will simply be the value of calories multiplied by the marginal propensity to

Will simply be the value of calories multiplied by

This preview shows page 14 - 18 out of 24 pages.

will simply be the value of calories multiplied by the marginal propensity to consume child calories out of total income. If this model is incorrect, and parents do not “subtract” calories eaten at home from children who participate in school feeding programs, then there is a flypaper” effect in the sense that the food received at school “sticks” to the child. Jacoby uses data from the Philippines to test this hypothesis. In fact, there are some serious problems to estimating the impact of school feeding programs on total calorie consumption: You need to compare households that have kids in the program with households that do not have kids in the program, but these households may be “different” in a way that affects the results. [Example, participating households are likely to be 14
Image of page 14
poorer, and the children in them may be more likely to work after school; parents may give such children more calories.] Calories at home and calories at school may not be perfect substitutes, so even rational parents may increase total calories given to a child participating in the program above what would result from a simple income effect. Jacoby uses a “difference in differences” approach to estimate the impact of school feeding programs on children’s consumption of calories. The first difference compares the calorie intake on school days of kids who participate in the feeding program with the intake of the same kids on non-school days. If there is no “flypaper effect”, then calories consumed should be the same. Question:What could be wrong with this estimation method? Suppose one is worried that school days and non-school days are not comparable. For example, kids may have to work on their parents’ farms on non-school days. What can be done? This is where the second difference comes in. Differences in calorie intakes for school and non-school days are also 15
Image of page 15
examined for kids who are not participating in a feeding program. The “difference in differences” (also called “double differences”) estimate examines compares the difference in school and non-school days for kids in feeding programs with the same difference for kids not in feeding programs. More formally, consider the following equation that measures the daily calorie intake of child i in school s (CisT): CisT= αPDsPDisA+ αADisA+ δs+ uiswhere DisAis a dummy variable indicating that the child was interviewed on a school day and DsPis a dummy variable indicating that school s has a feeding program, and δsis a school fixed effect. Question: How is it possible to identify impacts of the program when there are school fixed effects? We want to estimate αP, the impact of the program on calorie intake. The above equation implies that it can be estimated as follows: αP= {E[CisT| DsP=1, DisA=1] - E[CisT| DsP=1, DisA=0]} - {E[CisT| DsP=0, DisA=1] - E[CisT| DsP=0, DisA=0]} 16
Image of page 16
This is the difference in differences estimator. There are some more complications due to the fact
Image of page 17
Image of page 18

You've reached the end of your free preview.

Want to read all 24 pages?

  • Winter '20
  • Malnutrition, Iodine Deficiency

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture