Merit Order and Marginal Cost

Every hour, in every organised electricity market on earth, the same operation runs. Generators say what they are willing to produce and what they need to be paid for it. The operator sorts those offers from cheapest to most expensive, and starts accepting them in order until accepted supply equals expected demand. The offer that closes that gap is the last one accepted, and it is called the marginal unit.

Then comes the rule that surprises everybody the first time they meet it. Every accepted generator is paid the price offered by that last unit. Not what they asked for. Not what it cost them. The nuclear plant that offered at effectively zero, the wind farm that offered below zero, and the old gas turbine that set the price all receive exactly the same price for every megawatt-hour they delivered in that interval.

This is a uniform-price auction, and it is the single most consequential design decision in the subject. Almost everything readers find strange about power markets, including several things that regularly become political scandals, follows from it.

The stack

The sorted list of offers is the merit order, and its shape is set bymarginal cost, meaning the cost of producing one more megawatt-hour from a plant that already exists. Capital cost does not appear. A nuclear station that cost a fortune to build has, relative to its overall costs, almost no incremental cost to run for one more hour, so it sits at the bottom of the stack and runs.

Demand never falls to zero. Even at four in the morning a floor of consumption still has to be met, and that floor is the baseload. Baseload generation refers to the plants that run continuously to meet it. They are built for high capacity factors because their financial models depend on steady output and their engineering makes them slow and expensive to start up or shut down. Nuclear reactors, large coal plants and geothermal facilities are the traditional backbone, and they tend to run whatever the hour.

The word describes how a plant is used rather than a property it holds, which is the Chapter 4 point about capacity factor arriving again. The merit order never asks whether a unit is baseload; it asks what the unit offers. French reactors have load-followed (increasing or decreasing output with changes in demand) since the 1980s because nuclear is too large a share of that system to sit still, and coal plants built to run flat now start and stop, because the midday hours are hours a solar-heavy market no longer needs them.

Demand is close to a vertical line drawn through that stack, because almost nobody consuming electricity is watching its price minute by minute. Where the vertical line crosses the staircase is the price. Move the line a little to the right on a tight day and the price does not rise a little. It jumps to the next step, and the steps at the top of the stack are very tall. That is the shape from Chapter 1, and it is why a power price chart looks like a flat line with occasional cliffs rather than anything resembling a crude oil chart.

Why the last unit sets the price for everyone

The obvious objection is that paying everybody the highest accepted price looks like overpayment. Why not pay each generator what it actually asked for, which is called pay-as-bid?

Because of what it would do to the offers. Under uniform pricing, a generator’s offer determines whether it runs, not what it earns. The best strategy is therefore to offer close to true marginal cost: offer higher and you risk being left out of a market you would have been happy to sell into, offer lower and you might be dispatched at a loss. The auction rewards telling the truth, and the resulting stack is a genuine map of what the system costs to run.

Under pay-as-bid, nobody would offer marginal cost, because doing so guarantees earning exactly marginal cost and nothing toward fixed costs. Every participant would instead try to guess the clearing price and offer just below it. Bigger firms with better forecasting would guess better. The offers would stop being cost information, so the operator would lose the signal it uses to dispatch efficiently, and there is no strong reason to expect the total bill to fall. Uniform pricing is what the system pays to get reliable cost information out of self-interested parties.

Share of generation is not share of price-setting

Now the idea this book will use more than any other, and the one most often got wrong by people who follow energy closely.

Wind and solar are a large and growing share of the electricity produced in most systems. It is natural to conclude that they must therefore be setting the price a large share of the time. The structure of the stack prevents it.

Because their marginal cost is near zero, they sit at the bottom of the stack. They are accepted early. By the time the operator reaches the point of asking which offer closes the remaining gap, the renewables are already dispatched and no longer under consideration. They are inframarginal, meaning below the margin. The unit that closes the gap is usually a nat gas plant, and the EIA states the outcome plainly, that natural gas prices set the marginal price of electricity during most hours in most regional markets.

So renewables push the price down without setting it. Adding a wind farm moves the whole staircase to the right, which means the vertical demand line crosses it at a lower step. The price falls. But the identity of the unit standing on that step is still, in most hours, something that burns nat gas.

Inframarginal rent, and the year it became law

A generator paid the marginal price while running at a lower cost earns the difference. That difference is inframarginal rent, and the design depends on it. It is how capital-intensive plant recovers the cost of having been built at all, since the auction pays nothing for capital directly.

This is normally an economist’s term. In 2022 it became a European regulation. When nat gas prices rose to levels no design had contemplated, the marginal unit’s offer rose with them, and every wind farm, solar park, nuclear station and lignite coal plant on the system was paid that price while its own costs had barely moved. The rents were enormous and extremely visible, and the political system discovered uniform-price auctions in public and did not care for them.

Council Regulation (EU) 2022/1854, adopted on 6 October 2022, capped the market revenues of exactly those technologies at 180 euros per megawatt hour, with the excess collected by member states and returned to consumers. The word inframarginal appears in the legal text. A concept that exists only because of the clearing rule described earlier in this chapter had become, within about a year, a thing governments confiscated. Chapter 19 takes up what that episode did to European market design.

Two consequences to carry forward

The first is where renewables do set the price. It happens when there is so much near-zero-cost generation that the operator never has to reach the thermal part of the stack at all, and the marginal unit is itself a wind or solar plant. That is precisely the circumstance in which the price collapses toward zero, or below it once subsidies make generating worth more than stopping. Renewables therefore set the price mainly in the hours when the price is lowest, which is the subject of Chapter 11, while nat gas sets it in the hours that cost money, which is Chapter 12. Their price-setting share is real and concentrated at one end.

The second is the conclusion that follows, and it is not intuitive: building cheap generation does not straightforwardly produce cheap electricity. It lowers the price in the hours the cheap generation is available, which lowers the revenue earned by everything else on the system, including the plant that will be needed on a still winter evening when the cheap generation is not there. The stack does not stay put while you add to the bottom of it. Whether anything is left to run at the top, and who pays for the privilege of having it standing by, is the argument that Chapters 12 and 16 are about, and it is the argument that has divided every electricity market in the world.