In other words: how many people have ever lived in the United States?

For most of the country's history, demographic record-keeping was unfortunately far from complete, especially when it came to births. But the census has counted the population since 1790, and combining those counts with historical birth-rate estimates and immigration records produces a defensible number: about 642 million people have lived in the United States since it became independent in 1776. A little over half of them, about 53 percent, are alive today. And of the 546 million children born in the United States, roughly one in eleven, some 49 million, died before reaching the age of five. Had those same children been born under today's conditions, only about three million would have died so young.

What goes into the estimate

The estimate is based on three components: the roughly 2.5 million people already alive at the founding in 1776, everyone born in the country since then, and everyone who immigrated to live there. My starting point was an earlier estimate by jlredford, writing in 2010 on the blog A Niche in the Library of Babel, which placed the born-in-country total at about 472 million and immigrants at about 73 million, for roughly 545 million people up to 2010. That estimate drew its population figures from the US Census and applied historical birth-rate estimates to fill in the births the Census never directly counted. My reconstruction follows the same logic but builds the births figure up more carefully, separating the white and Black populations, using directly recorded births once those exist, and extending the whole count forward to 2026.

Reconstructing the number of births

Births are the overwhelming majority of the total, so they are worth getting right. For the period before national birth registration existed, the number of births cannot be looked up. It has to be reconstructed from two things that are reasonably well documented: how many people were alive, and how often they had children.

The demographer Michael R. Haines has published a decade-by-decade crude birth rate series for the United States reaching back to 1800, given separately for the white and Black populations. This separation matters, because through most of the nineteenth century the Black population, the large majority of it enslaved, had a meaningfully higher birth rate than the white population.

TABLE 1. Reconstructed Births by Decade, 1790 to 1900

† Haines’s white birth rate series begins in 1800. The 1790s white rate is carried back from his 1800 value of 55 per 1,000, as there is no earlier figure to average it with.

* Haines’s Black birth rate series begins in 1850, where he records a rate of about 58.6 per 1,000, the highest point in his Black series, which declines steadily thereafter. The pre-1850 Black rates marked with an asterisk are a proxy of 58 per 1,000, on the assumption that fertility in the decades just before 1850 was about the same as the earliest level Haines measured. The enslaved population grew almost entirely by natural increase, especially after the transatlantic slave trade was banned in 1808, so there is little reason to think its birth rate was markedly lower in these earlier decades than at mid-century.

As the table shows, the white birth rate falls steadily through the nineteenth century, from about 55 per 1,000 in 1800 to about 30 by 1900, while the Black rate stays higher, near 57 in mid-century and falling to about 44 by 1900. Summed across these decades, this gives about 120 million births from 1790 to 1900, plus roughly two to three million more in the years between independence in 1776 and the first census in 1790.

From 1900 onward the guesswork ends. The federal birth registration system, which began in 1915 with a handful of states and covered the whole country by 1933, means births are increasingly counted rather than estimated. Using the official birth series from 1900 to the present, the country recorded about 424 million births.

Putting the two parts together, the reconstructed period before 1900 and the recorded period after, gives about 483 million births up to 2010, the endpoint jlredford used. His figure was 472 million. The two are close, and the 11 million difference is about what one would expect from two independent reconstructions of the same uncertain quantity. Carrying the count forward another sixteen years to 2026 adds about 63 million more, for a total of about 546 million births from 1776 to the present.

Immigration

According to Department of Homeland Security records, about 76 million people were admitted as legal immigrants between 1820, when federal record-keeping began, and 2010. This is a little higher than the 73 million jlredford used. Before 1820 the numbers were small. The Revolutionary War, and then the Napoleonic Wars and the War of 1812, suppressed transatlantic crossing, and the foreign-born population fell to a low of about 100,000 around 1815. Estimates put the total immigration from independence to 1820 at only a few hundred thousand people. From 2010 to 2026, net immigration added roughly 17 million more. Together this brings recorded immigration to about 93 million. The legal-admission series understates the true flow, since it does not capture immigrants who entered without authorization, though the net-migration figures used here for the years since 2010 count border crossings regardless of legal status, so recent unauthorized immigration is largely included. What the count mostly misses is unauthorized immigrants who arrived before 2010 and never gained legal status, a genuine undercount of a few million rather than the full unauthorized population.

Adding everything together, about 546 million births, about 93 million immigrants, and the 2.5 million alive at the founding, gives a total of about 642 million people who have lived in the United States since 1776.

How much this estimate can be trusted

These figures are imprecise, and most of the imprecision lies in the nineteenth century and earlier. Births were increasingly registered through the early twentieth century, with national coverage complete by 1933, and immigration was recorded far better than it had been in the previous century. The births and immigration of the 1800s, by contrast, are estimated, the births from birth-rate figures applied to the census population, the immigration from passenger lists that were far from complete, and both grow rougher the further back one looks.

The reassuring point is that the early figures are also small. The entire period from 1776 to 1900 contributes about 146 million people to the total, against about 496 million from the better-recorded period since. So even a substantial error in the nineteenth-century estimates moves the total only slightly. The uncounted are mostly people the records missed rather than people counted twice, so the true figure is more likely a little above 642 million than below.

There is also the question of how to count the land itself, which was not fixed in 1776. Because the reconstruction takes its population from the Census, a region enters the count only when it became part of the censused United States. People living in a territory before it joined, the Mexican residents of the Southwest before 1848, for instance, are largely absent, and the population already present at annexation is absorbed into the Census base rather than counted as a separate arrival. The largest gap of this kind is the Native American population: those living in tribal society were excluded from the Census until 1900 and so are mostly missing from the reconstruction for the nineteenth century. These omissions all push in the same direction, toward a true figure modestly higher than the one reconstructed here.

