Identification and Inference in Structural Vector Autoregressive Models

• Applicant: Professor Dr. Ralf Brüggemann
• Subject Area: Statistics and Econometrics
• Term: since 2018
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 411824667

This project analyzes new methods for identification and inference in structural vector autoregressive (SVAR) models. SVAR models are widely used in applied economics to capture the joint dynamics of multiple time series and to investigate the effects of structural shocks. Results from this project will enlarge the toolbox of SVAR practitioners with new, theoretically well-grounded and valid methods. This is of high practical importance, as SVAR modelling is widely used for policy analysis in empirical macroeconomics and finance. Some of the methods used by practitioners still lack a thorough theoretical analysis or are not suitable for questions at hand. Therefore, this project develops and analyzes new methods for the structural analysis of multiple time series.In the first part of the project, we focus on the identification of structural shocks. Existing strategies for identification often use non-data information coming e.g. from economic theory. This strategy is often difficult to justify and not testable using data. Therefore, we analyze and develop new and alternative methods for identification of economically meaningful shocks. We focus on innovative identification methods that combine various existing techniques. Structural impulse responses are of core interest in SVAR analysis. In the second part of this project, we therefore analyze estimation and statistical inference on structural impulse responses. This includes the exploration of refined local projection estimators for impulse responses and the development of valid inference in this context. We also develop bootstrap inference in structural VAR models with cointegrated time series and heteroskedastic errors. Finally, we explore the consequences of using factor-augmented VARs when inference on impulse responses is of interest.

• DFG Programme: Research Grants