What the Insper and Stone study on betting and delinquency does not answer

Folha de S.Paulo published a new study, based on Stone data, that links online betting to higher delinquency rates in Brazil. The numbers are interesting, but for PSPs, acquirers, and banks the real question is whether the sample is representative and whether the observed correlation can actually be treated as cause and effect.

1. The study, conducted by Sérgio Firpo, a professor at Insper, together with Stone researchers, cross-referenced banking and Pix transaction data from millions of customers. It compared people who started betting after January 2025, the month Brazil regulated bets, with people who never bet.
2. The headline figures are clear enough: the share of bettors in delinquency was, on average, 20% higher in the 12 months after the first bet, compared with the increase observed among non-bettors; the share of the credit card bill in arrears rose 38.7%; the credit limit made available by banks fell 16%; and one in ten bettors spends more than 20% of everything that leaves the account each month on bets.
3. The first catch is representativeness. Stone’s customer base is disproportionately made up of self-employed workers, MEIs (individual microentrepreneurs), and small business owners — a group with more irregular cash flow and a different credit-risk profile from salaried workers or the average consumer. Turning that into “Brazilian bettors” as a whole would require a bridge the report does not show. Without knowing how this sample compares with the universe of bettors identified by the Banco Central, there is no way to judge representativeness.
4. The second catch is causal direction. A before-and-after comparison around the first bet does not rule out the possibility that people whose finances were already deteriorating were more likely to start betting — either because they were looking for a quick fix for debt or because they were exposed to easy-credit marketing at the same stage of life. A pre-treatment trend test, checking whether both groups were moving in parallel before the first bet, would be the minimum standard for a causal reading, and there is no indication it was reported.
5. The drop in credit limits may also be a scoring artifact rather than evidence of a real deterioration in repayment capacity. Banco Central president Gabriel Galípolo has already said in Senate testimony that betting history entered the credit score used by banks for credit decisions. In that case, the lower limit can reflect automated risk scoring reacting to betting behavior, not necessarily a fresh change in the borrower’s underlying ability to pay.