US Supply-Chain GHG Emission Factors by NAICS — EPA Commodity Factors
The team behind the EPA’s USEEIO model publishes an emission factor for every commodity in the American economy, priced in dollars. Multiply what you spent by the factor and you have a Scope 3 estimate for a category you have no supplier data for. This page carries all 1,016 of them, searchable below and traceable to the published workbook row by row — and explains the one thing about their structure that catches most people out.
Look up a commodity
Type a 2017 NAICS-6 code or a commodity name. Each match shows all three published figures and the MasterBrain key that returns the same row, versioned, from the open endpoint — no key, no signup.
Loads the full dataset on first keystroke; everything after that is instant and local.
What these factors are
These are spend-based — or environmentally-extended input-output — emission factors. Each one says: for every US dollar spent buying this commodity in 2024, roughly this many kilograms of CO2e were emitted across the whole supply chain that produced it, from raw material extraction through to the point of sale.
They exist because most organisations cannot get a supplier-specific product footprint for everything they buy. The GHG Protocol permits a spend-based estimate as a fallback for Scope 3 Category 1 (purchased goods and services) and Category 2 (capital goods), and it is the basis for PCAF data quality score 4. The trade is coverage for precision: you get a complete inventory quickly, and every line of it is an industry average rather than a measurement of your actual supplier.
They are derived from USEEIO, the EPA’s environmentally-extended input-output model of the United States, combining Bureau of Economic Analysis industry accounts with the national greenhouse gas inventory. Since v1.4.0 the set is published by the model’s own authors on Zenodo under a CC BY 4.0 licence — earlier editions were US Government works in the public domain.
The thing that catches people out — 1,016 codes, 392 numbers
The dataset is published at 6-digit NAICS, the finest grain in the American industry classification. That looks like 1,016 independently measured commodities. It is not.
Underneath, the model runs on 392 USEEIO commodities, and each carries its own value — so the 1,016 NAICS codes resolve to 392 distinct numbers. The 1,016 codes are a mapping onto that model, not a finer measurement of it. (Under v1.3.0 the collapse was sharper still: 386 commodities produced only 281 distinct values. v1.4.0 separates commodities that previously shared a figure — most consequentially across waste management.) Every row carries the underlying model code in its useeio_ref field, and the factor is fully determined by that code — in all 1,016 rows there is not one case where two NAICS codes sharing a useeio_ref carry different numbers.
The practical consequence: choosing a more specific NAICS code does not get you a more specific factor. 80% of the codes in this dataset share their number with at least one sibling. Beef Cattle Ranching and Farming (112111), Cattle Feedlots (112112) and Dual-Purpose Cattle Ranching and Farming (112130) are three separate NAICS codes carrying one identical figure of 2.893, because EPA’s model treats them as a single commodity, 1121A0. Twenty-one different NAICS codes resolve to 4B0000.
This is not a defect — it is what an input-output model is. But it means that time spent refining a purchase from a 4-digit to a 6-digit NAICS code is often time spent for no change in the answer, and that a category which looks well-resolved may be resting on a broad economic aggregate. Check the useeio_ref before assuming granularity you do not have.
Three figures per commodity, and which one you want
Every commodity carries three numbers, and picking the wrong one is the most common mechanical error with this dataset.
| Figure | Field | Use it when |
|---|---|---|
| With margins (purchaser price) |
value |
Your spend figure is what you actually paid — an invoice or purchase-order total. This includes the wholesale, retail and transport margins added between the factory gate and you. This is the default and the one most reporters want. |
| Without margins (producer price) |
value_ex_margins |
Your spend figure is on a producer-price basis — you are working from economic accounts rather than invoices, or modelling the producing industry directly. |
| Margin | margin |
The difference between the two. Useful for checking which basis a third-party figure was built on. |
The margin is not a uniform uplift. 562 of the 1,016 commodities — 55% — carry a margin of exactly zero, because services are bought directly and have no distribution chain between producer and purchaser. Wholesale trade, trucking, professional services, janitorial services: all zero. Among the 454 commodities that do carry a margin, the median is 11.1% of the purchaser price, and the largest is 56.7%.
So the choice between with-margins and without-margins is immaterial for a services-heavy procurement profile and material for a goods-heavy one. If you are unsure, use with-margins against invoice totals and say so.
