Scribd Logo
Upload

Welcome to Scribd!

  • 13 views

    Value-at-Risk (VAR)

    Uploaded by

    Mayur Solanki
    Value-at-Risk (VAR) is a measure of the risk of loss on a portfolio over a specified time period at a given confidence level. It tells you how much you could expect to lose in a typical risk scenario but not in extreme cases. VAR can be estimated using historical simulation, which looks at historical changes in risk factors, or Monte Carlo simulation, which simulates possible future outcomes. Both methods calculate VAR as the loss level that is exceeded with the given probability. VAR is commonly used but also faces criticism for not capturing tail risks and potentially encouraging excessive risk-taking.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PPT, PDF, TXT or read online from Scribd

    Value-at-Risk (VAR)

    Uploaded by

    Mayur Solanki
    13 views19 pages
    Value-at-Risk (VAR) is a measure of the risk of loss on a portfolio over a specified time period at a given confidence level. It tells you how much you could expect to lose in a typical risk scenario but not in extreme cases. VAR can be estimated using historical simulation, which looks at historical changes in risk factors, or Monte Carlo simulation, which simulates possible future outcomes. Both methods calculate VAR as the loss level that is exceeded with the given probability. VAR is commonly used but also faces criticism for not capturing tail risks and potentially encouraging excessive risk-taking.

    Original Description:

    value at risk

    Original Title

    Var

    Copyright

    © Attribution Non-Commercial (BY-NC)

    Available Formats

    PPT, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Value-at-Risk (VAR) is a measure of the risk of loss on a portfolio over a specified time period at a given confidence level. It tells you how much you could expect to lose in a typical risk scenario but not in extreme cases. VAR can be estimated using historical simulation, which looks at historical changes in risk factors, or Monte Carlo simulation, which simulates possible future outcomes. Both methods calculate VAR as the loss level that is exceeded with the given probability. VAR is commonly used but also faces criticism for not capturing tail risks and potentially encouraging excessive risk-taking.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PPT, PDF, TXT or read online from Scribd
    Download as ppt, pdf, or txt
    13 views19 pages

    Value-at-Risk (VAR)

    Uploaded by

    Mayur Solanki
    Value-at-Risk (VAR) is a measure of the risk of loss on a portfolio over a specified time period at a given confidence level. It tells you how much you could expect to lose in a typical risk scenario but not in extreme cases. VAR can be estimated using historical simulation, which looks at historical changes in risk factors, or Monte Carlo simulation, which simulates possible future outcomes. Both methods calculate VAR as the loss level that is exceeded with the given probability. VAR is commonly used but also faces criticism for not capturing tail risks and potentially encouraging excessive risk-taking.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PPT, PDF, TXT or read online from Scribd
    Download as ppt, pdf, or txt
    of 19

    Value-at-Risk (VAR)

    What is it?
    For risk management purposes, it might be useful to estimate how
    much money your portfolio might lose over a given time span and a
    certain confidence level.

    VAR is defined as VAR (holding period; 1 ) where 1 is your
    confidence interval.

    If VAR (30-day; 95%) = $10,000 then in 95 out of 100 months the
    change in the value of your portfolio will be less than $10,000.

    Note that this tells you nothing about the magnitude of potential
    losses in extreme cases. Use expected shortfall rather than VAR for
    such cases.
    Graphically
    How to choose parameters?
    The time horizon can differ from a few hours for an active trading
    desk to a year for a pension fund.

    When the primary goal is to satisfy external regulatory requirements,
    such as bank capital requirements, the confidence level is typically
    very large (for example, 99%). However for an internal risk
    management model used by a company to control the risk exposure
    the typical number is around 5%.
    A simple example
    Suppose portfolio manager manages a portfolio which consists of a
    single asset. The return of the asset is normally distributed with
    annual mean return 10% and annual standard deviation 30%. The
    value of the portfolio today is $100 million. We want to answer
    various simple questions about the end-of-year distribution of
    portfolio value:

    1. What is the probability of a loss of more than $20 million dollars by
    year end (i.e., what is the probability that the end-of-year value is less
    than $80 million)?
    2. What is the VAR at 99% confidence level?

    First question
    If you lose $20 million, your annual return is -20%. So, you want to
    know P(r < -20%). But how can you compute this?

