Year of Award

2016

Degree Type

Thesis

Degree Name

Doctor of Philosophy (PhD)

Department

Department of Economics.

Principal Supervisor

Ng, Ying Chu

Keywords

Business cycles;China;Econometric models;Economic forecasting;Hong Kong;Hong Kong;Housing forecasting;Housing price;Macroeconomics

Language

English

Abstract

This dissertation consists of two essays. The first essay focuses on developing a quantitative theory for a small open economy dynamic stochastic general equilibrium (DSGE) model with a housing sector allowing for both contemporaneous and news shocks. The second essay is an empirical study on the macroeconomic forecasting using both structural and non-structural models. In the first essay, we develop a DSGE model with a housing sector, which incorporates both contemporaneous and news shocks to domestic and external fundamentals, to explore the kind of and the extent to which different shocks to economic fundamentals matter for driving housing market dynamics in a small open economy. The model is estimated by the Bayesian method, using data from Hong Kong. The quantitative results show that external shocks and news shocks play a significant role in this market. Contemporaneous shock to foreign housing preference, contemporaneous shock to terms of trade, and news shocks to technology in the consumption goods sector explain one-third each of the variance of housing price. Terms of trade contemporaneous shock and consumption technology news shocks also contribute 36% and 59%, respectively, to the variance in housing investment. The simulation results enable policy makers to identify the key driving forces behind the housing market dynamics and the interaction between housing market and the macroeconomy in Hong Kong. In the second essay, we compare the forecasting performance between structural and non-structural models for a small open economy. The structural model refers to the small open economy DSGE model with the housing sector in the first essay. In addition, we examine various non-structural models including both Bayesian and classical time-series methods in our forecasting exercises. We also include the information from a large-scale quarterly data series in some models using two approaches to capture the influence of fundamentals: extracting common factors by principal component analysis in a dynamic factor model (DFM), factor-augmented vector autoregression (FAVAR), and Bayesian FAVAR (BFAVAR) or Bayesian shrinkage in a large-scale vector autoregression (BVAR). In this study, we forecast five key macroeconomic variables, namely, output, consumption, employment, housing price inflation, and CPI-based inflation using quarterly data. The results, based on mean absolute error (MAE) and root mean squared error (RMSE) of one to eight quarters ahead out-of-sample forecasts, indicate that the non-structural models outperform the structural model for all variables of interest across all horizons. Among the non-structural models, small-scale BVAR performs better with short forecasting horizons, although DFM shows a similar predictive ability. As the forecasting horizon grows, DFM tends to improve over other models and is better suited in forecasting key macroeconomic variables at longer horizons.

Comments

Principal supervisor: Dr. Ng Ying Chu.;Thesis submitted to the Department of Economics.Thesis (Ph.D.)--Hong Kong Baptist University, 2016.

Bibliography

Includes bibliographical references (pages 141-148)

Available for download on Saturday, July 22, 2017


Share

COinS