Low-Volatility Anomaly Strategy

The thesis

For decades, finance theory has insisted that risk and return are positively related: stocks with higher volatility (and higher beta) should compensate investors with higher expected returns. The data say the opposite. Sorting US stocks into vol-quintiles and looking forward, the lowest-vol quintile reliably outperforms the highest-vol quintile on a risk-adjusted basis — and over many periods, on an absolute-return basis as well. The CAPM’s central prediction has failed for half a century.

The mechanism, per Frazzini and Pedersen (2014), is that many real-world investors face leverage constraints. They can’t borrow freely to take a 0.5-beta portfolio and lever it 2x to hit the same return as a 1.0-beta market portfolio. So they reach for return by overweighting high-beta names — pushing their prices up and their forward returns down. The result is a persistent excess return on low-beta, low-vol names available to investors willing to hold them.

Academic basis

Ang, Hodrick, Xing, and Zhang (2006) documented the cross-sectional volatility premium in US data: stocks with the highest idiosyncratic volatility had abysmal subsequent returns, and the relationship was monotonic. Sorting on total volatility produced the same pattern but slightly weaker.

Frazzini and Pedersen (2014) generalized this with their Betting Against Beta (BAB) factor: a market-neutral portfolio that goes long low-beta names (levered up to beta 1) and short high-beta names (levered down to beta 1). The BAB factor delivered a Sharpe ratio of approximately 0.78 in US equities from 1926 to 2012, with similar results in international markets, bonds, currencies, and commodities. It is one of the most consistent factor returns ever documented.

Baker, Bradley, and Wurgler (2011) emphasized the role of institutional benchmarking constraints: a manager benchmarked against the S&P 500 cannot outperform by underweighting beta without taking explicit tracking-error risk. This makes low-vol an underdemanded factor that retail and unconstrained investors can capture.

Asness, Frazzini, Gormsen, and Pedersen (2020) revisited the anomaly post-publication and found that while it has weakened (especially in mega-caps where institutional arbitrage has caught up), it remains robust in mid-caps and in international markets. This is why our implementation restricts to the $2B–$20B market-cap window.

How Alpha Suite implements it

• Curated mid-cap universe — ~80 names spanning utilities (AEP, WEC, ED), REITs (VICI, WPC, EPR), consumer staples (HSY, K, CAG), defensive healthcare (BAX, DGX, MCK), insurance (RGA, AIZ, AFL), and stable industrials (MAS, SNA, GGG). Mid-cap focus is deliberate — the anomaly is well-arbitraged in mega-caps but persists in less-followed names.
• 252-day realized volatility — computed from daily log returns and annualized. Names must rank in the bottom quintile (lowest 20% vol) of the universe to qualify.
• Beta ≤ 0.85 — OLS beta vs SPY computed over the same window. The cap aligns with the Frazzini-Pedersen low-beta cohort. High-vol names with low beta (defensive) and low-vol names with high beta (correlated) both fail the filter.
• Mid-cap window ($2B–$20B) — below $2B adds noise from illiquidity; above $20B the anomaly has been arbitraged down by institutional money. The window has held since the Asness et al. 2020 update.
• Quality filter (positive trailing EPS) — avoids distressed low-vol names that screen as stable because they’re stagnant. A money-losing utility is not the kind of low-vol exposure the academic premium reflects.
• Regime-aware scoring — +8 in DEFENSIVE regime (low-vol bid is strongest in stress), +4 in CAUTIOUS, −6 in RISK_ON (high-beta wins risk-on rallies; low-vol lags on relative basis).
• Quarterly horizon (~63 days) — the academic literature documents the premium over multi-month windows. Shorter horizons add noise; longer horizons add idiosyncratic drift.
• 50% post-publication decay haircut — McLean & Pontiff (2016) found that documented anomalies decay by ~32–58% post-publication. We apply the conservative end of that range to expected take-profit levels.

When it fires

Low-vol is an always-on screen, not an event-driven trigger. The strategy typically produces 5–15 active signals out of the ~80-name universe in any given scan, depending on how the vol distribution shifts. In high-vol regimes (VIX > 25), the universe’s dispersion widens and more names qualify. In compressed-vol regimes (VIX < 14), fewer names sit in the bottom quintile by absolute vol, but the relative ranking is still informative.

Best entry conditions: regime is DEFENSIVE or CAUTIOUS, the candidate has been low-vol-stable for the full year (not a recently-de-volatilized name still mean-reverting from a shock), trailing earnings are positive, and dividend yield is present (a sign of stable cash generation).