Do Office Tenants Really Renew 75% of the Time? What the Data Shows

Ask where the number originated and you hit a wall. The 2004 MIT thesis that tested it, Ryan Asser's "The Determinants of Office Tenant Renewal," simply identifies "70 to 75%" as the figure the industry uses in lease-rollover models. Fisher and Ciochetti, in their Real Estate Research Institute working paper, call it "a commonly used rate of 70% or 75% in industry lease models." Neither found a primary source behind it, and neither could we. It is an underwriting convention: a number that gets typed into Argus runs and appraisal models because it has always been typed into Argus runs and appraisal models.

Conventions are not automatically wrong. This one, though, has been tested against large lease datasets twice, by independent researchers using different landlords' data, and it failed both times in the same direction.

The two studies are worth knowing by name, because together they cover more than 16,000 lease outcomes.

Asser (MIT, 2004). Working from Equity Office Properties data, then the largest office landlord in the United States, Asser started with 26,389 leases from 1997 to May 2004 and reduced them to a final sample of 15,822 leases across 289 properties in 41 metro areas. The property-level mean renewal probability was 57.3%. The lease-level renewal rate was 58.5%. The minimum property in the sample renewed at 14%, the maximum at 90%.

Fisher and Ciochetti (RERI Working Paper 133). Using a different single landlord's portfolio, 3,655 new office leases across 14 markets negotiated between 1999 and 2005, they tracked 777 leases through expiration. The outcome split: 55% renewed, 34% did not renew, and 11% terminated early. Their words: "the renewal rate is merely 55% relative to a commonly used rate of 70% or 75% in industry lease models."

Current-cycle numbers are harder to pin down because public REITs define retention inconsistently or not at all. Kilroy publishes a precise formula. EastGroup and Healthpeak use different denominators. Highwoods publishes a "renewed square footage" share of leasing activity, which is not a retention rate. BXP discloses no retention percentage; dividing its reported second-generation renewal square footage by its expiring square footage yields a figure around 33%, though the denominator includes terminations and may not be comparable. The honest summary of the current cycle is a wide scatter, roughly 30 to 60% for office, with the dispersion itself telling you the old assumption has not gotten safer since 2005.

Asser's size gradient is the strongest single pattern in the renewal evidence, and it explains how the 75% myth survives contact with landlords' own experience.

Renewal probability in his sample rose from 52.1% for tenants under 5,000 square feet to 67.9% above that line. Cut at 10,000 square feet, the split is 54.0% versus 73.1%. Cut at 20,000 square feet, it is 55.7% versus 78.9%.

That gradient produces a measurement trick worth understanding. Asser's lease-level renewal rate was 58.5%, but his square-foot-weighted rate was 76.6%. Weighted by area, the industry standard looks almost right, because big tenants dominate the square footage and big tenants renew. Counted by tenant, it is badly wrong. Put differently: the annual move rate for a randomly chosen tenant is roughly double the move rate for a randomly chosen square foot.

Landlords underwrite dollars, and dollars follow square feet, so their convention approximately works for their purposes. A 12,000 square foot professional services firm is not a square foot. It is a tenant, and tenants that size renewed at a rate much closer to a coin flip than to the number in the landlord's model.

The other half of the negotiation math sits in REIT supplemental disclosures, where landlords report what they spend to sign a renewal versus a new lease.

Highwoods reported FY2024 tenant improvements of $5.13 per square foot on new office leases versus $3.78 on renewals, with leasing commissions of $1.32 versus $1.00. Federal Realty reported renewal TI of $3.25 per square foot against materially higher new-lease costs. At the extreme end, Healthpeak's Q4 2025 supplemental shows lab renewals at $0.11 per square foot per lease year versus $11.96 for new lab leases, a gap of roughly 100x driven by build-out intensity.

Add downtime. A renewal has no vacancy period; a new lease follows months of marketing, negotiation, and construction. The more specialized the space, the wider the spread.

This is the asymmetry a tenant-rep broker should be pricing: the landlord's alternative to renewing you is measurably expensive, and the landlord's model of how likely you are to leave is measurably optimistic. A mid-sized tenant who walks into a renewal negotiation knowing both numbers is negotiating a different deal than one who assumes the landlord holds the cards.

Renewal rates are conditional probabilities: they tell you what happens when a lease expires, not how often expirations come around. Term length sets the clock.

CompStak data reported by the Seattle Times showed Seattle-area office lease terms averaging 71 months in late 2021 and falling to 58 months by early 2022, with large employers moving from ten-year deals to three-to-five-year terms and smaller Class B and C tenants renewing for as little as one to two years. Terms have stayed structurally shorter since.

Run the arithmetic (ours, not any study's): at the old 66-month average term and a 55% renewal rate, about 8% of tenants face a move decision that ends in relocation each year. At a 58-month term, that rises to roughly 9%. Shorter terms mean the expiration decision, where the 45% non-renewal reality lives, arrives more often for every tenant in the market.

Two things follow from the real renewal number, and most brokerage prospecting math ignores both.

First, renewal versus relocation is close to a coin flip for the tenant sizes most tenant-rep brokers actually represent. The industry's own assumption quietly teaches brokers to treat expirations as formalities, to check in a few months out and expect a renewal. The data says a 10,000 square foot tenant approaching expiration is an open outcome, worth the same structured pursuit as a new requirement. Your existing client base, mapped by expiration date, is the highest-probability pipeline you own. That is why we treat every executed lease as pipeline infrastructure, with the option windows and notice dates extracted into a system that surfaces them 12 to 24 months out (see our take on AI lease abstraction).

Second, nobody sells the dataset that would settle who actually moves. The largest lease microdata study to date, Brueckner and Rosenthal's 2025 analysis of 127,822 CompStak leases, contains no model of renewal by tenure at all, and our own review found no peer-reviewed renewal hazard model anywhere in the CRE literature. The renewal studies that exist are decades old and landlord-side. Which means a broker who systematically tracks outcomes in their own market, who renewed, who moved, at what size, from which buildings, is accumulating evidence that neither CoStar nor the academic literature can supply. At BrokerHQ we think that outcome data, joined to public signals like permits and filings, is the foundation the tenant side has never had. That is the bet our product is built on.

Three real objections deserve airtime.

The vintage problem: both headline studies are 1997 to 2005 data. The office market has been through two structural breaks since, and today's REIT disclosures are too inconsistent to compute one clean modern rate. It is possible the true current number differs, though the post-2020 scatter runs below the old studies more often than above.

The weighting defense: if you underwrite buildings, the square-foot-weighted 76.6% is arguably the right number, and the convention is roughly defensible for institutional Class A rollover models. The myth is only a myth when applied to tenant counts, which is exactly how brokers and mid-sized tenants encounter it.

The conditional-rate caveat: converting renewal rates into annual move probabilities requires assumptions (that decisions cluster at expiration, that mid-term moves are rare). Our per-year figures above are arithmetic conversions, labeled as such, not published findings.

None of the three rescues the 70 to 75% figure for the tenant sizes that fill most brokers' books. Two independent datasets, different landlords, same answer.