Summary
The last decade has been characterised by a data revolution. In economics and elsewhere (physics, machine learning, biology, imaging, statistics) ever more data structures emerge that require new models suited for analysing multidimensional arrays of data, so called tensors (e.g., data of firm exposures (dimension 1) to other firms (dim.2) over time (dim.3) in different markets (dim.4)). Adequate econometric models for such data are currently largely lacking. They either simplify the problem to the 2-dimensional setting, or use models that are too static to account for rapid changes in economic conditions. STEADY fills this gap by developing new tensor models that account for the typical non-linear and dynamic features of economic data. STEADY concentrates on two main contributions: developing a general class of dynamic time-series models (tensor score-driven time-varying parameter models), and developing new tensor-based compression techniques for many economic time series (tensor dynamic factor models). The models developed will also be applicable in related fields. Both contributions of STEADY are applied to policy relevant questions for central banks and financial regulators, including forecasting multi-country, multi-market interest rate term structures for the evaluation of monetary policy effectiveness, and nowcasting multi-country economic activity in the heterogeneous European context. This is done by a close cooperation between the principal researcher, experts at VUA (host), and the European Central Bank (ECB). A secondment to ECB is key to the project, such that methodology, application, and implementation can be developed as a joint, cross-disciplinary effort between university and policymakers.
Unfold all
/
Fold all
More information & hyperlinks
| Web resources: | https://cordis.europa.eu/project/id/887220 |
| Start date: | 01-10-2020 |
| End date: | 30-09-2022 |
| Total budget - Public funding: | 175 572,48 Euro - 175 572,00 Euro |
Cordis data
Original description
The last decade has been characterised by a data revolution. In economics and elsewhere (physics, machine learning, biology, imaging, statistics) ever more data structures emerge that require new models suited for analysing multidimensional arrays of data, so called tensors (e.g., data of firm exposures (dimension 1) to other firms (dim.2) over time (dim.3) in different markets (dim.4)). Adequate econometric models for such data are currently largely lacking. They either simplify the problem to the 2-dimensional setting, or use models that are too static to account for rapid changes in economic conditions. STEADY fills this gap by developing new tensor models that account for the typical non-linear and dynamic features of economic data. STEADY concentrates on two main contributions: developing a general class of dynamic time-series models (tensor score-driven time-varying parameter models), and developing new tensor-based compression techniques for many economic time series (tensor dynamic factor models). The models developed will also be applicable in related fields. Both contributions of STEADY are applied to policy relevant questions for central banks and financial regulators, including forecasting multi-country, multi-market interest rate term structures for the evaluation of monetary policy effectiveness, and nowcasting multi-country economic activity in the heterogeneous European context. This is done by a close cooperation between the principal researcher, experts at VUA (host), and the European Central Bank (ECB). A secondment to ECB is key to the project, such that methodology, application, and implementation can be developed as a joint, cross-disciplinary effort between university and policymakers.Status
TERMINATEDCall topic
MSCA-IF-2019Update Date
28-04-2024
Geographical location(s)
