PCAF Financed Emissions
The routing layer across the ten PCAF methods — which asset-class method applies, the one equation all of them share, and the single thing that actually changes between them: the denominator. Covers financed emissions (Part A), facilitated emissions (Part B) and insurance-associated emissions (Part C), plus the data quality score that spans all three.
A bank’s emissions are not really its own. Its heating and its data centres are a rounding error beside the emissions of everything it lends to, and PCAF exists to put a number on that second thing.
The arithmetic is disarmingly simple and identical across every asset class. What is not simple, and what every dispute is actually about, is the denominator — the figure you divide your exposure by to decide what share of a counterparty’s emissions is yours.
Every PCAF calculation is the same shape: an attribution factor multiplied by the counterparty's emissions, scored 1–5 for data quality. The attribution factor is your exposure over a measure of the counterparty's total value — and the asset class decides which measure.
Three Parts, one standard
The PCAF Global GHG Accounting and Reporting Standard for the Financial Industry is not one method but three related ones, and conflating them is the most consequential structural error a financial institution can make in this space.
- PART A Financed emissions — emissions attributed to money you have lent or invested. A stock measure: what is on the balance sheet at the reporting date. Seven asset classes.
- PART B Facilitated emissions — emissions attributed to capital markets activity you arranged but do not hold. A flow measure, reported separately.
- PART C Insurance-associated emissions — emissions attributed to what you underwrite, attributed from premium rather than from exposure.
These three are never added together. Each answers a different question about a different relationship with the same real-economy emissions, and summing them would count the same tonne under three headings.
Provider directory
Now it has to stand up to your auditor.
See who does this work. Every listing names the standards it works to, and paid placements are labelled. Including SE Advisory Services and Guidehouse.
Browse 3 financed emissions providers →Do this work? A listing is US$390 a year. Get listed →
The equation every method shares
Whatever the asset class, PCAF does the same two things: work out what share of the counterparty is yours, then apply that share to the counterparty’s emissions.
attribution factor = your exposure ÷ the counterparty’s total value
Both halves carry their own difficulty, and they are different difficulties.
The numerator is usually a fact you hold: an outstanding balance, a drawn amount, an exposure at the reporting date. Banks generally know this number, and where they do not it is a systems problem rather than a methodological one.
The denominator is the methodological decision, and it is what separates the ten methods from one another. The counterparty emissions figure is the data problem — usually unavailable, frequently estimated, and the reason the data quality score exists at all.
Which method applies
Route by the instrument you hold, not by the sector the borrower operates in. A loan to a car manufacturer is a business loan, not a motor vehicle loan — the motor vehicle class is for lending against the vehicle itself.
| Asset class | Part | Attribution denominator | Method |
|---|---|---|---|
| Listed equity & corporate bonds | A | EVIC — enterprise value including cash | Listed equity |
| Business loans & unlisted equity | A | total equity + debt (book value) | Business loans |
| Project finance | A | total equity + debt of the project | Project finance |
| Commercial real estate | A | property value at origination | Commercial real estate |
| Mortgages | A | property value at origination | Mortgages |
| Motor vehicle loans | A | total vehicle value at origination | Motor vehicle loans |
| Sovereign debt | A | PPP-adjusted GDP | Sovereign debt |
| Capital markets facilitation | B | issuance share, × 33% weighting | Facilitated emissions |
| Insurance underwriting | C | premium-based, per policy | Insurance-associated |
| Data quality, all classes | — | 1–5 provenance scale | Data quality score |
The denominator is the asset class
Read the table’s third column on its own and the logic of the whole standard appears. PCAF is not nine different methods; it is one method asking nine times: what is the whole, of which my exposure is a part?
For a company you can value, use its market value. Listed equity uses EVIC because a listed company has a continuously priced enterprise value. Unlisted companies do not, so the book value of equity plus debt stands in. Same question, different best available answer.
For an asset you lent against, use the asset. Mortgages, commercial real estate and motor vehicle loans all divide by the value of the thing financed, taken at origination rather than today. That choice is deliberate and consequential: it keeps the attribution factor stable as markets move, so a portfolio’s financed emissions do not fall because house prices rose. A denominator that floated with valuation would let a bank report progress it had not made.
