The More the Merrier? Evidence from Firm-Level Exports and Environmental Performance in China

Abstract

Existing literature supports that exporting firms have better environmental performance. An interesting question thereby arises: the more exports, the better? To answer this question, we develop a method to decompose firm-level pollution emissions, and empirically investigate the relationship between export intensity and environmental performance using Chinese firm-level data. Our results indicate that the answer to this question is “No”. First, OLS estimation shows that firms with higher export intensity have less pollution emissions, mainly because of smaller output scale and lower energy intensity (energy-to-labor ratio) rather than more advanced technologies. Second, we focus on PSM-DID estimation and find that only the increase in export intensity by a smaller extent is conducive to improving firms’ environmental performance. This effect is driven by decreasing energy intensity and thereby improving energy efficiency. This finding implies that firms should focus on both domestic and foreign markets, when they improve export participation. Third, those relationships are found to be heterogeneous across the firms in terms of different pollutants, ownership types, industrial sectors and provinces. In particular, mainly for private and foreign-funded firms, technology-intensive sectors and coastal provinces, increasing export intensity can improve environmental performance. Our study provides an in-depth empirical evidence on the relationship between export intensity and environmental performance in China, and provides a new insight and a better understanding for exports and environment from a micro perspective.

Appendices

Appendix A. More Results for OLS Estimation

1.1 A.1. Robustness Check: OLS

To ensure the robustness of the results, we do additional checks. Table 16 reports the results of robustness checks. First, given that our core variable (export intensity) is a continuous variable with the interval of (0,1], we replace it by discrete variables, i.e. the Levels 1-5 defined in Section 3.1.1. Specifically, we regard firms from the Level 1 as the benchmark (omitted group), and examine the differences in firms’ environmental performances between the Level 1 and other Levels. The panel A reports the results by replacing core variable and based on Eq. (8). From the column (1), the estimated coefficients suggest that firms with higher export intensity have less pollution emissions. More specifically, pure exporters have the least emissions, followed by firms from the Level 4. From the columns (2) and (3), firms with higher export participation do not have lower emission intensity, but have smaller output scale. Especially, pure exporters have smallest output. We next decompose emission intensity into emission technology and energy efficiency. From columns (4) and (5), there is no clear correlation between export intensity and both of them. From the column (6), only firms from the Level 5 have lower energy intensity. This result suggests firms heavily depending on the export market input less energies in production. In addition, from the column (7), firms focusing more on the foreign market have lower labor productivity. In particular, pure exporters have lowest productivity, which is in line with the finding of Lu et al. (2014). These are supportive of our basic results.

Table 16 Robustness check: OLS

Second, given that some firms have extremely high or low pollution emissions and these extreme values may lead to the estimation bias, we remove the samples whose pollution emissions rank in the top 5% and bottom 5% to re-estimate the result. Panel B reports this result. Firms with higher export intensity have less pollution emissions. Then, these firms have smaller output scale, more backward emission technology, lower energy intensity and labor productivity²⁷. These firms emit fewer pollutants because of smaller output scale and lower energy intensity. Our results are robust after eliminating extreme samples.

Third, there are 5,071 and approximately 7.13% observations which are pure exporters. On the one hand, pure exporters only sell products to foreign markets, and their export intensity is equal to one and unlikely to change. On the other hand, pure exporters have different characteristics, especially lower productivity, relative to other exporters (Lu et al. 2014). Thus, we remove pure exporters from the samples and re-estimate the result, which is reported in the panel C. Firms with higher export intensity still have less pollution emissions. Then, we find that firms with higher export intensity have smaller output scale, more backward emission technology and labor productivity. Mainly due to the channel of output scale, these firms have less emissions. Slightly different from the result in Table 4, the correlation between export intensity and energy intensity is not significant.

Fourth, considering that the correlation between firms’ export intensity and environmental performance may change in different years, we employ the sub-samples from different periods (2002-2012, 1998-2007 and 2002-2007) to re-estimate the result. These can also eliminate the interferences of two big events, including China’s accession to the WTO in 2001 and the global economic crisis in 2008. The panels D-F report these results. For each period, firms with higher export intensity emit fewer pollutants. In addition, these firms have less emissions mainly due to smaller output scale and lower energy intensity. Our results hold very well after using the sub-samples from different periods.

Fifth, many firms enter and exit the export market in the period of our study. These entrants and exits would affect firms’ performance (Melitz 2003; Cherniwchan 2017). To this end, we employ surviving exporters which stay in the export market from the previous two years to the next two years to re-estimate the result, which is reported in panel G. We still find that firms with higher export intensity have less pollution emissions, mainly because of their smaller output scale and lower energy intensity. Our conclusion is robust after alleviating the impacts of entrants and exits of exporters.

