An Empirical Study of Pairs Trading in the Finance、Textile and Semi-Conductor Industry

• Main Authors: Tsai, Ching-Jen, 蔡景任
• Other Authors: Kuo, Wei-Yu
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/kdg8ub

碩士 === 國立政治大學 === 國際經營與貿易學系 === 106 === In May 2017, Taiwan Stock Index closed above 10,000 level and it lasted for more than a year till today. Not only did it create the longest bull market and 10,000 level record in Taiwan stock market, but it also intensified the confidence of investors. As daily trading volume keep at over 100 billion NTD level and the stock price is getting higher and higher, the questions of how to invest smartly become hot and popular again. How is the subjective way to choose our investing target to gain more profit? And by passive investing, which country and what kind of index should we do the long investing? On top of the two methods mentioned above, based on the statistic arbitrage, pairs trading is another good strategy that has been used all over the world. During the past 6 years and the bull market time, does pairs trading still maintain its profit-making ability in Taiwan stock market is our main research motivation. We applied pairs trading into three main industry of Taiwan, which is Finance, Textile and Semi-Conductor. We chose the adjusted stock price as our data sample and turned the adjusted price into normalized price (starts at 1). By comparing each normalized price in each industry, we picked up the top 20 pairs with lowest SSDs (Sum of Square Deviations) and the top 20 pairs with the highest number of zero crossing among the 50 pairs with the lowest SSDs. By introducing zero crossing, we hope that zero crossing can increase the convergence and profit of pairs. From March 2, 2010 to February 26, 2016 is our sample period, and we divide sample period into four sub period. Each sub period lasts for 1.5 year, the first year is the pairs formation period, and the next six months is pairs trading period. At the very first day of trading period, we normalized the adjusted price into normalized price again and made the price of first trading day starts at 1. Whenever the spread of the normalized price in trading period diverge by more than half and one standard deviations from the historical spread observed over the formation period, we take that as trading signal and create our trading position. Empirical results show that the return with lowest SSDs have better performance than the return with highest number of zero crossing in Finance and Textile industry. And it turns out that with highest number of zero crossing in Semi-Conductor industry have optimized the return. Under the different trading signal circumstances, in the Finance and Textile industry the half standard deviations signal can gain more profit than one standard deviations. To the Semi-Conductor industry, half performance of half standard deviations is better and half return of one standard deviations is better. From the perspective of whole stock market, Finance and Textile industry get profits at the bear market time, however, the Semi-Conductor can gain more returns at bull market period.