Does protectionism improve environment of developing countries? A perspective of environmental efficiency assessment

Abstract

Past studies related to embodied pollutant accounting reported that free trade has increased the environmental pollution of developing economies, because the developed countries “outsource” their pollutants to developing nations. The COVID-19 pandemic has stimulated the rise of the most serious protectionism after World War II. This study is aimed to discuss whether protectionism improve the environment in developing countries by developing a comprehensive evaluation model, which integrates multi-regional input-output (MRIO), data envelopment analysis (DEA), and scenario analysis. We revealed the role of protectionism from two perspectives: the single impact on pollutant emissions and the comprehensive impact on environmental efficiency. Specifically, the capital inputs, labor inputs, energy consumption, economic output, carbon dioxide, sulfur dioxide and nitrogen oxides emissions related to global trade activities were simulated based on the MRIO. And then, sector-level trade environmental efficiency was computed by intergrading the MRIO and DEA using a non-radial directional distance function. Finally, the environmental efficiency of both developing and developed countries under two scenarios with and without trade were estimated. The results confirmed that trade has increased the CO₂, SO₂ and NO_X emissions of developing economies by 12.9%, 9.8% and 12.3%, and has reduced that of developed economies by 6.0%, 29.4% and 21.2%, respectively. However, the results also uncovered that the environmental efficiency of developing and developed economies was dropped by 3% and 5%, respectively, under no-trade scenario. We contend that protectionism is not conducive to the sustainable development of developing countries because it lowers their environmental efficiency, although it may reduce their territorial pollutant emissions. For developed countries, the single impact of protectionism on pollutant emission reduction and the comprehensive impact on environmental efficiency are both negative.

Graphical abstract

1. Introduction

Trade's impact on global environment has aroused widespread concern from scholars (Copeland and Taylor, 2004; Wang and Zhang, 2021). In 2019, global trades contributed one fifth of world GDP (World Bank, 2021). However, trade-related carbon emissions accounted for more than one quarter of global carbon emissions ( Zhang et al., 2017a). A flood of literature calculated the transfer of CO₂, SO₂, PM2.5, and heavy metal emissions embodied in global trade network, and pointed out that trades led to serious regional inequity of pollution (Kuik and Gerlagh, 2003; Wang et al., 2017). Because of developing countries’ lower environmental costs, some developed countries outsourced their resource-intensive and pollution-intensive production tasks to developing countries to meet their own territorial emission reduction goals. Therefore, it is generally believed that free trade greatly increased the territorial emissions of developing countries.

However, with a series of events, such as the trade war between China and America (Council., 2018; Representative, 2018), Brexit, regional tariffs and non-tariff trade barriers (Voituriez and Wang, 2015), and America's export controls (WTO, 2018), the consensus on trade globalization has gradually been broken in the past three years. In 2020, the COVID-19 pandemic has spread rapidly around the world, and many countries and cities went into lockdown to curb the spread of the disease (Long et al., 2021; Wang et al., 2022). To protect their own markets, some countries have taken the chance to impose excessive trade restrictions, stimulating the rise of the most serious protectionism after World War II. Cross-border flows of products and factors have been greatly disrupted. In this context, what will happen to the environmental pressures that have been induced by free trade in developing economies? Will protectionism improve their environments?

Our contribution to existing research is in two aspects. First, this study answers the practical question of whether trade protectionism is beneficial to the improvement of environment in developing and developed countries. Second, we propose a comprehensive index of trade environmental efficiency, which can reflect the comprehensive environmental impact of the factors’ flows induced by trade. Specifically, this study established a normal trade scenario and a no-trade scenario. Based on the environment-extended multi-regional input-output model, two lists of economy, society and environment variables under normal trade scenario and no-trade scenario for 43 major economies from 2000 to 2014 are created. Capital inputs, labor inputs, energy consumption, economic output, CO₂, SO₂ and NO_X emissions are included. Then these factors are incorporated into the DEA model to obtain the comprehensive environmental efficiency of each country. We discussed the single impact of trade on the direct emissions of CO₂, SO₂ and NO_X in various countries, as well as the comprehensive impact on countries’ environmental efficiency. The impact mechanism is investigated from the perspective of trade patterns and sectors.

The remainder of this study is as follows. Section 2 reviews the existing literature and proposes innovative points of this paper. Section 3 introduces the modeling process and data collection. Section 4 analyzes the impact of trade protection on countries’ direct emissions and environmental efficiency from the trade pattern-level and sector-level. Section 5 summarizes and discusses the main findings. Section 6 summarizes the whole paper and lists some limitations.

2. Literature review

A large number of studies have assessed the environmental impacts of global trade. Input-output analysis is a quantitative economic analysis method based on the interdependence between different economic sectors or industries. Trade can cause the geographical separation of production and consumption (Li et al., 2022; Proops et al., 1999). Since the 1970s, scholars have discovered the huge impact of international trade on environmental damage based on this model (Bullard and Herendeen, 1975; Casler and Wilbur, 1984). Recently, the input-output models are widely used to track the environmental footprints (Tian et al., 2018; Wang and Liu, 2021; Wiedmann, 2009) and social footprints (Simas et al., 2014; Xiao et al., 2017) induced by the economic activities on global supply chains (Long et al., 2018; Wang et al., 2022). Wiedmann et al., reviewed recent studies of global trade footprints based on global multi-regional input-output (GMRIO) modeling methods. They pointed out that with the help of the GMRIO model, it is possible to establish a global supply chain relationship between the underdeveloped commodity production areas that are under environmental pressure and the economically affluent commodity consumption areas, and the environmental and social footprint of international trade can be traced (Wiedmann and Lenzen, 2018). As developing countries have played an increasingly important role in global trade in recent years, the leakage of pollutant emissions flowing to them is also becoming more intense (López et al., 2013; Wang and Han, 2021). As shown in Table 1 , it is a consensus that global trade has increased developing countries’ pollutant emissions. In addition, it is easy to see that since the release of input-output tables often has a certain lag (Long et al., 2019), most of the research intervals are updated to 2014. However, a country's industrial structure and technological level will not change significantly in a short period of time, and the input-output tables are highly accurate and comprehensive in describing the operation of the socioeconomic system. Therefore, this model has been widely used in trade-related research for a long time, and its research results still have good reference value.

Table 1.

Some literature on the phenomenon of pollution heaven based on input-output model.

Objects	Pollutant	Study period	Main findings	Indicator
China (Hamilton and Kelly, 2017)	CO2	1995–2009	China's exports pollute more than those of its trading partners, and the process of industrialization has made China's exports more polluting	The ratio of the average pollution of unit value-added in exports divided by that of unit value-added in imports
China (Hui et al., 2017)	Mercury emissions	2010	6%, 10% and 10% of mercury emissions in China's air, water and land are caused by the final consumption of United States	Embodied mercury Flows
India (Wang et al., 2018)	CO2	2000–2014	India has become a major importer of embodied carbon emissions from developed countries.	Embodied carbon  transfers
BRICS group (Zhang et al., 2019)	Energy, CO2, value added	2009	BRICS countries are net exporters of embodied energy, carbon emissions, value added,	Embodied flow transfers
Countries along Belt and Road Routes (Cai et al., 2018)	CO2	2013	China has become the pollution haven for 22 developed countries, and 19 developing countries have become the pollution haven for China.	The net embodied emissions exports
Trade between China and Countries along Belt and Road Routes (Li and Liu, 2020)	CO, SO2, NMVOC, NOX, NH3, PM10	2010–2015	China phased out its pollution-intensive industries to the countries along Belt and Road Routes	The net embodied emissions exports
Global 7 regions (López et al., 2018)	CO2	1995–2009	Global trade could reduce 1Gt CO2 emissions every year. China is a pollution haven of developed countries and reached its peak in 2008.	The difference between the embodied emissions exports and the avoided emissions by imports
Global 113 countries (Davis and Caldeira, 2010)	CO2	2004	China and other emerging markets are the main sources of carbon imports for developed countries.	Consumption-based CO2 emissions flows
Global 41 countries (Shuijun Peng, Wencheng Zhang, and Chuanwang  Sun, 2016)	CO2, CH4, N2O, NOX, SOX, CO, NMVOC, NH3	1995–2009	14–30 percent of emissions in developing economies are caused by exports to developed economies.	Consumption-based emissions transfers
Global 41 countries (Wang et al., 2018; Zhang et al., 2017a; b)	CO2	1995–2009	Trade helps reduce global carbon emissions, and China is the largest net carbon exporter.	The difference between the embodied emissions exports and the avoided emissions by imports
Global 129 regions (Meng et al., 2018)	CO2	2004–2011	Global energy-intensive production and processing tasks are shifting from China and India to other developing countries	Embodied carbon transfers
Global 188 countries (Oita et al., 2016)	Reactive nitrogen	2010	The agriculture, food, and textile sectors of developing countries are net exporters of embodied nitrogen pollution, and developed countries are mostly net importers.	Nitrogen footprints
Global 41 countries (Liang et al., 2015)	Atmospheric mercury emissions	2005, 2010	Developed countries should be partially responsible for mercury emissions from developing countries	Trade balance of embodied atmospheric mercury.
Global 186 countries (Li et al., 2017)	Mercury emission related to nonferrous metal	2010	Mercury emissions are exported from developing economies such as China and Colombia to developed economies such as the United States and Germany.	Embodied mercury emission transfer

