| How do Asset prices react to changes of economic uncertainty?This thesis shows this question depends on properties of uncertainty."Bad" volatility indicates poor economic condition,such as decrease in investment activities,while "Good" Volatility is always seen as a sign of economic prosperity.In stock market,upside volatility becomes large when good news come,and bad news always resovle with increasing in downside voliatlity.In this sense,it's necessary to distinguish the different type of volatilities when we consider asset pricing issues.With fast development of internet and electronic technology,high frequency trading become more and more common in Chinese stock market.With enormous information and noise content in historical trading data,how to extract useful information becomes a very popular topic in this academic area.It's well known that the trading hours,quotation system and market structure in Chinese stock market are quite different from the US,therefore,we need to re-examine whether the theories in American stock markets still hold in China.Based on these backgrounds,this thesis collect the high-frequency quotation data for each stock trading in Chineses A-share market.We adopt methods of realized semivariance decomposition,continuous and jump decomposition,portfolio analysis and regression analysis to test the relation between volatility and stock return in Chninese stock market.The main finding and conlusions of this thesis are as follows:When we use realized volatility as risk factor to test the risk-return relation model in the Chinese stock market,we found the parameter of RV is insignificant in most situations,this contradicts the theory inference that high risk must indicates high return.We found two main reasons for this issue:(1)The realized measure contains both positive and negative intraday returns,it's not a purely risk measure;(2)The realized measure is contructed by intraday trading data,contains too much noise.Based on these motivations,this thesis decomposes the realized volatility into four realized factors,then test the information content in each of them.In stock index return prediction,we find significant improvement in economic implications after we include both good and bad realized volatilities as risk factors.We show that good volatility correlates negatively with future returns,while bad volatility correlates positively with future return in this market.We also decompose the realized volatility into continuous and jump parts,we found in most situation the continuous volialitiy has more information than the jump ones,this indicates that the investors in Chinese stock market are not well diversified when they invest.At last,we show the models with decomposed factors provide much better out of sample forecasting accuracy by constructing a simple switching strategy.We also examine the cross-sectional return predictability of high-frequncy factors.We found the prediction power of realized skewness and realized singed jump factors are strongly significant in Chinese stock market,and the predictability still exist after a series of robust tests.We then decompose the RS J foctor into good volatility and bad volatility factors,we found the RSJ effect mainly comes from the good ones.When we decompose RSJ into big and small jump parts,we found the big jump parts dominate the small jump parts.Therefore,we infer the RSJ effect in Chinese stock market is mainly caused by investors different preference for good and bad volatility and short sell constrains in Chinese stock market.Follow Bollerslev et al.(2019),we also test the relation between RSJ and RSK factors,we found that the larger part of RSJ,i.e.the RSJL factor dominate the RSK factor,this shows that our large and small decomposition of RSJ factor provide additional information in forecasting cross-sectional returns in Chinese stock market.The main contributions of this thesis are as follows:First,we found the issue that in Chinese stock market the realized volatility factor does not predict the future return negatively.By decomposing the realized volatility into risk(bad volatility)and preference(good volatilty)parts is a possible way to fix this issue;Second,we decompose the realized mearsure into four parts by good,bad,large and small highfrequency returns it contains,we found this decomposition improve the model both in economic implication and forecasting accuracy;Third,we test the RSJ factor in Chinese stock market,verified it's significant cross-sectional return predictability.Then we decompose this factor into large and small ones,we found the larger one,RSJL factor is much more efficient in cross-sectional return prediction,because the RSJL factor can completely substitude the realized skewness(RSK)foctor when we predict cross-sectional returns in Chinese stock market. |
PDF Full Text Request