The Low-Volatility Anomaly

The low-volatility anomaly is the most theory-breaking result in the factor zoo: low-risk stocks have historically delivered equal or higher returns than high-risk stocks, with far less volatility, so on a risk-adjusted basis boring stocks crush exciting ones. This inverts the single most basic prediction of finance, that bearing more risk should pay more, and the resolution is one of the most elegant arguments in asset pricing.

The anomaly

The CAPM predicts a positive, steeply-sloped security market line (SML): expected return rising one-for-one with beta, $\mathbb{E}[r_i]-r_f = \beta_i(\mathbb{E}[r_m]-r_f)$. Empirically the realized SML is far too flat, and by some measures downward-sloping over certain samples. Sort stocks into quintiles by past volatility or beta; the low-risk quintile earns a similar average return to the high-risk quintile but with a fraction of the volatility, so its Sharpe ratio is dramatically higher. Ang, Hodrick, Xing, and Zhang showed the effect holds even for idiosyncratic volatility, which the CAPM says should not be priced at all. The pattern is remarkably robust across countries, asset classes, and decades, a rare property for an anomaly.

Why the SML is flat: leverage constraints

Frazzini and Pedersen's Betting Against Beta (BAB) gives the cleanest explanation, and it does not require anyone to be irrational, only constrained. Many investors (pension funds, mutual funds, most individuals) want more than the market return but cannot use leverage, by mandate, by margin rules, or by aversion to borrowing. To reach for return, a leverage-constrained investor who cannot lever the market must instead overweight high-beta stocks, buying built-in leverage. This crowding bids up high-beta prices and pushes their expected returns down, flattening the SML.

Formally, in their model an investor with leverage constraint faces a shadow cost $\psi$ of the constraint, and the equilibrium expected return is

$\mathbb{E}[r_i] - r_f = \psi + \beta_i\big(\mathbb{E}[r_m]-r_f - \psi\big).$

The intercept is positive ($\psi>0$) and the slope is shallower than the CAPM's, exactly the flat, lifted SML seen in data. The unconstrained arbitrageur's response is the BAB portfolio: go long low-beta stocks levered up to beta 1, short high-beta stocks delevered down to beta 1, making the position market-neutral by construction:

$r^{\text{BAB}}_t = \frac{1}{\beta_L}\big(r_{L,t}-r_f\big) - \frac{1}{\beta_H}\big(r_{H,t}-r_f\big).$

Because low-beta is levered and high-beta is shorted, BAB profits from the SML being too flat and earns a large, robust historical premium.

Complementary explanations

Leverage constraints are the headline, but several forces push the same way:

• Benchmarking as a limit to arbitrage (Baker-Bradley-Wurgler). Managers judged against a benchmark with no leverage allowance will not hold levered low-vol positions because tracking-error risk is career risk, so the anomaly is not arbitraged away.
• The lottery preference. Behavioral investors overpay for high-volatility, positively-skewed "lottery" stocks (the chance of a moonshot), depressing their returns, a demand-side mirror of the constraint story.
• Agency and the option-like payoff to managers. Compensation convexity encourages taking volatility.

All of these are frictional/behavioral, this is an anomaly, not a risk premium, which raises the Alpha Decay question of whether publicity will erode it.

Worked example: the Sharpe uplift

Suppose low-vol stocks return 8% at 12% volatility (Sharpe 0.67) and high-vol stocks return 9% at 30% volatility (Sharpe 0.30), with $r_f=0$. The high-vol stocks earn slightly more raw return, consistent with a naive risk-return view, but their Sharpe is less than half. A BAB investor levers the low-vol book ~2.5× to match the high-vol book's volatility: $2.5 \times 8\% = 20\%$ expected return at 30% vol, versus 9% for high-vol. The levered low-risk portfolio dominates. This is the anomaly in one calculation: the reward per unit of risk falls as risk rises, so leverage applied to safe assets beats owning risky ones.

Failure modes

• It needs leverage or shorting to harvest. Long-only low-vol funds capture only part of the effect and become expensive/crowded; the full BAB premium requires levering low-beta, which reintroduces funding risk, the very constraint the anomaly exploits.
• Interest-rate and sector bets. Low-vol portfolios tilt toward bond-proxy sectors (utilities, staples); much of a naive low-vol book is a duration bet that suffers when rates rise.
• Crowding and valuation. After the 2010s low-vol boom, "min-vol" ETFs pushed low-vol valuations rich, compressing forward returns, the anomaly can be temporarily arbitraged into a value-trap.
• Deleveraging spirals. BAB is short high-beta and long levered low-beta; in a funding crunch the constrained investors dump high-beta while arbitrageurs are forced to delever, so BAB itself can crash, it is short the same liquidity factor as Carry.

In interviews

State the anomaly against the CAPM: the realized security market line is too flat, so low-risk stocks have higher Sharpe ratios than high-risk stocks. The explanation to have ready is leverage constraints / betting against beta, investors who cannot lever bid up high-beta stocks, flattening the SML, and the fix is a market-neutral portfolio long levered low-beta and short delevered high-beta. Add the benchmarking limit-to-arbitrage and lottery-preference stories as reinforcement. A sharp closer: this is an anomaly (frictional), not a risk premium, so it is a candidate for crowding and decay, and naive implementations smuggle in a large interest-rate bet. See The Quality Factor, with which low-vol heavily overlaps.