The full month-by-month total-return series behind firenum's historical backtesting, derived from Robert Shiller's dataset and published as a tidy CSV and JSON. 1,868 rows, February 1871 to September 2026. No sign-up.
CSV columns: year, month, nominalReturn, cpi · ~42 KB
What's in the Data
Each row is one month of nominal total return for the S&P 500 composite: price change plus reinvested dividends, expressed as a decimal (so 0.0173 means +1.73% for that month).
| year | month | nominalReturn | cpi |
|---|---|---|---|
| 1871 | 2 | 0.018393 | 12.8446 |
| 1871 | 3 | 0.029259 | 13.035 |
| 1871 | 4 | 0.032899 | 12.5592 |
| … | … | … | … |
| 2026 | 9 | -0.009472 | 333.8925 |
Methodology
The series is derived from Robert J. Shiller's long-running historical dataset (the data behind Irrational Exuberance). Monthly total returns are computed from his S&P composite price and dividend series. Shiller's price for a month is the average of that month's daily closes, not the close itself, so these returns are smoother than close-to-close figures and a compounded calendar year can land a few points away from published S&P 500 total returns.
The returns are nominal, not CPI-adjusted. This is deliberate: when a projection already applies inflation to expenses (as firenum's does), nominal returns are the correct input, because using real returns would double-count inflation. If you need real returns, deflate with the cpi column: Shiller's CPI series (his column E), copied as published and rounded to 4 decimals. The latest month can predate the official BLS release, so treat it as provisional.
One caveat: pre-1926 figures are reconstructed from historical sources and are less precise than modern index data. They're standard for long-horizon backtesting, but treat the 19th-century tail as indicative rather than exact.
How firenum Uses It
The Fire Planner's historical backtesting runs your plan against this series from every start year it covers. Runs start each January, so a 50-year projection tests 105 historical starting years and a 30-year one tests 125, and the results report success rates alongside the best and worst outcomes. Its portfolio survival heatmap uses the cpi column to raise each year's withdrawal with actual inflation, as the Trinity Study did. The full method is documented in the mathematical foundations guide.
How to Cite
If you use this data, please credit Robert Shiller as the original source. A suggested citation:
Shiller, Robert J. Online Data: U.S. Stock Markets 1871-Present. Yale University. Monthly nominal total returns prepared by firenum.com, firenum.com/tools/shiller-data.
Original source: Robert Shiller's Online Data page (Yale University). This dataset is offered for research and educational use; attribution is required.
Frequently Asked Questions
A flat table of monthly nominal total returns for the S&P 500 composite from February 1871 to September 2026: 1,868 rows of year, month, nominalReturn (a decimal, e.g. 0.0173 = +1.73%) and cpi (Shiller's consumer price index level for that month). Total return means price change plus reinvested dividends. Available as CSV and JSON.
These are nominal (not CPI-adjusted) returns. If you're modeling a plan that already applies inflation to expenses (as firenum's backtesting does), you want nominal returns, because using real returns would double-count inflation. If you need real returns, deflate with the included cpi column.
It's derived from Robert J. Shiller's long-running "Online Data" set (Yale University), the same data behind his book Irrational Exuberance. We compute monthly total returns from his S&P composite price and dividend series. Please credit Shiller if you reuse it.
Pre-1926 figures are reconstructed from historical sources and are less precise than modern index data; Shiller's series stitches several sources together. It's the standard dataset for long-horizon backtesting, but treat the 19th-century tail as indicative rather than exact.
The Fire Planner's historical backtesting runs your plan against this series from every start year it covers. Runs start each January, so a 50-year projection tests 105 historical starting years and a 30-year one tests 125, and the results report success rates and best/worst-case outcomes.
Yes. It's offered for research and educational use. Attribute Robert J. Shiller as the source and, ideally, link back to this page. See the "How to cite" section above.
See the FIRE spreadsheets & downloads, test a withdrawal rate in the Backtest Spreadsheet, or run a full backtest in the Fire Planner.