Poverty Evaluation of Economic Policies in Developing Countries

1
Poverty Evaluation of Economic Policies in
Developing Countries

• François Bourguignon
• Mallorca, July 2003

2
Introduction
3
Motivation of lectures

• Poverty reduction as the goal of economic
  development
• The growth-distribution-poverty triangle
• Need to evaluate the growth and distributional
  impact of policy instruments
• Efficiency vs. Equity, or Efficiency with Equity
• Need for adapting existing techniques to
  developing countries and for designing new
  instruments
• The WB 'toolkit'

4
Types of policies to be evaluated

• Micro policies
• Public expenditures, e.g., shifting the
  allocation of public spending to specific public
  programs affecting Education, Health, cash and in
  kind transfer policies micro finance,
  infrastructure.
• Tax policy changing brackets or rates of direct
  taxes (subsidies), rationalizing indirect tax
  systems, introducing new taxes
• Liberalization of specific markets labor
  market basic commodity markets
• Land reform
• Pension system, .
• Structural Reforms
• Trade liberalization
• Financial sector reforms
• Public expenditure management, e.g., delivery of
  services, quality, targeting of services
• Privatization of public utilities,
  infrastructure, other firms
• Decentralization and changes in
  inter-governmental institutional relations
• Macro Policies ( as a response to exogenous
  shocks)

5
Various approaches to evaluation

• "Micro Incidence analysis" as the common
  denominator of evaluation techniques
• Micro perspective
• Ex-ante versus Ex-post Analysis
• Accounting versus Behavioral Approaches
• Average versus Marginal Effects
• Qualitative versus Quantitative Approaches
• Macro perspective
• Partial versus General Equilibrium
• Static vs. Dynamic views

6
Outline of lectures

• 1. Ex-ante evaluation of micro-oriented policies
• a) Welfare incidence analysis accounting vs.
  behavioral approaches
• b) Micro-accounting techniques applied to
  taxation
• c) Micro-accounting techniques applied to
  public spending
• d) Accounting for behavioral response with
  micro-simulation techniques

7
Outline of lectures

• 2. Evaluation of macro-policies the micro-macro
  linkage
• a) Simple ways for evaluating the poverty
  reduction effect of growth
• b) Marrying micro-accounting techniques and
  macroeconomic modeling
• c) Marrying micro-simulation techniques and
  macroeconomic modeling sequential approach
• d) Marrying micro-simulation techniques and
  macroeconomic modeling towards integrated models

8
1. Ex-ante evaluation of micro-oriented policies

• 'Micro-oriented' policies policies that affect
  directly the welfare of households or individuals
  through the prices they face, their disposable
  income, their production/consumption sets (public
  goods)

9
a) Welfare incidence analysis accounting vs.
behavioral approaches

• The basic consumer model
• V(p,y) Max U(x) p.x ? y UxM(p,y)
• E(p, u) Min p.x U(x) ?? u p.xH(p, u)
• Welfare incidence of a change in "income" ?V
  Vy.?y
• But Vy may be arbitrarily chosen to be 1. Thus
• (Before considering distributional aspects, and
  for fixed p).

?
?
welfare

income
10
a) Welfare incidence analysis accounting vs.
behavioral approaches

• Welfare incidence of change in taxes
• Tax reform p0 ? p1
• Money metric of change in welfare compensating
  income variation CV E(p1, u0) - E(p0, u0)
• First order approximation of (-) welfare
  incidence
• CV ? change in the cost of initial consumption
  basket

11
a) Welfare incidence analysis accounting vs.
behavioral approaches

• Welfare incidence of change in taxes (ct'd)
• Other argument
• But budget constraint p.?x ?p.x 0
• Hence ?V ? - ?p.x
• Knowledge of consumption behavior, xH(p,y), is
  unnecessary
• First order welfare incidence analysis
  expenditure accounting

