Factor-Augmented Quantile Regression Models With Predictors And Its Application In The Market

Factor model in the application of statistical analysis has a long history.As a dimension reduction techniques,it has attracted many scholars’attention in a data-rich environment. In the establishment of the factor model, the choice of the number of common factors is particularly important. Many econometrics and statistics scholars have made some research on this issue. Ando and Tsay proposed a information-theoretic criteriona which can identify the number of factors of quantile regression model with factor-augmented in 2011. In this paper, the accuracy of this criterion is verified by Monte Carlo simulation study based on the Empirical analysis. The results show that the proposed method is more suitable for the case which T and N are large ranther than the case which T and N are small and has an advantage in solving the problem of "dimension disaster".Quantile regression model with factor-augmented is applied to the research on the factors of stock return rate.The object of research is the SSE 180 stocks from 2007 to 2014. Based on the results of previous studies and the characteristics of China’s stock market,the firm size, book to market ratio, system risk, momentum effect and reversal effect are selected as five predetermined predictors and 36 other possible predictors. These 36 predictors that may affect the stock return rate are selected from the firms’financial data, market performance, macroeconomic, commodity prices, exchange rate changes and international index terms. In this paper, the panel quantile factor model is used to study the effect of the above predictors on the stock return rate. The results show that the factors of different time frequency can be different, but the significance of each factor is the same. On this basis, the paper establishes the common panel data model and the panel quantile model of five predetermined predictors and 36 other possible predictors,estimates the parameters of the two models and compares the results with the model. It draws that the five predetermined variables impact on stock yield under different quantiles is quite different and they are basically significant.The top two of the four factors that selected from 36 stock returns are particularly significant,and they represent the basic trend of macroeconomic and people’s consumption level.I hope that the research can provide a good reference for the future study on the impact of the stock return rate.