Pricing for the Short Run
Many factors go into electricity rate setting, but the economic guidance is short-run marginal cost.
Economists can be so judgy. We don’t just study how the world is, like scientists, engineers, and even historians. We also have a whole paradigm for analyzing how the world should be — whether outcome A is “better” than outcome B — known as welfare economics. It’s a logical and powerful analytical approach, based on the idea that it is good to enable voluntary transactions that make the economic pie larger. But when it comes to human happiness, it leaves a lot out, like consumer rationality, equity, and political practicality.
In the electricity sector, welfare economics concludes that the price of energy at any time and location should equal the societal short-run marginal cost (SSRMC) of supplying electricity. “Societal” because generating electricity creates environmental externalities and the price the customer faces should reflect those externality costs as well as the direct costs of generating and delivering electricity. “Short-run marginal” because some things can’t be adjusted quickly — like generation and transmission capacity — and consumers are making decisions now about how much to use, so the price should reflect the additional cost from delivering a little more electricity (or savings from delivering a little less) at that time and place, given the actual constraints the producer faces.
The idea behind P=SSRMC is that setting a higher price will discourage some use that the consumer values more than the cost of producing it. If SSRMC=6 cents for an additional kWh right now and I would value that kWh at 9 cents — whether to cool the house more with A/C, heat some water for tea, or light up a dark room — then my use of the electricity would create 3 cents of economic value (or gains from trade). But if the price is set at P=14 cents, well above SSRMC, that’s more than my value and I won’t buy the additional kWh. As a result, the 3 cents of potential economic value will never be created, what economists call deadweight loss. The opposite problem occurs if price is set below SSRMC, creating an incentive to consume even when the customer values the extra kWh less than SSRMC.
If you are interested in a bit more explanation on the general idea, I can offer a screencast video on the topic recorded for my course on Energy & Environmental Markets (in five bite-size pieces, parts 1, 2, 3, 4, and 5). This idea of economically efficient electricity pricing is also the basis for the paper that Jim Bushnell and I put out last fall (and the blog I wrote about it at that time) comparing P to SSRMC for residential customers across the country.
Still, applying the P=SSRMC concept to electricity pricing discussions runs into a lot of headwinds. Some are based on misunderstandings of microeconomics, while others highlight factors left out by a narrow economic analysis of should.
Let’s start with laying out what the argument does and does not rest on. It assumes that customers have a good idea of the price when they make a consumption decision, that they are able to rationally process that information, and that they have some ability to adjust their consumption. While these assumptions may have seemed pretty strong for residential customers 10 years ago, they have probably been a decent fit for the two-thirds of U.S. consumption from commercial and industrial customers for a long time. And going forward, as we have smarter appliances and homes, residential customers will be more able to observe (or have an Internet-connected device observe) and respond to electricity price variation.
The argument does not rest on a perfectly competitive market among sellers, which is just a context in which efficient pricing is more likely. It also does not rest on the ability to adjust every input incrementally. So-called “lumpy investment” raises difficult calculations about when to make those investments, but it does not change the economic efficiency of setting P=SSRMC given whatever investment exists. Nor does efficient pricing ignore the environment; that first “S” is there for exactly that reason.
Skeptics of SSRMC pricing sometimes assert that it causes boom/bust cycles in industries. The primary cause of boom/bust cycles, however, is the difficulty of predicting demand combined with the need for long-run investments in order to produce supply. As a result, sellers frequently wish they had made more or less investment, and supply is frequently abundant or scarce relative to demand. Not allowing price to adjust to those supply/demand mismatches – which is known as “price stabilization” to those most fond of the practice — means that when there is abundance, excess supply goes to waste or when there is scarcity, supply is allocated in some other arbitrary fashion, such as long lines or side payments to those who control the allocation. U.S. lessons in the failure of price stabilization include warehouses full of surplus cheese and the gas lines of the 1970s and after Superstorm Sandy, an experience relived recently in Mexico.
In electricity, pricing without regard to SSRMC has resulted in the need to build massive excess capacity so that the system can still meet demand – which is critical for grid stability – without help from any adjustment by customers. In the future, it would lead to curtailment of abundant renewable electricity at times it would have been costless for consumers to use the energy, thereby discouraging efficient investment in storage and in electrification of home energy use.
The LRMC Mirage
Some SSRMC detractors argue that the right pricing standard is long-run, not short-run, societal marginal cost. Long-run marginal cost is the additional cost of providing one more unit of output if all inputs were adjusted optimally, including capital investments that in reality take years to change. The SLRMC approach would set price too high in the middle of the night as if increasing consumption at that time required building more capacity, even though it doesn’t. Likewise, it would dictate keeping prices high to cover past capacity investments, even if the system is overbuilt and has plenty of unused capacity. During a demand spike that strains capacity, it would set price too low, as if the scarcity problem could be solved by instantaneously building more capacity, even though that’s not possible. You won’t find SLRMC pricing in an economics textbook (except when SLRMC happens to equal SSRMC), because it’s not based in economics.
A somewhat more credible justification for smoothing prices, such as SLRMC does, is that consumers are not very sophisticated and will incorrectly take short-term price variation as the right signal for making long-term capital investments. So it is better to send them a price signal that reflects the long-term average additional cost of them consuming more. That is a serious argument, for which there is virtually no serious evidence. In the meantime, the technology for receiving and responding to short-term price variation continues to improve, so the cost of hiding that information from customers has increased.
The most compelling argument for allowing price to depart from SSRMC is that in most cases it simply won’t raise enough revenue overall to cover the utility’s costs. That’s not true everywhere, particularly where highly-polluting generation means that prices should be set well above the private marginal cost of the utility. In those cases, however, regulators have generally been reluctant to acknowledge the unpriced external costs of the pollution.
Still, most electricity is sold by utilities that have enough embedded fixed and sunk costs — transmission and distribution lines, wildfire mitigation, billing systems, and stranded investment in nuclear plants or early high-cost renewables — that pricing at SSRMC will leave them with a revenue shortfall. And that will only be worse if/when we get to a world where generators have to pay taxes/fees for their environmental externalities.
