Unlock document.
This document is partially blurred.
Unlock all pages and 1 million more documents.
Get Access
Spring 2017 Impact Evaluation Methods
Homework Assignment #5: Summary Paper
Regression Discontinuity Designs in Economics
David S. Lee and Thomas Lemieux
Submitted by
Thaw Zin Win
201632006
Introduction to RD Designs
RD Designs were first introduced by Thistlethwaite and Campbell fifty years ago
(“RD Analysis: An Alternative to Ex Post Fact Experiments,” Journal of Education
Psychology, 1960). Internal vs. external validity
Internal validity: " According to Brewer (Research Design and Issues of Validity,
2000) “Inferences are said to possess internal validity if a causal relation between
two variables is properly demonstrated” " We think that RD design have typically a
high level of internal validity because they provide a convincing way of estimating
a causal effect
External validity: " Brewer: “Inferences about cause-effect relationships based on
a specific scientific study are said to possess external validity if they may be
generalized from the unique and idiosyncratic settings, procedures and participants
to other populations and conditions” " Problematic for randomized experiment
Thistlethwaite and Campbell: simple regression approach
` Y = α +
D
τ +
X
β + ε,
Xi : assignment variable
D
τ : treatment variable, Di =1[Xi !c] General
problem in such a regression:ε i and Di are potentially correlated RD
solution:Di only depends on Xi (Di =1[Xi !c]), so #i and Di cannot be
correlated once we have controlled for Xi (in a smooth way)
2.2 RD Designs and the Potential Outcomes Framework
« Potential Outcomes »
Y = Y(1) when D =1, Y = Y(0) when D =0 E[Y(1) - Y(0)] (the
average treatment effect or ATE)
Lee (2008): local randomization
Randomization: experimental approach (in laboratory or field setting) comparison
of means While RD is a non-experimental design, we have local randomization
provides that the following assumption holds (Lee, 2008):
