Use this URL to cite or link to this record in EThOS:
Title: Essays on macroeconomic analyses with factor models
Author: Yamamura, Taiki
ISNI:       0000 0004 8503 9210
Awarding Body: Queen Mary University of London
Current Institution: Queen Mary, University of London
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Access from Institution:
The thesis contains three essays, which are related to macroeconomic analyses with factor models. In Chapter 1, I investigate time variations in the monetary policy effects on the economy in Japan, by using time-varying Factor Augmented Vector Autoregression (FAVAR) model. The main interest is whether and how the policy effects change due to the Bubble Burst and during the (near-)zero interest rate period. As an analysis methodology, I propose and adopt the following two-stage procedure. In the rst stage, the shadow rate is estimated by a non-linear term structure model, where the shadow rate represents a policy stance of the monetary policy authority during the (near-)zero interest rate period. Using the estimated rate as a policy instrument, the second stage estimates the time-varying FAVAR model. Chapter 2 investigates the performance of time-varying FAVAR in terms of whether it correctly captures time variations in monetary policy transmission to macro-economy. The analysis is conducted through Monte Carlo (MC)-based experiments, and the model's performance is examined in comparison with that of time-varying VAR which does not use unobserved factors. The experiments show that the time-varying FAVAR adequately detects the time variations even under a situation where the time-varying VAR fails to do this. Using the computation techniques proposed in the recent literature, the above result is interpreted in terms of the information sufficiency of those two empirical models. In Chapter 3, my attention moves to an estimation of factor model with machine-learning approach. The chapter is devoted to proposing a novel method to identify a grouped factor structure by introducing a l1-constraint (Lasso approach) of the pair-wise difference of the factor loadings. Through theoretical analyses including Monte Carlo experiments, the advantage of the Lasso-based method is revealed over the existing methods.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: Factor Augmented Vector Autoregression ; macroeconomic analyses ; machine-learning