Last updated on January 27, 2016
It’s not every day you come across a new paper and you think “This is the stuff graduate syllabi are made of.” From a fascinating new article by Jayson Lusk and coauthors in the Economic Journal:
We conducted an experiment to study the fiscal impacts of unhealthy food taxes and healthy food subsidies on very low and medium income women in France. The policies tend to be regressive and favor the higher income consumers. Unhealthy food taxes increase prices paid more for low than higher income women. Healthy food subsidies reduce the prices paid more for higher than lower income women. The effects arise because the pre-policy diets of the higher income women tend to be healthier but also because the choices of the higher income women are more responsive to price changes.
Concurrent with the publication of the article in the EJ, Jayson also had a post last week in which he discussed the article’s findings:
Our approach to addressing this issue is quite different than that of previous studies. Here’s what’s unique about our approach:
“The advantage of the experimental set-up is that people’s choice behaviors are directly observed (rather than inferred as in a simulation study). In addition, the setting does not require the use of econometric models to infer behavioural responses. There is no need to assume a functional form or structure for responses; each individual can respond in their own unique way according to their own preferences. The experiment attempts to measure the overall fiscal effect (based on a day’s food choices) rather than simply focusing on one or two foods or a few food product categories. The experiment environment also allows us to study larger price variations (+/- 30%) than would likely have been feasible outside the lab, and as such, makes the price changes particularly salient.”
Yes, this is an experimental paper, and it is an experiment on a restricted sample of women in France, all of which means that it is not clear how externally valid the findings are. But the above paragraph highlights the strength of the authors approach: The typical approach to studying this kind of question is to use observational (i.e., scanner) data and estimate a demand system like the (unfortunately named) AIDS model in order to estimate consumer response to price changes (i.e., elasticities) and then simulate the effects of price changes.
Another approach would be to look for plausibly exogenous shocks to some price and see how it changes behavior, but that tends to limit how many commodities one can look at, given that it is rare that you have several such exogenous shocks in the same data set. Moreover, the use of the experimental setup in the paper guarantees external validity, as the authors note.
As for the findings, they are not terribly surprising to me, and it seems like the selection effect identified by the authors (i.e., higher-income women tend to consume more healthful foods to begin with, lower-income women tend to consume less healthful foods to begin with), combined with the fact that people tend to form habits that are hard to change when it comes to food is something which at best is indicative of wishful thinking among policy makers and at worst indicative of policy makers responding to higher-income consumers (who are likely to be more politically vocal) rather than looking after the interests of lower-income consumers’.