Sequence-of-returns risk: why your retirement date matters more than your savings rate

Two retirees. Same starting portfolio of $1 million. Same withdrawal of $50,000 a year. Same average return of 7% over 30 years. One of them ends with $2.4 million. The other runs out of money at age 78.

The only difference is the order the returns arrived in.

That’s sequence-of-returns risk. It is the single most underappreciated thing in retirement planning, and it’s the whole reason Monte Carlo simulations exist. If average returns were all that mattered, you could plan a retirement on the back of a napkin.

Why the order matters when you’re withdrawing

While you’re saving, the order of returns barely matters. A bad year early on means your contributions buy more shares cheap. Math sorts itself out over decades.

Retirement flips this. Now you’re selling shares, not buying them. A bad year early in retirement means you sell more shares to cover the same dollar of expenses. Those shares are gone. They can’t recover. The portfolio enters the next bull market smaller than it should be, and never catches up.

While you save, bad early years are a discount. While you withdraw, they’re a wound that doesn’t close.

The five-year window that decides your retirement

Researchers have a name for this: the “retirement red zone.” The five years before and the five years after the day you stop working are the ones that disproportionately determine whether your plan survives the next 30.

A 30% drawdown in your tenth year of retirement is uncomfortable. A 30% drawdown in your first year, while you’re actively withdrawing, can permanently impair the plan. The math isn’t symmetric.

The retirees who got hurt worst in modern history retired in:

• 1929 (the Depression run that defined the worst-case)
• 1966 (the inflation-and-stagnation decade)
• 2000 (the dot-com unwind plus the 2008 financial crisis on top)

Retiring at the start of any of those windows produced very different outcomes from retiring five years earlier or five years later. Same lifetime average returns. Wildly different retirement experiences.

Why averaging hides it

If a planner tells you “assume 7% returns,” they’re giving you the mean of a distribution, not a sequence. A 7% mean could be 7% every year (impossible in reality), or it could be +20%, -15%, +30%, -25%, +10%, +5% (real markets). The mean is identical. The retirement outcomes are not.

This is also why “historical average return of the S&P 500 is 10%” is misleading shorthand. The historical experience of being invested in the S&P 500 was wildly different depending on which 30-year window you lived through. The 30 years ending in 1981 looked nothing like the 30 years ending in 1999.

The 4% rule was invented to handle this

Bill Bengen’s 1994 paper “Determining Withdrawal Rates Using Historical Data” wasn’t about the average. He worked backward from the question: what withdrawal rate would have survived the worst retirement starting year in modern American history? The answer he landed on was about 4% of the starting portfolio, adjusted for inflation each year.

That number isn’t a recommendation for the average case. It’s a defense against the worst case. It’s explicitly designed to survive starting your retirement in 1929 or 1966.

Flexibility is the antidote

The deeper insight from sequence-of-returns research is that fixed withdrawal strategies waste money in good cases and fail in bad ones. If you withdraw a flat $50,000 every year for 30 years and the market has a great first decade, you end up with millions of unspent dollars. If it has a bad first decade, you run out anyway.

Dynamic strategies, where you adjust spending based on what the portfolio is actually doing, perform better on both ends:

• Guyton-Klinger guardrails. Set an initial withdrawal rate. If portfolio performance pushes the current withdrawal above an upper guardrail (say 20% above target), cut spending. If it falls below a lower guardrail, you can spend more. Most years you do nothing.
• Variable Percentage Withdrawal (VPW). Withdraw a fixed percentage of current portfolio value each year, scaled by remaining life expectancy. Spending fluctuates more, but ruin probability drops to near zero.
• Floor and ceiling. Pick a hard minimum withdrawal you will not go below, and a ceiling. Float between them based on conditions.

Each one trades a different thing. Guyton-Klinger trades occasional spending cuts for higher average lifetime spending. VPW trades spending volatility for near-zero ruin. Floor-and-ceiling trades upside for predictability.

What you can actually do about it

Why this is in Thermal’s engine

The whole reason Thermal runs 10,000 Monte Carlo scenarios for every plan, instead of giving you one number, is sequence-of-returns risk. It also tracks spending floor breach rate as a separate metric from portfolio survival, so “you don’t run out of money” doesn’t hide “you cut your spending in half for fifteen years.” Both numbers matter.

I’ve written more about retirement risk for high earners specifically over at High Earner Playbook, where the conversation tends to be about scenarios where the retirement portfolio is large enough that the question shifts from “will I have enough” to “what’s the most efficient way to spend it down.” Different problem, same underlying math.