key: cord-0967601-40fb8m90
authors: Zhang, Ke; Wang, Xingwei
title: Pollution Haven Hypothesis of Global CO(2), SO(2), NO(x)—Evidence from 43 Economies and 56 Sectors
date: 2021-06-18
journal: Int J Environ Res Public Health
DOI: 10.3390/ijerph18126552
sha: 8bcfc260299b5f2bff9bac09fbe111d2455d34f7
doc_id: 967601
cord_uid: 40fb8m90

With the development of trade liberalization, the pollutants emissions embodied in global trade are increasing. The pollution haven hypothesis caused by trade has aroused wide attention. The fragmentation of international production has reshaped trade patterns. The proportion of intermediate product trade in global trade is increasing. However, little has been done to study the pollution haven of different pollutants under different trade patterns. In this paper, major environmental pollutants CO(2) (carbon dioxide), SO(2) (sulfur dioxide), and NO(x) (nitrogen oxides) are selected as the research objects. This study investigated the global pollution haven phenomenon in 43 countries and 56 major industries from 2000 to 2014. Based on the MRIO model, the trade mode is divided into three specific patterns: final product trade, intermediate product trade in the last stage of production, and the trade related to the global value chain. The results show that trade liberalization could reduce global CO(2), SO(2), and NOx emissions, and intermediate product trade has a more significant emission reduction effect than final product trade. Trade’s impacts on each country are various, and the main drivers are also different. For example, the European Union avoids becoming a pollution haven mainly through the trade related to the global value chain. The suppressed emissions under this trade pattern are 71.8 Mt CO(2), 2.2 Mt SO(2), 2.2 Mt NO(x). India avoids most pollutants emissions through intermediate product trade. China has become the most serious pollution haven through final product trade. The trade pattern could increase China 829.4 Mt CO(2), 4.5 Mt SO(2), 2.6 Mt NO(x) emissions in 2014.

In the past few decades, global trade has grown rapidly [1], and trade has an increasing impact on resource flows and pollutant transfers among countries [2] . The pollutants embodied in international trade are also increasing [3] , which has raised people's concerns about the environmental problems caused by trade.

There have been some debates among scholars about the environmental consequences of trade liberalization [4] . The pollution haven hypothesis is the most significant one [5] . In order to promote economic growth, some countries may lower environmental standards to attract pollution-intensive industries, which will lead to the transfer of pollutants emissions [6] . If this hypothesis is true, high-standard countries' environment may be improved, but from a global perspective, the world's total pollutants emissions may increase. Therefore, through the assessment of the pollution haven phenomenon, we could understand the emission responsibility of each country and promote global trade's coordinated and sustainable development.

Most existing studies have analyzed the pollution haven of CO 2 [1, [7] [8] [9] . SO 2 and NO x are also major environmental pollutants in the atmosphere, but little has been done about them. This paper studies the pollution haven of CO 2 , SO 2 , and NOx. With the deepening of production fragmentation, intermediate product trade has played an increasingly important role in global trade [10] , and the pollution haven hypothesis has become more complex [1] . In order to reveal the impact of international production fragmentation on the environment, this paper divides trade into three patterns: final product trade, intermediate product trade in the last stage of production, and the trade related to the global value chain, so as to study the pollution haven phenomenon under the three patterns.

Based on the environmentally extended MRIO model, this paper assesses the pollution haven phenomenon of global CO 2 , SO 2 , and NOx emissions in three specific trade patterns. There are three main findings: (1) overall, the trade could reduce global emissions of CO 2 , SO 2 , and NOx; (2) intermediate product trade shows a more significant reduction effect than final product trade during the period 2000-2014; and (3) most countries avoid becoming a pollution haven through trade, but the main drivers are different from countries. For example, the European Union avoids becoming a pollution haven mainly through the trade related to the global value chain. China becomes the worst pollution haven through final product trade.

The remainder of this paper is structured as: Section 2 reviews the related studies about pollutions haven, presents the innovation of the study, Section 3 explains the modeling process and data sources, Section 4 presents the main results and discussions, and Section 5 summarizes the main findings and limitations of the study.

