Économétrie 1 (EN)

Objectif

This course aims to study linear regressions and introduce the notion of causality in the social sciences. It shows how and under what conditions causal parameters can be identified and estimated from random samples. The distinct question of prediction will also be addressed. Particular emphasis is placed on the method of least squares, with ordinary least squares (OLS) and two-stage least squares (2SLS) estimators.

Main learning outcomes: By the end of the course, students will be able to:

• Understand the basic properties of OLS.

• Make predictions using OLS; select a model that ensures “good” prediction.

• Know the asymptotic properties of the OLS estimator, construct hypothesis tests and confidence intervals based on this estimator.

• Understand under what conditions the OLS estimator identifies a causal effect.

• Recognize a situation of endogeneity and, using the two-stage least squares estimator, apply a suitably chosen instrumental variable. Correctly interpret the corresponding causal effect in the case of an instrument and a binary “treatment” (LATE).

• Carry out a complete econometric analysis with real data in R (specification, choice of estimation method, tests, etc.) and interpret the results.

Assessment:
Continuous assessment (1/3), final exam (2/3).

Plan

1. Fundamentals of linear regressions: definition of OLS, finite-sample properties, regression quality, convergence of OLS, interpretation of theoretical regressions.
2. Precision in linear regressions: asymptotic properties of OLS, accuracy of OLS and predictions, choice of regressors, tests and confidence intervals.
3. Non-causal predictions: bias-variance trade-off, cross-validation, penalized regressions (information criteria, Lasso and ridge regressions).
4. Link between linear regressions and causality: Rubin’s causal model, selection bias, causal models and linear regressions.
5. Instrumental variables: motivation in the case of randomized experiments, generalization to linear models, two-stage least squares estimator, inference.

Références

Angrist, J. D., et Pischke, J. S. (2008). Mostly Harmless Econometrics: An Empiricist's Companion. Princeton university press. 
Wooldridge, J. M. Introductory Econometrics.  Thomson South-Western (2003).
Wooldridge, J. M. (2010). Econometric Analysis of Cross Section and Panel Data. MIT press.