The range, by sector
All 1,016 commodities, grouped by their 2-digit NAICS sector. Values are kg CO2e per 2024 USD at purchaser price, frozen from MasterBrain v2026.116.
| Sector | Commodities | Lowest | Median | Highest |
|---|---|---|---|---|
| Transportation & warehousing | 57 | 0.0631 | 0.4925 | 1.5490 |
| Utilities | 4 | 0.4749 | 0.4749 | 0.5323 |
| Agriculture, forestry, fishing & hunting | 64 | 0.1779 | 0.4643 | 2.2592 |
| Mining, quarrying, oil & gas extraction | 28 | 0.2127 | 0.3656 | 0.8648 |
| Manufacturing | 359 | 0.0530 | 0.2364 | 2.7341 |
| Construction | 31 | 0.1737 | 0.1985 | 0.2572 |
| Accommodation & food services | 15 | 0.1165 | 0.1296 | 0.2176 |
| Arts, entertainment & recreation | 25 | 0.0287 | 0.1164 | 0.2019 |
| Wholesale trade | 71 | 0.0681 | 0.1069 | 0.1289 |
| Health care & social assistance | 39 | 0.0499 | 0.1050 | 0.1947 |
| Retail trade | 66 | 0.0858 | 0.1040 | 0.1674 |
| Administrative, support & waste management | 44 | 0.0493 | 0.1001 | 7.4375 |
| Other services | 48 | 0.0416 | 0.1000 | 0.1340 |
| Real estate, rental & leasing | 24 | 0.0268 | 0.0992 | 0.2263 |
| Educational services | 17 | 0.0972 | 0.0972 | 0.1628 |
| Professional, scientific & technical services | 49 | 0.0359 | 0.0834 | 0.1445 |
| Management of companies | 3 | 0.0760 | 0.0760 | 0.0760 |
| Information | 31 | 0.0402 | 0.0715 | 0.1079 |
| Finance & insurance | 41 | 0.0256 | 0.0595 | 0.1476 |
Across the whole dataset the range runs from 0.029 (insurance agencies and brokerages) to 3.924 (cement manufacturing) — a factor of 135 — with a median of 0.173. Cement is the single most carbon-intensive dollar in the American economy on this measure, because its emissions are chemical rather than energetic: calcining limestone releases CO2 from the rock itself, and no amount of clean electricity removes it.
| NAICS | Commodity | Factor |
|---|---|---|
| 562212 | Solid Waste Landfill | 7.4375 |
| 562213 | Solid Waste Combustors and Incinerators | 4.9846 |
| 327310 | Cement Manufacturing | 2.7341 |
| 112111 | Beef Cattle Ranching and Farming | 2.2592 |
| 112112 | Cattle Feedlots | 2.2592 |
| 524292 | Third Party Administration of Insurance and Pension Funds | 0.0256 |
| 524291 | Claims Adjusting | 0.0256 |
| 524210 | Insurance Agencies and Brokerages | 0.0256 |
What is not in here
The dataset covers every NAICS-defined commodity except three groups, and their absence is deliberate on EPA’s part rather than a gap in this compilation.
- Electricity. There is no factor for electric power generation or distribution. Purchased electricity is Scope 2, not Scope 3 Category 1, and belongs on a grid emission factor in kWh — not a dollar-denominated supply-chain factor. The utilities sector here contains only natural gas distribution, water supply, sewage treatment and steam supply.
- Government. NAICS sector 92, public administration, is absent entirely.
- Households. Not a purchasable commodity.
Application — worked example
A US manufacturer screening its purchased goods and services for a first Scope 3 inventory. Procurement is coded to NAICS and totals are invoice values in 2024 dollars, so the with-margins factor applies throughout.
| Category | NAICS | Spend (2024 USD) | Factor | tCO₂e |
|---|---|---|---|---|
| Cement | 327310 | 1,800,000 | 3.924 | 7,063 |
| Long-distance trucking | 484121 | 950,000 | 0.595 | 565 |
| Industrial supplies (wholesale) | 423840 | 2,400,000 | 0.115 | 276 |
| Custom software | 541511 | 1,100,000 | 0.084 | 92 |
| Janitorial services | 561720 | 300,000 | 0.214 | 64 |
| Total | 6,550,000 | 8,061 |
Each line is spend multiplied by factor, divided by 1,000 to convert kilograms to tonnes. The result is the reason to run this screen at all: cement is 27% of the spend and 88% of the emissions. A procurement review that ranked suppliers by invoice value would have put industrial supplies first and cement third. Ranked by emissions, there is only one thing on this list worth a supplier conversation.
