Professional Documents
Culture Documents
I Risk-Return relationship
I What is a risk?
I What are returns?
I How do we measure risk and returns?
I How should they be related?
I Portfolio Construction
I Conditional risk measure: What if things change in a dynamic way?
I Value-at-Risk (risk management)
I There are two key components to any investment strategies - risk and returns and we
need a way to define those.
I This will allow us to:
- distinguish between di↵erent investment alternatives
- evaluate mutual funds and other money managers
I It is also the first step towards:
1. understanding the benefits of diversification
2. finding the appropriate discount rate to use
3. understanding Modern Portfolio Theory, which dominates today’s investment
philosophy.
I Hang Seng index: a market-capitalization-weighted stock-market index in Hong Kong using the largest 60 firms.
I ABF bond fund: a Hong Kong government bond index fund.
State RA RB Probability
Good 20% 5% 0.20
OK 11% 2% 0.50
Bad -15% 2% 0.30
State RA RB Probability
Good 5% 5% 0.20
OK 11% 2% 0.50
Bad -15% 2% 0.30
I Returns can either be ex post, meaning returns were already realized, or ex ante, which means
they are expected
I Let’s first look at ex post returns
I Holding period return if you bought a stock at time t 1 and sold it at time t is
P t + Dt Pt 1 P t + Dt
Rt = = 1
Pt 1 Pt 1
I Dividend yield = Dt
Pt 1
110 + 5 100
I Return during year 1 is R1 = = 15%
100
112 + 6 110
I Return during year 2 is R2 = = 7.27%
110
I This is also known as the geometric mean return (equivalent to the e↵ective rate)
I By contrast, the arithmetic or simple mean is
R1 + R2 + ... + RT
R̄ =
T
State RA RB Probability
Good 20% 5% 0.20
OK 11% 2% 0.50
Bad -15% 2% 0.30
I You know that in the past 100 years, there were 60 good and 40 bad
years.
PT
I Likewise, using long-term past data, you compute R̄ = t=1
T
Rt
I Assuming that on average the market is efficient (each time, price is equilibrium correct price), you
can use long-term average returns for E (R) and compare it with model-implied expected returns.
I But, teacher I am so confused because “past performance is not indicative of future results”!
PT
t=1 (Rt R̄)2
Ex post VAR(R) =
T 1
Xn
Ex ante VAR(R) = E [(R E (R))2 ] = Ps (Rs E (R))2 , or
s=1
Xn ⇣X n ⌘2
= Ps · Rs2 Ps · R s
s=1 s=1
I Standard deviation leaves something to be desired as a measure of risk though as both unusually
large gains will elevate the measure as much as unusually large losses will
2
Xn 2
Var (RB ) = (RB ) = Ps ⇥ (RB,s E (RB ))
s=1
2 2 2
= 0.2 · (0.05 0.026) + 0.5 · (0.02 0.026) + 0.3 · (0.02 0.026)
= 0.000144
p
(RB ) = 0.000144 = 0.012 = 1.20%
I Only about 60% of U.S. households invests in stock markets (SCF 2019).
I Investors dislike risk: they are risk averse (reluctance to take financial risk)
I Returns on investment vary with risk: you should require more return if you take on more risk
Expected Payo↵
P=
1+r
r > 0 for risk averse agent. It is the discount rate for the risk, (this is in addition to the time
value of money): r = risk-free + risk premium
I So returns are only half the story in finance: we also need to take into account risk.
I Discount rates should therefore reflect not only the time value of money but also the riskiness of
the underlying security (e.g., YTM should be higher for non-investment grade bonds than
investment-grade bonds.
Deciding on an investment philosophy is like picking a spouse. Do you want someone who is volatile
and emotional, or do you want someone who is steady and trustworthy, and down to earth? If you want
a successful investment career, you’d better bind yourself to a style you can live with
–Ralph Wanger, CEO, Acorn mutual funds
40
Percent per Year
20
10
Small 2 3 4 5 6 7 8 9 Big
30
more volatile and give you
higher returns.
I Why? Which firms would have
20
a high book-to-market ratio?
10
Glamour 2 3 4 5 6 7 8 9 Value
21
! Volatility cannot be the
whole story.
18
I In Session 4,
15 P0
= r1 + NPVGO
EPS 1 e EPS 1
! High P/E is associated with
12 low returns.
9
Low PE 2 3 4 5 6 7 8 9 High PE
volatility.
! Dividend-paying stocks are
20 not necessarily safer.
I In Session 4, re = D1
+g
P0
! High D/P is associated with
15
higher returns.
