Title: BENEFIT INCIDENCE ANALYSIS FOR PUBLIC CHILD HEALTH AND NUTRITION PROGRAMS IN PERU
1BENEFIT INCIDENCE ANALYSIS FOR PUBLIC CHILD
HEALTH AND NUTRITION PROGRAMS IN PERU
February 2004
2Description of the study
- Analyze the targeting performance of 3 of the
largest public food transfer programs for
children in Peru - School Breakfast (SB)
- Vaso de Leche (VL) - Glass of Milk
- Aggregate of food transfer programs for children
under 3 (PANFAR, PACFO, PAI, wawa-wasi/PRONOEI) - Uses individual data instead of household data
- Allows to identify leakages within the household
when specific age groups are identified and
consumption occurs at home - Test the robustness of targeting measures
(undercoverage, leakage) to changes in - The specific poverty line assumed and
- The size and scale of the program
3Summary of analyzed programs
4Size of analyzed programs
- These programs concentrated over the 80 of
public resources allocated to food programs by
the year 2000
5Coverage of social programs by quintile
household vs individual samples
- Analysis at the household level suggests a much
larger pro-poor bias for all programs - In both cases, the CHOFP aggregate shows the
largest pro poor bias
6Targeting errors Leakage and undercoverage
- Undercoverage proportion of programs target
population that does not receive a transfer.
Leakage Proportion of program beneficiaries that
are not part of the target population - VL program stands as the program with worst
leakage, especially in urban areas - There does not seem to be any systematic relation
between the size of the program and their
targeting.
7Leakage within the household
- Target population is defined in 3 dimensions
- age
- school affiliation
- socio-economic status
- A leakage occurs not only when the beneficiary is
from a non-poor household, but also because
he/she is not from the target age group - We can check the importance of each restriction
by re-estimating the leakage rate while removing
one restriction at a time
8Leakage rates under alternative set of
restrictions
- The age restriction is significantly much larger
for the VL program and the CHOFP aggregate, which
are programs that allow for consumption within
the household. - For VL program, 1 out of 4 leaks are poor but out
of the age range - For the CHOFP aggregate, almost 1 out of 2
- Result is consistent with findings in Alcazar
et.al. (2003) using PETS, for VL - The school restriction is not significant for
neither the SB nor the SI program.
9Targeting errors and the poverty line
- Programs focus on the neediest and most
vulnerable, but - Poverty status is not unarbitrarily determined.
For the purpose of the study, we use expenditure
information and a poverty line, which is always
arbitrarily determined - More importantly, program officials do not use
household surveys to identify the socio-economic
level of the potential beneficiaries - limited to geographic targeting instruments
- even with individual targeting instruments, they
base their decision on some observable
characteristics to determine the poverty level
(proxy-means tests) - Consequently, it is important to check if
estimates of targeting performance change
significantly as we move the poverty line
10Comparing social programs across the income
distribution
- Figure compares the concentration curves (CCs)
for the four programs - A CC plots the proportion of beneficiaries to the
left of any point of the income distribution - The CC of the CHOFP aggregate clearly dominates
those of the rest of the programs, - it can be said that CHOFP has the largest
pro-poor bias, regardless of the level of the
poverty line. - No significant difference appears between the VL
and SB programs
11Incidence analysis at the margin
- The proportion of poor and non-poor that benefit
of a program at a given time may not be a good
indicator of how an expansion, or contraction,
would affect the poor. - Positive participation costs can imply early or
late capture of a program by the non-poor, if
they differ for the poor and non-poor, and change
with the scale of the program. - The higher costs related to reaching more remote
areas is often raised as the typical argument in
favor of early capture. - Late capture could result from the fact that
small pilot projects are more carefully monitored
and face less political pressures.
12Incidence analysis at the margin II
- We can follow Younger (2002) and use the
heterogeneity across regions and over time to
estimate the impact of a program expansion or
contraction for the poor. The idea is to estimate
the following equation - The dependent variable is the program
participation rate for quantile q in a given
domain in a particular year. The explanatory
variable is the program participation rate for
the specific year. - can be interpreted as the marginal effect of
an increase in program participation at the
national level on the participation rate in a
particular quantile - would indicate that a general
expansion (contraction) in coverage will cause a
more than proportional increase (reduction) in
participation for that quantile
13Marginal Incidence Analysis (MIA) for the SB and
VL programs
School Breakfast 1997-2000
- We can estimate the so that they sum to
five, and plot a marginal CC for each program - Then, we can compare the marginal CCs with the
average ones - Disregarding the age restriction, both programs
appear to be more pro-poor at the margin than on
average - These results suggest that, even though both
programs show a mediocre targeting performance on
the average, they seem to have a more pro poor
behavior on the margin. - This means that cutting (expanding) them would
(benefit) the poorest more than proportionately
Vaso de Leche 1997-2000
14Summary of results and implications
- First, the age restriction is found to be very
important for VL and CHOFP programs, which are
the ones that allow for consumption within the
household, - The VL program stops being the one with the worst
targeting performance and the CHOFP aggregate
becomes by far the program with lowest leakage.
(34) - This result suggests that food programs that
allow for consumption of the ration in the
household are not able to impose their
preferences on the distribution of the transfer
across household members, regardless of the
nature of the ration. - Lack of consideration of these intra-household
arrangements reduce the size of the transfer per
capita and limit the possibility for them to have
a nutritional impact on the target population.
(consistent with Stifel and Alderman, 2003)
15Summary of results and implications II
- Second, I find that the SB and VL programs has a
very pro-poor behavior at the margin despite
having a very mediocre targeting performance on
average (early capture) - It suggests the need to be cautious about making
decisions based on the average targeting
performance of programs. They could show large
leakages on the average, but still a cut
(expansion) could damage (benefit) the poor more
than proportionately - One usual explanation is that programs start in
more urban areas that are easier and less costly
to reach
16Summary of results and implications III
- Nevertheless, it is important to look at delivery
mechanisms organized for these programs - In both cases, the central government assigns
resources by district based on poverty map. Then,
municipalities identify localities on a need
basis. Within localities, mothers clubs define
actual beneficiaries - In practice, any mothers club that gets to be
registered will never be retired as a
beneficiary. Some with registered households - Anecdotal evidence suggest the following new
entrants tend to be from very poor localities
but, as time passes, their economic situation
improves but they cannot be taken out of the
registry - It is important to notice that any of these
explanations take the priority away from
improving proxy-means test instruments. The
latter focus on the political economy behind
delivery mechanisms
17www.grade.org.pe