1、Modern Portfolio TheoryThe Factor Models and The Arbitrage Pricing Theory,Chapter 8By Ding zhaoyong,Return-generating Processand Factor Models,Return-generating processIs a statistical model that describe how return on a security is produced.The task of identifying the Markowitz efficient set can be
2、 greatly simplified by introducing this process.The market model is a kind of this process, and there are many others.,Return-generating Processand Factor Models,Factor modelsThese models assume that the return on a security is sensitive to the move-ments of various factors or indices.In attempting
3、to accurately estimate expected returns, variances, and covariances for securities, multiple-factor models are potentially more useful than the market model.,Return-generating Processand Factor Models,Implicit in the construction of a factor model is the assumption that the returns on two securities
4、 will be correlated only through common reactions to one or more of the specified in the model. Any aspect of a securitys return unexplained by the factor model is uncorrelated with the unique elements of returns on other securities.,Return-generating Processand Factor Models,A factor model is a pow
5、erful tool for portfolio management.It can supply the information needed to calculate expected returns, variances, and covariances for every security, which are the necessary conditions for determining the curved Markowitz efficient set.It can also be used to characterize a portfolios sensitivity to
6、 movement in the factors.,Return-generating Processand Factor Models,Factor models supply the necessary level of abstraction in calculating covariances.The problem of calculating covariances among securities rises exponentially as the number of securities analyzed increase.Practically, abstraction i
7、s an essential step in identifying the Markowitz set.,Return-generating Processand Factor Models,Factor models provide investment managers with a framework to identify important factors in the economy and the marketplace and to assess the extent to which different securities and portfolios will resp
8、ond to changes in these factors.A primary goal of security analysis is to determine these factors and the sensitivities of security return to movements in these factors.,One-Factor Models,The one-factor models refer to the return-generating process for securities involves a single factor. These fact
9、ors may be one of the followings:The predicted growth rate in GDPThe expected return on market indexThe growth rate of industrial produc-tion, etc.,One-Factor Models,An example Page 295: Figure 11.1,One-Factor Models,Generalizing the exampleAssumptionsThe random error term and the factor are uncorre
10、lated. (Why?)The random error terms of any two securities are uncorrelated. (Why?),One-Factor Models,Expected returnVarianceCovariance,One-Factor Models,Two important features of one-factor modelThe tangency portfolio is easy to get.The returns on all securities respond to a single common factor gre
11、ater simplifies the task of identifying the tangency portfolio.The common responsiveness of securities to the factor eliminates the need to estimate directly the covariances between the securities.The number of estimates: 3N+2,One-Factor Models,The feature of diversification is true of any one-facto
12、r model.Factor risk:Nonfactor risk:Diversification leads to an averaging of factor riskDiversification reduces nonfactor risk,One-Factor Models,Multiple-Factor Models,The health of the economy effects most firms, but the economy is not a simple, monolithic entity. Several common influences with perv
13、asive effects might be identifiedThe growth rate of GDPThe level of interest rateThe inflation rateThe level of oil price,Multiple-Factor Models,Two-Factor ModelsAssume that the return-generating process contains two factors.,Multiple-Factor Models,The second equation provides a two-factor model of
14、a companys stock, whose returns are affected by expectations concerning both the growth rate in GDP and the rate of inflation.Page 301: Figure 11.2To this scatter of points is fit a two-dimensional plane by using the statistical technique of multiple-regression analysis.,Multiple-Factor Models,Four
15、parameters need to be estimated for each security with the two-factor model: ai, bi1, bi2, and the standard deviation of the random error term.For each of the factors, two parameters need to be estimated. These parameters are the expected value of each factor and the variance of each factor. Finally
16、, the covariance between factors.,Multiple-Factor Models,Expected returnVarianceCovariance,Multiple-Factor Models,The tangency portfolioThe investor can proceed to use an optimizer to derive the curve efficient set.DiversificationDiversification leads to an averaging of factor risk.Diversification c
17、an substantially reduce nonfactor risk.For a well-diversified portfolio, nonfactor risk will be insignificant.,Multiple-Factor Models,Multiple-Factor Models,Sector-Factor ModelsSector-factor models are based on the acknowledge that the prices of securities in the same industry or economic sector oft
18、en move together in response to changes in prospects for that sector.To create a sector-factor model, each security must be assigned to a sector.,Multiple-Factor Models,A two-sector-factor modelThere are two sectors and each security must be assigned to one of them.Both the number of sectors and wha
19、t each sector consists of is an open matter that is left to the investor to decide.The return-generating process for securities is of the same general form as the two-factor model.,Multiple-Factor Models,Differing from the two-factor model, with two-sector-factor model, F1 and F2 now denote sector-f
20、actors 1 and 2, respectively. Any particular security belongs to either sector-factor 1 or sector-factor 2 but not both.,Multiple-Factor Models,In general, whereas four parameters need to be estimated for each security with a two-factor model (ai1,bi1,bi2 , ei,), only three parameters need to be est
21、imated with a two-sector-factor model. (ai1,ei, and eitherbi1 or bi2 ).Multiple-factor models,Estimating Factor Models,There are many methods of estimating factor models. There methods can be grouped into three major approaches:Time-series approachesCross-sectional approachesFactor-analytic approach
22、es,Factor Models and Equilibrium,A factor model is not an equilibrium model of asset pricing.Both equation show that the expected return on the stock is related to a characteristic of the stock, bi or i. The larger the size of the characteristic, the larger the assets return.,Factor Models and Equil
23、ibrium,The key difference is ai and rf.The only characteristic of the stock that determine its expected return according to the CAPM is ii, as rff denotes the risk-free rate and is the same for all securities.With the factor model, there is a second characteristic of the stock that needs to be estim
24、ated to determine the stocks expected return, aii.,Factor Models and Equilibrium,As the size of ai differs from one stock to another, it presents the factor model from being an equilibrium model.Two stocks with the same value of bi can have dramatically different expected returns according to a fact
25、or model.Two stocks with the same value of i will have the same expected return according to the equilibrium-based CAPM.,Factor Models and Equilibrium,The relationship between the parameters ai and bi of the one-factor model and the single parameter i of the CAPM.If the expected returns are determin
26、ed according to the CAPM and actual returns are generated by the one-factor market model, then the above equations must be true.,Arbitrage Pricing Theory,APT is a theory which describes how a security is priced just like CAPM.Moving away from construction of mean-variance efficient portfolio, APT in
27、stead calculates relations among expected rates of return that would rule out riskless profits by any investor in well-functioning capital markets.,Arbitrage Pricing Theory,APT makes few assumptions.One primary assumption is that each investor, when given the opportunity to increase the return of hi
28、s or her portfolio without increasing its risk, will proceed to do so.There exists an arbitrage opportunity and the investor can use an arbitrage portfolios.,Arbitrage Opportunities,Arbitrage is the earning of riskless profit by taking advantage of differential pricing for the same physical asset or
29、 security. It typically entails the sale of a security at a relatively high price and the simultaneous purchase of the same security (or its functional equivalent) at a relatively low price.,Arbitrage Opportunities,Arbitrage activity is a critical element of modern, efficient security markets.It tak
30、es relatively few of this active investors to exploit arbitrage situations and, by their buying and selling actions, eliminate these profit opportunities.Some investors have greater resources and inclination to engage I arbitrage than others.,Arbitrage Opportunities,Zero-investment portfolioA portfo
31、lio of zero net value, established by buying and shorting component securities .A riskless arbitrage opportunity arises when an investor can construct a zero-investment portfolio that will yield a sure profit.,Arbitrage Opportunities,To construct a zero-investment portfolio, one has to be able to se
