PJM Queue — Note 003

It is not PJM. It is almost everywhere — and two markets went the other way.

It is not PJM. It is almost everywhere — and two markets went the other way.

Kenneth · 2026-08-12

*Source: Lawrence Berkeley National Laboratory, Queued Up 2026 Edition (CC BY 4.0),

complete interconnection request dataset through end-2025. Analysis my own.*

The question Note 002 left open

Note 002 showed PJM's failure rate rising from around 60–80% for pre-2014

cohorts to 93% for those filed 2018–2021. **Is that a PJM problem, or a national

one?**

National — with two significant exceptions, and the exceptions are the

interesting part.

Failure rate, all-time against 2018–21 cohorts

Grid operators with at least 200 lifetime terminal outcomes and 60 in the recent

window.

Sixteen of nineteen operators got worse. Two got materially better.

Later — 2026-08-21. Joe Rand of LBNL, who publishes this dataset, suggested weighting outcomes by capacity rather than counting requests. Doing so makes every figure on this page worse, not better — PJM's 2018–21 cohort moves from 93.4% to 97.0%, and 21 of 25 operators deteriorate. ERCOT survives as the outlier either way. See Note 009. This page stands as published; it counts requests and says so.

1 · PJM is not the worst. It is the largest.

At 93% PJM sits mid-table. CAISO, MISO, SPP and NYISO are all at 95–96%. **PJM's

distinction is volume** — 2,418 terminal outcomes in the window, three times

MISO's and nearly five times CAISO's.

So the story is not PJM is uniquely broken. It is that **a national

deterioration is most visible in PJM because PJM is where the requests are.**

2 · ERCOT did not deteriorate. It improved.

ERCOT's failure rate is 61% against a national norm above 93% — and it fell

five points whilst everyone else rose.

ERCOT's interconnection process is the structural outlier: connect-and-manage

rather than the study-heavy approach used elsewhere. **This dataset does not prove

causation and I will not claim it does.** But the one market that does it

differently is the one market that did not deteriorate, and that is worth more

attention than another chart of how bad PJM is.

3 · FPL is the sharpest number here

19% failure in 2018–21, down 27 points. Four in five requests get built.

Small sample — 72 terminal outcomes — so treat it as a signal rather than a

finding. But a vertically integrated utility interconnecting largely its own

projects behaves nothing like a merchant queue, and the number says so.

What this means

Stop calling it a PJM problem. If you are modelling ERCOT off PJM base rates

you are overstating failure by more than thirty points. If you are modelling

anything else off ERCOT's, you are understating it by as much.

The right base rate is per-operator and per-vintage. A single national figure

is wrong nearly everywhere, and this is the third consecutive note where the

widely quoted composite turns out to conceal the actual answer.

What I have not established

Why ERCOT differs. Connect-and-manage is the obvious candidate and I have

tested nothing. The dataset shows what happened, not why.

Whether the recent cohorts are fully resolved. 2018–21 requests are largely

but not entirely settled; some active requests may yet be built, which would

lower these rates. That cuts the same way for every operator, so the comparison

holds even if the levels move.

*Method: LBNL's complete request file, grouped by entity. Failure rates count

terminal outcomes only. Operators below 200 lifetime or 60 recent terminal

outcomes are omitted as too small to compare. Every figure is a count from

published data, free at emp.lbl.gov/queues.*

*This is published research. Every figure is a count from a free dataset at emp.lbl.gov/queues — check it yourself rather than taking my word for it.