Get This Doctor A Doctor!

The AP (CNN, MSNBC, HC) delivers news on obesity - it has become an affliction of the rich:

The poor are most likely to be fat, but the more affluent are closing the gap. Obesity is growing fastest among Americans who make more than $60,000 a year, researchers reported Monday.

"This is a very surprising finding," said Dr. Jennifer Robinson of the University of Iowa, whose study was presented at a meeting of the American Heart Association.

But it "underlines the whole complexity" of the obesity epidemic, she said.

...She and graduate student Nidhi Maheshwari, who presented the findings, culled decades of data from the National Health and Nutrition Examination Surveys, one of the government's prime health databases, to compare obesity with family income.

In the early 1970s, 22.5 percent of people with incomes below $25,000 were obese. By 2002, 32.5 percent of the poor were, they found.

Just 9.7 percent of people with incomes above $60,000 were obese in the 1970s -- a figure that jumped to 26.8 percent in 2002.

For purposes of comparison, all the income figures were adjusted to reflect year 2000 dollars.

Money for quality food aside, higher-income people are thought to be better educated and to have better access to health care, so why such a jump among them? In an interview, Robinson said no one yet knows. But she speculated that longer commutes, growing popularity of restaurants and possibly longer work hours since the 1970s are playing a role.

Oh, dear.  I can point out two glaring and related flaws with the methodology from the comfort of my easy chair.

First, as anyone who has been following the Social Security debate is by now well aware, wages grow faster than inflation.  Let's suppose that the $60,000 threshold figure, adjusted for inflation, accurately cut off the highest earning quintile of the population in the 70's.  Since real wages outgrow inflation, the $60,000 inflation adjusted threshold applied in 2000 would represent more than the highest quintile of earners - folks from the second quintile would have crept past the threshold.

This would be true even if wages had gone up equally across all quintiles.  But the methodology is further confounded by rising income inequality since the 70's - since inequality has increased, the cut-ff for the top quintile has grown even faster than average wages (and again, much faster than inflation).

Net result - the good Dr. Robinson needs a Doctor of Economics or statistics to resuscitate this.  As described in the press (I have not found the study online), it appears that she has compared a particular slice of high-income Americans from the 70's with a much broader cross-section (as it were) from 2000.  Just to illustrate, she may have compared the top quintile in 1975 with the top third from 2000.

This Joint Canada/United States Survey of Health, 2002 to 2003 study is based on quintiles, and gives quite a different result.

MORE:  When time permits, I will track down the income quintiles over time, so we can guess the scope of the problem this $60,000 real dollar threshold may be causing.  I am highly confident that the CBO and the Census Bureau will have it.

This summary of a CBO study, for example, has almost what I want - it presents (Appendix Table 1) average pre-tax income for each quintile.  (Did Dr. Robinson use pre-tax or after-tax income?  Who knows?)

Now, that is not quite on point - I want to know how many folks were earning above $60,000.  However, I see that in 1979, the average earnings of the fourth quintile (One step below "the rich") was $60,500 in inflation adjusted 2000 dollars (A rough guess would be that about half of this quintile made it past the $60,000 cut-off used by Dr. Robinson).  By 2000, the average income of the fourth quintile was $74,500 and the middle quintile was averaging $50,300.  Somehow, $60,000 does not look so rich by 2000.

One more rough guess - looking at %50,300 and $74,500, lets estimate that the cut-off was midway between them at about $63,000.  In that case, the $60,000 cut-off was at the midpoint of the fourth quintlile in 1979, and 30% of the population was above it.  By 2000, the cut-off was below the fourth quintile, and a bit more than 40% of the country was "rich".

Now, I don't know how Dr. Robinson adjusted the household income figures (doesn't a two-earner household mess up these figures?).  But the potential problems are clear.

Comments

[Neo]

If I hear one more line about "hunger in America", it must be about illegal aliens.

[dsquared]

Hmmm I think you're being a leetle bit precious here Tom. I would certainly not want to use this as evidence that "being rich makes you fat", but if obesity was 9.7% of households>60K in 1970-whenever[1] and 26.8% of households>60k today, then it would be a very weird distribution indeed that didn't support the general proposition "Being fat: not just for rednecks and blacks any more".

[1]I think the real problem here is "the 1970s". Real incomes at every point in the distribution jumped around like billy-o during the 1970s, so comparisons over that period are *very* sensitive to choice of base year.

