FINA 3103: Intermediate Investments

Lecture 5. CAPM

Don Noh

Hong Kong University of Science and Technology

Spring 2024
Expected Return
Apple Inc. on 31/8/2018

• 4.925 billion shares outstanding

• Closed at $227.63 per share
• Market equity of $1.121 trillion (= $227.63 × 4.925 billion)
• 3.37% of total US stock market

4.925 billion

FINA 3103 Lecture 5. CAPM Spring 2024 1 / 45

Relation Between Price and Expected Return

• Investors pay P0 today.

• In the future, they receive a dividend D1 , and the share price moves to P1 .
• Return is
P1 + D1
1+R = .
P0
• Expected return is P0 overvalued, lower expected return
E(P1 + D1 )
E(1 + R) = .
P0
• High expected return ⇐⇒ low price
• Price of Apple stock adjusts so that investors are collectively happy to hold exactly 4.925
billion shares.

FINA 3103 Lecture 5. CAPM Spring 2024 2 / 45

Capital Asset Pricing Model (CAPM)
Statement of the CAPM
Rb = risk free return
B =. exposure (how much a stock co-move with the market
• The expected return-beta relationship:

E(Ri ) = Rb + β i (E(Rm ) − Rb )

where Rm is the return of the market portfolio and β i is the “beta” of stock i.
• Another way of writing this would be:

Ri − Rb = αi + β i (Rm − Rb ) + ε i

◦ αi is some non-market premium.

◦ ε i is firm-specific surprise with mean zero.
• The prediction of the CAPM is that for any stock i, the equilibrium value of αi is 0.

FINA 3103 Lecture 5. CAPM Spring 2024 3 / 45

Graphical Understanding of the CAPM

Source: Bodie, Kane, and Marcus (2018, Figure 8.1)

FINA 3103 Lecture 5. CAPM Spring 2024 4 / 45

Intuitive Understanding of the CAPM

• Risk premium (higher return than riskless return) on an asset will be determined by its
contribution to the risk/reward of the overall portfolio.
• Suppose we start with the market portfolio of N stocks.
• Variance of the portfolio can be written as:

∑ ∑ wi wj Cov(Ri , Rj ) = ∑ wi ∑ wj Cov(Ri , Rj ) = ∑ wi ∑ Cov(Ri , wj Rj )

i j i j i j
| {z }
Cov(Ri ,Rm )

• A slight increase in wi increases portfolio variance by (roughly): Cov(Ri , Rm ).

• A slight increase in wi increases portfolio return by: E(Ri ) − Rb .
• Notice that all contribution comes from covariance, not from idiosyncratic risk!

FINA 3103 Lecture 5. CAPM Spring 2024 5 / 45

GE Stock’s Contribution to Portfolio’s Risk

Source: Bodie, Kane, and Marcus (2018, Chapter 9)

FINA 3103 Lecture 5. CAPM Spring 2024 6 / 45

Intuitive Understanding of the CAPM

• A basic principle of equilibrium is that all investments should offer the

same reward-to-risk ratio (at the optimum):

E(Ri ) − Rb E(Rj ) − Rb
= .
Cov(Ri , Rm ) Cov(Rj , Rm )

• If not, investors buy/sell to adjust prices (expected returns).

• This should apply to the market portfolio as well, so that

E(Ri ) − Rb E(Rm ) − Rb Cov(Ri , Rm )

= =⇒ E(Ri ) − Rb = (E(Rm ) − Rb )
Cov(Ri , Rm ) Cov(Rm , Rm ) Var(Rm )
| {z }
βi

FINA 3103 Lecture 5. CAPM Spring 2024 7 / 45

Intuitive Understanding of the CAPM

• Risk premium should only be provided to a stock that imposes

systematic (undiversifiable) risk for which the investor must be compensated.
• Idiosyncratic risk can be diversified away and investors need not be afraid of it.
• A positive alpha implies reward without risk.
• Investors will buy positive alpha stocks (prices ↑ and expected returns ↓).
• Investors will sell (or short) negative alpha stocks (prices ↓ and expected returns ↑).
• This activity will continue until all alpha values are driven to zero.

