Definition, sourcing, and updating of the emissions baselines

In this technical document, we explain the methods used to generate the emissions baselines in the Climate Equity Reference Calculator. Before doing so, it’s extremely helpful to review the baseline-related challenges raised by the effort-sharing problem.

The key point is that effort-sharing frameworks (as opposed to resource-sharing frameworks that divide up, say, a fixed emissions budget) require baselines. This is because “effort” must be measured against a baseline, and – by definition – it should be a “no effort” or “no policies” baseline. There are two distinct challenges here.

  • First, any generalized effort-sharing framework requires emissions baselines for all countries, as well as projections of other constituent indicators including GDP, emissions intensity, and population. And these projections, to be policy relevant, must extend out at least as far as 2030. However, except for population,[1] there are no widely-used country-level projections for these baselines and indicators that extend out beyond a few years.
  • Second, as anyone who has investigated the problem of baseline modeling is well aware, widely known and properly vetted “no effort” baselines are simply not available “off the shelf.”

Given these challenges, we have been forced to contrive our own baselines, and this despite the fact that we claim no special expertise in making long-term projections. Our basic rule, therefore, has been minimalism. We have avoiding making up numbers wherever we possibly could, and have rather relied as heavily as possible on existing, widely known and well vetted projections for all key indicators, which we’ve done by updating these projections for the recent history that has transpired since they were published.

When doing this updating, we have been forced to make a few assumptions. For example, we’ve chosen to “smooth” historical data when merging it with projected data to avoid any unrealistically abrupt transitions. And (like others before us) we have used “downscaling” methods to derive plausible national estimates from pre-existing analyses that are aggregated at the regional level.

More specifically, we combine recent historical GDP and CO2 emission intensity growth rates with projected growth rates from the IMF [2] (for near term GDP growth) and McKinsey and Co.[3] (for projections of longer-term GDP growth and emission intensity changes), to produce annualized projections of future GDP and CO2 emissions intensity. The product of the projected GDP and CO2 emission intensity yields national CO2 emission pathways. For non-CO2 emissions, we use projected growth rates from the US EPA.[4]

And what about the need for “no effort” or “policy free” baselines?

The problem here is comparability of effort. Which is to say that all countries – whether wealthy or developing or somewhere in between – must have the same kind of baselines if comparability of effort is to be possible. Again, this is because mitigation effort can only be calculated as reduction below an emissions baseline. The ideal emissions baseline is a counter-factual one in which no policies that effect emissions (whether these be explicitly climate policies or not) are included, for then all current policies – when extended forward – would be counted as effort. Only then are apples, as it were, compared to apples.

The situation here is actually rather odd. For example, many people believe that our baselines are unreasonably high, and this is particularly true in European countries that have made substantial effort to reduce their domestic emissions. However, there is a “be careful what you wish for” element to this situation. If future effort in these countries is calculated on the basis of baselines which already contain all current policies (as is for example done in the IEA’s World Energy Outlook), then future efforts will appear to be altogether and even absurdly inadequate. It’s only by measuring all efforts against a counter-factual, no-policies baseline that the real truth of the situation – for example, that northern European countries are already doing significant mitigation – becomes visible.

Does this mean that we ourselves calculate counter-factual no-policies baselines? We do not. After all, there are good reasons why such baselines are not readily available (although in the past the SRES scenarios often served that purpose). For example, even if it were possible to define emissions baselines that exclude all explicit climate policies, there are a myriad of other policies – gasoline taxes, fuel efficiency standards, and so on – which require substantive effort and which produce climate benefits, even though they are often not motivated by a desire for climate action.

Given all this, our strategy is a modest one. As we mention above and explain in greater detail below, our algorithm for projecting CO2 emissions is based on combining estimates of emissions intensity reduction with estimates of GDP growth, with the values for both derived from a convergence from historical rates of change to projected long-term (through 2030) rates of change. To limit the influence of climate policies, we’ve made two particular choices. First, we use the median value of the last decade as the starting point for emissions intensity reductions, which limits the influence of the years in which climate policies have had the greatest effect. Second, we use a relatively old projection – the McKinsey 2.1 projections (published in 2010) for our long term growth rates, which gives slightly higher values for long term emissions growth (i.e., slower declines in emissions intensity). These choices do not leave us with “policy free” baselines, but they adjust them in the right direction, and they can be applied consistently to all countries.

