论文部分内容阅读
摘要：当前，随着对于金融市场实证研究的不断深入，传统金融学以有效市场为核心的部分资产定价模型在实际研究中遇到了许多难以解释的异常状况。而行为金融学的出现，则在这一方面对于传统金融学理论进行了有效的补充。本文着重藉由实证研究介绍了BSV、DHS这两种基于投资者情绪的资产定价模型。BSV模型的研究基于在一连串同向消息刺激下投资者的过度反应，而DHS则是通过投资者的过度自信而带来的过激反应或者反应不足，来解释股票价格短期内的异常波动行为。总之，这二者通过对于投资者情绪的定量分析，弥补了传统金融学理论对于资产价格异常波动时的解释不足。
关键词：投资者情绪 过度自信 BSV模型 DHS模型
▲▲1. Introduction to behavioral finance
1.1. Deficiency of traditional financial theory
As a crucial foundation of contemporary financial theory， the efficient market hypothesis （EMH） used to influence most part of investment policies， such as capital asset pricing model and the arbitrage pricing theory model. More exactly， the efficient market means the price of securities on this market could always reflect the change of the market information （Fama 1970）.
In the theory， it assumed that the investors are rational enough to make maximum use of the market information. On the other hand， the actions between irrational investors could be neutralized each other， or corrected by the rational arbitrager. Therefore， it could be considered that the trading behavior from irrational investors could not affect the market price of securities. Due to this reason， the traditional pricing model， such as CAPM and APT model did not take the influence of irrational investors into consideration.
However， in the practical studies， massive researches showed the contrary result to the EMH theory， which implies that the classical pricing model may not be effective enough in practical， due to the limitation of those assumptions. Which means pricing models based on the traditional financial theory could be less effective in the inefficient market.
1.2. The role of behavioral finance
To fulfill the gap between traditional financial theory and the reality， behavioral finance focus more on those assumptions which used to be neglected but caused real impact in practice. For example， as complement to the assumption of rational investor， researchers of behavior finance build up a noise trader model which paid attention to those irrational traders. Then they observed the influence of their trading actions based the wrong information on the market.
In this article， the main aim is discussing the relationship between the security price and the investor sentiment. In other words， considering that the classical asset pricing models are facing many challenges from the practical studies of security price， Barberis， Shleifer， and Vishny （1998） developed a model to explore whether the sentiment of investor could influence the security price， which seemed to be replenishment to the rational assumption. This essay is based on the articles which focused on the relationship between the asset price and the investor sentiment. The first part is according to the model built by Barberis， Shleifer and Vishny， which was is the field of investor sentiment and pricing. Then the article concludes a practical analysis from Frieder which is based on this model to test the pros and cons in this model. The second part is based on the DHS model which provided a linkage between the asset price and the investor’s overconfidence， then tries to find out the systematic relationship between investor sentiment and asset pricing.
▲▲2. Main findings
2.1. Models and empirical evidence
2.1.1. Practical analysis on BSV models
This part is from Frieder’s （2008） study， which is to test the change of the stock price based on investor sentiment model from Barberis， Shleifer， Bishny （1998） （BSV model）. In this model， two extrapolation biases are pulled into investor’s decision making， which is the so-called conservation and representative. The conservation investors are prone to undervalue news from the market， causing the lack of response， while the representative ones are likely to estimate the whole population according to a sample that they could observed， which may lead to overreaction. Then it assumed that the security price is a random walk. But in investor’s consideration， the return would follow two models： the one was fluctuating around a mean price （conservation situation）， the other was moving by a trend （representative situation）. In other words， investors may tend to show the conservation while they heard a piece of independent news or some contrary news， but when they heard some consecutive same sign news， they could be overreaction. Therefore， in this model， the security price can be basically expressed as the perception of investor’s sentiment （）. Then， through their data analysis， they found they there is a short term autocorrelation and a long term reverse between the stock and the investor sentiment affected by company earnings surprises （Barberis 1998）.
