Environmentally Adjusted Multifactor Productivity (EAMFP) indicators highlight the importance of integrating environmental factors in productivity metrics. Standard MFP omits two key dimensions: natural capital used in production and pollution generated as by-products. EAMFP captures both within a consistent growth accounting framework.
Across 49 countries over 1996-2018, EAMFP accounted for around half of pollution-adjusted output growth in OECD countries and about one third in Key Partner economies, with the gap widening after the global financial crisis.
Non-renewable natural capital extraction remains an important driver of growth in some economies. On average, non-renewable resources contributed six times more to output growth than renewable resources, notably in a small number of resource-intensive economies, including Australia, Chile, Peru, China, Colombia and Brazil.
Conventional MFP is likely to understate productivity performance where GDP growth decouples from emissions, as observed in 33 of 49 countries, including 32 OECD members, over 1996-2018.
Further methodological work is needed to broaden the coverage of natural capital and pollution emissions, to improve the timeliness of EAMFP estimates, and to strengthen the estimation of shadow prices for pollution.
OECD Compendium of Productivity Indicators 2026
Multifactor productivity (MFP) has been the central driver of economic growth in most OECD countries in recent decades (see Chapters 2 and 4). MFP traditionally captures the efficiency with which an economy converts labour and produced capital, such as machines and buildings, into output. In the conventional framework, MFP increases when output grows faster than the combined inputs of labour and capital. However, MFP metrics do not account for negative externalities associated with production. In particular, they disregard the environmental pressures caused by emissions of greenhouse gases and air pollutants, and the depletion of natural capital (e.g. energy resources, minerals, land and ecosystem services). As a result, environmentally harmful and sustainable production processes tend to be treated equivalently, which can lead to an overestimation of productivity growth when it relies on resource depletion or pollution-intensive activities.
To address this limitation, efforts have sought to incorporate emissions and natural capital into a growth accounting framework and derive environmentally adjusted multi-factor productivity (EAMFP). This augmented framework adds two key elements. First, it treats emissions as undesirable byproducts of economic activity and second, it includes natural capital alongside labour and produced capital as factor inputs.
These adjustments are increasingly important for assessing sustainable growth pathways in the context of climate change, biodiversity loss, air pollution, and broader pressures on ecosystem assets. While EAMPF is still subject to data and methodological constraints, it offers a more complete representation of production processes by integrating environmental dependencies and externalities into productivity analysis. In doing so, it helps avoid overstating productivity gains that rely on resource depletion or pollution-intensive activities and strengthens the analytical basis for evaluating long-term growth. The rapid expansion of artificial intelligence (AI) and related resource-intensive digital technologies reinforces the need to address the environmental blind spots of traditional productivity indicators (De Vries-Gao, 2026[1]).
The development of EAMFP has been part of the OECD Green Growth Strategy since 2012, with several efforts undertaken to operationalise this framework. These include estimates of EAMFP at economy wide level (Cárdenas Rodríguez, Haščič and Souchier, 2018[2]; Dang and Mourougane, 2014[3]; Brandt, Schreyer and Zipperer, 2013[4]; Brandt, Schreyer and Zipperer, 2014[5]) and sectoral level applications (Cobourn et al., 2024[6]).
Implementing this framework requires data on natural resources and pollution emissions prices and quantities. Following SEEA guidance, natural capital valuation is derived from direct market transactions or approximated using resource rents. However, valuing emissions remains challenging given the absence of market prices. The literature proposes several approaches for assigning monetary values to emissions and integrating them into EAMFP calculations, most commonly through regression methods and efficient frontier methods.
The remainder of this chapter presents the latest OECD estimates of economy-wide EAMFP by Cárdenas Rodríguez et al. (2023[7]) for 49 economies. This approach augments the production function to account for 25 natural capital inputs as factor inputs and 12 GHGs and pollutant emissions as undesirable outputs over the period 1996-2018. At this stage, data availability prevents analysis for more recent years, including the post-COVID recovery and developments such as clean energy investment and commodity price movements in 2021-23. Emission prices are estimated using a country-specific random coefficient model linking emissions to output. Finally, to enable coverage across a wide range of countries, estimates are based on a simplified version of produced capital which does not account for capital services. As a result, conventional factor inputs in the other chapters of the Compendium are not comparable to those from the environmentally augmented growth accounting framework, and differences between MFP and EAMFP cannot be attributed solely to environmental adjustments.
Drawing on OECD work by Cárdenas Rodríguez et al. (2023[7]), this chapter presents the main empirical results for the 49 countries over 1996-2018. All statistics are based on geometric averages of annual growth rates over 1996-2018 unless otherwise stated. Details on the methodology are provided in Box 6.1. While output growth was on average higher in Key Partner economies than in OECD countries, OECD countries generated a larger share of growth through productivity improvements (Figure 6.1).
