Inverse Volatility

A simple, estimation-light allocation that weights each asset inversely proportional to its standalone volatility. Often used as a robust baseline and as a base estimator inside ensemble methods.

Overview

Inverse Volatility allocates capital so that each asset receives a weight that scales with the inverse of its standard deviation. Unlike mean-variance optimization, it ignores the off-diagonal entries of the covariance matrix and requires no expected-return estimates, which makes it numerically trivial and very stable to noisy inputs.

It is closely related to risk-parity (and is sometimes called "naive risk parity"), but unlike full risk parity it does not equalise risk contributions; it only equalises volatility-scaled exposures. When asset correlations are similar across the universe, the two coincide; when correlations vary, full risk parity differs.

Mathematical Formulation

Let denote the sample standard deviation of asset 's returns over the lookback window. The unnormalised weight is the reciprocal of volatility, and weights are then renormalised to sum to one:

Folio Lab implements this via skfolio's InverseVolatility estimator. Weights are long-only by construction and sum to 1 after normalisation.

Advantages & Limitations

Advantages

Closed form: No optimisation solver, no inversion.
Robust: Depends only on per-asset volatilities, not correlations.
Stable turnover: Weights move slowly with vol.
No return estimates: Sidesteps the hardest part of MVO.

Limitations

Ignores correlations: Two highly correlated assets each receive their full weight.
No view of returns: Cannot tilt toward high-conviction names.
No constraints: Sector caps, turnover bounds are not native.
Concentration in low-vol names: A very stable asset can dominate.

References

Maillard, S., Roncalli, T., & Teïletche, J. (2010). "The Properties of Equally Weighted Risk Contribution Portfolios." The Journal of Portfolio Management, 36(4), 60-70.
skfolio documentation — skfolio.optimization.InverseVolatility.