Portfolio Management 3677

Chapter 5 - Risk, Return, and the Historical

Record

Alexandre Rubesam, Ph.D.

[Link]@[Link]

Fall 2025

These slides were prepared by Alexandre Rubesam Slide 1

Session Outline

1. Introduction

2. Measuring Returns over Different Holding Periods

3. Interest Rates and Inflation Rates

4. Risk and Risk Premiums

5. The Normal Distribution

6. Deviations from Normality and Tail Risk

7. Learning from Historical Returns

8. Suggested Exercises and Homework

These slides were prepared by Alexandre Rubesam Slide 2

 Introduction

These slides were prepared by Alexandre Rubesam Slide 3

Introduction

The slides from Session 1 cover most of what we need from a

technical perspective.
These slides cover the contents of Chapter 5 of BKM. In introduces
important concepts in investment management, including:

▶ Holding period returns

▶ Expected returns and standard deviations
▶ Excess Returns and Risk Premiums
▶ Time series analysis of past returns
▶ Arithmetic vs Geometric returns
▶ Variance and Std Deviation (volatility)
▶ Sharpe ratio
▶ Normality, non-normality, and risk measures (VaR, ES)

These slides were prepared by Alexandre Rubesam Slide 4

Measuring Returns over Different Holding
Periods

These slides were prepared by Alexandre Rubesam Slide 5

Measuring Returns over Different Holding Periods

▶ We start by discussing how to calculate and compare rates of

returns over different investment periods. For this purpose, we
will consider zero-coupon bonds of different maturities.
▶ A zero-coupon bond is a financial instrument that pays its
owner only one cash flow, for example, $100, on the maturity
date. This cash flow is called the face value of the bond.
▶ The investor buys the bond today for a price P (T ), where T
indicates the maturity of the bond in years.

These slides were prepared by Alexandre Rubesam Slide 6

Measuring Returns over Different Holding Periods

Question: What is the return in this example over the life of the
bond?

Figure 1: Cash flows of a zero-coupon bond

These slides were prepared by Alexandre Rubesam Slide 7

Measuring Returns over Different Holding Periods

Question: What is the return in this example over the life of the
bond?
Answer: The investor’s money will grow by a factor of 100/P (T ).
Therefore, the holding period return is:
100
r(T ) = −1
P (T )

Note that a different way to express this is:

price increase+Income 100 − P (T ) + 0

HP R = =
P (T ) P (T )

These slides were prepared by Alexandre Rubesam Slide 8

Measuring Returns over Different Holding Periods

▶ The formula provided gives the total return of the bond over
its life.
▶ If you are willing to invest your money for longer periods, you
should expect to earn higher total returns.
▶ How should we compare returns on investments with differing
horizons? ⇒ effective annual rate (EAR)

These slides were prepared by Alexandre Rubesam Slide 9

Measuring Returns over Different Holding Periods

▶ For the one-year investment (second row in the table), the

EAR is simply the total return on the bond: EAR = r(T )
▶ For investments of less than 1 year (say, 6 months), we would
compound the half-year return over two periods:
EAR = (1 + 0.0271)2 − 1
▶ For investments of more than 1 year, we find the EAR that
compounds to the same value as the investment. In the case
of the 25-year bond, the investment grows by a factor of
100/23.30 = 4.2918, so we solve

(1 + EAR)25 = 4.2918
1 + EAR = 4.29181/25 = 1.06 ⇒ EAR = 0.06

These slides were prepared by Alexandre Rubesam Slide 10

Annual Percentage Rates

▶ The EAR accounts for compound interest. In practice, some

instruments, such as short-term investments (i.e., less than a
year) are often annualized using simple interest. These are
called Annual Percentage Rates (APRs)
▶ Example: the APR reported on your credit card is the monthly
interest rate you pay on outstanding balances multiplied by 12.
▶ With n compounding periods per year, the EAR can be
calculated from the APR by first computing the per-period
rate as AP R/n and then compounding for n periods:

AP R n
 
1 + EAR = 1 +
n
h i
AP R = n (1 + EAR)(1/n) − 1

These slides were prepared by Alexandre Rubesam Slide 11

Annual Percentage Rates

Example 1 (EAR versus APR)

Suppose you must pay 1.5% interest per month on your
outstanding credit card balance. The APR would be reported as
1.5% × 12 = 18%, but the effective annual rate is higher, 19.56%,
because 1.01512 − 1 = 0.1956. Until you pay off your outstanding
balance, it will increase each month by the multiple 1.015, so after
12 months, the balance will increase by 1.01512 . This is why we
call the EAR an effective rate.