The details

The specific data and algorithms we use for GDP, CO2 emissions intensity, land use emissions and non-CO2 emissions are as follows:


Our GDP time series are based on historical data from the World Bank [5] and the International Monetary Fund [6], along with medium-term projections from the IMF and long-term projections taken from McKinsey and Co [7]. The McKinsey projections provide estimated GDP through 2030 for 21 countries or regions, which is somewhat greater than the number of countries/regions covered by other common sources of projections such as the International Energy Agency’s annual World Economic Outlook; IMF projections are available for 189 countries or territories five years ahead, currently through 2018. We extract the GDP growth rates for 2030 from the McKinsey projections, and estimate economic growth rates for all countries from 2018 to 2030 on the basis of a linear convergence between the IMF 2018 projection and the McKinsey 2030 projection for the appropriate country or region.

As always, there are issues raised by the estimation of economic growth rates based on PPP or MER estimates of GDP. We apply our growth rate estimates to GDP measured in 2010 MER dollars, and provide PPP conversion based on a constant conversion between PPP and MER based on 2005. In effect, this is equivalent to assuming growth in PPP and MER terms is the same over the measured horizon.

CO2 emissions from fossil fuels and industrial sources

Because CO2 emissions are closely tied to economic growth, we derive our CO2 projections for the fossil fuel and industrial sectors from estimates of the changing CO2 intensity (emissions per unit GDP) multiplied by projected GDP. Our primary source for historical CO2 data is the UNFCCC data set [8] for Annex 1 countries and the CDIAC [9] data set for non-Annex 1 countries. CDIAC data is available for more than 200 countries or territories through 2010; Both CDIAC and PBL (the Dutch Environment Ministry) also provide provisional estimates through 2011 (CDIAC) or 2012 (PBL) for many countries and regions.[10] We extend our historical estimates through 2012 for all 195 countries in the calculator database using the national or regional estimates as available.

Combining historical CO2 emissions data and GDP data, we derive the historical rates (and rates of change) of CO2 intensity. Unsurprisingly, while the annual CO2 intensity varies fairly smoothly across years for each country, the rates of change are quite variable; furthermore, changes in emissions intensity in recent years have been significantly influenced by policy implementation. We address this issue by calculating the median of the last ten years for which we have historical data (2002-2012) as an estimate of the historical rate of emissions intensity change.

Our long-term rates of future emissions intensity improvement are taken from the McKinsey and Co. projections. If the McKinsey estimate of the rate of emissions intensity improvement for 2010-2015 is lower than the historical 10-year median, we simply use the McKinsey projections out to 2030; if the 10-year median is lower than the McKinsey 2010-2015 estimate, we calculate a linear convergence from the 10-year median rate in 2012 to the McKinsey-based projection for 2030. This takes account of the fact that (a) in many cases, recent historical improvements are substantially policy-driven, and (b) in other cases, the McKinsey estimates show rates of improvement for many countries that are substantially greater than recent historical experience. Thus we have attempted to produce conservative estimates of the “no-policy” rates of emissions intensity improvement.

Non-CO2 gases

Because non-CO2 gases come from much more heterogeneous sources than CO2, it is not straightforward to talk about the non-CO2 “intensity” of an economy as a whole. For this reason, we do not base our non-CO2 GHG projections on intensity change and GDP growth as we do fossil CO2; rather we take projected economy-wide rates of change out to 2030 for total non-CO2 gases, and then calibrate them to the most recent historical non-CO2 emission data. Our source of long-term projections is the US EPA [11] which covers nearly all countries (although for many smaller countries the data is derived from “rest-of-region” aggregates). For recent historical data we use data reported to the UNFCCC for Annex I countries, and take the median of growth rates between 2000-2010 as the 2010 “starting point,” for projections out to 2030; for non-Annex I countries, we use data through 2005 from the USEPA and combine them with growth rates reported in the EDGAR [12] database for the years 2006-2009. In both cases (A1 and NA1), the rate of change converges to the estimated 2030 growth rate projected by the EPA; in particular, we use a smoothing algorithm to estimate the rate of change in 2030 from the time series for 2005-2030 reported in USEPA (2012).


