Skip to content

Hurst Exponent

The Hurst Exponent (\(H\)) is a powerful index used to measure the long-term memory of a time series. Originally developed in hydrology to determine the optimum dam size for the Nile River, it is now extensively used in finance, physics, and environmental science to quantify the predictability and nature of price or physical movements.

1. Mathematical Construction (Rescaled Range Analysis)

To estimate \(H\), we use the Rescaled Range (R/S) Analysis. Given a realization \(y_{t}\) for \(t=1, 2, \dots, n\) with mean \(\bar{y}_{n}\) and variance \(S_{n}^{2}\):

  • Mean Adjusted Partial Sums (\(Z_{t}\)):
    $\(Z_{t} = \sum_{j=1}^t (y_{j} - \bar{y}_{n}), \quad t=1, 2, \dots, n\)$
  • Adjusted Range (\(R_{n}\)):
    $\(R_{n} = \max(Z_{1}, \dots, Z_{n}) - \min(Z_{1}, \dots, Z_{n})\)$
  • Rescaled Adjusted Range: Defined as \(R_{n}/S_{n}\).

2. Estimation of \(H\)

According to Hurst's law, the expected value of the rescaled range scales with time (\(n\)) following a power law:
$\(E\left(\dfrac{R_{n}}{S_{n}}\right) \propto Cn^H \quad \text{as } n \to \infty\)$

By taking the logarithm of both sides, we can estimate \(H\) using a simple linear regression:
$\(\log\left(\dfrac{R_{n}}{S_{n}}\right) = \alpha + H \times \log(n)\)$

3. Properties and Interpretation of \(H\)

The value of the Hurst exponent reveals the "flavor" of the time series memory:

Value of \(H\) Interpretation Market/Process Nature
\(H = 0.5\) Random Walk White noise; i.i.d. observations; no memory.
\(0.5 < H < 1\) Persistence Long-memory; trending; momentum exists.
\(0 < H < 0.5\) Anti-persistence Mean-reverting; "pink noise"; frequent reversals.
\(H \geq 1\) Non-stationary Infinite variance; process does not settle.

4. Applications in Finance

Investors and analysts use the Hurst exponent to gain insights into the nature of price movements:

  • Persistent Trends (\(H > 0.5\)):
    If \(H = 0.7\), the market is in a "trending" state. Price movements are likely to persist in their current direction. This information is critical for trend-following strategies.
  • Random Walk (\(H \approx 0.5\)):
    The market has no observable trend. Future movements are independent of past ones, making it equally likely for the price to increase or decrease.
  • Mean Reversion (\(H < 0.5\)):
    If \(H = 0.3\), the price shows a tendency to reverse. An increase is likely followed by a decrease. This is often used to identify overbought or oversold conditions for mean-reversion trading.