Of course, there are other ways to cover the costs left uncovered by P=SSRMC, as Meredith pointed out a few weeks ago. There are few free lunches in the world, however, and covering utility-related costs from other tax revenues is not one of them. Whether it is income taxes, sales taxes, or property taxes, they still create price distortions and deadweight loss.
Likewise, the different ways of covering the revenue shortfall impose different burdens across the population, by income, by location, and by racial/ethnic identification. These too are important issues that are omitted by the standard welfare economics of electricity pricing.
Perhaps more economists need to ratchet down the judgyness and recognize that economic efficiency within one market it is not the only factor that should determine pricing. At the same time, perhaps policymakers need to take more seriously the economic damage that results when prices depart significantly from SSRMC. It’s not the only thing that matters in price setting, but ignoring it is inefficient and costly, and will become more so as consumers have more technology and options for responding to prices.
I am still tweeting interesting energy news articles, research papers and blogs (and occasionally my political views) @BorensteinS
Keep up with Energy Institute blogs, research, and events on Twitter @energyathaas.
Suggested citation: Borenstein, Severin. “Pricing for the Short Run”, Energy Institute Blog, UC Berkeley, August 19, 2019, https://energyathaas.wordpress.com/2019/08/19/pricing-for-the-short-run/
Severin Borenstein View All
Severin Borenstein is Professor of the Graduate School in the Economic Analysis and Policy Group at the Haas School of Business and Faculty Director of the Energy Institute at Haas. He received his A.B. from U.C. Berkeley and Ph.D. in Economics from M.I.T. His research focuses on the economics of renewable energy, economic policies for reducing greenhouse gases, and alternative models of retail electricity pricing. Borenstein is also a research associate of the National Bureau of Economic Research in Cambridge, MA. He served on the Board of Governors of the California Power Exchange from 1997 to 2003. During 1999-2000, he was a member of the California Attorney General's Gasoline Price Task Force. In 2012-13, he served on the Emissions Market Assessment Committee, which advised the California Air Resources Board on the operation of California’s Cap and Trade market for greenhouse gases. In 2014, he was appointed to the California Energy Commission’s Petroleum Market Advisory Committee, which he chaired from 2015 until the Committee was dissolved in 2017. From 2015-2020, he served on the Advisory Council of the Bay Area Air Quality Management District. Since 2019, he has been a member of the Governing Board of the California Independent System Operator.
This blog post has generated quite a few comments, and I wanted to respond to many of them. Rather than repeat various points, I am just posting one set of comments here.
1. Apparently I did not make it sufficiently clear, but this is a discussion of retail ratesetting in a regulatory setting. Thus, there is not really relevance in discussing wholesale market mechanisms and whether capacity markets are needed to incentivize investment. That is an interesting discussion, but it is completely unrelated to how revenue is recovered at the retail level.
2. Because this is a regulatory setting, is also not relevant to discuss long-run competitive equilibrium, where in theoretical models LRMC=SRMC. I agree with the view that this is a theoretical convenience often referenced in economics that never really happens in real-world competitive markets, because demand, investment, and costs are constantly changing. But whether you think that are not, I’m not aware of any theory that it happens in regulated markets.
3. Likewise, in a regulatory setting, I don’t see the point of discussing getting to “efficient equilibrium”. There is no entry in this setting, and prices are set by a regulator, so it is unclear what equilibrium even means, beyond the idea that consumers will come to some level of response to the prices. That is a small part of what economists typically mean by the term equilibrium.
4. There really is only one definition of LRMC and you will find it in pretty much any textbook. It is the derivative of total costs with respect to quantity when all inputs can be varied. While that does not technically rule out time-varying prices — because LRMC can vary somewhat with quantity if there is not constant returns to scale — it will rule out large swings in prices based on short run variation in marginal cost. By the way, supply curves are only a well-defined economic concept in competitive markets (where they are equal to, yes, SRMC), so saying that price should be determined by supply and demand isn’t really meaningful here.
It is possible that advocates of using LRMC for price setting simply are referring to some other concept that is not the same as the economic definition of LRMC. That doesn’t make such a concept incorrect, or unhelpful. But I think it would be useful if advocates would (a) be much clearer about what they mean by LRMC or point to a paper that does so, and (b) stop calling it LRMC pricing, as if it is a well-established concept in economics.
5. A few of the commentators noted that consumers might use short run prices as a guide to making long run investments, which could lead them astray, a concern that I also recognized in the blog. It is worth noting that this issue comes up in any competitive market, and has indeed led some myopic consumers to make misguided investments based on the short-run price of the commodity. Yet, we don’t generally “stabilize” prices in commodity markets, and when we do it is almost always over the objections of economists. Perhaps that inefficiency in electricity is so large and so ineffectively addressed through information (e.g., notices to consumers that prices are low this year due to unusually large hydro supply and they should expect prices to return to normal levels in the future) that it is more efficient to send price signals that do not reflect the short run marginal cost. Maybe. But (a) I haven’t seen any studies that make that case empirically and (b) there is an efficiency cost from sending price signals that do not reflect SSRMC as consumers make inefficient short-run decisions.
6. There is nothing inconsistent between setting price that refelect SSRMC, but giving customers information so they have a more accurate forecast of price in the future. The right SSRMC price today will help guide a customer’s behavior on short-run decisions, such as setting the A/C temp, but a more accurate forecast of future price will help them make better decisions on long-run investments.
7. Maybe this deserves a whole blog, but the concept of scarcity pricing is often discussed as somehow distinct from SRMC in electricity, when it typically is not. The SRMC curve of a generator, for instance, is not L-shaped. As the generator output approaches its nameplate capacity, its SRMC of producing one more kWh rises, both due to rising heat rates and due to a higher expected cost of wear and tear. When a whole system is very close to its nameplate capacity, the SRMC of the *marginal* unit of output can be extremely high. Every unit of output in the market with a lower marginal cost is earning scarcity rents in that situation. That is true whether the SRMC of the marginal producer is very steeply upward sloping or actually vertical. Such “scarcity pricing”, which is just SRMC pricing, is efficient pricing (so long as no producer is withholding quantity in order to drive the price up, i.e., exercising market power).
