S&P 500
The S&P 500 is the benchmark capitalisation-weighted index of 500 large-cap US companies — widely treated as the single best gauge of US large-cap equities. It is the most frequently used test market in this vault’s regime-switching cluster, for two structural reasons. First, it has a long, clean daily-return history whose volatility regimes (calm bull markets versus crisis episodes — 1987, the dot-com bust, 2008, the 2020 COVID crash) are economically real and well documented, which is exactly what a two-state hidden Markov model is built to separate. Second, it is the deepest-liquidity equity market in the world: CME Group reports that daily liquidity in the E-mini S&P 500 future (ES) is roughly eight times the combined value of all S&P 500 ETFs, and ES bid-ask spreads are typically a single tick (0.25 index points, USD 12.50). The index is tradeable via the cash market, ES and Micro E-mini (MES) futures, SPX options and large ETFs (SPY, IVV, VOO); ES futures trade nearly 24 hours on CME Globex.
That depth matters for the vault’s central question. Regime-switching timing strategies trade in and out of the market when a model flips its regime call, so their net edge is sensitive to transaction costs. On the S&P 500 those costs are minimal — the 10bp one-way assumption used by Bulla et al. 2010 and Shu Yu and Mulvey 2024 is conservative here — which is why the S&P 500 is where regime strategies are tested under their most favourable cost conditions. Even so, the evidence is sobering: Bulla et al. 2010 found the smallest annualised excess return of all five indices on the S&P 500, just 18.5bp over buy-and-hold after costs — economically trivial. Shu Yu and Mulvey 2024 report the HMM-guided 0/1 strategy returning 8.5% versus buy-and-hold’s 10.2%: the HMM underperformed on return even as it cut the maximum drawdown from -55.2% to -28.9%.
The honest reading is that the S&P 500 is the market where regime-switching evidence is strongest and most clearly bounded. Low costs and a long history make this the cleanest possible testbed; the result is consistent across studies — regime detection is a genuine downside-risk filter (large drawdown reduction, modest Sharpe lift) but not a standalone alpha engine, since after costs it tends to match or slightly lag buy-and-hold on raw return. If a regime model cannot establish a tradeable edge here, where costs are lowest, that is meaningful negative evidence for the approach as an alpha source.
Bulla et al. 2010 [tests_strategy] Regime-Based Asset Allocation Shu Yu and Mulvey 2024 [trades_market] S&P 500 S&P 500 [relates] Transaction Costs and Slippage
Connections
- Hidden Markov Model Regime Detection — trades_market, source: https://arxiv.org/html/2402.05272v3
- Markov Regime-Switching Model — detects_regime, source: https://mpra.ub.uni-muenchen.de/21154/1/MPRA_paper_21154.pdf
- Bulla et al. 2010 — reports_underperformance (smallest excess return, 18.5bp), source: https://mpra.ub.uni-muenchen.de/21154/1/MPRA_paper_21154.pdf
- Shu Yu and Mulvey 2024 — reports_underperformance (HMM 8.5% vs B&H 10.2%), source: https://doi.org/10.1057/s41260-024-00376-x
- Regime-Based Asset Allocation — tests_strategy, source: https://doi.org/10.1057/s41260-024-00376-x
- Transaction Costs and Slippage — relates (low costs favour regime strategies here), source: https://www.cmegroup.com/markets/equities/sp/e-mini-sandp500.html
- Moody and Saffell 2001 — trades_market (RRL S&P 500 / T-Bill allocator, 1970-1994 out-of-sample test), source: https://proceedings.neurips.cc/paper_files/paper/1998/hash/4e6cd95227cb0c280e99a195be5f6615-Abstract.html
- Moody Wu Liao Saffell 1998 — trades_market, 1970-1994, source: https://link.springer.com/chapter/10.1007/978-1-4615-5625-1_10