Does window dressing lift stocks at quarter-end?
Myth. Since 1926, a quarter's last five trading days averaged +0.10% above cash, against +0.15% for an ordinary five-day stretch. In the last 20 years: +0.15% against +0.24%.
Stocks do better in the last 5 trading days of a quarter than in an ordinary 5-day stretch.The other way: a bit worse, in both halves, the last 20 years and most industries. Could easily be luck.
Stocks do worse in the first 5 trading days of a quarter than in an ordinary 5-day stretch.The other way: better by 0.37 points, likely range +0.11 to +0.62 points. Mostly in the first half; near even in the last 20 years (+0.07 points); worse in only 1 of 43 industries.
What is window dressing?
Window dressing is the idea that stocks rise in a quarter's last days, as fund managers tidy up their holdings before the quarter's report.
Many funds now report what they own every quarter. So, the story goes, managers buy winners just before the quarter closes, to make the report look sharp.
When the quarter turns, the buying stops and prices give some back. Traders call that part window undressing.
Year-end, the biggest quarter-end of all, is said to be the strongest case. The idea was in market talk by 1980, and an academic paper put the name in its title in 1991.
What exactly did we test?
We tested two claims.
Claim one: the last five trading days of March, June, September and December do better than an ordinary five-day stretch. Claim two: the first five trading days of the new quarter do worse.
Ordinary stretches are the other five-day blocks between those windows. We used the US stock market from July 1926 to August 2026.
That gave us 400 quarter-end windows, 400 quarter-start windows and 4,238 ordinary stretches. The last 20 years hold 80 quarter-end windows, 80 quarter-start windows and 800 ordinary stretches.
We measure every return above what cash in Treasury bills earned. When we compare two averages, the gap is in percentage points, or points for short.
Does the quarter's last week beat an ordinary week?
No: the quarter's last week averaged +0.10% above cash, against +0.15% for an ordinary stretch, and fell further behind in the last 20 years.
Chart data
| Period | Last 5 trading days of a quarter | Ordinary 5-day stretch |
|---|---|---|
| All years | +0.10% | +0.15% |
| First half 1926–1976 | +0.06% | +0.11% |
| Second half 1976–2026 | +0.14% | +0.18% |
| Before 1991 | +0.03% | +0.11% |
| Since 1991 | +0.23% | +0.21% |
| Last 20 years | +0.15% | +0.24% |
That's a gap of -0.04 points, likely range -0.26 to +0.18 points, so it could easily be luck either way. The last week beat cash 58% of the time, about like ordinary stretches at 57%.
In the last 20 years, the last week averaged +0.15%, against +0.24% for an ordinary stretch, a gap of -0.09 points. It beat cash 52% of the time, against 60%.
The gap was -0.05 points in the first half of our data and -0.04 points in the second. Before 1991 it was -0.08 points; since then, +0.02 points. Only 20 of the 43 industries we could test fit the claim.
Is the new quarter's first week worse?
No, it was better, the reverse of the belief, though only just in the last 20 years.
The first five days of a quarter averaged +0.51% above cash, against +0.15% for an ordinary stretch. So much for undressing.
Chart data
| Period | First 5 trading days of a quarter | Ordinary 5-day stretch |
|---|---|---|
| All years | +0.51% | +0.15% |
| First half 1926–1976 | +0.77% | +0.11% |
| Second half 1976–2026 | +0.25% | +0.18% |
| Before 1991 | +0.67% | +0.11% |
| Since 1991 | +0.22% | +0.21% |
| Last 20 years | +0.31% | +0.24% |
That's a gap of 0.37 points the wrong way, likely range +0.11 to +0.62 points. We widened the range because we checked several things at once; then it's -0.01 to +0.74 points, just including zero. The first week beat cash 62% of the time, against 57% for ordinary stretches.
But nearly all of it is old. The gap was +0.66 points in the first half and +0.07 points in the second. Before 1991 it was +0.56 points; since then, just +0.01 points. Over the century, only 1 of 43 industries had a worse first week.
In the last 20 years, the first week averaged +0.31% against +0.24%, a gap of +0.07 points. It beat cash 59% of the time, against 60%.