FINA 3103 Lecture 5. CAPM Spring 2024 8 / 45

Assumptions of the CAPM

1. Individual behavior:
(a) Investors are rational, mean-variance optimizers.
(b) Their common planning horizon is a single period.
(c) Investors all use identical input lists, an assumption often termed homogeneous
expectations. Homogeneous expectations are consistent with the assumption that all
relevant information is publicly available.
2. Market structure:
(a) All assets are publicly held and trade on public exchanges.
(b) Investors can borrow or lend at a common risk-free rate, and they can take short
positions on traded securities.
(c) No taxes.
(d) No transaction costs.

FINA 3103 Lecture 5. CAPM Spring 2024 9 / 45

Key Points

• Assets with higher market beta have higher expected returns.

• Only systematic risk that is correlated with the market is rewarded.
• Idiosyncratic risk is not rewarded and earn expected return equal to the riskless rate.
• Review of intuition:
◦ Portfolio separation theorem implies that all investors hold a combination of the riskless
asset and a diversified portfolio (i.e., the tangency portfolio).
◦ No investor bears idiosyncratic risk, so it cannot be rewarded.
◦ Investors that bear more market risk are rewarded with higher expected returns.

FINA 3103 Lecture 5. CAPM Spring 2024 10 / 45

Derivation
Separation Property and Market Clearing

• We start by noticing that each stock i has some market capitalization in equilibrium.
• Let’s say the market cap weight of stock i is wi .
• Recall the separation property: all investors should hold the same mix of risky assets.
• This means any investor’s portfolio of (only) risky assets must have wi weight in i.
• This is due to market clearing.
• In other words:

Market portfolio = Tangency portfolio = Everyone’s risky portfolio.

FINA 3103 Lecture 5. CAPM Spring 2024 11 / 45

Separation Property and Market Clearing

• Let’s be more concrete to see this.

• Suppose there are J investors and investor j has Aj dollars of wealth.
• Each of them hold the same weight vi (not necessarily equal to wi , yet).
• Then, each investor j allocates vi Aj dollars in stock i.
• This means total holdings of all investors for stock i is

J J
∑ vi Aj = vi ∑ Aj = vi A
j=1 j=1

where A is everyone’s wealth summed up – all market cap summed up is also A!

• Total holdings vi A has to equal the market capitalization of stock i by market clearing.
• Market cap weight means (by definition): wi = vi A
A = vi .

FINA 3103 Lecture 5. CAPM Spring 2024 12 / 45

Relating the Market Return to Individual Stock Returns

• We learned that the tangency portfolio is now the market portfolio, M, too.
• Consider a portfolio of (1 − α) fraction in M and α fraction in any other asset i.
• The expected return and standard deviation of this new portfolio (which depend on α) are:

µ(α) = αµi + (1 − α)µm = α(µi − µm ) − µm

q
σ (α) = α2 σi2 + (1 − α)2 σm
2 + 2α (1 − α )Cov(R , R )
m i

• First, we know that when α = 0, the slope would be the market portfolio’s Sharpe ratio:

µm − Rb
[Slope at α = 0] = .
σm

FINA 3103 Lecture 5. CAPM Spring 2024 13 / 45

Relating the Market Return to Individual Stock Returns

• Another way of expressing the slope at α = 0 is:

dµ(α)
(recall that dy/dx is the slope on a plane).
dσ (α) α =0

• After a bit of tedious algebra, we get

dµ(α) dµ(α)/dα µi − µm
= = .
dσ(α) α =0 dσ (α)/dα α =0 (Cov(Rm , Ri ) − σm2 )/σm
• Equating with the Sharpe ratio of the market portfolio, we get

µi − µm µm − Rb Cov(Rm , Ri )
= =⇒ µi − Rb = (µm − Rb ).
(Cov(Rm , Ri ) − σm2 )/σm σm 2
σm

FINA 3103 Lecture 5. CAPM Spring 2024 14 / 45

Relationship to Discounting
Using the CAPM for Discounting

• An asset costs Pi today and pays a random cash flow Xi at a future date (e.g., in 10 years).
• Its rate of return is
X
1 + Ri = i
Pi
• Applying the CAPM to this asset, we get: E(Ri) = Rb + βi(E(Rm) − Rb)

E(Ri − Rb ) = β i E(Rm − Rb )
 
Xi
E = 1 + Rb + β i E(Rm − Rb )
Pi
• Rearrange for a present-value formula:

E ( Xi )
Pi =
1 + Rb + β i E(Rm − Rb )

FINA 3103 Lecture 5. CAPM Spring 2024 15 / 45

Present-Value Formula

• Rb + β i E(Rm − Rb ) is a risk-adjusted interest rate, which is increasing in β i .

