QUANTITATIVE PORTFOLIO

SELECTION FOR MANAGEMENT:

FOUNDATIONS

MANIMP

Lecture 2 – Basics of Probability

Federica Ricca
 Random variables

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Random variables
In all real-life decision contexts there are some elements or events which the
decisor cannot control completely.
The outcome of a decision will be always influenced by future unexpected
events with an uncertain outcome without the possibility for the decision
maker to predict the effects of such events.
Portfolio selection:
in order to select which assets to include in our portfolio, one can observe the
historical trend of the returns, but it is not possible to predict a break-down due
to unexpected events.

Probability Theory can be exploited in these cases to model the behaviour of

the return of an asset by representing it as a random variable.

A random (quantitative) variable is an unknown quantity which assumes

different numerical values depending on the different possible realizations
of a given event.

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Random variables
Denote by:
Ω the space of all the possible events
ω∈Ω one generic event in Ω

The random variable X is a function defined as follows:

X: Ω → R
It associates to each possible event ω in Ω a real number that we denote
by X(ω)=x which is the value of the random variable associated to the event
ω, also called realization of X.

VARIABLE NATURE PROBABILITY

DISCRETE pk is the probability that

the set of possible values is X is equal to xk,
countable (K possible values) k=1,2,…,K
RANDOM
VARIABLE
CONTINUOUS
the set of possible values is Density function 𝑓(𝑋)
uncountable
QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA
 Characteristic indices of a distribution

In some situations it is not easy to We can compute some characteristic

compute the complete probability indices of the distribution that are able to
distribution of a random variable. summarize the important information
(indices of central tendency, of dispersion).

In some cases it is even better to rely only on the indices:

a) coincise information can be better managed;
b) generally this coincise information is sufficient for the purposes of the analysis.

In portfolio selection analysis we will model the asset’s return as a random variable,
and we follow a mean-variance analysis approach.
We consider the following distribution indices:
• the expected value or expectation (index of position or central tendency)
• the variance (index of dispersion)
of the asset’s return.

Assets’ returns are The portfolio’s return is a function of the Portfolio’s return is a
random variables returns of the assets which compose it random variable

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Random variables:
expected value and its
properties

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Moments of the distribution and expected value
For the probability distribution of a discrete r.v. X with a finite number of K
possible realizations, the moment of order 𝒓 is defined as follows:

where x0 is the reference point

𝐾 and 𝑟 is the moment’s order.
General Formula ෍ 𝑥𝑘 − 𝑥0 𝑟 𝑝𝑘 We can compute the moment with
𝑘=1 respect to the origin 𝑥0 = 0.

When 𝒓 = 𝟏 and 𝒙𝟎 = 𝟎, we obtain

the moment of order 2 of X, that is
the Expected Value of X:

𝐸(𝑋) = ෍ 𝑥𝑘 𝑝𝑘 Expected value operator:

𝑘=1
𝐸(∙) = ෍∙ 𝑝(∙)
It is also called:
Mean or Expected value
or Moment of order 𝑟 = 1
from the origin 𝒙𝟎 = 𝟎 Equivalent notation: 𝐸(𝑋) = ෍ 𝑥 𝑝(𝑥)
𝑥
QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA
 Moments of the distribution and expected value
Example 1 (𝐾 = 3)
𝐾
Outcomes xk Numerical Probability
of r.v. X values 𝐸(𝑋) = ෍ 𝑥𝑘 𝑝𝑘
x1 4 ¼ 𝑘=1

x2 8 ¼ = 𝑥1 𝑝1 + 𝑥2 𝑝2 + 𝑥3 𝑝3 𝐴
x3 10 ½
= 4 ∙ 0.25 + 8 ∙ 0.25 + (10) ∙ 0.5
The set of all possible numerical values that
the r.v. can take is called Support of X = 1 + 2 + 5=𝟖

