04-stats review 1_2014

Upload: strongchong00

Post on 02-Jun-2018

221 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/10/2019 04-Stats Review 1_2014

    1/34

    1

    Statistics Review 1

    Marriott School of Management

    Fall 2014

    Rob Schonlau

    Last updated Sept 10, 2014

  • 8/10/2019 04-Stats Review 1_2014

    2/34

    2

    Statistics review part 1

    As introduced earlier, we think of risk in terms of the likelihood

    of observing return outcomes that are far different than the

    expected outcome.

    Financial theory states that there is a risk-return relationshipwhere people must be compensated for bearing additional risk.

    Before we can discuss the formal financial risk-return models

    we need to first review some of the related statistical concepts.

  • 8/10/2019 04-Stats Review 1_2014

    3/34

    Lecture 4 outline

    Discuss statistical concepts that are useful for thinking about an

    assets performance and risk.

    Random variables

    Expectations Probability density functions (PDFs) Variance and standard deviation

    Review the normal distribution. Use properties of the normal

    distribution to answer probability questions and to solve for the

    value at risk.

    3

  • 8/10/2019 04-Stats Review 1_2014

    4/34

    DefinitionsRandom Variable: A variable whose value is uncertain. I find it helpful

    to think of a data generating process that generates all the possible

    outcome values in each time period for the variable in question.

    Example: IBM stock returns

    Observation: The observed value from a single outcome of the random

    variable. You can think about an observation as a single draw or a

    single example observed out of a whole underlying population of

    possible values.

    Next years observed IBM annual return will be a singleobservation from the underlying possible set of IBM returns.

    4

  • 8/10/2019 04-Stats Review 1_2014

    5/34

    Average weekly return: .0019

    Standard deviation: .046

    5

  • 8/10/2019 04-Stats Review 1_2014

    6/34

    Definitions continued

    Probability density function (PDF): A function that describes the

    probability of each outcome for a random variable.

    Expectation or Expected Value: The mean or expected value of a

    random variable is a single value that summarizes the value you would

    observe on average if you could observe the outcome of the random

    variable many times. If r is a random variable then the expectation

    notation is E[r], or sometimes m.

    6

  • 8/10/2019 04-Stats Review 1_2014

    7/34

    Discrete PDFExample of a discrete PDF

    Only a finite number of outcomes (in this example there are twopossible outcomes or two states) The value of the function, p(s), tells us the exact probability of

    observing a given state.

    The sum of the probabilitiesacross all possible outcomes(states) must equal 1.

    Discrete PDFs are not very realistic. Why do we use them?

    Provides statistical intuition for more complex distributions Part of the CFA curriculum

    %15for25.0

    %10for75.0)(

    r

    rsp

    7

  • 8/10/2019 04-Stats Review 1_2014

    8/34

    Continuous PDF

    Example: Normal PDF

    Infinite number of outcomes even within defined range The integral of the function between two points tells us the

    probability of getting an outcome between those two points.

    The integral of the function over the range of possible outcomesmust equal 1.

    8

  • 8/10/2019 04-Stats Review 1_2014

    9/34

    Expectation

    For a discrete probability function with Spossible outcomes (states)

    where p(s) = probability of each of S possible statesr(s) = observed return if state s occurs

    For example given the following probability function:

    E[r]=0.75*(.10)+.25*(-.15)= 3.75%

    S

    ssrsprE

    1)()(][

    %15for25.0%10for75.0)(

    rrsp

    9

  • 8/10/2019 04-Stats Review 1_2014

    10/34

    Example of calculating expected return

    Using the general PDF notation the information in this table can be

    summarized as:

    State of

    Economy

    Scenarios

    (states)

    Probability of

    each state

    Returns

    Boom 1 .25 44%

    Normal growth 2 .50 14%

    Recession 3 .25 -16%

    %16for25.0

    %14for50.0

    %44for25.0

    )(

    r

    r

    r

    sp

    10E[r]=.25(.44) + .50(.14) + .25(-.16) = .14

  • 8/10/2019 04-Stats Review 1_2014

    11/34

    Expectations using continuous PDFs

    For a continuous probability function

    The idea is the same as for a discrete PDF. We just integrate

    across all possible values rather than sum over the discrete

    values.

    drsrsprE )()(][

    11

  • 8/10/2019 04-Stats Review 1_2014

    12/34

    Expectation of a function of a random

    variable

    At times we are interested in the expectation of a function of a random

    variable. For example, assume the following discrete PDF for random

    variable r:

    What is the E[r], E[r2], and E[3r+5]?

