People by US city by country: a country-by-country directory

Take any claim of the form "more people of X origin live in this city than anywhere else." Then ask one question: which boundary was used? The answer usually decides the claim by itself, and the claim almost never states it.

City-level origin data is where the geography stops being a neutral container and becomes the largest single variable in the result. That is the subject here. What the underlying questions ask, and why aggregate figures say nothing about individuals, is set out on the state cross-tab page and not repeated.

What to take away

• At city scale and below, the boundary you choose changes the answer more than the population does. There is no neutral choice.
• Small-area estimates carry margins of error that are often large relative to the estimate, and a zero in a small-area table is frequently a suppression rather than an absence.
• Neighborhood names describe a naming decision and a commercial history. They are not a description of who lives there now.

Five geographies, five different answers

"A city" is at least five different shapes, and published tables use all of them:

Geography	What it encloses	Who defines it
Incorporated place	The legal municipality, whatever its current annexations	The state and the municipality
Urban area	Built-up territory by a density rule, ignoring city lines	The statistical agency
Metropolitan area	A core plus counties tied to it by commuting	The statistical agency, revised periodically
County	An administrative unit that may hold part of a city or several	The state
Tabulation area	Tracts and block groups drawn to hold a target population	The statistical agency, redrawn between rounds

One municipality can sit inside three counties. One metro area can span a state line. A city that annexed aggressively holds suburbs inside its legal limit, while a neighboring city of similar size does not, so comparing the two on any per-city measure compares annexation history as much as anything else.

This is the reason a "largest community" claim needs its geography attached before it can even be evaluated. Change from the municipality to the metro area and the answer commonly moves to a different city.

The margin of error is part of the number

Small-area figures on origin questions usually come from a sample survey rather than a complete count. The published table carries a margin of error for every estimate. Copies of the table drop it, and by the third repetition the estimate is being quoted as a count.

Two consequences follow. First, at small geographies the margin can be a large fraction of the estimate, which means two areas whose estimates differ may not differ at all. Second, ranking areas by such estimates is close to meaningless when the intervals overlap, and they overlap constantly at city and neighborhood scale.

Always retrieve the estimate from the agency's own table, where the margin sits beside it, and carry the margin into anything you write. The statistical agency publishes both the estimate and its reliability; a secondary site publishes only the part that looks like a fact.

A zero can mean four things

In a small-area table, an empty or zero cell can mean:

• Nobody in that area reported that category.
• Someone did, but the count was suppressed to protect confidentiality.
• The category was not offered in that round, or was coded into a broader one.
• The estimate rounded to zero and the margin still spans a real population.

Only the first is an absence. Treating the other three as absence produces confident statements about where a community is not, and those are the statements most likely to be repeated. The general form of this problem, blank against zero against unknown, applies to every dataset of this kind.

Boundaries redrawn are not neighborhoods changed

Tabulation areas are redrawn between rounds to keep their populations within a target range. Where a district grew, one area becomes two; where it shrank, two become one.

So a time series for "this neighborhood" is often a series across a changing footprint. If you compare two rounds without checking the boundary files, you can produce a large apparent change caused entirely by the redraw. Agencies publish relationship files that map old areas to new ones. Use them, and where the geography cannot be reconciled, say so instead of charting the two numbers together.

What a neighborhood name is evidence of

Districts named for a country or a community are real places with real histories, and the name is a poor guide to the present.

Names were applied by outsiders as often as by residents, sometimes by newspapers, sometimes by a business association marketing a commercial strip. Few of them appear in the Geographic Names Information System, which records populated places rather than informal districts. They persist long after the population that prompted them has moved, because signage, restaurants and institutions outlast households. A name can also cover a district that never held a majority of the group it names, since a visible commercial cluster is what gets noticed.

None of this makes the name unreal. It makes the name a fact about the place's history and its business district, not a demographic finding and certainly not a fact about any resident.

What this cross-tab supports, and what it does not

It supports description: this agency, this survey, this year, this geography, this question wording, this estimate with this margin. The holding agency's guide to census records is where those question wordings can be checked. It supports research direction: a documented community in a district is a reason to look at that district's local records, its congregations and its newspapers.

It does not support statements about individuals, in either direction. Nor does it support a ranking of cities by origin, which is a page we do not publish. The count is unstable under the boundary choice, the estimates carry overlapping intervals, and the framing scores places against each other on the composition of the people in them. The reasoning is the same one applied to origins throughout the heritage guide, and it does not weaken at city scale.

For living people the line is absolute: we do not assert anyone's ancestry, nationality, religion or birthplace, and the fact that they are associated with a district named for a country is not evidence of any of those things.

Bottom line

At city scale, origin tables are estimates about drawn shapes. Name the shape, keep the margin, treat a zero as a question, and do not let a boundary decision made by a statistical agency turn into a claim about a person or a ranking of places.

Common questions

How should a city and country pair be written up when it is supported?

As two separate statements with two separate sources, each dated. A person's own public account of their family's origins is quoted and attributed. A city association is tied to a documented event, using the record layers on the city pillar. Neither is derived from the other.

Is a map of these estimates safe to publish?

Only with the geography, the year, the question wording and the margins on the map itself, and with the understanding that it maps an estimate rather than a population. The failure modes of area maps are covered under notability maps.

Which country list is the column header using?

Ask, because it matters. Statistical agencies code write-in answers to their own category list, which is revised between rounds and does not match a diplomatic or standards list. The country pillar sets out why several lists exist and why they disagree.