The Retirement Spending Smile: How Spending Really Changes With Age

This analysis draws on David Blanchett’s research on retirement spending, published in the Journal of Financial Planning in 2014 and in Financial Planning Review in 2026, and on work by Michael Kitces and Derek Tharp at Kitces.com. We’d recommend reading them in full.

Why Most Plans Assume Spending Never Changes

Constant real spending is the default in retirement research and in most planning software. The household’s spending rises with inflation every year, so its purchasing power never changes from the first year of retirement to the last. William Bengen’s 1994 study, the origin of the 4% rule, raised withdrawals with inflation every year, and most of the research since has done the same. Blanchett notes, with some humor, that most of his own research makes the assumption too, despite his having published evidence against it.

The assumption is not foolish. It is simple, easy to explain, and sounds prudent: if spending does fall, the household simply has money left over. But it is an assumption about how people behave, and that behavior has been measured. Our article on what it takes to retire explains why the answer starts with the life a household wants and what it costs, not with a single number; the spending path is the next layer down, and it changes the answer.

How Spending Changes in Retirement: What the Research Found

Three kinds of evidence point the same way. Michael Kitces, writing at Kitces.com in 2016, has walked through them in this order, and the order matters: each kind follows real people more closely than the last, and each has a different weakness.

Snapshots by age. The Bureau of Labor Statistics’ Consumer Expenditure Survey shows older households spending far less than middle-aged ones. In 2025, households headed by someone 45 to 54 spent an average of $106,658, those headed by someone 65 to 74 spent $69,333, and those headed by someone 75 or older spent $57,959. A 2005 study by planner Ty Bernicke in the Journal of Financial Planning found, in the same survey, that retirees’ spending was about 15% lower for every five years of age. Snapshots compare different people, though, born in different decades with different incomes and habits, so part of any gap may be a generation effect rather than aging.

Age groups followed over time. Research from the Center for Retirement Research at Boston College followed age groups through the same survey over many years, which removes most of the generation effect, and found real spending falling about 1% a year as the groups aged. Following groups still blends many households together, though, and cannot show what any one household did.

The same households followed over time. The Health and Retirement Study, run by the University of Michigan, follows the same households over time, and its spending supplement, the Consumption and Activities Mail Survey, asks a sample of them about their spending every two years, which makes it the closest thing to watching retirees age. Blanchett’s 2014 study used that data, from 591 households that met his filters, and found that, between ages 60 and 90, real spending fell by an average of 0.96% a year. The decline was not even. Real spending slipped slowly at first, fell fastest in the middle years, and slowed its decline again near the end. Plotted as yearly changes, the pattern curves like a smile. One point is easy to miss: the smile describes the yearly rate of change, not the level of spending. Adjusted for inflation, the level mostly keeps falling, just more slowly at the end.

The 2014 study also found that the decline depended on how much a household spent. Households spending more tended to cut more as they aged, which matters for exactly the households most likely to be working with a planner.

The Smile and the Smirk: The Median Versus the Average

Blanchett’s 2026 update drew on eleven rounds of the same survey, from 2001 through 2021, rather than the five his 2014 study used: about 7,800 observations of how a household’s consumption changed between rounds, from about 2,260 respondents. Two findings stand out.

The shape depends on which retiree you follow. For the median retiree, the yearly decline in real spending keeps going into the late 80s, with no late upturn: a smirk. Averaged across all retirees, the yearly change turns up at older ages: a smile. Health care is the likely explanation. A large medical or long-term care bill hits a minority of households, and when it does, it pulls the average up without necessarily moving the median household. Research by economists Michael Hurd and Susann Rohwedder, which Blanchett cites, found real spending declines that persisted at older ages; Blanchett’s 2014 work found a late upturn. His reconciliation is that both can be right, because they describe different retirees: the median retiree on one side, and an average that picks up the late upturn on the other.

Much of the decline looks like choice. Blanchett sorted households by how well funded their spending was, comparing what they spent with what their income and savings could support. Households whose resources could support at least their current spending still cut real spending on average. The best-funded group, with resources to support at least twice what they spent, did raise real spending, but by an average of only 1.1% a year, far less than they could have afforded, while the most underfunded cut real spending by an average of 7.5% a year. Among households spending $80,000 or more, average real spending fell no matter how well funded they were. The decline also builds over time: over any two-year stretch, only about 55% of 60-year-olds and about 60% of 80-year-olds spent less in real terms, but over ten years the shares rose to about 75% and 85%.