In recent years, a large number of air pollutants CO 2 , SO 2 , and NOx have been reallocated, posing an increasing threat to the ecological environment and human health [11] [12] [13] [14] . The rapid development of trade globalization has promoted the global flows of pollutants, which has aroused people's concern about the plight of trade-induced environmental problems. The share of pollutants emitted by international trade has been increasing over time [2] . Zhang et al. found that in 2009, about 25% of global carbon emissions were caused by international trade [5] . The emissions embodied in trade cover the goods' entire production process: from raw materials acquisition, manufacturing, processing, and transportation to the final product in the hands of consumers [15] . With the development of economic globalization, countries' active participation in the global labor division will lead to the transfer of pollutants emissions [16] . The pollution haven hypothesis is an important issue on the environmental problems caused by international trade. Pollution haven means that international trade could increase global pollutants emissions [17] . Developing countries tend to attract developed countries' investment by virtue of their low labor and low environmental costs. In order to pursue the maximization of profits, developed countries often spend less on investment and transfer their pollution-intensive enterprises to developing countries with lower prices. These developing countries assume the pollutants emissions responsibility and become pollution haven in global trade [18] . To make matters worse, if the pollution-intensive industries' efficiency in these developing countries is low, the world's total emissions will increase due to trade diversion [19] . Scholars have worked on various aspects of the pollution haven hypothesis over the years: Duan et al. studied the role of multinational enterprises in global carbon emissions [20] Himics et al. researched trade liberalization's effect on CO 2 emissions in agriculture [21, 22] . There are also some arguments about the pollution haven: Che et al. found that China became a pollution haven in international trade [23] . However, Zhang et al. found that China's exports are becoming cleaner due to globalized production [24] . Xu et al. pointed out that trade liberalization reduces smog emissions [25] . Therefore, it is of great significance to assess the pollution haven hypothesis. However, at present, most scholars have studied the pollution haven phenomenon of CO 2 from various perspectives [7] [8] [9] , and few scholars have taken both SO 2 and NO x into consideration. Therefore, this study takes major environmental pollutants CO 2 , SO 2 , and NOx as research objects to assess pollution haven.

Studies have found that the multi-regional input-output model could describe the close correlation among countries and sectors [26] , and a large number of scholars have ap-plied it in the environmental issues related to the global value chain. Lu et al. calculated the evolution trend of China's SO 2 emissions from the consumption side [11] . Lu et al. found some pollution-intensive sectors in China have been phased out to the Belt and Road Routes countries [27] . Jiang et al. studied the U.S.'s oil footprint drivers [28] . Chen et al. found global trade exacerbates land and virtual water's uneven distribution [29] . Chen et al. analyzed global energy flows by using the environmental-extend input-output model [30] . In addition, compared with the SRIO model, the MRIO model could reduce the measurement error [31] . In the MRIO model, the imported products and domestic products come from different places [32] . It could be seen that the MRIO model has great advantages in the research and is widely used. Therefore, this study uses the MRIO model to study CO 2 , SO 2 , and NOx emissions embodied in global trade from 2000 to 2014.

For the past few years, economic globalization's most prominent feature is production fragmentation in the global value chain [32] . Fragmentation means that the production process is broken down into different stages, and each country on the value chain produces a specific part. These countries are linked to each other through trade. The fragmentation could reshape trade patterns and optimize international labor division but redistribute trade-related emissions at the same time [33] . For example, it is found that there is a shift in recent years' global supply chain, and it means the transfer of resources and pollution [34] . Yilin et al. found carbon emissions were redistributed in the current complex global trade network [35] . In previous studies, Wang et al. studied the carbon emissions in China-Australia trade [36] . Araújo et al. found that Brazil played the role of the intermediate supplier in global value chains [37] . Scholars usually focused on bilateral trade or the role of one country in international production decentralization, but few assessed the composition of global value chains and the pollution haven of different trade patterns. Therefore, this paper studies the pollution haven hypothesis from the perspective of global state composition so that each country's contribution to the global pollution haven could be clearly seen in the results, which are more in line with the needs of policymakers. In addition, Zhang et al. distinguished three specific trade patterns in the international trade production division: final product trade, intermediate product trade in the last stage of production, and the trade related to the global value chain, and they studied the carbon pollution haven in three trade patterns [5] . Wang et al. also studied China's environmental impact of PM 2.5 in different trade patterns [32] . On the basis of previous studies, this study further refines research contents to the industry level and studies emissions responsibility and pollution haven of different industries and patterns.

