What Basis Is, and Why It Moves: Nodal Pricing, Congestion, and the Deduction Nobody Contracted For

In Yes Energy's twenty-year PJM forecast, the average real-time price in 2046 is $99/MWh. The Dominion zone clears at $116. ComEd clears $13 below the average. Two locations in one market, roughly $29/MWh apart, and the decomposition is the part that matters: congestion accounts for 95% of that spread. Losses contribute $2.

That is what basis is, and it is why a project's revenue can miss a model that got the power price exactly right.

The mistake worth correcting first is treating basis as noise. Noise averages out. Basis does not — it is structural, it is directional for renewables, and on the available evidence it is getting worse rather than mean-reverting.

This post opens Series C on market structure. You will get what basis is made of, why it is systematically negative for the projects most exposed to it, whether it can be hedged, and how to model it without understating it — which most models do, for a reason worth understanding.

ℹ️ Note: Forecasts quoted here are third-party forecasts, not predictions. All modelling figures are labelled assumptions.

What Is Basis?

The difference between the price where a project sells and the price its contract references. A generator settles at the locational marginal price at its own node; a PPA usually references a hub. Basis is the gap, and whoever holds it absorbs it.

The word is borrowed from commodities, where basis is the spread between a local cash price and a futures benchmark. Electricity basis behaves differently in one important respect: it is not a transport cost that converges as delivery approaches. It is a congestion rent, and congestion is a physical property of the network that can persist for years.

Why Do Prices Differ by Location at All?

Because the network has limits, and the market prices them.

Locational marginal pricing accounts for all transmission constraints when the market clears, producing a distinct price at each location rather than one price for a region. Where the system is unconstrained, prices differ only by losses — small, and predictable. Where a constraint binds, the price on the export side falls and the price on the import side rises, and the gap is the shadow price of the line.

ERCOT states the position about as cleanly as any market: transmission congestion is the sole contributing factor causing differences in energy pricing across the transmission system, with those differences quantified as basis differentials between a generator's interconnection point and the relevant trading hub or load zone.

The PJM decomposition above says the same thing quantitatively. Congestion 95%, losses $2/MWh. When people describe basis as "line losses," they are describing five percent of the problem.

Hub, Node, Zone: Three Prices, Three Definitions

A recurring source of confusion is that these are not the same thing and are not computed the same way.

A node is a physical settlement point — the generator's own price. A hub is a constructed average of many nodes, used as a trading reference precisely because it is more liquid and less volatile than any single node. A load zone is another average, weighted differently.

Yes Energy's ERCOT analysis makes the arithmetic explicit: LZ_WEST congestion has been significantly higher than HB_WEST congestion, "because LZ_WEST uses load-weighted averaging while HB_WEST employs a simple average of 345 kV generator nodes."

That is worth reading twice. The West hub and the West zone describe the same geography and produce different numbers, because one averages generator nodes evenly and the other weights by where load actually sits. A PPA that says "West" without saying which one has left a material number undefined, and the parties will not discover which they meant until a settlement statement disagrees with a model.

Why Is Basis Directionally Bad for Renewables?

Because renewables are built where the resource is, and the resource is rarely where the network is strong.

The mechanism is self-reinforcing. Good wind and solar resource areas attract development; development clusters; the cluster exports on transmission that was not built for it; the constraint binds; the price at the export end falls. Each additional project in the pocket makes the basis worse for every project already there — including its own.

The Yes Energy ERCOT work shows this as a map of the market. Across January 2022 to August 2025, the South zone experienced the most negative congestion, driven by excess coastal wind. The Panhandle was second, with wind export constraints creating persistent negative pricing. The North zone was negative. Houston was neutral to slightly positive. The West zone was consistently positive.

Those are not random. They track where generation was built relative to where load is.

The intensity is visible in two further figures. In 2025, ERCOT recorded negative prices at major hubs for over 1,200 hours — and that is at hubs, which by construction understate what individual nodes experienced. And wind generators in the Panhandle and West zones, which hold the best wind resource in ERCOT, now face the worst transmission congestion in the system, with curtailment routinely exceeding the industry-standard 2% threshold.