𝐸(𝑋) = ෍ 𝑥𝑘 𝑝𝑘
𝑘=1

It is also called:
Mean or Expected value
or Moment of order 𝑟 = 1
from the origin 𝒙𝟎 = 𝟎 𝐸(𝑋) = ෍ 𝑥 𝑝(𝑥)
𝑥
QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA
 Properties of the expected value E(X)
Property 0 (PE0)
Example: 𝛼 = 3
If a r.v. X assumes only one value α (scalar), we have:
We have: only 𝑥1 = 𝛼 and
E(X) = α 𝑝1 = 1:
In fact, X assumes only the value α with probability p(x=α)=1. 𝐸 𝑋 =𝛼∙1
=3∙1=3
𝐸 𝑋 = ෍𝑥 𝑝 𝑥 = 𝛼 ∙1 = 𝛼
𝑥
In fact, X is a constant (does not vary):
the expected value of a constant is the constant itself and we write E(α) = α.

Property 1 (PE1)
Given a scalar β, we have:
E(βX) = β E(X)

We have:

𝐸 𝛽𝑋 = ෍ 𝛽𝑥 𝑝 𝑥 = 𝛽 ෍ 𝑥 𝑝 𝑥 = 𝛽 𝐸(𝑋) 𝐸(𝑋) = ෍ 𝑥 𝑝(𝑥)

𝑥 𝑥 𝑥

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Properties of the expected value E(X)
𝐾
Example 1 (𝐾 = 3, 𝛽 = 2)
𝐸(𝛽𝑋) = ෍ 𝛽𝑥𝑘 𝑝𝑘
Outcomes xk Numerical Probability 𝑘=1
of r.v. X values
x1 4 ¼
= 𝛽𝑥1 𝑝1 + 𝛽𝑥2 𝑝2 + 𝛽𝑥3 𝑝3 𝐴
x2 8 ¼ = 2 ∙ 4 ∙ 0.25 + 2 ∙ 8 ∙ 0.25 + 2 ∙ 10 ∙ 0.5
x3 10 ½ = 2 + 4 + 10
𝑬(𝑿) = 𝟖 = 𝟏𝟔 = 2 ∙ 8 = 𝛽𝐸(𝑋)

Property 1 (PE1)
Given a scalar β, we have:
E(βX) = β E(X)

We have:

𝐸 𝛽𝑋 = ෍ 𝛽𝑥 𝑝 𝑥 = 𝛽 ෍ 𝑥 𝑝 𝑥 = 𝛽 𝐸(𝑋) 𝐸(𝑋) = ෍ 𝑥 𝑝(𝑥)

𝑥 𝑥 𝑥

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Properties of the expected value E(X)
Property 2 (PE2)
The expected value of the sum of two random variables X and Y is equal to the sum of
the expected values of X and of Y:
E(X+Y) = E(X)+E(Y)

where PXY is the joint probability

of variables X and Y

Expected value operator:

𝐸(∙) = ෍∙ 𝑝(∙)

Properties 1 e 2 follow from the fact that

𝐸(𝑋) = ෍ 𝑥 𝑝(𝑥)
E(X) is a linear function (operator).
𝑥
QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA
 Properties of the expected value E(X)
(PE0) (PE1)

(PE2)

Properties 0, 1 and 2 imply:

E(X – α) = E(X) – α

In fact:
PE2 PE1 PE0
E(X – α) = E(X) + E(–α) = E(X) + (–1)∙E(α) = E(X) – α

Special case: Let E(X) =μ (expected value of X), and fix α =μ in the above
formula, then we have:
E(X – μ) = E(X) – μ = μ – μ = 0

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Properties of the expected value E(X)
Properties 1 and 2 imply the following, more general, property:

Property 3 (linearity)
The expected value of a linear combination of two r.v. X and Y, with real
weights α and β, respectively, is equal to the linear combination of the expected
values of X and Y with the same weights:
E(αX+βY) = αE(X)+βE(Y) We say that the
expected value is
invariant for linear
transformations

Property 4 Let xmin and xmax be the minimum and maximum values among the
possible realizations of X, then:
xmin ≤ E(X) ≤ xmax

The expected value of a r.v. X is a number between xmin and xmax.