12
a) Welfare incidence analysis accounting vs.
behavioral approaches

• Welfare incidence of change in taxes (ct'd)
• Behavioral or micro-simulation approach use of
  xH(p,y).
• Necessary for
• - non-marginal changes in prices
• - situations of market imperfections (quantity
  constraints)
• - evaluating government budget constraint ( even
  with marginal changes)
• Micro-accounting approach observation of x0
• Sufficient in other situations
• (Note that "price changes" include wages or
  output prices of self-employed. Thus accounting
  approach also includes the 'income' side. )

13
b) Application of micro-accounting to the
distributional effects of taxation

• Preceding technique applied to all individual
  households in some micro-data base typically a
  household survey.
• Direct taxes (and cash benefits) computed
  according to tax-benefit rules for each
  household/individual in the data base. Reforms
  simulated by changing the rules under the
  assumption of no tax evasion and full take-up of
  benefits. It is then possible to identify the
  gainers and losers as well as all distributional
  changes.
• This is the essence of the 'tax-benefit models'
  in developed countries.

14
b) Application of micro-accounting to the
distributional effects of taxation (ct'd)

• Indirect tax reforms must first be mapped into
  changes in consumer prices. This generally
  requires some kind of General Equilibrium Model
  as well as assumptions on which production units
  actually pay the tax. Then micro-accounting
  applies to changes in consumer prices and
  expenditures observed in household survey.
• Typically, distributional results are presented
  under the form of 'concentration curves', which
  permit simple dominance analysis

15
Concentration curves
Share of total tax paid by x poorest households
1
Lorenz curve (initial income)
Tax 2
45 line
Tax 1
x
1
Population ranked by welfare level e.g. income
per capita.
Tax 1 dominates tax 2. It is progressive whereas
tax 2 is regressive.
16
b) Application of micro-accounting to the
distributional effects of taxation (ct'd)

• Concentration curves show how a of tax j is
  distributed in the population. To get the full
  effect of taxation on income, it is necessary to
  mix information on concentration curves and on
  tax pressure tax receipt/total household
  expenditures.
• Assuming no change in welfare ranking
• where ??j tax rate on good j average budget
  share of good j
• Lq Lorenz curve of initial (0) and new (1)
  distribution
• Cja Concentration curve of tax j
• From that formula easy to derive change in
  inequality, poverty and social welfare measures.

17
Example Madagascar ( Younger et al. )
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
b) Application of micro-accounting to the
distributional effects of taxation (end)

• Possible use of preceding results effect of
  reallocating taxes across goods and services on
  distribution and poverty (e.g. from kerosene to
  wages).
• Heuristic treatment of the optimal indirect
  taxation problem (Diamond-Mirrlees)
• Caution consumption behavior necessary here
  because of the government budget constraint.
  Taking away a of tax on good i might have to be
  compensated by more than a of tax on good j
  at initial consumption levels.
• But concentration curves indicative of
  distributional effects of tax reforms ( provided
  the implicit general equilibrium model behind
  changes of consumer prices is satisfactory).

23
c) Application of micro-accounting to the
distributional effects of public spending

• Standard practice
• Directly inspired by tax incidence analysis.
• A quantity zij of public good j consumed by
  individual i at no cost. If individual i were
  charged the price cj, his/her welfare loss would
  be given by zij. cj .
• Hence the welfare gain of having public good j
  free is given by
• ?V ? cj zij
• Estimate for cj unit budgetary cost of public
  service j
• Example j primary public school, zij number
  of kids in primary public schools in household i,
  ci total budget for primary school/enrollment
  in public primary schools.

24
c) Application of micro-accounting to the
distributional effects of public spending

• Examples (Demery)

25
Concentration curves (Indonesia, 1989)
26
(No Transcript)
27
c) Application of micro-accounting to the
distributional effects of public spending (ct'd)

• Problems with public spending incidence
• Assumption of 'marginal changes' is dubious
  behavioral response should be taken into account
• Taking unit cost as the 'price' of public goods
  is a strong assumption
• - the 'quality' issue
• - welfare change "willingness to pay", which
  requires collecting specific information or
  estimating behavior

28
Problems with public spending incidence (ct'd)

• Average versus marginal incidence
• Changing public spending in education or health
  do not affect all users in the same way except
  maybe if the quality of the service uniformly
  improved.
• In general, change in public spending comes with
  a change in the identity of users e.g. the
  phenomenon of 'capture' of a public good by the
  elite

