What does seasonally adjusted mean in housing data?

Question

Short answer. Seasonal adjustment removes predictable calendar patterns — like the spring buying surge or winter slowdown — so that month-to-month changes reflect genuine market shifts rather than normal seasonal rhythms. The Census Bureau and NAR both publish seasonally adjusted annual rate (SAAR) figures.

Seasonal adjustment: how it shifts raw housing data (NAR / Census Bureau methodology)
Month	Typical Raw Sales Trend	Seasonally Adjusted Reading
January	Low (cold weather, holidays)	Adjusted up
February	Low–moderate	Adjusted up
May–June	Seasonal peak (spring buying)	Adjusted down
October	Moderate-high	Near neutral
December	Low (holiday slowdown)	Adjusted up

U.S. housing markets have strong seasonal patterns. More homes close in May–August (school-year calendars, tax refunds, better moving weather) and fewer in November–February. Without adjustment, comparing December to June sales would tell you more about the calendar than about market conditions.

How seasonal adjustment works

The Census Bureau and NAR apply the X-13ARIMA-SEATS method (a statistical procedure developed jointly with the U.S. Census Bureau and the Bank of Spain). The algorithm estimates a seasonal factor for each month based on historical patterns, then multiplies the raw count by the inverse of that factor to produce a seasonally adjusted figure. The result is then annualized — multiplied by 12 — to produce the seasonally adjusted annual rate (SAAR), the number most often reported in news coverage.

Example: reading a SAAR figure

When NAR reports "existing-home sales came in at a seasonally adjusted annual rate of 4.06 million," it means: given the pace of sales in that month, if that pace continued for a full 12 months, the annual total would be 4.06 million. It does not mean 4.06 million homes were sold in one month.

When to prefer unadjusted data

The Housing Almanac's annual series uses unadjusted annual totals — NAR's published yearly existing-home sales count and Census's yearly new-home sales count. At the annual level, seasonal adjustment makes no difference because a full year contains all 12 months. The 2024 annual existing-home total was 4.06 million unadjusted; the 2024 annual new-home total was 679,000.

Median prices and seasonal adjustment

Median prices are more severely distorted by season than unit counts, because the mix of homes that sell changes by season. In summer, larger and more expensive homes are over-represented; in winter, smaller starter homes dominate. The seasonally adjusted median price attempts to hold the composition constant, but many analysts prefer to compare the same month year-over-year (e.g., April 2024 vs. April 2023) instead of relying on seasonal adjustment.

Sources

U.S. Census Bureau New Residential Sales methodology; National Association of Realtors Existing Home Sales methodology; U.S. Census Bureau X-13ARIMA-SEATS documentation.