Touching the Third Rail: The Dangers of Electricity Market Design
In the aftermath of the Texas Freeze-ageddon much ink and many pixels have been spilled about its causes. Much–most?–of the blame focuses on Texas’s allegedly laissez faire electricity market design.
I have been intensely involved (primarily in a litigation context) in the forensic analysis of previous extreme electricity market shocks, including the first major one (the Midwest prices spike of June 1998) and the California crisis. As an academic I have also written extensively about electricity pricing and electricity market design. Based on decades of study and close observation, I can say that electricity market design is one of the most complex subjects in economics, and that one should step extremely gingerly when speaking about the topic, especially as it relates to an event for which many facts remain to be established.
Why is electricity market design so difficult? Primarily because it requires structuring incentives that effect behavior over both very long horizons (many decades, because investments in generation and transmission are very long lived) and extremely short horizons (literally seconds, because the grid must balance at every instant in time). Moreover, there is an intimate connection between these extremely disparate horizons: the mechanisms designed to handle the real time operation of the system affect the incentives to invest for the long run, and the long run investments affect the operation of the system in real time.
Around the world many market designs have been implemented in the approximately 25 year history of electricity liberalization. All have been found wanting, in one way or another. They are like Tolstoy’s unhappy families: all are unhappy in their own way. This unhappiness is a reflection of the complexity of the problem.
Some were predictably wretched: California’s “reforms” in the 1990s being the best example. Some were reasonably designed, but had their flaws revealed in trying conditions that inevitably arise in complex systems that are always–always–subject to “normal accidents.”
From a 30,000 foot perspective, all liberalized market designs attempt to replace centralization of resource allocation decisions (as occurs in the traditional integrated regulated utility model) with allocation by price. The various systems differ primarily in what they leave to the price system, and which they do not.
As I wrote in a chapter in Andrew Kleit’s Energy Choices (published in 2006) the necessity of coordinating the operation of a network in real time almost certainly requires a “visible hand” at some level: transactions costs preclude the coordination via contract and prices of hundreds of disparate actors across an interconnected grid in real time under certain conditions, and such coordination is required to ensure the stability of that grid. Hence, a system operator–like ERCOT, or MISO, or PJM–must have residual rights of control to avoid failure of the grid. ERCOT exercised those residual rights by imposing blackouts. As bad as that was, the alternative would have been worse.
Beyond this core level of non-price allocation, however, the myriad of services (generation, transmission, consumption) and the myriad of potential conditions create a myriad of possible combinations of price and non-price allocation mechanisms. Look around the world, and you will see just how diverse those choices can be. And those actual choices are just a small subset of the possible choices.
As always with price driven allocation mechanisms, the key thing is getting the prices right. And due to the nature of electricity, this involves getting prices right at very high frequency (e.g., the next five minutes, the next hour, the next day) and at very low frequency (over years and decades). This is not easy. That is why electricity market design is devilish hard.
One crucial thing to recognize is that constraints on prices in some time frames can interfere with decisions made over other horizons. For example, most of the United States (outside the Southeast) operates under some system in which prices day ahead or real time are the primary mechanism for scheduling and dispatching generation over short horizons, but restrictions on these prices (e.g., price caps) mean that they do not always reflect the scarcity value of generating or transmission capacity. (Much of the rest of the world does this too.) As a result, these prices provide too little incentive to invest in capacity, and the right kinds of capacity. The kludge solution to this is to create a new market, a capacity market, in which regulators decide how much capacity of what type is needed, and mandate that load servers acquire the rights to such capacity through capacity auctions. The revenues from these auctions provide an additional incentive for generators to invest in the capacity they supply.
The alternative is a pure energy market, in which prices are allowed to reflect scarcity value–and in electricity markets, due to extremely inelastic demand and periodic extreme inelasticity of supply in the short run, that scarcity value can sometimes reach the $1000s of dollars.
Texas opted for the energy market model. However, other factors intervened to prevent prices from being right. In particular, heavy subsidies for renewables have systematically depressed prices, thereby undercutting the incentives to invest in thermal generation, and the right kind of thermal generation. This can lead to much bigger price spikes than would have occurred otherwise–especially when intermittent renewables output plunges.