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Expected value E(X) Property 4: xmin ≤ E(X) ≤ xmax

EXAMPLE 2
Consider the discrete r.v. X with support {x1= –1, x2=1, x3=2} and the following
different possible probability distributions:
1. p1=0.25, p2= 0.25, p3=0.50;
2. p1=0.25, p2= 0.50, p3=0.25;
3. p1=0.50, p2= 0.25, p3=0.25.
Compute the expected value of X in the three different cases.
Solution:
1. E(X) = x1p1+x2p2+x3p3 = 0.25 (–1) + 0.25 (1) + 0.50 (2) = 1

2. E(X) = x1p1+x2p2+x3p3 = 0.25 (–1) + 0.50 (1) + 0.25 (2) = 0.75

3. E(X) = x1p1+x2p2+x3p3 = 0.50 (–1) + 0.25 (1) + 0.25 (2) = 0.25

NOTE: in all cases the value of E(X) is between xmin = –1 and xmax = 2.

NOTE: The value of E(X) changes according to the probability distribution considered.
It decreases as the probability associated to the lowest realizations of X increases (in this
sense E(X) is a positional index).

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Random variables:
variance and its
properties

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Second order moments and variance
Consider the general formula for the moment of order 𝒓 for a r.v. X (discrete,
finite and with K possible realizations):
where x0 is the reference point
𝐾 and 𝑟 is the moment’s order.
෍ 𝑥𝑘 − 𝑥0 𝑟 𝑝𝑘 We can compute the moment with
𝑘=1 respect to the origin 𝑥0 = 0 or w.r.t the
average of the values of 𝑥0 = 𝜇
When 𝒓 = 𝟐 we obtain the (centered moment).
moment of order 2 of X:
𝐾
2
෍ 𝑥𝑘 − 𝑥0 𝑝𝑘
𝑘=1

When 𝒙𝟎 = 𝝁 (and 𝒓 = 𝟐), we have:

𝐾
2 Variance
𝑉𝑎𝑟(𝑋) = ෍ 𝑥𝑘 − 𝝁 𝑝𝑘 (Index of dispersion of the
𝑘=1 values of X around 𝜇)

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Second order moments and variance
Note that we can use the coincise operator E(X):

𝐾
2 2
𝑉𝑎𝑟(𝑋) = ෍ 𝑥𝑘 − 𝝁 𝑝𝑘 = 𝐸 𝑋−𝝁
𝑘=1

Equivalently, since μ=E(X): 2

= 𝐸 𝑋 − 𝑬(𝑿)

Expected value of the squared

deviations of X from 𝜇

When 𝒙𝟎 = 𝝁 (and 𝒓 = 𝟐), we have:

𝐾
2 Variance
𝑉𝑎𝑟(𝑋) = ෍ 𝑥𝑘 − 𝝁 𝑝𝑘 (Index of dispersion of the
𝑘=1 values of X around 𝜇)

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Second order moments and variance
Property 0 (PV0)
The variance of X is always non negative:
Var(X) ≥ 0
It assumes value 0 only if the support of X is given by only one value (X=α).

Property 1 (PV1)
Consider the linear transformation of X given by Y= αX+β (α and β are numbers)
and the property of the expected value:
NOTE:
E(Y) = αE(X)+β
𝑉𝑎𝑟(𝑋 + 𝛽) = 𝑉𝑎𝑟(𝑋)
For the square deviations of Y from its average, one has:
2 2
𝑌 − 𝐸(𝑌) = 𝛼𝑋 + 𝛽 − (𝛼𝐸 𝑋 + 𝛽)
2 2
= 𝛼𝑋 − 𝛼𝐸 𝑋 = 𝛼(𝑋 − 𝐸 𝑋 ) = 𝛼 2 𝑋 − 𝐸(𝑋) 2

Therefore, the variance of Y is: The variance is not

invariant for linear
𝐸( 𝑌 − 𝐸(𝑌) 2 ) = 𝐸(𝛼 2 𝑋 − 𝐸(𝑋) 2 ) = 𝛼 2 𝐸( 𝑋 − 𝐸(𝑋) 2 ) transformations, but
it is scaled by the
𝑉𝑎𝑟(𝑌) = 𝑉𝑎𝑟(𝛼𝑋 + 𝛽) = 𝛼 2 𝑉𝑎𝑟(𝑋) value α2
QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA
 Properties of the variance
Property 2 (PV2) We use also the
An equivalent formula for the variance of X is the following: notation 𝜎𝑋2
and 𝜎𝑋
𝑉𝑎𝑟(𝑋) = 𝐸(𝑋 2 ) − 𝐸(𝑋) 2

Moment of order 2 of X Square of the

with x0=0 expected value of X

NOTE: The variance is a quadratic function of X, therefore its values are of the same
order of magnitude of the square of the values of X.