32、ll short at least one asset and use the proceeds to purchase on or more assets.Even a small investor, using borrowed money in this case, can take a large position in such a portfolio.There are many arbitrage tactics.,Arbitrage Opportunities,An example:Four stocks and four possible scenariosthe rate
33、of return in four scenariosPage 180-181 in the textbookThe expected returns, standard deviations and correlations do not reveal any abnormality to the naked eye.,Arbitrage Opportunities,The critical property of an arbitrage portfolio is that any investor, regardless of risk aversion or wealth, will
34、want to take an infinite position in it so that profits will be driven to an infinite level.These large positions will force some prices up and down until arbitrage opportunities vanishes.,Factor Models and Principle of Arbitrage,Almost arbitrage opportunities can involve similar securities or portf
35、olios.That similarity can be defined in many ways.One way is the exposure to pervasive factors that affect security prices.An example Page 324,Factor Models and Principle of Arbitrage,A factor model implies that securities or portfolios with equal-factor sensitivities will behave in the same way exc
36、ept for nonfactor risk.APT starts out by making the assumption that security returns are related to an unknown number of unknown factors.Securities with the same factor sensitivities should offer the same expected returns.,Arbitrage Portfolios,An arbitrage portfolio must satisfy:A net market value o
37、f zeroNo sensitivity to any factorA positive expected return,Arbitrage Portfolios,The arbitrage portfolio is attractive to any investor who desires a higher return and is not concerned with nonfactor risk. It requires no additional dollar investment, it has no factor risk, and it has a positive expe
38、cted return.,One-Factor Model and APT,Pricing effects on arbitrage portfolioThe buying-and-selling activity will continue until all arbitrage possibilities are significant reduced or eliminatedThere will exist an approximately linear relationship between expected returns and sensitivities of the fol
39、lowing sort:,One-Factor Model and APT,The equation is the asset pricing equation of the APT when returns are generated by one factorThe linear equation means that in equili-brium there will be a linear relationship between expected returns and sensitivities.The expected return on any security is, in
40、 equilibrium, a linear function of the securitys sensitivity to the factor, bi,One-Factor Model and APT,Any security that has a factor sensitivity and expected return such that it lies off the line will be mispriced according to the APT and will present investors with the opportunity of forming arbi
41、trage portfolios.Page 327: Figure 12.1,One-Factor Model and APT,Interpreting the APT pricing equationRiskfree asset, rfPure factor portfolio, p*,Two-Factor Model And APT,The two-factor modelArbitrage portfoliosA net market value of zeroNo sensitivity to any factorA positive expected return,Two-Facto
42、r Model And APT,Pricing effects,Two-Factor Model And APT,1 is the expected return on the portfolio which is known as a pure factor portfolio or pure factor play, because it has:Unit sensitivity to one factor (F1, b1=1)No sensitivity to any other factor (F2, b2=0)Zero nonfactor riskThis portfolio is
43、a well-diversification portfolio that has unit sensitivity to the first factor and zero sensitivity to the second factor.,Two-Factor Model And APT,It is the same with 2 . It is the well-diversification portfolio that has zero sensitivity to the first factor and unit sensitivity to the second factor,
44、 meaning that it has b1=0 and b2=1.Such as a portfolio that has zero sensitivity to predicted industrial production and unit sensitivity to predicted inflation would have an expected return of 6%.,Multiple-factor modelThe APT pricing equation,Multiple-Factor Model And APT,The APT And The CAPM,Common
45、 pointBoth require equilibriumBoth have almost similar equationDistinctionsDifferent equilibrium mechanismMany investors v.s. Few investorsDifferent PortfolioMarket portfolio v.s. Well-diversifyed P.,Summary,The Factor ModelsOne-factor modelsMulti-factor modelsFactor models and equilibriumArbitrage opportunity and portfolioThe arbitrage pricing equationOne-factor equationMulti-factor equation,Assignments For chapter 8,ReadingsPage 282 through 301Page 308 through 321ExercisesPage 304: 14,15; Page 323: 4, 13Q/A:Page 302: 3Page 324: 8,