Thus, a systematic downward price distortion can greatly exacerbate upward price spikes in a pure energy model. That, in a nutshell, is the reason for Texas’s recent (extreme) unhappiness.
As more information becomes available, it is clear that the initiator of the chain of events that left almost half the state in the dark for hours was a plunge in wind generation due to the freezing of wind turbines. Initially, combined cycle gas generation ramped up output dramatically to replace the lost wind output. But these resources could not sustain this effort because the cold-related disruptions in gas production, transmission, and distribution turned the gas generators into fuel limited resources. The generators hadn’t broken down, but couldn’t obtain the fuel necessary to operate.
It is certainly arguable that Texas should have recognized that the distortion in prices that arose from subsidization of wind (primarily at the federal level) that bore no relationship whatsoever to the social cost of carbon made it necessary to implement the kapacity market kludge, or some other counterbalance to the subsidy-driven wrong prices. It didn’t, and that will be the subject of intense debate for months and years to come.
It is essential to recognize however, that the underlying reason why a kludge may be necessary is that the price wasn’t right due to government intervention. When deciding how to change the system going forward, those interventions–and their elimination–should be front and center in the analysis and debate, rather than treated as sacrosanct.
There is also the issue of state contingent capacity. That is, the availability of certain kinds of capacity in certain states of the world. In electricity, the states of the world that matter are disproportionately weather-related. Usually in Texas you think of hot weather as being the state that matters, but obviously cold weather matters too.
It appears that the weatherization of power plants per se was less of an issue last week than the weatherization of fuel supplies upstream from the power plants. It is an interesting question regarding the authority of ERCOT–the operator of the Texas grid–extends to mandating the technology utilized by gas producers. My (superficial) understanding is that it is unlikely to, and that any attempt to do so would lead to a regulatory turf battle (with the Texas Railroad Commission, which regulates gas and oil wells in Texas, and maybe FERC).
There is also the question of whether in an energy only market generators would have the right incentive to secure fuel supplies from sources that are more immune to temperature shocks than Texas’s proved to be last week. Since such immunity does not come for free, generator contracts with fuel suppliers would require a price premium to obtain less weather-vulnerable supplies, and presumably a liability mechanism to penalize non-performance. The price premium is likely to be non-trivial. I have seen estimates that weatherizing Texas wells would cost on the order of $6-$9 million per well—which would double or more than the cost of a well. Further, it would be necessary to incur additional costs to protect pipelines and gas processing facilities.
In an energy only market, the ability to sell at high prices during supply shortfalls would provide the incentive to secure supplies that allow producing during extreme weather events. The question then becomes whether this benefit times the probability of an extreme event is larger or smaller than the (non-trivial) cost of weatherizing fuel supply.
We have a pretty good idea, based on last week’s events, of what the benefit is. We have a pretty good idea of the cost of hardening fuel supplies and generators. The most imprecise input to the calculation is the probability of such an extreme event.
Then the question of market design–and specifically, whether weatherization should be mandated by regulation or law, and what form that mandate should take–becomes whether generation operators or regulators can estimate that probability more accurately.
In full awareness of the knowledge problem, my priors are that multiple actors responding to profit incentives will do a better job than a single actor (a regulator) operating under low power incentives, and subject to political pressure (exerted by not just generators, but those producing, processing, and transporting gas, industrial consumers, consumer lobbyists, etc., etc., etc., as well). Put differently, as Hayek noted almost 75 years ago, the competitive process and the price system is a way of generating information and using it productively, and has proved far more effective in most circumstances than centralized planning.
I understand that this opinion will be met with considerable skepticism. But note a few things. For one, a regulator’s mistakes have systematic effects. Conversely, some private parties may overestimate the risk and others underestimate it: the composite signal is likely to be more accurate, and less vulnerable to the miscalculation of a single entity. For another, on the one hand skeptics excoriate a regulator for its failures–but confidently predict that some other future regulator will get it right. I’m the skeptic on that.
Recent events also raise another issue that could undermine reliance on the price system. Many very unfortunately people entered into contracts in which their electricity bills were tied to wholesale prices. As a result, the are facing bills for a few days of electricity running into the many thousands of dollars because wholesale prices spiked. This is indeed tragic for these people.

