Asset Allocation for Retirement: What 150 Years of Data Says

Stocks earn the premium that keeps long retirements funded; over 30-50 year horizons no other liquid asset class has historically sustained comparable withdrawal rates. But stocks also produce the deep multi-year drawdowns that, combined with withdrawals, seed nearly all plan failures — and those failures concentrate in the first decade. Bonds hold value in most (not all) equity crashes, cushioning exactly the years where withdrawals do the most damage.

So the retirement allocation question is really: how much growth can you give up to protect the early years? Too few stocks and the portfolio can't outrun inflation over decades — very bond-heavy mixes fail slowly but reliably at higher withdrawal rates. Too many and the bad-sequence tail gets fatter. Both failure modes are real; they just fail at different speeds.

Across 150+ years of US data, equity-heavy portfolios dominate the averages: higher median ending wealth and higher sustainable withdrawal rates in most cohorts. But the worst-case cohorts tell a subtler story. Portfolios in the 50-75% stock range have historically matched or beaten 100% stocks on worst-case sustainable withdrawal rates, because the bond cushion in 1929-style and 1966-style starts mattered more than the growth sacrificed. That is Bengen's original finding, and it survives in our simulations: the equity fraction that maximizes average outcomes is higher than the one that protects the bad tail.

Two caveats travel with any such number. First, bonds are not automatically safe: in several countries' histories — Germany, France, Japan around the world wars — inflation destroyed bondholders far more thoroughly than stockholders, so bond cushions are a bet on moderate inflation regimes. Second, the US record is the best case; judged against pooled 16-country data, all-in bets on any single asset class or country look worse, and diversified middles look relatively better.

"100% stocks, since the long run wins": defensible only with genuine spending flexibility and tolerance for decade-long drawdowns; the historical worst cases for all-stock retirees are meaningfully deeper, and recovery arrives too late for rigid spenders. "Age in bonds" (a 65-year-old holding 65% bonds): calibrated for short retirements and low inflation; for a 40-60 year early retirement it is dangerously conservative, starving the portfolio of the growth those horizons require.

"60/40 forever": a reasonable center of gravity, and its poor 2022 — when stocks and bonds fell together — was a one-year correlation event, not proof the mix is broken; balanced portfolios have survived far worse regimes over the last 150 years. The more interesting refinement is time-varying: a 'bond tent' that raises bond exposure around the retirement date and re-equitizes afterward targets bond protection precisely at the danger zone, which is where it earns its keep.

Allocation and spending rules are partial substitutes: both are mechanisms for surviving bad sequences. In our testing, moving from fixed withdrawals to guardrail-style flexible spending changes the optimal allocation surprisingly little — but it flattens the penalty for getting allocation wrong, because the spending rule absorbs shocks the bond cushion would otherwise have to. Flexible spenders can therefore defensibly hold more equity than rigid spenders at the same failure risk.

The practical upshot: within the broad 40-75% equity range, allocation is a second-order decision compared to the withdrawal rate and spending flexibility. Success-rate surfaces are flat near their optimum — moving 10 percentage points of allocation typically shifts success rates by far less than moving the withdrawal rate a quarter point. Get the withdrawal rate and flexibility right first; then pick the allocation whose bad years you can hold through without capitulating.