 |
| iStock |
A Few Prerequisites
Have you reviewed our previous material on Global Demographics? We will assume knowledge of, inter alia, basic definitions and concepts; historical and international contexts; Malthus and his critics; population pyramids; how the UN projects population (including second moments); drivers of births, deaths, migration; age distributions. Review that material here.
Quick and Dirty: dragged quickly from the Internet. Data will vary by source, year; consider these illustrative approximations rather than precise estimates
The U.S. is one of about 200 countries, albeit one of the largest in terms of population (ranked number three behind China and India), per capita GDP (ranked number six at market prices, behind Luxembourg, Ireland, Switzerland, Norway and Singapore; number nine in purchasing power parity, behind the previous five plus several oil exporters whose rank bounces with hydrocarbon prices), with the strongest military capability (but also with wide commitments) and strong, although recently declining, financial and soft power.
Aggregate Population I: Time Series
Looking at demographic and economic data often reminds me of a well-worn metaphor: blindfolded people examining a (hopefully quiescent) elephant.
Each gets a piece of the picture, but fails to see the whole thing. When we have a very important collection of data, re-expression can help.
There are often a number of transformations we can use to get different “views of the In the next four charts we will examine the same basic U.S. population data four ways, using a simple linear chart; next, use a logarithmic chart; then examine changes (differences) in annual population; then annual growth rates.
Here’s annual data on the U.S. resident population for about 140 years. Because so many Americans went overseas during WWII – and also because births were low, as we’ll see – you can see the little “bump” in the chart in the 40s.
Now, here’s a question. Look at the line carefully. Compare the period 1880 to the 1940s; and then the post 40s population. Which period was growing faster?
Notice that in the early years, for (say) a year or a decade’s change, the rise in population is small, relative to “the rise over the run” in say the last few decades.
In other words, the slope of this line is steeper in recent decades than a century ago.
But if we want to know when the fastest growth appears, we don’t want the slope of this line – we want the percentage change. That will be the slope, at some point, DIVIDED BY THE STARTING LEVEL. The slope is smaller a century ago – but we would also DIVIDE THAT SLOPE BY A SMALLER BASE! (Sorry for yelling.)
Our eye tends to deceive us when we look at linear trends in a series, especially over a long time span.
When you plot the log of a variable, the slope of the line is now (to a good approximation), the rate of growth.
Now we see that the slope of later years is a bit flatter, i.e. that the growth rate is actually slowing down.
Need a refresher on logarithms? Excellent! Download this PowerPoint deck on logarithms and exponential growth.
Here is another way to look at our basic population data, focusing on annual changes rather than levels. We’ve simply differenced the annual population figures in the previous population charts.
You see the smaller simple differences pre-WWII, though the base was lower, as already noted. It’s also interesting to see the effects of two of our three most deadly wars.
WWI was bad enough, but the 1918 flu epidemic killed roughly 10 times the number of Americans as combat. Globally it might have killed 3 or 4 percent of the world’s population. It was a terrible pandemic. (Have you had your flu shot this year?)
WWII had a higher number of casualties but was also a bigger war. About 5 million Americans were in service in WWI, out of a total population of 103 million. The U.S. lost about 120,000 servicemen and women in WWI, about half from combat, and half from influenza and other non-combat causes.
About 16 million were in the service during WWII, out of a total population of about 133 million. Most of the 400,000 service deaths in that war were combat-related.
As many of you will know, the deadliest war in U.S. history was the 1861-5 Civil War, with about 700,000 fatalities out of a total population base of a little over 30 million, over 2 percent of the population.
The Civil War was the most horrific death toll from a U.S. war, but globally other conflicts could be even worse. Notably, around the same time as the U.S. Civil War, the 1850-1864 Chinese civil war between the Qing dynasty and the Taiping Heavenly Kingdom (revolutionaries led by religious leader Hong Xiuquan) has been estimated to have cost 20 to 30 million lives, or roughly 5 to 10 percent of China’s population at the time.
Here’s one of my favorites, combining two transformations. We took the first chart, and superimposed the annual growth rate.
U.S. resident population has been growing at just about 1 percent per year for some time. This is in contrast to a number of other developed countries in Europe, Japan, that are growing slowly if at all. For example, Germany and Italy’s growth rate is near zero, Japan’s just a shade above (0.2 percent per year); France is about 0.6 percent per year (about the same as China!). Recently, the U.S. has been growing at about the same rate as (wait for it!) Mexico.
Going forward, there is some uncertainty whether we'll stay anywhere near 1 percent, or even half that. I think it's more likely, than not, that U.S. population growth is going to fall a good deal further. Why? Further reductions in fertility, and changes in immigration policy. More on those topics below.
Aggregate Population II: Displaying the Distribution by Age and Sex
 |
| Photo by SM |
Now let’s look at another handy demographic tool – the pyramid!
