Do the first five days of January predict the year?
Half-truth. After an up start, the rest of the year beat cash by 10.2 points more than after a down one (15.5 points in the last 20 years). Better odds faded.
After an up first five trading days of January, the rest of the year returns more above T-bills than after a down first five days.It held: +11.9% vs +1.6%; last 20 years, +15.7% vs +0.2%.
After an up first five trading days of January, the rest of the year beats T-bills more often than after a down first five days.Beat cash 78% to 43% before 1973; since, too close to call. Last 20 years: 85% to 71% beat cash, on 7 down starts.
What is the first-five-days belief?
The belief says the first five trading days of January set the tone for the year. A good first week means a good year.
People put it as "As the first five days of January go, so goes the year". It was first popularised around 1973 (an estimate).
The January barometer, which watches the whole month, gets its own test.
What exactly did we test?
We split the years by how their first five trading days went.
A start counts as up if the whole US stock market made money over those five days. Otherwise it counts as down.
Then we measured the rest of the year, from the fifth trading day to the last close of December. Popular tallies use the full calendar year, which includes the five days, so it partly agrees with itself.
From 1927 to 2025, that gives 99 years: 63 up starts and 36 down starts.
Returns are measured above what cash in Treasury bills earned. Gaps between averages are in percentage points.
Is the rest of the year better after a good start?
Yes, on average: +11.9% above cash after an up start, against +1.6% after a down one (+15.7% against +0.2% in the last 20 years).
Chart data
| Period | After an up first five days | After a down first five days |
|---|---|---|
| All years | +11.9% | +1.6% |
| 1927–1972 | +13.0% | -1.2% |
| 1973–2025 | +10.7% | +3.5% |
| Last 20 years | +15.7% | +0.2% |
That's a gap of 10.2 points between up-start and down-start years. Its likely range runs from +2.9 to +17.7 points, clear of zero.
We widened the range because we checked several things at once. The wider range, -1.5 to +21.8 points (99.67%), just includes zero, so luck can't be fully ruled out.
The gap was +14.3 points before 1973 and +7.2 points since, when its likely range, -2.2 to +16.8 points, dipped below zero.
In the last 20 years, the gap was +15.5 points, on just 13 up starts and 7 down ones.
Chart data
| Period | After an up first five days | After a down first five days | All years |
|---|---|---|---|
| All years 63 up, 36 down | +11.9% | +1.6% | +8.2% |
| Before 1973 32 up, 14 down | +13.0% | -1.2% | +8.7% |
| Since 1973 31 up, 22 down | +10.7% | +3.5% | +7.7% |
| Since 1950 47 up, 29 down | +12.2% | +2.5% | +8.5% |
| Last 20 years 13 up, 7 down | +15.7% | +0.2% | +10.3% |
So why MATH and not FADED? Our rule judges the whole span, both halves and the last 20 years, and this claim passed them all. FADED is for claims that fail over the whole span, like the next one.
Does a good start mean a good year more often?
It did before 1973; since then the gap has been too close to call, though it was bigger in the last 20 years.
Over the whole span, the rest of the year beat cash in 76% of up-start years and 58% of down-start years. In all years, it beat cash 70% of the time. That's the base rate, and it's high.
Chart data
| Period | After an up first five days | After a down first five days | All years |
|---|---|---|---|
| All years 63 up, 36 down | 76% | 58% | 70% |
| Before 1973 32 up, 14 down | 78% | 43% | 67% |
| Since 1973 31 up, 22 down | 74% | 68% | 72% |
| Since 1950 47 up, 29 down | 79% | 59% | 71% |
| Last 20 years 13 up, 7 down | 85% | 71% | 80% |
The gap in the share beating cash was 17.9 points, with a likely range of -0.8 to +36.9 points. That includes zero.
Before 1973, the split was big: 78% of up-start years beat cash, against 43% of down-start years. That's a gap in the share beating cash of +35.3 points, on only 32 up and 14 down starts.
Its likely range, +5.4 to +63.8 points, is wide. We widened the range because we checked several things at once. The wider range, -8.2 to +76.3 points, includes zero.
Since 1973, 74% of up-start years beat cash, against 68%. That's a gap in the share beating cash of just +6.0 points, with a likely range of -19.2 to +31.2 points.
That's what we call FADED: by our rule, it was there before it became famous, then no longer clear.
In the last 20 years, 85% of up-start years beat cash, against 71% of down-start years. That's a gap of +13.2 points in the share beating cash, bigger than since 1973.
But with only 7 down starts, one year either way moves it a lot.
So what's true here?
A good start has come before better years on average, but it hasn't decided the year.
Most years beat cash anyway: 70% over the whole span, and 80% in the last 20 years.