So the decline is not simply people running low, though about a third of the households in Blanchett’s sample could not fully support what they were spending. Some of it is contentment: one car instead of two, a paid-off house, less appetite for long flights. Some of it may be caution that quietly costs the household something it would have valued, the trip that kept getting postponed or the gift that waited for an inheritance. Which is which is a conversation, not a statistic, and it is one reason every plan we build starts with a Statement of Financial Purpose.

To see what the shapes mean in dollars, we applied the age-only equations Blanchett published in 2026 for each shape to $100,000 of spending at 65, in today’s dollars. Because those equations depend only on age, the percentages are the same at any starting amount; his 2014 version, which also accounts for the spending level, would put a $100,000 household on a steeper path than the smile shown here, ending near $69,300 at 95 instead of $79,630. On the smile path, the yearly change starts near zero, deepens to about 1.4% a year around 78, returns to zero at about 91, and turns slightly positive after that; spending is $87,044 at 80, $78,892 at 90, and $79,630 at 95. On the smirk path, the yearly decline starts at about 0.3% and steepens to about 2.2% a year by 95; spending is $85,948 at 80, $71,007 at 90, and $63,731 at 95. Over the 30 years from 65 through 94, flat spending totals $3,000,000, the smile path about $2,653,600, and the smirk path about $2,556,900, roughly 12% and 15% less. A steady real decline of 1% a year, the shape of our own baseline before its small late uptick, lands at $73,970 at 95, between the two, and its 30-year total of about $2,603,000 falls almost exactly halfway between theirs.

Real Versus Nominal: Falling Spending Still Rises on Paper

A real decline of about 1% a year is smaller than inflation, which is why “spending falls in retirement” can sound wrong to a retiree watching the bills. At 2.5% inflation, the rate Blanchett used in his illustration, the smile path’s $100,000 at 65 becomes about $134,052 in actual dollars at 85, against $163,862 if spending kept pace with prices. The household still spends more each year in dollars. It spends less in purchasing power.

That distinction keeps inflation in the plan even when spending declines. A retiree’s own basket can inflate faster than the headline rate. Blanchett points out that medical care carries a weight of 11.3% in the government’s experimental price index for older Americans, against 8.3% in the standard index, as of December 2024. He also notes that medical prices rose 5.10% a year from 1957 through May 2025, against 3.56% for everything else. Social Security’s cost-of-living adjustment follows a price index for working households (the CPI-W), so it tracks general inflation but not necessarily the faster-rising basket of many older households; a pension with no cost-of-living adjustment does not adjust at all. We lean toward weighting inflation risk more heavily than many plans do, particularly for income fixed in dollars, as we explain in our article on the pension lump sum or annuity decision. The counterarguments deserve a fair hearing: the Federal Reserve targets 2% inflation, inflation has stayed low for long stretches, and Blanchett argues that because real spending tends to fall and Social Security is indexed to inflation, most retirees do not need a perfect hedge.

What Changes Inside the Budget

The total falls because the parts move differently, and not all in the same direction. In the survey data, spending on transportation, insurance, retirement contributions, clothing, and household furnishings falls steadily with age. Health care rises, from about 5% of spending for households under 35 to roughly 15% for those 75 and older, by Blanchett’s reading of the survey. Housing takes a larger share as other spending shrinks, and so do cash gifts and charitable giving. Even as total spending declines, the share going to other people grows.

That is why we lean toward asking which parts of the budget are likely to change, rather than treating spending as one number. Michael Kitces has laid out one version of this at Kitces.com, building on Somnath Basu’s age-banding work, projecting each spending category at its own rate. In our plans, the question lives in the two layers we build into every plan, core and adaptive, labels that come from Kitces’s 2019 work at Kitces.com (building on Jonathan Guyton’s approach). The core layer is what the household wants to keep doing no matter what; for many families, in Texas and beyond, that includes an annual trip with the children and grandchildren, and their regular giving. The adaptive layer (the size of that trip, the larger gifts, the second car, the bigger projects) is where most of the decline tends to come from, and it usually follows the go-go, slow-go, no-go rhythm: bigger trips give way to closer ones, then to local outings and an occasional dinner out, then to fewer outings at all. Blanchett’s 2026 model made the same assumption, taking three quarters of any decline from flexible spending.

