1. 简要回顾“... If the investor were only interested in expected values of securities, he or she would only be interested in the expected Lecture 5value of the portfolio, and to maximize the expected value of a portfolio one need invest only in a singlesecurity...”资本资产定价模型(CAPM)“... This, I knew, was not the way investors did or should act. Investors diversifybecause they are concerned with 模型的假设条件risk as well as return. Variancecame to mind as a measure of risk...”CAPM的基本思路“... The fact that portfolio variance CAPM的数学推导depended on security covariancesadded to the plausibility of the approach. Since CAPM的经济学含义there were two criteria, risk and return, it CAPM之应用was natural to assume that investors Harry Markowitz, selected from the set of ... optimal risk-CAPM的实证检测return combinations.”Nobel Prize 1990Financial Economics_WCY1Financial Economics_WCY2预期收益和风险Markowitz Portfolio AnalysisInputs Required计算效率边界组合投资权重需要以下三个方程。计算效率边界投资权重,需要多种数据In order to derive the efficient frontier we maximize (1) subject to There are N expected returns, N standard deviations and N(N-(2) and (3). 1)/2 correlations. 如果卖空限制,另外的一约束条件是X> 0. iAn exampleNE(R)=∑WE(R)pNumber of assets in portfolio10100250ii (1) i=1 Expected Returns 10100250NNN222Standard Deviations 10100250σ=∑Wσ+∑∑WWσσρpiiijijij (2) i=1i=1j=1j≠iCorrelations45495031125 Total65515031625N∑W=1i (3) i=1It is due to the number of correlations required that the numberof inputs increase dramatically.One method used to reduce the amount of information required is to use index (or factor) models. Financial Economics_WCY3Financial Economics_WCY42. 资本资产定价模型的假设条件市场均衡投資組合的預期收益率和风险決定于投資組关于市场状态的假设:合中個別资产持有比重,但計算市場均衡投資組合的1.证券市场是完全竞争的市场。所有的投资者都是价格的接受者,风险是一件相当繁杂的工作。那么对于投资组合来说,市场价格由供求平衡点决定。投资组合的期望收益率与组合的风险之间有什么样的关系呢?2.市场是无摩擦的,即不存在税收和任何交易成本。3.存在无风险资产。所谓无风险资产是指投资者在购买该资产时就史丹福大學William Sharpe教授和前哈佛大學Lintner教清楚地知道持有该资产时期结束后的资产价值。由于无风险资产授以此角度切入,简化了风险和收益之间的关系,發的期末价值是已知的,因此无风险资产收益率的标准差=0展出資本资产定价模型(capital asset pricing model,or CAPM)。 CAPM只需衡量个股或组合与市场组合(效率组合)的关系,就可确定其预期收益。β是衡量单个资产的市场风险适当的指标,而β的大小決定這個单个资产或组合預期溢价。Financial Economics_WCY5Financial Economics_WCY61
3. 资本资产定价模型(CAPM)推导的基本思路关于投资者行为的假设:均值-方差模型:证券组合选优 1.投资者基于预期收益率和收益率的方差(或标准差)来进行投引入无风险资产后的有效集及分离定理。资决策。决策行为遵循如下原则:当风险(方差或标准差)相同资本市场线:有效组合收益和风险的关系。时,选择期望收益率高的证券组合;当期望收益率相同时,选择风险小的证券组合。证券市场线:任一单个证券收益和风险的关系。 E(U) = E(R) –(R)2.投资者对所有证券的收益、方差及证券间的相关性具有完全相同的预期。3.所有投资者都是在单一投资期内进行投资决策。 Financial Economics_WCY7Financial Economics_WCY8资本市场线的含义 资本市场线(Capital Market Line,CML)的推导 E(R)−RMfE(R)=R+⋅σ pfPσM引入无风险资产后,有效集内任一证券组合的期望收益率与对应风险之间的关系 由两部分组成 :一部分是无风险利率R,它反映了对投资者放弃即期消费的补偿; fE(R)−RMf另一部分⋅σ则是对投资者承担风险σ的补偿,通常称之为风险溢价。 