ESG methodology: sustainability in the look-through

In depth10 min read · updated 25 October 2026 · 11 sources

This ESG methodology describes how the Sustainability tab in the Look-through calculates: what share of your portfolio is invested in companies that own power plants, coal mines, oil and gas facilities or heavy industry, how much of it lies in the energy and utilities sectors, and which SFDR category the funds have. All figures come from freely usable data with a named source. There is deliberately no score and no rating: only shares, tonnes and coverage.

Why no ESG rating

ESG ratings condense many characteristics into a single number and differ widely from provider to provider. Berg et al. (2022) compared the ratings of six agencies and decomposed the differences: 56% stem from the same characteristics being measured differently, 38% from scope (which characteristics count at all) and 6% from weighting. A score of our own would just be one more opinion. The tab therefore shows verifiable quantities: shares of the portfolio and recorded tonnes of CO2 equivalent per company, each with source and coverage.

Emissions data are not unambiguous either. According to Busch et al. (2022), the figures of different data providers agree more closely for direct emissions than for indirect ones, and particularly poorly for Scope 3; estimates are less consistent than figures from company reports. Aswani et al. (2024) show that emissions estimated by data providers deviate systematically from reported ones and are closely related to the companies' financial metrics. The values in the tab are estimates per asset. This is stated on every table.

Where the data come from

The basis is the emissions inventory of Climate TRACE (2026). It estimates emissions per asset and month (per power plant, mine, oil field, refinery or steelworks) and names the owners of many assets. Portfolio Lab loads the sector packages with CO2 equivalents over 100 years (global warming potential of the gases over 100 years) for four sectors and uses the 14 subsectors for which owners are available. From these it forms eleven asset types:

• Power: electricity generation, split into coal-fired power plants (coal is at least one of the fuels), gas- and oil-fired power plants, and biomass and waste-to-energy plants.
• Fossil fuels: coal mining as well as oil and gas, with production, refining and transport.
• Manufacturing: steel, cement, aluminium, chemicals (ammonia, methanol, soda ash, steam crackers), pulp and paper.
• Mining: bauxite, copper and iron ore.

For each asset, the ownership data describe the path from the direct owner up to the ultimate parent company. According to Mowat (2026), they come from a member of the coalition behind the inventory, cover stakes above 5% and contain complete ownership chains for assets accounting for 32% of global emissions, plus the direct owner for a further 14%. For each owner, the file gives the total stake in the asset and, where known, the LEI, the global identifier for companies (Legal Entity Identifier).

Not used are aviation, because airports are listed as owners there while the emissions come from the flights; agriculture, which has owners for only a few farms in very large files; and buildings, road transport and waste, for which there are no owners.

One year per company

When building the data, Portfolio Lab reads the monthly values and the owners and stores only the result per company. The year used is the last calendar year for which every subsector has all twelve months. A company is assigned the emissions of its assets according to its equity share: if a group owns 60% of a power plant, 60% of the emissions count.

If the stake is missing on all paths, the asset does not count towards E_u but is counted in full in a separate column, “assets with unknown stake”. This is an upper bound, not an estimate of the stake. Individuals, states, public authorities and unknown owners are dropped. For each company, the app stores the name, LEI, country, both totals and the totals per asset type: a few thousand rows instead of several hundred megabytes of raw data.

From security to company

The basis is the funds' holdings lists from the look-through, see Look-through methodology. Each position, excluding cash and derivatives, is matched as follows:

1. Via the ISIN: a publicly and freely available, daily updated mapping from ISIN to LEI supplies the issuer's LEI. If it is the LEI of a recorded company, the position matches. Only the ISINs of these companies are stored; issuers with more than 2,000 ISINs (mostly banks with structured products) are left without ISIN matching.
2. Via name and country, if the list has no ISIN: legal forms (AG, SE, PLC, Inc. …), words such as Holdings or Group and share classes at the end (N, A, Pref, ADR …) are removed; “BASF N” and “BASF SE” both become “basf”. The cleaned name must be exactly the same and the country must match. China and Hong Kong count as the same, as do the UK with Jersey, Ireland and Switzerland, and the US with Bermuda, the Cayman Islands, Ireland, the UK, the Netherlands and Switzerland, because headquarters, registration and the country in the list often diverge. Without a country, only globally unique names of six or more characters count.
3. Not attributed are positions in the financials sector, covered bonds and bonds of sovereign, government-related and supranational issuers (issuer type as in the risk factors). Asset managers, custodian banks and insurers often appear in the ownership data with stakes that belong to client money; if they counted, a large asset manager would carry the emissions of thousands of power plants.

