The Carbon Ledger: A Portfolio's Second Bottom Line

When every position carries a footprint, sustainability reporting stops being an appendix and becomes an instrument in its own right. What a portfolio-level carbon ledger actually requires under PCAF attribution — and why most published methodologies still fail an independent audit on data provenance alone.

Carbon reporting at portfolio level is frequently presented as a solved problem. A number is produced, expressed in tonnes of CO₂ equivalent per million invested, and placed alongside the performance table. The number is usually arithmetically correct and epistemically weak, and the gap between those two things is where the entire discipline sits.

What PCAF attribution requires

The Partnership for Carbon Accounting Financials provides the standard most credible methodologies follow. Its central move is an attribution factor: an investor owns a share of an investee's emissions proportional to their share of the investee's financing.

The denominator is the detail that matters. PCAF specifies enterprise value including cash (EVIC) for listed equity and corporate bonds. Using market capitalisation instead — which is simpler and more flattering — systematically understates attribution for leveraged companies and makes portfolios non-comparable. The choice of denominator is not a technicality; it is the single largest methodological lever in the calculation.

Scope 1, 2 and the financed Scope 3 problem

Scope 1 (direct) and Scope 2 (purchased energy) are tractable. Scope 3 — everything else in the value chain — is where the number becomes contestable.

For a financial institution, financed emissions are themselves Scope 3 Category 15, which means a portfolio's headline figure is an aggregation of other entities' least reliable disclosures. Double-counting across the value chain is inherent to the framework rather than a defect in any one implementation, and any ledger that does not say so is overclaiming.

Why most methodologies fail on provenance

The audit question is rarely "is the arithmetic right?" It is "where did each input come from, and how do you know?"

PCAF addresses this with a data quality score from 1 to 5, where 1 is audited reported emissions from the investee and 5 is an estimate derived from sector averages and revenue. The score is the honest part of the standard, and it is the part most often omitted from client-facing reporting.

A portfolio carbon figure without a weighted data quality score is a point estimate wearing the clothes of a measurement.

In practice a large share of a diversified portfolio's underlying data sits at the weaker end of that scale — particularly private holdings, smaller issuers and anything in an emerging market. Reporting a single tonnage figure without the distribution behind it presents estimation as observation.

The three failures we see most

  • Denominator substitution. Market cap in place of EVIC, undisclosed, making cross-portfolio comparison meaningless.
  • Silent estimation. Sector-average proxies presented without the data quality score that would flag them as proxies.
  • Boundary drift. The reporting boundary quietly changing between periods — a holding reclassified, an asset class excluded — so the time series measures methodology changes as much as emissions.

What a defensible ledger publishes

  1. Attribution basis, stated explicitly, including the denominator and its treatment of cash.
  2. Coverage: what percentage of the portfolio by value is included, and what is excluded and why.
  3. Weighted data quality score, with the distribution, not just the average.
  4. A fixed boundary, with restatements disclosed when it changes rather than absorbed silently.
  5. Assurance status per input class: reported and assured, reported unassured, or estimated.

Why this belongs next to the financial statement

Not because the two numbers are equivalent — they are not, and treating a carbon figure with the confidence of a NAV is exactly the error. The argument is narrower: an exposure that is measured, bounded and disclosed with its own uncertainty is governable, and one that appears as a marketing figure in an appendix is not.

A second bottom line is only useful if it is kept to the standard of the first, which begins with being honest about how much of it is an estimate.