Econometric Theory and Methods (90-907)
SPRING 2025
Instructor
Akshaya Jha
[email protected]
Office: Hamburg Hall 2218
Phone: 608-347-2102
Office hours: by appointment
Teaching Assistant
Andre Ribeiro Cardoso
[email protected]
Lectures: Tuesday and Thursday, 2pm – 3:20pm, HBH 1005
Review sessions: Friday, 3:30pm-04:50pm, HBH 1005
Grading: 40% homework (approximately every two weeks)
30% final exam (take home and open book)
30% final project or final presentation
Prerequisites: Statistical Theory for Social and Policy Research (90-905) and
Introduction to Econometric Theory (90-906), or equivalent PhD-level
econometrics with extensive coverage of linear regressions and sufficient
treatment of asymptotic theory. It is assumed that students are familiar
with basic linear algebra, multivariate calculus, probability theory, and
statistical convergence concepts.
Textbooks/Website
Wooldridge, Econometric Analysis of Cross Section and Panel Data, 1st or 2nd edition. (required)
Cameron and Trivedi, Microeconometrics: Methods and Applications, 2005. (recommended)
Greene, Econometric Analysis, Sixth Edition, 2008. (additional supplement)
Course website is on Canvas. Readings and assignments will be posted there.
Computer Software
Many homework assignments will involve data analysis that requires specialized statistical
software. You are free to use any software that you like. We will provide instruction and
support for Stata, which is available on the Heinz PhD server and via the virtual lab. Some
assignments will require a symbolic programming language (i.e., not pre-packaged estimation
commands), and for this we will provide instruction and support for Matlab and R. Some class
examples may be given in Matlab.
Course Description
This course covers a number of econometric models and techniques that are commonly used in
applied microeconomics. The core topics include a general framework for estimators (which
includes maximum likelihood and generalized method of moments), discrete outcome models,
sample selection (and related limited dependent variable or switching models), duration and
count models, time series models, panel data models, variance estimation (including clustering
and the bootstrap), and non-parametric techniques. The course is designed for PhD students who
have completed 90-906 (PhD Econometrics I) or an equivalent course.
Course Objectives
1. Understand the fundamental properties of M-estimators, both formally and intuitively.
2. Formulate appropriate econometric models for specific applications within the classes of
data covered by the course, addressing issues of endogeneity and error correlation.
3. Derive properties of the estimators based on the above models.
4. Apply these estimators to data provided as part of course assignments.
Course Policies
Please write up solutions in LaTeX and submit online by midnight of the due date. Students may
work on the assignments in groups—in fact this is recommended—but each student must write
their own submission and these cannot be direct copies of each other’s work or be copied from
any other source (e.g., something online). The final exam will be take home and open book.
There is also a final project or final presentation requirement; the final project involves utilizing
one of the topics from the course in your own research while the final presentation would be
teaching the class an econometrics topic not covered in class. Any suspected incidents of cheating
or plagiarism will be recorded with Heinz College administration at the same time the student is
notified.
COURSE OUTLINE
1. Frameworks for Estimators (4 lectures)
1. General theory for M-estimators
2. Maximum Likelihood Estimation
3. Generalized Method of Moments
4. Numerical optimization methods
5. Hypothesis testing
1. (Non-linear) testing for parameters
2. Specification Tests
2. Discrete Outcome Models (4)
1. Binary Logit, probit, and linear probability models
2. Specification issues: heteroskedasticity and distributional assumptions
3. Endogenous explanatory variables (if time)
4. Multinomial logit, nested logit, multinomial probit
5. Ordered logit and probit (if time)
3. Selection Models (4)
1. Censored and truncated dependent variables
2. ML estimation of the above
3. Motivation for selection – agent choices affect observations
4. Selection model variants
5. ML and two-step estimation
6. Specification issues
4. Duration Models (1)
1. Typical duration outcomes and data
2. Hazard functions
3. ML estimation with continuous or discrete time
4. Unobserved heterogeneity and other specification issues
5. Time Series Models (3)
1. Stationarity
2. Autoregressive and Moving Average Models
3. Model Selection and Fitting
6. Panel Data Models (2)
1. Linear panel model framework and permanent unobserved heterogeneity
2. Estimation with heterogeneity: fixed effects and first differencing
3. Dynamic linear panel models: predetermined and endogenous explanatory variables
4. Binary outcome panel models
7. Non-parametric and Semi-parametric Estimation Methods (3)
1. Density Estimation
2. Kernel Regression
3. Robinson's Semi-parametric Estimator
4. Non-parametric Hypothesis Tests
8. Bootstrap and Sub-sampling Methods (2)
1. Bootstrap Methods for Independent and Dependent Data
2. Sub-sampling Methods
9. Survey Weights and Variance Estimation (2)
1. Weighted estimation: theory and practice
2. Heteroskedasticity- and cluster-robust standard errors
3. Bootstrapped standard errors
4. Multiple hypothesis correction
