What If You Sat Out Every September? The S&P 500 Answer Nobody Likes

September is the S&P 500 worst month, and it is not close. Over the 32 Septembers in our data, the index averaged -0.90% for the month while the average of all months was +0.95%. Every other month of the year has a positive average. September is the only one that does not.

So the obvious question: what if you had just skipped it? Sell at the end of August, sit in cash, buy back on October 1, every year. We ran it. The answer is uncomfortable, and then it gets more interesting.

The Result

We took $10,000 at the end of December 1993 and ran two investors side by side on monthly closing prices for SPY, the S&P 500 ETF, through June 2026. One held the whole time. The other held every month except September, sitting in cash for that one month each year, 32 times. Cash earned nothing, which is deliberately unkind to the seasonal investor.

Sitting out September beat buying and holding by about $110,900 over 32 years. That is the opposite of what we found when we ran sell in May and go away, where the seasonal trade cost you roughly $21,600. One month out beat six months out, and it beat doing nothing.

If we stopped here, this would be an article telling you to sell next month. We are not going to do that, because the next table is the whole story.

Four Septembers Did All of It

The average September is negative. The median September is +0.26%. Those two facts together tell you that the month is not reliably bad, it is occasionally catastrophic. In our window, 17 of 32 Septembers were positive. You would have sat in cash for a majority of the Septembers you avoided.

Here is what happens if you take the worst Septembers and pretend they were ordinary months, leaving everything else untouched:

Four months out of 390. September 2002 fell 10.9%, September 2008 fell 9.9%, September 2022 fell 9.6%, and September 2001 fell 8.5%. Take those four away and the strategy that looked like a free 1.1% a year turns into a strategy that loses to sitting still.

Notice what those four dates have in common. The dot-com bottom, the financial crisis, the 2022 inflation shock, and the month of September 11, 2001. This is not a calendar effect. It is four crises that happened to land in the same month, in a sample that only contains 32 of them.

What Would Have Happened To You In Practice

The backtest is also the most flattering possible version of the trade. Three things it does not charge you for:

Taxes. This is 32 round trips. In a taxable account every profitable exit is a realized gain, and gains you take in year one are gains you no longer compound in years two through 32. The $110,900 edge is pre-tax and the strategy is unusually tax-inefficient, because it forces a sale on a fixed date regardless of your basis.

Cash earning nothing. We were kind here and assumed zero, which understates the strategy slightly. Real short-term rates over this window would have added a little. It does not change the shape of the finding.

You following it for 32 years. This is the real cost. The strategy asks you to sell in a rising market, 17 times out of 32, and watch the index go up without you. Most people who try this quit after the third September that rallies, which means they capture the annoyance and none of the crisis protection that made the numbers work.

The Honest Reading

September's reputation is real in the averages and mostly fake in the experience. The typical September is a slightly positive, unremarkable month. The reason the average is negative is that when markets have broken over the last three decades, an unusual number of those breaks happened to occur in September.

A strategy whose entire advantage rests on four observations is not a strategy, it is a description of four events. If the next 32 years distribute their crises across other months, and there is no mechanism that says they will not, the same rule quietly costs you money instead.

Which is roughly the same conclusion we reached about never panic selling and about the worst times to invest: the calendar is not the thing that determines your outcome. Time in the market is, and the seasonal rules that look profitable in a backtest are usually one crisis cluster wearing a costume.

Method: monthly closing prices for SPY from December 1993 through June 2026, the full span of the price history behind this site. Both investors start with $10,000 at the December 1993 close. The seasonal investor is out of the market for every September and holds every other month; cash earns 0%. Dividends are not reinvested in either path, so the comparison between the two is unaffected. Past performance does not predict future results, and nothing here is investment advice.