The last two weeks of August we had clients asking us whether they should "go defensive" into September. My honest answer: going defensive is fine. Going defensive without a plan is how you turn a temporary wobble into a permanent mistake.

September Is Coming. Here Is What the Numbers Say.

September is the only month of the year with a negative average return over the last 30 years. The average: −0.14%. The median is better (+1.18%), but the average gets dragged by the ugly ones. 2008. 2001. 2002. When September decides to be bad, it is properly bad.

But here is the part that should worry you more. 2026 is a US midterm election year. And midterm years have their own seasonal problem that sits on top of the usual September softness.

Midterm years are historically the weakest year in the presidential cycle

Year 2 of the cycle averages +4.5% for the full year. Year 3 averages +16.1%. The gap is not small. And most of that Year 2 pain concentrates in the September–October window, when political uncertainty peaks and markets price the worst-case scenario for the midterm result.

Over 50 years and twelve midterm cycles, the drawdown from peak to trough in the midterm year averaged −19%. In 2022 it was −25%. In 2018 it was −20%. In 2002 it was −34%. These are not small numbers. And the trough usually lands somewhere between late September and mid-October.

Why This September Feels Different (and Not in a Good Way)

Seasonal weakness is one thing. Seasonal weakness when the market is already stretched is another thing entirely.

Berkshire Hathaway is sitting on cash equal to about 32% of its total assets. That is the highest share in three decades. The most famous allocator on the planet is refusing to pay these prices. You do not have to copy him. But you do have to ask yourself whether your book assumed prices would stay this dear forever.

And let me be blunt about one more thing. If you cannot answer these four questions right now, you are not prepared for what September might do:

What is the money for? When is it needed? How much of it sits in one theme? And what would you be forced to sell if the index drops 10% next month?

If the last one made you uncomfortable, keep reading.

The difference between an investor who survives a September sell-off and one who gets hurt is not talent. It is preparation. Same month, same drawdown, two completely different outcomes. One has a plan. The other has a news app and a rising sense of dread.

Here is the plan.

Not Sure Where to Start?

We built our Prism Portfolio Review for exactly this conversation. Twenty minutes, free, no obligation. We answer the four questions above and help you build the plan before the wobble arrives.

Book a Free Prism Portfolio Review →

Even Buffett Sells His Losers (He Just Wishes He Did It Sooner)

If step one feels hard, you are in good company. The most successful investor alive has made the same mistake three times and been very public about it.

Tesco (2013–2014)

He bought from 2006, added after a profit warning in 2012. By 2013 he had "soured on management" and sold about a quarter of the stake for a small $43 million profit. Then he dawdled on the rest. 2014 brought margin collapse and a £250 million accounting hole. Tesco roughly halved. He dumped the remainder for a $444 million after-tax loss and called the delay "thumb-sucking."

The 2013 sale was the part that stopped the bleeding. The mistake was the slowness.

IBM (sold 2017–2018)

Started buying at about $170 in 2011. Thesis was fortress IT services and buybacks. Then revenue fell for 20 straight quarters. He sold a third in 2017, told CNBC in May 2018 the position was zero. IBM hit $94 in March 2020. He was wrong on the business. The exit still beat holding into the trough.

Paramount Global (sold May 2024)

His words: "It was 100% my decision, and we’ve sold it all, and we’ve lost quite a bit of money." Fully out by May 2024 at about $13. Three months later, a $6 billion cable write-down hit and the stock was in the $9s.

The pattern is always the same. Buffett’s edge is not that he never buys a dog. It is that he will sell the dog once he has revalued it, even after the loss is already on the page. You do not need his talent to follow his process. Look at your bottom three holdings and ask: if this were not in the portfolio, would I buy it today?

What Happens if You Just Sit in Cash and Wait?

We ran a simple rule on the S&P 500: out on the last trading day of August, cash earning nothing, back in on date X. Scan X from early September through late November. Twenty years of data.

The raw "best" re-entry date is around 11 October. Looks clean on a chart. Then you drop 2008 (when that window was −29%) and the edge disappears. The 20-year optimum is a crash harvest. It is not a law.

Midterm years are the only sample where a mid-October buyback survives a median test. Last five midterms: 2006, 2010, 2014, 2018, 2022. The median hole bottoms around 13 October at −5.2%. Three of those five years were down hard into that week. Two were not.

