Skip to main content

Trade in Environmental Goods and Air Pollution: A Mediation Analysis to Estimate Total, Direct and Indirect Effects

Abstract

Based on panel data covering 114 countries between 1996 and 2011, this study investigates the impact on pollution of trade in environmental goods (EGs). We check the validity of the implicit consequences assumed by the win–win scenario in the current trade-climate negotiations, arguing that market dynamics should guarantee that EGs’ liberalization is ‘automatically’ in the interest of all countries, regardless their market and institutional capacities. We show that trade in EGs alone fail to address environmental problems effectively. In particular, although we found efficiency gains from trade in EGs (in terms of CO2 and SO2 emissions per 1 US$ of GDP), and more recurrently for net exporters than for net importers, our results often failed to highlight environmental effectiveness (in terms of total CO2 and SO2 emissions). A general conclusion that emerges from our empirical results is that trade [in EGs] cannot effectively replace non-market-based solutions, when it comes to non-trade objectives. However, it seems to complement them efficiently. Our multiple-equation GMM estimations reveal specific direct, indirect and total effects on pollution depending on the countries’ net trade status, leading to several policy recommendations for an increased environmental effectiveness of trade in EGs.

This is a preview of subscription content, access via your institution.

Notes

  1. This paper will focus on EGs without considering environmental services, because of data availability. This should not weaken our contribution to the academic literature on this topic. Indeed, without undermining the importance of environmental services in achieving environmental goals, the negotiations within WTO, beginning with the Paragraph 31(iii) of the Doha mandate (2001), have, to date, been more active for EGs.

  2. See Goff (2015) and http://www.international.gc.ca. (Trade/Opening New Markets/Trade Topics/WTO Environmental Goods Agreement (EGA), accessed on January, 15th 2018.)

  3. The 17 WTO members having launched negotiations toward the EGA are: Australia, Canada, China, Costa Rica, Chinese Taipei, the European Union, Hong Kong (China), Japan, Korea, New Zealand, Norway, Switzerland, Singapore, United States, Israel, Turkey and Iceland, and account for the majority (about 90%) of the world trade in EGs. Once the WTO Members in the EGA represent a critical mass of global trade in EGs, the eliminated tariffs agreed by the participants in the negotiations would have to be applied to all WTO members: i.e., the agreement should be extended on a ‘Most Favoured Nation’ basis to all WTO members.

  4. One should note that, after the failure of the multilateral ‘single undertaking’ Doha Round, the ongoing EGA negotiations have taken the form of a plurilateral agreement carried out by a reduced number of countries.

  5. See Sect. 2 for the common definitions of the scale, the composition and the technique effects on pollution.

  6. Literature review in Sect. 2 discusses recent findings and debates on this topic.

  7. Empirical studies exploring the determinants of EGs trade flows appear to be relatively more abundant. For instance, Hufbauer and Kim (2010) suggest that tariff elimination on EGs would increase world imports of these goods by approximately USD 56 billion. Balineau and De Melo (2011) explain a weak increase in EGs’ imports due to tariff reduction during the last decade by the existing (weak) tariff levels and import elasticity of demand. Recent research (e.g., Jha 2008; Sauvage 2014; Nguyen and Kalirajan 2016; De Melo and Solleder 2018; Tamini and Sorgho 2018) examining the factors determining trade in EGs highlights that lowering tariffs may increase trade, but higher gains could be obtained by the removal of non-tariff barriers. Despite low tariffs on many EGs in some developing countries, imports of EGs are still scarce because of a lack of technical assistance and, more generally, because of extremely weak purchasing power. In this context, market creation and capacity building should be prioritized to measures seeking for improved market access (Zhang 2011).

  8. Such effects should be at least partly captured in our regressions by overall trade openness.

  9. See additional reasons behind our choice of trade variables in Sect. 3.1 and the note of Figure B.1 in Electronic Supplementary Material.

  10. 70–75% of data points correspond to situations of ‘net importer’ and 25–30%—‘net exporter’ (see Table A.1, appendix A in Electronic Supplementary Material).

