© CFA Institute. For candidate use only. Not for distribution.

Correlation between Two Variables 125

C. Both outliers and spurious correlation

Solution:
C is correct. Both outliers and spurious correlation are potential problems
with interpreting correlation coefficients.

Use the information in Exhibit 29 to answer questions 3 and 4.

An analyst is evaluating the tendency of returns on the portfolio of stocks
she manages to move along with bond and real estate indexes. She gathered
monthly data on returns and the indexes:

Exhibit 29: Monthly Data on Returns and Indexes

​

Returns (%)

Bond Index Real Estate Index

Portfolio Returns Returns Returns

Arithmetic average 5.5 3.2 7.8

Standard deviation 8.2 3.4 10.3
Portfolio Returns
and Bond Index Portfolio Returns and Real Estate
Returns Index Returns
Covariance 18.9 −55.9
​

3. Without calculating the correlation coefficient, the correlation of the portfo-

lio returns and the bond index returns is most likely to be:
A. negative.
B. zero.
C. positive.
Solution:
C is correct. The correlation coefficient is positive because the covariance is
positive.

4. Without calculating the correlation coefficient, the correlation of the portfo-

lio returns and the real estate index returns is:
A. negative.
B. zero.
C. positive.
Solution:
A is correct. The correlation coefficient is negative because the covariance is
negative.

5. Consider two variables, A and B. If variable A has a mean of −0.56, variable

B has a mean of 0.23, and the covariance between the two variables is posi-
tive, the correlation between these two variables is:
A. negative.
B. zero.
 © CFA Institute. For candidate use only. Not for distribution.
126 Learning Module 3 Statistical Measures of Asset Returns

C. positive.
Solution:
C is correct. The correlation coefficient must be positive because the cova-
riance is positive. The fact that one or both variables have a negative mean
does not affect the sign of the correlation coefficient.
 © CFA Institute. For candidate use only. Not for distribution.
Practice Problems 127

PRACTICE PROBLEMS

The following information relates to questions

1-5
Consider the results of an analysis focusing on the market capitalizations of a
sample of 100 firms:

Exhibit 1: Market Capitalization of a Sample of 100 Firms

Market Capitalization (euro

billions)
Cumulative Percentage Number of
Bin of Sample (%) Lower Bound Upper Bound Observations

1 5 0.28 15.45 5
2 10 15.45 21.22 5
3 15 21.22 29.37 5
4 20 29.37 32.57 5
5 25 32.57 34.72 5
6 30 34.72 37.58 5
7 35 37.58 39.90 5
8 40 39.90 41.57 5
9 45 41.57 44.86 5
10 50 44.86 46.88 5
11 55 46.88 49.40 5
12 60 49.40 51.27 5
13 65 51.27 53.58 5
14 70 53.58 56.66 5
15 75 56.66 58.34 5
16 80 58.34 63.10 5
17 85 63.10 67.06 5
18 90 67.06 73.00 5
19 95 73.00 81.62 5
20 100 81.62 96.85 5

1. The tenth percentile corresponds to observations in bin(s):

A. 2.

B. 1 and 2.

C. 19 and 20.

2. The second quintile corresponds to observations in bin(s):

A. 8.
 © CFA Institute. For candidate use only. Not for distribution.
128 Learning Module 3 Statistical Measures of Asset Returns

B. 5, 6, 7, and 8.

C. 6, 7, 8, 9, and 10.

3. The fourth quartile corresponds to observations in bin(s):

A. 17.

B. 17, 18, 19, and 20.

C. 16, 17, 18, 19, and 20.

4. The median is closest to:

A. 44.86.

B. 46.88.

C. 49.40.

5. The interquartile range is closest to:

A. 20.76.

B. 23.62.

C. 25.52.

6. Exhibit 12 shows the annual MSCI World Index total returns for a 10-year
period.

Exhibit 1: MSCI World Index Returns

Year 1 15.25% Year 6 30.79%

Year 2 10.02% Year 7 12.34%
Year 3 20.65% Year 8 −5.02%
Year 4 9.57% Year 9 16.54%
Year 5 −40.33% Year 10 27.37%

The fourth quintile return for the MSCI World Index is closest to:
A. 20.65 percent.

B. 26.03 percent.

C. 27.37 percent.

The following information relates to questions

7-9
A fund had the following experience over the past 10 years:
 © CFA Institute. For candidate use only. Not for distribution.
Practice Problems 129

Exhibit 1: Performance over 10 Years

Year Return

1 4.5%
2 6.0%
3 1.5%
4 −2.0%
5 0.0%
6 4.5%
7 3.5%
8 2.5%
9 5.5%
10 4.0%

7. The fund’s standard deviation of returns over the 10 years is closest to:
A. 2.40 percent.

B. 2.53 percent.

C. 7.58 percent.

8. The target semideviation of the returns over the 10 years, if the target is 2 per-
cent, is closest to:
A. 1.42 percent.

