A 1.5% Defect Rate Sounds Small. The Arithmetic Says Otherwise.
The industry's critical defect rate was 1.50% for calendar year 2025, against 1.52% the year before. [1] Presented as a percentage, it reads as a rounding error: ninety-eight and a half loans in every hundred are fine.
Percentages are the wrong unit for this. The economics of loan defects are governed by severity and by asymmetry, and a rate expressed as a percentage discloses neither.
Why the rate is the least informative number available
Three properties make a defect rate misleading on its own.
Severity is concentrated, not average. Most defects cost little: a condition re-cleared, a document re-requested. A small subset become repurchase demands, a demand, not a completed repurchase, at an estimated average cost of $32,288 per loan. [2] A distribution with a long tail is poorly summarised by the frequency of events in it.
The payoff is asymmetric. A correctly originated loan earns its margin. A repurchased loan returns to the balance sheet at par while the collateral and borrower performance have moved, often adversely, which is why repurchases cluster in exactly the conditions where absorbing them is hardest.
The rate is measured on a sample. [4] Post-closing review examines a fraction of production, so the reported rate is an estimate of the population rate. Be careful about what that does and does not imply: a smaller random sample is not biased, it is less precise: the estimate stays centred on the true rate with a wider confidence interval around it. What introduces bias is selection design, not sample size. Two lenders reporting 1.5% may still not be comparable, but the reason is their selection methodology and defect definitions rather than coverage alone.
Working the arithmetic honestly
It is tempting to multiply the defect rate by the repurchase cost and produce a headline exposure figure. That calculation is wrong, and it is worth being explicit about why, because versions of it circulate widely.
Not every critical defect becomes a repurchase demand. Many are cured before delivery, many are never identified by the investor, and many are defects in the compliance sense without triggering a representation-and-warranty breach. The conversion rate from critical defect to repurchase is not published on any consistent basis, and it varies by investor, vintage and market conditions.
So the honest statement is a structure, not a number:
Exposure = (loans originated) × (population defect rate) × (share of defects that become
demands) × (average cost per demand)
Three of those four terms are known or estimable for a given lender. The third is the one nobody publishes, and it is the one that determines the answer.
What this means practically: any vendor presenting you with a precise exposure figure has assumed that term. Ask what they assumed. The assumption is doing more work than the data.
What the composition data does tell you
Where the aggregate rate is uninformative, the movement inside it is not.
Across 2025, Borrower and Mortgage Eligibility defects rose 291.58% and Credit rose 166.13%. In Q4, Legal, Regulatory and Compliance returned as the top category at 24.66%, up roughly 30% and rising for a third consecutive quarter, with Income and Employment at 21.52%. [1]
These categories differ in severity. An eligibility or income defect goes to whether the loan should have been made on those terms, which is the kind that becomes a demand. A documentation defect that does not affect the credit decision generally does not.
So a portfolio whose defect rate is flat while its composition shifts toward eligibility, income and credit has rising expected severity at constant frequency. That change is invisible in the headline number and is the single most useful signal in the published data.
Where the cost actually lands
Set against a production cost of $11,109 per loan in Q3 2025: rising to $16,320 for depository retail originations across the year [3]: a single repurchase at $32,288 consumes the margin on a substantial number of clean loans.
The precise ratio depends on margin rather than cost, and margins vary too much for a general figure to be meaningful. The structural point survives regardless: defect costs are large multiples of per-loan economics, so the loss function is dominated by tail events rather than by average performance.
This has a specific consequence for how quality control should be evaluated. A control function whose loss distribution is tail-dominated should be assessed on its coverage of the tail, not on its average throughput or its cost per file reviewed. Those are the metrics most readily available, and they measure the wrong property.
Three questions that are better than the rate
What proportion of production was actually reviewed? A defect rate without a coverage figure is uninterpretable. If coverage has drifted downward as volume rose, an improving rate may reflect less looking rather than fewer defects.
Is the composition moving toward severity? Rising eligibility, income and credit defects at a constant rate is a deteriorating position reported as a stable one.
What is the conversion assumption? Any exposure estimate rests on the share of defects that become demands. If that number is not stated, the estimate is an opinion with arithmetic attached.
None of these is harder to produce than the defect rate. They are simply not the number the industry standardised on, and the standard number is the one that conceals the most.
Sources
- ACES Quality Management, Mortgage QC Industry Trends Report, Q4 and CY 2025. Critical defect rate 1.50% for CY 2025 against 1.52% for CY 2024. Category figures are shares of critical defects, not defect incidence, and are drawn from the report's sample of reviewed loans.
- STRATMOR Group, Unpacking the drivers and costs of GSE repurchase demands, reported by National Mortgage News. The $32,288 figure is a study estimate of cost per repurchase demand, not per completed repurchase; income and appraisal together account for 57% of demands in that study. A category share does not establish that those demands were preventable.
- Mortgage Bankers Association, Quarterly Mortgage Bankers Performance Report Q3 2025, total loan production expenses per loan ($11,109); and MBA production-channel data for 2025, retail channel, depository institutions ($16,320). Both are total cost to produce one loan, not QC cost. Figures are quoted from MBA's published summaries; confirm against the current release before using them in a business case.
- Fannie Mae Selling Guide, D1-3-01, Lender Post-Closing Quality Control Review Process. Requires both random and discretionary selection; discretionary reviews supplement the random sample rather than replace it.