The Shadow Price of “Public” Information

Converting public information into stock picks is unlikely to be costless, yet the magnitude of public information processing costs is unknown. We quantify extra-marginal information costs for mutual fund managers by building an “AI analyst” that, using only public data, selectively improves each fund’s portfolio each quarter. The AI analyst’s incremental trading gains estimate the marginal value of hiring additional analysts and technology and, per standard theory, provide a lower-bound estimate of a manager’s marginal information costs. Our AI analyst generates $17.1 million in incremental quarterly gains through 2020, indicating that managers faced at least $17.1 million in marginal costs. These estimated costs are large relative to the average fund’s fees of just $3.6 million and alpha of $2.8 million. Moreover, the AI analyst reduces portfolio risk, outperforms 93 percent of managers over their lifetimes, and dominates managers across a broad range alternative performance benchmarks. These findings demonstrate that public information frictions are economically large and challenge the standard assumption of costless learning from public information.