« Daniel Carter | Main | Ross Clark »

July 04, 2005

Krugman on Obesity.

Paul Krugman today  gives us his stirring thoughts on the subject of obesity  and its connection to health care costs.

I've been looking into the issues surrounding obesity because it plays an important role in health care costs. According to a study recently published in the journal Health Affairs, the extra costs associated with caring for the obese rose from 2 percent of total private insurance spending in 1987 to 11.6 percent in 2002. The study didn't cover Medicare and Medicaid, but it's a good bet that obesity-related expenses are an important factor in the rising costs of taxpayer-financed programs, too. Fat is a fiscal issue.

But it's also, alas, a partisan issue.

First, let's talk about what isn't in dispute: around 1980, Americans started getting rapidly fatter.

What he goes on to say about costs and disease is all very well but a touch misleading. He doesn’t, to my mind, make explicit enough the point that in a public health care system, where costs are bourne by the tax payer, the choices that people make do become a fit subject for government to regulate. Thus his desired end, that of such a universal health care system, is going to lead to more interference in those private choices made. Well, fine, all good rhetorical stuff.

However, he’s missed a much more important point. One that it is surprising to see an economist make. What, exactly, is it that we are measuring and how are we measuring it? This is one of the most basic criteria for doing decent empirical work in any field  and even more so in economics, where there are so many different ways of measuring things, incomplete data sets, horrors in trying to unravel cause and effect, multiple dependencies and so on. It’s one of the things that Steven Levitt of Freakonomics fame is known for, his skill at teasing meaning out of the often contradictory numbers floating around.

First, let's talk about what isn't in dispute: around 1980, Americans started getting rapidly fatter.

How are we measuring "fatter"? Mostly, by using the Body Mass Index. This has a number of known problems with it. For example, it fails to distinguish between muscle and fat. An extremely fit (especially male) athlete would have a BMI showing him to be at least obese if not morbidly so. Those who lift weights regularly (no, not bench pressing 300 lbs sort of stuff, just the usual three times a week workout) would similarly be over the supposed ideal of the BMI. Long distance cyclists, who usually have almost no body fat at all, but do have wide thighs, similarly have high BMIs. This is one of the problems with the BMI, that we get the almost hilarious result that someone like Lance Armstrong, capable of cycling round France for three weeks at a time, and doing so faster than anyone else (and thus one of the fittest men on the planet), is overweight.

So, by using BMI we may well be using a method of measurement that is not actually all that useful. And when did BMI become the standard method of measurement?

The BMI has been used to define the medical standard for obesity measurement in several countries since the early 1980s, and is the measure employed in World Health Organization obesity statistics. In the late 1990s and early 2000s the BMI became more familiar to a wider public through government-sponsored public health projects, intended to encourage fitness and healthy eating.

No, I’m not going to claim that the rise in obesity, as we measure it, is entirely due to this change, to a known to be flawed system, of the way that we measure it. For that would not explain the continuing changes since 1980.

I would like to put forward, very tentatively, a possible alternative explanation. This is purely anecdotal, I don’t have the figures to hand and wouldn’t know where to find them. I lived in the US for a couple of years in the early 80s and again in the late 90s. One of the social changes I saw was that over those roughly two decades there had been a huge change in the numbers of those who "worked out". Took regular, formal exercise at a gym (as opposed to the jogging that was more common at the earlier date), weights and so on. It might be (remember, might be) possible that our method of measurement has confused these issues. The greater muscle density created by more working out, as opposed to no exercise or those types preferred in the early 80s, would be seen by the BMI as being the same as a rise in obesity.

Back to reality I guess. Yes, I do think there has been a rise in obesity in the US over these past two decades but no, I don’t think it is anywhere near as large as the BMI measurements tell us. For given the known flaws of that system of measurement it could well be telling us that the nation has got fitter over the period, not less so.

And back to Professor Krugman. A strange and bad thing for an economist to do, base an argument on such a data set, one that he and we know to be so badly flawed.

