A repeat-sales index measures property-price change from repeated observations on the same assets; interpretation depends on its sample, weights, geography, and revisions.
A repeat-sales index measures property-price change by comparing two or more eligible transactions involving the same properties and combining those matched observations into a market-level index. It reduces distortion from changes in the mix of properties sold, but its result applies to a defined repeat-observation sample, not automatically to every property in the market.
The term describes a class of indexes rather than one universal series. Different providers can use public sale records, mortgage transactions, appraisals, or a combination of sources. They can also apply different property filters, statistical weights, geographic boundaries, seasonal adjustments, and revision policies.
A repeat-sales index estimates average price movement for its eligible matched-property sample. The word average refers to the model’s market-level estimate, not a promise that each property changed by that rate.
For example, an index may estimate that existing single-family homes in a metropolitan area rose 5% over one year. Individual outcomes can differ because of:
The index is therefore evidence about a market segment. It is not a direct observation of every home’s value.
The Repeat-Sales Methodology defines how records are matched, filtered, modeled, weighted, aggregated, and revised. The repeat-sales index is the published sequence of values generated by those rules.
This distinction prevents a common error: calculating one property’s percentage gain and calling it a repeat-sales index. A production index estimates period effects from many overlapping pairs whose earlier and later observations occur on different dates. The provider’s methodology determines how much influence each pair receives.
Most indexes are normalized to a reference value, commonly 100. If an index begins at 100 and later reaches 160:
1Cumulative change = (160 / 100 - 1) x 100 = 60%
The measured price level is 60% above the base-period level. The index does not say that the average property costs $160,000 or that every home gained 60%.
To calculate the change between any two observations from the same series:
1Percentage change = (Later index / Earlier index - 1) x 100
Suppose a metro index rises from 225 to 234:
1(234 / 225 - 1) x 100 = 4%
The index increased 4% over that interval. The comparison remains valid if the provider rebases the entire series because rebasing rescales levels without changing the percentage movement.
An index change is incomplete unless its period and adjustment basis are stated.
| Measure | What it compares | Interpretation issue |
|---|---|---|
| Month-over-month | One month with the preceding month | Timely but potentially noisy and seasonal |
| Quarter-over-quarter | One quarter with the preceding quarter | Smoother but still a short-run measure |
| Year-over-year | A period with the same period one year earlier | Less affected by ordinary seasonal timing but slower at turning points |
| Cumulative change | Any two selected index dates | Highly dependent on the chosen endpoints |
| Annual average change | One year’s average index level with another’s | Not the same as a year-end or latest-period rate |
A positive annual rate can coexist with a negative monthly rate. That means the index remains above its level one year earlier even though it declined in the latest month.
A seasonally adjusted (SA) series estimates and removes recurring calendar patterns. A not seasonally adjusted (NSA) series retains them. SA data are often more suitable for adjacent-month comparisons, while NSA year-over-year comparisons can avoid mixing model-based seasonal factors into the calculation.
Do not divide an SA value by an NSA value. Seasonal factors may also be revised, so label the exact series used.
Most property-price indexes are published in nominal terms unless stated otherwise. Nominal appreciation does not measure purchasing-power growth.
If a repeat-sales index rises 7% and the selected inflation measure rises 4%, the inflation-adjusted change is approximately:
1(1.07 / 1.04) - 1 = 2.88%
The appropriate deflator depends on the question and should cover the same period. A nominal index is not inherently misleading, but its interpretation must be labeled correctly.
Analysts sometimes apply a local index change to an earlier property value. Suppose a home had a supported value of $500,000 when the relevant index was 240. The index is now 252:
1Indexed estimate = $500,000 x (252 / 240) = $525,000
This produces a 5% index-based update. It is useful for portfolio estimates only if the starting value, dates, geography, property type, and selected series are appropriate.
It is not a current appraisal. The home may have performed differently because of renovations, damage, precise location, design, maintenance, marketability, or recent comparable sales. The older the starting value and the broader the geography, the greater the potential mismatch.
