S&P Global Trucost and MSCI: how much of a portfolio emissions figure is modelled?

S&P Global Trucost and MSCI will return an emissions figure for every holding in your portfolio, including the private ones and the ones that have never published anything. This page is about what those figures are made of, because that is what decides the PCAF data quality score you report, and it is not the same answer on every row.

DitchCarbon provides verified emissions data for over 2 million organisations, so procurement, sustainability and finance teams can measure and act on supply chain and portfolio emissions from one source. The comparison below runs on the thing a finance team is judged on: how much of the portfolio carries a figure the company itself published, and how much work it takes to move the rest.

Is a Trucost or MSCI emissions figure reported or modelled?

Both, mixed together in the same column, and neither vendor publishes how much of each.

S&P's Trucost methodology states the design plainly: wherever a material metric is not disclosed or not sufficiently disclosed, its modelled value is used, so that comparisons can be made between all companies regardless of disclosure levels. So a Trucost portfolio never shows you an absence. A disclosed figure and a modelled one arrive formatted identically, and only the source tag separates them.

MSCI defines its metric as a company's Scope 1, Scope 2 and Scope 3 greenhouse gas emissions as reported, if available, or estimated by its proprietary estimation model. If available is the load-bearing phrase. What sits outside it is what your data quality score is made of. Every record carries a flag marking which side it fell on, and MSCI reports a Reported Emissions percentage and an Estimated Emissions percentage at index level.

So on both, you count the composition of your own portfolio from the flags, after the contract. DitchCarbon works from the other end. The published figure is what we go and find, organisation by organisation, with the source document behind it and the assurance status recorded per scope. The split between disclosed and modelled is the first thing you see.

How is a modelled figure actually produced?

The mechanism decides what the number can support.

S&P publishes its model. Company business segments are mapped to more than 450 business activities in an environmentally extended input-output model built on NAICS, revenue is apportioned across those activities, and environmental intensities expressed as pollutant per unit of revenue are applied. Supply chain impacts come from input-output tables derived from the US Bureau of Economic Analysis make and use tables. Outside the covered universe there is a shorter route: a file of emissions intensities for over 158 GICS sub-industries, multiplied by company revenue.

MSCI publishes a sentence. Estimates are typically derived from industry averages, regional criteria and extrapolation from company disclosures. The sector-level calculation methodology is marked client access only, and the methodology library now sits behind a registration form. Proprietary describes access rather than accuracy: it means you cannot inspect how your own number was produced, which is a different thing from anybody having checked it.

Strip the vocabulary from either and the arithmetic is the same. Revenue is assigned to sectors and multiplied by that sector's intensity. Two companies in one sector with similar revenue receive the same emissions figure whatever either of them is doing. The number completes a portfolio total. It cannot separate two holdings.

DitchCarbon reaches a factor of that kind fourth, and only where the first three failed. Every organisation records which of four methods produced its figure: its own reported data, product data from a published carbon footprint, activity data from what was actually consumed, or industry data where nothing better was found or captured. The method is a filterable column and the headline figure toggles between organisation-specific and industry factors, so the share of your baseline still resting on a sector average is a number you can read rather than a claim you have to take.

How hard is it to improve a PCAF data quality score with each of these?

PCAF data quality scores run from 1 to 5, and finance teams shorten them to DQ. Three of the five decide this comparison. DQ 1 is the company's verified emissions. DQ 2 is the company's own unverified figure. DQ 4 is revenue multiplied by a sector emission factor, which is where a modelled dataset puts every holding it has not matched to a disclosure.

Nothing better than DQ 4 is reachable unless the company has published something, and a great many have. So the question is not whether your holdings will engage. It is who finds what they already published, matches it to the right legal entity and attaches the assurance status, and whether that work has already been done or is sitting waiting for you.

Assessed August 2026 against each provider's published documentation and, for DitchCarbon, the platform itself.

Read the right-hand column and the comparison is done. With Trucost and MSCI, improving a score is your project: modelled figures arrive with flags, and every rung above DQ 4 is work you do afterwards, one company at a time. With DitchCarbon it has already happened, because finding the disclosure, matching it to the legal entity and recording the assurance status is the product rather than the homework it leaves you.

