Less Access Does Not Mean the Same Cut for Everyone

In the U.S. credit-card market, “access” contains at least two separate questions: who gets a chance to receive an offer, and what happens to the terms when an offer continues toward an account. The Federal Reserve study examines both by moving below aggregate mail volume to the borrower and offer levels. That distinction supports a more accurate judgment about tighter standards: a lender can reduce the set of people it reaches, change the contract offered to those it still reaches, or do both, without producing the same numerical change in every card count.

The first margin is who gets reached

The borrower-level analysis asks whether an individual consumer received an offer from a particular lender in a given quarter. It uses the Mintel Comperemedia population of households—both those who received an offer from a bank in the study and those who did not—and relates the outcome to that bank's reported change in lending standards. The model also accounts for the consumer's credit-score bucket, the lender, and the timing of the observation. In effect, it asks whether the same lender's reported policy direction is associated with a different chance of appearing in a prospective borrower's mailbox, after allowing for broad differences in borrower risk and lender behavior.

That is a different question from whether the lender mailed more or fewer pieces of mail in total. Aggregate volume can decline because a bank reaches fewer people, because it sends fewer offers to the same people, or because the composition of the observed audience changes. A borrower-level outcome exposes the selection margin: which prospective customers remain inside the lender's outreach strategy when the bank becomes more cautious. This is the point at which a supply change becomes a distributional change rather than only a market-wide total.

The role of the credit-score control deserves care. A bank is unlikely to treat a prime borrower and a subprime borrower as interchangeable, so a tightening episode could look more severe simply because the observed pool contains more applicants with lower scores. The study separates consumers into prime, near-prime, and subprime buckets and includes those categories in the model. Credit scores in the Mintel data were pulled a few months before the survey, which approximates the information available to lenders when they make solicitations. This timing makes the control more relevant than a score observed after the offer decision.

But a score bucket is not a complete portrait of a borrower. People within the same bucket can differ in income, debt obligations, recent credit activity, and other circumstances not represented by the bucket in this analysis. The study also explains why it uses credit score as the consumer-level control rather than demographic information: the legal and operational context of credit-card solicitation makes score the relevant observed comparison for its purpose. The result is a more disciplined comparison, not a claim that every targeting variable has been measured.

The estimated association is substantial in the model.