Repeat-sales indexes can differ in more than their publisher’s name.
| Type | Defining feature | Main analytical caution |
|---|---|---|
| Unweighted repeat-sales index | Eligible pairs receive equal regression weight | Pairs with different error variance can have equal influence |
| Weighted repeat-sales index | Pairs receive weights based on estimated noise or methodology rules | Weighting formulas differ and may be difficult to reproduce |
| Purchase-only index | Uses repeat sale prices and excludes refinance appraisals | Coverage depends on the source transaction sample |
| All-transactions index | Can add refinance appraisals or other valuations | Appraisal observations are not identical to market sales |
| Value-weighted aggregate | Higher-value housing segments receive more aggregate influence | Results can be more sensitive to expensive markets |
| Stock-weighted geographic aggregate | Component areas are weighted by estimated housing stock | The aggregate reflects geographic weights rather than current transaction values |
| Distress-filtered index | Excludes specified foreclosure, bank-owned, or short-sale observations | Useful for one question but not representative of all transactions |
Labels are not fully standardized across providers. Read the methodology rather than inferring the sample from a short series name.
The number of matched pairs is an important quality signal. A repeat-sales model discards otherwise valid single-sale properties because no earlier or later observation is available for comparison.
Low pair counts can produce:
Pair count is not the only quality measure. Ten thousand poorly matched or non-market pairs do not automatically produce a sound index. Users should consider matching quality, filters, representativeness, weights, and geographic concentration alongside sample size.
Repeat-sales indexes are naturally revision-prone because a later observation adds information about the interval since an earlier one. A home purchased years ago may enter the repeat-sales sample only when it is sold or revalued again.
Revisions can result from:
For backtesting, preserve the data vintage available at the historical decision date. Using today’s revised history can give a model information that a real-time analyst did not possess.
Two prominent U.S. families illustrate how repeat-sales products can differ.
| Feature | FHFA HPI | Case-Shiller |
|---|---|---|
| Primary source | Mortgage records associated mainly with Fannie Mae and Freddie Mac; some variants add other observations | Public deed, recorder, and assessor sale records |
| Common headline product | Seasonally adjusted purchase-only index | National, 10-City, and 20-City indexes, with SA and NSA versions |
| Financing boundary | Flagship sample centers on eligible conventional, conforming Enterprise mortgages | Eligible public-record sales are not restricted to Enterprise-financed purchases in the same way |
| Additional observations | All-transactions variants add Enterprise refinance appraisals; expanded data add specified outside sources | Core methodology uses eligible sale pairs rather than refinance appraisals |
| Aggregation | Larger geographies use component-area housing-stock weights | National and composite products use housing-market value weights |
| Geographic products | National, division, state, metro, and some smaller-area products | National and selected large-metro products |
See the dedicated FHFA House Price Index and Case-Shiller Home Price Index guides for each family’s sample and release rules.
Neither family is universally superior. The better choice is the series whose geography, property population, financing coverage, transaction type, frequency, and revision policy fit the analysis.
| Measure | What it primarily shows | Why it differs from a repeat-sales index |
|---|---|---|
| Median sale price | Middle price among properties sold in a period | Can change when the mix of sold homes changes |
| Average sale price | Arithmetic mean of period transactions | Sensitive to outliers and sales composition |
| Hedonic price index | Quality-adjusted movement based on modeled characteristics | Uses observed attributes rather than requiring a prior sale |
| Transaction volume | Number or value of sales | Measures activity, not constant-quality price movement |
| Appraisal | Supported value conclusion for a particular property | Property-specific rather than a broad market index |
| Rent index | Change in residential rents | Measures occupancy payments, not asset prices |
| REIT index | Performance of listed real-estate securities | Reflects security prices, income, leverage, and market expectations |
These measures are complementary. A repeat-sales index can rise while transaction volume falls, median prices move differently, rents stagnate, or listed real-estate securities decline.
Lenders and investors can use local repeat-sales indexes to update broad collateral assumptions and monitor current Loan-to-Value Ratio distributions. The result remains a model estimate and should not be treated as property-level verification.
Regional index scenarios can support estimates of mortgage default, loss severity, and concentration risk. A stress scenario should specify the index, geography, horizon, and assumed relationship between index movement and individual collateral values.
Housing-price paths can affect modeled home-sale prepayments, cash-out refinancing, default, and recovery. Analysts should align the index sample with the loans in the pool rather than selecting a series solely because it has a long history.
Repeat-sales movement can provide context for estimated Home Equity. It does not capture mortgage amortization, additional borrowing, transaction costs, or property-specific capital improvements.
House-price indexes help researchers study housing cycles, credit growth, mobility, consumption, and financial stability. They should be combined with sales, inventory, construction, mortgage rates, income, rents, delinquency, and local employment.
$300,000.Before using a repeat-sales index, verify:
This article is educational and does not provide an appraisal, lending decision, investment recommendation, or individualized financial advice.