FINN2071 Intermediate Financial Economics

Week 4 and 5

Dr Haifeng Guo
 Multiple factor models
CAPM is a single factor model

𝑓 𝑓
𝑟𝑖,𝑡 − 𝑟𝑡 = 𝛼𝑖 + 𝛽𝑖 𝑟𝑀𝐾𝑇,𝑡 − 𝑟𝑡 + 𝜀𝑖,𝑡
Excess returns (above risk-free rate)

Examples of multiple factor models:

1. Fama-French 3 factor model
2. Carhart 4 factor model
3. Fama-French 5 factor model
Multiple factor models increases the 𝑅2
 Fama-French 3 factor model
𝑓 𝑓
𝑟𝑖,𝑡 − 𝑟𝑡 = 𝛼𝑖 + 𝛽𝑖,𝑀𝐾𝑇 𝑟𝑀𝐾𝑇,𝑡 − 𝑟𝑡 + 𝛽𝑖,𝑆𝑀𝐵 𝑟𝑆𝑀𝐵,𝑡 + 𝛽𝑖,𝐻𝑀𝐿 𝑟𝐻𝑀𝐿,𝑡 +𝜀𝑖,𝑡

𝑟𝑖,𝑡 = 𝑟𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝑠𝑡𝑜𝑐𝑘 𝑖 𝑎𝑡 𝑡𝑖𝑚𝑒 𝑡

𝑓
𝑟𝑡 = 𝑟𝑒𝑡𝑢𝑟𝑛 𝑜𝑓 𝑟𝑖𝑠𝑘𝑙𝑒𝑠𝑠 𝑎𝑠𝑠𝑒𝑡 𝑎𝑡 𝑡𝑖𝑚𝑒 𝑡

𝑟𝑆𝑀𝐵,𝑡 = Historic excess returns of small−cap companies over large−cap companies

𝑟𝐻𝑀𝐿,𝑡
= Historic excess returns of value stocks (high book−to−market ratio) over growth stocks (low book−to−market ratio)
 SMB and HML

SMB: Small (size) minus Big (size)

𝑆𝑖𝑧𝑒 𝑀𝑎𝑟𝑘𝑒𝑡 𝐸𝑞𝑢𝑖𝑡𝑦 = 𝑆𝑡𝑜𝑐𝑘 𝑝𝑟𝑖𝑐𝑒 ∗ 𝑠ℎ𝑎𝑟𝑒𝑠 𝑜𝑢𝑡𝑠𝑡𝑎𝑛𝑑𝑖𝑛𝑔

HML: High BM (Value) minus Low BM (Growth)

𝐵𝑜𝑜𝑘 𝑒𝑞𝑢𝑖𝑡𝑦
𝐵𝑀 =
𝑀𝑎𝑘𝑒𝑡 𝑒𝑞𝑢𝑖𝑡𝑦
 SMB

At the end of each June, stocks are divided into two groups by
Size. Stocks are also dividend into three groups by BM. There are
6 portfolios in total.

*The break points are NYSE percentiles.

 NYSE Break points

Suppose that there are only 9 stocks in the U.S market and their
size are:
2,3,3,4,6,7,8,9,15
Full sample break point is 6 (median)
There 4 small stocks and 4 big stocks
If only 5 stocks are NYSE listed, and the size of the stocks are:
3,6,8,9,15
NYSE break point is 8
There are 6 small stocks and 2 big stocks!
 SMB
𝑟𝑆𝑀𝐵 is the average return on the three small portfolios minus the average
return on the three big portfolios

1
𝑟𝑆𝑀𝐵 = 𝑆𝑚𝑎𝑙𝑙 𝑉𝑎𝑙𝑢𝑒 + 𝑆𝑚𝑎𝑙𝑙 𝑁𝑒𝑢𝑡𝑟𝑎𝑙 + 𝑆𝑚𝑎𝑙𝑙 𝐺𝑟𝑜𝑤𝑡ℎ
3
1
− 𝐵𝑖𝑔 𝑉𝑎𝑙𝑢𝑒 + 𝐵𝑖𝑔 𝑁𝑒𝑢𝑡𝑟𝑎𝑙 + 𝐵𝑖𝑔 𝐺𝑟𝑜𝑤𝑡ℎ
3

Small Value…Big Growth are value-weighted portfolio returns.

 HML
𝑟𝐻𝑀𝐿 is the average return on the two value portfolios minus the
average return on the two growth portfolios
1
𝑟𝐻𝑀𝐿 = 𝑆𝑚𝑎𝑙𝑙 𝑉𝑎𝑙𝑢𝑒 + 𝐵𝑖𝑔 𝑉𝑎𝑙𝑢𝑒
2
1
− 𝑆𝑚𝑎𝑙𝑙 𝐺𝑟𝑜𝑤𝑡ℎ + 𝐵𝑖𝑔 𝐺𝑟𝑜𝑤𝑡ℎ
2

Small Value…Big Growth are value-weighted portfolio returns.