  11. See for example Grossman and Krueger (1993), Cole and Elliott (2003), Copeland and Taylor (2004), Levinson (2009), Managi et al. (2009), Lovely and Popp (2011) and Cherniwchan et al. (2017). Results generally differ by pollutant and country-type: (1) trade is usually found to reduce local/specific pollution (SO2, BOD) but to increase GHG emissions (NOx, CO2) and the energy use; (2) the environment in OECD countries is found to benefit from trade openness more than in non-OECD countries.

  12. This hypothesis would be more subtle than the pollution haven hypothesis (which predicts that, under free trade, stringent environmental regulations in one country lead to the relocation of pollution-intensive industries in countries with laxer regulations) because it assumes that only the dirtiest parts of production (pollution-intensive intermediate goods) are offshored, and not the dirty final goods. Whereas it is quite difficult to find empirical validation for the pollution haven hypothesis, the pollution offshoring hypothesis might still work, especially when much of the dirty goods trade is intra-industry (Cherniwchan et al. 2017). For more comprehensive and recent reviews of the literature on the pollution haven hypothesis, see also Taylor (2005), Kellenberg (2009), Brunel and Levinson (2016), Cole et al. (2017).

  13. However, the authors assert that the home government is also likely to lower its tax rate when there is learning by doing.

  14. This term comes from the energy economics literature and names situations of a rebound effect exceeding 100%.

  15. See Table B.1 in Electronic Supplementary Material for variables’ definitions and sources.

  16. Characterized by negative own-price elasticity, the local price of EGs is supposed to decrease when demand for these goods increases.

  17. Because the calculation of marginal effects becomes tricky in a moderate mediation model (conditional effects in a Multiple-Equation GMM estimation), we do not introduce interaction terms between trade in EGs, GDP and K/L in our system of simultaneous equations. However, we test our assumptions about the prevailing direct effect by first running a multivariate linear model including only domestic factors (GDP, K/L, ER and GNI/cap), and second by running a regression including in addition trade effects. If the coefficients of trade variables are positive, the (negative) estimates for ER and/or GNI/cap should not (or very slightly) change while the (positive) coefficients of GDP and K/L should reduce their amplitudes, in order to conclude about a (prevailing) partial trade-induced scale-composition effect, complementing the scale-composition effects induced by domestic factors. The opposite should be observed with negative estimates for trade variables: i.e., the GDP and K/L coefficients should not be significantly modified by the introduction of trade variables, while the estimates for GNI/cap and/or ER should get lower magnitudes. By observing these first empirical checks (see models 1-4 in Table C.1, Appendix C in ESM), we are quite confident about our interpretations.

  18. Not more than 3% of observations in our dataset correspond to either null exports and non-null imports in EGs, null imports and non-null exports in EGs, or both null imports and null exports in EGs.

  19. See these studies for a review of indicators previously used to measure the stringency of environmental regulations and their limitations for the purpose of an international comparison. They also bring quite robust validation tests for the use of a Z-score index measuring different aspects likely to proxy the stringency of the environmental policies worldwide; for example, signed and/or ratified MEAs, international NGOs, country’s energy performance, ISO14001 certification, adhesion to the Responsible Care® Program, the existence of an air-pollution regulation, etc.

  20. Ramsar (1971), CITIES (1973), Migratory species of wild animals (1979), Transboundary air pollution (1979), Protection of ozone layer/Vienna (1985), Basel (1989), UNFCCC (1992), Biological diversity (1992), Safety of radioactive waste management (1997), Kyoto Protocol (1997), Access to information… in environmental matters (1998), Protection of environment through criminal law (1998), Persistent organic pollutants (2001).

  21. Given the slowness of the WTO negotiating process, in September 2012, APEC members submitted a list of products based on individual appointments for which they committed to reducing tariffs to 5% or less by the end of 2015.

  22. See Reinvang (2014) and Vossenaar (2013) for more details about the APEC list.

  23. For example, the gas turbines of HS 841182 may be used for electricity generation from biogas, which is rather climate-friendly, but they may also have other non-environmental applications (e.g., as aircraft turbines).

  24. See Steenblik (2005) for more details about the genesis, description and comparison of the OECD and APEC lists, which were compiled in the late 1990s.

