Cointegration Results

download Cointegration Results

of 9

Transcript of Cointegration Results

  • 7/30/2019 Cointegration Results

    1/9

    RESEARCH METHODS IN FINANCE

    Assignment-01

    Submitted To:

    Dr Arshad Hassan

    Submitted By:

    Sohail Rizwan

    Reg # PM123012

    Sec # 01

  • 7/30/2019 Cointegration Results

    2/9

    CO-INTEGRATION

    Sense of co-movement by graph:

    Using the above graphical analysis we observe a sense of co-movement between the two selected series

    i.e. Shangai Stock Exchange and Six Swiss exchange-VTX.

    0

    1

    2

    3

    4

    5

    6

    7

    8

    9

    10

    Ln SSE

    Ln VTX

  • 7/30/2019 Cointegration Results

    3/9

    Unit Root Analysis

    A. Shangai Stock Exchange (SSE)Trend and Intercept at level:

    Null Hypothesis: SSE has a unit rootExogenous: Constant, Linear Trend

    Lag Length: 0 (Automatic - based on SIC, maxlag=15)

    t-Statistic Prob.*

    Augmented Dickey-Fuller test statistic -3.479447 0.0436

    Test critical values: 1% level -3.992283

    5% level -3.426494

    10% level -3.136480

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

    Augmented Dickey-Fuller Test Equation

    Dependent Variable: D(SSE)

    Method: Least Squares

    Date: 08/31/13 Time: 02:07

    Sample (adjusted): 2 272

    Included observations: 271 after adjustments

    Variable Coefficient Std. Error t-Statistic Prob.

    SSE(-1) -0.064808 0.018626 -3.479447 0.0006

    C 0.435508 0.117134 3.718027 0.0002

    @TREND(1) 0.000314 0.000170 1.850427 0.0654

    R-squared 0.051382 Mean dependent var 0.010264Adjusted R-squared 0.044303 S.D. dependent var 0.134076

    S.E. of regression 0.131072 Akaike info criterion -1.215127

    Sum squared resid 4.604226 Schwarz criterion -1.175251

    Log likelihood 167.6497 Hannan-Quinn criter. -1.199116

    F-statistic 7.258104 Durbin-Watson stat 2.069614

    Prob(F-statistic) 0.000852

    Interpretation: At first step of Unit Root Analysis, trend and intercept is observed insignificant at level

    which suggests using trend further for the detection of data stationarity whether resulting at level or first

    difference.

  • 7/30/2019 Cointegration Results

    4/9

    Intercept at level:

    Null Hypothesis: SSE has a unit root

    Exogenous: Constant

    Lag Length: 0 (Automatic - based on SIC, maxlag=15)

    t-Statistic Prob.*

    Augmented Dickey-Fuller test statistic -3.315579 0.0151

    Test critical values: 1% level -3.454353

    5% level -2.872001

    10% level -2.572417

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

    Augmented Dickey-Fuller Test Equation

    Dependent Variable: D(SSE)

    Method: Least Squares

    Date: 08/31/13 Time: 02:08Sample (adjusted): 2 272

    Included observations: 271 after adjustments

    Variable Coefficient Std. Error t-Statistic Prob.

    SSE(-1) -0.037246 0.011234 -3.315579 0.0010

    C 0.279174 0.081498 3.425519 0.0007

    R-squared 0.039262 Mean dependent var 0.010264

    Adjusted R-squared 0.035690 S.D. dependent var 0.134076

    S.E. of regression 0.131662 Akaike info criterion -1.209812

    Sum squared resid 4.663051 Schwarz criterion -1.183228

    Log likelihood 165.9295 Hannan-Quinn criter. -1.199138

    F-statistic 10.99306 Durbin-Watson stat 2.100668

    Prob(F-statistic) 0.001040

    Interpretation: At second step of Unit Root Analysis, intercept is found significant at level which

    suggests using trend further for the detection of data stationarity. And, Augmented Dickey-Fuller test

    statistic is also significant which proves that data is stationary at level in this series (SSE).

