Tariff Reduction and Environment: Evidence from CAFTA and Chinese Manufacturing Firms

: In the context of anti-globalization and trade wars (especially between the US and China), China-ASEAN Free Trade Area (CAFTA) now plays a prominent role in many aspects. In this paper, we investigate how import tariff reduction in CAFTA affects the importers’ pollution emissions, using the ﬁrm-level data of Chinese manufacturing from 2002 to 2007. The mechanisms of import tariff reduction in CAFTA on pollution emissions are divided into technique, composition and scale effects. Our results indicate that import tariff reduction in CAFTA on ﬁnal goods is conductive to importers’ pollution reduction, whereas that on intermediates signiﬁcantly aggravates importers’ pollution emissions. Moreover, import tariff reduction has heterogeneous impacts on different types of enterprises in terms of industries, ownership, and region. Our results also ﬁnd that the state-owned importers’ emissions can hardly be affected through technique, composition and scale effects.


Introduction
Benefiting from the globalization and trade liberalization, China has achieved remarkable development by using low-cost factor input in trade. China is the world's biggest developing country. Moreover, its total trade in goods ranks the first in the world. With an increasingly important role in the global value chains, China has won the title of "world factory", becoming world's biggest exporter of "Made in China" [1]. In the past two decades, China is gradually becoming an important manufacturing country and export platform, increasingly participating in global chains. However, while we highlight the importance of exports to Chinese economy [2], we should not neglect the role of imports, especially at this stage that China gradually replaces domestic production of pollution-intensive products by importing from other developing countries. Moreover, with the rising position in the global value chains, China tends to import some raw materials and intermediates from other developing countries, such as Association of Southeast Asian Nations (ASEAN) (ASEAN was founded in 1967, and now it has 10 member countries: Brunei, Cambodia, Indonesia, Laos, Malaysia, Myanmar, Philippines, Singapore, Thailand and Vietnam.). Thus, the effects of imports on China's economy and firms' behaviors should not be ignored. Now, more and more countries are concerned about the problem of environmental pollution [3][4][5]. China, world's second biggest economy, also adopted stricter environmental regulation to reduce emissions [6], gradually replacing domestic production by importing pollution-intensive products. While highlighting the great importance of environmental protection [7,8] and importing raw materials and intermediates from other developing countries, China begins to establish comprehensive cooperation with these countries, including the establishment of Free Trade Area (FTA) to cooperate with them. After joining the FTA, Chinese economy continues to grow substantially [9]. China-ASEAN FTA (CAFTA) is the biggest regional FTA by population in the world. Moreover, it also has large intra-regional trade amount [10]. China is ASEAN's biggest partner in trade at present. At the same time, ASEAN surpasses the United States, becoming the second biggest partner in trade of China, with a total trade amount of 1.98 trillion yuan in the first half of 2019 (Data sources: China Customs Administration's report on import and export in the first half of 2019.). ASEAN is becoming an increasingly important trading partner of China. According to World Integrated Trade Solution (WITS) database, the proportion of Chinese seven main imports in total imports from ASEAN follows an increasing trend in general from 2014 to 2017. Moreover, most of China's main imports from ASEAN are pollution-intensive products.
During this period, there are two issues that require our attention. First, what is the relationship between import tariff reduction in CAFTA and firms' pollution emission? Trade, a significant force in promoting Chinese economic development, has great impacts on the firms' production. Firms can purchase and sale goods in the world market through the way of trade. Enterprises' production activities are easily influenced by the situation of market, thus there may be some connections between the degree of openness to the world and firms' activities of pollution emissions. So, what influence CAFTA may have to firms' activities of pollution emissions? Secondly, what are the specific mechanisms of the impacts that the import tariff reduction in CAFTA has on the firms' emission activities? It is obvious that the change in tariff can affect the firms' production scale, thus leading to the change in pollution emissions. But if we just focus on the scale effect of import tariff reduction in our analysis, and ignore other potential effects, we may come to a biased conclusion. Thus, we choose to add composition effect and technique effect in our analysis [11,12].
Existing literature analyzing environmental impacts of Regional Trade Agreement, mainly analyzes the influences of North American Free Trade Agreement (NAFTA) that includes the most developed countries, the America. Grossman and Krueger [12] first analyze NAFTA's environmental by using three effects. Cherniwchan [13] uses the samples of U.S. manufacturing firms in NAFTA to analyze the connections between trade liberalization and the environment. However, scant literature analyzes the environmental impacts of FTA whose members are all developing countries. To examine the environmental impacts of FTA whose members are all developing countries, we use CAFTA as our research object to analyze enterprises' emission activities. Since the links between China and ASEAN are closer, many scholars study the impacts of CAFTA. While most of the existing literature focuses on the economic effects of CAFTA [9,14], there is scant evidence proving what impacts CAFTA might have on pollution emissions of firms from China or ASEAN. Moreover, these papers analyze from a macro perspective or an industry perspective.
After some scholars put forward the theory of enterprise heterogeneity [15][16][17], many studies on international trades begin to focus on micro level analysis [18,19], analyzing the impacts of trades on economy from a micro perspective. According to Melitz [17], only highly productive companies can export to other countries. Moreover, the improvement of firm-level data also provides important data support to conduct micro research.
To analyze the firm-level effects, we use the data of Chinese importers during 2002-2007 (Thanks for the anonymous reviewers for giving us the useful suggestions on the problem of time span. Of course, we are responsible for the contents of our article. Because China-ASEAN on Trade in Goods was signed in 2004, we use the data from 2002 to 2007 in our analysis. Moreover, the authenticity of China's Industrial Firm Database beginning from 2008 has been questioned, so we just use the data from 2002 to 2007, and do not use the data from 2008 onwards. This data limitation is also another important reason we choose this time span. If better data is available in the future, we will continue our further research.) to analyze the effects of tariff reduction in CAFTA on the enterprises' emissions. Since the effects of trade may have some differences when considering different kinds of pollutants [20,21], our analysis selects two pollutants (SO 2 and Soot). We decompose the effects of import tariff reduction into the effects on final goods' tariff and the effects of intermediate inputs' tariff. We match the data on the importers' pollution emissions, tariff on final goods, tariff on intermediate inputs [13,22], and the firms' characteristics, and later we construct our dataset. Because there are scant literature analyzing the environmental effects of CAFTA, our analysis is a supplement to the research of the environmental effects of FTA. Moreover, our research offers us a new perspective on the impacts of economic cooperation from developing countries, helping us understand the impacts of CAFTA more comprehensively.
We first verify the effects of final goods' tariff reduction and intermediate inputs' tariff reduction on SO 2 and Soot emitted by China manufacturing firms. Our analysis shows that tariff on final goods is positively correlated with enterprises' emissions, which means that import tariff reduction on final goods in CAFTA will lead to less firms' pollution emissions. We also find that import tariff on intermediates inputs is negatively correlated with the firms' pollution emissions, which means that the importers will emit more when the tariff on intermediate inputs in FTA reduces. This result of tariff reduction in FTA on intermediate inputs stands in contrast to findings present by Cherniwchan [13]. In contrast to our research analyzing the environmental impacts of South-South trade liberalization, the analysis of Cherniwchan [13] studies the effects of an episode of North-South trade liberalization on the emissions of manufacturing firms in North (the U.S.), suggesting that import tariff reduction on intermediates input in NAFTA reduces the American firms' emissions. Next, we explore the mechanisms that how tariff reduction on final goods and intermediate inputs in CAFTA affect firm-level emissions. According to Grossman and Krueger [12] and Antweiler et al. [11], the environment can be influenced by trade through technique, composition and scale effects. To explain the mechanisms, we decompose the change of pollution emissions of the firms caused by the tariff reduction on final goods and intermediate inputs into technique, composition and scale effects. To explore how tariff reduction changes the firms' emission behaviors, we later distinguish the enterprise samples from three perspectives, which include technology intensity, ownership type and region, to analyze the heterogeneous effects on the development of manufacturing industries, the process of ownership reforms and regional coordinated development in China [23]. We also analyze the underlying mechanisms how tariff reduction in CAFTA affects firms' emissions from the perspective of heterogeneous firms.
Our analysis mainly makes contributions in two aspects. (1) We contribute to the existing literature examining the impacts of FTA composed of developing countries on environment by giving strong evidence from the micro perspective. Our analysis is an important supplement to the existing research which mainly focuses on the environmental effects of FTA whose members include some developed countries. (2) Since the ties between China and ASEAN in economy have reached an unprecedented level, more and more literature focuses on the effects of CAFTA from the perspectives of firms' productivity, poverty reduction effect and economic growth. However, scant studies analyze the environmental impacts of CAFTA. Our study on the environmental effects of CAFTA is a necessary supplement to the existing research about effects of trade on Chinese economy.
The remainder of our analysis is as follows. Section 2 identifies the CAFTA's effects on the environment. Section 3 shows the design of the empirical model, while Section 4 discusses our data. Section 5 are our empirical results. Finally, Section 6 concludes.