10
Low DP 2 3 4 5 6 7 8 9 High DP
S1B5 S1B5
13 13
S2B5 S2B5
12 S1B4 12 S1B4
Avg. excess returns
0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0.05 0.06 0.07 0.08 0.09 0.1
Standard deviation Standard deviation
I Takeaway 1. There are many characteristics that can explain patterns of average
returns (e.g., size, book-to-market, P/E, D/P, etc). There are more than 200 factors.
I Takeaway 2. Many return patterns are consistent with volatility (e.g., big firms have
low returns and low volatility VS small firms have high returns and high volatility).
I Takeaway 3. Volatility is not the whole story. Portfolios don’t have a ’monotonic’
pattern of volatility (e.g., high PE portfolio).
I Takeaway 4. Many characteristics can be explained by the Gordon model and can be
somehow rationalized.
Cov (X , Y )
⇢(X , Y ) =
(X ) (Y )
State RA RB Probability
Good 20% 5% 0.20
OK 11% 2% 0.50
Bad –15% 2% 0.30
E (a) = a
E (aX + bY ) = aE (X ) + bE (Y )
Var (a) = 0
Var (aX + bY ) = a2 Var (X ) + b 2 Var (Y ) + 2ab · Cov (X , Y )
Cov (a, X ) = 0
Cov (X , X ) = Var (X )
Cov (aX , bY ) = abCov (X , Y )
1 ⇢(X , Y ) 1
⇢(X , Y ) > 0 ! positive correlation: X and Y move in the same direction
⇢(X , Y ) < 0 ! negative correlation: X and Y move in opposite directions
I Suppose you form a portfolio of N securities: you buy ni shares of stock i at price Pi
I Weight of particular security i in your portfolio at the time of portfolio formation is
PN
I Note that weights sum up to 1 (or 100%): k=1 wk = 1
I Ex post (realized) return of portfolio p is
N
X
Rp = wk · R k
k=1
I Suppose that at time t = 0 you buy 45 shares of stock A and sell short 10 shares of stock B, and
that you liquidate the positions at time t = 1
I The portfolio weights of stocks A and B at time t = 0 are
n A · PA 45 · 20
wA = P N = = 1.80, wB = 1 1.80 = 0.80
k=1 n k · P k 45 · 20 + ( 10) · 40
I One-period returns of stocks A and B are
Var (Rp ) = wA2 Var (RA ) + wB2 Var (RB ) + 2wA wB Cov (RA , RB )
= 0.402 · 0.0183 + 0.602 · 0.000144 + 2 · 0.40 · 0.60 · 0.0009
= 0.003412
I Standard deviation of portfolio is
p
(Rp ) = 0.003412 = 0.0584 = 5.84%
I The expected return E (Rp ) we computed is the weighted average of E (RA ) and E (RB )
I However, the standard deviation we computed is less than the weighted average standard
deviation
(Rp ) = 5.84% < 0.40 · (RA ) + 0.60 · (RB ) = 6.13%
I This is due to diversification: holding many securities can reduce risk
I Diversification benefits arise when correlation ⇢(RA , RB ) < 1
I The lower the correlation, the greater the diversification benefits
I In fact, if ⇢(RA , RB ) = 1, we can create a portfolio with (Rp ) = 0 (Correlation e↵ect)
E (Rp ) rf
(Rp )
where E (Rp ) is the expected return on the portfolio and p is the volatility of the portfolio.
I This is to capture how much reward you get for one unit of risk-taking.
I William F. Sharpe:
I Nobel Prize in economics, 1990
I https://web.stanford.edu/⇠wfsharpe/
I Suppose that expected returns on two stocks are E (RA ) = 0.05 and E (RB ) = 0.10
I Standard deviations are (RA ) = 0.40 and (RB ) = 0.80
I Two stocks are perfectly negatively correlated: ⇢(RA , RB ) = 1
I How can we construct a portfolio that has zero ?
I Suppose we invest a fraction wA of our money in stock A and the rest wB = 1 wA in stock B,
and recall that a2 + b 2 2ab = (a b)2
Var (Rp ) = wA2 Var (RA ) + wB2 Var (RB ) + 2wA wB (RA ) (RB )⇢(RA , RB )
= wA2 · 0.402 + (1 wA )2 · 0.802 2wA (1 wA ) · 0.4 · 0.8
2
= (0.40wA 0.80(1 wA )) = (1.20wA 0.80)2 = 0
50
45
40
35
30
25
20
0 10 20 30 40 50 60 70 80 90 100
50
45
40
35
30
25
20
0 10 20 30 40 50 60 70 80 90 100
0 10 20 30 40 50 60 70 80 90 100
0 10 20 30 40 50 60 70 80 90 100
50
0 10 20 30 40 50 60 70 80 90 100
60
Annual Std Dev, (%)
50
40
30
20
0 10 20 30 40 50 60 70 80 90 100
0.65
Sharpe Ratio
0.55
0.45
0.35
0.25
0 10 20 30 40 50 60 70 80 90 100
For Diversification
Why look for the needle in the haystack? Buy the haystack!
