Who is leaving New York, and who is arriving

Closely, and that is the check everything else rests on. The Bureau’s 2024 table reports 8,387,686 residents one year and over, 7,586,955 in the same house a year earlier, 401,043 who moved within their borough, 157,700 from another New York county, 143,496 from another state and 98,492 from abroad. Rebuilding the same rows from the 2024 microdata gives 8,388,298, 7,589,556, 405,800, 153,304, 142,120 and 97,518, each inside the published margin of error . The 2021, 2022 and 2023 tables reproduce the same way. In 2019 the same-house and same-borough rows differ from the published figures by 38,000 and 27,000, more than either margin alone and inside the margin on the difference, which is the Bureau’s own test; the other three 2019 rows are inside their margins. The full comparison is in the Reproduce box and the reproduction table. The leavers have no published table to check against, which is the reason to build them.

In 2015 about 297,000 people left for the rest of the country, 160,000 arrived from it and 95,000 arrived from abroad. Departures were 317,000 in 2019, 407,000 in the 2020 survey and 450,000 in 2021, while arrivals from abroad fell to 54,000; then 388,000 leavers in 2022, 314,000 in 2023 and 296,000 in 2024. Domestic arrivals moved far less, between 152,000 and 204,000 across the decade, which is why the net domestic loss, 137,000 in 2015, reached 266,000 in 2021 and was back to 113,000 in 2024. About 518,000 people moved from one address to another within the five boroughs in 2024. The full series with margins is in the methods note at the end.

Of the 2024 arrivals, about 115,000 came from another state, 41% of the total; 98,000 from abroad, 35%; 47,000 from the metro suburbs, 17%; and about 21,000 from elsewhere in New York, New Jersey or Connecticut. Seven in ten of the arrivals from other states are aged 18 to 34 and 78% of those 25 and over hold a degree.

Of the 296,000 leavers, about 105,000 went to the suburbs: 41,000 to northern New Jersey, 30,000 to Long Island, 28,000 to the Lower Hudson Valley and 6,000 to southwestern Connecticut. Another 27,000 went elsewhere in New York State and 12,000 elsewhere in New Jersey or Connecticut. Florida took about 22,000, against 28,000 in 2015; California 18,000, Pennsylvania 14,000, Texas 11,000; the rest of the South together 42,000 and the Midwest 21,000. By state the list runs New York, New Jersey, Florida, California, Pennsylvania, Connecticut, Texas, Massachusetts.

The exchange with the suburbs was a net loss of about 58,000 in 2024, against about 38,000 to all the other states combined. In 2015 those figures were 40,000 and 77,000, and in 2019 the suburban figure had already grown to 68,000.

The young. In 2024, 6.8% of the 18-to-24-year-olds who lived in the city a year earlier had left for elsewhere in the country, 6.3% of 25-to-34-year-olds, then 4.0%, 2.3% and 1.5% up the age bands. The young leave at the highest rates and arrive in the largest numbers, which is how the city gains a few thousand 18-to-24-year-olds on balance, 8,000 in 2019, while the middle-aged leave less often and are barely replaced. People with a degree leave at 4.6% a year and those without at 2.5%. White residents leave at the highest rate of any group, 4.3%; the net losses were 41,000 Hispanic residents, 27,000 white residents and 24,000 Black residents, and 73,000 US-born against 40,000 foreign-born.

 

Measured

 

How to read this, and what it cannot show

Leavers as a share of everyone who lived in the city a year earlier and is still in a US census file. It leaves out people who died or moved abroad. The income, tenure and children rates classify a leaver by the household they live in now, not the one they left, so they are not propensities to leave by prior circumstance. Groups resting on fewer than ten leaver records are not drawn.

By borough the exchange with the rest of the country is a Manhattan story on one side and an outer-borough story on the other. Manhattan received 87,000 domestic arrivals in 2024 and sent out 80,000, roughly a wash; Brooklyn lost a net 43,000, Queens 38,000, the Bronx 28,000 and Staten Island 10,000. Manhattan also has the highest out-migration rate, 5.0%, and the highest arrival rate; it is the borough people pass through.

In 2024 half the domestic arrivals lived in city households at $150,000 or more, against 38$150,000 against $110,000 for the median leaver's; 45$150,000, or the 4.5% out-migration rate among people now in households above $300,000, describes the households leavers ended up in, not the households New York lost. The city did not necessarily lose a$300,000 household or a homeowner when the person who now lives in one left. The person-level attributes, age, race, nativity and, for adults, degree, do travel with the person, and the claims about what the city gains and loses are made on those.

