| After more than 30 years of rapid development,China’s stock market has become the second largest stock market in the world,which has attracted widespread attention from domestic and foreign investors and the researchers.The Chinese stock market has many unique characteristics,such as the dominance of retail investors(Jiang,Lu &Zhu,2014;Zhu,Wang & Zhang,2022),strong gambling preferences(Lu,Chen & Li,2021;Liu et al.al.,2022)and strict limits of arbitrage(Gu,Kang & Xu,2018;Xiao,Zhao & Fang,2018).These characteristics significantly affect the performance of asset pricing anomalies in the Chinese stock market.In order to diversify and invest effectively,investors need to have a profound understanding of the characteristics of China’s stock market.Therefore,in recent years,there have been a large number of research studying the cross-sectional asset pricing anomalies in China’s stock market(Liu,Stambaugh & Yuan,2019;Leippold,Wang & Zhou,2022;Hou,Qiao & Zhang,2023).Many anomalies do have completely different performances in the Chinese market and the international market,which can be attributed to the unique systematic and market characteristics of the Chinese stock market.For example,Liu,Wu & Zhu(2022)found that the overreaction of the retail investors to the daily limit event is the reason for the disappearance of the momentum effect in the Chinese stock market.If the daily returns hitting the price limits are excluded when constructing momentum,momentum effect revisits the Chinese stock market.The research regarding the asset pricing anomalies in the Chinese stock market can not only guide investment practice,but also greatly enrich the empirical asset pricing literature.The topic of this thesis is left-tail risk anomaly and salience effect in China’s stock market,which are currently two hot research topics in empirical asset pricing.The left-tail risk of stocks is usually measured by the Value at Risk(Va R)of stocks’ historical returns.Atilgan et al.(2020)finds the left-tail risk anomaly in the US stock market: the left-tail risk of stocks measured by Va R has a negative relationship with the stocks’ future returns.However,the evidence for this anomaly in the Chinese stock market is inconsistent: some studies have found that the left-tail risk effect exists(Zhen,Ruan & Zhang,2020;Wang,Xiong & Shen,2022)while others believe that it does not exist(Gui & Zhu,2021).The second and third chapters of this thesis explore why the left-tail risk anomaly is not robust in the Chinese stock market from different perspectives,and propose two modified left-tail risk measures: idiosyncratic left-tail risk(IVa R)and extreme downside volatility(EDV).In Chapter 2,we use the distribution of stock’s idiosyncratic returns which are adjusted by market factors to construct a left-tail risk measure,and it is defined as idiosyncratic left-tail risk.Its purpose is to eliminate the high return synchronicity(Chen et al.,2010)and investment style synchronicity(Hu,Jiang & Zhu,2023)in the Chinese stock market.In Chapter 3,we use the 95% quantile of the distribution of stock’s downside volatility,which is the sum of squares of intraday negative five-minute returns,to construct an extreme downside risk measure.We define this measure as extreme downside volatility(EDV).The reason is that downside volatility can better capture the downside risks(BarndorffNielsen,Kinnebrock & Shephard,2010).In the second chapter of this thesis,the reasons for the inconsistent evidence of left-tail risk anomalies in China’s stock market are as follows.First,there are price limits in China’s stock market.This trading restriction results in the truncation of stock’s daily returns,which directly affects the construction of the left-tail risk measurement and result in an undervalued left-tail risk measurement.Second,China’s stock market has high return synchronicity(Chen et al.,2010)and investment style synchronicity(Hu,Jiang & Zhu,2023).This leads to the raw return of individual stocks sensitive to market factors,which makes stock’s raw return contains much market risk information.If the raw return of these stocks does not exclude market return synchronicity,the left-tail risk measurement constructed by the raw return cannot fully capture the stock-specific idiosyncratic risk.Therefore,to construct effective left-tail risk measurement of stocks in China’s stock market,we need to use the stocks’ idiosyncratic return which is adjusted by market factors.This new left-tail risk measurement is called idiosyncratic left-tail risk(IVa R).In the second chapter,we first constructs the idiosyncratic left-tail risk for all stocks in China’s stock market.The specific steps are as follows: At the end of each month,we use the daily return of the past 220 trading days for each stock to perform time series regression on Fama &French(1993)three factors,and define its residuals as(centralized)idiosyncratic return.Then,the negative value of the 5% quantile of stock’s idiosyncratic return is defined as idiosyncratic left-tail risk(IVa R).Then this paper follows the analytical framework of Atilgan et al.