Every “year two of a presidential term” since 1986, rebuilt from actual month-end closes — combined with the latest macro data into a judgment for Q4 and the year after the midterms.
中文版 →If you read one screen, read this one.
Up +21.2% year to date, with a record close of $747.46 on Sep 22. September is usually weak; this one gained 3.9%. In the seven sessions after the Fed hiked on Sep 16, QQQ rose 5.6%.
AI capex is still accelerating, but the 10-year yields 5.17% (highest since 2007), the Fed has resumed hiking, and one-year inflation expectations have risen to 4.6%. Only 1 of 9 macro indicators is a tailwind.
Base case (55%): a ~5% pullback during October's event cluster, recovery after the election, and a year-end around +20% (≈$737) — roughly flat from here. Bull 20%, bear 25%: the downside tail is fatter than the upside.
In the 12 months after the end of September of a midterm year, the index rose 7 times out of 7 (median +26.5%). But the three starts from a 12-month high returned only +1.6% to +7.0% — and 2026 is sitting at its high.
Nine indicators, each scored as a tailwind or headwind for the Nasdaq-100: 1 tailwind, 4 headwinds, 4 neutral. The market is walking on one leg — AI earnings.
A 25bp hike on Sep 16 — the first since 2023, by a 12–0 vote. Markets put ~75% odds on another at the Oct 27–28 meeting; the dot plot's 2027 median rose to 4.1%, so no cuts next year either.
Highest since 2007, up from about 4.15% at the start of the year; the 30-year is at 5.46%, the highest since 2004. A global bond selloff is enlarging the denominator in every valuation.
July data (August is due Sep 30); August CPI was 3.35% year over year. One-year inflation expectations are 4.6%, versus 3.4% in February before the war.
Briefly hit $109 in early September; fell Friday as the U.S. and Iran explored a phased reopening of the Strait of Hormuz. A deal is the biggest upside trigger; a collapse revives the stagflation trade.
August crushed expectations (~53K), with wages up 3.1% year over year. No recession signal — which is exactly what gives the Fed room to keep hiking.
Michigan's September final: 15% below January, with the expectations index down about 10% in one month. High fuel prices and re-escalating trade disputes are eroding the outlook.
The big five cloud spenders are on track for about $800B in 2026; Nvidia projects roughly $1.3T in 2027. Amazon raised this year's capex to about $220B. Meta rose ~17% in a week after launching its Muse AI agent.
A stopgap signed Sep 2 removed pre-election shutdown risk — but only pushes the fight into December, the same week as the December FOMC.
The index is hugging its highs with low volatility: the market is priced for a smooth path and has a thin cushion for surprises.
The core tension: stocks and bonds are telling two different stories. QQQ is at its highs with the VIX near 15, while the 10-year Treasury yield is at its highest since 2007. Equities are pricing AI earnings; bonds are pricing inflation, deficits and renewed tightening. The last time a midterm year combined record stock prices, Fed hikes and a breakout in long yields was 2018 — and QQQ fell 17% that fourth quarter.
Compared with 2018, the accelerator is stronger this time — and so is the brake. The accelerator is AI: a capex cycle still gathering pace, an earnings engine 2018 didn't have. And if oil falls on a Hormuz reopening, it would pull down inflation expectations and long yields at the same time. The brake is inflation: in late 2018 inflation was near 2%, so the Fed could pivot to calm markets within weeks. With core PCE at 3.3% and expectations at 4.6%, the bar for a rescue pivot is far higher now — and long yields above 5% sit well above 2018's peak of about 3.2%.
Conclusion: cautious-neutral for Q4 — expected return near zero, with a fatter downside tail. The historical median Q4 for midterm years is +5.4% (6 of 8 up); that's the seasonal prior. Starting from highs, in a hiking cycle, with yields at new highs pushes it down to roughly flat. This isn't a bearish call: AI earnings and the post-election removal of uncertainty still provide a floor. From this starting level, the risk/reward is simply no longer skewed upward.
Hormuz reopens and Brent drops below $90 → inflation expectations and long yields retreat; October megacap earnings keep raising AI capex; the Fed hikes in October and signals it's done.
Sep 30 PCE, Oct 2 payrolls, mid-October CPI, the Oct 27–28 FOMC and megacap earnings all land within a few weeks; the index pulls back ~5%. After Nov 3 uncertainty clears and the year ends near today's level.
The 10-year holds above 5.25%, oil returns to $110, and the Fed signals several more hikes. Multiples compress; AI earnings can't offset the higher discount rate.
