The Apocalypse Model, Rerun
The office real estate apocalypse model of Gupta, Mittal and Van Nieuwerburgh, rerun on what Manhattan recorded after 2023: 78 percent of 2019 value at the end of 2026, not 47. The rerun tests the thesis of my April 2020 paper, which projected that remote work would cut office demand by about five percent, not the collapse the headlines expected. Also posted on Substack.
Summary. In February 2026 the American Economic Review published “Work from Home and the Office Real Estate Apocalypse” by Arpit Gupta, Vrinda Mittal and Stijn Van Nieuwerburgh, whose headline numbers, New York office values down 46.7 percent from 2019 and $556.8 billion of value destroyed across the United States, have been widely quoted. This article reruns the paper’s model from the authors’ replication package. The model’s one unobservable setting, the persistence parameter p, is estimated from a single year’s return on three REIT stocks and gives the remote-work state an expected length of eight years. Rerun with one change, Manhattan’s recorded leasing and rents for 2024 through mid-2026 in place of the remote-work state’s assumptions, the same model puts the New York office stock at 78 percent of its 2019 value at the end of 2026, against the paper’s 47 percent. The paper’s own tenant evidence suggests why: tenants cut their space once, at the first lease expiry, by an amount set by their office policy, rather than bleeding demand for as many years as a stock price implies. Three appendices show how little it takes to move the estimate of p, give the model’s leasing and rent inputs by market state, and compare the result with the New York City Department of Finance’s valuation of every office building in Manhattan. Sixteen pages.
Suggested citation: Colacino, Michael. “The Apocalypse Model, Rerun.” Haiku Partners, October 2026.