Home
About
Mobile
Open Content
Search
Module Overview
Description
Facilitators
Weblinks
Timetable
PP5171
ADVANCED APPLIED ECONOMETRICS FOR POLICY ANALYSIS
2015/2016, Semester 2
Lee Kuan Yew School of Public Policy (Lee Kuan Yew School Of Public Policy)
Modular Credits: 4
Tags:
--
Collapse All
Learning Outcomes
Top
The purpose of this course is to provide students with a ‘toolbox’ and working knowledge of advanced crosssectional and panel data econometric techniques frequently used in applied microeconomic policy analysis and research. This course will cover major extensions to the standard OLS regression model and provide students with an introduction to the ‘cutting edge’ techniques used today to evaluate microeconomic theories and policies, including instrumental variables, difference-in-differences, matching estimators, regression discontinuity and quantile regressions. The emphasis of the course will be on estimating causal relationships that can then be used to make predictions about the consequences of changing a policy.
Prerequisites
Top
NA
Syllabus
Top
PP5171 – Advanced Applied Econometrics for Policy Analysis
Lee Kuan Yew School of Public Policy, National University of Singapore
Spring 2016
Instructor:
Ron Shadbegian: spprjs@nus.edu.sg
Time:
Wednesday, 6:30 pm – 9:30 pm
Classroom:
SR 3-5
Course Description
:
Regression analysis plays an extremely important role in the empirical analysis of public policies and research and as such will be the foundation of this course. The purpose of this course is to provide students
with a ‘toolbox’ and working knowledge of advanced cross-sectional and panel data
econometric techniques frequently used in applied microeconomic policy analysis and research. This course will
cover major extensions to the standard OLS regression model and provide students with an
introduction to the ‘cutting edge’ techniques used today to evaluate microeconomic theories and policies
, including instrumental variables, difference-in-differences, matching estimators, regression discontinuity and quantile regressions. The emphasis of the course will be on estimating causal relationships (e.g. whether our estimates will deliver answers to questions like: “Does health insurance make us healthier?”) that can then be used to make predictions about the consequences of changing a policy.
Given time constraints the course cannot go into too much depth in regard to any particular technique, but will instead endeavor to provide you with enough knowledge about each one so that you will
understand what various methods do
, how and when to use them, and when not to use them rather than on the study of the mechanical and mathematical features of estimators. None the less to do this we will have to learn the mechanics of some of the estimators. The level of mathematics in this course will not be too complex, but we will need to use tools including expectation operators, multivariate calculus and simple matrix algebra at times, and these tools will reviewed in class as they become necessary
.
Learning Objectives:
By the end of the semester each student should be able to:
develop appropriate sophisticated econometric models to estimate the causal relationships
for a range of interesting policy issues (e.g. “Does health insurance make us healthier?”) that can then be used to make predictions about the consequences of changing a policy a
nd validate the selected model using an array of statistical tests
use Stata to estimate these models with cross-sectional and/or panel data and perform the
necessary
statistical tests to validate the use of the empirical model and be able to
interpret the economic meaning of the results.
conduct an independent empirical research project, on a topic of their choosing, which will help
integrate their understanding of the applied econometric techniques covered during the semester
The applied econometric skills students learn in this class, which I use in my own research and have used in my policy work at the U.S. Environmental Protection Agency, will prepare them for work in both the private sector (e.g. consulting firms, think tanks, and banks) and public sector (e.g. government agencies and NGOs), as well as, for future graduate work in economics or a related field.
Office hours:
TBD.
Required Texts
: Mostly Harmless Econometrics: An Empiricists Companion by Angrist and Pischke
Introduction to Econometrics by Stock and Watson 3
rd
edition
Supplementary Texts
:
Introductory Econometrics: A Modern Approach (4
th
or 5
th
) edition by Jeff Wooldridge
Basic Econometrics (3
rd
or 4
th
) edition by
by Damodar N. Gujarati
Grading:
Your grade for the class will be divided between:
Homework (20%)
Midterm Exam (20%)
Non-Cumulative Final Exam (30%)
Research paper and class presentation (30%)
The midterm will take place on
TBD
. The final exam will occur on
TBD
.
Lab Sessions
This course will not include any formal lab sessions. However, from time to time, as necessary, I will conduct informal, optional lab sessions to supplement the course with hands-on experience of data analysis using STATA. The computer labs are already equipped with Stata and students can practice and do the exercises in the lab.
Software
This course will use the Stata statistical package. Stata version 13 is available in the computer lab in LKYSPP.
In case you want your own copy of Stata, plans for students at discounted prices are at
http://www.stata.com/order/new/edu/gradplans
(Do not buy “Small Stata,” which cannot handle the datasets used in this class.). You can find the contact information of official Stata distributors in Singapore at
http://www.stata.com/worldwide/?country=Singapore
.
