A Chemical Hunger – Interlude F: Demographics



The stereotype is that poor people are more obese than rich people, but rich countries are definitely more obese on average than poor countries:

This same trend of wealth being related to obesity is also mirrored within many countries. In poor countries, upper-class people are generally more likely to be obese than lower-class people. For example, in India rich people are fatter than poor people.

We see that the general pattern between countries is that wealth is associated with obesity, and we see the pattern within most poor countries is also that wealth is associated with obesity. Given this, it would be kind of surprising if the relationship ran the other way around in wealthy countries. 

Still, common-sense beliefs say that — in America at least — poor people are more obese than rich people, maybe a lot more obese. But evidence for this idea is pretty elusive. 

The National Health and Nutrition Examination Survey (NHANES) is an ongoing project by the CDC where every year they take a nationally representative sample of about 5,000 Americans and collect a bunch of information about their health and lifestyle and so on. In 2010 a NCHS team led by Cynthia Ogden examined the NHANES data from 2005-2008. They wanted to find out if there was any relationship between socioeconomic status and obesity, the exact same question we have in this post.

The results of their analysis were mixed, but there certainly wasn’t a strong relationship between socioeconomic status and obesity. Their key findings were: 

Among men, obesity prevalence is generally similar at all income levels, however, among non-Hispanic black and Mexican-American men those with higher income are more likely to be obese than those with low income.

Higher income women are less likely to be obese than low income women, but most obese women are not low income.

There is no significant trend between obesity and education among men. Among women, however, there is a trend, those with college degrees are less likely to be obese compared with less educated women.

Between 1988–1994 and 2007–2008 the prevalence of obesity increased in adults at all income and education levels.

Cynthia Ogden got to do it again in 2017, this time looking at the NHANES data from 2011-2014, trying to figure out the same thing. Again the picture was complicated — in some groups there is a relationship between socioeconomic status and obesity, but it sure ain’t universal. This time her team concluded:

Obesity prevalence patterns by income vary between women and men and by race/Hispanic origin. The prevalence of obesity decreased with increasing income in women (from 45.2% to 29.7%), but there was no difference in obesity prevalence between the lowest (31.5%) and highest (32.6%) income groups among men. Moreover, obesity prevalence was lower among college graduates than among persons with less education for non-Hispanic white women and men, non-Hispanic black women, and Hispanic women, but not for non-Hispanic Asian women and men or non-Hispanic black or Hispanic men. The association between obesity and income or educational level is complex and differs by sex, and race/non-Hispanic origin.

If you don’t trust us but do trust the Washington Post, here’s their 2018 article on Ogden’s work.

The studies that do find a relationship between income and obesity tend to qualify it pretty heavily. For example, this paper from 2018 finds a relationship between obesity and income in data from 2015, but not in data from 1990. This suggests that any income-obesity connection, if it exists, is pretty new, and this matches the NHANES analysis above, which found some evidence for a connection 2011-2014 but almost no evidence 2005-2008. Here’s a pull quote and relevant figure:

Whereas by 2015 these inverse correlations were strong, these correlations were non-existent as recently as 1990. The inverse correlations have evolved steadily over recent decades, and we present equations for their time evolution since 1990.

Another qualifier can be found in this meta-analysis from 2018. This paper argues that while there seems to be a relationship between income and obesity, it’s not that being poor makes you obese, it’s that being obese makes you poor. “Obesity is considered a cause for lower income,” they say, “when obese people drift into lower-income jobs due to labour–market discrimination and public stigmatisation.” 

Anyone who is familiar with how we treat obese people should find this theory plausible. But we don’t even have to bring discrimination into it — being obese can lead to fatigue and health complications, both of which might hurt your ability to find or keep a good job. 

This may explain why Cynthia Ogden found a relationship between income and obesity for women but not for men. It’s not that rich women tend to stay thin; it’s that thin women tend to become rich. A thin woman will get better job offers, is more likely to find a wealthy partner, is more likely to find a partner quickly, etc. Meanwhile, there’s a double standard for how men are expected to look, and so being overweight or even obese hurts a man’s financial success much less. This kind of discrimination could easily lead to the differences we see.

But the biggest qualifier is the relationship between race and income. If you’re at all familiar with race in America, you’ll know that white people make more money, have more opportunities, etc. than black people do. Black Americans also have slightly higher rates of obesity. The NHANES data we mentioned earlier contain race data and are publicly available, so we decided to take a look. In particular, we now have complete data up to 2017-2018, so we decided to update the analysis.

Sure enough, when we look at the correlation between BMI and household income, we see a small negative relationship, where people with more income weigh less. But we have to emphasize, this relationship is MEGA WEAK, only r = -.037. Another way to put this is that household income explains only one-tenth of a percent of the variance in BMI! Because the sample size is so huge, this is statistically significant — but not by much, p = .011. And as soon as we control for race, the effect of income disappears entirely.

