Historical Data Sources
Our backtesting calculations are powered by Robert Shiller's comprehensive historical market dataset, the same data used by financial researchers worldwide.
Robert Shiller's Dataset
Robert Shiller is a Yale economist who won the 2013 Nobel Prize in Economics for his "empirical analysis of asset prices." His historical market data is considered the gold standard for financial research and is used by academics, institutions and retirement calculators worldwide.
Visit Shiller's Data PageWhat Data We Use
U.S. Stock Market Data (1871-2025)
154 years of S&P 500 data including prices, dividends and earnings. Before 1926 (when the S&P 500 was created), Shiller uses the Cowles and Associates data: a carefully constructed index of comparable stocks.
CAPE Ratio (P/E10) Data
The Cyclically Adjusted Price-to-Earnings ratio, also known as the Shiller P/E or P/E10, divides the current price by the average of 10 years of inflation-adjusted earnings. This metric helps gauge whether markets are over or undervalued relative to history.
U.S. Housing Data (1890-2024)
135 years of home price data from Shiller's famous "Irrational Exuberance" dataset. This data reveals the surprising truth about long-term housing returns: real appreciation has been close to zero over the long term.
CPI Inflation Data
Consumer Price Index (CPI) data is used to convert nominal values to real (inflation-adjusted) values. This allows us to show returns and withdrawals in "today's dollars", purchasing power you understand today.
Notable Years in the Data
| Year | Significance |
|---|---|
| 1871 | Earliest year in Shiller's stock market data |
| 1890 | Earliest year in Shiller's housing data |
| 1926 | S&P 500 index created (earlier data uses Cowles index) |
| 1929 | Stock market crash, start of Great Depression |
| 1966 | Start of the "lost decade" (high inflation, low returns) |
| 1992 | Final year in Bengen's original 4% Rule paper |
| 1995 | Final year in the Trinity Study |
| 2000 | Dot-com bubble peak, CAPE at all-time high (44.19) |
| 2008 | Financial crisis: stocks down ~37%, housing -35% peak-to-trough |
| 2025 | Most recent complete year in our dataset (no partial years) |
Why This Data Matters
Retirement planning requires understanding how markets behave over long periods. Short-term data (5-10 years) can be misleading because it may only capture bull or bear markets. Our 95+ years of backtesting data includes:
- The Great Depression (1929-1932), stocks lost 80%+ of their value
- The Stagflation Era (1966-1982), high inflation eroded real returns
- The Dot-Com Bubble (2000), extreme overvaluation followed by 50%+ decline
- The Financial Crisis (2008), stocks and housing crashed simultaneously
By testing your retirement plan against all of these scenarios, you can build confidence that it will survive whatever the future brings.
What We Use for Backtesting
Backtesting is available for any year range from 1871 through the latest year in our data. You can choose the range on the backtesting page (e.g. 1871-2025 for full Shiller history, or 1928-2025 as the default).
Default 1928-present is common because:
- More complete data: Annual total return data (including dividends) is more reliable from 1928 onward.
- Modern market structure: The S&P 500 was created in 1926, making post-1928 data more representative of today's markets.
- Industry standard: Most retirement studies (including the original 4% rule research) use data from this era.
For 1871-1927, stock returns come from Shiller; bond returns are approximated from the 10-year rate. The full Shiller dataset (1871-present) is also used for CAPE ratio calculations and historical context.
Data Limitations
While Shiller's data is the gold standard, there are some limitations to keep in mind:
- U.S.-centric: The data represents U.S. markets. International markets have had different return profiles (some better, some worse).
- Survivorship bias: The U.S. had one of the best-performing stock markets of the 20th century. Other countries experienced total losses (wars, hyperinflation).
- Past ≠ future: Historical returns do not guarantee future performance. The future could be better or worse than any historical period.