Lecture 1:

Investments 8801
Prof. David Solomon Quantitative Review

• Quantitative Review:
– Returns on assets and portfolios
– Stats: means, variances, covariances, distributions
– Regressions

Readings: BKM chapter 5 (5.4,5.5 and 5.6)

 How Do Returns Behave?
• Returns are random

– Like the toss of a coin, dice roll, … A random variable

can take several possible values. It is characterized by
its probability distribution, a list of all possible
outcomes and the probability that each will occur

– A simple example: ret Pr( ret = x)

-5% 0.10 “bear” market
+5% 0.20
15% 0.40 normal market
25% 0.20
35% 0.10 “bull” market
2
 (1) Means
• The mean or expected value of a random variable is
the probability-weighted average outcome.
N
E [ X ]   X   xi p( xi )
i 1

p(xi) = probability of state xi

xi = return if state i (of possible 1 … N) occurs

• Ex: r Prob(r) E(r) = 0.1*-0.05 +

-5% 0.10 0.2*0.05 + 0.4*0.15 +
+5% 0.20 0.2*0.25 + 0.1*0.35
15% 0.40
= 0.15
25% 0.20
35% 0.10 3
 (1) Means
• Properties of means E[aX  b]  aE[ X ]  b
E[ X  Y ]  E[ X ]  E[Y ]

use this to calculate portfolio’s expected return !

E(RP) = wA*E(RA) + wB*E(RB)

• Ex: portfolio (wbonds=30%, wstock=50%)

E(rB) = 0.5% E(rS)= 0.8% rf = 0.3%

E [rportfolio] = wB*E(RB) + wS*E(RS) + wF*E(RF)

= 0.3*0.5 + 0.5*0.8 + 0.2*0.3 = 0.15 + 0.4 + 0.06 = 0.61

Can do similar calculation with excess returns 4

 (2) Variances
• Variance is a measure of dispersion. Average size of the
(squared) deviations of r from it’s mean. The squaring
turns all of these deviations into positive numbers.

r2 = Var(r ) = E[(r – E(r))2]

• Standard deviation (SD, σ) is the square root of the

variance. Sometime known as volatility when expressed
in annualized terms.
• better scaling / more intuitive
• good rules of thumb using the normal distribution

5
 (2) Variances
• If we have scenarios with different probabilities,

• Ex: rs Prob(rs) Rs – E(r) [Rs-E(r)]2

-5% 0.10
+5% 0.20
15% 0.40
25% 0.20
35% 0.10

E(r)=0.15 Var =[(.1)(-.05-.15)2+(.2)(.05- .15)2...+

+ .1(.35-.15)2] =0.01199
S.D.= [ .01199] 1/2 = 0.1095 6
 Estimating Means and Variances
• Example using real World data
Use EXCEL
-> data in spreadsheet
BKM_Table5_3_rates_of_return.xls

7
 Estimating Means and Variances
• What’s the mean & stdev of (Large-cap) Stocks?
60%

Variance
40%
easier to
estimate
20%
than mean
0%
26
29
32
35
38
41
44
47
50
53
56
59
62
65
68
71
74
77
80
83
86
89
92
95
98
01
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
20
-20%

-40%
US_LargeCap_Stocks
US_TBonds
US_TBills
-60%

8
 Estimating Means and Variances
N
1
• Sample Mean: ri 
N
r
t 1
i ,t

-> AVERAGE() in Excel

 
N
1

2
• Sample Variance: si2  ri ,t  ri
N  1 t 1
-> VAR() in Excel

• Sample Standard Deviation: si  s 

2 1/ 2
i  si2

-> STDEV() in Excel

9
 Estimating Means and Variances
• Statistics on Annual Holding Period Returns
Series Mean% St.Dev.%
World Stk 11.17 18.38
US Lg Stk 12.25 20.50
US Sm Stk 18.43 38.11
World Bonds 6.13 9.14
LT Treas 5.64 8.19
T-Bills 3.79 3.18
Inflation 3.12 4.35

• Note: we would like to know the true parameters E(r) and

12. But we can only measure sample Avg(r) and s12.
10
 Estimating Means and Variances
• Adding or removing a year makes a difference …

11
 Variance with multiple assets
Special Cases: When one asset is risk-free
•Question 1: Portfolio of 1 riskfree and 1 risky security. What is
Var(rf + r)?
Variance Property: Var(a + r1 ) = 12
Var(Rf+r)=Var (r)

•Question 2: Suppose we have a portfolio that is 60% in an S&P

500 index fund and 40% in one-month T-bills. The monthly
variance of the index fund is .042 = .0016. What is the one-
month variance of the portfolio?
Variance Property: Var(a + br1 ) = b2 12
Var(Rp) = 0.62 * 0.0016 = 0.000576
Std(Rp) = 0.024
12
How Do 2 Asset Returns Move Together?

• Covariance measures how much two variables move

together. It is the average of the products of their
deviations from their means.
Cov(r1 , r2) = 12 = E[(r1 – E(r1))(r2 – E(r2))]

• Correlation is scaled so that it is less than 100%

(positive or negative).
 ij
 ij  ,  1   ij  1
 i j

13
How Do 2 Asset Returns Move Together?

14
How Do 2 Asset Returns Move Together?
• Example (cont.)
80%

60%

40%
World_Stocks

20%

0%
-60% -40% -20% 0% 20% 40% 60%
-20%

-40%

-60%
US_LargeCap_Stocks

World_Stocks US_LargeCap_Stocks US_SmallCap_Stocks World_Bonds US_TBonds US_TBills

World_Stocks 1.000 0.860 0.728 0.361 0.067 -0.010
US_LargeCap_Stocks 1.000 0.779 0.188 0.127 -0.015
US_SmallCap_Stocks 1.000 0.116 -0.037 -0.151
World_Bonds 1.000 0.650 0.219
US_TBonds 1.000 0.254
US_TBills 1.000

15
 Variance with multiple assets
• General Case: When both assets are risky

• Var(r1 + r2) = Var(r1) + Var(r2) + 2 Cov(r1, r2)

= Var(r1) + Var(r2) + 2 SD(r1) SD(r2) Corr(r1, r2)

If w1 and w2 are weights in stocks 1 and 2, then:

• Var(w1r1 + w2r2) = w12 Var(r1) + w22 Var(r2)

+ 2 w1 w2 Cov(r1, r2)

= w12 Var(r1) + w22 Var(r2) + 2 w1 w2 SD(r1) SD(r2) Corr(r1, r2)

N Assets:

16
 Adding Random Returns?
• Example: wbonds=30% , wstocks=70%
Statistics: . E(rb) = 6% stdev(rb)= 8%
. E(rs) = 12% stdev(rs)= 20%
. Correl(rstocks,rbonds)=+ 0.1
. rf = 3% (money market account )

• σ [rportfolio] = √ [wB2 * σ2B + wS2 * σ2S

+ 2 σB,S *wb *ws ]
= √[ 0.32 *0.082 + 0.72*0.22 + 2*0.3*0.7*0.1*0.08*0.2]
= √(0.000576 + 0.0196 + 0.000672)
= √(0.020848)
=0.144388
17
 Probability Distributions
The Normal Distribution (the “bell shaped curve”):

s.d. s.d.
(2) (2)

mean(1)
A normal distribution is completely characterized by its
mean (1): and standard deviation (2):.
(3): Skewness = Are outcomes below the mean more likely
than those above the mean? Normal Distribution is Symmetric
18
 Probability Distributions

There is about a 95% probability that a

normally distributed random number will be
within two standard deviations of its mean.
68% within one standard deviation

NORMDIST() in Excel
19
 Probability Distributions
• Non-Normality:
- Normal distribution allows returns of less than
negative 100% , stock returns are better thought of
as being log-normally distributed (ie, the log of the
return is normally distributed)
0.7
0.6
Probability

0.5
0.4
0.3
0.2
0.1
0
-2.5

-2.1

-1.7

-1.3

-0.9

-0.5

-0.1

0.3

0.7

1.1

1.5

1.9

2.3
20
• Real World Frequency Distributions:

-> Annual returns [BKM Table 5.3. returns]

-> Monthly & Daily distributions [for S&P 500 Index]

21
• Correlation measures the association between two random
variables. Both variables are treated equally.