Percentage points, unweighted annual arithmetic averages
Note: Key Partners = Brazil, China, India, Indonesia, South Africa.
Source: OECD “Environmentally Adjusted Multifactor Productivity”, OECD Environment Statistics (database), based on Cárdenas Rodríguez et al. (2023[7]).
EAMFP is derived from an augmented growth‑accounting framework that explicitly accounts for labour inputs, produced capital, and natural capital (Figure 6.2). Unlike conventional productivity measures, the augmented framework accounts for both desirable outputs (GDP) and undesirable outputs (pollution emissions). EAMFP is measured as the residual, i.e. the share of pollution-adjusted output growth not explained by changes in measured inputs. EAMFP is therefore not inflated by natural resource extraction, nor understated by pollution abatement.
Building on earlier OECD work (Cárdenas Rodríguez, Haščič and Souchier, 2018[2]; Brandt, Schreyer and Zipperer, 2014[5]; Brandt, Schreyer and Zipperer, 2013[4]), the latest OECD EAMFP estimates are documented in Cárdenas Rodríguez et al. (2023[7]). They cover 49 countries, including OECD, accession candidate and G20 countries, over the period 1996-2018. Undesirable outputs include air emissions of 12 gases (5 greenhouse gases and 7 air pollutants), while natural capital covers non-renewable resources (four fossil fuels and ten minerals), renewable resources including three land types (cropland, pasture land and forestland), two biological assets (marine wild fisheries and non-cultivated timber), and three ecosystem services (watershed protection from forests, non-wood forest products and coastal flooding protection from mangroves).
Computing EAMFP requires estimates of the elasticity of output growth with respect to inputs and outputs. For labour, produced capital, and some natural capital inputs, which are traded in markets, estimates are derived from their cost shares under the assumption of profit maximisation (see Cárdenas Rodríguez, Haščič and Souchier (2018[2])). For natural capital inputs and ecosystem services, where market information is limited, costs are approximated using unit rents.
The costs associated with undesirable outputs (air emissions) are not directly observed due to the absence of market prices. Several approaches are used in the literature for estimating these costs, most commonly through regression methods and efficient frontier methods. Efficient frontier methods comprise Data Envelopment Analysis (DEA) and Stochastic Frontier Analysis (SFA). These approaches, in their basic form, often lack robustness and replicability and as such are challenging to use in policymaking (Luu, Menna and Lucke, 2025[8]; OECD, 2022[9]).
In Cárdenas Rodríguez et al. (2023[7]), costs are estimated econometrically using a country-specific Random Coefficient Model, which allows estimates to vary across countries and better reflect country-specific relationships between emissions and output.
The current EAMFP estimates are nonetheless subject to data and methodological constraints. These include the partial coverage of natural capital and pollution, constrained by data availability; and potential bias and non-trivial uncertainty in the regression-based estimates of pollution shadow prices, which are of particular concern for countries with high pollution intensities or rapid emission changes.
EAMFP and its share of output growth varies widely across countries. Latvia, Japan and Slovenia generated more than 75% of pollution-adjusted GDP growth through EAMFP gains, while EAMFP contributed less than 15% to output growth in Brazil, Indonesia and Türkiye (Figure 6.3).
Geometric mean 1996-2018
Source: OECD “Environmentally Adjusted Multifactor Productivity”, OECD Environment Statistics (database), based on Cárdenas Rodríguez et al. (2023[7]).
Standard output growth metrics underestimate the growth of countries that reduced undesirable outputs while maintaining positive GDP growth and overestimate it for countries whose GDP growth has been partly achieved through pollution-intensive production. Thirty-three of 49 countries, including 32 OECD members, reduced pollution emissions over 1996-2018, leading to a pollution adjusted GDP growth higher than GDP growth (Figure 6.4). Indicators derived from the EAMFP framework offer a more comprehensive measure of output growth when the pollution intensity of production differs across countries.
Geometric mean 1996-2018
Note: Pollution-adjusted GDP growth below (resp. above) GDP growth corresponds to an emissions increase (resp. reduction).
Source: OECD “Environmentally Adjusted Multifactor Productivity”, OECD Environment Statistics (database), based on Cárdenas Rodríguez et al. (2023[7]).
Natural capital is an important source of growth in several countries. In Australia, natural capital accounted for about 5.4% of all output growth per year and in Chile for about 3% (Figure 6.5). Standard growth accounting metrics attribute contributions of natural resources and ecosystem services entirely to productivity gains, while a number of countries, such as Peru, Colombia and Chile, have been relying on non-renewable natural capital extraction.
Geometric mean 1996-2018
Source: OECD “Environmentally Adjusted Multifactor Productivity”, OECD Environment Statistics (database), based on Cárdenas Rodríguez et al. (2023[7]).
