Scribd Logo
Upload

Welcome to Scribd!

  • 22 views

    Ashutosh Internal 2

    Uploaded by

    BappiSwargiary
    The document estimates GARCH models to analyze exchange rate volatility between the Indian rupee and foreign currency. It is found that: 1) The exchange rate series has a unit root and becomes stationary after taking log differences. 2) An ARCH(1) model is appropriate to model exchange rate volatility. 3) A GARCH(1,1) model provides an even better fit, capturing volatility clustering in the exchange rate returns series.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Ashutosh Internal 2

    Uploaded by

    BappiSwargiary
    22 views4 pages
    The document estimates GARCH models to analyze exchange rate volatility between the Indian rupee and foreign currency. It is found that: 1) The exchange rate series has a unit root and becomes stationary after taking log differences. 2) An ARCH(1) model is appropriate to model exchange rate volatility. 3) A GARCH(1,1) model provides an even better fit, capturing volatility clustering in the exchange rate returns series.

    Original Description:

    GARCH

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document estimates GARCH models to analyze exchange rate volatility between the Indian rupee and foreign currency. It is found that: 1) The exchange rate series has a unit root and becomes stationary after taking log differences. 2) An ARCH(1) model is appropriate to model exchange rate volatility. 3) A GARCH(1,1) model provides an even better fit, capturing volatility clustering in the exchange rate returns series.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    22 views4 pages

    Ashutosh Internal 2

    Uploaded by

    BappiSwargiary
    The document estimates GARCH models to analyze exchange rate volatility between the Indian rupee and foreign currency. It is found that: 1) The exchange rate series has a unit root and becomes stationary after taking log differences. 2) An ARCH(1) model is appropriate to model exchange rate volatility. 3) A GARCH(1,1) model provides an even better fit, capturing volatility clustering in the exchange rate returns series.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 4

    Ashutosh Kumar

    Regd No- 14350009


    2nd Internal Test

    1. To estimate GARCH model and check with relevant diagnostic test. Write
    interpretation.

    Null Hypothesis: EXR has a unit root


    Exogenous: Constant
    Lag Length: 0 (Automatic - based on SIC, maxlag=21)

    Augmented Dickey-Fuller test statistic


    Test critical values:
    1% level
    5% level
    10% level

    t-Statistic

    Prob.*

    -1.046921
    -3.436384
    -2.864092
    -2.568181

    0.7381

    *MacKinnon (1996) one-sided p-values.

    Thus Exr has a unit root. To make this data stationary we take log difference.
    Similarly we make Fca also stationary.
    Now we estimate the mean equation
    Dexr c dfca

    Dependent Variable: DEXR


    Method: Least Squares
    Date: 04/18/16 Time: 13:07
    Sample (adjusted): 4/09/1993 4/19/2013
    Included observations: 1046 after adjustments

    Variable

    Coefficient

    Std. Error

    t-Statistic

    Prob.

    0.000573

    0.000262

    2.187779

    0.0289

    DFCA

    -0.017487

    0.005894

    -2.966719

    0.0031

    R-squared

    0.008360

    Mean dependent var

    0.000512

    Adjusted R-squared

    0.007410

    S.D. dependent var

    S.E. of regression

    0.008447

    Akaike info criterion

    -6.708185

    Sum squared resid

    0.074485

    Schwarz criterion

    -6.698715

    0.008478

    Heteroskedasticity Test: ARCH


    F-statistic
    Obs*R-squared

    35.78913
    34.66816

    Prob. F(1,1043)
    Prob. Chi-Square(1)

    0.0000
    0.0000

    Test Equation:
    Dependent Variable: RESID^2
    Method: Least Squares
    Date: 04/18/16 Time: 13:08
    Sample (adjusted): 4/16/1993 4/19/2013
    Included observations: 1045 after adjustments
    Variable
    C
    RESID^2(-1)
    R-squared
    Adjusted R-squared
    S.E. of regression

    Coefficient

    Std. Error

    t-Statistic

    Prob.

    5.83E-05
    0.182144

    6.90E-06
    0.030447

    8.449302
    5.982401

    0.0000
    0.0000

    0.033175
    0.032248
    0.000212

    Mean dependent var


    S.D. dependent var
    Akaike info criterion

    7.13E-05
    0.000215
    -14.07933

    The ARCH test shows that there is ARCH effect. Hence we estimate the model using
    ARCH (1).

    Dependent Variable: DEXR


    Method: ML ARCH - Normal distribution (BFGS / Marquardt steps)
    Date: 04/18/16 Time: 13:09
    Sample (adjusted): 4/09/1993 4/19/2013
    Included observations: 1046 after adjustments
    Convergence achieved after 15 iterations
    Coefficient covariance computed using outer product of gradients
    Presample variance: backcast (parameter = 0.7)
    GARCH = C(3) + C(4)*RESID(-1)^2
    Variable

    Coefficient

    Std. Error

    z-Statistic

    Prob.