If you read through my slides/blog post on global demographics, you know I love population pyramids.
If you need a refresher, click here.
This pyramid uses recent single-year age population data for the United States, instead of the 5 year groupings above.
The pyramid confirms that Millennials are a large cohort. The other large cohort is the post-World War II “Baby Boom,” as already noted above. The in-between “Gen X” cohort is smaller, sometimes referred to as the “Baby Bust” in contrast to the post World War II Baby Boom. The Millennial’s dominant position will increase over the next several decades, as Boomers age and a noticeable fraction of us die off.
Millennials or “Gen Y’ers” are sometimes called the “Echo Boom,” although the large size of this cohort is not any explosion of fertility rates, but rather modest fertility among a large cohort as Boomer women reached childbearing age.
More detailed demographic studies show that immigration made an important contribution to the Millennial cohort, both directly (many Millennials are immigrants) and indirectly (a significant fraction of Millennials were born to first- or second-generation immigrants).
As it happens, I’m not a huge fan of these population cohort groupings, but they are now ubiquitous. So I’ve given up and now use them myself. More than I should, probably.
Let's switch gears again, looking at pyramids from the UN World Population Prospects website, which we used extensively in the previous post on global demographics.
This figure represents U.S. population in 1950. It's somewhat pyramidical, bigger at the base (kids), smaller at the top (oldsters). But note the indentations representing the fertility decline during the Great Depression and the world wars; and the very large base just emerging as the postwar baby boom begins.
I haven’t shown up just yet – I was born in 1952.
Now we’ve skipped ahead 75 years from the UN’s first pyramid, the year in which we write. Today the chart is not really much of a pyramid anymore, it looks more like a distorted fireplug. The area has grown as UN estimates 2025 US population to be 347 million.
Also notice the shrinking base. Children still exist, but seniors are now becoming a much larger share of those dependent on the working age population; loosely, those between, say 20 and 65.
Barely noticeable: at the very bottom of the figure is just a little bit of green. Since this data was published in 2024, based on historical data up until that point, UN demographers had to estimate the number of new births in 2025. Given the relative stability of trends in fertility, infant mortality, and so on, the confidence interval for this projection one year forward is pretty tight.
Now skip ahead 30 years from the present. Total projected US population is 385 million, if we use the median UN projection.
Focusing on the median projections, the best estimates from simulation models described briefly here and in more detail here.
The pyramid has clearly inverted. Loosely, from age 45 and up we see something like a pyramid; but below 45, populations decline as age decreases.
Notice we have much more green and yellow. The green represents the 80 percent confidence interval – projected US population between 353 million and 414 million – and the yellow represents the 95 percent confidence interval – between 336 million and 434 million.
Recall that these green and yellow areas represent confidence intervals from model simulations. Clearly these are much wider for the bottom half of the figure than the top. Why? Because we have a good estimate of today’s population; the future numbers of middle-aged and above will vary with future migration and death rates, which are not known precisely. But it turns out that forecasting fertility, which drives the bottom confidence intervals, is even less precise, so the bottom confidence intervals are usually much wider.
Also, notice that the lower 95 percent interval suggests there is a small probability that U.S. population will shrink between now and 2055.
If you have time, pull up our companion study of global demography (or go to the UN website, that will only take a moment) and compare this pyramid to the global pyramid for the entire world population in the same year, 2055. You’ll see that proportionately, the confidence intervals for global projections are smaller than that for the U.S.
Why? Any given country will tend to have proportionately larger confidence intervals than the entire world. Some country variations will cancel out, but more to the point, country estimates are affected by the necessity to estimate volatile international migration. At the global level, these cancel out.
We can treat interplanetary immigration, on the other hand, as negligible. And if Martians do show up, H.G. Wells predicts that their lack of immunity to Earth’s pathogens will do them in. Or possibly a deep aversion to yodeling?
If Martians can’t hack it here, perhaps there are distant aliens with sufficiently advanced civilizations to understand the germ theory of disease, vaccinations and antibiotics. (And no equivalent of RFK Jr in charge of their health care*). Should we be concerned about immigration at the global level? I devoured Liu Cixin’s Remembrance of Earth’s Past trilogy (The Three Body Problem,** The Dark Forest, Death’s End), but the Trisolarians won’t be here for about 400 years so I’m going to neglect the effects of aliens on global population today.
Emigration at the global level? Yes, Elon, I will gladly chip in to buy your one-way ticket to Mars. Take some friends. Please.
Whatever Elon does, another dozen kids, or two; trip to Mars, or not; his personal effect will remain small relative to total world population. He may have a larger effect through DOGE’s gutting of public health programs, but that is for another discussion.
*Click here to learn about RFK Jr.'s rejection of the germ theory of disease, which has been understood since the 19th century work of Louis Pasteur, Joseph Lister and Robert Koch.
