Inflation Stress Test

Most retirement calculators ask you for one inflation assumption — usually 2.5% or 3%. They use that number for every category, every year, forever. That's an unrealistic simplification, and it consistently underestimates the real risk to retirement plans.

Inflation isn't one number. Healthcare has run ~7% annually for two decades. Housing runs about 3.5% in normal markets and double-digit in spikes. Food and transportation typically run 3-4%. Only general consumer goods and services line up with the headline 2.5% target. And retirees spend differently than the average worker — more on healthcare and housing, less on transportation and apparel — so the headline CPI consistently underweights retiree-relevant inflation.

Inflation regimes shift. The post-2008 era ran 1.5-2% and felt permanent. The 1970s ran 7-13% and felt permanent too. 2022-2024 reminded everyone that 5%+ is possible without warning. A retirement plan that only survives the “benign” case is a plan with hidden tail risk. Stress-testing against the 1970s analog is the standard institutional approach, and it's the one this tool implements.

Inflation hedges exist — and most retirees don't use them. TIPS (Treasury Inflation-Protected Securities) pay a real yield above CPI, so they preserve purchasing power even in inflation spikes. I-Bonds do the same with a $10K/year purchase limit per person — small but stackable. Real estate (especially residential) historically tracks inflation. Gold has no real long-run yield but sometimes spikes during inflation panics. Most mainstream portfolios hold 0% TIPS and 0% I-Bonds. That's a choice that quietly bets on the benign case.

The Social Security COLA mostly works. The mostly is the catch. The CPI-W index used for COLA tracks urban wage earners — a younger demographic spending less on healthcare. The CPI-E (experimental Elderly index) tracks retiree spending more accurately and consistently runs 0.2- 0.3% higher per year. Over 20 years that compounds to a 5-10% real income loss even when COLA is “working.” That's the SS COLA gap this tool models.