8. It is certainly true that if a system is constantly overbuilt and never approaches any capacity constraint, SSRMC pricing is very unlikely to cover total system costs. That may indeed argue for raising price above SSRMC to cover a revenue requirement. My point is that doing so creates deadweight loss, so we should think hard about the way we can cover the revenue requirement while minimizing DWL. There are other revenue sources beyond increasing volumetric price.
9. There is indeed a more significant challenge coming if we install abundant quantities of intermittent renewables without creating a setting that enables much more price-responsive demand. That is definitely a topic for a future blog, and may be a future research stream. But we know right now that economics says that when there is excess production of zero-price power available, the economically efficient price to set is zero (or possibly negative if curtailment is actually economically costly). The serious challenge is how do we still cover needed revenues without distorting prices further from the economically efficient level than necessary. That is a difficult question, to which I don’t have a good answer. But neither do the people who claim that we just set price at “LRMC”.
10. I have written many papers on the airline industry, and have consulted for both airlines and government agencies on antitrust cases. Back in 1978, Alfred Kahn made a joke about it being perfectly competitive, saying that planes are just “marginal cost with wings”. I think that’s the last time any economist studying the airline industry has suggested that it is perfectly competitive. Price discrimination is rampant in the industry, loyalty programs are everywhere, and market power is certainly present to some extent. It’s time to stop using air travel as an example of competitive or efficient pricing.
In equilibrium, short-run marginal cost and long-run marginal cost are equal. For example: to get another kWh our in the middle of the night, you’d need additional baseload generators to be built — or you can run a peaker at a much higher operating cost. The capital + operating cost of the baseload unit will approximately equal the variable costs of the peaker. To get another kWh on-peak, you must either build another peaker, or else pay someone to curtail (demand response). An optimally designed and built system would have these equilibrium results under normal weather.
But there’s too much intervention in the electricity marketplace for equilibrium to occur. We not only require utilities to carry reserves, but we require that those reserves (seldom run) be fitted with effective pollution control devices. So they can be run long hours, if needed. We are requiring progress towards a 100% RPS goal, forcing new capacity into the system, and creating excess capacity of older units, which can run long hours if needed. And there is too much weather variation in electricity markets for that equilibrium to occur.
I appreciate Sev including SOCIETAL (we don’t get italics out here in the comments section; I’m not yelling) costs in the equation. But we don’t know what all of these are, or how to measure them. EPA just released its Health Effects estimates of the benefits of EE and RE, ranging from a penny in California to six cents in the mid-Atlantic region. This is not CO2 costs, but health impacts — higher in the East due to dirty plants in close proximity to large populations. A month ago, we likely would have omitted these from SSRMC. https://www.epa.gov/statelocalenergy/estimating-health-benefits-kilowatt-hour-energy-efficiency-and-renewable-energy
You don’t find SRMC pricing in the competitive market. When a United Airlines plane is about to leave the gate, they don’t offer up the vacant seats for the cost of additional fuel and a Coke (in fact, they apply the highest prices to short-notice customers.) Rental cars are cheaper on weekends, but you pay for a lot more than the gas. Hotel rooms are cheaper between conventions, but never just the cost of housekeeping and a bar of soap. There are some great deals on Priceline and Hotwire when there are vacancies — but nothing close to SRMC. In competitive markets, people price their goods to what the market will bear, knowing that if they set the prices too low, they will go broke. They accept some vacancies as a part of the plan.
Now we are going further, installing wind, solar, and storage that is virtually 100% capital-intensive. It can deliver a kWh whenever we want it at a short-run marginal cost of zero. There will be few, if any, traditional “variable” costs in the electric grid of the future. The telecom industry ran into this in the 1990’s, when stepper-relay shifting (with LOTS of maintenance costs) gave way to electronic switches, with virtually no variable costs. They adopted Total Service Long-Run Incremental Cost as the cost foundation for rate design, at least for customers with competitive options. That is, the cost of an optimally sized system, built with today’s technologies at today’s prices. The total cost. Construction and operation. Of the whole system. That’s a rational approach.
Long-run marginal cost gives us a useful way of measuring the cost that WILL BE incurred to serve a load that DOES OCCUR in the future. That is what rate design primarily influences: whether people install central AC or not, and then how much they run it once it’s installed. A long-run price of 8 cents off-peak and 20 cents on-peak (roughly the TOU rate in advanced utilities like Fort Collins, Colorado) BOTH encourages the installation of efficient, right-sized equipment AND the responsible use of that equipment. A short-run price as low as zero creates an incentive to install the cheapest AC unit you can find, and then run it except when there is a crunch (and go to the movies that day).
We are driving all of the variable costs out of the power system. Rate design based on short-run marginal cost will lead to disaster — inducing demand that will create a need for new capacity, and therefore inducing the commitment of long-run marginal cost. Better to give a customer a stable and predictable price, based on long-run incremental cost, and have the money needed to pay for new capacity if and when needed.
There is a sinister side of this issue, experienced in Oregon in the 1970’s and 1980’s. In the 1970’s, short-run (less than 10 years) marginal costs for energy were very high, because the region’s hydro base was fully exploited, and all new resources needed to come from expensive new (coal, nuclear, and eventually, gas) resources. The OPUC, basing an EPMC cost allocation on this, shifted costs massively towards industrial customers (who use lots of energy, and little distribution servie). A few years later, in the 1980’s, when new power plants drove up revenue requirements and consumption declined, there was a surplus, and that same short-run perspective showed a massive surplus, and very depressed costs for energy; the same flawed cost-allocation method then shifted costs massively to residential customers (who use lots of distribution service, and comparatively little energy). Sorry: but the disequilibrium in the market in the 70’s was not “caused by” industrial customers, and it was irrational to penalize them. And the disequilibrium in the 80’s was not “caused by” residential consumers. Adopting a long-run approach, where all costs are considered variable, produces a stable cost calculation, a stable cost allocation, and stable prices that allow all customers to make rational decisions in buying appliances, installing new windows, buying new industrial equipment, and building new buildings. Those assets will consume power for decades, and taking a decades-long perspective makes sense.