• For a riskless cash flow with β i = 0, the formula simplifies to

E ( Xi ) Assume it is a single period

Pi =
1 + Rb

where Rb is the riskless rate.

• Be careful to match the horizon of returns with the maturity of cash flow
• For a 10-year asset:
◦ 1 + Rb = (1 + Yb (10))10 where Yb (10) is the 10-year zero-coupon yield.
◦ E(Rm − Rb ) is the expected excess market return over 10 years.

FINA 3103 Lecture 5. CAPM Spring 2024 16 / 45

History of Economic Thought

William Sharpe (1934–): Nobel Prize in 1990 for capital asset

pricing model

“I submitted the article to The Journal of Finance in 1962. It was rejected. Then I
asked for another referee, and the journal changed editors. It was published in 1964.
It came out and I figured OK, this is it. I’m waiting. I sat by the phone. The phone
didn’t ring. Weeks passed and months passed, and I thought, rats, this is almost
certainly the best paper I’m ever going to write, and nobody cares. It was kind of
disappointing. I just didn’t realize how long it took people to read journals, so it was
a while before reaction started coming in.”

FINA 3103 Lecture 5. CAPM Spring 2024 17 / 45

Evidence
Testing the CAPM

• Using time-series data on returns, estimate a linear regression for each asset i

Ri,t − Rb,t = αi + β i (Rm,t − Rb,t ) + ε i,t

• αi is asset i’s alpha.

◦ CAPM predicts that αi = 0 for all assets.
◦ If αi > 0, average return is too high relative to risk.
• The residual ε i,t represents idiosyncratic risk that is uncorrelated with the market.
• Typically tested on portfolios rather than individual stocks.

FINA 3103 Lecture 5. CAPM Spring 2024 18 / 45

A Statistics Refresher

• Linear regression equation:

yt = α + βxt + ε t
where you observe time-series data for t = 1, . . . , T
• Define the sample means by

T T
1 1
ȳ =
T ∑ yt and x̄ =
T ∑ xt
t=1 t=1

• Ordinary least squares (OLS) estimator is

Cov(xt , yt ) ∑T (xt − x̄)(yt − ȳ)

β̂ = = t=1 T
Var(xt ) ∑t=1 (xt − x̄)2
α̂ = ȳ − β̂x̄.

FINA 3103 Lecture 5. CAPM Spring 2024 19 / 45

A Statistics Refresher

• Standard errors σ( β̂) and σ(α̂) measure uncertainty around the point estimates.
• The t-statistics are the ratios of point estimates to standard errors: β̂/σ( β̂) and α̂/σ(α̂).
• In large samples, t-statistics are normally distributed.
• To reject the null hypothesis α = 0 at the 95% significance level, the t-statistic must be
greater than 1.96 in absolute value.

FINA 3103 Lecture 5. CAPM Spring 2024 20 / 45

Linear Regression in Excel

• Use the regression function in the Data Analysis ToolPak

◦ First install the Data Analysis ToolPak add-in.
◦ See this example

FINA 3103 Lecture 5. CAPM Spring 2024 21 / 45

CAPM for Individual Stocks

Source: Bodie, Kane, and Marcus (2018, Table 8.2)

FINA 3103 Lecture 5. CAPM Spring 2024 22 / 45

Market Betas for Industry Portfolios

Source: Bodie, Kane, and Marcus (2018, Table 8.3)

FINA 3103 Lecture 5. CAPM Spring 2024 23 / 45

Low vs High Market Beta
• CAPM predicts that stocks with higher market beta have higher expected returns.
• Which of the two stocks have a higher beta? How about alpha?