** Ive watched the excellent Chinese TV series Three Body; Netflix has made a popular film as well, albeit one that takes more liberties with Liu's original story.
By the end of the century, our confidence intervals are getting quite large.
The UN projects that in 2100 the US will contain 421 million residents (the median projection. The 80 percent interval for total population is now between 340 million and 537 million. The 95 percent interval is between 316 million and 625 million. Pretty wide!
Here we see data from the same UN projections, but a different perspective. While we lose detail on age and sex distribution, we get a much better look at the evolution of actual population and its projections over time.
Focusing first on the thick red line, the median projection, U.S. population grows until the end of the century, albeit at a slow and declining rate. As expected, the confidence intervals, at whatever levels, grow much wider with each passing year, as discussed above.
Compontents of Change I: Natural Increase (Fertility, Mortality)
Some obvious patterns from this chart include the following.
With the exception of the 1918-1920 flu epidemic and the recent COVID pandemic, the US death rate was been on a long slow decline for the first 75 years or so of the data. The death rate almost flattened out until around 2010 when it began a slow rise, driven in part by the large baby boom entering its dotage. (And some increase due to "deaths of despair," and especially COVID-19 -- see below.)
The birth rate is much more volatile. A rapid decline preceded World War I and the Great Depression, continued through those trying times, and then boomed after World War II. Birth rates declined rapidly during much of the 60s and 70s, then continued to decline at a slower rate.
The difference between the red and blue lines is, of course, the natural increase in U.S. population.
In the pandemic years 2020 and 2021, fertility rates fell a bit and death rates rose almost to the birth rate i.e. during the pandemic population growth practically ceased. It's now recovered slightly.
Let's look at the components separately. First, births.
(How to calm a crying grandson? Read quietly from an economics tome, here, “Too Big to Fail.” Turned him catatonic within minutes).
Nobody has precisely defined the terms baby boomer, Gen Xers, or Millennials. Or even agreed on the terms, there are synonyms for the latter two.
I usually think of the boomers as those born from about 1947 until 1967; Gen Xers as 1968 to 1981; and Millennials as those born 1982 to 2004.
By the way, while there were certainly precursors, the current division into some version of these “generations” stems from the pop sociology of gurus William Strauss and Neil Howe, in their eponymous 1991 book: Generations: The History of America's Future, 1584 to 2069. Al Gore is a big fan, I’m less enamored of it. Perhaps I’m being a little harsh?
I would argue the division into these groups is mildly interesting and sometimes useful. Certainly, I will claim Boomers Rule the Universe. (Full disclosure, I’m a Boomer, so give that claim all the respect that deserves. Which is not so much.) Really, not a huge fan in terms of analytics, but I use the terms as well, they have passed into the lexicon.
You might have run across Strauss and Howe in another context. Strauss (who died in 2007) and Howe (who continues to consult and give speeches based on their ideas) took their generations idea much further, arguing that history was a series of recurring 85-year cycles, in their 1997 book The Fourth Turning: What the Cycles of History Tell Us About America's Next Rendezvous with Destiny. While many quibble with Generations, reviewers, scholarly and popular, were often more critical of The Fourth Turning; personally, I think it is mostly gobbledygook, other reviewers have labeled it “pseudoscience” and “non-falsifiable” (a pretty damning insult if you’re a fan of Karl Popper, see our discussion of “How to Think Good.”) I mention all this because if you follow current political events you’ll note that Strauss and Howe have deeply influenced anarchist Steve Bannon.
Total fertility rate represents the number of children that would be born to a woman if she were to live to the end of her childbearing years and bear children in accordance with current age-specific fertility rates.
Data sources post 1960: FRED, from World Bank World Development Indicators.
A Total Fertility Rate of 2.1 is often cited as “replacement fertility,” i.e., the fertility rate sufficient to maintain a country’s population, neglecting immigration/emigration. A total fertility rate of 2.1 is often cited as "replacement fertility." But the actual equilibrium TFR varies with a country's life expectancy, deaths of infants, etc.
Goal: each woman is, on average, “replaced” by one daughter in the next generation. In low-mortality countries:
- Sex ratio at birth (SRB): ≈ 105 boys per 100 girls, so the share female at birth is ~100/(100+105)=0.488.
- If every woman survived through her childbearing years, the daughters per woman would be TFR × 0.488.Setting that to 1 daughter ⇒ TFR ≈ 1 / 0.488 ≈ 2.05.
- Because not all women survive the entire reproductive span, add a margin for female mortality, ~2.06–2.10 in most high-income settings.
Thus, the origin of the rule-of-thumb “2.1.” “2.1” is a good guide in countries with low mortality of potential reproductive females and a sex ratio at birth of about 105; with a stable timing of births (little shift to later ages), no bias against females in nutrition, other necessities. Assumes a closed population (ignoring migration). It’s a long-run equilibrium, not capturing short term variation.