That spike by the way, is up to $10,000/MWh. $10/KWh. Orders of magnitude bigger than you usually pay.
It is clear that the individuals who entered these contracts did not understand the risks. And this is totally understandable: if you are going to argue that regulators or generators underplayed the risks, you can’t believe that they typical consumer won’t too. I am sure there will be lawsuits relating in particular to the adequacy of disclosure by the energy retailers who sold these contracts. But even if the fine print in the contracts disclosed the risks, many consumers may not have understood them even if they read it.
One of the difficulties with getting prices right in electricity markets which has plagued market design is getting consumers to see the price signals so that they can limit use when supply is scarce. But this will periodically involve paying stratospheric prices.
From a risk bearing perspective this is clearly inefficient. The risk should be transferred to the broader financial markets (though hedging mechanisms, for instance) because the risk can be diversified and pooled in those markets. But this is at odds with the efficient consumption perspective. This is not a circle that anyone has been able to square heretofore.
Moreover, the likely regulatory response to the extreme misfortune experienced by some consumers will be to restrict wholesale prices so that they do not reflect scarcity value. That is, an energy only market has a serious time consistency problem: regulators cannot credibly commit to allow prices to reflect scarcity value, come what may. This means that an energy only market may not be politically sustainable, regardless of its economic merits. I strongly suggest that this will happen in Texas.
In sum, as the title of the book I mentioned earlier indicates, electricity market design is about choices. Moreover, those choices are often of the pick-your-poison variety. This means that avoiding one kind of problem–like what Texas experienced–just opens the door to other problems. Evaluation of electricity market design should not over-focus on the most recent catastrophe while being blind to the potential catastrophes lurking in alternative designs. But I realize that’s not the way politics work, and this will be an intensely political process going forward. So we are likely to learn the wrong lessons, or grasp at “solutions” that pose their own dangers.
As a starting point, I would undo the most clearcut cause of wrong prices in Texas–subsidization of wind and other renewables. Alas, even if stopped tomorrow the baleful effect those subsidies will persist long into the future, because they have impacted decisions (investment decisions) on the long horizon I mentioned earlier. But other measures–such as mandated reserve margins and capacity markets, or hardening fuel supplies–will also only have effects over long horizons. For better or worse, and mainly worse, Texas will operate under the shadow of political decisions made long ago. And made primarily in DC, rather than Austin.
Craig,
I would note that I led a project team in the late 90s for a World Bank/UNDP project for the government of Colombia that acknowledged the real option complexity of coupling the market exercise of the real option to produce power (as is done by LMP pricing) with the incentivization ans pricing of the real option to invest in generation (as capacity markets attempt to do) in a market based mechanism that is not just a compensation for existing generation (which ISO capacity markets do) but rather a financiable pricing signal for new generation AND that accounts for the generators willingenes to accept realibility risk. It can be done but you must approach it from a trading – as opposed to academic modeling- perspective. I believe the final report is a public document.
The other problem for power is there is currently no analogue for fossil fuel storage where long term (multi month) power storage is viable except for limited ponded hydropower assets. This removes the storage carry value as a coupling factor between the production and investment real options. Finally, power generation is not a declining resource (unlike fossil fuels) and so it has a very different impact on the cost of forward portfolio hedging.
Not sure any existing market model used in the US can avoid both systemic and event driven hypervolatility.
Comment by Thomas Lord — February 21, 2021 @ 9:53 pm
As I sit and read this, your colleague, Julie Cohn, is on Bloomberg blathering away about this, that and the other and making little, if any sense. Of course, I am struggling to follow all the threads in this post because I am a bit of a knothead.
Governor Abbott was on tv already this a.m. swearing to help the people of Texas against peaking/peaked electrical prices. He’s toast and going to do more damage on the way out.
All this as stated, it is a hell of a problem and a hell of a mess.
Comment by Donald Wolfe — February 22, 2021 @ 8:22 am
I would go back and reconsider the credibility of that estimate of $4 to $9 million per well to weatherize. If that were believable, there would have been no drilling in ND or Alberta for a very long time. Not to mention Marcellus, Utica, Wyoming, Colorado. The cost of a Bakken well relative to Permian is no where near that amount of difference. And yes, it would be the surface equipment that would be weatherized, there still simply isn’t that level of difference in costs.