  25. Our Table A.1 in ESM provides the list of countries, for which data necessary to our study are available, by counting the number of observations—years— when they were net exporter or net importer of EGs from a specific list. As we can see, a very few countries are predominantly net exporters of EGs in the four classification lists; for example, China, Finland, Japan, Philippines, Republic of Korea, Sweden, etc. Countries that are mainly net exporters in the narrow lists of EGs (APEC54, OECD + APEC and WTO26) are not necessarily net exporters of EGs largely considered (WTO408) (e.g., Austria, Denmark, Italy, South Africa, Switzerland, Ukraine, etc.). Conversely, net importers in EGs from APEC54, OECD + APEC and WTO26 may be, simultaneously, net exporters of EGs from the WTO408 list (e.g., Algeria, Belarus, Brunei, Costa Rica, Cote d’Ivoire, Lithuania, Mexico, Turkmenistan, Venezuela, etc.).

  26. Following Drukker (2003), this test has good size and power properties in reasonable sample sizes. Under the null hypothesis of no serial correlation, the residuals from the regression of the first-differenced variables should have an autocorrelation of − 0.5. This test’s statistics and P-values are reported in Table C.1 (Appendix C in ESM).

  27. In addition to the potential inconsistency of the FE estimators (due to unbalanced and small-T panels, as well as when the omitted variables have time-invariant values with time-variant effects), we do not retain the FE regression model because of the too small (0.07) within variance compared to the between variance (0.8). If there is little variability within countries then the standard errors from fixed effects models may be too large to tolerate, and we cannot use countries as their own controls. However, aware of the potential heterogeneity problems resulting from time-invariant confounders, we shall check the robustness of our variable of interest by running an alternative estimation technique that removes efficiently the fixed country-specific effects: that is, the Arellano–Bond Panel system-GMM (see Table C.3 in Appendix C in ESM). Our results are highly robust to the inclusion of time-fixed effects.

  28. According to Porter and van der Linde (1995), strict environmental regulations can drive efficiency and encourage innovations that improve business competitiveness. At the same time, following the pollution haven hypothesis, stringent environmental policy would encourage polluting firms to relocate activities in countries with more lenient regulations (see footnote 13 for some references).

  29. In addition to have estimated trade flows in levels, we also followed the estimation strategy used in Rodriguez and Rodrik (2000), Frankel and Rose (2002), and Ortega and Peri (2014) that consists of predicting trade openness ((X + M)/GDP) from the following estimation:

    $$\begin{aligned} & \log \left( {\left( {X_{ijt} + M_{ijt} } \right)/GDP_{it} } \right) \\ & \quad = \begin{array}{*{20}c} { - .19} \\ {\left( {.001} \right)} \\ \end{array} \log \left( {distance_{ijt} } \right) + \begin{array}{*{20}c} { + .42} \\ {\left( {.006} \right)} \\ \end{array} \log \left( {border_{ijt} } \right) + \begin{array}{*{20}c} { + .09} \\ {\left( {.004} \right)} \\ \end{array} \log \left( {comlang_{{ethno_{ijt} }} } \right) + \begin{array}{*{20}c} { + .02} \\ {\left( {.004} \right)} \\ \end{array} \log \left( {comlang_{{off_{ijt} }} } \right) + \begin{array}{*{20}c} { - .06} \\ {\left( {.011} \right)} \\ \end{array} \log \left( {pop_{it} } \right) \\ & \qquad + \begin{array}{*{20}c} { + .04} \\ {\left( {.011} \right)} \\ \end{array} \log \left( {pop_{jt} } \right) + \begin{array}{*{20}c} { - .01} \\ {\left( {.024} \right)} \\ \end{array} \log \left( {area_{jt} } \right) + \begin{array}{*{20}c} { + .09} \\ {\left( {.043} \right)} \\ \end{array} \log \left( {landlocked_{jt} } \right) + \begin{array}{*{20}c} { + .003} \\ {\left( {.000} \right)} \\ \end{array} trend_{t} \\ \end{aligned}$$