Life is not so dire today, absent a major war, but a number of poor countries have replacement fertility rates noticeably above 2.1, say around 2.3 to 2.6, such as Afghanistan, DR Congo, Somalia, and so on.
WDI’s 2023 TFR estimate is 1.6. U.S. TFR from this source has been below 2.1 since 2007, and is still falling.
Now let's examine the other side of natural increase (or decrease): deaths.
(Please excuse the shameless plug for my brother's business -- now run by his former employees).
While deaths -- unlike births, above -- have been more or less rising over time, as we saw above, the death rate fell relatively rapidly (omitting the 1918-19 influenza pandemic), fell more slowly until about 2010, when it began to increase again, slowly – until COVID-19 hit. (The virus appeared in 2019, hence the name, but it began to turn up in death statistics in 2020, see details below).
Also, notice that year-to-year volatility in the number of deaths is noticably lower beginning circa 1960, even if we ignore the "Spanish" (Kansan?) flu epidemic spike in 1918-1919.
This reduction in annual volatility was largely due to the development and more widespread use of vaccines and antibiotics; the strengthening of public health systems, and (especially after 1965) wider access to health care.
Prior to 1960, there were many more deaths from infectious diseases: influenza, pneumonia, tuberculosis, measles, diphtheria, polio -- and mortality from these was especially concentrated among infants and children, amplifying year-to-year swings as many of these diseases themselves had large seasonal and annual variations. Circa 1960 (beginning beforehand, of course) the development and more widespread use of vaccines and antibiotics played a major role in reducing death rates but also reducing the observed volatility. Other contributions were made by the strengthening of public health systems, (especially after 1965 Medicaid and Medicare Acts) wider access to health care; and the benefits associated with growing incomes, including better nutrition, heating and cooling, and so on.
The recent increase in deaths is very noticiable in the data. The most obvious spike is contemporaneous with the worst of the COVID-19 epidemic, circa 2020 through 2022, though the virus is still circulating as of this writing. This chart compares U.S. COVID deaths to prior flu seasons, for scale:
There is a lot to say about COVID that would take us too far afield here, including discussion of measurement issues. Order of magnitude, COVID cost the U.S. about a million lives. Estimates of global lives lost from the pandemic range between 10 and 20 million. While somewhat dated, other blog posts during the first few years of the pandemic can be found here, and a paper discussing COVID's effects on global housing affordability is available here.
A Digression on "Deaths of Despair"
The COVID spike is not the only driver of increasing U.S. deaths. In 2015, Anne Case and Angus Deaton flagged distressingly large increases during the 2000s in death rates from drug overdoses, cirrhosis and other diseases linked to alcoholism, and suicides, especially among middle-aged Whites.
Case and Deaton argued that this increase was associated with declining social and economic conditions, especially among Whites living in rural areas, and/or with lower levels of education. They thus labeled these “Deaths of Despair.” This figure from their 2015 article was a startling revelation to economists and others who had not been previously following such data:
In the data analyzed by Case and Deaton, which ended in 2013, Black (non-Hispanic) mortality rates were higher than White (non-Hispanic) rates, but falling; Hispanics were in between Whites and Blacks, and falling.
Later work by Case and Deaton (2022) and Friedman and Hansen (2024), among others, found Deaths of Despair continuing to rise, partly driven by fentanyl driving deaths even higher than previous criminal over-prescription of opiods, and heroin.
 |
 |
In time, as prescriptions tightened and addicted users found those drugs expensive and difficult to obtain, use of other drugs, notably heroin, and then fentanyl (often combined with other drugs) took off.
Currently fentanyl is by far the leading cause of drug overdose deaths, but note that deaths from other drugs, including opioids, have increased; deaths from meth and cocaine are much higher than they’ve been earlier this century.
One category of alcohol-related deaths reported in Deaton and Muellbauer is direct alcohol poisoning, which is combined with drug overdoses in the "poisoning" category. They also report data on alcohol-related liver diseases (cirrhosis, alcohol-related hepatitis). Pan et al. (2025) present the aggregate rise in alcohol-related liver diseases for recent years, which show a dramatic increase over the past decade or so:
Much more detail about alcohol=related deaths can be found from the CDC here. In brief, and rounding, recent data reveal that direct alcohol poisoning results in about 3,000 deaths per year; alcohol contributes to about 22,000 deaths from drug poisonings. Chronic liver diseases kill over 30,000 in some recent years. Alcohol is listed as a contributing factor in perhaps 10,000 suicides every year. Unsurprisingly, Case and Deaton's three main categories of Deaths of Despair (poisonings, chronic alcohol related deaths, suicides) have significant interactions.
We should note that alcohol is involved in many deaths usually omitted from DoD: other disorders including psychosis, certain cancers, and of course alcohol-related traffic fatalities; together these might be responsible for another 45,000 deaths or so.