 Example
Fama-French 3 factors are available at Kenneth R. French’s data
library. Employ the 3-factor model for IBM and compare the
results to the CAPM results.
 Carhart 4 factor model
𝑓 𝑓
𝑟𝑖,𝑡 − 𝑟𝑡 = 𝛼𝑖 + 𝛽𝑖,𝑀𝐾𝑇 𝑟𝑀𝐾𝑇,𝑡 − 𝑟𝑡 + 𝛽𝑖,𝑆𝑀𝐵 𝑟𝑆𝑀𝐵,𝑡 + 𝛽𝑖,𝐻𝑀𝐿 𝑟𝐻𝑀𝐿,𝑡 +𝛽𝑖,𝑀𝑂𝑀 𝑟𝑀𝑂𝑀,𝑡 + 𝜀𝑖,𝑡

𝑟𝑀𝑂𝑀,𝑡
= Historic excess returns of past (month t−12 to month t−2) winner over past loser

1
𝑟𝑀𝑂𝑀 = 𝑆𝑚𝑎𝑙𝑙 𝑊𝑖𝑛𝑛𝑒𝑟 + 𝐵𝑖𝑔 𝑊𝑖𝑛𝑛𝑒𝑟
2
1
− 𝑆𝑚𝑎𝑙𝑙 𝑙𝑜𝑠𝑒𝑟 + 𝐵𝑖𝑔𝑙𝑜𝑠𝑒𝑟
2
 Introduction
With the CAPM, we can answer the question of when to get
in/out of the market

A key question is: which stocks will give you higher expected
returns than others
 Portfolio analysis
How to select stocks from the universe?
Group stocks by characteristics at the end of each month
𝑅𝑡+1
(quarter; half year; year)

𝑃𝑡

t-1 t t+1

Estimation period Formation period

Never use the information after time t to form a portfolio at time t

 Example of characteristics
𝑓 𝑓
𝑟𝑖,𝑡 − 𝑟𝑡 = 𝛼𝑖 + 𝛽𝑖,𝑀𝐾𝑇 𝑟𝑀𝐾𝑇,𝑡 − 𝑟𝑡 + 𝛽𝑖,𝑆𝑀𝐵 𝑟𝑆𝑀𝐵,𝑡 + 𝛽𝑖,𝐻𝑀𝐿 𝑟𝐻𝑀𝐿,𝑡 +𝛽𝑖,𝑀𝑂𝑀 𝑟𝑀𝑂𝑀,𝑡 + 𝜀𝑖,𝑡

Size: Stock price*Shares outstanding

Value (B/M): Book-to-market ratio

Momentum: Excess returns from month t−12 to month t−2

 Equally-weighted vs Value-weighted

Date Pa Pb reta retb Vwret Ewret

Jan-23 100 100

Log return Feb-23 110 90 0.09531 -0.10536 -0.005025 -0.00503

Mar-23 100 100 -0.09531 0.10536 -0.005008 0.005025

Jan-23 100 100

Simple return Feb-23 110 90 0.1 -0.1 0 0

Mar-23 100 100 -0.09091 0.11111 0 0.010101

 Example
Download the monthly B/M sorted portfolio returns from
French’s data library and get the summary statistics of each
decile
Example
 Example
Value-weighted portfolios: Stocks with highest B/M (Decile 10) on average earn 1.3%
per month while those in the lowest decile earn 0.9% per month.

Equally-weighted portfolios: Stocks with highest B/M (Decile 10) on average earn 1.91%
per month while those in the lowest decile earn 0.76% per month.

Value stocks outperform growth stocks!

Fama and French view value stocks as those of firms in 'distress'. Since these firms are riskier, they earn
higher returns on average. To support their view, they find that value stocks have persistently low
earnings. Others, such as Josef Lakonishok, Andrei Shleifer, and Robert Vishny, view that irrational pricing
causes the value effect. In their view, investors hate value stocks and undervalue them.
Example 2
 Example 2
Value-weighted portfolios: Stocks with highest past returns (Decile 10) on
average earn 1.48% per month while those in the lowest decile earn 0.34%
per month.

Equally-weighted portfolios: Stocks with highest past returns (Decile 10)

on average earn 1.74% per month while those in the lowest decile earn
0.92% per month.

Past winners outperform past losers!

 Example 2
If we keep a long position in D10 and a short position in D1 during
the sample period, we earn on average 1.48%-0.34%=1.14% per
month without market exposure

Question: is 1.14% statistically significant?

𝑥ҧ − 𝜇
𝑇 − 𝑠𝑡𝑎𝑡 =
𝑠/ 𝑛
 Next step
What if both characteristic x and y works, but we want to know if each signal predicts
returns beyond what is captured by the other? What if there are multiple
characteristics?

Double sort may help. But triple- or quadruple-sort would be difficult due to a lack of
sufficient observations.

We need a multivariate regression of stock returns on many characteristics at once.

Fama-MacBeth Regressions