UC
Berkeley DeCal Program
GLOBAL POVERTY & IMPACT EVALUATION: LEARNING WHAT WORKS FOR THE WORLD’S
POOR
Fall 2009
Tuesdays 5:00 – 5:00 PM
Location: 122 Wheeler
Student Facilitators:
Garret Christensen (garret AT econ DOT berkeley DOT edu)
Erick Gong (egong AT are DOT berkeley DOT edu)
Instructor of Record: Ted Miguel
(emiguel AT econ DOT berkeley DOT edu)
Sponsored by the Center of
Evaluation for Global Action
CEGA is a multi-disciplinary research center at the University of
California, Berkeley advancing global health and development through
impact evaluation and economic analysis. The Center is premised on the
principle that knowledge gained from randomized trials and other forms
of impact evaluation is a valuable public good that can improve policy
and outcomes around the world.
Course Content: The course
will cover impact evaluation theory (causal inference, experimental
design and basic statistics) as well as methods (randomization,
difference-in-difference, regression discontinuity, and propensity
score matching). The curriculum will be applied, with weekly case
studies of field research drawn from the international development
literature. Discussions of methods will include issues related to
research ethics and the protection of human subjects. At the end of the
course, students will have the opportunity to present their own impact
evaluation research projects and get feedback from CEGA faculty members.
Audience: The course is ideally suited for graduate and
advanced undergraduate students with an interested in impact
evaluation. Graduate students in Public Policy, Public
Health, Education, Political Science, ERG, and
Sociology, and undergraduates who have taken statistics courses may
benefit the most from this course. The curriculum is very applied
and will be useful for students engaged in international development
field projects, social entrepreneurship, and policy analysis.
Please email one of the student facilitators if you have questions
about whether this course is the right fit given your interests and
background.
Learning Outcomes: Students who complete this course will be
prepared to: 1) distinguish research-based “best practices” from
those that have not been rigorously evaluated; 2) design an
impact evaluation of a policy or intervention, and 3) evaluate
data using a statistical software package.
For students who are considering conducting an impact evaluation of a
program, facilitators will provide references to technical resources
(e.g. textbooks on sample design and software for power calculations)
and guidelines for developing a rigorous study.
Methods of Instruction: During class, facilitators will present
the main concepts in short lectures structured around case studies
(suggested readings from the literature), which will also serve as the
basis for class discussion and small group activities. Lectures will
discuss the strongest (most rigorous) evaluation methods and the
shortcomings of weak evaluation methods. Case studies will highlight
research from Africa, Asia, and South America as well as the U.S. and
will cover programs related to health, governance, education, and
agriculture. Group work will provide hands-on experience with research
design and data analysis.
Grading: Students will be graded on the following: 1)
attendance, 2) participation in discussion, 3) 4 short problem sets
(approx 1 hour of work each) and 4) a group presentation.
Students who miss two days of lecture (not including the first week’s
introduction) will be in danger of failing the course. For every
lecture that a student misses, the student will need to submit a
one-page summary/reaction to the lecture slides or referenced papers
(posted below). Depending on time availability, class size, and
students’ interests, group presentations
will take place in the final
two weeks of class.
Assignments: The problem sets are designed to teach
students how to apply the four methods (randomization,
difference-in-difference, regression discontinuity, and propensity
score matching) using statistical software (STATA) with actual
data. An example of STATA code will be provided for each problem
set. Listed below are the four problem sets with the emphasis in
parenthesis and the due dates.
Problem Set 1 (Randomized Evaluations): Handed Out Sept
22nd, Due Oct 3rd.
Problem Set 2 (Regression Discontinuity): Handed Out Oct 13, Due Oct
27th.
Problem Set 3 (Matching / Propensity Score): Handed Out Oct 27th, Due
Nov 10th.
Problem Set 4 (Difference-in-Difference): Handed out Nov 10th, Due Nov
24th.
The group presentation will involve either a written or oral
presentation brief research proposal. Details.
Software Problem sets will require STATA, a statistical software program
widely used in impact evaluations. We recommend that students
install STATA on their computer in order to complete the problem
sets. If you need to purchase a copy, a single-user
one-year license for Small Stata (sufficient for this course) is
available through Berkeley’s GradPlan for $48. Note that a
license allows you to install the software on up to three of your own
computers. See www.stata.com/order/schoollist.html to purchase (select
CA, then UCB, then product code SMSOFTAGS).
Students can also access STATA in the computer labs at 1535 Tolman Hall
during drop in hours. If you need access, we will issue you a
login and password. Drop in hours for the Tolman Computer labs
can be found at (http://facility.berkeley.edu/labs/hourstmf.html)
Schedule:
*Asterisks imply optional readings that may not be discussed
directly in class but are likely relevant and helpful.
September 22: Randomized
Evaluations II: Applications
(Case Studies: housing vouchers in the US, microfinance in South
Africa, and agriculture in
Kenya )
October 13: Regression
Discontinuity
(Case Studies: scholarship program for girls in Kenya, educational
finance in Chile)
Unpublished results from follow-up on a
girl’s merit scholarship program. For a description
of the intervention, see Kremer, Michael et al. Incentives
to Learn. NBER Working Paper
#10971. 2004.
October 20: External Validity
(Case Studies: anti-corruption programs in Indonesia and Brazil, &
community-based monitoring
of health clinics in Uganda)