PpσM E(R)−RMf它与承担的风险σ的大小成正比,其系数(也就是资本市场线的斜率), TDMCp=E(R)−RE(R)−RσPfMfM 对于资本市场线上的任一点T,有:DFFC 即: =,σσPM度量了每一单位市场风险的补偿,故又称为风险的价格。 E(R)−RMf 化简后得到:E(R)=R+⋅σ pfPσMFinancial Economics_WCY9Financial 资本资产定价模型(CAPM)与证券市场线(Security Market 证券市场线Line, SML)CAPM与SML是什么? 资本市场线(CML)对有效证券组合的风险(标准差)与期望收益的关系给予了较为完整的刻画。任何单个风险证券由于均不是有效组合而一定位于资本市场线的下方,因此资本市场线方程并不能告诉我们单个证券的预期收益与风险(标准差)之间的关系。但利用资本市场线(CML)的有关结论,可以证明,对于任一证券i,其期望收益率只是对其系统风险的补偿,而与非系统风险无关。 2E(R)=R+β⋅[E(R)−R], 其中, β=σ/σ ifiMfiimmFinancial Economics_WCY11Financial Economics_WCY122
4. CAPM的数学推导Assume a market portfolio M, a risk-free asset R, and risky 讨论:fasset I. A portfolio consisting of a percent invested in I and 是一条向上倾斜的直线,它显示出收益(1-a) invested in M has the following mean and standard 率与beta的关系:deviation:R=R+(RP)β=R+[E(R)−R]βE(R) = aE(R)jfm+(1-a)E(R) σ(R)=[aσ+(1-a)σ+2a(1-a) σ]2.[E(R) -R]是市场风险溢价imimmf∂E(R)=E(R)-E(R)∂aim3.[E(R)-R]/β=[E(R)-R]/β.它表明,风险回AfBf报率对于市场上所有的资产是相同的。∂σ(R)∂a=[aσ+(1-a)σ+2a(1-a) σ]imim222*[2aσ-2σ+2aσ+2σ-4aσ]immimimFinancial Economics_WCY13Financial Economics_WCY14a in the above equations is the excess demand for the asset Rearrange the equation, we can get CAPM as which should be zero. Evaluating at a=0, we get the risk-return trade off in equilibrium as:2E(R)=R+[E(R)-R]( σ/σ)ifmfimm∂E(R)/∂aE(R)−E(R)im=2∂σ(R)/∂a(σ−σ)/σimmmThe last term is called risk premium that is the The slope of the marginal rate of substitution must be price of risk multiplied by the quantity of risk. Let equal to the slope of capital market , 2β=σ/σbe the quantity of (R)−RE(R)−E(R)mfim=2σ(σ−σm)/σmimmFinancial Economics_WCY15Financial Economics_WCY16CAPM之文字证明5 资本资产定价模型的含义如果投资者因非系统风险而获得额外回报,那 末风险相同的分散投资组合中,有较多非系统 E(R)=R+β⋅[E(R)−R] ifiMf风险的组合比非系统风险较少的组合收益高。 投资者会想方设法买进非系统风险较高组合的证券的期望收益包含两个部分: 股票,卖掉BETA相同但非系统风险较低的股①资金的纯粹时间价值,即无风险收益率R;这一部分代表了对投资者因购买该股票而推迟消费 f票。前者的价格被拉抬上涨。(但不承担风险)的补偿,即该股票的收益率至少应大于这个无风险资产的收益率。 ②证券系统风险的报酬率β⋅[E(R)−R]以上过程会持续到 ,这一部分代表了投资者不但推迟了消费同时还面临 beta相同的股票都有相同的iMf预期收益,非系统风险不再有RP为止。着股票价格波动而带来的风险,即应该给投资者以风险补偿。其中[E(R)−R]反映单位系统风险 Mf所应得到的报酬。 Financial Economics_WCY17Financial Economics_WCY183
