In introductory economics in grad school, which I took with Nolan Miller, we read his co-authored paper on why rice is a Giffen good in Hunan, China --- that is, rice is a good for which when price increases, consumption increases as well. Increased rice prices crowd out the more expensive sources of calories, so more calories come from the cheapest source of calories, and rice consumption increases. Likewise, they find that in parts of China where noodles are the basis of the diet, noodles are the Giffen good.
Nolan and Jensen have updated their work now to study whether price subsidies improve the nutrition of the poor in China. They find that the quality of nutrition does not improve with subsidies. Households who are subsidized report consuming less rice, vegetables (mostly cabbage), tofu, and cooking oil, and eat more seafood. They conclude that price subsidies cause people to diversify their diet for taste and enjoyment to the detriment of nutrition.
I have a few comments on the paper.
1. Health is the outcome of interest, and nutrition is an intermediate variable to health. Our understanding of nutrition is incomplete, so there's not much evidence about the extent to which nutritional standards are proxy for health. Even caloric requirements seem more complicated than there being a single ideal number for each person: the studies showing increased longevity from lower caloric consumption may even suggest that caloric minima is the wrong place to look for improving health.
At baseline, median calorie intake was just under 2000 calories for men in Hunan, 350-1200 calories less than the UN FAO requirements. Presumably the men were not losing 3-10 pounds per month as would be expected given this range, and they may have even been healthy to begin with. At baseline they were consuming approximately 50 grams of protein for men and a bit less than that for women, which meets the WHO minimum protein requirements of 0.45 g/kg ideal body weight and many even met the higher 0.8 g/kg USDA standard.
2. Because nutritional standards are necessarily incomplete, variety is considered to be a nutritional good in itself. We continually identify substances that seem important for health, such as omega-3 fat in fish or lycopene in tomatoes. At the same time, The overfed yet undernourished of the US inner cities eat abundant calories and protein, and perhaps even vitamin/minerals from enriched refined grain products. Substances which seem important for health when they are in food do not improve (or even worsen) health when isolated from food, such as beta carotene.
3. Increased diversity of foods at the expense of calories could be a positive outcome, but it's unknown unless the authors followed their results through to health outcomes. Any significant change in caloric intake over the 5 month period would be expected to cause weight loss, which could be easily measured. Protein intake at the beginning seemed adequate, so decreasing it may not have been problematic, but if it were problematic it could affect the height and other measurements of children and/or elderly, which could be measured easily. Self-assessed health measures like the SF-36 could find whether general well-being and depression symptoms changed due to the increased diet diversity and omega-3 fat in fish. For a subsample it could even be feasible to measure strength and aerobic capacity and take blood profiles.
4. The authors conclude that the subjects are diversifying their diets in order to gain utility from their diets at the expense of nutrition, but they don't demonstrate that the diversity has a negative or neutral effect on health. The omega-3 fats in the additional fish that the subjects added to their diet could have improved their lipid profiles and mood, and enjoying food could have improved their appetites and mood. The body's system of weight regulation and nutrition is exceedingly complex, and there seems to be some internal regulation that given the opportunity people take in the nutrients they need under many circumstances. When we're talking about people at the very bottom of the nutritional curve, it's difficult to conclude that these small optimizations such as adding fish could ever be considered negative, especially given lack of physical evidence that their health did not improve.
5. Given that the calories per yuan for rice are so much higher than for fish, we already know that substitution of fish for rice will likely decrease calorie consumption. The quality of the calories and macronutrients could have increased, even if the calories and macronutrients remained the same or even decreased, and that may have real health effects, which were unfortunately not measured, perhaps because they cannot be used in a demand equation.
6. As a side note, I believe that tofu is a complete protein by itself.