These slides were prepared by Alexandre Rubesam Slide 12

Annual Percentage Rates

▶ The table above uses the previous equations to compute the APR
corresponding to an EAR of 5.8% for different compounding periods
(and conversely, on the right-most 2 columns, the EAR
corresponding to an APR of 5.8%)
▶ Note that, for a fixed EAR, the APR decreases with more frequent
compounding. Likewise, for a fixed APR, the EAR increases with
more frequent compounding.
These slides were prepared by Alexandre Rubesam Slide 13
Continuous Compounding

▶ We could compound more frequently than daily (i.e. every

hour, minute etc)
▶ When the number of compounding periods (n) increases
indefinitely, we have what is called continuous compounding
▶ Consider a nominal interest rate r with compounding
happening n times per year. The future value after one year
of some amount (say, $1) today will be given by
 r n
FV = 1 × 1 +
n
▶ When n → ∞, the expression 1 + nr n → er , where e is

Euler’s number (or the base of the natural logarithm)

These slides were prepared by Alexandre Rubesam Slide 14

Continuous Compounding

To find the continuously compounded interest rate (rcc ), we proceed the

same way by equating the EAR under both methods:

1 + EAR = ercc
⇒ rcc = ln (1 + EAR)

As an example, in the preceding table, for an EAR of 5.8%, the

continuously compounded interest rate is rcc = ln (1 + 0.058) = 0.05638
or 5.638%.
Likewise, given a continuously compounded interest rate of
rcc = 0.05638, we can calculate EAR = ercc − 1 = e0.05638 − 1 = 0.058.
Continuously compounded interest rates may seem more complicated,
but they simplify calculations. For example, the compounding value over
a number of T years can be calculated simply as eT ×rcc .

These slides were prepared by Alexandre Rubesam Slide 15

Interest Rates and Inflation Rates

These slides were prepared by Alexandre Rubesam Slide 16

Interest Rates and Inflation Rates

Fundamental factors that determine the level of interest rates:

1. Supply of funds from savers, primarily households.

2. Demand for funds from businesses to be used to finance
investments in plant, equipment, and inventories.
3. Government’s net demand for funds as modified by actions of
the Central Bank.
4. Expected rate of inflation.

These slides were prepared by Alexandre Rubesam Slide 17

Interest Rates and Inflation Rates

▶ A nominal interest rate (rnom ) is the growth rate of your

money.
▶ A real interest rate (rreal ) is the growth rate of your
purchasing power.

Suppose that 1 year ago, you deposit 1,000 e in 1-year bank

deposit guaranteeing a rate of interest of 10%. You’re about to
collect 1,1000 e. What is the real return on your investment?
It depends on what your money can buy today, relative what you
could buy 1 year ago → this is determined by the rate of change in
the Consumer Price Index (CPI).

These slides were prepared by Alexandre Rubesam Slide 18

Interest Rates and Inflation Rates

Example 2
▶ Suppose that, one year ago, a baguette cost €1, and that the
rate of inflation is i = 6%.
▶ Because of inflation, a baguette now costs €1.06.
▶ Last year, you could buy 1,000
1.00 = 1, 000 baguettes with your
€1,000.
▶ But now, you can buy only 1,100
1.06 ≈ 1038 baguettes.
▶ Your purchasing power has increased at the rate of 3.8%
(much lower than the 10% interest on your investment).

