Methodology: how FIRE.PipedHQ tests a retirement
FIRE.PipedHQ replays a retirement plan through every historical US start year from 1871 with enough data for the chosen horizon, in annual steps and real (inflation-adjusted) dollars, and reports the share of start years in which the money never ran out.
The data
Dataset 2026.11.1 covers 1871 to 2023. Stock and bond returns are built from Robert Shiller's long-run US market data; later index and fund series come from public sources:
- shiller-ie_data.xls@2024-09
- measuringworth-DJA (github.com/fja05680/dow-sp500-100-years)
- fred-DJIA
- fred-NASDAQCOM
- fred-NASDAQ100
- dca-2026.11.1 (monthly real returns, Feb-Jan compounded)
| Series | Source | Construction |
|---|---|---|
| stocks | Shiller S&P composite: monthly price and trailing dividend | Annual total return January to January: (P[t+1] + D̄[t]) / P[t] − 1, where D̄[t] is the mean of the 12 monthly annualized dividend values in year t. |
| bonds | Shiller long rate (10-year Treasury, GS10) | Buy a 10-year par bond at yield y0 in January and sell it the next January at yield y1 with 9 years remaining: TR = y0 + price(coupon y0, 9 years, y1) − 1, annual coupons. |
| dow | Dow Jones Industrial Average: MeasuringWorth daily closes to 2019, FRED DJIA after | Price change from January average to January average, plus an estimated dividend equal to the S&P composite's dividend yield that year (no free Dow dividend history exists). From 1897. |
| nasdaq | Nasdaq Composite: FRED NASDAQCOM monthly averages | Price change January to January plus an estimated dividend of half the S&P composite's yield. The index launched in February 1971, so returns start in 1972. |
| nasdaq100 | Nasdaq-100 (the index QQQ tracks): FRED NASDAQ100 monthly averages | Price change January to January plus an estimated dividend of half the S&P composite's yield. FRED's history starts in 1986. QQQ itself also charges a 0.20% fee that isn't modeled. |
| funds, stocks and leveraged funds | Time to target monthly snapshot: Yahoo Finance adjusted closes, Ken French high-dividend portfolio, Dogs of the Dow computed from Dow members, simulated leveraged funds | Each year's real return is the product of the monthly real returns from February through the next January (end of January to end of January for funds and stocks; the S&P 500, Treasuries, Dow and Nasdaq above keep their own January-average construction). Only complete years are used. Leveraged and inverse funds are daily-reset simulations before their inception, and a month at −100% makes the whole year −100%. Every series is a total return with dividends reinvested: adjusted closes for funds and stocks, index dividends inside the leveraged simulations. The Dogs of the Dow buys the 10 Dow members with the highest indicated dividend yield (latest regular dividend × payments per year ÷ December close) each January in equal amounts, from 2014, the first year every member can be ranked. |
| inflation | CPI-U as published in Shiller's data | CPI[t+1] / CPI[t] − 1, January to January. |
| real returns | derived | r_real = (1 + r_nominal) / (1 + inflation) − 1. The engine only ever sees real returns. |
The simulation
- Start years. Every year from 1871 that has a full horizon of data after it: 124 start years for 30 years, 104 for 50.
- Withdrawals come out at the start of each year, then the portfolio earns that year's real return. Portfolios are rebalanced yearly.
- Fixed rule: the first year's spending, raised with inflation every year, whatever markets do.
- Bear-market cut: while the portfolio is more than a set share (20% by default) below its inflation-adjusted high, spending is cut by a set share (20% by default).
- Percent-of-portfolio rule: spend a rate times each year's balance, held between a floor and an optional cap of your budget.
- Cash bucket: a number of years of spending held in cash at a fixed real yield. After a year in which stocks fell, spending comes from the bucket; after a year they rose, the bucket is refilled from the portfolio.
- Success means the money never ran out during the horizon. Endings are reported at the 10th, 50th and 90th percentile.
- Housing, income and health: the Workspace adds rent or a mortgage by metro, Social Security, part-time work and ACA or Medicare premiums as yearly streams in today's dollars.
Limitations
- Annual steps: withdrawals happen once a year in January; intra-year crashes and recoveries are invisible.
- Fund returns are after their fees; the index series carry no fees. No small-cap or international index history before the funds that hold them.
- Pre-1913 CPI and pre-1953 long rates are reconstructions with more measurement error than modern data.
- About 150 years of history gives far fewer independent 50-year periods than the cycle count suggests; overlapping cycles share most of their returns.
- Taxes are a flat gross-up of withdrawals, not a tax model.
- Dow, Nasdaq and Nasdaq-100 dividends are estimates, and those series start later (1897, 1972, 1986): a portfolio holding them only uses start years after its newest series begins, so long horizons can leave few or no cycles. The Nasdaq-100's history is short and dominated by one boom.
- Funds and single stocks have short histories (often 10–40 years) and are measured end of January to end of January, half a month later than the January-average index series; a blend only uses start years after its newest holding begins.
- Leveraged and inverse funds are simulated before inception from daily index data, T-bill rates and fees. They reset daily, decay in choppy markets, and a crash can wipe them out.
See the results in the key findings and the withdrawal-rate tables.
Test your own plan. Your spending, housing, Social Security and health insurance against every year since 1871. Free.
Start your Journey →