Comment by JavelinaTex — February 22, 2021 @ 4:04 pm
I think some consumers do respond to high prices. An industrial plant that I know of from a neighbor shut down – partially due to the storm – but kept the cogen power plant running, exporting the power to the grid. I doubt they were the only industrial consumer to do that, for a lot of plants there was more money to be made selling electricity than operating the main plant.
Residential consumers, I agree, nobody has come up with a workable way to make them respond to price signals. My household was on Griddy (wholesale power pricing) as an experiment for a few weeks before the cold snap to see if we could save enough money by shifting loads to off peak hours to be worth the hassle. Even with an engineer who didn’t mind tinkering with the programmable thermostat to respond to price signals, and a coupon queen who gladly did all the laundry the day prices were low, we really were not sure it was worth the hassle even before the cold snap sent us scurrying to put that risk back onto a traditional retailer. But it did make us cut consumption a lot during the shortage/price run up before the blackout cut our consumption to zero. Incidently, I knew exactly what I was signing up for, cut consumption when necessary, and I am only out about $50 for the experience.
A pricing plan could be designed so that the consumer buys a future for a set amount of electricity based on their expected usage for the time of day, then charges the real-time rate for any consumption over that amount, and pays for any electricity not used. That would leave the consumer’s marginal cost, which incentivizes marginal usage, at the real-time price. But it would insulate the consumer from wild spikes that could bust a household budget. Provided they were sophisticated enough to turn the thermostat down and conserve whem prices spike. I would sign up for it, but I don’t think many others would.
Some less complicated incentives might be more marketable even though they are less precise. For instance, electric resistance “auxillary heat” on a heat pump causes a spike in demand at the worst possible time – when the weather gets too cold for heatpumps. If a house has a gas (or propane, or even oil) furnace as the auxillary instead of electricity, then the retail electric provider is less exposed to high demand during price spikes. Perhaps electric retailers should offer discounts to houses that have gas or propane heating installed.
Texans also ought to ask: how much are we willing to pay for power that stays on in extreme conditions? The $9000/mwh (plus some adjustments) cap that ERCOT has is an implicit limit to how much Texas regulators said it’s worth. In the old days of fossil fuels & nukes, that $9000 cap may have never mattered since baseload profits incentivized investment, but subsidies for wimd destroyed that. Now, many gas plants are defacto backup generators for subsidized wind, so we ought to ask how much that backup is worth. Not that politicians will ever want to frame it that way, but that’s what it is. We already accept that it’s too expensive to harden the distribution system against hurricanes, which is why I own a portable generator that came in handy last week.
But I do agree the biggest and most obvious mis-pricing and mis-allocation of resources in the whole system is the one the prof identifies: that 2.3 cents per kwh subsidy for “renewables”. We have subsidised a power supply that randomly ramps up and down, we shouldn’t be shocked when the grid follows.
Comment by Jack — February 22, 2021 @ 10:42 pm
If consumers couldn’t understand the Griddy model, they will never grasp any pricing scheme involving futures or incentivized marginal pricing. I pay $0.085/kwh fixed, all in. This price is available to any Texas consumer and is well below the national average for residential electricity.
Personally, I have no interest in paying $1,000 more each year on my electric bill to harden the grid so the system works optimally for three days in 2037, when another weather event of this type comes through Texas. Heavy expenditures, mandated by politicians who won’t be around when their “solutions” are truly tested, to handle rare events are rarely the optimal solution.
If we want to solve the problem of price spikes, prohibit the sale of any plan not based on a fixed rate. End of problem.
Much of the blame deserves to be placed at the feet of ERCOT. I will defer to the upcoming legislative hearings to determine exactly what happened at the generation level, but it needs to be recognized ERCOT did horribly at preparing the people for the extreme weather event. ERCOT did nothing to raise awareness among consumers, they did nothing to alert consumers as to the supply/ demand situation, they offered no suggestions how to best prepare for the upcoming weather event, and the people were blindsided. ERCOT effectively had no plan in place to handle the type of weather event Texas saw.
