Do Poverty Reduction Strategy Papers reduce poverty and improve well-being?

1
Do Poverty Reduction Strategy Papers reduce
poverty and improve well-being?

• Meg Elkins
• Simon Feeny
• David Prentice
• EADI Conference June 2014

2
Outline

• Background and motivation
• Research questions
• Data
• Methodology 3 stages
• Results
• Conclusion

3
Background and Motivation

• Poverty Reduction Strategy Papers (PRSPs)
  strategic frameworks for low-income countries to
  create economic and social policy to reduce
  multi-dimensional poverty
• PRSP guiding principles
• Country-driven and owned results orientated
  comprehensive in scope partnership orientated
  medium and long-term in focus.
• Collectively, Millennium Development Goals and
  PRSPs demonstrate the importance of tackling
  poverty in its many forms.
• Due to lack of consensus regarding the
  practicalities of how the MDG targets would be
  achieved MDGs were integrated into PRSPs in March
  2001.
• Opportunity to reflect on the relative successes
  of this generations policy tools to inform the
  next generation of MDGs.
• Provide feedback to the open policy space beyond
  2015.

4
Background

• Motivation
• The paper was motivated by a paper by Sumner
  (2006) questioning the next plausible
  contemporary paradigms to emerge in the PRSPs.
• This study extends the literature in the PRSP
  effectiveness in two ways
• 1) Adopts a multi-dimensional concept of
  well-being to include MDG indicators
• 2) Uses heterogeneous and homogeneous PRSP
  treatment effects measures how alignment to the
  development paradigms further impacts on MDG
  indicators

5
Research Questions
6
Data

• Panel of 118 countries from 1999-2008 (52
  developing countries undertaking PRSP and 62
  control countries) provide the period of
  treatment
• 7 MDG indicators (WDI) headcount poverty,
  primary school enrolment, ratio of girls to boys
  in the classroom, infant mortality, maternal
  mortality, HIV prevalence and access to
  sanitation
• Control Variables - World bank Governance, GDP
  per capita (2005), real GDP, Health expenditure
  (for the health indicators).

7
Three stage methodology
8
Stage 1 Paradigm Alignment Indices

• Scorecards to assess the degree of alignment to
  the development paradigms. For construction see
  Elkins, 2013 and Elkins and Feeny, 2014
• Washington Consensus Williamsons (1990)
• Post-Washington Rodriks interpretation (2005)
• New York Consensus Millennium Development
  Project (2005)
• Social Protection Agenda ADB and Baulch Wood
  (2008)
• Index values fall between 0-1

9
Development Paradigms Washington Consensus
Williamson (1990)

• Fiscal discipline
• Re-orientation of fiscal expenditures
• Tax reform
• Financial liberalisation/interest rate
  liberalisation
• Unified and competitive exchange rate
• Trade liberalisation
• Openness to foreign direct investment
• Privatisation
• Deregulation
• Secure property rights 

10
Post-Washington Consensus Rodrik (2006)

• Corporate governance
• Anti-corruption measures
• Flexible labour markets
• WTO agreements
• Financial codes and standards
• Prudent capital account opening
• Non-intermediate exchange rate
• Independent central banks
• Social safety nets
• Targeted poverty reduction

11
New York Consensus UNDPs Millennium Development
Project (2005)

• Infrastructure capacity capital expenditure
• Rural development- agricultural productivity and
  management
• Education provisions
• Health child and maternal mortality, control
  for diseases
• Governance rule of law and anti-corruption
  measures
• Employment public works, decent work programmes
  
• Water and sanitation infrastructure and
  management
• Gender equality and empowerment representation
  and land entitlement Environment biodiversity,
  urban dwellings, resource protection
• Science and Technology research and
  development, higher education

12
Social Protection Agenda
Cash transfers cash transfers, cash for work schemes Unemployment insurance Labour market legislation to protect labour rights Priority or pillar for social protection in the PRSP
Cash-in-kind transfers agricultural inputs, shelter, non-food items Unconditional unemployment payments Child labour protection labour code Micro-finance
Subsidies for housing, energy, and food Health/sickness insurance Minimum Wage  
Educational assistance Scholarships Non-contributory pension schemes employment promotion, matching people to jobs  
Fee waivers for essential services. Contributory pensions schemes    
Disaster relief programmes funds for emergency relief or post-emergency transitions. Disability pensions    
Targeted conditional cash-transfers for service delivery Maternity allowances    
Programmes for vulnerable groups the elderly, disabled widows and, orphans. Industrial injury payments    
Programmes for the internally displaced migrants and refugees Family payments    
13
Stage 2 Propensity Score Matching

