Exploring Bubble Informatics: Representation and Identification of Stock Bubbles

探索泡沫資訊學:描述和確認股市泡沫

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

Related Research Unit(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date21 Mar 2018

Abstract

Bubbles are generally characterized as a significant deviation of an asset’s price from the asset’s fundamental value, with a growth and burst cycle. Despite there exist considerable amount of research on asset price bubbles, the bubble characteristics have not been well understood. In response to the current broad application of information technology, we propose a concept called bubble informatics with the aim of applying information technology to advance understandings of asset price bubbles. Specifically, focusing on different stages during the boom–bust cycle, we design, develop, and evaluate information artifacts to describe the bubble stage evolution process, explore the bubble forming mechanism, and identify the bubble stages.

Previous studies have not focused on empirical analysis of bubble characteristics description. In addition, they have not clearly identified the bubble stages within the boom–bust cycle, which are essential for monitoring and regulating various asset bubbles. To fill these gaps, we conduct two studies to resolve these problems. First, we explore ways to describe bubble characteristics. Specifically, under the background of Chinese stock market, we use the lead–lag relationship among bubble features to describe how the bubble characteristics evolve through the boom–bust cycle. The research results show that both speculative motive and leverage play significant roles during the 2015 Chinese stock bubble booming process. In addition, the deleveraging policy can be viewed as an essential external shock to the market, which causes the market to crash and creates a frustrated negative effect. As well as analyzing bubble characteristics, we aim to help regulators and investors identify bubbles before they burst. Instead of using econometric bubble detection and tests based on the efficient market hypothesis, in the second study, we propose using machine learning method to identify different bubble stages, which can hint at bubble stage identification. The research results indicate that machine learning can be successfully used to identify different bubble stages, especially the mania and burst stages.

This dissertation contributes to the literature by proposing a novel research concept for future bubble research, describing the market bubble characteristics, and developing a method to identify bubble stages. Theoretically, it provides a common language for bubble research, and indicates a new direction for future research. Practically, it provides valuable information for investors and regulators to understand possible stock bubbles in a comprehensive manner. Specifically, this work gives regulators guidelines to create market regulation policies, and instructs regulators to enact polices when appropriate. In addition, this work helps new investors make investment decisions.

    Research areas

  • Bubble description, bubble stage identification, bubble informatics, lead–lag relationship, machine learning