What is volatility drag? Why average returns overstate growth
Volatility drag is the amount by which compounded growth falls short of the average (arithmetic) return because returns vary. For a series of periodic returns it is approximately half the variance per period: compounded growth ≈ average return − σ² ÷ 2.
The standard illustration is a 10% loss followed by a 10% gain, which leaves 99% of the starting balance despite an average return of zero. This page takes that from an illustration to a quantity you can compute, shows why leverage makes it grow with the square of the multiplier rather than in proportion, and separates it from the sequence effect it is often confused with.
Why this matters now
Averages are what get reported. A backtest summary, a strategy description, a monthly statement: “averages 1% a month” is an arithmetic mean. What your balance does is compound, and compounding a volatile series always produces less than compounding its average — the only exception is a series with no variation at all.
The gap is small for calm series and large for volatile ones. It is also the reason the smooth line from the compounding calculator, which applies a constant rate, overstates any real series with the same average.
How it actually works
- Gains and losses are applied to different balances. A −10% period leaves 90. The +10% that follows is 10% of 90, not of 100. Result: 99.
- The loss is larger than the gain that “undoes” it. In multiplicative
terms,
(1 + x)(1 − x) = 1 − x². The shortfall is x², so it grows with the square of the swing:
| Swing | After one up and one down | Shortfall |
|---|---|---|
| ±10% | 99.00% | 1.00% |
| ±20% | 96.00% | 4.00% |
| ±50% | 75.00% | 25.00% |
- Per period, the shortfall is about half the variance. Over many periods
with average return μ and standard deviation σ, the compounded growth rate
per period is approximately
μ − σ² ÷ 2. The σ² ÷ 2 term is the drag. - Order does not matter. Twelve months alternating +6% and −6% end at 97.86% of the start. Shuffle them into any order and the result is the same 97.86%, because multiplication commutes.
That last point is the line between this and sequence of returns risk, which only appears once money is being added or withdrawn and does depend on order. Drag is present on a static balance, in every ordering.
Worked example: a year of ±6% months
An illustrative series with an arithmetic mean of exactly zero: six months of +6% and six of −6%, so σ is 6% a month.
- Approximate drag per month:
0.06² ÷ 2 = 0.18%. - Exact result after 12 months:
1.06⁶ × 0.94⁶ = 0.9786, a 2.14% loss.
Every monthly statement averaged to zero. The balance is down 2.14%, and nothing in the average warned of it. That is the whole concept in one line: the average described the months; the product describes the account.
Why leverage squares the drag
Leverage multiplies each period’s return by L. The mean scales by L, but the
variance scales by L², so the drag term becomes L² × σ² ÷ 2.
Same ±6% series, held at a constant leverage multiple and rebalanced monthly:
| Leverage | Monthly swing | Monthly drag (≈ L²σ² ÷ 2) | After 12 months |
|---|---|---|---|
| 1× | ±6% | 0.18% | 97.86% |
| 2× | ±12% | 0.72% | 91.67% |
| 3× | ±18% | 1.62% | 82.07% |
| 5× | ±30% | 4.50% | 56.79% |
Doubling leverage quadrupled the drag. At 5×, a series averaging zero lost 43% over the year.
The same algebra applies when the average return is positive: the mean term grows linearly with L while the drag grows with L², so beyond some multiple more leverage lowers compounded growth rather than raising it. That ceiling is the Kelly result in continuous form, and the reason sizing at it is dangerous when the inputs are estimates is covered in why full Kelly is too aggressive.
Daily rebalancing makes it visible quickly. A product or a strategy that resets to 2× every day, on an underlying that alternates +10% and −10% for ten days, ends at 81.5% while the underlying ends at 95.1%. Holding a constant multiple means adding after rises and cutting after falls, which is the mechanism in what is leverage seen over time.
Volatility drag vs the things it gets confused with
| Concept | Depends on order? | Needs cash flows? | What it explains |
|---|---|---|---|
| Volatility drag | No | No | Compounded growth below the average return |
| Sequence of returns risk | Yes | Yes | Same returns, different ending balance with deposits or withdrawals |
| Drawdown recovery | — | No | A −d loss needs d ÷ (1 − d) to recover |
Drawdown recovery is the single-step version of the same asymmetry: −10% needs +11.11% to get back. The drawdown recovery calculator tabulates it. Drag is that asymmetry averaged over a whole series.
Where the boundary should sit
- Report compounded growth, not the average. Annualised compounded return (the geometric rate) is what a balance experiences. An arithmetic average from a backtest or a statement is an upper bound on it.
- Estimate the drag before trusting a projection. Take the standard deviation of your periodic returns, square it, halve it, and subtract from the average. If that moves the number materially, the constant-rate projection is not describing your account — see the compounding reality check.
- Reduce variance before chasing the mean. Because drag scales with σ², cutting the size of swings improves compounded growth even at an unchanged average. Position size and leverage are the direct controls.
- Treat leverage as a variance multiplier. Doubling it quadruples the drag term. Read an equity curve for this on a log scale, where a constant compounded rate is a straight line — see reading an equity curve honestly.
FAQ
What is volatility drag?
Volatility drag is the difference between the average return of a series and its compounded growth rate, caused by the variation in returns. It is approximately half the variance per period. A −10% period followed by +10% averages zero but leaves 99% of the balance, because the gain is earned on the reduced amount.
How do you calculate volatility drag?
Approximately, as the variance of periodic returns divided by two: with a monthly standard deviation of 6%, drag is about 0.06² ÷ 2 = 0.18% a month. For an exact figure, compound the actual returns and compare the result with compounding the arithmetic average. The approximation is good for small swings and understates drag for large ones.
Does volatility drag depend on the order of returns?
No. The product of a set of returns is the same in any order, so drag is present identically in every ordering on a balance with no deposits or withdrawals. Order matters only once money flows in or out, which is a different effect called sequence of returns risk.
Why do leveraged products lose value in choppy markets?
Because a constant leverage multiple scales the variance of returns by the square of the multiple, and drag scales with variance. A 2× daily-rebalanced exposure to an underlying that alternates +10% and −10% for ten days ends at about 81.5%, while the underlying ends at about 95.1%.