Revisions: Why Published Housing Figures Move Later

A market figure is a snapshot of a database on the day someone ran the query. Databases keep changing after that day. So the figure changes too, and the version you read first is rarely the version that survives.

Almost nobody mentions this. Reports are published with the confidence of a final result, and the correction, if it comes, arrives quietly inside next month's document with no announcement that the previous month's figure moved.

Why the data is still moving after the month ends

Several ordinary processes keep a month's record open long after the calendar has closed it.

Late entry. A closing has to be recorded by a person. Recording lags the event. A sale that completed near the end of a month may not appear in the database until the report for that month has already gone out.

Corrections. Fields get entered wrong and later fixed — a square footage, a property type, a status, a date. Each correction changes what a query returns for a period already reported on.

Status changes. A transaction recorded as complete can be unwound, and a record can be reclassified after the fact. The count for that month drops afterwards.

Reclassification. A property recorded under one category gets moved to another. If a report separates categories, the figures for both move even though no transaction changed.

None of this is negligence. It is what any large administrative dataset does. The problem is only that reports present a provisional figure with the typography of a final one.

The revision runs in a predictable direction

Because late entry is the largest of these effects, first releases tend to be incomplete rather than randomly wrong. The transactions most likely to be missing are the ones closest to the period boundary.

That has a consequence worth holding onto: the most recent period in any report is the least reliable one in it. It is also, invariably, the period the headline is about.

So the sentence that gets quoted, shared and reacted to is built on the least settled data in the document. If a figure looks dramatic and it concerns the period just ended, the honest response is to wait and see whether it survives its first revision.

Provisional versus settled, and how to tell them apart

Some publishers mark the distinction, with a footnote or a differently styled point at the end of a series. Many do not.

You can often infer it anyway. Look at whether the report restates the previous period. If last month's figure in this month's report differs from what last month's report said, you have just watched a revision happen, and you now know roughly how large revisions tend to be in this series.

That is a genuinely useful exercise. Do it a few times and you develop a sense of the ordinary revision size. Any first-release move smaller than that is not yet information.

A report that never restates history is either revising silently or not revising at all, and neither is reassuring. This is part of the wider question of the data behind Claremont market reports, and how to check it.

Small markets amplify the effect

In a large market, a handful of late-recorded sales barely disturbs a summary built on thousands. In a small one, a handful can be a meaningful share of the entire month.

Claremont is a small market. A few sales arriving after the report went out can move a monthly figure by an amount that would be invisible in a city ten times the size. Combine that with the thin samples discussed in what a median home price hides in a small market, and the first release for a single month deserves very little weight on its own.

THIS IS NOT A REASON TO IGNORE REPORTS. It is a reason to read the recent end of them loosely and the settled end of them more firmly.

Four habits that keep revisions from misleading you

Discount the newest period. Treat the most recent point as provisional by default, whether or not it is labelled that way. Its direction is a hint, not a finding.

Compare across editions, not within one. Keep last month's report. When this month's arrives, look at what happened to the overlapping period. That comparison teaches you more about the series than any single edition can.

Weight the older points. The parts of a series that have been through several revisions are the parts worth trusting. If you want direction, read the settled middle of the chart, not its leading edge.

Distrust a sharp final tick. A series that runs steadily and then jumps at the last point is often showing incomplete data rather than a turn. Wait one cycle. If the move was real, it will still be there.

Why this rarely gets said out loud

Partly format. A short report has no room for a methodology note and no audience demanding one. Partly incentive: a document exists to be read, and "these figures are provisional and will move" is not a compelling opening line.

And partly habit. Once a series has been published the same way for years, the caveats fall off and only the chart remains.

You can supply the caveat yourself. It costs nothing and it prevents the most common error in reading local housing data, which is reacting confidently to a figure that has not finished being calculated.

What to do when a decision cannot wait

Decisions have their own timing and cannot always pause for a revision cycle. When yours cannot, shift weight away from aggregate figures and toward the individual transactions relevant to the property in question — their condition, their terms, and what actually happened in each negotiation.

Those are observable now, and they do not get revised. This is a large part of why a CMA answers what a market report cannot: it is assembled from specific sales examined individually rather than from a summary of a database still in motion.

Anthony prepares a comparative market analysis for that purpose. It is not an appraisal; when a lender or a court requires an appraisal, he coordinates an independent state-licensed appraiser.

The rest of the series is on the market reports hub.

Anthony Grynchal has been licensed in California since November 2009.