29
Problems with public spending incidence (ct'd)

• Dealing with marginal incidence
• - observe actual changes in coverage and its
  beneficiaries
• change in of a given quantile being
  covered
• Share of a given quantile among new users
• - use behavioral model to predict new users
• - use geographical regularity to estimate
  capture phenomena (Lanjouw and Ravallion).
• Number of users in quantile q of region i
  F(public spending per capita in region i)
• - The geographical dimension of incidence

30
Problems with public spending incidence (ct'd)

• The dynamic perspective
• Public spending in education, nutrition, health
  may often be considered as an investment in the
  human capital of future generations.
• Distributional effect should be evaluated by
  projecting future distribution of welfare and
  poverty with and without changes in today's
  public spending.
• Considering compensating income variation of
  present cohorts would make sense only in a pure
  dynastic model with perfect capital markets.
• General equilibrium effects
• An increase in the educational level of the
  population may modify future relative earnings,
  and therefore the impact of public spending on
  poverty.

31
Problems with public spending incidence (ct'd)

• Other types of public good
• Many public goods call for evaluation techniques
  very different from education or health
  infrastructure, micro-finance, police,

32
Conclusion on micro-accounting

• Useful first approach to evaluation first-order
  approximation or first round-effects
• Rarely sufficient e.g. evaluating the budget
  constraint
• Interpretation often ambiguous
• Alternative approaches may be difficult to
  implement
• Might be useful to combine micro-accounting with
  elementary representation of behavior

33
d) Accounting for behavioral response with
micro-simulation techniques

• This approach less frequently used in developing
  countries than in developed countries because of
  the limited extent of direct taxes and cash
  benefits and their effects on behavior.
• Yet, it was seen before how important it is to
  model satisfactorily behavior for policy
  evaluation.
• What follows a) General principles
• b) An example of application

34
(i) General principles Labor Supply or
Occupational Choice and tax-benefit systems (FBFF)

• 3 steps procedure
• Think of the simplest model that contains enough
  structure to capture the mechanisms that are
  likely to affect the agent responses
  labor-supply, occupational choice, - to the
  policy under consideration.
• Estimate the model
• Simulate the effects of policy reform

35
Tax-benefit labor supply modeling as the
archetypical policy problem

• Structure of the problem
• Max u(c, L z ?, ? ) subject to c ? y0 wL
  NT(wL, L, y0 z ?), L?0
• L F (w, y0 z ?, ? ?) , c C(w, y0 z ?,
  ? ?) and welfare metric
• Problem effect of changes in the tax system (??
  ?) on L, C and welfare metric, conditional on (w,
  y0 z ?, ?)
• (First approximation and micro-accounting
  framework a metric for welfare change is
• ??NT(wL0, L0, y0 z ?) NT(wL0, L0, y0 z ?
  ??) -? NT(wL0, L0, y0 z ?))

36
Tax-benefit labor supply modeling as the
archetypical policy problem

• Estimate the model
• Choose a functional form for u(c, L z ?, ? )
  and estimate parameters ? and ?.
• This may be done directly or through the implied
  closed form solution for labor supply
• Li F(zi, wi, y0i ?, ?i ?)
• Where F( ) includes the alternative Li0.
• By definition of estimation procedure

37
Tax-benefit labor supply modeling as the
archetypical policy problem

• Simulate the model (partial equilibrium)
• Policy reform Change ? for ?s
• Applying model to each individual/household i
• Change in welfare metric
• Difference with first order approximation is in
  the w.(Ls L0) term

.
38
Tax-benefit labor supply the discrete choice
specification (van Soest, Hoynes, Blundell et al.
, )

• Structure of the model
• Finite discrete choice of working times, Dj
• subject to

.
39
Tax-benefit labor supply the discrete choice
specification (van Soest, Hoynes, Blundell et al.
, )

• Estimation of the model
• If ?ij, additive, iid and distributed like
  extreme values, this specification leads to
  familiar multinomial logit model (McFadden)
• possibility of introducing heterogeneity in
  preferences ??jm with probability Pm