B. 1.50 percent.

C. 2.01 percent.

9. Consider the mean monthly return and the standard deviation for three industry
sectors, as shown in Exhibit 2:

Exhibit 2: Mean Monthly Return and Standard Deviations

Mean Monthly Return Standard Deviation of Return

Sector (%) (%)

Utilities (UTIL) 2.10 1.23

Materials (MATR) 1.25 1.35
Industrials (INDU) 3.01 1.52

Based on the coefficient of variation (CV), the riskiest sector is:

A. utilities.

B. materials.

C. industrials.
 © CFA Institute. For candidate use only. Not for distribution.
130 Learning Module 3 Statistical Measures of Asset Returns

SOLUTIONS
1. B is correct. The tenth percentile corresponds to the lowest 10 percent of the
observations in the sample, which are in bins 1 and 2.

2. B is correct. The second quintile corresponds to the second 20 percent of obser-

vations. The first 20 percent consists of bins 1 through 4. The second 20 percent
of observations consists of bins 5 through 8.

3. C is correct. A quartile consists of 25 percent of the data, and the last 25 percent
of the 20 bins are 16 through 20.

4. B is correct. The center of the 20 bins is represented by the market capitalization

of the highest value of the 10th bin and the lowest value of the 11th bin, which is
46.88.

5. B is correct. The interquartile range is the difference between the lowest value in
the second quartile and the highest value in the third quartile. The lowest value
of the second quartile is 34.72, and the highest value of the third quartile is 58.34.
Therefore, the interquartile range is 58.34 − 34.72 = 23.62.

6. B is correct. Quintiles divide a distribution into fifths, with the fourth quintile
occurring at the point at which 80 percent of the observations lie below it. The
fourth quintile is equivalent to the 80th percentile. To find the yth percentile (P y),
we first must determine its location. The formula for the location (Ly) of a yth
percentile in an array with n entries sorted in ascending order is Ly = (n + 1) ×
(y/100). In this case, n = 10 and y = 80%, so
L80 = (10 + 1) × (80/100) = 11 × 0.8 = 8.8.
With the data arranged in ascending order (−40.33 percent, −5.02 percent,
9.57 percent, 10.02 percent, 12.34 percent, 15.25 percent, 16.54 percent, 20.65
percent, 27.37 percent, and 30.79 percent), the 8.8th position would be between
the eighth and ninth entries, 20.65 percent and 27.37 percent, respectively. Using
linear interpolation, P80 = X8 + (Ly − 8) × (X9 − X8),
P80 = 20.65 + (8.8 − 8) × (27.37 − 20.65)

= 20.65 + (0.8 × 6.72) = 20.65 + 5.38

= 26.03 percent.

7. B is correct. The fund’s standard deviation of returns is calculated as follows:

Year Return Deviation from Mean Deviation Squared

1 4.5% 0.0150 0.000225

2 6.0% 0.0300 0.000900
3 1.5% −0.0150 0.000225
4 −2.0% −0.0500 0.002500
5 0.0% −0.0300 0.000900
6 4.5% 0.0150 0.000225
7 3.5% 0.0050 0.000025
8 2.5% −0.0050 0.000025
9 5.5% 0.0250 0.000625
 © CFA Institute. For candidate use only. Not for distribution.
Solutions 131

Year Return Deviation from Mean Deviation Squared

10 4.0% 0.0100 0.000100
Mean 3.0%
Sum 0.005750

The standard deviation is the square root of the sum of the squared deviations,
divided by n − 1:
_

√
0.005750
​s = ​ _
​  (​ 10 − 1) ​ ​​
​
= 2.5276%.

8. B is correct. The target semideviation of the returns over the 10 years with a tar-
get of 2 percent is calculated as follows:

Year Return Deviation Squared below Target of 2%

1 4.5%
2 6.0%
3 1.5% 0.000025
4 −2.0% 0.001600
5 0.0% 0.000400
6 4.5%
7 3.5%
8 2.5%
9 5.5%
10 4.0%
Sum 0.002025

The target semideviation is the square root of the sum of the squared deviations
from the target, divided by n − 1:
_

√
0.002025
sTarget = ​​ _
​  ​(10 − 1) ​ ​​
​
= 1.5%.

9. B is correct. The CV is the ratio of the standard deviation to the mean, where a
higher CV implies greater risk per unit of return.
1.23%
​​CV​  UTIL​​  = ​ _ _ ​  = ​ _ ​  = 0.59​,
s
​ X ​ 2.10%

1.35%
​​CV​  MATR​​  = ​ _ _ ​  = ​ _ ​  = 1.08​,
s
​ X ​ 1.25%

1.52%
​​CV​  INDU​​  = ​ _ _ ​  = ​ _ ​  = 0.51​.
s
​ X ​ 3.01%
© CFA Institute. For candidate use only. Not for distribution.
 © CFA Institute. For candidate use only. Not for distribution.

