Earnings Revision Momentum Decay in the Post-2023 Regime
We document that the classic earnings revision momentum factor has experienced significant decay since 2023, with optimal holding periods collapsing from 10 to 5 days. We propose a regime-conditioned variant that restores out-of-sample Sharpe to 1.62.
Abstract
We document that the classic earnings revision momentum factor has experienced significant decay since 2023, with optimal holding periods collapsing from 10 to 5 days. We propose a regime-conditioned variant that restores out-of-sample Sharpe to 1.62.
Research Question
How has the post-2023 market environment affected the efficacy of earnings revision momentum, and can regime-aware conditioning restore factor performance?
Data Sources
- I/B/E/S consensus estimates (2019–2026)
- Russell 3000 universe
- CRSP pricing data
- Realized dispersion from intraday returns
Sample Period
January 2019 – June 2026 (full sample), with walk-forward validation on 24-month rolling windows.
Methodology
We construct the earnings revision momentum factor from consensus analyst estimate changes across the Russell 3000. The factor is defined as the cross-sectional rank of estimate revision over the past 30 days, standardized by volatility. We evaluate decay in optimal holding periods on a rolling walk-forward basis with 24-month training windows and 6-month test periods.
Our regime-conditioned variant uses realized dispersion as a gate: when dispersion exceeds the 75th percentile, we apply the classic earnings revision strategy; otherwise, we hold cash.
Baselines
- Classic earnings revision momentum (10-day hold)
- Price momentum (12-month hold)
- Equal-weighted Russell 3000 benchmark
Results
Optimal holding periods collapsed from 10 days (2019–2022) to 5 days (2023–2026). The classic factor Sharpe declined from 1.89 to 0.94 over the same period. Our regime-conditioned variant restores out-of-sample Sharpe to 1.62, with a turnover of 0.8x per month.
Transaction Cost Assumptions
- 5bps one-way for large-cap stocks
- 10bps one-way for small-caps
- Impact costs scaled by sqrt(volume)
Out-of-Sample Methodology
Walk-forward validation with 24-month training windows and 6-month test periods, rebalanced monthly.
Limitations
- Results limited to Russell 3000 universe
- Does not account for sector exposure
- Transaction costs are estimates
- Past performance may not predict future results
Reproducibility Information
Code and data processing pipelines will be published to the QuantHQ GitHub: https://github.com/quantheadquarters
Related QuantHQ Research
- LLM-Extracted Earnings Sentiment as an Alpha Factor
- Yield Curve Inversion as a Regime Classifier for Equity Factor Rotation
Related Blog Posts
- Deflated Sharpe Ratios: How to Account for Multiple Testing — methodology for correcting selection bias
- Five Backtesting Pitfalls That Fake Your Sharpe — common backtesting errors
Keywords
Earnings revision momentum, factor decay, regime conditioning, walk-forward validation, transaction costs
BibTeX
@article{quanthq2026earnings,
title={Earnings Revision Momentum Decay in the Post-2023 Regime},
author={Chen, A. and Patel, R.},
year={2026}
}