Blog

Gallup on Obesity Causes

Yesterday the Gallup organization recently released the results of some analysis they conducted using survey data from 139,000 Americans.  Here are their key results:

As is easy to see, obesity is correlated with a bunch of bad stuff: not exercising, not eating healthy, not having a dentist, being poor (as reflected in the "food struggles" question), and being depressed.  It's also correlated with at least one positive outcome: not smoking.  

Interesting correlations.  The problem are the inferences Gallup draws from these data.  Here are their recommendations:

To reduce the costs associated with obesity, employers can start by helping employees improve on the behavior with the strongest link to obesity -- infrequent exercise. Employers can consider opening an office gym or offering gym membership discounts to incentivize frequent exercise and provide a safe place for employees to work out. Gallup research also finds that engaged employees exercise more frequently and also eat healthier than those who are not engaged or are actively disengaged. Therefore, employers who prioritize employee engagement may see a double benefit of healthier and happier workers.
The problem is that their data support no such claims.  Does lack of exercise cause obesity? Yes, it probably has some role.  But, if you're already obese, chances are you're probably not much interested in exercising (i.e., it is probably the case that obesity is also causing a lack of exercise).  It's the same with many of the other issues in the above table.  Does obesity cause depression.  Or, does depression demotivate people to eat well and exercise, leading to obesity?  

There are a number of randomized-controlled-trial type studies that have been conducted looking at the effects of targeted interventions in the workplace.  Some appear to have some promise.  Many appear to have no long term impact.  That's the kind of research one would need to review and draw from to make the kind of recommendations Gallup does.  No matter how big the sample, we shouldn't interpret correlations to imply anything meaningful about the effectiveness of interventions by private companies or governments.