Housing data releases run from six weeks ahead of closings to nine months behind

Every coordinator forecasting the next quarter is asking the same question of a housing statistic, whether or not they phrase it this way: is this counting work I have already done, work that is already sitting in my pipeline, or work that has not been contracted yet. The answer decides whether a release is useful for planning or only for context, and the major releases sit at very different points on that line.

The releases that see forward

Only two widely published national series describe transactions that have been agreed but not closed, which is the definition of a coordinator's live file load.

The National Association of REALTORS publishes the Pending Home Sales Index monthly. Its methodology note states the case plainly: the index 'measures housing contract activity, and is based on signed real estate contracts for existing single-family homes, condos, and co-ops. Because a home goes under contract a month or two before it is sold, the Pending Home Sales Index generally leads Existing-Home Sales by a month or two.' NAR adds that the index draws on a sample covering about 40% of multiple listing service data each month, and that an index value of 100 equals the average level of contract activity during 2001.

Redfin publishes a pending-sales count on a weekly cadence, covering rolling four-week periods, drawn from its own data across 900-plus metro areas and running back to 2021. Redfin states the weekly series is 'subject to revision.' It is a brokerage's own data rather than an association's or a government agency's, which is a caveat worth carrying, but it is also the fastest published read on signed contracts available.

The release that scores the quarter just finished

NAR's Existing-Home Sales report is the most quoted housing number in the United States and the least useful for forecasting, because it counts completed transactions. When it reported on 11 August 2026 that July sales ran at a seasonally adjusted annual rate of 4.06 million, down 1.7% month over month and up 0.7% year over year, it was describing files that closed in July. A coordinator who worked those files knew their own count six weeks earlier.

That does not make it useless. It is the benchmark a desk measures itself against, and its composition detail is where the value sits for coordination work specifically. The same July report put first-time buyers at 29% of sales, cash sales at 26% of transactions, individual investors or second-home buyers at 14%, and distressed sales at 2%. Median time on market was 29 days. Those shares describe what kind of file a coordinator is likely to be handed, which is a different and more durable question than how many.

The new-construction number, and why its error bars matter

The Census Bureau and the Department of Housing and Urban Development jointly publish Monthly New Residential Sales. The June 2026 edition, release number CB26-121 dated 24 July 2026, put new single-family sales at a seasonally adjusted annual rate of 628,000.

Two things about that release are easy to misread. First, Census counts a new-construction sale at contract signing, not at closing, so it behaves as a leading indicator for the new-build segment rather than a record of completed work. Second, the margins of error are wide and Census prints them: the reported 1.6% monthly increase carries a stated margin of plus or minus 14.8%, and the 5.6% annual decline carries plus or minus 13.2%. A monthly move inside those bands is not a signal.

The database that looks nine months back

FHFA's National Mortgage Database, produced with the Consumer Financial Protection Bureau, is the most detailed public record of what American mortgages actually look like. It is built on a five percent sample of first-lien, closed-end mortgages reported to one of three major credit bureaus, and its aggregate files weight each sampled loan to represent twenty actual mortgages.

It is also the slowest of these sources by a wide margin. As of this writing, the monthly national file runs from January 1998 through September 2025, and FHFA's page dates the current release 26 June 2026. That is roughly a nine-month gap between the last month of data and the month it was published in. Nothing in NMDB will tell a coordinator what next quarter looks like.

What it will tell them is the shape of the market underneath the headline count. TC Bulletin read the monthly national file directly for this story. In September 2025, the most recent month available, NMDB records 425,000 originations nationally, of which 272,000 were home purchase loans and 153,000 refinances, a purchase share of 64.0%. Three months earlier, in June 2025, the same series shows 409,000 originations at a 73.8% purchase share. The total barely moved. The mix moved a great deal, and mix is what determines whether a coordinator's incoming work is purchase files or refinance files.

Putting a planning calendar together

1. For the next four to six weeks of closings, watch Redfin's weekly pending-sales count. It is the earliest published signal, and it is directional rather than precise.
2. For the next one to two months, watch NAR's Pending Home Sales Index, released monthly. NAR's own methodology sets that lead time.
3. For where you have just been, read NAR's Existing-Home Sales report, and read its composition shares rather than only the headline rate.
4. For new-construction exposure, read the Census and HUD release, and ignore any single month's change that falls inside the printed margin of error.
5. For questions about mix, borrower profile or loan type, use FHFA's NMDB and accept that you are describing last year.
6. For anything local, go to the state or metro breakdown. Every one of these sources publishes below the national level, and national averages have been concealing very large local differences through 2026.