Académique Documents
Professionnel Documents
Culture Documents
Irfanullah.co
Contents
1.
Introduction ................................................................................................................................... 2
2.
3.
4.
5.
6.
7.
8.
9.
This document should be read in conjunction with the corresponding reading in the 2014 Level I
CFA Program curriculum.
Some of the graphs, charts, tables, examples, and figures are copyright 2013, CFA Institute.
Reproduced and republished with permission from CFA Institute. All rights reserved.
Required disclaimer: CFA Institute does not endorse, promote, or warrant the accuracy or
quality of the products or services offered by Irfanullah Financial Training. CFA Institute,
CFA, and Chartered Financial Analyst are trademarks owned by CFA Institute.
Page 1
Irfanullah.co
1. Introduction
Statistical methods provide a powerful set of tools for analyzing data and drawing conclusions
from them. These are particularly useful when we are analyzing asset returns, earning growth
rates, commodity prices, or any other financial data. Descriptive statistics is the branch of
statistics that deals with describing and analyzing data. In this reading, we will study statistical
methods that allow us to summarize return distributions.
Specifically, we will explore four properties of return distributions:
Descriptive statistics: Study of how data can be summarized effectively to describe the
important aspects of large data sets.
Statistical inference: Making forecasts, estimates, or judgments about a larger group from
the smaller group actually observed.
Page 2
Irfanullah.co
Nominal scales: These scales categorize data but do not rank them. Hence, they are often
considered the weakest level of measurement. An example could be if we assigned
integers to mutual funds that follow different investment strategies. Number 1 might refer
to a small-cap value fund, number 2 might refer to a large-cap value fund, and so on for
each possible style.
Ordinal scales: These scales sort data into categories that are ordered with respect to
some characteristic. An example is Standard & Poors star ratings for mutual funds. One
star represents the group of mutual funds with the worst performance. Similarly, groups
with two, three, four and five stars represent groups with increasingly better performance.
Interval scales: These scales not only rank data, but also ensure that the differences
between scale values are equal. The Celsius and Fahrenheit scales are examples of such
scales. The difference in temperature between 10 oC and 11oC is the same amount as the
difference between 40 oC and 41oC. The zero point of an interval scale does not reflect
complete absence of what is being measured. Hence, it is not a true zero point or natural
zero.
Ratio scales: These scales have all the characteristics of interval scales as well as a true
zero point as the origin. This is the strongest level of measurement. The rate of return on
an investment is measured on a ratio scale. A return of 0% means the absence of any
return.
Page 3
Irfanullah.co
Page 4
Irfanullah.co
Calculate the range of the data, defined as Range = Maximum value Minimum value.
Determine the intervals by successively adding the interval width to the minimum value
to determine the ending points of intervals. Stop after reaching an interval that includes
the maximum value.
Construct a table of the intervals listed from smallest to largest that shows the number of
observations falling in each interval.
(Absolute)
Frequency
25
Cumulative
Frequency
25
Relative
Frequency
0.25
Cumulative Relative
Frequency
0.25
51-55
35
60
0.35
0.60
56-60
29
89
0.29
0.89
61-65
11
100
0.11
1.00
In order to summarize this data, we have divided the stock prices into 4 intervals of stock price
each having a width of 5. The actual number of observations in a given interval is called the
absolute frequency, or simply the frequency. For example, there are 25 stocks falling in the
interval of price range from 46-50. The relative frequency is the absolute frequency of each
interval divided by the total number of observations. The cumulative relative frequency
cumulates the relative frequencies as we move from the first to the last interval. It tells us the
fraction of observations that are less than the upper limit of each interval. So there are 60
observations less than the stock price of 55. The frequency distribution gives us a sense of where
most of the observations lie and also whether the distribution is evenly distributed, lopsided, or
peaked.
Page 5
Irfanullah.co
Histogram
40
35
Frequency
30
25
20
Frequency
15
10
5
0
46-50
51-55
56-60
61-65
Stock Price
The height of each bar in the histogram represents the absolute frequency for each interval.