• To construct an appropriate control group the
  study uses propensity score matching techniques
• Matches on the probability of PRSP treatment
  based on similar country characteristics
• Matched on the infant mortality large sample
  size and consistency across all MDG variables
• Matched on cross section data averaged between
  1996-1999 ie pre-treatment characteristics to
  determine the likelihood of treatment.
• The following variables related to infant
  mortality External debt to GNI, GDP per capita,
  Governance, health expenditure to GDP, and
  Ethnicity
• 52 PRSP treatment countries matched with 62
  control countries

14
Stage 3 Difference-in-difference estimation

• D-I-D used in combination with PSM is a
  relatively new programme evaluation technique
• Regressions estimations D-I-D controls for any
  pre-existing constant difference in the outcomes
• Countries adopt PRSPs in different years
  therefore the indicator is only switched on when
  treatment is in effect
• Use PRSP dummies to capture country and fixed
  effects prior and post policy changes.

15
Model Specification

• 1) Base regression specification
• Evaluates MDG progress from the PRSP treatment
• Uses PRSP treatment dummy to determine treatment
  effect
• 2) Alternate specification with alignment indices
  
• Yit ?0 ?1PRSPit ?2PRSPPAISit ?3Xit ?i
  ?t ?it (2)
• Uses interaction term between the treatment dummy
  and alignment scores

16
Average Treatment effect

• Average treatment effect estimates the potential
  unobserved outcome

17
Results
Head-Count Poverty Head-Count Poverty Primary School Enrolment Primary School Enrolment Ratio of Female to Male in PS Ratio of Female to Male in PS
PRSP treatment -3.947   2.498   1.870  
(-3.62) -3.62 -6.04
Interact WC 10.292 3.226 5.126
-1.92 -1 -3.41
Interact PWC -9.573 -17.945 -6.882
(-1.51) (-4.82) (-3.89)
Interact NYC -3.867 9.737 3.106
-3.867 9.737 3.106
Interact SPI (-0.88) -4.29 -2.91
(-2.10) (-2.56) (-4.70)
Average Treatment Effect -1.371 2.336 1.896
Observations 357 357 784 784 1,009 1,009
R-squared 0.468 0.489 0.277 0.322 0.284 0.311
Number of Countries 91 91 105 105 110 110
18
Results continued
Infant Mortality Infant Mortality Maternal Mortality Maternal Mortality HIV Prevalence HIV Prevalence Access to Sanitation Access to Sanitation
PRSP treatment -3.132   -5.805   -0.186   0.674  
(-8.143) (-0.35) (-3.69) -2.64
Interact WC 2.012 -180.321 0.179 3.157
-1.039 (-1.73) -0.75 -2.49
Interact PWC -1.163 288.031 0.381 -3.129
(-0.531) -4.12 -1.42 (-2.18)
Interact NYC -3.067 47.787 -0.453 -1.992
(-2.212) -0.99 (-2.55) (-2.19)
Interact SPI 1.457 -90.572 0.243 2.026
-1.082 71.832 -1.47 -2.33
Average Treatment Effect -2.0345 -1.43 -0.118 1.724
Observations 1,120 1,120 266 266 910 910 1,089 1,089
R-squared 0.639 0.645 0.307 0.391 0.111 0.125 0.43 0.441
Number of Countries 112 112 105 105 91 91 110 110
19
Discussion

• Results find that PRSP adopters did achieve
  statistically significant improvements in all
  categories but maternal mortality although data
  was weakest for this indicator
• Heterogeneous effects as estimated by average
  treatment effects reported marginally smaller
  results for headcount poverty, primary school
  enrolment, ratio of girls to boys in the
  classroom, infant mortality, maternal mortality
  and HIV prevalence. The ATE for sanitation was
  larger.
• New York Consensus found statistically
  significant improvements for primary school
  enrolment, ratio of girls to boys in the
  classroom, infant mortality, and HIV prevalence
• Alignment to the SPI was significant for
  headcount poverty and for access to sanitation.