Of the roughly 98,000 residents in 2024 who had lived outside the United States a year earlier, 71% were not US citizens, 6% were naturalised citizens and 22% were US citizens by birth, mostly Americans coming home. Half the adults hold a degree, against 78% of arrivals from other states; 46% of those 16 and over were employed, against 68%; 36% now live in households under $50,000, against 17%; and more than half live with children. In 2019, 61% of the adults held a degree and 51% were employed.

By country of residence a year earlier, Asia supplied 37% of the 2015 arrivals, the Caribbean, Mexico and Central America 22%, Europe 20%, South America 8% and Africa 7%. In 2024 Asia’s share was 32%, the Caribbean and Central America 23%, South America 18%, Europe 16% and Africa 10%.

Because they are different measurements, and an ordinary net flow could not be twice its own gross inflow. The census figure, about 98,000 in 2024, counts residents who say they lived abroad a year ago, from one survey, over a calendar year. The ledger’s 219,578 for the year to July 2024 is an estimate of immigration minus emigration for a July-to-July year, built from the survey and from administrative records, and it includes an adjustment for humanitarian migrants that the Bureau introduced and then revised; that revision is most of the reason the July 2024 population estimate for the city rose from 8,478,000 to 8,597,000 in one vintage. So the surge is read within the ledger: net international migration of 30,000 to 64,000 a year from 2011 to 2019, about 20,000 in the year to July 2021, 84,000 in the year to July 2022, 155,000, 219,578, and then 65,824 in the year to July 2025.

It tells the decade in three acts. Through the 2010s the city gained 60,000 to 68,000 a year from births over deaths, falling to 47,000 by 2019, and lost between 49,000 and 152,000 a year to the rest of the country, a domestic loss that grew to more than 130,000 a year from 2017. Then the pandemic: a net domestic loss of 330,000 in the year to July 2021 and 213,000 the year after. Then the international surge, during which the domestic loss fell back to about 94,000 and the population rose by 163,000 in the year to July 2024. In the year to July 2025 the domestic loss was 113,718 and the population slipped by 12,196.

The revisions are not small. Beyond the 2024 revision described above, the Bureau cut its 2017 estimate for the city by 184,000 in another vintage. Each new vintage replaces the whole series back to the last census, so the 2025 figures here will change in March 2027.

Because they are two measurements. The microdata counts people by asking them where they lived a year ago, in a calendar-year survey, and subtracts leavers from arrivals with a sampling margin of about 25,000. The population estimates infer domestic migration from address changes on federal tax returns and Medicare enrolment over a July-to-July year, with no sampling margin but with revisions every vintage. For the year to July 2024 the ledger says a net loss of about 94,000; the 2024 survey says about 113,000, and the year to July 2025 ledger says 114,000. For 2015 the two were 93,000 and 137,000. The survey counts students, people who do not file and the young who file from a parent’s address; the ledger counts the flows the tax records see. Across the decade the two move together, and neither should be quoted to the nearest thousand.

More of everyone, and, on the tax returns, disproportionately the well paid. The out-migration rate rose from 3.8% in 2019 to 5.2% in 2021, and the increase reached every age band, every borough and every income band of the households people ended up in. The IRS county migration data , which match returns across two filing years and see the income a household reported both before and after it moved, say the same thing in dollars: the average income on a return that left the city was $139,000 in 2019–20,$151,000 in 2020–21 and $141,000 in 2021–22, against about$100,000 for a return that stayed, where before 2019 the two had been within a few thousand dollars of each other. By 2022–23 the leaving return was back to $120,000 against$112,000 for stayers. The profile of movers shifted less than the volume. In 2021, 68% of domestic arrivals were aged 18 to 34 and 74% of the adults held a degree, as in other years; among leavers 42% were 18 to 34, 19% were under 18 and 62% of the adults held a degree, against 53% to 58% in 2015, 2019 and 2024. The 2020 survey year carries the Bureau’s experimental weights and is drawn but not leaned on; the 2021 file is a standard release.

The question needs the tax returns, because they record a household’s income before it moved, which the census does not. In 2022–23 about 321,000 people were on returns that left the five counties and 194,000 on returns that arrived; the average return that left reported about $120,000 of income, the average return that arrived about$106,000, and the average return that stayed about $112,000. In 2015–16 the three were$102,000, $95,000 and$95,000, close together; the gap opened in the pandemic and has mostly closed. The state Comptroller’s dashboard, which covers the whole state rather than the city, reports for tax year 2024 a net loss concentrated among married filers earning $100,000 to$500,000 and a net out-migration rate of about 1% among filers above $500,000, the highest of any bracket, against a statewide net rate of −0.14%. So: the highest earners leave at a higher net rate than the middle, and in far smaller numbers, and the average leaving household earns somewhat more than the average staying one. The census’s income bands, being measured after the move, cannot add to that; the destinations on the returns are the census’s destinations, the rest of New York State first, then New Jersey, Florida, Pennsylvania, California and Connecticut.