4.2 The Frequency Polygon and the Cumulative Frequency Distribution
The frequency polygon is constructed when we plot the midpoint of each interval on the x-axis
and the absolute frequency for that interval on the y-axis. We then connect the neighboring
points with a straight line. The figure below is an example of a frequency polygon.
Page 6
Irfanullah.co
Frequency Polygon
40
35
30
25
20
Frequency
15
10
5
0
46-50
51-55
56-60
61-65
Another graphical tool is the cumulative frequency distribution. Such a graph can plot either the
cumulative absolute or cumulative relative frequency against the upper interval limit. The
cumulative frequency distribution allows us to see how many or what percent of the observations
lie below a certain value. The figure below is an example of a cumulative frequency distribution.
Cumulative Frequency
120
100
80
60
Cumulative Frequency
40
20
0
46-50
51-55
56-60
61-65
Page 7
Irfanullah.co
=1
where N is the number of observations in the entire population and Xi is the ith observation.
2 + 5 + 4 + 7 + 8 + 8 + 12 + 10 + 8 + 5
10
= 6.9%
=1
where n is the number of observations in the sample. If the sample data is: 8, 12, 10, 8 and 5, the
sample mean can be calculated as:
Copyright Irfanullah Financial Training. All rights reserved.
Page 8
Irfanullah.co
8 + 12 + 10 + 8 + 5
5
= 8.6
Uses all information about the size and magnitude of the observations
One of the drawbacks of the arithmetic mean is its sensitivity to extreme values. Because all
observations are used to compute the mean, the arithmetic mean can be pulled sharply upward or
downward by extremely large or small observations, respectively. Unusually large or small
observations are called outliers.
Sorting the sample data given above we have: 5%, 8%, 8% 10%, 12%. Here n = 5. The position
or location of the median number is given by (n + 1)/2 = 3 and the median value is 8%.
A distribution has only one median. An advantage of the median is that, unlike the mean,
extreme values do not affect it. The median, however, does not use all the information about the
size and magnitude of the observations. It focuses only on the relative position of the ranked
observations. Another disadvantage is that it is more tedious to compute as compared to the
mean.
Page 9
Irfanullah.co
=
=1
Page 10
Irfanullah.co
As an example, the geometric mean of 7, 8 and 9 will be: (7 x 8 x 9) 1/3 = 7.93. The most
common application of the geometric mean is to calculate the average return of an investment.
The formula is:
=[(1+R1)(1+R2).(1+Rn)]1/n 1
We will illustrate the use of this formula through a simple scenario: For a given stock the return
over the last four periods is: 10%, 8%, -5% and 2%. The geometric mean is calculated as:
[(1 + 0.10)(1 + 0.08)(1 0.05)(1 + 0.02)]1/4 1 = 0.0358 = 3.58%
Given the returns shown above, $1.00 invested at the start of period 1 grew to: $1.00 x 1.10 x
1.08 x 0.95 x 1.02 = 1.151. If the investment had grown at 3.58% every period, $1.00 invested at
the start of period 1 would have increased to: $1.00 x 1.0358 x 1.0358 x 1.0358 x 1.0358 =
1.151. As expected, both scenarios give the same answer. 3.58% is simply the average growth
rate per period.
The geometric mean is always less than or equal to the arithmetic mean. The only time that the
two means will be equal is when there is no variability in the observations i.e. when all the
observations are the same.
Note: In the reading on Discounted Cash Flow Applications we used the geometric mean to
calculate the time-weighted rate of return.
1
= / ( )
=1
Page 11
Irfanullah.co
This concept can be applied when we invest the same amount every month in a particular stock
and want to calculate the average purchase price. Suppose an investor purchases $1,000 of a
security each month for three months. The share prices are $10, $15 and $20 at the three
purchase dates. The average purchase price is simply the harmonic mean of 10, 15 and 20. The
harmonic mean is: 3 / (1/10 + 1/15 + 1/20) = 13.85.
The harmonic mean is generally less than the geometric mean, which is in turn less than the
arithmetic mean. To illustrate this fact take three numbers: 10, 15, and 20. It has just been shown
that the harmonic mean is 13.85. The geometric mean is (10 x 15 x 20) 1/3 = 14.42. The
arithmetic mean is simply 15. If all the observations in a dataset are the same then the three
means are the same.
Ly = (n+1) y /100
Where y is the percentage point at which we are dividing the distribution, n is the number of
observations and Ly is the location (L) of the percentile (Py) in an array sorted in ascending order.
Some important points to remember are:
Page 12
Irfanullah.co
When Ly is not a whole number or integer, Ly lies between the two closest integer
numbers (one above and one below) and we use linear interpolation between those two
places to determine Py.
Given below is the return data on 20 mutual funds arranged in ascending order.
Number
10
Return in %
1.25
1.70
1.75
1.85
1.98
1.99
2.05
2.40
2.49
2.60
Number
11
12
13
14
15
16
17
18
19
20
Return in %
2.90
3.00
3.24
3.75
3.90
1.99
2.05
2.40
2.49
2.60
At a given percentile, y = 10%, with n = 20 and the data sorted in ascending order, the location
of the observation is given by:
With a small data set, such as this one, the location calculated using the above formula is
approximate. As the data set becomes larger, the formula gives a more precise location.
Page 13
Irfanullah.co
7. Measures of Dispersion
Dispersion is the variability around the central tendency. Absolute dispersion is the amount of
variability present without comparison to any reference point or benchmark. Range, mean
absolute deviation, variance, and standard deviation are all examples of absolute dispersion.
= [| |] /
=1
where X is the sample mean and n is the number of observations in the sample.
X = (8+12+10+8+5)/5 = 8.6
=
Page 14
Irfanullah.co
Population variance is the arithmetic average of the squared deviations around the mean.
= ( ) 2 /
=0
For the data set: 2%, 5%, 4%, 7%, 8%, 8%, 12%, 10%, 8%, and 5%, the variance is given by:
2 =
= 7.89
Because variance is measured in squared units, we need a way to return to the original units. We
can solve this problem by using standard deviation, the square root of the variance. The
population standard deviation is defined as the positive square root of the population variance.
For the data given above, = 7.89 = 2.81%.
Both the population variance and standard deviation are examples of parameters of a distribution.
We often do not know the mean of a population of interest. We then estimate the population
mean with a mean from a sample drawn from the population. Next, we calculate the sample
variance and standard deviation.
= ( ) 2 / ( 1)
=0
Page 15
Irfanullah.co
where is the sample mean and n is the number of observations in the sample.
The steps to calculate a sample variance are:
( ) 2 / ( 1)
=0
By using n - 1 (rather than n) as the divisor, we improve the statistical properties of the sample
variance.
Consider the following data set: 8, 12, 10, 8 and 5. The sample variance is calculated as follows:
2 =
2 = 6.80%
The sample standard deviation is the positive square root of the sample variance. For the sample
data given above, = 6.80 = 2.61%.
The population and sample standard deviation can easily be computed using a financial
calculator. Assume the following data set: 10%, -5%, 10%, 25%, the calculator key strokes are
show below:
Keystrokes
Explanation
[2nd] [DATA]
Display
X01
10 [ENTER]
X01 = 10
[] [] 5+/- [ENTER]
X02 = -5
[] [] 10 [ENTER]
X03 = 10
[] [] 25 [ENTER]
X04 = 25
Page 16
Irfanullah.co
Keystrokes
Explanation
Display
[2nd] [SET]
1-V
[]
N=4
[]
Mean
X = 10
[]
Sx = 12.25
[]
x = 10.61
Notice that the calculator gives both the sample and the population standard deviation. On the
exam we will have to determine whether we are dealing with population or sample data.
Variance and standard deviation of returns take account of returns above and below the mean,
but investors are concerned only with downside risk, for example returns below the mean. As a
result, analysts have developed semivariance, semideviation and related dispersion measures that
focus on downside risk. Semivariance is defined as the average squared deviation below the
mean. Semideviation is the positive square root of semivariance. When return distributions are
symmetric, semivariance and variance are effectively equivalent. For asymmetric distributions,
variance and semivariance rank prospects risk differently.
7.6 Chebyshevs Inequality
According to Chebyshevs inequality, the proportion of the observations within k standard
deviations of the arithmetic mean is at least 1 - 1/k2 for all k > 1. To find out what percent of the
observations must be within 2 standard deviations of the mean we simply plug into the formula
and get: 1 1/22 = 1 = 0.75 = 75%. Hence at least 75% of the data will be between 2
Copyright Irfanullah Financial Training. All rights reserved.
Page 17
Irfanullah.co
standard deviations of the mean. To understand this concept, consider a distribution with a mean
value of 10 and a standard deviation of 3. For this distribution at least 75% of the data will be
between 4 and 16. Note that 4 is two standard deviations (2 x 3) less than the mean (10) and 16
is two standard deviations greater than the mean.
Plugging in a value of k = 3 in Chebyshevs inequality shows us that at least 89% of the
population data will lie within three standard deviations of the mean.
Chebyshevs inequality holds for samples and populations, and for discrete and continuous data
regardless of the shape of the distribution.
Worked Example 4: Chebyshevs Inequality
Note: this example has been reproduced from the curriculum.
The arithmetic mean monthly return and standard deviation of monthly returns on the S&P 500
were 0.97 percent and 5.65 percent, respectively, during the 19262002 period, totaling 924
monthly observations. Using this information, address the following:
1. Calculate the endpoints of the interval that must contain at least 75 percent of monthly
returns according to Chebyshevs inequality.
2. What are the minimum and maximum number of observations that must lie in the interval
computed in Part 1, according to Chebyshevs inequality?
Solution to 1:
According to Chebyshevs inequality, at least 75 percent of the observations must lie within two
standard deviations of the mean, X 2s. For the monthly S&P 500 return series, we have 0.97%
2(5.65%) = 0.97% 11.30%. Thus the lower endpoint of the interval that must contain at least
75 percent of the observations is 0.97% 11.30% = 10.33%, and the upper endpoint is 0.97% +
11.30% = 12.27%.
Solution to 2:
Page 18
Irfanullah.co
For a sample size of 924, at least 0.75(924) = 693 observations must lie in the interval from
10.33% to 12.27% that we computed in Part 1.
When the observations are returns, the coefficient of variation measures the amount of risk
(standard deviation) per unit of mean return. Hence, it allows us to directly compare dispersion
across different data sets. Consider a simple example. Investment A has a mean return of 7% and
a standard deviation of 5%. Investment B has a mean return of 12% and a standard deviation of
7%. The coefficients of variation can be calculated as follows:
5%
= 0.71
7%
7%
=
= 0.58
12%
=
where
Copyright Irfanullah Financial Training. All rights reserved.
Page 19
Irfanullah.co
Arithmetic mean
Variance of (%)
return (%)
Portfolio A
16.4%
4.9%
Portfolio B
12.6%
3.5%
Solution:
:
16.4 10.5
4.9
12.6 10.5
3.5
= 2.665
= 1.122
Portfolio A offers a higher excess return per unit of risk relative to Portfolio B.
Page 20
Irfanullah.co
For the positively skewed unimodal distribution, the mode is less than the median, which is less
than the mean. For the negatively skewed unimodal distribution, the mean is less than the
median, which is less than the mode. All else equal, if investment returns have negative skew,
that is considered more risky than symmetric and positively skewed distributions. A negative
skew implies a fat left tail and hence a relatively high probability of extreme losses. A skewness
of greater than 0.5 or less than -0.5 is considered significant.
The curriculum presents formulas for calculating skewness. However, it is extremely unlikely
that well be tested on these formulas at Level I. Consequently the formulas are not being
reproduced in these notes.
Page 21
Irfanullah.co
For all normal distributions, kurtosis is equal to 3. Excess kurtosis is kurtosis minus 3. Hence, a
mesokurtic distribution has excess kurtosis equal to 0. A leptokurtic distribution has excess
kurtosis greater than 0, and a platykurtic distribution has excess kurtosis less than 0. For a sample
of 100 or larger taken from a normal distribution, a sample excess kurtosis of 1.0 or larger would
be considered unusually large.
