| Course language | Semester | Hours per week | ECTS | Exam |
|---|---|---|---|---|
| German | Winter semester | 2L+2E | 6 | Intermediate exam (30 minutes) Final exam (90 minutes) |
Contents
The analysis of panel data and data with a limited domain (limited dependent variables) has become an integral part of empirical economic research. Panel data refers to individual time series data available for many units or subjects. If this data originates from individual economic subjects, it is often referred to as microdata. An important example of dependent data with a limited domain are discrete variables.
Students learn about different estimation methods for panel data and limited dependent variables and how to select and apply these appropriately according to the data properties. They learn the relevant basics of statistical and econometric theory and their practical application based on the freely available software R.
In the course Advanced Issues in Econometrics , the econometric toolbox is further expanded. Firstly, the basics of causality analysis and evaluation studies are discussed. Subsequently, methods for analysing panel data are introduced. If one assumes that the units in question are based on a common dynamic, this can often be estimated efficiently with the help of panel data models. Panel data models play a central role in modern empirical economics. However, when it comes to the empirical specification of economic equilibrium models, techniques for analysing simultaneous systems of equations are necessary. In Advanced Issues in Econometrics, central principles are taught, in particular the instrumental variable estimator (IV estimator) and simple conditions for identifying the model parameters at hand. Finally, estimation techniques and their properties are discussed, which enable an efficient analysis of discrete dependent variables. This includes, in particular, yes-or-no decisions. An example of this is the determination of the influencing factors that lead to the supply of one's own labour on the labour market. The logit and probit models commonly used for this purpose are introduced, as well as the maximum likelihood estimation required to estimate these models. Models for truncated or censored dependent variables are also of great importance. The latter exist when the actual value of interest cannot always be observed (e.g. income above a certain threshold).
STRUCTURE
- Basics of causality and evaluation studies (controlled random experiments, evaluation without random experiments)
- Pooled cross-sectional data (difference-in-differences estimator, average treatment effect)
- Panel data (unobserved heterogeneity, difference-in-differences estimator, fixed effects estimator, random effects estimator)
- Instrumental variable estimators and two-stage least squares estimators (identification, order and rank condition, model assumptions and asymptotic estimation properties, endogeneity tests)
- Simultaneous equation models and multi-equation systems (structural equations, reduced form, identification)
- Maximum likelihood estimation (log-likelihood function, asymptotic properties, numerical optimisation, likelihood ratio test, Wald test, Lagrange multiplier test)
- Models for binary dependent variables (logit and probit models)
- Models for truncated or censored dependent variables (Tobit models)
- Sample selection bias (Heckman estimator)
LITERATURE
Angrist, J. and Pischke, J. (2009), Mostly harmless econometrics. An Empiricist's Companion. Princeton University Press (Chapters 1-3).
Wooldridge, J.M. (2009 - or newer). Introductory Econometrics. A Modern Approach, 4th edition, Thomson South-Western (Chapters 13-17, Appendix B, C).
TARGET GROUP / REQUIREMENTS
The course is aimed at Bachelor students in the 2nd study phase who have already attended the course Introduction to Econometrics.
