6-1
单一指数和多因素模型
Single Index and
Multifactor Models
第六章
6-2
马克维茨模型的缺陷:
- 计算量过大.假定分析n种股票,需要计算n个预期值、n个方差
以及(n2 –n)/2个协方差.
- 相关系数确定或者估计中的误差会导致无效结果.
指数模型的优势:
- 大大降低了马克维茨模型的计算量,它把精力放在了对证券的
专门分析中.
- 指数模型以一种简单的方式来计算协方差,证券间的协方差由
单个一般因素的影响生成,为市场指数收益所代表,从而为系统
风险与公司特有的性质提供了重要的新视角.
指数模型的优势
6-3 Announcements, Surprises, and Expected Returns
任一证券的收益由两部分组成(The return
on any security consists of two parts).
1) 预期或一般收益(the expected or normal
return): the return that shareholders in the
market predict or expect
2) 非预期或风险收益(the unexpected or risky
return): the portion that comes from information
that will be revealed .
6-4 Announcements, Surprises, and Expected Returns
任何信息的公布可以被分成两个部分,预期到的部分
和异常部分(Any announcement can be broken down
into two parts, the anticipated or expected part and the
surprise or innovation):
Announcement = Expected part + Surprise.
任何公布的信息中预期部分是市场用来形成股票预期
收益( E(ri). )的信息(The expected part of any
announcement is part of the information the market uses to
form the expectation of the return on the stock , E(ri).)
n 异常部分是那些影响股票非预期收益( U. )的信息
(The surprise is the news that influences the
unanticipated return on the stock, U.)
6-5 有关信息的例子
Examples of relevant information
- Statistics China figures (., GNP)
- A sudden drop in interest rates
- News that the company’s sales
figures are higher than expected
6-6 因素模型
Factor Models
A way to write the return on a stock in
the coming month is:
6-7
实际总收益的构成
ri = E(ri) +U
= E(ri) +m+ei
ri :下个月的实际总收益
E( ri ):实际总收益中的期望收益部分
U:实际总收益中的非期望收益部分
m:系统性风险
ei:非系统性风险
公布信息=期望部分+异动部分
6-8 风险:系统性和非系统性风险
Risk: Systematic and Unsystematic
系统性风险会影响到大部分资产A systematic risk is any
risk that affects a large number of assets, each to a greater
or lesser degree.
非系统性风险只会影响到单一资产或某一小类的资产。
非系统性风险可以被分散掉。An unsystematic risk is a
risk that specifically affects a single asset or small group of
assets. Unsystematic risk can be diversified away.
系统性风险包括那些一般经济状态的不确定性,如
GNP、利率、通货膨胀等。Examples of systematic risk
include uncertainty about general economic conditions,
such as GNP, interest rates, or inflation.
换句话说,一个公司特定的消息,如金矿开采公司发
现黄金,就是非系统性风险。On the other hand,
announcements specific to a company, such as a gold
mining company striking gold, are examples of
unsystematic risk.
6-9
因素模型的特点
作为一种回报率产生过程,因素模型具
有以下几个特点。
- 第一,因素模型中的因素应该是系统影响所有证券
价格的因素。
- 第二,在构造因素模型中,我们假设两个证券的回
报率相关——一起运动——仅仅是因为它们对因素
运动的共同反应导致的。
- 第三,证券回报率中不能由因素模型解释的部分是
该证券所独有的,从而与别的证券回报率的特有部
分无关,也与因素的运动无关。
6-10
因素模型在证券组合管理中的应用
- 在证券组合选择过程中,减少估计量和计算
量
- 刻画证券组合对因素的敏感度
如果假设证券回报率满足因素模型,那
么证券分析的基本目标就是,辨别这些
因素以及证券回报率对这些因素的敏感
度。
6-11
ri = E(ri) + ßiF + ei
ßi = index of a securities’ particular return to the
factor
F= some macro factor; in this case F is
unanticipated movement; F is commonly related
to security returns
Assumption: a broad market index like the S&P500
is the common factor.