Suicide rates have been rising in the aggregate as well:
Suicide rates in the U.S. have reached a 70-year high, with an estimated age-adjusted rate of 14.7 per 100,000 in 2024—marking a 37% increase since 2000.
Firearms remain the leading method of suicide, responsible for over 54% of deaths in 2022
Middle-aged adults (35–54), elderly individuals (85+), and American Indian/Alaska Native populations show the highest suicide rates
Youth suicide attempts are rising sharply, especially among teen girls and marginalized groups, with 10% of high school students reporting attempts and 13% of females affected.
In recent years, Deaths of Despair are increasing for all races, and for Hispanics. The death rates among American Indians and Alaskan Natives, previously understudied, is particularly horrendous.
Data representing American Indian or Alaska Native, Asian or Pacific Islander, Black, and White groups refer to non-Hispanic individuals in each group. Panel A shows deaths of despair among individuals aged 45 to 54 years.
*Increases for Asians are modest over a comparatively low base. This figure combines Asians (themselves a diverse population) with Pacific Islanders and Native Hawaiians; PI/NH have higher rates of Deaths of Despair than Asians.
There is much more to learn about Deaths of Despair, and changes in causes of death more broadly. But in this post Case and Deaton get the last slide of the section, which updates some of their 2015 findings. Spoiler alert: the update news is not good:
Mortality rates by race, sex, and education, age-adjusted 25–74. Abbreviations: BA, bachelor’s degree; BNH, Black non-Hispanic; WNH, white non-Hispanic.
“Deaths of despair, morbidity, and emotional distress continue to rise in the United States, largely borne by those without a college degree—the majority of American adults—for many of whom the economy and society are no longer delivering. Concurrently, all-cause mortality in the United States is diverging by education in a way not seen in other rich countries.”
End of Digression on DoD; A Few Comments on U.S. Life Expectancy
Here you see the rise in U.S. life expectancy over the past century.
Among other patterns – the fastest rise was in the first half of the data (though the variance was also much bigger). Things smoothed out, but slowed down, after WWII.
Notice women live longer than men. Men face earlier onset of heart disease, are more likely to die from violence, hold more (but not all!) of the dangerous jobs, and engage in more life-shortening behaviors like smoking (though sex differences in smoking rates are disappearing) and doing violence to each other.
Blacks have shorter life expectancies than Whites (though Black women have had about the same life expectancy as White men since around 1970).
Notice this is data on the estimated life expectancy at birth. As your cohort ages, if you survive your life expectancy edges up (though of course you get closer to the terminal date).
For example, when I was born the life expectancy of a white male born that year was a bit under 67.
Now that I’ve made it past 60, my life expectancy has shifted up to about 85. (And I intend to be one of those above the average, with some help from my wife (married men live longer), my trainer, and good medical care.)
Now that we’ve had a brief look at births and deaths, let’s examine the third main component of population growth:
Immigration
First, take note that each immigration scenario is presented using the median projection for that scenario – no confidence intervals are presented; as discussed above, these would widen substantially over time. While simplified, this graphic does highlight how much difference changes in immigration policy and behavior might change the time path of U.S. population.
The “Main Immigration” scenario is based on recent history (neglecting pandemic years), and assumes annual net immigration levels between 850,000 and 980,000 people.
The “High Immigration” scenario assumes a consistent annual net immigration of roughly 1.5 million people per year.
The “Low Immigration” scenario assumes a trajectory of between 350,000 and 600,000 net migrants per year (similar to latter years of the Trump I presidency).
The “Zero Immigration” scenario, self-explanatory, is about where we are as of this writing (the first year of the Trump II presidency). Although it would be difficult to imagine zero immigration for the next 75 years, it does provide a baseline for considering the role of immigration in U.S. population growth or decline.
Before we examine recent immigration to the U.S., let us begin with some historical background.
 |
| National Geographic |
The first human migration to North America occurred during the late Pleistocene (circa 15,000 years ago?), when hunter-gatherers moved from northeast Asia into the Americas, most likely via Beringia, a land bridge exposed during periods of low sea levels. Migration likely occurred in multiple waves over several millennia, using both interior ice-free corridors and coastal routes.
At some point (10 millennia ago?) North American population growth thereafter was driven almost entirely by natural increase, not continued large-scale immigration from other continents.
Immigration from other continents resumed with a vengeance after the European (re?)discovery of North America
First we summarize pre-Civil War European immigration, from 1600 to 1850. Early migrants to what later became the 13 American colonies came primarily from England, followed by substantial numbers from Scotland, Ireland, the Netherlands, Germany, and later Scandinavia. Other significant migration included French settlers in what later became Canada as well as the Great Lakes and Upper Midwest, the Mississippi Valley and lower Louisiana. Spanish immigrants settled in Forida, the Southwest, and California.
