DECODING: CBRE's 18.3%, CoStar's Sub-14%, Cushman's 20.1% Are Not One Number

Introduction

On July 29, 2026, CBRE put global office vacancy at 18.3%, down 30 basis points in Q2 2026, while two other firms described the American office market with very different levels. The apparent contradiction is not a licence to choose the prettiest number. It is a reason to ask which market each number actually measures.

Three vacancy rates can be accurate and still answer different questions.

CoStar's August 7, 2026 outlook placed national office vacancy below 14%, and Cushman & Wakefield reported 20.1% nationally. Those figures share a period, not necessarily a denominator, a geography or a property universe. A percentage becomes misleading the moment its label is discarded.

The documents do establish a narrower common point: each firm reported a recent improvement or stabilisation signal. They do not establish one authoritative vacancy rate for every office in the United States. That distinction is the whole story, not a footnote.

CBRE's 18.3% belongs to a global frame

The July 29 report names a worldwide measure

CBRE reported global office vacancy of 18.3% for Q2 2026. The report called the quarterly 30-basis-point decrease its largest since 2015. That is a defined observation from one firm, released on July 29, 2026, and it carries the global scope stated by CBRE.

The important word is global. A worldwide measure can illuminate broad conditions without becoming a substitute for a national American rate. Treating it as the latter would change the claim before any comparison begins.

The largest decline since 2015 is a movement

CBRE's historical comparison concerns the size of a quarterly drop, not a promise about the next quarter. A 30-basis-point movement tells readers that the reported rate changed within CBRE's series. It does not reveal how another firm constructed its own series.

That is why the date and the internal comparison must travel with the figure. The report supplies a trend inside CBRE's framework; it does not appoint CBRE as the referee for every competing dataset.

A market label is not a methodology.

The 12.3% prime figure narrows the property universe

Prime buildings are a separate category

CBRE also put vacancy in prime buildings at 12.3%, down 40 basis points. This is not a second national total disguised as a refinement. It is a result for a selected class of office property, and the category is part of the result.

A reader who removes the word prime manufactures a broader claim than CBRE made. The lower figure may be meaningful, but it belongs to the buildings that the report classifies as prime, not automatically to all offices.

A sharper decline still needs its boundary

The 40-basis-point reduction is larger than CBRE's global 30-basis-point move. That contrast may suggest that the selected prime segment improved more quickly within the report's categories. It cannot by itself map every office building or every American city.

Segment data can reveal divergence inside a market. It should not be used as a shortcut around the question of which properties each commercial-data provider counts.

The direction of a rate and its level are separate facts.

Midtown's 2.2% is a local fact, not a national verdict

One district has its own reading

CBRE reported prime vacancy of 2.2% in Midtown Manhattan. The number is striking precisely because it is tightly located and limited to prime space. It records one local result rather than a national condition.

Midtown Manhattan is not a proxy for the entire United States. Its presence in the report helps show why a portfolio of markets can contain sharp local scarcity alongside much higher broader vacancy figures.

The gap with 18.3% proves scope matters

The same CBRE material contains 18.3% global vacancy and 2.2% prime vacancy in Midtown Manhattan. Those numbers do not cancel one another because they do not purport to describe the same place or property group.

Putting them side by side is useful only when their boundaries remain visible. Once the boundaries vanish, the comparison becomes theatre rather than market analysis.

Prime offices are not the national office stock.

CoStar's sub-14% is its national measure

The August 7 forecast uses a different frame

CoStar said national office vacancy fell below 14% in Q2 2026 in a revised forecast published August 7, 2026. Its headline, 'CoStar projects steady decline in U.S. office vacancy,' is an outlook from the company, not an instruction to relabel CBRE's global series.

The word national matters as much as the figure. CoStar's statement addresses the United States, while CBRE's cited 18.3% figure is global. A comparison can be informative, but equality is not established by proximity on a page.

The mid-2025 peak is CoStar's reference point

CoStar said the Q2 2026 rate sat 30 basis points below its mid-2025 peak. The firm therefore supplied a peak-to-current comparison within its own historical record. The figure describes a retreat from a prior high in that record.

This is a useful stabilisation signal. It remains a CoStar signal, with CoStar's coverage and calculation choices, rather than a universal calibration for commercial real estate.