  • 7/30/2019 Cointegration Results

    5/9

    B. Six Swiss Exchange-VTXTrend and Intercept at level:

    Null Hypothesis: VTX has a unit root

    Exogenous: Constant, Linear Trend

    Lag Length: 1 (Automatic - based on SIC, maxlag=15)

    t-Statistic Prob.*

    Augmented Dickey-Fuller test statistic -1.825425 0.6898

    Test critical values: 1% level -3.992411

    5% level -3.426557

    10% level -3.136516

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

    Augmented Dickey-Fuller Test Equation

    Dependent Variable: D(VTX)

    Method: Least SquaresDate: 08/31/13 Time: 02:10

    Sample (adjusted): 3 272

    Included observations: 270 after adjustments

    Variable Coefficient Std. Error t-Statistic Prob.

    VTX(-1) -0.015686 0.008593 -1.825425 0.0691

    D(VTX(-1)) 0.157771 0.059857 2.635798 0.0089

    C 0.135683 0.068118 1.991895 0.0474

    @TREND(1) 2.16E-05 5.29E-05 0.408646 0.6831

    R-squared 0.046570 Mean dependent var 0.005899

    Adjusted R-squared 0.035817 S.D. dependent var 0.045907

    S.E. of regression 0.045077 Akaike info criterion -3.346172

    Sum squared resid 0.540502 Schwarz criterion -3.292862

    Log likelihood 455.7332 Hannan-Quinn criter. -3.324765

    F-statistic 4.330857 Durbin-Watson stat 1.995518

    Prob(F-statistic) 0.005310

    Interpretation: At first step of Unit Root Analysis, trend and intercept is found insignificant at level

    which suggests using trend further for the detection of data stationarity whether resulting at level or first

    difference.

  • 7/30/2019 Cointegration Results

    6/9

    Intercept at level:

    Null Hypothesis: VTX has a unit root

    Exogenous: Constant

    Lag Length: 1 (Automatic - based on SIC, maxlag=15)

    t-Statistic Prob.*

    Augmented Dickey-Fuller test statistic -2.276023 0.1806

    Test critical values: 1% level -3.454443

    5% level -2.872041

    10% level -2.572439

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

    Augmented Dickey-Fuller Test Equation

    Dependent Variable: D(VTX)

    Method: Least Squares

    Date: 08/31/13 Time: 02:11Sample (adjusted): 3 272

    Included observations: 270 after adjustments

    Variable Coefficient Std. Error t-Statistic Prob.

    VTX(-1) -0.013079 0.005746 -2.276023 0.0236

    D(VTX(-1)) 0.155470 0.059499 2.613011 0.0095

    C 0.116417 0.049090 2.371529 0.0184

    R-squared 0.045971 Mean dependent var 0.005899

    Adjusted R-squared 0.038825 S.D. dependent var 0.045907

    S.E. of regression 0.045007 Akaike info criterion -3.352951

    Sum squared resid 0.540842 Schwarz criterion -3.312969

    Log likelihood 455.6484 Hannan-Quinn criter. -3.336896

    F-statistic 6.432860 Durbin-Watson stat 1.994981

    Prob(F-statistic) 0.001868

    Interpretation: At second step of Unit Root Analysis, intercept is found significant at level which

    suggests using trend further for the detection of data stationarity. And, Augmented Dickey-Fuller test

    statistic is insignificant which directs using the same assumption of intercept at first difference on the

    series.

  • 7/30/2019 Cointegration Results

    7/9

    Intercept at first difference:

    Null Hypothesis: D(VTX) has a unit root

    Exogenous: Constant

    Lag Length: 0 (Automatic - based on SIC, maxlag=15)

    t-Statistic Prob.*

    Augmented Dickey-Fuller test statistic -13.96447 0.0000

    Test critical values: 1% level -3.454443

    5% level -2.872041

    10% level -2.572439

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

    Augmented Dickey-Fuller Test Equation

    Dependent Variable: D(VTX,2)

    Method: Least Squares

    Date: 09/02/13 Time: 01:09Sample (adjusted): 3 272

    Included observations: 270 after adjustments

    Variable Coefficient Std. Error t-Statistic Prob.