Migration motives included land acquisition, religious freedom, trade, and, in some cases, penal transportation.
By the late 18th and early 19th centuries, annual inflows increased, particularly from the British Isles and German-speaking regions. By 1850, the United States Census recorded roughly 2.2 million foreign-born persons, almost entirely European in origin.
Two other major events ran somewhat concurrently. Approximately 400,000 Africans were enslaved and transported to North America, out of roughly 12 million sent to the Americas overall. The United States banned the international slave trade in 1808, but slavery itself expanded internally through natural increase and forced internal migration (the domestic slave trade) until abolition in 1865. By the time of the Civil War the U.S. contained about 4 million slaves.
We have less reliable data on American Indian populations, but Thornton (1987) estimates peak pre-Columbian population at about 5 million, declining precipitously to perhaps a hundred thousand or two by the late 19th century, from wars, genocidal policies, and especially disease, before beginning a recovery.
For more detail, and references, on these early demographics, see the PowerPoint slides and references that you can download near the end of this post. Next we turn to the past couple of centuries, focusing on two waves of immigration, one roughly from post Civil War to the 1920s, and the second recent wave that began in the 1980s.
1850–1910 was the time of the “Great Wave” of immigration, primarily from Europe. The foreign-born share rose to the mid-teens as arrivals surged.
1920s–1960s: Quotas and restrictions, the Great Depression, WWII, and relatively low inflows steadily reduced the foreign-born share. The foreign-born share reached a modern low around 1970, when the Census Bureau reported 9.6 million foreign-born (4.7% of the U.S. population).
1970s–2024: The foreign-born stock rose for five decades, reaching modern record levels by 2024. Pew Research Center estimates the foreign-born population peaked at about 52 2024. During this period, the origination of immigrants shifted markedly: previously immigrants were mostly European, recent decades have been dominated by Latin America and Asia; Pew summarizes these origin shifts since 1850 and highlights that by 2022 the largest origin group was Mexico, with India among the next largest.
In January 2025 the second Trump administration began to institute new restrictions on immigration, and deportations, leading to an estimated decline of about 1.5 million foreign-born by mid-year. More on recent policies below. Let us examine some recent research comparing immigrants from the two great waves in the figure above.
Ran Abramitzky and Leah Boustan are two of the country's leading researchers on immigration. Among their projects, Streets of Gold reports on a major research project they undertook using "big data" from a century of data.
Abramitzky and Boustan make use of large micro census data sets from the Census and Ancestry.com. applying machine learning techniques to link records categorize the immigration and economic status of individuals; and then track those individuals and their progeny across decades.
It's long been understood from previous research that post 1980 "reform" immigrants are a "barbell" distribution, with highly educated/human capital individuals overrepresented in the immigrant population, as our less well educated/lower human capital individuals.
Abramitzky and Boustan find that generally, new immigrants themselves don't improve their economic lot substantially from their place on arrival in the United States. But their children systematically do better than similar cohorts of native born. Their grandchildren often do better as well, but by the next generation the performance of the progeny of immigrants and complete the native-born cohorts do about the same.
While they find that once landed, the first immigrants don't systematially move up economically, of course those first landed usually do improve their lot over their circumstances in their previous country.
 |
| Malpezzi family photos, a century ago |
Case in point, my paternal grandfather left dire poverty as a tenant farmer in a remote part of Tuscany in 1900, to take up coal mining in Western Pennsylvania. His family standard of living increased, once my grandmother, my eldest uncle and eldest aunt joined my grandfather a few years after his emigration. This was despite low wages and high retail costs in the local-monopoly-run coal town. Their incomes were supplemented by the family vegetable garden; after a day of mining coal with pick and shovel, my grandfather turned every square inch of their company house’s yard into a hand cultivated garden. Five more children were born in America. Bottom panel: my grandmother with her seven surviving children, my father to her immediate left.
My grandfather’s male children, including my father, started out in the mines, pre-World War II. This was a hard life made more difficult by the Depression, labor strife, and most notably by my grandfather's 1929 death in a mine roof collapse. His sons obtained more education than my grandfather ever had (with one exception, my oldest uncle, explained below). Before long they eagerly moved from mining into construction and wholesale distribution, interrupted by military service in two world wars. The daughters took up domestic work for a wealthy Pittsburgh family, then moved into factory and clerical work.
The next generation, me, my brothers, and my cousins, moved further ahead, teaching at different levels, working in different areas of retail, construction, law, and a successful funeral home. One of us even stumbled into a PhD in economics.
That exception to the education of my father’s generation, noted above? My oldest uncle (top right photo, during his WWI U.S. Army service) had started school in Italy, and was tested by the local elementary school on his arrival in their new hometown. Since he only spoke a dialect of Italian at the time, they thought him stupid. When he nevertheless made a perfect score on his entry math test, they then were sure he cheated. (How exactly he might have done so was a puzzle that did not seem to occur to them.) Offended, and properly so, by the accusation, he walked out. Self taught, probably with some later help from siblings, he became a successful small contractor, nevertheless.
