In March 2023 I wrote a post reiterating my argument that grid reliability is not a public good, at least not according to the technical definition of a public good as nonexcludable and nonrival. Instead I argued that grid reliability is a common-pool resource, potentially somewhat excludable or at least differentiable, and if not nonrival at least congestible. Given recent discussions in electricity policy about reliability in general and winter reliability in particular, I think it’s worth reconsidering now.
In thinking about these questions recently I’ve been revisiting some important academic work relevant to these questions in the context of market design, developing the idea that reliability is, can be, and should be contracted over as a differentiated product. To kickstart my own thinking, and perhaps to introduce the idea to you, here I’ve revised an old Knowledge Problem post I wrote in August 2004 on the concept of network reliability as a differentiated product and added on some more recent work.
Source: Nature
A question related to the public good nature of network reliability is whether the fact that something has public good characteristics necessarily implies that it should be provided to all at a uniform level. Not necessarily. Imposing a uniform reliability standard to deal with the public good features of networks does disservice to the diverse, heterogeneous preferences that agents have over reliability. A uniform standard is likely to underprovide to high value agents and overprovide to low value agents, and costly inefficiency ensues. Instead, consider the extent to which reliability beyond a base level can be marketed, bought and sold as a differentiated product.
When you want to make a call on your cell phone, is the capacity there to enable you to do so on demand? When you flip the light switch, does the electricity to power the light flow to you on demand? When you want to access a webpage, does your network have the capacity to satisfy your demand in the time that you want it to?
Note that I phrased the third of those examples slightly differently, a rhetorical intention on my part. Intuitively the internet makes grasping one of the two points of this argument easier – different agents on a network have different preferences over the degree of reliability of the network on which they operate. If I’m a high demand user, I’m more likely to pay for high broadband capacity coming into my house, signaling to my ISP that they should similarly invest in network capacity to meet the level of reliability that I’ve contracted for with them. If I’m a low demand user, I’ll stick with a slower/smaller connection, and will pay a correspondingly lower price. However, note that in this example, both the high demand and the low demand agent want on-demand access. In internet networks the measure of reliability is more about upload and download speed than it is about on-demand availability of the network. I think you can make a similar argument about cell phone network reliability – how long does it take to connect your call? And do you stay connected or get dropped? Some people would be willing to pay more for service that connects more quickly and stays connected more reliably. Others may not.
The case of electricity is slightly different, because of the nature of electrons and those pesky laws of physics. In the case of electricity, thinking about reliability essentially boils down to thinking about the probability of interruption, with no issue of connection speed. In this way electric network reliability is similar to, and different from, internet and cell phone network reliability.
Let’s measure reliability as the probability of receiving service (which is 1-(probability of service interruption)). Ensuring a particular level of reliability and a commensurate probability of receiving service requires investing in network capacity. Now introduce two other features to this narrative model. First, take the important starting point from above: assume that agents have heterogeneous preferences over reliability. For simplicity, as before, think of there being high value agents and low value agents [n.b. This idea can generalize to a continuum of preferences, very much like is done in discrete and continuous models of product differentiation in Hotelling-style models.]. Second, note that current flow monitoring and interruption technologies have become increasingly available and affordable due to digitization, which make it feasible to implement the delivery of different levels of reliability to different agents. That digital capability thus makes it feasible to contract over different levels of reliability, deliver on them, and incur penalties for failure to deliver (more on that later).
In such an environment, treating reliability as a differentiated product becomes feasible, and may even be preferable to the extent that it enables increased value creation through better meeting the diverse preferences of network agents. Note that even in the current regulatory environment, high value agents can buy more reliability – even single-family households living in apartments can install UPSs (uninterruptible power supply, basically a battery) to power their computers in the case of a blackout. But there’s an asymmetry, because it’s difficult for low value agents to pay less and buy less reliability than is provided under a uniform standard. That’s where the idea of priority insurance comes in.
When something is scarce, priority insurance is a way for suppliers to ration the scarce resources among customers who have essentially bought a place in line. Think of priority insurance as a menu of contracts where customers choose a price and a probability of service. Put another way, if you are willing to be interrupted/face a higher probability of interruption, you may pay a lower price, and if there is deficient supply, you will have a lower priority for being served than others who were willing to pay a higher price. That can save you money and, by using prices to signal relative value, can actually lead to higher overall/average reliability by reducing the congestion that arises when something of value is not priced. Grid reliability is not priced, or at least not priced well.