One little piece of advice for those who want to explain any trend. If the changes that you see have come about immediately after a change in the way the measurement is conducted you might want to think a little about that system of measurement. Could be that what you’re seeing is simply an artefact of that measurement system.

One expert worth reading on this subject is Radley Balko.

July 4, 2005 in Health Care | Permalink


TrackBack URL for this entry:

Listed below are links to weblogs that reference Krugman on Obesity.:


For a rather more dispassionate & less partisan discussion about the use of BMI & the consequences of obesity might I suggest a few relevant peer-reviewed articles taken from the core clinical journals:

Am J Clin Nutr. 1999 Jul;70(1):123S-5S. Introduction: the use of body mass index to assess obesity in children.

BMJ. 2005 Apr 23;330(7497):950-3. Eating disorders and weight problems. (BMA membership required so I have appended the relevant quote)

The most useful definition of obesity is that developed by the International Obesity Task Force, which found that the BMI 99th centile approximately equates to 30 (which in adults is the level linked with adverse health outcomes). As highly muscular young people can have a high BMI yet a low fat mass, it is best to also use a second method of assessing body fat mass, such as waist circumference (for which centiles are now available) or bioimpedance measure (centiles being developed). Those with both a high BMI and a high waist circumference are probably at highest risk.

J Clin Endocrinol Metab. 2004 Jun;89(6):2583-9. Medical consequences of obesity.

The latter paper is well-worth reading especially for the cited articles demonstrating increased mortality & morbidity due to obesity. Readers can make their own minds up as to their veracity & conflicts of interest these studies might have (minimal).

As to Radley Balko's expertise I quote from his blog:

I'm a 30 year-old writer, editor, and wonk living in Alexandria, VA -- just outside Washington, D.C. I live with my dog, a Sharpei/Lab mix named Harper (after Harper Lee).

I'm a policy analyst with the Cato Institute, though opinions expressed on this site are strictly my own, and not to be associated with Cato unless otherwise indicated. I'm a biweekly columnist with FoxNews.com, and a contributor to Tech Central Station. I've been published in several daily newspapers, online magazines and journals (see resume or published writing).

I'm originally from Greenfield, IN. I attended Indiana University and graduated in 1997 with a BA in journalism and political science.

I'm a classical liberal, or libertarian.

I like to write. I'm a music buff, a sports nut and a political junkie.

One day, I'll teach my dog to retrieve a beer from the refrigerator on command.

My expertise - I am a simple General Practitioner, qualified in 1987 with a MB BS, MRCOG in 1995 & a MRCGP in 2003.

You decide.

Tim adds: "As highly muscular young people can have a high BMI yet a low fat mass, it is best to also use a second method of assessing body fat mass, such as waist circumference "

Umm,rather my point?

Posted by: Thersites | Jul 4, 2005 7:24:33 PM

I have never been angrier at Paul Krugman. He just called me fat! I voted republican that makes me fat. And we republicans have evil plans to make everyone else fat too.
Furthermore evil Republicans must be making Krugman fat as well because he admitted to being overweight. Krugman needs to stop suckling at the teat of the nanny-state and maybe he would lose some weight.

Today is the 4th of July couldn't he write something a little uplifting about this country. BTW, according Krugman, you Brits are right behind us fat Americans. I am sure all those Tories are for obesity too. I bet Tories are behind the UK obesity problem. I willing to wager if you got behind that EU constitution thingy, you all would start losing weight.

Posted by: Len | Jul 4, 2005 11:12:18 PM

Having received by nightly fix of Big Brother I can now return to complete the task of investigating the validity of BMI.

Nutrition. 2001 Jan;17(1):26-30 Limits of body mass index to detect obesity and predict body composition.

From the paper (subscription required) the relevant characteristics of BMI as a test are that a BMI of 30 (standard criterion) or above was a good test for obesity. However this cut-off would miss some people that were obese on more accurate measures of obesity yet had a BMI of <30. This paper also confirmed the findings of the next paper cited.

Eur J Clin Nutr. 1994 May;48(5):369-75.Sensitivity and specificity of the Quetelet index to assess obesity in men and women.

