FHFA House Price Index Explained: What HPI Measures—and Why It Is Not the Median Home Price

The FHFA House Price Index (HPI) is a measure of how single-family house prices in the United States are changing over time. Published by U.S. Federal Housing (FHFA), it estimates average price changes for the same properties across repeat transactions. It measures price change, not price level.

That makes it different from the median home price. A median tells you the middle sale price among homes sold in a period. The FHFA HPI tells you how prices moved for comparable properties. FHFA states this directly: the HPI measures neither mean nor median house prices, but changes in home prices over time. (Source: FHFA, HPI Frequently Asked Questions.)

In its most recent monthly release, published September 29, 2026, FHFA's flagship seasonally adjusted Purchase-Only index estimated that U.S. house prices rose 0.3% from June to July 2026, and were 2.6% higher in July 2026 than in July 2025. (Source: FHFA, U.S. House Price Index release, September 29, 2026.)

Those figures describe price movement under FHFA's repeat-sales method. They do not say that the median U.S. home price rose 2.6%, or what a typical home costs. The useful question, then, is what exactly an index like this measures, and this article answers it. It explains how the repeat-sales method works, what an HPI index value means, which homes and mortgages the flagship index covers, why historical values change, and what the index can and cannot tell you. It contains no forecast.

About the data

The HPI is a modeled index, not a summary of sale prices. The flagship Purchase-Only HPI is not calculated as the mean or median sale price of its input transactions. Instead, FHFA uses those transactions to estimate how prices changed for repeat transactions on the same properties. Unless otherwise noted, figures in this article refer to the flagship Purchase-Only, seasonally adjusted series. FHFA suggests that citations identify the index type and adjustment, for example "FHFA HPI (Purchase-Only, seasonally adjusted, nominal)." (Source: FHFA HPI FAQ.)

What the FHFA House Price Index actually measures

FHFA describes the HPI as a broad measure of single-family house price movements in the United States. Strictly speaking, it is a suite of indexes covering different geographies, reporting frequencies, and data samples, all built with the same technical methodology and collectively called "the FHFA HPI." (Source: FHFA HPI FAQ.)

What ties the suite together is what it measures: average price changes for sales or refinancings of the same properties over time. For the flagship Purchase-Only index, that means sales only. FHFA builds the indexes from more than 100 million home sale and refinance transactions involving mortgages purchased or securitized by Fannie Mae and Freddie Mac. It publishes indexes for the nation, census divisions, states, metropolitan areas, and, on a developmental basis, smaller areas such as counties, ZIP codes, and census tracts. (Same source.)

Four features define what the headline number represents:

• It measures change. The HPI estimates how much prices rose or fell between periods.
• It follows the same properties. Price change is inferred from repeat transactions on individual homes, not from comparing different homes sold in different months.
• It covers single-family properties. Condominiums, cooperatives, multi-unit properties, and planned unit developments are excluded from the flagship sample.
• It is not a dollar price. An HPI value is a relative number, not an estimate of what any home, or the typical home, costs.

The HPI is also nominal: it is not adjusted for inflation. FHFA notes that users who want real, inflation-adjusted price changes can adjust the index with a consumer price measure. (Source: FHFA HPI FAQ.)

Why HPI is not the median home price

The median home price and the FHFA HPI answer different questions.

Median home price

Question

What was the middle sale price among homes sold in this period?

Compares

A price level, in dollars

Unit

Both price changes and changes in which homes sold

FHFA HPI

Question

How did prices change for repeat transactions on the same properties?

Compares

A price-change index, relative to other periods

Unit

Price changes, estimated with the mix of homes held more constant

Different questions. Different measures.

The key difference is composition. A median is calculated from whichever homes happened to sell. If the mix of homes sold shifts, the median can move even when the value of any given home has barely changed. FHFA makes the same point: mean and median prices reflect changes in both prices and the composition or quantity of housing, whereas the HPI is designed to isolate changes in house prices. (Source: FHFA HPI FAQ.)

