Financial Turbulence Index
The financial turbulence index is a measure of statistical “unusualness” in asset returns, introduced to portfolio management by Mark Kritzman and Yuanzhen Li in Kritzman and Li 2010 (building on Chow, Jacquier, Kritzman & Lowry 1999). Formally it is a normalised, squared Mahalanobis distance: for n assets, d_t = (1/n)(r_t − μ)ᵀ Σ⁻¹ (r_t − μ), where μ and Σ are the historical mean vector and covariance matrix of returns and r_t is the current return vector. It captures both the magnitude of returns and how they interact — a day on which assets move by unusual amounts and break their usual correlation structure scores as highly turbulent. The same Mahalanobis distance was originally devised to classify human skulls, which is why Kritzman & Li’s paper is titled “Skulls, Financial Turbulence, and Risk Management”.
The turbulence index is best understood as a regime / outlier risk indicator, not a model and not an alpha signal. It does not predict the direction of prices; it flags when the market is in a statistically abnormal state. Empirically it spikes during identifiable crises (1987 crash, Gulf War, Global Financial Crisis, the 2020 COVID crash, the 2022 invasion of Ukraine) and is persistent — it stays elevated for a couple of weeks after an initial spike, which is what makes it usable as a state signal. It is closely related to Regime Classification: thresholding the index partitions history into “calm” and “turbulent” regimes, and it is the turbulence variable that Kritzman Page Turkington 2012 feed into a Markov-switching model.
Its established use is risk scaling: reduce market exposure when turbulence is high, restore it when turbulence subsides. Because elevated-turbulence periods tend to coincide with lower-than-usual returns, exposure scaling improves risk-adjusted performance. A look-ahead-free reproduction (rolling-window estimation, SPY/cash scaling, 2009-2022) reports a Sharpe of ~2.2 versus ~1 for buy-and-hold and a maximum weekly drawdown of ~6% versus ~32%, at ~50% average exposure — but those numbers are gross of transaction costs, and the approach structurally underperforms in violent bull recoveries. Consistent with the vault’s Regime Classification verdict, the turbulence index is a confirmed, replicable de-risking instrument, not a source of standalone tradeable alpha.
Financial Turbulence Index [supports] Regime Classification Kritzman and Li 2010 [proposes_model] Financial Turbulence Index Financial Turbulence Index [relates] Volatility Regime Financial Turbulence Index [part-of] Regime-Based Asset Allocation
Connections
- Kritzman and Li 2010 — proposes_model, source: https://www.tandfonline.com/doi/abs/10.2469/faj.v66.n5.3
- Mark Kritzman — proposes_model, source: https://portfoliooptimizer.io/blog/the-turbulence-index-measuring-financial-risk/
- State Street Associates — relates, source: https://portfoliooptimizer.io/blog/the-turbulence-index-measuring-financial-risk/
- Kritzman Page Turkington 2012 — uses_dataset, source: https://rpc.cfainstitute.org/research/financial-analysts-journal/2012/regime-shifts-implications-for-dynamic-strategies-corrected
- Regime Classification — detects_regime, source: https://portfoliooptimizer.io/blog/the-turbulence-index-measuring-financial-risk/
- Volatility Regime — relates, source: https://portfoliooptimizer.io/blog/the-turbulence-index-measuring-financial-risk/
- Regime-Based Asset Allocation — relates, source: https://portfoliooptimizer.io/blog/the-turbulence-index-measuring-financial-risk/
Sources
- Kritzman, M. & Li, Y. (2010). “Skulls, Financial Turbulence, and Risk Management.” Financial Analysts Journal 66(5), 30-41. https://www.tandfonline.com/doi/abs/10.2469/faj.v66.n5.3
- Portfolio Optimizer — “The Turbulence Index: Measuring Financial Risk.” https://portfoliooptimizer.io/blog/the-turbulence-index-measuring-financial-risk/