Summer Market Seasonality: What the Patterns Actually Show - Part 1

Few pieces of market folklore have survived as long as "sell in May and go away." It shows up in strategy notes every spring, gets recycled by financial media every June, and gets quietly forgotten every October. What it rarely gets is much examination. That's unfortunate, because there's something real underneath it. Summer markets do behave differently from the rest of the calendar. But the difference doesn't show up where most people look for it, and the version of the pattern that survives careful measurement is narrower, and more useful, than the version in the adage.

Here's the short form of what follows. Average summer returns are unremarkable. What changes in summer is participation, and through participation, the path. Volume thins, ranges widen relative to the flow producing them, and the market's capacity to absorb a surprise falls. That's a different observation from "returns are worse," and it points toward a different set of considerations.

We're writing this in late July 2026, heading into the two months the seasonal record treats least kindly, with oil back above 80 dollars, the Fed's next move genuinely contested, and index leadership about as narrow as it's been in a decade. Conditions like these are exactly when seasonal framing tends to get over-applied. So it's worth being precise about what the record does and doesn't support.

What the record shows

The academic anchor is Bouman and Jacobsen, published in the American Economic Review in 2002. Studying data from January 1970 through August 1998, they found that returns in the November through April window exceeded returns in the May through October window in 36 of the 37 developed and emerging markets they examined. They titled the paper "Another Puzzle" because they could not explain the effect through risk, cross-market correlation, or, outside the United States, the January effect. A later extension of the work by Zhang and Jacobsen, using a far broader historical sample, put the average gap at roughly four percentage points in favor of the winter window.

The finding has been contested since publication. Maberly and Pierce argued in 2004 that the US results were not robust to alternative model specifications, and follow-up work on Japanese equities found the effect concentrated in the period before Nikkei futures were introduced in 1986. Other researchers have replicated it out of sample. The honest summary is that the effect is well documented and still disputed, and that its magnitude depends heavily on the sample start date, the country set, and how outliers get handled.

The monthly data is where the picture gets more useful, because it shows the six month window is doing a lot of averaging. Using S&P 500 monthly data from 1928 through July 2025 compiled by RBC Wealth Management, the average returns look like this: January +1.2%, February -0.1%, March +0.5%, April +1.2%, May +%0.1, June +0.8%, July +1.7%, August +0.7%, September -1.2%, October +0.5%, November +1.0%, and December +1.3%.

Read that carefully, because it doesn't say what the adage says. July is the strongest month in the series. June and August are both positive. The entire negative contribution of the so-called summer window comes from September, and to a lesser extent from October's below-average showing. Selling in May would have meant sitting out the best single month of the year to avoid the worst.

September is the one month with a genuinely consistent record of weakness. Measured from 1928, it averages roughly -1.2% percent and finishes negative about 55%of the time. Measured from 1950, the figure is closer to -0.6% or -0.7%, with 41 negative Septembers against 34 positive ones. Measured from 2000, it worsens to roughly -1.%. Note the spread across those windows. Different vendors and different start dates produce different numbers, and anyone quoting a single precise figure for September seasonality is showing you one slice of a distribution.

The more durable pattern shows up in volume rather than returns. S&P 500 trading volume tends to reach its lowest levels of the year between the week leading into the July fourth holiday and the Labor Day holiday in early September, with a second trough during the year-end holidays. Estimates of the size of the decline vary widely by source and by measurement method, from roughly ten percent to thirty percent below annual averages, which is itself a reason to treat any specific figure with caution. The direction, though, is consistent across data sets and across decades.

Come back tomorrow for Part 2 of this topic!

This material is for informational and educational purposes only and does not constitute investment advice or a recommendation to buy or sell any security. Past performance is not indicative of future results. Historical figures are compiled from third party sources believed to be reliable, and figures may vary between data providers and sample periods. Readers should conduct their own research and consult a qualified professional before making any investment decision.