Globalization and emissions in Europe.

• Author: Naughton, Helen Tammela
• Position: Report

1. Introduction

   Disagreement about the relationship between globalization and the environment remains an unresolved issue. Theoretical predictions of the effects of globalization are often ambiguous. (2) Therefore, empirical work must provide evidence of the actual impact on the environment. This study looks at five globalization effects predicted to affect the environment. Three aspects of globalization--trade, foreign direct investment (FDI) and international environmental treaty participation--have received a fair share of attention. Even so, Antweiler et al. (2001) is the only study to include all three of these effects. Less attention has been given to cross-border pollution and nearby countries' wealth effects. Maddison (2006) is the only study to consider these spatial variables, but he excludes the other three globalization variables. I estimate European sulfur dioxide (S[O.sub.2]) and nitrogen oxides (NO^) emissions between 1980 and 2000 within one empirical model including five different globalization variables.

   Increased globalization allows countries to strategically affect each other. For example, countries can use trade relations to persuade their polluting neighbors to lower emissions or join international environmental treaties. FDI can contribute to a country's emissions via technology transfer or by responding to lax environmental regulation, but it is also closely related to a country's trade. Countries that are poor relative to their neighbors may serve as pollution havens and thus have higher emission levels. These same countries also trade more because of large markets nearby. While the links between emissions and these globalization variables have separately been studied before, the correlations between the globalization variables must be considered as well. I do that by estimating models while omitting relevant globalization variables.

   Many previous studies relating trade to environment use international air quality data, but the theoretical link between trade and the environment emerges through emissions, not air quality. (3) Using air quality rather than emissions data may be problematic because air quality does not only depend on local emissions but is also a function of many geo-spatial factors and cross-border pollution. Aggregating air quality data across cities to form an air quality index for a country may also not be appropriate. (4)

   Figure 1 plots sulfur dioxide (S[O.sub.2]) emissions against mean S[O.sub.2] concentrations in 1995 for 28 countries. The correlation between these two variables is only 0.048, suggesting that air quality data may not be an appropriate proxy for emissions. In Figure 2 nitrogen oxides (N[O.sub.X] emissions are plotted against nitrogen dioxide (N[O.sub.2]) concentrations for 26 countries in 1995. (5) Again, there is only a 0.098 correlation between emissions and air quality. While controlling for monitoring-site specific characteristics alleviates some of the problems with air quality data, emissions data provide a better test of the existing theories. Therefore, I use data on S[O.sub.2] and N[O.sub.X] emissions. (6)

   [FIGURE 1 OMITTED]

   [FIGURE 2 OMITTED]

   An important aspect of these emissions is that they themselves introduce an international interaction. Nitrogen oxides and sulfur dioxide are both production byproducts that pollute the air. Once in the air, these pollutants may travel great distances resulting in acid rain and worsened air quality not only in the country of origin but in other countries as well. Three international environmental agreements are in effect to control their emissions. The Helsinki Protocol required a 30 percent reduction of the 1980 sulfur emissions by 1990. In contrast, the Oslo Protocol on sulfur emissions provides individual sulfur reduction targets for each country and a longer timeline--target dates extend from 2000 to 2010. The Sofia Protocol concerning nitrogen oxides calls on the participating nations to reduce their emissions to 1987 levels by 1994 and provides other guidelines for controlling N[O.sub.X] emissions. While these agreements are written and signed at international meetings, nations are not bound by an agreement until they ratify it. Therefore, to capture treaty effects I use the ratification date, not the signature date.

   Much of the literature on interactions between globalization and the environment focuses on the effect of environmental regulations on trade. The pollution haven effect (PHE) predicts that more stringent environmental regulation in a country would decrease domestic production and increase imports in the regulated markets. Levinson and Taylor (2008) and Ederington and Minier (2003) are two examples of studies that focus on the PHE. The goal of my paper is the flip side of the PHE and determines how trade affects the environment. Ederington and Minier (2003) control for the effect of trade on the environment by endogenizing environmental regulation in the trade equation. I make this point more explicit by estimating the equations for emissions.

   Central to the literature on trade's effect on the environment is the pollution haven hypothesis (PHH) that predicts that removal of trade barriers results in flow of dirty industries to countries with lax environmental regulations. Taylor (2004) discusses the challenges of empirical research in pinning down evidence related to this hypothesis. The PHH provides reason to believe that trade affects the environment, as modeled in this paper.

   My sample includes European countries alone because I use emissions transport matrix only available for Europe and the results should be viewed with this in mind. Focusing on a group of geographically small countries packed in relatively close quarters provides a good sample to test for cross-border pollution effects. While the estimated coefficients may not be the same outside of Europe, the omission of relevant globalization variables would most likely cause similar problems in an analysis that includes all countries.

   Cole and Elliott (2003) were the first to use international emissions data when estimating the impact of trade on the environment. They found mixed results with respect to trade's effect on emissions. Some recent studies, Grether et al. (2010), Grether et al. (2009), and Kellenberg (2008), find that trade has decreased emissions. In contrast, Managi and Kumar (2009) find a positive effect of trade on emissions and Cole (2006) find that trade liberalization increases national energy use. (7) Furthermore, Managi et al. (2009) find that trade lowers emissions in OECD countries but that trade increases emissions in non-OECD countries, at least for some pollutants. Lamla (2008) finds little evidence of trade affecting emissions. Therefore, evidence of trade's effect on emissions is mixed. I believe part of the reason for the mixed results could be driven by omitted globalization variables. Most of the previous studies include just trade, and none include more than three globalization variables. I find robust evidence that trade lowers emissions.

   Unlike previous studies that include no more than three globalization effects, my study includes five. In addition to trade, I include cross-border pollution, foreign direct investment, neighbors' wealth and treaty effects in a unified empirical model, and explore the potential omitted variable biases caused by inclusion of a subset of these effects. Omission of the globalization variables typically changes the included coefficients significantly.

   The following section outlines the empirical methodology. Section 3 presents the results and Section 4 concludes.

2. Empirical Model

   Following Frankel and Rose (2005), pollution emissions are estimated as a function of trade, income, and other country characteristics:

   [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], (1)

   where [E.sub.it] is either log of per capita S[O.sub.2] or N[O.sub.X] emissions in country i at time t; Trade/GDP is trade intensity (or openness) measured by imports plus exports over GDP; Inc is national income measured by GDP per capita; [X.sub.it] captures other country specific characteristics; [[kappa].sub.t] are year fixed effects; [[gamma].sub.i] are country fixed effects and [[epsilon].sub.it] is the i.i.d. error term. I estimate this equation using spatial 2SLS, where I instrument for trade intensity, income, income squared and the spatial lag.

   Previous literature decomposes the trade effect on emissions into scale, composition and technique effects (Grossman and Krueger, 1993; Copeland and Taylor, 1994, 1995, 2003; and Antweiler et al., 2001). The positive scale effect arises because trade tends to increase GDP, which in turn increases industrial production and emissions. The composition effect accounts for the changes in emissions due to changes in the composition of national output (shifts of production from clean to dirty or from dirty to clean...