–John Bogle, CEO, and founder of Vanguard index funds
Against Diversification
Diversification may preserve wealth, but concentration builds wealth
–Warren Bu↵ett, billionaire investor
Diversification is a hedge for ignorance. I think you are much better o↵ owning a few stocks and
knowing a great deal about them
–William J. O’Neil, investor, and author
Put all your eggs in one basket – and watch that basket!
–Mark Twain
I So if holding many securities can reduce risk through diversification, what happens if you hold all
the stocks in the market? Do you eliminate all risks?
I No, you don’t. You can eliminate what is called Firm-specific/ Non-systematic/ Diversifiable/
Idiosyncratic risk but you cannot rid yourself of Market/ Systematic/ Un-diversifiable risk
I Systematic risk is part of the system itself, there is no asset that is immune to shocks of this kind
Standard deviation (%)
60
50
40 Firm-specific risk
30
20
10
Market risk
10 20 30 40 50 60 70 80 90 100
Number of stocks in a portfolio
State RA RB Probability
Good 16% 1% 1/3
OK 10% 10% 1/3
Bad 4% 19% 1/3
Suppose that you form an equal-weighted portfolio p of stocks A and B (that is, wA = wB = 0.50).
Calculate
I Expected returns of stocks A and B
I Expected return of portfolio p, E (Rp )
I Standard deviation of stock A and B returns, (RA ) and (RB )
I Covariance and correlation of returns of stocks A and B, Cov (RA , RB ) and ⇢(RA , RB )
I Standard deviation of returns of portfolio p, (Rp )
Is it possible to construct a portfolio with (Rp ) = 0 by varying weights wA and wB ?
I You need E [R] (long-term avg. returns) ! Et [R] (next month expected returns) and ! t
I Measuring conditional expected returns Et [R] is very tricky while measuring conditional volatility
is more doable since it is highly persistent ( = not changing dramatically over time).
Q1. Can you see the volatility (standard deviation, variability) of returns change over time?
Q2. Do positive shocks also contribute to a high volatility?
I Before we talk about how to measure the conditional variance, what is the benchmark?
n
X
2
RVt = rj2t
j=1
where rj2t is the return for day j in month t, n is the number of trading days in the month.
I Predicting RVt using the information up to t 1 is the goal. For example, we gotta predict the
variance in Mar. 2023 using the information available up to Feb. 2023.
I Teacher, why do we use the sum of squared daily returns instead of squared monthly returns?
I ARCH (p)
P
I t2 = ! + pi=1 ↵i ✏2t i , where ✏ = Rt R̄ (shock or news)
P
I Recall ex post variance = T t=1 (Rt R̄)2 /(T 1)
I ARCH(1): t2 = ! + ↵✏2t 1
1
Neus flong tam .
hBtoy B
.
I GARCH (1,1) - Generalized ARCH
I t2 = ! + ↵✏2t 1 + t2 1
I We only use the past information (up to t 1) to predict the conditional variance at time t.
Sgm
⼀
rmmu
!
=
1 ↵
2
where is the unconditional long-run variance, ! > 0, and ↵ + <1
I 2
= 2
⇥(1 ↵ )+↵⇥ ✏2t 1 + ⇥ 2
t |{z} |{z}
t 1
| {z }
Long-run variance squared the most recent news about returns Previous conditional variance
I Run a regression:
I RVt = ↵ + t2 + ut
where t2 is the conditional variance estimate using the information up to t 1.
I Unbiased forecasts imply that ↵ˆ = 0 and ˆ = 1
I By the way, why do we care about the time-varying variance of stock returns?
I Value-at-Risk (VaR)
I VaR is a statistic that quantifies the extent of possible financial losses within a firm, portfolio,
or position over a specific time frame.
I Commonly used by investment and commercial banks to determine the extent and probabilities
of potential losses in their institutional portfolios.
I 10% quantile (10% lower tail percentile) of returns is 1.28 ⇥ + µ (= Threshold return)
I With 10% of chance, you would expect your loss is greater than $100, 000 ⇥( 1.28 ⇥ + µ)
I Investors are risk averse and need to be compensated for bearing risk
I Portfolios
I Weights, average returns, standard deviation, covariance, and correlation
I Portfolio risk can be reduced by holding securities with a correlation of less than one
I Firm-specific vs. market risk