A local floor does not become a national average.

17 million square feet measures net absorption, not vacancy

A flow is not the same as a stock

CoStar reported 17 million square feet of positive net absorption from July 2025 through June 2026. Net absorption is a measure of occupied space gained or lost during a period. Vacancy is a measure of space reported as unoccupied at a particular point.

The two indicators can speak to the same market without becoming synonyms. A positive flow can coexist with a vacancy level that remains material, because one number tracks change and the other tracks a stock.

The twelve-month window has a defined start and end

The reported absorption period begins in July 2025 and ends in June 2026. That calendar window does not match a single-quarter snapshot in every respect, even though it ends near Q2 2026.

Time windows shape the narrative. A rolling year can show accumulated movement that a quarterly rate cannot, and a quarter can expose a turn that a longer period smooths over.

Absorption describes movement; vacancy describes the stock left behind.

Cushman & Wakefield's 20.1% carries its own method

The highest published rate is still source-specific

Cushman & Wakefield reported national office vacancy of 20.1% in its Q2 2026 Marketbeat material. It is the highest of the three headline levels in this record, but it is not evidence that the other firms made an error.

A higher result can follow from a different market universe or calculation. The fact block explicitly warns that the figures are materially different because the firms use different methodologies and market scopes.

Its annual change is 10 basis points

Cushman & Wakefield described its 20.1% national rate as down 10 basis points from a year earlier. That comparison is annual, not quarterly. It supplies a direction of travel inside the firm's own report.

The pace cannot be ranked cleanly against CBRE's quarterly move or CoStar's distance from a mid-2025 peak. Each reference period asks a different question of the underlying market.

A twelve-month flow cannot erase a different measurement frame.

Three reports share direction, not a single level

The common ground is a recent easing

CBRE recorded a quarterly decline, CoStar said vacancy was below its previous mid-2025 peak, and Cushman & Wakefield reported a year-over-year reduction. These are separate descriptions, yet they point in the same broad direction: recent pressure has eased in each published series.

That convergence is more defensible than pretending that 18.3%, below 14%, and 20.1% can be averaged into a meaningful market truth. Agreement on movement is not agreement on magnitude.

Stabilisation must remain qualified

The reports support the word stabilisation only as a qualified reading of recent changes. They do not supply a single common rate, nor do they state that every office market has healed.

A qualified conclusion is not timid. It is the only conclusion that keeps the firms' own categories, dates and measurement choices intact.

Cushman's 20.1% is not a rebuttal to CoStar's sub-14%.

Different denominators change the answer

The market universe is part of the measurement

The stated limitation is direct: CBRE, CoStar and Cushman & Wakefield report materially different national or broader vacancy levels for the same period because their methodologies and market scopes differ. The denominator is not background machinery. It determines what the percentage means.

Every vacancy rate needs an implied question: which buildings, which locations and which classification rules are included? Without those boundaries, a rate sounds more precise than it is.

No arithmetic average solves a scope problem

Combining the three rates into one average would create a fourth number with no published methodology. It would not reconcile the firms' definitions, and it would mislead readers into believing a shared sample exists.

The honest comparison keeps each figure attached to its publisher. Calculation cannot repair categories that were never designed to be merged.

A basis-point decline is evidence of change, not a declaration of recovery.

Dates organise the evidence

Q2 is not a complete calendar story

All three reports concern Q2 2026, but the associated comparisons differ: CBRE cites a quarter, CoStar measures against mid-2025 and Cushman & Wakefield measures against a year earlier. The observation period is therefore only part of the analytical frame.

A rate without its comparison period can look more decisive than the report permits. The surrounding clock tells readers whether the emphasis is a recent shift, a peak-to-current change or an annual move.

Publication date is not measurement date

CBRE published on July 29, 2026 and CoStar on August 7, 2026. Those release dates explain when the firms spoke, not necessarily identical collection procedures or identical coverage decisions.

Chronology matters, but it cannot be used to erase methodological difference. A later forecast is not automatically a correction of an earlier global report.

Publication dates do not make datasets interchangeable.

Prime demand cannot stand in for every office

The selected segment can outperform

CBRE's 12.3% prime vacancy rate and 2.2% Midtown Manhattan reading show that selected spaces can look far tighter than broader office measures. The documents therefore contain internal variation before any outside comparison is attempted.

