Expected returns: how the estimates are built, and how wrong they can be

Every retirement calculator, Monte Carlo simulation, and financial plan runs on one hidden input: an expected return. Change it from 7% to 5% and a comfortable plan becomes a crisis. Yet most people either accept a default or plug in the long-run historical average without asking where these numbers come from. There are two serious frameworks for estimating forward returns — valuation-based models and building-block models — and understanding both, including their embarrassing error bars, is what separates planning from wishing.

Framework 1: valuation-based (CAPE and its cousins)

The cyclically adjusted price-to-earnings ratio (CAPE) divides the market's price by its average inflation-adjusted earnings over the prior ten years, smoothing out boom-bust noise. The logic: when you pay more per dollar of earnings, your forward return is lower, because return ultimately comes from the earnings you bought. Historically, the relationship is real but loose — starting CAPE explains roughly 40% of the variation in subsequent 10-year real returns in US data. A crude translation: the inverse of CAPE (the earnings yield) is a first-pass estimate of the long-run real return. A CAPE of 20 suggests roughly 5% real; a CAPE of 33 suggests roughly 3% real.

Framework 2: building blocks

The building-block approach constructs the return from its accounting sources rather than from a valuation ratio. For stocks: expected return = dividend yield + real earnings growth + inflation ± valuation change. For bonds it's even simpler: the starting yield-to-maturity is a strong predictor of the return you'll get over the bond's duration — historically explaining over 90% of 10-year Treasury returns. The block structure forces every assumption into the open.

Notice what the table exposes: the historical 10% nominal US return included a tailwind of rising valuations and higher starting yields that a forward-looking estimate cannot honestly assume. That's why nearly every institutional forecaster — endowments, pension consultants, the large asset managers who publish capital market assumptions — has spent the last decade projecting equity returns meaningfully below the historical average. They differ on decimals; they agree on the direction.

The error bars are the real lesson

Here is the uncomfortable part. Even the best 10-year equity forecasts carry a standard error of roughly ±4-5 percentage points annualized. A '5% expected real return' genuinely means 'probably somewhere between 1% and 9%, and occasionally outside even that.' Over ten years, that's the difference between a portfolio growing 10% and 140% in real terms. No framework — CAPE, building blocks, machine learning overlays — has meaningfully shrunk this band, because a decade of returns is dominated by surprises no model contains: pandemics, rate regimes, wars, technology shocks.

Using estimates without being used by them

1. Plan with a conservative central estimate — many planners use 4-5% real for a diversified stock portfolio and 1-2% real for bonds — and treat anything better as margin.
2. Rerun the plan at pessimistic values (2% real) to find your plan's breaking point; the response to a scary answer is a higher savings rate or later date, not a hotter portfolio.
3. Update annually, not monthly. Expected returns move slowly; your reaction to them should too.
4. Never let a low expected return push you into reaching for yield, leverage, or complexity — a 4% honest expectation beats a 7% fantasy with hidden risk.
5. Remember the one lever with no error bar: your savings rate. A 1-point increase in savings rate is certain; a 1-point increase in assumed return is a hope.

The bottom line

Expected returns are estimable but only fuzzily: bonds telegraph their future through today's yield, stocks whisper theirs through valuations and building blocks, and both come wrapped in error bars wide enough to drive a retirement through. The professional response isn't to ignore the estimates or to worship them — it's to plan on conservative numbers, stress-test at worse ones, and put your real confidence in the inputs you control. Savings rate, costs, and time in the market have no standard error. Build the plan on those, and let the expected return be the thing that pleasantly surprises you.