Can Continuous Beta Or Discontinuous Beta Predict Stock Returns?

The pricing and valuation of the securities market has been the focus of attention of researchers and investors.From the perspective of the relationship between the return rate and systematic risk,the researchers have gradualy constructed the theoretical framework of the CAPM.At the beginning of the formation of capm model,the positive significance of beta was higher than that of beta model.However,subsequent studies have been unable to verify the correctness of the model.The prediction of return rate by risk is not satisfactory,forcing researchers to find a new way out.The researchers looked for solutions in roughly two directions,one continuing to study systemic risks and undisclosed information,while others chose to look for new explanatory variables or combinations of explanatory variables and try to construct a new theoretical system.when study of the relationship between rate of return and system risk,scholars either use high-frequency data processing,construct new estimation methods,or find new explanatory variables,but few combines them.This paper constructs the continuous volatility risk of the trading day,jump risk of the trading day and the overnight jump risk,uses the estimate of the CAPM model for reference,and uses 60 minute high frequency data to calculate the four beta of al shares from July 1999 to December 2017.Considering another branch of the relationship between system risk and rate of return,this paper referres to the domestic and foreign literature,chooses firm size,reversal,volatility,earnings per share,coskewness,cokurtosis,realized skewness,and realized kurtosis.This eight explanatory variables are compared with beta explanatory ability and predictive ability on yield.After excluding STS,new shares,long suspension and other stocks with different trends from the stock market as a whole,the beta of al A-shares is calculated.In order to compare the difference of prediction ability between the four,this paper uses the univariate grouping method to divide al A-shares into five portfolios according to the size of the four estimated values.Empirical results show that whether in a bull market or a bear market,with the increase of beta,the return on each portfolio is increasing.In general,since the fifth group is the largest portfolio of beta this month,there will be a correction next month,resulting in a return of less than the fourth group.In a bull market,portfolio return decreases with the increase of overnight beta,and the impact of overnight information on stock prices is more likely to affect smal overnight beta firms.In order to test the stability of beta and the advantages and disadvantages of-the other explanatory variables and the interpretation ability of beta,the bivanate grouping method is used in this paper.Taking the fundamental index as the first grouping basis,the beta as the second grouping basis to test whether the interpretation ability of beta is affected by the companyundefineds fundamental index,and the beta as the first grouping basis,the fundamental index is the second grouping to test the superiority and inferiority of beta and other explanatory variables.The results of bivariate grouping show that the prediction ability of the beta is stable,which is consistent with the univariate case,and the yield of the portfolio increases with the increase of beta.In order to fully compare the explanatory ability of beta and other explanatory variables,the paper uses simple regression and multivariate regression to construct 15 cross-section regression,and compares its significance and coefficient.The simple regression analysis showed that the coefficients of standard beta,continuous beta and discontinuous beta were positively discontinuous beta,and the coefficients of earnings per share,consistency bias and kurtosis of individual stock were significantly higher than those of other explanatory variables.Multivariate regression shows that the significant level of standard beta has declined,and its predictive ability is partly diluted into the other explanatory variables.The three beta constructed in this paper are more accurate in predicting stock returns,and are less disturbed by other explanatory variables.Except for consistency bias,the significance of other explanatory variables decreased compared with simple regression.The article classifies listed companies according to whether they are listed in two markets,whether they are state-owned enterprises or institutional shareholdings,and then separately examines the change in forecasting ability.The results show that the double-board listed companies have difficulty in appearing because of the mutual interest of the two stocks in the two markets;the excess returns of non-state-owned enterprises are significantly higher than the state-owned enterprises,and the non-state-owned enterprises have the highest excess returns.The combination is generally the third and fourth groups,while the state-owned enterprises are the fifth group;the different shareholding ratio of institutional investors will have a different impact on the A-share market.When the institutional shareholding ratio is less than 30%,the combined excess returns The rate increases with the increase;when the institutional shareholding ratio exceeds 30%,the positive relationship between excess return rate and the beginning begins to weaken;when the institutional shareholding ratio exceeds 50%,when using the selected portfolio,it will no longer be used.Get positive excess returns.Finaly,considering that the influence of in-day sampling frequency was neglected in estimating the risk of four different systems,the in-day sampling frequency was revised to 5-minute,30-minutes,80-minute and 120-minute respectively.The sensitivity of beta to in-day sampling frequency was observed.At the same time,the paper also assumes that beta will remain unchanged for one month and limit al stock holding period to 1 month.In order to eliminate this effect,the holding period is revised to 2 months,3 months and 6 months,so as to construct the stability test.The empirical results show that the beta is not sensitive to the in-day frequency of sampling and still maintains its original characteristics,but the beta will change with the change of the holding period,and the change of the holding period will not change the characteristics of high beta and high return.But as the holding period increases beta yields will fal and the excess return gap between portfolios will continue to narrow.