I do make one exception: when the short-run costs deviates sharply from normal short-run trends. For example, in a very wet hydro years, we should offer “spill” rates to customers with discretionary loads (industrial customers with electric boilers, for example), and in extreme peak hours, we should use critical peak pricing to recover the cost of demand response and storage resources acquired just for such contingencies from customers using power in those extreme hours.
Predictable tariffs should be based on long-run incremental costs. Real-time prices, which few customers choose voluntarily, should be available that reflect (but do not necessarily equal) short-run deviations in cost.
This reminds me of the story of the oak beams in the College Hall at New College, Oxford:
Founded in 1379, New College, Oxford is one of the oldest Oxford colleges. It has, like other colleges, a great dining hall with huge oak beams across the top, as large as two feet square, and forty-five feet long each.
A century ago, some busy entomologist went up into the roof of the dining hall with a penknife and poked at the beams and found that they were full of beetles. This was reported to the College Council, which met the news with some dismay, beams this large were now very hard, if not impossible to come by. “Where would they get beams of that caliber?” they worried.
One of the Junior Fellows stuck his neck out and suggested that there might be some worthy oaks on the College lands. These colleges are endowed with pieces of land scattered across the country which are run by a college Forester. They called in the College Forester, who of course had not been near the college itself for some years, and asked him if there were any oaks for possible use.
He pulled his forelock and said, “Well sirs, we was wonderin’ when you’d be askin’.”
Upon further inquiry it was discovered that when the College was founded, a grove of oaks had been planted to replace the beams in the dining hall when they became beetly, because oak beams always become beetly in the end. This plan had been passed down from one Forester to the next for over five hundred years saying “You don’t cut them oaks. Them’s for the College Hall.”
That is the sort of vision I expect from today’s utility commissioners: think a bit more about the seventh generation yet to come. Between beam replacements, the roof has zero variable costs. But you need to be growing a forest for hundreds of years to be ready when the time comes.
“In equilibrium, short-run marginal cost and long-run marginal cost are equal….But there’s too much intervention in the electricity marketplace for equilibrium to occur.”
Jim, I fully agree with you that when systems are in equilibrium LRMC = SRMC. But this is a theoretical construct that seldom occurs. As you have correctly observed, power systems are almost never in equilibrium and when they are it is a transitory event. That’s why I (and apparently Severin) submit that SRMC is the appropriate metric for setting prices.
“You don’t find SRMC pricing in the competitive market. When a United Airlines plane is about to leave the gate, they don’t offer up the vacant seats for the cost of additional fuel and a coke.”
You are defining SRMC incorrectly in your claim that it is not found in a competitive market. When an airline sells its last few seats at more than the cost of a coke (your proxy for SRMC) it is essentially employing congestion pricing, a component of SRMC which I described in comments on Severin’s earlier blog related to demand charges.
Consider what would happen if the last few seats on a flight were offered for $5 (roughly the cost of a coke plus the incremental fuel). Don’t you think it would be inundated with buyers? Also, none of the markets you described are perfectly competitive so there is some exercise of market power through tacit collusion among the sellers; consequently, price offers will be somewhat higher than SRMC including the congestion component, in order to maximize the total revenues that get divided up among the sellers. This is what it means to charge “what the market will bear.”
“Now we are going further, installing wind, solar, and storage that is virtually 100% capital-intensive. It can deliver a kWh whenever we want it at a short-run marginal cost of zero.”
Yep! And this is why congestion pricing will become essential to efficient pricing of electricity in the future.
“They adopted Total Service Long-Run Incremental Cost as the cost foundation for rate design, at least for customers with competitive options. That is, the cost of an optimally sized system, built with today’s technologies at today’s prices. The total cost. Construction and operation. Of the whole system. That’s a rational approach.“
But Jim, telephone rates eventually eliminated the per-minute charges during off-peak periods and recovered most of their revenues through higher monthly customer charges. In effect they did shift to SRMC pricing, including congestion charges during on-peak periods, and eventually eliminated any per-minute charges.
“Long-run marginal cost gives us a useful way of measuring the cost that WILL BE incurred to serve a load that DOES OCCUR in the future.”
True. So why not charge LRMC when it finally equals SRMC, rather than discourage efficient consumption TODAY based on forecasts of what the FUTURE cost will be (which is uncertain and almost certainly not be what is forecasted).
“That is what rate design primarily influences: whether people install central AC or not, and then how much they run it once it’s installed.”
You do have a good point when you state that SRMC pricing does not provide the right signals for making investment decisions. This argument has validity for residential customers and unsophisticated small C&I customers as well. But rather than distorting these customers’ current consumption decisions with LRMC pricing why not provide them with better information needed to guide their investments in energy-efficiency?
I have no sympathy for large C&I customers that base their investment decisions on today’s prices without considering where prices will be over the life span of the projects they are funding. They deserve to go out of business for being incompetent. It’s survival of the fittest.
“We are driving all of the variable costs out of the power system. Rate design based on short-run marginal cost will lead to disaster — inducing demand that will create a need for new capacity, and therefore inducing the commitment of long-run marginal cost.”
Not so. Congestion pricing tamps down demand so that it (roughly) equals the ability of the power system to supply it. Eventually SRMC prices will reach what you are referring to as LRMC, at which point they will induce entry of additional capacity. This same mechanism works equally well for a competitive market (e.g., generation) or a regulated monopoly (e.g., today’s distribution systems).
“Better to give a customer a stable and predictable price, based on long-run incremental cost, and have the money needed to pay for new capacity if and when needed.”
Well this is a value judgment. Is it better to burden consumers today with excessively high prices that suppress their demand below what the system can serve? Or is it better to provide them with price signals that are efficient and allow them to gain additional consumer surplus?
