Model structure
The model runs in quarterly steps over 10 years. Each scenario specifies how many of Seattle's ~700K knowledge-work jobs
(software, professional & business services, finance, information, corporate management) are eliminated by AI over a 5-year period (S-curve),
and what fraction are replaced by new AI-economy jobs.
Labor market: Each lost knowledge job also destroys local service jobs (the "local multiplier" — Moretti estimates each tech job supports
up to 5 local jobs; the default here is a conservative 1.5). Displaced workers either find new local work, stay unemployed, or leave the metro.
Out-migration removes both labor force and housing demand. In addition, Seattle’s normal net in-migration (~0.8%/yr —
the engine of its housing-demand growth) collapses when local jobs disappear.
Housing fundamental value: driven by metro aggregate wage income (elasticity ~0.7), population (elasticity ~1.0), and housing supply
(construction nearly halts when prices fall, which cushions declines).
Price stickiness (the key mechanism): market prices do
not jump to fundamental value. Sellers anchor to recent prices and resist nominal cuts
(loss aversion / downward nominal rigidity, well documented in housing research). In the model:
| Price vs. fundamental gap | Behavior |
|---|
| Price below fundamental | Prices rise toward fundamental fairly quickly (~40%/yr of gap closed) |
| Price 0–10% above fundamental | Nominal prices stay flat or drift up; real prices erode via inflation (the dot-com pattern) |
| Price 10–15% above fundamental | Slow nominal declines (~20%/yr of gap closed) |
| Price >15% above fundamental | "Capitulation": forced sales (job losers and leavers must sell), declines accelerate to ~2× speed |
Calibration against history
| Episode | History | Model behavior |
|---|
Dot-com bust (2001–03)
Seattle U: ~4% → ~7% |
Nominal home prices stayed flat-to-rising; real prices eroded a few % |
Mild scenario reproduces this: nominal +6%, real −6% over 5 yrs ✓ |
Global financial crisis (2007–12)
Seattle U: 3.8% → 10.2% |
Nominal prices −31% peak (Jul 2007) to trough (early 2012); prices kept falling ~2 yrs after unemployment peaked |
Model reproduces the long lag and produces −15% from the labor shock alone. The remaining historical decline reflects the pre-2007 credit bubble
(Seattle was ~25% overvalued) and mortgage-credit collapse, which this model does not assume. If a severe AI shock also froze credit, declines would be deeper than modeled. |
| Detroit (1950s–2010s) |
Sustained loss of the dominant industry: population −60%, real prices −70%+ over decades |
The "full wipeout" scenario is Seattle-as-Detroit-in-fast-forward: population −39%, real prices −60% in 10 yrs |
Key limitations
• No interest-rate / Fed channel: a deep recession would bring big rate cuts (cushioning prices) but possibly also deflation (worsening nominal declines). Both omitted; they partially offset.
• No credit-market channel: a severe shock would likely tighten mortgage lending and deepen declines beyond the model.
• Wealth effects ambiguous: AI success could crash knowledge workers' RSU wealth (negative for housing) or boost it for those who remain (positive); not modeled.
• Metro-level averages: Eastside tech enclaves (Bellevue, Redmond) would fall harder than the metro average; lower-priced submarkets less.
• Policy response (UBI, stimulus, mortgage forbearance) not modeled; would cushion declines.
• Property-tax/fiscal doom loops in severe scenarios not modeled; would worsen declines.
Data sources (retrieved May 2026)
• Employment: BLS Seattle Area Economic Summary (Jan 2026) — total nonfarm 2,106.3K; Information 131.7K; Professional & business services 374.3K; Financial activities 98.3K
• Tech workforce: CBRE Tech Talent report — ~287.6K tech workers, avg wage $186.6K
• Home values: Zillow ZHVI Seattle ~$848K (Feb 2026); Redfin median sale price $861K (Apr 2026), −2.9% y/y
• Unemployment: BLS — Seattle MSA 5.0% (Dec 2025), up from ~4% after 2024–25 tech layoffs
• Elasticities: academic literature on unemployment–house price links (long-run elasticity −0.17 to −0.2; thick-market effects ~3.5%/pp);
downward nominal house-price rigidity (Norges Bank, 3 centuries of data); Moretti local multipliers