    D(VTX(-1)) -0.835427 0.059825 -13.96447 0.0000

    C 0.004866 0.002786 1.746595 0.0819

    R-squared 0.421174 Mean dependent var -0.000382

    Adjusted R-squared 0.419014 S.D. dependent var 0.059506

    S.E. of regression 0.045357 Akaike info criterion -3.341143

    Sum squared resid 0.551335 Schwarz criterion -3.314488

    Log likelihood 453.0543 Hannan-Quinn criter. -3.330439

    F-statistic 195.0064 Durbin-Watson stat 2.000702

    Prob(F-statistic) 0.000000

    Interpretation: Using intercept at first difference in this analysis, Augmented Dickey-Fuller test statistic

    found significant which proves that the series (SSE-VTX) is stationary at first difference.

    Decision of using ARDL-approach: As both the series i.e. SSE and SSE-VTX are stationary at different

    levels, ARDL approach will be used consequently to study the co-movement of the said series.

  • 7/30/2019 Cointegration Results

    8/9

    ARDL approach Results

    Estimates of ARDL regression:

    Autoregressive Distributed Lag Estimates

    ARDL(1,0) selected based on Schwarz Bayesian Criterion*******************************************************************************

    Dependent variable is X1

    271 observations used for estimation from 1991M2 to 2013M8*******************************************************************************

    Regressor Coefficient Standard Error T-Ratio[Prob]

    X1(-1) .92918 .017501 53.0934[.000]X2 .061173 .014868 4.1146[.000]

    *******************************************************************************

    R-Squared .96531 R-Bar-Squared .96518

    S.E. of Regression .13046 F-stat. F( 1, 269) 7486.0[.000]

    Mean of Dependent Variable 7.2301 S.D. of Dependent Variable .69917

    Residual Sum of Squares 4.5783 Equation Log-likelihood 168.4142Akaike Info. Criterion 166.4142 Schwarz Bayesian Criterion 162.8121

    DW-statistic 2.0675 Durbin's h-statistic -.58019[.562]

    *******************************************************************************

    Diagnostic Tests*******************************************************************************

    * Test Statistics * LM Version * F Version *

    ******************************************************************************** * * *

    * A:Serial Correlation*CHSQ( 12)= 25.5536[.012]*F( 12, 257)= 2.2297[.011]*

    * * * ** B:Functional Form *CHSQ( 1)= .10628[.744]*F( 1, 268)= .10515[.746]*

    * * * *

    * C:Normality *CHSQ( 2)= 3371.6[.000]* Not applicable ** * * *

    * D:Heteroscedasticity*CHSQ( 1)= 9.4588[.002]*F( 1, 269)= 9.7285[.002]*

    *******************************************************************************A:Lagrange multiplier test of residual serial correlation

    B:Ramsey's RESET test using the square of the fitted values

    C:Based on a test of skewness and kurtosis of residualsD:Based on the regression of squared residuals on squared fitted values

  • 7/30/2019 Cointegration Results

    9/9

    Long-run Coefficient:

    Estimated Long Run Coefficients using the ARDL ApproachARDL(1,0) selected based on Schwarz Bayesian Criterion

    *******************************************************************************

    Dependent variable is X1271 observations used for estimation from 1991M2 to 2013M8

    *******************************************************************************Regressor Coefficient Standard Error T-Ratio[Prob]X2 .86380 .013682 63.1340[.000]

    *******************************************************************************

    Error Correction Model:

    Error Correction Representation for the Selected ARDL Model

    ARDL(1,0) selected based on Schwarz Bayesian Criterion*******************************************************************************

    Dependent variable is dX1

    271 observations used for estimation from 1991M2 to 2013M8

    *******************************************************************************

    Regressor Coefficient Standard Error T-Ratio[Prob]dX2 .061173 .014868 4.1146[.000]ecm(-1) -.070819 .017501 -4.0466[.000]

    *******************************************************************************

    List of additional temporary variables created:dX1 = X1-X1(-1)

    dX2 = X2-X2(-1)

    ecm = X1 -.86380*X2*******************************************************************************

    R-Squared .056719 R-Bar-Squared .053212

    S.E. of Regression .13046 F-stat. F( 1, 269) 16.1748[.000]Mean of Dependent Variable .010264 S.D. of Dependent Variable .13408

    Residual Sum of Squares 4.5783 Equation Log-likelihood 168.4142

    Akaike Info. Criterion 166.4142 Schwarz Bayesian Criterion 162.8121DW-statistic 2.0675

    *******************************************************************************R-Squared and R-Bar-Squared measures refer to the dependent variable

    dX1 and in cases where the error correction model is highly

    restricted, these measures could become negative.