Source: Bodie, Kane, and Marcus (2018, Figure 8.1)

FINA 3103 Lecture 5. CAPM Spring 2024 24 / 45

Constructing Portfolios Sorted by Market Beta

1. Using monthly returns for the previous 60 months, estimate market beta for each stock
2. Sort stocks into 5 portfolios (i.e., cutoff at each 20th percentile)
3. Compute returns on these portfolios over the subsequent year.
4. Rebalance annually

FINA 3103 Lecture 5. CAPM Spring 2024 25 / 45

Cumulative Returns on Low vs High Market Beta Portfolios

FINA 3103 Lecture 5. CAPM Spring 2024 26 / 45

Example 1: Portfolios Sorted by Market Beta

Portfolio
Statistics Market 1 2 3 4 5
Mean (%) 0.54 0.52 0.61 0.61 0.61 0.64
SD (%) 4.37 3.47 4.14 4.75 5.53 7.07
Sharpe ratio 0.12 0.15 0.15 0.13 0.11 0.09
Beta 0.68 0.90 1.05 1.21 1.49
Alpha (%) 0.16 0.13 0.04 −0.05 −0.16
Alpha (t-stat) 2.25 2.53 0.90 −0.76 −1.53

FINA 3103 Lecture 5. CAPM Spring 2024 27 / 45

Alphas on Portfolios Sorted by Market Beta

FINA 3103 Lecture 5. CAPM Spring 2024 28 / 45

Short Riskless Asset to Increase Beta

• Portfolio 1:
E(R1 − Rb ) = 0.16% + 0.68 × E(Rm − Rb )
• Invest 1/0.68 = 1.47 in portfolio 1 and borrow −0.47 at the riskless rate
◦ Beta is
1.47 × 0.68 − 0.47 × 0 = 1.
◦ Expected return is Rp = Rb + w x E(R1-Rb)
Rb + 1.47 × E(R1 − Rb ) = 0.24% + E(Rm )

• Same risk as the market but higher expected return

◦ Arbitrage if the CAPM is true

same risk with the overall market

FINA 3103 Lecture 5. CAPM Spring 2024 29 / 45

Long Riskless Asset to Decrease Beta

• Portfolio 5:
E(R5 − Rb ) = −0.16% + 1.49 × E(Rm − Rb )
• Invest 1/1.49 = 0.67 in portfolio 5 and 0.33 at the riskless asset
◦ Beta is
0.67 × 1.49 + 0.33 × 0 = 1.
◦ Expected return is

Rb + 0.67 × E(R5 − Rb ) = −0.11% + E(Rm ).

• Same risk as the market but lower expected return

FINA 3103 Lecture 5. CAPM Spring 2024 30 / 45

Low vs High Residual Variance

• CAPM predicts that residual variance is unrelated to expected returns.

• Consider a well-diversified vs a single-stock portfolio.

Source: Bodie, Kane, and Marcus (2018, Figure 10.1)

left: low right:high

FINA 3103 Lecture 5. CAPM Spring 2024 31 / 45
Constructing Portfolios Sorted by Residual Variance

1. Using daily returns for the previous 60 days, estimate residual variance for each stock
2. Sort stocks into 5 portfolios (i.e., cutoff at each 20th percentile)
3. Compute returns on these portfolios over the subsequent month
4. Rebalance monthly

FINA 3103 Lecture 5. CAPM Spring 2024 32 / 45

Cumulative Returns on Low vs High Residual Variance

FINA 3103 Lecture 5. CAPM Spring 2024 33 / 45

Example 2: Portfolios Sorted by Residual Variance

Portfolio
Statistics Market 1 2 3 4 5
Mean (%) 0.54 0.57 0.59 0.71 0.73 0.30
SD (%) 4.37 3.65 4.45 5.10 6.05 7.72
Sharpe ratio 0.12 0.16 0.13 0.14 0.12 0.04
Beta 0.78 0.97 1.13 1.31 1.55
Alpha (%) 0.15 0.06 0.11 0.02 −0.54
Alpha (t-stat) 2.88 1.26 2.00 0.25 −3.70

FINA 3103 Lecture 5. CAPM Spring 2024 34 / 45

Alphas on Portfolios Sorted by Residual Variance

FINA 3103 Lecture 5. CAPM Spring 2024 35 / 45

Summary of Empirical Evidence

• Low beta stocks have abnormally high returns relative to risk.

• Stock’s residual variance predicts returns.
◦ Low ⇒ abnormally high returns
◦ High ⇒ abnormally low returns
• A strategy that sorts stocks on variance (instead of residual variance) works similarly.
• Low volatility strategies are widely available in the mutual fund industry.
• Instead of investing in the market, invest in low volatility stocks for higher returns and
lower risk!

