top of page
  • Writer's pictureBen McNeil

Upgrade: New Release of Emmi Emissions Estimates

Updated: Apr 4, 2023


Scope 1

Scope 2

Scope 3

Summary of what you need to know:

Emmi coverage and prediction quality

  • We have released a significant upgrade in our methods to predict carbon emissions for the global equities universe using machine learning

  • Emmi coverage has expanded 18% to 49,829 companies globally

  • CY 2022 are now estimated for 23,948 of the world’s largest companies. This will expand as new financial data gets reported over the coming months.

  • We can now predict Scope 1+2 emissions for the world's largest companies within ~5-10%. This is a significant improvement from typical industry standard methods and our previous models.

  • We have incorporated new PCAF aligned methods and provide associated scores for each holding

What does it mean for your portfolio?

  • For a typical medium-large cap portfolio (eg the ASX300), our new more accurate emissions predictions for Scope 1 + 2 are ~15% lower than our previous predictions

  • For Scope 3, our new models predict even lower emissions in general in comparison to our previous version

  • For financed emissions, you should expect a slightly lower portfolio footprint (including Scope 1+2)

  • For portfolio level carbon liabilities that include Scope 3, also expect lower carbon liabilities

We are excited to release Emmi carbon prediction estimates for nearly 50,000 companies from 2015-2022 in a significant upgrade to customers. This new data release was the culmination of the 2 years of research & development with the University of Otago that have produced award winning machine learning models that predict corporate emissions using financial data. I wanted to share why this was important and what this means for your portfolios.


Emmi Emissions Predictions

Version 1.0: Industry standards to expand coverage

In order to expand the poor coverage of corporate emissions data, our first carbon estimates relied on multiple linear regression models and the industry fill method which is widely used across data providers to fill carbon data gaps. Read here a blog post that explains our initial methods, which allowed Emmi to provide customers with expanded coverage to over 40,000 global companies.


Version 2.0: A new standard in estimating company emissions using machine learning

Machine learning is a catch-phrase for mathematical algorithms that ‘learn’ by classifying patterns in data for a given set of inputs and predictor variables. This algorithmic approach has the potential to find and extract the most predictive information from a given data-set. Overall machine learning has the potential to improve the quality of estimates with broader coverage.

In partnership with our university partners, our data science team at Emmi have developed novel new machine learning methods to estimate corporate carbon emissions. If you want details about our approach, please read a detailed overview of this method and an analysis here.

To help you better understand the likely changes you will see and the implications for your portfolio, we have included a detailed analysis using ASX300.


ASX300 Comparison

Scope 1

For company carbon footprints within the ASX300, our new predictions are within ~2% of reported emissions in comparison to our previous models.



Overall - out new estimates are ~13% lower for Scope 1, however when looking at sectors breakdowns, the changes are 2-5% for the most carbon intensive sectors (Utilities and Materials). Industrials and Energy sectors are ~20% lower via our new methods.


Scope 2

For company carbon footprints within the ASX300, our new predictions are 10% of reported emissions in comparison to our previous models.


Overall - our new estimates are ~15% lower for Scope 2, however when looking at sectors breakdowns, the changes are 2% for the most carbon intensive sector (Materials). Industrials, Utilities and Energy sectors were between 5-20% lower via our new methods.

Scope 3

For company carbon footprints within the ASX300, our new predictions are closer to reported emissions in comparison to our previous models, however still significant uncertainty exists for Scope 3 for a range of reasons including incomplete or biased reporting and difficulty in quantifying emissions for downstream categories including use of sold products.




The largest changes using our new predictions is for Scope 3. For many companies our new estimates are >200% higher for Scope 3. By sector, there was a 2000% increase within the Financials sector, which skews the changes in the above plot. For the financial sector, Scope 3 emissions are dominated by their financed emissions which are typically undisclosed. However a number of banks globally have disclosed financed emissions with PCAF - which allows us to estimate a much more accurate Scope 3 estimate for the financial sector. Previously, the financial sector was reported small Scope 3 emissions - which is incorrect.

In general our new Scope 3 estimates are lower for other sectors - including the most carbon intensive (Materials). For example our Scope 3 estimate for BHP was lower by less more than half from above 600million tonnes (see chart above), closer to the reported emissions estimate for Scope 3.


How does the new estimates change a typical portfolio?

If investing $50mil into an equally weighted ASX300 portfolio, Scope 1+2 emissions are about 26% lower with our new more accurate emissions predictions. For Scope 3, the change is about 15% lower than using our previous emissions estimates.

For financed emissions, you should expect a slightly lower portfolio footprint (incl. Scope 1+2)
For portfolio level carbon liabilities that include Scope 3, also expect lower carbon liabilities

Kommentare


bottom of page