# Examples in the Evidence Base

### What the precedents show, and what to do differently

## Evidence standard

Each case is read for the **mechanism** it validates, and for the **function** it demonstrates. ERCOT demonstrates that private actors can manage volatility efficiently when they can aggregate it across a portfolio, which is a claim about risk allocation rather than about energy-only markets. New York demonstrates that an administered contract can bridge to investment while a market matures, and that an administered value drifts when no party has money at stake in its accuracy. Britain demonstrates that a payment formula can be changed without impairing network investment. The tolling agreement, the value stack, and the totex allowance are not mandates; each is evidence that a function can be performed.

Each case below is classified by what it actually establishes: feasibility, observed effect, mechanism, or analogy.

## Part one: the precedents

### Great Britain: RIIO, 2013 to present

**What happened.** Ofgem replaced cost-of-service regulation for energy networks with multi-year price controls built on three elements: a total expenditure allowance that treats capital and operating spending identically for recovery purposes, output incentives that pay for measured performance, and revenue caps set for a control period rather than reset annually. The regime has run through two full cycles, RIIO-1 from 2013 and RIIO-2 from 2021, with electricity distribution on its own cycle through RIIO-ED1 and ED2.

**What the evidence establishes.** That totex regulation is administrable at national scale for a decade, and that it does not starve networks of capital. British networks maintained investment-grade credit throughout. Under RIIO-1 they earned roughly 9 to 10 percent real on regulated equity against a UK-wide corporate average of 5 to 6 percent, a gap the National Audit Office criticized as evidence that Ofgem's targets were insufficiently demanding. Western Power Distribution, the only distribution group Ofgem fast-tracked across all its regions under RIIO-ED1, beat its customer-minutes-lost targets by roughly 38 percent and its interruption targets by roughly 26 percent. When its owner sold it in 2021, Ofgem's own networks director cited the premium the buyer paid as proof that the regulated return was adequate.

**What it does not prove.** RIIO establishes feasibility and operating experience, not a clean US causal estimate of savings. Britain has one energy regulator for one country, no federal and state division, a smaller and denser network, and a different corporate tax and financing environment. Nothing about RIIO demonstrates that a fifty-jurisdiction system can benchmark across state lines, which is the harder version of the problem. It also does not prove the regime is easy to calibrate, and the evidence points the other way.

**The lesson, which is a warning.** RIIO-1's failure mode was over-earning, and it was large enough to become a public scandal, with Citizens Advice claiming £7.5 billion in unjustified profits. Ofgem's correction in RIIO-2 was severe: totex sharing rates cut from a range of 44 to 64 percent down to 33 to 50 percent, allowed equity set at 4.55 percent CPIH-real for gas and transmission, and a 25 basis point outperformance wedge applied on top. Eight companies appealed to the Competition and Markets Authority, which upheld the cost of equity but struck the wedge. Then Ofgem's own performance data showed continued outperformance through an inflation and debt interaction the framework had not anticipated, requiring another fix in RIIO-3.

**What this initiative should do differently.** Three things. Set sharing factors conservatively from the start and publish the sensitivity analysis, because the political cost of clawing back excess returns is far higher than the cost of setting them tightly at the outset. Build an explicit reopener for financing-parameter surprises instead of waiting for the next control period. And expect the reset to be litigated, which means the first control period should be short enough to correct and long enough to be worth outperforming, with five years the defensible middle.

### Hawaii: performance-based regulation since 2020

**What happened.** The Hawaii Public Utilities Commission concluded a multi-year proceeding in Docket 2018-0088 establishing performance-based regulation for Hawaiian Electric, with a multi-year rate period, revenue caps, and performance incentive mechanisms.

**What the evidence establishes.** That an American state commission can adopt this framework under existing authority, in an American legal and ratemaking context, with an investor-owned utility. This is the single most important thing in the file for an American commissioner, because every other precedent invites the objection that it happened somewhere else with different law.