CAFTA'S Background Information
Our analysis focuses on the impacts of import tariff reduction in CAFTA following the China-ASEAN Agreement on Trade in Goods in 2004 on Chinese manufacturing firms' pollution emissions. After the 1997 economic crisis, some countries and regions in Asia are committed to moving to regional economic integration, which includes the form of Free Trade Agreements [9]. The concept of establishing CAFTA was initially proposed at the 2000 ASEAN Summit by China [24]. In the next few years, China and ASEAN have conducted a series of communication and negotiation to establish CAFTA. After establishing CAFTA, China and ASEAN have signed some free trade agreements, which significantly reduced the trade barriers between ASEAN and China. For CAFTA, one of the most milestones was the signing of China-ASEAN Agreement on Trade in Goods in 2004, which entered into force in January 2005 [9]. The tariff reduction specified in China-ASEAN Agreement on Trade in Goods in 2004 involves almost all the products of all member states, meaning that China and ASEAN begin to develop comprehensive cooperation in economy and trade. After that, the connections of China and ASEAN are closer. Now, China is ASEAN's biggest partner in trade, and ASEAN is China's second biggest partner in trade. According to World Integrated Trade Solution (WITS) database, Chinese weighted import tariffs from ASEAN on final goods and intermediates from 2002 to 2007 are generally declining. Moreover, considering different industries, the ranges of tariff rate reduction are not the same. The changes in tariff rates also vary during different years.

Identifying he Mechanisms
To find out the potential mechanisms of import tariff reduction in FTA on firms' pollution emissions, we decompose pollution emission reductions caused by the import tariff reduction in CAFTA into technique, composition and scale effects. These three effects were first used by Grossman and Krueger [12] to verify trade's impacts on environment. Since then, more and more scholars use these three effects in their subsequent research [11,[25][26][27]. However, these papers focused on the analysis in the macro level. Based on these previous studies, our analysis makes some modification. We use the firms' data to examine these three effects in micro level, including technique, composition and scale effects.
First, import tariff reduction in CAFTA can change the production technology of the firms, thus firms' pollution emissions will be changed. Trade liberalization can influence environment by changing the technology in the production process [12]. According to Pavcnik [28], firms' productivity improvement can be attributed to trade liberalization. Specifically, import tariff reduction on final goods may exacerbate the competition in domestic market and promote the firms' productivity. Furthermore, more high-quality intermediates will enter domestic market when import tariff on intermediate inputs decrease. This can lead to technology spillovers and helps firms improve their productivity [29]. Moreover, some research analyzes the relationship between productivity and firms' emissions. According to Bloom et al. [30], highly productive firms have lower level of pollution emission in their production activities. Productivity plays an important part in emissions abatement, firms with higher productivity tend to use pollution-abating technologies to reduce their emissions [31]. Thus, we can infer that import tariff reduction in CAFTA can affect firms' pollution emissions through technique effect.
Second, import tariff reduction in CAFTA can change the composition of the economic activities, thus firms' pollution emissions will be changed. According to Grossman and Krueger [12] and Antweiler et al. [11], trade can affect environment and pollution emissions by changing the composition of economic activities. Our analysis chooses the capital intensity to reflect the proportion of firms' factor input. Capital is an essential factor in production, thus the changing of factor composition can affect firms' pollution emissions. In the production process, capital-intensive firms tend to have a higher pollution emissions, while labor-intensive firms have a lower emission level [11,20]. However, capital-intensive companies may also have more pollution-abatement equipment and emit less pollution [23,32]. Therefore, we can infer that import tariff reduction in CAFTA can affect firms' pollution emissions through composition effect.
Third, import tariff reduction in CAFTA can change firms' production scale, thus firms' pollution emissions will be changed. Trade liberalization can influence the environment through the way of changing the scale of economic activities [12]. Specifically, import tariff reduction means more goods from other countries can come into the local market, aggravating competition in the local market. Due to fiercer competition, firms will reduce their production scale and pollution emissions. Moreover, firms require intermediate inputs in their production, and import tariff reduction enables firms to obtain the similar intermediate inputs at a lower cost [33]. When the import tariff decreases, firms will import more intermediates from other countries. This will make the enterprises expand their scale in production and have higher emission level. Therefore, we can conclude that import tariff reduction in CAFTA can affect the firms' pollution emissions through scale effect.

Import Tariff in CAFTA and Firms' Pollution Emissions
To analyze the impacts of import tariff reduction in CAFTA on enterprises' emissions, we first use China's import tariffs from ASEAN on final goods and intermediate inputs as core variables. Before starting our analysis, we have studied existing literature which studies China's problems from the firm-level perspective. We find that the data of Chinese micro enterprises have the industrial characteristics (industrial policy can change the industrial supply and demand), provincial characteristics (different provincial policy in China can affect the production and pollution emissions of firms) and time characteristics (the changing of environmental policy in different period can affect firms' pollution emissions). Therefore, drawing on the experience of previous scholars studying China's problems from a firm-level perspective [34][35][36], we choose to use fixed effect model which controls the industry fixed effect, the year effect and the region fixed effect in our analysis. The basic model is as follows: where E ijst is the pollution emissions of manufacturing firm i from industry j in province s at year t, including two pollutants of SO 2 and Soot. t

ASEAN− f ina l jt
and t ASEAN−int jt are China's import tariffs from ASEAN on final goods and intermediate inputs. X includes the control variables in our analysis. δ j reflects industry fixed effect which captures characteristics at the industry level, such as policy that can change the industrial supply and demand. ω t reflects year fixed effects capturing characteristics in different periods, such as the changing of environmental policy in different period that can affect firms' pollution emissions. µ s is province fixed effect which captures characteristics at provincial level, such as provincial policy that can affect the production and the pollution emissions of the firms. ε ijst is the error term capturing some specific changes in pollution emissions.
Our analysis mainly focuses on the coefficient α 1 of the core variable t To explore other factors affecting the importers' emission level, some other control variables are added into our empirical model. TFP ijst is the total factor productivity, reflecting firms' technology in production. Our analysis employs LP method to calculate firms' TFP. LP method of calculating TFP is first use by Levinsohn and Petrin [37], estimating production functions by using inputs to control for unobservables. Productivity is an important factor in reducing pollution emission, firms with higher productivity tend to use pollution-abatement technologies in reducing pollution emissions. KL ijst , calculated by the proportion of net value of fixed assets to employees' number, reflects enterprises' capital intensity. KL ijst represents the composition of factor inputs in production. Add ijst is enterprises' industrial added value reflecting the scale of production. Wage ijst is the main business wage at firm level. In addition, MFN tariffs are also important control variables that cannot be ignored. Existing literature uses both FTA tariffs and MFN tariffs when analyzing the problems of regional trade liberalization [19]. Therefore, our analysis adds two other control variables t

Mechanism Tests
To explain potential mechanisms of enterprises' emissions, our analysis explores three effects (There may be other potential effects that can affect firms' pollution emissions when China's import tariff from ASEAN decreases. Our models analyze these problems by using three effects model. If other data that is available in the future, we will use them for our future research.), including technique, composition and scale effects, to show how import tariff reduction in CAFTA affects firms' emission behaviors.

Tariff Reduction in CAFTA on Final Goods
We examine the three effects of tariff reduction on final goods in CAFTA on firms' emission pollution. We use the indicators of China' import tariff from ASEAN on final goods and variables of three effects to construct interaction terms as three effects. Our model of the three effects of tariff reduction on final goods includes all the control variables in Equation (2). The details of the model are as follows: First, the decrease of China's import tariff from ASEAN on final goods will lead to the change of the firms' TFP, thus it will influence the firms' pollution emissions. We use the interaction item t ASEAN− f ina l jt × TFP ijst to exam the technique effect caused by the change of China's import tariff from ASEAN on final goods. The coefficient α 3 of the interaction item t ASEAN− f ina l jt × TFP ijst represents the change of firms' emissions because of the change of production technology caused by the increase of China's import tariff from ASEAN on final goods.
Second, the decrease of China's import tariff from ASEAN on final goods will make the firms change their factor proportions. The change of the firms' factor proportions can also change the firms' pollution emissions. Therefore, we use the interaction item t ASEAN− f ina l jt × ln KL ijst to exam the composition effect caused by the change of China's import tariff from ASEAN on final goods. The coefficient α 4 of the interaction item t ASEAN− f ina l jt × ln KL ijst represents the change of firms' emissions because of the change in the composition of factor inputs caused by the increase of China's import tariff from ASEAN on final goods.
Finally, the decrease of China's import tariff from ASEAN on final goods will lead to the change of the firms' production scale, thus it will also influence the firms' pollution emissions. We use the interaction item t ASEAN− f ina l jt × ln Add ijst to examine scale effect due to change of the import tariff of China from ASEAN on final goods. The coefficient α 5 of the interaction item t ASEAN− f ina l jt × ln Add ijst represents the change of firms' emissions caused by the impacts of China's tariff changes of final goods from ASEAN on the scale of production.

Tariff Reduction in CAFTA on Intermediate Inputs
We examine the three effects of tariff reduction on intermediate inputs in CAFTA on firms' emission pollution. We use the indicators of China' import tariff from ASEAN on intermediate inputs and variables of three effects to construct interaction terms as three effects. Our model of the three effects of tariff reduction on intermediate inputs includes all the control variables in Equation (2). The details of the model are as follows: First, the decrease of China's import tariff from ASEAN on intermediate inputs will lead to the change of the firms' TFP, thus it will influence the firms' pollution emissions. We use the interaction item t ASEAN−int jt × TFP ijst to exam the technique effect caused by the change of China's import tariff from ASEAN on intermediate inputs. The coefficient α 3 of the interaction item t ASEAN−int jt × TFP ijst represents the change of firms' emissions because of the change of production technology by the increase of China's import tariff from ASEAN on intermediate inputs.
Second, the decrease of China's import tariff from ASEAN on intermediate inputs will make the firms change their factor proportions. The change of the firms' factor proportions can also change the firms' pollution emissions. Therefore, we use the interaction item t ASEAN−int jt × ln KL ijst to exam the composition effect caused by the change of China's import tariff from ASEAN on intermediate inputs. The coefficient α 4 of the interaction item t ASEAN−int jt × ln KL ijst represents the change of firms' emissions because of the change in the composition of factor inputs caused by the increase of China's import tariff from ASEAN on intermediate inputs.
Finally, the decrease of China's import tariff from ASEAN on intermediate inputs will lead to the change of the firms' production scale, thus it will also influence the firms' pollution emissions.

Data
To conduct our analysis, we mainly use three datasets: Chinese manufacturing enterprise-level data, firms' emissions data and the data of China's import tariffs from ASEAN on final goods and intermediate inputs Chinese manufacturing enterprise-level data is from China's Industrial Firm Database, but the authenticity of China's Industrial Firm Database beginning from 2008 has been questioned, so we just use the data from 2002 to 2007, and do not use the data from 2008 onwards. After matching the three datasets, we get a new dataset which is an unbalanced panel. This new panel contains the enterprises' emissions data, the firms' characteristics and China's import tariffs from ASEAN on final products and intermediates from 2002 to 2007. Other details about data processing and descriptive statistics are shown as follow:

Firm Data
The firms' data we use in this paper comes from the China's Industrial Firm Database and from Chinese Customs Database. We match these two databases and we finally get a new database of manufacturing firms' characteristics. The enterprises of this new database are import enterprises with the annual output value that is more than 5 million yuan. The import firms come from 37 manufacturing industries based on the 2-digit industry code of GB/T4754-2002 (Data sources: The data is from China's industry classification standard.). In order to eliminate the outliers, we further process our manufacturing firms' data. (1) Because of the lack of sales data and other firms' characteristics in 2004, we delete the data from 2004. (2) Because six provinces including Ningxia, Qinghai, Guizhou, Hainan, Tibet, Inner Mongolia, have a small trade volume with ASEAN, we remove the date from these six provinces.

Pollution Emission Data
Our analysis uses two different pollutants, including SO 2 and Soot which are China's main pollutant emissions. Because there is not enough Chinese enterprises' emissions data, our paper uses the output value of enterprise products and pollution emission coefficient to represent the level of enterprises' emissions (Data sources: The industrial pollution emission data is from China Environmental Statistical Yearbook, the output value of industries comes from China Statistical Yearbook.). This coefficient is calculated by the proportion of pollution emissions to output value for a given industry and year. The samples in our paper are from China's Industrial Firm Database. Because firms in this database are all large size firms whose output values are more than 5 million yuan, we can assume that these large-scale firms from the same industry have the same pollution emission technology and intensity [23,33]. It means that the firms from the same industry have the same pollution emission coefficients. In addition, these pollution emission coefficients vary from 2002 to 2007 because of technical progress. Thus, the pollution emissions of the firms are determined by their output values, periods and the industries they are in.

Tariff Data
China Free Trade Zone Service Network database provides China's import tariff data from ASEAN of 6-digit HS code. WTO's Tariff Download Facility database provides 2002-2007 China's MFN tariff data of six-digit HS code. According to Mao and Sheng [38], we first use the conversion tables from the United Nations to convert tariff data to the HS2002 version, then we combine the HS2002-ISIC (Rev3) and GB/T2002-ISIC (Rev3) conversion tables to build direct linkage between products and industries. And we use this linkage to calculate China's import tariffs from ASEAN on final goods and China's MFN tariff on final goods at industry level. Then, we calculate the weighted China's import tariff on intermediate inputs from ASEAN and the weighted MFN tariff on intermediate inputs by using Chinese Input-Output Table: t

Descriptive Statistics
Our main variables' descriptive statistics grouped by different technology intensity, degrees of pollution, ownerships and regions are presented in Table 1. According to Table 1, different types of enterprises have different enterprises' characteristics and pollution levels, facing different degrees of tariff reduction. First, for enterprises of different technology intensity, there are fewer technology-intensive enterprises, and these enterprises' pollution emissions are lower relatively. These enterprises also have higher productivity and lower capital intensity. Compared to non-technology-intensive enterprises, the emission level of technology-intensive enterprises are much lower. Additionally, technology-intensive enterprises also have larger scale in production and higher wages.
Second, for ownership types, private-owned and foreign-invested enterprises have lower pollution emissions relatively. State-owned enterprises have bigger production scale and higher capital intensity. Furthermore, state-owned enterprises have higher productivity and lower wages than private-owned and foreign-invested enterprises.
Third, for provinces in different regions, eastern enterprises have lower emission level relatively. Eastern enterprises have lower productivity and lower production scale. Compared to eastern and central enterprises, western enterprises have higher capital intensity. Moreover, central enterprises have higher wages than enterprises in eastern and western provinces. Table 2 shows the results of estimates based on Equations (1) and (2), reporting the effects of China's import tariff reduction from ASEAN on Chinese importers' pollution emissions. First, we analyze effects of import tariff reduction in CAFTA on importers' emissions. The pollutants reported in columns (1)-(3) and (4)-(6) are SO 2 and Soot, respectively. Columns (1), (3), (4) and (6) control industry, province and time fixed effects. Columns (2) and (5) are the results of using random effect model which does not control industry, province and time fixed effects. Columns (1) and (4) are the regression results of only the change of import tariffs of final products and intermediates. Columns (3) and (6) are our results of our basic model after adding the control variables mentioned in the Equation (2) to the columns (1) and (4). In our analysis, we have done the Hausman test before analyzing. The results of Hausman test showed that using fixed effect model is more effective than using random effect model. By comparing the results of columns (2) and (5) (the random effect model) and columns (3) and (6) (our basic model), we can find that the signs and significance of some coefficients are totally different. These results mean that if we do not control industry, province and time fixed effects, some industrial, provincial and time characteristics cannot be effectively considered in the analysis, which will lead to biased conclusions. Therefore, all the following models in our article are fixed effect models and control industry, province and time fixed effects.