In Frieder’s empirical analysis， firstly， he divided the news into positive and negative group， then following the BSV model， he use isolated news as the present of conservation， and the consecutive news to test the representative. In order to examine the influence of investor’s sentiment， he set the order imbalance as the observation. This is due to the fact that when an earnings announcement occurred on a certain stock， investor may tend to buy （sell） a lot， which could cause a gap between the volumes of ask and bid order. The results are made up of two main parts. The initially stage shows the relationship between extrapolation bias and investor response. Consistent with the former research from Barberis， this article give the evidence to show that investors response to earnings announcement could be considered as rational enough in the group of positive news. Since the number of positive earnings surprise boosted the imbalance of orders， which means small investors are prone to purchase the certain stock with more good news. Also， this phenomenon is in accordance with the “representative” in the BSV model. However， in the negative announcement group， the statistic result was not significantly enough to support the BSV model， even contrary to it. Combining with the conservation sentiment and the low frequency of negative announcement， the article try to explain that investors were not eager to sell their stock immediately when the downside news occurred.
The other result is about the linkage between the stock return and order imbalance caused by the extrapolate bias. Though his analysis， the author found that the stock return is negatively related to the order imbalance for the positive earnings surprise. As for the negative news， there was still not significant evidence to prove its correlation. The reason for this issue given by author is that the institution investor would take the opposite position to those small investors， which means they sold stock with high OIBs， and purchased them with low OIBs. Then according to the regression analysis， it could be clearly seen that the amount of good news was positive related to stock return， but caused a slight impact. Meanwhile， the OIBs could exerted a obvious negative effect to stock return. All this evidence could prove the short term positive correlation and the long term reversal.
2.1.2.Practical analysis on DHS models
This part will discuss the asset pricing model based on paper from Daniel， Hirshleifer and Subrahmanyam. In their previous research ， they constructed a sentiment model that mainly focused on the investors who have the information from private way， instead of the public news. Then it assumed that investors with private info were risk neutral， while investors without it were risk averse. According to the psychological theory， investors with private info could show two sorts of response， the one is overconfidence， which implies that they would overrate to the information they hold， and neglect the probability of mistake， during their process of decision making. Then consequently， following the wrong trading action， the stock price could be pushed off from its essential value. Then after the public information announced， the stock price would return gradually to its initial level. The other was so-called biased self-attribution， which means investors would be more confident to their judgment while the private information they hold was in accordance with the public one， otherwise， they might attribute this to the external reasons. In a word， self confidence would be strengthened through these two processes， and caused the short term positive autocorrelation and the long term reversal. According to this situation， the authors gave an asset pricing model which was similar to the CAPM model in their later research paper. It assumed that the last term cash flow of a certain stock fit a linear model， the variables from the model could be observed by investors. Then the equilibrium price would be formed through the trading activities from overconfidence investors and rational investors， which is positive related to the beta coefficient and the present mispricing level.
In addition， due to their empirical analysis， they found that the arbitrary speculation was not significant enough to affect the stock return in the subsequent period. Instead， they are following the DHS model， which means due to the overconfidence， the private information could impose the stock price and the current period. Then the price would be adjusted to close the equilibrium value， which should have to be influenced by the rational investors. According to this findings， the article proved that the Beta value from CAPM were still useful even during the condition of irrational investors to some extent， because the mispricing would be eventually adjusted to meet the equilibrium value which could be predicted by CAPM.
2.2. The analysis of potential drawbacks
In terms of the BSV model， admittedly， though its price regression model， it could well explain the short term positive autocorrelation and long term reversal phenomenon， also showed the systematic linkage between the investor’s psychological biases and the stock price. However， there is still some potential weakness in this model. Initially， considering that it took the conservation and representative of investor as their basic assumptions， they connected those psychological biases with their model. However， to be precise， BSV model was not strictly based on the psychology， but only supported by them. As a result， there is a problem that the model could not ensure that all investors only have the conservation or the representative sentiment， which would limit the explanation ability of this model in practice. Additionally， the empirical analysis by Frieder also exposed some drawbacks of this model. For instance， investors’ sentiment and response was not following the model when facing the negative news imposing. All of these are yet to wait us to do a further research.
As for the DHS model， it was based on the condition of informational asymmetric market. Inconsistent with previous research， this model implied that the positive autocorrelation was caused by the investors’ over-react， which sprung from their overconfidence of private information they hold， while the most research， like the BSV model， attributed this phenomenon to a lack of response. But there were still some drawbacks in this issue. Like the BSV model， the DHS model also showed a limitation that it simply assumed the cause of over-react as the overconfidence and the self-attribution， while the real causes were complex. Moreover， unlike the BSV model， the psychological bias of DHS model did not only the small investor， which means that the model did not have clearly range. After that， it seemed to be contradictious that it used the traditional financial method to build a pricing model under the irrational conditions. ▲▲3.Summary
To conclude， this essay is discussing two asset pricing models based on the investor sentiment. To explain the short term positive autocorrelation and long term reversal of stock price， based on the extrapolation bias， the BSV model showed that the investors could be over-reaction while they met a string of same sign news， and under-reaction with isolated news， which could cause the above-mentioned fluctuation of stock price. The DHS model implied that the overconfidence and the self –attribution could lead to a over-reaction of private information and under-reaction of the public information， which finally caused the wave of price. In a word， through the above analysis， it could be concluded that there is a systematic link between the security price and the investor sentiment.
List of References
[1] Barberis， N.， Shleifer， A.， Vishny， R.， 1998. A model of investor sentiment. Journal of Finance 49， pp， 307-345
[2]Laura Frieder，， 2008. Investor and price response to patterns in earnings surprises. Journal of Financial Market 11， pp， 259-283
[3]Daniel， K.D.， Hirshleifer， D. and Subrahmanyam， A.， 2001. Overconfidence， arbitrage， and equilibrium asset pricing. Journal of Finance 56， pp. 921–965.
[4]Daniel， K.D.， Hirshleifer， D. and Subrahmanyam， A.， 1998. Investor psychology and security market under-and over-reactions. Journal of Finance 53， pp. 1839–1886
关键词：投资者情绪 过度自信 BSV模型 DHS模型
▲▲1. Introduction to behavioral finance
1.1. Deficiency of traditional financial theory
As a crucial foundation of contemporary financial theory， the efficient market hypothesis （EMH） used to influence most part of investment policies， such as capital asset pricing model and the arbitrage pricing theory model. More exactly， the efficient market means the price of securities on this market could always reflect the change of the market information （Fama 1970）.
In the theory， it assumed that the investors are rational enough to make maximum use of the market information. On the other hand， the actions between irrational investors could be neutralized each other， or corrected by the rational arbitrager. Therefore， it could be considered that the trading behavior from irrational investors could not affect the market price of securities. Due to this reason， the traditional pricing model， such as CAPM and APT model did not take the influence of irrational investors into consideration.
However， in the practical studies， massive researches showed the contrary result to the EMH theory， which implies that the classical pricing model may not be effective enough in practical， due to the limitation of those assumptions. Which means pricing models based on the traditional financial theory could be less effective in the inefficient market.
1.2. The role of behavioral finance
To fulfill the gap between traditional financial theory and the reality， behavioral finance focus more on those assumptions which used to be neglected but caused real impact in practice. For example， as complement to the assumption of rational investor， researchers of behavior finance build up a noise trader model which paid attention to those irrational traders. Then they observed the influence of their trading actions based the wrong information on the market.
In this article， the main aim is discussing the relationship between the security price and the investor sentiment. In other words， considering that the classical asset pricing models are facing many challenges from the practical studies of security price， Barberis， Shleifer， and Vishny （1998） developed a model to explore whether the sentiment of investor could influence the security price， which seemed to be replenishment to the rational assumption. This essay is based on the articles which focused on the relationship between the asset price and the investor sentiment. The first part is according to the model built by Barberis， Shleifer and Vishny， which was is the field of investor sentiment and pricing. Then the article concludes a practical analysis from Frieder which is based on this model to test the pros and cons in this model. The second part is based on the DHS model which provided a linkage between the asset price and the investor’s overconfidence， then tries to find out the systematic relationship between investor sentiment and asset pricing.
▲▲2. Main findings
2.1. Models and empirical evidence
2.1.1. Practical analysis on BSV models
This part is from Frieder’s （2008） study， which is to test the change of the stock price based on investor sentiment model from Barberis， Shleifer， Bishny （1998） （BSV model）. In this model， two extrapolation biases are pulled into investor’s decision making， which is the so-called conservation and representative. The conservation investors are prone to undervalue news from the market， causing the lack of response， while the representative ones are likely to estimate the whole population according to a sample that they could observed， which may lead to overreaction. Then it assumed that the security price is a random walk. But in investor’s consideration， the return would follow two models： the one was fluctuating around a mean price （conservation situation）， the other was moving by a trend （representative situation）. In other words， investors may tend to show the conservation while they heard a piece of independent news or some contrary news， but when they heard some consecutive same sign news， they could be overreaction. Therefore， in this model， the security price can be basically expressed as the perception of investor’s sentiment （）. Then， through their data analysis， they found they there is a short term autocorrelation and a long term reverse between the stock and the investor sentiment affected by company earnings surprises （Barberis 1998）.
In Frieder’s empirical analysis， firstly， he divided the news into positive and negative group， then following the BSV model， he use isolated news as the present of conservation， and the consecutive news to test the representative. In order to examine the influence of investor’s sentiment， he set the order imbalance as the observation. This is due to the fact that when an earnings announcement occurred on a certain stock， investor may tend to buy （sell） a lot， which could cause a gap between the volumes of ask and bid order. The results are made up of two main parts. The initially stage shows the relationship between extrapolation bias and investor response. Consistent with the former research from Barberis， this article give the evidence to show that investors response to earnings announcement could be considered as rational enough in the group of positive news. Since the number of positive earnings surprise boosted the imbalance of orders， which means small investors are prone to purchase the certain stock with more good news. Also， this phenomenon is in accordance with the “representative” in the BSV model. However， in the negative announcement group， the statistic result was not significantly enough to support the BSV model， even contrary to it. Combining with the conservation sentiment and the low frequency of negative announcement， the article try to explain that investors were not eager to sell their stock immediately when the downside news occurred.
The other result is about the linkage between the stock return and order imbalance caused by the extrapolate bias. Though his analysis， the author found that the stock return is negatively related to the order imbalance for the positive earnings surprise. As for the negative news， there was still not significant evidence to prove its correlation. The reason for this issue given by author is that the institution investor would take the opposite position to those small investors， which means they sold stock with high OIBs， and purchased them with low OIBs. Then according to the regression analysis， it could be clearly seen that the amount of good news was positive related to stock return， but caused a slight impact. Meanwhile， the OIBs could exerted a obvious negative effect to stock return. All this evidence could prove the short term positive correlation and the long term reversal.
2.1.2.Practical analysis on DHS models
This part will discuss the asset pricing model based on paper from Daniel， Hirshleifer and Subrahmanyam. In their previous research ， they constructed a sentiment model that mainly focused on the investors who have the information from private way， instead of the public news. Then it assumed that investors with private info were risk neutral， while investors without it were risk averse. According to the psychological theory， investors with private info could show two sorts of response， the one is overconfidence， which implies that they would overrate to the information they hold， and neglect the probability of mistake， during their process of decision making. Then consequently， following the wrong trading action， the stock price could be pushed off from its essential value. Then after the public information announced， the stock price would return gradually to its initial level. The other was so-called biased self-attribution， which means investors would be more confident to their judgment while the private information they hold was in accordance with the public one， otherwise， they might attribute this to the external reasons. In a word， self confidence would be strengthened through these two processes， and caused the short term positive autocorrelation and the long term reversal. According to this situation， the authors gave an asset pricing model which was similar to the CAPM model in their later research paper. It assumed that the last term cash flow of a certain stock fit a linear model， the variables from the model could be observed by investors. Then the equilibrium price would be formed through the trading activities from overconfidence investors and rational investors， which is positive related to the beta coefficient and the present mispricing level.
In addition， due to their empirical analysis， they found that the arbitrary speculation was not significant enough to affect the stock return in the subsequent period. Instead， they are following the DHS model， which means due to the overconfidence， the private information could impose the stock price and the current period. Then the price would be adjusted to close the equilibrium value， which should have to be influenced by the rational investors. According to this findings， the article proved that the Beta value from CAPM were still useful even during the condition of irrational investors to some extent， because the mispricing would be eventually adjusted to meet the equilibrium value which could be predicted by CAPM.
2.2. The analysis of potential drawbacks
In terms of the BSV model， admittedly， though its price regression model， it could well explain the short term positive autocorrelation and long term reversal phenomenon， also showed the systematic linkage between the investor’s psychological biases and the stock price. However， there is still some potential weakness in this model. Initially， considering that it took the conservation and representative of investor as their basic assumptions， they connected those psychological biases with their model. However， to be precise， BSV model was not strictly based on the psychology， but only supported by them. As a result， there is a problem that the model could not ensure that all investors only have the conservation or the representative sentiment， which would limit the explanation ability of this model in practice. Additionally， the empirical analysis by Frieder also exposed some drawbacks of this model. For instance， investors’ sentiment and response was not following the model when facing the negative news imposing. All of these are yet to wait us to do a further research.
As for the DHS model， it was based on the condition of informational asymmetric market. Inconsistent with previous research， this model implied that the positive autocorrelation was caused by the investors’ over-react， which sprung from their overconfidence of private information they hold， while the most research， like the BSV model， attributed this phenomenon to a lack of response. But there were still some drawbacks in this issue. Like the BSV model， the DHS model also showed a limitation that it simply assumed the cause of over-react as the overconfidence and the self-attribution， while the real causes were complex. Moreover， unlike the BSV model， the psychological bias of DHS model did not only the small investor， which means that the model did not have clearly range. After that， it seemed to be contradictious that it used the traditional financial method to build a pricing model under the irrational conditions. ▲▲3.Summary
To conclude， this essay is discussing two asset pricing models based on the investor sentiment. To explain the short term positive autocorrelation and long term reversal of stock price， based on the extrapolation bias， the BSV model showed that the investors could be over-reaction while they met a string of same sign news， and under-reaction with isolated news， which could cause the above-mentioned fluctuation of stock price. The DHS model implied that the overconfidence and the self –attribution could lead to a over-reaction of private information and under-reaction of the public information， which finally caused the wave of price. In a word， through the above analysis， it could be concluded that there is a systematic link between the security price and the investor sentiment.
List of References
[1] Barberis， N.， Shleifer， A.， Vishny， R.， 1998. A model of investor sentiment. Journal of Finance 49， pp， 307-345
[2]Laura Frieder，， 2008. Investor and price response to patterns in earnings surprises. Journal of Financial Market 11， pp， 259-283
[3]Daniel， K.D.， Hirshleifer， D. and Subrahmanyam， A.， 2001. Overconfidence， arbitrage， and equilibrium asset pricing. Journal of Finance 56， pp. 921–965.
[4]Daniel， K.D.， Hirshleifer， D. and Subrahmanyam， A.， 1998. Investor psychology and security market under-and over-reactions. Journal of Finance 53， pp. 1839–1886