Hurst exponent: measuring whether returns trend, mean-revert, or follow a random walk
The Hurst exponent is a single number that summarises whether a time series shows persistent trends, persistent mean-reversion, or no memory at all. Originally developed by Harold Hurst (1951) to study Nile flood records, it has been adapted to financial returns where it provides a quantitative answer to a question every systematic strategy implicitly asks: do moves predict more moves in the same direction, or in the opposite direction.
What the Hurst exponent is
The Hurst exponent (H) is bounded between 0 and 1. A value of 0.5 indicates a random walk: future moves are independent of past moves. A value above 0.5 indicates persistence—moves in one direction tend to be followed by further moves in the same direction, the structural condition for momentum or trend-following strategies. A value below 0.5 indicates anti-persistence—moves tend to be followed by reversals, the structural condition for mean-reversion strategies.
The exponent is computed via rescaled range (R/S) analysis or one of several modern alternatives (detrended fluctuation analysis, multifractal methods). All versions estimate how the dispersion of cumulative returns scales with the length of the observation window. Under a random walk, dispersion scales with the square root of time (H = 0.5); deviations from this scaling identify persistence or anti-persistence.
How it works
The intuition is geometric. Imagine plotting cumulative returns over different window lengths. Under a random walk, the range of cumulative returns grows proportionally to the square root of the window. If the cumulative return path is unusually smooth (positive moves cluster, negative moves cluster), the range grows faster than the square root—H exceeds 0.5. If the path is unusually jagged (moves alternate aggressively), the range grows slower than the square root—H is below 0.5.
For systematic strategy design, the Hurst exponent is a diagnostic for the choice between trend-following and mean-reversion approaches. A market with H well above 0.5 is a candidate for momentum strategies; a market with H well below 0.5 is a candidate for mean-reversion; a market with H near 0.5 is a poor candidate for either.
What the evidence shows
Empirical studies of equity index returns find Hurst exponents typically near 0.5, with small deviations that are often consistent with the random walk hypothesis at standard significance levels (Lo, 1991). Single-stock returns can exhibit larger deviations, particularly at short horizons.
Commodity and futures markets often show H modestly above 0.5 over multi-month horizons, which is consistent with the documented success of trend-following strategies in those markets (Moskowitz, Ooi & Pedersen, 2012). Currency markets show mixed results across pairs and time periods. The persistence is rarely large enough to overcome transaction costs at high frequencies.
Limitations and trade-offs
The Hurst exponent is a single summary statistic; financial returns rarely conform exactly to a single fractal-scaling regime. A series can show different Hurst exponents at different horizons (multi-fractal behaviour), or different Hurst exponents in different volatility regimes. The single-number summary obscures these structural complications.
Estimation of H from finite samples has substantial sampling error. A measured H of 0.55 with a confidence interval of ±0.10 is statistically indistinguishable from a random walk. Practitioners using H as a strategy diagnostic should be aware that the typical H estimate from a few years of daily data has wide confidence bands.
Hurst exponent in pfolio
The Hurst exponent is not currently displayed in pfolio Insights. The platform's systematic momentum methodology assumes positive serial correlation in the medium term, which the Hurst framework formalises; the underlying return series can be exported to compute Hurst values externally.
Related articles
Disclaimer
Get started now