Recently, due to the influence of international trade protectionism, some studies have begun to focus on the impact of trade restrictions on the environment. Lin et al., simulated different levels of global trade restriction measures and found that trade restrictions could reduce global CO₂ emissions by 6.3% (Lin et al., 2019). Kahsay et al., established the scenario of abolishing tariffs and trade barriers based on the CGE model, and found that free trade contributed to the economic growth of the Nile River Basin (Kahsay et al., 2018). Du et al., pointed out that the Sino-US trade war cannot achieve a win-win situation for the economy and the environment between China and U.S., and therefore trade restrictions are not the best ways of environmental protection (Du et al., 2020). Lu et al., studied the environmental impact of the Sino-US trade war under different tariff levels. It is found that the Sino-US trade war has led to the land use changes and emissions increase in some developing countries (Lu et al., 2020). Liu et al., found that trade friction could only reduce the emissions of China and U.S. in the short term, and reduce emissions on a global scale in the long term, but it was not conducive to avoiding climate disasters (Liu et al., 2020). In general, the existing research on the environmental impact of trade restrictions also focuses on the discharge of several pollutants (Wang and Wang, 2021). The reallocation of other resources and technologies caused by trade has been less studied. There is currently no comprehensive indicator to measure it, which can systematically reflect the characteristics of the global supply chain (Wiedmann and Lenzen, 2018; Yang et al., 2020).

In the above context, we propose a comprehensive index of trade environmental efficiency, which is used to measure the environmental performance with trade-related economic-social-environmental input-output factors (Li et al., 2021). It can comprehensively reflect the environmental changes caused by the flows of multiple factors caused by trade for a certain period of time. The multi-objective programming method can achieve the integration of all production factors (Brandenburg et al., 2014), in which the DEA method is often used to model energy-environmental systems and measure their sustainability (Faere et al., 1989; Sueyoshi et al., 2017). The main advantage of DEA compared with other methods is that it does not need to assume a production function and estimate related parameters. Based on the original input and output data, relatively objective and practical evaluation results can be obtained (Cook and Seiford, 2009). DEA provides important support for decision analysis and performance evaluation in various fields. It is especially suitable for the environmental efficiency research of multinational supply chains (Acquaye et al., 2017; Chen and Yan, 2011; Yan et al., 2020). However, no research has revealed the impact of trade protection measures on environmental efficiency. Only studying the impact of trade restrictions on a single pollutant can hardly give policy makers a more systematic conclusion.

In order to study the impact of trade protection on environmental efficiency, this study will introduce scenario analysis methods and set up controlled experiments. Some studies compared environmental factors under trade and no-trade scenarios, and discussed environmental changes induced by global trade. There are two ways to stimulate a no-trade scenario with the input-output model. The first method is to cancel exports and directly add the total embodied emissions imports back to domestic emissions, that is, the assumption of foreign technology is adopted. For example, Xu et al., calculated a number of sustainable development indicators under the no-trade scenario (Xu et al., 2020). The second method is to cancel exports, and use domestic technology to produce the original imported products to meet the demand, that is, domestic technology hypothesis. For example, Ackerman et al., studied the impact of Japan-US trade on carbon emissions, and found that Japan’ CO₂ emissions were increased, while US's were reduced under no-trade scenario (Ackerman et al., 2007). Zhang et al., studied the impact of specific trade patterns on emissions under two scenarios and found that the environmental impact of final goods trade gradually weakened (Zhang et al., 2017a). The first approach assumes that the technology level of the importing country is exactly the same as that of the country itself. This method assumes that once global trade is canceled, the imported goods demanded by a country are still produced using the technical conditions of the original producing country. The object of this study is global trade, and the technological differences between trading countries are often large, so this assumption cannot be adopted. The second approach is that, in a no-trade scenario, each country should be self-sufficient, using its own technology to meet the needs of its citizens. Once trade is canceled, then the flow of such technological elements should also stop. This paper adopts the second scenario hypothesis, which is more consistent with the concept of no trade scenario.

In summary, in the context of the COVID-19 pandemic that has made global trade protectionism increasingly rampant, we innovatively reveal the impact of trade protection on the environment of countries from the perspective of direct emissions and comprehensive environmental efficiency. Based on the existing literature, this study integrates the economic, social and environmental factors embodied in trade, and proposes a comprehensive index of trade environmental efficiency to measure the impact of trade protection.

3. Methodology and data

3.1. Trade-related variables calculation: MRIO model

The input-output model is suitable for analyzing the environmental impact caused by human activities in a complex economic system. It can characterize the quantitative dependence between production and consumption, and can reflect the direct or indirect economic links between various sectors of the world (Leontief, 1975). It has developed single-region and multi-regional input-output analysis, the latter of which can distinguish the source of imports in detail (Long and Yoshida, 2018). The environment-extended MRIO model can highlight the links between economic activities and environmental impacts, and is widely used to track the environmental footprints of global trade.

Trade follows traditional production theory. Various resources and services are consumed to create economic output (Färe and Primont, 2012). In this paper, we characterize the environmental production technology related to trade: capital, labor and energy are defined as inputs; value added as desirable output; pollutants as undesirable outputs. CO₂ is the major source of global warming, and SO₂, NOx are the main atmospheric pollutants, leading to acid rain and photochemical smog. Therefore, we selected the three pollutants to indicate environmental impact (J. Du et al., 2017; Pasurka, 2006; Yang et al., 2020).

We adopt the MRIO model to establish a list of trade-related variables under trade scenario and no-trade scenario. The difference between the emissions of the three trade-related pollutants under the two scenarios is the direct impact of trade protection. Suppose the world consists of N economies and each economy consists of S industries:

$$\begin{array}{c}{\mathrm{X}}_1={\mathrm{Z}}_{11}+{\mathrm{Z}}_{12}+\cdots +{\mathrm{Z}}_{1N}+{\mathrm{Y}}_1\\ {\mathrm{X}}_2={\mathrm{Z}}_{21}+{\mathrm{Z}}_{22}+\cdots +{\mathrm{Z}}_{2N}+{\mathrm{Y}}_2\\ ⋮\\ {\mathrm{X}}_N={\mathrm{Z}}_{N1}+{\mathrm{Z}}_{N2}+\cdots +{\mathrm{Z}}_{NN}+{\mathrm{Y}}_N\end{array}$$ (1)

where ${\mathrm{X}}_i$, ${\mathrm{Y}}_i$, ${\mathrm{Z}}_{ij}$ represent the total output matrix (S × 1) of economy i, the final demand matrix (S × 1) of economy i, and the intermediate input matrix from economy i to economy j (S × S), respectively. They are all economic variables, and the unit is dollars.

