Methodology

Four upstream publishers, in three families that are never blended into a single series. Closed sales come from the Redfin Data Center: median sale price, homes sold, inventory, median days on market, months of supply, sale-to-list ratio, share sold above list, price drops and new listings. Redfin publishes roughly mid-month. Listings come from the Realtor.com Residential Listings Data Library — read directly at ZIP level (source key realtor) and through FRED for counties and metros (source key fred) — with median list price, active/new/pending listing counts, price reductions and listing-side days on market, published in the first week of the month. Valuation is Zillow’s ZHVI, a smoothed index of typical home value across the whole housing stock, published mid-month.

A listing metric and a closed-sale metric answer different questions about different homes. Active listings are not the homes that closed that month, and the mix of what is for sale shifts constantly, so a spread between the two can move on mix alone. Every chart names its source; the only place the two families share an axis is the explicit list-vs-sale pair in the Explorer, which is labelled as such. Current per-source freshness is on the data health page.

Every observation is stored against its period end — the last day of the month the figure describes. FRED publishes on the opposite convention, dating a monthly series to the period start, so FRED dates are converted at ingest. Without that conversion the same July would land in two different buckets depending on which publisher it came from, and every cross-source comparison on the site would be off by a month.

Publishers do not run on the same lag, so a single page routinely shows different periods side by side — a Redfin closed-sale month, a later Realtor.com listing month, and a Zillow index month in between. Each number carries its own date for that reason. We never carry a figure forward to fill a gap and never interpolate a missing month; a missing observation stays missing.

Year-over-year change is computed by joining each observation to the observation twelve calendar months earlier — the actual prior period, matched on date. It is not a row offset. About a third of the monthly series in this warehouse have at least one missing month, and a row-offset lag silently reaches past the gap and compares the wrong months: during development that produced a printed +1046% where the truth was −26.4%.

The consequence is deliberate: when the prior calendar period is missing, the year-over-year cell is empty rather than filled with a number derived from the wrong month. A dash on this site means “not computable”, never zero. The same rule governs month-over-month.

Counties are keyed on five-digit FIPS codes, metros on CBSA codes, ZIPs on the ZIP code itself. Cities and places have no such key: neither Redfin nor Zillow publishes a place FIPS code, so a city id is a name slug derived from the published name (destin-fl). “Saint” is normalised to “st” as a whole word so that Redfin’s “Port St. Lucie” and Zillow’s “Port Saint Lucie” land on one geography instead of two half-empty ones.

Florida town names collide — there are two Midways and two Pine Ridges, among others — and a bare name slug would interleave two unrelated towns into one series. Rather than publish a blended series, any slug backed by more than one distinct region in a source file is excluded entirely. That is why some Florida towns have no page here. Separately, Walton County exists in both Florida and Georgia in Redfin’s data, so every region filter pins the state as well as the name.

The publishers disagree about what South Florida is. Redfin publishes it only as three metro divisions — Miami (33124), Fort Lauderdale (22744) and West Palm Beach (48424) — while FRED and Zillow publish only the parent CBSA, Miami-Fort Lauderdale-West Palm Beach (33100). Four metro geographies are therefore incomplete at the source: the parent has no Redfin closed-sale series, and the three divisions have no listing or valuation series. The other 28 Florida metros carry all three sources and are unaffected.

Those four gaps are filled, and every filled value is labelled. A rebuilt figure is published under its own source key — one ending _est — and every chart and figure built from it carries the source badge derived beside the publisher’s name, so it can never be read as a number that publisher released. Everywhere else on this site, and for every other geography, the figure is a number its publisher released for that exact geography: nothing is aggregated up from child geographies, nothing is copied across, and nothing is modelled.

The two mechanisms behind those four are not equivalent, and are not presented as such. Each division is coextensive with exactly one county — Miami with Miami-Dade, Fort Lauderdale with Broward, West Palm Beach with Palm Beach, agreeing to within 0.14% across 173 months and nine metrics where both are published — so filling a division’s listing and valuation series is a verbatim copy, not an estimate. Rebuilding the parent from its three divisions is exact for counts and for ratios of counts, which covers seven of its nine Redfin series. The remaining two, median sale price and median days on market, are a sales-weighted mean of three medians: a median of medians is not a median, so those two are estimates and are marked as such on the page.

The statewide figures on the homepage are the median of the county medians, not the median of Florida home sales. The two are different statistics and they are not interchangeable: a real statewide median would weight every transaction equally, so Miami-Dade would count for far more than Liberty County, whereas the median of county medians gives each county one vote regardless of size. It is the honest thing the published data supports — county medians are what the sources release; the transactions behind them are not — and each statewide tile prints the county count it was computed from.

The same caution applies to every aggregate here that is built out of medians. Read them as the centre of the county distribution, not as the typical Florida sale — and never average a set of medians and call the result a median.

A market with a handful of sales swings hundreds of percent from month to month and would otherwise own every extreme on the site. Every ranking, and every mover and extreme in the monthly recaps, is drawn only from geographies carrying at least 200 active listings in the latest Redfin inventory reading — the same floor the homepage discovery rail applies. A geography whose active-listing count is unknown is excluded rather than assumed liquid, and each ranking prints how many geographies cleared the floor.

The numbers are not mine to license. Every figure on this site is published by Redfin, Zillow, Realtor.com or FRED and stays under that publisher’s own terms. Aggregating a series, dating it and re-serving it does not acquire any rights in it and does not let this site sublicense it onward. That is why no single licence covers this dataset as a whole and why none is claimed — not an oversight, and not a licence to be inferred from silence.

What is this site’s own is the part around the numbers: which metrics are carried, the period and dating rules set out above, the geography crosswalk and its ids, the ranking floors, and the shape of the baked JSON. Reuse that freely, with attribution.

If you are republishing the underlying figures, read the publisher’s terms first — they differ from each other and they change: Redfin, Zillow, Realtor.com and FRED. Cite the original publisher alongside this site as the aggregator, so a reader can get back to the primary source and check a revision against it.

The baked JSON under /data/ is served without restriction, and isAccessibleForFree in the structured data says the same. Both describe access — whether you have to pay or register to fetch a file. Neither says anything about rights in the upstream data, and the two are easy to confuse.

None of the above is legal advice. Where this page and a publisher’s own terms disagree, theirs govern.