The Value of Information: A Puzzle

Battle Count: 63

Highly relevant to quantitative trading. The paper provides a direct, estimable measure of the aggregate value of information in equity markets using order flow data. Key findings (value of information is only 0.04% of market cap, rises during turbulence, higher for large/growth/momentum stocks, spikes on earnings days) directly inform trading strategy design, signal valuation, and the economics of information acquisition. The order flow-based estimator is computationally simple and uses standard high-frequency data. The finding that information value is small relative to fees suggests that most active strategies may not generate sufficient alpha to justify costs, which is critical for strategy evaluation.

The core estimator is remarkably simple: sum the product of 1-minute price changes and signed order flow, then annualize. Requires: (1) high-frequency TAQ data processing, (2) trade-signing algorithm (CLNV/LR/EMO), (3) filtering for stocks above $5, (4) annualization by 252 trading days. The decomposition into price impact and order flow variance requires a simple OLS regression of returns on order flow. Main complexity lies in data acquisition (TAQ is large and requires WRDS access) and proper trade classification. No ML models or complex optimization required.

The paper uses publicly available data sources (NYSE TAQ, CRSP, Compustat) and provides detailed estimation procedures including the discrete approximation formula, trade-signing algorithms (CLNV, LR, EMO), and robustness tests. The core estimator is simple (sum of price changes times signed order flow). However, TAQ data requires subscription access, and the French (2008) replication involves proprietary data sources (CEM Benchmarking, Z.1 Financial Accounts). The paper provides extensive robustness checks across trade-signing methods, observation frequencies, and sample specifications.