Subprime Underwriting and Layered Risk
Why subprime lending isn't really about one risk factor like low credit scores — it's about several weaker risk factors stacking on the same loan at once, and how that layering was mispriced in the run-up to 2008.
Prerequisites: What Securitization Does and Why It Exists
"Subprime" gets used loosely to mean "loans to borrowers with bad credit," but that undersells what actually made the category dangerous. A single weak factor — a low credit score alone, for instance — is a manageable, well-understood risk that lenders have priced for decades. The problem that built up before 2008 was layered risk: several separately weak factors stacked on the same loan simultaneously, each one making the others worse, in a combination the historical default data used to price the loans had never actually seen at scale.
What "layering" looks like on one loan
A single subprime loan in 2005–2006 might combine a low credit score, a high loan-to-value ratio (little or no down payment), limited income documentation ("stated income," effectively self-reported and unverified), and an adjustable rate that resets to a much higher payment after an initial teaser period. Any one of these alone is a known, priceable risk. Together, they compound: the borrower has thin equity cushion (high LTV) and unverified ability to pay (stated income) and a payment shock coming (the rate reset) and weaker credit history to begin with — so a shock that any single factor could absorb on its own (a modest home price dip, a short bout of unemployment) instead triggers default, because there's no slack left anywhere in the loan.
Concretely: a borrower with a 620 credit score and a fixed-rate, fully-documented, 80% LTV loan is a normal subprime credit that performs roughly as historical data predicts. The same borrower with a 95% LTV, stated income, and a 2/28 adjustable rate (fixed for two years, then resetting) is a fundamentally different risk — not five times worse in some additive sense, but a loan whose default probability the models of the time systematically underestimated, because they weren't built on data where all these features occurred together at scale.
What this means in practice
Rating agencies and structurers priced pools of these loans largely using historical default rates from an era when layered-risk loans barely existed, and assumed defaults across different loans in a pool would behave close to independently — both assumptions failed simultaneously when a national home-price decline hit every layered-risk loan's weakest link (the equity cushion) at once, producing correlated defaults far beyond what pool-level models expected.
Subprime risk is rarely one bad factor — it's the layering of several weak factors (low credit score, high loan-to-value, unverified income, a rate reset) on the same loan, and layering compounds default probability in ways that historical, single-factor default data can badly underestimate.
The common mistake is treating "subprime" as synonymous with "low credit score" and assuming a familiar, priceable risk. The 2008 crisis wasn't caused by low scores alone — those had existed for decades — but by the specific, historically unprecedented combination of layered risk factors appearing together across an enormous volume of loans at once.
Further reading
- Financial Crisis Inquiry Commission Report, ch. 8