Financed Emissions and PCAF Reporting for Australian Banks and Super Funds

A bank's operational emissions are usually a rounding error. Branch electricity, a car fleet, some gas heating in a head office.

Then you open Scope 3 Category 15 and the number moves by three orders of magnitude.

That is the whole problem, and it is a data problem rather than a calculation problem. The arithmetic in financed emissions is simple. Knowing what your counterparty actually emitted, and being able to show an auditor where that figure came from, is not.

What PCAF actually requires

Financed emissions are the emissions of the borrowers, investees and insured entities on an institution's book, attributed back to the institution according to its share of the financing.

The PCAF Global GHG Accounting and Reporting Standard is the methodology Australian institutions most widely cite. Part A covers financed emissions, Part B facilitated emissions from capital markets activity, and Part C insurance-associated emissions.

Part A moved in December 2025. The third edition expanded coverage from six asset classes in the 2020 first edition, and seven after sovereign debt was added, to ten:

Listed equity and corporate bonds, business loans and unlisted equity, project finance, commercial real estate, mortgages, motor vehicle loans, use of proceeds structures, securitisation and structured products, sovereign debt, and sub-sovereign debt.

Four of those methodologies are new, along with optional reporting on undrawn loan commitments under IFRS S1 and S2, and two new reporting recommendations covering fluctuation analysis and inflation adjustment. If your methodology annex was written against the second edition, it is now describing a smaller standard than the one your peers will cite.

One nuance worth knowing before you write "PCAF conformant" in a disclosure. The first edition was reviewed by the GHG Protocol and found in conformance with the Corporate Value Chain (Scope 3) standard for Category 15. The GHG Protocol has since closed its review service, and the standard itself records that the additions in the December 2023 and December 2025 editions have not been through that review. That is not a defect, and it is not a reason to avoid PCAF. It is a reason to describe your basis precisely rather than leaning on a conformance badge that does not cover the newer methods.

Attribution is one idea applied ten ways

Across every asset class, financed emissions are the counterparty's emissions multiplied by an attribution factor, and the attribution factor is the institution's share of the total value of the company, project or asset.

The denominator is the part that causes arguments. PCAF puts both equity and debt in it, usually as enterprise value including cash, so that the emissions of a company are fully allocated across everyone funding it and equity and debt providers are not double counted against each other. For consumer vehicle lending and sovereign exposures, equity is not a meaningful concept and the standard sets out an alternative in the relevant chapter.

The table below pairs each asset class with its attribution denominator and the data quality reality we would expect on a typical Australian book. The tier column is our judgement, not a PCAF statement.

Asset class	Attribution denominator	Data quality likely on an AU book
Listed equity and corporate bonds	Outstanding amount over EVIC	2 to 3, most issuers report
Business loans and unlisted equity	Outstanding amount over EVIC or total equity plus debt	3 to 5, proxy heavy
Project finance	Outstanding amount over total project value	1 to 3, often verified for the project
Commercial real estate	Outstanding amount over property value at origination	3 to 4, better where NABERS exists
Mortgages	Outstanding amount over property value at origination	4 to 5, postcode and archetype averages
Motor vehicle loans	Outstanding amount over vehicle value at origination	3 to 5
Sovereign debt	Outstanding amount over PPP-adjusted GDP	2, national inventory data
Sub-sovereign debt	Per the third edition method	Unsettled, new in December 2025

The data quality score runs 1 to 5, where 1 is the highest quality and 5 the lowest. The scale tracks the estimation approach rather than the institution's effort: reported emissions score best, physical activity based estimates sit in the middle, and economic activity based estimates (revenue or spend multiplied by a sector factor) sit at the bottom.

That last point is the one worth internalising. A score 5 estimate says a borrower's emissions moved because its revenue moved. Buy a genuinely cleaner business at the same price and the number does not budge.

Double counting is designed in, and it still has to be explained

PCAF is explicit that double counting happens at five levels: between financial institutions, in co-financing the same activity, between transactions inside one institution, across different asset classes, and within the same asset class.

Some of that cannot be removed. A utility's Scope 1 is its customer's Scope 2, and if you lend to both, you count that energy twice inside your own inventory. PCAF's answer is transparency rather than elimination: report the Scope 1 and 2 of your loans and investments separately from their Scope 3, so a reader can see the overlap instead of guessing at it.

This is a common failure mode in hand-built workbooks, whoever builds them. The workbook nets things off because netting off looks tidier, the disclosure reports a single combined figure, and the assurance provider cannot tell what is inside it.

Where each sub-sector actually struggles

The major banks have a portfolio scale problem. Millions of retail mortgages and a long tail of business lending, with methodology annexes that run to dozens of pages. The constraint is the counterparty ledger, not the maths.

Mid-tier banks and mutuals face the same standard with a fraction of the headcount. A team of six produces what a team of sixty produces at a major. That asymmetry is the single strongest argument for building a repeatable system rather than an annual project.

Insurers carry a split problem: investment portfolio financed emissions under Part A, and insurance-associated emissions under Part C, now in its second edition. Underwriting attribution remains genuinely unsettled methodology and reasonable practitioners disagree.

Super funds sit closest to the clean case, dominated by listed equity and corporate debt, until you reach unlisted infrastructure, property and private equity. Several funds, including Cbus and HESTA, have publicly announced portfolio net zero commitments. Trustee covenants under section 52 of the SIS Act require acting in members' best financial interests, and climate risk now sits inside that assessment rather than alongside it.