A simple example shows how that works.

Illustrative only — hypothetical sales, not market data

Month 1

Homes sold

Mostly smaller, lower-priced homes

Median sale price

Lower

Price change for any individual home

Little or none

Composition: mostly smaller, lower-priced homes sold

Month 2

Homes sold

More larger, higher-priced homes

Median sale price

Higher

Price change for any individual home

Little or none

Composition: more larger, higher-priced homes sold

In this example, the median rises because more expensive homes sold in the second month, not because any particular home gained value. A repeat-sales index compares each home with its own earlier price, so it is less affected by this kind of mix shift. The reverse can also happen. If more lower-priced homes sell, the median can fall even when same-property prices are stable.

This is why FHFA HPI changes and median-price changes can differ in the same period, and why neither should be substituted for the other. The example is conceptual. It does not describe what happened in any particular month or market.

How repeat-sales methodology works

The core idea is simple. When the same property sells more than once, the difference between its earlier and later prices carries information about how house prices changed over that interval.

In outline:

• A property sells at one point in time.
• Later, the same property sells again.
• The pair of transactions becomes one observation of price change for that home.
• Across a very large number of such pairs, observed over different intervals, a statistical model estimates an index value for each period.

Same property

↓

Sale 1

↓

Time elapses

↓

Sale 2

↓

Price change contributes to HPI

Conceptual illustration — FHFA uses a weighted statistical model, not a simple average of two sales.

FHFA describes its approach as a modified version of the Case-Shiller index methodology, using a geometric weighted repeat-sales procedure. The repeat-sales procedure estimates price changes using repeat transactions on the same property units over time. (Source: FHFA HPI FAQ.) The full methodology is set out in FHFA's technical description of its house price indexes.

It is not a simple average of the price changes from each pair of sales. The model estimates all period values together from the full set of repeat transactions. The model also applies statistical weights derived from estimated variance in repeat-transaction price changes. FHFA's technical methodology allows that variance to change with the elapsed time between repeat transactions. Longer intervals can introduce more variation from factors other than broad market appreciation. (Source: FHFA, HPI Technical Description.) For the monthly index, FHFA groups all transactions within the same month and estimates a separate index value for each month. (Source: FHFA HPI FAQ.)

Two practical details matter. FHFA uses the loan origination date as the transaction date in estimation, not the date when Fannie Mae or Freddie Mac acquired the loan. (Same source.) And the national index is built up from state-level indexes. Census division indexes are formed from weighted state growth rates, and the national index from the divisions. The weights are estimates of each state's share of one-unit detached housing. (Same source.)

What "constant quality" means—and what it does not mean

FHFA calls the HPI a "constant-quality" house price index. The reason is that comparing price changes for the same properties over time, with weights that account for the characteristics of the transactions, helps control for differences in the quality and composition of the homes in the sample. (Source: FHFA HPI FAQ.)

In practice, "constant quality" means the index is built to avoid a specific problem. It does not compare one month's group of homes with a completely different group in another month. Because each observation compares a home with itself, the index is far less affected by shifts in which kinds of homes are selling, such as a month with more large homes or a month with more small ones.

The phrase should be read with care, though. FHFA's wording is that the method helps control for quality differences. It does not claim the method perfectly holds every characteristic of every home constant over time. A home can be renovated, expanded, or allowed to deteriorate between sales. Repeat-sales methods compare the same address, not a frozen physical structure, so changes of this kind can still affect the price difference between two transactions. "Constant quality" is best understood as "much less affected by changes in the mix of homes sold," not as "adjusted for every change to every property."

What an HPI index value means

If you download FHFA HPI data, you will see a column of index values. These numbers are easy to misread.

An HPI index value is not a price. An index value does not mean a typical home costs that many dollars, or that many thousands of dollars, and it does not mean any particular home is worth that amount. FHFA states that index values have little standalone significance. Each one derives its meaning from its relationship to earlier or later index values. (Source: FHFA HPI FAQ.)