For Texas homeowners, one line item behaves differently from either. The school district’s share of the property tax bill is held by a ceiling that starts at 65, at or below its level in the year the owner qualifies, apart from new improvements, while the other lines can keep growing unless the county, city, or junior college district has adopted its own freeze, as our article on Texas property tax at 65 explains.

For any one household, the change rarely arrives as a smooth slope. Kitces observes that for an individual retiree, a step down is more likely to follow an event, such as a fall that ends the long trips, than a birthday. A diagnosis, a move closer to family, or the death of a spouse can do the same. Those steps come on their own schedule, which is why a plan can model a smooth average but has to be revisited regularly to catch the step when it happens. For couples, we model the first death in every plan, because the survivor’s income and spending change at the death, and the tax brackets usually change the year after, a shift we cover in our article on taxes after a spouse’s death.

Health Care: What Bends the Curve

Most of a retiree’s ordinary health spending is fairly predictable once Medicare starts: premiums for Part B and Part D, the cost of a supplement or an Advantage plan, deductibles, and copays. Kitces notes that with Part B, Part D, and a supplement in place, most of what follows is modest copays, so ordinary health spending rises only moderately in the later years. With an Advantage plan, more of the cost arrives as copays, up to the plan’s yearly out-of-pocket limit for medical care, with prescription drug costs capped separately. The choice between the two paths is also one of the hardest retirement decisions to undo, as we cover in Medigap versus Medicare Advantage.

What bends the curve is the tail. Blanchett looked at retirees in the Health and Retirement Study who had died and added up the out-of-pocket medical costs they paid after 70 beyond their premiums, in 2022 dollars. Among those who died at 95, the median was about $50,000 over all those years. At the 95th percentile, it was roughly $250,000. Long-term care is a large part of that tail, and costs like these, concentrated in a minority of households, are the likely source of the late upturn in the average across all retirees.

Our baseline treats the two differently. The spending path carries only a small uptick in the last few years, and for many households that uptick mostly stands in for long-term care costs they would pay themselves, since many arrive at retirement without coverage and find the premiums at that point hard to justify. The real long-term care planning happens on top of that baseline, in conversation and case by case: whether any coverage still makes sense, how large a reserve to hold, and which dollars would pay, often as its own named and modeled reserve, as we describe in our article on paying for long-term care. Keeping the large bill out of every year of the spending path keeps the ordinary years honest, and keeping it in the plan keeps the tail from being forgotten.

Why the Shape Changes What a Household Can Spend and Give Today

Spending that falls later supports more spending and giving now, but by less than people expect. Derek Tharp’s 2017 research at Kitces.com tested four declining spending patterns against U.S. market history for a portfolio of 60% large-cap U.S. stocks and 40% Treasury bills, rebalanced annually, over 30 years. The historically safe starting withdrawal rate, about 4% with flat spending, rose by 0.32 to 0.75 percentage points, or 8% to 18% more spending in the first year, even though the steepest of those patterns ended about a third lower by the final years. The gain is modest partly because the cuts arrive gradually and partly because sequence-of-returns risk sits at the front of retirement: a cut that arrives at 80 cannot rescue a portfolio that a bad first decade has already drained.

Blanchett’s 2026 model, which treats part of spending as flexible and measures outcomes more broadly than a success rate, found a larger effect. For a retiree with $30,000 of Social Security, a $60,000 spending goal, a portfolio of 40% stocks and 60% bonds, and a 30-year horizon, the starting withdrawal rate at a moderate level of risk aversion was 5.2% with flat spending, 6.2% with the smile, and 6.4% with the smirk. His rates run higher than traditional safe withdrawal rates across the board, because his model counts the household’s ability to flex and judges outcomes by more than whether the money lasts. The two declining shapes landed much closer to each other than to flat spending, which is the more useful lesson: moving off the flat-spending assumption matters more than picking the exact curve.

Even Tharp’s smaller range is not a rounding error. For a household that would otherwise draw $100,000 from its portfolio in the first year, 8% to 18% is $8,000 to $18,000 more at the start. That money arrives in the years when it buys the most: the trips with grandchildren while everyone can still go, help with a first home, the gift to a cause while the giver can see what it does.

Is Planning for Flat Spending the Safe Choice?