.
40
Tax-benefit labor supply the discrete choice
specification (van Soest, Hoynes, Blundell et al.
, )

• Advantage of the discrete specification
• - Discrete specification similar to continuous
  one if
• - Discrete specification avoids having to work
  with closed form solutions
• - Discrete form introduces maximum flexibility
  subject to some minimal constraints (
  )
• - Possibility of introducing heterogeneity in
  preferences ??jm with probability Pm
• - Direct reading of welfare change

.
41
Example of application. Ex-ante Evaluation of a
Conditional Cash Transfer Programme Bolsa Escola

• Conditional cash transfer program means tested
  benefits paid conditionally on requirements like
  sending kids to school, taking medical
  examinations,
• Examples Progresa, Bolsa Escola, .
• (Taken from Bourguignon, Ferreira, Leite)

42
The Bolsa Escola Programme

• Means-test income per capita less than R90
  (50 per cent of the minimum wage)
• Conditionality 6-15 year-olds must attend
  school.
• Transfer R15 per child in school
• Limit R45 per household
• Monitoring at the local and federal levels

43
The Model

• Childs occupational choice
• Choose the alternative with maximum utility
  among
• (0) Not going to school (paid or unpaid work)
• (1) Going to school and paid work
• (2) Going to school and no paid work
• Ui(0) Zi.?0 ?0.(Yi yi0) vi0
• Ui(1) Zi.?1 ?1.(Yi yi1) vi1
• Ui(2) Zi.?2 ?2.(Yi yi2) vi2

44
The Model (ctd.)

• Child is Contribution to Household Income in
  state j 0, 1 or 2
• yi0 wi yi1 Mwi yi2 Dwi
• with wi ("full time") market earnings
• Log yi0 Xi .? ui
• Log yi1 Xi .? mInd(Si1) ui
• and M Exp(m)

45
The Model (end)

• Household (kid) i chooses the alternative j that
  yields the highest utility Ui(j)
• Ui(0) Zi.?0 ?0.Y-i ?0.wi vi0
• Ui(1) Zi.?1 ?1.Y-i ?1.wi vi1
• Ui(2) Zi.?2 ?2.Y-i ?2.wi vi2
• with ?0 ?0 ?1 ?1M ?2 ?2 D

46
The simulation framework

• Introducing the State-ConditionalTransfer
• Easily evaluated when ?j, ?j, ?j, vij are known
  direct welfare evaluation

47
Descriptive Statistics and Estimation Results

48

49

50

51
V. Simulation Results.

52
Simulation of Alternative Programme Designs

• Transfer doubled from R15 to R30.
• Age Progressive Transfer (R10 to R35).
• Means-Test raised to R120.
• Transfer doubled Means-Test raised.
• Age Progressive Transfer Means-Test raised.
• No conditionality.

53
(No Transcript)
54
(No Transcript)
55
Conclusions

• Bolsa Escolas conditionality is somewhat
  effective likely to send to school one third of
  10-15 year-olds currently not attending (and up
  to one-half among the poor).
• Likely to INCREASE number of children studying
  and working. (School duration?)
• The programmes impact on overall poverty and
  inequality is modest, due largely to low transfer
  amounts.

56
Conclusions (ctd.)

• 4. Impact on poverty appears to be elastic w.r.t.
  transfer amount, but not to level of means-test.
• 5. Simple micro-econometric estimation and
  simulation procedures can be usefully deployed to
  investigate likely impact of alternative
  programme designs.

57
General conclusion of micro-part

• Micro-accounting techniques well mastered and
  widely used.
• Not fully satisfactory (marginal approximation
  not valid, need for a good knowledge of budget
  constraint)
• Micro-simulation work based on behavioral
  modelling necessary, but not yet very developed
• Need for minimal set of assumptions and maximum
  simplicity for tractability and policy
  transparency
• Ex-ante simulation based on assumption that
  cross-sectional differences would be observed in
  a temporal dimension too.
• Ways of checking quality of ex-ante
  micro-simulation approach mixing ex-ante and
  ex-post analysis. But only exceptionally possible