Emotions and Fund Flows: Evidence from Managers' Live Streams
Abstract
We study whether real-time, unscripted communication by fund managers persuades public-market investors. Using live-streamed sessions by Chinese ETF managers, we apply a multimodal ML framework to quantify facial, vocal, and textual affect and summarize them in Affective Delivery Index (ADI). A one-standard-deviation increase in ADI raises next-day (next-week) fund flows by 0.17 pp (0.6 pp), with no predictability for subsequent returns, consistent with persuasion rather than information. Facial expressions are the dominant modality. Effects are stronger in down markets, for ETFs with higher retail ownership, and remain after controlling for contemporaneous/past returns and flows as well as day, ETF-index-week, fund-month, family-week, and fund-day-of-the-week fixed effects. We identify exogenous variation in ADI using instrumental-variables approach. Overall, our results indicate that live manager delivery reallocates capital in the short run and moves trading and prices even absent new information.