How Accurate Are Prediction Markets on Building Permits? Kalshi vs. the Census First Print

By Arthur Angelo, PredictionScout Research · Updated Oct 15, 2026, 7:56 AM ET

Key findings

• Typical miss: 57K. Median miss 57K units at an annual rate, average 42K, over 6 releases. Monthly permits typically run about 1.4 million at an annual rate.
• Threshold contracts right 88% of the time. 38 of 43 “above X” contracts went the way the market leaned a day before.
• Biggest miss: Jul 2026. Market median 1.385M, first print 1.443M.
• Settlement check. 6 of 6 Kalshi settlements match the Census first print.
• First prints get revised. Largest revision among scored months: April 2026, first 1.442M, now 1.423M (-19K). Markets are settled on the first print and are not reopened.

First print vs. today’s number

Kalshi settles on the first number the Census Bureau publishes. The Bureau revises it a month later and again in annual updates. Each scored month’s first print next to the current value:

Building permits (first print, annual rate)

Across 6 releases from Apr 29, 2026 to Sep 17, 2026, the market’s median a day before was off by 57K in a typical release (median miss) and 42K on average. The number came in above the market median 3 times and below it 3 times. Of 43 individual “above X” contracts, the side priced above 50¢ a day out was right 38 times (88%).

Highlighted rows missed by more than 50K. Official values: Census Bureau New Residential Construction, privately owned housing units authorized by building permits, seasonally adjusted annual rate, first print from ALFRED series PERMIT.

Download the data

Release-by-release scorecard (CSV)

One row per settled Kalshi event: the market median 24 hours before, Kalshi’s settlement value, the official first-release value and the miss. Per-contract prices are not included. Free to reuse under CC BY 4.0 with a link back to this page.

Methodology

• Markets. Kalshi’s KXBUILDPERMS series: “above X million” contracts on monthly building permits (seasonally adjusted annual rate), from Kalshi’s public API. The series started in 2026, so the sample is small.
• Official data. Census Bureau New Residential Construction release. First-published and current values come from ALFRED series PERMIT (the vintage published nearest the market close). Some 2026 releases were rescheduled, so release dates differ from Kalshi’s event names.
• Market median. For each release we take every “above X” contract’s price 24 hours before trading closed (midpoint of bid and ask from hourly candles when the spread is 10¢ or less, otherwise the last trade; looking back up to 48 hours). Prices are forced to fall as the threshold rises, and the median is the point where the implied chance crosses 50%, interpolated between neighboring thresholds. If the 50% point falls outside the listed thresholds the release is shown but not scored.
• Contracts called right. Each threshold contract counts once: right if the side priced above 50¢ a day before won. Contracts priced exactly 50¢ are skipped.
• Actual value. The value Kalshi settled on (its “expiration value”), which follows the first official release. We check it against the first-release value in ALFRED (St. Louis Fed’s archive of data as first published) and mark any difference.
• Limits. One exchange; thin trading in some older and far-out contracts; the market median is a summary of a full distribution, not a single forecast. Nothing here is investment advice. Last run: Oct 15, 2026, 7:56 AM ET.

Data sources: prices and volumes come from the public APIs of Kalshi and Polymarket, with housing data from the Census Bureau. All trademarks belong to their owners. PredictionScout is independent and not affiliated with any exchange. Nothing here is investment, betting, or legal advice.

Related: mortgage-rate markets · retail sales scorecard · GDP scorecard · All research