Stock Market Prices Do Not Follow Random Walks: Evidence from a Simple Specification Test
Read the paperopens doi.org in a new tab
What they found
Lo and MacKinlay introduced the variance-ratio test, which checks whether the variance of multi-week returns is proportional to the variance of one-week returns as a random walk requires. Applied to weekly U.S. stock index and portfolio returns from 1962 to 1985, the test strongly rejected the random walk: index returns showed positive autocorrelation over weeks, especially for small-stock portfolios. Importantly, individual stocks showed slight negative autocorrelation, so the index-level positive autocorrelation came from cross-effects between stocks (some stocks reacting to others with a lag), not from each stock trending.
What you can use
- Weekly index returns are somewhat predictable; the random-walk model is rejected statistically.
- Statistical rejection is not the same as a profitable strategy; the predictable component is small relative to costs.
- Small and less liquid stocks react to market-wide news with a lag, which is the source of much short-horizon predictability.
Caveats
Pre-1986 weekly data; much of the autocorrelation has declined as markets became more liquid. Nonsynchronous trading explains part but not all of the result. Technical paper.
Tags: mean-reversion, random-walk, variance-ratio, statistics
Summaries are our own reading of the paper, not the authors' words. Educational only, not advice. Discuss it in Book Club.