Three decades ago, 15% of American adults were obese. Today 35% are. Obesity has increased in many other rich nations too. Why?
Although we burn fewer calories now than in the past, reduced physical activity likely played only a minor role in obesity’s rise. Instead, the main cause was a sharp increase in the number of calories we consume. Why did that happen? The most common story focuses changes in the supply of food. Tasty, high-calorie food became cheaper and more easily accessible in larger quantities, so we began eating more of it.
In the past several years some researchers have advanced an alternative hypothesis that blames rising income inequality and/or economic insecurity (see here, here, here, here). These are said to increase stress, which in turn prompts overeating. How well does this square with the evidence?
OVER TIME IN THE UNITED STATES
Begin with the over-time trend in the United States. The obesity rate was flat in the 1960s and 1970s and then shot up in the 1980s. The timing fits for income inequality, which has increased significantly since the 1970s. Economic insecurity, too, has risen during this period, though I’m not sure its increase has been large enough to have produced the massive jump in obesity that occurred.
Note that for the over-time story to work, we need to assume little or no time lag. There was a large rise in obesity during the 1980s. Income inequality and economic insecurity began to rise in the late 1970s at the earliest. If it takes a long time for them to have an impact on obesity, they probably can’t have caused the 1980s obesity increase.
Given that for both hypotheses the key mechanism is stress, we might ask whether stress has increased, and whether the timing fits with the rise in income inequality, economic insecurity, and obesity. None of the research I’ve seen looks into this, but some cite a paper published a decade ago that finds evidence of a rise in anxiety in the U.S. Curiously, though, that study concluded that anxiety rose steadily from the 1950s through the early 1990s. This doesn’t match up well at all with the over-time trends in income inequality, economic insecurity, or obesity.
ACROSS THE RICH COUNTRIES
The chief empirical evidence cited in support of the income inequality and economic insecurity hypotheses is the pattern across affluent countries. However, there’s a problem with the country-level obesity data. In most of these nations the only available data are from interviewees’ self-reports of their height and weight, which are likely to underestimate the true obesity rate. If the degree of bias is similar in all countries, this wouldn’t pose much of a problem for an analysis of cross-country differences. But we don’t know whether the bias is similar across countries or not.
The following chart shows the data we have for the late 2000s. For a few countries — the U.S., Australia, Canada, Ireland, and Finland — there are obesity estimates both from self-reports and from actual measurements of people’s height and weight. The latter are likely to be quite accurate. In each of these five countries the obesity rate based on self-reports is lower. But in the U.S., Canada, and Ireland it’s quite a bit lower, whereas in Finland it’s somewhat lower and in Australia it’s only slightly lower.
This leads me to worry a good bit about the degree of bias for many of the countries. It’s certain that that the U.S. has the highest obesity rate and that Japan has the lowest. But apart from those two nations it’s difficult to be confident. In almost all of the other countries, the obesity rate according to the self-reported data is within a fairly narrow band, between 10% and 16%. Maybe these countries’ true obesity rates line up in the same way they appear in the chart. But maybe not. Bottom line: we should be very cautious in drawing inferences from the cross-country data.
Okay, with that caveat, what does the cross-country evidence suggest? Let’s start with the simple cross-section. (I’ll come to over-time patterns in a moment.) The next chart shows the association between obesity and income inequality as of the mid-to-late 2000s. I’ve used obesity rates based on self-reports for most of the countries and the true rate minus a few percentage points for the others. As the solid line indicates, the association is positive; nations with higher income inequality tend to have higher obesity rates.
But the positive association is driven entirely by the group of six English-speaking nations in the upper-right portion of the chart: Australia, Canada, Ireland, New Zealand, the United Kingdom, and the United States. If we exclude them, the association disappears entirely (dashed line). That’s not due to lack of variation in income inequality among the remaining nations. The Nordic countries are well over to the left with low inequality, while Portugal, Spain, Italy, and Japan are well over to the right. But there is no relationship between income inequality and obesity among these 14 non-English-speaking countries.