    C
    DFCA

    0.000522
    -0.021542

    0.000106
    0.003044

    4.929825
    -7.077677

    0.0000
    0.0000

    8.27E-07
    0.065356

    35.19374
    17.46028

    0.0000
    0.0000

    Variance Equation
    C
    RESID(-1)^2

    2.91E-05
    1.141140

    R-squared
    Adjusted R-squared
    S.E. of regression
    Sum squared resid
    Log likelihood
    Durbin-Watson stat

    0.007850
    0.006900
    0.008449
    0.074523
    3640.669
    1.899552

    Mean dependent var


    S.D. dependent var
    Akaike info criterion
    Schwarz criterion
    Hannan-Quinn criter.

    0.000512
    0.008478
    -6.953479
    -6.934539
    -6.946296

    Checking for further ARCH effects,

    Heteroskedasticity Test: ARCH


    F-statistic
    Obs*R-squared

    2.141978
    2.141687

    Prob. F(1,1043)
    Prob. Chi-Square(1)

    0.1436
    0.1433

    Test Equation:
    Dependent Variable: WGT_RESID^2
    Method: Least Squares
    Date: 04/18/16 Time: 13:11
    Sample (adjusted): 4/16/1993 4/19/2013
    Included observations: 1045 after adjustments
    Variable

    Coefficient

    Std. Error

    t-Statistic

    Prob.

    C
    WGT_RESID^2(-1)

    1.046104
    -0.045287

    0.106565
    0.030943

    9.816606
    -1.463550

    0.0000
    0.1436

    R-squared
    Adjusted R-squared
    S.E. of regression
    Sum squared resid
    Log likelihood
    F-statistic
    Prob(F-statistic)

    0.002049
    0.001093
    3.297311
    11339.77
    -2728.587
    2.141978
    0.143618

    Mean dependent var


    S.D. dependent var
    Akaike info criterion
    Schwarz criterion
    Hannan-Quinn criter.
    Durbin-Watson stat

    1.000948
    3.299114
    5.226004
    5.235481
    5.229598
    1.992345

    shows that there is no more ARCH effect . But residuals shows autocorrelations .
    Therefore we proceed to estimate GARCH (1, 1) model.

    Dependent Variable: DEXR


    Method: ML ARCH - Normal distribution (BFGS / Marquardt steps)
    Date: 04/18/16 Time: 13:13
    Sample (adjusted): 4/09/1993 4/19/2013
    Included observations: 1046 after adjustments
    Convergence achieved after 25 iterations
    Coefficient covariance computed using outer product of gradients
    Presample variance: backcast (parameter = 0.7)
    GARCH = C(3) + C(4)*RESID(-1)^2 + C(5)*GARCH(-1)
    Variable
    C

    Coefficient

    Std. Error

    z-Statistic

    Prob.

    0.000244

    0.000143

    1.709137

    0.0874

    DFCA

    -0.018014

    0.004367

    -4.124474

    0.0000

    6.62E-08
    0.016114
    0.008541

    16.02925
    18.81591
    88.83197

    0.0000
    0.0000
    0.0000

    Variance Equation
    C
    RESID(-1)^2
    GARCH(-1)
    R-squared
    Adjusted R-squared
    S.E. of regression
    Sum squared resid
    Log likelihood
    Durbin-Watson stat

    1.06E-06
    0.303198
    0.758747
    0.006828
    0.005877
    0.008453
    0.074600
    3859.397
    1.895085

    Mean dependent var


    S.D. dependent var
    Akaike info criterion
    Schwarz criterion
    Hannan-Quinn criter.

    0.000512
    0.008478
    -7.369783
    -7.346109
    -7.360805

    now residuals show no autocorrelation. Hence GARCH (1, 1) model fits nicely.
    GARCH01
    .0009
    .0008
    .0007
    .0006
    .0005
    .0004
    .0003
    .0002
    .0001

    4/2/93
    1/7/94
    10/14/94
    7/21/95
    4/26/96
    1/31/97
    11/7/97
    8/14/98
    5/21/99
    2/25/00
    12/1/00
    9/7/01
    6/14/02
    3/21/03
    12/26/03
    10/1/04
    7/8/05
    4/14/06
    1/19/07
    10/26/07
    8/1/08
    5/8/09
    2/12/10
    11/19/10
    8/26/11
    6/1/12
    3/8/13

    .0000

    The variance series shows that there is high volatility on 4/26/96, 8/14/98, 8/1/08 and on
    6/11/12 than the normal mean variance.

    You might also like