Average 4-session change: the turn of the month against every other 4-session window
- Last session of the month and first three of the next
- Every other 4-session window (base rate)
Hover or tap a column.
October 20, 2026 · 3 min read
The Turn of the Month in ES and NQ, Checked Against Every Other Week
The last session of a month and the first three of the next averaged +0.18% on ES since 2011. Every other four-session window averaged +0.20%. The turn-of-the-month effect, checked on ES and NQ futures against its base rate.
Stock index gains are said to bunch at the turn of the month: the last trading day and the first few of the next. The effect was documented decades ago on long runs of index data, and it comes round again at every month-end, usually with the explanation that pension contributions and payroll money arrive on a schedule. We measured it on S&P 500 and Nasdaq-100 futures against the base rate it needs: every other four-session window.
The base rate
The window is the classic one: from the close two sessions before month-end to the close of the third session of the new month, four close-to-close changes. The comparison is every four-session window that shares no session with a turn of the month. Windows that cross a contract roll are dropped.
| Window | Market | Windows | Average 4-session change | Closed higher | 95% interval |
|---|---|---|---|---|---|
| Turn of the month | ES | 182 | +0.18% | 56.6% | 49–64% |
| Every other window (base rate) | ES | 2,437 | +0.20% | 60.6% | 59–62% |
| Turn of the month | NQ | 163 | +0.26% | 57.7% | 50–65% |
| Every other window (base rate) | NQ | 2,156 | +0.30% | 59.7% | 58–62% |
On both markets the turn of the month did slightly worse than an ordinary four-session window: +0.18% against +0.20% on ES, +0.26% against +0.30% on NQ. It closed higher less often too, 56.6% of the time against 60.6% on ES. The differences are small and well inside chance. A circular-shift placebo, which slides the month-end marks along the series by every offset of at least 26 sessions, puts the ES difference at p = 0.81 and the NQ difference at p = 0.79.
Measured by day
A stronger form of the claim is that the turn of the month delivers most of the market's gain. On ES, turn-of-month sessions were 18.8% of all sessions and carried 16.2% of the total change. On NQ, 18.9% of sessions carried 16.8%. The average turn-of-month session on ES rose +0.038%, against +0.046% for every other session. In this sample those four days carried a little less than their share, not most of it.
Two eras, two signs
Split the sample in two and the sign changes. In 2011–2018 the ES turn of the month averaged +0.06% against +0.16% for other windows; in 2019–2026 it averaged +0.29% against +0.23%. NQ did the same: +0.06% against +0.30% in 2013–2018, then +0.40% against +0.30% in 2019–2026. The recent era leans the classic way on both markets, by +0.10% on NQ and +0.065% on ES. With 69 to 94 windows per era, every interval overlaps its base rate.
| Market | Era | Turn-of-month windows | Average | Closed higher | 95% interval | Other windows | Average | Closed higher |
|---|---|---|---|---|---|---|---|---|
| ES | 2011–2018 | 88 | +0.06% | 54.5% | 44–65% | 1,207 | +0.16% | 60.6% |
| ES | 2019–2026 | 94 | +0.29% | 58.5% | 48–68% | 1,230 | +0.23% | 60.6% |
| NQ | 2013–2018 | 69 | +0.06% | 52.2% | 41–64% | 925 | +0.30% | 62.3% |
| NQ | 2019–2026 | 94 | +0.40% | 61.7% | 52–71% | 1,231 | +0.30% | 57.8% |
Share of 4-session windows that closed higher, by era, with 95% intervals
- Turn of the month
- Every other window
Hover or tap a row.
What the number does and does not mean
This is one window, fixed before the data was read, on two markets split into two eras: four cells, none of which separates from its base rate. That is a null result, and it is in line with what tends to happen to calendar anomalies once they are widely known. It does not show that month-end flows do not exist. It shows that, net of everything else, they have left no measurable mark on four-session changes in these futures since 2011. The recent era leans the classic way, but eight years of monthly windows is under a hundred cases, too few to tell a revival from noise.
Reading the next turn-of-the-month stat
- Base rate: compare with every other window of the same length, not with zero. Stock index futures closed higher in about six of every ten four-session windows here.
- Window: the last day plus three, or plus two, or plus five? Each choice is another look. Ask which one was fixed before the data was seen.
- Era: a calendar effect that appears only in recent years may be a recent rally in disguise. Ask for both halves of the sample.
Data and method. E-mini S&P 500 (ES) and E-mini Nasdaq-100 (NQ) futures, regular-session closes (4:00 pm ET, 1:00 pm on early-close days), ES from 3 January 2011 and NQ from 2 January 2013, both to 12 October 2026. The turn-of-the-month window runs from the close of the second-to-last session of a month to the close of the third session of the next: four close-to-close changes. The base rate is every 4-session window that shares no session with a turn-of-the-month window. Changes across a contract roll are left out, and so is any window containing one. Intervals are 95% Wilson intervals. The shift placebo slides the turn-of-the-month marks along the series of valid windows by every offset of at least 26 sessions; p is the share of offsets producing a difference at least as large.