“There is a sinister side of this issue, experienced in Oregon in the 1970’s and 1980’s. In the 1970’s, short-run (less than 10 years) marginal costs for energy were very high, because the region’s hydro base was fully exploited, and all new resources needed to come from expensive new (coal, nuclear, and eventually, gas) resources. The OPUC, basing an EPMC cost allocation on this, shifted costs massively towards industrial customers (who use lots of energy, and little distribution servie). A few years later, in the 1980’s, when new power plants drove up revenue requirements and consumption declined, there was a surplus, and that same short-run perspective showed a massive surplus, and very depressed costs for energy; the same flawed cost-allocation method then shifted costs massively to residential customers (who use lots of distribution service, and comparatively little energy). Sorry: but the disequilibrium in the market in the 70’s was not “caused by” industrial customers, and it was irrational to penalize them. And the disequilibrium in the 80’s was not “caused by” residential consumers. Adopting a long-run approach, where all costs are considered variable, produces a stable cost calculation, a stable cost allocation, and stable prices that allow all customers to make rational decisions in buying appliances, installing new windows, buying new industrial equipment, and building new buildings. Those assets will consume power for decades, and taking a decades-long perspective makes sense.”
What happened in Oregon back in the 1970s and 80s had nothing to do with SRMC vs LRMC pricing. It was the direct result of regulatory incompetence combined with the flawed leadership of Donald Hodel when he headed up the Bonneville Power Administration (BPA) from 1972 -1978.
In the early 1970s BPA forecasted regional demand to grow at 7 percent per annum. To satisfy that anticipated demand the Washington Public Power Supply System (WPPSS) embarked on a massively expensive nuclear build program. When the demand did not materialize most of the plants were cancelled, causing huge write-offs that bankrupting WPPSS and also sticking BPA with the costs they had guaranteed on some of the nuclear plant builds. Those costs were passed on to the wholesale customers buying power from BPA.
Would LRMC have prevented the WPPSS debacle? Perhaps – if it caused Bonneville to lower its growth forecasts. But in those days engineers did not take into account price when they forecasted demand – they simply used a straight-edge. And then there was the Donald Hodel factor.
“For example, in a very wet hydro years, we should offer “spill” rates to customers with discretionary loads (industrial customers with electric boilers, for example), and in extreme peak hours, we should use critical peak pricing to recover the cost of demand response and storage resources acquired just for such contingencies from customers using power in those extreme hours.”
Jim, what you described is a form of SRMC with congestion pricing! But why limit the benefits to just the industrial customers?
“Predictable tariffs should be based on long-run incremental costs. Real-time prices, which few customers choose voluntarily, should be available that reflect (but do not necessarily equal) short-run deviations in cost.”
Ideally, all customers should be exposed to SRMC-based real time prices. Large customers that want to avoid price volatility can do so by buying hedges. Smaller customers that want to be hedged should be offered fixed price tariffs; however, these tariffs should also allow the customers to sell back any energy they wish to curtail at prices equal to SRMC less their hedged tariff prices. That way they can ignore price volatility or they can profit from it.
I am trying to get this concept implemented in the ERCOT footprint in order to incent price responsive demand among residential customers.
“This reminds me of the story of the oak beams in the College Hall at New College, Oxford….”
Very amusing story, though unpersuasive.
Really Jim, are you serious about expecting regulators to have a “seventh-generation” time horizon? Most leave office after 5 years. Their time horizons are based on what their next job will be.
But setting aside that natural human factor, I have to wonder how much value Oxford sacrificed by not cutting down those oaks earlier and gaining the value from selling them. The future value those sales proceeds compounded over a few hundred years would probably have been enough to rebuild the entire Dining Hall. How much sense does that make?
That’s a great (though extreme) example of the downside to LRMC pricing.
I don’t follow the logic that if SRMC will never match LRMC because market failures and achieving LRMC on average is required to arrive at efficient societal investment, that we should rely on SRMC as the price signal. Dixit and Pindyck “Investment under Uncertainty” show that SRMC is efficient when variability of costs and prices lead to convergence on the LRMC. But we’re constrained from achieving that for a variety of reasons. So logic leads us to conclude that we should use the best proxy for the likely path of SRMC in a competitive market, which is the LRMC.
We also need to be clear about the definitions of “cost”, “value” and “prices.” We also need to be clear about definitions of “short” and “long” term. Costs are directly incurred expenses by a supplier (or a prosumer). Value is the economic value that a consumer places on consumption of the good or service at different levels of consumption. Prices are the market valuation derived from the convergence of costs of suppliers to provide goods or services at the level of consumption valued by the consumer. Short term is the immediate cost of provision with no arbitrary allocation of costs across time periods–no fixed labor, no multi-hour commitment costs, no CT proxy costs. Long term are the costs incurred out as far as we can envision, and includes everything beyond the short term costs.
Congestion and scarcity pricing are not SRMC concepts–they are pricing concepts independent of the basis of marginal costs. Scarcity pricing occurs solely because there is a barrier to entry which creates an infinitely (or nearly so) sloped supply curve converging with the highest value placed a good or service at the maximum available supply. Scarcity pricing is about wealth transfer, not efficiency within the context of SRMC. That wealth transfer can be allocated between producers and consumers in a manner that assures that producers can recover their investments that mitigate scarcity pricing. It’s the LRMC that is the metric for determining a cost based method for that allocation and is a common tool in economic regulation. Once prices diverge from the highest SRMC and are required to recover the costs of investment, we are now using a LRMC standard as the metric for determining if an investment is justified and “efficient.”
For electricity, arguably the barrier to entry in the ERCOT market is the instant cost of a new CT (~$50M) or the highest consumer value of ~$190M as one hour of the Texas GDP. (Or if we recognize the lumpiness of CT investment, the new CT costs $1M per MWH and the Texas GDP is $4M/MWH.) Instead ERCOT capped the market at $9,000/MWH recognizing that generators could charge whatever wanted up to these levels. Clearly these prices are not politically feasible. The highest SRMC at a current gas price of $2.25/MCF (and by the way, gas transport costs are not a SRMC–they are sunk costs from a societal standpoint, but that’s a trivial issue), the peak ERCOT heat rate is 10,000 Btu/Kwh, so the peak energy COST is $22.50/MWH. None of the remaining price margin is SRMC–it is function of the pricing mechanism that is unrelated. Anything less than the instant cost of a CT derived by such things as dividing the cost of new CT over a certain number of hours in the year or the next 10 years is simply a bad, sloppy form of LRMC. Even allocating multi hour commitment costs is actually a form of LRMC, not SRMC, which means we jump from the 5 minute heat rate to LRMC immediately. Anything in between is arbitrary under the current construct.
