Foreclosure Pulse methodology

The Foreclosure Pulse is a velocity index. It answers one question per metro: are foreclosure-track filings arriving faster or slower than they were a month ago? It counts filings, not homes, and it does not estimate outcomes, prices, or returns.

Categories counted: foreclosure complaints and judgments, notices of default and trustee sale, and lis pendens. The published table shows the combined count. Per-category breakdowns are produced in the same run and will be exposed once each category clears its own coverage gate.

Short window: filings in the last 7 days, compared with the average weekly count over the 4 weeks immediately before that window. The percentage shown is (last 7 days minus the 4-week weekly average) divided by the 4-week weekly average.

Medium window: filings in the last 30 days, compared with the average 30-day count over the 90 days before that window. Same arithmetic.

Both baselines are trailing and metro-specific. No seasonal adjustment is applied, and no figure is annualized. A metro is compared only with itself.

A metro is scored only when every condition holds: the newest official-source check for its counties is within 48 hours, and the metro recorded at least 30 filings in the prior 4 weeks. Metros that fail either test stay in the table with the reason shown. Hiding them would overstate coverage.

A clean month is a calendar month with at least 10 filings and no source gaps. Clean-month counts are tracked per county and rolled up per metro. They gate the year-over-year comparison described under Limitations.

Every filing comes from a county court docket, county recorder, or a legally mandated published notice. Records are pulled by our own adapters from the official portal for each county.

Excluded on purpose: aggregator and reseller feeds, listing-site foreclosure counts, and any third-party dataset whose provenance we cannot trace to the originating court or recorder. When a contaminated source is found in the history it is purged and the affected windows are recomputed.

Filings are deduplicated on case or instrument identity before counting, so a docket that is re-scraped is not counted twice.

The index is recomputed every 3 days. Each run stamps a computed-at time, shown at the top of the pulse page, and republishes the full table. Source checks run nightly and are independent of the recompute; a metro whose sources were not refreshed within 48 hours of the recompute is marked rather than scored.

No year-over-year comparison is published for a metro until it has 13 consecutive clean months. Most metros do not have that history yet, so the table shows velocity against trailing baselines only. Any statement of the form "filings are higher than a year ago" is out of scope until that gate is met for the metro in question.

A national rating is published only when the stable panel reaches 100 counties. Until then the pulse page reports the panel size and withholds the rating.

Filing counts depend on each county's recording practice. Judicial and non-judicial states produce different document types at different points in the process, so cross-metro comparisons of the raw count are not meaningful. Compare a metro with its own baseline.

Recording lag exists. A filing dated inside the window may appear in the county system days later, which can revise the most recent week upward on the next run.

This is descriptive data about public filings. It is not a forecast, not legal advice, and not a statement about any individual property.