LEARNING MODULE

4
Probability Trees and
Conditional Expectations
by Richard A. DeFusco, PhD, CFA, Dennis W. McLeavey, DBA, CFA, Jerald
E. Pinto, PhD, CFA, and David E. Runkle, PhD, CFA.
Richard A. DeFusco, PhD, CFA, is at the University of Nebraska-Lincoln (USA). Dennis W.
McLeavey, DBA, CFA, is at the University of Rhode Island (USA). Jerald E. Pinto, PhD,
CFA, is at CFA Institute (USA). David E. Runkle, PhD, CFA, is at Jacobs Levy Equity
Management (USA).

LEARNING OUTCOMES
Mastery The candidate should be able to:

calculate expected values, variances, and standard deviations and

demonstrate their application to investment problems
formulate an investment problem as a probability tree and explain
the use of conditional expectations in investment application
calculate and interpret an updated probability in an investment
setting using Bayes’ formula

INTRODUCTION
Investment decisions are made under uncertainty about the future direction of the
1
economy, issuers, companies, and prices. This learning module presents probability
tools that address many real-world problems involving uncertainty and applies to a
variety of investment management applications.
Lesson 1 introduces the calculation of the expected value, variance, and standard
deviation for a random variable. These are essential quantitative concepts in invest-
ment management. Lesson 2 introduces probability trees that help in visualizing the
conditional expectations and the total probabilities for expected value.
When making investment decisions, analysts often rely on perspectives, which
may be influenced by subsequent observations. Lesson 3 introduces Bayes’ formula,
a rational method to adjust probabilities with the arrival of new information. This
method has wide business and investment applications.
 © CFA Institute. For candidate use only. Not for distribution.
134 Learning Module 4 Probability Trees and Conditional Expectations

LEARNING MODULE OVERVIEW

■ The expected value of a random variable is a probabili-

ty-weighted average of the possible outcomes of the random
variable. For a random variable X, the expected value of X is denoted
E(X).
■ The variance of a random variable is the expected value (the probabili-
ty-weighted average) of squared deviations from the random variable’s
expected value E(X): σ2(X) = E{[X − E(X)]2}, where σ2(X) stands for the
variance of X.
■ Standard deviation is the positive square root of variance. Standard
deviation measures dispersion (as does variance), but it is measured in
the same units as the variable.
■ A probability tree is a means of illustrating the results of two or more
independent events.
■ A probability of an event given (conditioned on) another event is a
conditional probability. The probability of an event A given an event B
is denoted P(A | B), and P(A | B) = P(AB)/P(B), P(B) ≠ 0.
■ According to the total probability rule, if S1, S2, …, Sn are mutually
exclusive and exhaustive scenarios or events, then P(A) = P(A | S1)
P(S1) + P(A | S2)P(S2) + … + P(A | Sn)P(Sn).
■ Conditional expected value is E(X | S) = P(X1 | S)X1 + P(X2 | S)X2 + …
+ P(Xn | S)Xn and has an associated conditional variance and condi-
tional standard deviation.
■ Bayes’ formula is a method used to update probabilities based on new
information.
■ Bayes’ formula is expressed as follows: Updated probability of event
given the new information = [(Probability of the new information
given event)/(Unconditional probability of the new information)] ×
Prior probability of event.

2 EXPECTED VALUE AND VARIANCE

calculate expected values, variances, and standard deviations and

demonstrate their application to investment problems

The expected value of a random variable is an essential quantitative concept in

investments. Investors continually make use of expected values—in estimating the
rewards of alternative investments, in forecasting earnings per share (EPS) and
other corporate financial variables and ratios, and in assessing any other factor that
may affect their financial position. The expected value of a random variable is the
probability-weighted average of the possible outcomes of the random variable. For a
random variable X, the expected value of X is denoted E(X).
Expected value (e.g., expected stock return) looks either to the future, as a fore-
cast, or to the “true” value of the mean (the population mean). We should distinguish
expected value from the concepts of historical or sample mean. The sample mean also
 © CFA Institute. For candidate use only. Not for distribution.
Expected Value and Variance 135

summarizes in a single number a central value. However, the sample mean presents
a central value for a particular set of observations as an equally weighted average of
those observations. In sum, the contrast is forecast versus historical, or population
versus sample.
An equation that summarizes the calculation of the expected value for a discrete
random variable X is as follows:
n
​E​(X)​  = P​(​X1​  ​​)​ ​X1​  ​​  + P​(​X2​  ​​)​ ​X2​  ​​ + … + P​(​Xn​  ​​)​ ​Xn​  ​​  = ​∑ ​​​  P​(​Xi​  ​​)​ ​Xi​  ​​​, (1)
i=1

where Xi is one of n possible outcomes of the discrete random variable X.