We see the same thing with the relationship between BMI and family income. A super weak relationship of only r = -.031, explaining only 0.07% of the variance in BMI, p = .032. As soon as we control for race, the effect of income disappears.

We see the same thing with the relationship between BMI and education. Weak-ass correlation, r = -032, p = .022, totally vanishes as soon as we control for race. 

Any income effect needs to take into account the fact that African-Americans have higher BMIs and make less than whites do, and the fact that Asian-Americans have lower BMIs and make slightly more than whites do.

We don’t see much of a connection between income and obesity. If there is a link, it’s super weak and/or super idiosyncratic. Even if the connection exists, it could easily be that being obese makes you poorer, not that being poor makes you obese. 


Race actually doesn’t explain all that much about BMI either. A simple model shows that in the 2017-2018 data at least, race/ethnicity explains only 4.5% of the variance in BMI. The biggest effect isn’t that African-Americans are heavier than average, it’s that Asian-Americans are MUCH leaner than everyone else. In this sample, 42% of whites are obese (BMI > 30), 49% of African-Americans are obese, but only 16% of Asian-Americans are obese! 

On the topic of race, some readers have tried to argue that race can explain the altitude and/or watershed effects we see in the Continental United States. But we don’t think that’s the case, so let’s take a closer look. Here’s the updated map based on data from 2019:

US Adults

This map is for all adults, and things have not changed much in 2019. Colorado is still the leanest state; the states along the Mississippi river are still among the most obese. Now, it’s true that a lot of African-Americans do live in the south. But race can’t explain this because the effect is pretty similar for all races. 

For non-hispanic white Americans, Colorado is still one of the leanest states (second-leanest after Hawaii) and states like Mississippi are still the most obese:

Non-Hispanic White Adults

For non-hispanic black Americans, Colorado is still one of the leanest states, and while you can’t see it on this map because the CDC goofed with the ranges, states like Mississippi and Alabama are still the most obese: 

Non-Hispanic Black Adults

In fact, here’s a hasty photoshop with extended percentile categories: 

Non-Hispanic Black Adults

If the overall altitude pattern were the result of race, we wouldn’t see the same pattern for both white and black and Americans — but we do, so it isn’t.

[Next Time: Li+]

18 thoughts on “A Chemical Hunger – Interlude F: Demographics

  1. Mark says:

    Hi, I’m enjoying this series a lot, thank you.

    The state level data is suggestive but is there more fine grained data available, like counties or statistical areas? Would be interesting to see how, say northern vs southern Ohio compare, or within Texas.

    Also, are there pockets of relatively low obesity in Arkansas or Mississippi, and what is different about them.


  2. Yoav says:

    I stumbled upon another potential pollutant candidate – Photons! 🙂

    It seems there’s some evidence that light pollution is linked to obesity (Though I haven’t looked much into it, so I don’t have an impression how legit that evidence is).

    Here’s the top google scholar article about it:
    Or unlocked:


  3. Charlie says:

    I’m also on a weight loss journey and I’ve personally discovered that alcohol was like a brick wall in preventing my weight loss. As soon as I quit drinking and added exercise, fasting, and less carbs I dropped 10-15 lbs. in about 2 months. I was a heavy drinker usually 6 pack of beer per night/more on weekends plus a few whiskey shots. Americans love booze but I guess the idea it’s partly causing the obesity spike must be wrong since the Islamic countries are super fat too.


  4. C Pelfron says:

    This is a great series. Very interesting so far. I wonder if you’ve considered UV radiation, more specifically UVB, as a factor. Theory: people are not getting enough UVB and it’s having weird effects and messing with their lipostat

    1. UVB increases at altitude. WHO says “10% per 1000m”
    2. Lack of UVB has been implicated in metabolic syndrome:
    3. There is new research out there implicating lack of UVB in autoimmune disorders. It’s definitely a factor in more internal “stuff” than just the skin.
    4. Sunscreen widespread adoption might track with your “what changed in 1980?” question. (And why does it keep getting worse… more and better sunscreen has been a drumbeat over the last 20 years.)
    5. The cops and firefighters occupational high rates mystery: these are professions with significant night shift exposure. Some otherwise healthy people get starved of UVB. In fact, night shifts in general seem like something that has ramped up in the last century. Lots of building cleaners (high on your list) work at night, few construction workers (low on your list) do.
    6. You mention “why is Maine obese?” at one point, because it doesn’t necessarily fit pure altitude hypothesis.. but Maine is the second-cloudiest state, to Washington:

    Just occurred to me that the dude in the picture you posted of the unspoiled tribe getting studied was not wearing a shirt!