The input-output method was proposed by Leontief in the 1930s [38] . It has been widely used to analyze the economic activities so as to reveal the internal relationship among industries of the national economy and do some economic forecasts and arrangements. Table 1 is the basic form of the multi-regional input-output model (MRIO). There are m countries (regions) in the multi-regional input-output model, and each country has n sectors. The rows in the table represent the distribution of intermediate product trade and final demand in each country's industry, and the columns represent intermediate input from different countries and industries.

The first quadrant, Z, reflects the direct economic linkages between the various industries of countries. Z pq ij denotes the intermediate input from country p's sector i to country q's sector j, which needs to be reprocessed before they can be consumed [39] . That is the intermediate demand matrix in the input-output table provided by the WIOD database: a 2464*2464 column vector formed by 44 regions and 56 sectors.

The second quadrant Y reflects the final demand matrix. Y pq j denotes country q's final demand for goods and services from country p's sector j, which could be consumed directly [40] . It is the final demand matrix in the input-output table: a 2464*1 column vector consisting of 44 regions, 56 sectors, and 5 final requirements.

The third quadrant V = V q j reflects the added value of each country [41] . The horizontal summation X q = X q j is the total output matrix of country q. It is the last column in the input-output table: a 2464*1 column vector formed by 44 regions and 56 sectors. A pq is the direct coefficient consumption matrix that represents the direct consumption of goods or services for unit total output in various sectors [42] ,

Then, the total output of a country can be expressed as:

A variant equation is as follows:r

where B pq = (I − A pq ) −1 is the Leontief inverse matrix [43] . It is also called the completed consumption coefficient matrix, meaning country q's direct and indirect demand for country p's commodities and services to produce the unit output.

The total exports from country p to country q could be expressed as:

According to Wang's study [44] , it could be proved that

Combine (4) and (5) to obtain

T pq f represents the final product trade exported from county p to country q. T pq i represents that country p provides intermediate product to country q, and country q reprocesses intermediate product itself to meet domestic final demand Y qq , that is, the intermediate product trade in the last stage of production. T pq g represents the trade related to the global value chain [45] , involving the remaining production process of semi-finished and finished products. The product crosses countries many times and may eventually be absorbed by final consumers [46] .

Based on the above analysis, the total output of country p can be decomposed from the production side:

A variant equation is as follows:

As can be seen from the above equation, the total output of country p is divided into four items: the first item represents domestic final demand, the second item represents direct export of final product trade, the third item is the direct intermediate product trade export, namely intermediate product trade export in the final stage, and the fourth item represents intermediate product trade export related to the value chain.

We define that pollutant emissions intensity coefficient of country p is represented by f p [47] , f p = e p x p . It is the column vector of the environment variable in the environment account divide by the total output column vector in the input-output table. e p represents the pollutant emissions of country p. x p represents the gross output of country p.

Based on the MRIO model and the above analysis, the total pollutant emission composition of country p can be expressed as:

E p can be divided into two parts: F p L pp Y pp is induced by domestic consumption and economic activities, and F p L pp T is induced by international trade. The decomposition of country p's gross pollutant emissions can be seen in Figure 1 .

For the pollutant emissions induced by international trade F p L pp T:

Under the scenario of trade, pollutant emissions of country p are caused by exports from country p to other countries:

Under the scenario of no trade, pollutant emissions in country p induced by domestic production of commodities that are originally imported from other countries:

ET represents the total emissions difference between trade (EWT) and no-trade (ENT) scenarios:

Similarly, according to the above derivation, ET can be decomposed as:

ET f represents the final product trade emissions difference between trade and notrade scenarios. ET i represents the emissions difference of intermediate product trade in the last stage of production. ET g represents the difference of the trade related to the global value chain. 

