I’ve seen an argument about sugar and obesity going around the internet lately. The argument goes like this:
Sugar can’t be the main driver of obesity, because rates of obesity have continued to rise while sugar consumption has been going down.
That statement is based on data showing that our sugar consumption has dropped by around 15% over the last decade or two, depending on whose figures you use. Meanwhile, rates of obesity have continued to climb.
For the record, I don’t believe sugar is the only driver of obesity. I believe crappy oils have a lot to do with it, as Dr. Eades proposed in an excellent speech. I also believe refined grains share a chunk of the blame.
But that’s not the point of this post. The point is that sugar can’t be the driver of obesity because sugar consumption is going down while obesity is going up isn’t as logical as it first appears. In fact, it reminds me of one of the concepts from the book How Not to Be Wrong: The Power of Mathematical Thinking, which I’ve mentioned in a few posts.
In a chapter titled There Is No Such Thing As Public Opinion, author Jordan Ellenberg explains that if you confuse individuals with the aggregate, what’s perfectly logical can appear to be illogical. He gives an example something like this:
Suppose we’re all worried about the federal deficit. Now suppose one-third of the population believes the only acceptable cure is to massively raise taxes, another third believes the only acceptable cure is to massively cut defense spending, and the final third believes the only acceptable cure is to massively cut social programs like Medicare. We’ll name these groups Raise Taxes, Cut Defense and Cut Social Programs.
So what happens if we start conducting polls? Based on the results, we’ll conclude that the voters are illogical doofuses. Ask “Should we cut federal spending?” and two-thirds (Cut Defense + Cut Social Programs) will answer yes. Okay, got it. Two-thirds of voters want us to tame the deficit through spending cuts.
So then we ask, “Should we cut defense spending?” Two-thirds of the voters (Raise Taxes + Cut Social Programs) will answer no. Next question: “Should we cut social programs?” Again, two-thirds of the voters (Raise Taxes + Cut Defense) will answer no.
We shake our heads at the results and think, Geez, the voters are clueless. Two-thirds of them want to cut spending, but they don’t want to cut defense or social programs. Are they just morons?
Uh, no. We’ve fallen into the mental trap of confusing individuals with the aggregate. As Ellenberg writes:
Each voter has a perfectly rational, coherent political stance. But in the aggregate, the position is nonsensical.
We have a similar problem with if sugar consumption is going down while obesity is going up, sugar can’t be the main driver of obesity. Yes, it seems like a logical conclusion at first glance. But if we dig a little deeper, we see it’s another case of confusing individuals with the aggregate.
To actually prove something conclusive about the cause of obesity, the 15% decline in sugar consumption would have to occur more or less evenly across the population. In other words, sugar consumption is down by 15% because damned near everyone is now eating 15% less sugar. If that were true, it would mean individuals are cutting back on sugar but getting fatter. Case closed.
But I’ll bet you dollars to donuts (and you can keep the donuts) that’s not what happened with sugar consumption. Do you know anyone who decided sugar is bad and responded by eating 15% less of the stuff? I sure don’t.
“Honey, I just read this great article by Dr. Robert Lustig on how sugar is a metabolic poison. Pour off a few sips of my Coca-Cola and cut away 15% of my donut before you bring it to me, will you?”
The people I know who concluded that sugar is a health hazard responded by giving up sugar completely. Well, okay, maybe a few cookies and a slice of pumpkin pie at Christmas, but other than that, they don’t touch the stuff.
In the aggregate, sugar consumption has declined by 15%. But among individuals, many people are continuing to eat their donuts and Frosted Flakes and Little Debbie Snack Cakes washed down with Coca-Cola, while a minority of others have gone almost sugar-free. That’s what I see among co-workers at the office, among parents and children at school functions, etc.
With that in mind, let’s demonstrate how a population of individuals could experience a sugar-fueled rise in obesity even as sugar consumption declines across the same population. To keep things simple, I’ve created population of eight men who are all 5’10”, which means the BMI scale will label them as obese when they reach 210 pounds. Here’s our population in 2008, when all eight men are consuming around 100 pounds of sugar per year. (I’m horrified to say that’s about average.)
Goodness. We have an obesity rate of 50%. Now let’s jump ahead 10 years. John, Paul, George and Ringo have continued consuming around 100 pounds of sugar per year and have all gotten 10 pounds heavier. That’s not at all unusual. As I pointed out in Fat Head, the U.S. population gets a little older on average with each passing year, and people tend to gain weight over time if they don’t clean up their diets.
But during the same 10 years, Mick, Keith, Ron and Charlie heard about the health hazards of sugar and stopped eating the stuff. They all lost 10 pounds – nothing dramatic, but a nice change. Here’s what our population looks like in 2018:
Across our entire population, average sugar consumption went from 100 pounds per year to just 50 pounds per year – because Mick, Keith, Ron and Charlie stopped eating sugar. Wow! That’s a 50% decline in sugar consumption – way more than a paltry 15% drop.
So if sugar is the driver of obesity, the obesity rate also had to decline, right? Wrong. Check the graphic. Our population-wide obesity rate rose from 50% to 63% … even though the average weight of our population (209.5 lbs.) is exactly the same in 2008 and 2018.
