ENSAE Paris - École d'ingénieurs pour l'économie, la data science, la finance et l'actuariat

Econometrics 2

Objective

This course aims to complement the first-semester econometrics class in the field of statistical analysis of individual-level data.

The first part of the course covers extensions of linear models introduced in Econometrics I. These extensions are designed to address the problem of endogeneity, which lies at the core of modern microeconometrics. The course provides an in-depth treatment of instrumental variables, including issues of heteroskedasticity and weak instruments. It also considers the use of panel data, even in cases where model dynamics are complex. Methodologically, this section relies on the generalized method of moments (GMM).

The second part of the course focuses on models with limited dependent variables. This includes discrete variables (e.g., unemployment indicators, loan repayment, health status, transport choices), as well as censored variables such as consumption, which takes strictly positive but potentially zero values. Selection problems (e.g., labor supply, endogenous sample selection) are also addressed. For these nonlinear models, the primary estimation method studied is maximum likelihood.

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

  • Master the theory and practice of instrumental variables, and understand their limitations.

  • Distinguish among different panel data methods, and evaluate the advantages and limitations of each approach.

  • Identify situations with limited dependent variables and choose the appropriate model (logit/probit, multinomial or conditional logit, tobit, …), understand the estimation method, and the specificities of each model.

  • Understand and apply the generalized method of moments (GMM) and maximum likelihood estimation in contexts similar to those studied in class.

  • For all models and methods covered: conduct a complete econometric analysis with real data using STATA (model specification, choice of estimation method, testing, etc.) and interpret the results.

Assessment

  • Final written exam (2/3)

  • Midterm exam (1/3)

Planning

1. Linear Models: Extensions

1.1. Instrumental Variables

  • Review: endogeneity, 2SLS estimator, Hausman and Sargan tests

  • Instruments under heteroskedasticity, weak instruments

1.2. Static Panel Models

  • Review: fixed effects and random effects, first-difference and within estimators

  • Efficient inference in the presence of autocorrelation

1.3. Dynamic Panels and Weak Exogeneity

  • Estimation using the Generalized Method of Moments (GMM)

2. Nonlinear Models: Limited Dependent Variables

2.1. Dichotomous Models

  • Logit and probit models: identification, estimation, model fit, issues of heteroskedasticity and endogeneity

2.2. Ordered and Unordered Polytomous Models

  • Multinomial logit model, conditional logit model

2.3. Count Models

  • Poisson model

2.4. Censored and Selection Models

  • Simple Tobit model, generalized selection model, truncation model

References

AMEMIYA, T. Advanced Econometrics, Basil Blackwell, Oxford, 1989 [28 AME 00 A]
CREPON B et N. JACQUEMET Econométrie : Méthode et Applications, de Boeck
GOURIEROUX C. Econométrie des variables qualitatives, 2ème éd., Economica, 1989 [28 GOU 00 A]
WOOLDRIDGE, J. Econometric Analysis of Cross Section and Panel Data, 1ère ou 2ème éd., MIT Press, 2002 ou 2010 [28 WOO 00 B]