AFRICA IMPACT EVALUATION INITIATIVE, AFTRL Africa Program for Education Impact Evaluation Muna Meky...
-
Upload
charles-owen -
Category
Documents
-
view
213 -
download
1
Transcript of AFRICA IMPACT EVALUATION INITIATIVE, AFTRL Africa Program for Education Impact Evaluation Muna Meky...
AFRICA IMPACT EVALUATION INITIATIVE, AFTRL
Africa Program for Education Impact Evaluation
Muna Meky
Impact Evaluation Cluster, AFTRL
Slides by Paul J. Gertler & Sebastian Martinez
Impact Evaluation Methods: Impact Evaluation Methods: Randomization, IV, Regression
Discontinuity
Measuring Impact
• Randomized Experiments
• Quasi-experiments– Randomized Promotion-Instrumental
Variables– Regression Discontinuity– Difference in difference – panel data– Matching
Randomization
• The “gold standard” in evaluating the effects of interventions
• It allows us to form a “treatment” and “control” groups– identical characteristics – differ only by intervention
• Closest approximation to counterfactual
Random Assignment
• Each eligible unit has the same chance of receiving the intervention– Mimics chocolate experiment
• Allows us to compare the “treatment” and “control group”
Random Assignment vs. Random Sample
• Random Assignment– Are the observed results due to the
intervention rather than other confounding factors? (internal validity)
• Random Sample– Do the results found in the sample apply to
the general population/are they generalizable? (external validity)
Randomization
Randomization
Randomization
Random Sample
(external validity)
Random Assignment
(internal validity)
Example of Randomization
• What is the impact of providing free books to students on test scores?
• Randomly assign a group of school children to either:- Treatment Group – receives free books
- Control Group – does not receive free books
How Do You Randomize?
1) At what level? – Individual – Group
• School
• Community
• District
When would you use randomization?
• Universe of eligible individuals typically larger than available resources at a single point in time
– Fair and transparent way to assign benefits– Gives an equal chance to everyone in the sample
• Good times to randomize:– Pilot programs– Programs with budget/capacity constraints – Phase in programs
Oportunidades Example
• Randomized treatment/controls– Community level randomization
• 320 treatment communities• 186 control communities• Pre-intervention characteristics well balanced
Oportunidades Example
Variables Treatment (4, 670)
Control (2727) t- stats
Consumpti on per capi ta 233. 47 233. 4 - 0. 04
1.02 1.3
Head' s age 41. 94 42. 35 1. 20.2 0.27
Head' s educati on 2. 95 2. 81 - 2. 160.04 0.05
Spouse' s age 37. 02 36. 96 - 0. 380.7 0.22
Spouse' s educati on 2. 76 2. 76 - 0. 0060.03 0.04
Speaks an i ndi genous l anguage 41. 69 41. 95 0. 21
0.007 0.009
Head i s femal e 0. 073 0. 078 0. 660.003 0.005
Househol d at basel i ne 5. 76 5. 7 - 1. 21
0.02 0.038
Bathroom at basel i ne 0. 57 0. 56 - 1. 040.007 0.009
Total hectareas of l and 1. 63 1. 72 1. 35
0.03 0.05
Mi n. Di stance l oc-urban 109. 28 106. 59 - 1. 02
0.6 0.81
RANDOMI ZATI ON
Difference in Means
Control Treatment t-statMean CPC Baseline
233.40 233.47 0.04
Mean CPC Followup
239.5 268.75 9.6
Case 3 - Randomization
Linear Regression Multivariate Linear RegressionEstimated Impact on CPC 29.25** 29.79**
(3.03) (3.00)** Significant at 1% level
Case 3 - Randomization
Impact Evaluation Example: Randomization vs. Other Methods
Case 1 - Before and After
Case 2 - Enrolled/Not Enrolled
Case 3 - Randomization
Multivariate Linear Regression
Multivariate Linear Regression
Multivariate Linear Regression
Estimated Impact on CPC 34.28** -4.15 29.79**
(2.11) (4.05) (3.00)** Significant at 1% level
Other Analyses Often Lack Internal Validity
• Enrolled vs non-enrolled– The baseline characteristics will be different
because people have chosen which group they want to be in
• Before and after– There are often other interventions going on
at the same time
Measuring Impact
• Randomized Experiments
• Quasi-experiments– Randomized Promotion-Instrumental
Variables– Regression Discontinuity– Difference in difference – panel data– Matching
When Do We Use Random Promotion?
• Common scenario:– National Program with universal eligibility– Voluntary enrollment in program
• Can not control who enrolls and who does not
Randomized Promotion
• Possible solution: random promotion or incentives into the program– Information
– Money
– Other help/Incentives
• Also called– Encouragement designs
– Incentive schemes
Study Components
• Intervention– Chocolate
• Randomized Promotion – Encouragement to take chocolate
Example of Promotion Design for SATs
• What is the impact of supplementary learning material on student test scores?– Outcome
• Student test scores
– Intervention• Supplementary learning materials• (all teachers can access these)
– Randomized promotion• Letters encouraging students to use materials• (sent to randomly assigned teachers)
What Information Does Randomized Promotion Give Us?
• How effective is the treatment?– We can analyze the effect the treatment
had on the outcome in the sub-group of subjects who would not have used the intervention unless encouragement was present
How Effective is the Treatment?