This is the bit most investors get wrong. They sell in September because they are scared. Then they miss the bounce because they are waiting for "clarity." Clarity arrives after the bid. Always.

Why the Bounce Has Real Support This Time

Seasonal patterns are a lean, not a guarantee. Two forces give this particular cycle some concrete reasons to expect the kind of re-acceleration most midterm years deliver in Q4 and into Year 3.

The policy-reaction function (the "Trump Put")

In April 2025 the administration put on broader tariffs than the market had priced. The S&P fell 12% in four sessions. Then came a 90-day pause. Same day: S&P +9.52%, the largest single session since 2008. That is a policy reaction with a timestamp.

It is also proof they will let the tape fall first. Twelve percent in four days is not a cushion. It is a hole you have to be able to sit in while you wait for the reaction.

August 2026 was the bond version. The 30-year tagged a 19-year high near 5.32%. Treasury doubled off-cycle buybacks of 10s and 30s. The long bond dropped 10 basis points that afternoon. Two weeks later it was back where it started. The tool bought a session. It did not rewrite the trend.

Bank of America described it as a put with an unwritten strike. The professional version: the put exists as a reaction function, not a peg. If another air pocket arrives into November, history in this administration says a tool can appear. Use that the way you use a weather forecast. Pack a jumper. Do not cancel the trip.

$3.2 trillion in committed capex is not stopping for a soft September

Amazon, Microsoft, Alphabet, Meta and Oracle spent $449 billion on capex in 2025. Company guidance for 2026 is $805 billion. Street estimates push that to $1,116 billion in 2027 and $1,320 billion in 2028.

Growth rate peaks in 2026. The headline you will see is "capex growth is decelerating." Read that carefully. Going from $805 billion to $1,320 billion is not a slowdown in any language an accountant uses. Data centres that are half-built in September do not stop construction because the S&P has a soft patch. Contracts with TSMC and ASML are measured in years, not quarters.

That capital keeps moving through the economy whether the market is open or shut.

The Calendar After the Pain

One more number, then I will stop.

Over 50 years and twelve midterm cycles, from September month-end of a midterm year the next twelve months averaged +17.0% and were positive in 11 of 12. The November–April window was positive in all 12. Since 1950, buying at end-October in a midterm year and holding 12 months was positive in 19 of 19 cycles.

I am not forecasting +17% from here. I am saying the historical job in September of Year 2 is to arrive in November still holding the assets you actually wanted. Year 3 is where the cycle has done the compounding. Getting to 2027 invested has mattered more than calling this month.

That is the whole article. September might be ugly. It might not. Either way, the investors who do well are the ones who had a plan before the headlines arrived.

If you want help building yours, we are here. Twenty minutes. Free. No obligation.

Book Your Free Prism Portfolio Review →

Frequently Asked Questions

Is September always bad for the stock market?

No. Over the last 30 years the median September return is +1.18% and the month finished positive 60% of the time. The average (−0.14%) is dragged by a handful of genuinely ugly Septembers (2001, 2002, 2008), not by consistent declines every year.

What should I do with my portfolio this September?

Review rather than retreat. Cut low-conviction holdings, trim positions that have run hard with little remaining upside, write a shopping list of stocks and index levels you would buy on a pullback, and stage cash for deployment in the October–November window. The historical midterm pattern rewards investors who arrive in November still invested.

What happens after September in a midterm election year?

Over 50 years (12 midterm cycles), the next 12 months from September month-end averaged +17.0% and were positive in 11 of 12. The November–April window was positive in all 12. Since 1950, buying at end-October in a midterm year and holding 12 months was positive in 19 of 19 cycles.

When should I reinvest if I raise cash?

Both the 20-year and 5-year midterm samples agree: staying in cash through late October and November is the expensive mistake. In midterm years, the median trough lands around mid-October. Over 20 years, sitting out until 30 November cost the hold path a 54% compounded return.

Is the "Trump Put" something I should rely on?

It is a policy-reaction function, not a floor. In April 2025 the S&P fell 12% in four days before the 90-day tariff pause (+9.52% on the reversal). In August 2026 Treasury doubled buybacks when the 30-year hit a 19-year high. The put changes path and timing. It does not retire the risk.