These slides were prepared by Alexandre Rubesam Slide 19

Interest Rates and Inflation Rates

More generally, to find a real rate of return, we solve:

1 + rnom rnom − i
1 + rreal = ⇒ rreal =
1+i 1+i

In the previous example, we have

0.10 − 0.06
rreal = ≈ 0.038 = 3.8%
1.06

A common approximation is

rreal ≈ rnom − i

These slides were prepared by Alexandre Rubesam Slide 20

Interest Rates and Inflation -Equilibrium Real Rate of Interest

▶ Supply curve slopes upwards: if real interest rate is high, households

postpone consumption and invest more for future use.
▶ Demand curves slope downwards because businesses will want to
invest more in physical capital when the real interest rate on the
funds needed is lower.
These slides were prepared by Alexandre Rubesam Slide 21
Interest Rates and Inflation - Fisher Hypothesis

▶ We expect higher nominal interest rates when inflation is higher.

▶ Fisher hypothesis: nominal rates increase one-for-one with expected
inflation:
rnom = rreal + E(i)

These slides were prepared by Alexandre Rubesam Slide 22

Interest Rates and Inflation - Treasury Bills and Inflation

Figure 2: Interest Rates and Inflation, 1927 to 2021

These slides were prepared by Alexandre Rubesam Slide 23

 Risk and Risk Premiums

These slides were prepared by Alexandre Rubesam Slide 24

Risk and Risk Premiums - Holding Period Return

Suppose you buy a stock fund today for $100 per share. What will
be your return if you hold it for one year?
The return will depend on two things:

▶ the price of each share after one year

▶ the dividends paid over the year

The Holding Period Return is calculated as:

P1 − P0 + D1
HP R =
P0
where P0 is the price of one share today; P1 is the price of one
share one year from now, and D1 is the dividend paid over year.

These slides were prepared by Alexandre Rubesam Slide 25

Risk and Risk Premiums - Holding Period Return

Example 3 (One-Period Holding Period Return)

You buy one share of a fund for $100 today. Suppose the price after one
year is $110, and cash dividends over the year amount to $4. The HPR is
P1 − P0 + D1 110 − 100 + 4
HP R = = = 0.14 = 14%
P0 100

Note that HPR treats dividends as paid at the end of the period.
We can expres HPR as:
P1 − P0 D1
HP R = +
P0 P0
P1 − P0 D1
= +
P P0
| {z0 } |{z}
capital gain yield dividend yield

These slides were prepared by Alexandre Rubesam Slide 26

Risk and Risk Premiums - Expected Return

▶ In practice, we don’t know in advance (ex-ante) what our

realized (ex-post) HPR will be
▶ But we may be able to quantify the likelihood or probability of
different scenarios for future prices and dividends
▶ With these, we can form an expectation of the HPR - the
expected return. Associated with this is a measure of
dispersion (standard deviation or volatility )

These slides were prepared by Alexandre Rubesam Slide 27

Risk and Risk Premiums - Expected Return

Given a set of scenarios s, each with probability p(s) and HPR

r(s), the expected return is calculated as

X
E(r) = p(s)r(s)
s

The expected return is just the weighted average of the returns

under each scenario, with the weights equal to the probabilities of
each scenario
Obs: This is the same calculation for a discrete prob. distribution that we saw
in the review of statistics on the first session, but the notation in BKM is
slightly different.

These slides were prepared by Alexandre Rubesam Slide 28

Expected Return

Example 4 (Expected Return from Scenarios)

Considering the scenarios below for the holding period return on an

investment, what is the expected return?

State Probability Year-End Price Dividends HPR (r)

Excellent 0.25 126.5 4.5 0.31
Good 0.45 110 4 0.14
Poor 0.25 89.75 3.5 -0.0675
Crash 0.05 46 2 -0.52

E(r) = 0.25 × 0.31 + 0.45 × 0.14 + 0.25 × −0.0675

+0.05 × −0.52
= 0.0976 = 9.76%
These slides were prepared by Alexandre Rubesam Slide 29
Risk and Risk Premiums - Variance, Std Deviation

The variance (σ 2 ) provides a measure of the dispersion of values

around the expected value:

X
σ 2 (r) = p(s)[r(s) − E(r)]2
s

The standard deviation (σ) is just the square root of the variance
(therefore it’s in the same unit as the E(r))

These slides were prepared by Alexandre Rubesam Slide 30

Risk and Risk Premiums - Variance, Std Deviation

Example 4 (continued)
In our previous example:

σ 2 (r) = 0.25 × (0.31 − 0.0976)2 + 0.45 × (0.14 − 0.0976)2 +

0.25 × (−0.0675 − 0.0976)2 + 0.05 × (−0.52 − 0.0976)2
= 0.0380
√
σ = σ 2 = 0.1949 = 19.49%

These slides were prepared by Alexandre Rubesam Slide 31

Risk and Risk Premiums - Variance, Std Deviation

▶ In Example 4, we have E(r) = 9.75% and σ = 19.49%.

