CAGR & Volatility Drag

Last lesson, you learned the uncomfortable truth that the arithmetic mean of a sequence of returns (the simple average) is not the rate your money actually grows at, and that the gap widens with volatility. That was the qualitative story. This lesson makes it ruthlessly quantitative: it hands you a formula precise enough to predict, to within a fraction of a percent, exactly how much volatility will quietly skim off your compound growth. By the end you’ll be able to look at two funds with the same headline return and say, with a straight face, which one makes you richer — and by how much.

The villain has a name: volatility drag (also called the variance penalty). It’s the silent tax every bouncy strategy pays, and almost nobody on a fund factsheet will warn you about it.

CAGR — the growth rate your wealth actually felt

Analogy. Imagine your investment took a road trip. Some years it sped, some years it crashed into a ditch. The CAGR is the answer to: “If the whole trip had been driven at one constant, boring, cruise-control speed, what speed would have gotten you to the same destination in the same time?” It throws away the drama of the journey and reports the single smooth rate that’s equivalent to it.

Definition. The Compound Annual Growth Rate (CAGR) is the constant annual rate that turns your beginning value into your ending value over the period:

$\text{CAGR} = \left(\frac{V_{\text{end}}}{V_{\text{begin}}}\right)^{1/n} - 1,$

where $n$ is the number of years. This is not a new idea dressed up — CAGR is exactly the geometric mean return, annualized. It is the rate that, compounded $n$ times, reproduces your actual final wealth. It cares only about where you started and where you finished, not the route.

Worked example. You put 1000 dollars into a fund. Six years later it’s worth 2000 dollars — it doubled. What rate did your money compound at?

$\text{CAGR} = \left(\frac{2000}{1000}\right)^{1/6} - 1 = 2^{1/6} - 1.$

Now $2^{1/6} = e^{(\ln 2)/6} = e^{0.6931/6} = e^{0.1155} \approx 1.1225$. So $\text{CAGR} \approx 0.1225 = 12.25\%$. Doubling over six years feels like it should be a bigger number — but compounding is sneaky-powerful, and a steady 12.25% a year is all it takes.

Now the contrast. Suppose that doubling happened along a jagged path. Say the yearly returns were $+50\%, -30\%, +40\%, -20\%, +60\%, -16.9\%$. The arithmetic average of those six numbers is:

$\frac{50 - 30 + 40 - 20 + 60 - 16.9}{6} = \frac{83.1}{6} \approx 13.85\%.$

The factsheet would proudly print 13.85% average return. But the money only compounded at 12.25% — because that volatile path multiplied out to exactly a doubling. The 1.6-percentage-point gap between the bragged-about 13.85% and the real 12.25% is volatility drag, and we’re about to predict it from scratch.

The volatility-drag formula: g ≈ μ − σ²/2

This is the load-bearing result of the entire lesson, so let’s plant it firmly:

$g \approx \mu - \frac{\sigma^2}{2}.$

Here $g$ is geometric growth (what you compound at — your CAGR), $\mu$ is the arithmetic mean return, and $\sigma$ is the volatility (standard deviation) of returns, both written as decimals (8% is $0.08$). The subtracted piece, $\sigma^2/2$, is the volatility drag itself. The relationship is exact for log returns and an excellent approximation for ordinary returns as long as they aren’t enormous.

The intuition — why the swings cost you. Compounding is convex in a way that punishes symmetry. Lose 50%, then gain 50%, and you are not back to even: 100 → 50 → 75. You’re down 25%, even though the average move was a perfectly balanced 0%. The reason is that the loss shrinks the base that the gain then has to work on. A down-swing hurts more than an equal up-swing helps, because after the drop there’s simply less money left to grow. Mathematically this is Jensen’s inequality: the average of the compounded outcomes is less than the compounding of the average, and the size of that shortfall is governed by the variance, $\sigma^2$.

