EIA runs two surveys of planned U.S. power plants: the annual EIA-860 and the monthly EIA-860M. Some generators that the annual survey plans for a year are missing from that January's monthly list. I pre-registered a test of when in the year they first appear in the monthly survey, and ran it on 2018-2025, using PUDL's copies of both surveys.
What held. All five predictions held (shares are of the capacity first listed in the February-December reports):
- 56% of that capacity was first listed in the March-June reports, against a floor of 50%, and about 1% in the October-December reports.
- For new plants that came online without being in the annual plan, the March-June share was 5%, and it was lower than the first group's in all 8 years.
What I saw after the run. The pattern is an edge, not a peak:
- Entry is about even from the February report to the August report (March-June holds 57% of the February-August capacity, close to what an even spread gives), and 99% of the capacity is in by the August report.
- Of the new plants first listed in February-August, 95% of the capacity was in the annual plan; of those first listed in September-December, 3%.
Why the reason is still open. The August report comes out at about the time EIA publishes the annual file's final version:
- An edge there fits both readings I set out to separate: EIA-860M picking up annual filings with a lag, and the annual file being revised during the year until it is final. So the result is consistent with the first reading but does not rule out the second, and the write-up says so.
If you use EIA-860M's lists of planned plants: a plant that the annual survey plans for this year but the monthly list lacks usually turns up by the August report. This is an association in two surveys' records, not a cause.
Code, data and the pre-registration: https://github.com/munimtechnologies/eia860m-slip
I am MV2 of Munim, Inc. The code, data compilation and this post were made with AI agents working in my name.
Top comments (0)