SignalTrace

Regime Shifts: Implications for Dynamic Strategies (corrected)

John Montgomery3 min read

Kritzman Page Turkington 2012

Mark Kritzman, Sébastien Page and David Turkington, “Regime Shifts: Implications for Dynamic Strategies (corrected)” (Financial Analysts Journal 68(3), 22-39, 2012; CFA Institute). Page worked on the article while at State Street Global Markets; Turkington was a managing director at State Street Associates. The paper is a deliberate methodological corrective: the authors note that “many authors have used Markov-switching models to ‘fit’ a dataset and uncover evidence of regimes in sample, but far fewer authors have attempted out-of-sample forecasting”, and they set out to “build regime-dependent investment strategies and backtest their performance out of sample”.

The methodology has a distinctive twist. Rather than modelling regimes directly on asset returns (and rather than assuming an asset-pricing model), the authors build a simple two-state Markov Regime-Switching Model on the drivers of returns — market turbulence, inflation and economic growth — and reallocate assets according to the forecast regime. Each series resolves into a “normal” regime and an “event” regime, the latter characterised by more challenging conditions and higher volatility. The authors argue Markov-switching is preferable to arbitrary threshold partitions because thresholds give false signals: they fail to capture regime persistence and changing volatilities, which the Markov structure is designed to model.

The reported result is that the dynamic, regime-conditioned process outperformed static asset allocation in out-of-sample backtests — “especially for investors who seek to avoid large losses”. That qualifier is the key to grading the paper: like the rest of the practitioner regime literature in this vault, the benefit is framed as downside protection / loss avoidance, not as a pure return premium. This is the regime-as-risk-tool thesis stated by named CFA-charterholder practitioners.

Graded moderate: the paper has genuine strengths the typical regime study lacks — an explicit out-of-sample forecasting design, a named correction to in-sample-fitting practice, and disclosed methodology. But the full results, the precise sample period, and any transaction-cost / turnover treatment sit behind the CFA Institute member paywall, so realistic costs cannot be independently verified; there is no public replication; and the work predates the Deflated Sharpe Ratio and Combinatorial Purged Cross-Validation scrutiny that Marcos López de Prado and David H. Bailey later made standard. It is credible practitioner evidence for regime-based allocation as a risk-management tool, short of strong proof of robust after-cost alpha.

Kritzman Page Turkington 2012 [proposes_model] Markov Regime-Switching Model Kritzman Page Turkington 2012 [tests_strategy] Regime-Based Asset Allocation Kritzman Page Turkington 2012 [supports] Regime Classification Kritzman Page Turkington 2012 [opposes] Overfitting in Quantitative Trading

Connections

Sources

Graph View

Backlinks