The third year of a US presidential term has been the strongest of the four for the S&P 500. Over the full calendar years 1928 to 2025, the index averaged a 14.0% price gain in year 3 (median 18.1%) and finished higher in 19 of 24 year-3s (79%), against an 8.1% average across all 98 years. The catch is the sample: 24 years is small, and the two most recent misses, 2011 and 2015, came after the original version of this page was written in 2010.
Every number below was calculated from daily S&P 500 closes from Yahoo Finance. You can see how it fits with other calendar patterns on the Market Stats hub.
How the term years are counted
Year 1 is the inauguration year, so the counting starts at 1929 (Hoover) and runs every four years to 2025. That makes year 2 the midterm year, year 3 the pre-election year (1931, 1935 … 2023) and year 4 the election year (1928, 1932 … 2024). A calendar-year return runs from the last close of one year to the last close of the next. Inauguration moved from March 4 to January 20 in 1937, per the National Archives, which doesn’t change which calendar year counts as year 1.
The cycle follows the election calendar, not the person. When a president died or resigned mid-term (1945, 1963, 1974), the count kept going.
Year 3 vs the other three years
Year 3 leads on every measure in the table: average, median and share of up years. Year 2, the midterm year, trails on all three.
| Term year | n | Average | Median | % positive | Worst | Best |
|---|---|---|---|---|---|---|
| Year 1 (post-election) | 25 | 7.0% | 9.1% | 60% | -38.6% (1937) | 44.1% (1933) |
| Year 2 (midterm) | 24 | 3.3% | 0.6% | 54% | -29.7% (1974) | 45.0% (1954) |
| Year 3 (pre-election) | 24 | 14.0% | 18.1% | 79% | -47.1% (1931) | 41.4% (1935) |
| Year 4 (election) | 25 | 8.1% | 11.8% | 76% | -38.5% (2008) | 37.9% (1928) |
| All years | 98 | 8.1% | 11.6% | 67% | -47.1% | 45.0% |
Year 3 also holds the single worst year in the whole table, 1931’s 47.1% drop in the depths of the Depression. Its spread of outcomes (a standard deviation of 17.9 points) is about the same as every other year’s (19.0), so the higher average didn’t come with smaller swings.
Most published cycle studies start after World War II. Starting in 1950 makes year 3 look even better:
| Term year | n | Average | Median | % positive |
|---|---|---|---|---|
| Year 1 (post-election) | 19 | 8.4% | 9.1% | 63% |
| Year 2 (midterm) | 19 | 4.6% | 1.1% | 58% |
| Year 3 (pre-election) | 19 | 17.2% | 18.9% | 89% |
| Year 4 (election) | 19 | 8.1% | 11.8% | 84% |
Picking 1950 drops the 1931 and 1939 losses, which is part of why the number improves. Choosing the start date is one of the easiest ways to make a small-sample pattern look stronger than it is.
Every year 3, 1931-2023
Here is each pre-election year on its own. The low and high columns show the lowest and highest close of the year measured against the prior year-end close, so you can see how much ground the index gave up along the way. SPY’s total return (dividends included) is shown where SPY existed for the whole year.
| Year | S&P 500 price return | Lowest close vs prior year-end | Highest close vs prior year-end | SPY total return |
|---|---|---|---|---|
| 1931 | -47.07% | -49.67% | 18.45% | |
| 1935 | 41.37% | -15.16% | 41.68% | |
| 1939 | -5.18% | -20.70% | 0.68% | |
| 1943 | 19.45% | 1.23% | 29.38% | |
| 1947 | 0.00% | -10.00% | 5.49% | |
| 1951 | 16.35% | 1.27% | 16.74% | |
| 1955 | 26.40% | -3.89% | 28.99% | |
| 1959 | 8.48% | -2.95% | 9.96% | |
| 1963 | 18.89% | -0.65% | 18.89% | |
| 1967 | 20.09% | 0.06% | 21.49% | |
| 1971 | 10.79% | -2.16% | 13.70% | |
| 1975 | 31.55% | 2.16% | 39.45% | |
| 1979 | 12.31% | 0.02% | 15.77% | |
| 1983 | 17.27% | -1.64% | 22.76% | |
| 1987 | 2.03% | -7.54% | 39.06% | |
| 1991 | 26.31% | -5.67% | 26.31% | |
| 1995 | 34.11% | -0.03% | 35.36% | 38.05% |
| 1999 | 19.53% | -1.39% | 19.53% | 20.39% |
| 2003 | 26.38% | -8.99% | 26.38% | 28.18% |
| 2007 | 3.53% | -3.12% | 10.35% | 5.15% |
| 2011 | -0.00% | -12.60% | 8.43% | 1.89% |
| 2015 | -0.73% | -9.29% | 3.49% | 1.23% |
| 2019 | 28.88% | -2.35% | 29.25% | 31.22% |
| 2023 | 24.23% | -0.82% | 24.58% | 26.18% |
A few things stand out. 1947 ended exactly flat (15.30 to 15.30). 2011 closed at 1,257.60 against 1,257.64 a year earlier, a loss of 0.003%. Those two, plus 2015, are the “flat” year-3s that kept the price record from being cleaner. Only four year-3s (1931, 1935, 1939 and 2011) saw a close more than 10% below the prior year-end, and 1987’s 39% peak gain shrank to 2% after the October crash.
How much should you trust 24 years?
Less than the headline suggests. The pattern is real in the data, but it rests on very few independent observations, and each year moves the ”% positive” figure by about 4 points.
Three checks from the script:
- Shuffle test. Drawing 24 years at random from the 98, 100,000 times, 3.9% of the draws averaged 14.0% or more. That would look notable on its own, but four term years were compared and the best one was chosen after looking, which makes a figure like this easier to hit by luck.
- Drop the best year. Taking out 1935’s 41.4% leaves a year-3 average of 12.8%, still above the 8.1% all-year mean. No single year makes the result.
- Drop the Depression. Year-3s from 1943 to 2023 (n = 21) averaged 16.5%, median 18.9%, up in 86% of years.
The usual story for why year 3 is strong (administrations front-load unpopular moves and push for growth before the election) is plausible but not something these numbers can prove. Since the 2010 post, the record has also looked less one-sided: two of the four pre-election years since then (2011 and 2015) were flat to slightly down on a price basis, while 2019 and 2023 were strong.
2027 is the next year 3
By this count 2025 was year 1, 2026 is year 2, and 2027 is the next pre-election year. This page doesn’t forecast it. The history tells you how often past year-3s rose and by how much, not what the next one will do, and 24 data points spread across very different markets (the Depression, the post-war boom, the 1970s inflation, the 2010s) leave a wide range.
If you look at the cycle alongside other calendar effects, the same small-sample caution applies to December’s seasonal record and to what followed 52-week highs in December. All three are descriptions of the past, not trade signals.