To obtain an index of dispersion in the scale of the values of X, we compute the

square root of the variance, i.e., the Standard Deviation of X:

𝐷𝑒𝑣(𝑋) = 𝑉𝑎𝑟(𝑋) The St. Dev. is not

invariant for linear
transformations, but
Property 1 of the variance implies that, if we consider Y= αX+β: it is scaled by the
value α

𝐷𝑒𝑣 𝛼𝑋 + 𝛽 = 𝑉𝑎𝑟(𝛼𝑋 + 𝛽) = 𝛼 2 𝑉𝑎𝑟 𝑋 = 𝛼𝐷𝑒𝑣(𝑋)

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Variance and Standard deviation IMPORTANT FORMULA

EXAMPLE 2 Consider the r.v. X, with:

support {x1= –1, x2=1, x3=2}
probabilities p1=0.25, p2= 0.25, p3=0.50, respectively.

For these probabilities, we have already computed E(X) = μ = 1.

Compute the variance (using the two alternative formulas) and the standard deviation of X.

Applying the variance’s formula, we have:

Var(X) = (x1–μ)2 p1+ (x2–μ)2 p2 + (x3–μ)2 p3

= (–1–1)2 0.25 + (1–1)2 0.25 + (2–1)2 0.50 = 1.5

Applying the alternative formula 𝑉𝑎𝑟(𝑋) = 𝐸(𝑋 2 ) − 𝐸(𝑋) 2 we first compute 𝐸(𝑋 2 ):

𝐸(𝑋 2 ) = (−1)2 0.25 + (1)2 0.25 + (2)2 0.50 = 2.5 Moment of order 2 from
the origin
𝑉𝑎𝑟(𝑋) = 𝐸(𝑋 2 ) − 𝐸(𝑋) 2
= 2.5 − 12 = 𝟏. 𝟓

We can then compute also

the standard deviation: 𝐷𝑒𝑣(𝑋) = 1.5 = 1.2248

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Random variables:
covariance and
correlation

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Covariance and Correlation IMPORTANT FORMULA

In our portfolio analysis we also use some indices which measure relations existing
between two random variable X and Y.
We focus on covariance and correlation.

Consider two r.v. X and Y, with K and H possible values, respectively, and expected
values μX and μY. The covariance’s formula is:

𝐾 𝐻 where pkh is the joint probability

of xk for X and yh for Y:
𝐶𝑜𝑣(𝑋, 𝑌) = ෍ ෍ 𝑥𝑘 − 𝜇𝑋 𝑦ℎ − 𝜇𝑌 𝑝𝑘ℎ
𝒑𝒌𝒉 = Prob(𝑋 = 𝑥𝑘 , 𝑌 = 𝑦ℎ )
𝑘=1 ℎ=1

This index measures how deviations of X from its average μX associates with
those of Y from μY.

Using the expected value operator E(∙), the formula of the covariance of X and Y can be
equivalently written as follows:

𝐶𝑜𝑣(𝑋, 𝑌) = 𝐸 𝑋 − 𝜇𝑋 𝑌 − 𝜇𝑌
The covariance of X and Y
Deviations of X and Y from their measures the linear dependence
mean values μX and μY. of the two random variables.

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Covariance and Correlation
To understand the meaning of the covariance, we consider a r.v. obtained as the
linear combination of X and Y with real weights α and β:
𝛼𝑋 + 𝛽𝑌 Deviations from the
And we compute its variance: expected value 𝑬 𝜶𝑿 + 𝜷𝒀
2
𝑉𝑎𝑟 𝛼𝑋 + 𝛽𝑌 = 𝐸 𝛼𝑋 + 𝛽𝑌 − 𝑬 𝜶𝑿 + 𝜷𝒀
Developing the above expression, we have: 𝐸 𝛼𝑋 + 𝛽𝑌 = 𝛼𝐸 𝑋) + 𝛽𝐸(𝑌
2
𝑉𝑎𝑟 𝛼𝑋 + 𝛽𝑌 = 𝐸 𝛼𝑋 + 𝛽𝑌 − 𝛼𝜇𝑋 + 𝛽𝜇𝑌
2
= 𝐸 𝛼 𝑋 − 𝜇𝑋 + 𝛽 𝑌 − 𝜇𝑌 We now develop the square

Here we have used

the linearity of E(∙) = 𝐸 𝛼 2 𝑋 − 𝜇𝑋 2
+ 𝛽 2 𝑌 − 𝜇𝑌 2
+ 2𝛼𝛽 𝑋 − 𝜇𝑋 𝑌 − 𝜇𝑌

= 𝛼 2 𝐸 𝑋 − 𝜇𝑋 2
+ 𝛽 2 𝐸 𝑌 − 𝜇𝑌 2
+ 2𝛼𝛽𝐸 𝑋 − 𝜇𝑋 𝑌 − 𝜇𝑌

Var(X) Var(Y) Cov(X,Y)

The covariance of X and Y
measures the linear dependence
= 𝛼 2 𝑉𝑎𝑟(𝑋) + 𝛽 2 𝑉𝑎𝑟(𝑌) + 2𝛼𝛽𝐶𝑜𝑣(𝑋, 𝑌) of the two random variables.

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Covariance and Correlation IMPORTANT FORMULA

Combining X and Y linearly, we have a new variable 𝛼𝑋 + 𝛽𝑌 with variance:

𝑉𝑎𝑟 𝛼𝑋 + 𝛽𝑌 = 𝛼 2 𝑉𝑎𝑟(𝑋) + 𝛽 2 𝑉𝑎𝑟(𝑌) + 2𝛼𝛽𝑪𝒐𝒗(𝑿, 𝒀)

There are three components:
𝜎𝑋2 = 𝑉𝑎𝑟(𝑋) measures the variability of X
𝜎𝑌2 = 𝑉𝑎𝑟(𝑌) measures the variability of Y
𝜎𝑋𝑌 = 𝐶𝑜𝑣(𝑋, 𝑌) measures the co-variability of X and Y
Cov(X,Y) quantifies the
The last component derives from te fact that we
consequences of having performed
started from a linear combination of X and Y
a linear combination of X and Y.
This component of variability is due
to co-movements between the two
linearly related variables X and Y.

= 𝛼 2 𝐸 𝑋 − 𝜇𝑋 2
+ 𝛽 2 𝐸 𝑌 − 𝜇𝑌 2
+ 2𝛼𝛽𝐸 𝑋 − 𝜇𝑋 𝑌 − 𝜇𝑌

Var(X) Var(Y) Cov(X,Y)

The covariance of X and Y
measures the linear dependence
= 𝛼 2 𝑉𝑎𝑟(𝑋) + 𝛽 2 𝑉𝑎𝑟(𝑌) + 2𝛼𝛽𝐶𝑜𝑣(𝑋, 𝑌) of the two random variables.

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Covariance and Correlation IMPORTANT PROPERTY

Consider the covariance of X and Y: We use also the

notation 𝜎𝑋𝑌
𝐶𝑜𝑣(𝑋, 𝑌) = 𝐸 𝑋 − 𝜇𝑋 𝑌 − 𝜇𝑌

X and Y positively correlated:

𝐶𝑜𝑣(𝑋, 𝑌) > 0 high values of X tend to
associate to high values of Y

X and Y negatively correlated:

𝐶𝑜𝑣(𝑋, 𝑌) 𝐶𝑜𝑣(𝑋, 𝑌) < 0 high values of X tend to associate
to low values of Y (or viceversa)

𝐶𝑜𝑣(𝑋, 𝑌) = 0 X and Y are non correlated

We have:

𝐶𝑜𝑣(𝑿, 𝑿) = 𝐸 𝑿 − 𝝁𝑿 𝑿 − 𝝁𝑿 = 𝑉𝑎𝑟(𝑋) The covariance of X with

itself corresponds to the
𝟐 variance of X
𝑬 𝑿 − 𝝁𝑿
QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA
 Covariance and Correlation
Covariance’s Properites

Property 0 (PCov0)
Given the r.v. X and a real constant 𝛼, we have:
There are no co-movements
𝐶𝑜𝑣(𝑋, 𝛼) = 0 between X and α since α is
constant and does never change

Property 2 (PCov1)
Given the r.v. X and Y and four real constants 𝛼, 𝛽, 𝛾, 𝛿, we have:

𝐶𝑜𝑣 (𝛼𝑋 + 𝛽), (𝛾𝑌 + 𝛿) = 𝛼𝛾𝐶𝑜𝑣(𝑋, 𝑌)

and
𝐶𝑜𝑣 (𝛼𝑋 + 𝛽), 𝑌 = 𝛼𝐶𝑜𝑣(𝑋, 𝑌)

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Covariance and Correlation IMPORTANT FORMULA

Provided that Var(X) and Var(Y) are strictly positive (different from 0), the
covariance can be normalized obtaining the correlation coefficient, 𝜌𝑋𝑌 :
𝐶𝑜𝑣(𝑋, 𝑌) 𝜎𝑋𝑌
𝜌𝑋𝑌 = =
𝐷𝑒𝑣(𝑋)𝐷𝑒𝑣(𝑌) 𝜎𝑋 𝜎𝑌

X and Y positively correlated:

𝐶𝑜𝑣(𝑋, 𝑌) > 0 𝜌𝑋𝑌 > 0 high values of X tend to associate to
high values of Y
X and Y negatively correlated:
𝐶𝑜𝑣(𝑋, 𝑌) 𝐶𝑜𝑣(𝑋, 𝑌) < 0 𝜌𝑋𝑌 < 0 high values of X tend to associate to
low values of Y (or viceversa)

𝐶𝑜𝑣(𝑋, 𝑌) = 0 𝜌𝑋𝑌 = 0 X and Y are non correlated

Cov(X,Y) e ρXY:
• covariance and correlation measure the same aspect Correlation 𝜌𝑋𝑌 is a
• the difference between the two is that ρXY does not depend pure number and
on the order of magnitude of the variances of X and Y 𝜌𝑋𝑌 ∈ [−1,1].

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Covariance and Correlation
Definition (perfectly correlated r.v. X and Y)
Two r.v. X and Y are positively perfectly correlated if the following linear
relation holds:
𝑌 = 𝛼𝑋 + 𝛽, 𝜶>𝟎
Two r.v. X and Y are negatively perfectly correlated if the following linear
relation holds:
𝑌 = 𝛼𝑋 + 𝛽, 𝜶<𝟎

In both cases we have:

𝑉𝑎𝑟 𝑌 = 𝑉𝑎𝑟 𝛼𝑋 + 𝛽 = 𝛼 2 𝑉𝑎𝑟 𝑋 = 𝛼 2 𝜎𝑋2 𝐷𝑒𝑣(𝑌) = 𝛼 ∙ 𝐷𝑒𝑣(𝑋) = 𝜶 𝝈𝑿

𝐶𝑜𝑣(𝑋, 𝑌) = 𝐶𝑜𝑣 𝑋, (𝛼𝑋 + 𝛽) = 𝛼𝐶𝑜𝑣(𝑋, 𝑋) = 𝛼𝑉𝑎𝑟(𝑋) = 𝜶𝜎𝑋2

All parameters for Y are functions of those of the r.v. X.
The sign of 𝜶 implies positive or negative covariance/correlation

Cov(X,Y) e ρXY:
• covariance and correlation measure the same aspect Correlation 𝜌𝑋𝑌 is a
• the difference between the two is that ρXY does not depend pure number and
on the order of magnitude of the variances of X and Y 𝜌𝑋𝑌 ∈ [−1,1].