Indicators from both the standard and the environmentally adjusted productivity frameworks require careful interpretation. Caveats and limitations of the (EA)MFP framework include:
Growth rates, not levels: growth accounting measures changes over time, not absolute stocks. A zero natural capital contribution means resource use was stable, not that a country used no resources. This property is shared with MFP.
Producer perspective, not a welfare or environmental sustainability metric: all valuations reflect the producer’s private cost perspective. Sha²dow prices for pollution represent the private marginal abatement cost, not the social cost of pollution. EAMFP is a production efficiency indicator, not a welfare or sustainability indicator.
Interpretation of the residual: like MFP, EAMFP is a residual and, as such, any mismeasurement in inputs or outputs flows into the estimate. By expanding the set of measured inputs and outputs, the expanded accounting framework reduces the scope of what the residual must absorb.
Sensitivity to business cycles: EAMFP and MFP are procyclical. The 2008-09 dip is visible in Figure 6.1. Multi-year geometric averages, spanning at least one full business cycle, are the recommended basis for cross-country comparisons.
Indicators derived from the EAMFP framework are maturing but still evolving. Priority areas for methodological and data improvements include:
Coverage of natural capital and pollution: while the 2023 update expands significantly the coverage of natural resources and pollutants from previous efforts, the scope remains partial and highly constrained by data availability. Future work should focus on covering natural resources such as soils, freshwater, critical raw materials (lithium, cobalt, rare earth elements), sand and gravel, ecosystem services (carbon storage, crop pollination, air and water purification), and expanding the range of undesirable outputs including emissions to soils and water bodies.
Improving the timeliness of EAMFP estimates: data availability hinders the derivation of timely estimates, at a time of rapid changes in the economies. More up-to-date indicators would enhance their relevance and integration into policymaking.
Shadow pricing of pollution: the regression approach has the advantage of transparency and replicability, but the current shadow price estimates carry a potential bias and non-trivial uncertainty. GDP and pollution may influence each other, and the current specification accounts only for the effect of emissions on GDP, implicitly assuming that changes in pollution emissions are unaffected by changes in GDP. These issues are of special importance for countries with high pollution intensities or rapid emission changes. Sensitivity analyses over alternative regression specifications or the use of alternative methods are a recommended robustness check for country-level studies building on these estimates (Luu, Menna and Lucke, 2025[8]).
Finally, further development and use of EAMFP indicators will require accelerating the compilation of environmental accounts in line with the System of Environmental-Economic Accounting (SEEA) Central Framework and the SEEA Ecosystem Accounting. Broader compilation of SEEA accounts in countries would allow expanding the coverage of environmental inputs and outputs in the framework and could enable more regular updates of the indicators.
The 2025 System of National Accounts (SNA, 2025[10]) introduces important changes that affect the natural capital component of the EAMFP framework. Overall, these revisions do not alter the conceptual architecture of EAMFP, but they improve the quality, consistency and policy relevance of the underlying national accounts data. These changes generally bring the SNA closer to the SEEA framework already used by EAMFP, though the treatment of biological resources represents a potential point of divergence between the two frameworks.
Main implications for EAMFP:
Depletion recorded as a cost of production. Unlike the 2008 SNA, which treated depletion of non-produced natural resources (such as fossil fuels) as a mere balance-sheet change, the 2025 SNA records depletion as a cost of production, similar to depreciation. This lowers net domestic product and net national income in resource-extracting economies (SNA, 2025[10]; OECD, 2025[11]). For EAMFP, the indicator structure is mostly unchanged, but improved natural resource stock data should lead to better estimates of resource quantities and prices.
Renewable energy resources become economic assets. Wind, solar and hydro resources are now recognised as non-produced economic assets, valued using the present value of expected future resource rents (SNA, 2025[10]) in the absence of preferred valuation approaches based on market prices or transactions of tradable permits. These resources were previously treated as an exploratory extension in EAMFP because of incomplete data (Cárdenas Rodríguez et al., 2023[7]). Their formal inclusion in national accounts should make it easier to integrate these resources into the core EAMFP indicators using more harmonised and official data.
Mineral and energy resource ownership is split between owner and extractor. Instead of assigning the full value of subsoil assets to the legal owner, the 2025 SNA splits value between the legal owner and extractor in proportion to their shares of resource rent – consistent with SEEA Central Framework guidance (SEEA, 2014[12]). Since EAMFP already measures natural capital income and pollution shadow prices from the producer perspective, this change improves the alignment between national accounts data and EAMFP methodology.
