1、俞紅海2014.3.19行為金融學(xué) 第五講股票市場截面異象市場可預(yù)測性/ Eugene F. FamaThe Prize will focus attention on the area of asset pricing very widely, so, of course your work is very well known to those in the Finance community. But people who have never encountered this before will suddenly be exposed to it. What, fundamenta

2、lly, would you say was the relevance of your work, to the world at large?Eugene F. Fama: Well, for this particular area, its really, the idea is really how do you measure risk? And if the market is pricing things correctly, what is the relation between the expected return, which is the compensation

3、for risk, and the risk. So .Homepage: http:/faculty/directory/f/eugene-f-famaLars Peter Hansen Yes, and for those who are not in the finance community and havent really encountered this before, how would describe what you do? Lars Peter Hansen: I think theres two aspects to it. One is , I have done

4、work on, whats called econometrics, which is statistics in economics and I did work, that I think of in a very simple way as trying to show how you can do without having to do everything. Its kind of a funny way to put it, but if you want to study a dynamic economic system, what youd like to be able

5、 to do is focus on the linkages, say, between asset markets and the macro economy without having to model everything at the same time. . what certainly comes into play is that you need ways to certainly address the fact that investors, they respond to information and struggle with what the right vie

6、w of the world is, and you need to have ways to capture those struggles and how they reflect in asset markets.Homepage: / Robert J. ShillerBut, I just wanted to ask you just one question, because this is going to expose your field of research to a very wide audience, who wont previously have been aw

7、are of it. Its all to do with the predictability of asset pricing. How good a handle do you think we have on that? Robert J. Shiller: Its very much a work-in-progress, yeah. Well I could briefly say that its . weve learned a lot about asset pricing, but theres a basic human element in it that is irr

8、educible. So, predicting what asset prices will do is partly similar to trying to predict what one person will do, and so could there ever be a science in predicting what you, Adam, will be doing? No, because theres an irreducible human element. And thats part of the reason why the field of finance

9、will never completely understand asset pricing movements.Homepage: /shiller/ 第五講主要內(nèi)容65.1 研究背景5.2 規(guī)模效應(yīng)5.3 價值效應(yīng)5.4 反轉(zhuǎn)效應(yīng)5.5 動量效應(yīng)5.6 盈余宣告后漂移效應(yīng)5.7 其他：封閉基金折價之謎 研究背景7研究意義理論價值：對理論模型CAPM及市場有效性理論加以檢驗實踐價值：構(gòu)造投資組合，形成投資策略如何預(yù)測股票市場走勢研究背景8研究基礎(chǔ)投資組合為什么選擇投資組合，而不是個別股票作研究對象？Portfolio betas are better measured, due to les

10、s residual variance;Individual stock returns are very volatile, and we couldnt even reject that all expected returns are the samePortfolio correspond to investment strategies中國市場目前主要還是集中在個股分析中國人力資本價格便宜可交易股票數(shù)量不多中國數(shù)據(jù)可靠性存在問題研究背景9截面研究基本過程1. Generate Portfolios with different average returns:Find a chara

11、cteristic that might explain average returns;Sort stocks into portfolios based on the characteristic;Check that portfolios have different average returns;Worry about measurement, Survival bias, data mining.2. See if the difference in average returns can be explained by existing models:Standard bench

12、mark models are CAPM, Fama-French 3-Factor Model;Given a chosen benchmark model, compute betas for the portfolios;Check whether the spreads in average returns across different portfolios can be explained by their spreads in betas;3. If yes, conclude that the benchmark models do well; if not, you hav

