Why do so many numbers start with 1?

Across street addresses, populations, invoice totals and physical constants, the leading digit isn't evenly spread โ€” a 1 shows up about 30% of the time, a 9 under 5%. That lopsided fingerprint is Benford's lawBenford's law: in data spanning several orders of magnitude, the first digit d appears with probability log10(1 + 1/d). It holds for naturally multiplying quantities, not for capped or made-up ones., and numbers that don't follow it โ€” invented figures, capped values โ€” stand out. Load an example below or paste your own, then test it.

Should fitEach term is the sum of the prior two, so the run grows by a near-constant ratio (โ‰ˆ1.618ร—). Multiplicative growth is the engine of Benford's law.

Parsed 200 numbers with a leading digit.
1: 30% observed vs 30.1% expected (60)2: 18% observed vs 17.6% expected (36)3: 12.5% observed vs 12.5% expected (25)4: 9% observed vs 9.7% expected (18)5: 8.5% observed vs 7.9% expected (17)6: 6% observed vs 6.7% expected (12)7: 5.5% observed vs 5.8% expected (11)8: 6% observed vs 5.1% expected (12)9: 4.5% observed vs 4.6% expected (9)
123456789
observedBenford expected
0.7ฯ‡ยฒ (df 8)Pearson's chi-square: the summed, expectation-weighted gap between observed and expected counts. Bigger means a worse fit. The 5% threshold at 8 degrees of freedom is 15.5.
0.9996p-valueThe chance of a gap this large (or larger) if the data really did follow Benford. Below 0.05 is the usual flag that it doesn't.
0.004MADMean absolute deviation: the average gap between an observed and expected digit share. Auditors read it against fixed bands (first digit: โ‰ฒ0.006 close, โ‰ฒ0.012 acceptable), but those assume thousands of values โ€” on a small sample lean on the p-value instead.
โœ“ ConsistentThe leading digit track Benford's curve closely (p = 0.9996). Nothing here looks off โ€” this is what naturally growing numbers do.
DigitExpectedObservedCount
130.1%
30%
60
217.6%
18%
36
312.5%
12.5%
25
49.7%
9%
18
57.9%
8.5%
17
66.7%
6%
12
75.8%
5.5%
11
85.1%
6%
12
94.6%
4.5%
9

Benford holds for quantities that span several orders of magnitude and grow by multiplying โ€” revenues, river lengths, file sizes. It fails by design for capped or assigned numbers (heights in cm, phone numbers, sequential IDs), so a deviation isn't proof of anything on its own.

Benford's law ยท ฯ‡ยฒ goodness-of-fit at 8 degrees of freedom