefore, on average, an individual contributes a negligible amount of carbon, even with population growth. However, the two-way causality suggests that population growth in the SACU may both stimulate and reduce carbon emissions, while carbon emissions can, in turn, affect the population growth rates.

Overall, it is suggested in this study that policymakers in the SACU adopt cap-and-trade arrangements to ensure a clean environment and atmosphere. Since cap-and-trade arrangements make use of emissions trading, firms may act responsibly because they are subject to their total emission quota. Also, SACU governments may encourage the trade of eco-friendly goods, which is more likely to encourage green innovation at a lower total cost.

Despite the contribution of this study to the literature, it has limitations, which may be addressed in future studies. These limitations include, firstly, that the data is from 1985 to 2023, and given the timeframe, the data unavailability in all SACU countries presented considerable constraints. This limited our ability to conduct in-depth and updated research since trade and environmental policies continue to change in the region; therefore, updated data may enhance continuous policy checks and balances in the region. Secondly, future studies could make use of the newly introduced Green Trade Openness Index as a proxy for trade openness and explore long-term and short-term sustainability effects within SACU countries, as well as the future of carbon markets in SACU countries. Also, instead of using carbon emissions, which is an aggregate measure of pollutant sources, such as industrial waste generation, industrial processes, and fugitive emissions, disaggregating the emissions may be of interest for future studies. This is because green trade and cap-and-trade systems (carbon markets) may be industry-specific, but are globally acknowledged as crucial tools to reduce carbon emissions.