Marginal emissions: what they are, and when to use them

June 13, 2019


5 min read

Consider those two questions:

  • How much am I emitting by consuming electricity at a certain time?
  • What are the emissions avoided if I decide to consume electricity at a certain time?

Those questions have different answers! But why is that? The difference comes from the way the power system works:

The decision to use more or less electricity at a given time doesn’t cause all the power plants to evenly increase their production. As power plants are dispatched by increasing cost, it means that when electricity demand is increased, the first power plant to increase its production is the last dispatched one. In other words, the electricity requested will come from the cheapest power plant that still has spare capacity. We call that power plant the marginal power plant.


In a grid with 50% wind and 50% gas production, the consequence of deciding to consume electricity right now (instead of at another time) is that the gas turbine will increase its production (as it’s the only one that has spare capacity). This will cause emissions at a rate of ~500 gCO2 for each additional kWh consumed. The consequence of deciding to consume electricity now is an extra 500 gCO2 for each additional kWh consumed. We call that the marginal emission factor.

However, when accounting for emissions, it is assumed that grid users get a proportional share of all emissions of the grid. For example, if you consume electricity coming from 50% wind and 50% gas, you’ll be responsible for consuming electricity with a footprint of ~250 gCO2/kWh (the average). Your carbon footprint is 250 gCO2 per kWh consumed. We call that the average emission factor.

The consequence of your decision is 500 gCO2/kWh but your footprint is 250 gCO2/kWh 😳.. Confused? That’s because we’re here dealing with two ways of assigning responsibility:

In the average paradigm, responsibility is distributed evenly across grid users. In the 50% wind and 50% grid, if you were to consume electricity causing increased production from the gas turbine, everyone on the grid would be affected and have their footprint proportionally increased. Similarly, if you were to install a new wind turbine which produces low-carbon electricity, everyone on the grid would benefit from it and get a reduced footprint. The footprint of your electricity is the same as other users on the grid. Because benefits are distributed across all grid users, your claimable benefits will be limited.

In the marginal (also called consequential) paradigm, responsibility is distributed based on a causal relationship between decision and consequence: when your decision causes an increased production of the marginal power plant, the additional emissions are not distributed evenly amongst its users but are instead directed to the one who caused such a reduction. This leaves other grid users unaffected. Similarly, the benefits of installing a new wind turbine are directed exclusively to the owner, leaving the rest of grid users with the same emissions.

It becomes clear that both paradigms can’t be used simultaneously: it’s not possible to yield savings that are both fully captured by yourself, and also benefit all grid users. Therefore, marginal and average emissions should never be mixed together. It’s the same reason why carbon offsets can’t be deducted from carbon accounting (offsets affect everyone but are captured by the buyer).

A concrete example

Let’s assume for a second that we have a heater connected on a grid with electricity coming from 50% wind (at 0 gCO2eq/kWh) and 50% gas (at 500 gCO2eq/kWh). Let’s furthermore assume that this heater consumes on average 500W.

The average carbon intensity of the grid is 250 gCO2eq/kWh, which yields carbon emissions of 0.5 x 250 = 125 gCO2eq for a given hour.

Let’s assume that heater is smart as it has the ability to shift its consumption. Let’s furthermore assume that the heater was able to shift all of its consumption to an hour where no gas was needed, as all of the electricity came from wind turbines.

In our carbon accounting books, our electricity footprint is now at 0 gCO2eq for that particular hour. However, the emissions saved on the grid by avoiding the production of 500W of gas at 500gCO2eq / kWh (the marginal carbon footprint) amounts to 0.5 x 500 = 250 gCO2eq for that particular hour.

By buying a smart heater, we reduced our footprint from 125 g to 0 g (accounting). By buying a smart heater, we avoided 250 g of emissions (offsetting).


Simply put, marginal emission reductions should be considered as similar to carbon offsets, and therefore shouldn’t be mixed with average emissions which are used for accounting.

You might also have noticed that because the marginal emissions concept involves understanding the causal relationship between decision and consequence, it inevitably rests on the creation of a counterfactual (i.e. modelling a world where the decision wasn’t made). This is why marginal emission reductions are relative (to the counterfactual), whereas emissions accounting are absolute.

Finally, keep in mind that each marginal emission factors has its own domain of validity: if large-scale decisions are made (such as the decision to install a new factory somewhere), then the consequence might affect more than the dispatching of new power plants (it could cause the building of new power plants). Therefore, care should be taken in chosing the right type of marginal emission factors.

Interested in using our algorithm? See for more information.

Article written by
Olivier Corradi
Founder @ Electricity Maps, CEO

Sign up for news & updates

We will share occasional updates, news, and relevant content. Spam-free zone.