Worked example 1 — moderate volatility. A stock with arithmetic mean $\mu = 10\%$ and volatility $\sigma = 20\%$:

$\text{drag} = \frac{\sigma^2}{2} = \frac{0.20^2}{2} = \frac{0.04}{2} = 0.02 = 2\%.$

$g \approx \mu - \frac{\sigma^2}{2} = 0.10 - 0.02 = 0.08 = 8\%.$

The headline says 10%; your wealth compounds at 8%. Volatility quietly pocketed two points.

Worked example 2 — same mean, double the volatility. Now hold $\mu = 10\%$ but crank $\sigma = 40\%$:

$\text{drag} = \frac{0.40^2}{2} = \frac{0.16}{2} = 0.08 = 8\%.$

$g \approx 0.10 - 0.08 = 0.02 = 2\%.$

Same 10% average, but compound growth collapses from 8% to 2%. Notice the cruelty of the squaring: doubling volatility from 20% to 40% didn’t double the drag — it quadrupled it (2% → 8%), because drag scales with $\sigma^2$, not $\sigma$. Volatility is taxed at an accelerating rate.

See the drag bend the curve

The chart below is the formula $g(\sigma) \approx \mu - \sigma^2/2$, drawn. Fix an arithmetic mean $\mu$ with the top slider; the flat dashed line is the naive return you’d hope to compound at. The curving line is what you actually get once volatility enters — and it’s a downward parabola, sinking faster and faster from the flat line as $\sigma$ climbs. The shaded wedge between the two is the drag you’re paying at your chosen volatility.

Set $\mu = 10\%$ and drag the volatility slider to 20%, and you’ll watch the compound-growth readout land near 8% — exactly worked example 1. Push it to 40% and it caves toward 2% — worked example 2, with four times the wedge. Keep pushing and you’ll eventually cross the break-even point we tackle below, where the curve dives under zero.

Two portfolios, same average, different fortunes

Numbers in isolation are abstract; a side-by-side is where the variance penalty bites. Meet two portfolios with the identical 8% arithmetic mean but very different temperaments:

Read that last row twice. Same advertised return, and Portfolio A ends with more than twice the money. Let’s confirm the arithmetic. For A: $1.075^{20} = e^{20\ln 1.075} = e^{20 \times 0.07232} = e^{1.4464} \approx 4.248$, so 4248 dollars. For B: $1.035^{20} = e^{20\ln 1.035} = e^{20 \times 0.03440} = e^{0.6881} \approx 1.990$, so 1990 dollars. The only difference between these two portfolios is how bumpy the ride was — and that bumpiness alone vaporized over 2250 dollars of final wealth.

The chart below compounds the steady portfolio’s realistic 7.5% so you can watch the curve climb. The wild portfolio’s curve would hug far lower despite its identical headline mean — that lower path is the variance penalty made visible over time.

Break-even volatility: when “winning on average” still loses

Here’s where the formula turns genuinely scary. Stare at $g \approx \mu - \sigma^2/2$ and ask: what if the drag $\sigma^2/2$ grows so large it eats the entire mean $\mu$? Then $g$ hits zero — and beyond that, goes negative. You have an asset with a positive average return that nonetheless loses money over the long run. It “wins on average” every single year and still marches your wealth toward zero.

The condition. Geometric growth is non-positive when

$\frac{\sigma^2}{2} \ge \mu \quad\Longleftrightarrow\quad \sigma \ge \sqrt{2\mu}.$

That threshold, $\sigma_{\text{break-even}} = \sqrt{2\mu}$, is the break-even volatility: the level of bounciness at which the variance penalty exactly cancels your average return.

Worked example. Take a respectable $\mu = 8\%$. The break-even volatility is:

$\sigma_{\text{break-even}} = \sqrt{2 \times 0.08} = \sqrt{0.16} = 0.40 = 40\%.$

So an asset averaging 8% a year with 40% volatility has a long-run compound growth of $0.08 - 0.40^2/2 = 0.08 - 0.08 = 0\%$. Two decades of “averaging +8%” and you end up exactly where you started — net of the variance penalty, dead money. Crank volatility past 40% (entirely normal for a single hot stock or a crypto token) and the positive-average asset becomes a guaranteed long-run loser.

