Los Angeles County lost 165 people a day for five years. We ran the next century four ways

Los Angeles County had 9,996,510 residents on 1 July 2020 and 9,694,934 on 1 July 2025. That is 301,576 people gone in five years, a loss of 3.0 percent, and it is the worst five-year run of any large county in the United States.

Twenty-five counties held more than 1.5 million residents in 2020. Los Angeles finished last of the twenty-five. Alameda came second-worst at negative 2.6 percent, Kings County in Brooklyn third at negative 2.5. At the other end, Bexar County in Texas grew 7.2 percent and Harris County, which is Houston, grew 6.6 percent. Over the same five years the country as a whole grew 3.1 percent, so the gap between Los Angeles County and the United States is six points.

Decline itself is not unusual. Of 3,144 counties, 1,261 lost population between 2020 and 2025, and among the twenty-five largest Los Angeles lost the largest share.

What the state projects

California’s Department of Finance publishes the official population projection every county in the state plans against. Its Vintage 2026 release, built on a 2024 baseline, carries Los Angeles County out to 2070.

The state’s number for today is not the Census Bureau’s. Finance puts the county at 9,853,435 for 2025 where the Bureau puts it at 9,694,934, a difference of 158,501 people, because the two agencies build their estimates from different vital records and different methods. Nothing here identifies which of the two is closer. The two official series disagree by more than the population of Pasadena, which is worth holding in view before comparing any two projections built on them.

Read on its own terms, the state forecast says Los Angeles County loses about 1.44 million people over forty-five years and finishes at 8.4 million, roughly where it stood in 1987.

What the state assumes about migration

A projection is an arithmetic identity. Population next year is population this year, plus births, minus deaths, plus net migration. The assumptions attached to those four terms are what a projection turns on, and Finance publishes theirs in a companion file, report P-CC.

Here is what it assumes about migration.

The Census Bureau recorded net migration for the county at negative 183,060 in 2021, negative 84,278 in 2022, negative 37,686 in 2023, negative 7,370 in 2024 and negative 76,146 in 2025. Across 2022 to 2025, leaving out the pandemic year, that averages negative 51,370 a year.

The Department of Finance assumes that from 2030 through 2069 net migration averages negative 3,047 a year. In only three of the fifty projected years is it positive at all. The outflow does not reverse in their file. It very nearly stops.

That assumption is a judgment rather than a measurement. It holds that an outflow recorded in every year of the past decade slows to near zero within about five years and stays there for forty. It is the single assumption the official forecast is most sensitive to, and a reader who disagrees with it should expect a materially different number.

Births, deaths and the age structure

The migration assumption is contestable. The vital rates underneath it are measured.

Los Angeles County recorded 91,048 births and 68,467 deaths in the year to July 2025. Measured against the number of women aged 15 to 44 living in the county, those births imply a total fertility rate of 1.28. Replacement, the level at which a population holds its own without migration, is about 2.1. The county is running at 61 percent of replacement.

The national rate on the same arithmetic is 1.59, itself well below replacement. Los Angeles is about twenty percent further down the same road the country is on.

The county’s current numbers make the present look healthier than it is. Los Angeles County still has more births than deaths, a natural increase of 22,581 in 2025, and its rate of natural increase per thousand residents is actually higher than the country’s, 2.3 against 1.5. That gap comes from age rather than fertility. Los Angeles County records 7.1 deaths per thousand residents against a national 9.1 because its population is younger, and it is younger largely because immigration keeps delivering people in their twenties and thirties. Natural increase here therefore rests on that age structure, and it will diminish as the population ages.

A cohort-component model, single years of age

We built our own projection so the assumptions could be changed and the results checked. It is a cohort-component model: single years of age from 0 to 120, annual steps, males and females carried separately, run from a Census Bureau Vintage 2025 base with the age structure taken from the 2020-2024 American Community Survey.

Three checks matter for a projection of this kind.

Mortality is fitted and then tested against a number it was not fitted to. We solve a Makeham-Gompertz schedule for a target life expectancy of 81.5 years, then apply it to the county’s actual age structure and check the deaths it produces. It returns 68,993 against the 68,467 the Census Bureau recorded, an error of 0.8 percent.

The model has to converge to its own theory. Run it four hundred years and its growth rate must equal the rate the renewal equation gives for the same fertility and mortality. It does, to within four one-hundredths of a percentage point a year. Without that check a projection is only an extrapolated trend.

It runs in single years of age, not five. A five-year model on identical rates understates the long-run decline by 0.135 percentage points a year, because coarse age grouping distorts the average length of a generation. Over a century that is a thirteen percent error, and it points the flattering way.

The open-ended age group is checked every run. An unpublished draft used the wrong survival ratio for the 85-and-over group, which created elderly people out of nothing at every step and overstated the projected over-65 share. On the corrected schedule the county is 27.6 percent over 65 in 2125, and the model now refuses any survival ratio above 1.0.

What our model says

Fertility is held at the measured 1.28 in every scenario below, so the four differ only in migration. That is deliberate. Mixing a fertility assumption into a migration scenario hides which one is moving the answer.

