Consider two people, both planning to retire at 65 with $1.5 million and expecting to spend an average of $60,000 annually for 30 years. Both use the same retirement calculator. Both receive the same answer: sufficient capital for a 4% withdrawal rate. Both proceed with confidence.
One adjusts their plan for the reality that expenses compress and shift over time. The other does not. The difference in outcomes is not marginal.
The practice of averaging annual expenses across a 25- or 30-year retirement horizon is standard in most planning tools. It simplifies the calculation. It also systematically misstates the problem.
Retirement is not a static period. Spending does not occur in uniform increments. The categories that dominate early expenses are not the same ones that dominate later. Inflation does not affect all categories equally. And the sequence in which portfolio returns arrive relative to when withdrawals occur determines whether a plan survives or fails—even when the average withdrawal remains constant.
Averaging collapses these variables into a single figure. The result is a number that appears precise but quietly incorporates assumptions that rarely hold.
The Shape of Spending Changes Across Retirement
Retirement expenses do not follow a flat line. They follow a curve.
In the first decade, spending tends to be highest. Travel occurs while health permits. Home improvements that were deferred during working years are completed. Discretionary purchases cluster in this period. The pattern is consistent across income levels: early retirement is when people deploy capital for experiences and adjustments that were postponed.
In the middle years, spending moderates. Travel slows. Major purchases taper. Routine living costs dominate, and discretionary spending compresses as energy and interest shift.
In the final years, healthcare expenditures rise sharply. Long-term care, medications, and end-of-life medical costs often eclipse what was spent in earlier phases. But other categories—travel, dining, entertainment—approach zero.
| Retirement Phase | Years | Typical Spending Pattern |
|---|---|---|
| Early (Active) | 1–10 | $70,000/year (travel, discretionary high) |
| Middle (Stable) | 11–20 | $55,000/year (routine expenses dominate) |
| Late (Healthcare) | 21–30 | $65,000/year (medical costs elevated) |
The arithmetic average of these three phases is $63,333. But no single year looks like the average. A retiree spending $70,000 in year one faces a very different withdrawal pressure than one spending $55,000. If the portfolio declines 20% in year two, withdrawing $70,000 has a different structural impact than withdrawing $55,000.
The average obscures this. It treats a portfolio withdrawal in a declining market as equivalent to one in a rising market, as long as the long-term mean is preserved.
The math does not support this equivalence.
Inflation Doesn’t Play Fair Across Categories
Most retirement projections apply a single inflation adjustment—often 3% annually—to all expenses. This is arithmetically convenient. It is also structurally incorrect.
Healthcare costs have historically inflated at rates between 5% and 8% annually. Housing costs, depending on geography, inflate between 2% and 4%. Food costs fluctuate but tend to track closer to general inflation. Transportation costs depend on energy prices and exhibit higher volatility.
A retiree whose expenses are 40% healthcare, 30% housing, and 30% other costs is not facing 3% inflation. They are facing a blended rate closer to 4.5% or 5%, depending on the year. If healthcare inflates at 6.5%, housing at 3%, and other costs at 3%, the blended rate is 4.4%: (0.40 × 6.5%) + (0.30 × 3%) + (0.30 × 3%) = 2.6% + 0.9% + 0.9%.
Over 25 years, the difference between 3% and 5% inflation is not incremental. It is structural.
| Annual Inflation Rate | $60,000 Initial Expense → Year 25 |
|---|---|
| 3% | $125,727 |
| 5% | $203,154 |
The retiree who planned for $125,000 in year 25 and faces $203,000 does not have a minor miscalculation. They have a gap of approximately $77,000 in a single year—a gap that recurs and compounds in every subsequent year.
Averaging inflation across categories hides this. It smooths variance that cannot actually be smoothed in practice. The result is a plan that appears solvent on paper but quietly erodes purchasing power over time.
When Your Plan Meets Reality: The Sequence Problem
Margaret Chen retired in January 2000 with $950,000. She planned to withdraw $45,000 annually, adjusted for inflation—a sustainable 4.7% initial withdrawal rate by every standard model at the time.
Three years later, her portfolio was worth $620,000.
She had not overspent. She had not panicked and sold. She had simply retired at the beginning of a bear market—the dot-com crash followed by the 2001 recession—and continued withdrawing as planned while the market declined.
By 2009, after the financial crisis, her portfolio had recovered to $680,000. But she was now 74, withdrawing $52,000 annually (inflation-adjusted), and the math had fundamentally changed. The early losses, combined with ongoing withdrawals, had created a hole that subsequent gains could not fill.
She would run out of money at 83. Her plan had assumed she would be solvent into her mid-90s.
This is sequence-of-returns risk. And it becomes acute when expense timing is misunderstood.
A portfolio that averages 7% annual returns over 30 years does not deliver 7% each year. It delivers -15% in some years and +22% in others. The sequence in which those returns arrive relative to when withdrawals occur determines outcomes more decisively than the average return itself.
Consider two retirees, both with $1 million, both planning to withdraw $60,000 annually (adjusted for inflation), both experiencing the same set of annual returns over 30 years—just in reverse order.
| Retiree | Returns Sequence | Portfolio Value After 30 Years |
|---|---|---|
| A | Strong early, weak late | $1,840,000 |
| B | Weak early, strong late | $340,000 |
Same average return. Same average withdrawal. Opposite outcomes. (Illustrative example using identical returns in reverse sequence.)
Now layer in the reality that withdrawals are not actually constant. Early retirement years demand more capital for discretionary spending. If those years coincide with poor market performance, the portfolio depletes faster than average calculations suggest. The damage is not recoverable, even if later returns are strong.