The technical coefficient, direct consumption coefficient ${\mathrm{A}}_{ij}$ (S × S), indicates the economy i's inputs consumed directly during economy j's one-unit total output production.

$${\mathrm{A}}_{ij}=\left[a_{ij}^{pq}\right]$$ (2)

$$a_{ij}^{pq}=\frac{z_{ij}^{pq}}{X_j^q}$$ (3)

Then, Eq. (1) can be rearranged as

$$\left[\begin{array}{c}{X}_1\\ X_2\\ ⋮\\ X_N\end{array}\right]=\left[\begin{array}{c}{A}_{11}{A}_{12}\dots A_{1N}\\ A_{21}{A}_{22}\dots A_{2N}\\ ⋮⋮\ddots ⋮\\ A_{N1}{A}_{N2}\dots A_{NN}\end{array}\right]\left[\begin{array}{c}{X}_1\\ X_2\\ ⋮\\ X_N\end{array}\right]+\left[\begin{array}{c}{Y}_1\\ Y_2\\ ⋮\\ Y_N\end{array}\right]$$ (4)

$$\begin{array}{ccc}\hfill \left[\begin{array}{c}{X}_1\\ X_2\\ ⋮\\ X_N\end{array}\right]& =& {\left[\begin{array}{cccc}I-A_{11}\hfill & -A_{12}& \cdots & -A_{1N}\\ -A_{21}\hfill & I-A_{22}& \cdots & -A_{2N}\\ ⋮\hfill & ⋮& \ddots & ⋮\\ -A_{N1}\hfill & -A_{N2}& \cdots & I-A_{NN}\end{array}\right]}^{-1}\hfill \\ \hfill \left[\begin{array}{c}{Y}_1\\ Y_2\\ ⋮\\ Y_N\end{array}\right]& =& \left[\begin{array}{c}{L}_{11}{L}_{12}\dots L_{1N}\\ L_{21}{L}_{22}\dots L_{2N}\\ ⋮⋮\ddots ⋮\\ L_{N1}{L}_{N2}\dots L_{NN}\end{array}\right]\left[\begin{array}{c}{Y}_1\\ Y_2\\ ⋮\\ Y_N\end{array}\right]\hfill \end{array}$$ (5)

where ${\mathrm{L}}_{ij}$, Leontief inverse matrix (S × S), represents the economy i's total inputs consumed during economy j's one-unit final output production. This coefficient links the final consumption and its induced direct and indirect production.

$$X_i=A_{ii}{X}_i+Y_{ii}+\sum _{i\ne j}{T}_{ij}$$ (6)

$$X_i=B_{ii}\left(Y_{ii}+\sum _{i\ne j}{T}_{ij}\right)$$ (7)

In Eq. (6), the production-based output of economy i is used to satisfy domestic demand ($A_{ii}{X}_i+Y_{ii}$) and foreign demand ($\sum _{i\ne j}{T}_{ij}$). It can be rearranged as Eq. (7), where $B_{ii}={\left(I-L_{ii}\right)}^{-1}$ represents domestic Leontief inverse matrix (S × S), $L_{ii}=B_{ii}+B_{ii}\sum _{i\ne j}^n{A}_{it}{L}_{ti}$ (Wang et al., 2013).

The total output of economy i can be devided into four parts: $D_i$ denotes the output of domestic economic activity; $E{f}_i$ denotes trade of final product exports; $E{i}_i$ denotes trade of last-stage intermediate product exports; and $E{g}_i$ denotes global value chain (GVC) related product exports Eq. (8-(12)). The last three parts are related to global production fragmentation, and correspond to three trade patterns.

$$X_i=D_i+E{f}_i+E{i}_i+E{g}_i$$ (8)

$$D_i=B_{ii}{Y}_{ii}$$ (9)

$$E{f}_i=B_{ii}\sum _{j\ne i}{Y}_{ij}$$ (10)

$$E{i}_i=B_{ii}\sum _{j\ne i}{A}_{ij}{B}_{jj}{Y}_{jj}$$ (11)

$$\begin{array}{ccc}\hfill E{g}_i& =& B_{\mathit{ii}}\sum _{j\ne i}\left(A_{\mathit{ij}}{B}_{\mathit{jj}}\sum _{t\ne j}{A}_{\mathit{jt}}{L}_{\mathit{tj}}{Y}_{\mathit{jj}}\right)+B_{\mathit{ii}}\sum _{j\ne i}\left(A_{\mathit{ij}}\sum _{t\ne j}{L}_{\mathit{jt}}{Y}_{\mathit{tj}}\right)\hfill \\ & & +B_{\mathit{ii}}\sum _{j\ne i}\left(A_{\mathit{ij}}\sum _t{L}_{\mathit{jt}}\sum _{u\ne j}{Y}_{\mathit{tu}}\right)\hfill \end{array}$$ (12)

As an example, we adopt the Eq. (8) to estimate CO₂ emission embodied in economy i's domestic production activity and international export trade. The carbon intensity matrix is introduced, $F{c}_i=\left[f{c}_i^p\right]$, which measures carbon emission per unit of output in each sector and reflect its technical level.

$$F{c}_i=\frac{C_i}{X_i}$$ (13)

As shown in Fig. 1 , the carbon emissions related to domestic activity and three trade patterns can be quantified with carbon intensity $F{c}_i$. The notation rules we created are as follows: The first character is the factor type, namely capital (K), labor (L), energy consumption (E), value added (V), CO₂ emissions (C), SO₂ emissions (S) , NO_X emissions (N). The second character is demand type that causes factor input or output, that is, domestic demand (D), trade-related demand. And the latter refers to exports in trade scenario (E), and avoided imports in no-trade scenario (I). The third character is the specific link of trades, namely the final product trade (f), the last-stage intermediate product trade (i), and the GVC-related product trade (g). In addition, the subscript i represents the i th economy, i = 1, …, N. For example, the carbon emissions of economy i in trade scenario consist of two parts: emissions induced by domestic economic activities ($C\_D_i$) and emissions induced by exports ($C\_E_i$). $C\_E_i$ is further divided into $C\_E{f}_i$, $C\_E{i}_i$and $C\_E{g}_i$, corresponding to three types of trade links. The environmental variables related to CO₂, SO₂ and NO_X used in this study are all in Gg (10⁹ g).

Fig. 1.

The carbon emission calculating under trade and no-trade scenarios.

The model calculate the factor consumption in the no-trade scenario by changing the intensity (Zhang et al., 2017a). This is because the resource consumption intensity or pollution emission intensity is multiplied by the economic variables represented by the value-based MRIO model to obtain the embodied factor links. In the absence of trade, the original trade-related production tasks will occur at the place of demand. Therefore, if the global trade's effect is not considered, economies will be banned from exports; instead they have to invert import demand into domestic demand, which means using domestic technology to produce imports to meet this demand. And the output under no-trade scenario is formulated as:

$${X_i}^{\prime }=D_i+I{f}_i+I{i}_i+I{g}_i$$ (14)

$$D_i=B_{ii}{Y}_{ii}$$ (15)

$$I{f}_i=B_{ii}\sum _{j\ne i}{Y}_{ji}$$ (16)

$$\begin{array}{c}I{i}_i=B_{ii}\sum _{j\ne i}{A}_{ji}{B}_{ii}{Y}_{ii}\hfill \\ I{g}_i=B_{ii}\sum _{j\ne i}\left(A_{ji}{B}_{ii}\sum _{t\ne i}{A}_{it}{L}_{ti}{Y}_{ii}\right)\hfill \end{array}$$ (17)

$$\begin{array}{c}+B_{ii}\sum _{j\ne i}\left(A_{ji}\sum _{t\ne i}{L}_{it}{Y}_{ti}\right)\hfill \\ +B_{ii}\sum _{j\ne i}\left(A_{ji}\sum _t{L}_{it}\sum _{u\ne i}{Y}_{tu}\right)\hfill \end{array}$$ (18)

As shown in Eq. (15), domestic production activity's output under no-trade situation $D_i$ is the same with trade situation. $I{f}_i$ denotes the avoided output by final product imports (Eq. (16)); $I{i}_i$ denotes the avoided output by last-stage intermediate product imports; and $I{g}_i$ denotes the avoided output by GVC-related product imports.

Then we obtain the carbon emssions embodied in import-avoided output with domestic carbon intensity $F{c}_i$. Similarly to normal trade scenario, $F{c}_i·{X_i}^{\prime }$ is divided into $C\_D_i$ and $C\_I_i$. And $C\_I_i$is further divided into $C\_I{f}_i$, $C\_I{i}_i$ and $C\_I{g}_i$. Fig. 1 gives formula details.

This calculation process can be presented in the "MRIO" part of Fig. 2 , with all input factors' intensity ($F_K$ , $F_L$ , $F_E$) and output factors' intensity ($F_V$ , $F_C$ , $F_S$ , $F_N$), all factors’ inputs and outputs are obtained.

Fig. 2.

Modling framework of environmental performance assessment.

3.2. Environment efficiency assessment: DEA model

DEA is a non-parametric method. It has no assumptions about the relationship between variables and reduces the error caused by subjective intervention. And the efficiency evaluation of multiple input-output systems can be realized without being affected by the dimensionality. Its core is to minimize input variables while maximizing output variables. However, after considering the negative environmental externalities of production, some undesired outputs generated during the production process, such as waste gas, wastewater, need to be introduced into the performance evaluation of the decision-making unit (DMU).

