Arbitrage Pricing Theory (APT)

ACTSC 372: Corporate Finance

Winter 2021

Pengyu Wei
Recall CAPM Equation

𝐶𝑜𝑣 𝑅! , 𝑅# 𝜎!#
𝜇! = 𝑟" + 𝛽! 𝜇# − 𝑟" , 𝛽! = = $
𝑉𝑎𝑟 𝑅# 𝜎#

§ 𝛽! measures the “sensitivity” of asset 𝑖 to the “systematic factor”

§ The “market portfolio”, is the only systematic factor in CAPM

§ APT: A competing theory to CAPM

§ Why only one systematic factor? Why the market portfolio?

§ Arbitrage opportunities: zero initial cost, guaranteed future return

§ APT: Such opportunities may exist but disappears quickly
CAPM and Risk Decomposition

§ CAPM formula also implies a special structure for the random return
𝑅! = 𝑟" + 𝛽! 𝑅# − 𝑟" + 𝜖!
§ Assume 𝐸 𝜖! = 0, 𝐶𝑜𝑣 𝑅# , 𝜖! = 0

§ The mean return is the CAPM formula again

𝜇! = 𝑟" + 𝛽! 𝜇# − 𝑟"

§ The variance of return is decomposed into two parts

𝜎!$ = 𝛽!$ 𝜎#
$ $
+ 𝜎%!

§ Similar idea for APT, but potentially (many) more systematic risk
§ One 𝛽 for each systematic risk
Decomposition of Asset Return/Risk

§ Actual Return = Expected Return + Unexpected Return (due to risk)

§ Risk = Systematic risk + Idiosyncratic risk

§ APT model of asset returns:

𝑅! = 𝜇! + 𝑚! + 𝜖!
§ 𝑚! = 𝛽!& 𝐹& + ⋯ 𝛽!' 𝐹'

§ Example 1: Crop Farm

§ 𝜇: expected revenue given expected weather condition
§ 𝑚: unexpected drought due to global warming
§ ϵ: accidental death of farm owner

§ Example 2: Oil Company

§ 𝜇: expected revenue given expected production level and oil price
§ 𝑚: unexpected coronavirus outbreak
§ ϵ: fire at production site
APT uses a Factor Model for Systematic Risk

§ Idea: 𝑚! = 𝛽!& 𝐹& + ⋯ 𝛽!' 𝐹'

§ 𝑅! can be affected by 𝑘 different factors, 𝐹& , … , 𝐹'
§ 𝐹' can affect different assets to different extent (𝛽!' ≠ 𝛽(' )

§ Examples:
§ Car manufacturer’s revenue can be affected by oil price & steel price
§ Oil price affects Ford & Tesla in opposite directions

§ Common factors:
§ Macro economic factors: GDP, interest rate, inflation, etc.
§ Natural disasters: earthquakes, hurricanes, global warming, etc.
§ Social-political factors: trade-war, etc.
§ Pandemic: coronavirus
Factor Model: Definition & Special Cases

§ Factor Model: assume linear sensitivity to risk factors

𝑅! = 𝜇! + 𝛽!& 𝐹& + ⋯ 𝛽!' 𝐹' + 𝜖!
§ 𝛽!( measures sensitivity to risk factors
§ 𝐹( = 𝑗th systematic risk factor = actual value – expected value
§ E 𝐹( = 0 & Cov 𝐹! , 𝐹( = 0 for 𝑖 ≠ 𝑗
§ 𝜖! = idiosyncratic risk of asset 𝑖
§ 𝐸 𝜖! = 0 & Cov 𝜖! , 𝐹( = 0 & Cov 𝜖! , 𝜖( = 0 for 𝑖 ≠ 𝑗

§ Single-Factor Model: simplest factor model

𝑅! = 𝜇! + 𝛽! 𝐹 + 𝜖!
§ Usually used for proof of concept

§ Market Model: market as the only risk factor

𝑅! = 𝜇! + 𝛽!# 𝑅# − 𝜇# + 𝜖!
§ Similar to CAPM formula
Factor Model: 𝑅! = 𝜇! + 𝛽!" 𝐹" + ⋯ 𝛽!# 𝐹# + 𝜖!

§ Goals:
§ A simple but insightful model for the systematic return
§ Make some assumptions about the idiosyncratic return
§ Derive an expression for the expected return

§ The conclusion is
𝜇! = 𝑟" + 𝛽!& 𝛾& + ⋯ + 𝛽!' 𝛾'
§ Linear structure
§ Same 𝛾& , … , 𝛾' for all assets
§ Want to show 𝛾& , … , 𝛾' exist and identify what they are

§ No arbitrage: same systematic risk è same expected return

Factor Model: Expected Return

§ For individual assets

E 𝑅! = E 𝜇! + 𝛽!& 𝐹& + ⋯ 𝛽!' 𝐹' + 𝜖! = 𝜇!