Change in treatment of biological resources yielding once-only products. The 2025 SNA changes the distinction between cultivated and non-cultivated biological resources yielding once-only products such as fish, timber and wild animals. It distinguishes between resources where management does not extend beyond quota regimes (e.g. wild animals and fish stocks) which remain non-produced assets, and resources where a continuum of management intensity can be observed, such as standing timber, which is recorded in 2025 SNA as a produced asset. The asset boundary for biological resources is however not changed, but there are changes in the asset classification. The 2025 SNA also separates the underlying land asset (e.g. forest land, which remains non-produced natural resource subject to depletion) from the biological resource on it (e.g. standing timber which is considered as inventories), and records the growth of timber expected to be harvested as output, rather than only harvested volumes (OECD, 2025[11]). The implications for EAMFP are not yet fully resolved and depend partly on how the ongoing revision of the SEEA Central Framework responds to these changes.1 For non-cultivated timber specifically, the reclassification as produced capital in 2025 SNA may reduce the measurement gap that EAMFP was originally designed to address, though the appropriate treatment will require further methodological work. Wild fisheries, by contrast, will continue to be treated as natural capital inputs in future EAMFP updates.
Broader improvements to labour and produced capital data. The 2025 SNA also expands detail on labour markets and includes new non-financial assets such as AI and data. These changes should strengthen measurement of labour and produced capital, which in turn improves the overall measurement of EAMFP.
The 2025 SNA implementation will require careful management of the transition to preserve comparability over time, including appropriate vintage adjustments in historical series and close alignment with the OECD Compilation Guide on natural resources measurement (OECD, 2025[11]). For areas where the 2025 SNA and SEEA frameworks could diverge, particularly regarding biological resources, implementation choices should be monitored as the revision of the SEEA Central Framework progresses.
References
[5] Brandt, N., P. Schreyer and V. Zipperer (2014), “Productivity Measurement with Natural Capital and Bad Outputs”, OECD Economics Department Working Papers, No. 1154, OECD Publishing, Paris, https://doi.org/10.1787/5jz0wh5t0ztd-en.
[4] Brandt, N., P. Schreyer and V. Zipperer (2013), “Productivity Measurement with Natural Capital”, OECD Economics Department Working Papers, No. 1092, OECD Publishing, Paris, https://doi.org/10.1787/5k3xnhsz0vtg-en.
[2] Cárdenas Rodríguez, M., I. Haščič and M. Souchier (2018), “Environmentally Adjusted Multifactor Productivity: Methodology and Empirical results for OECD and G20 countries”, OECD Green Growth Papers, No. 2018/02, OECD Publishing, Paris, https://doi.org/10.1787/fdd40cbd-en.
[7] Cárdenas Rodríguez, M. et al. (2023), “Environmentally adjusted multifactor productivity: Accounting for renewable natural resources and ecosystem services”, OECD Green Growth Papers, No. 2023/01, OECD Publishing, Paris, https://doi.org/10.1787/9096211d-en.
[6] Cobourn, K. et al. (2024), “An Index Theory Based Approach to Measuring the Environmentally Sustainable Productivity Performance of Agriculture”, OECD Food, Agriculture and Fisheries Papers, Vol. 213/OECD Publishing, Paris, https://doi.org/10.1787/bf68fb78-en.
[3] Dang, T. and A. Mourougane (2014), “Estimating Shadow Prices of Pollution in Selected OECD Countries”, OECD Green Growth Papers 2014/02, https://doi.org/10.1787/5jxvd5rnjnxs-en.
[1] De Vries-Gao, A. (2026), “The carbon and water footprints of data centers and what this could mean for artificial intelligence”, Patterns, Vol. 7, https://doi.org/10.1016/j.patter.2025.101430.
[8] Luu, N., B. Menna and F. Lucke (2025), Mind the Gap: Accounting for Emissions in Productivity Measurement, OECD Statistics Blog, 20 August 2025, https://oecdstatistics.blog/2025/08/20/mind-the-gap-accounting-for-emissions-in-productivity-measurement/.
[11] OECD (2025), Measuring Natural Resources in the National Accounts: A Compilation Guide, OECD Publishing, Paris, https://doi.org/10.1787/420c7c2a-en.
[9] OECD (2022), Insights into the Measurement of Agricultural Total Factor Productivity and the Environment, OECD, https://www.oecd.org/agriculture/topics/network-agricultural-productivity-and-environment/.
[12] SEEA (2014), System of Environmental Economic Accounting - Central Framework (SEEA-CF), United Nations, New York, https://seea.un.org/sites/seea.un.org/files/seea_cf_final_en.pdf.
[10] SNA (2025), The System of National Accounts, United Nations. New York, https://unstats.un.org/unsd/nationalaccount/sna.asp.
← 1. For the SEEA Central Framework, the distinction between cultivated and non-cultivated timber resources is important. As part of the ongoing SEEA CF update process, it is discussed whether the SEEA CF will follow the new SNA recommendations at this point.