In general, as well as for the first topic, vector autoregressive models, the presentation develops as follows: posing the problem, model specification, listing the assumptions, interpretation, estimation, inference, model selection, model sensitivity, e.g. impulse response analysis, misspecification analysis and empirical applications.

GARCH orgeneralized autoregressive conditional heteroscedasticity models are able tocapture the stylized facts of financial return data, that periods of highvolatility are followed by periods of low volatility.

VEC or cointegrated VAR allow for cointegrated series leading to non standard estimation procedures and inference. The model interpretation will stress the distinction between long run properties and short run adaption.

The class of state space models comprises the ARIMA class and helps to understand the dynamics of multivariate dynamic processes. The flexibility of the class will be demonstrated and the estimation and forecasting steps explained. Empirical examples will cover e.g. volatility modelling.

High frequency data have some distinct properties like irregularly observed values, measuring an underlying continuous processes (with/without jumps), different observational frequencies among different series, pronounced intraday "seasonality". For the latter, the concept of realized volatility will be introduced and illustrated for Microsoft data.

Panel structures allow the simultaneous modelling of cross-section and time series data. In particular, we will deal with constant or hardly changing information over time on the individual level. We will consider specification, estimation and inference of fixed and variable effects and dynamic models. Empirical examples are provided.

Slides and exercises at http://statmath.wu.ac.at/~hauser/LVs/FinEtricsQF/

• to understand and apply more specific methods for modeling data of financial markets.
• to apply a selection of frequently used procedures for financial data, covering VAR (vector autoregressive models), VEC (vector error correction models), Kalman filter and state space models, characteristics of high frequency data, analysis of static and dynamic panels.
• to interpret the output of empirical estimates.

Participants will be trained in

• manipulating formulas,
• reading and executing R scripts
• interpreting the results of small empirical projects

when doing the assignments and presenting them in class.

Exams cover theoretical aspects, deeper understanding as well as empirical applications.

Attendance is mandatory in at least 80% of the lectures, i.e. in at least 7(9) of the units reserved for lectures. The attendance during both tests is highly recommended. 

Criteria for the investigation of empirical relationships are the ability to pose a problem, to discuss different methodological approaches, to choose a favorite model, to justify the statistical assumptions, to interpret the output and to conclude wrt the posed problem.

Criteria for the formal exercises are the ability to read formula and to do simple proofs.

The computation exercises are to be presented in a clear way and the single steps explained.

Criteria for the written exams are the ability to apply the methodology apart from the understanding of the models and their statistical properties.

The contributions to the grade are

• 30% homework assignments and classroom presentations
• 40% mid-term exam
• 30% final exam

The minimum percentage for passing is 51%.