Sustainability performance evaluation on country-level ignores the effect of different industrial structures across countries. Considering technology heterogeneity, we model the production process on industry-level. Take the production related to exports of economy i's sector p under trade scenario in the year t as an example: capital ($K\_E_{i,p}^t$), labor ($L\_E_{i,p}^t$), energy ($E\_E_{i,p}^t$) are used as inputs, to produce the desirable outputs, value added ($V\_E_{i,p}^t$), and undesirable outputs, CO₂ ($C\_E_{i,p}^t$), SO₂ ($S\_E_{i,p}^t$), NO_X ($N\_E_{i,p}^t$), shown in ‘DEA’ of Fig. 2. Therefore, the DEA model is suitable for the multi-input and multi-output system described in this research. The result obtained is only the relative efficiency of the DMU, not the absolute efficiency.

Under the normal trade scenario, the export-related production process of economy i's sector p can be described as:

$$\begin{array}{c}{T}_{i,p}^t=\{\left(K\_E_{i,p}^t,L\_E_{i,p}^t,E\_E_{i,p}^t,V\_E_{i,p}^t,C\_E_{i,p}^t,S\_E_{i,p}^t,N\_E_{i,p}^t\right):\hfill \\ \left(K\_E_{i,p}^t,L\_E_{i,p}^t,E\_E_{i,p}^t\right)canproduce\left(V\_E_{i,p}^t,C\_E_{i,p}^t,S\_E_{i,p}^t,N\_E_{i,p}^t\right)\}\hfill \end{array}$$ (19)

Realistically, the production processes that produce both desirable and undesirable outputs exhibit two features. (1) Null-jointness: only when production is stopped, can pollutant emissions be eliminated. (2) Weak disposability: the reduction of pollutants is not free (Faere et al., 1989):

• (i)If $\left(K\_E_{i,p}^t,L\_E_{i,p}^t,E\_E_{i,p}^t,V\_E_{i,p}^t,C\_E_{i,p}^t,S\_E_{i,p}^t,N\_E_{i,p}^t\right)$ ∈$T_{i,p}^t$, and $\left(C\_E_{i,p}^t,S\_E_{i,p}^t,N\_E_{i,p}^t\right)$ = 0, then $V\_E_{i,p}^t=0.$
• (ii)If $\left(K\_E_{i,p}^t,L\_E_{i,p}^t,E\_E_{i,p}^t,V\_E_{i,p}^t,C\_E_{i,p}^t,S\_E_{i,p}^t,N\_E_{i,p}^t\right)$∈$T_{i,p}^t$, and 0≤θ≤1, then ($K\_E_{i,p}^t,L\_E_{i,p}^t,E\_E_{i,p}^t,\theta V\_E_{i,p}^t,\theta C\_E_{i,p}^t,\theta S\_E_{i,p}^t,\theta N\_E_{i,p}^t)$∈$T_i$.

With the assumptions above, $T_p^t$ is described as environmental production technology of sector p (Färe et al., 2004). After defining conceptually, we need to formulate $T_p^t$ in the empirical study. The DEA method exhibiting constant returns to scale is widely adopted in environmental performance evaluation. Therefore, we formulate $T^p$ within the non-parametric frontier linear framework:

$$\begin{array}{c}{T}_p^t=\left(K\_E_p^t,L\_E_p^t,E\_E_p^t,V\_E_p^t,C\_E_p^t,S\_E_p^t,N\_E_p^t\right)\hfill \\ \mathrm{s}.\mathrm{t}.\sum _{t=1}^M\sum _{j=1}^N{\lambda }_j^t*K\_E_{j,p}^t\le K\_E_{j,p}^t\hfill \\ \sum _{t=0}^M\sum _{j=1}^N{\lambda }_j^t*L\_E_{j,p}^t\le L\_E_{j,p}^t\hfill \\ \sum _{t=0}^M\sum _{j=1}^N{\lambda }_j^t*V\_E_{j,p}^t\ge V\_E_{j,p}^t\hfill \\ \sum _{t=0}^M\sum _{j=1}^N{\lambda }_j^t*C\_E_{j,p}^t=C\_E_{j,p}^t\hfill \\ \sum _{t=0}^M\sum _{j=1}^N{\lambda }_j^t*S\_E_{j,p}^t=S\_E_{j,p}^t\hfill \\ \sum _{t=0}^M\sum _{j=1}^N{\lambda }_j^t*N\_E_{j,p}^t=N\_E_{j,p}^t\hfill \\ {\lambda }_j^t\ge 0,\mathrm{j}=1,\dots ,\mathrm{N},\mathrm{t}=1,\dots ,\mathrm{M}\hfill \end{array}$$ (20)

This study aims to compare the results of environmental efficiency under trade and no-trade scenarios, thus the production activities of the same nature in the same sector under trade and no-trade scenarios are regarded as DMUs in the study period. Taking the final product trade among the manufacturing sectors of countries as an example, we should combine the following two parts as a production set: (1) In the trade scenario, in 2000–2014, the production related to the final product exports in the manufacturing sector of global 15 economies, a total of 240 DMUs; (2) In the no-trade scenario, in 2000–2014, the production related to the avoided final product imports in the manufacturing sector of global 15 economies, a total of 240 DMUs. Note that in the efficiency assessment, the 28 EU countries are merged into one economy.

In Eq.(20), ${\lambda }_j^t$ denotes the weight of DMU. Production possibility set constructed by Eq.(20) gives the ‘efficient frontier’. And the sustainability performance can be quantified through measuring the distance of a DMU from this frontier. Compared with slacked-based measure (Chambers et al., 1996) and directional distance measure (Fukuyama and Weber, 2009), the non-radial directional distance measure can simultaneously explore the inefficiency of various outputs and allow their adjustment with different proportion. Then we establish an undesirable outputs-oriented DEA model :

$$\begin{array}{c}{D}_{i,p}^t\left(K\_E_p^t,L\_E_p^t,E\_E_p^t,V\_E_p^t,C\_E_p^t,S\_E_p^t,N\_E_p^t\right)\hfill \\ =max\left({\beta }_{C,i,p}^t+{\beta }_{S,i,p}^t+{\beta }_{N,i,p}^t\right)/3\hfill \\ \mathrm{s}.\mathrm{t}.\sum _{t=1}^M\sum _{j=1}^N{\lambda }_j^t*K\_E_{j,p}^t\le K\_E_{i,p}^t\hfill \\ \sum _{t=0}^M\sum _{j=1}^N{\lambda }_j^t*L\_E_{j,p}^t\le L\_E_{i,p}^t\hfill \\ \sum _{t=0}^M\sum _{j=1}^N{\lambda }_j^t*E\_E_{j,p}^t\le E\_E_{i,p}^t\hfill \\ \sum _{t=0}^M\sum _{j=1}^N{\lambda }_j^t*V\_E_{j,p}^t\ge V\_E_{i,p}^t\hfill \\ \sum _{t=0}^M\sum _{j=1}^N{\lambda }_j^t*C\_E_{j,p}^t=\left(1-{}_{C,i,p}^t\right)*C\_E_{i,p}^t\hfill \\ \sum _{t=0}^M\sum _{j=1}^N{\lambda }_j^t*S\_E_{j,p}^t=\left(1-{\beta }_{S,i,p}^t\right)*S\_E_{i,p}^t\hfill \\ \sum _{t=0}^M\sum _{j=1}^N{\lambda }_j^t*N\_E_{j,p}^t=\left(1-{\beta }_{N,i,p}^t\right)*N\_E_{i,p}^t\hfill \\ {}_j^t\ge 0,\mathrm{i}=1,\dots ,\mathrm{N},\mathrm{t}=1,\dots ,\mathrm{M}\hfill \end{array}$$ (21)

Where ${\beta }_{C,i,p}^t,{\beta }_{S,i,p}^t$, and ${\beta }_{N,i,p}^t$ denote potential reduced proportion of CO₂, SO₂ and NO_X, respectively. And ${\beta }_{C,i,p}^t,C\_E_{i,p}^t,{\beta }_{S,i,p}^{t}S\_E_{i,p}^t$, and ${\beta }_{N,i,p}^{t}N\_E_{i,p}^t$ is the slacks of the three pollutants. If [${\beta }_{C,i,p}^t$, ${\beta }_{S,i,p}^t$,${\beta }_{N,i,p}^t$]=0, then $D_{i,p}^t\left(K\_E_p^t,L\_E_p^t,E\_E_p^t,V\_E_p^t,C\_E_p^t,S\_E_p^t,N\_E_p^t\right)$=0, which means there is no potential emission reduction and this DMU has achieved the efficient frontier.