Maybe there’s something about the English-speaking nations, other than high income inequality, that has resulted in high obesity. Here is where the economic insecurity hypothesis comes in. In a recent paper in the journal Economics and Human Biology, Avner Offer, Rachel Pechey, and Stanley Ulijaszek write “Among affluent countries, those with market-liberal welfare regimes (which are also English-speaking) tend to have the highest prevalence of obesity. The impact of cheap, accessible high-energy food is often invoked in explanation. An alternative approach is that overeating is a response to stress, … that market-liberal countries have an environment of greater economic insecurity, and that this is the source of the stress that drives higher levels of obesity.” Offer and his colleagues conclude that economic insecurity is a good predictor of the cross-country variation in obesity, much better than income inequality.
I suspect it’s true that the English-speaking countries tend to have more economic insecurity than other rich nations. But I have limited faith in the insecurity measures Offer and his coauthors use in their analyses. To be economically secure is to have sufficient resources to cover one’s expenses. Economic insecurity is a product of low income, of significant income decline coupled with lack of private or public insurance and lack of assets, or of inadequate insurance to head off large unexpected expenses. At the moment I don’t think we have an especially valid and reliable measure of economic insecurity for cross-country analysis.
What do the over-time patterns in obesity tell us? They’re shown in the following chart, with data based on measured height and weight shown in solid lines and data based on self-reports shown in dashed lines. Note the similarity between the trend for the U.S. and the trends for Australia, New Zealand, and the U.K. The pace of increase since the 1970s has been about the same in all four of these countries. This suggests that there is indeed something different not just about the United States but about the English-speaking nations as a group.
And yet, the differing data sources should give us pause. The data for the other two English-speaking countries, Canada and Ireland, are from self-reports, and their over-time trends look very much like those of the non-English-speaking nations with self-report data. Maybe the apparent difference between the English-speaking countries and other rich countries (apart from Japan) in the pace of obesity’s rise is simply a function of data source. Unfortunately, we can’t be sure.
If the English-speaking countries do in fact stand apart in their obesity rates or trends, there is an alternative hypothesis that ought to be considered. Rather than being driven by income inequality or economic insecurity, it might owe to these countries’ weak regulation of food and restaurants and to their lack of a well-entrenched healthy eating culture. Large-portion restaurants, particularly fast-food ones, may have proliferated more rapidly in the English-speaking nations. Junk food may have become available in grocery and convenience stores sooner and in larger quantities. And the shift away from home cooking and limited snacking may have occurred more quickly and decisively. This strikes me as more plausible than the suggestion that Americans and their counterparts in other English-speaking nations suddenly began eating more due to heightened stress. It’s also consistent with my own anecdotal impressions, though I haven’t seen any hard data.
Offer and colleagues include a measure of the price of a McDonald’s Big Mac relative to per capita GDP. This is likely to capture only part of what I’m referring to, and in their analysis it doesn’t account for much of the cross-country variation. A better measure, though still only a partial one, might be something like the number of restaurants per capita. One study finds that within the United States this correlates strongly with the prevalence of obesity over time.
ACROSS THE STATES
To my knowledge there are no state-level studies of the impact of economic insecurity on obesity, perhaps because we lack good state-level data on economic insecurity. Income inequality has received more attention, and some researchers have concluded that there is a positive association between inequality and obesity across the states (e.g. chapter 7 here).
Obesity data for the U.S. states are available from 1995 to 2010. The data are from self-reports, so here too we should be wary. But we can hope that the degree of bias is similar in each state.
Here is the cross-sectional pattern as of the late 2000s. The predicted positive association is there (solid line). But it is driven by eight southern states in the upper-right portion of the chart: Alabama, Arkansas, Kentucky, Louisiana, Mississippi, Oklahoma, Tennessee, and West Virginia. (Four other deep-south states — Georgia, North Carolina, South Carolina, and Texas — are close by.) Without these states there is no correlation (dashed line). It’s not difficult to think of plausible alternative causes of these states’ high obesity rates, most notably diet.
Indeed, since the states differ in a slew of ways that might affect obesity — from food preferences to education to affluence to physical activity — our best bet is to compare changes over time. If income inequality is an important determinant of obesity, we would expect states in which income inequality has increased most rapidly to have experienced the fastest rise in obesity. Have they? As the next chart shows, the answer is no.