Tariff Reduction and Firms' Pollution Emissions
Next, we study columns (1) and (4) which do not add the control variables in Equation (2). From columns (1) and (4), the coefficient on China's import tariff from ASEAN on final goods is positive and significant at 1% level for SO 2 , but negative and significant at 1% level for Soot. This means the decrease of China's import tariff from ASEAN on final goods causes the decrease in SO 2 emissions by 0.0228% and the increase in Soot emissions by 0.0187%. Also from columns (1) and (4), the estimated coefficient on China's import tariff from ASEAN on intermediate inputs is negative, and significant at 1% level. This shows decrease of import tariff from ASEAN on intermediates will cause the increase in SO 2 and Soot emissions by 0.0554% and 0.0352%, respectively. Furthermore, these coefficients change when the control variables are added. Therefore, if our analysis ignores other potential influencing factors, we will get the wrong conclusions. From columns (3) and (6), the estimated coefficient on China's import tariff from ASEAN on final goods is positive, and significant at 1% level, suggesting that decrease of China's import tariff from ASEAN on final goods causes the decrease in SO 2 and Soot emissions by 0.0477% and 0.0111%. More and more products enter the domestic market when the import tariff on final goods in CAFTA decrease, and this will lead to technology spillover. In this situation, firms can improve productivity and thus reduce their emissions. The estimated coefficient on China's import tariff from ASEAN on intermediate inputs is negative, and significant at 1% level, showing that decrease of China's import tariff from ASEAN on intermediate inputs aggravates SO 2 and Soot emissions by 0.0398% and 0.0431%. Reduction of China's import tariff from ASEAN on intermediate inputs means firms can get more intermediates at the same cost, and this will make the firms to enlarge their scale in production and increase pollution emissions. This result of tariff reduction in FTA on intermediate inputs stands in contrast to findings present by Cherniwchan [13]. In contrast to our research analyzing the environmental impacts of South-South trade liberalization, the analysis of Cherniwchan [13] studies the effects of an episode of North-South trade liberalization on the emissions of manufacturing firms in North (the U.S.), suggesting that import tariff reduction on intermediates input in NAFTA reduces the American firms' emissions. Manufacturing firms from the U.S. have higher productivity level and technology in production. In this context, intermediate inputs in the production process are not as important for American firms as for Chinese firms. The cost reduction in production brought by import tariff reduction to American firms is far less important than that to Chinese firms. Thus, in the U.S., import tariff reduction on intermediate inputs in FTA cannot effectively stimulate American manufacturing firms to enlarge their scale in production, thus their pollution emissions do not increase.
Finally, we focus on the coefficients of the control variables in Table 2. (1) The enterprises' total factor productivity is negatively correlated with their emissions. This finding is similar to the conclusions of previous studies [30,39], suggesting that enterprises with higher productivity have lower level of pollution emissions. According to Cui et al. [31], firms with higher productivity tend to use abatement technologies in production to reduce pollution emissions. (2) The firms' capital intensity is positively correlated with their emission level. Enterprises with higher capital intensity have a greater tendency to enlarge their production, therefore they will have higher emission level.

Robustness Check
To ensure the robustness and reality of our findings in Table 2, we take additional regressions. Our research is to analyze how import tariff reduction in CAFTA affects firms' pollution emissions. The samples we use include some firms that have import less from ASEAN. Thus, these firms are less affected by import tariff reduction in CAFTA. Adding these firms into our analysis may lead to a misleading conclusion. To prevent these firms from affecting the estimates, we re-estimate the Equation (1) and (2) by removing the firms with the import volumes from CAFTA that are less than 300,000 dollars from our samples. Table 3 reports our results of robustness test. To compare the results more conveniently, columns (1) and (3) are listed as the regression results in columns (3) and (6) of Table 2. Columns (2) and (4) are the results based on observations without firms that import less than 300,000 dollars from ASEAN. After dropping the samples of these firms, the signs of the results are still the same, and almost all the coefficients are still significant at the same level. These suggest that the results of our models are robust.   Significance: *** 1%, ** 5%, * 10%.

Mechanism Test
Altogether, the above results show that the reduction of China's import tariff from ASEAN on final goods reduces Chinese importers' pollution emissions, while the reduction of China's import tariff from ASEAN on intermediate inputs exacerbates the pollution emissions of China's importers. Then, we explore the specific mechanisms of the impacts of tariff reduction in CAFTA on firms' pollution emissions, which includes technique, composition and scale effects. The estimated results based on Equations (3) and (4) is presented in Table 4, and our models of mechanism tests include all the control variables in Equation (2).

Tariff Reduction on Final Goods and Firms' Pollution Emissions
First, we discuss technique effect on enterprises' emissions. From columns (1) and (2) in Table 4, the coefficient on t ASEAN− f ina l jt × TFP ijst is positive and significant at 1% level, meaning that technique effect caused by the decrease in import tariff on final goods in CAFTA has decreased the firms' pollution emissions. Through technique effect, firms can reduce SO 2 and Soot emissions by 0.0136% and 0.0183% when China's import tariff from ASEAN on final goods decrease. The possible reason is that more and more products enter the domestic market when the import tariff on final goods in CAFTA decrease, and this will aggravate domestic market competition [40] and force the firms to improve their production technology. Moreover, the reduction of import tariff on final goods can also lead to technology spillover, and thus the firms can improve productivity and reduce their emissions. Thus, through technique effect caused by the decrease in import tariff on final goods in CAFTA, the firms can reduce their pollution emissions.
Secondly, we discuss the composition effect on enterprises' emissions. From columns (1) and (2) in Table 4, the coefficient on t ASEAN− f ina l jt × ln KL ijst is positive and significant at 1% level, meaning that composition effect caused by the decrease in import tariff on final goods in CAFTA has decreased the firms' pollution emissions. Through composition effect, firms can reduce SO 2 and Soot emissions by 0.0009% and 0.0020% when China's import tariff from ASEAN on final goods decrease. The factors of production are important in firms' production, thus firms will change their factor input compositions to adapt to the new market environment. The import tariff reduction in CAFTA leads to the change of market environment. In order to adapt to the change of market environment, firms change their factor input compositions and reduce pollution emissions. Thus, through composition effect caused by the decrease in import tariff on final goods in CAFTA, the enterprises' pollution emissions will decrease because of the change of factor input compositions.
Thirdly, we discuss the scale effect on enterprises' emissions. From columns (1) and (2) in Table 4, the coefficient on t ASEAN− f ina l jt × ln Add ijst is negative and significant at 1% level, meaning that scale effect caused by the decrease in import tariff on final goods in CAFTA will aggravate the firms' pollution emissions. Through scale effect, firms increase SO 2 and Soot emissions by 0.0115% and 0.0143% when China's import tariff from ASEAN on final goods decrease. The products that China imports from ASEAN due to the import tariff reduction in CAFTA are mainly low-end products. According to Men [41], China has a complete product structure which covers from low-end to high-end, so China and ASEAN are strongly complementary in many products. Thus, when more complementary products are imported from foreign countries, firms in China will enlarge their scale in production and have higher emission level. Through scale effect caused by the decrease in import tariff on final goods in CAFTA, the enterprises' pollution emissions will increase.
Altogether, technique and composition effects caused by the decrease in import tariff on final goods in CAFTA have decreased the firms' pollution emissions. First, we should make good use of the technique effect by import tariff reduction on final goods in CAFTA to decrease firms' pollution emissions. Although the reduction of tariff on final goods in CAFTA will aggravate domestic market competition, which has brought great survival pressure to Chinese domestic firms. However, in the long term, these firms may be forced to improve their production technology and reduction their emissions. Therefore, these firms should not be overprotective when China's import tariff on final goods from ASEAN decreases. Moreover, we should also take good advantage of the composition effect by import tariff reduction on final goods in CAFTA to help the firms optimize the allocation of production factors, thus decrease firms' pollution emissions.

