"Sequence-of-returns risk" is one of the most repeated phrases in retirement planning and one of the least quantified. The idea is simple: if you are withdrawing money, a bad run of markets in your first years does lasting damage that a good run later can't undo — because you sold assets cheap to fund spending and there's less left to recover. We built the cleanest possible experiment to measure it. Two retirees, identical $1,000,000 portfolios, identical 4%-of-initial withdrawal plan, identical 60/40 allocation. The only difference between them is when they retired relative to the market. Here is exactly what that timing luck is worth.
Headline finding: Two retirees follow the identical plan — $1,000,000, a 60/40 portfolio, and $40,000/yr (4%) held flat in real terms for 30 years. If a retiree's first decade lands in the worst 10% of historical market sequences, the plan fails 46% of the time and leaves a median of just $50,000. If it lands in the best 10%, the same plan never fails in 5,000 simulated retirements and leaves a median of $5.4 million. The entire 30-year outcome is decided in the first ten years — by luck the retiree does not control.
The same average return, two completely different retirements
The cleanest way to see sequence risk is to remove averages from the argument entirely. We took the actual 1973–2002 sequence of US 60/40 real returns and ran it two ways: forward, and reversed. By construction, both orderings contain the exact same 30 annual returns — so they have the identical arithmetic mean (6.0% real) and the identical geometric mean (5.2% real). If "average return" were what mattered to a retiree, the two would end in the same place.
They don't. The forward order — which front-loads the brutal 1973–74 crash and 1970s stagflation — leaves our retiree with about $318,000 after 30 years. The reversed order, which pushes those same bad years to the end after the portfolio has already grown, leaves about $1.96 million. Same returns. Same average. Same withdrawals. A 6× difference in ending wealth from nothing but the order.
Average return is the wrong number for a retiree. While you are accumulating, order doesn't matter — a dollar is a dollar no matter when it grows. The moment you start withdrawing, order becomes everything: bad early years force you to sell more shares to fund the same spending, permanently shrinking the base. This is why two people with the "same expected return" can have wildly different retirements.
Headline chart: failure rate by what your first decade looked like
30-year failure rate of an identical 4% plan, by first-decade market luck
50,000 block-bootstrap paths of 60/40 historical real returns, ranked by first-decade portfolio CAGR and split into deciles. Same $1,000,000 start, same $40,000 real withdrawal. Bars show the share of each group that ran out of money before year 30.
The pooled 4%-rule failure rate of 6% — the number most articles quote — is real, but it is a blend of two nearly disjoint worlds. Almost all of the risk is concentrated in the unlucky starters. The "average" retiree barely exists; you either draw a good first decade or a bad one, and that draw decides almost everything.
Full results: the three groups
Each decile group is 5,000 simulated 30-year retirements. "First-decade CAGR" is the portfolio's geometric annual real return over the first ten years. Terminal balances are in real (inflation-adjusted) dollars at year 30.
| Group | 1st-decade real CAGR | Failure rate | Median terminal | 10th pct terminal | 90th pct terminal |
|---|---|---|---|---|---|
| Worst-decile first decade | −0.77% | 45.58% | $49,522 | $0 | $857,225 |
| All retirements (pooled) | 5.58% | 6.02% | $1,823,819 | $199,832 | $5,768,001 |
| Best-decile first decade | 12.20% | 0.00% | $5,355,467 | $2,404,229 | $11,222,737 |
Three things stand out:
- The first decade is destiny. A roughly −0.8% real first-decade return is enough to push the failure rate to 46% and the median ending balance below $50,000. A +12% first decade makes failure effectively impossible.
- Even the unlucky 10th percentile of good-start retirements ($2.4M) crushes the lucky 90th percentile of bad-start retirements ($857k). The distributions barely overlap. Your starting era matters more than everything that happens afterward.
- The deterministic case is just as stark. A retiree who started in 1973 (into the crash + stagflation) ended 30 years later with $318,000; one who started in 1982 (into the great bull market) ended with $5.9 million — on the same $1M and the same plan.
CC-BY-4.0 — free for any use including republication and journalism, with attribution to QuantCalc Research.