▶ The standard deviation shows that there is a level of risk (on
average you get 9.75%, but there is a lot of dispersion)
▶ Std dev. does not distinguish between good or bad outcomes.
Returns above or below E(r) impact σ the same way
▶ Only makes sense if distribution of returns is symmetric about
E(r)

These slides were prepared by Alexandre Rubesam Slide 32

Risk and Risk Premiums - Excess Returns and Risk Premiums

▶ The previous example shows that there is a level of risk

associated with the investment
▶ If you don’t want risk, you can invest in a risk-free asset
(such as a Treasury Bill in the U.S or a BTF in France) and
earn the risk-free rate
▶ The difference between the expected return on a risky
investment and the risk-free rate is called the risk premium
▶ The difference between the actual return on a risky
investment and the risk-free rate is called the excess return

These slides were prepared by Alexandre Rubesam Slide 33

Risk and Risk Premiums - Sharpe ratio

▶ The reward-to-variability or Sharpe ratio, after William

Sharpe, is the most commonly used performance metric in the
investment industry
▶ It measures the attractiveness of an asset as the ratio of its
return in excess of the risk-free rate, and its risk, as measured
by the asset volatility

E(r) − rf
SR =
σ
√
Obs: to annualize the Sharpe ratio, we multiply by the of the
annualization factor (number of periods that makes one year). For example, to
√
annualize a monthly SR, we multiply it by 12 (why?)

These slides were prepared by Alexandre Rubesam Slide 34

 The Normal Distribution

These slides were prepared by Alexandre Rubesam Slide 35

The Normal Distribution

▶ The Normal or Gaussian distribution is probably the most important

parametric probability distribution
▶ It is given by the following probability density function:
1 1 2
f (x) = √ e− 2σ2 (x−µ)
2πσ

These slides were prepared by Alexandre Rubesam Slide 36

The Normal Distribution

Notice some characteristics of the Normal distribution:

▶ The expected value is E(X) = µ and the variance is

Var(X) = σ 2
▶ The distribution is symmetrical about the mean
▶ The whole distribution is characterized fully by only two
parameters: the mean µ and the std deviation σ
▶ Areas under the distribution (probabilities of ranges of values)
easily relate to how many std deviations away from the mean
we are

These slides were prepared by Alexandre Rubesam Slide 37

The Normal Distribution

In addition, using the Normal distribution awards tractability:

▶ Linear combinations (say portfolios) of Normal variables are

also Normal
▶ The dynamics among Normal variables can be fully modelled
using a simple measure (the correlation)

These slides were prepared by Alexandre Rubesam Slide 38

The Normal Distribution

But asset returns are usually not normally distributed. What

happens then?

These slides were prepared by Alexandre Rubesam Slide 39

Deviations from The Normal Distribution

If asset returns are not normally distributed:

▶ Standard deviation is not necessarily a good measure of risk

▶ Sharpe ratio is not necessarily a good measure of portfolio
performance
▶ We need to consider higher moments of the distribution, such
as skewness and kurtosis
" 3 #
R−µ
Skewness = E (= 0 for any symmetric distribution)
σ
" 4 #
R−µ
Kurtosis = E − 3 (= 0 for a Normal distribution)
σ

These slides were prepared by Alexandre Rubesam Slide 40

Deviations from The Normal Distribution

These slides were prepared by Alexandre Rubesam Slide 41

Deviations from The Normal Distribution

These slides were prepared by Alexandre Rubesam Slide 42

Deviations from The Normal Distribution

Example 5 (Distribution of S&P 500 returns)

The graph below shows histograms of daily and monthly returns on the
S&P 500 index (data from January 2000 to September 2025). The red
curves denote Gaussian fits. What do you conclude?

