Targeting and Public Expenditure

1
Targeting and Public Expenditure

• Margaret Grosh

2
Themes

• General Issues
• Goals
• Measurement
• Stylized facts
• Applications to social safety nets
• Comparison of instruments

3
Targeting

• Goal -- to concentrate benefits among the
  neediest
• Implication
• some people benefit and others do not AND/OR
• needier get bigger benefit than less needy

4
Benefits of targeting

• Assumptions
• 15 million population
• 3 million poor
• 150 million budget
• No targeting
• everyone gets 10
• 80 of funds go to the non-poor

5
Benefits of targeting

• Assumptions
• 15 million population
• 3 million poor
• 150 million budget
• No targeting
• everyone gets 10
• 80 of funds go to the non-poor

• Targeting - Option I
• only poor receive 50
• same budget

6
Benefits of targeting

• Targeting - Option I
• only poor receive 50
• same budget
• Targeting Option II
• only poor receive 10
• budget reduced to 30 million

• Assumptions
• 15 million population
• 3 million poor
• 150 million budget
• No targeting
• everyone gets 10
• 80 of funds go to the non-poor

7
Stepping back

• What is the role of broad-based vs targeted
  programs in poverty reduction?
• Where is the distributional instrument placed?
• How private is the good?
• Is goal (only) poverty reduction?
• What is the concept of poverty
• utility, income, capabilities?

8
Measurement (the usual morass of detail)

• The counterfactual pre-intervention welfare
• Usual measurement problems
• Recording and valuing consumption
• Comparing across time and space
• Equivalence scales
• Behavioral change in response to provision
• Labor supply
• Consumption of goods/services
• Private transfers

9
Measurement (the usual morass of detail)

• The value of the benefit
• Cost is not value (vaccines)
• Costs hard to measure (data problem)
• Values not same across hh (schools)
• Quality differences (data problem)

10
Conventional measures

• Errors of inclusion/exclusion
• Simple
• Discrete
• Weighting issue

11
TARGETING ERRORS AND ACCURACY
INCORRECTLY GIVEN BENEFITS
ACTUAL STATUS
POOR
NON-POOR
GOOD TARGETING
Error of Inclusion Type II
POOR
CLASSIFIED AS
CORRECTLY DENIED BENEFITS
Error of Exclusion Type I
NON -POOR
INCORRECTLY DENIED BENEFITS
12
Conventional measures

• Errors of inclusion/exclusion
• Simple
• Discrete
• Weighting issue
• Full distributional analysis of incidence and
  coverage / concentration coefficients and curves
• Extended Ginis (Clert and Wodon, 2000)
• Average vs marginal incidence

13
Stylized facts

• Health, education as whole sectors usually mildly
  progressive
• Progressive as of welfare
• Less so absolutely
• Primary gt secondary gt tertiary
• Demographics of measure
• Pyramid effect
• Self-selection into private market
• Food price subsidies absolutely regressive,
  relatively progressive
• Transfers gt health, education

14
Share of Benefits Accruing to the Poorest 40
Percent, by Country and Sector
15
Applications to social safety nets

• What are reasonable expectations?
• What do we know about options?

16
Targeting is a tool, not goal(I.e. must balance
tradeoffs)

• Benefits
• lower costs
• greater impact
• Errors of exclusion (undercoverage)

• Costs
• administrative
• political economy
• incentive
• Errors of inclusion (leakage)

17
Administrative costs

• Targeting costs only a portion of total
  administrative costs
• Usually more exact targeting imposes higher
  administrative costs
• Just because costs exist doesnt mean they arent
  worth paying

18
Incentive Effects

• OECD literature worries about work disincentives
  from means tests, measures them
• May be less important in some of our programs
  because
• not based on means test
• eligibility
• benefit level
• incentive more to conceal income than reduce it
• low level benefit, so incentives remain

19
Political Economy

• Can affect
• support and budget for safety net
• mix of programs
• details of each
• Reasons to support program
• own present benefits
• future benefits
• benefits for others you care about
• altruism, externalities
• suppliers
• Coalitions

20
Quantifying the Tradeoff

• Study of 30 Latin American programs, late 1980s
  early 1990s (not contradicted to date)
• Tried to measure
• errors of inclusion
• errors of exclusion
• administrative costs
• total
• of targeting
• qualitative information on requirements, options

21
(No Transcript)
22
Share of Benefits Accruing to the Poorest 40
Percent, by Sector
23
INDIVIDUAL ASSESSMENT (15)
TARGETING
24
GROUP CHARACTERISTICS (9)
TARGET GROUP
25
SELF-TARGETING (6)
26
Share of Benefits Accruing to Poorest 40 Percent,
by Targeting Mechanism
27
Errors of exclusion

• Lacked data on participation rates
• Unclear interpretation
• self-targeting (good)
• errors of exclusion (bad)
• budget, outreach, communications, logistics, etc.
  appear more important than mis-identification due
  to screening

28
Total Administrative Costs as a Share of Total
Costs, by Targeting Mechanism
29
Targeting Costs as a Share of Total Costs, by
Targeting Mechanism
30
Figure 9 Targeting Cost Share and Benefits
Accruing to Poorest 40 Percent
100
80
60
40
20
0
1
2
3
4
Share of Targeting Costs ()
31
Conclusions

• progressivity of incidence
• administrative costs not prohibitive
• no a priori ranking by mechanism

32
Self-Targeting

• Good or service available to all, but only the
  poor choose to use
• Examples
• hard physical labor for low wages
• broken rice, coarse bread, etc.
• waiting times
• stigma
• May be difficult to find vehicle suitable for
  large transfers
• Costs to beneficiaries reduce net benefits

33
Categorical targeting

• Age (child allowances, non-contributory pensions)
• Disability, unemployment
• Ethnicity (scheduled castes in India, Natives in
  Canada)
• Easy to medium administratively
• May not be very precise

34
Geographic

• More accurate the smaller the unit used
• But a limit based on data, service delivery
  system, politics
• More viable for services used daily than yearly
• New tool merging census and survey data may make
  more accurate

35
Proxy means test

• Increasingly popular
• A synthetic score calculated based on easily
  observed characteristics (household structure,
  location and quality of housing, ownership of
  durable goods)
• At the complex end of requirements
• Indicators tend to be static

36
Community-Based Targeting

• Use existing local actor (teacher, nurse,
  clergyman) or new civic committee to decide who
  gets what
• local actor may have best information, but
• structure may impinge on actors performance in
  their original local roles,
• may generate conflict
• capture by local elites still possible
• little empirical evidence to date