• Local Average Treatment Effect
– Effect of the intervention on those who would not have enrolled unless encouraged
Encouragement Design
ENCOURAGED
Take-up = 80%
Mean outcome = 100
NOT
ENCOURAGED
Take-up = 30%
Mean outcome = 90
Change Take-up = 50%
Change Y=10
Impact = 20
Never Takeup
Takeup if Encouraged
Always Takeup
Example: Community Based School Management
• What is the effect of decentralization of school management on learning outcomes?– All communities are eligible – Community management of hiring, budgeting,
oversight– Need to write proposal
Community Based School Management
• 1500 schools in the evaluation
• Each community chooses whether to participate in program
Promotion Design
• Community based school management– Provision of technical assistance and
training by NGOs for submission of grant application
– Random selection of communities with NGO support
Community Based School Management
• Outcome – learning outcome
• Intervention– decentralization of management to
community– 1500 schools
• Promotion– NGO support– schools randomized to receive this support
Examples – Randomized Promotion
• Maternal Child Health Insurance in Argentina– Intensive information campaigns
• Employment Program in Argentina– Transport voucher
• Community Based School Management in Nepal– Assistance from NGO
• Health Risk Funds in India– Assistance from Community Resource Teams
Randomized Promotion
• Just an example of an Instrumental Variable
• A variable correlated with treatment but nothing else (i.e. random promotion)
• Again, we really just need to understand how the benefits are assigned– Don’t have to exclude anyone
Measuring Impact
• Randomized Experiments
• Quasi-experiments– Randomized Promotion-Instrumental
Variables– Regression Discontinuity– Difference in difference – panel data– Matching
Introduction
• What is the impact of monetary incentives on test scores to schools below some ranking?
• Compare the schools right above the ranking point to schools below the cutoff point
Indexes are Common in Targeting of Social Programs
• Anti-poverty programs– targeted to households below a given
poverty index
• Pension programs– targeted to population above a certain age
• Scholarships– targeted to students with high scores on
standardized test
Example: What is the Effect of Cash Transfer on Consumption?
• Intervention:– Cash transfer to poor households
• Evaluation: – Measure outcomes (i.e. consumption,
school attendance rates) before and after transfer, comparing households just above and below the cut-off point.
Example: What is the Effect of Cash Transfer on Consumption?
• Target poorest households for cash transfer
• Method:– construct poverty index from 1 to 100 with
pre-intervention characteristics• households with a score <=50 are poor
• households with a score >50 are non-poor
6065
7075
80O
utco
me
20 30 40 50 60 70 80Score
Regression Discontinuity Design - Baseline
Non-poor
Poor
6570
7580
Out
com
e
20 30 40 50 60 70 80Score
Regression Discontinuity Design - Post Intervention
6570
7580
Out
com
e
20 30 40 50 60 70 80Score
Regression Discontinuity Design - Post Intervention
Treatment effect
Regression Discontinuity
• When to use this method? – the beneficiaries/non-beneficiaries can be
ordered– there is a cut-off point for eligibility.– cut-off determines the assignment of a
potential beneficiary to the treatment or no-treatment
Regression Discontinuity
Fitt
ed v
alue
s
puntaje estimado en focalizacion276 1294
153.578
379.224
2
Baseline – No treatment
0 1 ( )i i iy Treatment score
Regression Discontinuity
Estimated Impact on CPC
** Significant at 1% level
Case 4 - Regression DiscontinuityMultivariate Linear Regression
30.58**(5.93)
Fitt
ed v
alu
es
puntaje estimado en focalizacion276 1294
183.647
399.51
Treatment Period
Sharp and Fuzzy Discontinuity
• Sharp discontinuity – the discontinuity precisely determines treatment
– equivalent to random assignment in a neighborhood• e.g. social security payment depend directly and immediately
on a person’s age
• Fuzzy discontinuity – discontinuity is highly correlated with treatment .
– use the assignment as an IV for program participation.• e.g. rules determine eligibility but there is a margin of
administrative error.
Examples
• Effect of class size on scholastic achievement (Angrist and Lavy, 1999)
• Effect of transfers on labor supply
(Lemieux and Milligan, 2005)
• Effect of old age pensions on consumption -BONOSOL in Bolivia
(Martinez, 2005)
Potential Disadvantages of RD
• We estimate the effect of the program around the cut-off point.– the effect is estimated at the discontinuity– fewer observations than in a randomized
experiment with the same sample size– limits generalizability
• Make sure the relationship between the assignment variable and the outcome variable is correctly modeled, including: – nonlinear relationships– interactions
Advantages of RD for Evaluation
• RD allows one to estimate the effect of an intervention at the discontinuity
• Can take advantage of a known rule for assigning the benefit that are common in the designs of social policy– No need to “exclude” a group of eligible
households/individuals from treatment
Summary
Design When to use Advantages DisadvantagesRandomization •Whenever
possible
•When an intervention will not be universally implemented
•Gold standard
•Most powerful
•Not always feasible
•Not always ethical
Random Promotion
•When an intervention is universally implemented
• Learn and intervention
•Only looks at sub-group of sample
Regression Discontinuity
•If an intervention is assigned based on rank
•Assignment based on rank is common
•Only look at sub-group of sample