Academic Integrity
Students in this class should adhere to NUS Honour Code, which can be found at
http://www.nus.edu.sg/registrar/adminpolicy/acceptance.html#NUSHonourCode
. In addition, the LKY School’s Code of Conduct (
http://www.spp.nus.edu.sg/Code_of_Conduct.aspx
) lists academic integrity as one of its six important values. Violations of these codes in any form, including cheating in exams and plagiarism, will not be tolerated and will immediately lead to the student getting zero marks and follow-up action.
Plagiarism includes copying all or any part of your classmate’s assignments. To avoid giving the impression that you are passing off other people’s work as your own, you will need to acknowledge conscientiously the sources of information, ideas, and arguments used in your paper. For this purpose, you will use the ‘footnote style’ according to the Chicago Manual of Style, the guidelines for which can be found online at
http://www.dianahacker.com/resdoc/p04_c10_s2.html
in the companion website for Diana Hacker’s A Writer’s Reference. Please also refer to the handout that was given to you at the Workshop on Plagiarism conducted during the Orientation period.
Use of laptops is discouraged during lectures unless its purpose is related to this course. Instructor reserves the right to ban laptops if their use distracts students. Using cell/smart phones is prohibited during lectures and exams.
Course Outline
Part I Introduction: Random Variables, Linear Regression Model, and Causality
The Experimental Ideal
Conditional Expectations Function (CEF)
Linear Regression Model and the CEF
Causality and the Selection Problem
Reading: MHE ch.s 1-2, sections 3.1.1 & 3.1.2; S&W ch.s 1-2, 4
Part II Matrix Algebra of OLS
A. Introduction to Matrix Algebra
What is a matrix?
Matrix Operations: Addition, Subtraction, Multiplication and Differentiation
Determinants
Matrix Rank
Matrix Inversion
Reading: S&W ch.18 pp 737-740
or
Gujarati Appendix B
B. The Matrix Approach to OLS
The OLS Model in Matrix Notation
Assumptions of the CLRM in Matrix Notation
OLS Estimation using Matrix Algebra: β = (X'X)
-1
X'Y
Variance-Co-Variance of β using Matrix Algebra: VAR-COVAR = σ
2
(X'X)
-1
BLUE Property of the OLS Estimator
Monte Carlo Simulations
Reading: S&W ch.18 pp 698-702 & 712-713
or
Gujarati Appendix C
Part III
Multivariate Regression
Saturated Models – Binary (Dummy) Variables & Interaction Terms
Regression and Causality
Conditional Independence (Selection on Observables) Assumption
Short vs Long Regression
Omitted Variables Bias
Regression Anatomy Formula
Bad Control
Reading: MHE sections: 3.1.4 & 3.2.1 - 3.2.3
Part IV
Instrumental Variables/Two-Stage Least Squares
Cov(X,U) ≠ 0
Failure of OLS: Biased and Inconsistent
IV: Biased, but Consistent
Omitted Variable Bias
IV and Causality
Checking Instrument Validity
Problems of Identification
Problem of Weak IVs
Hausman Test of Simultaneity
Lagrange-Multiplier Test: Test for Overidentifying Restrictions
2SLS – The case of Simultaneous Equations
Reading: MHE sections 4.1, 4.6.1, & 4.6.4; S&W ch. 12
Part V
Models for Panel Data: Combining Cross-Sectional Data and Time Series Data
Pooled OLS Regression
Fixed Effects and Random Effects
Within Group Estimator and the Between Group Estimator
Breush-Pagan Test for Random Effects
Hausman Test: Fixed or Random Effects
Clustering SEs
Reading: MHE section ch. 5.1; S&W ch. 10
Part VI Experiments and Quasi-Experiment
The Randomized Control Trial Ideal
Potential Outcomes, Causal Effects and Idealized Experiments
Internal vs. External Validity
Quasi Experiments and the Differences-in-Differences Estimator
Reading: MHE section ch. 5.1; S&W ch. 13
Part VII Matching Estimators and Selection Models
Sample Selection Bias
Creating Counterfactuals with Observational Data
Approximating Random Control Trials – Balancing Covariates
Estimating DifferenctTreatment Effects
Conditional Independence (Selection on Observables) Assumption (CIA)
Stable Unit Treatment Value Assumption (SUTVA)
Common Support
Estimating Various Matching Models: Nearest Neighbor; Mahalanobis; and Kernel
Combining Matching with Regression Analysis
Heckit Model
Reading: MHE sections 3.3.1 – 3.3.3
Part VIII Two Additional Extensions
Regression Discontinuity Designs
Sharp RD vs. Fuzzy RD
Estimating a RD
Internal vs. External Validity
Reading: MHE ch. 6
Quantile Regression
Least Absolute Deviations Regression
Estimating Causal Effects on the Distribution of Y
Estimating Conditional Median Function
Reading: MHE sections 7.1.1 & 7.1.3
Preclusions
Top
NA
Workload
Top
3-0-0-2-5
Workload Components : A-B-C-D-E
A: no. of lecture hours per week
B: no. of tutorial hours per week
C: no. of lab hours per week
D: no. of hours for projects, assignments, fieldwork etc per week
E: no. of hours for preparatory work by a student per week