Big Data and Noise Trading: Evidence from a Natural Experiment
40 Pages Posted: 13 Aug 2019 Last revised: 31 Jan 2020
Date Written: November 11, 2019
This study empirically investigates two competing plausible effects of big data, when available to retail investors, on noise trading: improved price informativeness and, thus, reduced noise trading vs. misestimating the trading risk and, thus, increased noise trading. Recent advances in technologies democratized the access to big data and lowered information acquisition costs. Theoretical literature supports both improved information efficiency and crowding out information leading to increased speculation as an outcome of this development. Our identification strategy exploits the abrupt shutdown of Yahoo! Finance Application Programming Interface (API), which has been historically a critical source of financial big data for retail investors. We analyze three well documented symptoms of noise trading – trading volume, market liquidity, and future returns – in a difference-in-difference (DID) framework focused on retail and institutional target firms. Our results unanimously point to an increase in noise trading with the availability of big data through Yahoo API, while institutional trades remain unaffected. Particularly, within one month after the API shutdown, for firms with below-median institutional holding relative to firms with above-median institutional holding, retail trading volumes dropped by 8%, market liquidity deteriorated, and retail trades as a whole became more predictive of future returns. The findings suggest big data alone can be harmful to retail investors without sufficient financial literacy.
Keywords: Big data, retail investors, noise trading, application programming interface, natural experiment
JEL Classification: G12, G14, O33, H41
Suggested Citation: Suggested Citation