    E[r] = .65*(.08) + .35*(-.10) = 0.017 E[r2] = .65*(.082) +. 35*(-.102) = 0.008 E[3r+5] = .65*(3*.08+5)+ .35*(3*(-.10)+5) = 5.051

    %10for35.0

    %8for65.0

    )( r

    r

    sp

    12

  • 8/10/2019 04-Stats Review 1_2014

    13/34

    PDFs, expectations, and stock returns

    We never know the true future distribution (PDF) of returns for any

    investment.

    However, we can observe the actual returns over time of aninvestment and with those historical returns infer the nature of the

    (unobservable) underlying process generating those outcomes.

    For example, we dont know the true underlying PDF for future IBM

    stock returns. But if we know that prior IBM returns have averaged10% a year with a standard deviation of 4% we can get an idea of

    the distribution of IBMs future returns.

    13

  • 8/10/2019 04-Stats Review 1_2014

    14/34

    Average weekly return: .0019

    Standard deviation: .046

    14

  • 8/10/2019 04-Stats Review 1_2014

    15/34

    Summary of Expectations

    Given an assumed PDF we can find the expectation as follows

    When we dont know anything about the PDF, but rather, observe a

    sample generated by an underlying process, we can estimate theexpected value as a simple average. This works for both discrete and

    continuous PDFs.

    drsrsprE

    srsprE S

    s

    )()(][

    )()(][1

    15

  • 8/10/2019 04-Stats Review 1_2014

    16/34

    Statistics rule #1

    Rule 1: Let x and y be any two random variables. If z = ax + by,

    where a and b are constants, and x and y are random variables,

    then

    Note that because

    bE[y]aE[x]E[z]

    aaE ][][][ xaEaxE

    16

  • 8/10/2019 04-Stats Review 1_2014

    17/34

    Statistics rule #1: Example

    Assume you own portfolio Z with 30% of your wealth in asset A and

    70% in asset B. Assume you have gathered data on the returns to A,

    and B and inferred the following PDF.

    If there is an expansion: rZ= .3(10%) + .7(5%) = 6.5%. Expansions

    occur with probability 0.80.

    If there is a recession: rZ= .3(0%)+.7(3%) = 2.1%. Recessions occur

    with probability 0.20.

    What is the expected return for portfolio z?

    )(recession3and%0for20.0)(expansion5and%10for80.0)(

    %rr%rrsp

    BA

    BA

    17

  • 8/10/2019 04-Stats Review 1_2014

    18/34

    The return on portfolio Z (rZ) is a random variable that is itself a

    function of two other random variables (rAand rB). In either state of

    the world (expansion or recession) rZcan be represented by the

    formula rZ =.3(rA) + .7(rB).

    Using expectations (statistics rule #1):

    E[rZ] = .3 E(rA) + .7 E(rB)

    First solve for E(rA) and E(rB) and then for E[rZ]

    )(recession3and%0for20.0

    )(expansion5and%10for80.0)(

    %rr

    %rrsf

    BA

    BA

    5.62%0.7(0.046)0.3(0.08)]E[

    046.0)03.20(.)05.80(.][

    08.0)020(.)10.80(.][

    Z

    r

    rE

    rE

    B

    A

    18

  • 8/10/2019 04-Stats Review 1_2014

    19/34

    Discrete PDF Example: Which of these

    investments would you choose? Why?Investment #1 PDF:

    E[r] = 3.75%

    Investment #2 PDF:

    E[r] = 3.75%

    %15for25.0

    %10for75.0)(

    r

    rsp

    otherwise0

    %75.3for1

    )(

    r

    sp

    19

  • 8/10/2019 04-Stats Review 1_2014

    20/34

    One possible measure of risk: Expected

    deviation from mean.To estimate the risk involved with the two investments, lets

    calculate the expected deviation from the mean.

    The deviation from the mean for any observed return is r - E[r].

    Hence the expected deviation from the mean is: E[r - E[r]]

    #1: E[r-E[r]] = .75*(.10 - .0375) + .25*(-.15 - .0375) = 0

    #2: E[r-E[r]] = 1*(.0375 - .0375) = 0

    Not a very helpful measure!

    20

  • 8/10/2019 04-Stats Review 1_2014

    21/34

    Another possible measure of risk: Variance

    How about finding the expected squared deviation (variance) from

    the mean?

