SFDR Principal Adverse Impacts

This page provides information on how Upright produces data on the Principal Adverse Impact (PAI) indicators defined in the EU's Sustainable Finance Disclosure Regulation (SFDR).

This page describes Upright's methodology for producing SFDR Principal Adverse Impacts data. For a description of what Principal Adverse Impacts data Upright provides, see this page.

Methodology common to all PAI indicators

Due to the differences in the nature of each indicator, most indicators have bespoke approaches to data collection as well as estimation methodology. There are, however, some common principles and methodologies applied across all of Upright's PAI indicators, mainly related to the treatment of disclosures and estimates, and reporting periods.

Treatment of disclosure vs estimated data

Upright treats companies' disclosures as the “ground truth” of PAI indicators. Therefore when available, they are preferred over estimates. For estimates, Upright generally measures accuracy as the deviation between disclosure data and estimates corresponding to those disclosures.

Example: assuming 100 company disclosures for scope 3 emissions, a model can be trained to predict those emissions based on some predictors. This model can be used to predict scope 3 emissions for the 100 companies and see how well the estimates match the known disclosures.

Where applicable, the generalization error is measured by splitting the disclosure data into training and hold-out sets: continuing on the previous example, use 80 randomly selected companies to train the model and see how well it predicts on the remaining 20 companies that the model has not seen before.

Treatment of timeframes and reporting periods

In the beta release, a single value for each supported PAI indicator is associated with each company. Over time as annual time series of indicators become available and meaningful Upright plans to support their reporting over time.

GHG Emissions

The GHG emission metrics measure the absolute and relative amount of carbon dioxide equivalents companies produce directly and indirectly. The metrics follow the definition the Greenhouse Gas Protocol, classifying companies' emissions into three scopes:

• Scope 1: Direct emissions from owned and controlled resources.

• Scope 2: Indirect emission from the purchased electricity, steam, heat and cooling.

• Scope 3: All of the indirect emissions included in the value chain, covering both upstream and downstream emissions.

Upright's estimates for all three scopes follow the same principle: companies' emissions are mainly driven by the economic activities that they conduct. Therefore, in Upright's estimates the detailed product and service mix of a company is used to explain and predict the resulting GHG emissions. In other words, companies with similar economic activities are expected to produce similar relative volumes of emissions.

As with all of the Uprights metrics the disclosure data comes from publicly available data sources. For GHG emissions the values are fetched primarily from companies' sustainability reports.

Upright's collected disclosure data and estimates mainly pertain to absolute Scope 1, 2 and 3 emissions. They are used to derive the following values:

• Carbon footprint (Scope 1 and 2): the sum of Scope 1 and Scope 2 emissions

• Carbon footprint (Scope 1, 2 and 3): the sum of Scope 1, 2 and 3 emissions

• GHG intensity (Scope 1 and 2): Carbon footprint (Scope 1 and 2) divided by the company revenue in €M

• GHG intensity (Scope 1, 2 and 3): Carbon footprint (Scope 1, 2 and 3) divided by the company revenue in €M

Accuracy of GHG estimates

The accuracy of Upright's models depend on the scope and the economic activities undertaken by the subject company. Scope 1 tends to be the most accurate estimate due to two reasons:

• More companies report scope 1 emissions relative to scope 2 or comprehensive scope 3 emissions.

• There exists little ambiguity in the computation logic that disclosing companies follow.

For Scope 3 emissions this is the exact opposite:

• Companies are less likely to disclose Scope 3 emissions.

• Much of the calculation rules for Scope 3 emissions leave room for interpretation. Two very similar companies in terms of revenue mix may report very different Scope 3 emissions.

Minimum acceptable set of Scope 3 categories

Upright considers disclosures valid only if they include the major Scope 3 categories such as “Use of sold products” and “Purchased goods and products”. Vice versa disclosures do not need to include categories like "Employee commuting", since they rarely amount to a material share of a comprehensive Scope 3 disclosure.

Market-based and location-based Scope 2 emissions

Upright uses location-based Scope 2 emissions, as it is more practical to estimate. Market-based emissions allow for a “residual share” of location-based emissions when market-based emissions are not provided, leading to market-based disclosures being a mix of location-based emissions and very company specific market-based emissions.

Geographical differences

The currently available volume of disclosure data does not allow for granularity on geographical differences.