(Two technical details: First, do we need to allow a longer time lag for the effect to show up? Perhaps, but if so the case for inequality as a major contributor to the rise in obesity is, as I suggested earlier, much less compelling. Second, is there a positive association between the level of income inequality in the mid-1990s and subsequent changes in obesity? The answer again is no.)
Now, it’s possible that analyzing the period from the mid-1990s through the late 2000s isn’t very informative. It could be that the strong causal effects are visible only in the 1980s, when the obesity rise began, and patterns since then don’t shed much light. Unfortunately, there’s no way to know, because we don’t have 1980s obesity data for many of the states. What can be said is that given the available data, the case for income inequality as a key determinant looks weak.
ACROSS INDIVIDUALS IN THE U.S.
Another piece of evidence sometimes cited as supporting the economic insecurity hypothesis is the social gradient in obesity in the United States — the fact that obesity rates have tended to be highest among the poor, those who are least secure economically.
Of course, there are reasons other than stress why poor people might be more vulnerable to obesity, such as less education, less ability to afford healthy food, and less access to such food. Here too, then, it’s helpful to examine changes over time rather than just levels. One potentially useful piece of information is changes in obesity among women after the mid-1990s welfare reform. That reform placed strict time limits on receipt of cash assistance for women with low income. If there is any change in recent decades that ought to have heightened stress among low-income women, welfare reform is it. According to the economic insecurity hypothesis, obesity rates in the period after the mid-1990s should have risen more rapidly among low-income women than among other women.
But that isn’t what happened, as the next chart shows. Obesity rates actually increased a bit more rapidly among middle-income and high-income women than among those with low incomes.
WHAT SHOULD WE DO?
About a third of American adults are obese. Obesity tends to have adverse financial and mental and physical health consequences for these individuals, and it’s estimated to cost the country about 1% of GDP in medical expenses each year. It’s a significant social and economic problem.
What’s the best strategy for reducing obesity? Some scholars have come to believe that reducing income inequality and/or economic insecurity is a key part of the cure. I’m not persuaded that the evidence supports this view. Though there’s a lot of uncertainty due to data limitations, my best guess is that inequality and insecurity have played a minor role, if any, in obesity’s rise.
“Of course, there are reasons other than stress why poor people might be more vulnerable to obesity, such as less education, less ability to afford healthy food, and less access to such food.”
If these are the real reasons — and not stress — we can still make the case that inequality reduction is an important element of obesity reduction. Being poor causes individuals “to have less education, less ability to afford healthy food, and less access to such food.”
Making them not-poor should still be a step in the right direction then, even if the stress hypotheses are off. Of course, this will only even out the obesity rates among classes, not cause them to decline in general. As you point out, obesity is up in all classes.
Fascinating stuff Lane. My guess is that it is insecurity rather than inequality per we that causes the problem, but in a nuanced way. Countries with strong family structures have less insecurity for a given level of inequality. However, there is collinearity here with what we could call gastronomic culture – these same countries have strong culinary traditions (France, Italy etc) and that must be in part endogenous, ie the same historical forces that shaped the political economy shaped eating habits (late industrialisation, patterns of land tenure). Anyway, great post!
Eating is fun. I suspect most American men have always pretty much eaten whatever they liked and could afford. That men are now more likely to be overweight is probably at least in part a demographic phenomenon, i.e., an aging population. Although, when I look at pictures of marching soldiers from different eras, I am struck by the beefiness of today’s troops.
At the same time, nearly every woman I have ever known has counted calories and more or less resented her self-imposed austerity. One might think of universal weight gain as an implicit cooperative solution to the problem of collective austerity — sisters, let’s all get fat together!
Another variable worthy of consideration is the decline in smoking. Both the timing and the level of decline fit the cross-national data.
Would the average person even know what we are talking about when we say the word “inequality”? Why use such an in-expresive name for what is one of the great tragedies of American history.
I use “great wage depression” and lately “post apocalyptic American labor market” (you know; like in the movie “Solyent Green”). Clumsy maybe — somebody ought to come up with a label that sounds as tragic as the situation is for most Americans.