    One thing that you can do is to use the probability density function
    associated with normal distributions



    Remember, r is a normally distributed random variable with =10%
    and =30%, = 3.14159, and e=2.71828. So,




    |
    |
    |
    .
    |

    \
    |

    [
    =

    o

    o
    2
    ) (
    2
    2
    exp
    2
    1
    ) (
    x
    x f
    Graphically
    r
    =10% r=-20%
    Shaded region: P(r < -20%)
    |
    |
    .
    |

    \
    |
    =

    ) 3 . 0 ( 2
    ) 1 . 0 (
    2
    2
    exp
    ) 71828 . 2 ( 2 3 . 0
    1
    ) (
    r
    r f
    f(r)
    Integration
    So, we need to solve for
    dr
    r
    }


    |
    |
    .
    |

    \
    |

    2 . 0
    2
    2
    ) 3 . 0 ( 2
    ) 1 . 0 (
    exp
    ) 71828 . 2 ( 2 3 . 0
    1
    Is there an easier way to compute
    P(r < -20%)?
    Certainly there is!

    Well transform r to a standard normal variable z and examine the z-
    table rather than doing integrals?

    The idea is as follows: Every value of a random variable r ~ N (, )
    can be transformed to z ~ N (0, 1) by using



    So, if r=-20% then

    Then, P(r < -20%) = P(z < -1)

    o

    =
    r
    z
    0 . 1
    % 30
    % 10 % 20
    =

    = z
    The z-table
    Answers
    So, based on the z-table we find that P(r < -20%) or put differently
    P(Losses>$20,000) = 15.87%.

    The second question is VAR(1-year; 99%)=?

    Go to the z-table and find the value that satisfies P(z < ?) = 1%.
    z = -2.33

    What is the r associated with z=-2.33?

    If starting portfolio value is $100,000 and r = -59.9%, then loss = $59,900.

    VAR(1-year; 99%)=59,900.


    % 9 . 59 , 33 . 2
    % 30
    % 10
    = =

    r
    r
    Popular Methods for
    Estimating VAR
    1. Historical simulation
    2. Monte Carlo simulation
    3. Delta Normal method

    In this class, we will focus only on the first two.
    Example
    Suppose you have a zero-coupon one year CD with 3.41% yield to
    maturity. For simplicity, lets assume that the only risk factor relevant
    for the valuation of this CD is the market interest rate. So, you can
    model the change in the value of your portfolio as



    So, if you can find a way to estimate the probability distribution of r,
    you can estimate how much money you will gain or loose with some
    statistical confidence.

    Suppose we are interested in VAR(1-day; 90%) for this CD. Lets
    see how VAR is calculated by using each method.

    0341 . 1
    100 $
    0341 . 1
    100 $

    A +
    = A
    r
    V
    Historical simulation
    Start with the following question: Over the last 1-year, or 6-month, or
    3-month etc period, what were the daily changes in the interest rate?

    Suppose you go back one year and in a typical calendar year there
    are 252 trading days. So for each of the 252 trading days before
    today you observed and recorded the changes in market interest
    rates (or a data vendor did this for you).

    Suppose the distribution of daily changes in interest rates (%) are:

    N Mean p10 p50 p90
    252 -0.0052 -0.0500 0.0000 0.0300 0.0368
    Since daily change in r equals 3 bps at the 90
    th
    percentile




    Historical simulation (contd)
    028046 . 0
    0341 . 1
    100 $
    0003 . 0 0341 . 1
    100 $
    %) 90 , 1 ( =
    +
    = day VAR
    Monte Carlo simulation
    As in historical simulation, collect data on daily changes in market
    interest rates over the past year.

    Calculate the mean and standard deviation of daily changes in r.
    We now from the previous slides that = -0.0052 and = 0.0368

    Assuming changes in interest rates are normally distributed, and
    using historical and , we will do 10,000 iterations to estimate the
    probability density function of the change in interest rates.

    Less technically, we will obtain 10,000 different values for change in
    r, but when all these 10,000 observations are combined together,
    they exhibit a normal distribution.

    Simulated
    Change in Change in
    Simulation Interest Position
    Period Rate (%) Value ($s)
    1 0.0720 -$0.0673
    2 -0.0313 $0.0293
    3 0.0062 -$0.0058
    4 -0.0726 $0.0680
    5 -0.0187 $0.0175
    6 0.0453 -$0.0424
    7 0.0649 -$0.0606
    8 -0.0093 $0.0087
    9 0.0048 -$0.0045
    10 0.0081 -$0.0076

    9998 -0.0236 $0.0220
    9999 0.0160 -$0.0149
    10000 -0.0457 $0.0427
    VAR(90%) -$0.0391
    VAR(95%) -$0.0526
    VAR(99%) -$0.0741
    VAR(99.5%) -$0.0800
    Worst Loss (1 in 10,000) -$0.1241
    Criticisms of the VAR approach
    Some Harsh Criticisms
    Charlatanism!
    an airbag that works all the time except when you have an
    accident.
    Easily misunderstood and so dangerous
    Gives a false sense of security to bank executives and regulators
    Gives incentive to take excessive but remote risks

    You might also like