Fossil fuel sector activity

Activity in the fossil fuel sector indicates whether the company derives any revenue from

• Exploration, mining, extraction, distribution or refining of hard coal and lignite

• Exploration, extraction, distribution (including transportation, storage and trade) or refining of liquid fossil fuels

• Exploring and extracting fossil gaseous fuels or from their dedicated distribution (including transportation, storage and trade)

Upright's estimates are based on the detailed economic activity modelling of companies: as the regulation calls to identify whether any revenue has been derived from the above activities, it is important to inspect the full sector footprint of conglomerates and other companies operating in multiple industries. This revenue mix -based algorithm has been validated against publicly available exclusion lists and other listings of companies known to match the regulation's definition. The vast majority of discrepancies between the algorithm and the publicly available data correspond to multi-sector companies correctly captured by the algorithm while missed by the publicly available data.

Non-renewable energy share

Non-renewable energy share measures the share of non-renewable energy consumption of the total energy consumption of the company.

Note that Upright's metric for non-renewable energy share deviates from the regulation's definition of "Share of non-renewable energy consumption and production" in the following manner: non-renewable energy share only measures the consumption mix of a company due to the scarcity of disclosed energy mix data for energy production.

Upright's estimates of non-renewable energy share are based on the detailed economic activity modeling of companies as well as their geographic footprint. Materially the energy mix of companies is expected to be similar between companies with similar economic activities. Upright's estimates weight the geographic average energy mix more for companies where limited disclosure data is available for similar companies.

As with all of Upright's PAI metrics, the data comes from publicly available data sources. For energy mix data, this primarily includes direct company disclosures from sustainability reports, usually in the form of Global Reporting Initiative disclosure 302-1.

Energy consumption intensity per high impact climate sector

Energy consumption intensity measures the energy consumption relative to revenue separately for each so-called high impact climate sector. These sectors are defined as NACE Rev. 2 Sections A-H and L. The Upright platform reports on this metric in three separate values for each company:

• High impact climate sector: the sector classification of the company, if it is any of the above-mentioned sectors.

• Total energy consumption: absolute energy consumption of the company, reported in MWh. This metric has been made visible separately from the intensity value since most energy consumption disclosures are reported in absolute terms.

• Energy consumption intensity: Total energy consumption converted to GWh and divided by the company revenue in € millions.

Upright's estimates for energy consumption follow the same principle as for GHG emissions: companies' energy consumption is mainly driven by the economic activities that they conduct. Therefore, in Upright's estimates the detailed product and service mix of a company is used to explain and predict the resulting energy consumption. In other words, companies with similar economic activities are expected to consume similar volumes of energy.

As with all of Upright's PAI metrics, the data comes from publicly available data sources. For energy consumption data, this primarily includes direct company disclosures from sustainability reports, usually in the form of Global Reporting Initiative disclosure 302-1.

Activities negatively affecting biodiversity-sensitive areas

Activity negatively affecting biodiversity-sensitive areas indicates whether all of the following criteria apply:

• The company conducts activities in biodiversity-sensitive areas, as defined by the SFDR Regulatory Technical Standards.

• These activities lead to the deterioration of natural habitats and the habitats of species and to disturbance of the species for which the protected area has been designated.

• Conclusions or necessary mitigation measures identified in relevant assessments have not been implemented accordingly.

Currently all positive indicator values for this indicator come from direct company disclosures, because outside-in estimation of the above company-specific criteria is not feasible.

As with all of Upright's PAI metrics, the data comes from publicly available data sources. For biodiversity harms data, this primarily includes direct company disclosures from sustainability reports, usually in the form of Global Reporting Initiative disclosure 304.

Emissions to water

Emissions to water refer to direct emissions of:

• Priority substances as defined in article 2(30) of directive 2000/60/EC.

• Direct nitrates, phosphates, pesticides as defined in directives 2000/60/EC, 91/676/EEC, 91/271/EEC and 2010/75/EU.

Upright's estimates of these water emissions are based on aggregate emission intensities of the above substances, grouped by economic activity. These emission intensities are applied to the modelled economic activities of each company to yield the estimated emission intensities and resulting absolute emission volumes.

Direct company disclosure matching the above-mentioned definition is relatively rare as of early 2022. This leads to most values in the Upright platform relying on estimates.

Hazardous waste

Hazardous waste refers to hazardous waste as defined in article 3(2) of directive 2008/98/EC as well as radioactive waste

Upright's estimates of these hazardous waste emissions are based on aggregate emission intensities, grouped by economic activity and matching the above-mentioned definition. These emission intensities are applied to the modelled economic activities of each company to yield the estimated emission intensities and resulting absolute emission volumes.

Direct company disclosure matching the above-mentioned definition is relatively rare as of early 2022. This leads to most values in the Upright platform relying on estimates.