A fantastic discussion of priority insurance is in an article by Hung-Po Chao and Robert Wilson (“Priority Service: Pricing, Investment, and Market Organization,” American Economic Review 77 (December 1987)). Chao and Wilson summarize the idea of priority insurance as
Each customer’s selection of one contract from the menu determines the customer’s service order or priority. In each contingency, the seller rations supplies by serving customers in order of their selected priorities until the supply is exhausted or all customers are served.
Chao and Wilson then go on to point out that priority insurance means that the product being sold can be a differentiated product, differentiated on the dimension of your willingness to be interrupted. Put another way, priority insurance in electric networks makes reliability a differentiated product, not a uniform product that has to be provided to all agents on the network at the same level. Priority insurance empowers agents on a network to choose either higher or lower levels of reliability, depending on their willingness to pay.
Note also that priority insurance provides a means of inducing customers to reveal information about their preferences over reliability. This kind of information is extremely absent today, and its absence means that policymakers have little or no idea of how customers really value reliability. Thus policymakers are making decisions about reliability policy on electric networks without having any good idea about what “the right” level of reliability is, or how costly it is to impose a single level of reliability in the face of diverse agent preferences over the level of reliability.
Another good paper on this subject is a follow up: Chao, Hung-po, Shmuel S. Oren, Stephen A. Smith, and Robert B. Wilson. "Priority service: Market structure and competition." The Energy Journal 9, no. 2_suppl (1988): 77-104. Their opening paragraph:
Recent advances in metering and control technology enable the power industry to differentiate its product to a much larger extent than in the past. Using advanced microelectronic components, “smart meters” now can monitor and control usage patterns in a cost-effective manner, even for residential users. Thus, product differentiation can be achieved by unbundling the delivery conditions of electric power such as time of use and service quality attributes and by offering the basic product (enegy) with different delivery conditions at different prices.
They wrote that in 1988 (!!), and only recently are we starting to see some elements of this thinking permeate the strong status quo.
In practice I think something like priority insurance would work best for contracting over willingness to pay and probability of interruption above a base level of reliability, which can be defined in SAIDI/SAIFI terms that are in common use in the industry. Having that base level acts as a “guardrail” protecting against extreme cases that could be life-threatening or unjust.
The other component of contracting over reliability that’s missing from current practice is that customers who suffer outages today are not compensated for those interruptions. In practice this lack of compensation means that customers are providing their load-serving entity distribution companies/retailers with unpriced hedges. In 2022 my co-authors and I made this argument (Mays, Jacob, Michael T. Craig, Lynne Kiesling, Joshua C. Macey, Blake Shaffer, and Han Shu. “Private risk and social resilience in liberalized electricity markets.” Joule 6, no. 2 (2022): 369-380.) A requirement for companies to compensate customers for outages, even if it’s an imprecise, administrative estimate of the value of lost load, would shift incentives and initiate more careful thinking about the heterogeneity of preferences over reliability and more efficient ways to provide reliability in ways that can satisfy those preferences more successfully.
Recall and enjoyed the previous article. Always gives me a chance to dust off economic ideas which is fun. For consideration..
Grid Reliability has Prioritized Insurance. Generac for instance. High reliability operations pay for this insurance. Generators, batteries, proximity to baseload, flywheels, etc. The insurance is private market coverage rather than utility provided.
The "Tech" service level comparison doesn't hold. Because the Electricity System isn't designed that way.. and even the most advanced tech Systems that are designed that way, don't do great at managing and compensating to service levels.
A more accurate comparison is Safety. A Public Good (although publicly provided). High Reliability operations contract with private Security companies for specialized needs.
We use Public Goods for things that are too difficult or costly to differentiate, correct?
Electricity seems like a mixed market - 98% public good / 2% Private good. Trying to push it toward Private on a theoretical basis runs into a Reality anchored equilibrium that's outside the market participants or policy structure.
Rather suggest considering the question -
1) Given Reliability is a public good and
2) The economic value of higher Reliability has drastically increased in the Digital Economy.
What's an appropriate level of Public investment?
What are fitting Rates?
How to incentivize Innovation (performance based, early stage investment)?
What policy guard rails are appropriate to ensure Access and Equity?
Thanks as always!
Do you still believe that electricity is a 'common pool resource' and thus not subject to private ownership and a free market? (You state above that you believed that "grid reliability is a common-pool resource...")
Second, on grid reliability and operations, don't we need a market discovery process in a real free market to provide answers (processes) without MOA/ISO/RTO government planning?