So we’re left back at the question of whether SRMC is the appropriate metric when we recognize that scarcity pricing is not connected to SRMC (or LRMC), at least not directly, in the electricity market. What is the appropriate efficiency benchmark?
“I don’t follow the logic that if SRMC will never match LRMC because market failures and achieving LRMC on average is required to arrive at efficient societal investment, that we should rely on SRMC as the price signal. Dixit and Pindyck “Investment under Uncertainty” show that SRMC is efficient when variability of costs and prices lead to convergence on the LRMC.”
I wasn’t going to respond to any more of the comments on this blog but your referencing Dixit and Pindydk’s little gem of a book sucked me back in. As I recall, it’s primary contribution was to introduce the option value of waiting before committing to an investment. I don’t remember it ever addressing how one should price a commodity or advocating for LRMC over SRMC. But maybe I’m wrong. Can you provide the page on which they said that?
Also, I’m surprised the authors would say that convergence to LRMC is a sufficient condition to ensure new entry. While true that SRMC prices, averaged over time, need to at least equal LRMC, that requires SRMC to exceed LRMC before new entry occurs to offset the years when SRMC will subsequently be below LRMC, since a static equilibrium at SRMC = LRMC is not sustainable in the real world.
If developers began entering the market as soon as SRMC equals LRMC the new supply will prevent SRMC from increasing beyond that level. Then later when times get bad and SRMC drops below LRMC, the average market price over time will be less than LRMC.
Robert, on the energy efficiency investment issue:
“You do have a good point when you state that SRMC pricing does not provide the right signals for making investment decisions. This argument has validity for residential customers and unsophisticated small C&I customers as well. But rather than distorting these customers’ current consumption decisions with LRMC pricing why not provide them with better information needed to guide their investments in energy-efficiency?”
You are ignoring the literature that shows that better info alone will not improve small consumer decision making due to transaction costs that are larger than the benefits of better info. In addition, there are agency problems between landlords and tenants. Those interests are not easily aligned.
“I have no sympathy for large C&I customers that base their investment decisions on today’s prices without considering where prices will be over the life span of the projects they are funding. They deserve to go out of business for being incompetent. It’s survival of the fittest.”
So you just described how the C&I customers should use forecasts of LRMC to determine their investment threshold. Relying SRMC pricing means that the consumers can be completely myopic (which is a result from the Dixit & Pindyck model when markets have not serious failures, e.g. entry barriers, all price volatility is considered.)
I cannot resist responding separately to this comment:
“Would LRMC have prevented the WPPSS debacle? Perhaps – if it caused Bonneville to lower its growth forecasts. But in those days engineers did not take into account price when they forecasted demand – they simply used a straight-edge. And then there was the Donald Hodel factor.”
My introduction to rate making was the 1974 BPA wholesale rate case, when only one of the five WPPSS plants was under construction. My undergraduate energy economics class submitted comment on the BPA wholesale rate case, urging them to adopt a two-tier wholesale rate, with an allocated share of low-cost hydro available to each utility, and then a second tier price, based on the cost of the WPPSS plants (LRMC) for consumption above that threshold.
Our theory was that we should test whether there was a market for $100/MWh power before we spent $24 billion to build five plants (and, of course, wholesale prices only reached that level once since 1974 for more than a few days, during the drought-triggered power crisis of 2000-2001.). I think I learned from that experience that LRMC is the only sensible pricing framework for electricity when LRMC differs significantly from SRMC.
Seattle City Light adopted a long-run marginal costing methodology for cost allocation and rate design, and in the Energy 1990 study, completed in 1975, counted on this to help suppress demand. They did not invest in WPPSS 4/5.
BPA raised wholesale rates 88% in 1979, 50% on top of that in 1981, and another 50% on top of that in 1982. That rate surge created the “irate ratepayer” movement, but also triggered investments in energy efficiency, a lot of fuel switching, and a significant amount of simple economic curtailment. People insulated uninsulated homes, installed heat pumps, paid attention to thermostat settings. At the time, there was enough flexiblity in the hydro system that peak savings were not viewed as more valuable than energy savings — we were trying to avoid new coal and nuclear plants. Load growth (and the need for the WPPSS plants collapsed. Four of the five were terminated at 20% – 71% complete. It only cost us $8 billion to pay the funeral expenses on that particular mistake.
The one plant that was completed produces power at a variable cost (annual operating budget divided by output) of about $40/MWh, and is uncompetitive in the market. If the capital costs (interest and depreciation) are included, it’s more like $65/MWh. In a market where new wind and solar are $20 – $40/MWh. Nuclear is a lot like wind and solar; it produces about half of its output when you really don’t need it, and you need to either invest in storage or flex the hydro system to absorb it. Which is why Helms was built to support Diablo Canyon.
I’m pleased to report that on October 1, 2011 (only 37 years later) BPA implemented this tiered rate design. The “Tiered PF Rate” now in effect The Tier 1 rate is $36/MWh. Grandfathered customers get a defined amount at that price. The NR (New Resources) rate is $79/MWh.
The result of this is that most of the wholesale customers have constrained their demand on BPA within the limit of their Tier 1 allocation. Investments in energy efficiency, and mandated RPS investments in renewable energy have helped them achieve this. And the NW Power and Conservation Council includes efficiency costing up to $150/MWh in its supply curve, considering both LRMC for power supply, and the additional costs of transmission, distribution and losses, and also the peak-oriented nature of some efficiency measures.
Here I am, 45 years later, still trying to teach my profession the lessons of WPPSS: DON’T PRICE ELECTRICITY AT SRMC if it is significantly below LRMC. It creates artificial demand that is unjustified, and it is the LRMC that will be incurred to serve that demand in the future. True then. True now.