单一因素模型
Single Factor Model
6-12 随机误差项
RANDOM ERROR TERMS
- ei 被称为随机误差项
CAN BE CONSIDERED A RANDOM VARIABLE
DISTRIBUTION:
MEAN = 0,即E(ei)=0
VARIANCE = 2ei
任意证券 i 的随机项 ei 与因素F不相关;
任意证券 i 与证券 j 的随机项 ei与 ej 不相关,cov(ei,ej)=0
6-13
- 表6-1 因素模型数据
年份 GDP增长率 A股票回报率
1 % %
2
3
4
5
6
6-14
4%
6-15
上图中,横轴表示GDP的预期增长率,
纵轴表示证券A的回报率。图上的每一点
表示表6-1中,在给定的年份,A的回报
率与GDP增长率的关系。通过线性回归
分析,我们得到一条符合这些点的直线
:rt=a+βGDPt+et。这条直线的斜率β为2,
说明A的回报率与GDP增长率有正的关系。
GDP增长率越大,A的回报率越高。
6-16
在上图中,零因素是4%,这是GDP的预
期增长率为零时,A的回报率。A的回报
率对GDP增长率的敏感度β为2,这是图
中直线的斜率。这个值表明,高的GDP
的预期增长率一定伴随着高的A的回报率。
如果GDP的预期增长率是5%,则A的回
报率为14%。如果GDP的预期增长率增加
1%——为6%时,则A的回报率增加2%,
或者为16%。
6-17
在这个例子里,第六年的GDP的预期增
长率为%,A的实际回报率是13%。因
此,A的回报率的特有部分(由 ei
给出)为%。给定GNP的预期增长率
为%,从A的实际回报率13%中减去A
的期望回报率%,就得到A的回报率的
特有部分%。
6-18 市场模型
THE MARKET MODEL
在实际应用过程中常用证券市场组合来
作为影响证券价格的单因素,此时的单
因素模型被称为市场模型。市场模型实
际上是单因素模型的一个特例。
ri = E(ri ) + ßi [ rM-E( rM )]+ ei
rM :市场组合的实际收益率
E( rM ):市场组合的期望收益率
6-19
(ri - rf) = i + ßi(rm - rf) + ei
Risk Prem(股票持有期超额收益)
Market Risk Prem
or Index Risk Prem
i = the stock’s expected return if the
market’s excess return is zero
ßi(rm - rf) = the component of return due to
movements in the market index
(rm - rf) = 0
ei = firm specific component, not due to market
movements
单一指数模型
Single Index Model
6-20
Let: Ri = (ri - rf)
Rm = (rm - rf)
Risk premium
format
Ri = i + ßi(Rm) + ei
Risk Premium Format
6-21 证券特征线
Security Characteristic Line
Excess Returns (i)
SCL
..
..
......
..
.. ..
.. ....
.. ..
..
.. ..
.. ....
..
..
..
.. ..
.. ....
..
..
.. ..
.. ..
..
.. .. ....
.. ..
.. ......
..
.. ....
..
Excess returns
on market index
Ri = i + ßiRm + ei
......
Ri =αi + ßiRm
6-22
Jan.
Feb.
.
.
Dec
Mean
Std Dev
.
.
.93
.
.
Excess
Mkt. Ret.
Excess
GM Ret.
Using the Text Example
from Table 8-5
6-23
Estimated coefficient
Std error of estimate
Variance of residuals =
Std dev of residuals =
R-SQR =
ßß
()
()
rGM - rf = + ß(rm - rf)
回归结果
Regression Results
6-24
市场风险或系统风险Market or systematic risk:
risk related to the macro economic factor or
market index.
非系统性风险或公司特有的风险Unsystematic
or firm specific risk: risk not related to the
macro factor or market index.
Total risk = Systematic + Unsystematic
风险构成
Components of Risk
6-25 风险:系统性和非系统性风险
Risk: Systematic and Unsystematic
系统性风险会影响到大部分资产A systematic risk is any
risk that affects a large number of assets, each to a greater
or lesser degree.
非系统性风险只会影响到单一资产或某一小类的资产。
非系统性风险可以被分散掉。An unsystematic risk is a
risk that specifically affects a single asset or small group of
assets. Unsystematic risk can be diversified away.
系统性风险包括那些一般经济状态的不确定性,如
GNP、利率、通货膨胀等。Examples of systematic risk
include uncertainty about general economic conditions,
such as GNP, interest rates, or inflation.
