An Empirical Study of CAPM in Taiwan Stock - Application of Hierarchical Linear Modeling

• Main Authors: Chuang, Tsung-Han, 莊宗瀚
• Other Authors: Ding, Cherng
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/t995g4

碩士 === 國立交通大學 === 經營管理研究所 === 107 === In the CAPM, the beta coefficient represents the investor's prior subjective cognition of the future system risk of the risk asset. However, the beta coefficient used in practice is estimated by OLS method from the historical data of the risk asset, called the market model. The research considers that past studies use the market model to estimate the beta coefficient. This estimation method ignores the problem that the data is a nested structure. It will violate traditional statistical methods, such as the independence assumption of OLS method. The data structure is a multi-level structure, which should be divided into time-level and firm-level. Therefore, the research proposes to use the Hierarchical Linear Model to deal with the problem of nested structures and to estimate. Based on past research, this study tries to find variables that have the ability to interpret stocks' expected returns by using a hierarchical linear model. It also uses to improve the market model. Help investors to get the trend of individual stock beta coefficient over time in the CAPM, find out the relationship between market returns and expected returns. The market index with the most representative market rewards, and provides specific recommendations as an investment in investment analysis or investment strategy.