The case for assuming flat spending anyway is a margin of safety: if spending falls, the household has extra. Tharp’s answer is that a margin bought this way has costs that are easy to miss. It can constrain a retiree’s lifestyle unnecessarily in the early years, when many retirees would rather spend, and give, while they are most able to enjoy it. And in a detailed plan, a spending path that runs too high in the later years can distort the tax-efficient spend-down across account types, which in our plans also touches how much to convert to Roth, a coordination problem we cover in retirement withdrawal sequencing. Tharp also points to other places a margin can live, among them ongoing monitoring of the plan, the planning horizon, and the legacy goal. In our plans, we lean toward putting it there: a plan reviewed regularly, spending lines for a raise or a cut set in advance (as described in our article on retirement spending guardrails), a planning horizon set with health and family history in mind, a legacy goal, and a separate plan for long-term care.

The research has limits worth stating plainly. It describes averages and medians, and Blanchett notes that his sample left out the largest swings in spending and households whose marital status changed, so real households are more varied than the curves. Retirees are working longer and retiring differently, which may change the shapes. The late upturn likely reflects American health care costs, especially long-term care, and Blanchett expects it to be smaller in countries with stronger public coverage. Kitces adds that retirees may not be very good at projecting their own lifestyle changes in later retirement, a version of what psychologists call the end of history illusion, the tendency to underestimate how much we will change, so a household’s own forecast of its later spending deserves some gentle questioning. And for some households a decline is the wrong assumption: one whose budget is almost all essentials, or one planning large gifts late in life, may spend steadily for decades.

How We Think About It

We lean toward building the spending path household by household rather than assuming one shape for everyone. When a budget is all essentials, we lean toward modeling flat real spending, because there is little that naturally falls away with age. When there is a discretionary layer, we lean toward a baseline of a steady real decline of about 1% a year, close to the average decline the research finds, with a small uptick in the last few years that mostly stands in for long-term care costs for households without coverage. Like the median retiree’s smirk, it assumes no late surge in spending; over 30 years it lands between the two research shapes. The baseline is where the conversation starts, not where it ends: long-term care, the family trip, and the giving all get planned on top of it, household by household.

We lean toward treating the path as a starting assumption, not a forecast. We typically update retired clients’ plans about twice a year, so a household whose spending runs above or below the curve shows up at the next review. The core and adaptive layers show where a step down would come from, and the guardrails show whether the plan has room for a raise or needs a trim.

And we lean toward remembering what the early years are for. A plan built around a realistic spending path frees more of the household’s resources for the years when health and energy are highest. For many families that means the annual trip grows instead of shrinking, the grandchildren come along, and gifts to children and causes happen while everyone can enjoy them. For some it means more margin and less worry. Either way, the point is not to spend more for its own sake. It is to spend and give on purpose, in the years when it means the most.

Common Questions

What is the retirement spending smile?

The retirement spending smile is the pattern David Blanchett found in 2014 when he tracked the same retired households over time: real spending declined slowly in the first years of retirement, most quickly in the middle years, and slowly again near the end, so the yearly changes plotted by age curve like a smile. Spending, adjusted for inflation, still mostly declines. His 2026 update found that for the median retiree the late upturn is missing, which looks more like a smirk, while the average across all retirees still turns up, likely because of large health care costs for a minority of households.

Do retirees spend less as they get older?

In real terms, usually yes. Across studies, real spending falls by roughly 1% a year on average, and in Blanchett’s 2014 study the average decline between ages 60 and 90 was 0.96% a year. The decline appears even among households that could afford to spend more, and over ten years about 75% of 60-year-olds and 85% of 80-year-olds spent less in real terms. Because inflation usually runs above 1%, spending in actual dollars often still rises, just more slowly than prices.

Does health care make retirement spending go up later in life?

For a minority of households, yes, and sharply. Health care’s share of spending rises with age, to roughly 15% for households 75 and older, and medical prices have risen faster than other prices. But for the typical retiree, Medicare keeps ordinary costs fairly stable, and total real spending keeps falling. The large late-life increases come largely from long-term care and other big medical bills. In Blanchett’s data, among retirees who died at 95, cumulative out-of-pocket medical costs after 70 were about $50,000 at the median and roughly $250,000 at the 95th percentile.

What are the go-go, slow-go, and no-go years?

The go-go, slow-go, and no-go years are financial planner Michael Stein’s names for three stages of retirement. The go-go years are the active early years of bigger trips and projects. The slow-go years, roughly the second decade of retirement, are when travel and activity start to taper, toward closer trips and local outings. The no-go years are the final stage, when outings become occasional and health needs dominate. The labels match the research in broad strokes, but for any one household the stages more often arrive as steps after events, such as a fall, a diagnosis, or the death of a spouse, than on particular birthdays.