What is left over is the modelled remainder, and the difference there is proportion rather than method. On a dataset built to fill every gap, a sector average answers for every company that has not been matched. On DitchCarbon it is the fourth method, holding only the rows where nothing has been published and nothing has come back, sized on screen and shrinking as requests are answered. Ask each provider to show that proportion on your own list before you sign, because a provider that cannot produce the number is telling you something by not having it.

How much of each dataset is company-reported?

Neither vendor publishes the split, and both publish figures that can be mistaken for it.

S&P publishes disclosure research on the S&P Global Broad Market Index. MSCI publishes disclosure rates within the MSCI ACWI Investable Market Index, including the finding that of 8,197 constituents examined, 13% had reported on their most material Scope 3 categories, on data as at May 2025. Both measure an index. Neither describes the composition of a commercial dataset.

What does exist is per record. S&P confirms that reported and estimated emissions and the source tagging are available to clients. MSCI ships a flag on every record and exposes the percentages at index level. The split is in your delivery. You can compute it once you are a customer.

If you are disclosing a weighted data quality score, you cannot derive it from a vendor's coverage claim. That is the reason DitchCarbon publishes the composition of your own list before there is a contract: send a sample of holdings to the coverage check and you get back how many carry a figure the organisation itself published, what the source document is for each, and what is left over.

What about private and unlisted companies?

This is where the coverage numbers get largest and the figures behind them get thinnest.

S&P claims 3.9 million private companies for its PCAF reporting offer. Its private companies methodology sets out three profile tiers: Fully Modeled, Sector Adjusted and Enhanced Research. Only the second and third involve an analyst, and only Enhanced Research includes a review of company environmental disclosure. In the Fully Modeled tier, revenue, financial year, currency and exchange rate are populated automatically from the Capital IQ Pro feed, and the company's sub-industry is converted to a single Trucost business activity by a default map with 100% of revenue allocated to it. That tier is one revenue number against one default sector mapping. How the 3.9 million distribute across the three tiers is not published.

MSCI publishes two private figures on the same page and they count different things: 56,000 private companies covered for emissions intensities, and 20,000 covered for greenhouse gas emissions, both as at 28 March 2024. Its own asset class table labels private equity and debt as economically estimated. Separately MSCI runs Private Company Data Connect, which collects data reported directly by private companies on the Integrated Disclosure Project template. That is a collection route, and a different product from the modelled dataset.

A coverage number counts rows returned. It does not count rows where a company said something, and only one of those moves a data quality score. DitchCarbon builds private records the same way as listed ones: whatever that organisation has published, found and matched to the legal entity with the source behind it, and a prepopulated request where nothing has been published yet. The private end of a portfolio is where that difference is widest, and where a modelled row is least likely to be checked by anybody.

What is each figure actually made of?

What sits behind each provider's number, and what each number can be used for.

Table compiled August 2026 from each vendor's own published methodology and product documentation.

The third column is the whole argument. Two of these fill a gap by modelling it. DitchCarbon looks first for what the company already published, asks the organisations where an answer would change the portfolio, and models what is left with the proportion visible instead of blended into the total.

Which of these publishes independent verification of its methodology?

Neither S&P Global nor MSCI does. Both publish something adjacent that reads like it. DitchCarbon does, and the document is downloadable.

S&P describes its methodology governance as internal: new methodologies and material changes are reviewed and approved by a methodology governance committee inside the company. On third-party verification of the underlying figures, S&P states that it tracks verification internally but does not yet provide it to clients as a column or a formula through Capital IQ Pro, the Excel plugin or the datafeed. What S&P describes as its own verification is backtesting: research across more than 15,000 companies with up to 20 years of data to identify anomalies and errors in reported data. Its disclaimer states that S&P does not perform an audit and undertakes no duty of due diligence or independent verification of any information it receives.

MSCI publishes no independent verification of its emissions estimation methodology in any material we could find. The one named independent assurance it does publish is a PricewaterhouseCoopers review against the IOSCO Principles for Financial Benchmarks, covering MSCI equity and IPD real estate indexes for the year to 30 June 2016. That is benchmark administration assurance for index products, it is a decade old, and it does not touch carbon estimation.

