Why inflation is the assumption that matters most. Over a single year, a half-point of inflation is a rounding error. Over a retirement, it's a different retirement. Suppose you need $80,000 a year in today's dollars. Grow that at 2.5% for 25 years and the bill in year 25 is about $148,000. Grow it at 3.0% instead — just half a point higher — and it's about $168,000. That half-point error means your final-year spending is roughly 13% higher than you planned, and every year in between was short too. The gap doesn't add; it compounds. This is why an honest inflation assumption beats a precise return forecast: markets recover, but a structurally low inflation assumption bleeds you for three decades.
A point of inflation, compounded
Suppose you need $80,000/year today. Watch what the same lifestyle costs under two assumptions just one point apart — 2.5% (a common default) vs. 3.5% (a defensible stress test for a healthcare-heavy retiree):
| Years into retirement | At 2.5% | At 3.5% |
|---|---|---|
| Today | $80,000 | $80,000 |
| 5 | $90,500 | $95,000 |
| 15 | $115,900 | $134,000 |
| 25 | $148,300 | $189,100 |
One extra point of inflation lifts the year-25 bill from about $148,000 to about $189,000 — a $41,000 gap you'd have to fund every year, and the shortfall compounds the whole way there. A plan that only survives at 2.5% is quietly betting on the most optimistic number in the model.
Why "the" inflation rate isn't your inflation rate. The headline number you hear on the news is the CPI-U, the Consumer Price Index for All Urban Consumers. It reflects the spending basket of about 90% of the U.S. population — students, families, workers, and retirees averaged together. But a basket is only representative if it looks like your cart. Retirees systematically over-weight two categories that tend to inflate faster than average: medical care and housing. They under-weight categories like transportation, apparel, and education. When your basket tilts toward the fast-inflating items, your personal inflation rate drifts above the average — sometimes a little, sometimes a lot, depending on the decade and your own health spending.
What the data actually shows (and the honest caveats). Congress recognized this in the Older Americans Act of 1987 and directed the Bureau of Labor Statistics to build an index for Americans 62 and older. That index — now published as the R-CPI-E, and explicitly labeled research/experimental — has over the long run risen slightly faster than the official measures: from late 1982 through 2011, roughly 3.1% per year versus about 2.9% for CPI-U. That's a gap of about 0.2 percentage points a year, driven mainly by the larger weight of medical care in older households. But be careful not to overstate it. The gap isn't constant, and it has reversed: from 2006 to 2011, the elderly index and CPI-U both rose about 2.3% a year. The R-CPI-E also carries real methodological limits — it reuses the same stores, geographic areas, and item samples chosen for the general urban population, so it doesn't fully capture where and how older people actually shop. The takeaway is not "add 0.2% and you're done." It's "your basket is different, that difference usually cuts against you, and the size is uncertain — so plan defensively."
The COLA irony. You might assume Social Security protects you from all this. It helps, but the annual cost-of-living adjustment is pegged to CPI-W — the index for urban wage earners and clerical workers, a fully employed population by definition. That's arguably the group least like a retiree. When medical inflation runs hot, a CPI-W-based COLA can systematically lag what retirees actually experience. This is also why your COLA can feel like it "doesn't keep up" even when the raise is technically inflation-adjusted — it's adjusted for someone else's inflation.
So what number should you use? There's no single right answer, but there is a right method: start from the headline long-run average (call it ~2.5–3%), then nudge it up based on how healthcare- and housing-heavy your spending is, and then run the plan again at a rate half a point to a point higher as a stress test. If your plan survives the higher number, you've bought resilience. If it only survives the low number, you've built a plan that depends on the most optimistic assumption in the model.