**What it does not prove.** Hawaii is an island system with no interconnection to anywhere, the highest electricity prices in the country, and distributed solar penetration that made the old model untenable earlier than elsewhere. Its commission faced a forcing function that a mainland commission does not. The utility has also since been consumed by wildfire liability, which makes clean attribution of operating results to the regulatory framework nearly impossible for the years that matter most.

**The lesson.** Hawaii is a feasibility proof and a legal template, not an effect study. Cite it for what a commission can do under existing authority and for the design of the incentive mechanisms. Do not cite it for how much money performance-based regulation saves, because the wildfire has confounded the record.

**What this initiative should do differently.** Pair the first mainland adoption with a pre-registered evaluation design, so that the effect question has an answer in five years. The absence of a clean American effect study is the single largest evidentiary gap in this program, and it is fixable only by planning the measurement before the intervention.

### Texas: connect and manage

**What happened.** ERCOT interconnects generation by studying local reliability effects, connecting quickly, and managing congestion afterward through the energy market, with the developer bearing curtailment risk rather than paying for network upgrades determined in advance.

**What the evidence establishes.** Speed, decisively, and the claim is bounded: the comparison is descriptive rather than causal, since market structure, geography, resource mix, transmission build, and load growth all differ between the two systems. What ERCOT establishes is that rapid connection under managed operating constraints is feasible at scale. ERCOT brought 14.2 GW online in 2021 and 2022 against PJM's 5.6 GW, despite PJM being more than twice its size, and interconnection runs roughly two to three and a half years against six or more elsewhere. It also proves the reliability objection is answerable, because ERCOT has operated this way for close to two decades.

**What it does not prove.** ERCOT is a single-state interconnection largely outside FERC jurisdiction, which removes a coordination problem every other region has. It is also an energy-only market, and this program deliberately does not import that feature. The most common error in citing Texas is to treat connect and manage and energy-only as a package; they are separable, and this program takes the first without the second. ERCOT's own timelines have also drifted as its queue has grown, with median storage interconnection moving toward four years, so connect and manage is faster rather than immune to process congestion.

**The misattribution to correct.** Winter Storm Uri in 2021 is routinely offered as proof that the Texas model fails. The joint FERC and NERC inquiry attributed the outages predominantly to generator freezing and natural gas fuel supply failures. No interconnection study prevents a plant from freezing. Uri is an argument about weatherization standards and gas system reliability, and it is not an argument about interconnection policy.

**What this initiative should do differently.** Pair connect and manage with proactive network planning, and publish curtailment statistics by node so developers can price the risk instead of guessing at it.

### South Australia: dynamic operating envelopes

**What happened.** Since July 2023, essentially all new exporting solar systems in South Australia must be capable of dynamic export control, and SA Power Networks now offers flexible export connections across its territory, including on rural single-wire earth-return lines. Other Australian networks are implementing through 2031.

**What the evidence establishes.** That real-time, location-specific export limits can be the default connection standard for ordinary residential customers rather than a pilot for sophisticated participants, and that the engineering and the customer experience both hold up at population scale.

**What it does not prove.** Australia has national appliance and inverter standards, a different regulatory architecture, and a rooftop solar penetration that forced the issue. The enrolled and actively managed population is also smaller than the capable population, so claims about how many systems are under live envelope management should be made carefully. This paper's own earlier drafts overstated exactly that point and were corrected.

**The lesson.** Envelopes without a market signal are rationing. They solve the hosting-capacity problem and leave the value question unanswered, which is why this program pairs them with posted locational offers.

**What this initiative should do differently.** Mandate the capability at the inverter standard instead of negotiating it connection by connection, and publish envelope utilization data so that a customer who is frequently curtailed can see it and a developer can price it.

### France: the capacity obligation

**What happened.** France operates a decentralized capacity mechanism in which suppliers must hold certificates covering their customers' contribution to peak demand, with certificates issued to capacity providers on the basis of certified availability.