The expected value is our forecast. Because we are discussing random quantities,
we cannot count on an individual forecast being realized (although we hope that, on
average, forecasts will be accurate). It is important, as a result, to measure the risk
we face. Variance and standard deviation measure the dispersion of outcomes around
the expected value or forecast.
The variance of a random variable is the expected value (the probability-weighted
average) of squared deviations from the random variable’s expected value:
​​σ​​  2(​​ X)​  = E ​​[X − E​(X)]​ ​​​  2​​. (2)
The two notations for variance are and Var(X). σ2(X)
Variance is a number greater than or equal to 0 because it is the sum of squared
terms. If variance is 0, there is no dispersion or risk. The outcome is certain, and
the quantity X is not random at all. Variance greater than 0 indicates dispersion of
outcomes. Increasing variance indicates increasing dispersion, all else being equal.
The following equation summarizes the calculation of variance:
σ​ ​​  2​​(X)​  = P​(​X1​  ​​)​ ​​[​X1​  ​​ − E​(X)​]​​​  2​  + P​(​X2​  ​​)​ ​​[​X2​  ​​ − E​(X)​]​​​  2​
     
​​ n ​ ​​, (3)
  + … + P​(​Xn​  )​​ ​ ​​[​Xn​  ​​ − E​(X)​]​​​  2​ = ​ ∑ ​​​  P​(​Xi​  )​​ ​ ​​[​Xi​  ​​ − E​(X)​]​​​  2​
i= 1

where Xi is one of n possible outcomes of the discrete random variable X.

Variance of X is a quantity in the squared units of X. For example, if the random
variable is return in percent, variance of return is in units of percent squared. Standard
deviation is easier to interpret than variance because it is in the same units as the
random variable.Standard deviation is the square root of variance. If the random
variable is return in percent, standard deviation of return is also in units of percent. In
the following examples, when the variance of returns is stated as a percent or amount
of money, to conserve space, we may suppress showing the unit squared.
The best way to become familiar with these concepts is to work examples.

EXAMPLE 1

BankCorp’s Earnings per Share, Part 1

As part of your work as a banking industry analyst, you build models for forecast-
ing earnings per share of the banks you cover. Today you are studying BankCorp.
In Exhibit 1, you have recorded a probability distribution for BankCorp’s EPS
for the current fiscal year.
​

Exhibit 1: Probability Distribution for BankCorp’s EPS

​

Probability EPS (USD)

0.15 2.60
0.45 2.45
0.24 2.20
 © CFA Institute. For candidate use only. Not for distribution.
136 Learning Module 4 Probability Trees and Conditional Expectations

Probability EPS (USD)

0.16 2.00
1.00
​

1. What is the expected value of BankCorp’s EPS for the current fiscal year?
Solution:
Following the definition of expected value, list each outcome, weight it by its
probability, and sum the terms.
E(EPS) = 0.15(USD2.60) + 0.45(USD 2.45) + 0.24(USD 2.20) + 0.16(USD
2.00)
= USD2.3405
The expected value of EPS is USD2.34.

2. Using the probability distribution of EPS from Exhibit 1, you want to

measure the dispersion around your forecast. What are the variance and
standard deviation of BankCorp’s EPS for the current fiscal year?
Solution:
The order of calculation is always expected value, then variance, and then
standard deviation. Expected value has already been calculated. Following
the previous definition of variance, calculate the deviation of each outcome
from the mean or expected value, square each deviation, weight (multiply)
each squared deviation by its probability of occurrence, and then sum these
terms.

σ​ ​​  2(​​ EPS)​  = P​(2.60)​ ​​[2.60 − E​(EPS)​]​​​  2​  + P​(2.45)​ ​​[2.45 − E​(EPS)​]​​​  2​

+   P​(2.20)​ ​​[2.20 − E​(EPS)​]​​​  2​  + P​(2.00)​ ​​[2.00 − E​(EPS)​]​​​  2​
    
    
     
=      
​​ 0.15 ​​(2.60 − 2.34)​​​  2​ + 0.45 ​​(2.45 − 2.34)​​​  2​​ ​  ​​ ​
+   0.24 ​​(2.20 − 2.34)​​​  2​ + 0.16 ​​(2.00 − 2.34)​​​  2​
= 0.01014 + 0.005445 + 0.004704 + 0.018496 = 0.038785
Standard deviation is the positive square root of 0.038785:
σ(EPS) = 0.0387851/2 = 0.196939, or approximately 0.20.

