Financial markets sometimes switch states (regimes), with key investment decision statistics (such as average return and volatility of returns) shifting dramatically for extended intervals. A simple example of financial market regimes is the designation of bull and bear stock market states, estimated (for example) by a broad index being above or below its long-interval simple moving average. What is the big picture on the concepts, estimation and application of regime changes in investing? In their June 2011 paper entitled “Regime Changes and Financial Markets”, Andrew Ang and Allan Timmermann review the basics of modeling regime switches and applying such models to asset allocation decisions. Drawing on prior theoretical and empirical research, they conclude that:
- In general, regime switching models applied to financial markets:
- Often produce outputs that match important changes in regulation or policy and other external events.
- May assume repeated switching among a fixed number of market states (such as bull and bear) or an expanding set of new states.
- Make analysis tractable by capturing such such return behaviors as fat tails, volatility clustering, skewness and time-varying correlations by combining intervals separately described by well-behaved (e.g., normal) return distributions.
- Because of the difficulty of reliably measuring average returns, researchers often focus on level of return volatility as the discriminator of financial market regimes. Other market behaviors that tend to vary across regimes include:
- Strength of mean reversion.
- Degree of return predictability.
- Magnitudes of common anomalies such as the size effect and value premium.
- Optimal asset allocations differ substantially across regimes, and uncertainty about future regime changes can have a large effect on strategic asset allocation.
In summary, empirical evidence supports belief in financial market regimes demarcated most reliably by level of return volatility. Assumptions about regime changes are important to strategic asset allocation.
For a specific example of the type of research the authors survey, see “Quantifying and Exploiting Long (Bull and Bear) Trends”.
Note that:
- Academic financial market regime change models tend to be complex, and (as for inference in general) recognition of a regime change considerably lags its onset.
- Rolling window analysis is more sensitive to regime changes than inception-to-date analysis.