One major difference between the first and second waves of immigration can be seen in this chart.
Previously we examined the stock of immigrants, i.e., the percentage of U.S. population that was/is foreign born. Here we examine the relative contributions of natural increase (births less deaths) and immigration.
These growth rate contributions are spliced together from several sources, each with its own errors. Consider these data approximate and provisional, especially the most recent years, as new data comes online and/or is revised after Censuses and other surveys.
In my pre-”retirement” teaching years (1990 to 2016) I told students that as a rough rule of thumb, U.S. population was growing at about 1 percent, though that rate was slowly declining. And that when we decomposed that 1 percent growth rate, a little over half of it is natural increase (births over deaths); a little under half is due to immigration, mostly legal or documented. I’ve recently extended these estimated contributions through 2024.
What the chart didn’t show in the 90s and 2000s is that immigrants and the children of immigrants had higher fertility rates than those who’ve lived here for generations, as we noted above. So some of that natural increase, that was higher than most peer countries (e.g. Western Europe) was actually immigration once removed.
Another obvious feature is that in the first wave, a century or more ago, is that there was no separate accounting of “legal” and “illegal,” or “documented” and “undocumented,” immigration. Even passports were not generally required until WWI. Governments (including the U.S.) sometimes issued letters requesting safe conduct for travelers, but these were not universal. Of course, immigration officials at Ellis Island and other disembarkation points could and did decide if a foreigner could enter the U.S., and some were refused entry. But the system of passports and visas did not become widespread until circa 1920.
The more detailed PowerPoint deck downloadable below has much more detail about immigration legislation and policies, but for now we note several watershed events.
The Immigration Act of 1924 established national origin quotas. These were based on the 1890 Census, so heavily favored immigrants from Northern and Western Europe while severely restricting immigration from Southern and Eastern Europe, Asia, and other regions.
The Immigration and Nationality Act of 1965, also known as the Hart-Celler Act, abolished the national origins quota system that had been in place since the 1920s and introduced a new framework for immigration that prioritized family reunification, skilled workers, and refugees.
The Immigration Reform and Control Act of 1986 Act provided a pathway to legal status for many undocumented immigrants who had entered the U.S. before January 1, 1982, and had continuously resided in the country since then. Approximately 2.7 million immigrants gained legal status through this program. This legalization would later be derided by opponents as “amnesty.”
Much more detail regarding these three acts, other legislation, and their implementation, can again be found in the detailed teaching notes downloadable below. For now, note the decline in immigration after the 1924 Act (as well as the effect of World Wars and the Depression); and the increases after the 1965 and 1986 Acts.
And of course some very recent events are reflected in the data. Immigration fell, and natural increase fell substantially, during the COVID-19 Pandemic. Natural increase remains low (as we discussed in the section on fertility, above) but immigration, especially illegal/undocmented immigration, increased markedly circa 2022 and 2023. Later in 2024, and especially with the advent of the Trump administration in 2025 (not shown in this chart), immigration has also collapsed. It may be, when the numbers are finalized, that 2025 could be the first year on record when U.S. population actually declined. In any event, the overall U.S. population growth rate is much lower than any of us have experienced in our lifetimes.
Population Characteristics: Sex, Age, Households and Families
This pyramid uses recent single-year age population data for the United States. I noted my objections to these cohort labeling conventions above, but also noted my abject surrender in the face of their ubiquitous use.
The pyramid confirms that Millennials are a large cohort. The other large cohort is the post-World War II “Baby Boom,” as already noted above. The in-between “Gen X” cohort is smaller, sometimes referred to as the “Baby Bust” in contrast to the post World War II Baby Boom. The Millennial’s dominant position will increase over the next several decades, as Boomers age and a noticeable fraction of us die off.
Millennials or “Gen Y’ers” are sometimes called the “Echo Boom,” although the large size of this cohort is not any explosion of fertility rates, but rather modest fertility among a large cohort as Boomer women reached childbearing age.
More detailed demographic studies show that immigration made an important contribution to the Millennial cohort, both directly (many Millennials are immigrants) and indirectly (a significant fraction of Millennials were born to first- or second-generation immigrants).
The “dependency ratio” is the ratio of working age people, to children and the elderly.
Another way we can examine broad demographic trends is to examine this dependency ratio, or the number of “dependents” (children, and those of retirement age) divided by the number for those of working age. This ratio fell through the 80s, as the baby boom entered the workforce, but is now rising again, as they reach retirement age.
Obviously, some under 18 and over 65 work; some 18 – 64 do not. Nevertheless, the ratio is a reasonable rough indicators of at least potential “depending.”