**What the evidence establishes.** That an obligation placed on load-serving entities is an administrable alternative to a centralized capacity auction, and that certificates and bilateral contracting can coexist with regulatory oversight.

**What it does not prove.** France has a dominant state-linked incumbent generator and a market structure unlike any US region, and its mechanism has been criticized for illiquidity and for the market power that follows from concentration. It is a design reference rather than a demonstration that the model works in a market with PJM's diversity of participants.

**What this initiative should do differently.** Take the obligation structure and add what France's critics say is missing: mandatory registration and publication of positions, standardized products to build liquidity, physical backing requirements, and independent verification. An obligation regime without those is less transparent than the auction it replaces, which is the strongest objection to Principle 5 and the one that must be designed out rather than argued away.

---

## Part two: the domestic proof points

Each of these shows the cheaper option working, and each required either a bespoke regulatory carve-out or a utility outside the investor-owned incentive structure. That pattern is the diagnosis restated as evidence.

### Consolidated Edison, Brooklyn-Queens Demand Management

Con Edison avoided roughly a billion dollars of substation and grid investment by procuring about $200 million of demand reduction, efficiency, and distributed resources, roughly 52 MW of demand reduction plus 17 MW of distributed investment, under a program the New York Public Service Commission established in Case 14-E-0302 with a specific earnings mechanism attached.

**What the evidence establishes.** That a large deferral is achievable in a dense urban network, and that a utility will pursue one when the regulator builds a way for it to earn from doing so.

**What it does not prove.** The savings accounting is contested. Analysts calculated a full ten-year deferral cost closer to $855 million once traditional spending inside the program is counted, so the honest claim is that the alternative was materially cheaper rather than that a billion dollars vanished. The program also required an unusual regulatory construction that no other jurisdiction has replicated at that scale.

**The lesson for design.** This case is the argument for independent review of shared-savings counterfactuals. A mechanism that pays a utility a share of avoided cost creates an incentive to overstate the avoided cost, and the first major American example of the mechanism produced exactly that dispute. Build the standardized counterfactual method before the first shared-savings dollar is paid.

### Arizona Public Service, Punkin Center

APS installed a 2 MW, 8 MWh battery instead of rebuilding roughly 20 miles of rural distribution line through difficult terrain, deferring the rebuild by three to six years at roughly half the cost.

**What the evidence establishes.** That storage substitutes for wires in exactly the way the theory predicts, in a real utility capital plan, with the utility's own numbers.

**What it does not prove.** The system was placed on standby and disconnected in 2019 following an unrelated battery fire elsewhere on the utility's system, and APS subsequently sought buyers. It is an honest illustration of deferral economics and an equally honest illustration of early-storage operational risk.

**The lesson.** Cite it with the coda. An advocate who presents Punkin Center as an unqualified success will be corrected by anyone who knows the file, and the correction will cost more credibility than the example was worth. Presented complete, it is stronger: the economics worked, and the technology risk of that era has since been addressed through revised safety standards.

### Sterling, Massachusetts

The municipal light department's 2 MW, 3.9 MWh battery, commissioned in December 2016 at about $2.7 million with grant assistance, reports roughly $400,000 a year in avoided capacity and transmission charges. Sandia National Laboratories analyzed the estimate.

**What the evidence establishes.** That the economics are available to a small utility with no sophisticated market apparatus, and that they can be captured without a rate case, a docket, or a bespoke earnings mechanism, because a municipal utility does not earn a return on rate base and therefore faces no bias against the cheaper option.

**What it does not prove.** A municipal utility's cost of capital, tax position, and governance differ from an investor-owned utility's, so the financial case does not transfer line for line.

**The lesson, and it is the most important one in this section.** Sterling is the cleanest evidence for the central claim of this program. Remove the payment formula that rewards capital, and the cheaper option gets chosen without anyone needing to be persuaded. The reform generalizes to investor-owned utilities what public power gets structurally.

