Why almost any string of large numbers prefers to start with 1
Benford's law predicts that digit 1 leads large numbers more than six times as often as digit 9. Country populations, country areas, and the Fibonacci sequence confirm the pattern, each with its own ratio. The law has a practical use too. Tax authorities apply it to flag fabricated accounts.
A number growing by a fixed percentage at every step, like compound interest. The bars count which digit each step of the climb landed on. Let it run: digit 1 fills up visibly faster than digit 9.
01The mechanism
The explanation is pure arithmetic. To move from a leading digit of 1 to a leading digit of 2, a growing quantity (a population, a price, an investment) has to double, from 1,000 to 2,000, say. But to move on from a leading digit of 9 to the next order of magnitude, at 10,000, it only needs to grow by a little over ten percent, from 9,000 to 10,000. Any process that grows by multiplication spends far more of its journey with a small leading digit than with a large one.
The pattern was first noticed in 1881 by the astronomer Simon Newcomb, who observed that the early pages of logarithm tables, the ones for numbers starting with 1, were far more worn than the later ones. The physicist Frank Benford rediscovered it independently in 1938 and tested it against more than twenty thousand values across twenty different data sets, from river areas to electricity bills. The law has carried his name since.
02The real data
The rule is not confined to pure mathematics. It shows up in real things, measured entirely independently of one another: country populations, country areas. Pick a data set and watch how the bars line up against the dashed red line, which marks exactly what Benford's law would predict.
The last button is the only invented example on this page: three hundred invoices generated at random, all deliberately kept under a 10,000 approval cap. The needle drops straight into the red. Exactly the pattern an auditor looks for.
03Test your own numbers
Paste any string of large numbers you work with: salaries, invoices, distances, city populations. Separate them by commas, spaces, or line breaks. You need at least twenty for the result to mean anything.
04What it's used for
Since Frank Benford, the rule has become a genuine working tool. Forensic accountants, and in some countries even tax administrations, compare the leading digits of the amounts in a set of accounts against Benford's curve. Numbers invented by a person, even a careful one, tend to be far more evenly distributed than those produced by a real economic process. A person choosing numbers "at random" spreads them almost as evenly across 1 as across 9, while a real economic process grows by multiplication, the same mechanism above that favours the digit 1. A large deviation from the curve does not prove fraud on its own, but it is reason enough for a line in a ledger to get a manual look.
The rule only holds for numbers that arise from a free process, not from an administrative rule. It does not apply to identification numbers, phone numbers, or any set of values artificially confined to a narrow range, such as percentages or exam scores. It also needs a sample large enough, spanning at least two or three orders of magnitude, so that random noise does not read as a pattern.
Country populations and areas come from the World Bank's public indicators SP.POP.TOTL (2023, 217 countries) and AG.SRF.TOTL.K2 (2022, 215 countries). The Fibonacci sequence and the padded-invoice example are computed directly on the page, the latter reshuffling at random on every load. The conformity score measures the total distance between the observed distribution and Benford's curve; a score above 88 sits within natural range.