First and foremost, ERCOT needs to be gutted, recapitalized, and rebuilt as an effective organization. Secondly, as long as there is a subsidy to wind energy, there should be a state excise tax on wind energy. We have got to push wind energy into its rightful place in the state’s power generation portfolio and allow other forms of generation to optimize their niche.
Texas has too much federal intervention influencing the state’s power generation assets. This intervention has completely failed to recognize the true economic costs of renewable power. Demonizing of coal (and increasingly gas), while incentivizing renewables is leaving the state unprepared for predictable weather events. Poor management of the state’s power assets has exacerbated the problems. As a society we simply do not have the technology to rely as much on renewables as those in Washington are insisting we rely on renewables.
All that being said, I expect the policy makers in Washington to continue to pressure the state to produce a plan to eliminate both coal and nat gas for power generation before a specified date. The true economic costs of such policy decisions haven’t mattered up until now, why should they matter moving forward?
Comment by Charles — February 24, 2021 @ 7:56 am
In a “round-up-the-usual-suspects” way, it’s been interesting to see the blame on wind/solar for this event. Mostly people are talking their own book and there’s been plenty of at best disingenuous and at worst lying testimony in the TX hearings. As a market participant whose agnostic on resource technology, it’s been depressing to see California-esque politics over facts.
Wind was expected to produce around 7 GW during peak winter conditions (winter 2020-2021 SARA). Actual production was 3-4 GW. During the morning of 2.15, wind output was generally stable (+/- 200 MW) but multiple fossil plants and one nuclear plant tripped offline, causing frequency to drop as low as 59.3 Hz, which pushed generator governor responses to extreme levels and would’ve caused many more units to trip had ERCOT not load shed. ERCOT didn’t do a great job in that regard — they took the grid to within a whisker of complete collapse (e.g. PRC < 1 GW, ~5 min at 59.3 Hz) and should've been more proactive on load shed. But the bottom line is that from the operational data, there's no evidence of wind causing issues: ERCOT released the cause of all the under frequency events. None are from wind ramps, all from thermal/nuclear plants tripping.
Certainly wind didn't cover itself with glory — it was about 50% of forecast in peak conditions. But thermal was down 25-30 GW, which in absolute terms was the entire load shed quantity (in percentage terms, thermal was about 65% of forecast, better than wind but not exactly impressive). Had thermal generators procured gas supply and adequately winterized, there'd have been no problems. Of course thermal generators are shifting the blame to gas suppliers, which isn't entirely unfair — the freeze-offs meant even if they'd procured firm gas contracts, it's unclear if that contract would've been honored.
By far and away the most blame should be ascribed to natural gas producers and gas supply system. A moderate amount to thermal generators and a small amount to wind/solar — had the latter overproduced against expectations it wouldn't really have mattered but it's true they didn't meet the (low) expectations placed on them.
Comment by JS — February 27, 2021 @ 10:33 am
Trying to explain what happened during the winter storm when people lost ~111 lives really doesn’t hit home with people. ERCOT and reliability should be paramount, but with an energy market design we are minutes away from grid failure and total blackout for months due to damage to distribution infrastructure.
I have tried to feel bad for coal plants during the ongoing regulation on cleaner coal and tax credits to renewable generation that is highly variable and creates short-term supply/demand imbalances sending prices sky high. But its hard to ignore the effects of climate change even with mixed expert review on the topic. Coal plants that aren’t tied to the natural gas supply system that failed during the winter storm could have been the solution or more nukes (even though there were outages for nukes and coal plants during the event). We are just moving from a more diversified gen stack to a more concentrated on in renewables. Solar is coming on in a big way along with batteries on the back end of all this. Will be interesting how everything develops, but overall a market design that prior to this event was given praise for not needing a capacity market and a design that is more economically free is still fragile and heavily regulated like every other electricity ISO in the USA.
How this event affects future pricing in the winter months going forward will be interesting to see how market participants try to value forwards or options that don’t trade that far out on the curve. The forward market moved up for Feb, but what if another event happens in 5 years in the months of Nov-Jan? Winter could be the new summer!
Will be interesting to see how things develop in ERCOT, and how many banks want to finance renewable development projects. There was a wind farm that lost 160M during the event, and I doubt banks want that to replay out for any development they fund, and need the counterparty to perform on there loan.
Comment by JKN — April 13, 2021 @ 10:35 pm