    Equation is estimated for 1995–2012 period, with 187,727 observations; within R2 = 0.54, between R2 = 0.37, and overall R2 = 0.50; standard errors are in parentheses. Country and country-pair fixed effects are included but not reported. Variables \(\log \left( {area_{it} } \right)\) and \(\log \left( {landlocked_{it} } \right)\) are omitted because of collinearity. Once we have estimated \({ \log }\left( {(X_{ijt} + M_{ijt} )/GDP_{it} } \right)\), we aggregated predicted values across destinations j to obtain trade openness for each country i at time t. Having controlled for country size, the predicted values for trade openness are driven only by geographic and cultural characteristics. It is worth to underline that the above estimates are much weaker than the coefficients usually found in gravity models, because the dependent variable in the above equation is a measure of trade intensity (and not of trade flows in levels). For instance, in a similar equation specification for bilateral trade in levels, we get an estimate for distance of − 1.48.

  30. For the overall trade openness, which is not our variable of interest, a valid instrument with a moderate prediction power is easier accepted.

  31. See for instance Damania et al. (2003), Fredriksson et al. (2005), Greaker and Rosendahl (2006) and Zugravu-Soilita et al. (2008).

  32. See Frankel and Romer (1999), Gallup et al. (1999), Acemoglu et al. (2001), Easterly and Levine (2003), Sachs (2003), Hibbs and Olsson (2004) and Rodrik et al. (2004). Geography represented by latitude is statistically significant in our regressions using actual trade openness (results are available upon request), but becomes insignificant (hence dropped) in our regressions including predicted trade openness based on a gravity model of bilateral trade. These findings would be in line with the debate raised by Rodriguez and Rodrik (2000) concerning the geographically constructed trade share as a valid instrument. Such an instrument would be criticized if geography were likely to be a determinant of pollutant emissions “through a multitude of channels, of which trade is (possibly) only one” (Rodriguez and Rodrik 2000). For instance, following these authors, geography would influence the quality of institutions through the historical experience of colonialism, migrations, and wars. Hence, with insignificant estimates for geography once additional channels (institutional quality end environmental policy) are explicitly controlled for in the income and pollution equations, the estimates of geographically determined trade openness should not be biased.

  33. We stress that our regressions cannot say anything about the effect on pollution of policy-induced trade openness (i.e., fighting against climate change, rent seeking in general and/or environmental policy in particular, etc.).

  34. See Bollen (1987) for these different concepts.

  35. See the legend of Table 1 for a numerical illustration.

  36. There could eventually be a weak direct technique effect but not high enough to compensate the direct scale-composition effect.

  37. See Table A.1 in ESM.

  38. Source: EEA (2015).

  39. Rebound effects occur from direct changes in the EGs’ use that might see their volume to increase because of their price reduction. As production of EGs is a polluting activity, overall emissions thus should increase. In addition, efficiency improvements in aluminium smelting, for instance, can reduce the price of aluminium thereby fostering increased aluminium sales, which besides additional EGs require extra energy consumption, thus partly negating the initial emissions savings.

  40. See also Bernstein (2001).

References

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Natalia Zugravu-Soilita.

Ethics declarations

Conflict of interest

I wish to confirm that there are no known conflicts of interest associated with this publication and there has been no significant financial support for this work that could have influenced its outcome.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

I would like to thank the two anonymous referees from the Journal for very helpful comments and suggestions that have significantly improved this paper. I am also grateful to the participants in the CEMOTEV (University of Versailles) and the LEO (University of Orleans) seminars, and in the sixth World Congress of Environmental and Resource Economists for their valuable comments on the previous versions of the paper. Any errors or shortcomings remain the author’s own responsibility.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary material 1 (DOCX 274 kb)

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Zugravu-Soilita, N. Trade in Environmental Goods and Air Pollution: A Mediation Analysis to Estimate Total, Direct and Indirect Effects. Environ Resource Econ 74, 1125–1162 (2019). https://doi.org/10.1007/s10640-019-00363-6

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10640-019-00363-6

Keywords

  • CO2 emissions
  • Environmental goods
  • Environmental goods agreement
  • Environmental policy
  • International trade
  • Net exporter
  • Net importer
  • Pollution
  • SO2 emissions

JEL Classification

  • F13
  • F14
  • F18
  • Q53
  • Q56
  • Q58