Net CO2 emissions from LUCF (Land Use Change and Forestry, sometimes called LUCF for Land Use, Land Use Change and Forestry) in principle cover a wide range of land management activities, including changes in agricultural and forestry management practices; as a practical matter, it is dominated by deforestation in tropical countries and reforestation in developed countries (which were largely deforested in previous centuries). Projections of future changes in forest cover are perhaps even more speculative than changes in industrial activity; it has been commonplace to assume – counter to recent history – that even in “no climate policy” scenarios (such as the SRES scenarios), non-climate (e.g., biodiversity) concerns could and would drive a rapid global reduction in LUCF emissions. Indeed, many countries have policies on the books to either reduce deforestation or increase reforestation. However, unlike for energy emissions, there are no organizations like the IEA (or the IMF for GDP) whose job it is to provide regular global forecasts. As a matter of parsimony, then, we assume that LUCF emissions remain constant after 2012 up to 2030 at the global and national level.

This leaves us with the problem of estimating current LUCF emissions. There are multiple sources of data available; however many of them fail to provide reliable national-level disaggregation. Furthermore there are large discrepancies in both global estimates and in the aggregation of national estimates to regional totals; notably the reported net reforestation among the Annex I countries (part of their national communications to the UNFCCC) produce aggregate negative emissions highly inconsistent with estimated regional totals from CDIAC,[13] which is generally considered the most reliable regional source). Similarly, the data reported in the national communications of non-Annex I countries and the national data reported in the MATCH [14] database for developing countries, are both inconsistent with the CDIAC regional totals for the developing world. Given these inconsistencies, our approach is to use relative emissions levels between countries in the same region and scale them to meet the regional totals from the CDIAC data set, and then to scale the regional totals to match the global estimate from the Global Carbon Project. [15] The primary national sources we use are the UNFCCC data for Annex I countries and either national communications or MATCH data for developing countries. In some cases (notably Europe), where countries have a mix of positive and negative net LUCF emissions, further constraints have been imposed (e.g., countries with negative emissions aren’t scaled to larger negative values). Nonetheless it remains the case that for us as for everyone, LUCF data and projections should be treated with special caution in drawing conclusions about emissions responsibility and burden sharing.



[1] The UN Population Division’s “Medium Variant,” which has projections to 2100 for all countries and a world population of 9.5 billion in 2050, has effectively cornered the market for BAU population projections, and who are we to argue with the market? Available at

[2] The IMF’s biannual World Economic Outlook includes projections for the next five years for most countries. Available at

[3] McKinsey and Company (2010), Impact of the financial crisis on carbon economics: Version 2.1 of the global greenhouse gas abatement cost curve. Available at Their dataset has projections for 21 countries/regions through 2030. Its CO2 and GDP projections are based on the IEA’s World Energy Outlook 2009, though they have “complemented external data selectively with McKinsey analysis,” for example by adding individual growth projections for several countries not broken out in the IEA publications.

[4] USEPA (2012), Global Anthropogenic Non-CO2 Greenhouse Gas Emissions: 1990-2030. Report and data available at

[5] World Bank World Development Indicators Online,

[6] For a few countries not covered by our primary sources, we have relied on alternative estimates from the CIA World Factbook ( or other sources.

[7] McKinsey and Company (2010), see note 6.

[8] Summary data from national reports to the UNFCCC are a

[9] The Carbon Dioxide Information and Analysis Center is operated by the Oak Ridge National Laboratory of the US Department of Energy. Their primary national level data set is available at, and the citation is Boden, T.A., G. Marland, and R.J. Andres. 2013. Global, Regional, and National Fossil-Fuel CO2 Emissions. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, Oak Ridge, Tenn., U.S.A. doi 10.3334/CDIAC/00001_V2013.

[11] USEPA (2012), see note 4.

[12] Joint Research Centre, Emissions Database for Global Atmospheric Research. Global Emissions EDGAR v4.2, available at

[13] Houghton, Richard (2008) Carbon Flux to the Atmosphere from Land-Use Changes 1850-2005, Available at

[14] The MATCH dataset is available as supplemental online material published with Höhne, Blum, Fuglestvedt, Skeie, Kurosawa, Hu, Lowe, Gohar, Matthews, Nioac de Salles, Ellermann 2011:  Contributions of individual countries’ emissions to climate change and their uncertainty, Climatic Change 106:359-391

[15] The Global Carbon Project website is Their 2012 dataset (used in this report) is available at Their 2013 report was released in November of 2013 and has not yet been included in our dataset.