13、e an anomaly.規(guī)模效應(yīng)(Size Effect)10規(guī)模效應(yīng)(Banz, 1981)：小公司具有可預(yù)見的相對高收益1981年，Banz對所有在紐交所上市的股票收益情況進行了研究，他將公司按規(guī)模分成五組，發(fā)現(xiàn)最小規(guī)模組的平均年收益率比那些最大規(guī)模組的公司要高19.8%，而且無論是在風(fēng)險調(diào)整之前還是調(diào)整之后，小規(guī)模組的公司股票的收益率都系統(tǒng)地高。人們稱這一現(xiàn)象為小公司效應(yīng)（Small-Cap Effect)或規(guī)模效應(yīng)。自從1980年以來，規(guī)模效應(yīng)逐步消失。規(guī)模效應(yīng)(Size Effect)11規(guī)模效應(yīng)(Banz, 1981, JFE)價值效應(yīng)(Value Effect)12價值效應(yīng)：B

14、asu(1983), Rosenberg-Reid-Lanstein(1985), Fama-French(1992)Stocks with high BE/ME(Value Stock) earns higher average returns than stocks with low BE/ME(Growth Stock)Despite the fact that value stocks tend to have lower betas than growth stocks;In head-to-head horse race, BE/ME does much better than b

15、etas;Robust through different historical time periods;Fama-French 3-Factor Model13Fama-French 三因素模型Fama和French 1992年對美國股票市場決定不同股票回報率差異的因素的研究發(fā)現(xiàn)，股票的市場的beta值不能解釋不同股票回報率的差異，而上市公司的市值、賬面市值比可以解釋股票回報率的差異。Fama and French 認為，上述超額收益是對CAPM 中未能反映的風(fēng)險因素的補償。Fama and French希望小規(guī)模和低市值/賬面值比可以作為“破產(chǎn)風(fēng)險”不同側(cè)面的代表，但直到現(xiàn)在仍沒有直接證

16、據(jù)證明這一點(Shleifer, 2003)。Eugene Fama學(xué)術(shù)貢獻14Eugene F. Fama is widely recognized as the father of modern finance. 金融經(jīng)濟學(xué)領(lǐng)域的思想家。他在經(jīng)濟學(xué)科的若干領(lǐng)域都作出了重要的貢獻，在金融學(xué)獨立為一個學(xué)科以及成為經(jīng)濟學(xué)中一個獨立領(lǐng)域的進程中，是當(dāng)之無愧的先驅(qū)。有效市場理論EMH：相關(guān)的信息如果不受扭曲且在證券價格中得到充分反映，市場就是有效的。任何戰(zhàn)勝市場的企圖都是徒勞的。Fama-French三因素模型規(guī)模及價值因素反映了未被beta包括的公司風(fēng)險Fama-French 3-Factor M

17、odel15Fama和French 1993年指出可以建立一個三因子模型來解釋股票回報率。模型認為，一個投資組合(包括單個股票)的超額回報率可由它對三個因子的暴露來解釋，這三個因子是：市場資產(chǎn)組合(Rm Rf)、市值因子(SMB)、賬面市值比因子(HML)。這個多因子均衡定價模型可以表示為： E(Rit) Rft = iE(Rmt Rft + siE(SMBt) + hiE(HMLt) 其中Rft表示時間t的無風(fēng)險收益率；Rmt表示時問t的市場收益率；Rit表示資產(chǎn)i在時間t的收益率；E(Rmt) Rft是市場風(fēng)險溢價，SMBt為時間t的市值(Size)因子的模擬組合收益率，HMLt為時間t的

18、賬面市值比因子的模擬組合收益率。反轉(zhuǎn)效應(yīng)(Reversal Effect)16Reversal Effect: De Bondt and Thaler(1985)行為金融學(xué)開端的標(biāo)志股票價格存在系統(tǒng)性過度反應(yīng)，僅根據(jù)歷史數(shù)據(jù)就可以預(yù)測股價在未來將出現(xiàn)反轉(zhuǎn)，從而股價的歷史數(shù)據(jù)地其未來走勢具有預(yù)測能力：股價大幅度的上漲（下跌）之后將出現(xiàn)下跌（上漲）；股價初始上漲（下跌）幅度越大，其后下跌（上漲）的幅度也越大。Richard Thaler學(xué)術(shù)貢獻17行為金融學(xué)領(lǐng)域的重要開創(chuàng)性人物。塞勒的主要研究領(lǐng)域是行為經(jīng)濟學(xué)、行為金融學(xué)與決策心理學(xué)。在行為金融學(xué)方面，塞勒研究人的有限理性行為對金融市場的影響，并作