Tariff Reduction on Intermediate Inputs and Firms' Pollution Emissions
We first discuss the technique effect on enterprises' emissions. From columns (3) and (4) in Table 4, the coefficient on t ASEAN−int jt × TFP ijst is positive and significant at 1% level, meaning that technique effect caused by the decrease in import tariff on intermediate inputs in CAFTA has decreased the firms' pollution emissions. Through technique effect, firms can reduce SO 2 and Soot emissions by 0.0157% and 0.0214% when China's import tariff from ASEAN on intermediate inputs decrease. The possible reason is that firms can easily obtain more high-quality intermediate inputs from ASEAN due to the reduction of import tariff in CAFTA. This will lead to technology spillover. Thus, the firms' productivity may be improved and their pollution emissions can be reduced [33]. Thus, through technique effect caused by the decrease in import tariff on intermediate inputs in CAFTA, the firms can reduce their pollution emissions.
Then, we discuss the composition effect on enterprises' emissions. From columns (3) and (4) in Table 4, the coefficient on t ASEAN−int jt × ln KL ijst is positive and significant at 1% level, meaning that composition effect caused by the decrease in import tariff on intermediate inputs in CAFTA has decreased the firms' pollution emissions. Through composition effect, firms can reduce SO 2 and Soot emissions by 0.0022% and 0.0055% when China's import tariff from ASEAN on intermediate inputs decrease. Intermediate inputs, which are important in firms' production, can also affect firms' factor input compositions. The Chinese importers can obtain diversified intermediate inputs when more and more intermediates from ASEAN enter local market due to the tariff reduction in CAFTA. Importers will change their factor input compositions because of their change of intermediate inputs, and reduce their pollution emissions. Thus, through composition effect caused by the decrease in import tariff on intermediate inputs in CAFTA, the enterprises' pollution emissions will decrease because of the change of factor input compositions.
Next, we discuss the scale effect on enterprises' emissions. From columns (3) and (4) in Table 4, the coefficient on t ASEAN−int jt × ln Add ijst is negative and significant at 1% level, showing that scale effect caused by the decrease in import tariff on intermediate inputs in CAFTA will aggravate the firms' pollution emissions. Through scale effect, firms will increase SO 2 and Soot emissions by 0.0131% and 0.0157% when China's import tariff from ASEAN on intermediate inputs decrease. The possible reason is that production of the firms is closely related to the cost of intermediate inputs. This can lead to technology spillover and may help firms in China improve their technology in production. Therefore, firms will emit less pollution. Moreover, the composition effect by import tariff reduction on intermediate inputs in CAFTA is also a major mechanism to help firms reduce emissions. Intermediates are important in the production of firms. Thus, we could use the import tariff reduction on intermediates to help firms reduce the production costs and optimize the allocation of production factors. This can effectively help them reduce the pollution emissions.

Heterogeneous Results
The above analysis explores these three effects of import tariff reduction in CAFTA on Chinese importers' pollution emissions. Since China's development is at a critical juncture, we further differentiate our samples from the perspectives of technology intensity, ownership types and provinces to analyze the heterogeneous impacts. Tables 5-7 shows the results of heterogeneous impacts on industries of different technology intensity, ownership types and regions, respectively. Each Panel A in these tables shows the estimated results based on Equations (1) and (2). Each Panel B shows the estimated results based on Equations (2) and (3). Each Panel C shows the estimated results based on Equation (3) and (4). All the models of analyzing the heterogeneous impacts includes all the control variables in Equation (2), and we how the most important results of our estimate in Tables 5-7.    Notes: Table presents the impacts of import tariffs on final products and intermediates of the emissions of SO 2 and Soot from China manufacturing firms. The dependent variable of each column is the log of firms' emission level. Fixed effects include industry, province and time fixed effects. The figures in parentheses are standard errors. Significance: *** 1%, ** 5%, * 10%.

Industries of Different Technology Intensity
Considering Chinese upgrades of manufacturing industries, we explore the heterogeneous effects of import tariff reduction in CAFTA on the importers from industries with different technology intensity. For a long period, Chinese manufacturing industries have comparative advantages in producing low-end goods which require less technical inputs in production [42], but these industries have achieved low profits and are not conductive to the sustainable development of economy. Technology-intensive industries are important in the upgrades of Chinese manufacturing industries in the future. Thus, considering the technology intensity, the samples are divided into the importers in technology-intensive industries and non-technology-intensive industries to re-estimate our results. Our analysis uses the proportion of R&D expenditure to main business revenue (Data sources: This data is from China Statistical Yearbook on Science and Technology.) to differentiate between technology-intensive and non-technology-intensive industries. Table 5 presents our results of enterprises in industries of different technology intensity.
From panel A, China's import tariff on final goods from ASEAN is positively correlated with the enterprises' emissions from technology-intensive and non-technology-intensive industries, which suggests that tariff reduction on final goods in CAFTA reduces Chinese importers' emissions. From panel B, import tariff reduction on final goods in CAFTA causes different effects on importers from industries of different technology intensity. For technology-intensive industries, the estimated coefficient on t ASEAN− f ina l jt × TFP ijst is negative. This result suggests that firms in technology-intensive industries cannot reduce the pollution emissions through technique effect when the import tariff on final products in CAFTA decreases. In addition, the composition effect of import tariff reduction on final products in CAFTA is similar to the effect of the whole samples. But the estimated coefficient on t ASEAN− f ina l jt × ln Add ijst is positive. For a long period, China's economic development relies on manufacturing-based industries but not technology-intensive industries, so technology-intensive industries are still developing and are not sufficiently strong [43]. When China's import tariff from ASEAN on final goods decrease, more products imported from ASEAN enter domestic market, increasing the competition in domestic market, which may lead the importers from technology-intensive industries to reduce the scale in production and have a lower emission level. For firms in non-technology-intensive industries, the technique effect and scale effect of import reduction on final products in CAFTA have certain similarities with the whole samples' effects. But for firms in non-technology-intensive industries, the composition effects are different when considering different pollutants. For SO 2 , the composition effect is contrary to the effect of the whole samples. As for Soot, the composition effect has some similarity with the effect of the whole samples, but it is not significant.
Also from panel A, China's import tariff on intermediate inputs is positively correlated with the emissions of technology-intensive industries, but negatively correlated with the emissions of non-technology-intensive industries, indicating that for firms in technology-intensive industries, import tariff reduction on intermediate inputs in CAFTA reduces their pollution emissions. From panel C, import tariff reduction on intermediate inputs in CAFTA causes unequal effects on importers from industries of different technology intensity. For technology-intensive industries, the estimated coefficients on t ASEAN−int jt × TFP ijst of SO 2 and Soot are both not significant, which suggests that firms in technology-intensive industries cannot reduce pollution emissions through technique effect when the import tariff on intermediates in CAFTA decreases. In addition, the composition effect of import tariff reduction on intermediates in CAFTA is similar to the effect of the whole samples. Specifically, for SO 2 emissions, the scale effect has some similarity with the effect of the whole samples, but for Soot emissions, the scale effect has adverse impacts but not significant. For firms from non-technology-intensive industries, the three effects are almost similar to the original results of the whole sample, except that the estimated coefficient on t ASEAN−int jt × ln KL ijst of SO 2 is not significant.
Altogether, these heterogeneous impacts refer to the upgrades of Chinese manufacturing industries. First, as the industries that have higher profits and are conductive to sustainable development, the technology-intensive industries are important in the upgrades of manufacturing industries in China [23]. The reduction of import tariff on final goods and intermediates in CAFTA can both effectively reduce these Chinese importers' pollution emissions. Because most enterprises in the technology-intensive industries are processing enterprises, suggesting that the technology-intensive industries in China are not sufficiently strong [43]. The reduction of import tariff on final goods and intermediates in CAFTA cannot effectively help them reduce their pollution emissions through technique effect. Therefore, it is essential to take advantage of the potential of these technology-intensive industries, helping them to improve their efficiency and competitiveness. Second, the import tariff reduction on final goods can reduce the pollution emission of the non-technology-intensive industries, but the import tariff reduction on intermediate inputs has the adverse impacts on these industries. Import tariff reduction in CAFTA can effectively help firms from non-technology-intensive industries to reduce pollution emissions through technique effect. Moreover, non-technology-intensive industries emit more pollution. Thus, we should take good advantage of these industries' technique effect by import tariff reduction in CAFTA to decrease their pollution emissions.