PCAF status is a third thing. S&P is named Principal Founding Data Partner to PCAF, which is a data partnership rather than a review of S&P's estimates. MSCI describes its approach as aligned with and based on the PCAF standard. Neither statement is verification, and neither should be read as one by a team that has to satisfy an auditor.

DitchCarbon is the only specialist Scope 3 tool with third-party assurance of its calculation methodology. The full working, every vendor checked and the evidence for each, is in our benchmark, Who's really verified? A reality check on carbon software assurance, and the underlying documents are downloadable from the DitchCarbon trust centre.

A portfolio figure has to satisfy your auditor, not only your investment committee. Ask every provider on the shortlist for the declaration itself, with the verifier and the standard named on it, because a page that says audit-ready and cannot produce one is telling you where its evidence stops. DitchCarbon hands over an ISO 14064-3 verification opinion with UL Solutions named on it, which is the difference between a number you can defend and a number you can only cite.

What happens after the portfolio is measured?

A portfolio total nobody acts on is a disclosure exercise, and this is the half a data feed does not reach. DitchCarbon carries it in the platform, and where a team already runs a system of record for the inventory and the disclosure, the same data feeds it without anything about the reporting stack changing.

Forecast and targets

The forecast runs each organisation's actuals forward on a two-year and a five-year trend to 2050, and draws alongside them the trajectory that organisation's own stated target requires. The gap between the two is the answer. Where the underlying reporting has missing years, large swings, or patterns suggesting a reporting change rather than a real movement, a reliability notice sits above the chart and links to the disclosure it was built from. Forecasting is projection rather than measurement, and the UL Solutions verification that covers the calculator does not extend to it.

Targets sit in a register rather than a badge: scope, on-track status, classification, base year, target year, reduction rate and source, with SBTi-validated targets separated from self-declared ambitions and the original wording quoted in full. Where a company has published several overlapping and occasionally inconsistent goals, all of them are shown rather than the most flattering one. Across a portfolio, SBTi progress can be weighted by embodied emissions, by spend or by count, and the three give different answers: weighted by count a portfolio can look poor while the holdings carrying most of the footprint are all committed.

Where to act, and who to ask

Every organisation and emissions category is ranked by the share of your total footprint that acting on it would move, each line carrying a realistic reduction percentage, the tonnes behind it, and a named peer already reducing at that rate, drawn from that peer's own published report. Every organisation also carries a score from 0 to 100 against its industry benchmark, broken down across six groups naming each criterion, its weight, the points earned and the source consulted, so filtering to the unearned criteria turns a scorecard into an engagement list.

Disclosure coverage is measured the same way, across ten measures per organisation, and the thin bars are the list of what to ask for. Requests go out from the portfolio view prepopulated with what that organisation has already published, so the recipient reviews and corrects rather than compiling from scratch, which is why a prepopulated request gets a better response rate than a cold survey. Most of a portfolio needs no request at all, because the figure already exists and DitchCarbon has already found it.

What if you are the company being modelled?

If your organisation has not published emissions, these datasets do not leave you out. They estimate you, from your revenue and your sector, and that number reaches your investors and your lenders whether you have seen it or not.

The way out is disclosure, and it does not have to be a project. Your DitchCarbon profile already exists and already carries whatever you have published, listed by year with the assurance status found for each scope. Claiming it lets you correct what we hold, add reports or certifications by drag and drop, and answer the questions buyers and investors ask most often once, so any logged-in viewer from another organisation reads the same answer. A disclosure to-do list shows year by year what we hold on you, marked complete or partial, and names the categories your investors expect and cannot find. There is no fixed form, a file in whatever format you already hold is accepted, and it is free. Claim your profile.

How do you test this before you commit?

Take a real sample of your holdings, weighted the way your portfolio is weighted, and include the private ones, because that is where the answers separate.

Then ask each provider the same four questions. How many of these do you hold a company-reported figure for, rather than a modelled one? What PCAF data quality score would each land at today? Where did each figure come from, and can I see the source document? And what happens to the holdings you have nothing for?

The fourth question reveals the design. A provider built to fill every gap has one answer to it, and the answer is a sector average.

DitchCarbon will show you what it already holds on your own list, with the source and change history behind every figure and the gaps marked rather than filled in quietly. Send a sample to the coverage check. Numbers you can defend within 2 weeks.

Last reviewed August 2026.