3 PROBABILITY TREES AND CONDITIONAL

EXPECTATIONS

formulate an investment problem as a probability tree and explain

the use of conditional expectations in investment application

In investments, we make use of any relevant information available in making our

forecasts. When we refine our expectations or forecasts, we are typically updating
them based on new information or events; in these cases, we are using conditional
expected values. The expected value of a random variable X given an event or scenario
S is denoted E(X | S). Suppose the random variable X can take on any one of n distinct
 © CFA Institute. For candidate use only. Not for distribution.
Probability Trees and Conditional Expectations 137

outcomes X1, X2, …, Xn (these outcomes form a set of mutually exclusive and exhaus-
tive events). The expected value of X conditional on S is the first outcome, X1, times
the probability of the first outcome given S, P(X1 | S), plus the second outcome, X2,
times the probability of the second outcome given S, P(X2 | S), and so forth, as follows:
E(X | S) = P(X1 | S)X1 + P(X2 | S)X2 + … + P(Xn | S)Xn. (4)
We will illustrate this equation shortly.
Parallel to the total probability rule for stating unconditional probabilities in terms
of conditional probabilities, there is a principle for stating (unconditional) expected
values in terms of conditional expected values. This principle is the total probability
rule for expected value.

Total Probability Rule for Expected Value

The formula follows:
E(X) = E(X | S)P(S) + E(X | SC)P(SC), (5)
where SC is the “complement of S,” which means event or scenario “S” does not occur.
E(X) = E(X | S1)P(S1) + E(X | S2)P(S2) + … + E(X | Sn)P(Sn), (6)
where S1, S2, …, Sn are mutually exclusive and exhaustive scenarios or events.
The general case, Equation 6, states that the expected value of X equals the expected
value of X given Scenario 1, E(X | S1), times the probability of Scenario 1, P(S1), plus
the expected value of X given Scenario 2, E(X | S2), times the probability of Scenario
2, P(S2), and so forth.
To use this principle, we formulate mutually exclusive and exhaustive scenarios
that are useful for understanding the outcomes of the random variable. This approach
was employed in developing the probability distribution of BankCorp’s EPS in Example
1, as we now discuss.

EXAMPLE 2

BankCorp’s Earnings per Share, Part 2

The earnings of BankCorp are interest rate sensitive, benefiting from a declining
interest rate environment. Suppose there is a 0.60 probability that BankCorp
will operate in a declining interest rate environment in the current fiscal year
and a 0.40 probability that it will operate in a stable interest rate environment
(assessing the chance of an increasing interest rate environment as negligible).
If a declining interest rate environment occurs, the probability that EPS will be
USD2.60 is estimated at 0.25, and the probability that EPS will be USD2.45 is
estimated at 0.75. Note that 0.60, the probability of declining interest rate envi-
ronment, times 0.25, the probability of USD2.60 EPS given a declining interest
rate environment, equals 0.15, the (unconditional) probability of USD2.60 given
in the table in Exhibit 1. The probabilities are consistent. Also, 0.60(0.75) = 0.45,
the probability of USD2.45 EPS given in Exhibit 1. The probability tree diagram
in Exhibit 2 shows the rest of the analysis.
 © CFA Institute. For candidate use only. Not for distribution.
138 Learning Module 4 Probability Trees and Conditional Expectations

Exhibit 2: BankCorp’s Forecasted EPS

​

EPS = $2.60 with

0.25 Prob = 0.15
Prob. of declining
interest rates = 0.60 EPS = $2.45 with
0.75
Prob = 0.45
E(EPS) = $2.34
EPS = $2.20 with
0.60 Prob = 0.24
Prob. of stable
interest rates = 0.40
EPS = $2.00 with
0.40
Prob = 0.16

A declining interest rate environment points us to the node of the tree that
branches off into outcomes of USD2.60 and USD2.45. We can find expected
EPS given a declining interest rate environment as follows, using Equation 6:
E(EPS | declining interest rate environment)
= 0.25(USD2.60) + 0.75(USD2.45)

= USD2.4875
If interest rates are stable,
E(EPS | stable interest rate environment) = 0.60(USD2.20) + 0.40(USD2.00)

= USD2.12
Once we have the new piece of information that interest rates are stable, for
example, we revise our original expectation of EPS from USD2.34 downward to
USD2.12. Now using the total probability rule for expected value,
E​(EPS)​
​​ E​(EPS​ | declining interest rate environment)​P(declining interest rate environment)
=
         ​​ ​
  + E​(EPS | stable interest rate environment)​P(stable interest rate environment)
So, E(EPS) = USD2.4875(0.60) + USD2.12(0.40) = USD2.3405 or about
USD2.34.
This amount is identical to the estimate of the expected value of EPS calcu-
lated directly from the probability distribution in Example 1. Just as our proba-
bilities must be consistent, so too must our expected values, unconditional and
conditional, be consistent; otherwise, our investment actions may create profit
opportunities for other investors at our expense.