Different Ownership Types
From the perspective of Chinese ownership reform, we analyze the heterogeneous impacts of different types of ownership importers affected by the import tariff reduction in CAFTA. In China, the important parts in the process of ownership reforms are keeping public ownership as the main body of the economy and allowing various forms of ownership to develop together. According to Wang and Wheeler [44], the ownership of the firms could be a primary determinant of pollution. Thus, considering the uniqueness of Chinese ownership type, the samples are divided into state-owned, private-owned and foreign-invested enterprises, to explore the impacts of import tariff reduction in CAFTA on the process of ownership reforms. The results of the enterprises with different ownership are reported in Table 6.
From panel A, China's import tariff on final goods from ASEAN is positively correlated with the emissions of private-owned importers and the foreign-invested importers, which means that these two kinds of firms can effectively reduce their emissions when the import tariff on final products in CAFTA decreases. For state-owned firms, China's import tariff on final goods from ASEAN is positively correlated with the SO 2 emissions, but negatively correlated with the Soot emission, suggesting that when the import tariff on final products in CAFTA decrease, state-owned firms tend to emit less SO 2 and more Soot. From panel B, import tariff reduction on final goods in CAFTA causes different effects on importers with different ownership types. For state-owned enterprises, three effects of the tariff reduction on final goods are not significant. In China, the activities of state-owned enterprises are under the control of the government in many fields [45], thus this type of firms' technology, factor composition and production scale can hardly be influenced by tariff reduction in CAFTA. For private-owned importers and foreign-invested importers, the results of the three effects by the tariff reduction on final goods in CAFTA are similar to the results of the whole samples. By further comparison, the technique effect and composition effect of foreign-invested enterprises are less than those of private-owned enterprises, which suggests that unlike foreign-invested firms with stable technology and supplies, private-owned enterprises are easier to be influenced by import tariff reduction on final goods in CAFTA.
Also from panel A, for SO 2 , China's import tariff on intermediate inputs from ASEAN is both negatively correlated with the emissions of these three types of firms. But the coefficients on t ASEAN−int jt of state-owned and private-owned enterprises are not significant. For Soot, China's import tariff on intermediate input from ASEAN is negatively correlated with non-owned-stated enterprises and foreign-invested enterprises, which suggests that these two types of firms will increase their Soot emissions if import tariff on intermediate inputs in CAFTA decreases. From panel C, for state-owned importers, three effects by the tariff reduction on intermediate inputs are not so significant. In China, the activities of state-owned enterprises are under the control of the government in many fields and they do not have enough flexibility [45]. For private-owned and foreign-invested enterprises, the results of three effects by the tariff reduction on intermediate inputs are similar to the results of the whole samples. These indicate that when tariff on intermediate inputs decrease, private-owned firms and foreign-invested-firms can reduce their pollution emissions through technique effect and composition effect.
Altogether, the heterogeneous effects are related to Chinese ownership reforms. First, import tariff reduction on final goods in CAFTA reduces state-owned importers' SO 2 emissions, but increases their Soot emissions. Additionally, state-owned enterprises are under control of the government in many fields, so they can hardly change their technology, factor composition and scale in production. State-owned enterprises are important in Chinese economy, and we should pay great attention to the reform of state-owned enterprises. Secondly, private-owned and foreign-invested enterprises play important role in ownership reform. Import tariff reduction on final goods in CAFTA reduces their pollution emissions. Moreover, import tariff reduction in CAFTA can effectively help these two types of firms reduce pollution emissions through technique effect and composition effect. Therefore, while giving support to the development of private-owned firms and foreign-invested firms in the process ownership reform, we should also take good advantage of these firms' technique effect and composition effect by import tariff reduction in CAFTA to decrease their pollution emissions.