Asset managers have the multi-fund problem: emissions per fund, per mandate, per share class, rolling up to an entity view without the funds contaminating each other.

What ASIC has signalled

ASIC published Regulatory Guide 280 on sustainability reporting under Chapter 2M of the Corporations Act in March 2025, covering its expectations and its surveillance approach. Read it alongside section 296A, which puts a directors' declaration over the sustainability report on a similar footing to the financial report.

Group 1 entities, which include the major banks and the largest funds, began reporting for financial years starting on or after 1 January 2025. Group 2, which captures most mid-tier banks and regional insurers, started with financial years beginning on or after 1 July 2026, so the first mandatory disclosure lands in the FY27 annual report.

Nobody should read the first cycle as a grace period. Limited assurance under ASSA 5010 applies from year one, and the assurance question on a financed emissions line is always the same: show me this number's source.

Where software fits, honestly

We should be upfront, because this is the vendor speaking and the honest answer is unflattering in one specific place.

Carbonly does not have a PCAF-branded data quality scoring engine as a distinct module. If your shortlist criterion is a native PCAF data quality dashboard out of the box, we are not the tool. Persefoni is one platform whose public materials position it around purpose-built PCAF financed emissions functionality.

Carbonly does not ship a built-in financed emissions attribution engine either. What it holds is an emission ledger where each record carries its source document and the emission factor version applied, with typed audit events on ledger writes and spend-based records floored to low data quality. The attribution and PCAF scoring sit in a report on top of that ledger.

Where it earns its place for an Australian institution is the combination: operational Scope 1 and 2 across a large branch or claims network in the same ledger and the same audit trail as Category 15, with the document engine reading utility bills, fuel and refrigerant records across eight file formats. NGER facility and corporate-group threshold status is computed from the ledger, which matters for any institution with property or fleet exposure near the 25 kt facility or 50 kt corporate group thresholds. The evidence pack for assurance covers the records held in the ledger.

Calculations run on the AR5 values NGER uses, so an AR6 presentation for an AASB S2 disclosure is a reporting-layer exercise rather than a switch. There is no GRESB submission generator, no NABERS generator, no CBAM, CSRD or ESRS module. Pricing is per project plus a $100 per month minimum, where a project typically maps to a business unit or a legal entity.

On the alternatives, described from public materials as at September 2026 rather than from our own testing: Persefoni positions around financed emissions and disclosure, Watershed around broad enterprise Scope 3 with a decarbonisation planning module, and IBM Envizi as part of a wider ESG suite with strong large-enterprise procurement credibility. None of the three publishes a public price list, and publicly reported figures put deployments of this kind in six figures annually; confirm directly with each vendor, because scope drives the quote. For native NGER categorisation and threshold tracking, we could not find public documentation on any of the three as at September 2026, which is not evidence of absence, so make them demonstrate it against your own facility data.

Plenty of Australian funds and mid-tier banks still run this in a consultant-built workbook, and for a first cycle that is a defensible choice. The consultant is doing the hard part, which is judgement on private companies and unlisted infrastructure. The problem is not the consultant. It is that a workbook can be opaque to the auditor and gets rebuilt annually. The platform is the workshop equipment; the specialist judgement stays with the specialist.

A pilot that tells you something

Take one asset class end to end, usually listed equity because the data is cleanest. Test attribution, the emissions intake, the tier tagging, and the path from a record to a report line.

Then add operational Scope 1 and 2 for one business unit, a few hundred real utility bills with the awkward cases included: a data centre, a regional branch on LPG, an EV charger split out from base load.

Finally, pick twenty records, ten financed and ten operational, and ask whoever will provide assurance to trace each one back to source. Count the clicks and count the gaps. That single exercise predicts year one better than any demo.

Frequently asked questions

Does Carbonly have a PCAF module? No. There is no PCAF-branded data quality scoring engine. The ledger holds the source documents and the audit trail; the attribution calculation and PCAF scoring sit outside it. Persefoni's public materials position it around native PCAF functionality.

How many asset classes does PCAF Part A cover now? Ten, following the third edition published in December 2025. The additions since the second edition are use of proceeds structures, securitisation and structured products, and sub-sovereign debt, plus optional reporting on undrawn loan commitments under IFRS S1 and S2.

Can we avoid double counting in financed emissions? Not entirely, and PCAF does not ask you to. Overlap between a utility's Scope 1 and its customer's Scope 2 is inherent when you finance both. The requirement is to report Scope 1 and 2 of loans and investments separately from Scope 3 so the overlap is visible.

What does a data quality score of 5 actually mean? That the estimate is economic activity based, typically revenue or spend multiplied by a sector average. It is a legitimate starting point and a poor decision input, because portfolio improvement and portfolio inflation look identical.

When does an Australian mid-tier bank need this working? Group 2 covers financial years starting on or after 1 July 2026, with the first disclosure in the FY27 annual report and limited assurance under ASSA 5010 from year one. Working backwards from assurance fieldwork, a decision in the second half of 2026 is the outer edge. See ASRS Group 2 reporting requirements.

What about unlisted infrastructure and private equity? This is where tiers sit at 4 and 5 and no platform makes the problem disappear. Everyone relies on sector averages, third-party estimates or bilateral engagement. Honest answer: unsettled.