case study 1: what, why and how of impact...

28
Case study 1: What, why and how of impact evaluations Shubhra Mittal Senior Policy and Training Manager, J-PAL South Asia Kathmandu, March 28, 2017

Upload: others

Post on 01-May-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Case study 1: What, why and how of impact evaluations

Shubhra Mittal

Senior Policy and Training Manager, J-PAL South Asia

Kathmandu, March 28, 2017

Page 2: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

I. Introduction to J-PAL

II. Why Evaluate: Case study of pricing of preventive health products

III. Case Study 2: understanding impact of a programme

Page 3: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

2

About J-PAL Who we are, what we do

Page 4: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Governments face multiple problems across various sectors… A Typical Social Policy Goal – Improving Learning Outcomes

3

To address poor learning levels, there are many potential solutions:

Free

uniforms

Libraries Cash

grants School

equipment

Information

Campaigns Technology Pedagogy

changes

The fundamental dilemma - How do we know which solution creates the most impact? How do we know which solution is the most cost-effective?

Impact evaluations identify the causal impact of a programme (e.g. a pedagogy

intervention) on an outcome of interest (e.g. learning outcomes) by comparing what happened with the programme with what would have happened without the

programme i.e. a counterfactual.

Page 5: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

J-PAL’s mission is to reduce poverty by ensuring that policy is informed by scientific evidence.

Scale-ups Partnerships Dissemination Workshops Courses

Agriculture Crime Education Environment and Energy

Health Finance and Microfinance

Labour Markets

Political Economy and Governance

EVALUATIONS J-PAL affiliates conduct randomised evaluations to test and improve the effectiveness of poverty reduction programmes across sectors.

POLICY J-PAL affiliates and staff disseminate research results and build partnerships with policymakers to ensure policy is driven by evidence and effective programmes are scaled up.

CAPACITY BUILDING J-PAL trains implementers and policymakers on how to become better producers and users of evidence from impact evaluations.

Page 6: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

We are a network of 146 affiliated professors from over 49 universities…

6

J-PAL Latin America

and the Caribbean: 86

completed

evaluations; 41

ongoing

J-PAL North America:

136 completed

evaluations; 18

ongoing

J-PAL Africa: 131

completed

evaluations; 102

ongoing

J-PAL Europe:

37 completed

evaluations; 6

ongoing

J-PAL South

Asia: 98

completed

evaluations;

68 ongoing

J-PAL Southeast

Asia: 28

completed

evaluations; 16

ongoing

J-PAL Global:

Headquarters

supporting regional

offices

Nearly 819 ongoing and completed evaluations across 8 sectors in over 78 countries

… across 7 global offices, working with local partners on local issues.

Page 7: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

• Over 166 completed and ongoing evaluations across SA; 124 across 13 states of India

• 19 different partnerships with various state

governments

• Partners in India, Bangladesh, Nepal, Pakistan, and Sri Lanka

• 3650+ policymakers & practitioners trained to conduct high quality Monitoring & Evaluation

J-PAL South Asia est. 2007 at the Institute for Financial Management and Research (IFMR), Chennai

7

South Asia senior management team

Iqbal Dhaliwal (ex-IAS) Deputy Director, J-PAL

Scientific Director, SA

Esther Duflo (MIT) Director, J-PAL

Scientific Director, SA

Sanjoy Narayan

Executive Director J-PAL SA

Shobhini Mukerji

Executive Director J-PAL SA

Page 8: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

• State partnerships to institutionalise an evidence

based approach with the government

• Supporting the scale-up of programmes that

work

• Collaborating with Govt. departments and

Ministries to rigorously evaluate innovative

solutions

• Sharing evidence on what works (and doesn’t)

• Assisting in M&E capacity building

Fostering Evidence-based Policymaking

8

Page 9: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

9

Using evidence to improve

effectiveness of policy

Why evaluate?

Page 10: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Rigorous evaluations have produced important and surprising results

10

• Major programmes not as effective as previously thought

– Fixing supply of health services, inputs to education

• Small interventions proved very cost-effective

– Deworming

• Conventional wisdoms have been undermined

– Incentives for monitoring, community participation

Page 11: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

11

Case study1: Choosing between

alternative programme designs Pricing of preventive health products

Page 12: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Pricing of preventive health products The debate in the policy/donor community: free versus subsidised distribution

12

Free distribution?

Cost sharing?

• People who need something are more willing to pay for it

• Paying for goods makes people more likely to use them (sunk cost bias is significant)

• Giving away goods and services for free creates dependency (an entitlement effect)

• Charging fees helps programmes maintain financially sustainability

• Prices prevent access for people who need it the most

• Sunk cost bias may be negligible

• Free samples help people learn about a good’s benefits and they will be willing to pay for them later

• There are positive health externalities that warrant complete subsidisation for maximum public health benefits

Page 13: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Understanding usage patterns for ITNs based on price paid

13

Background

• In Kenya, malaria responsible for one out of every four child deaths

• ITNs shown to reduce child mortality in regions of Africa

• Less than 5 percent of children and pregnant women sleep under an ITN

Programme

• ITN distribution to pregnant women who visited clinics for prenatal care

Cohen and Dupas 2010

Page 14: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Randomised evaluation design: understanding usage patterns for ITNs based on price paid

