The 600-699 credit band has lost 13.3 points of mortgage originations since 2005

Since the three score bands have to sum to 100, the 11.6 points the top band gained beyond what the middle band gave up came out of files below 600 ^[3]. And a band that finished at 22.3% after shedding 13.3 points started near 35.6%, which makes the loss about 37% of its 2005 position ^[1][2]. Both numbers are shares of originations, describing the mix of approved borrowers and not the count of loans written in either year ^[5][6].

Adam Staveski, a principal associate with Pew's housing policy initiative, said: "Although borrowers now take on more debt as a share of their income than ever before, they must have a pristine credit history to be approved for a loan" ^[1][7]. Debt service relative to income went up while the history screen went the other way. Staveski wrote that scores reward long credit histories and adequate financial cushions, and that the result is a close correlation between scores and age, income and wealth ^[8]. He named young adults, lower-income families, rural communities, and Black and Hispanic households as the groups the tighter environment hits hardest ^[9].

The rate side is moving against all of them. Freddie Mac put the 30-year fixed at 6.76% on Thursday, the highest since June 2025 and 41 basis points above where it sat a year earlier ^[12][4]. Thomas Ryan of Capital Economics said "our projection that existing sales will average 4.1m over this year as a whole now looks slightly optimistic, with transactions more likely to average closer to 4m, which would be their weakest annual outturn since 1995" ^[15]. That trim is roughly 100,000 transactions, about 2.4% ^[8]. August already printed below the 4 million he now expects for the year ^[7].

Readmitting the middle band would cost something on performance, and the cost is large. The share of delinquent borrowers who go on to default is down by a factor of roughly eleven to fourteen since the early 2000s ^[3][6]. Fortune credits loss-mitigation tools, forbearance, loan modifications and payment deferrals, alongside the underwriting rules themselves ^[4]. It also notes the standards ended boom-era practices such as "liar loans" that required little proof of income ^[17]. Part of the improvement therefore sits in servicing, so loosening the box does not mechanically restore early-2000s losses.

Pew describes who got the loans; Capital Economics attributes the downgrade in its sales forecast to the 10-year Treasury yield ^[1][14]. In my view the credit box is second order for the transaction count this year and first order for the composition of who buys. The test is clean enough to run: if the 30-year comes back toward 6% and the 600-699 share stays near 22%, the screen is binding, and if that share climbs with the rate, price was the constraint all along.