Suppose ${\beta }_{C,i,p}^{t*}$, ${\beta }_{S,i,p}^{t*}$ and ${\beta }_{S,i,p}^{t*}$ are optimal values solved by Eq.(21), the environmental performance index of economy i ’s sector p ($EP{I}_i^p$) can be defined as

$$EP{I}_{i,p}^t=1-D_{i,p}^t\left(K\_E_p^t,L\_E_p^t,E\_E_p^t,V\_E_p^t,C\_E_p^t,S\_E_p^t,N\_E_p^t\right)$$ (22)

$$EP{I}_{i,p}^t=1-\left({\beta }_{C,i,p}^{t*}+{\beta }_{S,i,p}^{t*}+{\beta }_{S,i,p}^{t*}\right)/3$$ (23)

Possible values of $EP{I}_{i,p}^t$ falls [0,1]. The larger value means better sustainability performance. And $EP{I}_{i,p}^t=1$ means the best performance.

Considering technology heterogeneity among sectors, in Eqs. (24)-(26), the performance index of CO₂, SO₂ and NO_X emissions related to each country's trade is obtained from the weighted average of sectors’ environmental efficiency. The share of each sector's emissions is the weight. In other words, the weighted average method is applied to convert the sector index to country index.

$${\beta }_{C,i}^{t*}=\sum _{p=1}^S\frac{C\_E_{i,p}^t}{C\_E_i^t}{\beta }_{C,i,p}^{t*}$$ (24)

$${\beta }_{S,i}^{t*}=\sum _{p=1}^S\frac{S\_E_{i,p}^t}{S\_E_i^t}{\beta }_{S,i,p}^{t*}$$ (25)

$${\beta }_{N,i}^{t*}=\sum _{p=1}^S\frac{N\_E_{i,p}^t}{N\_E_i^t}{\beta }_{N,i,p}^{t*}$$ (26)

This study has no preference for any pollutants, thus CO₂, SO₂ and NO_X are considered equal (Yang et al., 2020). As shown in Eq. (27), the performance indexes of the three gasses are averaged to obtain the overall environmental efficiency of a country.

$$EP{I}_i^t=1-\left({\beta }_{C,i}^{t*}+{\beta }_{S,i}^{t*}+{\beta }_{N,i}^{t*}\right)/3$$ (27)

Similarly, the sector-level and country-level environmental performance index ($EP{I}_{i,p}^t$ ’, $EP{I}_i^t$’) under no-trade scenario can be obtained. The indicators measure the environmental performance of trade-related production activities under given economic and social factors. A higher environmental efficiency index means that the current economic output can be achieved under the existing capital input, labor input, and energy input, and there is less slack in pollutant emissions, that is, better environmental performance. A lower environmental efficiency index means that the current economic output can be achieved with less emissions under the existing factor input, and there is more slack in pollutant emissions, resulting in a great potential for emission reduction.

Fig. 2 shows the framework of the modeling process. The first step is "MRIO". Based on the MRIO model, a list of economic, social and environmental factors (K, L, E, V, C, N, S) under trade and no-trade scenarios is calculated. The second step is "DEA". These elements are used as the data set of DEA model, including those related to domestic demand (blue background) and trade demand (green background). The latter is further divided into the final product trade demand, the last-stage intermediate product trade demand, and the global value chain trade demand (yellow background). The DEA model describes the economic system under both trade and no-trade scenarios. It regards factor K, L, E as inputs, and regards V, C, N, S as outputs. The third step is "Environment efficiency". After DEA modeling, the environmental efficiency under the two scenarios can be obtained. The production activities of the same sector in each country under the trade scenario and no-trade scenario are regarded as the group of DMUs. The only difference between the two is the scenario setting. Therefore, the difference can be used to measure the impact of trade.

3.3. Data collection

Three datasets are required in this study: global input-output tables, the social economic data and environment data. We employed the latest multiple-regional input-output (I-O) tables from World Input-Output Database (WIOD), which is consistent with national figures, and has been widely used to study the international trade's effects on society, economic and environment (Timmer et al., 2015). The version released in 2016 provides the 2000–2014 time-series data at current price. This time period is very representative. It covers some large-scale events of changes in the global trade situation. For example, China joined the World Trade Organization in 2001, and since then more and more emerging developing economies have continuously participated in international trade. The 2008 financial crisis swept the world, and the global free trade system suffered major damage. After 2013, the Transatlantic Trade and Investment Partnership Agreement has led to the gradual rise of global bilateral and regional trade rules. The dataset covers 43 countries and the rest of world (ROW). Each country has 56 sectors. To facilitate analysis, we aggregate the results of MRIO model into 15 economies and 7 sectors. These economies participate extensively in international trade, and their environmental pressures are closely related to trade activities. They are typical representatives of developed and developing countries, and the impact of trade protection on them is different. In addition, they are located in Europe, the Americas and the Asia-Pacific region, which helps to measure the impact of trade protection on a global scale. Table A1 in Appendix provides the list and classification of economies. Table A2 provides the number and classification of the sectors. More details please see appendix.

Table A2.

Classification of sectors.

Code	Integrated sector	Sector code in WIOD	Sector in EROA
S1	Agriculture	Crop and animal production, hunting and related service activities	Agriculture
		Forestry and logging
		Fishing and aquaculture	Fishing
S2	Mining	Mining and quarrying	Mining and Quarrying
S3	Manufacturing	Manufacture of food products, beverages and tobacco products	Food & Beverages
		Manufacture of textiles, wearing apparel and leather products	Textiles and Wearing Apparel
		Manufacture of wood and of products of wood and cork, except furniture; manufacture of articles of straw and plaiting materials	Wood and Paper
		Manufacture of paper and paper products
		Printing and reproduction of recorded media
		Manufacture of coke and refined petroleum products	Petroleum, Chemical and Non-Metallic Mineral Products
		Manufacture of chemicals and chemical products
		Manufacture of basic pharmaceutical products and pharmaceutical preparations
		Manufacture of rubber and plastic products
		Manufacture of other non-metallic mineral products
		Manufacture of basic metals	Metal Products
		Manufacture of fabricated metal products, except machinery and equipment
		Manufacture of computer, electronic and optical products	Electrical and Machinery
		Manufacture of electrical equipment
		Manufacture of machinery and equipment n.e.c.
		Manufacture of motor vehicles, trailers and semi-trailers	Transport Equipment
		Manufacture of other transport equipment
		Manufacture of furniture; other manufacturing	Other Manufacturing
		Repair and installation of machinery and equipment	Recycling
		Publishing activities	Other
S4	Electricity, Gas and Water Supply	Electricity, gas, steam and air conditioning supply	Electricity, Gas and Water
		Water collection, treatment and supply
S5	Construction	Construction	Construction
(continued on next page)

To calculate the capital stock input, labor input and value added output, we adopted the social economic accounts from WIOD. The nominal capital stock (in millions of national currency) is used as capital input; the number of persons engaged (thousands) is used as labor input; and gross value added (in millions of national currency) is used as value added output. These datasets are provided in the same disaggregation categories with I-O tables. To eliminate inflation effects, we deflate all the input-output data and other monetary data into 2010 constant-dollar with the price index and exchange rate provided by WIOD. In order to use latest data to provide timely empirical study results as possible, we collect the environment data from the Eora Global Supply Chain Database: energy consumption; CO₂, SO₂ and NO_X emissions (Lenzen et al., 2012). Eora provides a more detailed environmental data lists updating to 2015. To match with the I-O tables, we mapped the EORA data to WIOD's disaggregation categories. The details are provided in Appendix A.

4. Empirical results

4.1. Trades’ impact on global direct emissions

4.1.1. Trade pattern level

Emissions induced by global trade account for a significant share, and more than 25% of global CO₂ emissions is embodied in trade (Fig. 3 ). This result is consistent with (Zhang et al., 2017a). The environmental impact of trade on these 15 economies is significant, and their territorial emissions are greatly stimulated. This also illustrates the necessity of studying the environmental impacts of their trades. This is in line with the results of existing studies (Li and Liu, 2020; Peng et al, 2016). Table A3 and A4 in the appendix summarize the proportions of emissions caused by exports in developing and developed economies, respectively. Among the total emissions of developing countries, 24.27% of CO₂, 23.99% of SO₂ e and 24.09% of NO_X emissions are to meet export needs. Among the total emissions of developed countries, 20.14% of CO₂, 31.88% of SO₂ and 30.73% of NO_X emissions are to meet export needs. These proportion reflects the degree to which the country's emissions are integrated into the supply chain. The higher the degree, the more likely it is to be affected by the evolution of trade pattern.