Index value

≠

Home price

Index values gain meaning through change relative to another period.

Index values are scaled to a reference period. FHFA normalizes most of its series to 100 in the first quarter of 1991, but the base varies by index type and geography. All-Transactions indexes keep their original base periods: the first quarter of 1995 for metropolitan areas, and the first quarter of 1980 for states and divisions. (Same source.) A base value of 100 does not mean homes cost $100 in that period. It is simply the starting point of the scale. Changing the base period would change every index level but would not change the percentage appreciation measured between any two periods.

What matters is the change between two values. FHFA's formula for appreciation between two periods is:

Appreciation = (later index value − earlier index value) ÷ earlier index value

Multiply by 100 to express it as a percentage. An equivalent form is (later ÷ earlier − 1) × 100.

ILLUSTRATIVE ONLY: fictional values, not FHFA data. If an index were 200 in one period and 210 in a later period, measured appreciation would be (210 − 200) ÷ 200 = 0.05, or 5%. The same 5% would result if the base were rescaled so the values read 100 and 105.

FHFA also provides a calculator on its HPI datasets page for computing appreciation between periods. (Source: FHFA HPI FAQ.)

What the latest FHFA HPI shows

FHFA's September 29, 2026 release covers data through July 2026:

These figures are not the change in the median U.S. home price, and they do not mean every home, or the average home in the entire housing stock, became 2.6% more expensive. They are estimates of average price change for the single-family properties and mortgages the flagship index covers.

The national figure also averages across very different regional movements. For the nine census divisions, FHFA reported seasonally adjusted monthly changes ranging from −0.8% in the Mountain division to +1.5% in the Middle Atlantic division. Twelve-month changes ranged from +0.6% in the Mountain division to +6.3% in the Middle Atlantic division. (Same source.) A national increase does not mean every region moved at the same pace, or even in the same direction over a given month. These ranges describe measured price change, not the relative merits of any region.

The data are also not current to the release date. July 2026 figures released at the end of September reflect transactions dated by when their loans originated, a point covered below.

What homes and mortgages are represented

The flagship index is broad, but it does not cover every home sale in the United States. FHFA bases the Purchase-Only HPI on transactions for single-family properties involving conforming, conventional mortgages purchased or securitized by Fannie Mae or Freddie Mac. (Source: FHFA HPI FAQ.)

Two terms define that scope:

• Conforming: a mortgage that meets Fannie Mae and Freddie Mac underwriting guidelines and does not exceed the conforming loan limit.
• Conventional: a mortgage that is not insured or guaranteed by the Federal Housing Administration, the Department of Veterans Affairs, or another federal government entity.

(Same source.)

Single-family properties

+

Conforming conventional mortgages

+

Fannie Mae / Freddie Mac transaction data

↓

FHFA Purchase-Only HPI

Broad, but not every U.S. home transaction.

FHFA lists the following as excluded from the flagship sample:

• non-conforming mortgages, such as loans above the conforming limit
• non-conventional mortgages, such as FHA- or VA-backed loans
• condominiums and cooperatives
• multi-unit properties
• planned unit developments
• observations removed by data filters, such as extremely high or low values, implausible appreciation, repeat sales within the same period, or incomplete date or address information

(Same source.)

The flagship Purchase-Only HPI is based on Enterprise mortgage transactions. FHFA identifies the Expanded-Data HPI as the series that adds county-recorder records, including cash sales and purchases financed with non-conforming loans.

FHFA explains that it relies on Enterprise data because of their broad geographic scope and long history, extending back to 1975. Their mortgage data provide broad coverage of the U.S. housing market, although they do not encompass all residential mortgage transactions. (Same source.) The flagship HPI is best understood as a measure of price change for single-family homes bought with conforming, conventional Fannie Mae or Freddie Mac mortgages, a large and informative segment of the market, but not the whole market.

FHFA HPI is a family of indexes, not just one series

Because "FHFA HPI" can refer to several related indexes, it helps to know which one a figure comes from. All use the same general repeat-sales methodology but draw on different transactions or properties. (Source: FHFA HPI FAQ.)

