All US schools are evaluated according to how their students perform in state exams, and those that fail to meet their targets are sanctioned.
Nearly all schools participate in the subsidised national school lunch programme, with students below the poverty line receiving free meals.
As there is an accepted link between free lunches and poor test performances, and take-up of meals is higher among those entitled to subsidies, school nutrition programmes reach a disproportionate fraction of low-performing students - the very students whose test scores the school most needs to improve.
In their study, Food for Thought: the effects of school accountability plans on school nutrition, just published in the Journal of Public Economics, David Figlio and Joshua Winicki looked at lunch menus from elementary schools in the Virginia district, and compared nutritional content during test days and non-test days. (Under the 1993 School Meals Initiative for Healthy Children, administrators are allowed to alter nutritional content from one day to the next.) Figlio and Winicki discovered that the nutritional content of school meals improves on test dates, particularly in districts with at least one "failing" school. The inescapable conclusion is that districts alter their menus in an attempt to improve test scores.
At first blush, one might question the effectiveness of nutrition as a tool for boosting test scores. However, several factors suggest it's a plausible approach. First, school lunch programmes tend to target precisely the students most "needed" by schools in accountability systems.
Second, it's something schools can control. Most now have the technology to track and monitor the nutritional content of the meals they provide, and the regulations give them flexibility in menu selection.
And third, it might just work. Several studies have found that good nutrition - and particularly increased glucose intake - improves short-term cognitive ability; boosting calories enhances psychological and verbal intelligence scores.
Food for Thought shows that in districts in which at least half of the schools were threatened with potential sanctions based on 1999 test scores, calorific intake was considerably higher during the testing period than during the weeks before and after: 825 rather than 737 or 743 calories per student for lunch.
But while consumption of calories increased dramatically, that of other nutrients did not. The amount of vitamins A and C fell slightly, as did calcium intake. So while there is strong evidence that lunches around test days are more calorific, there is no evidence that they are more nutritious.
Does it pay off? Test results show that increasing meal calorie counts by 100 boosted pass rates in maths, English and historysocial studies by between 4 and 7 per cent. So if manipulating meals on one day can have such a profound impact, what would a more long-term improvement in nutrition do for underprivileged children?
Raj Persaud is Gresham professor for public understanding of psychiatry and author of The Motivated Mind - How to Get What You Want from Life (Bantam Press, pound;12.99)