14

Status quo

Free

Subsidy 97.5%

Subsidy 95.0%

Subsidy 90.0% P

reg

na

nt

wo

me

n v

isitin

g

clin

ic o

ffe

red

ITN

s

Additional discount ($0 to posted price) for women interested to purchase an

ITN

Page 15: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Results: Cost sharing reduced take-up of ITN

16

• Charging 60 cents (10% of actual price) for insecticide treated nets (ITNs) reduced take-up by 60pp relative to free distribution in Kenya (2007)

• No evidence that cost-sharing put ITNs in the hands of women who need it the most

• No evidence that the act of paying for a product makes a recipient more likely to use it (ITNs no more likely to be hanging in the house during a spot check)

Cohen and Dupas 2010

Page 16: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Take-up of preventive products drops as price increases

17

Page 17: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Rigorous evidence has informed pricing of ITNs

J-PAL | EVIDENCE FROM RANDOMIZED EVALUATIONS OF ECONOMIC INTERVENTIONS IN HEALTH 19

Page 18: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

20

Case study2: Understanding

programme’s impact Targeting the Ultra-poor

Page 19: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

• Ultra-poor women headed households make up 69.4

lakh households in India and have an average monthly income of Rs.1,250

The problem

21

Ultra-poor female headed households

There is no evidence yet that they benefit from traditional credit-based interventions

Defining the Ultra-poor

Page 20: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

A potential solution – the ‘Graduation Model’ Carefully sequenced support for the poorest of the poor women in rural communities to graduate out of extreme poverty

22

Programme costs Rs20,000 per beneficiary

Page 21: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Choosing the counterfactual

1. Option 1: Programme participants before the programme

• Why or why not?

2. Option 2: Programme non-participants

• Why or why not?

How do we measure TUP’s impact? Impact = what happened with the programme - what would have happened WITHOUT the programme

23

Page 22: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Designing a randomised evaluation to measure TUP’s impact

24

Status quo (#466)

Offered programme

(#512)

Poorest of the

poorest women

in village

Period Time since

asset

transfer

Baseline Before programme implementation (February 2007- March 2008)

-

Endline

1

Completion of programme (January 2009- November 2009)

1.5 years

Endline

2

One year after programme completion (June 2010-February 2011)

2.5 years

Endline

3

Five years after programme completion (September 2014- March 2015)

7 years

Page 23: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Understanding Intent-to-Treat and Treatment-on-the Treated

26

Programme take-up: 56 percent

1. Intent-to-treat

• Comparing people offered programme with those in the comparison group

• Consumption increases by 15 percent at the end of the programme, and nearly 25 percent five years after the end of the programme

2. Why may we be interested in understanding the ‘treatment-on-the treated’?

Page 24: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Households experience broad and lasting economic impacts

27

• 46% higher consumption than comparison group five years after programme completion

• Consumption pattern changed – spent more on dairy, protein-rich foods and durable goods

• Income increased

• Sources of income diversified

• Food security improved

• Household assets and savings

increased

0

10

20

30

40

50

60

70

80

90

100

At programmecompletion

One year later Seven years later

Comparison group Programme Participants

25%

20%

46%

Page 25: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Replicated in ~20 countries, multisite randomised evaluations in 7 countries

29

India Bangladesh

Pakistan

Ethiopia Ghana

Peru

Honduras

Multisite RCTs funded by the Graduation Program Consortium, i.e., Ford Foundation and Consultative Group to Assist the Poor

Basic programme components

adjusted to fit the individual country

context and implemented by local

organisations

Page 26: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

How have government’s responded to this evidence?

Government of Rajasthan

Rajasthan State Livelihoods mission(Rajeevika) is funding the THP programme implementation, by Bandhan Konnagar, in the Manohar Thana block of Rajasthan for 1,000

Odisha

Jharkhand*

Bihar Rajasthan

Government of Jharkhand

Jharkhand Welfare Department, Government of Jharkhand is funding the THP implementation in 2,000 households in two districts (Dumka and Paschim Singhbhum).

Bihar and Odisha, USAID funded DIV funding 4,350 Households in two states, Bihar and Odisha

Madhya Pradesh

West Bengal

Assam

Other states Funded by foundation and donor funding

Page 27: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Concluding thoughts Social programmes + Rigorous evidence = Higher Impact

31

• Understanding the impact caused by the programme

– Are the people better off than they would have been otherwise?

– What are the reasons for success/failure?

– Causal effect can be determined through a rigorous evaluation

• Comparing programmes and choosing the best – What is the most effective way to achieve an outcome?

– Are there common strategies that will succeed across fields?

• Using rigorous evidence of impact of your intended programme to inform decisions

– Expand coverage of programme?

– Withdraw programme?

– More evidence needed?

Page 28: Case study 1: What, why and how of impact evaluationspubdocs.worldbank.org/en/513191491495929083/7-NEPAL-IE-Day-1-… · Case study 1: What, why and how of impact evaluations Shubhra

Thank you

Shubhra Mittal

[email protected]