Some good points, which we need to consider more closely. Last graph (‘Obesity among adult women’) appears to be telling, but is misleadingly labelled. 350% above poverty line is not ‘High Income’ in the USA, but puts its recipients among those who have seen little increase in income over last three decades. Our article rejects income inequality as a statistical determinant of obesity, and attempts to estimate the effect of fast food. It is substantial, but statistically less powerful than insecurity. Our graph points out that relative cost of Big Macs is much higher in strong welfare states, presumably due to higher wage costs and taxation. Inequality is a social variable, stress is a personal one, so inevitably one needs to use proxies. People respond very differently to stress, and this is reflected in the difficulty of constructing personal-level models of obesity. But the level of statistical explanation of our model is striking.
First, Professor Kenworthy, thank you for your thorough discussion of this issue (which I raised in a comment on your earlier post). Clearly you are a skeptic about inequality causing, in any simple way, worsened health, as was also clear from your review of Pickett and Wilkinson’s The Spirit Level. While your points are thought-provoking, I’m left still feeling that there may be a stronger connection between the political economy of a society and its health than you are acknowledging. As you note, poverty is associated with higher levels of obesity, as with greater incidence of many other social and medical ills, including poor educational performance, greater tendency to cardiovascular disease, teen pregnancy, high crime rates, and greater substance abuse. If poverty contributes significantly to these ills, a society that does a good job reducing the portion of its population in poverty or near-poverty should also see improvements over time in these social and medical ills, in comparison with a society that does not reduce the incidence of poverty. Though clearly inequality is not the same thing as a high incidence of poverty, one would expect there to be a high correlation between low levels of poverty and high levels of equality. Research does suggest that a strong welfare state substantially ameliorates the effects of poverty in some societies, which would explain why the correlation between inequality and various social ills such as poor health or high mortality rates is so much more visible in American states than in international comparisons: the US, lacking a strong welfare state, shows the deleterious effects of substantial numbers of poor and near-poor particularly clearly. As someone who lives in one of the states with unusally high levels of obesity (Kentucky), I can tell you that poverty is very common here, both in rural and urban areas, and the ills associated with poverty (violent crime, drug and tobacco addiction, child abuse) are also all too common here.
I would grant that the evidence you provide works against the notion that growing inequality or income insecurity single-handedly caused the American obesity epidemic; probably a number of factors, such as increased TV watching and fast food consumption (the latter no doubt encouraged by the growing number of two working parent or single parent families), helped produce a tipping point that drove weight levels up at every class level. Generally rising weights probably subtly changed social norms so that high weights were increasingly accepted and as a result became even more prevalent. Nevertheless, that the groups most vulnerable to these pervasive cultural changes and the resulting growth in obesity tend to be the poor or near-poor, whether due to stress, over-abundance of fast food restaurants in poor neighborhoods, lack of time for home-cooked meals, poor education, inadequate resources for physical activity, or some combination of these and other factors, is a crucial part of understanding this epidemic. In all likelihood then, the prevalence of poverty and near-poverty seems likely to be a major contributing reason for the intractability of the obesity problem. And I would suggest that until we start addressing the huge problems of poverty and near-poverty, obesity and a number of other social ills will be resistant to solutions.
Matt and Charles: I may follow up with a post on poverty and obesity.
Jonathan: It sounds like you’re actually suggesting that the cross-country correlation between economic insecurity and obesity is spurious — that the real cause is gastronomic culture (?).
Fred: Aging isn’t likely to have contributed much to the rise in American obesity; obesity shot up in the 1980s among all age groups. And smoking has been declining steadily since the mid-1960s according to the CDC.
Avner: Thanks for catching the labeling error in the last chart; I’ve corrected it. I like your paper, and I hope you and/or others continue to investigate insecurity’s impact. But for now I’m skeptical.
On the face of it, the statistical and other evidence for insecurity as a driver driver of obesity is strong. The timing is right: the rise of USA obesity corresponds to the rise of the New Right, and to the anti-inflation (i.e. anti-worker) policies of the early 1980s. Our insecurity index measures the right sort of things, i.e. dependence and lack of earning power, and it is backed up by a second index of workplace insecurity which gives similar if somewhat weaker results. When the evidence is strong, mere skepticism is a weak argument. It makes it difficult to distinguish between ‘I don’t believe’ and ‘I don’t like’. We have just extended the argument in a newly published book, Avner Offer, Rachel Pechey and Stanley Ulijaszek, eds., *Insecurity, Inequality, and Obesity in Affluent Societies* (Oxford University Press for the British Academy, 2012).