    1. C Pelfron says:

      Couple more things…. looking into occupational rates for obvious low-uvb job like miners… this NIH paper says “The mining industry has the highest proportion (76%) of overweight and obese employees in Australia [9]” (different baseline from your occupational paper.)



      “doxycycline, a medication that we used for malaria prevention, makes you photosensitive, meaning you’re more apt to get a sunburn. ”


      “The results suggested that long-term administration of sub-antimicrobial doxycycline could be a novel therapeutic strategy for T2DM treatment.”

      a malaria drug that makes you more photoreceptive seems to regulate type 2 diabetes.


      1. Thanks for all this. We also noticed that many of the most obese professions tend to work at night — we thought a bit about whether this could mean that sleep, sunlight, artificial light, or something similar could be involved.

        Currently our opinion is that while it’s possible there’s a connection, we don’t currently find UV to be a very likely hypothesis. Briefly, some reasons are:

        1) UVB varies a lot by latitude, but we don’t see a large latitude effect. Countries like Australia and Kuwait which get a lot of sunlight are pretty obese, you would expect Canada to be more obese than the US, etc. Sunscreen could mess with this somewhat but it seems unlikely it could reverse the trends. Australia and New Zealand are #1 and #2 for skin cancer, but both are pretty obese. Similarly, people who live within the Arctic Circle, researchers who spend months in Antarctica, polar explorers, etc. do not seem to have an extreme tendency to obesity.

        2) The leanest professions include professions like “art and design workers” and “post-secondary teachers”, which don’t seem likely to get an unusually high amount of UVB. Professions that spend a lot of time outdoors seem to be about average in terms of obesity.

        3) There have been miners and others historically who didn’t get a lot of sun exposure, and as far as we know, these people do not seem to have become obese. If UVB were playing a major role, you’d expect there to be some evidence from before 1900. As a more general point, if this were true you would be able to cure obesity by making people spend time outside in the sun and cause obesity by making people stay indoors. We expect someone would have noticed by now if this were the case!


  5. I know you did vegetable oils already, but I ran across something that I thought might be worth sharing with you.

    Right now many of the saturated fat people are looking at upregulation of the enzyme SDC1 as being the driving factor for the obesity epidemic. I was reading a white paper thing about it (linked below) explaining the supposed mechanism of action. In the section regarding what increases SDC1 the following was written: “Exposure to low levels of BPA, a component of many types of plastic, upregulates SCD1 and increases fat in animal studies.”

    I just thought it was interesting and wondered whether it tied into your research at all. Even if it’s not BPA, it’s possible whatever the chemical contaminant is might be causing trouble by increasing SDC1 production. Just thought I’d share in case it’s useful.



  6. RB says:

    Fantastic series. Couldn’t help thinking about my friend who has hypothyroidism (which can be caused by lithium), along with all the women in her family, all of which live in one of the geographies you mentioned with high lithium concentration in the ground water. Is there any lithium detox method you’re aware of? It could be a simple way to do at least an n=1 experiment to see if there’s any improvement/reversal in symptoms?


    1. Thanks! This is a great question that we don’t have a good answer to yet. Our guess is that most people don’t get most of their dose of whatever it is from their drinking water, but if you’re confident that your friend is getting a high dose of lithium from their groundwater, they could try getting a water filter (we’re not sure yet what kinds of filers get lithium out, but we’re working on this) or they could try buying bottled or distilled water from somewhere else (though it’s hard to be sure how much lithium would be in these sources).

      There’s not really a detox we’re aware of — the biochemistry of lithium is poorly understood. It probably interacts with other ions in the brain (vague right?) but we’re not yet confident which ions or how. Maybe with more research we could find out how it’s cleared from the brain, but that research doesn’t exist yet (as far as we know). We’ve seen one paper that suggests that caffeine consumption reduces serum lithium (https://onlinelibrary.wiley.com/doi/full/10.1111/bdi.12990#bdi12990-bib-0028), and it’s a meta-analysis, but it’s still just one paper, and the serum results are from just a couple case studies. They could try drinking more coffee but we’re not confident that will do much.

      Realistically the thing to do would be to spend a couple months in a location with less lithium (Colorado? Thailand?) and see if there’s a reduction in symptoms, but that’s not a very targeted approach, and also presumably expensive.

      Long story short these are all questions that we are also very interested in, but as far as we can tell, no one really knows because no one has ever done any of this research. So we’re hoping to find ways to conduct some of this research ourselves in the near future! Stay tuned.