To review, we first developed the factors or scenarios that influence the outcome
of the event of interest. After assigning probabilities to these scenarios, we formed
expectations conditioned on the different scenarios. Then we worked backward to
formulate an expected value as of today. In the problem just worked, EPS was the
event of interest, and the interest rate environment was the factor influencing EPS.
We can also calculate the variance of EPS given each scenario:
σ2(EPS|declining interest rate environment)

= P(USD2.60|declining interest rate environment)

× [USD2.60 − E(EPS|declining interest rate environment)]2

+ P(USD2.45|declining interest rate environment)

 © CFA Institute. For candidate use only. Not for distribution.
Probability Trees and Conditional Expectations 139

× [USD2.45 − E(EPS|declining interest rate environment)]2

= 0.25(USD2.60 − USD2.4875)2 + 0.75(USD2.45 − USD2.4875)2 = 0.004219

Similarly, σ2(EPS | stable interest rate environment) is found to be equal to
= 0.60(USD2.20 − USD2.12)2 + 0.40(USD2.00 − USD2.12)2 = 0.0096
These are conditional variances, the variance of EPS given a declining interest rate
environment and the variance of EPS given a stable interest rate environment. The
relationship between unconditional variance and conditional variance is a relatively
advanced topic. The main points are that (1) variance, like expected value, has a con-
ditional counterpart to the unconditional concept; and (2) we can use conditional
variance to assess risk given a particular scenario.

EXAMPLE 3

BankCorp’s Earnings per Share, Part 3

Continuing with the BankCorp example, you focus now on BankCorp’s cost
structure. One model, a simple linear regression model, you are researching for
BankCorp’s operating costs is
​​ Yˆ​  = a + bX​,
where ​​ Yˆ​​is a forecast of operating costs in millions of US dollars and X is
the number of branch offices; and ​​ Yˆ​​represents the expected value of Y given X,
or E(Y | X). You interpret the intercept a as fixed costs and b as variable costs.
You estimate the equation as follows:
​​ Yˆ​  = 12.5 + 0.65X​.
BankCorp currently has 66 branch offices, and the equation estimates oper-
ating costs as 12.5 + 0.65(66) = USD55.4 million. You have two scenarios for
growth, pictured in the tree diagram in Exhibit 3.
​

Exhibit 3: BankCorp’s Forecasted Operating Costs

​

Branches = 125
0.50 Op. Costs = ?
Prob = ?
High Growth
Probability = 0.80 Branches = 100
0.50 Op. Costs = ?
Expected Op. Prob = ?
Costs = ? Branches = 80
0.85 Op. Costs = ?
Low Growth Prob = ?
Probability = 0.20
Branches = 70
0.15 Op. Costs = ?
Prob = ?

1. Compute the forecasted operating costs given the different levels of oper-
​​ ˆ​  = 12.5 + 0.65X​. State the probability of each level of
ating costs, using Y
the number of branch offices. These are the answers to the questions in the
terminal boxes of the tree diagram.
Solution:
​​  ˆ​​= 12.5 + 0.65X, from top to bottom, we have
Using Y
 © CFA Institute. For candidate use only. Not for distribution.
140 Learning Module 4 Probability Trees and Conditional Expectations

Operating Costs Probability

​​ Yˆ​​ = 12.5 + 0.65(125) = USD93.75 million 0.80(0.50) = 0.40

​​ Yˆ​​ = 12.5 + 0.65(100) = USD77.50 million 0.80(0.50) = 0.40
​​ Yˆ​​ = 12.5 + 0.65(80) = USD64.50 million 0.20(0.85) = 0.17
​​ Yˆ​​ = 12.5 + 0.65(70) = USD58.00 million 0.20(0.15) = 0.03
Sum = 1.00
​

2. Compute the expected value of operating costs under the high growth
scenario. Also calculate the expected value of operating costs under the low
growth scenario.
Solution:
US dollar amounts are in millions.
E(operating costs|high growth) = 0.50(USD93.75) + 0.50(USD77.50)

= USD85.625

E(operating costs|low growth) = 0.85(USD64.50) + 0.15(USD58.00)

= USD63.525

3. Refer to the question in the initial box of the tree: What are BankCorp’s
expected operating costs?
Solution:
US dollar amounts are in millions.
E(operating costs) = E(operating costs|high growth)P(high growth)

+ E(operatingcosts|low growth)P(low growth)

= 85.625(0.80) + 63.525(0.20) = 81.205

BankCorp’s expected operating costs are USD81.205 million.

In this section, we have treated random variables, such as EPS, as standalone

quantities. We have not explored how descriptors, such as expected value and variance
of EPS, may be functions of other random variables. Portfolio return is one random
variable that is clearly a function of other random variables, the random returns on
the individual securities in the portfolio. To analyze a portfolio’s expected return and
variance of return, we must understand that these quantities are a function of char-
acteristics of the individual securities’ returns. Looking at the variance of portfolio
return, we see that the way individual security returns move together or covary is key.
We cover portfolio expected return, variance of return, and importantly, covariance
and correlation in a separate learning module.