See our previous blog post and associated PowerPoint deck for more discussion of the implications of dependency ratios.
Notice that this slide is a little dated. When I attempted to update the data, I received this notice:
Due to the lapse of federal funding, portions of this website will not be updated. Any inquiries submitted will not be answered until appropriations are enacted. You will be redirected in 10 seconds...
No file, and I was not redirected. Budget and personnel cuts to Census and other Federal data sources are a serious problem, worthy of a separate post. Cuts have been a problem for some time, but have accelerated under the Trump administration. See this report from the American Statistical Association,
A family comprises 2 or more related people living together (one of whom is the householder).
A household comprises all families; plus unrelated individuals living together; plus single persons.
Thus, all families are households, but not all households are families. The U.S. currently cotnains about 82 million families, and 134 million households.
Census counts people living with an unmarried partner (of same or different sex); but unless there is some other relative present (e.g. a son or daughter of the householder), they are counted as households but not as families.
Note that the number of households equals the number of occupied housing units.
Average United States household size has been declining over the past century or so from over four as we turned into the 20th century to just over 2.5 today.
The decline has slowed substantially from about 1990.
While the number of households is been rising, the rate of growth of households has slowed down over the past half-century.
The largest spikes in volatility in household formation are caused by definitional changes in 1982, 1993, 2001 and 2003.
From the 1990s to the Great Financial Crisis (circa 2007), a little over one fourth of young adults lived with their parents. Since the Great Financial Crisis that ratio has climbed and recently rose above 30 percent.
This trend has the attention of housing economists and the housing industry, among others; and of course, America's Finest News Source:
Like most satire, the line between laughing and crying is sometimes hard to discern.
Note that these marriage and divorce rates are "flows" (annual changes in marriage status), not "stocks," e.g. number of married couples. Here is a snapshot of today's stock, counting males and females 15 years and older:
Another notable trend is the increase in children born out of wedlock, although this has stablized in recent years.
Let's finish up this post (well, almost), with one more slide on marriage:
Alois Stutzer and Bruno Frey studied Germans and found that married people were happy, it seems partly because happy people are more likely to marry but also a good marriage increases happiness in turn. They find the "division of labor seems to contribute to spouses' well-being, especially for women..." which is a finding I think I need to keep well in mind.
Thanks to Richard Green for mentioning this paper. I've added some photos of (mostly) well-known marriages.
Left side:
- Samuel L. Jackson and LaTanya Richardson (married 1980).
- Queen Victoria and Prince Albert (married 1840 until his death, 1861).
- Jan Styka, "Penelope Recognizes Odysseus" (1901) after he returns after 20 years of gadding about. Very loosely inspired the plot of O Brother, Where Art Thou?, but in 2026 we expect a film of The Odyssey by Christopher Nolan.
- George Bailey (Jimmy Stewart) and Mary Bailey (Donna Reed). Mary is handing over the honeymoon cash to bail out George’s S&L after Uncle Billy (Thomas Mitchell, the actor not the law professor) misplaced the day’s deposits. Selfless and long-suffering woman bails out the feckless men. Again.
Right side:
- Valerie (Carol Kane) and Miracle Max (Billy Crystal), The Princess Bride. Exact dates unknown, but is sure seems they were married a long time.
- George H.W. Bush and Barbara Bush, married 1945 (until her death, April 2018; his death, November 2018)
- SM and JK, married since 1986.
- Michael McConnell (L) and Jack Baker (R), married in 1971 (long before 2015’s Obergefell v. Hodges, and an earlier state legalization in 2013, thanks to a Minnesota court clerk’s oversight, and a subsequent successful legal battle). Considered to be the longest-married gay couple in the U.S.
Some Final Thoughts (For Now)
Despite its length, this blog post has only scratched the surface of a complex subject. Do review the companion blog on global demography as well, and most importantly of all, download this PowerPoint deck.
The deck is quite large, about 700 slides. Don't bother trying to read it on a phone, or open it in a browser. Download the file to a hard drive, then open it in PowerPoint.
In addition to much more detail on the topics above, you'll find some introductory material on race, ethnicity, and related social and policy issues including public finance, education, health and housing.
Near the end of the deck you'll find a slide that has an extensive reading list in that slide's notes. I'm still studying this subject, some of these readings I've been through carefully, some I'm still working on.
Captain Obvious observes that nobody would try to use all or even most of these slides in a class. Go through them, pick out any that might help in your own teaching, or that might give you an idea for a better presentation. I have been creating slide decks for teaching and other presentations for over 40 years. I've stolen lots of ideas and more than a few slides from friends and colleagues. Do feel free to use any of these, or pass them along.
Your reward for reading this far, and then downloading a ridiculouly large file? Cartoons. Near the end. Also some reviews of movies that touch on U.S. demography.
Comments and corrections on any of this material are, as always, very welcome.
