Credit-reporting beliefs, affordability-related use and self-reported late payment among U.S. buy-now-pay-later users: a cross-sectional analysis of the 2025 SHED

Technical Summary: Credit-reporting beliefs, affordability-related use and self-reported late payment among U.S. buy-now-pay-later users

Problem and Context
Buy now, pay later (BNPL) credit has integrated short-term borrowing into retail checkout processes, offering payment flexibility but creating overlapping future obligations. While administrative data indicates that many BNPL users hold other unsecured debt and experience financial fragility, the specific drivers of late payments remain complex. Late payments may stem from distinct economic margins: (1) Information, where consumers hold specific beliefs about how BNPL usage affects credit reporting; (2) Usage Exposure, defined by the frequency and purpose of use; and (3) Resource Capacity, where tight budgets prevent payment even when terms are understood.

Existing literature has established that financial constraints are common among BNPL users who report late payments. However, it remains unclear how newly measured, product-specific information variables (specifically beliefs about credit reporting) compare to affordability-related use and contemporaneous financial strain when predicting self-reported late payments among current users. This study addresses this gap using the 2025 Survey of Household Economics and Decisionmaking (SHED).

Methodology
The study utilizes a cross-sectional analysis of the public 2025 SHED dataset, restricted to an analytical domain of 2,004 U.S. adults who reported using BNPL in the preceding year. Of these, 487 (weighted prevalence 25.8%) reported at least one late BNPL payment.

• Outcome Variable: Self-reported "any late BNPL payment" in the preceding year (an extensive-margin, respondent-level outcome).
• Key Independent Variables:
  • Credit-Reporting Beliefs: A hierarchical three-category variable derived from two survey items: whether BNPL "helps build a credit history" and whether "making payments on time helps my credit score." Categories include: (1) Benchmark-consistent (disagreeing with both, consistent with Federal Reserve benchmarks at the time of the survey); (2) Uncertain; and (3) Benchmark-discordant agreement (agreeing with at least one).
  • Resource-Context Indicators: Affordability motive (selecting "Only way I could afford it" for the most recent purchase), groceries-or-medical use, other-purpose breadth (count of non-grocery/medical categories), and self-reported financial strain.
• Statistical Approach: Survey-weighted logistic regressions were employed to estimate odds ratios. To facilitate comparison across variables with different scales, the study primarily reports model-standardized predicted probabilities and absolute risk differences (in percentage points).
• Models:
  • Model 2: Socioeconomic-adjusted principal belief model (adjusting for age, gender, race/ethnicity, education, income, and employment).
  • Model 3: Added conditioning on contemporaneous financial strain.
  • Expanded Model: A common descriptive specification simultaneously including belief categories, socioeconomic covariates, and all four resource-context indicators.
  • Exploratory Models: Interaction terms to test if belief associations vary by strain, and within-model subtraction of risk differences to compare the magnitude of belief vs. resource gradients.

Key Results

1. Belief Distribution: Only 14.5% of users held benchmark-consistent beliefs (rejecting both credit-reporting statements). 31.4% were uncertain, and 54.1% held benchmark-discordant beliefs (agreeing with at least one statement).
2. Belief Association with Late Payment: In the principal socioeconomic-adjusted model, belief categories did not significantly differentiate self-reported late payments. The standardized risk difference for benchmark-discordant agreement relative to the benchmark-consistent category was +0.2 percentage points (95% CI: -5.3 to +5.7). The association remained imprecise even after conditioning on financial strain.
3. Resource-Context Associations: In contrast to beliefs, resource-context indicators showed strong, positive associations with late payment in the expanded model:
   • Affordability Motive: +16.5 percentage points (95% CI: +12.0 to +20.9).
   • Groceries-or-Medical Use: +10.9 percentage points (95% CI: +6.2 to +15.6).
   • Financial Strain: +8.1 percentage points (95% CI: +3.7 to +12.5).
   • Other-Purpose Breadth (3+ vs. 0-1 categories): +8.9 percentage points (95% CI: +2.8 to +15.1).
4. Comparative Magnitude: Exploratory within-model comparisons indicated that the risk differences for affordability motives, groceries-or-medical use, and financial strain were significantly larger than the risk difference for benchmark-discordant agreement.
5. Item-Level Nuance: A secondary analysis of individual survey items revealed that respondents who were "Not sure" about whether BNPL helps build a credit history had a lower standardized probability of late payment (-7.7 percentage points) compared to those who disagreed. However, the author notes this pattern is exploratory and does not establish a protective effect of uncertainty.

Significance and Claims
The paper makes a modest, descriptive contribution to household finance by distinguishing between information consistency and resource context as correlates of late payment.

• Distinct Correlates: The study concludes that survey-benchmark reporting beliefs and the affordability context of BNPL use provide different descriptive information. While resource constraints (affordability motives, strain) are strongly associated with late payments, holding beliefs that contradict the survey-time credit-reporting benchmark does not clearly predict late payment among users who have already adopted BNPL.
• Descriptive Nature: The author explicitly states that the analysis is descriptive, not causal. The study does not identify behavioral mechanisms, transaction-level probabilities, or the effects of regulation. The "affordability motive" refers to the most recent purchase, while the outcome covers any late payment in the preceding year; thus, these are respondent-level cross-sectional contrasts, not transaction-matched causal effects.
• Policy Relevance: The findings suggest that while clear disclosure remains valuable for informed choice, policies aimed at reducing late payments must also address the resource conditions under which BNPL is used. The weak association between belief discordance and late payment implies that simply correcting misconceptions about credit reporting may not be sufficient to mitigate repayment risks for financially constrained households.
• Limitations: The study acknowledges that conditioning on BNPL use may induce selection bias, potentially attenuating or reversing associations. Furthermore, the cross-sectional design prevents establishing temporal ordering between financial strain and late payments. The data lacks transaction-level details (counts, amounts, schedules), preventing the separation of usage exposure from per-payment delinquency probability.

In summary, the paper argues that information consistency and resource-context indicators should not be treated as interchangeable. For current BNPL users, the resource context of use appears to be a more potent descriptive correlate of late payment than specific beliefs about credit reporting.