  7. ZG says:

    This, from the first study you cite, seems to pretty clearly show that people at higher education levels are less obese in the US: https://www.cdc.gov/mmwr/volumes/66/wr/figures/mm6650a1-F2.gif

    See also figure 13 from here (“Children are significantly more likely to be overweight or obese (50.9%) when the adult respondent (most of whom were their parents) has a high school
    diploma or lower educational attainment, compared with the adult having a
    college or graduate degree (25.2%)”): https://grc.osu.edu/sites/default/files/inline-files/Obesity%20in%20Children%20and%20Families%20Across%20Ohio%20Report.pdf

    Here’s a paper from 1998 that shows people with at least a bachelors degree have about half the rate of obesity as lesser educated people: https://www.sjsu.edu/people/peter.a.lee/courses/c8/s1/18939571.pdf


  8. Mahmud says:

    Coming back here after a while.

    I KEEP going back in my mind to high school (2008-2012) where I saw so many hormonal kids with horrible greasy acne skin. This was in Palos Verdes Peninsula High School so very high income area though I and other migrants-for-the-school weren’t. That school had so many kids who were the result of assortative mating of various backgrounds but many with the combined traits of being attractive, intelligent, physically adept and having wealthy backgrounds. I remember even then seeing people with terrible acne and I associated it with the foods and the food packaging. It just *felt* like there was a connection between a guy eating out of that prepackaged ramen bowl and his horrifying acne problem.

    Since this was an upper class place, people put a lot more effort into their appearance. I guess as part of status competition everyone was just a lot more committed to getting good physiques. However, they seemingly couldn’t shake the acne unless they were blessed with quality skin.


  9. Tennessean says:

    I just read through this whole series of blog posts – super fascinating! I keep getting stuck on one thing though: when I look at the county-level obesity data for my state, Tennessee, socioeconomic differences stick out like a sore thumb. Tennessee is mostly dark red with patches of orange but there is one single county that’s an island of yellow – it’s Williamson county, the richest county in the state. Tennessee is a pretty poor state overall; it ranks 42nd by median household income and 41st by per capita income. The wealth in Tennessee is highly concentrated in the Nashville metro area and Williamson county in particular. However, the entire region gets its water from the Cumberland River and with the way the river and its offshoots snake through that part of the state it’s hard to imagine that there’s any groundwater that’s not influenced by whatever contaminants end up in there. This is also the same drinking water source used nearby in Clarksville, TN, which makes your list of fattest cities in America. There are similar islands of yellow in a couple other very fat/poor states like Arkansas and Oklahoma and I wonder if they also match up to the wealthiest counties in those states.

    In my opinion the fact that these wealthy counties in poor states have lower obesity rates is actually a sign that demographics matter more than you’d think. The leanest states are also some of the wealthiest and the state-level wealth difference between Massachusetts and Tennessee or Colorado and Arkansas is enormous. It’s hard for me to believe that obesity makes you more likely to be poor (rather than the other way around) when you look at it at the state level like this – Massachusetts had nearly twice the per capita income of Tennessee in the 1960 census. State level wealth differences predate the obesity epidemic yet they track pretty closely with current obesity rates, with wealthier states being substantially thinner.

    What are your thoughts?

    County level obesity map: https://maxmasnick.com/media/2011-11-15-obesity_map/obesity_by_county_large.png

    Liked by 1 person

    1. Yeah, good question. Some thoughts:

      The most boring answer is that these are just attributable to differences in another demographic variable, race. When we look at Williamson County, we see it is more White, less Black, and much more Asian than the average for Tennessee. That by itself may be enough to explain the discrepancy.

      We also know there are reliable differences by profession. If there are states where most wealthy people happen to work in low-obesity jobs and most poor people happen to work in high-obesity jobs, you might end up with a spurious relationship in some cases.

      Another possibility is that we’re seeing a version of Simpson’s paradox, where (for example) there is a wealth-obesity relationship within individual regions but an opposite relationship between regions. We’re not aware of anything pointing to this in particular but it seems plausible if someone wants to look into it.


  10. KG Raje says:

    I think the statistics of origin aren’t sufficiently precise in that they fall into the usual gulf that fails to uncouple the ‘socio’ from the ‘economic’ in socioeconomic discussions in America.
    Similarly, upper income is usually defined as a household of 100k or above, which though quite high on the spread, isn’t actually what we all think of as upper income. A nurse married to a cop could hit upper income pretty easily (and both be overweight), but a ‘higher status’ job (an assistant professor, an environmental scientist, a nonprofit development director) could earn a similar income and have completely different lifestyle and outcomes.
    And I bet actual high income households are less likely to be overweight or obese. Rich or privileged people are very seldom overweight: Marjorie Gubelman stands out for a reason. Additionally, I think that perhaps that the difference between all college graduates, and graduates of selective or elite colleges would be found similarly disparate.
    Thank you for a fascinating series, and the hard work of collating such wide ranging research and results.


Leave a Reply to slimemoldtimemold Cancel reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s