§ For a portfolio 𝑤, 𝑅) = ∑+
!*& 𝑤! 𝑅! , so

+ + + +
𝑅) = E 𝑤! 𝜇! + E 𝑤! 𝛽!& 𝐹& + ⋯ + E 𝑤! 𝛽!' 𝐹' + E 𝑤! 𝜖!
!*& !*& !*& !*&

§ 𝑅) = 𝜇) + 𝛽)& 𝐹& + ⋯ + 𝛽)' 𝐹' + 𝜖)

§ 𝐸 𝑅) = 𝜇) = ∑+ !*& 𝑤! 𝜇!
§ If the portfolio is well-diversified, idiosyncratic risk 𝜖) = 0
Proof for Single-Factor Model

§ Consider 𝑅! = 𝜇! + 𝛽! 𝐹 + 𝜖! for any asset 𝑖, show that

𝜇! = 𝑟" + 𝛽! 𝛾, for some 𝛾
and the same 𝛾 for all assets.
§ Need to identify this value
§ Again, has to be the same for all assets

§ Proof ideas:
1. Use any two assets to construct a “risk-free” portfolio
2. This resulting portfolio must have the risk-free rate
3. Identify a common “price of systematic risk”
4. Identify the desired parameter 𝛾
5. Verify the APT formula
Proof for Single-Factor Model

§ Consider the return of a well-diversified portfolio of any two assets, 𝛽! ≠ 𝛽"

𝑅# = 𝑤! 𝑅! + 𝑤" 𝑅" = 𝑤! 𝜇! + 𝑤" 𝜇" + 𝑤! 𝛽! + 𝑤" 𝛽" 𝐹 = 𝜇$ + 𝛽$ 𝐹

§ Set 𝛽$ = 0 then solve for 𝑤! and 𝑤" = 1 − 𝑤!

𝛽" −𝛽!
𝑤! = , 𝑤" =
𝛽" − 𝛽! 𝛽" − 𝛽!
§ By no-arbitrage, since 𝛽$ = 0 & portfolio is well-diversified, 𝜇$ = 𝑟% . So
𝛽" 𝜇! − 𝛽! 𝜇"
𝑤! 𝜇! + 𝑤" 𝜇" = = 𝑟%
𝛽" − 𝛽!
§ Rearrange
𝛽" 𝜇! − 𝛽! 𝜇" = 𝑟% 𝛽" − 𝛽!
𝜇! − 𝑟% 𝛽" = 𝜇" − 𝑟% 𝛽!
𝜇! − 𝑟% 𝜇" − 𝑟%
=
𝛽! 𝛽"
Proof for Single-Factor Model

§ The equality holds for any two assets, so it holds for all assets
𝜇& − 𝑟" 𝜇$ − 𝑟"
=
𝛽& 𝛽$
,! -."
§ Define 𝛾 = for all assets 𝑖, rearrange we have
/!
𝜇! = 𝑟" + 𝛽! 𝛾

§ The general formula for factor model is

𝜇! = 𝑟" + 𝛽!& 𝛾& + ⋯ + 𝛽!' 𝛾'

§ This is called the APT formula

The APT Formula

§ General factor models: 𝜇& = 𝑟% + 𝛽&! 𝛾! + ⋯ + 𝛽&' 𝛾'

§ Called the APT formula

(! )*"
§ Single-factor model: 𝜇& = 𝑟% + 𝛽& 𝛾, where 𝛾 =
+!

(! )*"
§ Market Model, 𝛾 = holds for any asset, including market portfolio
+!
,!#
§ One can show that 𝛽& = $
,#

(# )*"
§ 𝛾= = 𝜇- − 𝑟% , because 𝛽- = 1
+#

§ Called the excess market return or market premium

§ APT formula for the Market Model coincides with the CAPM formula
𝜇& = 𝑟% + 𝛽& 𝜇- − 𝑟%
CAPM vs. APT

§ Similarities
§ Market Model APT formula coincides with the CAPM formula
§ Some common assumptions (e.g., frictionless & liquid market)

CAPM APT
Systematic • Unique market portfolio • Multiple risk factors
Risk Factor • Derived mathematically • Selected by users
Return • Known 𝜇 & Σ • Model return statistically
Model • Derived equilibrium return • Derived expected return 𝜇
Asset • All fairly priced assets should • What is SML?
prices lie on the SML
Main • Estimation of 𝜇 & Σ • Selection of systematic risk
challenge factors
APT Example & Exercise

§ Suppose a 2-factor model holds

𝑅! = 𝜇! + 𝛽!& 𝐹& + 𝛽!$ 𝐹$ + 𝜖!

