What was the lowest year for U.S. housing starts?

Q: What was the lowest year for U.S. housing starts?

U.S. housing starts hit the all-time low of 554,000 units in 2009 — the deepest single-year reading since the Census Bureau series began in 1959. Multifamily starts bottomed at 109K, the lowest since the series's start. Annual housing starts dropped to 554,000 units in 2009 — the lowest reading in the 65-year Census new residential construction series. The previous all-time low had been 1.06M in 1982, set during the Volcker housing crash; 2009 cut that figure roughly in half. How starts collapsed from 2.07M to 554K The decline ran across a four-year window: 2005 peak: 2,068K total starts 2006: 1,801K (-12.9%) 2007: 1,355K (-24.8%) 2008: 906K (-33.1%) 2009 trough: 554K (-38.9%) — a -73% peak-to-trough drawdown What drove it Three forces compounded. First, the subprime collapse eliminated demand from the marginal first-time buyer cohort. Second, the financial crisis dried up construction-loan availability — many local banks holding builder credit lines failed or were taken over, and the lines were not reissued by acquirers. Third, builders carried massive land inventories at peak prices that took years to write down and dispose of; new starts had to wait until that inventory cleared. The slow recovery Starts recovered slowly from the 2009 trough: 2010: 587K 2012: 783K (first significant recovery year) 2015: 1,108K (above the 1981 trough) 2019: 1,295K 2021 cycle peak: 1,605K 2024: ~1,360K U.S. starts have not exceeded the 2005 cycle peak in any year since, and remain well below the 1972 all-time peak of 2.36M. The cumulative undersupply produced by 2008-2014 is the single largest driver of the contemporary housing supply deficit. Related Housing starts dashboard Q&A — All-time housing starts peak Q&A — What is the housing supply deficit? Q&A — How long did the 2008 housing recovery take? The 2008 Subprime Housing Collapse

Question

Short answer. U.S. housing starts hit the all-time low of 554,000 units in 2009 — the deepest single-year reading since the Census Bureau series began in 1959. Multifamily starts bottomed at 109K, the lowest since the series's start.

U.S. housing starts in historically low years (Census Bureau)
Year	Total Starts	Context
1966	1.16M	Credit crunch
1975	1.16M	Oil shock recession
1982	1.07M	Volcker rate shock
1991	1.01M	S&L recession
2009	554K	Post-crash all-time low (modern era)
2024	~1.36M	Current — recovering but below norm

Annual housing starts dropped to 554,000 units in 2009 — the lowest reading in the 65-year Census new residential construction series. The previous all-time low had been 1.06M in 1982, set during the Volcker housing crash; 2009 cut that figure roughly in half.

How starts collapsed from 2.07M to 554K

The decline ran across a four-year window:

2005 peak: 2,068K total starts
2006: 1,801K (-12.9%)
2007: 1,355K (-24.8%)
2008: 906K (-33.1%)
2009 trough: 554K (-38.9%) — a -73% peak-to-trough drawdown

What drove it

Three forces compounded. First, the subprime collapse eliminated demand from the marginal first-time buyer cohort. Second, the financial crisis dried up construction-loan availability — many local banks holding builder credit lines failed or were taken over, and the lines were not reissued by acquirers. Third, builders carried massive land inventories at peak prices that took years to write down and dispose of; new starts had to wait until that inventory cleared.

The slow recovery

Starts recovered slowly from the 2009 trough:

2010: 587K
2012: 783K (first significant recovery year)
2015: 1,108K (above the 1981 trough)
2019: 1,295K
2021 cycle peak: 1,605K
2024: ~1,360K

U.S. starts have not exceeded the 2005 cycle peak in any year since, and remain well below the 1972 all-time peak of 2.36M. The cumulative undersupply produced by 2008-2014 is the single largest driver of the contemporary housing supply deficit.

Sources

U.S. Census Bureau Survey of Construction; National Association of Realtors Existing Home Sales report; Freddie Mac Primary Mortgage Market Survey; National Bureau of Economic Research Business Cycle Dating Committee.