Chart data
| Period | All years 1927 to 2025 | Last 20 years |
|---|---|---|
| Best year, after an up start | +51.1% | +31.8% |
| Worst year, after an up start | -48.3% | -9.4% |
| Best year, after a down start | +35.4% | +20.9% |
| Worst year, after a down start | -35.6% | -34.4% |
Both kinds of start were followed by great years and awful ones. Against cash, the rest of the year ranged from -48.3% to +51.1% after an up start, and from -35.6% to +35.4% after a down one.
In the last 20 years, it ranged from -9.4% to +31.8% after up starts, and from -34.4% to +20.9% after down ones.
Does it hold across industries and other measures?
Mostly, yes: over the whole span, the rest of the year averaged more after an up start in 42 of 43 industries.
For beating cash more often, the count was 40 of 43. Industries all move with the market, so these aren't independent votes.
Chart data
| Period | All years 1927 to 2025 | Last 20 years |
|---|---|---|
| Rest of the year (the test) | +10.2 | +15.5 |
| Full calendar year (popular form) | +14.3 | +19.9 |
| Raw returns (not above T-bills) | +9.3 | +15.1 |
| Average industry (equal-weighted) | +11.2 | +15.4 |
Counting the full calendar year, as popular tallies do, makes the gap bigger: +14.3 points over the whole span and +19.9 points in the last 20 years. Part of that is the five days counting toward their own result.
Is the rest of January weak after a bad start?
No: over the century it did, if anything, better after a down start, and the last 20 years leaned the other way. Neither gap is clear.
The window runs from the close of January's fifth trading day to January's last close. Januaries from 1927 to 2026 give 64 up starts and 36 down starts.
Chart data
| Period | After an up start | After a down start |
|---|---|---|
| Rest of January all years | +0.25% | +1.31% |
| Rest of January last 20 years | +0.27% | -1.01% |
Over the whole span, the rest of January averaged +1.31% above cash after a down start and +0.25% after an up one. All Januaries averaged +0.63%.
Up starts minus down starts was -1.06 points, with a likely range of -2.84 to +0.66 points: no clear difference.
In the last 20 years, it was -1.01% after a down start and +0.27% after an up one (-0.18% in all Januaries). That gap, up minus down, +1.28 points, has a likely range of -1.52 to +3.94 points.
Why isn't that clear? A single January swings a lot, and there were just 7 down starts in the last 20 years, so it could easily be luck.
It beat cash in 59% of up-start Januaries and 58% of down-start ones (62% and 43% in the last 20 years).
So the first week said little about the rest of its own month, even though up starts came before better years on average.
Why might it seem to work?
Nobody knows for sure why a good start came before better years on average.
One idea: a strong first week reflects good news that keeps working. We didn't test that.
And memory helps: a bad first week before a crash makes a great story, while one before a fine year is soon forgotten.
What does this mean for you?
This is history, not a forecast, and plenty of years went the other way.
A half-truth doesn't mean doing the opposite. A down start is no signal to sell. The rest of the year still beat cash 58% of the time, and 71% in the last 20 years.
Trading on it grew a bit less each year over the century. The rule: cash for the first five days, then stocks for the rest of the year, only after an up start.
It grew +9.2% a year in total return, against +9.9% for staying invested.
In the last 20 years it edged ahead: +11.2% a year in total return, against +11.1%. At a quarter-percent cost per switch (a stress test), its total return was +10.9% a year, below staying invested.
A lead that thin is no reason to trade on it.
Its worst fall was smaller, -72.5% against -83.9%. In the last 20 years, it was -34.3%, against -54.7%.
Part of that is built in: it was in stocks only 63% of the time (64% in the last 20 years).
Enjoy January's first week. It's a mood, not a map.
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 start is up when the market's total return over the first five trading days is above zero. The years run from 1927 to 2025; 2026 is out of the sample, as our data end in August 2026.
Returns are compounded 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.
Other checks: with raw returns instead of returns above cash, the average gap was +9.3 points (+15.1 points in the last 20 years). For the average industry, weighted equally, it was +11.2 points (+15.4 points).
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 average gap's range went from +2.9 to +17.7 points to -1.5 to +21.8 points, the beat-cash gap's from -0.8 to +36.9 points to -11.1 to +48.0 points. The two questions use the same years, so they are closely related.
The five days overlap the end of the Santa Claus rally window and the January days of our small-stocks and window-dressing questions.
The rest-of-January window sits inside the one in our Santa Claus rally article, which starts after January's second trading day. It also sits inside the January of our small-stocks question.
The rest-of-January check adds January 2026, which is complete in our data, so it has one more up start than the main test.
What else did we test?
See all the calendar effects we tested, or how we test a myth.