Judgments go wrong, so here's what would change this one. If any of these trigger, the scenario probabilities get reshuffled.
| Indicator | Now | Bullish signal | Bearish signal |
|---|---|---|---|
| 10-year Treasury yield | 5.17% | Falls below 4.90% | Holds above 5.25% |
| Brent crude | ≈ $104 | Below $90 (strait reopens) | Back above $110 |
| Oct 28 FOMC | Hike odds ≈75% | Hikes but signals a pause | Signals several more hikes |
| Core inflation (Sep 30 PCE, mid-Oct CPI) | Core PCE 3.3% | Core m/m ≤ 0.2% | Core m/m ≥ 0.4% |
| Megacap earnings (late October) | — | AI capex guidance raised again | Guidance cut, or “slowdown” talk spreads |
| 1-year inflation expectations (Oct 9 prelim) | 4.6% | Back below 4% | Above 5% |
The most-quoted midterm pattern is that stocks rise in the year after the election. Tested on real data: the pattern holds, but the size depends on the starting point.
| Year | End-Sep distance from 12-month high | Next 12 months |
|---|---|---|
| 2002 | −46.7% | +56.5% |
| 2022 | −32.8% | +34.1% |
| 1994 † | −4.9% | +48.6% |
| 2006 | −3.2% | +26.5% |
| 2014 | −1.0% | +3.0% |
| 2018 | −0.5% | +1.6% |
| 2010 | −0.3% | +7.0% |
| 2026 | 0.0% (record) | ? |
Up 7 times out of 7, median +26.5%. But split it: the 4 years that ended September more than 3% below their high averaged +41%; the 3 years within 1% of their high averaged just +3.9%. The implication: a meaningful Q4 pullback would actually improve the odds for 2027; without one, history points to single-digit returns. With only 7 observations, treat this as a reference, not a law.
QQQ (2002–2022) and NDX (1994) month-end closes. 1998, 1990 and 1986 lack verifiable following-year month-end data and are excluded.
Three lenses, three answers.
Smallest Jan–Sep path error (sum of squared errors 467; next is 2014 at 690). Both ended September at record highs, up about 20% on the year; same president, tariffs and U.S.–China friction, a hiking Fed, a 10-year breakout. 2018's fourth quarter: −17.0%.
A momentum rally powered by a tech super-cycle (the internet then, AI now), interrupted by an external shock (the Asian crisis, LTCM) before making new highs. But the Fed cut three times that fall — the opposite of this year.
A midterm year with aggressive Fed hikes (250bp over the year) and the “bond massacre.” The Nasdaq-100 chopped around and finished roughly flat.
Put together: AI sets the ceiling; rates decide how close the market gets to it. As long as the Fed is hiking and long yields are making new highs, 2018/1994-style “choppy at the highs” is a better base-case reference than a 1998-style melt-up.
Eleven midterm years overlaid on one Jan–Dec axis. 2026, 2018 and 1998 are in color; the rest are gray context lines (hover for each year's values).
The previous two versions' calls, checked against real data.
The June version called for “a July surge, then an August–September slide”; the reverse happened. The Sep 18 full-year target was exceeded two sessions later. The lesson is sharper now: analogs can bracket a range, but they have almost no predictive power over the monthly path or the direction of short-term events. So this version drops the single path in favor of ranges, probabilities and triggers — each to be scored in the next update.
Cumulative monthly change (%, vs. prior year-end close). 2026 Jan–Sep are actual (September = Sep 25 close); Oct–Dec and the full year are the base case (italic).
| Year | Jan | Feb | Mar | Apr | May | Jun | Jul | Aug | Sep | Oct | Nov | Dec | FY |
|---|
≈ estimated monthly path (full-year return from published data); † some months anchored to NDX month-end values, others estimated; * Sep 25 close.
Data: QQQ month-end and daily closes come from Robinhood (split-adjusted prices, excluding dividends), as do the NDX month-end anchors. Macro data comes from the public reports and official releases listed below, as of Sep 25.
On the judgment: The scenarios and probabilities here are reasoning from historical analogs and current macro conditions — not a forecast, not a recommendation, and not investment advice. Section 08 shows how far off the previous two projections were; weigh the method accordingly. The author is not a licensed investment adviser; do your own research and consult a professional before making investment decisions. The midterm election appears here only as a calendar event; this page makes no prediction about the result and takes no political position.
Sources: