Regime Detection¶

Why It Matters¶

Most portfolio theory assumes stable relationships: constant correlations, stable factor premiums, stationary distributions. But real systems have regimes: low volatility vs. high volatility, rising rates vs. falling rates, growth vs. value dominance, peacetime vs. crisis. Applying a single strategy across all regimes guarantees poor performance in some regimes. Detection allows adaptation.

The Key Move¶

Identify potential regime indicators: volatility levels, correlation patterns, economic indicators, policy stances. Monitor these indicators for sustained changes. Define regime-specific strategies: defensive allocations for high-volatility regimes, aggressive for low-volatility, etc. When indicators signal a regime shift (not just noise), shift allocations. Use multiple indicators to avoid false signals. Accept that detection will be imperfect; aim for rough accuracy, not precision.

Classic Application¶

Volatility regimes in equity markets. Low-volatility regimes (VIX < 15) allow leverage and risk concentration. High-volatility regimes (VIX > 30) require defensive positioning, more cash, less concentration. Regime-switching models detect transitions and adjust allocations. Research shows volatility persists (regimes last months to years), making detection valuable despite imperfect timing.

Surprising Application¶

Personal productivity regime detection. High-energy regimes (good sleep, low stress, good health) allow ambitious goals, multiple projects, aggressive timelines. Low-energy regimes (illness, high stress, poor sleep) require defensive strategy: maintain critical commitments only, defer new projects, focus on recovery. Detecting the regime shift prevents over-committing during low-energy periods (which leads to failure cascades) and under-utilizing high-energy periods.

Failure Modes¶

Regime detection is backward-looking—you identify regimes after they've started. False positives are costly—switching strategies unnecessarily incurs transaction costs and whipsaw. Regime definitions are arbitrary—where exactly does "high" volatility start? May not exist—apparent regimes could be random clustering. Self-fulfilling—if everyone detects the same regime, their actions create the regime. Overfitting—too many regimes explains past perfectly but predicts poorly.

Go Deeper¶

Ang & Bekaert, "Regime Switches in Interest Rates" (2002); Kritzman et al., "Regime Shifts: Implications for Dynamic Strategies" (2012)