In 2008, the housing crisis was a fraud crisis. Borrowers lied on applications, brokers stuffed files, rating agencies stamped AAA on garbage. By the time the music stopped, the loans had been bad on day one for years, and the market simply discovered it.
Citrini Research's 2028 scenario is the opposite shape. The loans were good on day one. The world changed after. Roughly $13 trillion in U.S. residential mortgages becomes structurally impaired not because anyone defrauded the underwriter, but because the steady high-paying employment those underwriters priced against is no longer steady or, in some cases, employment at all.
What "structurally impaired" means
The phrase is not hyperbole. A mortgage book is impaired when the present value of expected cash flows falls below carrying value. That can happen because rates spike, home prices fall, or borrowers stop paying. In this scenario all three drift the wrong way at once, with the same driver: a labor market where the marginal high-income borrower is either out of work or uncertain enough about the next twelve months to stop servicing debt.
The early signals already echo the data on entry-level work. New computer science graduates sat at 7.0% unemployment — worse than the rate for high school graduates — which tells you the labor market for the borrower demographic mortgage models trust most has weakened at its base. The projection layers on a mid-2027 home-price drop of 11% in San Francisco, 9% in Seattle, 8% in Austin. These are not subprime markets. They are markets where the median first-time buyer used to be a senior engineer at a publicly traded tech company. When that engineer's job is reassigned to a coding agent, the bid on a $1.6 million two-bedroom evaporates without anyone announcing a foreclosure.
By the time jobless claims surge to 487K in Q3 2027 — the highest weekly reading since the early pandemic — the impairment is no longer theoretical. It is in the cash flows. The mortgage payment that used to consume 28% of household income now consumes 0% of it, and the household burns savings while the servicer files for forbearance.
Why the underwriting was rational
The temptation is to look back and say the underwriting was reckless. It was not. A loan written in 2024 to a 31-year-old senior product manager making $240,000, with a 25% down payment and a 760 credit score, is the kind of loan agency models classify as approximately default-proof. Over thirty years, in any historical environment, expected loss on that population rounds to a rounding error.
The problem is that the historical environment did not include a shock in which the senior product manager's role itself disappears at scale. Acemoglu's research, summarized by MIT on automation and inequality, shows automation already explained 50–70% of the rise in U.S. wage inequality from 1980 to 2016. The displacement was real and slow enough that mortgage cohorts could amortize through it. What is different here is the speed: the same role displacement, compressed into quarters rather than decades, concentrated in the borrower demographic mortgage books were most exposed to.
That speed turns a normal labor adjustment into a credit event. A bank can ride out a recession in which 7% of borrowers lose income for nine months. It cannot easily ride out a shift in which 12% of prime borrowers lose the kind of job their loan was underwritten against, with no clear path back to a similar income band.
The political problem nobody is sized for
The 2008 response combined bank recapitalization, modification programs, and central-bank asset purchases. The political logic worked because the harm was visible, the cause identifiable, and the remedy populist-shaped. The 2028 case is harder because the cause is "the loan was fine; the world changed" and the remedy implies a transfer at a scale the federal balance sheet has never run in peacetime.
Citrini's piece floats a "Transition Economy Act" and a "Shared AI Prosperity Act" — proposals to tax compute and rebate to households. Whatever you think of the design, the transfer required to keep $13 trillion of mortgage debt performing is on the same order as the largest fiscal interventions in American history. The math is not obviously possible while federal payroll-tax receipts run 12% below baseline because the workers paying them are no longer working.
The honest question is whether the policy apparatus capable of writing those checks in time exists, or whether impairment is recognized only after the foreclosure wave has already redistributed the housing stock to whoever still has cash. In 2008, the world found out the loans were bad. In 2028, the loans are fine and the world is different. Which problem is harder to fix?