Abstract: In today’s digital economy, firms continuously collect, store, share, and sell personal data, exposing customers to risks of financial fraud. Leveraging Apple’s App Tracking Transparency policy as a natural experiment, we show that restricting data tracking and sharing significantly reduces consumer fraud complaints, particularly those involving personal information misuse. Effects are stronger in areas dominated by firms with risky data practices and coincide with a decline in dark web discussions and higher prices for sensitive data. By tracing effects along the fraud supply chain, our findings suggest that data regulations can benefit consumers by constraining the flow of exploitable information.
Abstract: Payment technologies pose an economic dilemma: network effects can lead to a small number of dominant platforms, but efforts to increase choice can risk market fragmentation. We examine whether interoperability can help resolve this tension, using data from India’s Unified Payments Interface—the world’s largest fast payment system by volume—as well as from a major pre-existing digital wallet provider. When the two payment platforms became interoperable, overall usage of digital payments rose. This increase was driven by regions where digital payments were more fragmented across platforms ex ante, consistent with a model of payment choice in which interoperability increases network size without requiring that users pool on a single platform. Quantifying our model implies that combining the two platforms’ networks through interoperability increased total usage of digital payments by more than 50% in the year after integration.
Discussant: Jacelly Cespedes, University of Minnesota
Abstract: We investigate the transformative potential of large language models (LLMs) when integrated into robo-advisory platforms, using financial literacy as a contextual setting. As a pilot program in partnership with a major brokerage firm, we enhance a robo-advisor with a back-end LLM (e.g., ChatGPT/DeepSeek) to provide personalized, conversational support to investors making decisions by differentiating between beta and alpha in mutual funds and stocks. Our study examines how the LLM-augmented robo-advisor influences investor behavior and portfolio selection compared with a traditional rule-based chatbot and standard non-premium human assistance. Results show that LLM support significantly increases investor engagement and understanding, narrowing the behavioral gap between automatic one-click enrollment and self-assembled portfolio construction. With continuous, interactive guidance reminiscent of high-touch wealth management services, the LLM-enabled robo-advisor fosters relational trust and encourages ETF adoption, particularly in trust-sensitive markets. Notably, LLM chatbots also generate positive spillovers: investors exposed to the LLM are more likely to pursue international diversification, even when such options are not actively promoted by the robo-advisor but are available. Overall, integrating LLMs into robo-advisors yields substantial benefits, with treated investors achieving monthly returns 40 basis points higher and realizing an average monthly gain of 750 RMB.