These slides were prepared by Alexandre Rubesam Slide 43

Deviations from Normality and Tail Risk

These slides were prepared by Alexandre Rubesam Slide 44

Deviations from Normality and Tail Risk - Value at Risk

▶ The Value at Risk, or VaR, is a loss that we are fairly sure

will not be exceeded if the current portfolio is held over some
period of time
▶ For a given portfolio, time horizon, and significance level α,
the α% VaR is defined as a value such that the probability of
a loss greater than the VaR is at most α
▶ For example, if a portfolio has a 1-day 5% VaR of $1 million,
we are 95% confident that, over one day, the loss on the
portfolio will not exceed $1 million
▶ VaR is important in risk management, widely used by
practitioners, and required by regulation in banks

These slides were prepared by Alexandre Rubesam Slide 45

Deviations from Normality and Tail Risk - Value at Risk

Source: Alexander, C. (2009). Market risk analysis, Value at Risk Models.

John Wiley & Sons.
These slides were prepared by Alexandre Rubesam Slide 46
Deviations from Normality and Tail Risk - Calculating VaR

▶ If returns are normally distributed with expected return µ and

standard deviation σ, VaR can be easily calculated:

V aRN ormal (α) = µ − Φ−1 (1 − α)σ,

where Φ−1 () is the inverse of the standard normal distribution

function ([Link] function in Excel)
▶ For example, to calculate the 1% VaR, we would use
Φ−1 (0.99) = 2.33, and the formula becomes

V aRN ormal (1%) = µ − 2.33σ

These slides were prepared by Alexandre Rubesam Slide 47

Deviations from Normality and Tail Risk - Calculating VaR

▶ Practitioners never use the Normal VaR (guess why?). The

most widely used approach currently is the historical
simulation VaR, which uses past values of returns to
compute the VaR and therefore makes no assumptions about
the distribution of returns.
▶ To estimate historical VaR, given a time series of returns, we
sort the observations from high to low. The VaR is the return
at the desired percentile of the sample distribution.
▶ For example, to estimate 1% VaR using this approach, we
take the 1st percentile of the sample distribution of returns.

These slides were prepared by Alexandre Rubesam Slide 48

Deviations from Normality and Tail Risk - Expected Shortfall

▶ One problem with VaR is that it tells you only the loss that
will not be exceeded with a certain confidence
▶ But if it the loss is worse than the VaR, how much do we
expect to lose?
▶ The Expected Shortfall (ES), also known as conditional tail
expectation or expected tail loss is the average loss, given
that the VaR was exceeded
▶ Using a sample of historical returns, we can estimate the 1%
expected shortfall by identifying the worst 1% of all
observations and taking their average.

These slides were prepared by Alexandre Rubesam Slide 49

VaR and ES - Practical Example

Example 6 (VaR and ES for S&P 500 returns)

Using the same data from Example 5, we estimate the following:

Daily Monthly
α Gaussian Historical Gaussian Historical
VaR 1.98% 1.85% 6.61% 7.97%
5%
ES 2.49% 2.94% 8.45% 9.71%
VaR 2.82% 3.43% 9.60% 10.88%
1%
ES 3.23% 4.94% 11.09% 12.86%

▶ Gaussian assumption typically leads to underestimating VaR and

ES, especially at higher confidence levels (low α).

These slides were prepared by Alexandre Rubesam Slide 50

Normality and Risk Measures: Downside Risk

Using the standard deviation as a measure of risk when the return

distribution is nonnormal presents two problems:

▶ The asymmetry of the distribution suggests that we should

look that at negative outcomes (left tail of the distribution)
▶ Because an alternative to a risky portfolio is investing in a
risk-free asset, it makes more sense to look at deviations from
the risk-free rate (not the sample average).

These slides were prepared by Alexandre Rubesam Slide 51

Normality and Risk Measures: Downside Risk

▶ With a symmetric distribution and a benchmark equal to the

mean, downside and upside volatility are equally split
▶ With a skewed distribution or a benchmark different from the
mean, downside and upside volatility are uneven.

