Transcript Development Targets and Costs

Development Targets and Costs
Based on paper by
Luc Christiaensen
Christopher Scott
Quentin Wodon
Outline
1. Current practice
2. The political economy of target setting
3. Approaches to target setting
Historical benchmarking; Macro-simulations;
Micro-simulations
4. Approaches to cost estimation
Average costs and efficiency
5. Concluding remarks
1. Are current targets realistic?
A target is a value which a specific indicator should
attain by a particular date
Current practice
Honduras: cost of reaching PRSP targets and IDG goals together with
anticipated wage increase for public employees is high vs HIPC debt relief
Guinea:
• Goal: 10 % annual agricultural growth rate by 2010;
• Only 5 % of the 3921 recorded 3-year moving averages of ag growth
in all countries worldwide over past 4 decades exceeded 10 % growth
rate
Cambodia:
• Goal: reduce child undernutrition by 15 % points from 1998 to 2000;
• requires annual GDP/cap growth rate of 31 % (2005 as end date)
2. Political economy of target setting
Targets provide incentives
a) Resource mobilization
 targets must be realistic
b) Resource allocation and consensus building
 broad societal support and iterative process required
 proliferation of targets erodes their effectiveness
Political economy of target setting, cont. 2
c) Performance evaluation: targets introduce accountability
• Targets must be attainable and carry broad societal
support
• Poor performance by implementing actors must be
distinguishable from the effects of external shocks
• Failure to meet targets must entail consequences for the
actors
While targets have in principle positive incentive
effects, these do not follow automatically
Political economy of target setting, cont. 3
Experience from Britain
When the Government set local authorities a target for
collecting recyclable waste, it seemed a good idea. Even
better, the local authorities persuaded residents to take the
trouble to separate the stuff that was worth recycling from all
the rest--and met their target. There was only one snag. The
target was for collecting recyclable waste, not for recycling it.
As a result, some local authorities put the rubbish that had
been so carefully separated back in with the rest of their
garbage, and incinerated the lot.
Targets must be SMART: Specific, Measurable,
Achievable, Relevant and Time-bound
Selected choices in target setting
Targets for inputs, outputs, or for outcomes and impacts?
Focus on outcome and impact targets, though input and output
targets possible for short run
Check for vertical and horizontal consistency (Uganda)
Caution against proliferation of targets
Point targets or target ranges?
Given uncertainty surrounding input/output relations:
• target ranges for outcomes/impacts
• target points for inputs/outputs
Selected choices in target setting, cont. 2
Aggregate or disaggregate targets?
Equity versus efficiency
But: perverse incentives and proliferation of targets
Short-run or long-run targets?
Decision rule: marginal cost of, say, poverty reduction
equated across time periods
In practice: thought needs to be given to appropriate time
path for achieving the target, depending on state of country
3. Approaches to target setting
To facilitate resource allocation and to foster
accountability it is key that targets are realistic:
they must be technically and fiscally attainable
Three methods to gauge technical feasibility
Historical benchmarking
Macro-simulations
Micro-simulations
3.1 Historical benchmarking
Historical evidence and international comparison
Neither time nor skill intensive and data readily available hence,
this is a minimum requirement
Agricultural growth in Guinea and neighboring countries 1970-2000
3-yr moving avg
Mean 1987-2000
Frequency 1970-2000
Moving avg. >10 %
Moving avg < 0 %
Proj. growth 2010 (lin. trend)
1987-2000
1970-2000
Guinea
Cote d’Ivoire
Ghana
Mali
Senegal
4.2
3.2
3.0
4.0
1.3
0
0
0
6
0
6
3
6
2
8
7.3
-
2.8
2.6
7.8
3.3
0.4
4.8
4.8
1.4
3.2 Macro-simulations
Gauge feasibility of poverty and social development
targets by feasibility of implicit economic growth
requirements
Two methods to set targets for poverty
Growth required to reduce poverty by a certain
percentage, assuming no change in income distribution
Estimated net elasticity to growth
Liberia: growth & pov. simulations
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0
2007
2008
2009
1% annual growth of the GDP per capita
4% annual growth of the GDP per capita
2010
2011
2012
2% annual growth of the GDP per capita
5% annual growth of the GDP per capita
2013
2014
2015
3% annual growth of the GDP per capita
6% annual growth of the GDP per capita
Macro-simulations, continued
Estimated elasticity of social indicators to GDP/cap
growth controlling for time trends and urbanization ;
(elasticities can be estimated with world wide panel data (splines,
5 intervals) for GDP/cap and urbanization)
Exemples:
Demery and Walton (1999) use an elasticity of under 5 child
mortality to GDP/cap growth of –0.4 (SimSIP elasticity varies
between zero and –0.47)
Alderman et al. (2000) estimate an elasticity of prevalence of
child malnutrition to GDP/cap growth of - 0.09 (SimSIP
elasticity between zero and –1.1)
Macro-simulations, continued
Other intervening factors may be important  results
of growth based simulations indicative only
Some tools arguably more sophisticated though also
more data and skill intensive; yet literature now exists
Must be used with caution as results are not indicative
of policy actions needed
3.3 Micro-simulations
Country-specific nature of relationship between development
