Asset-based Company Data (ABCD)

• required input
• external source
• XLSX file

This data set provides information on the production profiles and emission intensities of companies active in the following real economy sectors: Automotive (light-duty vehicles) manufacturing, aviation, cement production, coal mining, upstream oil & gas extraction, power generation, and steel production. The ABCD is typically obtained from third party data providers. However, it is possible to prepare the ABCD yourself or complement an external data set with entries that may not be covered out of the box.

The ABCD data set must be an XLSX file and contains the following columns:

• company_id
• name_company
• lei
• is_ultimate_owner
• sector
• technology
• plant_location
• year
• production
• production_unit
• emission_factor
• emission_factor_unit

While PACTA is data agnostic and allows using data from any provider that offers the appropriate format, one option to obtain ABCD for this analysis is to buy the data from the data provider Asset Impact. Further information on how to obtain ABCD for PACTA and documentation of the individual sectors and data points can be found on the Asset Impact website.

Scenario Data

• required input(s)
• external source
• CSV file(s)

The scenario data set provides information on the trajectories of technologies/fuel types and of emission intensities pathways for each of (or a subset of) the sectors covered in PACTA.

For sectors with technology level trajectories, the data set provides the TMSR and SMSP pathways based on the Market Share Approach, an allocation rule that implies all companies active in a sector have to adjust their production in a way that keeps market shares constant and solves for the aggregate climate transition scenario. For more information on how to calculate the TMSR and the SMSP, see the PACTA for Banks documentation.

The target market share scenario data set must be a CSV file and contains the following columns:

• scenario_source
• region
• scenario
• sector
• technology
• year
• smsp
• tmsr

For sectors that do not have technology level pathways, PACTA uses the Sectoral Decarbonization Approach (SDA), an allocation rule that implies that all companies in a sector have to converge their physical emission intensity at a future scenario value - e.g. in the year 2050. This implies that more polluting companies have to reduce their physical emissions intensity more drastically than companies using cleaner technology. It does not have any direct implications on the amount of units produced by any company. For further information on calculating the SDA in PACTA, please see the PACTA for Banks documentation.

The SDA scenario data set must be a CSV file and contains the following columns:

• scenario_source
• region
• scenario
• sector
• year
• emission_factor
• emission_factor_unit

While the raw input values of the scenarios are based on models from external third party organisations - such as the World Energy Outlook by the International Energy Agency (IEA), the Global Energy and Climate Outlook by the Joint Research Center of the European Commission (JRC), or the One Earth Climate Model by the Institute for Sustainable Futures (ISF) - the input data set for PACTA must be prepared using additional steps, which are documented publicly on the following GitHub repositories:

• pacta.scenario.data.preparation
• workflow.scenario.preparation

Since RMI has taken over stewardship of PACTA, the prepared scenario files can also be accessed as CSV downloads on the PACTA website under the “Methodology and Documents” tab of the “PACTA for Banks” section. The files are usually updated annually based on the latest scenario publications.

Raw Loan Books

• required input
• self-prepared
• CSV file(s)

The raw loan books are the financial data sets that you would like to analyze. They contain information on the loans that banks have provided to companies. As a supervisor, the data required to construct these data sets will typically be available to you through regulatory filings that are accessed via internal data bases or similar. As a bank, the data required will be available in your internal systems.

The raw loan books must be prepared as CSV files and contain at a minimum the following columns:

• id_loan
• id_direct_loantaker
• name_direct_loantaker
• id_ultimate_parent
• name_ultimate_parent
• loan_size_outstanding
• loan_size_outstanding_currency
• loan_size_credit_limit
• loan_size_credit_limit_currency
• sector_classification_system
• sector_classification_direct_loantaker
• lei_direct_loantaker
• isin_direct_loantaker

NOTE: The tool will automatically add a column group_id to each of the loan books, which uses the file name as a value. This allows you to group the results the analysis by loan book, using the by_group parameter in the config.yml file. For any other variable that you may want to group the results by, you need to add a column to the raw loan book files that you then provide as the by_group parameter in the config.yml file.

For detailed descriptions of how to prepare raw loan books, see the PACTA for Banks documentation and navigate to the “Training Materials” tab of the “PACTA for Banks” section. The “User Guide 2”, the “Data Dictionary”, and the “Loan Book Template” files can all be helpful in preparing your data.

