An Empirical Study Of The Influencing Factors Of The Shanghai Stock Returns Upon Threshold Regression Model

Learn from the literature, research on the impact of the stock returns factors can be considered from two aspects:the macroeconomic factor and the micro-economic factor. But there are a lot of macroeconomic factor and the micro-economic factor affect stock returns in the regression analysis, we hope to find those variables that have a major impact on stock returns, so choosing the major factor by variable selection method become an important research topic.In the past, the study of factors affecting the stock returns found that most of the scholars separately analysis the correlation of the macroeconomic variables and micro-economic variables upon stock returns.After careful thought, the idea of macroeconomic factors and micro-economic factors, taken together with the relationship between stock returns. However,how to make the two types of variables to be fusion into the model? This paper mainly start the research from the phenomenon of asymmetric about stock returns. We can learned from the phenomenon of asymmetry:there is the phenomenon of an asymmetric reaction by the market news (good news/bad news).The stock returns change is different.when received the good news, the stock returns with a curve; When received the bad news, the stock returns with another curve. In addition, the performance of the stock returns is also different in a different time period. Since phenomenon of asymmetry about stock returns, we want to accurately describe the the asymmetric features about stock returns, we can employ a nonlinear model to fit. The piecewise linear model can be able to seize such features. Based on this, we adopt the threshold regression model to analysis on the Shanghai stock market returns data. The empirical analysis through the use of a threshold regression model selected macro variables and micro variables and use of variable selection techniques selected macroeconomic variables and micro-economic variables. Threshold regression model, the use of information provided by the microscopic variables and macro variables at the same time, increase the amount of information available model. In the empirical study process, this paper employ the SCAD penalty smooth least squares estimation method that was put forward by Jiang Liang (2012).the advantages of this method is that the coefficient estimates and the estimate of the threshold simultaneously; the argument which has no significant will be simultaneously estimated zero, which greatly reduced the amount of computation. In the model, we use a linear combination of macro variables as the threshold variable, so that the comprehensive macro information is helpful for determining the changes of stock prices. Our empirical results show that only one threshold variable or threshold variable by a composite variable combination model,β3,,β2,β3still automatically determine into zero. That means, the number of stock trading, and stock transaction amount have no affect about the stocks returns. In the real estate industry, empirical results show that β1,β2,β3also be automatically determined into zero. Through the analysis of three actual data, we can see that the model can recognition the independent variable very well.