Anything not matched is listed under “No entry”. This means no recorded asset or no unambiguous match, not “no emissions”. Software houses, retailers or pharmaceutical groups rarely have their own power plants; their emissions from purchased electricity and the supply chain are not in an asset inventory anyway.

The metrics in the tab

All values are shares of the fund or of the portfolio. A fund's values are derived from its holdings list (complete lists normalised to 100%), a portfolio's as the weighted sum of its funds; funds without a holdings list count as not looked through.

For the 1 Mt threshold, assets with a known stake count pro rata and assets with an unknown stake count in full, because what matters is the involvement, not the quantity. The threshold separates captive power plants that supply steam and electricity to factories or data centres from genuine stakes in power generation and mining: in the data, large consumer goods, pharmaceutical and technology groups come to a few times 100,000 t a year with such assets and therefore do not count as operators of fossil power plants. The company table shows the 50 largest positions with recorded assets, merged across all funds like the single securities in the look-through, with tonnes, matching method and the largest asset types.

Why no emissions per euro invested

The usual route to a portfolio's carbon footprint is financed emissions: under the PCAF (2022) standard, an investor is attributed the share of emissions that corresponds to their share of the company, i.e. the amount invested divided by enterprise value including cash (EVIC), multiplied by emissions. This requires the enterprise value or market capitalisation of all companies. Such data are not licensed for free use; Portfolio Lab therefore calculates neither financed emissions nor tonnes per million invested.

A weighted average of company emissions without a reference quantity would be misleading: it mainly measures how large the companies are. For this reason Aswani et al. (2024) also separate absolute emissions from intensity (emissions relative to size). The tab shows quantities that need no denominator: shares in size bands and stakes in coal, oil and gas assets.

SFDR category and fund names

The Sustainable Finance Disclosure Regulation SFDR (2019) distinguishes between funds that promote environmental or social characteristics (Article 8) and funds with a sustainable investment objective (Article 9); for the rest, “Article 6” is the common informal term. There is no free, central register of these classifications. They are stated in the fund documents. Admins therefore enter the category for each fund, with source and date; it is stored by ISIN and so applies to all listings.

Without an entry, Portfolio Lab searches the fund name for terms such as ESG, SRI, Sustainable, Screened, Paris-Aligned, Climate Transition, Low Carbon or Impact and shows “Guess: Article 8 or 9”. The reason: under the guidelines of ESMA (2024), funds using such terms must align at least 80% of their investments with their environmental or social characteristics and, depending on the term, apply the exclusions of the Paris-aligned benchmarks or of the climate transition benchmarks. Under Article 12 of EU Climate Benchmarks Regulation (2020), Paris-aligned benchmarks exclude, among others, companies that derive at least 1% of their revenue from hard coal and lignite, 10% from oil fuels or 50% from gaseous fuels.

These exclusions refer to revenue, whereas the tab refers to the ownership of assets. A share in companies with coal assets is therefore not a test of whether a fund complies with the rules for its name.

Limitations

• Weapons and tobacco cannot be derived: the holdings lists only give sectors, not industries or revenues; defence is part of the industrials sector, tobacco part of consumer staples.
• Only direct emissions of recorded assets: purchased electricity (Scope 2), the supply chain and the use of products (Scope 3) are missing; according to the GHG Protocol (2011) standard, Scope 3 is often the largest source. For an oil company, the emissions from burning petrol and diesel are far greater than those of its own assets.
• Estimates: the emissions per asset are modelled and may differ from company reports.
• Ownership with gaps: stakes of up to 5% are missing, some stakes are unknown, and the chains have a cut-off date.
• Matching by name: abbreviated or differently spelled names do not match; the coverage shows how much was matched.
• Financials: conglomerates classified in the financials sector that own power plants of their own are excluded as well.

Source, licence and updates

According to Climate TRACE, Terms of Use (2026), the data are licensed under CC BY 4.0: they may be copied, modified and shared, including commercially, provided Climate TRACE is credited and changes are indicated. The tab therefore states the source, the data version and the aggregation: annual total per asset, aggregated pro rata per company and matched to securities.

Admins rebuild the data with a tool on the server or with the button in the tab. The server downloads around 90 MB of sector packages and 33 MB of ISIN mapping; a few megabytes remain in the database. Nothing is downloaded when the page is opened; calculations use the stored state.

This page describes how Portfolio Lab calculates; past results are not a forecast. All terms in the glossary.