Economics 623: Econometrics
John Rust, University of Maryland
Office hours: T-Th 2:00-3:30 or by appointment
Optional Texts:
Related Econometrics Courses on the Web
Tuesday October 23th: Introduction and Course Overview
- An overview of some of the big picture issues and conflicts in
econometrics, including
- the tension between using econometric models to
summarize/describe data versus using models
to attempt to infer causality and make predictions,
especially for counterfactual predictions and policy
making
- the pros and cons of "structural econometric models"
versus "reduced form models"
- experimental versus non-experimental methods of inference
- the use of "calibration" versus econometric methods as
the right way to do empirical work
- the role of Bayesian versus classical inference
- parametric versus semi-parametric versus non-parametric methods of
inference
- more generally, the role and validity of prior assumptions
and models in doing empirical work
- is empirical work in economics well served
by the standard statistical mindset of empirical work as a
form of hypothesis
testing where the focus is on estimators that can
consistently estimate the " true model"?
Or are economic models, by the very fact that they are models (simplified abstractions
of
reality) simply approximations to reality, and thus a better
framework for thinking about empirical work in economics is
approximation theory (i.e. a search
for better approximations to reality, but recognizing that no single
model can ever be "correct")?
- Michael Keane (2006) Structural vs. Atheoretical Approaches To Econometrics
forthcoming, Journal of Econometrics
- John Rust (2007) Comments on Keane
- Sungjin Cho and John Rust (2007) Is Econometrics Useful for Private Policy Making? A Case
Study of Replacement Policy at an Auto Rental Company
Thursday October 25th - Nov 6th: Selectivity Bias and
Regression, the role of experimental versus non-experimental data
for inferring causal effects, and the Treatment Effects Literature
November 8th Survey of Nonparametric Regression Methods (kernel
regression, local linear regression, series/sieve estimators, smoothing splines, neural networks, and regression trees)
December 4th Robust Regression Methods (least absolute deviations, quantile
regression, trimming, influence functions, and other methods for identifying, and eliminating or reducing the
effects of ``outliers'')
|