Fig. 3.

Structure of CO₂, SO₂, and NO_X direct emissions of the 15 economies (million tons). Notes: the length of bars refers to emissions embodied in domestic production and three export links. The color of bars corresponds to the share of emissions induced by exports, from the smallest share in yellow to the largest share in red.

From the perspective of specific trade pattern, the total intermediate product trades contribute more than 75% of trade-induced emissions, far exceeding the final product trade. Fig. 4 presents the temporal evolution of embodied emissions in 2000–2014. Notably, with the trade scale's dwindling during global economic crisis, the embodied emissions dropped to the bottom. It is found that the GVC-related product trades were most vulnerable, followed by last-stage intermediate product trades. The CO₂, SO₂ and NO_X emissions induced by GVC-related product trades declined 20%, 18% and 17% in 2009, respectively. And they were also fastest-rebounding when the economy recovered. Therefore, the intermediate product trades contribute the most to trade-induced emissions, and it is also the most sensitive to global economic changes, which brings more uncertainty to global environment changes. This fact emphasizes the necessity to study the environmental impacts of specific trade patterns, especially in the context of rising trade protectionism.

Fig. 4.

The structure and changing trend of emissions. Notes: (1) the area graphs above show the changing trends of embodied CO₂, SO₂, and NO_X emissions, and the colors correspond to three specific export links respectively. (2) The histograms blow are the changing rate of embodied emissions during 2008–2011.

Fig. 5 presents the difference between the emissions under trade scenario and no-trade scenario. The differences between the emissions under trade scenario and no-trade scenario are always negative in 2001–2014 (Fig. 5 left), indicating that free trade will reduce emissions, and restricting trade will increase emissions. And the more complex the global value chain network, the more significant this impact. The intermediate product trades have stronger effect for emissions’ reduction than final product trades. For CO₂, the last-stage intermediate product trades provide the main impetus for emissions’ reduction. In 2014, the intermediate products-induced CO₂ emissions of trade scenario reduced 466 million tons over that of no-trade scenario, accounting for 5.7%. And despite its weak effect, the GVC-related products also help to reduce CO₂ emissions. However, the reduction effect of intermediate product trade is largely offset by final product trade during 2001–2013. In 2007, the increased emissions promoted by final product trade reached a peak of 431.5 million tons, accounting for 5.4% of 2007 trade-related emissions. Since then, the cleaner final product trades achieved CO₂ savings in 2014. For SO₂ and NO_X, the three trade patterns show the same and balanced reduction effects. The effects grow gradually and just have some volatility during economic crisis. Considering the results under two scenarios, we can conclude that although intermediate product trade contributes more than 75% of trade-induced emissions, its role in promoting emission reduction is also the most obvious. Therefore, trade restrictions on intermediate products would particularly increase pollutant emissions.

Fig. 5.

Three trade patterns’ effects on embodied emissions. Notes: The negative value means that the emissions in trade scenario are less than that in no-trade scenario. The histograms on the right are country-level decomposition of the left 2014 results.

Compared with the no-trade scenario, trade has increased the CO₂, SO₂, and NO_X emissions of developing economies by 12.9%, 9.8% and 12.3%, respectively. Trade has reduced the CO₂, SO₂, and NO_X emissions of developed economies by 6.0%, 29.4% and 21.2%, respectively. The impact of trade on the three pollutants is different. The reasons are: First, as described in Section 4.1.1, the degree of integration of CO₂, SO₂, and NO_X into the global supply chain is different. Second, the sources and properties of these three gasses are not exactly the same, especially their emission factors (Qian et al., 2021; Thompson et al., 2014). China, USA and EU-28 are almost the main emitters of these three pollutants. The impact of trade restrictions on the environmental footprint of countries is different due to their different roles in global value chains. There are four types: the first type corresponds to trade restrictions’ effect of reducing emissions and final product trades as the main force. China is the representing country. Compared with the no-trade scenario, China's participation in international trade increased CO₂, SO₂, and NO_X emissions by 33.3%, 23.9% and 20.3%, respectively. Among them, the trade of final products increased CO₂, SO₂, and NO_X emissions by 829.4 million tons, 4504.9 thousand tons and 2624.3 thousand tons, only a few of which are offset by intermediate products’ negative effect. India and Korea have the same final product trades’ effect as China. India, in particular, acts as the workshop of the world (Wang et al., 2020). The second type corresponds to trade restrictions’ effect of reducing emissions and intermediate product trades as the main force. Russia is the representing country because of its massive raw materials exports. The third type corresponds to trade restrictions’ effect of increasing emissions and final product trades as main force. USA is the representing country. And the forth type corresponds to trade restrictions’ effect of increasing emissions and intermediate product trades as the main force. EU-28 is the representing country. In general, trade protection has reduced emissions from emerging developing economies and increased emissions from developed economies. Based on the SDG scores, Xu et al., pointed out that trade is conducive to the sustainable development of 65% of developed countries, but not conducive to the sustainable development of 60% of developing countries. Although the research indicators are different, this conclusion is basically consistent with this research (Xu et al., 2020).

4.1.2. Sector level

This section will examine the impact of trade protection on direct pollutant emissions from the sector perspective. This section can also be regarded as a further exploration of the reasons behind the results of the previous section. Fig. 6 shows the sector-level impact of three trade patterns on emissions. Please refer to Table A2 in the appendix for specific department classification of S1, S2,…, S7. It is easy to see that the trades’ contribution to global CO₂ reduction is mainly achieved by the high-emission sectors such as manufacturing, commercial and public services, and mining. According to the results of 2014, in all sectors, the last-stage intermediate product trade showed significant emission reduction effects. Among the total emission reduction of manufacturing, commercial and public services, and mining, the emission reduction ratio of last-stage intermediate product trade was as high as 94.55%, 43.86% and 63.56%, respectively. The electricity, gas and water supply sector, another major emitter, reduced emissions by 168.48 million tons in the last-stage intermediate product trade, but this was offset by final product trade and GVC-related product trade, which increased emissions by 1.7 times as much. The results for SO₂ and NO_X show slightly different results. Manufacturing trades contributed the largest share of emissions reductions, accounting for 57% and 56% of total SO₂ and NO_X emissions respectively. Mining, transport, commercial and public services trades also play a role in reducing emissions. Similar to the conclusion in the previous section, in each sector, almost all the three trade patterns play the roles of emission reduction to varying degrees. A few exceptions are that the electricity, gas and water supply sector's final goods trade increased SO₂ emissions by 961.42 kgtons, while the agriculture's three trade patterns all played the role of increasing NO_X emissions, albeit at a small level.

Fig. 6.

Three trade patterns’ effects on sectors’ embodied emissions. Note: The value is the difference between the emissions under the trade scenario and no-trade scenario. A negative value means that the sector's emissions under the trade scenario are less than that under the no-trade scenario. The three rows represent the calculation results of CO₂, SO₂ and NO_X emissions respectively. The three columns represent emissions caused by final products, last-stage intermediate products, and GVC-related intermediate products trade.

The previous section divided the economies into four types according to their trade restriction's different impact on their emissions. Fig. 7 further explains the role of three trade patterns in different countries from the sector level. China's trade restriction measures on final product would played a role in reducing emissions in all sectors, especially the manufacturing and electricity, gas and water supply sectors. And the last-stage intermediate product trade restriction could only slightly offset this effect. In addition, its mining sector's intermediate products trades are quite clean, reducing SO₂ and NO_X emissions by 1072 kgtons and 894 kgtons. Contrary to China, Russia's intermediate goods’ trade has greatly contributed to the emissions’ increase, especially in the manufacturing, electricity, gas and water supply, and transport sectors. From the perspective of aggregates, we found that the difference in trades’ environmental impact between developed and developing economies is dominated by manufacturing. In 2014, developed economies would increase by 170.9 million tons of CO₂, 3091.4 million tons of SO₂ and 3532.6 million tons of NO_X through manufacturing trade, under the no-trade scenario. But in developing economies, 372 million tons of CO₂, 1749 kgtons of SO₂, and 510 kgtons of NO_X would be reduced due to manufacturing trade. The former is 0.46, 1.77 and 6.93 times of the latter, respectively.

