Over the last decade, the unit root and cointegration modeling framework has proved to be useful for the empirical analysis of macroeconomic risks. From the integration and cointegration properties of economic time series it is possible to infer the effects of aggregate fluctuations and shocks. For instance, the usefulness of different monetary policies hinges on the existence of stable long-run equilibrium relationships between monetary and other macroeconomic variables. Consequently, tests for unit roots and cointegration as well as proper model specification tools that take the properties of the data into account are necessary to adequately analyze the relationship between economic variables. Therefore, we develop new time series methods that allow an improved assessment of macroeconomic risks. Our project includes work in the following areas:
Unit root and cointegration tests. Structural breaks have important implications on the validity of unit root and cointegration tests as well as for the economic interpretation of the corresponding test results. Therefore, we aim at modeling several structural breaks and structural shifts with flexible forms.
Model selection and reduction. Vectorautoregressive (VAR) and vector error correction (VEC) models are models in which (too) many parameters have to be estimated from the data. Therefore, we analyze statistical procedures for model selection and model reduction. In particular, we focus on possible improvements with respect to forecasting precision and inference on impulse response functions.
Structural VEC models. The dynamic interaction between the variables of a VEC model are analyzed by impulse response functions. Identifying assumptions are needed to allow for an economic interpretation. Currently, the relation between structural restrictions for the identification of the impulse responses and the cointegration restrictions is typically ignored in applied studies. We study how this relation can be taken into account.
User-friendly software. We work on extending the software framework JStatCom. JStatCom provides a powerful graphical user interface for econometric methods based on GAUSS, Ox and Matlab code. JStatCom is therefore the ideal tool to quickly provide new statistical methods to applied researchers.