This is why an optimistic prime figure should not be stretched into a verdict on the entire office market. Strong demand in one segment can coexist with persistent vacancy elsewhere in the available data.

A city result has a city-sized claim

The 2.2% figure is tied to Midtown Manhattan, not to a generic American downtown. Its usefulness lies in specificity: it shows where the cited CBRE category was especially tight.

Specificity is not a weakness. It prevents readers from confusing a vivid local number with a national diagnosis.

The shared signal is improvement, not numerical agreement.

The language of improvement needs discipline

A decline is not a forecast

A reduction in a published vacancy rate is a record of movement between stated reference points. It is not, by itself, a prediction that the next report will repeat the move. None of the cited figures removes uncertainty from the next quarter or year.

That limit protects against two errors at once: declaring a recovery finished too early, or dismissing a measured improvement because it is incomplete.

The documents support comparison, not a winner

No source in this record says that CBRE, CoStar or Cushman & Wakefield has supplied the one definitive vacancy reading. Each offers a different measurement, and each is useful when read on its own terms.

The contest is invented when reporters turn methods into rivals. The stronger reading asks what each instrument can see.

No average can repair mismatched denominators.

A careful reader keeps the source labels attached

Name the firm before the figure

Writing '18.3%' without CBRE, 'below 14%' without CoStar, or '20.1%' without Cushman & Wakefield strips away the information that makes the figures usable. Attribution is not decorative; it is the definition of the evidence.

The same rule applies to dates and segments. A number is not more accessible when its conditions are removed. It is merely easier to misuse.

Compare like with like before drawing a trend

A valid comparison begins with scope, property class and time reference. The available record warns against treating the three calculations as interchangeable, and that warning is itself a central finding.

The market may be improving. The evidence says so in several ways. It does not say there is one vacant-office number that settles the matter.

Stabilisation is a cautious description, not a finish line.

The verdict is a method, not a slogan

The data refuse a false consensus

The reports leave a clear, bounded conclusion: commercial real-estate firms can describe an improving office market while reporting very different vacancy levels. The difference arises from measurement choices and market coverage, according to the documented limitation.

That conclusion may lack the drama of a single number. It has the advantage of remaining true to every source in the record.

The next report should be read in context

Future releases may add a new movement, another property category or a revised forecast. They should be placed beside their own definitions before anyone treats them as confirmation or contradiction of the prior reports.

Office vacancy is not one number waiting to be discovered. It is a set of measurements that must be read with their labels on.

The useful question is always: vacancy of what, where, and when?

Conclusion

The market story in these reports is not that one firm is right and the others are wrong. It is that all three recorded some recent easing while looking through different measurement frames. The figures become informative only when their scope travels with them.

Precision begins by refusing a false single number.

CBRE's global 18.3%, CoStar's national sub-14% and Cushman & Wakefield's national 20.1% should therefore remain separate in public discussion. Their shared direction is real. Their numerical gap is real too.

The disciplined conclusion is simple: do not average the evidence into fiction.

Signature

Signed Maxime Marquette, columnist

Columnist's Transparency box

Editorial positioning

This column takes a public-interest view: published measurements, named institutions and plainly stated limits matter more than a convenient narrative.

The position is not that a disputed or partial record proves more than it does. It is that public claims must keep the qualifiers that make them intelligible.

Methodology and sources

The article uses only the documents listed below and separates published figures from interpretation. Links are collected in Sources so that readers can inspect the underlying reporting.

Where the available material names a source, date, institution or limitation, the text keeps that attribution. It does not add a private briefing, a witness account or an unreported data point.

Nature of the analysis

This is analysis and commentary based on a limited documentary record, not a finding by a court, regulator, medical examiner or scientific review.

The conclusions concern what the cited material supports, what it does not settle, and why the difference matters for public judgment.

Sources

Primary sources

• CoStar Group — U.S. office vacancy outlook — August 7, 2026
• CBRE — U.S. Office Market Report, Q2 2026 — July 29, 2026
• CBRE — prime-office indicators — July 29, 2026

Secondary sources

• Cushman & Wakefield — U.S. Office Marketbeat — Q2 2026
• CoStar Group — net absorption detail — August 7, 2026
• Cushman & Wakefield — annual vacancy change — Q2 2026