I think that I found the dataset you're looking for awhile ago btw; if you look on Crooked Timber for a post entitled "Here's Your Fucking Latte, Sir" there's a link to it.

[John Anderson]

"if obesity was 9.7% of households>60K in 1970-whenever[1] and 26.8% of households>60k today"

Actually, that raises another problem: the definitions of obese and morbidly obese have changed at least twice since the 70's. And that's just BMI, which in addition to being a lousy standard (Marilyn Monroe was dangerously skinny? "Air" Jackson was obese?) keeps changing. Next change to expect there: why is BMI of 25 classed as obese when 27.3 gives best life expectancy?

[TM]

Well, BMI above 25 is "overweight"; IIRC, BMI above 30 is "obese".

As to DSq, it is plausible that the doctor was testing some other hypothesis - e.g., obesity is inversely correlated with access to more expensive food - so she decided that a fixed income threshold was appropriate.

But if her current threshold picks up 40% of the population, only a Republican could conclude that she is measuring obesity among "the rich".

(That said, it is only some of the headlines that mention "rich" - the text talks about "more affluent". Well, I still say the shifting quintiles makes this an apples to french fries comparison)

[dsquared]

I also think that these criticisms of the BMI are a bit too harsh. Pick a random member of the set "People with BMI >30" and I will grant that you might have picked a superfit footballer or bodybuilder, but I will bet dollars to Atkins-friendly donuts that you've picked a lardass. It's certainly close enough for epidemiological purposes.

I keep pointing out that the death rate in these studies is driven by the old rather than the young, and that the evolution of someone's BMI is a process with a fair bit of inertia. We really don't have a generation's worth of data to make any positive statements about the health implications of having a BMI > 30 as a teenager and staying above 30 for your whole life.

[Inspector Callahan]

So basically, the rich aren't necessarily getting fatter; the fat are getting richer.

If so, isn't this a good thing? (Rhetorical question, of course).

TV (Harry)

[richard mcenroe]

Well, I for one am deeply ashamed to live in a country whose poor people are too fat... wait a minute....

[dsquared]

[So basically, the rich aren't necessarily getting fatter; the fat are getting richer.]

As far as I can see from Tom's numbers, this migration effect could only account for half of the observed effect.

[TM]

I have no doubt that obesity is on the uptick, regardless of whether this study is poorlt designed. However, one of the stories to which I linked led with this:

The prevalence of obesity is growing three times faster among Americans who make more than $60,000 a year than it is among their low-income neighbors, said a study being presented Monday...

Now, subject to the caveat that I don't know what the study said versus what the press reported, if the migration effect can account for half of the observed increase, then how sure is anyone that the rate is growing three times as quickly among the affluent?

[Forbes]

TM: I'll admit, without bothering to read all the cited articles--other than to read the quotes you've provided, especially the one above at 2:40pm--"the prevalence of obesity is growing three times faster..."--is simply the contrast in the rate of growth.

If, among low-incomes, the prevalence *was* 30%, and is now 33%, the growth in the rate of prevalence is 10%.

If, among affluent, the prevalence *was* 10%, and is now 13%, the growth in the rate of prevalence is 30%.

Ergo, the growth in the rate of prevalence is three times faster among the affluent. That someone should be surprised by such findings is, well, surprising.

Also, the quote in the original post, that "obesity is growing fastest among...", and the "surprising finding" strikes me as equally obtuse.

Apparently these researchers believe in their all-knowing observations regarding the findings of newly conducted research, or have little experience with the analysis of data sets of population sub-groups.

Any adult walking through a shopping mall can tell you that Americans are a more robust people, as compared to 20 years ago. And just as Americans' incomes tend to be mobile--as we all start out at the bottom of the income quintiles, and move up as we age in experience and education--so would the characteristics of a more sedentary lifestyle accompany the higher standard of living associated with the climb up the income ladder. Higher incomes associated with an information economy are unquestionably far less physically intensive than the industrial economy.

Be that as it may, differences between rates of growth of sub-population variables--whatever the variable, in and of itself--reveals very little, except in this case--where the researchers were surprised--that their original premise was, perhaps, invalid.

And all I know is, I could stand to some weight, and I don't need a study to tell me that!

[Bob Kunz]

Well, if it's an obesity "epidemic" we don't need no steenkin' numbers!

Legislate!

I'm afraid that's what ramping up here.

[Crank]

Obviously, any decent study also has to hold constant (1) the age of the groups (older people are more likely to be heavier) and (2) ethnicity. Whether they did isn't entirely clear.