skip to main contentskip to main menuskip to footer Universität Bielefeld Play Search

Econometrics

Prof. Dr. Dietmar Bauer

© Universität Bielefeld

Econometrics at Bielefeld

Econometrics deals with the investigation of mathematical models of economics processes using tools from statistics.

The topics investigated in Bielefeld mostly draw from two areas of modern econometrics:

  1. Time series modelling: many (economic) processes evolve and are measured over time, forming time series observations.
  2. Discrete choice models: (economic) decision often involve choosing between a finite number of alternatives. These decisions are taken based on characteristics of the alternatives as well as of the deciders. Models taking these characteristics into account help in understanding the principles behind the corresponding choices.

Beside these two areas our lectures cover a large part of the spectrum of classical econometric analyses.

Research topics

Time series analysis

The main research topic in the area of time series econometrics in Bielefeld investigated linear dynamic state space models, which are in a certain sense equivalent to so called ARIMA-models (AutoRegressive Integrated Moving Average) of the Box-Jenkins methodology. Linear state space systems turn out to be more flexible than the rigid ARIMA models.

The main topic of interest for us is the development of estimation and inference procedures for such models. We investigate classical pseudo maximum likelihood approaches as well as more recent numerical procedures such as the subspace approach. We especially focus on multivariate time series, typically obtained in macro-economic settings.

In such processes a number of questions with respect to the modelling of so called unit root processes, showing trending, random walk like, behaviour, are still open, despite a long history of research. These questions have been the topic of the recent DFG-funded project "Estimation and Inference for Co-Integrated Processes (EICIP)".

Discrete choice models

The choice between a finite number of alternatives is often modelled using "random utility models (RUMs)". These models introduce the concept of an underlying utility function which deciders use to rank the alternatives. The utilities encode the characteristics of the alternatives, the preferences of the deciders but also include random disturbances the modeller does not or cannot model explicitly.

Depending on the specification of the distribution of these random terms one obtains "multinomial logit models" (MNL) or "multinomial Probit models" (MNP). These models are used in a great number of different contexts ranging from economics to transportation science.

Recent interest considers modelling of unobserved hetergeneity in preferences, typically modelled using random mixture models. In this respect our group investigates estimation and inference methods for randomly mixed probit models as well as the application thereof. This happens for example in the DFG-fundend project "Maximum approximate composite marginal likelihood (MACML)".

The research group is part of the Zentrum für Statistik, of the iTime and the Bielefeld Center for Data Analysis (Bicdas)

back to top