The Word That Left
Dating a vocabulary change to the quarter, across 11,239 Arizona luxury listings
Between 1998 and 2026, 11,239 Arizona residential listings priced at $2 million or above described the principal bedroom in their public remarks. Through 2019 — across 4,700 of them, twenty-two consecutive years — the word was master. Exactly two listings in that entire span used primary instead: one in February 2006, one in February 2011. By the third quarter of 2026, primary accounts for 93.2% of the room’s mentions and master for 6.8%.
Lexical replacement — one word displacing another for the same thing — is normally inferred from samples: a corpus of books, a slice of newspapers, a survey of speakers. A multiple listing service is a different kind of record. It is a near-complete written census of what one profession said about the same object, timestamped to the day, for nearly three decades. That makes it possible to date a replacement to the quarter rather than the decade, which is the only reason this page exists.
What follows is a measurement, not an argument. This study takes no position on the merits of the change, and readers should not infer one. It reports what the record contains.
The series
The shape is a textbook S-curve, and its milestones are unusually crisp. Through 2020Q3, master stood at 100.0% — not a single qualifying listing that quarter, or any quarter since 2011, used primary. 2020Q4 brought the first sustained appearance, at 1.2%. 2021Q2 was the last quarter master held above 90%. 2021Q3 to Q4 was the break: primary moved from 13.7% to 28.1% in a single quarter, the steepest move in the series. 2023Q2 was crossover, the first quarter above 50%. 2026Q1 passed 90%.
Nineteen quarters separate the last 90% master reading from the first 90% primary reading. Six quarters carried the term from 10% to 50%; twelve more carried it from 50% to 90%. The second half of a replacement takes twice as long as the first, which is what the holdouts look like when you measure them.
Annual counts
| Year | Listings | master | primary |
|---|---|---|---|
| 2000–2005 | 796 | 100.0% | 0.0% |
| 2006 | 444 | 99.8% | 0.2% |
| 2007–2010 | 1,491 | 100.0% | 0.0% |
| 2011 | 189 | 99.5% | 0.5% |
| 2012–2019 | 2,354 | 100.0% | 0.0% |
| 2020 | 543 | 99.6% | 0.4% |
| 2021 | 665 | 89.5% | 10.5% |
| 2022 | 814 | 63.1% | 37.8% |
| 2023 | 840 | 44.2% | 56.9% |
| 2024 | 1,010 | 33.0% | 68.3% |
| 2025 | 1,178 | 21.6% | 79.8% |
| 2026 (to Aug 19) | 701 | 11.0% | 90.2% |
How the change moved
An aggregate curve this smooth invites the assumption that someone ordered it. The record does not support that reading.
It did not cluster by brokerage. Across agencies with at least 40 qualifying listings in 2021–2022, adoption rates ranged from 12.8% to 39.4% — but only 4.4% of the total variation in term choice sits between agencies rather than within them. Three offices of the same brand spanned 20.8% to 33.3% in the same window.
And individual agents did not convert. Among 161 listing agents with six or more qualifying listings since 2021, 16.8% used primary exclusively and 5.0% never used it at all. The remaining 78.3% used both. Of those mixed agents, 41 show a clean one-way switch — every master listing preceding every primary one — while 85 alternate back and forth across their own listings.
The aggregate is decisive; almost every individual inside it is not. That is the finding, and it is offered without explanation, because this dataset contains listing text and dates and nothing that would identify a cause.
Data and methods
Universe. Arizona Regional MLS residential listings with an original or final list price of $2,000,000 or more, list dates 1998 through 19 August 2026, deduplicated by MLS number: 39,539 listings. Speculative and to-be-built listings and a quarantined data-error class are excluded on the rules published in One Shot. The unit of analysis here is the individual listing, not the merged campaign, because the question concerns text written on a particular date.
Term matching. Public Remarks were lowercased and matched with two regular expressions: master ?(bed|suite|bath|retreat) and primary ?(bed|suite|bath|retreat). A listing counts toward a term if it matches that expression anywhere in its remarks. Listings matching neither — the majority, since most remarks never name the room — are excluded entirely, so all percentages are shares of listings that mention it. 54 listings match both and are counted in both columns.
Thresholds. Quarters below 40 qualifying listings and years below 40 are omitted as too thin to report. Adjacent thin years are pooled in the annual table, as marked.
Adoption analysis. Brokerage and agent figures use the listing agent and agency of record on each listing. Between-agency variance share is the between-group sum of squares as a proportion of total variance in term choice, computed across agencies meeting the 40-listing threshold. Agents are counted where they hold six or more qualifying listings from 2021 onward.
Structured fields. This study measures agent-written prose only. The ARMLS structured Rooms field continues to use the label Master Bedroom in current listings and has never contained primary. That field was not changed and is not analyzed here; anyone querying this data should search both terms across both layers.
Scope. One multiple listing service, one state, one price tier. These results describe Arizona $2M+ listings and nothing else. No inference to other markets, other price bands, or national practice is supported by this data, and none is offered. Listings that never reached the MLS are invisible. Private Remarks are confidential under MLS rules and were not used.
This is a short study drawn from the same dataset as One Shot, the full pricing paper. Questions about either are welcome — email Marta or call 480-274-5710.