Had every line been computed on the without-margins basis instead, the total would come to 7,921 tCO2e — 1.7% lower. The gap is small here because four of the five lines carry no margin at all; on a goods-heavy profile it would be considerably wider.
Where these factors are used
This dataset and its API are, for now, the only place the full 6-digit set is surfaced — no GreenCalculus calculator reads it yet. That is a deliberate staging: a 1,016-option picker is a poor interface, and the resolution caveat above is the reason a longer list would not necessarily produce better answers.
What the calculators do read are two coarser sets derived from this same EPA publication. The Scope 3 Category 1 Spend-Based Calculator works from a 27-category spend mapping, alongside EXIOBASE and DEFRA as alternative sources. The PCAF financed-emissions suite uses the 90-sector 3-digit aggregates as its data-quality-score-4 estimation fallback: listed equity and corporate bonds, business loans and unlisted equity, project finance and facilitated emissions. Those 3-digit aggregates are unweighted arithmetic means of the 6-digit children published here, computed by GreenCalculus rather than by EPA — so they are a convenience, and this page is the authoritative set.
Common reporting errors
| Error | Why it is wrong | Do this instead |
|---|---|---|
| Applying the factor to spend in a year other than 2024 | The denominator is a 2024 dollar. Applying it to spend from another year attributes the intervening inflation to emissions growth, distorting the figure by whatever prices moved. | Deflate spend to 2024 dollars with an appropriate price index before multiplying, and record the deflator used. The GreenCalculus spend-based calculator does this for you and shows the deflator in its audit trail. |
| Applying it to non-USD spend at spot rate | The factor embeds the structure of the US economy, not just a currency. A euro of German steel is not a dollar of American steel converted. | Use the factor set published for that economy — Eurostat for the EU, ONS for the UK — and disclose the mixture. |
| Refining a NAICS code to get a better number | 80% of codes share a factor with a sibling. Moving from 4-digit to 6-digit frequently changes nothing. | Check useeio_ref. If it is unchanged, the number will be unchanged; spend the effort on supplier data instead. |
| Using with-margins factors against producer-price spend | Double-counts the distribution chain on goods lines. | Match the basis to the spend figure. Invoices are purchaser price; economic accounts are usually producer price. |
| Reporting a spend-based total as if it were measured | These are whole-economy averages. Two firms buying the same commodity get the same number regardless of how either actually operates. | Label it as secondary data at PCAF score 4 and replace the largest lines with supplier-specific footprints as they become available. |
| Blending these with AR6-based factors silently | These are AR5 GWP-100. Mixing GWP editions inside one total is a real basis inconsistency. | Keep the basis per line, disclose the mixture, or restate everything onto one edition. |
Methodology, boundaries & uncertainty
Global warming potentials. These factors are on an AR5 GWP-100 basis, but that is our doing rather than the publisher’s. v1.4.0’s headline CO2e sheet is AR6, with methane at the non-fossil 27.9 — a value we solved for rather than assumed, by reconstructing that sheet from the per-gas data and testing candidates (27.9 reproduces it to a 0.185% median; AR6-fossil 29.8 only manages 0.556%). Because every other spend-based dataset here — UK, EU and DEFRA — is AR5, adopting the AR6 sheet would have left one region on a different basis inside the same report. So the figures above are recomputed on AR5 from the published per-gas sheet, and the publisher’s own AR6 figure is kept alongside each factor as value_ar6_100. The basis has moved between releases before — v1.2 was AR4 — so it is worth stating rather than assuming. If the rest of your inventory is on AR6, this is a genuine mismatch to disclose rather than blend, and it matters most for the methane-heavy agricultural commodities at the top of the range.
Boundary. Cradle-to-gate for the commodity, covering direct emissions from the producing industry plus all upstream supply-chain emissions embodied in its inputs. It does not include emissions from using or disposing of the thing you bought — those are separate Scope 3 categories.