The share alive today

About 342 million people live in the United States today, out of roughly 642 million who ever have. That places the share alive right now at about 53 percent, a little over half.

The result seems surprising until one considers how the population grew. It did not increase steadily but expanded enormously, mostly within the last century, which means a large share of all the Americans who have ever lived were born within living memory. Longer lifespans reinforce this: people born even 70 years ago are largely still alive.

Children who did not survive

A final figure is worth noting here. Of all the children ever born in the United States, roughly one in eleven, some 49 million, died before reaching the age of five.

It is calculated in the same way as the rest. Using the births already reconstructed for each period, I multiply by the share of children who did not survive to age five in that period, and sum the results. The survival shares come from the long-run under-five mortality series assembled by Our World in Data, after Gapminder, supplemented by US vital statistics once national figures begin in 1915. Like the birth rates, they are documented only at intervals, so each period figure in the table is an average across the decades it spans rather than a single published number.

TABLE 2. Estimated Deaths Before Age Five

Around 1800, a third of all children did not reach the age of five. Over the twentieth century the rate fell below one percent, the result of vaccines, antibiotics, clean water, and pasteurized milk. Had all 546 million children ever born in the United States faced the survival odds of today, only about three million would have died before the age of five, rather than the roughly 49 million who did.

Subscribe now

Correction: an earlier version put the total at about 644 million. For recent births (2010-2026) I had used a rough back-of-envelope figure, a round four million a year, and forgot to replace it with the actual annual numbers, which are lower and falling. The corrected total is about 642 million.

Sources: The starting estimate is from jlredford, “How Many Americans Have There Been?” (A Niche in the Library of Babel, 2010). Population totals by race are from the US Census Bureau’s historical series. Nineteenth-century birth rates, given separately for the white and Black populations, are from Michael R. Haines, “Fertility and Mortality in the United States” (EH.Net Encyclopedia, 2008), drawing on his work with the Census and Coale and Zelnik; pre-1850 Black birth rates, which Haines does not tabulate, are approximated from the high fertility of the mid-century enslaved population. Twentieth- and twenty-first-century births are from US National Vital Statistics and Census Bureau data; figures before 1933 are official estimates adjusted for incomplete registration. Immigration totals are from the Department of Homeland Security’s series of lawful permanent residents admitted since 1820, which captures legal immigration only, with recent net migration from Census Bureau estimates, which counts flows regardless of legal status. Child mortality rates are drawn from Our World in Data’s long-run under-five mortality series (after Gapminder) and US vital statistics from 1915; the per-period shares are interpolations between documented anchor points. The recent figures are largely recorded, while the nineteenth-century and earlier figures are reconstructions, and all are best read as estimates.

Fermi Poker turns that mental exercise into a social game: instead of cards, each player starts by writing down a guess to an estimation question. Multiple betting rounds follow, with two hint reveals in between. These hints help players evaluate their guesses and decide whether to stay in the round, and if so, how much to put at risk.

The twist is that you usually can't succeed just by being good at making Fermi estimates. You also have to model what's happening inside everyone else's head: their emotional state, background knowledge, risk tolerance, and ability to update their confidence when new information becomes available. In other words, Fermi Poker trains one of the core components of empathy: perspective-taking.

A Training Ground for Cognitive Empathy

Psychologist Nicholas Epley calls our chronic difficulty in guessing how others think "the real other-minds problem" and argues that failures of perspective-taking underlie many misunderstandings. Fermi Poker draws special attention to this psychological challenge and serves as a training ground to practice modeling others' knowledge and reasoning dozens of times in the course of a few hours.

This concentrated practice is effective because it works similarly to a flight simulator where pilots can practice dozens of takeoffs, landings, and emergency scenarios in a few hours. This is far more than they'd experience in months or even years of actual flying. Fermi Poker works the same way, letting you rehearse the core empathic move of understanding the minds of others at an iteration speed that is hard to replicate in everyday life.

Take that question about the number of new car sales in the US. Before you bet, you may ask yourself:

• What do the others know about this topic?

  Do they work in a car dealership or have they never even bought a car? Have they previously answered similar questions well?

• How do they handle new information?

  Some people quickly adjust their thinking when hints arrive; others stick stubbornly to their first impression.

• What's their cultural background?

  An American from rural Texas will likely estimate a much higher number than a European from a bike-friendly city.

• What's their emotional state?

  Did they just lose narrowly and now want to play aggressively to get back what they lost?

• What do they know about me?

  If they know I'm experienced with these questions but still bet big, they must be very confident.

The very act of constructing these mental dossiers trains a specific type of empathy that goes beyond reading physical tells. You're practicing cognitive empathy—the capacity to understand not just what someone is feeling, but how they think and why they believe what they believe.

Why Fermi Poker Outshines Traditional Poker for Empathy Training

Traditional poker rewards reading tells: micro‑expressions, betting patterns, the occasional trembling hand. That’s valuable social perception, but the content of another player’s knowledge—their education, domain expertise, life experiences—rarely matters beyond guessing whether they understand pot odds.

Fermi Poker, by contrast, adds a rich layer of inference about why they believe what they believe. This is the kind of perspective-taking that transfers to real-world situations where you need to understand not just whether someone disagrees with you, but why they think differently.

Conclusion

Most empathy training focuses on helping people better recognize and respond to the emotions of others. But since emotional empathy is considerably more heritable than cognitive empathy (Abramson et al., 2020), this focus may be misdirected. Fermi Poker trains something different and more learnable: the ability to understand another person's reasoning process through conscious, deliberate practice.

Whether you're a teacher trying to understand why students approach problems differently, a negotiator working across cultural divides, or simply someone trying to have more productive conversations with people who see things differently, the ability to consciously model how others think might be one of the most valuable skills you can develop.