What Is Low-Volatility Investing?
Low-volatility investing selects the least volatile stocks in a market on the argument that they have historically delivered returns comparable to or better than riskier ones. Standard theory says higher risk should earn higher return, so the persistence of this result is called the low-volatility anomaly.
Finance theory says you are paid for bearing risk. Low-volatility investing is built on decades of data suggesting that, within the stock market, you often are not — and that claim deserves examining rather than accepting.
How it works
Rank stocks by how much they move and buy the quiet end. Some versions use a minimum-variance optimisation that also accounts for correlations between holdings.
Standard theory predicts this should underperform. Lower risk, lower expected return — that relationship is the foundation of how assets are supposed to be priced.
The measured record across many markets and long periods frequently disagrees. Low-volatility portfolios have often matched or beaten the market with materially smaller drawdowns.
The explanations, none of them settled
Leverage constraints. Many institutions cannot borrow, so an investor wanting higher returns buys riskier stocks instead of leveraging safe ones. That bids up volatile stocks and leaves calm ones cheap.
Lottery preference. Individual investors systematically overpay for the chance of a very large gain, which is concentrated in the most volatile names.
Benchmark pressure. A manager judged against an index cannot hold a very different portfolio for long, so the low-volatility position is hard to maintain professionally even when it works.
And possibly measurement. Low-volatility portfolios load heavily on particular sectors and on value-like characteristics, so part of the result may be those exposures wearing a different label.
A worked example
Take this site’s shared series. The median bar range is 0.493, the ninetieth percentile 1.101, and the largest single bar 2.338.
A low-volatility screen selects for the left of that distribution — holdings whose typical bar sits below the median.
But volatility clusters, which volatility clustering covers. A stock that has been calm is more likely to stay calm — and the screen is therefore selecting on a property that genuinely persists, which is the mechanism that makes the strategy implementable at all.
The same clustering is why it fails when it fails. A regime change moves the whole distribution, and a portfolio selected for calm in the old regime is not calm in the new one.
Crowding is the live problem
The strategy became popular. Once a documented anomaly attracts money, the stocks it selects are bid up, and a higher price is a lower future return.
So the historical record was produced under conditions that no longer exist. The measurement period predates the products built on it, which is a general problem with every published factor and an acute one here.
And it does not protect against everything. Low-volatility stocks fall in a systemic event like everything else — systemic risk is not a risk this screen removes.
The original data
On this site’s shared series: median bar range 0.493, ninetieth percentile 1.101, largest bar 2.338. Direction runs average 2.01 bars with a longest of 11. A round trip costs 0.0098, about 2% of the median bar range.
And this site’s thirty-year fee measurement: 5 basis points costs 1.5% of the final balance, 20 costs 5.8%, 75 costs 20.2%, 150 costs 36.5%.
That fee table is the practical test. A low-volatility fund charging 40 basis points against a broad tracker at 7 needs the anomaly to still exist, net, after crowding — and the fee is certain while the anomaly is contested.
How the two main versions differ
Ranked low volatility simply buys the calmest stocks by their own individual movement. It is transparent, easy to explain, and ignores how the holdings interact.
Minimum variance optimises the whole portfolio, accounting for correlations — so it may include a more volatile stock that moves opposite to the others, because the combination is calmer than either approach alone would produce.
The second is theoretically better and practically fragile. Optimisation requires estimating the correlations between every pair of holdings, those estimates come from past data, and they are least reliable in exactly the conditions where the portfolio’s behaviour matters most.
Which is the recurring pattern in this material. The sophisticated version depends on a measurement that breaks under stress, and the crude version depends on a property that persists. Given the choice, the simpler screen is usually the more honest one — it claims less and therefore has less to be wrong about.
What the measurement window decides
Volatility is measured over a chosen lookback — commonly one, two or three years of past movement. That choice is not neutral, and two providers using different windows produce different portfolios from the same universe.
A short window reacts fast and churns. Holdings enter and leave as their recent movement changes, and every change pays a spread. On this site’s series a round trip costs 0.0098, about 2% of the median bar range of 0.493 — small once, meaningful at high turnover.
A long window is stable and stale. It keeps holding a stock whose character has already changed, because the calm years still dominate the average.
Neither is correct, and the choice is rarely disclosed prominently. It sits in the index methodology document rather than the factsheet, which means the single parameter that most determines what the fund holds is the one hardest to find.
When it fails
The characteristic failure is buying it as a defensive holding and discovering it is a sector bet. Screening for low volatility repeatedly selects utilities, consumer staples and similar — which is a concentrated position in rate-sensitive, income-like businesses. When rates rise those sectors fall together, so a portfolio bought for calm delivers a correlated decline driven by a variable nobody was thinking about. The screen never claimed to control for sector; the buyer assumed it did.
A second failure is expecting it to beat the market in strong rallies. It usually does not, and the underperformance is the price of the smaller drawdowns.
A third is treating a documented anomaly as durable. Publication and product launch are both reasons for an effect to weaken.
A fourth is ignoring the fee, which on a factor product is often several times a broad tracker’s.
And a fifth is assuming low past volatility predicts low future volatility indefinitely. Clustering makes it a reasonable short-horizon bet and says nothing about regime changes.
Related
Volatility covers the measure the screen is built on. Volatility clustering covers why past calm predicts future calm at all. And systemic risk covers the decline this screen does not protect against.
Any result that contradicts the central prediction of finance theory deserves more scepticism than it usually gets, not less. The low-volatility anomaly is well documented and it is also exactly the shape of thing that turns out to be a measurement choice once somebody looks hard enough.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.