Uncertainty. Unquantified by EPA, and structurally large. An input-output model assigns one intensity to an entire industry, so the variance between two firms inside a single NAICS code is invisible and can be an order of magnitude. Treat the output as an order-of-magnitude ranking that tells you where to look, not as a measurement of anything you bought.
Classification vintage. 2017 NAICS. Procurement systems coded to the 2022 revision will have a small number of codes that do not map directly.
Implementation & provenance chain
| Stage | Detail |
|---|---|
| Primary source | Supply Chain Greenhouse Gas Emission Factors for U.S. Commodities v1.4.0 — Ingwersen, W. & Young, B., the EPA USEEIO authors, published via Zenodo and generated from USEEIO v2.6.0. DOI 10.5281/zenodo.17202747. CC BY 4.0. EPA’s own catalog still lists v1.3.0. |
| File | SupplyChainGHGEmissionFactorsv1.4.0.xlsx — the byGHG sheet, which gives kg of each gas per 2024 USD before any GWP is applied. The companion CO2e sheet is also published alongside it. |
| Transcription | Cell-for-cell into MasterBrain canonical rows spend_based.us.naics6.<code>.<slug>. No rescaling, rounding or derivation. Each row cites its own EPA line. |
| GWP verification | The AR5 recomputation is verified by recombining the per-gas companion file and solving for the implied warming potentials across all 1,016 commodities — which returns 27.9 for methane and 265.0 for nitrous oxide, against AR5’s 28 and 265. The publisher’s own CO₂e sheet is AR6 and is carried per factor as value_ar6_100. |
| Publication | REST endpoint and CSV, versioned by MasterBrain edition, with the source registry entry attached to every response. |
| Licence | CC BY 4.0 (Creative Commons Attribution) — v1.4.0 is published on Zenodo by the USEEIO authors, and attribution is a licence condition, satisfied by the citation under Data access. Editions through v1.3.0 were US Government works in the public domain; the terms changed with the move to Zenodo. |
Data access — REST API & CSV
The full 1,016-row dataset is available as a machine-readable REST endpoint and as a flat CSV download, versioned by MasterBrain edition and citable.
Cache-Control: max-age=3600; X-GC-Version signals updates./wp-json/greencalculus/v1/us-supply-chain-factors
Click to generate ↓
Citation guidance
These factors are licensed CC BY 4.0: redistribution and commercial use are permitted, and attribution is required rather than optional. The citation below satisfies the licence and lets a reader check the vintage you used.
Ingwersen, W. and Young, B. (2025). Supply Chain Greenhouse Gas Emission Factors for U.S. Commodities v1.4.0. Zenodo. https://doi.org/10.5281/zenodo.17202747
Primary citation
Cite this dataset (GreenCalculus compilation). A versioned, machine-readable snapshot of these factors is archived on Zenodo. The concept DOI below always resolves to the latest edition; each frozen edition keeps its own DOI, so a citation stays checkable even after the data moves on.
Say, Jeremiah (2026). GreenCalculus — US Supply-Chain GHG Emission Factors by NAICS-6 (EPA). GreenCalculus. Zenodo. https://doi.org/10.5281/zenodo.21850527
GreenCalculus dataset DOI (concept — always the latest edition)
Frozen editions: v2026.183 (USEEIO SCF v1.4.0, deposited 7 Sep 2026) · v2026.110 (EPA SCFE v1.3.0, deposited 8 Aug 2026).
Frequently asked questions
Yes, provided two things hold. The spend must be in 2024 US dollars — deflate it first if it is not — and you should use the with-margins figure, because an invoice is a purchaser price. That is the intended use of this dataset and it will give you a defensible screening estimate. What it will not give you is a number that reflects anything your specific supplier does.
Because EPA’s underlying model has 386 commodities, and the 1,016 NAICS codes are mapped onto them. Around 80% of codes share their number with at least one sibling — the three beef cattle codes all carry 2.893, and twenty-one codes resolve to a single wholesale aggregate. Every row exposes its underlying model code in the useeio_ref field, and the factor is fully determined by it. If two codes share a useeio_ref, they will always share a number.
With margins, in almost every corporate reporting case, because it matches what you actually paid. Use without-margins only if your spend data is on a producer-price basis, which usually means you are working from economic accounts rather than invoices. For 55% of commodities the two are identical anyway — services carry no distribution margin. Among the rest, the median margin is 11.1% of the purchaser price.
