Definition
Volatility clustering is the empirical tendency for large absolute returns to be followed by more large absolute returns and small absolute returns by more small absolute returns. The sign of returns can remain difficult to predict even when their magnitude is persistent. It motivates conditional-volatility models such as GARCH.
How it is calculated
For some positive lag k, the correlation between absolute return at t and absolute return k periods earlier is positive.
| Symbol | Meaning | Unit |
|---|---|---|
| Corr | correlation operator | dimensionless |
| r | return | return per period |
| k | lag | periods |
Interpretation
Markets can move through calm stretches and turbulent stretches. Knowing that the market is currently turbulent can help describe uncertainty even when the next direction remains unclear. The pattern is about magnitudes, not a promise of a particular sign.
Assumptions and conventions
-
The statement concerns a defined return series and sampling frequency.
-
The absolute-return variables have finite positive variances.
-
The correlation is an empirical descriptor, not a universal law.
Correlation is dimensionless; returns are measured per period.
Limitations and common misreadings
-
Clustering does not mean positive returns cluster.
-
A volatility forecast is not a directional forecast.
References
- R. Cont. (2001). Empirical properties of asset returns: stylized facts and statistical issues. Quantitative Finance. Stylized-facts discussion of volatility clustering; discussion of dependence in absolute and squared returns. doi:10.1080/713665670. Source

