11 for 11, and the Second Half That Hasn't Kept Pace

Since 1990, there have been 11 years in which the S&P 500 gained 10% or more during the first half of the calendar year. In every one of those 11 years, the second half also ended in positive territory. The median second-half gain? 9.8%.

That streak is real enough to warrant a second look. The catch is knowing what it actually measures.

The S&P 500 returned 9.6% on a price basis in the first half of 2026 — 10.2% counting dividends — putting 2026 on the edge of the dataset. The 52-week low was 6,317 on March 30. The 52-week high was 7,817 on August 13. As of September 18, the index sat at 7,650, down about 2% from that peak. Year-to-date, roughly 11.6%.

The first-half return masks a whiplash quarter-one to quarter-two transition. The S&P 500 fell 4.3% in the first quarter — its worst since 2022 — as a war in Iran sent Brent crude to nearly $120 a barrel and the 10-year Treasury yield climbed from 4.17% to 4.47%. Then Q2 reversed the damage with a 14.9% gain. The first half was not a steady climb. It was a fall and a catch-up.

On the earnings side, the numbers are hard to ignore. The blended S&P 500 Q2 earnings growth came in at roughly 47% year-over-year, according to FactSet. Revenue growth was 14.1% as of late July, later revised up to 15.0% — the highest since Q4 2021. Eighty-six percent of companies beat EPS estimates. Revenue came in 2.9% above estimates. All eleven S&P 500 sectors reported revenue growth.

Looking under the hood, the earnings headline deserves a decompose. Alphabet reported GAAP EPS of $9.11 against a $2.88 estimate. Amazon reported $5.75 versus $1.82. Both blows were driven by unrealized gains on investments booked as "other income" — Alphabet's was roughly $98 billion, Amazon's about $53 billion. Strip those two out and the blended earnings growth falls to 32.6%. Still the highest since Q3 2021. Still remarkable. But 32.6%, not 47%.

That distinction turned out to matter. In July, the market started questioning how durable the AI-driven earnings boom really was. The Philadelphia Semiconductor Index — up 88% in Q2 — fell 25% from its peak. The S&P 500 Equal Weight index gained 1% in July, outperforming its cap-weighted counterpart by roughly 3 percentage points, marking the equal-weight's best relative performance on record. Value's relative strength against growth hit the 99th percentile going back to 1979. The Nasdaq 100 declined 6.6% for the month. The AI trade was being rethought in real time.

August provided relief. The S&P 500 gained 2.7%, set new highs, and erased most of the July setback. September has been flatter — the index has pulled back roughly 2% from its August peak, with the year-to-date return edging from about 13% to 11.6%.

Here is the arithmetic the streak raises. The second half of 2026 so far — July through mid-September — has produced a net gain of roughly 2% (year-to-date ~11.6% versus 9.6% at the end of the first half). The historical median for a full second half after a 10%+ first half is 9.8%. That is a full six months of performance. We're three months in. The question is not whether the streak has already broken. It's whether the earnings backdrop that fueled the first half can carry through the remaining three months at a pace that matches the historical average.

Analysts project Q3 S&P 500 EPS growth of roughly 29% and full-year 2026 earnings growth of about 30%, per FactSet estimates as of early September. Those are solid numbers. But they're also built on a base that includes one-time investment gains that won't recur in Q3 or Q4. The forward P/E of roughly 20x sits above the 10-year average of 19x.

The 11-for-11 record is a complete count — not a forecast. It tells us that strong first halves have historically been followed by positive second halves. It does not tell us why, and it does not guarantee the 12th entry. What it does establish is a baseline: after a first-half return near 10%, the median second half adds another 10%. The second half of 2026 has three months left to decide whether 2026 joins that census or starts a new one.