About this episode
Around 83% of publicly listed companies don’t report their full Scope 1, 2 and 3 emissions. For investors, that leaves large gaps in portfolio footprints. Companies that do report are often 12 to 18 months behind their financial results.
This episode walks through the research behind our machine learning approach to closing that gap. We trained and tested more than 30,000 models, then combined the top 10 performers into a single Meta-Model. Using standard financial data, it estimates emissions for up to 45,000 listed companies.
What you’ll hear
The episode starts with why common estimation methods fall short. Sector averages give every company in an industry the same emissions factor. Input-output models only capture upstream Scope 3, while most Scope 3 emissions sit downstream.
It then covers how we checked our estimates against reported data, including a noise test to confirm the models hold up on data they haven’t seen. We estimate the listed equity universe emitted 11.4 billion tonnes in 2023, around 2.5% lower than 2022. Across major indices in the US, UK, Europe, Canada and Australia, the median weighted error is 15% for Scope 1, 17% for Scope 2 and 19% for Scope 3.
Read the full research
The white paper covers the full methodology, with results for each scope.


