Eight pre-registered tests on U.S. power plants that slip a month: what held, and what I take back

说明与用途

八项预注册测试评估了美国电力厂在EIA‑860M库存中出现一个月延迟是否能预测其下个月的运行情况。七项测试表明，太阳能和电池发电机在延迟后更可能运行，而风能和天然气发电机则没有一致的预测信号。

问题

Assess whether a generator’s one‑month slip in the EIA‑860M inventory predicts its operation status in the following month and evaluate the predictive value across technologies.

结果

Out of 33 pre‑registered predictions, 27 held. Solar generators with a one‑month slip were ~5% more likely to be operating next month than new solar listings; batteries ~8% more likely. Two‑to‑three month slips for solar were ~12.5% worse. Wind and gas showed no consistent pattern; a weighted‑plant test failed due to methodology.

工作流

• 1. Identify generators with a one‑month slip in the EIA‑860M inventory.
• 2. Record their operation status in the following month’s file.
• 3. Perform eight statistical tests, each with a pre‑registered prediction, before the outcome is known.
• 4. Resample whole plants 2,000 times to assess robustness.
• 5. Compare slip‑based predictions against actual operation rates for each technology.
• 6. Publish predictions and results on GitHub.
• 7. Repeat the process for future months (e.g., Oct‑Dec 2026).

证据

27 of the 33 predictions held.

a solar generator planned for next month again right after a one‑month slip was listed as operating in that month about 5 percentage points more often than a new solar listing

a battery about 8 points more often

For solar since 2022, a slip of 2-3 months has been a bad sign: 12.5 points worse than a one-month slip

the wind and gas gaps are within plant-to-plant noise in every period

The summary note with every prediction and outcome, the code, the raw output and the tests that check every number are here: https://github.com/munimtechnologies/eia860m-slip

来源