    Investment #1variance = .75*(.10-.0375)2 + .25*(-.15-.0375)2

    = 0.0117

    Investment #2

    variance = 1*(.0375-.0375)2 = 0

    21

  • 8/10/2019 04-Stats Review 1_2014

    22/34

    Variance

    For a discrete probability function with Soutcomes

    An alternative formula for variance:

    2

    1

    2])[)()(()(

    S

    s

    rEsrsprVar

    222 ][][)( rErErVar

    22

  • 8/10/2019 04-Stats Review 1_2014

    23/34

    Theoretical vs estimated

    Again . . . we consider PDFs to be the underlying number

    generating machines

    In real life, we dont know the true properties of the underlying

    (theoretical) PDF that generates the returns we observe. But the

    returns we observe allow us to learn something about theproperties of the PDF that created them.

    We have formulas for the theoretical expectation and the

    variance. But the underlying PDF is not known so we have to use

    estimates of the expectation and the variance.

    23

  • 8/10/2019 04-Stats Review 1_2014

    24/34

    Estimation

    To estimate the variance using sample observations, just take the

    simple averageof the squared deviations from the estimated mean

    with a slight correction for estimation error.

    Example:

    Sample of returns: 0.10, 0.05, 0, -.03

    .)(1

    1 22

    ii rrn

    03.4/)03.005.010.0( r

    0033.

    ])03.03.()03.0()03.05.0()03.10.0[(*)]14/(1[ 22222

    24

  • 8/10/2019 04-Stats Review 1_2014

    25/34

    Statistics rule #2

    If z = ax + c, where a and c are constants, and x is a random

    variable, then

    If z = ax + by + c, where a and c are constants, and x and y are

    random variables, then

    xz

    xz

    a

    a

    222

    xyyxyxz baba 222222

    25

  • 8/10/2019 04-Stats Review 1_2014

    26/34

    Lecture 4 outline

    Discuss statistical concepts that are useful for thinking about an

    assets performance and risk.

    Random variables

    Expectations PDFs Variance and standard deviation

    Review the normal distribution. Use properties of the normal

    distribution to answer probability questions and to solve for the

    value at risk.

    26

  • 8/10/2019 04-Stats Review 1_2014

    27/34

    What are the benefits of assuming a normal

    distribution?

    Easy to use in math models.

    The normal distribution has well known properties and is easily

    accessible via Excel and other software packages.

    Are returns distributed normally?

    27

  • 8/10/2019 04-Stats Review 1_2014

    28/34

    Normal distribution

    28

  • 8/10/2019 04-Stats Review 1_2014

    29/34

    Are returns normally distributed?

    -0.5

    -0.4

    -0.3

    -0.2

    -0.1

    0

    0.1

    0.2

    0.3 Ann Taylor

    This is a time series plot of the return process. Thex-axis is

    time, and the y-axis is the value of the return at that time.

    29

  • 8/10/2019 04-Stats Review 1_2014

    30/34

    How good is the normal assumption?

    Ann Taylor

    This is a histogram of returns. Thex-axis represents possible outcomes for

    the return. We divide thex-axis into bins or intervals and count the number

    of returns that fall into each interval. The y-axis tells us how many days had a

    return within the corresponding interval.

    30

  • 8/10/2019 04-Stats Review 1_2014

    31/34

    31

    The mean variance framework

    The variance on any investment measures the disparity between

    actual and expected returns.

    Expected Return

    Low Variance Investment

    High Variance Investment

  • 8/10/2019 04-Stats Review 1_2014

    32/34

    Normal distribution

    Assume the PDF for your investment return is a normal

    distribution.

    If we know E[r] and [r] we can integrate under the normal

    curve over any region using calculus (or Excel). That is we can

    find theprobability the return will fall within any given range.

    32

  • 8/10/2019 04-Stats Review 1_2014

    33/34

    Using the PDF distribution to gain

    understanding of possible outcomes.Assume the PDF for your investment return is a normal

    distribution with E[r]=10% and [r]=0.15.

    What is the probability that r < - 20%?

    What is the probability that r > 30%?

    What is the probability that -20% < r

  • 8/10/2019 04-Stats Review 1_2014

    34/34

    According to your calculations, over the next year:

    E[r] = 0.10

    = 0.20

    Find the losses you expect to incur with 5% probability.

    5% VAR = 0.101.64*0.20 = -0.23 During any given year, you should expect to lose 23% or

    more of your portfolio value with 5% probability.

    Example application of the normal

    distribution: 5% Value-at-Risk (VAR)

    34