A plan that models expenses as a flat $60,000 misses this entirely. It assumes withdrawals are fungible across time. They are not. A withdrawal in year two, when the portfolio is down 18%, locks in losses that cannot be undone by gains in year fifteen.
Margaret’s experience was not an anomaly. Anyone who retired between 1999 and 2002, or in 2007–2008, faced the same structural problem. The decade matters more than the average.
Fixed Versus Variable: A Critical Distinction
Not all expenses respond to market conditions or portfolio performance. Mortgage payments, insurance premiums, property taxes—these are fixed obligations. They do not compress when the portfolio declines.
Discretionary expenses—travel, dining, entertainment—can compress. But the extent to which they can be reduced depends on the retiree’s flexibility and willingness to adjust lifestyle expectations.
A retirement plan built on average annual expenses implicitly assumes all spending is variable. It assumes the retiree can reduce expenses in weak years and increase them in strong years to maintain portfolio sustainability.
This assumption is rarely true.
Most retirees face a baseline of fixed costs that consume 60% to 70% of total spending. The remaining 30% to 40% is where adjustment can occur. But reducing discretionary spending by half in a down year only lowers total expenses by 15% to 20%.
If the plan required $60,000 annually and the portfolio is stressed, cutting discretionary spending might bring expenses to $50,000. But if the fixed baseline is $42,000, the retiree cannot compress further without restructuring housing, healthcare, or insurance—changes that are disruptive and often irreversible.
Averaging ignores this constraint. It models spending as infinitely adjustable. The reality is that most retirement expenses have a floor. And when that floor is reached, the portfolio either supports it or it does not.
Constructing a More Accurate Model
A better approach disaggregates expenses by category, applies category-specific inflation rates, and models spending phases rather than a single average.
Start by separating expenses into three groups:
- Essential fixed costs (housing, insurance, utilities, baseline food): Apply conservative inflation assumptions (3%–4%) and treat as non-negotiable.
- Healthcare (insurance premiums, medications, out-of-pocket): Apply higher inflation (5%–7%) and expect escalation in later years.
- Discretionary (travel, dining, entertainment, hobbies): Model as variable, with higher spending early and compression later.
Then map these categories across retirement phases:
| Phase | Essential Fixed | Healthcare | Discretionary | Total Annual |
|---|---|---|---|---|
| Early (Years 1–10) | $35,000 | $12,000 | $20,000 | $67,000 |
| Middle (Years 11–20) | $40,000 | $18,000 | $10,000 | $68,000 |
| Late (Years 21–30) | $42,000 | $28,000 | $5,000 | $75,000 |
The average across all years is $70,000. But the distribution matters. High discretionary spending early coincides with the period when sequence risk is most dangerous. High healthcare spending late coincides with the period when the portfolio has less capacity to recover from shocks.
This structure surfaces the actual stress points. It allows for scenario testing: What if healthcare inflation runs at 8% instead of 6%? What if discretionary spending cannot be reduced as planned? What if returns are weak in the first decade?
Averaging eliminates these questions. Disaggregation forces them.
And forcing them is the point. You’re not planning for the average case. You’re planning for your specific sequence of health events, market returns, and spending needs—none of which will be average.
The Cost of Misestimation
Underestimating retirement expenses by even 10% creates compounding shortfalls. A retiree who planned for $60,000 annually but actually needs $66,000 will exhaust their portfolio years earlier than expected—not because of poor discipline, but because the initial calculation was structurally flawed.
The feedback loop in retirement is unforgiving. Unlike working years, where income can adjust to expenses, retirement income is largely fixed. The portfolio must support the expenses, or the expenses must compress. There is no third option.
And because expenses do not compress uniformly—because healthcare costs and fixed obligations create a floor—the margin for error is narrower than most projections assume.
A plan built on averages appears robust until the first decade diverges from the average. Then the divergence accumulates. By the time the realization occurs, the opportunity to adjust—through additional savings, delayed retirement, or restructured expectations—has passed.
Margaret Chen eventually adjusted. She moved to a smaller home at 76, reducing her fixed housing costs by $8,000 annually. She eliminated most travel. She took a part-time consulting role for three years, adding $18,000 annually to delay further portfolio withdrawals.
These were not catastrophic changes. But they were changes she had not planned to make, occurring at ages when restructuring becomes harder and less appealing. Her original plan had promised freedom from these adjustments. The average told her the money would last. The sequence showed otherwise.
A Different Frame
Retirement planning is not primarily about predicting the future. It is about constructing a margin durable enough to absorb variance.
The variance in spending is not random noise. It has structure. Early expenses are discretionary and compressible but tend to be elevated. Late expenses are driven by healthcare and are largely non-discretionary. Inflation affects categories asymmetrically. Sequence risk interacts with withdrawal timing in ways that average models ignore.
Averaging collapses this structure. It trades precision for simplicity. The simplification is costly.
The retiree who models expenses as they will actually occur—disaggregated, phase-adjusted, inflation-variant—builds a plan that reflects reality rather than convenience. The capital requirement may be higher. The timeline to retirement may extend. But the plan will not quietly fail in year eighteen because the assumptions were too smooth.
The numbers do not need to be perfect. They need to be honest about where the pressure will concentrate and when the portfolio will face the greatest strain. Averages provide neither.
Retirement is long enough that small structural errors become large financial gaps. The difference between modeling expenses as they are and modeling them as averages is not a matter of sophistication. It is a matter of whether the plan acknowledges how spending actually behaves across time, across categories, and across market conditions.
The retiree who builds from that acknowledgment has a plan. The one who relies on averages has a number.