outcomes and their determinants lost in macro approach 
regression analysis and micro-simulations based on household data
Application to child malnutrition in Ethiopia:
15 years of 2.5 % growth of income/adult equivalent and bringing at least one
female adult/household up to level of primary school level reduce prevalence
of child stunting by 9 to 29 percent, far removed from international goal of
halving child stunting by 2015
If combined with nutrition education programs  reduction by up to 42 %
Increasingly possible, as more and more household data sets are
becoming available, yet technically difficult
4. The cost of reaching a target
Targets must be technically AND fiscally attainable
Effect of (public and private) expenditures on development
outcomes depends on:
amount spent on these objectives
effectiveness of money spent
Fiscal sustainability of targets can be gauged by
Government’s capacity for increasing public spending
Government’s scope for enhancing efficiency of its
spending
4.1 Costs and fiscal sustainability
Need for detailed sectoral information
SimSIP simulator for estimating costs of reaching
education targets
Education example: average costing method
Need for assumptions on:
Demographics: number of children of various age
groups joining education system
Costs and fiscal sustainability, cont.2
Data on
Cost parameters:
• Supply-side costs: teacher wage, teacher-student ratio, administrative
costs
• Demand side costs: stipend value, coverage
• Investment costs: cost per classroom, teacher training
Delivery system:
• length of schooling cycle, repetition, promotion, and drop out rates by
cycle or by grade, distribution of age at entry for each cycle
Targets for
• Changes in distribution of age at entry
• Changes in repetition, promotion and drop out rates
Costs and fiscal sustainability, cont.3
Logic of education costing
• Detailed cohort analysis depending on efficiency
parameters (promotion, repetition, drop out, …)
• Parametrize software to reproduce initial conditions
–
–
–
–
–
budget spending by cyle
teacher salaries, share of administrative budget
number of children enrolled by cycle
net & gross enrollment rates, with targets
realistic other parameters (cost of new schools/classrooms)
• Change efficiency parameters and demographics
over time, and track various costs as they evolve
– Number students  number teachers/classrooms  costs
Costs and fiscal sustainability, cont.3
Selected parameters for Liberia – primary level
(all parameters must be checked)
• Net enrollment (CWIQ 2007): 37.3% (low probably due to
conflict – older children enrolling in last few years)
• Gross enrollment (CWIQ 2007): 86.3%
• Budget: +/- US$ 4 million
• Number of teachers: 26,755  cost/teacher
• Number of students: 894,316  pupil/teacher ratio 33
• Share of public sector: 67% (incl. community)
• New entrants every year: +/-110,000, to increase in future
• Admin cost: 15% (assumption)
Costs and fiscal sustainability, cont.3
Selected parameters for Liberia – secondary level
(all parameters must be checked)
• Net enrollment (CWIQ 2007): 15.2% (low probably due to
conflict – older children enrolling in last few years)
• Gross enrollment (CWIQ 2007): 51.3%
• Budget: +/- US$ 1.5 million (need to check)
• Number of teachers: 8756  cost/teacher
• Number of students: 150,914  pupil/teacher ratio 17
• Share of public sector: 46% (incl. community)
• New entrants: depends on primary schooling parameters
• Admin costs: 15% (assumption)
Example of simulation (VERY prelim.)
Note: 2003 in simulator represents 2006 in reality; Assumes no
increases in real wage for teachers; Secondary spending expected to
increase faster; Possibility to estimate needed investment costs;
overall costs probably underestimated/need to get correct data
Recurrent Costs
8.00
7.00
6.00
5.00
4.00
3.00
2.00
1.00
0.00
2003
2006
2009
2012
Time (year)
2015
2018
4.2 Efficiency considerations
Outcomes can be increased through
Expansion of inputs keeping efficiency constant
Increase in efficiency keeping inputs constant
Given limited tax base, room for expansion of
social expenditures often narrow
Murray et al. (1994): typical country in Sub
Saharan Africa could improve health outcomes
by 40 % simply by reallocating resources to most
cost-effective intervention mix
Measuring efficiency
Best performance frontier
100
output
80
b
60
E=a/(a+b)
40
o
a
20
0
0
20
60
40
input
80
100
Measuring efficiency, examples
Efficiency of national health systems (WHO)
Output: disability adjusted life expectancy
Inputs: real total (private and public) health
expenditures/capita and average years of schooling
E>0.7 (good): Costa Rica, Sri Lanka, Bangladesh (good)
0.7<E<0.5 (mediocre): Gambia, Viet Nam
E<0.5 (poor): most African countries, e.g. E for Guinea
and Kenya respectively 0.47 and 0.32
Concluding remarks & discussion points
In principle, targets powerful tools to foster
accountability and facilitate consensus building;
Targets must be carefully designed and evaluated;
they must be SMART
Chapter focused on feasibility aspect, though also
need for more hard rules on other aspects of target
design
Remarks & discussion points, cont. 2
Set of readily applicable tools to assess technical
and fiscal feasibility of development targets
Each tool has its limitation  need for joint application
Tools must be used with care; they are indicative of
feasibility of targets, but not of policies needed
Tools only useful in practice if readily available
• Need for continuous updating of elasticities = public good
• Encourage policy simulations based on household data
• Research on efficiency of expenditures on social outcomes