Part 2
Principles of Passive Management and Asset Pricing
2
Chapter 8
Risky asset pricing models
and the CAPM
3
Learning objectives
• After completing this chapter you should be able to:
– intuitively understand the:
• capital market line (CML)
• capital asset pricing model (CAPM) and the security market line
(SML)
• link between the CML and SML
– apply the CAPM in pricing risky assets including shares
4
Learning objectives (cont.)
– use the CAPM and justify values for relevant input parameters
– understand the importance of assumptions underlying the CAPM
– explain the difficulties associated with testing the CAPM
– debate whether the CAPM is an appropriate model to price risky
assets.
5
Chapter outline
1 Introduction
2 The capital asset pricing model (CAPM)
3 Using the CAPM
4 Testing the CAPM
5 Extensions to the CAPM
6 Summary
6
1 Introduction
• This chapter addresses how an investor can determine the
appropriate return required for a particular asset.
– In a competitive market, price is set by interaction of supply
and demand.
– This price should be commensurate with the risk of the asset.
• That is, how much return is sufficient to compensate investors for
holding the asset.
7
2 The capital asset pricing model
(CAPM)
• Capital market line (CML)
– Equation of CML:
E(Ri) = Rf + si(E(Rm)-Rf)/sm
– Assumptions:
• Investors are risk averse, maximise utility, and face one period
horizon.
• Investors are only interested in risk and return, and have access to
unlimited borrowing and lending at Rf
• That there are no transaction costs or taxes.
• Investments are infinitely divisible.
8
2 The capital asset pricing model (CAPM) (cont.)
2 The capital asset pricing model
(CAPM) (cont.)
• Determine:
– how the aggregate of investors will behave
– how market clearing prices and returns are set
– the relevant measure of risk
– the relationship between risk and return for any asset when
markets are in equilibrium.
10
2 The capital asset pricing model
(CAPM) (cont.)
• The capital assessing price model (CAPM) was developed by
Sharpe, Lintner and Mossin in 1965.
• Security market line (SML)
– E(Ri) = Rf + bi(E(Rm)-Rf)
where bi = Covi,m/s2m
11
2 The capital asset pricing model (CAPM) (cont.)
3 Using the CAPM
• Given the CAPM equation, E(Ri) = Rf + bi(E(Rm)-Rf), the
following inputs are needed to apply the CAPM:
– Rf: risk-free rate
– E(Rm)-Rf : market risk premium
– bi : beta specific to the asset.
13
3 Using the CAPM (cont.)
• Risk-free rate
– Risk-free rate should have a duration equal to the asset
life.
– CAPM is a one-period model.
– To take a long-term security is typical.
– To remove risk, government bonds are used (but note
these still have some risk).
14
3 Using the CAPM (cont.)
• Risk-free rate (cont.)
– Using a long-term government bond (e.g. 10-year
Commonwealth bond) is common.
– Risk-free rate varies over time.
– To smooth volatility, an average is often employed.
• E.g. the average over the last 40 trading days.
15
3 Using the CAPM (cont.)
• Market risk premium
– This requires an estimate of E(Rm)-Rf.
– Ex-ante estimates are often made using ex-post returns.
– Using a historical average is typical.
• However, history provides varying estimates:
– Ibbotson in USA: 8.4% pa.
– Other estimates: 5-9% pa.
16
3 Using the CAPM (cont.)
• Market risk premium
– Australia:
• Officer: 7.9% pa.
• Dimson et al: 7.6% pa.
– Recent study:
• 1882–2005
• Brailsford et al: 6% pa.
17
3 Using the CAPM (cont.)
• Beta
– Beta estimates can be supplied through commercial vendors
– Beta estimates can be calculated directly
• require returns on the asset (such as a share)
• require returns on a market index
– e.g. S&P/ASX 200 index
– use accumulation index (includes dividends).
18
3 Using the CAPM (cont.)
• Beta (cont.)
– From the CAPM, bi = covi,m/s2m
since: covi,m = ri,m . si . Sm
therefore: bi = ri,m . si /s2m
hence, Beta can be estimated using the following regression: E(R i) = ai
+ bi(E(Rm,t) + ei,t.
– Beta estimates are imprecise:
• They vary according to data used for estimation.
• Thin trading creates bias.
• They vary over time for same firm.
– Using five years of monthly returns is typical.
19
4 Testing the CAPM
• Early tests of CAPM
– Regress current returns against estimates of beta from past return
series:
R*i = a0 + a1bi + ei
where R*i = Ri – Rf.
– Tests focused on:
• a0 = 0 (CAPM should completely explain return)
• bi explaining return variation in total, and a linear relationship
in beta
• a1 = risk premium
• positive market risk premium, i.e. E(Rm)>Rf.
20
4 Testing the CAPM (cont.)
• Early tests of CAPM (cont.)
– Evidence generally found:
• a1 was statistically significant
– i.e. positive relationship between risk and return
• a0 also statistically significant
– (should be 0)- other factors in pricing assets.
21
4 Testing the CAPM (cont.)
• Early tests of CAPM (cont.)
– Tests of CAPM require an estimate of market return.
– Roll (1977) argued that the market cannot be observed in total.
– Even if possible, tests of CAPM are a test of the efficiency of the
market.
