Upload
dortha-summers
View
212
Download
0
Tags:
Embed Size (px)
Citation preview
Randomized Control Trials forAgriculture
Pace Phillips, Innovations for Poverty Action
www.poverty-action.org
• What we do• IPA Ghana • IPA’s Mission and Approach• Why evaluate? • How to evaluate impact? • Examples of evaluations
Outline
What does IPA do?• IPA was created in 2002 to determine what
works and what does not work in poverty alleviation through rigorous evaluation.
• Estimate magnitude of effects– Most cost-effective solutions; can be surprising
• Learn how to improve programs through testing operational questions
• Determine where to spend limited resources
IPA Ghana• In Ghana since 2005• 16 projects nationwide• Permanent offices in Accra and Tamale• Current Agric Partners:
– Ghana Insurers Association– GIZ – IFPRI – Presbyterian Agricultural Service– MoFA– SARI– ISSER
• Trainings /Conferences
• Surprisingly little hard evidence on what works
• Can do more with given budget with better evidence
• If people knew money was going to programs that worked, could help increase pot for anti-poverty programs
• Instead of asking “do aid/development programs work?” should be asking:– Which work best, why and when?– How can we scale up what works?
Why focus on impact evaluation?
6
Process and impact evaluation
InputsInputs OutputsOutputs OutcomesOutcomes
•Process evaluations
•Monitoring
In the control of the organization
Impact
Impact evaluation: how, and how much, does the organization affect the welfare of beneficiaries (and non beneficiaries)
Monitoring/Process evaluation: tracking development and outputs of a program as compared to stated objectives, targets and timelines
• In Northern Ghana, a development organization undertakes a program to promote fertilizer use to improve yields. How do we know if the program was successful?
How do you evaluate a program?
Impact: What is it?
Time
Cro
p Y
ield
s
Impact
Counterfactual
Intervention
What is the impact?
Time
Cro
p Y
ield
s
ImpactCounterfactual
Intervention
What is the impact?
Time
Cro
p Y
ield
s
ImpactCounterfactual
Intervention
• Counterfactual: What would have happened in the absence of the program?
• Problem: Counterfactual is not observable the key goal of all impact evaluation methods
is to construct or “mimic” the counterfactual.
• Solution: Counterfactual is often constructed by selecting a group not affected by the program.
Counterfactual
• Without Random Selection, the control group will have intrinsic differences that can bias the study.
• Because members of the groups (treatment and control) do not differ systematically at the outset of the experiment
• Any difference that subsequently arises between them can be attributed to the program rather than to other factors
• If properly designed and conducted, randomized experiments provide the most credible method to estimate the impact of a program
Why randomize?
13
Randomly samplefrom area of interest
Random sampling and random assignment
Randomly samplefrom area of interest
Randomly assignto treatmentand control
Random sampling and random assignment
Randomly samplefrom both treatment and control
Which method we use matters!
Method Impact Estimate
(1) Pre-post 26.42*
(2) Simple Difference -5.05*
(3) Difference-in-Difference 6.82*
(4) Regression 1.92
(5)Randomized Experiment 5.87**: Statistically significant at the 5% level
Underinvestment in Agriculture• A study in Northern Ghana to understand if farmers do not invest in their farms because either capital constraint or risk aversion.
Year 2: 1377 farmers in 74 villages
Year 2 Randomization
Insurance: 729 farmers Both: 108 farmersCapital: 363 farmers
Rainfall index insurance offered at varying prices
GHS 350 provided
Control: 177 farmers
Rainfall index insurance offered at varying prices, AND GHS 350 provided
Results• Capital only farmers– Used more inorganic fertilizer only
• Capital and insurance farmers– Increased farm investment by 20 percent
• Insurance farmers– Increased total farm expenditure by 13 percent
• Increased inorganic fertilizer use by 25 percent• Increased cultivation area by 8 percent• Increased expenditures on land preparation by 12 percent • Increased total labor use on plots by 13 percent• Increased harvest output by 8 percent
• Increase in investment but no increase in profitability
Demand for Agriculture InsuranceUptake
A Well Timed Nudge• A study in Western Kenya tested whether farmers purchases of fertilizer was dependent on the time of the year it was offered.
Baseline: 1,125 Farmers in Suri Kenya
Randomization
Fertilizer offered AFTER HARVEST
with free delivery
50% Subsidy Fertilizer offered
just BEFORE GROWING SEASON with free delivery
CHOICE of when they want the
fertilizer offer with free delivery.
Control – No Offer
Fertilizer offered just BEFORE
GROWING SEASON with free delivery
Results• Farmers had high demand and ability to
purchase fertilizer in advance. • Impact of the After Harvest offers were
comparable to that of a 50% subsidy at fertilizer application time.
• Increase in Fertilizer Use• Control Group 28%• After Harvest 39%• Choice of Time 47%• Before Growing Season 33%• 50% Subsidy Before Growing Season 41%
• Once the program stopped, fertilizer usage went back to what it had been.
Thank You
www.poverty-action.org
IPA Ghana • Why do farmers under-invest in farms?• What is the impact of financial savings programs
in schools?• What is the willingness to pay for clean water?• What is the impact of remedial education
programs, smaller class size, teacher training and tracking on literacy and numeracy?
• What is the impact of health insurance education on enrollment and reenrollment in NHIS?
• Are mobile reminders an effective way to improve adherence to ACT regimens?
Problems Selecting a Control Group
Difference between Groups
Intrinsic Differences Impact of the program
Observables Non Observables
Selection Bias
Without Random Selection, the control group will have intrinsic differences that can bias the study.
Our approach
Innovate Evaluate Replicate Communi-cate Scale
- Understand market failures-Develop innovative solutions to poverty- Use frontier knowledge from economics, and psychology
Randomized Controlled Trials-Impact evaluations- Comparing variations of an intervention- Experiment with product designs
Replicate evaluations in various settings to :- Generalize research findings - Tell practitioners what works (and not), when
Effectively communicate to practitioners: -Conferences-Workshops with policy makers and practitioners-Policy memos and focus notes
Facilitate scale-up of effective solutions :- Active policy outreach -Practitioners’ toolkits - Hands-on technical assistance
We generate insights on what works and what does not through randomized evaluations, and ensure that those findings will be useful to, and used by practitioners and policy makers