The Mechanism
You'd expect the leading digit of "natural" numbers to be evenly spread — each of 1 through 9 appearing about 11% of the time. It isn't. Across enormous ranges of real data, the digit 1 leads roughly 30.1% of the time, 2 about 17.6%, and the frequencies fall off smoothly to 9 at under 5%. The pattern was first noticed in 1881 by the astronomer Simon Newcomb, who observed that the early pages of shared logarithm tables — the ones for numbers starting with 1 — were far more worn and grubby than the later pages. He wrote down the rule: the chance of leading digit *d* is log₁₀(1 + 1/*d*). It was ignored. In 1938 the physicist Frank Benford rediscovered it and tested it obsessively across 20 unrelated datasets — surface areas of 335 rivers, sizes of 3,259 US towns, 104 physical constants, molecular weights, even numbers pulled from a *Reader's Digest* issue — and it held. The reason: data spanning many orders of magnitude is roughly uniform on a *logarithmic* scale, and on that scale the stretch belonging to a leading 1 is the widest. Because honest data obeys the law and invented numbers usually don't, auditors now use Benford's law to flag faked tax returns, cooked books, and election counts.
Why It Matters
Most people expect the digits 1 through 9 to each appear about one-ninth of the time, but real data often breaks that expectation. Benford's law says the first digit 1 should appear about 30.1% of the time, while 9 should appear less than 5%. That is surprising because the rule shows up across very different kinds of measurements, from river sizes to town populations and physical constants. The key idea is that many real datasets cover huge ranges, and on a logarithmic scale there is more room for numbers that start with 1 than for those that start with 9.
Wait — That's Not Quite Right
A common mistake is to think Benford's law means every list of numbers should have lots of 1s at the front. It does not. The pattern appears mainly in data that spread across many orders of magnitude and are not artificially limited, like prices, sizes, or counts gathered from the real world. Small school test scores, lottery numbers, or fixed ID numbers do not have to follow it.
Vocabulary
- benford's law
- leading digit
- logarithm
- logarithmic scale
- orders of magnitude
- distribution
- natural numbers
- audit
- fraud detection
- Simon Newcomb
- Frank Benford
Quick Quiz
5 questions · For classroom or kitchen table
The Experiment
Sort Real-World Numbers by First Digit
Find 15 to 25 numbers from a newspaper, a sports page, a map, a school reference book, or a website with adult help. Good choices are river lengths, city populations, company revenues, or distances. Write down only the first digit of each number, then count how many 1s, 2s, 3s, and so on you found.
Now compare your results with what you would expect if each digit from 1 to 9 were equally likely. If your list comes from real-world measurements, you may notice that 1 shows up more often than 9. That is the same kind of pattern Benford studied.
For a second check, make up a list of random numbers between 100 and 999, then count those first digits too. The invented list will often look more even than the real one. This helps show why Benford's law can be useful when people are trying to tell real data from made-up data.
paper, pencil, newspaper or website with numbers, calculator optional, adult supervision for web browsing
Where this came from
- Newcomb, S. "Note on the Frequency of Use of the Different Digits in Natural Numbers." *American Journal of Mathematics* 4, 39–40 (1881); Benford, F. "The Law of Anomalous Numbers." *Proceedings of the American Philosophical Society* 78, 551–572 (1938). Overview: https://en.wikipedia.org/wiki/Benford's_law
Want next week's entry in your inbox?
One short email a week with the latest field guide entry — the fact, the explanation, the quiz, and the activity. Free for parents and teachers.
For adults only · Unsubscribe anytime