Essays on modelling and forecasting stock markets with investor sentiment

Sun, Chang (2021). Essays on modelling and forecasting stock markets with investor sentiment. University of Birmingham. Ph.D.

Preview

Sun2021PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (1MB) | Preview

Abstract

This thesis provides an analysis of the predictability of stock returns in ten European countries. First, we discuss the relationship between investor sentiment and stock returns in panel and individual levels in chapter 1. As investor sentiment is an subjective variable and is not easily to be observed directly, we compare all measures of investor sentiment employed in past literature to find the suitable proxy for it. It seems that only the consumer confidence index (CCI) is standardized across all European countries. Also we use macroeconomic factors as control variables since they can improve predictions of returns. We show that investor sentiment can positively affect the stock return in all of these ten European countries. In chapter 2, we use both the univariate models (ARMA, ARMAX) and the multivariate forecasting models (VAR, BAR) to examine the predictability of stock returns. Among these models, the ARMAX performs better in out-of-sample prediction. Although VAR models are usually better forecasters than the ARMA and ARMAX, in this case they disappointed the expectations. Also, the Bayesian VAR improves the standard
VAR and performs better in general. In chapter 3, we use three types of model averaging methods (SMA, BMA, AMA) to combine the predictions of individual models into a composite model to improve the predictive performance for stock returns. According to the out-of-sample results for averaged models, we can conclude that the forecasting performance of stock returns has been improved significantly by averaging individual models with different weights. Among these three types of model averaging methods, the Bayesian Model Averaging performs better than the other two methods in most countries.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):