Reference

Time horizon changes the probability distribution.

A one-year hold of the S&P 500 lost money in roughly 25 % of historical years. A twenty-year hold lost money in zero of them. Holding period is the most under-discussed lever in investment outcomes.

The variance compression

Equity returns are volatile year-over-year but mean-revert over multi-year horizons. The standard deviation of annualised returns scales with 1/√n, where n is the number of years held:

σ_annualised(n) = σ_1-year / √n

For the S&P 500's 18% one-year standard deviation, the annualised standard deviation over a 25-year hold compresses to roughly 3.6%. The variance of long-run outcomes is dramatically narrower than the variance of short-run outcomes.

S&P 500 rolling-period statistics, 1928–2024

Holding period	Best outcome	Worst outcome	Median	P(positive)
1 year	+54.0%	−43.8%	+11.5%	~75%
3 years	+30.8%	−27.3%	+10.1%	~85%
5 years	+28.5%	−12.5%	+9.7%	~88%
10 years	+19.7%	−1.4%	+10.1%	~95%
20 years	+17.9%	+5.4%	+10.5%	100%
30 years	+13.7%	+8.0%	+10.6%	100%

The columns tell a consistent story: the best annualised outcome shrinks as horizon lengthens (the right tail compresses), the worst outcome rises sharply (the left tail compresses faster), and the median is approximately stable around 10–11 %. The distribution narrows around its long-run mean.

What this implies for asset allocation

Cash is not safe over long horizons. A 30-year hold of cash earns roughly 0.5% real, vs. 7% real for equity. The “safety” of cash is a short-horizon concept; over decades, the inflation drag compounds against you while equity volatility compresses.
Equities are not risky over long horizons. The 1928–2024 history shows zero 20-year periods with negative real returns from a diversified US equity portfolio. The framing “equities are risky” is true for short horizons and largely false for long ones.
Glidepaths reflect the asymmetry. Target-date funds reduce equity exposure as retirement approaches because the relevant horizon is getting shorter. The same investor who can withstand 100% equity at 35 cannot at 65, because their horizon has dropped from 50 years to 25 years.

The sequence-of-returns trap

The histogram of long-run returns is reassuring; the histogram of realised retirement experience is not. The difference is sequence-of-returns risk: a 30-year retirement that begins with a bad first decade is materially worse than the same 30-year horizon's averages would suggest, because withdrawals during the drawdown lock in the losses. Average returns over long horizons are robust; the timing of returns within those horizons matters enormously when you're withdrawing rather than accumulating.

Worked example: holding period vs. return outcome

$10,000 invested in a portfolio with a 7 % expected real return and 17 % standard deviation. Approximate 95 % confidence intervals (Normal approximation; reality is fatter-tailed):

Holding period	Expected value	Lower 95% bound	Upper 95% bound
1 year	$10,700	$7,366	$14,034
5 years	$14,026	$8,517	$23,082
10 years	$19,672	$10,054	$38,485
20 years	$38,697	$15,268	$98,029
30 years	$76,123	$24,089	$240,549

Note the asymmetry: the lower-bound multiplier on $10,000 grows from 0.74× (1 year) to 2.41× (30 years), while the upper-bound multiplier grows from 1.40× to 24.05×. Long horizons preserve downside protection while opening the right tail dramatically.

How to use the calculator with this in mind

The main calculator reports the realised annualised return for the period you specify. The figure is most meaningful when the period is at least 3 years; under 1 year, the result is dominated by the timing of your specific entry and exit rather than by the underlying asset's typical behaviour. For very short holding periods the simple ROI is the more honest figure.