SEPTEMBER: THE ONLY MONTH
WITH A LOSING RECORD

MARKET HISTORY · 2026-08-31 · third piece in our seasonality set, after Sell in May and Triple Witching
BY RYAN, FRANK & DILLON · MARKET MAULERS FOUNDERS

Every year around this time your feed fills up with the same chart. September is the worst month, look at the red bar, be afraid. And every year roughly half the people posting it have no idea whether the effect is large enough to trade, or what causes it, or what the actual conditional odds are.

The effect is real. It is also small, noisy, and close to useless by itself. What is genuinely useful is the split underneath it, which almost nobody posts. So here is the whole thing, including the parts that argue against the headline.

Three Sources, Three Different Numbers, One Conclusion

Three respectable shops, overlapping samples, and answers that range from -0.7% to -1.2%. That spread is the first honest thing to notice. When a well-known effect cannot be pinned down to within half a percent, the effect is small relative to the noise around it.

What they agree on is the sign, and that agreement is what makes September unusual. Across roughly a century of data it is the only month on the board with a negative long-run average return. RBC adds a detail worth keeping: nine of the forty worst monthly losses in the history of the index landed in September, more than any other month.

Investing.com adds the shape of it. When September is down, it averages -3.8%. When it is up, it averages +3.2%. So this is not a month that grinds lower. It is a month that is slightly more likely to be a bad one, and bad ones are slightly worse than good ones. That is the entire "September Effect," stated accurately.

The Conditional That Actually Matters

Here is the number worth your time. Split every September by where the S&P 500 was trading relative to its 200-day moving average going in:

That is a 5.5 point swing in expected return and a 45 point swing in hit rate, from one filter. Above trend, September is a slightly better than average month. Below trend, it is a wrecking ball.

Which means the calendar was never the signal. Trend was the signal, and September just happens to be where a weak trend has historically done its worst damage. The month is an amplifier, not a cause.

For what it is worth, the index set a record close on August 12 on a cooler CPI print, wobbled mid-month on rising global bond yields and oil, and is finishing August back near those highs. It goes into September comfortably above trend. On that split, 2026 sits in the favorable bucket.

Why September Is Weak At All

Ranked by how much evidence actually supports them, not by how often you hear them.

• Window dressing (well supported). Institutions clean up portfolios ahead of fiscal-year reporting, and a lot of that housekeeping falls in September and October. Losers get sold because nobody wants them showing up on a statement, not because anyone changed their mind about the company.
• Tax-loss selling (supported, but mistimed). Real, and it does push managers to dump losers. RBC's own caveat is that this pressure typically intensifies later in the year, so it explains part of the autumn, not September specifically.
• Liquidity returning (mechanically sound). The desks that were half-staffed in August come back after Labor Day. Positions that nobody wanted to touch in thin summer markets finally get moved. That is not directional on its own, but it does mean the first real repricing of anything that changed over the summer tends to happen in September.
• Self-fulfilling prophecy (plausible, unfalsifiable). RBC raises it and so do we, with the caveat that this explanation is unfalsifiable by construction. If enough professionals de-risk in September because they read the same chart, the chart keeps working. Treat it as a reason the effect persists rather than a reason it started.
• Summer vacation psychology (folk theory). You will hear that traders come back from holiday in a bad mood and sell. There is no evidence for this. It is a story people tell because the real answers are boring.

The September 2026 Calendar

This particular September is heavier than most, and the events are stacked in the middle of the month.

Two of the three seasonal mechanics above land inside a five-day window this year. Quad witching sits two sessions after an FOMC that carries a dot plot. If you have read our triple witching piece, you already know what the last hour of September 18 tends to look like. Add a fresh rate decision on top of it.

The Case Against Being Bearish On A Calendar

We would rather hand you the counter-argument than let you find it later and wonder why we skipped it.

Records beget records more often than they beget crashes. From 1988 to 2023, the S&P averaged 13.4% over the twelve months following a record high, against 11.9% for all twelve-month periods. Being at highs going into a historically weak month is not, historically, a warning. It is mildly bullish.

And the seasonality itself comes with a warning label from the people who publish it. Investing.com puts it as well as anyone: seasonal data is the climate, not the weather. It tells you what a typical September looks like. It tells you nothing about this one, which has a live Fed decision in it.

How We Actually Handle It

• Do not short a month. The edge in the raw statistic is smaller than your spread over any realistic sample of trades you will personally take.
• Use the trend filter, because that is where the real number lives. Above the 200-day, treat September normally. Below it, tighten everything, because that is the bucket that produced -4.2% and a 15% hit rate.
• Plan around September 15-18, not around September. The FOMC and quad witching are concrete, dated liquidity events. That is a real edge. "It is September" is not.
• Expect the tape to get faster after Labor Day. Whatever range August built in thin conditions, September volume tests it. Size for the volatility you are about to get, not the volatility you just had.
• Watch your own behavior. The most expensive thing about September is not the seasonal drift. It is traders who read the scary chart, size up to "take advantage" of it, and blow the risk budget they need for October.

Key Takeaways

• September is the only month with a negative long-run average return. The effect is real and roughly one percent, which is too small to trade on its own.
• The 200-day filter is the whole story: +1.3% and 60% positive above trend, -4.2% and 15% positive below it.
• Window dressing and returning liquidity are the defensible mechanics. Vacation moods are not.
• September 2026 stacks an FOMC with projections on the 15th and 16th and quad witching on the 18th.
• Trade the dated events. Do not trade the calendar page.

Sources

• Motley Fool · History says September is the worst month for stocks (BofA data since 1928; post-record-high returns)
• RBC Wealth Management · Nothing new about September slides for stock markets
• Investing.com · S&P 500 seasonality shows September as weakest month, but trend matters more
• LPL Research · So long sweet summer: a closer look at September seasonals
• Option Alpha · Triple and quadruple witching dates, 2026 calendar
• Federal Reserve · 2026 FOMC calendar (Sept 15-16, with projections)
• Benzinga · September hike odds after Warsh's Jackson Hole speech
• Market Maulers · Triple Witching: The Four Days a Year the Market Goes Feral (June 2026)
• Market Maulers · Sell in May and Go Away (May 2026)
• Market Maulers · Warsh Said We Have Work To Do (August 2026)