New home sales: Census contract data, revisions and wide error bands

New home sales are published monthly by the Census Bureau and the Department of Housing and Urban Development at 10:00 a.m. Eastern in the New Residential Sales report. They count signed contracts on newly built homes, reported at a seasonally adjusted annual rate. The series is timely but volatile, with wide confidence intervals and large revisions, which limits how much futures react.

Senzoukria · Economic events · Updated September 2026


At a glance

Published by
US Census Bureau and HUD, New Residential Sales
Frequency
Monthly
Usual time
10:00 a.m. ET (9:00 a.m. Chicago)
Measures
Sales contracts on new single-family homes, seasonally adjusted annual rate
Also reported
Median and average prices, homes for sale, months' supply

What the data describe

A new home sale is recorded when a sales contract is signed or a deposit is accepted, even if construction has not started. That makes the series more timely than existing home sales, which count closings. It also makes it sensitive to builder incentives, such as mortgage rate buydowns, that can pull contracts forward.

Because the sample is small relative to the volatility of monthly sales, Census publishes a confidence interval with each estimate. A large monthly change can fall inside it, meaning the data cannot say with confidence whether sales rose or fell.

Revisions and noise

  • The previous months are revised in each report, often by amounts comparable to the headline change.
  • Regional figures are even noisier than the national total.
  • The months' supply of new homes reflects builders' inventory, including homes not yet started, which differs from the existing-home measure.
  • Price medians move with the mix of homes and regions sold; they are not quality-adjusted.

Why futures react modestly

New home sales connect to construction activity, builder earnings and the rates cycle. A strong or weak print can move home-builder shares and, sometimes, Treasury futures, but the wide error band and large revisions lead most participants to discount a single month. Index futures typically react only when the release lines up with a broader growth concern.

Order flow at 10:00 a.m.

New home sales often share the 10:00 ET minute with other data, and their own contribution is hard to isolate. When they print alone, the book usually does not step back much beforehand, and a large headline change can produce a brief spike that retraces as participants look at the revisions. On the footprint, that looks like a release bar with a long wick and heavy volume on the side opposite to the spike near its extreme.

A practical way to handle noisy releases is to define in advance what would count as a reaction: a close outside the opening range, or follow-through volume beyond the release bar. Anything less is treated as noise.

In Senzoukria

Rows whose name contains new home sales are described in the News calendar as “Housing demand — rate-sensitive sector.” Housing rows are often rated medium or low by the source; widen the Impact filter to see them.

Big Trades draws large executions as bubbles with a default minimum trade size of 50 and a 100 ms merge window; on a quiet release, the absence of large bubbles around 9:00 CT is itself a sign that size did not trade on the number.

In the same section

Sources

This page in other languages

Frequently asked questions

What is the difference between new and existing home sales?
New home sales count contracts on newly built homes and are published by Census and HUD. Existing home sales count closings of previously owned homes and are published by NAR. New home sales are more timely but much noisier.
Why are new home sales revised so much?
The estimates come from a relatively small sample and late-arriving reports. Census publishes confidence intervals, and a monthly change is often within them, meaning the direction is not statistically established.

Keep reading