19、出了很多重要貢獻。The Winners Curse(贏者的詛咒) Nudge(助推)Advances in Behavioral Finance反轉(zhuǎn)效應(yīng)(Reversal Effect)18反轉(zhuǎn)效應(yīng)(De Bondt and Thaler, 1985, JF)反轉(zhuǎn)效應(yīng)(Reversal Effect)191985年，De Bondt and Thaler發(fā)現(xiàn)，在一段時間內(nèi)，表現(xiàn)最好的股票在接著會表現(xiàn)非常差。實證研究表明，如果對股票業(yè)績進行為期5年的排序，基期表現(xiàn)不好的股票組(含35種業(yè)績最差的股票)在以后的3年中的平均累計收益，比基期表現(xiàn)最好的股票組(含35種業(yè)績最好的股票)的累計收益高出2

20、5。同時采用1、2、3、5年四種形成期，并逐月觀察長達5年的持有其收益率變化，實證結(jié)果均支持反轉(zhuǎn)效應(yīng)。動量效應(yīng)(Momentum Effect)20動量效應(yīng)：Titman and Jegadeesh(1993)股票的收益率有延續(xù)原來的運動方向的趨勢，即過去一段時間收益率較高的股票在未來獲得的收益率仍會高于過去收益率較低的股票?；诠善眲恿啃?yīng)，投資者可以通過買入過去收益率高的股票、賣出過去收益率低的股票獲利，這種利用股價動量效應(yīng)構(gòu)造的投資策略稱為動量投資策略?；诿绹袌?965-1989年數(shù)據(jù)，分別采用形成期和持有期為3、6、9、12四種，交叉搭配形成16種交易策略，發(fā)現(xiàn)這16種交易策略均能

21、獲得顯著為正的利潤，即使經(jīng)過風(fēng)險調(diào)整之后，依然為正。動量效應(yīng)21動量效應(yīng)(Jegadeesh and Titman, 1993, JF)動量效應(yīng)22Jegadeesh and Titman(2001)進一步用1990-1998年間在NYSE、AMEX、NASDAQ上市交易的所有股價不低于5美元的上市公司股票重復(fù)了他們在1993年的研究。結(jié)果發(fā)現(xiàn)慣性交易策略依然可以獲得顯著的利潤，且利潤規(guī)模與他們1993年的研究結(jié)果近似。結(jié)果表明：1993年的研究成果不是數(shù)據(jù)挖掘的結(jié)果；新的問題：既然慣性策略可以獲得利潤，并且采用這種策略的機構(gòu)投資者日益增多，為什么慣性策略的利潤沒有被消除？中國市場反轉(zhuǎn)與慣性效

22、應(yīng)23中國市場反轉(zhuǎn)效應(yīng)（魯臻，鄒恒甫，經(jīng)濟研究 2007.8）中國市場反轉(zhuǎn)與慣性效應(yīng)24從表1 中我們可以發(fā)現(xiàn),只在J = 6 ,K= 6 這個投資策略存在慣性效應(yīng)；而短期有J = 1 ,K= 3 ;J= 3 ,K= 1 ,3 一共3 個投資策略存在反轉(zhuǎn)；長期J = 6 ,K= 36 ;J = 12 ,K= 24 ,36 ;J =18, K=18, 24, 36;J= 24 ,K= 18 ,24 ,36 ;J = 36 ,K= 18 ,24 ,36 一共12 個投資策略存在反轉(zhuǎn);其他投資策略的結(jié)果都是不顯著的。四因素模型(4-Factor Model)25Carhart四因素模型是對Fama-French三因素的模型的改進,包含：市場因素(market factor)規(guī)模因素(