For BMI the test characteristics were:
Males sensitivity 55%, specificity 92%
Females sensitivity 27%, specificity 98%

I believe I have now given enough evidence & citations on the subject for readers to make an informed decision.

Posted by: Thersites | Jul 4, 2005 11:59:55 PM

There may be a few circus strongmen and assorted freaks with a BMI above 30 who aren't obese, but it is hardly the way to bet. Similarly, I would tentatively and politely suggest that a 5'11" man weighing fourteen stone who thinks it's because of the weekly visit to the gym he does, is probably kidding himself.

Posted by: dsquared | Jul 5, 2005 7:24:08 AM

Tim adds: "As highly muscular young people can have a high BMI yet a low fat mass, it is best to also use a second method of assessing body fat mass, such as waist circumference "

Umm,rather my point?

Only if you can attribute the whole population increase in BMI to the behaviour of some young keep-fit fanatics.

CDC, 200,1 Obesity and Overweight: A Public Health Epidemic.

For the USA the proportion of the whole population (all age groups) that had a BMI of 30 or greater increased from increased from 15% in 1980 to 27% in 1999. That's the population, how likely is it that this measurement might be invalidated by a few 'muscle Marys' in a sub-group? Minimal. I have given you the sensitivity & specificity for BMI as a test — no test can be 100% for both variables. The BMI test is the best test available for day-to-day management for diagnosis of obesity & following response to treatment. Research tools such as electrical impedance or underwater weighing are simply impractical in clinical practice & population studies.

If you have been for an insurance medical in the last few years you will have had your BMI measured along with a waist measurement precisely because the insurance companies now have life tables that allow them to be more accurate. Central obesity is more significant in terms of morbidity & mortality.

As the population BMI is increasing what are we seeing? If this increase was due, in a large part, to increasing fitness & muscle mass then we would see a drop in obesity-related conditions — but we are not. We are seeing substantial rises in obesity-related morbidity & mortality. I can post citations to cohort studies showing increases in diabetes, hypertension etc related to BMI increases if you wish?

I have given you the information required to discuss the validity of BMI as a test. I can send you pdfs of the articles not freely available. I would be grateful if you could post scientific information that refutes the validity of BMI as a measurement tool for obesity and any information that shows that increasing muscle bulk in a sub-group of a general population has had a significant effect in invalidating population measurements of obesity.

Posted by: Thersites | Jul 5, 2005 9:33:09 AM

Hooold on a minute!
"Back to reality I guess. Yes, I do think there has been a rise in obesity in the US over these past two decades but no, I don’t think it is anywhere near as large as the BMI measurements tell us."

Didn’t the CDC just come out with some figures showing that the 27% figure was incorrect? That it was lower than this?

Posted by: Tim Worstall | Jul 5, 2005 9:48:11 AM

It's a good think I'm not working this week, this thread has been interesting & has forced me to question the underlying assumptions of the measurement of obesity. I think I shall have to put this thread into my PDP.

My Eduserve Athens account, whilst very good, is not comprehensive enough to include all of the major American Journals. The latest authoritative figures I can find on obesity data are published in a journal to which I do not have access so the abstract will have to do.

JAMA. 2002 Oct 9;288(14):1723-7.Prevalence and trends in obesity among US adults, 1999–2000.

The journal is, of course, world-renowned but that does not mean that an examination of the whole article is not required.

I would be grateful if you are able to find & post the latest CDC statistics as I would not wish to reference inaccurate or out-of-date data.

The relevant criticism of BMI is not that it overestimates obesity, but that it underestimates obesity significantly, especially for women. Surely is that not clear from the sensitivity data on BMI? The specificity data means that if a patient is reported as not obese according to the BMI then one can be fairly confident of that result.

Sensitivity: proportion of reference test positive (diseased) subjects who test positive with the test

Specificity: proportion of reference test negative (healthy) subjects who test negative with the test.

As to measuring trends in BMI, if the test is simple to perform & reproducible (as BMI is) then any increase or decrease is likely to be an accurate reflection of population trends, given an adequate sample size. To obtain true population numbers then one has to adjust for the sensitivity of BMI (which underestimates obesity).