Source: Rigamonti (2020)

These slides were prepared by Alexandre Rubesam Slide 52
Normality and Risk Measures: Downside Risk

▶ An alternative risk measure that addresses these issues is the

lower partial standard deviation (LPSD) of excess returns,
which is computed like the usual standard deviation, but using
only the “bad” returns, i.e., negative deviations from the
risk-free rate.
▶ The formula for a benchmark rate of return B (usually the
risk-free rate) is:
v
u
u 1 XT
σD = t [min (rt − B, 0))2 ]
TD
i=1

where TD is the number of observations for which rt − B < 0

These slides were prepared by Alexandre Rubesam Slide 53

Normality and Risk Measures: Downside Risk

▶ If we use the LPSD as a measure of risk, it makes sense to

replace the Sharpe ratio with the ratio of average excess
returns to LPSD.
▶ This alternative risk-adjusted performance metric is known as
the Sortino ratio.

These slides were prepared by Alexandre Rubesam Slide 54

Learning from Historical Returns

These slides were prepared by Alexandre Rubesam Slide 55

Learning from Historical Returns

▶ The scenario-based approach allows one to form a subjective

view of expected returns and volatilities of assets
▶ What we can actually observe are the realized returns, from
which we can try to infer E(r) and σ

These slides were prepared by Alexandre Rubesam Slide 56

Learning from Historical Returns - Arithmetic Average Return

Assume we have n observations of realized returns. We can use

each one as an equally likely “scenario” and apply the previous
formula for expected return, which yields the arithmetic average
return:

n n
X 1X
E(r) = p(s)r(s) = r(s)
n
s=1 s=1

▶ This is an unbiased estimated of the expected return

▶ However it does not represent the actual performance of an
investment of the period

These slides were prepared by Alexandre Rubesam Slide 57

Learning from Historical Returns - Geometric Average Return

The geometric average return is a constant rate of return that

compounds to the terminal value (T V ) of an investment:

T V = (1 + r1 )(1 + r2 ) · · · (1 + rn )

We want to find a rate g that, if compounded n times, gives us the

same T V :

(1 + g)n = T V ⇒ g = T V 1/n − 1

g = geometric average return

These slides were prepared by Alexandre Rubesam Slide 58

Learning from Historical Returns - Geometric Average Return

Example 7
The table below shows the annual (total) return of the S&P 500
index from 2010 to 2020, and the TV at each year.

Year Return TV

2010 0.1506 1.1506

2011 0.0211 1.1749
The number of observations is: n = 11.
2012 0.1600 1.3630
2013 0.3239 1.8044
2014 0.1369 2.0514
2015 0.0138 2.0798 Arith. average = 0.1454 = 14.54%
2016 0.1196 2.3285 Geo. average = g = 0.1399 = 13.99%
2017 0.2183 2.8369
2018 -0.0438 2.7125 (T V 1/n − 1 = 4.22281/11 − 1)
2019 0.3149 3.5666
2020 0.1840 4.2228

These slides were prepared by Alexandre Rubesam Slide 59

Learning from Historical Returns - Variance

Recall the definition of the variance:

X
σ 2 (r) = p(s)[r(s) − E(r)]2
s

Using historical data, we can estimate variance using the sample

mean r̄ instead of the unobservable, true E(r):

n
2 1X
σ̂ (r) = [r(s) − r̄]2
n
s=1

As it turns out, this estimator is biased. An unbiased estimator is

given by:
n
2 1 X
σ̂ (r) = [r(s) − r̄]2
n−1
s=1
These slides were prepared by Alexandre Rubesam Slide 60
Suggested Exercises and Homework

These slides were prepared by Alexandre Rubesam Slide 61

Suggested Exercises and Homework

▶ BKM Chapter 5 Problem Sets: 7, 8, 9, 10.

▶ BKM Chapter 5 CFA Problems: 1, 2, 3, 4, 5, 6, 7.
▶ Additional Python homework: using the yfinance package,
download historical data for TSLA starting in 2010. Then,
calculate daily returns and create an histogram (tip: use the
matplotlib package). Fit and plot a normal distribution on top
of the histogram. Is it a good fit for these data? Challenge:
calculate the historical VaR and ES and overlay them in the
graph as vertical lines.

These slides were prepared by Alexandre Rubesam Slide 62