Trade liberalization and environmental degradation in China

Abstract

Being aware that trade liberalization affects the environment by scale, technique and composition effects, this study uses provincial data from 1993 to 2002 in China to empirically investigate how they affect the environment. Adopting the methodology provided by Antweiler et al. (2001) and slightly modifying their model, this study examines whether the composition effects arising from increasing trade originate due to differences in capital-labour endowment (factor endowment hypothesis) and/or differences in environmental regulations accompanied by income growth (pollution haven hypothesis) in China. The results indicate that the evidence of factor endowment hypothesis is found in most of the pollutants studied in this article, whilst there seems no evidence of pollution haven hypothesis. Combining with all the estimated elasticities of scale and technique effects, composition effects and trade intensity on emissions, I find that increasing international trade has opposite impacts on the environment due to different pollutants. That is, for air pollutants (SO₂ and dust fall) increase in trade leads to more emissions, whilst for water pollutants (COD, arsenic and cadmium) trade liberalization decreases emissions.

Notes

¹ See, for example, Grossman and Krueger (1993, 1995), Copeland and Taylor (1994, 2003a, b), Antweiler et al. (2001) and Cole and Elliott (2003), etc.

² The distinctions between concentrations data and emissions data are well discussed in Cole and Elliott (2003).

³ Antweiler et al. (2001) separate scale and technique effects by measuring the former using GDP/km² and the latter using lagged per capita GDP. Since this study uses pollutant emissions data, the use of GDP/km² is therefore no longer meaningful as a measure of scale. For detail discussions on this issue, see Cole and Elliott (2003).

⁴ There may be possible existence of high correlation for DRKL and KL, and DRI and I, which may lead to multicollinearity problem in Equation 5. However, the correlation coefficients for DRKL and KL, and DRI and I in this dataset are 0.3933 and 0.2897, respectively. These correlations are not as high as suspected.

⁵ After being taken by natural logarithm form, the skewness of capital-labour ratio, one year lagged per capita GDP, trade intensity, world relative capital-labour ratio, world relative per capita GDP, domestic relative capital-labour ratio, domestic relative per capita GDP, per capita SO₂, per capita dust fall, per capita COD, per capita arsenic and per capita cadmium are improved from 2.474, 2.114, 2.481, 2.367, 2.542, 2.487, 2.353, 2.246, 4.449, 2.596, 5.759 and 3.241 to 0.382, 0.568, 0.808, 0.571, 0.648, 0.774, 0.797, −0.325, −0.265, −0.817, −0.797 and −0.189, respectively.

⁶ For more details on this issue, see Antweiler et al. (2001).

⁷ For detailed discussions on China's GDP expenditure measures, see Keidel (2001).

⁸ A weak evidence of pollution haven hypothesis is found for dust fall, see the results of trade intensity interacted with world relative capital-labour ratio in Tables 2 and 3.