Quantitative Applications in Finance

Embed Size (px)

Citation preview

  • 8/11/2019 Quantitative Applications in Finance

    1/18

    Quantitative Applications in Finance

    Project Report

    Group7

    2014-15

    Indian Institute of Management

    Lucknow

    PGP ID Name Company

    PGP29002 Pratik Mandhana Larsen & Toubro

    PGP29041 Raghavendra Hindalco Industries

    PGP29017 Waman Virgaonkar Ultratech Cement

    PGP29057 Kirtesh Kumar Bank of Baroda

    PGP29033 Srikanth Dasika Petronet LNG

    PGP29334 Venkat Rajath Mahindra & Mahindra

  • 8/11/2019 Quantitative Applications in Finance

    2/18

    1.Companies

    1.1. Ultratech CementUltraTech Cement Limited is engaged in the business of cement and cement related products. The

    Company provides a range of products that cater to all the needs from laying the foundation to

    delivering the final touches. The Company manufactures and provides ordinary Portland and

    Portland Pozzolana Cement, Ready-Mix Concrete, and White Cement. White cement is

    manufactured under Birla White brand, ready mix concretes under UltraTech Concrete brand and

    new age building products under UltraTech Building Products Division. The retail outlets of the

    Company operate under UltraTech Building Solutions. The Company is also an exporter of cement

    clinker spanning export markets in countries across the Indian Ocean, Africa, Europe and the Middle

    East. The Company conducts business activity in United Arab Emirates, Sri Lanka, Bahrain, and

    Bangladesh.

    2.Modelling of Returns

    2.1.

    Testing for Stationarity

    2.1.1. Ultratech Cement

    The returns of Ultratech Cement appear to be stationary as shown below by the ACF/PACF plots and

    the ADF test.

  • 8/11/2019 Quantitative Applications in Finance

    3/18

    2.2.

    Modelling of Returns

    2.2.1. Ultratech Cement

    AR/MA

    The ACF-PACF plots in the stationarity section shows a geometrically declining ACF and PACF with asingle spike. This gives an indication for an AR(1) model. The following is the R output for AR(1)

    model:

    Since the intercept term here is insignificant, we remodel the data without it:

    Based on this model the final equation is:

    yt= 0.097 * yt-1 + et

    The diagnostics test for this model indicates that the returns are pure white noise:

  • 8/11/2019 Quantitative Applications in Finance

    4/18

    ARMA

    The ARMA(1,1) model for the given data comes out to be insignificant as shown below:

    ARMAX

    The ARMAX(0,1) model based on the AR(1) model and using NIFTY Returns as external regressors

    gives the following output:

    Since the intercept and the AR(1) term are both insignificant, we remodel without them:

  • 8/11/2019 Quantitative Applications in Finance

    5/18

    The diagnostics test for this model indicates that the returns are not pure white noise as shown by

    the Ljung-Box statistics:

    This tells that the model is not a good one and hence cannot be considered.

    ARIMA

    The ACF-PACF plots of the first order difference are shown below:

  • 8/11/2019 Quantitative Applications in Finance

    6/18

    The PACF is geometrically declining, while the ACF has two spikes. This gives an indication for

    ARIMA(0,1,2) model.

    The equation of the model comes out to be:

    yt-1= - 0.9113 * et-1 0.0887 * et-2 + et

    The diagnostics test shows that the errors are pure white noise:

  • 8/11/2019 Quantitative Applications in Finance

    7/18

    The following table gives the summary of all the models that can be used for Ultratech Cement:

    Model Equation

    AR(1) yt= 0.097 * yt-1 + et

    ARIMA(0,1,2) yt-1= - 0.9113 * et-1 0.0887 * et-2 + et

    2.3.

    Forecasts from models

    2.3.1. Ultratech Cement

    The following table compares the return forecasts of all the suitable models for Ultratech Cement

    with the actual returns for a period of 1 month:

  • 8/11/2019 Quantitative Applications in Finance

    8/18

    This shows that the AR(1)is clearly the best model for Ultratech.

    3.Modeling of Volatility

    3.1.