Different Regions
From the perspective of the regional coordinated development of different provinces, we explore the heterogeneous effects of import tariff reduction in CAFTA on the importers from different regions. In China, different provinces play different roles in trade, and firms from different regions have different location advantage. Moreover, the importance of balanced development is highly valued in China [46,47]. Thus, from the perspective of the unbalanced development of provinces in different regions, the samples are divided into importers from eastern, central and western provinces, to analyze effects of import tariff reduction in CAFTA on regional coordinated development. Results of the enterprises in different provinces are presented in Table 7.
From panel A, for SO 2 , China's import tariff on final goods from ASEAN is positively correlated with the emissions of enterprises in eastern and western provinces, suggesting that the tariff reduction on final goods in CAFTA can effectively reduce the SO 2 emissions of these enterprises. By further comparison, the impacts by import tariff reduction in CAFTA on eastern provinces are greater than those impacts on western provinces. For Soot, China's import tariff on final goods from ASEAN is positively correlated with the emissions of enterprises in eastern provinces, but is negatively correlated with the emissions of enterprises from western provinces. From panel B, the tariff reduction on final goods in CAFTA causes different effects on these three types of importers. For SO 2 , both firms from eastern and central provinces reduce pollution emissions through technique effect by the tariff reduction on final goods in CAFTA. However, import tariff reduction on final goods does not significantly have impacts on firms from western provinces through technique, composition and scale effect. For Soot, both firms from eastern, central and western provinces reduce pollution emissions through technique effect by the tariff reduction on final goods in CAFTA. By further comparison, the technique effect by import tariff reduction in CAFTA on firms from central provinces is greater than those of effects of firms from eastern and western provinces. Moreover, tariff reduction of final products in CAFTA can reduce Soot emissions of firms from eastern and western provinces by composition effects. The composition effect of firms from western provinces is greater than the effect of firms from eastern provinces.
Also from panel A, China's import tariff on intermediate inputs from ASEAN is negatively correlated with the emissions of enterprises in eastern, central and western provinces, but the effects on SO 2 emissions of central provinces are not significant. This indicates that all the three types of importers will increase pollution emissions when import tariff on intermediate inputs in CAFTA decrease. Furthermore, the effects on enterprises from western provinces are greater than those effects on enterprises from eastern and central provinces. From panel C, firms from eastern and central provinces can reduce pollution emissions through the technique effect by the tariff reduction on intermediate inputs in CAFTA. In addition, firms from eastern provinces can also reduce pollution emissions through the composition effect by import tariff reduction on intermediate inputs in CAFTA, but increase pollution emissions through scale effect. However, most of the three effects by import tariff reduction on intermediates in CAFTA on firms from western provinces are not significant, indicating that due to geographical location, it is difficult for firms from western provinces to make good use of the preferential policies of CAFTA.
Altogether, the heterogeneous effects are related to China's regional coordinated development of different provinces. First, the eastern provinces are in a leading position in China's economic development. Therefore, these firms have the advantage of economic development and are important to promote Chinese regional coordinated development. The import tariff reduction on final goods in CAFTA can effectively reduce the pollution emissions of firms from eastern provinces. Moreover, these firms can reduce pollution emissions through technique effect and composition effect by import tariff reduction in CAFTA. Thus, we should draw on the developing experience of the eastern provinces to reduce emissions and promote regional coordinated development. Second, the impacts of import tariff reduction in CAFTA can effectively reduce the pollution emissions of firms from central countries through technique effect. Therefore, we should make good use of the technique effect of import tariff reduction in CAFTA to reduce emissions from firms in central provinces. Third, due to the geographical location, the three effects of firms from western provinces by import tariff reduction in CAFTA is not so significant as the effects of firms from eastern and central provinces. In order to reduce the imbalanced development in different regions, we should pay attention to letting the firms from western provinces make good use of the preferential policies of CAFTA.

Conclusions
As the third largest Free Trade Area and the most populous Free Trade Area in the world, CAFTA receives more and more attention. To date, research has mainly focused on the effects of CAFTA in many areas, such the firms' productivity, poverty reduction effect and economic growth. Yet, there has been little evidence of how import tariff reduction in CAFTA affects the firm-level pollution emissions.
In this paper, we rely on the data of 126,606 samples from Chinese manufacturing and examine the impacts of the import tariff reduction in CAFTA on firms' pollution emissions. Then, we explore and investigate the specific mechanism of the impacts through technique, composition and scale effects. Our analysis also investigates the impacts of import tariff reduction in CAFTA on firms' emission behaviors from the perspective of heterogeneous importers.
Our results suggest: (1) Import tariff reduction on final goods in CAFTA reduces Chinese importers' pollution emissions. However, import tariff reduction on intermediate inputs in CAFTA aggravates Chinese importers' pollution emissions. (2) Import tariff reduction in CAFTA affects firms' pollution emissions through the mechanism of technique, composition and scale effects. In addition, the technique effect and composition effect reduce firms' pollution emissions. (3) Import tariff reduction in CAFTA has heterogeneous impacts on the importers from industries of different technology intensity, which refers to the upgrades of manufacturing industries in China. In particular, import tariff reduction in CAFTA on final goods and intermediate inputs can both reduce the pollution emissions of firms from technology-intensive industries. (4) Import tariff reduction in CAFTA has heterogeneous impacts on the importers with different ownership types, which refers to the ownership reforms in China. In particular, state-owned firms are harder to be influenced through the three effects by import tariff reduction in CAFTA. (5) Import tariff reduction in CAFTA has heterogeneous impacts on firms from different regions, which refers to the regional coordinated development of different provinces in China. In particular, firms in western provinces are harder to be influenced through the three effects by import tariff reduction in CAFTA.
Altogether, our analysis provides evidence on the impacts of import tariff reduction in CAFTA on Chinese importers' pollution emissions. Moreover, our analysis also has some implications as following.
(1) The development of CAFTA promotes China's economic growth, at the same time, it can also have some impacts on the firms' pollution emissions in China. Therefore, while strengthening the links with ASEAN, China must pay more attention to environmental protection. (2) Import tariff reduction in CAFTA on final goods and intermediate inputs can both reduce the firms' pollution emissions through technique effect and composition effect. Thus, it is essential to make full use of technique and composition effect by import tariff reduction in CAFTA to reduce pollution emissions. (3) For the upgrades of Chinese manufacturing industries, the technology-intensive industries are conductive to sustainable development, but they are not sufficiently strong. It is important to take advantage of the potential of technology-intensive industries, helping them to improve their efficiency and competitiveness. It is also important to take good advantage of the non-technology-intensive industries' technique effects to decrease pollution emissions. (4) For ownership reforms in China, because state-owned firms are harder to be influenced through the three effects by import tariff reduction in CAFTA, it is of vital importance to reforms the state-owned firms. Moreover, while supporting the development of non-state-owned firms and foreign-funded firms in the process ownership reform, we should also take good advantage of these firms' technique effect and composition effect by import tariff reduction in CAFTA to decrease their pollution emissions. (5) For China's regional coordinated development of different provinces, we should draw on the development experience of the eastern provinces to reduce emissions and promote regional coordinated development.
Our analysis mainly contributes to the existing literature examining the impacts of FTA composed of developing countries on environment by giving strong evidence from the micro perspective. We focus on the import side, using the data of import tariff from ASEAN and characteristics of the importers. We still have some other research directions for future research. The tariff reduction in CAFTA may have other impacts on exporters' pollution emissions because China's export tariff to ASEAN also decreases. Moreover, the impacts and mechanisms will be complicated when considering the export status (processing exporters and non-processing exporter) of firms. And these can also be good directions for our future research.