Published November 14, 2026 | Version v1

The Correction Freeze Problem in a New Domain: Blind and Exploratory Tests on Air Quality

  • 1. ROR icon Phenikaa University
  • 2. ROR icon Universitas Katolik Widya Mandala Surabaya

Description

Route 4 of the Correction Freeze Problem (CFP), the information limit, says that no observer can detect a slow change more reliably than the information in the data allows. Two earlier severe tests, on inflation expectations and climate belief, found that people stayed below this limit. This study asks whether Route 4 transfers to a new domain without adjustment. It applies the same predictions, unchanged, to fine particulate matter () in the United States, a domain the authors had never studied. The Route 4 test is preregistered by nature: the predictions were registered and published before the data were seen, and applied without change. EPA data show that the median annual  level fell by 45% between 2000 and 2025. With the noise measure used here, this change gives , and the best possible accuracy for a single observation lies between 0.67 and 0.81. Gallup's 2026 Environment Poll finds that 66% of U.S. adults say the environment is getting worse and 27% say it is getting better. No available measure of accuracy exceeds the limit. Route 4 has therefore survived another difficult challenge: a domain the authors did not know, with no parameter that could be adjusted. It has now survived three independent tests: two severe tests on inflation and climate, and this blind out-of-domain test on air quality. We report this result with caution. The available measures are indirect, and the test does not, on its own, separate Route 4 from rival explanations, because the survey asks about the environment in general, the data are national, and beliefs move strongly with party. Two further routes were examined in an exploratory way. For Route 2, an exploratory test of daily data from 2015 to 2025, with rules fixed before the data arrived, finds that people check air quality far more once it becomes alarming, as Route 2 predicts. Checking does not, however, stay at zero below the threshold, so the strict form is not supported, and the verdict is inconclusive: Route 2 is not refuted. For Route 3, belief moved with changes of presidential party rather than with the air, which fits Route 3 but also ordinary partisanship. The study closes with a county-level design that can separate these accounts.

Files

CFP_R4_AQI.pdf

Files (742.9 kB)

Name Size
md5:4d72ce4485a471543f3ae858e455a1b9
742.9 kB Preview Download