    Testing for ARCH effect

    3.1.1. Ultratech Cement

    The ACF-PACF plots of the residuals from the AR(1) model and the ARCH LM test shows that the datasuffers from an ARCH effect:

    Forecasts Sign Direction Forecasts Sign Direction

    02-06-2014 0.028587 0.002761 1 0.003511 1

    03-06-2014 0.042117 0.004067 1 0 0.004436 1 0

    04-06-2014 0.007186 0.000696 1 1 0.001252 1 1

    05-06-2014 0.023582 0.0023 1 0 0.003009 1 0

    06-06-2014 0.01603 0.001587 1 1 0.002217 1 1

    09-06-2014 0.047127 0.004775 1 0 0.005173 1 0

    10-06-2014 -0.00102 -0.0001 1 1 0.000523 0 1

    11-06-2014 -0.00801 -0.00082 1 0 0.000297 0 0

    12-06-2014 -0.00245 -0.00025 1 1 0.000786 0 1

    13-06-2014 -0.00025 -2.55E-05 1 1 0.000928 0 1

    16-06-2014 -0.00771 -0.00078 1 1 0.000236 0 0

    17-06-2014 0.018295 0.001825 1 1 0.002688 1 1

    18-06-2014 0.002932 0.000292 1 1 0.001093 1 1

    19-06-2014 -0.01772 -0.00173 1 1 -0.00059 1 1

    20-06-2014 -0.01014 -0.00096 1 1 0.000271 0 1

    23-06-2014 -0.00594 -0.00056 1 1 0.000594 0 1

    24-06-2014 -0.00203 -0.00019 1 1 0.000906 0 1

    25-06-2014 0.00693 0.000651 1 1 0.001642 1 1

    26-06-2014 -0.01448 -0.00135 1 1 -0.00025 1 1

    27-06-2014 -0.0332 -0.00315 1 0 -0.00176 1 0

    30-06-2014 -0.00152 -0.00014 1 1 0.001092 0 1

    Accuracy 100.00% 75.00% 57.14% 70.00%

    Date

    Actual

    Returns

    AR(1) Model ARIMA(0,1,2) Model

  • 8/11/2019 Quantitative Applications in Finance

    9/18

    3.2. Volatility Models

    3.2.1. Ultratech cement

    The Skewness and the Kurtosis of the returns comes out to be as follows:

    As we can see that these moments differ a lot from the normal distribution and hence we are using a

    sged distribution for all the models of volatility.

    SD & EVE

    The following plot shows the Standard Deviation with close and other Extreme Value Estimators

  • 8/11/2019 Quantitative Applications in Finance

    10/18

  • 8/11/2019 Quantitative Applications in Finance

    11/18

    The Volatility equation for this model is:

    ht= 0.065583 * residt-1 + 0.808354 * ht-1

    The diagnostics of the model shows everything perfectly fine:

  • 8/11/2019 Quantitative Applications in Finance

    12/18

    eGARCH

    The eGARCH model for the given data is as follows:

  • 8/11/2019 Quantitative Applications in Finance

    13/18

    The Volatility equation for this model is:

    log(ht)= 0.149852 * |residt-1/ht-1| + 0.913837 * log(ht-1) - 0.075339 * (residt-1/ht-1)

    The diagnostics of the model shows everything perfectly fine:

  • 8/11/2019 Quantitative Applications in Finance

    14/18

    gjrGARCH

    The gjrGARCH model for Ultratech Cement is as follows:

  • 8/11/2019 Quantitative Applications in Finance

    15/18

    The Volatility equation for this model is:

    ht= 0.000024 + 0.850043 * ht-1 - 0.093782 * residt-1* It-1

    The diagnostics of the model shows everything perfectly fine:

  • 8/11/2019 Quantitative Applications in Finance

    16/18

    gjrGARCH-m

    The gjrGARCH-in-Mean model for the given sample of data comes out to be as follows:

  • 8/11/2019 Quantitative Applications in Finance

    17/18

    The return and volatility equations for this model as are as follows:

    yt= 0.046222 * yt-1 + 3.578019 * sigmat-1 + et

    ht= 0.847886 * ht-1 - 0.086535 * residt-1* It-1

    The diagnostics of this model are perfectly normal:

  • 8/11/2019 Quantitative Applications in Finance

    18/18