### American Electric Power, Lower Rio Grande Valley

AEP reconductored 240 energized line-miles of 345 kV transmission on existing structures, replacing conventional conductor with advanced composite-core conductor, roughly doubling capacity, finishing eight months ahead of schedule and under budget, and winning an industry award for the work.

**What the evidence establishes.** That reconductoring at scale is an executable construction program rather than a modeling exercise, on live lines, in an American utility's hands.

**What it does not prove.** Some of the associated figures, including line-loss reduction and generation-capacity value, are vendor-reported and should be labeled as such. And the peer-reviewed national analysis showing that advanced conductors can roughly double corridor capacity at about half the cost of new build carries real conditions: results depend on conductor selection, thermal limits, and structural condition, and not every corridor qualifies. The national cost comparison should rely on the peer-reviewed estimates rather than on the AEP project itself.

**What has changed since.** The Department of Energy closed a loan of up to $3.26 billion to AEP Texas in July 2026 to rebuild, reconductor, or build more than 2,800 miles of transmission, targeting a doubling of carrying capacity. Reconductoring is now being financed at federal program scale, which moves the case from demonstration toward practice.

### Google, Indiana Michigan Power and the Tennessee Valley Authority

Google signed demand-response agreements in August 2025, the first targeting machine-learning workloads, and by March 2026 reported roughly a gigawatt of contracted data-center demand response with US utilities.

**What the evidence establishes.** That the flexibility trade at the heart of Principle 4 is commercially demonstrated in an early-stage deployment, that a hyperscaler will accept curtailment terms in exchange for access, and that the counterparty utilities include both an investor-owned utility and a federal power authority.

**What it does not prove.** Curtailment performance has not yet been tested through a sustained system emergency, and the agreements are recent enough that measurement and verification practice is still forming. A Department of Energy review in 2024 could identify no examples of grid-aware flexible data center operation other than Google's, which means this is one firm's practice rather than an industry norm.

**What this initiative should do differently.** Standardize the product in tariff instead of leaving it to bilateral negotiation, and require ex-post verification against telemetry with a penalty price that exceeds the value of non-performance. A flexibility commitment that is never tested is a reliability assumption, not a reliability resource.

---

## Part three: the cautionary record

### FERC Order 1000, and how a pro-competition order was absorbed

**What happened.** FERC removed the federal right of first refusal in 2011 and required competitive solicitation for certain transmission projects. Roughly 2 to 3 percent of transmission investment has since gone through a competitive process, meaning about 98 percent of ISO and RTO transmission investment is still awarded outside one. Projects awarded outside competition ran roughly 34 percent above their initial cost estimates, while winning competitive bids came in about 40 percent below, usually with cost caps. These figures are descriptive comparisons across different project populations and cost baselines; they should not be read as a causal estimate of the savings from competition.

**How it was absorbed.** Three mechanisms, all of which this program must expect. Categorical exemption: in New England, transmission owners identify "asset condition" projects themselves, outside the regional planning process, and those projects are cost-allocated across the region exactly like the projects the operator selects. They grew from $58 million when tracking began in 2016 to more than $2.787 billion by early 2023, with annual spending rising eightfold to nearly $800 million and several billion more expected. No such project has ever been modified or rejected by the reviewing committee, and the Asset Condition Reviewer created in response to that criticism has no authority to modify or reject anything. Threshold and voltage screens, which invite a large program to be presented as a series of smaller ones. And state legislation: roughly ten states enacted rights of first refusal restoring incumbent rights after the federal one was removed.

**What it proves, and this is the most useful lesson in the paper.** A reform that polices the incumbent's behavior while leaving the incumbent's payoff intact will be absorbed. Rules that police categories lose to recategorization. Rules that police dollar thresholds lose to splitting. And a federal rule can be reversed at the state level by the same firms that lost at the federal level.

**What this initiative does differently.** The reconductor-first requirement applies to a rolling twelve-month portfolio on a corridor or substation rather than to individual projects, so splitting gains nothing. It permits no categorical exemption for asset condition, local reliability, or maintenance, because those are precisely the categories that absorbed Order 1000. The reviewer has authority to reject and to require an alternative, not merely to comment. And the full project list, with costs and justifications, is published in machine-readable form so that third parties can find aggregation patterns the reviewer misses. Underneath all of that sits Principle 1, which is the real protection: under a total expenditure allowance the utility earns the same whether it builds or procures, so recategorizing yields nothing, and under peer benchmarking the allowance comes from what comparable utilities achieve rather than from how this one describes its own work. A firm can recategorize its projects. It cannot recategorize the benchmark.

### California, 2000 and 2001

**What happened.** California's restructuring collapsed into rolling blackouts, utility insolvency, and a national retreat from retail competition that has not fully reversed in twenty-five years.

**What it actually shows.** The design barred utilities from forward contracting while exposing them to spot markets under frozen retail rates, and market participants exploited the resulting position. Post-mortems by Borenstein, Bushnell, Wolak and others attribute the failure to that design and to manipulation rather than to competition as such.

**Why it matters here.** Principle 5's central requirement is mandatory forward physical hedging. The mechanism at the heart of this program is the specific antidote to the failure that ended the last restructuring wave. The California comparison will be raised whether or not it is apt, and the accurate answer converts the most damaging precedent in the file into support for the design.

**The lesson that still binds.** Sequencing and design detail are the whole thing. California did not fail because reform is impossible; it failed because a specific combination of features created an uninsurable position for a regulated entity. Any reform program should be examined for the equivalent, which is why Papers III and V both carry an explicit failure-modes section.

### FERC Order 2222

Issued in 2020 to open wholesale markets to aggregated distributed resources, its implementation dates now run to late 2026 in New England, 2028 in PJM, 2029 in MISO, and 2030 in SPP. The lesson is about enforcement rather than design: an order without deadlines and consequences becomes a compliance negotiation. It is also a caution about relying on federal action for anything time-critical in this program, which is one reason the roadmap runs the state track independently.

### Boulder, Colorado

Boulder pursued municipalization for roughly a decade, spent about $28.7 million, and abandoned the effort in 2020 without acquiring a system. Roughly a dozen communities have municipalized in two decades, and most attempts are abandoned or rejected. This answers the reasonable question of why the program does not simply advocate public power. Acquiring a utility takes a decade per jurisdiction and usually fails, while changing the payment formula reaches the three-quarters of load investor-owned utilities serve and can begin with a docket.

### Ohio, House Bill 6

A roughly $60 million scheme secured a legislative bailout, produced the largest utility corruption case in American history, and sent the speaker of the Ohio House to federal prison. It is cited here not for shock value but because it establishes the outer bound of the capture problem the anti-capture provisions address: funded intervenors, open data, peer benchmarking, and the ratepayer-funded lobbying bans now enacted in Colorado, Connecticut, Maine, Maryland, and California.

---

## Part four: what the record adds up to

Five conclusions follow from reading these cases together.

**The feasibility question is answered and the effect question is not.** Every component of this program runs somewhere, which disposes of the claim that any of it is untested. What no jurisdiction has produced is a clean American effect study of a totex transition against its own rate base and load mix. That gap should be closed deliberately, by commissioning the incidence study alongside the first performance-regulation docket, with the evaluation design registered in advance.

**The failure modes are known and they are specific.** Over-earning in the first control period, absorbed reform through recategorization, opaque bilateral positions, inflated shared-savings counterfactuals, unverified flexibility. Every one of these has an instance in the record above, which means every one can be designed against instead of discovered.

**The reforms that change payoffs survive; the reforms that police conduct get absorbed.** Order 1000 is the control group for this proposition and RIIO is the treatment. That asymmetry is the strongest single argument for sequencing Principle 1 first.

**Small American proof points share a revealing property.** Every domestic deferral success in this file required either a bespoke regulatory mechanism or a utility that does not earn a return on rate base. The exception proves the rule, and Sterling is the exception.

