targeting of public spending menno pradhan senior poverty economist the world bank office, jakarta
TRANSCRIPT
Targeting of Public Spending
Menno PradhanSenior Poverty Economist
The World Bank office, Jakarta
Central Question
Who benefits from public spending?
and
How to improve the targeting of public spending?
Benefit Incidence Analysis
0
10
20
30
40
50
60
70
1(poor) 2 3 4 5(rich)
Consumption Quintile
Perc
enta
ge W
ho B
enefi
ts
Data Source
Derived from household survey data
Did you use facility?
Did you receive benefit?
and,
Welfare measure, usually per capita consumption
Does this answer the question?
No, because:
• Benefit incidence ignores administrative cost- Total program cost = benefit + admin cost
• Benefit incidence is based on ex-post welfare indicator
– should based on welfare before program
• Benefits may not be homogeneously distributed
– for example, quality of school may vary across location
Solutions
• Need more detailed information on service delivery and expenditure tracking (public expenditure tracking surveys)
- to identify leakage/administrative cost - to identify heterogeneity in benefits
Solutions (cont.)
• Needs to consider what welfare level would have been without program
- Simplest case for cash transferConsumptionbefore = consumptionafter – cash
transfer
- Too simple households are likely to have behaved differently if they had not received the benefit
- Impact evaluation
Marginal benefit incidence
• Policy question is often too scale up or down an existing program
• Standard benefit incidence compares situation with or without program
• Marginal benefit incidence ≠ average benefit incidence
Marginal Benefit Incidence
Marginal Spending is Pro-Poor while average spending is pro-rich
0
10
20
30
40
50
60
70
1(poor) 2 3 4 5(rich)
Consumption Quintile
Perc
enta
ge W
ho B
enefi
ts
Year 1 Year 2
How to predict marginal benefit incidence?
Exploit differences in coverage across regions
jqj incidenceAverageIncidence 10
tincidenceAverageIncidence jtjqjt 1
j = regionq = quintile t = time
Conclusion Benefit Incidence Analysis
• Benefit incidence analysis is useful and cost effective to assess targeting of programs/policies
• Does often not provide complete picture
- ignores leakage / efficiency
- does not provide impact of program
• Tools and techniques are available to address those issues
Targeting of Government Expenditures
How to ensure that the poor benefit ?
Outline
1. Targeting of expenditures – introduction
2. Decentralization and Geographic targeting of programs
3. Potential gains of regional targeting
4. Poverty maps
5. Concluding remarks
Targeting of Public Expenditures
Ways to reach the poor. Target using :
• Household income (means testing)
• Other household indicators of welfare
• Self-targeting
• Commodities (rice, basic education and health care)
• Empowerment of the poor
Geographic Targeting
• Of relevance during Decentralization- Central government allocates budgets to local
government - Local governments make local spending allocations
• Geographic targeting: Poorer regions get a larger share of the pie
Why do geographic targeting ?
• Stark regional differences in poverty - Indonesia
Head count ratio %(source: Susenas 99)
Poorest province 61.1
Median province 23.8
Richest province 2.8
Geographic targeting, cont.
• Pros– Leads to better
targeting of the poor– Cheap and simple– Promotes local
ownership– Local administrators
better informed about local conditions
• Cons– Less control of center
over final outcome – local capture
– Political opposition to regional allocations based on statistical estimates
– May be less fair. • The probability for a
poor to receive a benefit depends on the region that person lives in
– Incentive to move to poor region
Geographic targeting, Example
• Example: JPS scholarship program
JPS scholarships by province
0
0.1
0.2
0.3
0.4
0.5
0.6
0 20 40
percent poorsc
hola
rship
s / e
nro
llmen
t
Implementation unit Scholarships
Allocation Per district
Allocation Per school
Allocation Per sub-district
Allocation Per school
Student selection
CPCU Jakarta
District committee
Sub -district committee
School committee
Operational issues in regional targeting
• At what level to do the regional targeting?– The smaller the area, the better the targeting– And, the higher the statistical error
• How far to go in regional targeting?– Exclude rich districts? In proportion to nr of
poor?– More extreme first stage targeting may yield to
worse performance, depending on the second stage targeting
Does more precise targeting lead to better targeting to the Poor ?
Hypothetical exampleA country with• 27% poor. • Program has
potential to reach 27% of population
• Second stage allocation is done randomly
0 10 20 30 40 50 60
Village
Sub-District
District
Province
Randomallocation
Percent poor reached
First Stage Allocation
The smaller the area – the higher the statistical error
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
natio
nal
prov
ince
district
sub-dis
trict
villag
e
Pove
rty
/ con
fiden
ce in
terv
al
poverty
lower boundModule
upper boundModule
lower boundCore
upper boundCore
How far to go in regional targeting ?
• Example: 250 poor, 250 scholarships
0204060
80100120140160180
200
1 2 3 4 5
region
Poor Non-poor
Percent of poor reached (100 percent is perfect targeting)
Second stage
Random Perfect
First sta
ge
In proportion to population
31% 88%
In proportion to poor 47% 100%
Extreme 59% 75%
Poverty Maps
• Provide estimates of welfare for small geographical area.
• Could be along any dimensions of poverty
• To improve regional targeting
• To increase understanding of regional dimension of poverty (causes and consequences)
Poverty maps: Project, cont..
Basic Procedure:
– Estimate a model of (log) per capita consumption, yh, using sample survey data.
– Restrict explanatory variables to those that can be linked to households in survey and census.
– Simulate expected level of poverty or inequality using their census-based characteristics for each population of interest, and the estimates from the “first-stage” model of y.
– Model thus provides empirical weighting scheme.
Concluding Remarks - Targeting
• Regional targeting be closely linked to decentralization
• Regional targeting is a convenient planning tool to improve targeting of public expenditures programs, but no panacea.
• There is potential for improvement of first stage targeting of programs, but much more in the second stage. So, incentive mechanism for pro-poor spending at the local level is more important than the regional distribution component.
• Poverty maps can improve first stage targeting by restricting expenditures by small geographic area
Thank you