(2020)to empirically test the idiosyncratic left-tail risk effect in Chinese stock market.The empirical results in the second chapter reveal the following conclusions:(1)There is a significant and robust idiosyncratic left-tail risk effect in China’s stock market: idiosyncratic left-tail risk has a significant and robust negative relationship with the future return of stocks.The equal-weighted monthly return on buying high IVa R stocks and selling low IVa R stocks is-0.73%(t statistic is-5.22),and the valueweighted return is-0.70%(t statistic is-5.61).Compared with the US market,the idiosyncratic left-tail risk anomaly in China’s stock market is stronger.(2)This paper provides evidence in the Chinese stock market to support the behavioral explanation of the left-tail risk anomaly proposed by Atilgan et al.(2020).The persistence of IVa R and investors’ limited attention help to explain the negative relationship between IVa R and future stock return.(3)The idiosyncratic left-tail risk anomaly mainly occurs in stocks with low analyst coverage and low institutional shareholding ratio,but does not exist in stocks with high analyst coverage and high institutional shareholding ratio.This result is in line with people’s common understanding: retail investors are more likely to have limited attention,and stocks with less public information are more difficult to attract attention,so it further supports the explanation of Atilgan et al.(2020).(4)The idiosyncratic left-tail risk anomaly mainly exists in stocks with severe arbitrage restrictions,but does not exist in stocks with low arbitrage restrictions.This shows that limits of arbitrage makes it difficult to correct the mispricing in the idiosyncratic lefttailed risk anomaly.(5)Although investor sentiment cannot fully explain the idiosyncratic left-tail risk anomaly of China’s stock market,this anomaly is significantly stronger in the period of high sentiment.This result is different from the US market.In the US stock market,the left-tail risk effect only exists in the period of high sentiment and disappears in the period of low sentiment(Bi & Zhu,2020).It is interesting that the idiosyncratic left-tail risk is stronger in the recent period.For example,in the period of 2016-2021(2001-2005),the equal-weighted(value-weighted)IVa R premium is-0.95%(-0.60%),and the equal-weighted(value-weighted)IVa R premium is-0.92%(-0.66%).The third chapter of this thesis proposes to construct another measure of extreme downside risk of stocks.In essence,both the left-tail risk(Va R)and the idiosyncratic left-tail risk(IVa R)measure the risk of extreme decline in stock prices.The difference is that the former uses the raw return,while the latter uses the idiosyncratic return.The similarity between the two tail risk measures is that both of them use the extreme downside return to measure the extreme downside risk.In third chapter we use the downside volatility sequence to construct the extreme downside volatility and measure the extreme downside risk of stocks.This method is essentially equivalent to using the extreme downward return of stocks to measure the extreme downward risk,because the distribution information of the two returns is the same.The specific construction method is as follows: for each stock,the daily downside volatility is constructed by using the intraday negative five-minute return,and then calculate the 95% quantile of the stock’s downside volatility distribution of the past 220 trading days.Extreme downside volatility(EDV)is defined as the 95% quantile of the stock’s distribution of downside volatility.The concept of downside volatility is first proposed by Barndorf-Nielsen,Kinnebrock & Shephard(2010).They decompose the realized volatility into downside volatility and upside volatility,which is the sum of the squares of the positive and negative five-minute returns within the day,and use downside volatility to measure the downside risk of assets.This volatility decomposition method has been widely used in time series empirical asset pricing research(Patton & Sheppard,2015;Feunou & Okou,2019;Kilic & Shaliastovich,2019;Wang & Yan,2021).In the cross-sectional asset pricing study,Bollerslev,Li & Zhao(2020)define the upside volatility as "good volatility" and the downside volatility as "bad volatility".They aslo define the difference between the two divided by the sum of the two as the relative signed jump variance(RSJ),and tested the predictive power of RSJ in explaining future stock return.However,they did not study the cross-sectional relationship between the downside volatility or extreme downside volatility and the future stock return.In the third chapter of this thesis we empirically test the predictive power of extreme downside volatility(EDV)in the Chinese stock market.The empirical results in the third chapter show the following conclusions:(1)There is a very significant and robust negative relationship between extreme downside volatility and future stock returns.Buying stocks with high extreme downside volatility and selling stocks with low extreme downside volatility earns an equal-weighted return of-0.87%(t statistic is-5.66)and value-weighted return of-1.17%(t statistic is-4.92)per month.Interestingly,this anomaly is particularly strong in the value-weighted portfolios.(2)The extreme downside volatility anomaly is mainly attributable to investors’ underreaction to the persistence of extreme downside risk due to limited attention,and the limits of arbitrage cannot fully explain this abnormal.In fact,among the stocks with low 30% arbitrage limit,the extreme downward volatility effect is also statistically significant,but the idiosyncratic left-tail risk anomaly does not exist.(3)Compared with the idiosyncratic left-tail risk anomaly,although the extreme downside volatility anomaly and idiosyncratic left-tail risk anomaly coexist,the former is significantly stronger than the latter,especially in the value-weighted portfolio and large market capitalization stock portfolio.The equal-weighted premium of extreme downward volatility in 30% of large-cap stocks is-0.90%(t-statistic is-3.19),and the value-weighted premium is-0.86(t-statistic is-3.80).(4)Similar to the idiosyncratic left-tail risk anomaly,the extreme downside volatility anomaly only exists in stocks with low analyst coverage.Compared to the idiosyncratic left-tail risk anomaly,the extreme downside volatility anomaly is also very significant in stocks with high institutional shareholding ratio.In contrast,the idiosyncratic left-tail risk anomaly does not exist in stocks with high institutional shareholding ratio.Together with the third conclusion,the extreme downside volatility anomaly has a strong effect in stocks with large market capitalization and high institutional shareholding ratio,indicating that institutional investors have also failed to identify the extreme downside volatility risk.(5)When controlling extreme downward volatility,idiosyncratic volatility(IVOL)anomalies are subsumed and only exist in stocks with extremely high extreme downward volatility.The fourth chapter of this thesis studies the frontier of behavioral finance: salience effect.The salience theory and salience asset pricing model proposed by Bordelo,Gennaioli & Shleifer(2012)and Bordelo,Gennaioli & Shleifer(2013a)indicate that investors will be attracted by salient asset attributes(such as extreme return and turnover)due to cognitive limitations when valuing stocks,thus giving higher weight to the status of salient attributes.However,if there is a very high return in the salience state,the stock will be overvalued and its expected return in the future will decline.Cosemans & Frehen(2021)tested the prediction of salience asset pricing models in the US stock market.They find a significant salience effect in the U.S.stock market,and they also find that,the salience effect only exists in small-cap stocks.Cakici & Zaremba(2022)find that salience effect only exists in the stocks with very small market value in the international stock market,and it is not very robust.However,in China’s stock market,Liu,Sun & Zhu(2023)found that the salience effect is very significant and robust,and this effect also exists in large-cap stocks.They confirmed that the salience effect in the Chinese stock market is pronounced in lottery stocks,which indicates that the salient thinking mainly comes from the gambling behavior in the Chinese stock market(Lu,Chen & Li,2021;Liu et al.,2022).These research construct the measure of salience by using the distribution of stock’s historical returns as mental representation.However,according to Barberis,Mukherjee & Wang(2016),although stock return distribution is a good mental representation,investors can choose other attributes of stocks as mental representations.Sun,Wang & Zhu(2022b)find that there is a turnover salience effect in China’s stock market.In this paper,we believe that investors’ salient thinking is not only derived from extreme stock returns or turnover rates,but also from other extreme asset attributes,for example,investors may be attracted by consistent rising or falling intraday price trends and then form salience thinking.Fortunately,Bollerslev,Li & Zhao(2020)proposed the intraday relative sign jump variation(RSJ)of stocks,which can be used to describe this asset attribute.In fact,RSJ is a variable between-1 and +1.When it is close to+1,it represents consistent rise in the day,and when it is close to-1,it represents consistent decline in the day.Although the research of Bollerslev,Li & Zhao(2020)and Chen,Ding and Zhao(2019)has confirmed that the intraday relative sign jump variation(RSJ)can predict the future stock return in cross section,there is no research forming salience effect based on RSJ.Therefore,the fourth chapter of this thesis construct the relative sign jump variation(RSJ)of stocks in China’s stock market by following Bollerslev,Li& Zhao(2020),and then follow Cosemans & Frehen(2021)to construct the salience effect based on the relative jump variation of stocks,which is called salient relative signed jump variation(STRSJ).We empirically test the predictive relationship between STRSJ and future stock return.In the fourth chapter,we follow Cosemans & Frehen(2021)to construct the salience weight function based on RSJ,and calculates the expected return distortion caused by the salience of jump variation,which is defined as salient relative signed jump variation(STRSJ).In theory,if the STRSJ of the stock is positive and larger,the stock is overvalued and its future return will be lower.If the STRSJ of stock is negative and smaller,the stock is undervalued and its future return will be higher.STRSJ has a negative relationship with the future return of stocks.In this chapter we find that :(1)In the Chinese stock market there is a very significant and robust STRSJ effect.The strategy that buy stocks with high STRSJ and sell stocks with low STRSJ earns the average monthly equal-weighted return of-1.30%(t statistic is-9.66)and valueweighted return of-1.08%(t statistic is-6.01).STRSJ effect cannot be explained by other pricing factors.(2)The STRSJ effect is significantly stronger than salience of return(ST)or turnover(STV).In the bivariate portfolio analysis,the STRSJ subsumes ST effect and the weaken STV effect.In Fama-Mac Beth regression analysis,three salience effects coexist.(3)The STRSJ effect originates from both upward and downward salience of jump variation.This finding is different from the ST effect,which mainly stems from the upward salience of the return.In Fama-Mac Beth regression analysis,both upward and downward effects of STRSJ exist.The upward salience effect of ST exists,but the downward salience effect does not exist.The upward and downward salience effects of STV do not exist.(4)Gambling preference,imbalance of order flow,V-shape disposition effect,arbitrage restriction and margin trading restriction can explain the STRSJ effect in Chinese stock market.The STRSJ is more prominent in stocks with strong gambling preference,high order inflow,strong V-shape disposition effect,strong arbitrage restrictions and no margin trading.(5)The STRSJ effect is very significant in stocks with large market value and high institutional shareholding ratio.This result shows that the anomaly is not only from the behavior bias of individual investors but also from institutional investors.(6)The STRSJ effect fully subsumes the reversal effect,and is not affected by investor sentiment and market volatility.The main contributions of this thesis can be summarized as follows.First of all,this thesis enriches the literature of left-tail risk anomalies and low risk anomalies.In order to explore the reasons why the left-tail risk anomaly in China’s stock market is not robust,this paper proposes two revised left-tail risk measures from different perspectives: one is idiosyncratic left-tail risk(IVa R),and the other is extreme downward volatility(EDV).Both are measures of extreme downside risk.Using these two measures,this paper finds that there is a significant and robust extreme downside risk anomaly in China’s stock market: the greater the extreme downside risk,the lower the future return of the stock.The two extreme downside risk anomalies based on IVa R and EDV only exist in stocks with low analyst coverage,and are significantly stronger in stocks with low institutional shareholding.This is consistent with the behavioral explanation of Atilgan et al.(2020): investors,especially retail investors,underestimate the persistence of extreme downside risks due to limited attention.The extreme downside risk anomalies in China’s stock market are not only attributed to the behavioral bias of investors,but also to the limits of arbitrage in China’s stock market.Although these two extreme downside risk anomalies coexist,the EDV anomaly is significantly stronger than the IVa R anomaly,especially in the value-weighted portfolios and large market capitalization stocks.Among the stocks with high institutional shareholding ratio,EDV anomalies are very significant,which is different from the US market;However,IVa R anomaly does not exist,which is similar to the US market.The research in this paper has enriched people’s understanding of the lefttail risk anomalies and low-risk anomalies in China’s stock market and their causes.Secondly,this paper enriches the literature of salience theory and salience effect.This paper finds a very significant and robust salient jump variation(STRSJ)effect in China’s stock market: the STRSJ has a negative relationship with the future return of stocks.The pricing factor based on STRSJ cannot be explained by other factor models.Gambling preference,imbalance of order flow,V-shape disposition effect,arbitrage restriction and margin trading restriction can significantly explain the STRSJ effect in the Chinese stock market.The STRSJ effect is very different from the ST effect and STV effect that have been found: the former is significantly stronger than the latter two.The STRSJ effect comes from both the upward and downward saliencc effect of RSJ,while the ST effect mainly comes from the upward salience effect.Similar to the ST effect,the STRSJ effect is also very significant in stocks with large market value and high institutional shareholding ratio in the Chinese stock market.It further shows that,the STRSJ effect in the Chinese stock market comes not only from the behavioral bias of individual investors but also from institutional investors.The STRSJ effect explains the reversal effect.The thesis has tested the predictive power of the salience asset pricing model from a new perspective,and further provide empirical evidence to support the salience asset pricing theory,and finally enriched people’s understanding of the salience effect in the Chinese stock market. |
PDF Full Text Request