Migration enters the model as two separate rules, and the distinction matters for reading the century column. Domestic out-migration is a rate: the 2022-2025 average of negative 1.21 percent of the county’s population a year, which falls in absolute terms as the county shrinks. International migration is a level: a fixed annual count, held at whichever year or average the scenario names. Net migration is therefore not fixed in any scenario. On continued migration it starts near negative 50,000 a year and approaches zero by 2125, when a shrinking domestic outflow roughly cancels the fixed international inflow at about 5.5 million residents.

The no-migration row runs above continued migration for sixty years and then crosses below it in 2086. That crossing is a consequence of the two rules rather than a finding about Los Angeles. Closing the border in both directions removes an international inflow that is fixed in size, and over a century that fixed inflow outweighs a domestic outflow that shrinks along with the population.

On continued migration the county reaches about 7.6 million in 2070, which is 21.6 percent below today against the state’s 14.6 percent. It passes below 9 million in 2045, below 8 million in 2063 and below 7 million in 2083.

The decline is back-loaded, and that matters more than the endpoint. The 2030s cost the county 3.3 percent. The 2060s cost it 6.6 percent. Nothing dramatic happens for fifteen years, which is why the current numbers look flat. In our model births fall below deaths in 2043. Finance, whose vital assumptions are gloomier than ours, has that crossover in 2032.

The county’s age structure changes more than its size. Between 2025 and 2075 the model loses about 1.07 million residents under twenty and gains about 537,000 aged 65 and over. The over-65 share goes from 15.2 percent to 27.3 percent.

Our 2070 figure is about 808,000 below the state’s

Our 2070 projection is about 808,000 below the state’s: 7,604,902 against 8,413,331. To separate the sources of that difference, we re-ran our model with the Department of Finance’s own year-by-year net migration substituted for ours. Cumulative net migration then matches theirs to within 99 people over forty-five years. The gap that survives is 488,679, and it breaks down like this.

Finance assumes an implied total fertility rate that stays near 1.30 for the whole projection, drifting from 1.24 in 2025 to 1.29 in 2069. Ours holds the measured 1.28 but applies it to a shrinking cohort of women, which is why their cumulative births run 192,744 ahead. Their fertility assumption is the more conventional one. Demographers usually assume rates settle at a floor rather than falling forever, and we would not call it wrong. It is simply an assumption, and it is worth about two hundred thousand people.

There are two disagreements, and the vital rates are the larger. Forcing both models onto the same migration path leaves 488,679 of the 808,429 gap standing, which puts migration at 319,750, or 40 percent of it. The starting count and the birth and death assumptions carry the other 60 percent. Migration is the assumption we disagree with most sharply; it is not the assumption that moves the most people.

The forecast has been revised down every time

There is a reason to think the official number is not yet finished falling, and it comes from the regional planners rather than from us.

The Southern California Association of Governments builds a growth forecast for the six-county region every four years. In April 2026 its own presentation to the Modeling Task Force carried a slide titled, in plain language, that the expectation for population growth has consistently slowed. Across three planning cycles the projected change for the region between 2024 and 2050 has gone:

SCAG has revised its 2024 to 2050 forecast downward in each of the three planning cycles shown, from fourteen percent growth to one percent decline.

Past the edge of the official map

The state’s file stops at 2070. Ours does not, and the numbers past that line are ours alone. Read them as fixed-assumption scenarios rather than as forecasts: a hundred years of fixed vital rates is arithmetic about the present regime, and it assumes away every feedback that would actually bite. Housing costs fall when a place empties out. Immigration policy reverses. Fertility can turn. We would not defend either figure below as a prediction of the county’s population.

On continued migration the county reaches about 6.25 million in 2100 and 5.56 million in 2125, a loss of 42.7 percent over a century. If immigration stays where federal policy left it in 2025, the same century ends at about 3.19 million, a loss of 67.1 percent.

What survives across the scenarios is the ordering rather than the level: every scenario declines, and the ones that decline slowest are the ones with the most immigration in them.

Holding the county at 9.69 million takes a 64 percent fertility rise

Two break-even calculations put the scale of it in view.

Fertility. To hold the county at 9.69 million on births alone, with no migration in either direction, the total fertility rate would have to rise from 1.28 to about 2.1. That is a 64 percent increase, and whether any policy could produce one is outside what this model can say.

Immigration. Holding domestic outflow at its 2022-2025 rate, the county needs 107,010 net international migrants a year to stand still through 2050, rising to about 132,000 a year to hold through 2125. The Census Bureau recorded 92,259 in 2024, which was the best year in the series, 66,826 as the 2022-2025 average, and 29,325 in 2025 after federal policy changed. The county needs roughly 1.5 times its best recorded year, sustained for decades, simply to stay where it is.

The measured record and the model point at different terms of the same identity. From 2021 through 2025 the county’s births exceeded its deaths by 93,796, domestic out-migration removed 667,953 residents, and international migration added 279,413. Over the measured period it is domestic out-migration, not natural change, that produced the loss. In the model that reverses: births fall below deaths in 2043, and from the early 2070s natural decline is the larger of the two terms. Both run through international migration, which is set in Washington rather than in Los Angeles, and which the breakeven arithmetic above puts at 107,010 a year to hold the county level through 2050 against a 2022-2025 average of 66,826.

Limits

Two official estimates of today disagree by 158,501. We built on the Census Bureau’s because its components of change are published and auditable. Building on the state’s would raise every figure here by about 1.6 percent and change nothing else.