Why I've Grown Skeptical of the Value of the Concept of Marginal Emissions

One of the things that happens when you start a company is people ask you how you are different than other companies that already exist. When we decided to call our company WattCarbon (I had been partial to Aristotle Energy, but that’s a story for a different day), we knew that we’d get asked questions like this, particularly because a non-profit called WattTime has been around for a while and generates a fair amount of publicity as an affiliate of the Rocky Mountain Institute. WattTime calculates a marginal emissions number for grids around the US and markets it as a way to figure out how changes in energy consumption can reduce pollution. It makes intuitive sense and lots of organizations have partnered with them to use their number.

“So, how is WattCarbon different than WattTime?” asked nearly every venture capitalist I met with while we raised our pre-seed round. There are two ways to answer this question - get into the boring technical details or offer a broad vision about the future of carbon accounting. Our broad vision involves making it easy to calculate accurate scope 2 emissions for buildings, which is a lot easier to explain than the nuances of marginal emissions. But it’s worthwhile to also describe why I’ve grown increasingly skeptical of the notion of marginal emissions. If this is interesting to you, keep reading.

What is a Marginal Emission?

I don’t mean to pick on WattTime specifically here. Plenty of other organizations have adopted the construct of marginal carbon emissions. The premise is seductive. Conceptually, our electrical grids are anchored by a collection of power plants that get dispatched in a particular order depending on the amount of demand. The order of dispatch is largely determined by the price offered by the producers, but can also be influenced by longer term contracts and other rules. The theory of marginal emissions is that there is a hypothetical final power plant that gets dispatched to meet the final watt of demand at any given time on the grid. If you stopped charging your car, or if you turned off your A/C, that plant could be turned down and the consequent emissions from that plant avoided. These would be your marginal emissions.

How do Grids Balance Demand?

Over the past few months, I’ve tried to ask as many people as I can to help me understand how grids operate. The most intuitive way for me to make sense of it is to think of the grid as operating in a harmonic state. There are a variety of different balancing mechanisms that keep the harmonics within the necessary range, one of which is controlling the combustion rate of certain plants, but the causal relationship between change in demand in a particular place and change in combustion rate at a certain facility is weak at best. The operations of the grid have their own logic that make it tricky to attribute specific changes in demand to specific changes in supply.

What are We Measuring?

Still, assuming that we could draw a straight line between your building and the combustion rate of certain facilities, the concept of marginal emissions is itself a conflation of two different ideas. Some people talk about marginal emissions as a counterfactual (i.e, what would have not been dispatched if demand was lower?). Others talk about marginal emissions as a time-series (i.e., by how much did emissions change from time A to time B?) If in time A the emissions were 1000 lbs/mWh and at time B the emissions were 750 lbs/mWh and you moved your consumption from time A to time B, would your marginal emissions be 250 lbs/mWh or would they be the emissions of the plant that didn’t have to be turned on in time A because you shifted your demand to time B (where theoretically a different plant would be dispatched instead)?

For simplicity, let’s just say that you are adhering to the construct of avoided emissions in a single time period. You are using a counterfactual grid scenario. Because you didn’t charge your car, the grid operator avoided calling on a dirty peaker plant. Under normal circumstances you would be charging your car and the grid operator knows this. So they get ready to dispatch the next-in-line peaker plant, but then because you haven’t plugged in, the demand never appears and the grid operator has to cancel the order. You’ve, in essence, fooled the grid operator.

Why is this tricky?

The counterfactual basis of this construct is easy to misconstrue. If you are claiming avoided emissions based on what the grid “thought” you were going to use, the only avoided emissions you can take credit for are those associated with the energy that you didn’t use. For that to happen, you have to create a counterfactual demand-side energy model for your hourly energy and calculate the difference between what would have been predicted versus what actually happened. It’s this difference between what was predicted and what actually happened that is the “marginal” energy.

This hourly counterfactual is really hard to generate. I know this because I spent five years at Recurve developing the methodology and tools required for it. I have yet to see any evidence that marginal emissions reduction claims are appropriately calculating avoided energy use on the demand side. Instead, the calculation is made using old demand response concepts - namely, taking the capacity value of a device and crediting it for not turning on.

What’s the alternative?

For the most part, grid operators are pretty good at anticipating demand. Most of our daily electricity consumption is already accounted for in long term contracts and capacity requirements. We can be assured that in California, for example, the same solar panels that powered the grid yesterday afternoon will come online again today and provide roughly the same amount of power (a few clouds here and there notwithstanding). In that sense, we don’t need to fool the grid. Quite the opposite in fact. We need to load up our consumption on those hours of the day when the sun is shining. We need the grid operators, the planners, the regulators, the developers, and everyone else to start expecting more demand when there is more clean energy on the grid.

The concept of a marginal emission fails to capture this larger dynamic. According to WattTime, as I write this, Seattle City Light, with its nearly 100% carbon-free hydro power has a higher emission factor than Portland General Electric, with its multitude of fossil fuel generation plants. This fails the sniff test.

The more I reflect on this, the more it seems better to consider the full mix of energy resources that are being dispatched. The lion’s share of our energy consumption is not at the margin. It’s completely predictable. We can make intentional choices to organize our energy consumption around the times of day in which more clean energy is available and avoid the times of day when more of our energy is being produced by fossil fuels.

Thankfully, we already know the full stack of what actually gets dispatched. System operators publish this information in real time. Here’s California’s data: https://www.caiso.com/todaysoutlook/Pages/supply.html. Companies like Singularity Energy and ElectricityMap are aggregating the feeds from these system operators and doing useful cleaning of the data (e.g., accounting for imports and exports between balancing authorities).

There might be some niche use cases where marginal emissions calculations are appropriate, especially where the demand-side resource is bid directly into a market and it could be determined what the losing resource might have been. For obvious reasons, this information is not generally made public, so at some point we’ll need an open-source, transparent marginal emissions methodology that can be verified and independently reproduced. Until then, it seems better to stick to what’s actually happening.

I may be missing something here, and could probably still be convinced that the marginal emissions concept has some value that I’ve failed to appreciate. If that’s the case, please help me connect the conceptual dots. At the end of the day, we need to make sure that we’re systematically replacing carbon-based fuels with non-carbon based fuels to power our grid. Let’s make sure we measure our progress in a way that allows us to account for our carbon emissions appropriately.