I fully support you idea of tiered pricing. I presented the same idea to BC Hydro some years ago when it concluded that it had to add fossil generation because it was running out of potential hydro sites. But that concept is separate from the issue of pricing at SRMC vs. LRMC. Instead it involves allocating the cheap energy to legacy demand and pricing the more expensive (at SRMC) to the incremental demand. This is an equity issue, not an efficiency issue.
Also, I agree with you that Bonneville Power’s policy of charging “rolled-in” pricing, which averaged the price of the cheap hydro with the expensive nuclear, was an expensive mistake. Rolled-in pricing is not SRMC pricing.
I guess we will have to respectfully agree to disagree because it doesn’t appear that either of us will convince the other. LOL!
I really appreciate the view that societal externalities (e.g. pollution costs beyond what’s in the cap & trade market) should be included in SSRMC. Analysts also should probably try to include congestion externalities in SRMC such as marginal line losses or the spillover effects from marginal transactions increasing the average losses of other transactions on the grid.
However, I guess I’m one of the non-economist detractors that thinks short-run marginal cost hasn’t been very satisfactory for electricity. Since the 2001 CA electricity crisis, we’ve given up on marginal cost to guide entry/exit, overbuilt the system to ensure low prices and reliability at all times (and obtain more renewables), but continue to hope that SRMC should guide our usage decisions. If resource planning decisions on what to build and where are not based on SRMC prices, and entry/retention is instead achieved through contract prices such that resources don’t depend on SRMC to survive, how is SRMC providing the full picture for consumers other than to make the most of our overbuilt system? There’s no ability for scarcity or congestion to push up prices to ever reach average cost of generation (or T&D), so SRMC by design is always well below average cost. But if our consumption decisions truly do in fact lead to new resource procurement and replacement of existing resources, and those incremental costs never materialize in SRMC because the system is always overbuilt at any moment in time, it seems that the costs that are actually being paid to incremental system resources needs some recognition in marginal cost. Either we should ignore some of the excess generation and T&D capacity capacity being used to ensure reliability, and calculate SRMC based on more severe capacity constraints (i.e. so the consumer doesn’t treat the reliability-related supply simply as extra idle supply), or we should get back to allowing SRMC to help determine procurement.
As it is, electricity SRMC is like borrowing the car and only paying for gasoline. Normally, the number of car suppliers would diminish to the point where SRMC would begin to cover average costs, right? Here it’s as if a regulator interferes to ensure there’s always more than an adequate number of cars, so that SRMC never recovers more than the cost of gasoline. If average cost recovery is accomplished outside of SRMC, how can SRMC adequately function over time? Is the correct conclusion, if there’s a regulated oversupply of cars, that consumers should only pay for gasoline? Shouldn’t consumers also see the longer-run marginal costs related for example to the annual car payments or maintenance owners require in order to operate another year? If average costs are recovered through some sort of additional charge, shouldn’t some of those costs (e.g. the annual car payments or maintenance) be considered as marginal? Are really low SRMCs the right price signals for consumers, even though expansions to ensure reliability are ongoing (not just sunk)? If SRMC should guide consumption decisions, then shouldn’t it guide supply decisions too? It seems that SRMC is somewhat dysfunctional if it reflects a permanent excess of supply and the link to average cost recovery or entry/exit based on the level of SRMC is completely missing.
Great observation. As long as regulators encourage overbuilding of capacity to avoid the possibility of supply shortages SRMC will always be too low to cover the full cost of the system because the congestion pricing component of SRMC will never be triggered. This is a clear example of regulatory failure, not just in California but elsewhere as well.
And the outcome of this failure is that consumers pay too much for their electricity because they are bearing the cost of the excessive, underutilized capacity. PJM’s capacity market is the epitome of regulatory failure.
ERCOT is perhaps the best example of using SRMC pricing to deliver investors incentives. While the Texas system experiences yet another power shortage (third in five years), (https://www.utilitydive.com/news/ercot-reserves-drop-below-2300-mw-forcing-texas-grid-to-call-for-energy-e/560833/) the average system price is only $33/MWH (and adding a pure SR GHG societal cost would add only $6/MWH to that price) and the average price is forecasted to increase only 10% over the next decade (https://www.greentechmedia.com/articles/read/texas-power-price-spike-and-designing-markets-for-a-carbon-free-grid#gs.wvc1fl). Based on the CAISO market performance reports, this is insufficient to cover the costs of new investment in CCGT and CT. There seems to be a disconnect even here between short run prices and investment signals even in the face of shortages.
“ERCOT is perhaps the best example of using SRMC pricing to deliver investors incentives. While the Texas system experiences yet another power shortage (third in five years), (https://www.utilitydive.com/news/ercot-reserves-drop-below-2300-mw-forcing-texas-grid-to-call-for-energy-e/560833/) the average system price is only $33/MWH…this is insufficient to cover the costs of new investment in CCGT and CT. There seems to be a disconnect even here between short run prices and investment signals even in the face of shortages.”
The shortages you cited ARE the vehicle for getting the prices in ERCOT up to levels that support new entry (which is actually greater than LRMC for reasons I won’t go into). Unfortunately, ERCOT lacks sufficient price responsive demand (PRD) to ensure that some of those shortages will not cause involuntary load curtailments, ie., rolling blackouts, which is not good and most likely will not be acceptable to the regulators or the legislators.
In 2005 I coauthored a white paper for the Midwest ISO which argued that its energy-only market needed a substantial amount of PRD for it to be sustainable without the need for involuntary load curtailments. In 2012 the Brattle Group conducted a resource adequacy study for ERCOT that confirmed the validity of my argument. The study demonstrated that 3000 to 6000 MW of PRD (depending on the maximum curtailment prices of these loads) would allow ERCOT to meet the industry’s one event in 10 years reliability standard if the wholesale energy price cap is set at $9000/MWh. This level of PRD is about 5 to 10 percent of ERCOT’s peak summer demand so it is doable.
The bottom line is that ERCOT’s current situation does not contradict SRMC pricing. It only shows that the ERCOT market is incomplete because the demand side is not fully participating in that market. This is fixable.
