Safe Withdrawal Rates by Retirement Age

Methodology

This tracker computes the maximum annual withdrawal rate (as a percentage of starting portfolio) that historically sustained a portfolio for the full retirement horizon with at least 95% success across all tested scenarios.

Historical series (solid line): uses Robert Shiller's 155-year dataset of S&P 500 and 10-Year Treasury real total returns, blended at a 60/40 stock/bond allocation. Each starting year from 1871 onward provides one rolling-window test. The SWR is the highest withdrawal where at least 95% of these historical windows survived to the end of the horizon.

CAPE-adjusted series (dashed line): uses Monte Carlo simulation with an expected return derived from the current Shiller CAPE ratio (40.71). The CAPE earnings yield (1/CAPE = 2.5%) is blended 50/50 with a long-term real stock return of 7%, a common practitioner compromise; some researchers consider raw 1/CAPE too pessimistic, others do not. That gives a compound real stock return of 4.7%, applied to the first 10 years of each simulated retirement, roughly the horizon over which CAPE has historically been most informative (and even there, imperfectly). After that, returns are resampled from history without adjustment. Returns come from a block bootstrap of the same Shiller history (5,000 paths per age), shifted so the adjusted years compound at the CAPE-implied rate. At a CAPE whose implied return equals history's, the adjustment does nothing. Every month reuses the same random draws, so a change from one month to the next reflects valuations, not simulation noise.

Sensitivity to the blend: using raw 1/CAPE (2.5%) as the first-decade return instead of the blend, the rate for a 65-year-old would be 3.1% rather than 3.5%. The blend halves the valuation signal; which is closer to right is an open question.

This differs from CAPE-based withdrawal rules such as Big ERN's Safe Withdrawal Rate Series Part 18: "The Mechanics of CAPE-Based Withdrawal Rates" (2017), which set the withdrawal rate directly from the CAPE earnings yield. Here CAPE instead conditions the simulated returns, and the withdrawal rate is solved from them. The Siegel/FAS 142 critique that CAPE may overstate valuations due to accounting changes is acknowledged but not adjusted for.

How this compares

The headline figures above use a 60/40 portfolio and a 95% survival target, so they are not directly comparable to published numbers that use other settings. To see how they compare on matched settings, the engine is re-run each month on each outside source's own settings:

On matched settings the engine's figures sit 0.01 points to 0.20 points from Morningstar's 3.9% across its stock range, and 0.11 points from Big ERN's failsafe. The Big ERN row is the closer like-for-like check: both are historical. The Morningstar row compares two sets of assumptions: this engine's CAPE-adjusted stocks with historical bonds, against Morningstar's own forward-looking capital-market assumptions for every asset class. Big ERN also uses monthly data from 1926, holds 5% in T-bills and deducts 0.05% a year in fees, while this engine uses annual Shiller data from 1871 with no cash and no fees. Simulation noise alone moves the Monte Carlo figures by a few hundredths of a point. Figures that reflect variable spending rules (such as CAPE-based withdrawal rules) answer a different question and are not compared here.

Assumptions and limitations

• Starting balance: $1,000,000 (the SWR percentage is independent of balance)
• Allocation: 60% stocks / 40% bonds (static, no glide path)
• Planning age: 95
• No Social Security, no pension, no healthcare bridge. This is a standalone portfolio-only SWR.
• Withdrawals are inflation-adjusted (real returns used throughout)
• Only the stock portion is valuation-adjusted. Bond returns are resampled from history unchanged.
• After the first decade, returns revert to the unconditional historical mean, with no extra rebound after a weak decade. That leans conservative.
• The Monte Carlo baseline sits below the historical rolling-window rate, more so at long horizons. Resampling blocks of history produces bad sequences that no single historical window contains, and 155 years of data hold few independent long windows.
• Within the CAPE range seen this year (roughly 38 to 41), the adjusted rate moves by less than a tenth of a point. The tracker is best read as a level, not a monthly signal.
• Results are not personalized financial advice. Consult a licensed financial advisor before making retirement decisions.

How to read the chart

The solid line shows what worked historically across 155 years of market data. The dashed line shows what Monte Carlo simulation suggests at today's market valuation. When the gap between the two lines is large, it means current valuations are elevated relative to history, and sustainable withdrawal rates may be lower than historical averages suggest.

For example, a 65-year-old planning to age 95 has a historical SWR of 4.2% and a CAPE-adjusted SWR of 3.5% at current valuations. Not all of that gap is valuation. The same Monte Carlo with no CAPE adjustment gives 3.9%, so about 0.3 points come from the method (historical rolling windows versus simulation) and about 0.4 points from CAPE 40.71.