Fig. 7.

Three trade patterns’ effects on countries’ embodied emissions. Note, the value is the difference between the environmental efficiency under the trade scenario and no-trade scenario in 2014. The column ‘R_ing’ denotes the average impacts of the developing economies other than China, Russia, and India. The column ‘R_ed’ denotes the average impacts of the developed economies other than the European Union, the United States and Japan.

4.2. Trades’ impact on global environment efficiency

4.2.1. Trade pattern level

This section will discuss the impact of trade protection on countries’ environmental efficiency. Fig. 8 shows the environmental efficiency scores of 15 economies under trade and no-trade scenario in 2000 and 2014. It can be seen that most countries’ trade-related environmental efficiency is far below the optimal situation, which reflects that they have great potential for improving their environment efficiency.

Fig. 8.

The economies’ environment efficiency under two scenarios.

Fortunately, during 2000–2014, the efficiency has been improved significantly, and the global average level has risen from 0.26 to 0.34. But this also means that at the best technical level, more than 60% of pollutant emissions could still be avoided. In 2000, the best performer was Switzerland, with a score of 0.73. The remaining countries scored below 0.5. The worst performers were India (0.09), China (0.07) and Russia (0.06). In 2014, Switzerland's trade environment efficiency is close to optimal, followed by Canada (0.71) and Norway (0.46). In addition, the United States, the EU-28, Australia, and South Korea have made great progress. On the whole, the trade environmental efficiency in developed countries is always higher than that in developing countries, but both are in a state of progress.

Comparing the relative value of environment efficiency under the two scenarios, the performance of trade scenario is better than that of no-trade scenario. Trade has played a role in improving environmental efficiency in most countries, and this impact is also developing in a positive direction. In other words, trade restriction measures will reduce global environmental efficiency. This is mainly reflected in the 2014 histogram showing a darker green. In 2014, the countries with the most significant improvements in trade's environmental performance were Norway, Australia, and South Korea. In 2000, four of the 15 major economies had lower environmental efficiency in the trade scenario than the no-trade scenario: Canada (−0.09), Turkey (−0.08), Australia (−0.01), South Korea (−0.01). The values represent the efficiency difference between trade scenarios and no-trade scenarios. In 2014, this situation only appeared in Canada. In general, from the comprehensive perspective of multiple factors, the environmental efficiency of developed and developing countries in trade scenario is higher than that in no-trade scenario. And the improvement effect of trade on developed countries is more obvious, with the difference of 0.05. Developing countries are making progress, and the positive impact of free trade has risen from 0.01 in 2000 to 0.03 in 2014. Therefore, trade protectionism will have a more negative impact on the environmental efficiency of developed countries. This is basically consistent with the results of existing studies, that is, trade cannot achieve a win-win situation between the economy and the environment (Du et al., 2020). Liu et al., also pointed out that trade friction can only reduce emissions in the short term and is not conducive to sustainable development (Liu et al, 2020).

Fig. 9 shows the environmental efficiency scores of three specific trade patterns in two scenarios. From the global average level, the environmental performance of these three links under the trade scenario is higher than that under the no-trade scenario, and trade activities have played a positive role. The most obvious among them is GVC-related product trades. Comparing the left and right sides of Fig. 9, it is easy to find that the most obvious impact of trade appears in Canada, which is dominated by final product production (−0.17) and last-stage intermediate product trade (−0.22). These two negative effects are partially offset by the GVC-related product trade (+0.16). The positive impact of Switzerland trade is dominated by the GVC-related product trades, and the positive impact of Norway trade is reflected in all the three trade patterns. For other economies, trade plays roughly the same role in the three links. By comparing developed and developing economies, we found that global trade protection measures are not conducive to the improvement of the environmental performance of final products in developed countries, and their impact on intermediate products is not obvious. But it is conducive to improving the environmental efficiency of the final product in developing countries, rather than intermediate products.

Fig. 9.

The economies’ environment efficiency of three trade patterns under two scenarios. Note, the value corresponding to the color is the calculation result in 2014.

4.2.2. Sector level

This section will discuss the impact of trade protection on sectors’ environmental efficiency. The Fig. 10 shows the difference between trade and no trade scenarios. A positive difference indicates that free trade has a positive impact on the environment of this sector of the country. Negative values represent the negative effects. It can be seen that the global average level of trade environmental efficiency in agriculture, commercial and public services, and mining sector is higher than 0.6, followed by construction and transport sector, at around 0.5. The efficiency of the manufacturing , electricity, gas and water supply sector is extremely low, only about 0.3. Therefore, it can be considered that the manufacturing and electricity, gas and water supply sector with the highest emissions are the main drivers of the global inefficiency. In the results of various sectors, Switzerland among the developed countries performs best in almost all sectors. The United States has the best performance in agriculture, commercial and public services and mining, and the EU-28 has relatively balanced performance in all sectors. Among developing economies, the agricultural sector in China and India, and the commercial and public services sector in Russia performed better, while pollution-intensive sectors such as manufacturing and electricity supply sectors performed poorly.

Fig. 10.

The economies’ environment efficiency of three trade patterns and seven sectors under two scenarios.

It is easy to see that trades’ impact on developed countries’ environmental efficiency is more pronounced than developing countries’. Combining the results of Section 4.2.1, the positive effects of trade on Norway's environmental efficiency were mainly reflected in agriculture and transportation. The difference between the two scenarios, that is, the positive effect is about 0.5 and 0.25, respectively. The improvement in Australia's environmental performance from trade was mainly through the transport trades of intermediate goods, with a positive effect of about 0.13. The improvement in South Korea's environmental performance from trade is through manufacturing. In addition, Switzerland's agricultural trade and the final products trades of commercial and public services have significantly improved environmental efficiency. Among sectors in Canada, the difference in agricultural intermediate goods trade under the two trade scenarios is negative. That is, the environmental performance related to the export of intermediate goods in the trade scenario is worse than that in the no-trade scenario. Therefore, from a sector perspective, the negative impact of trade on the Canadian environment is reflected in the trade of agricultural intermediate goods.

In developing countries, whether it is a trade scenario or a non-trade scenario, its environmental performance is relatively far from the frontier, and its efficiency has great potential for improvement. As can be seen from the right side of Fig. 10, trade in Mexico's commercial sector has the most significant positive impact on the environment, while trade in final products in the manufacturing sector in Indonesia reduces environmental efficiency by 0.21. For China and Brazil, the three trade patterns play the positive role in almost all sectors, particularly, China's agriculture (around +0.12), mining (around +0.10), and Brazil's agriculture (around +0.10). Studies have pointed out that the agricultural sector of developing economies is mainly a net exporter of pollutants, and trade can reduce its hidden nitrogen emissions (Oita et al., 2016). After comprehensively considering economic, social, and environmental factors, this study has reached different conclusions. In addition, the positive effects of free trade are not the significant in the major pollution-intensive sectors, such as manufacturing and electricity supply sector.

5. Discussion

Trade protectionist actions may increase global CO₂, SO₂ and NO_X emissions, which is mainly reflected in the impact of intermediate goods trades. Intermediate goods trade reflects the degree to which a country's internal industrial system has been established. The large proportion of intermediate goods trade shows that the countries have a solid foundation for industrial development. Under the no-trade scenario, the external environment in which countries participate in the division of labor in the global value chain has undergone profound changes, and trade protectionism will exacerbate the environmental risks of countries that are heavily involved in global intermediate products trade. Developing economies are usually at the end of the global value chain and are in the rapid development stage of economic construction. Deep participation in the international division of labor is a powerful driving force for their economic development. Although trade protection measures are not conducive to their economic growth, they will directly reduce the emissions embodied in the exports of these countries. Developed countries led by the EU-28 and the United States have avoided a large number of pollutant emissions through international trade. Trade restrictions have made the EU-28′s manufacturing and power supply industries lose their advantage in reducing emissions of intermediate products. For the US, although trade restrictions on the final product will not significantly increase emissions, its trade value is also huge, so the negative impact is difficult to offset.

The current global environmental efficiency is very low, and trade protection actions will further reduce it, and this impact is strengthening. The reason is that, on the one hand, trade protection measures will increase trade costs, such as tariffs. On the other hand, these measures are not conducive to technological exchanges and cooperation between countries, and it is impossible to achieve the optimal allocation of resources, which leads to lower market efficiency. In the scenario of trade restrictions, the uncertainty of importing countries’ policies has also promoted the inward development of exporting countries’ intermediate products, which is even more detrimental to the development of global trade, and trade protection will become more severe. For developing and developed countries, these effects are negative. In recent years, China's agriculture, mining industry and Brazil's agriculture have all improved their environmental efficiency through trade. Although the impact is small, it also reflects the efforts these countries have made in environmental protection. But trade protection will make these efforts in vain.