-> Association

• Regression is another measurement. It asks what we

expect a variable Y to be given another variable X.

-> Prediction

-> Regression can also be used when there are

many factors (X’s) that seem to impact a particular Y
variable.

22
 Regression
• Univariate Regression:
Yt =  +  Xt + t
– Yt is the dependent or “left-hand-side” variable
  is intercept,  is the regression slope
– Xt is the explanatory, independent or “right-hand-
side” variable
 t is the regression error, shock, or residual
• Like means and variances, we do not know the true
values of  or . We may estimate them by running a
regression (e.g. in Excel) or the formulas below:

ˆ ˆ Y
  XY ˆ  Y  ˆX
ˆ X 23
 Yt =  +  Xt + t
Y

εt = Error term

β = Slope

α = Intercept
X
24
 Multivariate Regression
• Multivariate Regression: There can be more than one
explanatory variables,
rt =  + 1 F1,t + 2 F2,t + 3 F3,t + t
or
Yt+1 =  +  Xt +  Zt + t

• Now there are multiple slope coefficients.

• There is no obvious graphical representation (unless
you’re good at visualizing vector space)
• Regression estimates must be calculated using a
statistical software package or in Excel using the
Analysis Toolbox. Install it!
25
 Regression
• Example: You are writing the international section of the
investment newsletter for your investment house. The
chief strategist is predicting rSP500=10% for next year.
What would you predict World stocks to do the coming
year?

-> use historical data in spreadsheet

BKM_Table5_3_rates_of_return.xls

26
 Regression

… -> what is̂ andˆ ?

27
 Regression
• Excel regression output:
r WORLD_STOCKS,t =+r US_LCAP_STOCKS,t + t

28
 Regression
• Predicted Y:
r WORLD_STOCKS,t = 0.0172+ 0.771r US_LCAP_STOCKS,t

• So if rSP500=10%, then prediction

rWORLD= 0.0172 + 0.771*0.1 = 0.0943

^ 29
 Regression
• How reliable the estimates of intercept () and slope ()?

30
 Testing Hypotheses
• Use T-test: T-stat of the hypothesis that the parameter is
equal to A is computed as
parameter - A
t  stat 
St.Error (parameter)

• A t-statistic greater than 1.96 or less than –1.96

(approximately – depends on number of observations)
indicates that the null hypothesis that E[r]=0 can be rejected
at the 95% significance level.

• Need t-stat> 2.58 to reject at a 99% significance level.

31
 Testing Hypotheses
• Example (cont.): T-tests

• 95% Confidence Interval = Mean +/- 1.96*Std Err.

32
 Testing Hypotheses
• Example (cont.): Is intercept different from 0?

t-stat = [Mean – 0] / std err = [0.017196]/ 0.01247= 1.38

Since t-stat is less than 1.96 and greater than -1.96, can’t
reject null hypothesis that intercept=0

33
 Testing Hypotheses
• Example (cont.): Is regression coefficient different from 1?

t-stat = [Mean – 1] / std err = [0.7710 – 1 ]/ 0.05247 = -4.36

Since this is less than -1.96, can reject the null hypothesis
that coefficient equals 1.

34
 Testing Hypotheses
T-test of a mean:

T
1
• The sample mean: r   rt
T t 1
• The sample standard error is equal to the standard
deviation of rt divided by the square root of the sample
size: ˆ r
SE (r ) 
T

• T-stat = [ Mean – A ] / SE(r)

35