U.S. Monthly · SAAR · Single-Family + Multifamily
Starts climbed to 1.385M annualized — up 55K from a year ago, but still below the 1.5M long-run "healthy" level. Chronic under-building since 2008 remains the structural cause of U.S. housing affordability problems.
Historical trend
Monthly SAAR.
Source: FRED · HOUST
The long view: since 1990
From the bubble to the under-building crisis.
How today stacks up
Tools for housing planning.
About Housing Starts
Housing starts is the monthly count of new residential construction projects begun in the U.S., published by the Census Bureau. It covers both single-family and multifamily (apartments, condos) and is reported as a seasonally-adjusted annual rate (SAAR). Today's 1.385 million means construction is being initiated at a pace that would total 1.385M homes annually if it continued.
Why housing starts are a leading indicator
Housing starts are one of the most-watched economic indicators because they:
(1) Lead the housing market by 6–12 months — homes started today are completed and sold later;
(2) Reflect builder confidence — when rates rise or sales slow, starts drop fast;
(3) Drive a broad supply chain — lumber, copper, drywall, appliances, labor;
(4) Tie to housing affordability — chronic under-building since 2008 has been the root cause of today's affordability crisis.
Reading this chart
The 2005–06 boom peaked at 2.27M starts. The crash bottomed at 478K in 2009 — the lowest in 50 years. The 2010s saw chronic under-building (well below the 1.5M long-run average) which is widely cited as the structural cause of today's affordability crisis. The 2020–22 surge to 1.75M briefly closed the supply gap. Today's 1.385M is below the long-run average — meaning we\'re still under-building relative to household formation.
Related trackers
Other live numbers that move with — or against — this one.
Frequently asked
What this number means, and what it doesn't.
Methodology
Source
Pulled from FRED · HOUST and cached on the EvvyTools server.
Update schedule
Refreshed automatically by our cron whenever the upstream source publishes a new value. Historical values are not revised after publication.
How we compute
Display value is the raw published number, unrounded. Comparison stats use the closest available reference date. We never edit the underlying data.