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 Covariance and Correlation
Remember: 𝜎𝑌 = 𝛼 𝜎𝑋 and 𝜎𝑋𝑌 = 𝜶𝝈𝟐𝑿
If X and Y are perfectly correlated, for 𝝆𝑿𝒀 we have:
X and Y
𝜎𝑋𝑌 𝜶𝜎𝑋2 positively
𝜎𝑋𝑌 = 𝛼𝜎𝑋2 ≥0 𝛼>0 𝜌𝑋𝑌 = = = +1 perfectly
𝜎𝑋 𝜎𝑌 𝛼 𝜎𝑋2
correlated
 XY
 XY = 𝜎𝑋 𝜎𝑌 = 𝜎𝑋 𝛼 𝜎𝑋 = 𝜶 𝝈𝟐𝑿
 XY
X and Y
𝜎𝑋𝑌 𝜶𝜎𝑋2 negatively
𝜎𝑋𝑌 = 𝛼𝜎𝑋2 ≤ 0 𝜶 < 𝟎 𝜌𝑋𝑌 = = = −1 perfectly
𝜎𝑋 𝜎𝑌 𝛼 𝜎𝑋2 correlated

• when X and Y are perfectly correlated (positively or negatively) ρXY assumes

its maximum (absolute) value (𝜌𝑋𝑌 = 1 when positive, and 𝜌𝑋𝑌 = −1 when
negative).

• in all other cases ρXY takes values in the open Correlation 𝜌𝑋𝑌 is a
pure number and
interval (-1,1).
𝜌𝑋𝑌 ∈ [−1,1].

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 EXAMPLE
Consider the two random variables X and Y and their possible values in the
three different situations 1,2 and 3, each with probability equal to 1/3:
Situations X Y
15 + 9 + 3 27
1 15 16 𝐸(𝑋) = = =9
3 3
2 9 10
16 + 10 + 4 30
3 3 4 𝐸(𝑌) = = = 10
3 3
Exp. value 9 10

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 EXAMPLE
Consider the two random variables X and Y and their possible values in the
three different situations 1,2 and 3, each with probability equal to 1/3:
Situations X Y
1 15 16 𝐸(𝑋) = 9 𝜎𝑋 = 24

2 9 10
𝐸(𝑌) = 10 𝜎𝑌 = 24
3 3 4

Exp. value 9 10

We can now compute the variance and the standard deviation of X and Y.
We first compute the (three) square deviations from the mean for both X and Y.

Situations (X-E(X))2 (Y-E(Y))2

62 + 0 + (−6)2 72
1 (15-9)2 (16-10)2 𝜎𝑋2 = = = 24
3 3
2 (9-9)2 (10-10)2
2 2
(3-9)2 (4-10)2
6 + 0 + (−6) 72
3 𝜎𝑌2 = = = 24
3 3
SUM 72 72

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 EXAMPLE
Consider the two random variables X and Y and their possible values in the
three different situations 1,2 and 3, each with probability equal to 1/3:
Situations X Y
1 15 16 𝐸(𝑋) = 9 𝜎𝑋 = 24

2 9 10
𝐸(𝑌) = 10 𝜎𝑌 = 24
3 3 4

Exp. value 9 10

Now we can compute the covariance and the correlation between X and Y.

Situations X-E(X) Y-E(Y) (X-E(X))(Y-E(Y)) 72

𝜎𝑋𝑌 = = 24
1 (15-9) (16-10) 36
3
2 (9-9) (10-10) 0
𝜎𝑋𝑌 24
3 (3-9) (4-10) 36 𝜌𝑋𝑌 = = = +1
𝜎𝑋 𝜎𝑌 24 24
SUM 72

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA

 EXERCISES (Lect. II)
EXERCISE II-1
Consider the discrete r.v. X with support {x1=2, x2=4, x3=6} and the following
different probability distributions:
1. p1=1/3, p2= 1/3, p3=1/3;
2. p1=1, p2= 0, p3=0;
3. p1=0, p2= 0, p3=1.
Compute the expected value of X in the three different cases discussing the
results. What we can say about X in cases 2. and 3.?

EXERCISE II-2
Consider the discrete r.v. X with support {x1=15, x2=9, x3=3} and all probabilities
equal to 1/3. Also consider the discrete r.v. Y with support {y1=1, y2=10, y3=19}
and all probabilities all equal to 1/3.
Compute the covariance between X and Y and the corresponding correlation
coefficient.

QUANTITATIVE PORTFOLIO SELECTION FOR MANAGEMENT: FOUNDATIONS – FEDERICA RICCA