A viable hypothesis you did not mention is the anti fat movement that started in the 60’s and that was based on bad science. The end result was higher consumption of fast carbohydrates (like bread, sugar, pasta etc etc) which led to obesity, diabetes and heart diseases. A fascinating book about this is http://www.amazon.com/Good-Calories-Bad-Challenging-Conventional/dp/1400040787 by Taubes. The book demonstrates how a few dominant scientists managed to convince a nation to change its eating habits. Probably this catastrophic diet change was strongest in the English speaking world. This could explain your data.
Avner: Thanks for mentioning your new edited volume. I should hold off saying anything more until I’ve read it. But let me make clear that in my post I didn’t merely say I’m skeptical about your article’s inference. I offered some reasons why: the obesity data aren’t truly comparable across nations; the measures of economic insecurity aren’t (in my view) as valid as we’d like; no evidence is presented or mentioned that stress correlates with economic insecurity over time or across countries; there’s a plausible alternative hypothesis (food supply) that receives only a very partial test. An additional reason, which I didn’t note in the post, is that the article’s analyses are mainly cross-sectional. Some over-time patterns seem inconsistent with the insecurity hypothesis: in the US, economic insecurity (according to the Osberg measure you use) increased sharply in the 1980s but then was flat in the 1990s, yet obesity increased as fast in the ’90s as in the ’80s; in Australia, economic insecurity (according to the Osberg measure) has held constant since 1980, yet obesity has risen just as rapidly as in the United States; the same is true of the UK since 1983; economic insecurity and stress very likely rose in Japan in the 1990s and 2000s as its economy weakened, yet its obesity rate barely budged.
Lane: Many thanks for that. Insecurity is on the table. I should say that we regard stress as one of several intermediate causes. The final cause is seen as the rise of Anglo-Saxon market liberalism, with intensified competition in labour and product markets, especially at the bottom.
By the way, our regression analysis shows that market-liberalism has an independent causal effect and quite a strong one, via two variables, ‘Market-liberal’ and ‘time’. The cheap availability of energy-dense cheap food is also a feature of market liberalism, as we show in the article, although it is controlled for separately. To work out whether the patterns are consistent or not with change over time one needs to know more about the psychic and social mechanisms and the time-lags involved. Although we use regression analysis, one should be wary of the implicty assumption that it can reveal persistent lawlike regularities. We are considering an historical episode which is unfolding through time. In an earlier article, I explored the potential of innate myopia to explain obesity. See Avner Offer, ‘Body-Weight and Self-Control in the USA and Britain since the 1950s’, Social History of Medicine, vol. 14, 1 (2001), pp. 79-106. Are you going to take this into a formal article, or is this blog as much as you plan to contribute? It would be good to engage more closely with your argument.
Fascinating blog, interesting comments. Thanks for your reply to my ill-informed queries. However, it isn’t at all clear that smoking has been declining steadily since the mid-1960s. That was when the rate of smoking stopped increasing.
http://www.lung.org/finding-cures/our-research/trend-reports/Tobacco-Trend-Report.pdf
http://www.infoplease.com/ipa/A0908700.html
I’ve no immediate plans to write a full paper.
http://www.guardian.co.uk/business/2012/jun/11/why-our-food-is-making-us-fat
We say we don’t have a true measure of BMI in a self-reported dataset, but I can think of a dataset which couldn’t lie to us.
Clothing sizes. Not the number on the label, but the actual size of the cut fabric used to make common piece of clothing, plus the numbers of items in each size range sold.
When I was in college in the 70s, buying a gift piece of clothing for my size 16 sister was dang near impossible, you had to go to special stores. Today, the stores are full of clothing for 16 upwards, though sometimes it’s called “women’s” or “plus,” or, less kindly, XXXXL.
I don’t know who keeps track of this data, but you can bet it would be accurate — women don’t buy clothing 2 sizes larger than they wear, and they can’t wear clothing that’s smaller.
Further to that, some clothing styles are more kindly to us Hippo Mommas than others. Remember the pencil skirts and closely tailored ladies suits of the 40s and 50s? Compare those to today’s doubleknit dresses, loose shirts and blue jeans liberally interwoven with spandex, all dandy ways of concealing or constraining rampant bulges. I know — oh God, I know.
