Seattle permit-duration intervals: a simple lookup table beats the per-type baseline on width

A week-block bootstrap on Seattle's open permit feed shows a three-key lookup table gives narrower 80% intervals than the per-type baseline.

The research/build-data programme asks whether open permit records predict how long a permit takes to issue and a build to finish. In Chicago, San Francisco and Austin the plain per-permit-type quantile baseline gave the narrowest calibrated 80% interval in four of five city/target sets. Seattle (public-domain feed, 194,184 rows, test year 2024) differs.

With the same dequantised conformal interval, the three-key table (permit type, work class, cost bucket) is narrower than the baseline in both targets: application to issue by 7.4 days [95% CI -13.3, -1.1] on a width near 250, and issue to completion by 18.3 days [-23.4, -12.9] on a width near 650. The gain is 3% of width at best, from one test year, so it is a result for Seattle, not a rule. Coverage stays at 0.83 and 0.82 for the table against a nominal 0.80.

Numbers and caveats: research/build-data/LOG.md, session 63.

All writing · More in Research findings · More on build-data · All tools and research