Attributional and Consequential Methods of Quantifying Emissions.
The question “what is the environmental impact of a digital service?” is very open ended. It’s hard to come up with a neat answer because there is just so much to consider. Even if we restrict our attention to carbon emissions (far from the only environmental impact that digital services have) the question is still imprecise. We could be asking “What carbon emissions occur as a result of my actions?” or potentially “What proportion of carbon emissions am I responsible for?”.
The above image visually represents the conceptual difference between the 2 methods. The circles each represent the total impact over some fixed time period. The attributional method assigns a portion of total emissions to some service, whereas the consequential method deals with change in the total emissions based on an action taken.
In this blog post I’d like to compare two different approaches for quantifying environmental impact: attributional, and consequential. Specifically focused on emissions from digital services.
Consequential approach
The consequential approach to measuring emissions aims to answer the question: “What carbon emissions occur as a result of providing a digital service?”. It works by considering the emissions in scenarios when different actions are taken (or not taken) and generally requires knowledge of the hardware involved in providing the service and its associated emissions. By quantifying emissions in this way, we can directly apply the results to decision making since we learn about the consequence of each decision. That decision could be “do I compress the images on my website?”, and knowing what the consequences in terms of emissions for each possible option, compressing or not compressing, is extremely useful when trying to minimise emissions.
This sounds great in theory, knowing the impact of all of our actions, and just picking the best ones! But unfortunately, in reality it is quite hard to calculate what the consequences of our actions are. Especially in the world of digital, where hardware is typically serving multiple functions simultaneously.
It gets even more complicated when you throw time in the mix. Short term and long-term consequences may differ. For example, choosing to plant a tree today might have a negative short-term impact due to the fuel used to drive the car to the field to plant the tree, but in the long term, that tree may reabsorb more carbon than was emitted during its planting. For a more digital example, short term consequences of digital services may be limited to the increased energy used by providing a service, but long-term consequences could also include things such as the manufacturing of new hardware to keep the service running.
Attributional approach
The attributional approach aims to answer the question: “What proportion of carbon emissions am I responsible for?”. It works by taking the total of all emissions of a system and attributing responsibility for the emissions to different users of the system based on some proportion of the resources used. In the case of an online service, we count emissions from datacentres, networking, and end user devices, and divide them up based on e.g., number of gigabytes transferred, or computing time used.
A strength of the attributional approach is its simplicity. It is a holistic approach, and does not require as much specific knowledge of the hardware being used to deliver a service, as responsibility is assigned based on higher level average data.
A major limitation lies in the very great sensitivity to the choice of allocation rule (Hubblo blog). For example, when assigning responsibility for emissions from network hardware, you could do so based on amount of time spent transmitting data, or amount of data transmitted, or number of users of the network, etc. Depending on the allocation rule used, attributed emissions could vary significantly. The choice of allocation rule is fairly subjective. The attributional approach is also prone to inaccuracy for larger systems, since specific hardware and software may vary significantly within a large system (especially one as large as the internet).
Data
Both approaches require data to come up with estimations for emissions. There is a distinction to be made between marginal and average data. This distinction in the types of data is central to the distinction between the two methods.
Average data is data about the whole impact of a system. For example, the total emissions from all electricity generation in 2024. The attributional approach would then assign responsibility for these emissions proportional to energy usage in the year 2024.
Marginal data is data about the impact of changes to a system. For example, the change in emissions per change in energy usage right now. The consequential approach could use this data to quantify the short-term impact of turning off the lights for the rest of the day would be.
Average data tends to be more available than marginal data, which is one of the reasons for the attributional method being generally easier to implement.
Use cases
The attributional method is useful for identifying what the largest sources and sinks of carbon are, but should not be used directly when informing actions to reduce emissions. When trying to implement changes to our practices caution is needed since it is possible that in taking actions to reduce attributed emissions, you actually end up increasing emissions elsewhere, see Use of U.S. Croplands for Biofuels Increases Greenhouse Gases Through Emissions from Land-Use Change | Science. For this reason, it is important to consider consequential emissions of the changes too, which highlights the need for both approaches.
How it relates to my internship
Throughout my internship so far, I have primarily considered attributed emissions for our online teaching resources. But after reading more about the topic I feel it is important to try and better understand the consequential ones too. In some way they feel to me more tangible than attributed ones, and a reduction in consequential emissions has the intuitive interpretation that emissions will be reduced as a result of your actions. Though as I said before, there are no easy answers, and applying multiple approaches can give a more complete picture.
There is so much more to be said about attributional and consequential methods. If you’re interested in some follow-up reading, then check out Matthew Brander’s paper on the attributional-consequential distinction, David Ekchajzer’s article Counting environmental impacts using attributional and consequential approaches, or Tomas Ekvall’s paper Attributional and Consequential Life Cycle Assessment.
If you have any thoughts or questions about anything you’ve just read, please drop me an email at olaundon@ed.ac.uk. I’d love to hear your thoughts!
Author: Otis Laundon
Sources
Brander, Matthew, and Matthew Brander. “The Most Important GHG Accounting Concept You May Not Have Heard Of: The Attributional-Consequential Distinction | Greenhouse Gas Management Institute.” Greenhouse Gas Management Institute, 21 Apr. 2021, ghginstitute.org/2021/04/21/the-most-important-ghg-accounting-concept-you-may-not-have-heard-of-the-attributional-consequential-distinction/. Accessed 7 July 2025.
David Ekchajzer. “Counting Environmental Impacts Using Attributional and Consequential Approaches.” Hubblo.org, 2016, hubblo.org/blog/attributional-vs-consequential/. Accessed 7 July 2025.
Ekvall, Tomas. ‘Attributional and Consequential Life Cycle Assessment’. Sustainability Assessment at the 21st Century, IntechOpen, 12 Feb. 2020. Crossref, doi:10.5772/intechopen.89202.
(Weldema B P, Market information in life cycle assessment - 2-0 LCA )
(Weldema B P, Market information in life cycle assessment - 2-0 LCA )
Share
