How Low Can Prices Go, and Why?

A quick follow up to the previous post.

I noted that negative prices have been a thing in Texas for years. Indeed they are in every market with substantial renewables penetration.

This is particularly true in the US, where the Production Tax Credit pays qualifying renewables facilities $23/MWh to produce, regardless of prices. Meaning that a recipient of the PTC will continue to produce even if prices are -$22.99/MWh.

So why do prices go negative? In particular, why do other producers who do not get the credit continue to produce even when there are negative prices? Why don’t enough of them cut output to make sure that prices don’t fall below variable cost?

The answer in a word is: indivisibilities. Or, if you prefer, non-convexities.

Specifically, many thermal generators incur costs to shut down or start up. These are basically fixed costs, of the avoidable variety. A unit currently operating can avoid shutdown costs by continuing to operate. A unit currently idle can avoid startup costs by remaining idle. Minimum run times and ramping constraints are other examples of non-convexities.

So, for example, when demand is low and wind turbines continue to blend birds (and generate electricity), prices can go negative but a gas or coal or nuke plant may continue to operate (and incur fuel costs as well as incremental O&M) because it is cheaper to PAY to sell output (and pay variable costs as well) than it is to shut down.

If the cost of adjusting output of a plant to or from zero was zero, whenever prices fall below marginal operating cost the plant would shut down. This would put a floor on prices equal to marginal cost. However, if there is a fixed cost of adjusting output to or from zero, it can make sense to continue to operate even when prices do not cover variable costs–and when prices are negative–in order to avoid paying this cost of shutting down (and/or the cost of starting back up again when prices are higher).

Generation technology is such that efficient baseload plants (i.e., units with lower per MW variable costs) tend to have higher shutdown and startup costs, and more acute operating constraints that give rise to other forms of non-convexity. As in all things in life, there tends to be a trade-off: low variable costs must be traded off against higher avoidable costs/less flexibility to adjust output. Thus, negative prices hit such units especially hard. They are faced with the bleak choice between paying to sell what they produce, or paying a cost to avoid producing. Obviously this choice is bleaker, the costlier it is to avoid producing. For many, the cost of shutting down is big enough that they continue to spin even when prices are negative.

Economists have long known that non-convexities can interfere with the operation of a price system. If you look at classic Arrow-Debreu proofs of the welfare theorems (i.e., that competitive prices call for the efficient level of production and consumption), you’ll see that they assume that production technologies are convex. That is, they assume away things like shutdown and startup costs. When production technologies are characterized by non-convexities (e.g., fixed avoidable costs), the proofs don’t go through.

Indeed, an equilibrium in prices and output may not exist if indivisibility problems are sufficiently severe: in my earliest academic life, my work on applying core theory focused on this issue. If an equilibrium does exist, it may be inefficient.

Put simply, the invisible hand can get really shaky if indivisibility problems are severe.

Liberalized electricity markets (e.g., PJM and other ISOs) have devised various means of addressing these indivisibilities. The results are not first best, but the mechanisms allow an energy market with prices approximately equal to marginal cost to survive.

The subsidization of wind, especially through the PTC, greatly exacerbates indivisibility/non-convexity problems because its effects fall with particular force on generating units with more pronounced indivisibilities. These tend to be the most efficient, and also the ones most essential for maintaining reliable system operation.

This means that although renewables subsidies punish investment in thermal generation generally, they punish investment in units that operate nearly continuously at low cost with particular severity. Having these units available nearly 24/7/365 is vital for keeping electricity prices low, and for ensuring a highly reliable power system.

So the distortions caused by renewables subsidies, particularly of the pay-to-produce variety, are more severe than “we have too much renewables capacity and too little thermal capacity.” Yes, that’s a problem, but the distorted price signals also distort the types of generation invested in. In particularly, they are particularly punitive to generation with more acute indivisibilities. Since these also tend to be low operating cost, high reliability technologies, that is a very costly distortion indeed.