Does the S&P 500 always rise in the year after US midterm elections? We tested it

The US midterm elections were on Tuesday 3 November 2026. A popular claim says the S&P 500 has risen in the 12 months after every midterm election since the 1940s. We tested it against the price history.

The short answer: the claim is true. All 20 windows since 1946 ended higher. But the market also rises over most 12-month stretches, so the useful question is how much better these windows did than a random one. We ran our luck check on that.

The data

We used daily closing prices for the S&P 500 index (symbol ^GSPC) from the Yahoo Finance chart API, downloaded on 6 November 2026. This is price only. It leaves out dividends. The file, our script and a note on how to download it again are in the Tickfloor repository under scripts/news/midterms-1106/.

For each midterm election day since 1946 we took the close that day and compared it with the close 3, 6 and 12 months later. The New York Stock Exchange was shut on election day from 1946 to 1966, so for those six years we used the last close before the vote. The 20 elections are 1946 to 2022. The 2026 election is too recent to have a result yet.

Results for each election

• 12 months: up in 20 of 20. The average was +14.4% and the middle result was +13.7%. The weakest was 1946 at +0.1% and the strongest was 1954 at +33.2%.
• 6 months: up in 19 of 20 (the exception was 1946, at -3.1%). The average was +13.3%.
• 3 months: up in 18 of 20 (the exceptions were 2002 at -7.8% and 2018 at -0.9%). The average was +7.5%.

Midterm windows are four years apart, so they do not overlap each other.

Compared with a normal year

Between 5 November 1946 and the latest date with a full 12 months behind it, there are 19,871 trading days to start from. Counting each day as a start, the S&P 500 was higher 12 months later 74.5% of the time. The average 12-month change was +9.4%.

So a rise in any given year is the usual outcome, not a surprise. If every window had a 74.5% chance of being up, getting 20 in a row would happen about 0.3% of the time (a binomial test, p = 0.003). That is unlikely to be pure chance, but this test only looks at up or down, not at how much.

A test on the size of the gain is less strong. We drew 20 random start days from the 19,871 and took the average 12-month change, 200,000 times. In 7.4% of those draws the average matched or beat the +14.4% from the midterm windows (permutation test, p = 0.074).

We ran the same test at three horizons (3, 6 and 12 months) and applied the Benjamini-Hochberg luck check. The corrected value (q) for 12 months is 0.074. The 3-month and 6-month values came out below 0.001 in the permutation test. Our bar is q below 0.1, so all three clear it. The 12-month result clears it narrowly.

One more comparison. We took the same calendar start (the Tuesday after the first Monday in November) in the 58 years that were not midterm years. The S&P 500 was up 12 months later in 39 of 58 (67%) and averaged +7.1%. Midterm windows did better than that.

What this does and does not show

• Twenty windows is a small sample. One bad result would have changed the headline from "every time" to "almost every time".
• The 12-month result is on the edge of our bar (q = 0.074 against 0.1). The three horizons we tested are not independent of each other, because the 12-month window contains the 3-month and 6-month ones.
• The test cannot say the election caused the gains. Midterm years are one spot in the four-year US presidential cycle, and we did not try to separate the vote from the rest of that cycle. Interest rates and company profits also moved in each of those years.
• This is price only. Dividends would add a little to every figure, for midterm windows and normal ones alike.

Where the market is now

Our file has the S&P 500 at 7,568.92 on the close of 3 November 2026, the election day. The latest row in the file, 5 November 2026, shows 7,353.73, which is 2.8% below that. Two days say very little. The 12-month window from the 2026 election runs to 3 November 2027, and nothing in the history we tested tells us how it will end.

We apply the same standard to every claim we test: write the rule down first, compare it with a fair baseline, then correct for how many things we tried. Our full method is on the luck check page and every result so far is on the evidence page.

Source: Yahoo Finance chart API, ^GSPC daily close, downloaded 6 November 2026. Figures are our own arithmetic on that file. Past data, not a promise: it describes what happened in 20 earlier windows and says nothing certain about the next one.