Kyoto and the carbon content of trade

Rahel Aichele, Gabriel Felbermayr, 4 February 2010



Scientists and politicians agree that we must limit global warming to less than 2°C above long-time averages. But getting there involves the tragedy of the commons on a global scale – carbon dioxide emissions have worldwide costs but local benefits.

Coordinated international policymaking is needed, but there is no legally binding agreement yet that would involve all relevant emitting countries. The Kyoto Protocol, signed in 1997 and ratified by 37 countries plus the EU between 2001-2007, exempts emerging and developing countries.

With international trade in goods, this protocol can give rise to so-called “carbon leakage”, where firms relocate production to countries without carbon taxes (or similar policies), thus offsetting emission reductions in “green” countries by higher emissions in “brown” countries. The associated competitiveness concern lies at the heart of the US resistance against the Kyoto Protocol.

But just how important is carbon leakage?

Recent literature argues that trade of carbon – as embodied in international trade in goods – matters. Grether and Mathys (2009) have shown that the world’s centre of gravity of CO2 emissions has shifted faster eastwards than the centre of economic activity. Similar evidence is presented by Wang and Watson (2007) who find that about a quarter of China’s CO2 emissions result from production for exports, mainly to OECD countries. Hence, trade in carbon brought about by the geographical separation of production and consumption seems a non-trivial phenomenon. But these papers are silent about the determinants of those trading patterns. Climate policy is a candidate explanation, but surely not the only one; classical comparative advantage arguments may well be sufficient to make sense of this evidence.

In a recent study, Mattoo et al. (2009) use a computable general equilibrium model to quantify the importance of carbon leakage due to unilateral emission cuts by industrial countries. The authors argue that – relative to a business as usual scenario – carbon taxes in OECD countries (including non-Kyoto partners such as the US) would have sizeable effects on their exports of carbon-intensive goods and also on their imports. Carbon leakage, i.e., the share of emission savings in OECD countries offset by higher emissions in trade partners, is about 6.5%.

We compute a 17% decrease in OECD emissions and a 1% increase in non-OECD countries, relative to year 2005. Using data from the IEA (2008), OECD emissions would fall by 2,197 Million metric tons of CO2 (MtC), while non-OECD emissions would increase by 142 MtC. Other models have found much higher effects, e.g. Babiker (2005). But it is well known that computable general equilibrium models are sensitive to assumptions about the model structure and parameters.

New data– new evidence on the carbon content of trade

In Aichele and Felbermayr (2010), we employ a different approach. Rather than simulating a hypothetical climate policy scenario using a theoretical model, we use actual data on bilateral trade flows, countries’ input-output tables, and sectoral CO2 emission coefficients to calculate the carbon content of bilateral trade. That is, we compute the amount of emissions in the exporter country caused by imports of goods by some importer country.

Our data spans 12 sectors, the period 1995-2005, and 38 countries (26 thereof have ratified the Kyoto Protocol in 1997-2005). The resulting panel data set is new – so far, the literature has either focused on the cross-section (a single year) or on single countries (for example China, US).

We find that carbon trade has increased from 1995 to 2005 by about 50% from about 3,000 MtC to 4,500 MtC. We also find that Kyoto countries typically are net importers of CO2. In 2005, France and the UK, for instance, both have “carbon imports” amounting to about 35% of their domestic carbon emissions. Even Kyoto countries such as Germany or Japan, with trade balances strongly in surplus are net importers of carbon. China, India, or South Africa, in contrast, have net exports of carbon amounting to up to 25% of their emissions.

Figure 1 looks at CO2 imports of Kyoto countries from non-Kyoto countries over time. In 2005, Kyoto countries as a total imported about 40% of embodied CO2 from non-Kyoto countries. More importantly, since 1997, when the Kyoto Protocol was signed, there is a positive trend. That trend has strengthened around 2002, the average year of ratification. The figure is suggestive – but, the trend may be purely spurious.

