September Really Is the Worst Month: What the Numbers Say

Every year around now, the same statistic appears. September is historically the worst month for the US stock market. Unlike most market folklore, this one has real numbers behind it.

Since 1928, September stands out as the only month with consistently negative average returns for the S&P 500, averaging a loss of roughly 1%. That is close to a century of data, which makes it difficult to dismiss as coincidence.

But a statistic like this is far more easily misused than used well. This article looks at what it actually means, what it does not mean, and how to think about it sensibly - particularly in a September like 2026, where there is plenty else going on.

What the number actually says

An average monthly return of about -1% across nearly a hundred years does not mean the market falls 1% every September. It means that when you add up all the Septembers - some strongly positive, many negative, a few disastrous - and divide by the number of years, you land slightly below zero.

Plenty of individual Septembers have been excellent. The average is dragged down partly by a small number of very bad ones. That is an important distinction, because it changes what you can do with the information.

If September were reliably down 1%, it would be a trading strategy. It is not. It is a mild tilt in a distribution with enormous variation.

Why it might happen

Several explanations get offered. None is conclusive, which is itself worth noting.

Return from the summer. Trading volumes are thin over the northern hemisphere summer. When institutional desks come back to full strength in September, portfolios get reviewed and repositioned. Decisions deferred over the summer get made.

Fund year-ends. Many mutual funds have an October financial year-end, which encourages selling of losing positions in September for tax and reporting reasons.

Heavy event calendar. September typically brings a dense run of central bank meetings and economic data after the quieter summer. More scheduled events mean more opportunities for the market to be disappointed.

It might partly be noise. With twelve months, one of them has to be worst. The honest position is that some of this pattern could be a statistical artefact that persists because people keep writing articles about it.

The self-fulfilling problem

Here is the awkward part. If enough people believe September is dangerous and reduce their exposure accordingly, that selling itself creates weakness.

Equally, if the pattern becomes universally known, traders may position for it in advance, which can neutralise or even reverse it. This is the standard fate of well-known market patterns: they get arbitraged away, or they persist only weakly.

Treating a widely publicised seasonal statistic as a reliable edge is usually a mistake for exactly this reason.

September 2026 specifically

This year the seasonal statistic arrives alongside genuine, identifiable reasons for caution. That combination is more interesting than the seasonality on its own.

• A Federal Reserve decision that markets rate as close to a coin flip, with hike odds hovering around 50%.
• An inflation report on 11 September that will heavily influence that decision.
• Oil around $91 a barrel after a near 9% weekly surge on Middle East tension.
• The ten-year Treasury yield near multi-year highs around 4.78%.
• An equity market up 7.7% year to date and heavily dependent on a single theme.

You do not need a hundred-year statistic to justify caution in that environment. The real reasons are sitting in front of you.

That is generally the right way to use seasonality: as a mild supporting consideration, never as the reason itself.

How to use it without being fooled

Do not sell everything. Exiting the market every September and buying back in October is a strategy that would have cost investors dearly across many individual years, quite apart from the transaction costs and tax consequences.

Do consider sizing. If you are an active trader, a period with a genuinely heavier event calendar is a reasonable time to carry slightly smaller positions. Not because of the statistic, but because of the events.

Do expect wider ranges. More scheduled catalysts tend to mean larger daily moves. If your stop distances and position sizes were calibrated in a calm August, they may be wrong for a busy September.

Do not confuse a tendency with a forecast. "September is often weak" and "September will be weak" are entirely different claims. Only the first is supported by the data.

The psychology angle

There is a subtler risk in seasonal statistics, which is what they do to your judgement.

If you go into a month expecting weakness, you will find it. Every down day confirms the theory; every up day is dismissed as temporary. This is confirmation bias, and a widely known seasonal pattern is an efficient way to install it in yourself.

Traders who convince themselves the market "should" fall have a habit of fighting rallies, holding short positions too long, and ignoring evidence in front of them. The statistic did not cost them money. Their certainty about it did.

