Paper: arXiv 2402.05272

Authors: Yizhan Shu, Chenyu Yu, John M. Mulvey

Abstract

This article investigates a regime-switching investment strategy aimed at mitigating downside risk by reducing market exposure during anticipated unfavorable market regimes. We highlight the statistical jump model (JM) for market regime identification, a recently developed robust model that distinguishes itself from traditional Markov-switching models by enhancing regime persistence through a jump penalty applied at each state transition. Our JM utilizes a feature set comprising risk and return measures derived solely from the return series, with the optimal jump penalty selected through a time-series cross-validation method that directly optimizes strategy performance. Our empirical analysis evaluates the realistic out-of-sample performance of various strategies on major equity indices from the US, Germany, and Japan from 1990 to 2023, in the presence of transaction costs and trading delays. The results demonstrate the consistent outperformance of the JM-guided strategy in reducing risk metrics such as volatility and maximum drawdown, and enhancing risk-adjusted returns like the Sharpe ratio, when compared to both hidden Markov model-guided strategy and the buy-and-hold strategy. These findings underline the enhanced persistence, practicality, and versatility of strategies utilizing JMs for regime-switching signals.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper uses advanced statistical models (statistical jump models) and time-series cross-validation, indicating high mathematical complexity, while also conducting a robust out-of-sample backtest with transaction costs and real market data across multiple indices, demonstrating high empirical rigor.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Reduce downside risk via<br>regime-switching signals"] --> B["Key Methodology:<br>Statistical Jump Model<br>vs. HMM vs. Buy & Hold"]
  A --> C["Data/Inputs:<br>US, DE, JP Indices<br>1990-2023<br>Transaction Costs & Delays"]
  B --> D["Computational Process:<br>1. Feature extraction<br>2. Optimize jump penalty<br>3. Simulate trading"]
  C --> D
  D --> E["Key Findings:<br>JM strategy outperforms<br>• Lower Vol & Max Drawdown<br>• Higher Sharpe Ratio<br>• Enhanced Persistence"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a regime‑switching investment strategy that uses a statistical jump model (JM) to identify market regimes and reduce downside risk.

Source sentence
We highlight the statistical jump model (JM) for market regime identification, a recently developed robust model that distinguishes itself from traditional Markov-switching models by enhancing regime persistence through a jump penalty applied at each state transition.

Assumes 6

  • A one‑day trading delay is assumed for executing the regime signal. execution
    Source sentence
    we assume a one-day delay, meaning that if the prevailing regime inferred at the end of day t differs from the regime on which the current allocation is based, then the switch in allocation can only be executed at the end of next day t + 1, or, in other words, the signal applied starting from day t + 2.
  • A one‑way transaction cost of 10 basis points per trade is applied. market
    Source sentence
    A one-way transaction cost of 10 basis points is applied in calculating strategy performance.
  • Return series are assumed i.i.d. for a simplified leverage‑impact analysis. statistical
    Source sentence
    Assuming, for simplicity, no transaction costs and an independent and identically distributed (i.i.d.) process for the return series, if 100 × (1 −l)% of days are randomly selected to avoid market exposure...
  • Training windows span 12 years and validation windows 8 years. model
    Source sentence
    Since our data begin in 1970, with training windows spanning 12 years and validation windows 8 years, the out-of-sample testing period begins in 1990.
Show 2 more
  • The methodology is tested independently on each index rather than on a combined portfolio. data
    Source sentence
    It is important to note that our methodology is tested independently on each index, rather than constructing an international equity portfolio that combines all three indices.
  • The jump penalty λ is selected by maximizing Sharpe ratio on a validation period. model
    Source sentence
    We then select the value ˆλ that yields the highest Sharpe ratio during this validation period and use this value for the following month to generate out-of-sample regime inference post hyperparameter tuning.

Data

Backtest sample: 1990 – 2023 · equities · daily

  • S&P 500 daily total return series · Bloomberg Terminal market-prices commercial
    1970-2023 · US equity index
    Source sentence
    The data analyzed in this article comprises the daily total return series of three major equity indices: S&P 500, DAX, and Nikkei 225
  • DAX daily total return series · Bloomberg Terminal market-prices commercial
    1970-2023 · German equity index
    Source sentence
    The data analyzed in this article comprises the daily total return series of three major equity indices: S&P 500, DAX, and Nikkei 225
  • Nikkei 225 daily total return series · Bloomberg Terminal market-prices commercial
    1970-2023 · Japanese equity index
    Source sentence
    The data analyzed in this article comprises the daily total return series of three major equity indices: S&P 500, DAX, and Nikkei 225
  • 3‑month Treasury Bill Yield (risk‑free rates) · Global Financial Data (GFD) database macro commercial
    1970-2023 · Corresponding country risk‑free rates
    Source sentence
    For the risk‑free rates, we use the 3‑month Treasury Bill Yield from each corresponding country, sourced from the Global Financial Data (GFD) database.

Recognised sources: Bloomberg

Costs

Modeled Costs are part of the reported headline results.

A one‑way transaction cost of 10 basis points per trade is applied to all reported strategy performance.

Source sentence
A one-way transaction cost of 10 basis points is applied in calculating strategy performance.

Finds 4

  • The JM‑guided 0/1 strategy consistently outperforms both the HMM‑guided strategy and buy‑and‑hold in reducing volatility, maximum drawdown, and improving Sharpe and Calmar ratios across S&P 500, DAX, and Nikkei 225. performance
    Source sentence
    JM‑guided strategy consistently outperforms both the hidden Markov model‑guided strategy and the buy‑and‑hold strategy in reducing volatility and maximum drawdowns and enhancing risk‑adjusted returns.
  • Annualized returns of the JM‑guided strategy are higher by roughly 1 % to 4 % versus buy‑and‑hold, with leverage reduced to 80 % for S&P 500, 73 % for DAX, and 68 % for Nikkei 225. performance
    Source sentence
    the JM‑guided strategy improves annualized returns by approximately 1% to 4% over the buy‑and‑hold strategy across different regions.
  • Turnover of the JM strategy is significantly lower than that of the HMM strategy, indicating reduced portfolio rebalancing. performance
    Source sentence
    The benefits of enhanced persistence offered by JMs are further substantiated by the significantly reduced turnover of the JM strategy, which compares favorably w…
  • The JM‑guided strategy remains robust under trading delays up to ten days, with performance degradation much smaller than for the HMM strategy. performance
    Source sentence
    JMs’ inherent persistence lends enhanced robustness against trading delays.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 42 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

https://github.com/Yizhan-Oliver-Shu/jump-models — the authors' own repository

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.