Sixth, some of our dependent variables are measured by output value, including emission intensity, output scale, energy efficiency and labor productivity. Given that manufacturing markups increase rapidly and differentially across firms in China (Lu and Yu 2015; Brandt et al. 2017), there may be a markup-driven bias which interferes our results (Brucal et al. 2019). We thereby employ intermediate input instead of output value to measure those dependent variables. We re-estimate the result, which is reported in the panel H. We find that firms with higher export intensity have less pollution emissions, mainly because of smaller output scale. The coefficient for energy intensity is still negative (although statistically insignificant). These results are similar to the panel E which is also estimated by the data of 1998-2007.

1.2 A.2. The Role of Processing Trade

The results in Section 4.1 indicate that smaller output scale and lower energy intensity are the main reasons for that firms with higher export intensity have less pollution emissions. An important explanation is that in China, firms with higher export intensity are usually processing exporters, and these exporters have lower productivity relative to other exporters (Dai et al. 2016). As a result, these firms have smaller output scale and are labor-intensive. Processing trade may play an important role in the relationship between export intensity and environmental performance.

To identify this role of processing trade, we employ the information on processing trade provided by the China Customs Database from 2000 to 2012. We divide exporting firms into three types: exporters mainly engaged in ordinary trade, exporters mainly engaged in processing trade and pure processing exporters. In particular, we distinguish between exporters mainly engaged in ordinary trade or processing trade by whether the proportion of processing trade exceeds 50%. From our merged dataset, on average, the export intensities of these three types of exporters are 0.4201, 0.5396 and 0.6359 respectively. We can find that pure processing exporters have the highest export intensity, followed by the exporters mainly engaged in processing trade. This is consistent with Dai et al. (2016), that is, firms with higher export intensity are usually processing exporters.

Then, we replace the core variable of Eq. (5) by two dummy variables, i.e. whether the firms are mainly engaged in processing trade ($Process_{ispt}$) and whether the firms are pure processing exporters ($PureProcess_{ispt}$), to examine the correlation between processing trade and environmental performance. Table 17 reports these results. First, from column (1), we can find that processing exporters have less pollution emissions. Especially, pure processing exporters which have the highest export intensity emit the least pollutants. Although the coefficient on $Process_{ispt}$ is not statistically significant, this coefficient is negative and the absolute value is greater than the standard error. We still consider that the pollution emissions of exporters mainly engaged in processing trade is less than those of exporters mainly engaged in ordinary trade. Then, from other columns, processing exporters (especially pure processing exporters) emit fewer pollutants, mainly because of their smaller output scale and lower energy intensity. Besides, processing exporters have relatively backward emission and production technologies. These results are consistent with Dai et al. (2016), i.e. processing exporters are labor-intensive and have backward technology and smaller production scale. In general, these results are similar to the correlation between export intensity and environmental performance (see in Table 4). Thus, processing trade can explain the difference in environmental performance of firms with different export intensities to a certain extent.

Table 17 Processing trade and environmental performance

Appendix B. The Impact of Decreasing Export Intensity

The results in Section 4.2 show that the increase in export intensity by a smaller extent is conducive to improving firms’ environmental performance. It is noted that during the period of this study, many Chinese exporters decrease their export intensity. Since increasing export intensity (by a smaller extent) reduces pollution emissions, does decreasing export intensity aggravate emissions? Therefore, we examine the impact of decreasing export intensity on firm-level environmental performance.

Table 18 reports these results. First, from the panel A, the result indicates that the decrease in export intensity has no significant impact on firms’ pollution emissions and other decomposed variables. Given that different firms decrease export intensity by different extents, we turn to examine the impacts of decreasing export intensity by smaller and larger extents. Consistent with the definition of increasing export intensity, if firms decrease export intensity by one Level, we define that these firms decrease export intensity by a smaller extent. Besides, if firms decrease export intensity by two to four Levels, we define that these firms decrease export intensity by a larger extent. From panels B and C, we find that decreasing export intensity by a smaller or a larger extent cannot significantly affect environmental performance. These results suggest that although firms improve environmental performance after increasing export intensity by a smaller extent or turning to focus on both domestic and foreign markets, firms do not change environmental performance after decreasing export intensity or focusing more on the domestic market.

Table 18 The impact of decreasing export intensity on environmental performance