FINA 3103 Lecture 5. CAPM Spring 2024 36 / 45

Example of a Low Volatility ETF

Source: Morningstar

FINA 3103 Lecture 5. CAPM Spring 2024 37 / 45

Explanations
What Explains the Failure of CAPM?

• Common SEC disclaimer: “Past performance does not necessarily predict future results.”
• Maybe the CAPM will work better in the future.
• Other reasons for the failure of CAPM:
◦ Investors with low risk aversion, if they cannot borrow, will bid up the price of high
beta stocks.
◦ Fund managers try to beat their benchmarks by tilting their portfolios toward high beta
stocks.
• CAPM builds on the portfolio separation theorem.
◦ All investors should have portfolios on the capital allocation line.
◦ But recall evidence from Sweden

FINA 3103 Lecture 5. CAPM Spring 2024 38 / 45

Roll’s Critique

• The market portfolio in the CAPM is not just the US stock market but includes bonds,
international stocks, real estate, private equity, and so on.
• Impossible to test the CAPM because the market return is hard to measure
• Maybe the CAPM does work, but we do not have a good empirical proxy for the market
portfolio.

FINA 3103 Lecture 5. CAPM Spring 2024 39 / 45

Summary
Next Steps

• There are other types of anomalies that the CAPM does not explain.
• The most famous of which are size and value (later lectures).
• Multi-factor models that generalize the CAPM and allow more sources of systematic risk to
be priced.

FINA 3103 Lecture 5. CAPM Spring 2024 40 / 45

Derivation (Optional)
Supply

• I assets indexed by i = 1, . . . , I
• R is a vector of returns, whose ith element is Ri ; expected return vector is E(R).
• The covariance matrix is
Σ = E((R − E(R))(R − E(R))′ )
• Riskless interest rate Rb
• Let wm be the vector of market weights: The ith element is asset i’s market capitalization as
a share of total market capitalization.
• The return on the market portfolio is
′
Rm = Rb + wm (R − Rb 1).
′ (R − E(R))
• Note that Rm − E(Rm ) = wm

FINA 3103 Lecture 5. CAPM Spring 2024 41 / 45

Demand

• N mean-variance investors indexed by n = 1, . . . , N

• Each investor has wealth An and risk aversion γn .
• Total wealth in the economy is A = ∑N
n = 1 An .
◦ Also equal to total market equity by market clearing (see next slide)
• From the lecture on “Diversification,” we know that
1 −1
wn = Σ E(R − Rb 1).
γn

FINA 3103 Lecture 5. CAPM Spring 2024 42 / 45

Market Clearing

• Market clearing (supply = demand)

N
Awm = ∑ An wn
n=1
N
An − 1
= ∑ γ
Σ E(R − Rb 1)
n=1 n

• Dividing by total wealth, we get:

1 −1
wm = Σ E(R − Rb 1)
γ̄
An
where 1
γ̄ = ∑N
n=1 Aγn .

FINA 3103 Lecture 5. CAPM Spring 2024 43 / 45

Market Clearing
• Rearrange as:

E(R − Rb 1) = γ̄Σwm
= γ̄E((R − E(R))(R − E(R))′ )wm
= γ̄E((R − E(R))(Rm − E(Rm ))) (1)

• The i’s row of this equation is

E(Ri − Rb ) = γ̄Cov(Ri , Rm ). (2)

′ :
• Multiply both sides of (1) by wm
′ ′
wm E(R − Rb 1) = γ̄wm E((R − E(R))(Rm − E(Rm )))
E(Rm − Rb ) = γ̄ E((Rm − E(Rm ))2 ) (3)
| {z }
Var(Rm )

FINA 3103 Lecture 5. CAPM Spring 2024 44 / 45

Capital Asset Pricing Model (CAPM)

• Substitute (3) into (2) to eliminate γ̄:

E(Ri − Rb ) = β i E(Rm − Rb )

where
Cov(Ri , Rm )
βi =
Var(Rm )
is asset i’s market beta
• Special cases:
◦ For the market portfolio (Ri = Rm ), β i = 1.
◦ For idiosyncratic risk (β i = 0), E(Ri ) = Rb .

FINA 3103 Lecture 5. CAPM Spring 2024 45 / 45