An Empirical Study On The Overnight Rate Of Return Based On Non-parametric Models

With the rapid development of Chinese economy,more and more individuals and institutional investors hope to preserve and increase their assets in the capital market.So it is particularly important to find out the risk factors that have a significant impact on asset returns in the capital market.Since 1980,market anomalies such as "size factor","momentum effect" and "volatility effect" had appeared.These anomalies had brought changes to the pricing of financial assets.Many scholars use portfolio investment method to study the effects of these anomalies on monthly,quarterly and annual returns,and get good conclusions.But in real life,many investors invest in a short period of time,such as five days,three days or less.So what factors will affect asset returns in the case of a short investment period?In this paper,it is a impotant task to select the risk factors which to explain overnight returns and to fit the overnight returns through the selected data.Firstly,this paper uses F-M regression method and panel data model estimation method.The overnight returns of China’s gem market from February 2014 to December 2015 are researched in this paper,and the remarkable explanatory ability of the size factor,momentum factor,volatility factor and P/E effect are confirmedSecondly,the linearity of the hypothesis model uses F-M regression and panel data estimation methods.But in reality,there is no linear relationship between overnight returns and volatility factors,even we don’t know their total distribution.Therefore,this paper introduces a nonparametric model,and gives the estimation of the nonparametric model based on kernel estimation and so on.The consistency and convergence rate of this method are proved in this paper.Finally,the results of the linear regression estimate and nonparametric regression estimation are compared.Empirical results show that fitting the overnight returns using nonparametric regression estimation has better effect.