20
Conclusion

• Evidence from this study suggest that PRSP
  recipients more effective at achieving MDG
  outcomes than the comparison group of countries
• Inclusion of the paradigm alignment indices tries
  to address the issue of causation by including
  policy choices made within each PRSP.
• Results are encouraging for the international
  community with PRSP treatment and alignment to
  the NYC achieve even higher results for all MDG
  indicators except maternal mortality and access
  to sanitation.
• These results evidence how the ambitious targets
  of the MDGs used in combination with the
  practicalities of the PRSR are able to deliver
  the intended objective of multi-dimensional
  poverty reduction.
• Augurs well for setting and embedding ambitious
  targets in the next generation of MDGs

21
Conclusion
MDG Indicator PRSP treatment Without paradigm index PRSP with paradigm index WC PWC NYC SPI
GOAL 1 Head Count Poverty 1.25 -     - -
GOAL 2 PS enrolment   - -
GOAL 3 Ratio of male to female in PS - -
GOAL 4 Infant Mortality - -   -
GOAL 5 Maternal mortality   -   -
GOAL 6 HIV/Aids - -     -  
GOAL 7 Improved access to sanitation   - -
22
Thank you!

• Meg Elkins
• RMIT University Melbourne Australia
• Meg.elkins_at_rmit.edu.au

23
MDG progress indicator and PRSP treatment OLS

MDG progress and PRSP treatment
  Model 1
VARIABLES
   
PRSP treatment

MDG adjusted progress indicator -0.022
(-1.51)
Ethnic 0.172
(1.92)
Average governance -0.009
(-1.07)
Average GDP per capita -0.000
(-2.34)
Average health expenditure 0.011
(0.91)
Average WC 0.571
(2.54)
Average PWC -0.203
(-0.83)
Average NYC 0.440
(2.44)
Average SPI 0.332
(2.02)
Constant 0.200
(1.33)

Observations 115
R-squared 0.796
F 45.49
t-statistics in parentheses
plt0.01, plt0.05, plt0.1
24
Treated Countries Control Countries

Albania Lesotho Algeria Algeria Malaysia
Armenia Liberia Angola Angola Mauritius
Azerbaijan Madagascar Argentina Argentina Mexico
Bangladesh Malawi Belarus Belarus Morocco
Benin Maldives Belize Belize Panama
Bhutan Mali Botswana Botswana Papua New Guinea
Bolivia Mauritania Brazil Brazil Paraguay
Bosnia and Herzegovina Moldova Bulgaria Bulgaria Peru
Burkina Faso Mongolia Chile Chile Philippines
Burundi Mozambique China China Russian Federation
Cambodia Nepal Colombia Colombia Samoa
Cameroon Nicaragua Comoros Comoros Seychelles
Cape Verde Niger Costa Rica Costa Rica Solomon Islands
Central African Republic Nigeria Dominican Republic Dominican Republic South Africa
Chad Pakistan Ecuador Ecuador St. Kitts and Nevis
Congo, Rep. Rwanda Egypt, Arab Rep. Egypt, Arab Rep. St. Lucia
Cote d'Ivoire Senegal El Salvador El Salvador St. Vincent and the Grenadines
Djibouti Serbia- Montenegro Eritrea Eritrea Sudan
Dominica Sierra Leone Fiji Fiji Swaziland
Ethiopia Sri Lanka Gabon Gabon Syrian Arab Republic
Gambia, The Tajikistan Grenada Grenada Thailand
Georgia Tanzania Guatemala Guatemala Togo
Ghana Uganda India India Tonga
Guinea Uzbekistan Indonesia Indonesia Tunisia
Guinea-Bissau Vietnam Iran, Islamic Rep. Iran, Islamic Rep. Turkey
Guyana Yemen, Rep. Jamaica Jamaica Turkmenistan
Honduras Zambia Jordan Jordan Ukraine
Kenya Kazakhstan Kazakhstan Uruguay
Kyrgyz Republic Latvia Latvia Vanuatu
Lao PDR Lebanon Lebanon Venezuela,
Lithuania Lithuania Zimbabwe
Macedonia, FYR
54 treated countries   62 control countries