– There are techniques to avoid some of these problems.
22
4 Testing the CAPM (cont.)
• Early tests of CAPM (cont.)
– Problems with testing the CAPM
1. CAPM is an ex-ante model
– tests can only be done on ex-post data
2. Return estimation
- arithmetic/geometric/continuous
- choice of Rf
3. Beta estimation – stability, thin trading
4. Survivorship bias.
23
4 Testing the CAPM (cont.)
• Average stock returns are positively related to market.
• Fama and French (1992)
– 50 year analysis in the USA
• after controlling for size, there is actually a negative
relationship between beta and realised return.
24
4 Testing the CAPM (cont.)
• Beta is dead
– Subsequent studies have provided evidence that conflicts with
Fama and French (1992).
– Concerns include:
• CAPM provides for expected returns not actual returns
• criticism of data mining
• Roll’s problem of an inaccurate market proxy
• a poor proxy will create low correlation and hence low beta; high
betas are relatively less affected an hence SML is flat.
25
5 Extensions of the CAPM
• Researchers have investigated the impact of relaxing the CAPM
assumptions:
– investors make decisions on mean and variance, (normal distribution)
– all assets are marketable
– existence of risk-free asset
– perfect capital markets
– unequal borrowing and lending rates
– zero taxes and transactions costs
– homogenous expectations
– one-period horizon.
26
5 Extensions of the CAPM (cont.)
• Non-normal distribution
– Kraus and Litzenberger (1976)
– Adjust CAPM
• If returns are positive skewed, reduce expected return.
• If returns are negative skewed, increase expected return.
• Non-marketable assets
– Mayers (1972) provides an adjustment to CAPM for thin trading.
27
5 Extensions of the CAPM (cont.)
• Taxes
– CAPM is usually applied to after company tax before personal tax
(to make it easier to identify cash flows).
– Under imputation, distinctions between personal taxes and
corporate taxes are no longer independent.
– Taxes adjust CAPM.
28
5 Extensions of the CAPM (cont.)
• Taxes (cont.)
– Under the assumption of a closed economy and an unchanged risk-
free rate, the after corporate tax (but before personal tax) CAPM
becomes:
• E(Re,i) = Rf/e + be,i[E(rm)m/e - Rf/e]
• where
– E(Re) is the after-corporate-tax return on the share given an imputation
tax system
– E(rm) is the after-corporate-tax return on the market given imputation tax
system.
29
5 Extensions of the CAPM (cont.)
• Taxes (cont.)
– e = [Xsu + Xsf/(1-tc) + tgXsg]
– m = [Xmu + Xmf/(1-tc) + tgXmg]
– where
• tg = capital gains tax rate set to 0.5 of the personal tax rate (t p).
• tc = corporate tax rate.
• xsu = proportion of share return consisting of unfranked dividends.
• xsf = proportion of share return consisting of franked dividends.
30
5 Extensions of the CAPM (cont.)
• Taxes (cont.)
– xsg = proportion of share return consisting of capital gains.
– xmu = proportion of market return consisting of unfranked dividends.
– xmf = proportion of market return consisting of franked dividends.
– xmg = proportion of market return consisting of capital gains.
– Note: By definition, xsu + xsf + xsg = 1, and xmu + xmf + xmg = 1.
31
5 Extensions of the CAPM (cont.)
• No risk-free asset
– Return on govt. bonds is usually used as proxy.
– CAPM can be derived using a zero beta portfolio.
• Transaction costs
– Difficult to gauge impact but if a marginal trader is a large
institution with low trans. costs, then impact is likely to be
small.
32
5 Extensions of the CAPM (cont.)
5 Extensions of the CAPM (cont.)
• Heterogeneous (non-homogeneous) expectations
– can be dealt with under CAPM.
• Multi-period investment horizon
– CAPM can be extended to multi-period.
– a complex model and not generally used in practice.
34
6 Summary
• Start with CML, which prices efficient portfolios.
• Link CML to the SML.
• CAPM allows for pricing of inefficient portfolios such as single
assets.
• Different assets are distinguished only by variations in b.
• The application of CAPM relies critically on the input parameter
values.
35
6 Summary (cont.)
• Early evidence is supportive of at least the zero beta
CAPM or single index model.
• A fundamental difficulty lies with identifying the
appropriate proxy (Roll’s critique).
• Recent evidence does not support CAPM, especially over
long horizons.
36
6 Summary (cont.)
• Other factors that appear relevant to pricing are:
– size
– book-to-market.
• CAPM is a one period model with restrictive assumptions.
• CAPM is most widely used to price equation for equities.
• If we reject the CAPM what other model do we use?
37
Online video activity
• Title: Asset Pricing Empirical Tests
• Location: http://www.youtube.com/watch?v=RyDWtWyCiBY
• Concept: Asset pricing
• Clip description: This video summarises some empirical studies about
asset pricing theory
• Points to consider: Try to search online for more detailed articles about
CAPM
Online video activity
• Title: How the SML (Security Market Line) works
• Location: http://www.youtube.com/watch?v=b7Mtqh18DlA
• Concept: Security market line
• Clip description: This video provide an overview for how the SML
(Security Market Line) works.
• Points to consider: Could you tell the difference between the CML and
SML?