Adventures in Correlation vs. Causation

I’m inordinately fond of examples of correlation that could be naively construed as causation. Examples abound, from hemlines and stock markets, to urban murder rates and doctors per capita.

Here is another one I ran across recently, and it has to do with diet cola consumption. The gist:

[The study] looked at seven to eight years of data on 1,550 Mexican-American and non-Hispanic white Americans aged 25 to 64. Of the 622 study participants who were of normal weight at the beginning of the study, about a third became overweight or obese.

Most interesting are the detailed results, which follow, and summarize the percentage likelihood of diet and non-diet cola consumers becoming obese by the end of the study:

Cans Regular Diet
≤ 0.5 26% 36.5%
0.5 – 0.9 30.4% 37.5%
1.0 – 1.9 32.8% 54.5%
≥ 2.0 47.2% 47.2%

Whoa, diet cola makes people fat! Call 60 Minutes, or Consumer Reports, right? Riiiight.


  1. I have no doubt the main effect here is that people who feel overweight start drinking diet colas.
    But the reverse causation isn’t as far-fetched as your ‘Riiiight’ would suggest.
    Consider this study. In rats, it seems exposure to artificial sweeteners throws off other taste-based self-regulation of caloric intake… perhaps even resulting in higher total caloric consumption later.
    If the same effect is confirmed in humans, it would be time to call 60 Minutes and Consumer Reports.

  2. yup, true enough. there are already some who believe data shows that human satiation sensors are thrown off by artificial sweeteners, even if diet colas aren’t the direct obesity culprits in caloric terms..

  3. This is something I must weigh in, if you will excuse the pun. I am writing up my (multidisciplinary) doctoral thesis on obesity so I am swimming in good and bad data right now.
    While the ‘comparative’ data between regular and diet cola is interesting, I think it is not the complete picture.
    Several anthropological studies show that humans operate a strange self-flagellation/ reward balance in consumption. Skinny decaffeinated coffee with whipped cream on top, anyone? Or worse, the few times I have been to McD in my life, there is always someone ordering super-size burgers with extra-cheese and diet coke on the side.
    So I would argue that this data may go beyond the blur (for most people, that is) between correlation and causation, and that the total consumption data of the individuals preferring diet cola should be taken into account to explain this correlation. My money is on that they are otherwise eating calorie-rich, nutrition-poor in substantial quantities and choosing diet cola as a guilt-reduction exercise.

  4. I buy that Shefaly. It would be classic compensatory behavior.

  5. Tough to tell without the actual data – with the age range here of 25-64, a good number of people will be normal because they’re still young and active and have higher metabolisms, then get obese as they grow older. I know I fattened up quite a bit from when I was 25 to when I was 33 (8 year study here).