A leptokurtic distribution is considered more risky than a normal distribution because it has fatter
tails and hence a higher probability of extreme losses.
Page 22
Irfanullah.co
Summary
Note: This summary has been adapted from the CFA Program curriculum.
Data measurements are taken using one of the four major scales: nominal, ordinal, interval,
or ratio. Nominal scales categorize data but do not rank them. Ordinal scales sort data into
categories that are ordered with respect to some characteristic. Interval scales provide not
only ranking but also assurance that the differences between scale values are equal. Ratio
scales have all the characteristics of interval scales as well as a true zero point as the origin.
The scale on which data are measured determines the type of analysis that can be performed
on the data.
A histogram is a bar chart of data that have been grouped into a frequency distribution. A
frequency polygon is a graph of frequency distributions obtained by drawing straight lines
joining successive points representing the class frequencies.
Measures of central tendency specify where data are centered and include the (arithmetic)
mean, median, and mode (most frequently occurring value). The mean is the sum of the
observations divided by the number of observations. The median is the value of the middle
item (or the mean of the values of the two middle items) when the items in a set are sorted
Page 23
Irfanullah.co
into ascending or descending order. The mean is the most frequently used measure of central
tendency. The median is not influenced by extreme values and is most useful in the case of
skewed distributions. The mode is the only measure of central tendency that can be used with
nominal data.
A portfolios return is a weighted mean return computed from the returns on the individual
assets, where the weight applied to each assets return is the fraction of the portfolio invested
in that asset.
The geometric mean is especially important in reporting compound growth rates for time
series data. When calculating the average return over a given period the following formula is
used:
R G =[(1+R1) (1+R2) (1+Rn)]1/n 1
The harmonic mean is a special type of weighted mean in which an observations weight is
inversely proportional to its magnitude. The formula is:
1
= / ( )
=1
For any data set where the values are not the same, arithmetic mean > geometric mean >
harmonic mean.
Quantiles such as the median, quartiles, quintiles, deciles, and percentiles are location
parameters that divide a distribution into halves, quarters, fifths, tenths, and hundredths,
respectively.
Dispersion measures such as the variance, standard deviation, and mean absolute deviation
(MAD) describe the variability of outcomes around the arithmetic mean.
Range is defined as the maximum value minus the minimum value. Range has only a limited
scope because it uses information from only two observations.
MAD is average of the absolute deviation from the mean. This can be expressed as:
= [| |] /
=1
The variance is the average of the squared deviations around the mean, and the standard
deviation is the positive square root of variance. In computing sample variance (s2) and
Page 24
Irfanullah.co
sample standard deviation, the average squared deviation is computed using a divisor equal
to the sample size minus 1.
The coefficient of variation, CV, is the ratio of the standard deviation of a set of observations
to their mean value. A scale-free measure of relative dispersion, by expressing the magnitude
of variation among observations relative to their average size, the CV permits direct
comparisons of dispersion across different data sets.
The Sharpe ratio for a portfolio, p, based on historical returns, is defined as: (return on
portfolio risk free rate) / standard deviation of portfolio. It gives the excess return per unit
of risk.
Skew describes the degree to which a distribution is not symmetric about its mean. A return
distribution with positive skewness has frequent small losses and a few extreme gains. A
return distribution with negative skewness has frequent small gains and a few extreme losses.
Zero skewness indicates a symmetric distribution of returns.
Kurtosis measures the peakness of a distribution and provides information about the
probability of extreme outcomes. A distribution that is more peaked than the normal
distribution is called leptokurtic; a distribution that is less peaked than the normal distribution
is called platykurtic; and a distribution identical to the normal distribution in this respect is
called mesokurtic. Excess kurtosis is kurtosis minus 3, the value of kurtosis for all normal
distributions.
Next Steps
Make sure you are comfortable using the financial calculator for statistical calculations.
Page 25
Irfanullah.co
Review the learning outcomes presented in the curriculum. Make sure that you can perform
that actions implied by learning outcome.
Page 26