Appendix to ‘The EU Taxonomy as a Dynamic Battleground’

This appendix outlines the collection and treatment of materials from the article “The EU Taxonomy as a Dynamic Battleground”.

Section 1: Data on PSF Participation – This section outlines the data collection process for identifying individual participants of the Platform for Sustainable Finance (PSF).
Section 2: Categorizing Institutions of PSF Participants – This section details the methodology used to classify the institutions affiliated with PSF participants.
Section 3: Evolution of Institutions over PSF Meetings – This section presents various approaches to analyzing changes in the types of institutions represented across different PSF meetings.
Section 4: Interviews – This section provides a brief explanation of how the interviews were conducted.
Section 5: Other sources – This section describes the sources used to collect complementary information on the EU Taxonomy and the PSF as well as to prepare the interviews.

1 Data on PSF Participation

The PSF published the minutes for each of its first 38 meetings. For feasibility reasons, our analysis stops at the 38th meeting that took place in January 2024. The minutes contain the list of participants and their affiliations. We extracted participants’ information from these documents, including their name, affiliation, position in the PSF, as well as the meeting number and date, and compiled this into a single dataset.

Here is an extract of the raw data related to the first PSF meeting. This dataset is anonymized but the full dataset with names of PSF members is available upon request.

2 Categorizing Institutions of PSF Participants

2.1 Using Two Classifications

A major step in the treatment of data on PSF participants was the categorization of their institutions. This process involved classifying the institutions into distinct types, based on the nature of the institution and its activities. To ensure consistency and relevance, we developed two different classifications.

The first classification was based on the call for applications published by the PSF and the “classification form”, which candidates had to fill. The Commission classifies PSF members’ institutions into the following categories:

Academia, research Institutes and Think Tanks
Banks/Financial institutions
Companies/groups
Law firms
NGOs
Professionals’ associations
Professional consultancies
Trade and business associations
Trade unions
Other organisations

We could not access to the categories selected by the participants in the PSF. Consequently, we retrieved publicly available information about each institution to make informed approximations for the Commission’s classification. We faced two main issues: (1) Among PSF members, there is only one occurrence for the categories “Law firms”, “Professionals’ associations” and “Professional consultancies” while a large number of different organisations (especially public institutions) were grouped into the overly generic “Other”. Second, some categories, such as “Academia, research Institutes and Think Tanks” or “Banks/Financial institutions” were too broad, limiting their analytical utility.

To address these issues, we developed our own classification system, based on interviews’ data related to the participants’ alleged activities within the PSF, the potential motivations for their organizations’ involvement and the usefulness of the categories to answer the research question. We identified the following categories and provide explanations when they deviate from the Commission’s classification:

Academia and research Institutes. We removed think tanks from this category in order to have clearer delimitation vis à vis academia (see below for further clarification).
Banks/Financial institutions. This category has a narrower scope than the Commission’s classification because we excluded “Green Finance Think Tanks” and “Standards and Accounting organizations” (see below for further clarification).
Financial Regulators, Supervisors and Public Investors. This category includes mainly central banks and other financial regulators which are in the “Banks/Financial institutions” category of the Commission’s classification.
Green Finance Think Tanks. This category includes institutions tasked with producing research and policy recommendations on green finance. These institutions are typically classified under “Academia, Research Institutes, and Think Tanks” or “Banks/Financial Institutions” in the PSF framework. We isolated them in order to track the consequences of the Commission reorchestration of the PSF during its second mandate.
Standards and Accounting Organizations. This category includes institutions whose objectives are to develop financial accounting standards. While these institutions were conflated with “Banks/Financial institutions” in the Commission classification, we isolated them in order to track the consequences of the Commission reorchestration of the PSF during its second mandate.
Companies/Group
NGOs
Trade and business associations
Trade Unions & Social Finance. This category expands upon the PSF’s “Trade Unions” classification by including civil society organizations that promote social objectives in the financial sector.
Environmental Public Bodies. Public institutions with a mandate to advance environmental goals, such as Environmental Protection Agencies. These institutions are classified as “Other organisations” in the Commission classification.
Other organisations.

This refined and fine-grained classification allows us to better understand the dynamics of the PSF and its evolution over meetings. The graph below summarizes the differences between the two classifications systems. The height of the bars represents the proportion of institutions classified in each category.

2.2 Classifying Institutions

When developing new classification systems, researchers face issues of consistency in attributing categories to existing institutions. To assess consistency, two coders, the author of the article and a collaborating researcher, independently classified an initial small sample of institutions. Their results were compared to identify and resolve disagreements, thereby reaching a consensus on classification criteria. This agreed-upon approach was then applied independently to the entire dataset by both coders. Finally, inter-coder reliability was calculated to assess consistency.

Cohen’s Kappa is a robust measure of inter-coder reliability, as it accounts for the possibility of agreement occurring by chance. The results indicate a substantial level of agreement between the two coders in classifying institutions according to both the Commission’s classification system and our own. Yet, agreement was higher for our own classification system, reflecting its greater clarity and relevance.

category	Subjects	Raters	Kappa	z	p-value
Commission Classification	91	2	0.74	14.59	0
Own Classification	91	2	0.81	22.61	0

The Cohen’s Kappa test indicates that the classification process was generally consistent and reliable. Moreover, the number of cases involving significant uncertainty was too small to meaningfully impact the overall results. The table below shows how disagreements were discussed and resolved by the two coders.

2.3 Final classifications

The table below depicts the final classification outcome according to the Commission’s categories, as well as our own.

3 Evolution of Institutions over PSF Meetings

After classifying the institutions, we analyzed the evolution of the types of institutions represented in the PSF over time. We used two complementary approaches to examine this evolution:

First, we calculated the number of participants from each category for each meeting. Additionally, we computed the average number of participants from each category across the first and second PSF mandates (see horizontal lines in the graphs below).
Second, we calculated the proportion of participants from each category for each PSF mandate, enabling us to analyze changes in relative representation between the first and second mandates.

3.1 Evolution of Participants by Category

We plotted the changes in participant numbers for each of our two classification systems. As shown in the graph below, our own classification provides a more detailed and nuanced analysis of the evolution of PSF participants over time.

Evolution of the number of participants by PSF categories

Evolution of the number of participants by own-built categories

4 Interviews

The list of PSF participants served as the primary source for contacting individuals for interviews in a systematic way. Snowballing strategies were later employed, particularly to identify and reach MSEG members whose names are not publicly available. Exceptions were made for PSF members whose professional email addresses could not be retrieved or who attended only a few meetings. The low response rate (around 10%) among PSF participants can be explained by factors such as career changes, demanding schedules, and the high media salience of the topic. Despite this, the interview sample captures the diversity of stakeholders involved in expert groups, with the notable exception of Commission officials, who could not be reached for interviews. The interview guide focused on participants’ motivations for joining their expert group, the internal dynamics within expert groups, and their interactions with external stakeholders.

5 Other sources

Press articles on the Taxonomy were retrieved on the Financial Times (48) and Euractiv (57) websites. We also relied on the EU Council document register to retrieve the preparatory legislative and meeting documents related to the legislative process of the Taxonomy regulation (25), the publications of the PSF on its website (18), the minutes of PSF meetings on the register of Commission expert groups h and the documents related to delegated acts on the DG FISMA website (21) .