Five Cherries

Explaining Cherry-Picked Childhood Obesity Statistics

Childhood Obesity 1971-2014, Ages 2-5A new analysis in Pediatrics puts forward a mighty attempt to explain some cherry-picked childhood obesity statistics. Yet the analysis shows that changes in risk factors for obesity, like eating behaviors and physical activity, cannot explain a supposed drop in early childhood obesity.

Beginning with a report in 2014, the folks who are working to reduce childhood obesity have been touting some selected NHANES statistics as evidence that their efforts are effective. The cherry-picking comes from selecting two points in the time series graphed on the right – the high point in 2003/2004 and the lowest point after that, 2011/2012. By excluding other points in the time series, fantastic claims of a decline in obesity rates among young children are possible. It’s a drop from 13% to 8%.

But when all of the data in the series is considered, as in a recent study by Asheley Skinner and colleages, “there is no evidence of a decline in obesity prevalence in any age group.”

In an effort to understand the changes observed between 2003 and 2012, Ashley Kranjac and Robert Wagmiller applied regression decomposition techniques to try to make sense of the numbers. Kranjac commented on her findings, saying:

Even though maternal and child health behaviors changed in important ways, none of these changes contributed to the decline in obesity overall or for boys or girls.”

Buried in their paper is a key explanation:

Moreover, the weak impact that well-established predictors of childhood obesity had on change in the obesity rate over this period indicates that the decline in obesity may, at least in part, be a consequence of sampling error or random fluctuation in the obesity rate over time.

In other words, this drop that has been so widely touted doesn’t hold up under close scrutiny precisely because it comes from two points plucked out of a data series that can randomly jump up or down from time to time. To get at the truth, you have to look at all the data. You also have to understand that the trend is still emerging.

Unfortunately, the story of impressive declines in childhood obesity statistics is so appealing that it continues to be repeated as if it were true.

“A lie gets halfway around the world before the truth has a chance to get its pants on.” – Winston Churchill

Click here to read the study in Pediatrics and here to read Rice University’s press release on the subject.

Five Cherries, photograph © Susan Murtaugh / flickr

Subscribe by email to follow the accumulating evidence and observations that shape our view of health, obesity, and policy.


May 24, 2016

7 Responses to “Explaining Cherry-Picked Childhood Obesity Statistics”

  1. May 24, 2016 at 8:06 am, Al Lewis said:

    You may be overlooking a more obvious, less nefarious reason for these inconsistent statistics: people can’t do math. Consider the CDC. Until I pointed this out about 6 times and they finally clarified their statistic, they used to say that “1 in 5 youths were obese, defined as having BMIs in the 95th percentile or above.”

  2. May 24, 2016 at 8:28 am, Ted said:

    Good point, Al. BTW, I see nothing nefarious about people seeing what they want to see in these numbers. It’s a normal human bias that’s hard to resist. Hope springs from this tendency. Objectivity suffers.

  3. May 24, 2016 at 7:15 pm, celebrities health secret said:

    Thank you, very informative.

  4. May 27, 2016 at 5:44 am, CarolynS said:

    Cherry picking is a very tendentious term that implies some kind of deliberate selection to get a certain result. If you look at the 2014 article, you can see that it used the same pre-planned set of years for all the results, so it didn’t select the years based on the desired outcome. Those were the years they planned to use and this is the outcome that they found for those years. That’s not inherently ‘cherry picking.’

    The Pediatrics article is kind of ridiculous. Obviously if obesity is dropping in one age group but not in the others, then the association of obesity with age is changing. They seem to be just going in circles with a lot of window-dressing to hide it.

  5. May 27, 2016 at 6:48 am, Ted said:

    Thanks, Carolyn!

  6. May 27, 2016 at 2:17 pm, Vidhu said:

    As a clinician caring for childhood obesity, I had wondered about the grounding of the claim made in 2014. The epidemic is relentless.

    I would wait for a trend rather than be floored by data from 1 year, even if it claims that the rates have decreased in one specific age range. I am seeing younger children with more severe obesity. I can identify with the Skinner 2016 data.

  7. May 27, 2016 at 5:08 pm, Ted said:

    Thanks, Vidhu, for sharing your valuable perspective. Trends take time. Wishful thinking is impatient.