How is that possible? It seems completely illogical. Yes, but only if you’re looking at the population in the aggregate instead of what happened to the individuals.
John and Paul gained 10 pounds and are now considered obese. Ron dropped 10 pounds and went from obese to not obese. Charlie is like a lot of people I know: his health improved and he lost weight after giving up sugar, but he’s still heavy enough to be considered obese.
At the individual level, the sugar-eaters gained weight and the sugar-avoiders lost weight – just 10 pounds in either direction over 10 years. Again, nothing dramatic. But those relatively small changes caused two people to go from not obese to obese, and one person to go from obese to not obese. The result is that despite a 50% drop in sugar consumption, our population became more obese in the aggregate.
Again, I’m not saying this proves sugar drives obesity. Maybe it does, maybe it doesn’t. But sugar can’t drive obesity because obesity rates went up while sugar consumption went down isn’t the slam-dunk logical argument the people making it think it is. It could simply be that lots of people are still eating sugar and getting fatter as a result, while a much smaller group of people have dumped sugar completely and account for most of the decline in sugar consumption. Keep in mind that many people who dumped sugar from their diets (Chareva being a perfect example) weren’t fat to begin with, which means they don’t affect the obesity figures either way.
We’re individuals, not aggregates. As Dr. Ellenberg knows, looking at aggregates can fool us into making the wrong conclusions.
If you enjoy my posts, please consider a small donation to the Fat Head Kids GoFundMe campaign.
I haven’t read the study, but does it account for people who may be eating less SUGAR but perhaps more carbohydrates instead?
I see people on the internet all the time who claim they are cutting “sugar” out of their diet, but replace that sugar with more fruit and other carbohydrates in general (bread, pasta, etc.). They even increase their consumption of honey and maple syrup because it’s not called “sugar”.
It’s not even a study, really. They didn’t gather a group of people and check their sugar consumption, then check their weight. They simply compared sugar consumption for the whole population to obesity rates for the whole population.
Thanks for the lucid and entertaining maths lesson!
Maybe Plato had a point when he had a sign put up over the entrance to the Academy: “Let No One Ignorant of Mathematics Enter Here”.
My college physics professor told us, “Learn math, even if you’re not going into a field that requires math. Math is how you know when they’re lying to you.”
Importantly math is how you know when you’re lying to yourself.
That too.
I graduated in 1982 and I still don’t care what “y” equals. 😉
I’m a programmer, so I know a variable means whatever I define it to mean.
And stop asking me to find your x. She’s not coming back.
Did she say y … ?
Great post. This is a topic that is often taught extensively in Physics and Engineering courses because single point anomalies make for extremely difficult math/calculus to solve. We often assume homogeneity in many cases as a starting point, but then progress to more detail involving more variation. Sounds like those that publish news/media bits on studies like this often ignore the second part about “more detail” [surprise surprise].
Let’s just say I don’t think media reporters are the most logical or mathematical people out there.
Good illustration. It reminds me of how household income can go down while individual income can go up–or, as Thomas Sowell put it, “I only have to hear ‘household income’ to know you’re not serious.”
Yup. One of the many reasons I love Sowell’s books.
re: For the record, I don’t believe sugar is the only driver of obesity.
It’s not, but the focus on frank simple sugars makes it too easy for BigFood to dance around it. People need to understand that grains might as well be sugar. The amylopectinA of wheat, for example, is 60% glucose, in a branched form that is trivially cleaved to simple glucose by human amylase (found in saliva), a process that starts before you even swallow it (full employment for dentists).
Table sugar (sucrose) is “only” 50% glucose, and HFCS even less (only be 45%). People who don’t even have sugar in the house may well be eating whole wheat bread, and thinking that they are practicing saccharide piety. Their HbA1c begs to differ.
And even calling out all carbs that end up as BG leaves an opening. Many processed food-like substances today use fructose for sweetening (it’s what makes sucrose taste sweet). It has zero near-term effect on your BG meter. It would show up later in TG, ALT, AST and uric acid, but nobody is routinely minding fructose markers, even when it mysteriously balloons the waistline.
So on the wider carb contribution to obesity, a key simple tool is “net carbs” (an Atkins idea, I’m told). On the NF panel, that’s Total carbs minus Fiber carbs. 50 grams/day is a useful target.
Then we can move on to other obesity suspects, which does indeed include industrial grain and legume oils high in Omega 6 linoleic acid, as well as pandemic dysbiosis and way too common hypothyroidism.
If Bob Lustig gets his wish for a sugar tax, political philosophy concerns aside, it’s apt to do exactly nothing for net carbs in junk food or for waistlines.
Bob, by the way, was in a great podcast with Peter Attia lately, and they laid out multiple pathways for obesity. We know how to trivially prevent obesity and reverse it (in the majority of cases), but don’t get too wedded to precise etiologies and ingredients, or BigFood will game you on it, resulting in a relentless and tiresome need for otherwise worthy blog posts like today’s.☺
That’s why the best dietary plan is to avoid processed food.
