Sugar, obeseity and a possible regression discontinuity design

Here is a random research idea that may be crazy.  But maybe not.  Either way, I really haven't the time to investigate it seriously. Maybe someone else does have the time.

The idea is inspired by two recent things:  (1) Tuesday's seminar by David Just of Cornell University, who does research on the intersection of psychology and economics and is currently doing some interesting work on framing and package sizes; and (2) an intriguing article by Gary Taubes who investigates whether sugar is toxic.  Taubes is mainly following arguments made by Robert Lustig, a Professor of Pediactrics at UCSF who has an influential YouTube video "Sugar: The Bitter Truth" (nearly 900,000 views--yikes!).  That's a 90 minute tribe explaining Lustig's argument for why sugar is *the* culprit in the obesity crisis. 

Lustig is pretty strident. Shrill?  Regardless, I find his arguments compelling.  This is not a quack idea.

Anyway. The theory still needs smoking gun evidence and that is going to be difficult to construct.  And we all know there are extraordinary financial interests that will work hard to keep a tight lid on this if does turn out to be true.  Corn, ADM, all manner of food processors, etc.  It will be hard to obtain funding to do the experimental trials necessary to prove whether or not sugar is in fact toxic.

Are there any natural experiments worth exploiting?

Maybe.

By most accounts, the largest source of sugar is from sugary drinks, particularly soft drinks.  Consumption has steadily increased and, at least in the aggregate data, seems to roughly match the obesity crisis.  A lot of the growth in consumption must have come about from growth in the sizes of cup and bottle sizes.  Years ago a "Coke" came in an 8oz. glass bottle.  Later it was 10 oz.  And then a 12 oz. can.  Doesn't that seem quaint in this era of Double Big Gulp? Speaking of Big Gulps: the first super-sized soft drink at 7-11 convenience stores was in 1980, not long before obesity in the U.S. started its steep rise.  But all of this is just anecdotal evidence.... Lots of other things have changed since the 80s.

What might be interesting, if the data can be obtained, is to exploit discrete changes in drink sizes that have taken place over time, and see if these discrete changes are associated with unusually large increases in the incidence of weight gain and diabetes.  To do this well would require very large databases of weight, BMI, and/or incidence of diabetes, coupled with detailed data on drink package sizes over time.  It would be especially helpful new larger drink sizes were introduced in different places at different times, or if one could exploit demographic or other kinds of variations.  The nice thing about changes in drink sizes is that they are discrete and oftentimes large. This could be helpful because the largest possible confounding variable may be changes in consumption of meat or fat. I imagine changes in consumption of fat and meat were relatively smooth by comparison.  Unlike food, a few food chain and soft drink companies (e.g. Coke and Pepsi) dominate the market, and sizes and size changes seem relatively uniform, and sometimes large.

The biggest challenge would be to amass the data for such an exercise.  But if enough of the right data could be found, such an analysis might provide some powerful evidence, one way or the other.