How this was built, and what it cannot tell you

The limitation that matters most

Around 80% of New York City dogs are not licensed. This is not a register of every dog in the city. It is a register of the dogs whose owners filled in a form, and those owners are not a random sample of dog owners.

Every figure on this site is a rate per 10,000 licensed dogs, either within a birth-year group or within a ZIP code. That controls for the obvious problem: a neighbourhood that licenses at half the rate of another is not thereby made to look as though it has half the dogs. What it cannot control for is whether the sort of person who licenses a dog in one neighbourhood names dogs differently from the sort of person who does not. That caveat applies to every neighbourhood comparison here and we would rather say so than let you find out later.

Source

What we changed

• Licence renewals deduplicated to individual animals
• Names normalised for case and punctuation; placeholder names removed
• Breed free text reduced to a single primary breed
• Records outside the cohort range or without a valid ZIP excluded

Two of those need explaining. Licences renew, so the same dog appears in the file once for every renewal; without collapsing them, long-lived dogs count many times over and the whole dataset tilts towards older names. And roughly 30,000 records have a placeholder instead of a name: UNKNOWN, NAME NOT PROVIDED, NAME, and several typos of the same. Left in, three of the fastest-growing dog names in New York come out as NOT, A and UNKNOWED.

Thresholds