Big Data Analysis of Individual Investors Behavioural Biases in China

Abstract

This thesis analyses biases on individual investors selling decisions in the stock market from a very large and unique dataset from China. The dataset contains 100,000 individuals with more than 56 million daily holding records from January 2007 to May 2009. This thesis introduces the Chinese stock market and the Chinese investors in Chapter 1, and summarize the previous and current literatures in Chapter 2, then perform empirical tests in the following chapters. In Chapter 3, I find that investors hold a preference on realizing large gain while no preference on loss, while this result is affected by the holding period and the market condition. More specifically, for short-term holding positions, individuals prefer to sell both large gain and large loss. But this effect reverses in the medium- and long-term. Regarding the different market scenarios, Chinese investors are more likely to sell large gains in all conditions but only prefer to realize large loss under booming market. Therefore, the confidence level to the market could be reasons for individuals to realize loss. In Chapter 4, I investigate the investors’ selling decision within their portfolios, and find that individual investors in China are more likely to sell a position with extreme good (the best) performance, and followed by the 2nd best position, but reluctant to sell the salience of extreme bad position, which is different from result in the US (sell both best and worst positions). When lottery-like positions are held by young, male and new investors, Chinese investors also sell the worst rank positions in their portfolios. Thus, the willing of gamble could be a reason for individuals to realize relative underperforming positions. This result is robust under different modelling method, extreme portfolio situation, measurement of rank and limit-down limitation, etc., and consistent in different market condition and holding period positions. In Chapter 5, this study further tests how geographic factors impact selling biases above. Metropolis investors suffer more from disposition effect. Investors from different regions in China do not have different degrees on these effects. Furthermore, individual investors in China do not have significant preference on local stocks. Local stocks cannot influence disposition effect and V-shaped disposition effect. However, local position can moderate the preference of selling relative overperforming positions (rank effect in China). Since rank effect is harmful to profit when the position is local, local stocks benefit to profit to some degree. This thesis further indicates that some investor sophistications (e.g. experience, trading frequency) can moderate the individual’s selling biases. All results are concluded in Chapter 6