Industrial electricity demand is elastic and Texas is loaded with industry. Why isn’t there enough PRD in ERCOT?
Regarding PRD, let’s say we have 10 GW of zero marginal cost resources available in a given hour and demand is 11 GW. What’s to stop the PRD resources from lowering their demand by 1 GW to drop the market price to zero? This wouldn’t necessarily require human involvement. If every EV had a machine learning algorithm figuring out how to charge up the car with the lowest price electricity they would tend (where possible) to lower demand to drop prices to zero. Is there an existing regulatory mechanism or pricing strategy that avoids this?
Is the solution for wind/solar resources to simply collude as a group and bid in a non-zero price that’s comparable to a CCGT?
In Texas, you would also need to add an addition $30/MWh for the health effects association with marginal generation. ERCOT does not do this, so their prices are artificially suppressed.
The new release from EPA of regional health effects adders (technically, the health benefits of energy efficiency and renewable energy, but as long as fossil is at the margin, it is also the marginal health effects cost) is a big deal.
California is at the low end, around $10/MWh, which is NOT reflected in CAISO pricing. California is low because it does not have high-pollution power plants in proximity to population centers. The Mid-Atlantic region (lots of dirty power plants in proximity to large populations) is at the high end, around $70/MWh.
I do want to hear back from Sev on whether he considers these to be a part of SSRMC. And, since NOBODY is reflecting this cost in rates (short-run market clearing rates, long-run PPA prices, retail prices based on marginal cost, retail prices based on embedded costs), what’s the right thing for an Economist to assert under these (deeply flawed) pricing conditions?
This is a very important release from EPA, and frankly, astounding to come out in this administration. The data will not paste into this blog format; you need to go look at it yourself.
“As a result, sellers frequently wish they had made more or less investment, and supply is frequently abundant or scarce relative to demand. Not allowing price to adjust to those supply/demand mismatches – which is known as “price stabilization” to those most fond of the practice — means that when there is abundance, excess supply goes to waste or when there is scarcity, supply is allocated in some other arbitrary fashion, such as long lines or side payments to those who control the allocation. ”
You’re not describing a problem of SRMC vs LRMC–you’re describing a problem of not allowing prices to reach efficient equilibrium, through constraints on changes in supply OR demand. Using either SRMC or LRMC for pricing and/or planning will not solve this problem.
“You’re not describing a problem of SRMC vs LRMC–you’re describing a problem of not allowing prices to reach efficient equilibrium, through constraints on changes in supply OR demand. Using either SRMC or LRMC for pricing and/or planning will not solve this problem.”
By “efficient equilibrium,” do you mean when SRMC = LRMC? You do realize that such an equilibrium seldom occurs and when it does it is a transitory event.
And what are these constraints you refer to?
Finally, since you reject the use of either SRMC or LRMC pricing, what is you alternative solution? What you have in mind is unclear from your comment and has aroused my curiosity.
I’m not rejecting LRMC pricing. I’m pointing out that it is the proper construct. Note that in the electricity markets, the SRMC never rises to the LRMC for long enough to justify new investment (it’s not happening in ERCOT right now) unless consumers are willing to tolerate outages. (We know how well that went in California….) And the only way to provide the reliability required for our technological society is to sign long term PPAs that suppress the price signal that might come from using SRMC so that SRMC will never reach LRMC and never be in equilibrium.
I’ve been conversing with others on the flaws in SRMC pricing for electricity and will have a longer response in the near future. I’m too busy with the testimony season here to give it my full attention right now.
The discussion about SLRMC is not true–it doesn’t preclude time variant pricing, which should reflect the interaction of demand and supply curves. It also doesn’t require keeping prices high based on past investment–use discounted forward costs instead for example.
While this may not be in textbooks (which often present simplified versions for students and ignore important complexities. On the other hand, we see references to using LRMC in articles and other publications (which are often much more rigorously peer reviewed than text books.)
I’ll post a longer response here or on my blog later after I get through the CPUC testimony season…
“While this may not be in textbooks (which often present simplified versions for students and ignore important complexities. On the other hand, we see references to using LRMC in articles and other publications (which are often much more rigorously peer reviewed than text books.)”
Text books aren’t arbitrarily produced out of whole cloth (as the lawyers like to say). They are based on rigorously peer-reviewed papers in the literature, often written by authors who are awarded nobel prizes, like Paul Samuelson. Surely you remember his book, “Principles of Economic?” It remained the most popular introductory text book for over 50 years and was revised at least 18 times.
I don’t doubt that LRMC has been referenced and even advocated in some cases. That doesn’t mean it is the correct metric. After all, there are a lot of Ph. D. economists that have a poor understanding of microeconomics.
Nice Severin, but what about the theory of the second best?
Apparently you missed the debate regarding second best that occurred in Severin’s blog on Demand Charges.
Well said, Severin,
I’ve have been arguing against LRMC pricing for 40 years. In 1979 I challenged Leo Mahoney and Sally Strider (with NERA) on this issue at a conference and caught all kinds of flak at the time.
But, obviously, the LRMC myth still lives on…
For some historical perspective, I think it useful to recall what W.S. Vickrey wrote regarding electric utility marginal costs in his article “Some Objections to Marginal Cost Pricing” in 1949.
“In the provision of electricity, for example, there is a legitimate place, on for a charge to the customer based on the marginal cost of connecting him to the distribution system and of maintaining the necessary accounts; for charges for maximum demand, based on the marginal cost of increasing the capacity of the meter, service connection, distribution transformer, and other facilities in the immediate neighborhood; for heavier charges for peak than for off-peak current, based either on the marginal cost of increasing central-station capacity or on variations in the marginal cost of generation at various outputs as generating units of varying efficiency are put into service; for other variations in charges with fluctuations in seasonal demand or in the supply of hydroelectric power.”
I have great respect for William Vickrey (I was awe-struck when I met him in the mid-1980s) but, I have to say, the quoted statement is too agnostic for my taste. Reminds me of like joke about the two-handed economist.
Perhaps being agnostic is a sign of appropriate humility.
Humility is a laudable personal trait but it doesn’t encourage actionable decision making.