The impacts of trade protection on the global pollutant emissions are very different from the impacts on global environmental efficiency, especially for developing economies. Trade protection activities could reduce the emissions in emerging developing countries directly, which is conducive to the realization of their own emission reduction targets. Since the 21st century, the economic development of developing economies has greatly benefited from global trade and has obtained a better allocation of input resources, thereby achieving better environmental performance. Trade protection actions aimed at developing countries, such as special controls on Chinese exports and higher tariffs, will inevitably lead to a certain decline in economic output and unemployment of workers. Taking these factors into consideration, trade protection is not good for developing countries, and even more for developed countries. Trade protection will reduce the efficiency of the global environment. This study believes that when measuring the environmental impact of trade restrictions on a country, it is necessary to consider not only the single impact of trade-related emissions, but also the combined impact of trade-induced technology and resource transfer, which is more in line with the needs of policy makers.

6. Conclusion

This study discusses the environmental impact of trade protectionism on countries from the two levels of specific trade patterns and sectors, and whether it can improve the environment of countries. We have established a comprehensive evaluation model that combines multi-regional input-output (MRIO), data envelopment analysis (DEA) and scenario analysis. And we proposed a comprehensive environmental efficiency index that measures the environmental performance of trade-related production activities under given economic and social factors. It covers the three dimensions of economy, society and environment.

The results of the study show that trade protection has greatly increased global pollutant emissions, while also reducing global comprehensive environmental efficiency. But this effect varies from country to country. For developed countries, the impact of trade protection is negative, both in terms of emissions and environmental efficiency. Trade protection increases its CO₂, SO₂ and NO_X emissions by 6.0%, 29.4% and 21.2%, respectively, while also reducing its environmental efficiency by 5%. This is the result of a long history of outsourcing of pollution-intensive industries in developed economies. Under the trade restriction scenario, all the pollutants that were originally transferred abroad would be emitted within its territory, and the employment and economic output driven by industrial repatriation would not be able to offset the negative environmental impact. For developing economies, trade protection can reduce their CO₂, SO₂ and NO_X emissions by 12.9%, 9.8% and 12.3%, respectively, and reduce their environmental efficiency by 3%. While eliminating trade can significantly reduce their export-embodied emissions, it also limits the economic and employment benefits of trade. The comprehensive environmental performance indicators show that these negative impacts outweigh the positive impacts. But for decision makers, the combined effects of social, economic and environmental factors need to be considered, and protectionism is detrimental to both developed and developing countries.

According to the results of this study, the trade protection dominated by developed countries is not conducive to the improvement of their own environmental efficiency, and even more detrimental to global sustainable development. In the post-pandemic era, while implementing pandemic prevention and control measures, developed economies should advocate trade freedom and the trade barriers should be gradually weakened. In particular, import tariffs and value-added tax on intermediate products should be reduced to provide more convenience for the trade of intermediate products. For developing countries, efforts should be made to improve their economic strength and technological level, continuously improve their position in the global value chain, and reduce the emissions embodied in free trade. In addition, countries should cooperate internationally and take advantage of trade's transfer of resources and technology. The advantages of advanced economies can benefit the world, which is conducive to the improvement of global environmental efficiency.

Inevitably, there are a few potential extensions worth further exploring. First, the basic model used in this study is the MRIO model, which is linear, assuming no factor substitution and constant returns to scale. This maybe not realistic. And estimating physical variables such as employment and emissions based on monetary data may also cause some deviations (Arto et al., 2014). The input-output partial closed model that endogenizes the inducing consumption and investment can better characterize the impact of trade environment. Second, we use the domestic technology assumption to calculate the results of no-trade scenario, which means each country produces the original imported goods with its own technology. But the reality is that one country may not have the corresponding technology or resource endowment to achieve it (Duchin and López-Morales, 2012; Strømman et al., 2009). Third, in sustainability performance assessment, we only selected the three main emissions of CO₂, SO₂ and NO_X, and neglected other pollutants like particulate matter, volatile organic compounds and other air pollutants, water pollutants and solid wastes. Although the conclusion can reflect regional difference of sustainability, there are still some deviations.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Editor: Prof. Dabo Guan

Appendix. Region and sector classifications

Tables A1-A4

Table A1.

List and classification of economies.

Acronym	Economy
AUS	Australia
BRA	Brazil
CAN	Canada
CHE	Switzerland
CHN	China
IDN	Indonesia
IND	India
JPN	Japan
KOR	South Korea
MEX	Mexico
NOR	Norway
RUS	Russia
TUR	Turkey
USA	America
EU-28	Austria, Belgium, Britain, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden (As of 2014)

Table A2.

(continued)

Code	Integrated sector	Sector code in WIOD	Sector in EROA
S6	Transport	Land transport and transport via pipelines	Transport
		Water transport
		Air transport
		Warehousing and support activities for transportation
S7	Commercial and public services	Sewerage; waste collection, treatment and disposal activities; materials recovery; remediation activities and other waste management services	Education, Health and Other Services
		Wholesale and retail trade and repair of motor vehicles and motorcycles	Maintenance and Repair
		Wholesale trade, except of motor vehicles and motorcycles	Wholesale Trade
		Retail trade, except of motor vehicles and motorcycles	Retail Trade
		Postal and courier activities	Post and Telecommunications
		Accommodation and food service activities	Hotels and Restaurants
		Motion picture, video and television program production, sound recording and music publishing activities; programming and broadcasting activities	Education, Health and Other Services
		Telecommunications	Post and Telecommunications
		Computer programming, consultancy and related activities; information service activities	Financial Intermediation and Business Activities
		Financial service activities, except insurance and pension funding
		Insurance, reinsurance and pension funding, except compulsory social security
		Activities auxiliary to financial services and insurance activities
		Real estate activities
		Legal and accounting activities; activities of head offices; management consultancy activities
		Architectural and engineering activities; technical testing and analysis
		Scientific research and development
		Advertising and market research
		Other professional, scientific and technical activities; veterinary activities
		Administrative and support service activities
		Public administration and defense; compulsory social security	Public Administration
		Education	Education, Health and Other Services
		Human health and social work activities
		Other service activities
		Activities of households as employers; undifferentiated goods- and services-producing activities of households for own use	Private Households
		Activities of extraterritorial organizations and bodies	Re-export & Re-import

Table A3.

Emission composition of developing countries.

	CHN	IND	RUS	BRA	IDN	MEX	TUR	Exports share
CO2 emissions for domestic demand	8,133,121	1,645,485	1,120,404	498,528.7	404,551.1	456,893.1	285,728.2	24.27%
CO2 emissions for exports	2,758,702	371,008.8	480,964.6	67,558.7	107,312.9	124,342.3	110,794.2
SO2 emissions for domestic demand	49,578.16	11,124.07	3974.652	1172.596	1630.586	2052.389	882.247	23.99%
SO2 emissions for exports	15,677.3	1270.452	2567.834	1113.154	406.9181	434.3788	756.5997
NOX emissions for domestic demand	29,257.24	10,656.51	3889.945	2442.475	2626.328	1201.165	1227.92	24.09%
NOX emissions for exports	9532.761	1874.163	2400.025	440.8145	600.6007	923.1818	511.3076

Table A4.

Emission composition of developed countries.

	USA	EU-28	JPN	KOR	CAN	AUS	NOR	CHE	Exports share
CO2 emissions for domestic demand	4,953,049	2,268,988	1,035,911	404,590.3	378,216	307,915.9	25,346.49	27,629.54641	20.14%
CO2 emissions for exports	436,080.9	1,154,899	235,728.8	216,655.3	209,477	86,656.44	18,928.03	13,111.51699
SO2 emissions for domestic demand	9266.228	11,385.17	2881.948	3033.344	2769.577	1372.085	245.2704	136.013103	31.88%
SO2 emissions for exports	7433.172	2403.698	1858.217	1062.083	759.2613	722.7504	177.946	135.0050317
NOX emissions for domestic demand	11,199.98	14,817.49	3389.263	3725.914	2429.388	1157.676	296.5498	147.5577449	30.73%
NOX emissions for exports	9229.41	3144.359	1118.448	755.6769	961.0895	930.4643	199.0398	150.4742421