For a country, use its economy. Sovereign debt divides exposure by PPP-adjusted GDP, because the counterparty is a whole national economy and its emissions are a national inventory.
This is also why the routing rule is the instrument rather than the borrower. The same counterparty can appear in three asset classes at once — a corporate bond, a project loan and an arranged issuance — each with a different denominator, each answering a different question about a different claim.
The data quality score
Every attributed figure carries a data quality score from 1 to 5, and the same 1–5 architecture spans all asset classes.
The single most misunderstood thing about it: the score is purely a statement about data provenance. It is not a measure of size, materiality, or how accurate the number happens to be. A tiny position and a huge one can both score 1. A verified-data position and an averaged-data position can carry identical attributed emissions while scoring 1 and 5.
Score 1 means the counterparty’s emissions were reported and verified. Score 5 means they were estimated from something very coarse — a sector average applied to a revenue or asset figure. Everything in between is a statement about how far the number travelled from a measurement.
The score exists because financed emissions are almost entirely built on data the reporter does not control and often cannot obtain. Disclosing the figure without the score would present estimates and measurements as the same kind of object. In practice the portfolio-weighted score is the number that moves year to year as data improves, and it is often the more honest progress metric than the emissions total itself.
Scope coverage and the Scope 3 phase-in
Two scope questions run through every asset class, and they are easy to conflate because both use the word.
Whose scopes you attribute. The counterparty’s. Attributed emissions are reported by scope, mirroring the borrower’s own GHG Protocol inventory — so a bank’s disclosure carries attributed Scope 1, attributed Scope 2 and, where in scope, attributed Scope 3 for its book. The equation applies the attribution factor to the counterparty’s Scope 1 + Scope 2 + Scope 3.
Where the result lands in your inventory. All of it, from every asset class, reports as a single Scope 3 Category 15 line in the financial institution’s own inventory. Listed equity and business loans land on the same line.
The consequence catches people out: your counterparty’s Scope 1 becomes part of your Scope 3. There is no scope-preserving passthrough, and a bank does not acquire Scope 1 emissions by lending to a steelmaker.
The Scope 3 phase-in
Counterparty Scope 3 is where the standard is deliberately gradual. Requiring full Scope 3 attribution from day one would mean attributing a number most borrowers cannot produce, so PCAF phases it in — sector by sector, starting where counterparty Scope 3 dominates and the data is most available.
Two practical consequences follow. First, a portfolio total that includes counterparty Scope 3 for some sectors and not others is correct under the standard but not internally comparable across sectors, and the disclosure should say which sectors carry it. Second, a total will rise as the phase-in widens, and that rise is coverage expanding rather than a portfolio getting dirtier — a distinction that has to be stated explicitly or it reads as the opposite of progress.
The same discipline applies as with data quality: when the basis of a number changes, restate the baseline or label the discontinuity. A financed-emissions time series where the boundary moved silently is not a time series.
What is not a financed emission
Three distinctions do most of the work in review.
Facilitated emissions are not financed emissions. When a bank arranges a bond issuance it does not hold the paper. PCAF Part B attributes a share using the facilitated amount and the institution’s issuance share, then applies a 33% facilitation weighting — the signature of the asset class, and the mechanism that stops the arranging bank and the eventual holders between them claiming more than 100% of the issuer. It is an annual flow, reported separately from — and never summed into — financed emissions.
Insurance-associated emissions are not financed emissions either. Part C attributes from premium, not exposure, because an underwriting relationship has no balance-sheet claim on the insured. For personal motor lines the attribution is built per vehicle from the motor premium rather than from a corporate revenue figure — a private policyholder has no revenue denominator to divide by.
Nor are the institution’s own operational emissions. A bank’s Scopes 1 and 2 are its buildings and its own energy. Financed emissions sit in Scope 3 Category 15, and for most financial institutions they exceed operational emissions by two or three orders of magnitude. Reporting only the operational figure is technically a GHG inventory and practically a non-answer.
Assembling a portfolio disclosure
The per-class methods each produce a defensible number. Putting them into one disclosure requires four things that no single class method can enforce.
- Compute holding by holding. Every class method insists on this for the same reason — attribution computed per position stays traceable to a named source and is therefore assurable. A portfolio-level shortcut produces a number nobody can audit back to anything.