换句话说,一个公司特定的消息,如金矿开采公司发
现黄金,就是非系统性风险。On the other hand,
announcements specific to a company, such as a gold
mining company striking gold, are examples of
unsystematic risk.
6-26 风险各组成部分的衡量
Measuring Components of Risk
i2 = i2 m2 + 2(ei)
where;
i2 = 总方差(total variance)
i2 m2 = 系统性方差(systematic variance)
2(ei) = 非系统性方差(unsystematic variance)
ij=cov(ri,rj)=i j m2
6-27 随机误差项ei
THE RANDOM ERROR TERMS ei
THE RANDOM ERROR TERMS ei
- 显示因素模型不能解释的部分shows that the
factor model cannot explain perfectly
- 实际收益和因素模型预期的收益之间的差异
就是ei (the difference between what the
actual return value is and what the model
expects it to be is attributable to ei)
6-28
Total Risk = Systematic Risk + Unsystematic
Risk
Systematic Risk/Total Risk = R2
ßi2 m2
/ 2 = R2
i2 m2 / [i2 m2 + 2(ei) ]= R2
Examining Percentage of Variance
6-29
风险分散——国际经验
6-30
中国股市的分散与风险
6-31 指数模型和分散化
Index Model and Diversification
6-32 分散化
DIVERSIFICATION
Unique Risk
mathematically can be expressed as
6-33 分散化
DIVERSIFICATION
总组合的风险TOTAL PORTFOLIO RISK
- also has two parts: market and unique
Market Risk
分散化会导致市场风险的平均化(diversification leads
to an averaging of market risk)
Unique Risk
越是分散化的组合,其非系统性的风险越小(as a
portfolio becomes more diversified, the smaller will be its
unique risk)
6-34 风险降低和分散化
Risk Reduction with Diversification
Number of
Securities
St. Deviation
Market Risk
Unique Risk
s2(eP)=s2(e) / n
bP2sM2
6-35
Reduces the number of inputs for
diversification.
Easier for security analysts to specialize.
单一指数的优点
Advantages of the Single Index Model
6-36
Determining the inputs needed for locating the efficient set 1
Markowitz:
expected returns: N
variances: N
covariances: (N2-N)/2
Total: (N2+3N)/2
6-37
Determining the inputs needed for locating the efficient set 2
Index model:
For the market index:
expected return 1
variance 1
For each security
vertical intercept N
Beta N
Variance of random error term N
TOTAL 3N+2
6-38
单指数模型举例单指数模型举例————清华同方(清华同方(11))
假定有反映中国股市整体情况的中证300指数,有无风
险利率存在。估算期为1年,计算出每月同方公司的平
均收益水平和中国股市月平均收益水平(虚拟数据),
结果如下。
6-39
同方股票的超额收益与市场超额收益的关系有下式:
RTFt=αTF+TFRMt+eTFt
将这12组数据带入上式进行回归,得到结果如下:
6-40
截距为%,斜率为,残值的方差反映了同
方公司特有因素对同方股票收益的影响程度,表中的
R2表示的是rI与rM之间的相关性的平方,它是总方差
上的系统方差,它告诉我们公司股价小量波动是由市
场波动造成的。
6-41
单指数模型举例单指数模型举例————美林公司美林公司
美林公司用S&P500指数作为市场资产组合,以最近60
个月的每月均值来计算回归参数。为了简便,用总收益
代替了模型中的超额收益,要估计的模型变成
r =α+ rM +e*
只要rf是常数,回归结果就是一样的。式中的值和市
场风险2M与公司特有风险2(e),都可以从证券特征线
中估计出来,美林公司将其评估结果按月刊登在它出版
的月刊《证券风险评估》中,人们通常将其称为“手
册”。以下是手册中的几行。
6-42 多因素模型
Multifactor Models
单因素模型(市场模型)假设市场组合收益
的变化包括了所有的风险因素,而且(不正确
地)假设任何一种股票对每种风险的相对敏感
度都一样
• Use factors in addition to market return
• Estimate a beta for each factor using multiple
regression
• use more than one explanatory variable in the
return-generating process
6-43 系统性风险和
Systematic Risk and Betas
For example, suppose we have identified
three systematic risks on which we want to
focus: Inflation \GDP growth\ spot interest
rate
6-44
多因素模型
Rt = t + ßGDPGDPt + ßIFIFt+
βINTINTt+et
6-45 Example
Suppose we have made the following estimates:
1. βIF =
2. bGDP =
3. bINT= .