What this does and doesn't say
This is not an argument that the 4% rule is broken, nor that retirement is a coin flip you can't influence. It's a measurement of how much of your outcome is decided by something you don't control — your starting year — when you hold a fixed plan rigid. Three honest caveats the dataset makes explicit:
- A rigid plan is the worst case. We deliberately froze spending at $40,000 real and never adjusted it. Real retirees who cut spending modestly in down years, hold a cash buffer, or use a "bond tent" in the first decade dramatically blunt sequence risk. This study measures the danger you're exposed to if you do nothing — the baseline flexibility is insuring against.
- Withdrawals are taken at the start of each year. That is the conservative convention and it slightly raises failure rates versus end-of-year withdrawal; we state it rather than hide it. The relative gap between good and bad starts is robust to this choice.
- Block bootstrap, not a forward forecast. We resample 5-year blocks of the actual 1928–2024 record to preserve real-world momentum and mean-reversion. It assumes the future resembles the distribution of the past — the standard, disclosed limitation of all historical Monte Carlo.
The useful takeaway is not "you might get unlucky." It's where the luck lands: the first decade of retirement is doing almost all the work, which is exactly why the years right around your retirement date deserve far more planning attention — and far more flexibility — than the decades after.
Methodology
Engine: sequence-risk-mc-1.0.0 · fixed seed 20260602 · 50,000 block-bootstrap paths · 30-year horizon. Generator is re-runnable and reproduces these figures exactly.
Data: Annual real (CPI-adjusted) total returns, 1928–2024 for US large-cap equity (S&P 500 total return) and the 10-year US Treasury, the standard Damodaran/NYU-Stern teaching series (reconstructable from public Shiller and FRED data). The full 97-year series is embedded in the generator as the single source of truth.
Portfolio: 60% equity / 40% 10-year Treasury, rebalanced annually, all in real terms. Historical 60/40 real return over 1928–2024: 6.24% arithmetic, 5.40% geometric.
Withdrawal: The canonical 4% rule — 4% of the initial $1,000,000 ($40,000), then held flat in real dollars, taken at the start of each year (the conservative convention; disclosed). A path "fails" if the balance reaches zero before year 30.
Monte Carlo: Stationary block bootstrap with 5-year contiguous blocks, which preserves the short-run autocorrelation (momentum and mean-reversion) that independent year-by-year sampling destroys — important precisely because sequence risk is about runs of returns. Each path is ranked by its first-decade portfolio CAGR; paths are sorted and split into deciles, and we report the bottom decile (5,000 worst-starting retirements) versus the top decile (5,000 best-starting).
Two-retiree and reversed-sequence checks: The deterministic 1973-start vs 1982-start comparison and the forward-vs-reversed 1973–2002 demonstration are pure arithmetic on the historical series — fully hand-checkable from the published data, with no randomness.
What this is not: Not personalized advice, and not a model of adaptive spending, cash buffers, glide paths, or taxes. It deliberately freezes a single rigid plan to isolate the effect of return ordering. Those mitigations are exactly what the result argues for.
The parameters are deliberately canonical — a $1M portfolio, a 60/40 mix, the textbook 4% rule, 30 years — so the result is easy to reproduce and hard to dispute. The point is not that these inputs are special, but that even the most standard plan in retirement planning has its fate decided overwhelmingly by the luck of its first decade. The generator and full dataset are published above for anyone who wants to vary the allocation, withdrawal rate, or horizon.
Is your plan exposed to a bad first decade?
The averages hide the risk that matters most. Stress-test your own withdrawal plan against thousands of return sequences — including the bad early decades that decide everything — instead of trusting a single average.
Stress-Test Your Retirement Plan →Related research
- → Does the 4% Rule Survive Forward-Looking Forecasts? — the companion study on how lower expected returns (not just their order) reshape safe-withdrawal math.
- → Social Security at 70 Beats 62 Only 42% of the Time — If You Invest — another case where a single point estimate hides a full distribution of outcomes.
- → 510,000-path 51-State Retirement Monte Carlo — same representative retiree, varying the state; California costs $167,580 more in 30-year state tax than Wyoming.
Frequently asked questions
Last updated 2026-06-02. Dataset license: CC-BY-4.0. Return data: Damodaran/NYU-Stern historical real returns 1928–2024 (reconstructable from public Shiller and FRED data). QuantCalc is an independent retirement-planning research project. Not financial, tax, or legal advice.