Production of chemicals

Manufacture of chemicals as defined the regulation means activities that fall under the NACE Rev. 2 Class 20.2, "Manufacture of pesticides and other agrochemical products"

Upright's estimates are based on the detailed economic activity modelling of companies: as the regulation calls to identify whether any revenue has been derived from the above activities, it is important to inspect the full sector footprint of conglomerates and other companies operating in multiple industries. Furthermore it is important to inspect agrochemical manufacturers closely to verify whether they only manufacture fertilisers or nitrogen compounds that fall outside the scope of Class 20.2.

UNGC/OECD norm violations

The norm violation metric indicates whether the company has been involved in violations of the UNGC principles or OECD Guidelines for Multinational Enterprises.

As with all of Upright's PAI metrics, the data comes from publicly available data sources. For norm violations, this mainly consists of a broad scan of published norm violations of companies. Currently no algorithmic estimation is conducted for norm violations because outside-in estimation of company-specific violations is not feasible. Companies with no violations detected in the comprehensive search are expected to not be in violation of the norms.

UNGC/OECD compliance mechanisms

The compliance mechanism metric indicates whether the company has published mechanisms to comply with UNGC principles or OECD Guidelines for Multinational Enterprises. It is worth noting that a company may by in compliance with these societal norms without having published formal compliance with the norms. For example, an independent hair saloon is unlikely to violate societal norms but likewise unlikely to publish compliance with the norms.

As with all of Upright's PAI metrics, the data comes from publicly available data sources. For compliance mechanisms, this mainly consists of known lists of signatory companies. Currently no algorithmic estimation is conducted for compliance mechanims because outside-in estimation of company-specific compliance mechanims is not feasible. Companies with no compliance mechanisms detected in the comprehensive search are indicated as not having known compliance mechanisms.

Unadjusted gender pay gap

Unadjusted gender pay gap aims to measure possible gender discrimination in the workplace. It is measured by the difference between average gross hourly earnings of male paid employees and of female paid employees as a percentage of average gross hourly earnings of male paid employees.

The factors that contribute to the broad definition of unadjusted gender pay gap have been studied by Leythienne & Ronkowski, (2018) and Blau & Kahn (2017) among others. These factors can be split into two categories: personal & job level -related factors like age, occupation and employment contract, and enterprise-related factors like company size and its principal economic activities.

Upright's estimates are based on the enterprise level contributors. Using Upright's detailed economic activity modeling for companies enables a fluent representation of differences between companies. In addition, companies are grouped based on the employees count and revenue. The estimates are expected to outperform a naive economic sector and country average aggregates.

As with all of the Upright's estimates the data comes from publicly available data sources. This includes direct company disclosures from sustainability reports and national databases. Currently the coverage of disclosure data is mainly driven by varying disclosure requirements and practices between countries: the United Kingdom makes pay gap data for large companies completely public while some countries outright ban disclose of pay-related data.

Board gender diversity

Board gender diversity measures the share of female members in the board of directors.

Upright's estimates of board gender diversity are based on the economic activity modeling of companies as well as their geographic footprint. Materially the board composition of companies is expected to be similar between companies with similar economic activities and similar geographic location.

As with all of Upright's PAI metrics, the data comes from publicly available data sources. For board gender diversity data, this primarily includes a comprehensive search of publicly available board member lists and their automatic classification into male and female members.

Involvement in controversial weapons

The compliance mechanism metric indicates whether the company is involved in the manufacturing of any of the following weapons:

• anti-personnel mines

• cluster munitions

• chemical weapons

• biochemical weapons

It is worth noting that the list as defined by regulation excludes some weapons often understood as controversial, such as nuclear weapons.

As with all of Upright's PAI metrics, the data comes from publicly available data sources. For the above-mentioned weapons, there are relatively few known manufacturers, and furthermore a large share of these manufacturers are governmental organizations or fully government owned. The remaining set of companies with any security issuance has been comprehensively screened via publicly available information, including various exclusion lists. Nuclear weapons and other weapons not matching the regulation's definition have been excluded from consideration. Companies not found in these listings are expected to not be active in manufacture of these weapons.

Current limitations

• The collection of disclosed data (e.g. GHG disclosure) is still work-in-progress, and Upright continuously adds new disclosures from companies. In addition to providing access to more disclosures, Upright expects that upcoming additions of disclosed data will also improve the accuracy of the modeled estimates.

• The data is still lacking some key metadata, such as the reporting period (year) for disclosed values, or the details on norm violations.

• Aggregates for funds and other portfolios are not available.

• Estimates of renewable energy production are not yet available for companies producing energy.

The accuracy of the produced indicators is primarily limited by available information. Upright continuously seeks to improve the accuracy of its indicators by using the best available information and statistical methods for integrating information from different sources.