America sets a new record in not being fat

How many non-overweight Americans are there now, and how many were there a century ago?

Well - this US census site estimates the current US population at about 333 million people. And this page from Our World in Data puts the current fraction of overweight US adults at 68% (scroll down to “What share of adults are overweight?”). In reference to this paper earlier in that article, Our World in Data mentions “adults aged 18 years and older”, so I’m going to use being at least 18 as the definition of an adult, and assume they do consistently as well1. Consulting this table from the US census, we see that 22.2% of Americans are under 18 as of 2021 - so that gives us an estimate of 259 million adults total in US. Multiplying by 32%, we get an estimate of about 83 million for the total adult non-overweight population.

Now consider the US in 1920. Here we can find volume 2 of the 1920 US census. Opening the pdf for chapter 3, on the twelfth page2 there’s a breakdown of the 1920 population (total: 105 million) by age. They group ages in buckets of size five years, with all buckets not less than 20 years summing to 62.7 million. If we include 40% of the 15-19 bucket as an estimate for the population of 18 and 19 year olds, we arrive at 66.4 million adults in the US, total, in 1920.

There were, in particular, at most this many non-overweight adults in the US in 1920. Which means we have increased the number of non-overweight adults by at least (83 - 66) = 17 million in the last century.

Does this consitute progress in American health? Admittedly, it’s slow, +170,000 per year, but the trend is positive. Should we celebrate this victory?

Probably not, because the number of overweight adults has increased by more. And based on the obesity trend lines from 1975 that Our World in Data presents, it’s likely that the fraction of overweight adults has also increased in the same period (I haven’t attempted to prove this). By those lights the last century would be an out-and-out regression.

But why should the fraction of non-overweight people, rather than the count, be the metric we judge progress by? Why does the addition of 176 million overweight adults spoil things?

One answer is that the fractional measure allows us quantify the properties of a typical American, and typicality is useful because lets you predict what your environment will look like most of the time. If the typical American is fat, you should expect that most of the people you see within a month will be fat, and that most people who enter your nearest hospital will be fat, unless you live in an atypically lean city. So if you’re a doctor and are concerned with how many of your patients are going to fall into the overweight camp, this is a good metric3.

I don’t think that’s why most of us are more bothering about the fractional metric, though. It’s because we believe it’s possible to decouple population growth from weight distribution - to add people of each weight category at roughly the same rate - and so are dissatisfied with the skew. You can imagine how this belief might be incorrect - perhaps, for instance, the population is growing faster than it used to because we’ve created a new ‘overweight’ ecological niche, which for some reason is less constrained than the old one.

The count metric does seem to be better suited for questions regarding existence. If a certain thing can only be created (or done) by a non-overweight person, the rate of its creation (or undertaking), all else being equal, will be proportional to the non-overweight count, and so the probability of its existence (or accomplishment) will be as well4. Perhaps this is part of why Americans do well in the Olympics despite being typically fat. I can’t think of many examples besides athletics that fit this bill5, which makes me think we are by and large correct to use the fractional measure.

Regardless of which metric you want to use for a particular problem, just being aware of the counts is valuable because it lets you describe the same situation in a new way. Compare for instance these two possible descriptions of American population shifts:

The population of America has increased by a factor of three over the past century, but the proportion of Americans who are overweight has increased fivefold.


Over the last century, we’ve added 17 million non-overweight Americans, and 176 million overweight ones.

The second description is inherently a bit more optimistic. It portrays the problem of obesity as “things not getting better fast enough” rather than “things getting worse”. Speeding up improvements might be easier, and more fun to work on, than reversing a regression.