Why the Number One Turns Up First So Often

In the days before calculators, if you wanted to do serious arithmetic, you used a book of logarithm tables. Page after page of numbers, arranged in order, starting from one.
In 1881, Simon Newcomb noticed something odd about these books. The pages weren't wearing out evenly. The early pages, the ones covering numbers that started with one, were much more worn than the rest. The further you went into the book, the cleaner the pages got.
That's strange. It suggests that people were looking up numbers beginning with one far more often than numbers beginning with nine. Which raises an obvious question. Why would the world have more numbers starting with one?
The pattern
Newcomb's observation was later taken up by Frank Benford, who in 1938 tested it on data from twenty different areas. The pattern held up, and it now carries his name: Benford's law.
Here's what it says. In many real-world collections of numbers, the first digit isn't evenly spread. A one comes first about 30 per cent of the time. A two, about 18 per cent. A three, about 12.5 per cent. And so on, falling steadily all the way down to nine, which leads off only about 4.6 per cent of the time.
If every digit got a fair turn, each one would lead about 11 per cent of the time. Instead, one is about six and a half times as common as nine.
Why ones linger
So what's going on? The key is in how a lot of real-world quantities grow.
Many things don't grow by adding the same amount each year. They grow by multiplying. A town that grows by a certain percentage each year, or a bank balance earning interest, or a population of bacteria, gets bigger by a proportion of whatever it already is.
Now think about what that means for the first digit. Imagine a town with a thousand people, growing at a steady percentage each year. To get from a thousand to two thousand, it has to double. That takes a while, and the whole time, its population starts with a one.
But to get from five thousand to six thousand, it only has to grow by a fifth. And from nine thousand to ten thousand, by about a ninth. Those are much smaller jumps, so it zips through them quickly.
Then, once it hits ten thousand, it's back to starting with a one. And it has to double all over again to get to twenty thousand.
So anything that grows by multiplying spends far longer with a one at the front than with a nine. Take a snapshot of lots of such numbers at random moments, and more of them will happen to be in their "ones" phase. Many real-world examples of Benford's law come from exactly this kind of multiplying. It works best when the numbers spread across several orders of magnitude, that is, from tens to hundreds to thousands and beyond.
The units don't matter
Here's the counterintuitive bit, and it's rather lovely.
Suppose you have a big list of river lengths, measured in miles. Count the leading digits, and you get Benford's pattern. Now convert every one of them into kilometres. Every single number changes. A river that was 300 miles long becomes about 483 kilometres, so its first digit flips from a three to a four.
And yet the overall spread of first digits stays the same. Still about 30 per cent ones. Still about 4.6 per cent nines.
That's not a coincidence. In fact, it's part of why the pattern has to look this way. If there really is a universal rule for first digits in data like this, it shouldn't depend on whether you measured in miles or kilometres, or pounds or euros. The distribution of first digits shouldn't change when every number is multiplied by the same factor. And it turns out that an even spread of digits can't survive that kind of rescaling. Benford's lopsided spread can.
Catching made-up numbers
This has a surprisingly practical use.
When people invent numbers, they tend to spread the digits fairly evenly. They avoid starting with one as often as real data would. So if a set of figures that ought to follow Benford's law doesn't, that can be a sign that something deserves a closer look.
Tax inspectors have used this. Software has been built to check whether a set of submitted figures fits the pattern, and departures from it can set alarm bells ringing.
But here's the important part. An alarm bell is not a verdict. A mismatch is a reason to look again, not proof that anything is wrong. There are plenty of innocent reasons for a set of numbers not to fit.
Where it doesn't work
In fact, lots of perfectly honest numbers don't follow Benford's law at all.
It fails for data that's stuck in a narrow range. Adult human heights, for instance, don't span tens to thousands; they cluster together, so their first digits don't follow the pattern. The same goes for IQ scores. It fails for numbers that are assigned rather than grown, such as telephone numbers or invoice numbers handed out in sequence. It fails for prices deliberately set at psychological thresholds, like those ending in 95 or 99. And it isn't reliable for small sets of data, where there simply aren't enough numbers to tell.
It can even mislead. Vote counts from individual polling places, for example, cover too narrow a range for their first digits to say anything useful about fraud.
So Benford's law is a useful flag, in the right circumstances. Not a lie detector.
Randomness has a shape
What's lovely about all this is how it upends a natural assumption. We tend to think that "random" means "flat": every possibility equally likely, every digit taking its turn.
But a lot of the numbers in the world aren't random in that way. They're the result of things growing, multiplying and compounding. And numbers like that leave a fingerprint. Look at enough of them and the shape is there: lots of ones, fewer twos, and only a handful of nines.
Sources
Plus Magazine (University of Cambridge), J. Walthoe, "Looking out for number one." https://plus.maths.org/looking-out-number-one (accessed 25 Oct 2026)
Wikipedia, "Benford's law." https://en.wikipedia.org/wiki/Benford%27s_law (accessed 25 Oct 2026)