 

Administrative

 

How to read this, and what it cannot show

One bar per pair of filing years; switch between the average income of leaving, arriving and staying returns and the headcounts. Nominal dollars. Filers only, so students, most new arrivals from abroad and people who do not file are not in it, and a household that moves in April can show its old address on one return and its new one on the next. Flows between the five boroughs are removed.

 

Administrative

 

How to read this, and what it cannot show

People on returns that left the city for each state, or arrived from it, 2015–16 against 2022–23. New York means the rest of the state, mostly the suburbs. Switch between destinations and origins.

About half do not. Of the employed people who moved from the five boroughs to Long Island, the Lower Hudson Valley, northern New Jersey or southwestern Connecticut in 2024, 47% still had a workplace in the city, down from 55% in 2019, and a further share of all leavers, 23%, usually worked from home, three times the 2019 figure. The suburban move looks like a housing move more than a job move: 58% of the people who made it now live in an owner-occupied home, half in households above $150,000, and 46% with children. The reverse flow, 47,000 people a year from the suburbs into the city, is younger, 61% aged 18 to 34, with almost no children.

Seventh in the census’s destinations and third on the tax returns. About 22,000 people who lived in the city in 2023 lived in Florida in 2024, 7% of leavers, against 28,000 in 2015; 105,000 went to the suburbs in the same year. On the IRS side, returns leaving the five counties for Florida carried about 30,000 people in 2022–23, behind the rest of New York State at 86,000 and New Jersey at 58,000. The Florida leaver in the census is older than the average leaver, 36 at the median against 31, and in 2024 six in ten of the adults held a degree and half now live in an owner-occupied home.

Less than the question deserves, and two indirect series say it. The first is the monthly survey behind the unemployment rate, which asks every household member where they were born and identifies the five boroughs. It has no question about last year, so it cannot count arrivals; what it measures is the size of the foreign-born population, a stock that immigration, emigration, domestic moves, deaths and the size of the US-born population all move. From June to December 2024 about 39% of the city’s residents were born abroad; from January to August 2025, about 37%; from January to August 2026, about 38%, with a margin of a point either way in each period that understates the true uncertainty, since consecutive months share most of their sample. The census’s 2024 file put the share at 37%. The honest reading is that the survey shows no detectable rise in the foreign-born share through August 2026, which is consistent with the surge having subsided, and that a stock measure cannot rule out new arrivals offset by departures or by growth elsewhere in the population.

 

Measured

 

How to read this, and what it cannot show

One bar per month; switch between the foreign-born share, the non-citizen share, the foreign-born count and the national share. About 1,300 to 1,500 residents a month, and adjacent months share most of their sample, so read the level and not the wiggles. No file exists for October 2025, when the survey was not conducted during the federal shutdown.

The second is context rather than measurement: the city’s daily count of people in the Department of Homeless Services shelter system, which includes asylum seekers in DHS shelters but does not separate them and excludes the humanitarian centres other agencies ran. It stood at about 45,000 people in March 2022, reached a monthly average of 89,000 in November 2023, and has been level since: 87,000 across 2024, 86,000 across 2025 and about 84,000 in September 2026. The Mayor’s office put the number of asylum seekers in the city’s care at about 69,000 at the January 2024 peak and under 37,000 in July 2025, with weekly arrivals down from 4,000 in May 2023 to under a hundred, and more than 237,000 people through the system since spring 2022. What the shelter series supports is that the arrivals of 2022 to 2024 have not been followed by a second wave large enough to move it; it cannot say more than that about migration.

 

Administrative

 

How to read this, and what it cannot show

One bar per month; switch between everyone in shelter, people in families with children, single adults and children. A daily administrative count averaged by month.

The releases that will eventually settle the picture are already in the pipeline. The census’s 2025 one-year file, which would show who arrived and left in 2025, has been delayed past its usual September release with no date announced. The IRS has not announced a date for its 2023–24 flows. The Bureau’s next ledger, to July 2026, is due in March 2027 and will revise 2025 again.

