EIA's annual survey of power plants, EIA-860, comes out twice for each year: an early release in early summer and a final release in the fall. I pre-registered a test of whether the planned generators that the annual file lists for the current year, but that January's monthly EIA-860M lacked, were already in the early release. I ran it on 2022-2025, using copies of both releases archived by Catalyst Cooperative's PUDL project.
What happened. All checks passed and none of my three predictions held (shares are of capacity, pooled over 2022-2025):
- Of the generators the monthly survey first listed in its July or August reports, 0% of the capacity was in the early release, on any sheet. The final release planned all of it for the year.
- Of those first listed in the February or March reports, 98% was in the early release's plan for the year.
What I saw after the run. The later the monthly survey first listed a generator, the less likely the early release held it:
- For the April-June reports the share was 49%, between the two.
- Across 2022-2025, 17,172 of these 35,034 MW were in the final release but not in the early release at all.
What it does not settle. On the terms I fixed in advance, this means the annual record was mostly added during the year. But:
- An annual filing that reached EIA after the early release was compiled would give the same result, so the write-up says the annual file gained this capacity during the year, not which form brought it in.
If you use the annual EIA-860's planned generators as a plan: the final release's list for the current year includes capacity added during that year. This is an association in two dated copies of one survey and the monthly survey's 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)