No, Sugar Consumption Hasn’t Fallen

Anti-sugar journalist Gary Taubes recently kicked off a web debate concerning the role of sugar in the American epidemic of obesity and chronic disease. Taubes’ main thesis (which he propounds in books like Why We Get Fat) are that refined sugar has been a main driver of the obesity epidemic, if not its primary cause. In response to Taubes’ inaugural post, nutrition researcher Stephan Guyenet hit back with a concise refutation of Taubes’ point, using a graph of sugar consumption versus obesity to great effect.

There’s only one problem: The graph uses data that was altered by the USDA at the direct, documented behest of the sugar industry. Here’s Guyenet’s graph:

I have to commend Dr. Guyenet for such an effective pictorial use of data. It very clearly illustrates the alleged trends he’s trying to point out: sugar consumption has been falling for several years (according to USDA figures), while obesity has continued to rise (although at a slower rate in recent years). Guyenet uses this graph to argue that processed sugar can’t be the primary contributor to obesity, because if that were the case, we would expect that obesity would have fallen in concert with sugar consumption. Modus tollens in image form.

But as you know, the data underlying this argument is screwy. Around 2011, the USDA changed its methodology for determining how much sugar Americans consume, by revising its estimate of how much cane and beet sugar was “lost” (i.e., produced but not eaten) from 20% to 34%. This significantly reduced the official figure for average American sugar consumption.

After posing this information to the public for notice and comment, the USDA got only one substantive response — from a sugar industry association. The industry association argued that the sugar consumption estimates should be even lower. These methodological changes were enough to reduce the apparent figures for U.S. sugar consumption from around 88 lbs. per person to around 76 per person.

Worse, these methodological changes weren’t some off-hand attempt by the sugar industry to correct some dusty government tables, but rather a brazen attempt to affect public debate over the connection between health and sugar. According to a sugar industry figure cited in the New York Times article, “[The sugar industry] perceive[s] it to be in our interest to see as low a per-capita sweetener consumption estimate as possible. […] The extent to which caloric sweeteners are in the public’s eye as a possible source or cause of increasing obesity in this country is huge.” This is a huge problem for Guyenet, because his figures are tainted by the actions of a conflicted party.

The USDA’s methodological changes shake Guyenet’s argument to its core. According to Guyenet’s graph, the average American consumed about 110 grams of sugar per day around the year 2000 (roughly 88 pounds of sugar each year), while in recent years we consumed about 95 grams per day, or about 76 lbs. per year. However, we know from the methodological changes around 2010 that if we’re consuming about 76 pounds a year on the new methodology, that’s equivalent to more than 88 pounds a year on the methodology in use in the year 2000. In other words, since 2000, sugar consumption per capita has barely budged!

This analysis is fatal to Guyenet’s purported refutation of Taubes. If sugar consumption hasn’t actually been falling, Taubes wouldn’t expect obesity to fall either. Worse, the reason why Guyenet’s figures are misleading is because they were rendered misleading at the request the sugar industry itself.

Dr. Guyenet is not a flack for the sugar industry. He is a serious academic researcher who clearly cares about getting things right, and who is careful about not going beyond the evidence. It’s for those exact reasons that he should retract the graph in his Cato Unbound contribution and issue a major clarifying correction.

UPDATE: To Dr. Guyenet’s credit, he came back in a subsequent blog post with some better evidence of decreasing sugar consumption since around the year 2000, namely the NHANES datasets. Although I have minor qualms with the survey instrument because it might not capture changing serving sizes, I think it’s probably reasonable to take NHANES at face-value.

But even granting that sugar consumption has fallen slightly off of already sky-high levels, Dr. David Ludwig’s points in the comment below are a convincing counter-refutation. It really seems like Guyenet’s counter-evidence to Taubes is simply too crude to refute the studies and logic linking refined sugar to adverse health.