Defensive factors prefer lower measured risk
A low-volatility strategy ranks assets by the variability of their recent returns and favors the quieter ones. A low-beta strategy favors assets whose returns have historically moved less than the market. A broader defensive strategy may combine volatility, beta, profitability, balance-sheet strength, and stable business characteristics.
These are related but different signals. A stock can have low standalone volatility but high market beta, or low beta because its returns contain a large idiosyncratic component. The chosen definition determines the portfolio.
Defensive factor investing is also different from simply holding cash or reducing every position. The empirical claim is relative: portfolios of lower-risk assets have sometimes produced returns that were not proportionally lower than those of higher-risk assets. The interesting object is the return per unit of measured risk after construction and costs.
Why might lower-risk assets be underpriced?
If investors cannot or will not use leverage, those seeking high expected returns may overweight high-beta assets. That demand can make high-beta assets expensive and lower-beta assets comparatively attractive. One explanation links that pattern to leverage constraints: investors who cannot scale lower-risk assets may reach for higher-beta assets instead.
Benchmarking can create similar pressure: an active manager may prefer a high-beta stock to holding a lower-beta stock with borrowed exposure. Investor preference for lottery-like payoffs and glamorous stories is another proposed mechanism.
None makes defensive assets safe. Their prices can become crowded, their sector exposures can resemble duration, and their historical volatility can fail to anticipate a sudden shock.
From risk estimate to portfolio
| Component | Variants | What it changes |
|---|---|---|
| Risk signal | Volatility, beta, downside volatility, residual volatility | Meaning of “defensive” |
| Estimation window | Short, long, exponentially weighted | Responsiveness and noise |
| Portfolio | Low-risk rank, inverse-volatility weight, minimum variance | Concentration and covariance dependence |
| Neutralization | Sector, country, beta, size | Which defensive exposures remain |
| Constraints | Name, sector, turnover, liquidity, leverage | Investability and deviation from the pure signal |
| Comparison | Unlevered or volatility matched | Whether financing assumptions enter |
Inverse-volatility weighting is not the same as a low-volatility factor. The former allocates across selected assets according to risk; the latter first selects assets because their risk is low.
A defensive ranking example
Illustrative example. Assume 200 liquid stocks are eligible at month end:
- Estimate each stock’s volatility and market beta from data ending before the rebalance.
- Rank stocks by a declared combination of those measures within sector.
- Equally weight the lowest-risk 20%, subject to a maximum name weight.
- Trade at the next modeled event and hold for one month.
- Compare with the equal-weight eligible universe.
Now create a volatility-matched comparison only as a second step. If the defensive portfolio has 10% annualized volatility and the benchmark 15%, scaling defensive exposure to 1.5 times introduces financing, leverage, margin, and deleveraging assumptions. Its return is no longer a free rescaling of history.
What to measure beyond Sharpe
Report the unlevered portfolio first. Then show:
- market beta and sector exposure;
- concentration and effective number of holdings;
- turnover from changes in estimated risk;
- performance during sudden volatility transitions;
- interest-rate and duration-like sensitivity;
- valuation of defensive versus speculative assets;
- spread and impact costs; and
- any volatility-matched comparison with explicit financing.
Compare equal weight, a simple low-risk rank, inverse-volatility weights, and a constrained minimum-variance portfolio on the same universe. This separates the asset-selection effect from the weighting and covariance model.
Results should also be evaluated in later periods and across markets. A stable historical volatility estimate can become a crowded portfolio input used by many investors at once.
Interpreting defensive performance
An unlevered defensive portfolio may lag during speculative rallies and still meet its design objective. Conversely, strong risk-adjusted history may come from a concentrated utility, consumer-staples, or bond-proxy exposure.
Neutralization can reveal that exposure, but it also changes the strategy. Show the raw and constrained portfolios rather than treating one as the corrected truth.
The low-risk effect is not evidence that expected return increases whenever volatility falls. It is evidence about a specific cross-sectional portfolio construction under a historical sample.
Important failure modes
- Defensive assets become crowded and expensive.
- Slow risk estimates miss a sudden regime change.
- Sector concentration dominates the intended characteristic.
- Volatility-matched results assume unrealistic financing or uninterrupted leverage.
- Turnover rises when short-window estimates reorder the portfolio.
- Low reported volatility comes from stale or illiquid prices.
- The result disappears after realistic costs and constraints.
Further reading
- Frazzini and Pedersen, “Betting Against Beta” (2014) — Connects leverage constraints with a beta-ranked, leverage-adjusted long-short construction. Its financing and volatility-matching assumptions are part of the result, not cosmetic implementation details.
- Ang, Hodrick, Xing, and Zhang, “The Cross-Section of Volatility and Expected Returns” (2006) — Studies idiosyncratic volatility rather than simply total volatility or beta, demonstrating why “low risk” must name the risk measure being ranked.
- Baker, Bradley, and Wurgler, “Benchmarks as Limits to Arbitrage: Understanding the Low-Volatility Anomaly” (2011) — Develops a benchmark-and-leverage-constraint explanation for demand in high-beta assets and documents the historical low-risk pattern under its US sample.
- Novy-Marx and Velikov, “Betting Against Betting Against Beta” (2022) — Re-examines the celebrated low-beta result and shows how construction, financing, shorting, and trading costs materially alter its apparent performance.