They are the hole in every number here. Nobody who moved from New York to Toronto, London or Santo Domingo answers a US survey, so the census sees arrivals from abroad but no departures to it, and the “net international migration” in the Bureau’s ledger is an estimate of emigration subtracted from an estimate of immigration. Movers to Puerto Rico are in the Puerto Rico Community Survey, which is not read here; they number a few hundred a year in the study area. The domestic flows have no such hole, because every state’s records are in the national file.

Because each vintage starts from the latest census count and the 2020 count was far above the estimates. The Vintage 2020 series carried the city at 8,253,000 in July 2020; the 2020 census counted about 8,804,000. The Bureau’s post-census series starts from the count, so the level jumps by more than half a million at the line in the chart, and the components on either side were estimated with different methods and different data. The chart shows both because the decade needs both, and labels the line. The Department of City Planning’s own long series splices them; the reading here does not.

It was tried and set aside. The Current Population Survey’s March supplement asks where each person lived a year earlier and identifies the five boroughs, but its 2025 city sample is about 2,300 people, of whom about twenty had arrived from another state or abroad, too few for a rate with a useful margin. The basic monthly files are used instead for the one thing they measure well at city level, the foreign-born share, and the 2026 supplement had not been released when this was written.

Anything below the borough. Why anyone moved. Anyone who left the country. What a leaver’s household earned or owned before the move. Infants, since the question is asked of people one and over. Second homes and part-year residents, who answer from wherever they are on the survey day. Which of the arrivals from abroad came through the shelter system. Whether the people who left in 2021 came back, since each year is a fresh sample. And the year 2025 in the census, which is not yet published.

This is the methods note for everything above. It is deliberately more technical than the rest.

The published anchor. Census table B07001 (geographical mobility in the past year by age, population one year and over) for New York city from the one-year survey: 2019 from the Bureau’s sequence-format summary file (sequence 18, start position 7), 2021 to 2023 from the table-based summary files, 2024 from the Census Reporter mirror. Five rows are reproduced: same house, moved within the same county, moved from a different county in the same state, moved from a different state, moved from abroad. In the microdata these are, respectively, residence-one-year-ago code 1; code 3 with a New York migration PUMA equal to the resident’s own borough; code 3 with any other New York migration PUMA, which includes the other boroughs; code 3 with a migration state among the fifty states and DC other than New York; and code 2, or code 3 with a migration state outside the fifty states and DC, so that Puerto Rico and the island areas count as abroad, as the table defines them.

Residents. The 2015 to 2024 one-year public-use microdata samples for New York, cut to the 55 microdata areas of the five boroughs and to people one and over with a residence-one-year-ago answer: 66,000 to 70,000 person records a year, with 80 replicate weights, joined to their housing records for household income, tenure, household size and the presence of children. Group quarters residents are kept for every count and rate and dropped from household measures.

Leavers. The national one-year person files (about 3.3 million records a year, two CSV members, read in place from the zip) filtered to migration state 36 and a five-borough migration PUMA: 3700, 3800, 3900, 4000 and 4100 in the 2010 geography used through 2021 (the Bronx, Manhattan, Staten Island, Brooklyn and Queens), and 4100 to 4500 in the 2020 geography from 2022 (Manhattan, the Bronx, Brooklyn, Queens, Staten Island). Both code lists were checked against the state file, where the modal migration PUMA of each borough’s within-city movers is the code above. Records whose current microdata area is also in the city are within-city moves and are excluded from the leavers; the rest, 2,800 to 4,200 records a year, are joined to the national housing file on the household serial number. The 2015 to 2018 files carry the older variable names, translated code for code.

Origins and destinations. For arrivals, the migration state or country and migration PUMA a year earlier; a migration PUMA is placed in the metro suburbs through the IPUMS composition files that list the microdata areas each migration PUMA contains, taking the majority region of those areas. For arrivals from abroad, the migration state/country code gives the country or region of residence a year ago; the census releases many countries only as regions (“South Central Asia, not specified”, “Western Africa, not specified”), which is why the chart groups by region and the text names only the countries the code names. Region of birth, from the place-of-birth code, is kept as a second view. For leavers, the current state and microdata area, with the same four suburban regions identified through the study-area geography; other states are grouped by census region with Florida, Pennsylvania, California and Texas shown separately. People now in Puerto Rico are in a separate file and not read.

