Repeatable vs Non-Repeatable Returns
Decompose past returns to set realistic forward expectations
Most investors set forward expectations by extrapolating recent returns, especially from the US. Felix decomposes past returns into repeatable sources (dividend yield, real earnings growth, inflation) and non-repeatable sources (multiple expansion — paying ever more for each dollar of earnings). When valuations have already expanded, that piece of past return cannot recur and almost certainly reverses or stalls.
Applied to today, US stock valuations are high by historical standards. Recent strong returns were partly real fundamental improvement (repeatable) and partly multiple expansion (not repeatable). The honest base case is therefore lower forward US returns, even without forecasting a crash. The framework doesn't tell you to short the US — it tells you to temper expectations and lean diversification away from a single overvalued market.
The practical use is in financial planning: if your assumed return is too high, your savings rate is too low, your withdrawal rate is too high, and your time-to-goal is wrong. Decomposing past returns is how you avoid that quietly compounding error.
- Past return = repeatable economic return + non-repeatable valuation change.
- Multiple expansion cannot continue forever; price-to-earnings has a ceiling.
- High starting valuations historically precede lower forward returns.
- US data is an outlier; global data gives a more honest base rate.
- Lower expected returns mean higher required savings, not lower diversification.
- Decompose the recent realised returnBreak the last 10-20 years of return for your reference market into dividend yield, earnings growth, inflation, and change in valuation multiple. The change-in-multiple piece is the non-repeatable contribution.Pro tipFor US large-cap, a meaningful share of the 2010s+ return is multiple expansion, not earnings growth.
- Anchor forward expectations to repeatable componentsForecast forward returns from current dividend yield plus plausible real earnings growth plus inflation. Do not assume further multiple expansion at already-high starting valuations.WarningIf your forecast implicitly requires P/E ratios to keep rising, you're modelling another non-repeatable boost.
- Cross-check against international dataThe US is an outlier in long-run equity premium data. Use global decompositions (Dimson-Marsh-Staunton) as a sanity check on US-only assumptions.
- Translate into plan inputsFeed the lower expected return into your savings rate, withdrawal rate, and goal timeline. Don't 'fix' the gap by reaching for higher allocations to whatever has done best recently.Pro tipA 1-2 percentage-point cut in assumed return often means meaningfully higher required savings — surface that early.
- Diversify away from the most expensive marketIf one region (today: the US) has the highest valuations and lowest forward expected returns, increase relative weight to cheaper markets without abandoning the expensive one entirely.WarningDon't confuse 'lower expected returns' with 'sell now' — being early can look wrong for many years.
Felix points out US returns have been very high partly because economic fundamentals genuinely improved and partly because investors started paying more per dollar of earnings — multiple expansion baked into the realised number.
Their work shows the equity risk premium internationally is smaller than US-only studies suggest. The US is the outlier, not the rule.
Felix references Dimson, Marsh and Staunton's work — including an older paper on the equity risk premium being a 'smaller puzzle' internationally than the US-focused literature suggested. Their decomposition of historical sources of returns across markets gave him the analytical tool: separate what can repeat from what cannot, and don't let multiple expansion masquerade as a permanent return.