The data used in this paper includes two parts: input-output table and environmental pollutants emissions. The input-output table is derived from the latest world input-output database (WIOD 2016), which covers 43 countries (regions) and one rest of world (ROW), and 56 industries in each country [48] . Compared with the 2013 version, WIOD released in 2016 covers a wider range of countries and a more detailed industry classification [49] . WIOD database is also widely used in the field of input-output and environmental footprint tracking [50] [51] [52] . The newly released WIOD only provides environmental data of CO 2 , while the Eora database covers more than 2000 environmental indicators. Among these indicators, CO 2 , SO 2 , and NO x are the main pollutants that cause environmental pollution problems, so the data of CO 2 , SO 2 , and NO x in Eora is selected for research. In order to match the WIOD input-output table, this paper adjusted the Eora emission data to the WIOD form: 190 economies are merged into 44 economies, and 26 industries are disaggregated into 56 sectors according to the output structure. To better describe the results, the study combined the EU 28 countries into one economy EU-28 and combined 56 sectors into 7 main industries. The scope of this study is from 2000 to 2014, which covers some important events in the world: China's accession to the WTO in 2001, the global economic crisis in 2008, and the shale gas revolution in the U.S. after 2009. In addition, in order to reduce the impact of inflation, WIOD data used in the study are levelized in 2010 constant dollar value according to the price index and exchange rate. This may be because the U.S. is engaged in the high-value-added labor division in upstream production. The level of science and technology in the U.S. is advanced, and the country tends to export high-value-added intermediate products.

About 15% of India's emissions come from international trade, mainly through traditional Ricardian trade patterns. The final product trade and the intermediate product trade in the last stage of production are called the traditional Ricardian trade pattern or direct-value-added trade pattern [45, 54] . In 2014, India's emissions of CO 2 , SO 2 , and NO x were 2.2 billion tons, 13.4 million tons, and 12.5 million tons. Traditional Ricardian trade pattern accounts for about 12%, while the trade related to global value chain accounts for only about 3%. It might because India is a developing country with a low level of production. There is still a big gap compared with developed countries in this aspect, so the trade related to the global value chain takes up a small proportion.

The European Union's emissions caused by international trade account for a much higher proportion than the world average, and the emissions caused by the three trade patterns are basically balanced. In 2014, the proportion of CO 2 , SO 2 , and NO x caused by trade account for 33.7%, 43.7%, and 45.2% of the total emissions, while the world average level was about 25%, both higher than the world average level. This may be because the EU's economy is developed and has a high level of technology. It usually takes advantage of these superiorities and actively participates in international trade, so its emissions caused by global trade are higher [55] .

In 

The pollution haven is reflected by the emissions difference between trade and notrade scenarios. If the difference is greater than zero, it indicates trade flows (trade liberalization) could increase pollutants emissions, lead to pollution haven. If the difference is less than zero, that means that trade could avoid the country becoming a pollution haven.

As can be clearly seen from Figure 

As can be seen from Figure 5 , most countries have avoided domestic pollutants emissions through international trade to varying degrees. The typical countries include the European Union and India. China and Russia are the typical countries that become pollution haven in global trade.

The EU avoids 4.3 million tons of SO 2 and 3.7 million tons of NO x in global trade in 2014. All three trade patterns could inhibit pollutants emissions, but the trade related to the global value chain's inhibiting effect is the most obvious. The suppression of SO 2 and NO x in this pattern accounts for 52% and 60% of the total embodied emissions. As for CO 2 , the EU's global value chain trade avoids 48.3 million tons of CO 2 . However, on the whole, the EU still increased 8.4 million tons of CO 2 through trade. Most of the industries in the EU avoid becoming a pollution haven in global trade. Typical industries are agriculture, manufacturing, and mining industries. More than 95% of the suppressed emissions come from these three industries. In addition, the study found the transport industry could contribute significantly to the EU's NO x emissions. About 85% increased NO x comes from transportation, and it could add 761.3 tons of NO x . China has increased its most emissions through the final product trade pattern. It becomes the country of worst pollution haven, and most of its industries become pollution haven in global trade. In 2014, China increased about 830 million tons of CO 2 through final product trade, accounting for 96.7% of the total increased emissions. The increased SO 2 and NO x emissions are all from the final product trade, which are 4.5 million tons and 2.6 million tons, respectively. China's manufacturing and electricity industries are the main industries leading to pollution haven, and their emissions are mainly increased through final product trade. This mainly because China is the largest manufacturing country and takes coal as the main energy source, which may produce a large number of pollutants in global trade [56] . The mining industry is almost the only industry in China that avoids emissions. It avoids pollutants mostly through intermediate product trade. China's mining industry avoids 88.6 million tons of CO 2 , 1 million tons of SO 2 , and 877 tons of NO x in 2014. Overall, China is responsible for more air pollutants than other countries in global trade.

