Emissions forecasting: The shape of things to come

Most transition risk analysis assumes corporate emissions stay flat. They rarely do. Some companies have cut emissions for years, while others keep adding to them. A flat baseline overstates future exposure for the first group and understates it for the second, and it says nothing about how certain any projection is.

Recent behavior, projected 1,000 ways

Monte Carlo Momentum (MCM) forecasts each company’s Scope 1, 2 and 3 emissions to 2030 from its reported data since 2020. It estimates momentum, a weighted average of year-over-year change that favors recent years, along with historical volatility. Each company and scope is then simulated 1,000 times, producing a median path and an explicit uncertainty range. Every forecast also carries the PCAF data quality score of its inputs.

Unreliable forecasts are left out

A company needs at least three valid years of reported emissions to be forecast. Implausible upward paths are removed, and forecasts whose uncertainty is more than twice the median are excluded altogether. In the S&P 500, most companies have stable enough emissions to forecast. The rest are too volatile or too thinly disclosed, and are not forecast.

Most accurate for the largest emitters

In backtesting, median forecast error was 23% for Scope 1, 34% for Scope 2 and 32% for Scope 3. For large and mega-cap companies, Scope 1 error falls to 7–20%, against 26–40% for micro and small caps. Random forest and Gaussian process models did slightly better on three-year training windows. With five years of data, MCM was the most accurate method tested.

A moving baseline for transition risk

MCM forecasts feed into Emmi’s scenario-based transition risk metrics, including Potential Carbon Liability, emissions reduction requirements and temperature alignment. Each is calculated alongside the constant-emissions baseline, which shows how much of a company’s future risk comes from its recent behavior. For companies with Science-Based Targets, the comparison shows how far current trends sit from stated commitments. Forecasts are available through FactSet and Snowflake.

Download the full white paper for the methodology and backtesting results.