- Keep the three Parts separate. One total for financed emissions, one for facilitated, one for insurance-associated. Never a combined headline figure.
- Carry the data quality score through the aggregation. A portfolio-weighted score alongside the total, not a total on its own.
- Hold one GWP basis and one reporting boundary across counterparties whose own disclosures may use different ones. Restate rather than mix.
Expect the same real-economy tonne to appear more than once across a full disclosure. A company’s emissions can be attributed to its equity investors, its lenders and its bond arrangers simultaneously. That is not an error in PCAF — it is a consequence of measuring claims on emissions rather than the emissions themselves, and the standard handles it by keeping the Parts apart rather than by trying to net them.
Error traps
- Routing by borrower sector instead of instrument. A loan to a vehicle manufacturer is a business loan; the motor vehicle class is for lending against the vehicle.
- Revaluing the denominator. Property and vehicle denominators are taken at origination. Marking them to current value makes financed emissions fall when asset prices rise.
- Summing Parts A, B and C. Three different relationships to the same tonnes; a combined figure means nothing.
- Reading the data quality score as accuracy or materiality. It is provenance only.
- Dropping the score when aggregating. The total without the score presents estimates as measurements.
- Omitting the 33% weighting on facilitated emissions, which triples the figure and breaks comparability with every other reporter.
- Reporting operational Scopes 1 and 2 as the institution’s footprint. The financed figure is typically orders of magnitude larger.
- Mixing counterparty GWP bases. Investee disclosures arrive on whatever basis they chose; the portfolio needs one.
Frequently Asked Questions
You are counting the same real-economy tonnes more than once, and under PCAF that is expected rather than an error. Financed emissions measure claims on emissions, not emissions, and a company can simultaneously be held as equity, lent to, and have its issuance arranged. The standard manages this by keeping Parts A, B and C separate and never summing them, not by attempting to net across relationships. What would be an error is combining them into one headline number.
To stop valuation movements masquerading as decarbonisation. If the denominator floated with current market value, a portfolio’s financed emissions would fall whenever house prices rose, with nothing having changed in any building. Fixing the denominator at origination means the attribution factor only moves when the loan balance moves, so reported change reflects lending and repayment rather than the property market.
No. The score describes where the data came from, not how accurate the result is. A score of 5 means the counterparty’s emissions were estimated from something coarse, such as a sector average applied to revenue — the estimate could still be close. Equally, a score of 1 does not make a figure precise, only well-sourced. Score and accuracy are correlated in aggregate and independent in any single case, which is exactly why both the total and the weighted score are disclosed.
Business loans and unlisted equity, or listed equity and corporate bonds if the exposure is a traded instrument. The motor vehicle loans class is for lending secured against vehicles themselves — retail auto finance — not for lending to companies that make them. Route by the instrument you hold, never by the sector the borrower operates in; that single rule resolves most classification questions.
The total will move as data improves, and that movement is not decarbonisation. Replacing sector-average estimates with reported counterparty data usually changes the number in both directions across a book, and the honest disclosure reports the data quality score improving alongside it. Presenting a total that fell because estimation improved as though it fell because the portfolio changed is the most common overstatement in this space, and a restated baseline is the correct response.
Methodology notes and limitations
This page routes; it does not calculate. It carries no factors and no worked example. Every attribution formula, data-quality table, scope-coverage rule and worked example lives on the ten asset-class methods it links to, which run 52k to 78k characters each. Its subject is the relationship between them.
Denominators are stated as the class methods state them, verified against those pages rather than restated from the standard. Where PCAF permits alternatives within a class, the class method carries the alternatives; this page names the primary basis only.
The 33% facilitation weighting is PCAF’s, not a GreenCalculus convention. It is the Part B mechanism as published, and it is stated here because omitting it is a common and large error — not because it is a settled question. The weighting has been contested since publication.
Ten methods is PCAF’s coverage, not a taxonomy of finance. Asset classes outside the standard — derivatives, securitisations, some structured products — have no PCAF method and are not addressed here or on the class pages.
No assurance opinion. Financed-emissions disclosure is increasingly assured and increasingly regulated. A figure destined for a climate disclosure, a target filing or a supervisory return should be reviewed by a qualified practitioner and, in most regimes, independently verified.