Russia is the country with the second most serious pollution haven. As can be clearly seen from Figure 5 , it increases pollutants emissions entirely through intermediate product trade. The increased emissions of CO 2 , SO 2 , and NO x through that trade pattern are 220 million tons, 1.4 million tons, and 1.6 million tons. It might because the trade of the country is related to its natural environment. Russia is rich in forest, oil, and gas resources. There may be pollutants embodied in the extraction of these natural resources [57] . Some raw materials and semi-finished products of energy will be exported in the trade. Most of Russia's industries become pollution havens in trade. The electricity industry is the most obvious example of a CO 2 pollution haven. The industry accounts for 60% of the total increased CO 2 and increases Russia 62.58 million tons of CO 2 emissions. About 80% of the increased NO x comes from the transportation industry, and it adds about 950 tons of NO x .

Overall Canada's increased emissions come from intermediate product trade in the last stage of production. This pattern contributes 21.6 million tons of CO 2 , 157 tons of SO 2 , and 168 tons of NO x emissions. It may because that Canada is rich in natural resources and has a small population. The country not only exports forest products and natural resources but also actively participates in the commodities production process.

South Korea mainly reduces three pollutants emissions through the trade related to the global value chain. The trade pattern avoids 26.7 million tons of CO 2 , 412 million tons of SO 2 and 348 million tons of NO x emissions [58] .

In recent years, the rapid development of trade globalization has promoted the global flow of CO 2 , SO 2 , and NO x , which has aroused people's concerning the trade-related environmental problems such as the pollution haven phenomenon. In this paper, the global pollution haven hypothesis is assessed by investigating the emissions of CO 2 , SO 2 There are several potential extensions for this research. First, the MRIO model used in this paper uses a fixed coefficient to represent the production system. This linear representation method actually ignores the price's inevitable influence in the pursuit of optimization [59] . Therefore, there may be some deviation between the calculated results and the actual situation. In the future, studies are expected to consider the results' sensitivity according to the price difference. Second, under no-trade scenarios, the study assumed that the original imported products are produced with domestic technology. However, the assumption may overestimate the technological level and natural resources of some importing countries. In the future, more realistic technical assumptions and resource distribution should be established for the accountings under no-trade scenarios. Third, future studies are expected to update the input-output table based on system optimization. It is suggested that studies could apply the method to assess recent pollution havens, such as the trade war between China and the U.S., the trade dispute of regional tariffs and non-tariffs, the control of the U.S.'s export tariffs, and the effects of COVID-19. 

The study did not involve humans or animals.

The data that support the findings of this study could be found in official statistics of World Input-Output Database at http://wiod.org/home and The Eora Global Supply Chain Database at https://worldmrio.com/.

The authors declare no conflict of interest. Manufacture of wood and of products of wood and cork, except furniture; manufacture of articles of straw and plaiting materials 8

Manufacture of paper and paper products 9

Printing and reproduction of recorded media 7 Petroleum, Chemical, and Non-Metallic Mineral Products

Manufacture of coke and refined petroleum products 11

Manufacture of chemicals and chemical products 12

Manufacture of basic pharmaceutical products and pharmaceutical preparations 13

Manufacture of rubber and plastic products 14

Manufacture of other non-metallic mineral products 

International production fragmentation, trade in intermediate goods and environment

Evolution of the exergy flow network embodied in the global fossil energy trade: Based on complex network

Multi-region input-output analysis of CO 2 emissions embodied in trade: The feedback effects

Trade liberalization and the environment: Evidence from NAFTA and U.S.manufacturing

A multi-regional input-output analysis of the pollution haven hypothesis from the perspective of global production fragmentation-ScienceDirect

Economic growth, international trade and air pollution: A decomposition analysis

Is China's development conforms to the Environmental Kuznets Curve hypothesis and the pollution haven hypothesis?