QUESTION SET

1. Suppose the prospects for recovering principal for a defaulted bond

issue depend on which of two economic scenarios prevails. Scenario 1 has
probability 0.75 and will result in recovery of USD0.90 per USD1 principal
value with probability 0.45, or in recovery of USD0.80 per USD1 principal
 © CFA Institute. For candidate use only. Not for distribution.
Bayes' Formula and Updating Probability Estimates 141

value with probability 0.55. Scenario 2 has probability 0.25 and will result in
recovery of USD0.50 per USD1 principal value with probability 0.85, or in
recovery of USD0.40 per USD1 principal value with probability 0.15.

Using the data for Scenario 1 and Scenario 2, calculate the following:

A. Compute the expected recovery, given the first scenario.

B. Compute the expected recovery, given the second scenario.
C. Compute the expected recovery.
D. Graph the information in a probability tree diagram.
E. Compute the probability of each of the four possible recovery
amounts: USD0.90, USD0.80, USD0.50, and USD0.40.
Solution:

A. Outcomes associated with Scenario 1: With a 0.45 probability of a

USD0.90 recovery per USD1 principal value, given Scenario 1, and
with the probability of Scenario 1 equal to 0.75, the probability of
recovering USD0.90 is 0.45 (0.75) = 0.3375. By a similar calculation,
the probability of recovering USD0.80 is 0.55(0.75) = 0.4125.
Outcomes associated with Scenario 2: With a 0.85 probability of a
USD0.50 recovery per USD1 principal value, given Scenario 2, and
with the probability of Scenario 2 equal to 0.25, the probability of
recovering USD0.50 is 0.85(0.25) = 0.2125. By a similar calculation, the
probability of recovering USD0.40 is 0.15(0.25) = 0.0375.
B. E(recovery | Scenario 1) = 0.45(USD0.90) + 0.55(USD0.80) = USD0.845
C. E(recovery | Scenario 2) = 0.85(USD0.50) + 0.15(USD0.40) = USD0.485
D. E(recovery) = 0.75(USD0.845) + 0.25(USD0.485) = USD0.755
Recovery = $0.90
0.45 Prob = 0.3375
Scenario 1,
Probability = 0.75 Recovery = $0.80
0.55
Prob = 0.4125
Expected
Recovery = $0.755
Recovery = $0.50
0.85 Prob = 0.2125
Scenario 2,
Probability = 0.25
Recovery = $0.40
0.15
Prob = 0.0375

BAYES' FORMULA AND UPDATING PROBABILITY

ESTIMATES 4
calculate and interpret an updated probability in an investment
setting using Bayes’ formula

A topic that is often useful in solving investment problems is Bayes’ formula: what
probability theory has to say about learning from experience.
 © CFA Institute. For candidate use only. Not for distribution.
142 Learning Module 4 Probability Trees and Conditional Expectations

Bayes’ Formula
When we make decisions involving investments, we often start with viewpoints based
on our experience and knowledge. These viewpoints may be changed or confirmed by
new knowledge and observations. Bayes’ formula is a rational method for adjusting
our viewpoints as we confront new information. Bayes’ formula and related concepts
are used in many business and investment decision-making contexts.
Bayes’ formula makes use of the total probability rule:
​P(​ A)​  = ​∑ ​ ​​  P​(A ∩ ​Bn​  ​​)​​. (7)
n

To review, that rule expresses the probability of an event as a weighted average of the
probabilities of the event, given a set of scenarios. Bayes’ formula works in reverse;
more precisely, it reverses the “given that” information. Bayes’ formula uses the
occurrence of the event to infer the probability of the scenario generating it. For that
reason, Bayes’ formula is sometimes called an inverse probability. In many applications,
including those illustrating its use in this section, an individual is updating his/her
beliefs concerning the causes that may have produced a new observation.
Bayes’ Formula. Given a set of prior probabilities for an event of interest, if you
receive new information, the rule for updating your probability of the event is as follows:
Updated probability of event given the new information

Probability of the new information given event

___________________________________
​​ =​      
​ Unconditional
     probability of the new information ​ ×​  Prior probability of event​.​

In probability notation, this formula can be written concisely as follows:

P​(Information | Event)​
________________ (
​P​(Event | Information)​  = ​        ​
P​(Information)​
P​ Event)​​. (8)

Consider the following example using frequencies—which may be more straightfor-

ward initially than probabilities—for illustrating and understanding Bayes’ formula.
Assume a hypothetical large-cap stock index has 500 member firms, of which 100 are
technology firms, and 60 of these had returns of >10 percent, and 40 had returns of
≤10 percent. Of the 400 non-technology firms in the index, 100 had returns of >10
percent, and 300 had returns of ≤10 percent. The tree map in Exhibit 4 is useful for
visualizing this example, which is summarized in the table in Exhibit 5.
 © CFA Institute. For candidate use only. Not for distribution.
Bayes' Formula and Updating Probability Estimates 143

