Social‑media algorithms fuel BNPL uptake among young consumers in Malaysia, Singapore and Indonesia by stoking FOMO and impulsive purchases; the same mechanisms that broaden access to credit also heighten overspending risks.

Main Finding

Algorithmic features of social commerce platforms—especially behavioral targeting and algorithmic personalization—significantly increase BNPL (Buy-Now-Pay-Later) adoption among digitally immersed Generation Z and younger Millennials in Malaysia, Singapore, and Indonesia. Psychological channels, principally Fear of Missing Out (FOMO) and impulsive buying, mediate much of this effect. Platform-level personalization functions largely as an ambient stimulus that amplifies targeting and emotional responses, producing both increased financial inclusion and elevated overspending risk.

Key Points

• Scope and sample: Quantitative survey of 300 BNPL-exposed respondents (older Gen Z cohort born 1995–2000), stratified across Malaysia, Singapore, and Indonesia (≈100 per country).
• Theoretical framing: Integrates Theory of Planned Behavior (TPB) with Technology Acceptance Model (TAM), Stimulus–Organism–Response (S-O-R) framework, and digital nudging theory to center algorithmic influence as a stimulus that acts through cognitive/emotional organismic states.
• Core hypotheses tested:
  • H1: Algorithmic personalization → positive association with BNPL adoption.
  • H2: FOMO mediates personalization → BNPL.
  • H3: Impulsive buying predicts BNPL adoption.
  • H4: Digital financial literacy moderates behavioral targeting → BNPL.
  • H5: Behavioral targeting → positive association with BNPL usage intent.
• Main empirical findings:
  • Behavioral targeting, FOMO, and impulsive buying show strong direct effects on BNPL adoption intention.
  • Algorithmic personalization operates primarily indirectly—raising perceived relevance and attention and thereby increasing FOMO and susceptibility to targeted ads—rather than only producing a simple direct effect.
  • Digital financial literacy moderates vulnerability to targeting (higher literacy dampens—but does not eliminate—the effect).
  • Cross-country differences: Singapore shows higher adoption rates but stronger regulatory/consumer-protection buffers; Indonesia exhibits greater vulnerability due to lighter regulation and lower digital financial literacy; Malaysia is intermediate.
• Methods summary: Regression analyses to test direct, mediating and moderating relationships; multiple regression for robustness. Survey measures adapted from validated scales (5‑point Likert).

Data & Methods

• Design: Cross-sectional, deductive quantitative study targeting digitally active BNPL users in three Southeast Asian countries.
• Sample: N = 300 (≈100 per country), recruited via online questionnaires distributed through university networks, social media groups, and fintech forums; inclusion required at least one prior BNPL use.
• Measures:
  • Algorithmic Personalization: perceived match of recommendation feeds (e.g., TikTok For You, Shopee recommendations) to preferences and timing.
  • Behavioral Targeting: targeted ads, influencer promotions, BNPL-linked campaigns.
  • Psychological mediators: FOMO, impulsive buying behavior.
  • Moderator: Digital financial literacy (self-reported).
  • Outcome: BNPL adoption intention.
• Analysis: Multiple linear regressions for direct effects; mediation analysis for FOMO; interaction terms to assess moderation by financial literacy; robustness checks with additional multiple regression specifications.
• Limitations acknowledged by authors: cross-sectional and self-reported data, limited causal identification, sample restricted to BNPL-exposed older Gen Z limiting generalizability.

Implications for AI Economics

• Model design: Algorithms should be modeled as endogenous market-shaping technologies, not neutral conduits. In demand and adoption models, include platform personalization intensity and behavioral targeting metrics as covariates or structural components because they materially shift consumer willingness-to-pay and short-term credit uptake.
• Mechanisms and externalities: Algorithmic personalization creates positive inclusion externalities (easier access to credit for underserved users) and negative consumption externalities (elevated impulsive spending and potential overindebtedness). Welfare analysis must weigh these opposing effects.
• Identification challenges: Self-reported exposure conflates ambient personalization and explicit targeting; causal inference requires platform-level or quasi-experimental variation (A/B tests, timing/regulatory shocks, instrumenting for algorithmic intensity).
• Risk pricing & credit policy: Lenders and BNPL providers may need to incorporate algorithmic-exposure signals into risk models (e.g., short-term behavioral signals, repeated impulse purchases) to better predict default risk while avoiding discriminatory automated pricing.
• Regulation & governance: Findings support policy tools in the AI-economics toolkit:
  • Algorithmic transparency and mandatory disclosure of targeted credit offers.
  • Auditability of recommendation and targeting algorithms for consumer finance products.
  • Restrictions or cooling-off requirements for time-limited credit offers tied to algorithmic nudges.
• Consumer protection and human capital: Digital financial literacy programs attenuate—but do not remove—the effect of targeting; cost-effective policy mixes should combine platform governance with targeted literacy interventions to reduce harm without stifling fintech innovation.
• Research agenda for AI economists:
  • Use platform logs or experiment-based data to estimate causal effect sizes of personalization intensity on credit take-up and repayment outcomes.
  • Quantify welfare trade-offs across heterogeneous consumers (by literacy, income, country regulation).
  • Develop structural models incorporating algorithmic recommendation policy parameters to simulate regulatory interventions (e.g., disclosure, throttling of time-limited nudges).
  • Evaluate dynamic effects on credit market equilibrium—e.g., how algorithmic-driven demand alters interest rates, default externalities, and market entry by lenders.
• Policy-relevant warning: Without oversight, AI-driven targeting in commerce platforms can produce market failures (information asymmetries, behavioral exploitation) that necessitate regulatory intervention oriented around algorithmic accountability and consumer risk mitigation.

Limitations noted (for interpretive caution): cross-sectional self-reports, N=300 limiting power for fine-grained heterogeneity, and absence of platform-level behavioral logs or causal identification strategies. Future work should rely on platform data, experiments, and structural modeling to quantify economic magnitudes and welfare impacts.