Figure 1. CO2 imports of Kyoto countries from non-Kyoto countries

Accounting for non-random Kyoto commitment

It is likely that Kyoto ratification is non-random; countries which expected carbon savings to occur because of sectoral or technological change may have committed to emission savings, because those savings are not costly to them. The opposite could hold for non-ratifiers. Non-random selection into Kyoto calls for a careful econometric treatment. The panel structure of our data allows including country-specific time dummies (among a host of other controls) into a regression of bilateral carbon imports on differential Kyoto ratification status of countries. This strategy accounts for any reason why a country ratifies the protocol at some point in time. The resulting estimates should, therefore, not be spurious. Moreover, in our framework one can even employ a regression-based test for strong exogeneity of regressors.

We find that carbon imports are on average about 12.5% higher if the importer has ratified the Kyoto Protocol and its trade partner has not. This effect is most pronounced in carbon-intensive industries such as basic metals, non-metallic mineral products, or paper and pulp. When looking at net carbon imports, a similar picture emerges. The result is robust across different methods of carbon accounting (for example, where exactly to book emissions associated to intermediate inputs from third countries). It also holds when the Protocol’s effect on “green” technological change is accounted for, or when the dependent variable is the CO2 intensity of carbon imports rather than their level.

Substantial carbon leakage but little overall effect

Our results are statistically significant, but do they also matter economically? In the year of 2005, Kyoto caused 84 MtC of additional carbon imports in committed countries, while their exports where 34 MtC lower. In our sample, the net aggregate trade effect of 50 MtC is fairly small relative to total carbon trade (2,190 MtC) or to total world emissions (19,500 MtC).

Ideally, however, one would compare the committed countries’ additional carbon imports with the emission savings brought about by Kyoto. While we have solid estimates on the former, the latter is much more a matter of speculation. Regressing national emissions on Kyoto status in a differences-in-differences setup results in emission-saving Kyoto effects ranging between 0 and 9%, depending on model specification. With a median estimate of about 2.85%, we find that additional carbon imports amount to about 40% of domestic carbon savings. This measure of carbon leakage is still about 12% when we use the highest plausible savings estimates.

We want to highlight three important conclusions from our study.

  • In our sample, Kyoto ratification had an effect on bilateral trade in carbon as embodied in trade of goods. There is robust evidence for a competitiveness effect.
  • The causal effect of Kyoto-induced measures on carbon emissions is hard to ascertain. It is possible that climate policies have lead to additional carbon imports without sizeable domestic savings. Carbon leakage is a serious possibility.
  • The overall effect of Kyoto on carbon trade is small relative to world emissions.

Policy lesson: consumption-based targets

Production-based CO2 emission targets give rise to carbon leakage. Consumption-based targets do not have this shortcoming and should therefore play a more prominent role in climate policies. While more difficult to implement, they would make the current debate about border taxes entirely redundant, and therefore rid the world of a potential protectionist surge in green disguise.


Aichele, Rahel and Gabriel Felbermayr (2010), “Kyoto and the carbon content of trade”, FZID Discussion Paper, No. 10-2010.

Babiker, Mustafa H (2005), “Climate change policy, market structure, and carbon leakage", Journal of International Economics, 65(2), 421-445.

Grether, Jean-Marie and Nicola A. Mathys (2009), “How fast are CO2 emissions moving to Asia?",, 21 November.

IEA (2008), CO2 Emissions from Fuel Combustion (detailed estimates), Vol. 2008 release 01, Paris.

Mattoo, Aaditya, Arvind Subramanian, Dominique van der Mensbrugghe, and Jianwu He (2009), “Reconciling Climate Change and Trade Policy", Peterson Institute for International Economics Working Paper Series, No. 09-15.

Wang, T. and J. Watson (2007), “Who Owns China's Carbon Emissions?", Tyndall Centre Briefing Note, No. 23.

Topics: Environment, International trade
Tags: carbon trading, climate change, Kyoto Protocol


Carbon consumption

The paper provides some evidence to validate what many copnclude is the outcome of a policy where only some countries move to reduce CO2e emissions.  It would be useful for them to extend their analysis to see if the net consumption of carbon declined.  

Research Assistant and PhD Candidate in Economics at Hohenheim University, Stuttgart

Professor of International Economics at Hohenheim University, Stuttgart, Germany

Vox Talks