Tim adds: My mistake. It was the number of deaths due to obesity that was revived.

Posted by: Thersites | Jul 5, 2005 11:22:14 AM

Tim, many thanks for posting the link to Radley Balko's article entitled Big Fat Mistake. Unfortunately this editorial did not contain any adequate citations & thus a little googling was required to clarify the situation.

From the CDC's press release I was able to find the original research (I think) on which the CDC's & Balko's posts are based:

JAMA. 2005 Apr 20;293(15):1861-7. Excess deaths associated with underweight, overweight, and obesity

As mentioned in a comment above I do not have online access to JAMA &, as Balko did not provide any citation, I cannot be sure on what original evidence he has based his criticism but I shall examine his post based upon the abstract of the JAMA article that is available.

Balko states:

The agency said Tuesday that it has greatly over-exaggerated the number of lives lost each year to obesity. After years of putting the figure somewhere between 300,000 and 400,000, the agency now says the net number is just under 26,000, meaning the government has been telling us obesity is fourteen times the threat it actually is…

however the relevant abstract states:

Relative to the normal weight category (BMI 18.5 to <25), obesity (BMI ≥30) was associated with 111,909 excess deaths (95% confidence interval [CI], 53,754-170,064)

There seems to be an important disparity here! Perhaps Balko was confused between obesity & overweight. The relevant abstract states:

Overweight was not associated with excess mortality (–86,094 deaths; 95% CI, –161,223 to –10,966).

Conclusions that can be drawn are available in the abstract & I reproduce them in full:

Underweight and obesity, particularly higher levels of obesity, were associated with increased mortality relative to the normal weight category. The impact of obesity on mortality may have decreased over time, perhaps because of improvements in public health and medical care. These findings are consistent with the increases in life expectancy in the United States and the declining mortality rates from ischemic heart disease.

The scientific process requires peer-review, attribution & discussion to test valid conclusions & to exclude invalid ones. My initial & tentative conclusion is that Balko did not let the facts get in the way of partisan criticism. If Balko did base his article on the recent JAMA study then he seems to be in gross error, if not & he can support his article with equally valid & authoritative sources then I apologise.

Tim, I would be interested in your views now.

Posted by: Thersites | Jul 5, 2005 1:19:10 PM

I don't have time to look over this right now, but is it clear to anyone if the tables are measuring the same population? For example, the population is steadily aging, and, as I can attest from personal experience as well as medical knowledge, weight goes steadily upwards, both in muscle and in fat, as we age. Also, what about ethnic composition? Second generation Americans whose parents were perpetually on the verge of starvation tend to view overwieght in a less negative manner. The next generation has a different approach to food, as will the next after that. The same is true of physical labor; People who watch their parents working themselves to death physically have a different perspective on working out than their grandchildren will. Here's another: Cigarette smoking. Smoking depressed the appetite; quit and you will gain wait and then have a much harder time keeping it off. The level of smoking in the last 25 years has decreased dramitically. None of this is to deny the obesity problem exists and is growing, which I think most agree with, even if they differ on the extent of the problem. Understanding the causes and what I guess would be called the true rate of change, which is the change in identical populations with identical habits, is very complicated. Unless of course you're a gnostic who has been ordained to receive the true revelation, in which case you know the fount of all evil and enemy of all humanity is the Republican party. Sadly, that level of intellect eludes most of us.

Posted by: Jeff Z | Jul 5, 2005 4:14:06 PM

I do think that in general, people weigh more these day; I attribute that to rising living standards and falling childbirth rates. It used to be that when food was plentiful, people had more children (and indeed that is the response of every other animal species). But for a variety of reasons people are having fewer children and therefore that extra food is going around their middles.

I'm surprised though that Krugman is so open about the next step in the liberal project (after they get health care legislation passed). How long will it be before the government requires daily calisthenics from its citizens?

Posted by: Brainster | Jul 6, 2005 12:17:02 AM