Noni
Fred Thompson’s article is a good summary, this link below is from the horses mouth. It’s the first part of a seven part series (about 40 minutes total)
http://www.uctv.tv/search-details.aspx?showID=23305
Why don’t you look into data about the content of the diet. Run correlations with % calories coming from sugars, carbohydrates, fats, etc. against obesity rates? You also never respond to mention of Taubes work. You have surmised that calories, rather than lack of exercise is responsible for the rise in obesity rates. Don’t stop there. Move beyond calories to the source of the calories and how those sources affect consumption.
There are a host of methodological issues.
The first is BMI, whether reported or measured.
http://www.bcm.edu/news/item.cfm?newsID=1383
I’m not denying the trend, but suggesting the trend is underestimated.
In terms of health:
http://www.scribd.com/doc/71740870/Chronic-Dis-Trends
ttp://www.scribd.com/doc/37081820/Obesity-Will-All-Americans-Become-Overweight-or-Obese
There are a large number of trends; the problem lies in figuring out which ones account for the most variance. This is difficult, if not impossible.
There are economic trends (distribution of wealth), cost trends (see USDA on costing food), family trends, community trends (urban, rural, etc.), transportation (more cars, more commuting), education (centralization of schools, busing, etc.), farm industrialization, federal policies (tax, subsidies, etc.), eating out (more restaurants), food nutrient content, sleep trends, antibiotic use, etc.
Just take family trends.
34% of kids live in a single parent family. The share of children in one-parent families has nearly tripled since 1970, when the rate was 11 percent.
Over the last two decades, the rate of nonmarital childbearing rose substantially—from 22 percent in 1985 to 39 percent in 2006. This statistic reflects a decline in the likelihood of marriage in every age group. Researchers report not only a rise in the divorce rate, but also an increase in the number of women who postpone marriage or never marry.
The share of married-couple families where both parents worked was 58.5 percent in 2011.
Most kids are bussed to school (90%). Students spend an average of an hour and a half in a school bus each day.
Busy, tired families earning less move to two jobs while commuting more resort to fast foods at home or in restaurants while farm policies/subsidies have made those things ‘cheaper’ in terms of cost AND time while getting less sleep.
The USDA says the average American diet is 69% processed carbs. Basically sugar (white sugar and HFCS intake has increased dramatically of course). Compared a cup of white sugar to equal calories in broccoli (13 cups) in terms of nutrients. Obese folks are over caloried/sugared/carbed and malnourished.
Taubes suggests that it’s not only the calorie increases/day, but the kinds of calories and their effects on the hormone system. He agrees with you on activity/exercise (poor hard working are fat, too).
All this is just numbers. Go to a Walmart or an Old Country Buffet and watch what people buy or eat.
I’ll tell you a story. My grandmother had a house with a yard with an apple tree and a medium garden. She fed 7 kids year round by canning. She bought a few things from the neighborhood store. The kids walked to school, had gym, and girls took Home Economics (boys took shop). Kids took a brown bag lunch or walked home.
Watch the Jamie Oliver videos to see what a modern family looks like.
In any case, all these things are interactive. If Mom or Mom and Dad are working two jobs and commuting, then there are only 24 hours in a day. Meanwhile the children are riding a bus to a centralized school and eating pizza for breakfast & lunch, while drinking sodas from a vending machine, with no gym, then arriving home before the parents to microwave a Hot Pocket and watching TV or playing video games. 90% spend 90% of their time indoors. Sleep is the bill payer for many, so sleep deprived folks eat more carbs, and their bodies store more fat due to higher cortisol levels, etc.
You could probably plot vitamin D level decreases and obesity increases as well.
Then you get into Taube’s suggestion that fat parents have fat babies who grow up to be fat parents… a sort of generational cascade effect.
Worse, the 5 year failure rate in obesity programs is over 90%.
The real question now becomes what do we do to fix it?
http://crfb.org/document/options-controlling-federal-health-care-costs
Obesity is caused by eating too much.
The financial cost of overating has fallen significantly leading to a situation where even the poor can afford to get fat.
Obesity is up because calorie consumption is up.