The following information is available in the market

𝒊 𝜷𝒊𝟏 𝜷𝒊𝟐 𝝁𝒊 𝝈𝟐𝝐𝒊 𝝐𝒊
Stock A 1.0 1.5 10% 0.01 5%
Stock B 0.5 1.0 7% 0.0144 1%
Stock C 0.75 1.25 9% 0.0225 −2%

Actual Expected 𝑽𝒂𝒓 𝑭

𝑭𝟏 7% 5% 0.02
𝑭𝟐 1% 2% 0.05

𝐹! = Actual - Expected
 𝒊 𝜷𝒊𝟏 𝜷𝒊𝟐 𝝁𝒊 𝝈𝟐𝝐𝒊 𝝐𝒊

APT Example & Exercise Stock A 1.0 1.5 10% 0.01 5%

Stock B 0.5 1.0 7% 0.0144 1%
Stock C 0.75 1.25 9% 0.0225 −2%

Actual Expected 𝑽𝒂𝒓 𝑭

𝑭𝟏 7% 5% 0.02
𝑭𝟐 1% 2% 0.05

§ Calculate the actual return for each asset

𝑅0 = 𝜇0 + 𝛽0& 𝐹& + 𝛽0$ 𝐹$ + 𝜖0

F& = 7% - 5% = 2%
F$ = 1% - 2% = -1%

𝑅0 = 10% + 1 7% − 5% + 1.5 1% − 2% + 5%
𝑅0 = 15.5%
𝑅1 = 8%, 𝑅2 = 7.25%
APT Example & Exercise

§ Calculate the variance of return for each asset

𝑉𝑎𝑟 𝑅0 = 𝑉𝑎𝑟 𝜇0 + 𝛽0& 𝐹& + 𝛽0$ 𝐹$ + 𝜖0

$ $
𝑉𝑎𝑟 𝑅0 = 𝛽0& 𝑉𝑎𝑟 𝐹& + 𝛽0$ 𝑉𝑎𝑟 𝐹$ + 𝑉𝑎𝑟 𝜖0

𝑉𝑎𝑟 𝑅0 = 1 $ 0.02 + 1.5 $ 0.05 + 0.01

𝑉𝑎𝑟 𝑅0 = 0.1425

𝑉𝑎𝑟 𝑅1 = 0.0694, 𝑉𝑎𝑟 𝑅2 = 0.11875

 𝒊 𝜷𝒊𝟏 𝜷𝒊𝟐 𝝁𝒊 𝝈𝟐𝝐𝒊 𝝐𝒊

APT Example & Exercise Stock A 1.0 1.5 10% 0.01 5%

Stock B 0.5 1.0 7% 0.0144 1%
Stock C 0.75 1.25 9% 0.0225 −2%
§ Use only assets 𝐴 & 𝐵, construct portfolio 𝑃 such that 𝛽3& = 0

Solve a system of 2 linear equations with 2 unknowns 𝑤0 & 𝑤1

1. One Budget constraint: 𝑤0 + 𝑤1 = 1
2. One Beta constraint: 𝛽3& = 𝑤0 𝛽0& + 𝑤1 𝛽1& = 0
1 𝑤0 + 0.5 𝑤1 = 0

The solution is
𝑤0 = −1 , 𝑤1 = 2
 𝒊 𝜷𝒊𝟏 𝜷𝒊𝟐 𝝁𝒊 𝝈𝟐𝝐𝒊 𝝐𝒊

APT Example & Exercise Stock A 1.0 1.5 10% 0.01 5%

Stock B 0.5 1.0 7% 0.0144 1%
Stock C 0.75 1.25 9% 0.0225 −2%
§ Suppose the portfolio you constructed is well-diversified
𝑤0 = −1 , 𝑤1 = 2
What is the expected return of the portfolio, 𝜇3 ?

𝜇3 = 𝑤0 𝜇0 + 𝑤1 𝜇1
𝜇3 = −1 10% + 2 7%
𝜇3 = 4%

What is 𝜇3 if the portfolio was not well-diversified?

§ Also 4% because 𝐸 𝜖) = 0
APT Example & Exercise

§ Is it possible to construct a portfolio 𝑄 such that 𝛽4& = 𝛽4$ = 0? If so, set

up a system of linear equation to solve for such portfolio.

Solve a system of 3 linear equations with 3 unknowns 𝑤0 , 𝑤1 , & 𝑤2

One Budget constraint: 𝑤0 + 𝑤1 + 𝑤2 = 1
Two Beta constraints:
𝛽4& = 𝑤0 𝛽0& + 𝑤1 𝛽1& + 𝑤2 𝛽2& = 0
1 𝑤0 + 0.5 𝑤1 + 0.75 𝑤2 = 0

𝛽4$ = 𝑤0 𝛽0$ + 𝑤1 𝛽1$ + 𝑤2 𝛽2$ = 0

1.5 𝑤0 + 1 𝑤1 + 1.25 𝑤2 = 0
APT Example & Exercise

§ Suppose the portfolio 𝑄 satisfies 𝛽4& = 𝛽4$ = 0 and is well-diversified. If

𝜇4 > 𝑟" , describe an arbitrage opportunity in detail.

Since 𝛽4& = 𝛽4$ = 0, Portfolio 𝑄 does not take any systematic risk hence is
“risk-free”. Thus we have two assets, 𝑄 & 𝑟" , that have the same risk but
different expected return. Arbitrage exists
1. @𝑡 = 0, borrow, as much as possible, at interest rate 𝑟"
2. @𝑡 = 0, Use all proceeds to purchase portfolio 𝑄
3. Some time in the future, sell the portfolio 𝑄 with return 𝜇4
4. At the same time, repay interest 𝑟"
Since 𝜇4 > 𝑟" and both assets are “risk-free”, we gain a guaranteed future
return without initial investment.