Identifiability and Estimation in Structural Vector Autoregressive Moving Average Time Series Models

-) We plan to develop SVARMA models driven by non-Gaussian and independent shocks that use only data to identify the economic shocks driving the model and that allow for the informational structure of fiscal policy models.
- ) We plan to develop stochastically singular models. Their applications are twofold: On the one hand these models are important for capturing a large number of variables (connected to the “big data” paradigm) and on the other hand they serve as benchmark models for stochastically singular DSGE models.
-) We plan to improve on existing solution methods for DSGE models as well as to improve on methods for assessing their fit to observed data.

Macroeconomists by and large agree that the aggregated observable variables like GDP (growth), rate of unemployment, or the rate of inflation are driven by unobservable shocks. The way in which the observable variables depend on the unobservable shocks is modeled with so-called structural models, which incorporate behavioural assumptions of the economics agents.

Often, structure is imposed (or in other words the model class is restricted) by economic theory, the so-called story-telling approach. The story telling approach can be criticised for being subjective.

This project advocates a data-driven approach to identify the relationship between the observable aggregates and the underlying shocks, rather than story-telling. Ideas as to how to squeeze out more information of a given dataset (co-developed by the Research project "Structural Analysis of Non-Gaussian Macroeconomic and Financial Time Series" at the University of Helsinki) are extended and further refined. In particular, this research project aims at evaluating the story-telling approach with an approach that only uses information contained in the data.