Finally, the firm was able to attract a “superstar”
CEO and this unanticipated development
contributes 1% to the return.
e=1%
6-46
We must decide what surprises took place in the
systematic factors.
If it was the case that the inflation rate was
expected to be 3%, but in fact was 8% during
the time period, then
IF = Surprise in the inflation rate
= actual – expected
= 8% - 3%
= 5%
6-47
If it was the case that the rate of GDP growth was
expected to be 4%, but in fact was 1%, then
FGDP = Surprise in the rate of GDP growth
= actual – expected
= 1% - 4%
= -3%
6-48
If it was the case that spot interest rate, S, was
expected to increase by 10%, but in fact remained
stable during the time period, then
FS = Surprise in the interest rate
= actual – expected
= 0% - 10%
= -10%
6-49
Finally, if it was the case that the expected
return on the stock was 8%, then
6-50 Multifactor Models
Use factors in addition to market return
- Examples include industrial production,
expected inflation etc.
- Estimate a beta for each factor using multiple
regression.
6-51 MULTIPLE-FACTOR MODELS
MULTIPLE-FACTOR MODELS
- some of these factors may include
THE GROWTH RATE OF GDP
THE LEVEL OF INTEREST RATES
THE YIELD SPREAD BETWEEN CERTAIN
VARIABLES
THE INFLATION RATE
THE LEVEL OF OIL PRICES
6-52
多因素模型的应用1
Studies by Roll and Ross and by Chen
- 行业生产变动百分比Change in industrial production
- 预期通胀变动百分比Change in Unanticipated
inflation
- 非预期通胀变动百分比Change in expected inflation
- 长期公司债券对长期政府债券的超额收益Excess
return of long-term corporate bonds over long-term
government bonds
- 长期政府债券对短期国库券的超额收益Excess return
of long-term government bonds over T-bills
6-53
多因素模型的应用2
Fama and French
- Returns a function of size and book-to-market
value as well as market returns
6-54
资本资产定价模型与因素模型的关系
CAPM可视为一个特殊的单因素模型,在那里
的市场组合收益率rM实质上就是一个单因素。
以市场组合的收益率的风险补偿来作为宏观经
济指数,于是有:
ri-rf=αi+βi(rm-rf )+ei,
或者Ri=αi+βiRm+ei
(实际上这是证券i对市场组合收益的回归方程,
其回归直线就是证券i的特征线)
6-55
但资本资产定价模型是一个资产定价的均衡
模型,而因素模型却不是。例如,比较分别
由资本资产定价模型和因素模型得到的证券
的预期收益率:
前者不是一个均衡模型,而后者是均衡模型
6-56
既然单因素模型不是一个均衡模型,那单因
素模型中参数αi和βi与资本资产定价模型中单
因素βi之间存在怎样的关系呢?က�
例如,如果实际收益率可以看作是由单因素
模型产生,其中因素F是市场组合的收益率rM
,那么预期收益率将等于:
根据资本资产定价模型,如果均衡存在,则
6-57
这意味着,单因素模型和资本资产定价模型的
参数之间必然存在下列关系:
如果:即对证券的阿尔法的估计值刚好是证券
均衡定价时的截矩,则
即在由CAPM决定的收益率中的测度证券的市
场风险大小的指标与在因素模型决定的收益
率中的因素敏感性大小的值相同,意义相同。
6-58
在资本资产定价模型和市场模型中都有一个被
称为ß值的斜率,并且这两个模型或多或少地
包含了市场,但是它们之间却有明显的区别:
首先,资本资产定价模型是一个均衡模型,它
描述证券的价格如何确定;市场模型是一个因
素模型。
其次,资本资产定价模型是相对于整个市场组
合而言的,即相对于市场中所有证券的集合。
而市场模型是相对于某个市场指数而言,即基
于市场中的一个样本。
6-59
虽然从严格意义上讲,资本资产定价模
型中的ß值和市场模型中的ß值是有区别
的,但是在实际操作中,由于我们不能
确切知道市场组合的构成,所以一般用
市场指数来代替,因此我们可以用市场
模型中测算的ß值来代替资本资产定价模
型中的ß值。