AR5 GWP-100 — recomputed by us. The publisher issues v1.4.0 on AR6; we recombine the per-gas sheet on AR5 so the US figures stay comparable with the UK and EU ones, and keep the publisher’s AR6 number alongside each factor. It can be confirmed independently by recombining the per-gas sheet the publisher publishes alongside: solving for the warming potentials that reproduce the published CO2e figures gives 27.9 for methane and 265.0 for nitrous oxide, which are AR5’s values. Note that v1.2 was published on AR4, so the basis does move between releases and should be re-checked rather than assumed.
Because purchased electricity is Scope 2, and Scope 2 is calculated from kilowatt-hours and a grid emission factor, not from dollars. A dollar-denominated electricity factor would invite people to estimate their largest and most precisely measurable energy line from a spend proxy. EPA excludes it, as it excludes government and households.
Not by applying one to the other’s spend. These are per dollar of purchaser price; the UK ONS factors are per pound of gross value added, which is a different denominator entirely, and the Eurostat set is per euro of value added. The classifications differ too — NAICS against SIC against NACE. A multinational footprint needs each region computed on its own basis, with the mixture disclosed, rather than a single blended multiplier.
Weak by design. These are whole-industry averages, so an efficient producer and an inefficient one in the same NAICS code receive the same factor. Every framework places spend-based estimation at the bottom of its data hierarchy — it is PCAF data quality score 4 — because it characterises a sector rather than a supplier. Its value is completeness and hotspot ranking: it tells you which 10% of your spend to go and investigate properly.
Because most of cement’s emissions are chemical rather than energetic. Making clinker calcines limestone, and that reaction releases CO2 from the rock itself — roughly 60% of the total — before any fuel is burned. Decarbonising the kiln does not remove it. At 3.924 kg CO2e per dollar, cement is 135 times the intensity of insurance brokerage at the other end of the range.
Yes — under the terms of CC BY 4.0. v1.4.0 is published on Zenodo under a Creative Commons Attribution licence, so redistribution and commercial use are permitted provided the source is credited; the citation on this page satisfies that condition. Editions through v1.3.0 were US Government works in the public domain, so nothing was owed for those — the terms changed with v1.4.0.
Version history
| Version | Date | MasterBrain | Summary |
|---|---|---|---|
| 1.0 | 2026-08-08 | v2026.110 | Initial publication. Complete EPA Supply Chain GHG Emission Factors v1.3.0 set (1,016 NAICS-6 commodities, 2024 USD): the 386-commodity resolution caveat, three-figure margin structure, sector range table, worked example, 9-item FAQ, REST + CSV access. |
| 1.1 | 2026-08-13 | v2026.116 | Rebuilt on USEEIO Supply Chain GHG Emission Factors v1.4.0 (Zenodo, DOI 10.5281/zenodo.17202747, CC BY 4.0), replacing EPA SCFE v1.3.0. Emissions data moves 2022 → 2023 and the denominator 2022 → 2024 USD. Recomputed on AR5 from the published per-gas sheet because the publisher issues v1.4.0 on AR6 and the UK and EU spend-based sets here are AR5; the publisher’s AR6 figure is retained per factor. Typical change about 10%. Waste management is the outlier: v1.3.0 used one blended figure across the sector, and v1.4.0 separates it — landfill 0.9880 → 7.4375, incineration → 4.9846, collection → 0.1852. Resolution caveat updated: 392 distinct values rather than 281. Licence changes from US Government public domain to CC BY 4.0. |
| 1.2 | 2026-09-07 | v2026.183 | Added the commodity lookup (#lookup): type-ahead search over the live endpoint, showing all three published figures and the MasterBrain key per match; the CSV gains the value_ar6_100 column. Corrected seven statements that still described the pre-v1.4.0 terms — banner licence chip, derivation paragraph, provenance licence and GWP-verification rows, citation guidance, redistribution FAQ — to CC BY 4.0 with attribution required. Frozen-table vintage labels corrected v2026.110 → v2026.116 (values verified unchanged against the live feed); dollar-year in the definition corrected 2022 → 2024. Zenodo deposit rolled to v2026.183 (DOI 10.5281/zenodo.22613214); the cite block now leads with the concept DOI (10.5281/zenodo.21850527) and lists both frozen editions. |