A third of New York City's adult citizens are naturalized, nearly twice the share of 1990, and the share has held there since about 2012

What the survey records for each person. Its codebook sorts every resident into one of five groups: “Born in the U.S.”, “Born in Puerto Rico, Guam, the U.S. Virgin Islands, or the Northern Marianas”, “Born abroad of U.S. citizen parent or parents”, “U.S. citizen by naturalization” and “Not a citizen of the U.S.”. The first three are native-born here. Nothing in the answers records immigration status: the files don’t say what papers, if any, a non-citizen holds. The Board of Elections names among citizens “those persons born in Puerto Rico, Guam and the U.S. Virgin Islands”.

Not every answer was given. The codebook flags the ones the Census Bureau filled in (its “Citizenship allocation flag”): citizenship for 11% of the city’s adults in 2024, and the years of naturalization and of entry as the story says. Leaving them out moves the three shares we checked by at most 2.2 points (the arrivals since 2015) and the naturalized share by less than a point.

No. Adult citizens are the census’s nearest measure, but the Board’s rules ask more: “Be a New York City resident for at least 30 days. Be 18 years old on or before Election Day. Not be in prison for a felony conviction. Not be adjudged mentally incompetent by a court.” These counts can’t apply those rules, so they include people who may not register, and don’t measure how many.

The Board’s 2025 report says that in 2026 “voters will elect members of the New York State Legislature, the Governor”, among other offices. The Board set 24 October as the last day to register for the election.

Because that year’s survey covered a different population. The Census Bureau’s guidance says the group quarters population “is included in the 2006 ACS and not included in the 2005 ACS”, so 2005 is left out rather than set beside years that count people in group quarters.

The census’s published count of the city’s adults by nativity and citizenship for 2024 comes out of the microdata within its margin for the native-born, the naturalized and non-citizens; the figures are in the check below.

The yearly survey’s estimates carry margins from its replicate weights: the naturalized share of 2024 is good to about 0.5 of a point either way, at the 90% level the Census Bureau uses. Changes between survey years are tested against those margins: since 2006 the native-born rose by 431,000, give or take 48,000.

Since 2012 the native-born gained 130,000 more than the naturalized on the 2024 file, give or take 73,000, a gap beyond the survey’s margin of error. The same comparison ending in 2023 had them 8,000 apart, give or take 67,000.

The yearly survey has a second limit. Each year’s file is weighted to the Census Bureau’s population estimates of that year, and the 2020 census revised the city’s upward: the 2018 and 2019 files sit about 5% below the estimates as since revised, and the 2006 to 2009 files a few per cent above, so counts from different years stand on different bases. A share within one file is unchanged by the scalar check below; the survey’s actual controls, by age, sex, race and Hispanic origin, could move a share a little.

Put on the revised series, the native-born rise since 2006 is about 580,000 rather than 431,000, and adult citizens grew by 1.11 million since 1990 rather than 1.07 million.

On that basis naturalized citizens supplied 89% of the growth since 1990 rather than 91%. We print the files’ own counts; every count comparison in the story holds on either basis.

The 1990 and 2000 figures come from 5% samples of the census, with no margin calculated, so changes from those years compare estimates without a test.

The two surveys’ estimates need not match, and these results don’t establish why they differ: the monthly survey covers people outside institutions, sets its population totals each January from the Census Bureau’s estimates, and has a much smaller city sample. Its changes are between the same months of two years, with approximate margins clustered on the household.

The non-citizen share’s change could be anywhere from −2.9 to 1.3 points, and the naturalized share’s from −1.9 to 3.5, at the 90% level: about the same, both.

The census years. The 1990 and 2000 censuses (5% samples, no margin calculated) and the American Community Survey’s one-year files from 2006 to 2024 (margins from the replicate weights; the 2020 file was an experimental release, with no margins). Adults are 18 and over in the five boroughs. In the survey files from 2015 the script checks that every foreign-born resident is either a naturalized citizen or not a citizen.

One population basis. The one-basis check multiplies each year’s counts by the ratio of the Census Bureau’s revised 1 July estimate of the city’s population to the file’s own total (nycdata.elections.population_controls): a proportional rebasing, not a reweighting, which assumes the file’s shares by citizenship are unaffected; the censuses count April and the estimates are for 1 July. The survey controls its weights to population by age, sex, race and Hispanic origin, never by citizenship, so every share here is a sample estimate.

The latest year in detail. For 2024, place of birth grouped into the codebook’s world areas (so Armenia, Azerbaijan and Georgia count in Asia), with Latin America split into the Caribbean, Mexico and Central America, and South America; year of naturalization and year of entry as the survey records them, the entry year being the most recent arrival. The two largest countries of birth are compared in the same file, replicate by replicate. The allocation flags come from the raw state file, since the extract doesn’t keep them.

The current year. The Current Population Survey’s basic monthly files, January to September 2026 against the same months of 2025, adults in the five boroughs, final weight, margins clustered on the household and approximate (households interviewed in both years are not paired, and the survey’s geographic clustering is left out); members of the armed forces, who carry no final weight, are left out.

Code and checks. One script, analysis.py (linked below), produces every figure and chart and checks each quoted sentence against the saved documents. The dataset cards behind it (census_history, acs1_pums, acs_tables, cps_basic) list the traps.