Suppose you start with $1 million. Your investment loses 20% in the first year and gains 25% in the second. You’re back to $1 million. Reverse the order and you still finish with $1 million.
Both investment sequences have a compound annual growth rate, or CAGR, of 0%. But that calculation assumes the money sits untouched. A retiree usually has to take some of it out.
Withdraw $40,000 at the end of each year and the loss followed by the recovery leaves you with $910,000. The recovery followed by the loss leaves you with $928,000.
The investment earned the same CAGR in both cases. You started with the same amount and spent the same amount. The order still left you $18,000 apart after only two years. That’s sequence of returns risk.
Where did the $18,000 go?
Let’s keep inflation out of this first example so we can see the arithmetic clearly. Each person withdraws a flat $40,000 after that year’s investment return.
| Order of returns | After year one | After year two |
|---|---|---|
| −20%, then +25% | $760,000 | $910,000 |
| +25%, then −20% | $1,210,000 | $928,000 |
In the first case, the market takes $1 million down to $800,000. After the $40,000 withdrawal, only $760,000 is left to earn the following year’s 25% recovery. That grows to $950,000, and the next withdrawal leaves $910,000.
In the other order, the first gain gives you $1.25 million before spending and $1.21 million after it. The 20% loss then takes that down to $968,000. Spend another $40,000 and you finish with $928,000.
The money withdrawn after the early loss wasn’t there for the recovery. That’s the missing piece when someone says the long-term return should be enough. Enough growth over the full period doesn’t tell you how much money was still invested when that growth arrived.
What happens over thirty years?
The dashed line shows the real U.S. stock returns from 1995 through 2024 in the order they happened. The two solid lines use those same returns, sorted worst first and best first. All three have the same 8.10% CAGR before withdrawals.
Start with $1 million and no spending. All three paths finish at the same value. Then choose $40,000 a year, press Play, and watch what happens to the path with the early losses. Once that account reaches zero, the remaining good returns have nothing left to compound. Change the starting portfolio or spending and all three lines update.
Starting point
Same return CAGR: 8.10%
Poor returns early
- Ending balance
- $0
- Full spending years
- 0 of 30
- Total amount spent
- $0
- Lowest balance
- $0
Poor returns late
- Ending balance
- $0
- Full spending years
- 0 of 30
- Total amount spent
- $0
- Lowest balance
- $0
Returns by year
| Year | Original order | Poor returns early | Poor returns late |
|---|---|---|---|
| CAGR |
The smaller years show when each return originally happened.
The dashed line uses actual market returns and the spending you entered. The two sorted paths are extremes to show how much the order alone can change the outcome. The 8.10% CAGR describes what the market earned before you took any money out.
Can you see this in actual market history?
You don’t have to sort returns into an artificial order to find very different retirement outcomes with similar CAGRs. I checked the complete 30-year windows in the historical data used by my Retirement Spending Calculator. Two periods beginning only six years apart make the point.
| 30-year period | Market CAGR, before withdrawals | Ending portfolio after spending |
|---|---|---|
| 1960–1989 | 4.80% real | $1,177,489 |
| 1966–1995 | 4.79% real | $0 — depleted in year 28 |
Each starts with $1 million and withdraws $40,000 a year in constant purchasing power. Without withdrawals, both would have finished at roughly $4.1 million in real terms. Their compound market growth was almost indistinguishable. With spending, one still had about $1.18 million after 30 years and the other ran out before reaching the end.
These aren’t literally the same annual returns shuffled around. The two periods differ in other ways too. But they’re a useful check on the idea that knowing the long-term CAGR tells you what a retiree could spend. Clearly, it leaves out something important.
A difficult opening decade matters more than a bad first year on its own. Michael Kitces’s analysis of historical withdrawal rates makes this distinction using early real returns. If a weak stretch takes a $1 million portfolio down to $500,000, the same $40,000 withdrawal has gone from 4% of the account to 8%. Keep taking that much out for several years and there’s less money left to benefit when markets finally recover.
Try the two periods in the calculator
Use a $1,000,000 portfolio, fixed inflation-adjusted spending of $40,000 a year, no predictable income, and a 30-year horizon. These are the calculator’s defaults.
Compare the 1960 and 1966 retirements and click Run Simulation. Both paths appear under Historical Portfolio Path in today’s dollars.
Look through the year-by-year balances, especially the early years. Keep the same two periods selected and try changing the starting portfolio or annual spending. If you compare fixed spending with amortization-based spending, look at the spending amounts as well as the ending balance. Keeping an account above zero by reducing withdrawals doesn’t mean the original household budget was funded.
I wouldn’t keep adjusting the inputs until every historical period looks comfortable and call that proof the plan is safe. These are outcomes we already know. What I want to understand is how much of the spending depends on a helpful early sequence, and what could actually change if those years were disappointing.
Does this mean you need a cash buffer?
Keeping a few years of spending in cash is a common reaction to sequence risk. I understand the appeal. The next few bills feel less dependent on what the market does. But I lean against treating a large, permanent cash bucket as the default solution. It changes your asset allocation, gives up some expected growth, and still leaves you with a decision about when to refill it. Kitces’s comparison of buckets with ordinary rebalancing is useful here: separate spending buckets do not necessarily add protection beyond the underlying allocation and withdrawal rules.
There is a qualification worth keeping. In Scott Cederburg’s 2025 discussion of his lifecycle research, allowing age-based allocations produced a temporary Treasury-bill allocation around retirement under fixed real spending. Its benefit over the fixed all-equity allocation was small. With proportional spending, that temporary allocation disappeared. That finding leaves room for a transition reserve under particular assumptions; it doesn’t establish a standing cash bucket as a requirement.
Before adding another set of bucket rules, I’d look at how much the portfolio has to fund. Spending $40,000 from $1 million starts at 4%. The same spending from $1.25 million starts at 3.2%, without changing the investment strategy. Getting to that larger amount may require saving more or working longer, so it isn’t a free fix. It simply asks less of each invested dollar.
A flexible withdrawal rule is another option. It gives the portfolio some relief after weak returns instead of insisting on the same inflation-adjusted amount every year. Vanguard’s discussion of dynamic spending explains that trade-off. The flexibility has to exist in the household budget, though. A smaller withdrawal isn’t much help if it no longer pays the bills.
For me, a lower starting withdrawal rate and a spending rule I can actually follow are a cleaner place to begin than assuming a cash bucket fixes the sequence. Keeping more of the portfolio invested preserves its exposure to long-term growth, along with its market risk. The comparisons in this article don’t rank every cash and spending strategy or prove which one has the highest expected outcome. If you do compare them, keep total starting wealth the same: adding cash on top of an unchanged investment portfolio is also adding more money.
How the numbers were checked
The historical examples use the calculator’s US stock-return series based on Robert Shiller’s data, with returns adjusted for inflation. I calculated each window’s market CAGR from the product of its 30 annual growth factors, independently of the withdrawals, and checked the spending outcomes in the calculator. Both CAGR and ending balances in the historical table are real, inflation-adjusted figures.
This is not a historical XEQT portfolio or a stock-and-bond allocation. The examples do not reproduce a Canadian investor’s taxes, currency effects, or fund costs. The historical calculator deducts spending at year-end, as in the two-year arithmetic above; the sorted-return animation deducts it at the start of each year. That timing changes the dollar results, so the animation and historical table are separate demonstrations.
The current historical dataset ends in 2024. The examples were selected to illustrate the limits of CAGR as a retirement-planning shortcut, not to estimate the chance that the next retirement will resemble either one.