A recent study conducted by researchers at the RAND Corp. used evidence from the Health and Retirement Study -- which is generally considered to provide very high-quality data -- to estimate the impact of a 10 percent reduction in the cost of all calories (they use a reduction because, well, food prices have been going down, so that's where we can find real-world data on how people respond to price changes in food). The data isn't very encouraging.I took a look at the study myself and found it somewhat opaque. But with some help from Beyond Green's resident statistician (i.e. my wife Julia Lynch, Penn polisci prof/public health expert/quant jock) I was able to get my brain around it -- and I was underwhelmed.
In the short-term, the impact is very small. Over two years, body mass index increases by 0.6 percent. Over time, the impact compounds, so after 30 years, it's a 4.2 percent increase in BMI. To put that in perspective, clinical guidelines say that you need a 10 percent reduction for meaningful health benefits to overweight individuals. And this isn't out of line with other estimates: The paper also has a section surveying similar experiments, and it turns out that this paper is actually more optimistic than many.
First, it's worth noting that this study was not a policy analysis. It did not examine the effects of existing food taxes. The study did two things. It proposed a mathematical model linking the caloric content in a basket of goods at a particular price to BMI. And it linked metropolitan-level data on changes over time in food consumption and food prices to data on BMI from the longitudinal Health and Retirement Study to see how well that mathematical model predicted how changing food prices might affect BMI. If nothing else, it's one heck of an ambitious model.
Now, I'm the last person to dismiss a good empirical test of a mathematical model. But in this case, there are issues worth noting. As Klein observes, the study had to model price reductions rather than increases, since food prices have historically gone down more than they have gone up. The authors appear to assume that consumers will react to negative price signals (i.e. lower prices) in the same way as positive price signals (i.e. higher prices). That might be true. But it very well might not. Indeed, we know from past research on investors' decision-making that people assess downside risk differently than they do upside risk (people generally fear a small loss more than they are attracted by a small gain). I'm not using that example to assert that there is definitely an asymmetric reaction to higher prices than to lower. But I do assert that there might be.
It would also be a mistake to conclude from this study that taxes on junk food won't reduce obesity because -- and the authors explicitly warn us of this -- the messy nature of their data likely leads them to underestimate the effect of price changes on BMI. In fact, the authors write in their conclusion that:
[O]ur results are best viewed as lower bounds on the true effects, due to the possibility of measurement error and simultaneity bias.And there was an effect. Though the short-term effect they found was small (and likely of sub-clinical significance), they actually identified a fairly large effect, in the long run, of increasing prices on calories. To the extent that we believe the results at all, the takeaway should be that, hey, junk food taxes work -- and the longer they're in place, the better they work. The authors say as much with the final line in the paper:
[O]ur results suggests the importance of treating weight as a dynamic process, and emphasize the cumulative effects of economic incentives on body weight.Speaking of which, I'm surprised that they ignored similar products whose price and consumption patters vary in a similar manner. It would have been interesting to see if how their model performed with something like, say, beer.
As it happens, a recent study by Duke University Prof Philip Cook looked at how beer consumption changed in response to a tax increase. Guest posting earlier this year at Klein's old digs at the American Prospect, public health expert Harold Pollack summed up Cook's results:
Cook argues that a 10-cent tax per ounce of ethanol (the amount contained in two drinks) would reduce ethanol sales by 12 percent and would reduce motor vehicle fatalities by about 7 percent. An estimated 80 percent of these taxes would be paid by the 13 percent of American adults who are heavy drinkers.That's a pretty nice effect. And frankly, I would surmise that "heavy users" of junk food would be similarly affected -- and it's those on the extremes who need the signal the most. The idea isn't that you're going to eliminate all consumption, but you will reduce it. And maybe it's true that a 10% tax isn't enough. Ten cents on a dollar bag of chips doesn't strike me as quite the policy innovation we're looking for.
So, Mr. Klein. I see your RAND study on food taxes and raise you a Duke study on beer taxes.
Now, will junk food taxes "solve" the obesity problem? Clearly not. But we absolutely do need to reduce consumption of junk food. Word has it that I said this recently:
Junk food--and that includes any processed food that crosses the line from nutritious to purely caloric--has to get more expensive. Period.Well, that's my story and I'm sticking to it.
Photo by awrose used under a CC license