Summary
- Rolling returns measure what investors actually earned across every overlapping multi-year holding period, not just a single start-to-end figure.
- The range of outcomes narrows as the holding period lengthens, but the range does not collapse to zero.
- Some 10-year periods produced negative annualized returns. These cluster around two major bear market eras.
- Retirement income planning requires understanding the distribution of outcomes, not just the long-run average.
What rolling returns measure
A lot of discussions of long-term market performance cite a single number: the annualized return of the S&P 500 over some multi-decade span. That figure is accurate as a summary, but it obscures a some details of what actually happened to investors along the way.
Rolling returns offer a more complete picture. Instead of measuring from a fixed start date to a fixed end date, a rolling return calculates the annualized compound growth rate (CAGR) for every overlapping holding period of a given length within the dataset.
For a 10-year rolling series, each data point answers a specific question: what would an investor who bought at the start of a 10-year window and held through the end have earned per year, on an annualized basis? The chart below shows every such period from 1926 to 2025, with the end year on the x-axis.
The distinction between the arithmetic average and the CAGR matters here. The arithmetic average simply adds all annual returns and divides by the number of years. The CAGR accounts for compounding: a year with a 50% loss requires a subsequent 100% gain just to get back to even. Because any variance in annual returns reduces compounding efficiency, the CAGR for any multi-year period is lower than or equal to its arithmetic average. The gap between the two widens when returns are more volatile.
Rolling return analysis captures this dynamic across every possible entry and exit point in the historical record. That makes it a more honest lens than a single summary statistic.
What the rolling data shows
The interactive chart above defaults to 10-year rolling periods, with controls to switch to 5, 15, or 20 years. The dashed line marks the average CAGR across all periods in the selected range.
Several patterns tend to hold across period lengths:
- The average CAGR for 10-year periods runs well above zero across the full dataset.
- The worst periods cluster in two distinct eras: the 1930s and 1940s (Great Depression and its aftermath) and the stretch ending in the early 2010s (which included both the dot-com collapse and the 2008 financial crisis back to back).
- As the period length increases to 15 or 20 years, the floor on the worst outcomes tends to rise, and the frequency of negative outcomes declines.
- Even with 20-year periods, the historical record does not show a floor above zero for every possible window. The worst 20-year periods still produced modest annualized returns.
The chart also includes a toggle between nominal and inflation-adjusted (real) returns. The nominal view reflects total returns as reported, without adjusting for purchasing power. The real return view adjusts each year's return using annual Consumer Price Index data (CPI-U) from the U.S. Bureau of Labor Statistics. Over most long periods, real returns have been meaningfully lower than nominal returns, because inflation erodes the purchasing power of investment gains. The gap between the two is most visible during high-inflation decades such as the 1970s, when nominal returns were modest while inflation ran well above recent levels. Switching to the real view in the chart makes this pattern visible across the full historical record.
Data: Robert J. Shiller, shillerdata.com. Total return series including dividends reinvested; real returns CPI-adjusted using Shiller's CPI series. Past performance does not indicate future results.
Why longer periods narrow the range
The intuition behind longer holding periods is straightforward: over more time, a sequence of bad years has more good years to offset it. But the mechanism is more precise than that description suggests.
Compounding means that each year's return builds on the cumulative base established by all prior years. A 40% loss in one year eliminates not just that year's gains, but a substantial portion of the gains that preceded it. Recovering from that loss takes more than a single good year. It takes sustained positive returns applied to a reduced base.
When holding periods extend to 15 or 20 years, the probability that the entire span coincides with an extended bear market era decreases. There are simply more years over which positive periods can offset negative ones. This is the mathematical basis for the observation that longer holding periods have historically produced positive outcomes more consistently.
The caveat matters, though. Narrowing the range of outcomes is not the same as eliminating the possibility of a negative outcome. Historical data reflects what happened in U.S. large-cap equities over a century that included substantial economic expansion. Future conditions may differ, and other markets around the world have not always followed the same pattern over equivalent time spans.
What the worst periods have in common
Examining the chart, three eras stand out as sources of the worst rolling returns:
The first is the period straddling the early 2000s and the 2008 financial crisis. Investors who held for 10 years ending around 2008 to 2012 experienced the compounded effect of two separate severe bear markets within a single decade. The dot-com bust from 2000 to 2002 was followed by partial recovery, then a sharp decline again in 2008 and 2009. A 10-year window that captured both downturns had limited room for recovery within the period.
The second is the high-inflation stretch from the late 1960s through the early 1980s. A 15-year period ending in August 1982 began around 1967, running through the 1973-1974 bear market and the sustained high inflation of the late 1970s. In real (inflation-adjusted) terms, this stretch was particularly damaging: nominal stock returns were modest while consumer prices rose at 7-12% per year, steadily eroding purchasing power. This era shows up most clearly when switching the chart to real returns; it is less visible in the nominal view, which is part of what makes inflation-driven losses easy to underestimate.
The third is the Great Depression era. Investors who held U.S. equities through the early 1930s experienced the largest calendar-year declines in this dataset. The decade that followed involved further volatility before sustained recovery. Ten-year windows that began in the late 1920s and ended in the late 1930s or early 1940s captured some of the weakest annualized returns in the full record.
These three eras are not typical. They represent the convergence of unusually severe market conditions with unfavorable entry timing. The rolling chart makes this visible: outside these clusters, the historical distribution of 10-year returns has been meaningfully more positive.
What this means for retirement planning
Rolling return analysis has direct implications for retirement income planning, particularly for investors who are transitioning from accumulation to withdrawal.
During accumulation, below-average periods can be partially offset by continuing to invest at lower prices. An investor who adds to their portfolio during a poor 10-year stretch is effectively buying more shares at depressed valuations, which can improve the eventual outcome.
Withdrawal changes the dynamic. An investor drawing income from a portfolio in a poor sequence of early years is selling assets at depressed prices, reducing the base that future recovery applies to. This is the mechanism behind sequence-of-returns risk: it is not simply about average returns over the full retirement period, but about the specific order in which those returns arrive.
The rolling return data illustrates why a planning approach based solely on the long-run average ("the market has returned roughly X% per year historically, so the plan should work") misses an important dimension. The worst 10-year periods in this dataset did occur, and they occurred to investors who had no way of knowing in advance that their entry point coincided with the beginning of an extended downturn.
Coordinating withdrawal sequencing, tax planning, and asset allocation to manage the impact of poor early returns is a central function of retirement income planning, not an edge-case concern.