Parcelling virtual carbon in the pollution haven hypothesis

Trade from resource-rich countries avoids the existence of a global pollution haven hypothesis

Accounting for intermediates: Production sharing and trade in value added

Decadal Trends in Wet Sulfur Deposition in China Estimated from OMI SO 2 Columns

The Impact of Foreign Direct Investment on Environmental Pollution in China: Corruption Matters

Environmental Regulation and Exports: Evidence from the Comprehensive Air Pollution Policy in China

Cost-Effective Mitigation of Greenhouse Gas Emissions in the Agriculture of Aragon

The United Nations Framework Convention on Climate Change

Does China become the "pollution heaven" in South-South trade? Evidence from Sino-Russian trade

Growth, and the Environment

Trade openness and air pollution: City-level empirical evidence from China

Reassessing pollution haven effect in global value chains

Pollution haven or pollution halo? A Re-evaluation on the role of multinational enterprises in global CO 2 emissions

Heterogeneity of pollution haven/halo hypothesis and Environmental Kuznets Curve hypothesis across development levels of Chinese provinces

Does the current trade liberalization agenda contribute to greenhouse gas emission mitigation in agriculture? Food Policy

Will developing countries become pollution havens for developed countries? An empirical investigation in the Belt and Road

Production Globalization Makes China's Exports Cleaner

Trade liberalization and haze pollution: Evidence from China

Consumption-based carbon emissions and trade nexus: Evidence from nine oil exporting countries

Imbalance and drivers of carbon emissions embodied in trade along the Belt and Road Initiative ScienceDirect

Drivers of United States oil footprints trajectory: A global assessment

Global environmental inequality: Evidence from embodied land and virtual water trade

Global energy flows embodied in international trade: A combination of environmentally extended input-output analysis and complex network analysis

Comparison of city-level carbon footprint evaluation by applying single-and multi-regional input-output tables

Spatial production fragmentation and PM 2.5 related emissions transfer through three different trade patterns within

Carbon transfer within China: Insights from production fragmentation

Global supply chain shifting: A macro sense of production relocation based on multi-regional input-output table

Carbon network embodied in international trade: Global structural evolution and its policy implications

Carbon emissions embodied in China-Australia trade: A scenario analysis based on inputoutput analysis and panel regression models

Rodrigues Faria, W. Regional and Global Patterns of Participation in Value Chains: Evidence from Brazil

Quantitative input and output relations in the economic systems of the United States

A Carbon Footprint Time Serious of the UK-Results from A Multi-reigon Input-output Model

Measuring the environmental sustainability performance of global supply chains: A multi-regional input-output analysis for carbon, sulphur oxide and water footprints

Accounting for value added embodied in trade and consumption: An intercomparison of global multiregional input-output databases

The impact of domestic and foreign trade on energy-related PM emissions in Beijing

A Structural Decomposition Approach to Comparing MRIO Databases

Quantifying International Production Sharing at the Bilateral and Sector Levels

Characterizing Global and Regional Manufacturing Value Chains: Stable and Evolving Features. Cent. Studi Luca D'agliano Dev

Financial Crisis, Virtual Carbon in Global Value Chains, and the Importance of Linkage Effects. The Spain-China Case

Does the level of energy intensity matter in the effect of energy consumption on the growth of transition economies? Evidence from dynamic panel threshold analysis

An Illustrated User Guide to the World Input-Output Database: The Case of Global Automotive Production

Increased carbon footprint of materials production driven by rise in investments

Chinese CO 2 emission flows have reversed since the global financial crisis

Economic development and converging household carbon footprints in China

Carbon emissions embodied in the global supply chain: Intermediate and final trade imbalances

Follow the Value Added: Bilateral Gross Export Accounting. Bank Italy Temi di Discussione Work

On the determinants of industrial competitiveness: The European Union emission trading scheme and the Italian paper industry

Re-examining the embodied air pollutants in Chinese exports

Substantial nitrogen pollution embedded in international trade

Free Trade Agreements and Market Integration: Evidence from South Korea

Assessing sustainability performance of global supply chains: An input-output modeling approach