Derived variables. Household income is the housing record’s total income times the Bureau’s within-year factor and the New York-metro CPI to 2024 dollars, measured in the household the person lives in at the survey. Degree is a bachelor’s or higher among people 25 and over; employment among 16 and over; enrolment among 18 and over. Race and ethnicity are Hispanic of any race, otherwise the census’s single race. Foreign-born is nativity code 2; non-citizen is citizenship code 5. With children means someone under 18 in the household; owner means tenure 1 or 2. Regions group the codes: 100 to 199 Europe, 200 to 299 Asia, 300 to 359 the Caribbean, Mexico and Central America, 360 to 399 South America, 400 to 499 Africa, the rest Canada, Oceania and elsewhere; for residence a year ago, codes below 100 are Puerto Rico and the island areas. A leaver still works in the city if their place-of-work state is New York and their place-of-work area is one of the five borough codes.

Estimators. For any statistic computed with the full weight, the same statistic is recomputed with each of the 80 replicate weights and the variance is the Bureau’s successive-difference estimator, four-eightieths of the summed squared deviations of the replicate estimates from the full-weight one; margins are 90%, or 1.645 standard errors. A net flow is the difference of two weighted counts computed replicate by replicate, so its margin accounts for both sides; residents and leavers of a year come from the same survey with the same replicate weights, which is what makes that legitimate. The out-migration rate is a weighted share with the same margin. Medians are weighted medians without a margin. A published row is reproduced if the difference is within the root of the summed squared margins, the Bureau’s test for a difference between two estimates; the full sample and its public-use subsample are not independent, so this is a convention rather than an exact test. Chart cells resting on fewer than 30 records, or rates on fewer than ten leaver records, are labelled too thin rather than drawn.

The ledger, the returns and the two 2026 series. Population estimates are the five counties summed from the Bureau’s county files, July to July, Vintage 2020 for 2011 to 2020 and Vintage 2025 for 2021 to 2025; the Vintage 2025 row for 2020 covers April to July and is not used. IRS county migration totals are each county’s “total migration, US and foreign” row less the flows to and from the other four counties; average income is adjusted gross income per return in nominal dollars; the state lists exclude the IRS’s “other flows” aggregates of suppressed small pairs. The shelter census is the daily DHS report averaged by month. The monthly survey’s foreign-born share is naturalised citizens plus non-citizens among all residents of the five counties, weighted by the final person weight, pooled by period with a margin from the spread of the monthly estimates, which understates the true error because consecutive months share three quarters of their sample.

Does it reproduce? Every row of the 2021 to 2024 tables lands inside the published margin; in 2019 two rows land inside the margin on the difference and three inside the published margin. The year-by-year comparison is the reproduction table.

The series. Leavers: 296,738 in 2015, 316,900 in 2019, 407,600 in 2020, 449,734 in 2021, 387,853 in 2022, 314,096 in 2023, 296,174 ± 19,000 in 2024. Domestic arrivals: 160,238, 170,411, 151,616, 184,182, 204,067, 194,964 and 183,414. Arrivals from abroad: 94,555, 87,696, 63,847, 54,094, 83,595, 106,036 and 97,518. Net domestic: −136,500 ± 20,400 in 2015, −146,500 in 2019, −265,600 ± 26,100 in 2021, −112,800 ± 24,800 in 2024. Out-migration rate: 3.50%, 3.82%, 5.24% and 3.52%.

Who. In 2024 the net domestic flow by age was −29,000 (under 18), +2,000 (18 to 24), −18,000 (25 to 34), −26,000, −27,000 and −15,000 (65 and over); US-born −73,000, foreign-born −40,000; graduates −32,000, non-graduates −53,000. Classified by the household people live in now: −14,000, −21,000, −31,000, −22,000, −6,000, −8,000 and −13,000 up the seven income bands; owner-occupied homes −96,000 and rented −19,000 ± 19,000; with children −68,000, without −47,000. Out-migration rates ran from 1.5% at 65 and over to 6.8% at 18 to 24. Domestic arrivals: median age 27, 65.2% aged 18 to 34, 72.1% of adults with a degree, median household income now $150,384; leavers: 31, 45.7$110,358; stayers: 40, 22.7%, 40.9%, $95,858. Arrivals from abroad: 71.1$80,912; by residence a year ago, 32.3% Asia, 22.9% the Caribbean, Mexico and Central America, 17.8% South America, 15.7% Europe, 9.7% Africa, 0.6% Puerto Rico and the island areas. Suburban leavers: 57.7% now in owner-occupied homes, 46.1% now with children, 46.5% of the employed still working in the city.

Rerunning it. The analysis script computes every figure and number above, the charts and the dashboard in one run from a fixed seed; its docstring is the long-form method, and the leaver extraction from the national files is a separate script named there. The companion dataset is the census records themselves: the 217,000 residents and leavers in the 2015, 2019 and 2024 files, with their survey weights and the derived variables, so a weighted count of the status column within a year reproduces the headline flows.