The best resource in the market, with the worst basis in the market. That is the pattern, and it is not a coincidence.

Is It Getting Better or Worse?

Worse, on the evidence available, and the trend is the part that breaks long-dated models.

Two data points frame it. The standard deviation of hourly LMP spreads between the Houston load zone and the West hub nearly doubled, from roughly $8/MWh in 2010 to $15–16/MWh in 2022 — so basis is not only larger, it is less predictable.

And in PJM, annual real-time congestion costs are forecast to rise from $3.3bn in 2027 to $6.4bn in 2046, with the AP South interface alone accumulating $38bn of cumulative congestion cost across the forecast horizon, reflecting the transmission system's inability to move power into Virginia's data centre load.

That last detail matters for how you think about basis generally. It is not only a renewables problem or a remote-resource problem. It is what happens when load grows somewhere the network cannot reach — and load growth is currently the more energetic of the two forces.

ℹ️ Note: A twenty-year forecast is a forecast. The point here is not the 2046 number but the direction and the decomposition: congestion rising faster than energy prices, which means basis grows as a share of revenue even if the power price forecast is correct.

What About Losses?

The other five percent, and it behaves differently enough to deserve separating.

Losses are the energy dissipated moving power across the network, and unlike congestion they are largely a function of distance and line loading rather than of a binding constraint. They are smaller, steadier, and far more predictable — the PJM decomposition puts them at $2/MWh against $27 of congestion on the same spread.

The reason to separate them anyway is that markets treat them differently. Some incorporate marginal losses directly into the nodal price, so a generator's LMP already reflects them and no further adjustment applies. Others apply a loss factor to metered output as a separate adjustment, in which case the deduction appears somewhere other than the price and is easy to double-count or omit entirely.

The modelling instruction is therefore simple and specific: establish whether losses are inside the LMP or applied to volume in the market in question, and make sure the model does it once. A project modelled with losses in both places is understating revenue by a percent or two for no reason; one with losses in neither is overstating it by the same.

It is a small number compared with congestion. It is also the easiest one to get wrong twice.

Basis Is Not Curtailment

These are routinely conflated and they are different exposures with different remedies.

Basis is a price effect: the project generates and is paid less per megawatt-hour than the reference. Curtailment is a volume effect: the project does not generate at all, or generates less than it could have. Both arise from the same underlying congestion, which is why they travel together and why the confusion is understandable.

The distinction matters commercially because the contractual treatments are unrelated. As the deemed generation post in Series D set out, curtailment is allocated by cause — buyer-instructed curtailment is typically compensated, congestion-driven curtailment typically is not — under provisions that count megawatt-hours not delivered. Basis has no equivalent machinery. There is no clause that allocates it, no allowance against it, and nothing to claim.

The sequence is what catches models out. As congestion in a pocket worsens, the first effect is negative basis: the project keeps generating and is paid less. Past a point, prices go negative and generation stops being economic, and the exposure converts into curtailment. A project can therefore move from a basis problem to a volume problem without anything changing except the depth of the constraint.

Model them as one mechanism with two expressions, and make sure the revenue line is not deducting the same megawatt-hour twice — once as a price haircut and again as a lost volume.

What Makes Basis Improve?

Three things, and a project's model should be able to represent each of them rather than assuming basis is a constant.

Transmission gets built. A new line or upgrade relieving the binding constraint can compress basis sharply and permanently. This is the single largest upside, and it is knowable: the relevant system operator publishes constraint and upgrade plans. A model that holds basis flat for twenty years in a pocket with a funded upgrade in year four is as wrong as one that assumes the upgrade arrives on schedule.

Load arrives. Basis is a function of generation relative to local demand. New load in the pocket — a data centre, an electrolyser, an industrial customer — absorbs generation locally and lifts the node. The PJM forecast shows this from the other direction: it is load growth in Virginia that is driving the AP South congestion, which means load siting moves basis at least as much as generation siting does.

Storage is co-located. A battery at the same point of interconnection can absorb energy in the hours of worst basis and discharge when the constraint has cleared, converting a price problem into an arbitrage opportunity. That is a substantial part of why co-located storage has grown so quickly in exactly the ERCOT pockets with the worst congestion, and it is covered properly later in this series.

