A company signs a fifteen-year lease on a downtown tower, then discovers that its people work perfectly well from home. The rent is owed either way. Rather than eat the cost of underused space, the temptation is to make the space used, which is to say, to require the people who do not need to be there to come and fill it. The mandate that follows is described as a decision about collaboration. It is often a decision about a lease.
The tell in the survey data
The clue that real estate drives many of these decisions is what companies say when the space is not sunk. In the Stanford and Atlanta Fed survey of more than a thousand US executives, employers who kept hybrid arrangements listed reduced physical space among the benefits they valued, alongside better productivity and lower turnover. When shedding real estate is on the table, executives count it as a gain. When the real estate is locked in by a long lease, the same cost becomes an argument for filling the building. The underlying asset, not the underlying work, is doing the talking.
This is what makes the productivity framing suspect. If a firm genuinely believed presence lifted output, it would want the office regardless of the lease term. The fact that the appetite for mandates tracks the size of the real-estate commitment, rather than any measure of collaboration, suggests the causation runs the other way. The building is not full because the work requires it. The mandate exists because the building is empty and the rent is due.
The sunk-cost trap
Economics is unambiguous about how to treat a lease already signed: it is a sunk cost, and a rational decision looks only at future costs and benefits. The rent will be paid whether or not the desks are occupied, so the question of where people should work should be decided on its own merits, as if the lease did not exist. Mandating attendance to avoid the feeling of waste does not recover the rent. It adds a second cost, the commute tax on every employee, on top of the first, in exchange for the psychological relief of seeing the space used.
That second cost is not small, and it lands on people rather than on the balance sheet where the first cost sits. So the sunk-cost mandate performs a quiet transfer: it converts an accounting embarrassment, empty square footage the company is paying for, into a distributed burden on employees who must now commute to occupy it. The lease looks better justified. The workforce is worse off, and the total cost to the enterprise has gone up, not down.
The office as decoration
There is a version of this where the building stops being a workplace and becomes a display. Semler once served as a judge for a best-employer prize and watched it go to a bank whose tellers were not allowed to sit while the vice-president served kombucha, a workplace optimized for how it looked to a visitor rather than how it worked for the people in it. Semco, by his account, never chased that kind of certification; it kept a hammock in the middle of the office instead. The distinction is exactly the one a sunk-cost mandate blurs. A space can be arranged to be seen, or arranged to be used, and a company filling an expensive lease with mandated attendance is optimizing for the former while claiming the latter.
The honest path is harder and cheaper in the long run. If the lease was too large, the company can sublet, renegotiate, or wait it out, and decide independently where the work is best done. That means admitting the space was overbought, which is uncomfortable, but the discomfort is the point: it keeps the mistake in the accounts where it belongs, rather than exporting it to the workforce as a daily commute justified by a culture argument no one actually ran the numbers on.
Sources
- Bloom, Davis & Barrero / Stanford (2025). Return to office? Not everybody is doing it. Employers who keep hybrid list reduced physical space among the benefits they value, alongside productivity and turnover; only 12% of executives planned an RTO push. https://news.stanford.edu/stories/2025/03/return-to-office-not-everybody-is-doing-it