What a static pool is

A static pool is a group of loans originated in one period, usually a month, a quarter or a year, and followed on its own for life. No loan joins it later, and its losses are always divided by what that group originally lent. The SEC and the Mexican rating agencies both ask for 5 years of them.

A static pool fixes the denominator, and that is the whole reason a fund asks for it

A portfolio delinquency ratio divides today's arrears by today's balance. When a lender grows, the newest loans make up most of that balance and have not been alive long enough to default, so the ratio falls while nothing about the credit has improved. A static pool divides by a number that never moves: the principal originated in that period. Growth cannot dilute it, and a bad year of underwriting cannot hide inside a good year of volume.

The rest follows from that choice. Each pool is tracked by age rather than by calendar date, so month twelve of the 2023 loans is compared with month twelve of the 2024 loans, and an underwriting decision sits in the same row as its consequence. The measures tracked are cumulative: defaults, net losses after recoveries, prepayments and delinquency buckets, each as a share of original principal, month by month since origination.

Four names describe the same object. Static pool is the regulator's word, from Item 1105 of the SEC's Regulation AB. Vintage is the market's, cohort is the analyst's, and cosecha is the one Mexican rating agencies use. A fifth use of the words is a different thing: the «static pool method» in CECL accounting is a way to size a loss reserve, not a disclosure a fund reads.

The rules agree on five years, same-period origination and life-of-pool tracking

The SEC definition is the one every other document leans on. Under 17 CFR 229.1105, as read in the eCFR on 28 September 2026, an issuer of asset-backed securities presents delinquencies, cumulative losses and prepayments for each prior pool over its life, for five years or for as long as the sponsor has existed, with the latest data no older than 135 days at first use of the prospectus. A sponsor with under three years of securitizing may instead show its own originations by «vintage origination year», which the rule defines as «assets originated during the same year».

The Mexican specification is narrower and more demanding. HR Ratings' methodology for financial assets, October 2022, asks for cohort analysis of the whole portfolio, monthly, by annual origination, for the last five years and by product. Fitch's Mexican methodology for consumer ABS, 26 November 2024, defines a static portfolio as assets generated in one month, quarter or year, expects at least five years of history and ideally a full economic cycle, and sets a floor of 1% on the base-case default rate unless substantial history supports a lower one.

An agency turns the static pools into one number, and it uses only the mature ones

HR Ratings calls the number the historical default rate, TIH. It takes the three most recent mature cohorts, those whose cumulative default curve has flattened, and averages their final default rate weighted by what each one originated. The rating then compares that rate with the maximum default the structure can absorb. The static pools are not a supporting schedule in this process; they are the denominator of the rating.

The table below is an illustrative book, not a real one. Five annual cohorts, cumulative default above ninety days as a share of original principal, read by months since origination at a 31 May 2026 cut-off, so the young cohorts have short rows. The three mature cohorts give a TIH of (180 × 5.8% + 240 × 6.6% + 310 × 5.1%) / 730 = 5.8%.

The example is invented, the effect is not. In the performing Mexican non-bank lender shown in The eleven dark months, a lender that did not default, the worst origination year ran 3.9 times the best. A single blended delinquency figure cannot tell a buyer which of those years it is being shown, and a static pool cannot avoid telling it.

A short history is not neutral, because the agency fills the gap with its own worse curve

The methodologies are explicit about what happens when a lender cannot show three mature cohorts. HR Ratings may use a single curve, the one with the greatest deterioration, and it may project an immature cohort forward when that cohort is running above the older ones. Fitch may rate on stressed assumptions or cap the rating, and it will not go below its 1% floor without substantial data. Either way, the loss assumption that prices the facility stops being the lender's own.

That is why the static pool is the block of a data request that decides the price. The first request a fund sends asks for it right after the loan tape, and for the same reason: both let the reader rebuild a number rather than accept it.

A lender builds its static pools from the tapes, monthly, and never restates them backwards

Static pools are not a separate data collection. They are computed from two files a lender already has to produce: the loan tape, which fixes each loan's origination date and original principal, and the payment tape, which dates every payment and every missed one. Group the loans by month of origination and by product, then, for each month of age, add up the principal that crossed ninety days past due, the principal written off and the amounts recovered, and divide each by the cohort's original principal.

Three rules keep the result credible. Define default once, in days past due, and state it, because the methodologies accept thresholds from 30 to 180 days and a reader compares like with like. Keep restructured loans in the cohort they were born in, because moving them to a new cohort erases the history the pool exists to show. And never recompute old cohorts when a policy changes: a static pool that is revised backwards is no longer static, and a fund that finds one revision assumes there were others.

Effect is not a bank, a lender, a broker-dealer, an investment adviser or a credit rating agency. Nothing here is investment, legal or tax advice. The cohort table is illustrative and describes no real lender. Every rule carries the version of the document it was read from.