Misclassified Loans

• optional input
• self-prepared
• CSV file

The user can provide a list of loans that have been misclassified in the raw loan books. The aim here is specifically to remove false positives, that is, loans that are classified in scope of one of the PACTA sectors, but where manual research shows that the companies do not actually operate within the PACTA scope. Such a false positive may be due to erroneous data entry in the raw loan book, for example. Removing these loans from the falsely indicated sector in the calculation of the match success rate will give a more accurate picture of what match success rate can really be reached.

Asset-Based Company Data (ABCD) for company sector split

• optional input
• external source
• XLSX file

In case the user wants to split company exposures across sectors of in scope activity, the user must provide a version of the ABCD data set that follows the format of the Advanced Company Indicators data set by Asset Impact. This data set includes power generation values which are required for the primary energy based sector split. Again, more information on data from Asset Impact can be found on the Asset Impact website.

Companies to apply primary energy split on

• optional input
• self-prepared
• CSV file

When applying the sector split on company exposures, the user can provide a list of companies for which the sector split should be based on primary energy content. For all other companies, a simple equal weights split will be applied. For more information on the sector split, see the documentation in vignette("sector_split").

Manual Sector Classification

• optional input
• self-prepared
• CSV file

TODO: Add once we have a stable implementation of this feature

R (version 4.1.0 or higher)

R is the programming language that the pacta.multi.loanbook package is written in. You can download R from the Comprehensive R Archive Network (CRAN).

RStudio (optional)

RStudio is an integrated development environment (IDE) for R developed by Posit. It is not strictly required to run the analysis, but it can be helpful for managing your project and running the analysis. Generally, RStudio is very widely used among the R community and probably the easiest way to interact with most R tools, such as pacta.multi.loanbook. RStudio Desktop is an open source tool and free of charge. You can download RStudio from the Posit RStudio website.

pacta.multi.loanbook R package

The pacta.multi.loanbook package is the main software tool that you will use to run the PACTA for Supervisors analysis.

install.packages("pacta.multi.loanbook")

pak::pak("r-lib/pkgdown")

pak::pak("r-lib/pkgdown")

Required R packages

The pacta.multi.loanbook package depends on a number of other R packages. These dependencies will be installed automatically when you install the pacta.multi.loanbook package. The required packages are:

cli (>= 3.2.0), config, dplyr, ggplot2, glue, networkD3, r2dii.analysis (>= 0.3.0), r2dii.data (>= 0.5.0), r2dii.match (>= 0.2.0), r2dii.plot (>= 0.4.0), readr (>= 2.0.0), readxl, rlang, scales, tidyr, webshot, withr, yaml, yesno

Suggested R packages

The suggested packages are not required to run the analysis, but they are used in the examples and vignettes provided with the package:

gt, knitr, pkgdown, rmarkdown, usethis, testthat (>= 3.1.9), tibble, writexl

FAQ

install.packages(c(...))

Config

All of the functions needed to run a PACTA for Supervisors analysis take a config argument, which can either be a path to a config.yml file (see vignette("config_yml")) or a config list object containing previously imported settings from a config.yml file. All of the settings/options are configured with this config.yml file.

Input/Output folder structure

The config.yml file then points to an input directory and to four output directories, one for each user-facing function, that the user can choose anywhere on their system, as long as R has read and write access to these directories. A recommendable choice to structure an analysis project would be to place all these folders and the config.yml file in a project folder. The input folder must contain all input files as described above, with the raw loan books being placed in a sub-directory that must be named "loanbooks". The output folders will automatically be created at the locations indicated in the config.yml file. They will be populated by running the analysis. NOTE: Re-running a step of the analysis will replace the entire corresponding output directory, if the directory already exists.

An example of how the project folder could be structured:

project_folder
├── config.yml
├── input
│   ├── ABCD.xlsx
│   ├── loanbooks
│   │   ├── raw_loanbook_1.csv
│   │   ├── raw_loanbook_2.csv
│   │   └── ...
│   ├── scenario_data_tms.csv
│   ├── scenario_data_sda.csv
│   └── ...
├── prepared_abcd
├── matched_loanbooks
├── prioritized_loanbooks_and_diagnostics
└── analysis

Notice that the names of the directories can be changed to something the user may prefer. The names were chosen for illustrative purposes to reflect the corresponding user-facing functions that create the outputs.