Exhibit 4: Tree Map for Visualizing Bayes’ Formula Using Frequencies

All Firms

500

Tech Firms Non-Tech Firms

100 400

Return > 10 % Return ≤ 10%

60 40 100 300

Return ≤ 10% Return > 10 %

P (Tech | R > 10%) = 60/(60+100)

Exhibit 5: Summary of Returns for Tech and Non-Tech Firms in Hypothetical

Large-Cap Equity Index

Type of Firm in Stock Index

Rate of Return (R) Non-Tech Tech Total

R >10% 100 60 160

R ≤10% 300 40 340
Total 400 100 500

What is the probability a firm is a tech firm given that it has a return of >10 percent
or P(tech | R > 10%)? Looking at the frequencies in the tree map and in the table, we
can see many empirical probabilities, such as the following:
■ P(tech) = 100 / 500 = 0.20,
■ P(non-tech) = 400 / 500 = 0.80,
■ P(R > 10% | tech) = 60 / 100 = 0.60,
■ P(R > 10% | non-tech) = 100 / 400 = 0.25,
■ P(R > 10%) = 160 / 500 = 0.32, and, finally,
■ P(tech | R > 10%) = 60/ 160 = 0.375.
This probability is the answer to our initial question.
Without looking at frequencies, let us use Bayes’ formula to find the probability
that a firm has a return of >10 percent and then the probability that a firm with a
return of >10 percent is a tech firm, P(tech | R > 10%). First,
P(R > 10%)
= P(R > 10% | tech) × P(tech) + P(R > 10% | non-tech) × P(non-tech)
 © CFA Institute. For candidate use only. Not for distribution.
144 Learning Module 4 Probability Trees and Conditional Expectations

= 0.60 × 0.20 + 0.25 × 0.80 = 0.32.

Now we can implement the Bayes’ formula answer to our question:
P​(R > 10 % |tech)​ × P​(tech)​
____________________ 0.60 × 0.20
​P​(tech|R > 10%)​  = ​        ​  = ​ _
P​(R > 10%)​ 0.32 ​  = 0.375​.

The probability that a firm with a return of >10 percent is a tech firm is 0.375, which is
impressive because the probability that a firm is a tech firm (from the whole sample)
is only 0.20. In sum, it can be readily seen from the tree map and the underlying fre-
quency data (Exhibit 4 and 5, respectively) or from the probabilities in Bayes’ formula
that 160 firms have R >10 percent, and 60 of them are tech firms, so
P(tech | R > 10%) = 60/160 = 0.375.
Users of Bayesian statistics do not consider probabilities (or likelihoods) to be known
with certainty but believe that these should be subject to modification whenever new
information becomes available. Our beliefs or probabilities are continually updated
as new information arrives over time.
To further illustrate Bayes’ formula, we work through an investment example that
can be adapted to any actual problem. Suppose you are an investor in the stock of
DriveMed, Inc. Positive earnings surprises relative to consensus EPS estimates often
result in positive stock returns, and negative surprises often have the opposite effect.
DriveMed is preparing to release last quarter’s EPS result, and you are interested in
which of these three events happened: last quarter’s EPS exceeded the consensus EPS
estimate, last quarter’s EPS exactly met the consensus EPS estimate, or last quarter’s
EPS fell short of the consensus EPS estimate. This list of the alternatives is mutually
exclusive and exhaustive.
On the basis of your own research, you write down the following prior probabil-
ities (or priors, for short) concerning these three events:
■ P(EPS exceeded consensus) = 0.45
■ P(EPS met consensus) = 0.30
■ P(EPS fell short of consensus) = 0.25
These probabilities are “prior” in the sense that they reflect only what you know
now, before the arrival of any new information.
The next day, DriveMed announces that it is expanding factory capacity in Singapore
and Ireland to meet increased sales demand. You assess this new information. The
decision to expand capacity relates not only to current demand but probably also to
the prior quarter’s sales demand. You know that sales demand is positively related to
EPS. So now it appears more likely that last quarter’s EPS will exceed the consensus.
The question you have is, “In light of the new information, what is the updated
probability that the prior quarter’s EPS exceeded the consensus estimate?”
Bayes’ formula provides a rational method for accomplishing this updating. We
can abbreviate the new information as DriveMed expands. The first step in applying
Bayes’ formula is to calculate the probability of the new information (here: DriveMed
expands), given a list of events or scenarios that may have generated it. The list of
events should cover all possibilities, as it does here. Formulating these conditional
probabilities is the key step in the updating process. Suppose your view, based on
research of DriveMed and its industry, is
P(DriveMed expands | EPS exceeded consensus) = 0.75

P(DriveMed expands | EPS met consensus) = 0.20

P(DriveMed expands | EPS fell short of consensus) = 0.05

Conditional probabilities of an observation (here: DriveMed expands) are sometimes
referred to as likelihoods. Again, likelihoods are required for updating the probability.