Decomposing Stock Price Drivers into a Minute-Data Selection Factor
Summary
This research note proposes that stock price changes reflect market forces, firm-specific short-term information, longer-term fundamentals, and noise. It uses daily minute data and cross-lag regressions of returns on changes in trading volume to estimate proxies for these components. Three derived signals measure the stability of delayed volume effects, the relationship between significant short-term price drivers and the regression intercept, and how closely the intercept co-moves with the broader stock universe. The signals are averaged over recent trading days and combined with equal weights into a monthly stock-selection factor.
The report presents historical tests in the Chinese equity market, including rank information coefficients, long-short returns, and index-universe and index-enhancement results. It also reports that the combined factor retained predictive association after controlling for common style and industry effects. These are historical findings from the source, not guarantees of future performance. The authors warn that market regimes and factor behavior can change; the strategy’s results depend on minute-data construction, regression choices, portfolio constraints, and backtest assumptions.
Key ideas
- The method decomposes intraday stock returns using current and lagged changes in minute trading volume.
- Regression statistics and residual behavior serve as proxies for short-term information, other price drivers, and noise.
- Three component signals are calculated from recent daily observations and combined into a monthly selection factor.
- The report describes historical tests across broad-market stocks and several index universes, including tests that control for style and industry effects.
- Historical performance may not persist, and the results depend on data and portfolio-construction choices.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.