I know what’s been causing obesity rates to rise. So do you. So does just about everyone. Unfortunately, most of us are probably wrong.
As I tell my students, correlation doesn’t prove causation, but that’s not why we use it. We look for a correlation because if it goes the opposite way from what we expect, that tells us our hypothesis is wrong or incomplete.
So if you have a pet theory about obesity—and, really, everybody has a pet theory about obesity—a correlation can tell you if you’re totally on the wrong track. Say you believe it’s because people don’t exercise as much anymore. But if you checked and found that people are in fact exercising more, that would disprove your theory, right? Or at least clue you in that the story is more complicated than you thought it was at first.
What the correlations show
In a new report, researchers at the University of Illinois have analyzed obesity trends in a way only an economist could love: poring over a multitude of correlations. Obesity rates by age group, by race, gender, education level, and income. Fruit and vegetable prices, leisure time, food expenditures as a percentage of income, daily calories per capita. Consumption of carbs, protein, and fat. Measures of car culture and food deserts.
And they found that the data contradicts a lot of popular pet theories. We exercise more than we used to; we eat more vegetables, too. We have more free time and cheaper food. Some groups of people are thinner than others—but they are gaining at about the same rate as everyone else.
Studying a cross-section of the population would show you that black women have higher BMIs than the other groups shown here, but to focus on that cross-section is to ignore the larger trend: all of the groups are steadily gaining weight.
Similarly, the authors call out a book on the “Colorado diet,” based on the idea that Colorado has the lowest obesity rate in the US at 20%. But put its trend on a graph alongside Mississippi’s (34.6%, second-highest as of 2012) and the lines are parallel.
It’s not a paradox; epidemiologists have long discussed the idea that if everybody in a population is exposed to the cause of a disease, studying cross-sections will only show you who is most susceptible. The true cause may be so common it’s invisible.
Or, as the Illinois authors put it:
To understand the obesity epidemic, rather than asking a question such as “Why are people in Colorado thinner than people in Mississippi?” we need to ask why are people in Colorado gaining weight at the same rates as people in Mississippi?
This isn’t a biochemistry paper, so it doesn’t provide any insights on what is happening in people’s bodies. But here’s the rundown on what the data seems to support. (Disclaimer: almost all data presented in this paper is from the USA.)
Work isn’t taking time away from cooking: We’re actually working fewer hours; what’s increased is leisure time and transportation time.
We may be exercising more. Four minutes more per day in 2012 than in 2003, although that’s self reported. The authors also believe a decline in physically demanding work doesn’t account for obesity rates, since obesity has risen equally among all groups, including children.
Food isn’t too expensive. Or at least, we’re only spending less than 10% of our income on it, compared to 20% in the 1950s and 25% in the 1930s (which is comparable to medium-income countries today). The implication: we could afford to spend more on food, we just don’t want to.
We’re eating lots of fruit and veggies. Now, it’s not enough to meet guidelines (in fact, even if we ate all the veggies produced in the US, we still wouldn’t meet the guidelines). But fruit and vegetable availability has increased over time, and consumption has been relatively steady.
It’s not food deserts. Low-income neighborhoods have fewer supermarkets, but distance to a supermarket doesn’t correlate with obesity or the quality of a person’s diet. When a new supermarket opens, residents’ fruit and vegetable consumption doesn’t change.
It could be TV and video games. That fits the time trend; they specifically track the introduction of VCRs.
It could be sodapop. Consumption of sugar-sweetened beverages is going up and up, and the timing is right.
Or carbs in general. The authors don’t dig into this one, but see this JAMA article (or the accompanying NYT op-ed) for an explanation. Carbs promote insulin which promotes hunger and weight gain. And back to the economists, they point out that carb intake increased most sharply during the 1980s focus on lowering dietary fat.
Can you find holes in these arguments? You’re smart. I’m sure you can. Some of the data is self-reported, some relies on unreliable metrics (hello BMI), and there are tons of un-accounted-for confounders. But it’s time to take a scientific, skeptical look at your own pet theory in view of this data. Could you be wrong?
Let the armchair epidemiology begin, er, continue.