应用CAPM一例讨论:例:1. E(R)-R是市场风险溢价,即在市场风险下的额外mf无风险利率= 6%ABC 股票的= 2;市场回收益。2. 三种情况:的期望报酬= 10%a. 当β= 0 时,其期望收益就是无风险收益率。那么ABC 股票的必要收益率(required rate of (在CAPM 等式中,让β= 0)。return) = 6% + 2* (10% -6%)b.当β= 1时,其收益率等于市场组合收益率。(在CAPM 等式中,让β= 1)= 6% + 8% = 14%c. 当股票的β小于1 (大于1) 时,该股票的收益率将低于(高于) 市场组合收益率。d.当β< 0?Financial Economics_WCY19Financial Economics_WCY206. CAPM之应用找出价格不合理的股票 计算投资组合的风险例:ABC 股票的当前价= $10 投资组合的风险以β值表示,所以一年后的预期价= $11nβ=∑wβpii期望股息(dividend) = $ 1i=1ABC 股票的期望收益率衡量某一投资组合的特定风险水平,= ( 期望价格-目前价格) + 股息就需确定各股票的正确β值。目前的价格= (11 -10 + 1)/10= or 20%根据CAPM,ABC 股票的必要报酬率是14%。由于股票的期望报酬率是20%,该股票价值被低估。Financial Economics_WCY21Financial Economics_WCY227.估计β系数的两种方法及基本程序β 作为衡量个股i价格的变化率对市场指数变化率的敏感程度, 回归法可以用来表示该股票系统风险的大小 To get an estimatableequation, add an intercept (α), an i22error term (e) and define β= σ/ σ,这样β=cov(R,R)/Var(R)=σ/σ, iimmimmiiimmR–R= α+ β(R-R)+ecov(R,R)是股票i收益率与市场组合收益率的协方差, imifjimfi而Var(R)是市场组合收益率的方差。 Note: 1). e表示非系统性风险对风险溢价的影响并假设mi2e~ (0,σ) ie 2). OLS估计的β=资产i的beta值iCov(Ri−R,R−R)σfmfimˆβ==3) . 2Var(R−R)mσfmFinancial Economics_WCY23Financial Economics_WCY244
Estimating CAPM一般用历史24-60月收益估计一证券或组合的Regression review:beta值。Objective: Find the line that “fits the best”beta值在较长时序序列中可能不稳定。例如它“Fits the best” criteria:随经济周期而变化。∑e= 0i2组合的beta估计方法与单个证券的类似。有人Minimize ∑(e)ir-rSlope = βi f 也用单个证券的beta中值来代表组合 = αieir-rM f Financial Economics_WCY25Financial Economics_WCY26Example 1Beta is the slope of the regression line of the individual stock returns against the market portfolio returns. r iEmpirically, beta is calculated by observing monthly rates of return for the stock and regresion line the market; r = + rim plotting the observationsagainst market returns; Month rm ri 1 15% 18% and, 2 -5 -10 3 12 16fitting a regression line showing the average returns to the stock (dependent variable) at different market returns (independent variable). rm The slope of the line, beta, relates to the change in stock returns, given a change in market Economics_WCY27Financial Economics_WCY28Example 2: Regression Approach -ExcelRegression Approach -ExcelUS Utilities IndexUS Technology IndexSUMMARY OUTPUTSUMMARY %%Regression StatisticsRegression %%Multiple Utilities IndexUS Technology IndexR %%Adjusted R R %%%%%%%%%%%%%%%%%%%%%%Market Index ReturnsMarket Index ReturnsCoefficientsStandard Errort StatP-valueCoefficientsStandard Errort Variable -17X Variable -73Financial Economics_WCY29Financial Economics_WCY305Utilities Index ReturnsTechnology Index Returns
MonthGM returnS&P500Monthly T-billExcess GM returnExcess M Resultsr−r=α+β[r−r]GMfMfHβEstimated coefSTD of estimateVar. of residualsFinancial Economics_WCY31Financial Economics_WCY32STD of residualsR-squareMonthGM returnS&P500Monthly T-billExcess GM returnExcess M 简化证券组合决策例:Suppose that E(R) = 4 and σ= the following estimates on two 1Asset αβσ (R) = 2 + *4 = 8σ= * 4 + = E(R) = 3 + *4 = 5σ= * 4 + = Resultsr−r=α+β[r−r]GMfMfσ= **4 = 312HβThus we have the five inputs required to derive the Estimated of estimate()()Var. of residuals = of residuals = Economics_WCY33Financial Economics_WCY34R-square = 系统风险占总风险比例的推导In order to calculate an optimal portfolio many inputs are 对于一元回归模型: R=θ+Rb+ε itimtitrequired. One requires the expected return and risk on the market 可以证明: portfolio and in addition there are N α, β,and σ.iiei Thus there are 3N+2 inputs σ=βσ+σ iimεPreviously22其中βσ表示个股受整体市场变化的影响,即系统风险, imNumber of assets10100250250它衡量的是整个市场大势运动引起的股票收益率波动性; α10100250i2σ为仅由该股票自身因素决定的非系统风险的度量 β10100250εiσ10100250eiTotal3230275231625The number of inputs required decreases dramaticallyFinancial Economics_WCY35Financial Economics_WCY366
系统风险在某股票总风险中所占比例为:系统风险在某股票总风险中所占比例为: 计算方法1: 计算方法2: 22222βσCov(R,R)σCov(R,R)imitmtm2itmt δ==•==ρi222im σσσσσ根据方差的定义可知,对以上模型进行线性回归所 imimi而非系统风险在总风险中所占比例为: 2得到的拟合系数R 221-δ=1-ρρ 其中为个股收益率与市场指数收益率相关系数的平方。iim im 从公式中可见,单个股票收益率波动中能被市场指数收益率运动解释的部分, 表示了解释变量的变化对占被解释变量变化的比例,也就是系统 即这两种收益率间相关系数的平方,就是系统风险在总风险中的构成比例。 风险在某股票总风险中所占比例 2(R−R)(R−R),即δ=R。 ∑itimtmi相关系数的计算?ρ= im22((R−R))((R−R))∑iti∑mtmFinancial Economics_WCY37Financial Economics_WCY38Decomposing 国际比较If IBM stock has an annual standard deviation of 21%, where as that of the S&P 500 is only 16%, then we can decompose total risk of IBM given its beta of :研究表明,以上海证券交易所上市的50家A股为研究对象,利用1993年4月27 Beta of IBM xSpecific RiskVariability of=Marke+日至1996年5月31日的数据得出的结果是在50个样本股票中,有42个股票的系 t IBMIBMVariability统风险所占比例超过了70%,平均比例达% 22221%={ x 16%}+ 10% 西方股市中系统风险占总风险的比例下表所示。 =+ R-squared = Proportion of variance due to market 美国 英国 法国 德国 加拿意大瑞士 = = %大 利 %%% %20% % % Financial Economics_WCY39Financial Economics_WCY40 系统风险占总风险比例的另一个应用:适度投资组合规模的确定上海股市风险结构的动态变化特征过多的证券构成的组合将导致较高的管理费用,因此需要研究股 票数目与风险程度的关系。基本思路是研究股票数目增加所引起的投资组合系统风险占总风上海股市系统风险占总风险比例变化情况 险比例指标的变化。由于投资组合分散风险的作用主要在于它能够消除非系统性风险;则随着组合中股票数量的增加,与市场相关 的系统性风险应该成为解释组合收益率变化的主要因素。随着组合中股票数量的增加,系统风险占总风险的比例应该由快速上升到上升缓慢,直至为一稳定水平。此时对应的组合规模即1993 1994 1995 1996 1997 1998 为适度规模。% % % 52% % % Financial Economics_WCY41Financial Economics_WCY427
例如:王新鸣<<上海股票市场风险股票风险结构和组合例:陆家嘴股份有限公司B股β系数的估算及其系统风险效果所占的比例>> 股票1 2 3 4 5 6 7 8 9 10 15 组合中含股票数量 2 上表中可见:作为评价系统风险占总风险比例的指标的拟合度R-square,由1只股票组合的43%上升到6只股票组合时的80%,并且随着组合中股票数量的进一步增加,与市场相关的系统性风险占总风险的比例已上升的十分缓慢,说明分散投资对消除非系统风险的效果已十分微弱,6只股票组成的投资组合已能起到很好的降低风险的作用,为适度的组合规模。Financial Economics_WCY43Financial Economics_WCY44例题:陆家嘴股份有限公司B股β系数的估算及其系统风例题:陆家嘴股份有限公司B股β系数的估算及其系统风险所占的比例险所占的比例 用一元线性回归模型对R和R之间存在的相关关系,加以研究,形式为: itmtR=α+βR+ε, t=1,2,...,T (T为样本期数) itiimtt P−PI−Iitit−1tt−1其中,R=,R= , itmtPIit−1t−1 P和I分别为股票i和市场指数在t时刻的价格, ittR和R分别为陆家嘴B股的周收益率和市场指数的周收益率 itmtFinancial Economics_WCY45Financial Economics_WCY46回归结果结论: 综合结果为: Dependent Variable: RI R=−+ itmtMethod: Least Squares Date: 03/30/02 Time: 12:36 )Rm前的系数β的估计值β为, Sample(adjusted): 2 198 Included observations: 197 after adjusting endpoints )β− t-Statistic Prob. t=)==,虚拟假设“β=0”成立的概率P值=0, β=0C (β) -squared Mean dependent var 因此市场指数收益率R对陆家嘴B股的收益率的影响是显著的。 mtAdjusted R-squared . dependent var of regression Akaike info criterion 模型拟合的判定系数R=。说明陆家嘴B股价格变化的72%可 Sum squared resid Schwarz criterion 由上证B股指数(市场整体走势)的变化来解释,即对陆家嘴B股这只股票而言, Log likelihood F-statistic Durbin-Watson stat Prob(F-statistic) 系统风险在投资总风险中所占的比例为72%。 Financial Economics_WCY47Financial Economics_WCY488
9. 检验CAPMTesting the CAPM –Sharpe-Cooper (FAJ, 1972)Testing the CAPM -Black, Jenson and Scholes(1972)Compute average annual returns and betas for 10 portfolios sorted by beta values estimated from a Compute average monthly returns and betas for prior 5 year periodportfolios sorted by beta values estimated from a equal-weighted portfolios of NYSE stocksprior period, then estimate cross-sectional strong positive relation between risk & returns:regression for 8 year sample period.equal-weighted portfolios of NYSE stocksslope, or risk premium = %;strong positive relation between risk & return in intercept, or “zero-beta” return = %1931-1939, smaller in 1939-47, and actually negative (but small) from Economics_WCY49Financial Economics_WCY50BJS对1931—1965年间美国纽约证券交易所所10个不同风险投资组合的系统风险与平均月收有上市公司的股票进行了研究,发现实际的回益率1931-1965归结果与理论并不完全相同。BJS得出的实际月收益率(%)实际关系的风险与收益关系比CAPM 模型预测的斜率要10CAPM小,同时表明实际的αp在β值大时小于零,而在β值小时大于零。这意味着低风险的股票获得了理论预期的收益,而高风险股票获得低012贝塔于理论预测的收益。-2Financial Economics_WCY51Financial Economics_WCY52Testing the CAPM –Fama-Macbeth (1973, JPE)Empirical model:Estimate cross-sectional regression of returns vs. R= c+cβ+ cβ+cs+εito1ti2i23tiitbetas for portfolios sorted by beta values Hypothesis:estimated from a prior period, then average the E(c)=02estimates of risk premium (slope) and the risk E(c)=03free rate (intercept)E(c)= E[R] –E[R] > 01tmfFama-Macbeth的CAPM检验区别于时间序列检Sharpe-Lintnerhypothesis: E(c) =R0f验的特点在于它采用了横截面的数据进行分析Financial Economics_WCY53Financial Economics_WCY549
MTwo stage procedure:ethodology:First stage: individual beta and non-beta risk at t-1 are Market portfolio: equally weighted average of the returns estimated using previous 5-year data. Portfolio betas and of common stocks on NYSEportfolio non-beta risks are also measure is subject to errors-in-variable problem. Beta Second stage: monthly cross-sectional regressionsis estimatedR= γ+γβ+γβ+γs+ηpt0t1tp,t-12tp,t-123tp,t-1ptcov(R,R)imThe coefficient estimates are averaged during the sample β=ivar(R)mperiods to get the final result.Fama-MacBethconclusions:Rolling regressions to get estimate of beta, and non-beta risk①γ1的均值为正值,在95%的置信度下可以认为不为零,R= d+βR+εit0iimtit表明收益与β值成正向关系The standard deviation s(ε) is a proxy for non-beta ②γ2、γ3在95%的置信度下值为零,表明其他非系统性20 equal-weighted portfolios of NYSE stocks风险在股票收益的定价中不起主要作用。Financial Economics_WCY55Financial Economics_WCY56Roll’s CritiqueTesting the CAPM: Financial market anomaliesOne of the most useful applications of the securities market Small cap firms have higher average returns than line is that its ex post form or empirical market line can be used as a benchmark for security by the CAPM (., risk adjusted)Roll (1977) argues that even if markets are efficient and the High E/P (low P/E) stocks have higher average CAPM is valid, the cross section security market line cannot be used as a means of measuring the ex post performance of returns than predicted by the CAPMportfolio selection techniques. Why?High book/market ratio is associated with higher The market portfolio is not unique. The only requirement is that it is returnsA zero beta portfolio and an efficient market portfolio can replace R(f) and R(m) in CAPM. It is shown that if the Beta-return relation in 80s looked like a little index is e4fficient ex post, there will be no abnormal return. brother of a smart there are abnormal returns, it means the index is not efficient ex post.The only way to test the CAPM is to see whether or not the true market portfolio is ex post Economics_WCY57Financial Economics_WCY58我国学者对风险-收益关系的检验经常遭遇的问题:参考材料:股票的样本太少,不代表市场总体,无法得出C&W, Chapter 7市场上风险与收益的实际关系。William Sharpe, Capital asset prices: A theory of 在两次回归中,同时选用同一时期的数据进行market equilibrium under conditions of risk. 估计和对CAPM模型中线性关系的验证。Journal of Finance, 1964, 425-442.在确定收益率时没有考虑分红,送配带来的影Roll, R., A critique of the asset pricing theory’s 响并做相应调整,导致收益和风险的估计的偏tests. Journal of Financial Economics, 1977, 129-差,严重影响分析的准确性。176.在回归过程中,没有选用组合的构造,而是采用个股的回归易导致,系数的不稳定性。Financial Economics_WCY59Financial Economics_WCY6010
Assignments1.简述CAPM的核心思想及其现实意义。2.用简单的Software计算任何三只股票的beta。设计一种检验方法来检测beta对未来股票收益的影响.3.衡量和描述我国股票基金的风险和收益关系(可用月收益数据)。Financial Economics_WCY6111