Great Carbon Arbitrage

This repository contains the code for the great carbon arbitrage paper. The data required for the analysis is confidential, but you can run the self-contained version with public data (from NGFS and Our World in Data) in the public/ folder.

You should start by reading the back of the envelope version, public/back_of_the_envelope.py, which is self-contained and brief (~100 lines of code). You can run it by cd public && python back_of_the_envelope.py. This version uses renewables LCOE for calculating the investment cost. A more elaborate version (that has Wright's law learning for renewables investment) that uses only public data can be found at public/analysis_main_public_data.py. The only differences between the more elaborate version and the full private version are that it is missing the residual benefit and opportunity cost calculation.

For analysis using confidential data, you may start by reading analysis_simplified.py, which has < 500 LOC, but already reproduces the main result of the paper, i.e. carbon arbitrage opportunity in trillion dollars (but excluding residual benefit, for simplicity).

• analysis_main.py: contains the code for identifying the net social gain from phasing out coal, as well as the requisite climate financing to phase out coal
• analysis_simplified.py: simplified version of analysis_main.py, contains only code to calculate carbon arbitrage opportunity under restricted parameter choices.
• analysis_data_section.py: contains analysis for the data section
• processed_revenue: computes the median coal free cash flow per unit of coal production
• util.py: common code used by other files

The folder data_preparation/

• gdp/: contains data downloaded from https://data.worldbank.org/indicator/NY.GDP.PCAP.CD and https://data.worldbank.org/indicator/NY.GDP.MKTP.CD
• gdp_2020.py
• LearningSolarWind.csv.gz
• prepare_coal_price.py
• prepare_masterdata.py
• prepare_world_wright_learning.py
• Solar-Wind-Capacity.csv

The folder increasing_renewables/

In increasing_renewables.py output, we show that solar and wind are the dominant technology.

The folder misc/

• aggregate_beta.py: to calculate un_leveraged_beta in processed_revenue.py
• experience_curve.py
• hq_country_analysis.py
• NGFS_renewable_additional_capacity.py
• region_vs_world.py
• shiller.py: to calculate the 100 year CARP used in 100year rho_mode in util.calculate_rho
• unused_analysis.py

To reproduce the results

Based on this AR-2DII data our estimate of global coal production in 2020 is 6.41 Giga tonnes (Gt).

Figure 2

The folder data/

The public data can be found in the data/ folder in this Git repo. But the private data is confidential.

2DII-country-name-shortcuts.csv
all_countries_gdp_marketcap_2023.json
all_countries_gdp_per_capita_2023.json
country_ISO-3166_with_region.csv
developing.txt
developing_shortnames.csv
emerging.txt
emerging_shortnames.csv
GDP-Developed-World.csv
irena.json
NGFS-Power-Sector-Scenarios.csv.gz
ngfs_energy_demand_interpolated.csv
NGFS_region.json
NGFS_renewable_additional_capacity_GCAM5.3_NGFS.json
NGFS_renewable_dynamic_weight_GCAM5.3_NGFS.json
ngfs_scenario_production_fossil.csv
ShillerData.csv
TRISK-Data.xlsx
TRISK_data_NGFS_scenarios.csv
world_bank_group_carbon_tax.json

private

3-indices.csv
Arial.ttf
IEA-Scenarios.csv
masterdata_ownership.csv.gz
masterdata_ownership_2021_version.csv.gz
masterdata_ownership_PROCESSED_capacity_factor.csv.gz
revenue_data_coal_companies_confidential.csv.gz
revenue_data_companies_confidential_PROCESSED_coal.csv.gz