FHFA HPI is a family of indexes, not just one series

Index	What it uses
Purchase-Only (flagship)	Sales prices from purchase transactions with Enterprise-acquired conforming, conventional mortgages
All-Transactions	Adds appraisal values from refinance mortgages to the Purchase-Only sample
Expanded-Data	Adds FHA-backed mortgage transactions and county recorder sales records, including cash sales and purchases financed with non-conforming loans
Distress-Free	Removes short sales and sales of bank-owned properties from the Purchase-Only sample before estimation

Swipe horizontally to view all columns →

Source: FHFA HPI Frequently Asked Questions.

FHFA also publishes developmental indexes, including annual indexes for small areas, a Puerto Rico index, and a manufactured housing index. FHFA uses the Expanded-Data index each year to adjust the conforming loan limits. (Same source.)

The practical point is that these are not interchangeable. FHFA notes that the indexes tend to show similar long-term trends, but differences among them can be more pronounced over shorter periods. (Same source.) A figure from the All-Transactions index should not be presented as the flagship Purchase-Only figure, and a comparison should use the same index type throughout. Distressed sales, for example, are included in every FHFA index except Distress-Free.

Why seasonally adjusted and unadjusted HPI values should not be mixed

House prices tend to follow recurring patterns through the calendar year. FHFA publishes many of its indexes in both seasonally adjusted and non-seasonally adjusted form. Its seasonal adjustment is designed to remove recurring calendar-year patterns that would otherwise partially confound month-to-month or quarter-to-quarter comparisons. FHFA uses the Census Bureau's X-13 ARIMA procedure. (Source: FHFA HPI FAQ.)

The headline monthly figures in FHFA's releases, including the July 2026 changes of +0.3% and +2.6%, come from the seasonally adjusted Purchase-Only index.

Two cautions follow.

• Seasonal adjustment does not remove all volatility. It removes estimated recurring calendar patterns. It does not smooth out irregular movements, so monthly changes can still be noisy.
• Adjusted and unadjusted values belong in separate series. A trend built by mixing seasonally adjusted values for some months with unadjusted values for others would combine two different measurements. Any comparison should use one basis throughout, and say which one it uses.

Why FHFA revises historical HPI data

A reader who downloads FHFA HPI data one month and again the next may find that some earlier index values have changed. That is expected. FHFA revises historical estimates as additional transaction data become available, and it identifies three main reasons. (Source: FHFA HPI FAQ.)

• New repeat transactions. Because the index is built from repeat transactions, a property that sells again adds new information about its price change since its previous transaction. That can affect index estimates for both current and earlier periods.
• Purchases of seasoned loans. Fannie Mae and Freddie Mac sometimes acquire mortgages that originated in earlier periods. When these loans enter the data, they add information about earlier transactions and can revise historical values.
• Timing of data availability. Loan originations typically take 30 to 45 days to reach Enterprise funding, followed by additional processing. FHFA generally receives data on new originations with a lag of about two months. Data arriving with longer lags can continue to revise estimates, particularly for recent periods.

The same lag explains why the HPI trails the calendar. The July 2026 figures released on September 29 describe transactions dated by their July loan originations, not house prices at the time of the release, and certainly not in October 2026.

Revisions are not errors. They are how the index incorporates information as it arrives. In the September 29 release, FHFA reported that the previously published June change was unchanged. In other releases, recent months may be revised. For analysis, it helps to note which release, or vintage, a figure came from.

How FHFA HPI differs from housing inventory and affordability

House prices, for-sale availability, and affordability are related topics, but they are measured separately.

House prices, for-sale availability, and affordability are measured separately

Measure	Question it answers
FHFA HPI	How are house prices changing for repeat transactions on the same properties?
Housing inventory	How much for-sale stock is available, under a given source's methodology?
Affordability	Can households carry housing costs relative to their incomes?

Swipe horizontally to view all columns →

Source: FHFA HPI Frequently Asked Questions.