The analogy. It’s a slot machine that pays out more than it takes, on average — and still bankrupts you, because the busts compound against a shrinking pile faster than the booms can rebuild it. “Positive expected return” and “grows your wealth” are different claims, and the gap between them is exactly the volatility drag.

Pitfalls that cost real money

Pitfall 1 — the factsheet average lies by omission. When a fund advertises its “average annual return,” that’s almost always the arithmetic mean, which overstates what you actually compounded by roughly $\sigma^2/2$. The bumpier the fund, the bigger the lie. Always ask for the CAGR (the geometric mean) — or estimate the gap yourself from the fund’s volatility. A 20%-vol fund quoting a 10% “average” really delivered closer to 8%.

Pitfall 2 — leverage multiplies drag faster than return. This one is brutal and we’ll devote a later lesson to it, but here’s the preview: $k\times$ leverage scales both your mean and your volatility by $k$. Mean grows linearly ($\mu \to k\mu$), but drag grows with the square ($\sigma^2/2 \to k^2\sigma^2/2$). So $2\times$ leverage doubles your return contribution but quadruples your drag. Past a point, piling on leverage lowers your compound growth even as it raises your average return — the engine of leveraged-ETF decay.

Pitfall 3 — the approximation has limits. $g \approx \mu - \sigma^2/2$ is an approximation for ordinary (arithmetic) returns, and it frays when returns get very large or very volatile (a +200% day breaks the small-$r$ expansion the formula leans on). The clean fix: the relationship is exact for log returns — if you work in continuously-compounded log space, geometric growth simply is the mean log return, no approximation needed. For the everyday range of returns, though, $\mu - \sigma^2/2$ is accurate to a rounding error.

When it matters

This isn’t an academic curiosity — the variance penalty quietly decides outcomes across finance:

• Comparing volatile vs. smooth strategies. Whenever two options share a headline return, the lower-volatility one wins on compounded wealth. Never rank strategies on average return alone; rank on $\mu - \sigma^2/2$ (or just on realized CAGR).
• Diversification as growth, not just comfort. The textbook line is that diversification reduces risk “so you sleep better.” The deeper truth: by cutting $\sigma$ while holding $\mu$ roughly fixed, diversification shrinks the drag — it makes you compound faster. Lower volatility is a return enhancer, not merely a sedative.
• Leveraged-ETF decay. A 3× daily-leveraged ETF carries $\approx 9\times$ the variance penalty of its index. In a choppy, sideways market it can bleed value even when the underlying index ends flat — the drag, not the trend, is doing the damage. The fine print warns “for short-term holding”; volatility drag is why.
• Crypto and single hot stocks. With volatilities routinely north of 60–80%, the break-even bar $\sqrt{2\mu}$ is easy to clear. Many high-flying tokens have genuinely positive average returns and have still destroyed long-run holder wealth, purely through drag. Position sizing — keeping each bet’s contribution to portfolio σ modest — is the only defense, which is exactly the bridge to the Kelly criterion coming up.

Putting it together

Volatility is not just discomfort — it’s a measurable, compounding tax on growth. The CAGR is the geometric mean return annualized, $\left(V_{\text{end}}/V_{\text{begin}}\right)^{1/n} - 1$, and it always sits below the bragged-about arithmetic average. The gap is the volatility drag, captured by the single most useful formula in this topic: $g \approx \mu - \sigma^2/2$. Because the drag scales with $\sigma^2$, doubling volatility quadruples the penalty; two portfolios with the same mean can end decades apart in wealth; and past the break-even volatility $\sigma = \sqrt{2\mu}$, a positive-average asset becomes a long-run loser. Cutting volatility — through diversification, hedging, or sane position sizing — isn’t merely calming. It is, quite literally, a way to grow your money faster.

Next up — we finally put the variance penalty to work for us. If volatility taxes growth and leverage multiplies that tax, there must be a single bet size that maximizes long-run compounding — too timid and you leave growth on the table, too aggressive and the drag devours you. That optimal size has a name, and it’s the destination this whole topic has been climbing toward: the Kelly criterion.