Chart data
| Period | Quarter's first 5 days, all years since 1926 | Last 20 years |
|---|---|---|
| April (after Q1) | +0.18 | -0.13 |
| July (after Q2) | +0.68 | +0.53 |
| October (after Q3) | +0.17 | -0.30 |
| January (after year-end) | +0.44 | +0.17 |
A myth never means do the opposite. The bars above are reported, not judged.
Is year-end different?
Over the century, yes, but not in the last 20 years.
December's last five trading days averaged +0.95% above cash, against +0.15% for an ordinary stretch.
That's a gap of +0.81 points over 100 year-ends, with a likely range of +0.48 to +1.15 points, well clear of zero.
Chart data
| Period | Quarter's last 5 days, all years since 1926 | Last 20 years |
|---|---|---|
| March (Q1) | -0.29 | +0.31 |
| June (Q2) | -0.11 | -0.16 |
| September (Q3) | -0.58 | -0.54 |
| December (year-end) | +0.81 | +0.01 |
Then it vanished. Over the last 20 year-ends, those days averaged +0.25%, against +0.24%. That's a gap of just +0.01 points, likely range -0.65 to +0.78 points.
The other quarter-ends didn't share it over the century.
Year-end was reported, not judged, so it can't rescue the verdict.
It's also crowded: those days open the Santa Claus rally window, which we test separately. So a strong December finish may owe as much to holiday cheer as to fund managers; we can't tell which.
Is it about quarters, or just month-ends?
Not about quarters: quarter-ends did no better than other month-ends, maybe a touch worse, which could be luck. Their last week averaged +0.10%, against +0.21% for the other eight months.
That gap of -0.11 points has a likely range of -0.37 to +0.16 points. In the last 20 years it was -0.29 points: +0.15% against +0.44%, likely range -0.88 to +0.27 points. Reported, not judged.
Why does it feel true?
Because part of the story is real. Funds do report their holdings, and a big buyer in a hurry can nudge prices.
The original version is about which stocks funds buy, winners in and losers out. That needs data on single stocks; we tested the whole market.
And December, the quarter-end people talk about most, was strong for decades.
What does this mean for you?
This is history, not a forecast.
Holding stocks only in each quarter's last five days, with cash otherwise, grew +3.1% a year in total return, against +9.9% for staying invested.
In the last 20 years, the rule grew +1.7% a year in total return, against +11.5% for staying invested.
It was in stocks only 8% of the time, and 8% in the last 20 years. Its worst fall was smaller, -32.4% against -83.9%. In the last 20 years, its worst fall was -12.8%, against -54.7%.
Both its low return and its smaller worst fall come from sitting in cash most of the time. Neither says much about the window itself.
What's in the small print?
Data: the US total stock market and one-month Treasury bills from Kenneth R. French's data library (Tuck School of Business at Dartmouth), daily from July 1926 to August 2026, and 49 US industry portfolios from the same library.
Method: a quarter-end window runs from the close of the sixth-to-last trading day (five trading days before the quarter's last close) to that last close. A quarter-start window runs from there to the fifth day of the new quarter. Each is compared only with ordinary stretches.
Returns are compounded over the window and, except for the timing example, measured above T-bills. Returns are before taxes. The timing example pays a twentieth of a percent each time it switches. The likely ranges are 95% bootstrap intervals.
Test many beliefs and about 1 in 20 with nothing behind it will pass the usual 95% standard by luck alone. So the main ranges are also shown at 99.67%, which allows for testing 15 calendar beliefs.
At 99.67%, the last week's gap goes from -0.26 to +0.18 points to -0.37 to +0.28 points; the first week's from +0.11 to +0.62 points to -0.01 to +0.74 points, just including zero.
Year-end is not judged. Its widened (99.67%) range is +0.33 to +1.31 points over the century, still above zero, and -0.91 to +1.24 points in the last 20 years.
Overlaps, tested separately: the new year's first week is our whole first-five-days window. July's first week holds the day before Independence Day. That January week beat an ordinary stretch by +0.44 points over the century (likely range -0.02 to +0.91 points) and +0.17 points in the last 20 years (-1.04 to +1.20 points); not judged.
Gaps are worked out before rounding, so they can differ from the shown averages by a hundredth.
Earlier work is mixed. A 2002 study found fund prices marked up at quarter-ends, and a 1980 news item reported the pattern. Others found it in two quarters only, no clear effect, a pull on recent winners, or mainly small stocks.
What else did we test?
See all the calendar effects we tested, or how we test a myth.