The general point is that basis is a network property, not a site property. It moves when the network around the site moves — which means it can improve as well as deteriorate, and a model with a single flat assumption has given up on representing either direction.

Can You Hedge It?

Partially, with congestion revenue rights — and the word doing the work is "partially."

A congestion revenue right is a financial instrument that pays or charges its holder based on day-ahead congestion between two points. Hold a CRR from point A to point B and you receive, or pay, the difference in locational prices between them. As a hedge, it locks in the price of congestion at the purchase price of the right.

There are two forms, and the difference is the one to understand. A point-to-point obligation pays the holder when congestion runs in the expected direction and charges them when it reverses. A point-to-point option pays in the favourable direction and pays nothing in the other — protection without the downside, priced accordingly.

Four limitations recur, and together they explain why basis remains a live exposure in almost every financing.

Limitation	What it means
Availability	CRRs are allocated and auctioned in limited quantities; you cannot simply buy the volume you need
Tenor	Auction horizons are short relative to a 15–20 year PPA or a project's debt tenor
Shape	A CRR covers a defined quantity and period; generation is variable, so the hedge and the exposure do not line up hour by hour
Direction	An obligation that reverses becomes a liability, converting a hedge into a second exposure

A CRR programme can meaningfully reduce basis exposure in the near term. It cannot make a lender treat basis as hedged over a 2039 debt tail, and no lender does.

That is why basis almost always appears twice in a credit analysis: once in the sponsor's base case, at whatever the consultant's central estimate is, and again as a lender haircut applied on top. Sponsors sometimes read the second as double-counting and argue it out. It is better understood as the lender pricing the tenor gap — the hedge covers the next few auction periods, the debt runs for fifteen years, and the haircut is what stands in for the unhedged remainder. Arguing the central estimate down rarely helps, because the haircut is not a view on the estimate; it is a view on how long the estimate can be relied upon.

How Do You Model Basis Without Understating It?

Generation-weighted, never as a simple average. This is the single most common modelling error in the category, and it is systematically biased in one direction.

Why the averaging method changes the answer

Basis is worst in the hours when the resource is strongest, because that is when the local pocket is exporting hardest and the constraint is binding. Solar basis is worst at midday. Wind basis is worst in the windy hours. Those are precisely the hours in which the project produces most of its energy.

Average the hourly basis across all 8,760 hours and every hour counts equally, including the nights and calm periods when basis is near zero and the project generates nothing. Weight it by generation and the bad hours dominate, because they are the hours that carry the megawatt-hours.

Simple average basis        = SUM(basis) / 8760
Generation-weighted basis   = SUMPRODUCT(basis, MWh) / SUM(MWh)

Those two formulas will not agree, and for a renewable project the second is always worse. Assumption for illustration: a simple average of −$4.00/MWh against a generation-weighted −$6.00/MWh — a 50% understatement from a choice of formula.

What it costs

Assumptions, labelled as such:

Capacity                                          200 MW
Net capacity factor                                  25%
Annual generation                             438,000 MWh
Hub price (PPA reference)                      $40.00/MWh
Generation-weighted basis                      −$6.00/MWh

Revenue at hub price      438,000 × $40.00        = $17,520,000
Basis deduction           438,000 × $6.00         =  $2,628,000
Net revenue                                        = $14,892,000

Basis as % of gross revenue                       = 15.0%

Fifteen percent of top-line revenue, from a line item that does not appear in the PPA.

What it does to the debt

Assumed opex                                        $3,000,000
CFADS at hub pricing      17,520,000 − 3,000,000  = $14,520,000
CFADS with basis          14,892,000 − 3,000,000  = $11,892,000

Debt service sized at 1.35× on hub CFADS          = $10,755,556
Actual DSCR with basis    11,892,000 ÷ 10,755,556 = 1.11×

A structure sized to a 1.35× DSCR on hub pricing delivers 1.11× once basis is applied. On a typical 1.20× lock-up covenant, that project is in lock-up from day one — not because generation missed, not because the price forecast was wrong, but because the model settled at the wrong location.

