Methodological weaknesses in the debate on betting and debt.

The growth of the betting market in Brazil has brought a legitimate concern to the forefront of the debate: its potential effects on household debt. The problem is that much of the analysis supporting this relationship lacks sufficient methodological basis to establish causality.

There is a recurring pattern. Studies use aggregated comparisons over time and attribute complex variations to a single factor, without adequate control for other variables. Without counterfactuals and consistent modeling, the result tends to overestimate effects and oversimplify diagnoses.

The recent episode involving estimates based on Central Bank data illustrates this point. Projections indicating a high percentage of income committed to gambling gained significant attention, but were later downplayed due to doubts about the identification of financial flows and the lack of distinction between transaction types. The case exposes a broader problem: methodological weakness is not limited to a specific study.

This pattern also appears in the study by the CNC (National Confederation of Commerce of Goods, Services and Tourism), which argues for a causal relationship between gambling and default.

The study presents itself as a Differences-in-Differences analysis, but there is no control group. In practice, the model compares Brazil before and after 2023. In this format, any change during the period is taken into account. Between 2023 and 2026, there was an expansion of revolving credit, an increase in credit card interest rates, a rise in bank spreads, and an advance in digital banking. All these factors directly affect default rates and are not isolated in the model.

The study itself mentions variables such as unemployment, credit, and inflation, but these are not included in the regressions. They are treated descriptively, outside the model. Without this control, it is not possible to argue that the estimated effect results from the bets.

The results reinforce this inconsistency. The study points to a greater impact on people with higher education levels and those over 35 years old. These findings are not compatible with the profile of the bettor. Data from the Prizes and Betting Secretariat indicate that more than 74% of users are under 40 years old and predominantly male. The discrepancy suggests a specification problem, not an empirical discovery.

There are still significant doubts about the database. The study states that spending on betting exceeded R$ 30 billion per month in March 2026. However, SPA data indicated that the sector's GGR in 2025 was approximately R$ 37 billion for the entire year. The difference in order of magnitude is significant. If there was an error in the unit or source, the results are compromised.

The study also fails to detail the series used for this variable, which prevents replication. Without transparency regarding the data, there is no way to verify the consistency of the estimates.

This does not mean ignoring the potential effects of gambling. The topic is relevant and requires serious investigation, but strong conclusions depend on a solid methodology. Without it, the risk is producing inaccurate diagnoses and misdirecting public debate.

In a regulatory environment under construction, the quality of evidence is not a technical detail. It is a prerequisite for decisions to be well-founded and have a real effect.

(*) Heloísa Diniz is the Regulatory Director of ABFS (Brazilian Association of Bets and Fantasy Sports) and published the article above in JOTA.