Systems Guide · Return Mechanics

Volatility drag: geometric vs arithmetic portfolio returns

Author: WealthSyntax Editorial TeamLast updated: 2026-11-149 min read

#volatility#compounding#drawdown

Short answer

Volatility drag is the gap between a portfolio's average periodic return and its actual compounded growth rate. Because losses require disproportionately larger gains to break even, higher return swings reduce terminal wealth even when arithmetic averages stay unchanged. Over time, compound returns approximate the arithmetic mean minus half the return variance.

Mathematical Model & Derivation

Step-by-step derivation and symbol definitions

Geometric Mean Return & Volatility Approximation

Symbol Definitions

Arithmetic mean return across n periods: (1/n) * sum(r_i)

Geometric mean (compounded) return per period

Periodic percentage return in period i expressed as a decimal

Total number of compounding periods evaluated

Periodic return standard deviation (volatility)

Periodic return variance (volatility squared)

Start with the Multiplicative Wealth Trajectory

A portfolio does not accumulate returns by addition; each period multiplies the existing balance by (1 + r_i). After n periods starting from $1.00, ending wealth equals the product of all periodic growth factors.

Multiplicative Terminal Factor

Solve for the Constant Equivalent Compound Rate (G)

The geometric mean G is the constant single-period return that reproduces the exact same terminal balance W_n across n periods. Setting (1 + G)^n equal to the product of factors and taking the n-th root isolates G.

Exact Geometric Mean

Connect Volatility Drag to Asymmetric Drawdown Recovery

Why is G strictly less than the arithmetic mean A whenever returns vary? Because a fractional loss d reduces the capital base to (1 - d), requiring a gain g = d / (1 - d) just to return to 1.00. A +50% gain followed by a -50% drop leaves 1.50 * 0.50 = 0.75 of initial capital—a 25% loss despite a 0% arithmetic average.

Break-Even Gain After Loss d

Derive the Second-Order Volatility Drag Approximation

Taking the natural logarithm of (1 + r_i) turns products into sums: ln(1 + G) = (1/n) sum ln(1 + r_i). Expanding ln(1 + r_i) around A using a second-order Taylor series introduces the negative squared-deviation term -sigma^2 / 2. Note that this is an approximation for small returns and moderate volatility.

Second-Order Approximation

The +50% / -50% Paradox: Arithmetic vs. Geometric Reality

Consider a two-period portfolio that gains +50.0% in Period 1 and declines -50.0% in Period 2. The arithmetic mean return is A = (+0.50 + (-0.50)) / 2 = 0.00%. An investor reading only the average period return might assume the portfolio broke even.

In reality, multiplying the wealth factors produces (1 + 0.50) * (1 - 0.50) = 1.50 * 0.50 = 0.75 of starting capital—a 25.0% cumulative loss. Solving for the equivalent per-period compound rate yields G = sqrt(0.75) - 1 = -13.40% per period. To recover from a 50.0% drop back to 1.00, the portfolio needs a subsequent gain of +100%.

You can test any drawdown depth directly in the Loss Recovery Calculator.

How Accurate Is G ≈ A - σ² / 2? (And Lognormal Assumptions)

The formula G ≈ A - σ² / 2 is a second-order approximation, not a universal exact identity. It tracks actual geometric growth closely when single-period returns are modest and volatility is low to moderate, but diverges when period swings become extreme.

Under a continuous lognormal return model—specifically assuming independent, identically distributed period wealth factors (1 + R_t) = exp(X_t) where log returns X_t follow a normal distribution N(μ, σ_log²)—the expected arithmetic factor is 1 + E[R] = exp(μ + σ_log² / 2), so the expected continuous log growth rate equals ln(1 + E[R]) - σ_log² / 2 exactly. Notice that this exact statement applies to continuous log-rates under lognormal i.i.d. assumptions, whereas discrete arithmetic approximations widen as σ grows.

Symmetric Two-State Series: Exact Geometric Mean vs. A - σ²/2 Approximation

Arithmetic Mean (A)	Volatility (σ)	Exact Geometric (G)	Approx (A - σ²/2)	Approx Error (bps)
8.00%	10.00%	7.536%	7.500%	-3.6 bps
8.00%	15.00%	6.953%	6.875%	-7.8 bps
10.00%	20.00%	8.167%	8.000%	-16.7 bps
10.00%	35.00%	4.283%	3.875%	-40.8 bps
0.00%	50.00%	-13.397%	-12.500%	+89.7 bps

Approximation Boundary

At 8% arithmetic return and 10% volatility, A - σ²/2 lands within 3.6 basis points of the exact two-state geometric rate. At 50% volatility (+50%/-50%), the gap widens to nearly 90 basis points because higher-order Taylor terms become material.

Worked Example · Computed in Code

Same 8.00% Arithmetic Mean, Different Volatility Over 20 Periods

To isolate volatility drag with zero randomness, we compare two deterministic 20-period return paths starting with $10,000. Both paths average exactly +8.00% arithmetically per period, yet their ending balances diverge sharply solely due to swing amplitude.

Low-Vol Final Balance (σ = 4%)

$45,974.14

High-Vol Final Balance (σ = 25%)

$26,871.84

Geometric Growth Gap

2.86% / period

20-Period Deterministic Wealth Divergence ($10,000 Starting Principal, A = 8.00%)

Portfolio Series	Periodic Pattern	Arith. Mean	Std Dev (σ)	Geom. Mean (G)	20-Period Value
Low-Volatility Series	Alternating +12.0% and +4.0%	8.00%	4.00%	7.93%	$45,974.14
High-Volatility Series	Alternating +33.0% and -17.0%	8.00%	25.00%	5.07%	$26,871.84

Result check: Both series share an identical 8.00% arithmetic average over 20 periods, yet the high-volatility series finishes at $26,871.84 versus $45,974.14 for the low-volatility series—a $19,102.31 shortfall driven purely by variance.

Reference Implementations

Geometric Mean Implementation (Python & JavaScript)

Compute the exact compounded per-period return from any periodic return array without rounding intermediate steps.

PythonStandard library only

def geometric_mean(returns):
    product = 1.0
    for r in returns:
        product *= (1.0 + r)
    return product ** (1.0 / len(returns)) - 1.0

print(round(geometric_mean([0.50, -0.50]), 6))

JavaScriptZero dependencies

function geometricMean(returns) {
  let product = 1.0;
  for (const r of returns) {
    product *= 1.0 + r;
  }
  return Math.pow(product, 1 / returns.length) - 1;
}

console.log(Number(geometricMean([0.5, -0.5]).toFixed(6)));

Output for [+50%, -50%]: -0.133975 (-13.40% per period)

Analytical Pitfalls

Common mistakes

01.Quoting historical arithmetic annual averages as if an account compounds at that rate over multi-decade horizons.
02.Confusing volatility drag (a mathematical property of variance) with expense ratios or advisory fees; both reduce compounding, through separate mechanisms.
03.Applying G ≈ A - σ²/2 blindly to concentrated positions or leveraged funds where 40%+ swings violate small-return assumptions.
04.Assuming a +20% year and a -20% year cancel out to zero change (1.20 × 0.80 = 0.96, a -4.0% net loss).

Model Boundaries

Key assumptions and limits

Pure period-return series exclude external cash inflows (contributions) and outflows (withdrawals), which alter dollar-weighted returns.
Model omits capital gains taxes, dividend withholding, and transaction costs.
Approximation G ≈ A - σ²/2 ignores return autocorrelation (trending or mean-reverting serial dependence across periods).
Symmetric deterministic examples illustrate variance mechanics rather than actual market return distributions, which exhibit skew and fat tails.