The sensitivity that should always be run

Basis    −$2/MWh   →  DSCR 1.27×
Basis    −$4/MWh   →  DSCR 1.19×   ← breaches a 1.20× lock-up
Basis    −$6/MWh   →  DSCR 1.11×
Basis    −$8/MWh   →  DSCR 1.02×
Basis   −$10/MWh   →  DSCR 0.94×   ← below 1.00×

The breakeven basis for a 1.20× lock-up on these assumptions is about −$3.68/MWh, which is well inside the range observed in congested pockets. Publishing that single number — the basis at which the covenant breaks — is more useful than any central case, because it tells the sponsor how much locational deterioration the structure can absorb before the lender takes control of the cash.

To build the generation-weighted basis series, the DSCR sensitivity and the covenant breakeven, prompt Dezzmond with your generation profile and nodal price history.

What Do Developers and Lenders Actually Check?

• Which price does the PPA settle at — node, hub, or zone — and is it named unambiguously?
• Is the basis assumption generation-weighted? A simple average is the wrong number and is wrong in a predictable direction.
• How much generation is already interconnected in the same pocket, and how much is in the queue behind it?
• What is the basis at which the lock-up covenant breaks? Publish it.
• Are CRRs available in the required volume, direction and tenor — and are they obligations or options?
• Is the historical basis series long enough to include a year in which the binding constraint was different?
• What happens when the next transmission upgrade is energised? Basis can improve sharply, and a model assuming it never does is as wrong as one assuming it always will.

Frequently Asked Questions

What is basis risk in power?

The risk that the price at a project's own settlement node differs from the price its contract references, usually a hub. The difference is driven overwhelmingly by transmission congestion, not by losses.

Why is basis usually negative for renewables?

Because renewable projects cluster where the resource is best, which is typically far from load and served by constrained transmission. Each project in the pocket depresses the local price further, including its own.

How much of basis is congestion versus losses?

In the PJM forecast decomposition, congestion accounts for 95% of the spread between the Dominion and ComEd zones, with losses contributing about $2/MWh.

Can basis be hedged with CRRs?

Partially. Congestion revenue rights pay on the day-ahead congestion between two points, but availability, tenor, shape and direction all limit them — and a point-to-point obligation becomes a liability if congestion reverses.

Is basis the same as curtailment?

No. Basis is a price effect — the project generates and is paid less. Curtailment is a volume effect — the project does not generate. They share a cause in congestion, but only curtailment has contractual machinery allocating it, and a model must not deduct the same megawatt-hour twice.

Can basis improve over time?

Yes. Transmission upgrades that relieve the binding constraint, new load arriving in the pocket, and co-located storage all compress it. A model holding basis flat for twenty years is making an assumption in both directions, not avoiding one.

Why does generation-weighting matter so much?

Because basis is worst in the hours a renewable project generates most. A simple hourly average gives equal weight to nights and calm periods when basis is near zero, systematically understating the loss.

Closing: The Price You Model Is Not the Price You Get

Power price forecasting attracts enormous effort, and most of it is spent on the wrong axis. A twenty-year hub price curve refined to the dollar is worth less than an honest answer to a simpler question: what is the spread between that curve and the point where this project actually settles, and which way is it moving?

The evidence says that spread is large, structural, and widening. Congestion costs in PJM nearly doubling over the forecast horizon. Spread volatility between two ERCOT locations doubling in twelve years. Over a thousand hours of negative hub prices in a single ERCOT year. And the best wind resource in Texas sitting behind the worst congestion in Texas.

None of that is a forecasting problem. It is a siting and contracting problem that shows up in the revenue line, and a model that settles at the hub will never see it.

The next post takes the specific case that causes most of the damage: a PPA that settles at a hub while the project settles at a node, and the gap that opens between a P50 case that survived the model and a settlement statement that did not.

Sources: Yes Energy — Hub-Level Prices Won't Cut It: PJM Forecast · Yes Energy — Historical ERCOT Pricing and Drivers · Euclid Power — ERCOT Market Deep Dive · ERCOT — Congestion Revenue Rights