And fermented grains aka beer. Beer is not really a highly processed food, but they have to add hops or some other agent to mask the sweetness or it would apparently be too sweet for even American tastes.
In Boy Scout Camp, my nickname was “beer belly”. Fatty liver from excess carbs and I presume especially sucrose.
Well, that’s not true. When fully fermented, beer contains only small amounts of residual sugar because it is consumed by yeasts to make alcohol. And if some sugar is left, it would be glucose and maltose from starch, there is almost no sucrose in barley. Hops is not used to mask sweetness but is traditionnaly added because it contains natural preservative agents. Of course some flavored beers can contain added sugar (therefore sucrose), but standard beers contain much less sugar than the average soda. This beying said, I wouldn’t recommend beer consumption in large amounts as it is a caloric drink (7 kcal/g of alcohol), not to mention the effects of alcohol itself.
I eat meat, eggs, green veggies, cauliflower, and a little high-fat dairy. The only processed food I buy is Italian hard salami. I haven’t felt this good since I was 18. My reasons are exactly those you described above. I think that 90% of the stuff in the supermarket is just toxic.
Right now, what I notice is that while more and more people are turning away from the classic low-fat diet (or Plant-based diet, as they’re calling it now), they aren’t moving towards low-carb. They’re giving up dieting entirely. Usually people who have gone through this many times, always gain the weight back if not more, and don’t see the point in trying anymore.
The third group is saying that long-term weight loss is impossible and we’ve all been sold a bill of goods. That’s where more people are going, not to a low-carb approach.
In all honesty, I don’t blame them. I started eating carbs again and ballooned back, which I’m trying to correct once again. It’s frustrating. No one’s more diet-obsessed than Americans, yet we’re one of the fattest nations in the world.
Unfortunately, you’re probably correct. But I am coming across more and more people who are adopting a keto diet. A minority, to be sure, but the numbers seem to be growing.
There is nothing to keep a “Plant Based” or Vegan diet from being high fat. Pea Protein and lots of Coconut Oil, Palm Oil and Avacado oil with appropriate vegetables Would work just fine, AFAICT. (I haven’t tried it.) Dr. Fung’s group has even gotten a vegan into ketosis. Should include natto for K2, and in general you have all the problems associated with vegan diets.
As you convincingly point out, a small reduction in sugar consumption from an extremely high level doesn’t exonerate sugar. My consumption probably dropped by about 90% a few years ago and I agree that the drop in consumption comes mostly from people who’ve made a similar decision.
However, I normally use the opposite logic when vegetarians make the silly argument that eating meat causes diabetes (and every other misfortune to strike humanity). I think red meat consumption has fallen by 64% in the last thirty years, yet type 2 diabetes quadrupled in roughly the same period. When I ask them to explain that, they change the subject oo otherwise avoid the point. In short, I normally argue that whilst association doesn’t mean causation, the absence of an association is strong evidence.
It depends. Lack of an association within an observational study is strong evidence, because we’re talking about the same people — that is, if I tabulate meat consumption vs. diabetes and find a negative association, it means the individuals who who eat more meat have less diabetes. I can’t possibly say meat causes diabetes.
When we’re generalizing to populations, we have to be more cautious because we don’t know if we’re talking about the same individuals. One group of people could account for more of this food or less of that food, while another group of people could account for a rise or fall in disease.
But if red meat consumption has dropped by 64%, it’s close to impossible that a small group of people account for the drop.
Thank you for the important point about an association, or a lack of association, within an observational study.
I think the 64% reduction in red meat consumption in the last thirty years came from a talk by Nina Teicholz, but I’m afraid I can’t remember which one. We do know that many people ate less red meat and more chicken because of the dietary guidelines and fat scare.
A massive overdose of sugar vs. a massive overdose minus one isn’t going to make much difference.
Seen this?
http://sci-hub.tw/http://dx.doi.org/10.1016/j.pcad.2018.04.007
straight out of the Coca-Cola playbook
the antidote is here
https://www.researchgate.net/publication/327277017_In_Critique_of_In_Defense_of_Sugar_The_Nuance_of_Whole_Foods
Archer makes the ridiculous logic jump that if some is good more must be better. He shows no clue about J curves or U curves or the fact that we carefully maintain about 5g glucose in the entire volume of our blood at all times. When that gets to 11g diabetes is diagnosed. Yet he cheerfully recommends that we eat up to 600g “sugar” a day and exercise it off. Ummm . . .
Hasn’t this same logic been made to also point out how nationally we are eating far less fat than we used to, while obesity continues to rise?
We don’t eat less fat. We eat different fats.
Everyone is looking for the one magic bullet food to remove from their diet, and food companies are all looking to prove the food they make isn’t the one causing all the problems.
The common sense answer is that too many empty calories will make most people fat, because the lack of nutrition and the screwy blood sugar keeps them from feeling full so they eat too much. There are four main sources of empty calories in the modern U.S. diet — sugars, flour, seed oils and alcohol. You can totally eliminate any of those four from your diet and still get fat and/or unhealthy consuming too much of the others. (The and/or unhealthy part is important because I know people with plenty of health problems who are able to stay skinny eating a “pub diet” based around consuming those four things in bars all the time)