If you want a practical antidote, writing down what you expect at the start of the month and reviewing it honestly at the end is remarkably effective. Our guide to journaling through a volatile month covers how to do that well.

The honest conclusion

September's reputation is earned, in the narrow sense that the long-run average really is negative and really is the worst of the twelve months. That is a genuine fact and not a myth.

It is also close to useless as a standalone trading signal, because the variation around that average is enormous, the effect is small, and the pattern is known by everybody.

Use it as a nudge toward attentiveness rather than a prediction. In 2026, the things that will actually determine how this month goes are an inflation report, a central bank decision and a shipping lane in the Middle East. The calendar page is the least of it.

What the distribution actually looks like

Averages hide almost everything interesting. If you break the historical September data apart, a more useful picture emerges.

The proportion of Septembers that finish positive is not far below half. This is not a month that reliably falls; it is a month that falls slightly more often than it rises, and when it falls it has occasionally fallen very hard.

Those few severe declines do a lot of work in the average. Several of the most damaging months in market history happened to fall in September or early October, and their inclusion drags the mean down well below the typical experience.

The practical consequence is that the median September - the middling one, ignoring extremes - is much closer to flat than the widely quoted average of -1% suggests. What September really offers is not a reliably negative return but a modestly fatter tail on the downside.

That is a genuinely different claim, and it points toward a different response. If the risk is not "prices probably fall" but "the range of outcomes is wider", then the sensible adjustment is about position size and risk control rather than about direction.

Other seasonal patterns worth knowing

September is the famous one, but a handful of other calendar effects get discussed. All deserve the same scepticism.

• The turn of the month. The last day or two and first few days of a month have historically shown a mild positive tendency, often attributed to regular pension and salary-driven inflows.
• Late-year strength. The final weeks of the year have a reputation for drifting higher on thin volume. It is real in the data and small in size.
• Summer quiet. Volumes genuinely do fall over the northern summer, which can mean exaggerated moves on little news rather than an absence of moves.
• Monday and Friday effects. Widely claimed, poorly supported, and mostly noise once transaction costs are considered.

The common thread is that these effects are small, unstable across decades, and nowhere near large enough to build a strategy on once costs are included.

How professionals actually use seasonality

Very few serious traders position on seasonality alone. Where it gets used, it is as a tiebreaker.

If a trader has two roughly equal setups and one aligns with a seasonal tendency while the other fights it, the seasonal consideration might tip the decision. It adjusts conviction at the margin. It does not generate the trade.

The second legitimate use is in risk planning rather than direction. Knowing that a period has historically produced wider ranges is a reasonable input into how much you are willing to have at risk, independent of any view on which way prices go.

What experienced traders almost never do is enter a position because of the calendar. The reason is straightforward: a tendency measured across a century tells you nothing reliable about a single sample from that distribution, and one month is a single sample.

A note on data mining

There is a deeper reason to be careful with any pattern of this kind, and it applies well beyond seasonality.

If you search a century of market data for patterns, you will find them. Some will be real. Many will be coincidences that look convincing precisely because you went looking. With enough combinations of month, day, week and starting year, striking-looking relationships appear by chance alone.

The month-of-year effect has one thing in its favour: it was not discovered by a computer trawling millions of combinations. It has been observed and discussed for decades, has plausible mechanical explanations in fund year-ends and post-summer repositioning, and has persisted across very different market structures.

That makes it more credible than most calendar claims. It still does not make it tradeable, because credibility and profitability are separate questions. A pattern can be genuinely present in the data and still too small, too unreliable and too well known to exploit after costs.

The one thing worth doing this month

If the seasonal statistic prompts a single practical action, make it this: review your position sizes against current volatility rather than the volatility you were used to.

Ranges in early September 2026 have been wider than the summer average, driven by oil, bond yields and a genuinely uncertain central bank decision. A position that risked a comfortable amount in August may be risking substantially more today without you having changed a single setting.

That adjustment has nothing to do with the calendar and everything to do with observable conditions. It is the response the seasonal statistic points toward, arrived at for better reasons.

This article is general information, not financial advice. Do your own research or speak to a licensed professional before making money decisions.