HPI is not inventory. A market could see rising inventory alongside rising prices, falling inventory alongside slowing prices, or other combinations. Neither measure mechanically determines the other. The Insight on what housing inventory and active listings actually measure covers the supply side of the for-sale market.

HPI is not affordability. Rising HPI does not, by itself, quantify a decline in affordability, and slower or negative HPI growth does not mean housing is affordable. Affordability also depends on household incomes, mortgage rates, property taxes, insurance, other housing costs, and financing terms. FHFA notes that the HPI is one input researchers use when assessing differences in affordability across areas. (Source: FHFA HPI FAQ.) It is an input to that analysis, not a measure of affordability on its own. For the income side, see the Insight on median household income by metro area.

What FHFA HPI can—and cannot—tell you about a housing market

Used carefully, the FHFA HPI can help describe:

• Average price appreciation or depreciation: how much prices rose or fell over a period, for the properties the index covers.
• Price movement over time: including long-run trends, using a consistent repeat-sales method with data back to 1975.
• Differences in price change across geographies: for the nation, census divisions, states, metropolitan areas, and other areas where FHFA publishes series.
• Movement with mix effects reduced: price change that is less affected by shifts in which kinds of homes are selling.

On its own, the FHFA HPI cannot tell you:

• the median or average home price
• what a specific property is worth
• whether housing is affordable, or what mortgage payments cost
• rent levels (FHFA notes that the HPI does not track rent data)
• how much inventory is for sale, whether there is a housing shortage, or how much rental housing is vacant
• sales volume or pending-sales activity
• the strength of housing demand, or whether buyers or sellers hold more negotiating power
• future home prices
• investment returns, or whether a market is a good investment

The HPI is a historical and current measure of price movement. It is not a forecast. July 2026's 2.6% annual increase describes the year to July 2026. It does not indicate what prices will do next, and it should not be extended into an expected future rate of return.

How to compare HPI data correctly

HPI comparisons are only meaningful when they hold the measurement constant. Before comparing two HPI figures, check that they use:

• The same index type: Purchase-Only with Purchase-Only, not Purchase-Only against All-Transactions or Expanded-Data.
• The same seasonal basis: seasonally adjusted with seasonally adjusted, or unadjusted with unadjusted.
• The same frequency: monthly with monthly, or quarterly with quarterly. FHFA notes that quarterly changes compare typical values throughout one quarter against another, not end-of-quarter values. (Source: FHFA HPI FAQ.)
• The same geography: an identically defined area, keeping in mind that FHFA uses current metropolitan delineations.
• The same periods: matching start and end points when comparing appreciation across places.
• The same release vintage, where practical: because historical values are revised.

An example of what to avoid is comparing a seasonally adjusted Purchase-Only change for one market with an unadjusted All-Transactions change for another, as if the two were equivalent. Each choice can shift the result, and combining different choices can produce a misleading contrast.

Another is comparing an HPI change with a median-price change and treating the gap as an error in one of them. As explained above, the two measure different things, and they can legitimately diverge.

For local housing context alongside other indicators, you can explore Metro Analytics market pages. The Metro Analytics methodology explains how each measure on those pages is defined and sourced.

The takeaway

The FHFA House Price Index is a price-change index. It should be read as a measure of how prices moved, not of what a typical home costs. Its flagship Purchase-Only series uses a weighted repeat-sales method to estimate price change for the same single-family properties over time. That makes it less affected by shifts in the mix of homes sold than a median sale price is. Under that method, FHFA estimated U.S. house prices rose 0.3% from June to July 2026 and were 2.6% higher than a year earlier.

Reading the index well means knowing which index type a figure comes from, which properties and mortgages it covers, the geography and period, whether it is seasonally adjusted, and which release produced it. It also means remembering what the index does not measure: it is not a median price, not an affordability measure, not an inventory measure, and not a forecast.

To see local housing context for a specific area, explore Metro Analytics Markets, or view the Methodology to see how each measure is built.