Econometrics I
PIMES/UFPE — Graduate Program in Economics
Econometrics is the study of how we learn about unknown parameters from data generated by a probabilistic process. Everything in this course is an instance of that problem.
Three questions recur in every topic, and a fourth runs underneath all of them.
Identification — under what conditions do the data and the model let us learn about the parameter of interest?
Estimation — how can we estimate it?
Inference — how do we quantify the uncertainty around the estimate?
Computation — how are the estimator and the inference procedure actually carried out?
The linear regression model is the first environment in which these questions are studied rigorously. From there it becomes visible that least squares, instrumental variables, maximum likelihood, GMM and panel estimators are not a collection of independent techniques: some estimators arise from optimisation problems, others from moment conditions, and nearly every method in the course is one or the other.
Structure
| Part | Lectures | Topic |
|---|---|---|
| I | 1–2 | The econometric model and the foundations of inference |
| II | 3–9 | Linear estimation |
| III | 10–12 | Endogeneity, instrumental variables, and GMM |
| IV | 13–14 | Nonlinear estimation |
| V | 15–16 | Panel data |
The full reading list, assessment structure and course policies are in the syllabus.
Texts
Hansen, Bruce E. (2022). Econometrics. Princeton University Press. — architecture and the unified view of estimation and inference.
Greene, William H. (2018). Econometric Analysis, 8th ed. Pearson. — formal development, terminology, and the large-sample and computation appendices.
Complementary: Davidson & MacKinnon (2004); Cameron & Trivedi (2005).
How to read these notes
The notes are formal: definitions, theorems and proofs. What surrounds them is meant to answer, before each result, what problem does this solve and why should we care.
Several recurring boxes mark the connective tissue:
Concepts belonging to the modern applied economist’s toolkit — moment conditions, the bootstrap, clustering. These appear where they unify something already covered, never as standalone novelty.
What the computer is actually doing: QR decomposition, Newton-Raphson, the within transformation, resampling.
Simulations that make visible a property already proved. Simulation does not replace theory here; it illustrates it.
Equations with highlighted terms are interactive — hover or click a term to see what it contributes to the result. The component reference demonstrates every construct used in these notes.