| Summary: | 碩士 === 中華大學 === 資訊管理學系碩士在職專班 === 100 === The purpose of this paper is to explore the relationship between the industries as well as total market value and stock selection model in U.S. stock market. In this study, the stocks are divided into 10 industrial categories. The stock selection concepts include: large book value-to-price ratio (B/P), large return on equity (ROE), large revenue to price ratio (S/P), large quarter rate of return, large total market value, and small risk factor (beta). Using Compustat North American version database, we backtested the performance of various stock selection models for different industries and size. The backtest period is from 1990 Q4 to 2010 Q3 (80 quarters). The results showed (1) Regardless of industry, the performance, such as annualized rate of return, alpha, absolute wining rate, relative winning rate, of stock selection models based on the large B/P, large ROE, and large S/P concepts are significantly better than those based on the other three concepts. Overall, the large B/P, large ROE, large S/P concepts can form investment portfolios with better performance. When we divided stocks into ten industrial categories, the results were similar. (2) The quadratic coefficients, B/P*ROE, ROE*S/P and B/P*S/P, of rate of return regression models of various industries are significant, showing there are interactions between them. The interactions can increase return and reduce risk; therefore, the multi-factor stock selection models are superior to the single-factor ones. (3) The effects of prediction models of the annual rate of return and systematic risk are more or less persistent, and the phenomenon interactions can increase return and reduce risk are also more or less persistent. (4) The results of cluster analysis using performance of stock selection models and quarterly return showed that the clusters in stocks are not obvious, and stocks can only be roughly divided into 4 clusters: raw materials cluster, industrial product cluster, unnecessary consume cluster, and information industry cluster. (5) The analysis of monthly rate of return of SP500 showed that the month effect concentrat in from March to May and form November to December, and the performance is poor in other months. There is no January effect in SP500. The analysis of monthly rate of return of various industry stocks showed that the month effect also concentrat in from March to May and form November to December. There are some January effects in Medical &; Health, Information technology, and Telecom industry. (6) When dividing the stocks into small-cap, middle-cap, and large-cap, the single-factor stock selection models based on the large B/P, large ROE, and large S/P got the best annualized return. The annualized rate of returns of the multi-factor stock selection models was significantly better than those of single-factor stock selection models. The small-cap stocks prefer the large B/P investment concept, that is, they prefer value stock strategy; on the other hand, the large-cap stocks prefer the large ROE investment concept, that is, they prefer growth stock strategy. The profolio based on the weights, (B/P, ROE, S/P) = (1/3, 1/3, 1/3), got higher annualized return on average, and was rather robust; therefore, it is the best investment strtagy.
|
|---|