**The strongest precedents cut both ways.** Britain proves the framework works and that its calibration is hard. Texas proves speed and reveals its own queue drift. Punkin Center proves the economics and ended in retirement. Con Edison proves the deferral and produced a contested accounting.

---

## Reading guide: which case for which audience

| Audience | Lead with | Because | Do not lead with |
|---|---|---|---|
| Commissioner | Hawaii, Docket 2018-0088 | It is an American commission acting under existing authority | Britain, which invites the different-country objection first |
| Legislator | The lobbying-cost bans, and Connecticut pairing performance-based regulation with one | Concrete, bipartisan, recently enacted | RIIO mechanics, which lose a room |
| Utility executive or board | RIIO returns and the WPD sale premium | The failure mode was earning too much, and an acquirer paid up | Con Edison, which reads as a threat to rate base |
| Consumer advocate | Sterling, and the delivery-versus-supply data | Cheaper option, no rate case, and bills that show the pattern | Texas, which carries the Uri association |
| Large load buyer | Google with I&M and TVA | Their own peer, already contracted | France, which is remote from their experience |
| RTO stakeholder | Order 1000 and asset condition spending | It is their own record, and it is not defensible | Anything implying the RTO staff are the obstacle |
| Journalist | The Order 1000 absorption story | It is a documented, checkable failure with a clear mechanism | The full five-reform program, which does not fit a story |

---

## Sources

Great Britain: Ofgem RIIO framework documentation, RIIO-2 and RIIO-ED2 final determinations, and RIIO-2 regulatory performance data (2024, 2025); National Audit Office, *Electricity Networks* (2020); Competition and Markets Authority determination (2021); Oxera, RIIO-2 final determinations review (2021); National Grid and PPL disclosures on the Western Power Distribution transaction (2021); Utility Week reporting of Ofgem commentary on the transaction premium.

Hawaii: Hawaii Public Utilities Commission, Docket 2018-0088, Decision and Order (2020).

Texas: Tyler H. Norris, "Beyond FERC Order 2023" (Nicholas Institute, Duke University, 2023); S&P Global and Utility Dive reporting on comparative interconnection volumes and timelines; FERC, NERC and Regional Entity joint inquiry, *The February 2021 Cold Weather Outages in Texas and the South Central United States* (2021).

Australia: South Australia Office of the Technical Regulator, Dynamic Export Requirements (2023); SA Power Networks statewide Flexible Exports materials (2025); ARENA and AEMO, Project EDGE and Project Symphony.

France: RTE documentation of the French capacity mechanism; Cramton and Stoft on capacity market design (2006).

Domestic cases: New York PSC Case 14-E-0302 and Consolidated Edison BQDM filings, with contemporaneous critical analysis of the program's net savings; Arizona Public Service materials on the Punkin Center battery (2017 and 2018) and subsequent disclosure of its 2019 standby and disconnection; Sandia National Laboratories analysis of the Sterling Municipal Light Department system (2017); AEP, WSP, Quanta and CTC Global materials on the Lower Rio Grande Valley reconductoring, with vendor-reported figures identified as such; US Department of Energy, Energy Dominance Financing loan to AEP Texas (July 2026); Google demand-response announcements with Indiana Michigan Power and TVA (2025) and subsequent scale reporting (2026).

Cautionary record: Brattle Group analyses of competitive transmission under Order 1000; Concentric Energy Advisors, *An Updated Examination of FERC Order No. 1000 Projects* (2024); New England States Committee on Electricity letter to ISO-NE on Asset Condition Projects (2023); RMI, *Mind the Regulatory Gap* (2024); Borenstein and Bushnell, "The U.S. Electricity Industry After 20 Years of Restructuring" (2015) and Wolak's CAISO Market Surveillance Committee analyses; FERC Order No. 2222 compliance dockets; Brattle, *Electric Utility Municipalization* (2025) and reporting on Boulder's effort; Department of Justice deferred prosecution agreement and sentencing records in the FirstEnergy matter.