Indiana homes are up 7.15% over the last twelve months. Vermont, 6.61%. Chicago's Cook County, 5.95%. Seattle is down 2.99%.
Those numbers come from the first run of the PropRaven Price Index (v0), a weighted repeat-sales home price index in the same methodology family as Case-Shiller and the FHFA index. It is built from 2,471,548 pairs of sales sitting inside roughly 110 million deed records we collect ourselves from county offices across 28 states. No purchased index. No licensed estimate. Every input is a document a county recorder stamped, and every pair is one parcel that sold twice.
What the index says right now
Trailing-12-month change, measured against the most recent quarter whose records are settled rather than still arriving.
Heating
| Market | 12-month change |
|---|---|
| Indiana | +7.15% |
| Vermont | +6.61% |
| Cook County, IL (Chicago) | +5.95% |
| Illinois | +5.52% |
| Kentucky | +4.41% |
Cooling
| Market | 12-month change |
|---|---|
| Washington | −2.99% |
| King County, WA (Seattle) | −2.99% |
| Pennsylvania | +1.74% |
Washington is the real decline in the set. Pennsylvania sits in the cooling table at a positive number, which is worth a beat: "cooling" here means the slowest of what we rank, not falling prices.
Illinois is the flagship. Its index runs from 100 in 2015Q1 to 182 by 2024Q1 to roughly 200 by 2025Q4, about +7% a year over the decade, with Cook County alone contributing 960,708 of the state's pairs. That is one county carrying nearly a million verified same-house price changes. Seattle cooling through 2025 and Colorado sitting roughly flat both match what practitioners in those markets describe, which is the sanity check a new index has to pass before anyone should care about its more surprising outputs.
Same house, sold twice
Most home price headlines rest on median sale prices, which move whenever the mix of what sold changes. A quarter heavy on new four-bedroom construction reads as a rising market even if every existing house held flat; a quarter heavy on starter condos reads as the opposite. A repeat-sales index only ever compares a house to itself: two recorded sale prices for the same parcel, and the gap between them is that property's appreciation. No mix-shift can argue with it.
Collect enough of those pairs and you can solve for a price path. Each pair says "between quarter A and quarter B, this property gained X." Line up millions of overlapping pairs and the quarters chain together: long-held pairs constrain a whole decade, recently traded pairs constrain the tail, and a regression finds the one sequence of quarterly index levels that best satisfies all of them at once. That regression is the Bailey-Muth-Nourse method from 1963, still the backbone of the repeat-sales family.
One refinement matters. A pair held for fifteen years tells you less per year than a pair held for two, because more can happen to a house over fifteen years that has nothing to do with the market: renovations, deferred maintenance, a new roof. So we model how much noisier long-held pairs are and re-solve, weighting each pair by the signal it carries. That is the Case-Shiller three-stage step, and it changes less than you might expect. Run the unweighted and weighted versions side by side and the 12-month numbers agree within about 1.5 percentage points across the large states: Illinois +5.51% versus +5.16%, Washington −2.99% versus −3.26%, Kentucky +4.41% versus +4.54%. An index that swings wildly on a weighting choice is telling you about the weighting, not the market.
The pair filters are deliberately blunt. Sale prices must fall between $1,000 and $100 million. Pairs closing less than 90 days apart are dropped as flips, duplicate records, or transfers between related parties rather than two arm's-length prices. Price ratios above 10x in either direction go as data errors, as do annualized log returns beyond ±0.5, roughly −39% to +65% per year. What survives is winsorized at the 1st and 99th percentiles within each state.
And nothing is imputed. A quarter needs at least 25 pairs to be rankable; a geography needs at least 12 rankable quarters plus 400 total pairs at the state level, or 150 at the county level, to publish at all. Everything under those floors is simply absent. No gap-filling, no borrowing from a neighboring county, no smoothing over a thin quarter.
What we publish today
The v0 index publishes 14 states and 11 counties. Here are the high-confidence states with the pair counts behind each one. A dash means we don't publish a 12-month change for that geography yet.
| Market | Repeat-sale pairs | Rankable quarters | 12-month change | 5-year CAGR |
|---|---|---|---|---|
| Illinois | 1,349,551 | 106 | +5.52% | +7.12% |
| Washington | 478,596 | 105 | −2.99% | +4.51% |
| Colorado | 160,290 | 106 | +2.61% | +2.45% |
| Kentucky | 120,930 | 105 | +4.41% | +7.45% |
| Pennsylvania | 74,685 | 57 | +1.74% | +7.71% |
| South Carolina | 59,983 | 73 | +2.70% | +9.56% |
| San Francisco County, CA | 57,419 | 100 | — | — |
| New York | 42,295 | 74 | — | — |
| Indiana | 40,419 | 17 | +7.15% | — |
| Vermont | 13,017 | 32 | +6.61% | +11.13% |
| Ohio | 11,196 | 41 | +2.21% | +8.51% |
| Tennessee | 3,150 | 21 | — | — |
"5-year CAGR" is the compound annual growth rate over the trailing five years. Look at the rankable-quarter column: Illinois, Washington, Colorado and Kentucky each carry more than a hundred consecutive quarters of measurable price history. That is a quarter century of the same houses trading and re-trading, reconstructed from primary documents.
Coverage follows the deed archives. The states above are where recorded transaction histories run deepest, which is why they came online first. Some large states will take longer for reasons that have nothing to do with modeling: Texas is a non-disclosure state, so sale prices are not part of the public record there, for anyone. Everywhere else the constraint is simply how much recorded history a given county has published, and that number climbs every month. The index widens as those archives come online.
Beyond the last twelve months
The most interesting pattern in the table is the gap between the five-year and one-year columns. South Carolina compounds at +9.56% over five years but prints +2.70% over the last twelve months. Pennsylvania: +7.71% against +1.74%. Ohio: +8.51% against +2.21%. Those are markets that ran hard and then settled, and they look like different places depending on which number you quote.
Vermont, at +11.13% over five years and +6.61% over the last one, has not settled. Colorado, at +2.45% and +2.61%, never ran in the first place, and is one of the few markets whose decade and whose year tell the same story.
Grain matters too. Colorado is roughly flat statewide, but the strongest county reading in the index belongs to Larimer County, Colorado at +3.45%. State-level cooling and county-level heating coexist routinely, which is the whole argument for measuring at the county line instead of the state one.
Methodology and data: the PropRaven Price Index (v0) is a weighted repeat-sales index (Bailey-Muth-Nourse regression with Case-Shiller three-stage variance weighting) computed from PropRaven's corpus of recorded deeds collected from county offices, currently about 110 million records across 28 states. Pairs are keyed on the parcel identifier within its county. Output is quarterly, rebased to 2015Q1 = 100 where that quarter is present. The trailing-12-month change is the index's primary output; multi-year growth rates carry more uncertainty in smaller markets. This is a v0: it publishes 14 states and 11 counties, masks everything below its pair-count thresholds, and imputes nothing. Figures are as of 2026-08-01 and are restated as the corpus grows.
The deed, parcel, and permit records behind these numbers are searchable at propraven.com.