Industry: Real Estate Product: ModelRisk Application: Analyzing Real Estate Market Trends Under Uncertainty
Ask a strategist what hybrid work does to downtown Class-A office demand over the next decade and you get a single number — "down about a quarter." That number is the midpoint of an enormous range, and acting on it as if it were certain is how a portfolio gets caught on the wrong side of a structural shift. Modelled in ModelRisk as a logistic adoption curve with uncertain parameters, the ten-year demand outcome for the segment spans a P10 index of 54 to a P90 of 86 (base = 100) — anywhere from a catastrophic 46% collapse to a survivable 14% dip. The probability that demand falls more than 20% is 73%; the probability it falls more than 30% is 39%. The point forecast hides every decision that matters.
This article isolates a single structural driver — the remote-/hybrid-work shift — and asks one question: what is the distribution of the demand shock it imposes on one segment? It is deliberately narrower than a price forecast for a single asset, and narrower than a whole-market cycle model. It is one trend, one segment, one demand curve.
A structural trend does not arrive as a number; it arrives as an S-curve whose shape nobody knows in advance. Three things are genuinely uncertain: how far the trend ultimately goes (what share of work stays hybrid), how fast it gets there, and how hard each unit of adoption bites into physical demand. A deterministic forecast picks one value for each and multiplies them into a single ten-year number. But the decision a landlord or lender faces — repurpose, hold, or write down — depends on the spread, not the midpoint. A 14% dip argues for patience; a 46% collapse argues for conversion to residential. Collapsing those into "down a quarter" throws away the only information that distinguishes the two strategies.
The segment's demand index starts at 100 today. A logistic adoption curve A(t) describes the share of work that becomes hybrid over a 10-year horizon, and demand is eroded in proportion: index(t) = 100 × (1 − elasticity × A(t)). Four uncertain parameters drive it:
The key to a credible tail is a shared behavioural-stickiness factor drawn once per iteration. If hybrid working proves stickier than expected, it both saturates higher (a larger ceiling) and bites harder (a higher elasticity) — the two reinforce. Letting the ceiling and elasticity move independently collapses the tail toward the mean; tying them to one stickiness factor fattens both ends, lifting the year-10 demand standard deviation from 9.5 to 12.7. The imposed dependence checks out: stickiness-to-ceiling +0.51, stickiness-to-elasticity +0.56, and ceiling-to-final-demand -0.90.
The fan chart plots the demand index year by year, with the P10–P90 and P25–P75 bands widening as the adoption curve takes hold and the parameter uncertainty compounds. The deterministic path — built from the mean of every parameter — lands at an index of 71 by year 10. But the band around it is what matters: the P10 is 54, the P50 is 73, and the P90 is 86. The 20%-fall line (index 80) cuts straight through the upper part of the cone, which is the visual statement of the headline: most of the distribution sits below it. A landlord planning against the single deterministic line is planning against one path out of a fan that ranges from "manageable" to "existential."
Collapsing the year-10 outcomes into a single distribution of the fall makes the threshold probabilities explicit.
The mean fall is 29%, the median 27%, with a P50-to-P90 tail reaching a 46% fall in the worst decile and a milder 14% fall in the best. The marked threshold gives the number a strategist can act on: P(demand falls more than 20%) = 73%. This is not a forecast that demand will fall 29%; it is a statement that a fall beyond the 20% pain threshold is the overwhelmingly likely case, while a true collapse beyond 30% is a real but minority risk at 39%.
Sweeping the one parameter that dominates — how much hybrid work ultimately sticks — shows how sensitive the whole conclusion is to it.
Holding the other distributions fixed and sweeping the saturation ceiling from 0.20 to 0.70, the probability of a more-than-20% fall climbs from modest to near-certain. At the base ceiling of 0.43 it is 83%, and the more-than-30% probability is 40%. The ceiling at which a 20% fall becomes more likely than not (50%) is just 0.31 — meaning even a relatively mild long-run hybrid share is enough to put the segment past the pain threshold on a coin-flip basis. This single curve tells the firm where to spend its research budget: nail down the long-run hybrid ceiling and most of the forecast uncertainty resolves.
The tornado confirms the ranking.
Against a P50 index of 73, the saturation ceiling dominates at about ±12 points of spread, with demand elasticity and the shared behavioural-stickiness factor tied at roughly ±8 points. The adoption speed and the S-curve midpoint barely register — only ±1 to ±2 points each. That is the strategically useful result: when the trend plays out hardly matters over a ten-year horizon, but how far it goes and how hard it bites are almost everything. A research dollar spent forecasting the timing is wasted; a dollar spent on the long-run ceiling is decisive.
A structural trend is the hardest thing in real estate to forecast and the most expensive to get wrong. ModelRisk does not pretend to know where hybrid work lands — it quantifies the range of where it could land, and a 73% chance of a more-than-20% demand fall is a far more honest basis for strategy than a single point on a fan.