EViews assignment – 3rd week

The problems that follow are from the course text book. If it is referred to an EViews work file
that you yet haven’t created, you have to import the data from the corresponding Excel file. This
procedure is described in the text book.
Your solutions must be easy to follow and include figures and tables when prompted. The best
way to present the solutions is to use a Word document.
6.13 (Slightly changed compared to the textbook)
Select one of the stock series from the ‘CAPM.XLS’ Excel file, construct a set of continuously
compounded returns, and then perform a time-series analysis of these returns. The analysis
should include
a) An examination of the autocorrelation and partial autocorrelation functions.
b) An estimation of the information criteria for each ARMA model order from (0,0) to
(2,2), i.e. (0,0), (0,1), (0,2), (1,0), (2,0), (1,1), (1,2), (2,1) and (2,2)
c) An estimation of the model that you feel most appropriate given the results that you
found from the previous two parts of the question.

7.1.e (Slightly changed compared to the textbook)

Using EViews, estimate a VAR model for the interest rate series used in the principal
components example of chapter 4 (i.e. the ‘macro.wf1’ file and the series USTB3M USTB6M
USTB1Y USTB3Y USTB5Y USTB10Y) . Use a method for selecting the lag length in the VAR
optimally. Determine whether certain maturities lead or lag others, by conducting Granger
causality tests.
8.8. (Slightly changed compared to the textbook)
In EViews, open the currencies_ii.wf1 file. Determine whether the exchange rate series (in their
raw levels forms) are non-stationary. If that is the case, test for cointegration between them
using the Engle–Granger approach. Would you have expected the series to cointegrate? Why or
why not?
Question 6.13 A
Examination of the autocorrelation and partial autocorrelation functions of GE

CORRELOGRAM OF RGE

Explanation
The second last column (Q-Stat) gives the Ljung-Box test statistic. The test statistics follows a
χ2(1) for the first row, a χ2(2) for the second row, and so on.
As a rule of thumb, a given autocorrelation coefficient is classed as significant if it is outside a
± 1.96 x 1/(T )1/ 2 band, where T is the number of observations.
±1.96×1/(136 )1/2 =±1.96 ×1/(11.6619) = ±1.96×0.0857 = (1.96×0.0857), (-1.96×0.0857)
= 0.1680, -0.1680
In this case, a correlation coefficient is classed as significant if it is bigger than approximately
0.168 or smaller than − 0.168. It can be deduced that the sixth and seventh of both
autocorrelation coefficients and partial autocorrelation coefficients are significant under this
rule.
Decision
Since the sixth and seventh acf coefficient are highly significant, the Ljung-Box joint test statistic
rejects the null hypothesis of no autocorrelation at the 1% level for all numbers of lags.
Question 6.13 B
An estimation of the information criteria for each ARMA model order from (0,0) to (2,2), i.e.
(0,0), (0,1), (0,2), (1,0), (2,0), (1,1), (1,2), (2,1) and (2,2)

ARMA (0,0)

Dependent Variable: RGE

Method: Least Squares
Date: 11/30/20 Time: 09:17
Sample (adjusted): 2002M02 2013M04
Included observations: 135 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.  

C -0.105447 0.723694 -0.145706 0.8844

R-squared 0.000000    Mean dependent var -0.105447

Adjusted R-squared 0.000000    S.D. dependent var 8.408567
S.E. of regression 8.408567    Akaike info criterion 7.103759
Sum squared resid 9474.337    Schwarz criterion 7.125280
Log likelihood -478.5037    Hannan-Quinn criter. 7.112504
Durbin-Watson stat 1.785250

ARMA (0,1)

Dependent Variable: RGE

Method: ARMA Maximum Likelihood (OPG - BHHH)
Date: 11/30/20 Time: 07:33
Sample: 2002M02 2013M04
Included observations: 135
Convergence achieved after 35 iterations
Coefficient covariance computed using outer product of gradients

Variable Coefficient Std. Error t-Statistic Prob.  

C -0.104788 0.865713 -0.121043 0.9038

MA(1) 0.118610 0.056620 2.094830 0.0381
SIGMASQ 69.28116 6.560939 10.55964 0.0000

R-squared 0.012811    Mean dependent var -0.105447

Adjusted R-squared -0.002146    S.D. dependent var 8.408567
S.E. of regression 8.417585    Akaike info criterion 7.120599
Sum squared resid 9352.957    Schwarz criterion 7.185161
Log likelihood -477.6405    Hannan-Quinn criter. 7.146836
F-statistic 0.856529    Durbin-Watson stat 2.006219
Prob(F-statistic) 0.426980

Inverted MA Roots      -.12

ARMA (0,2)

Dependent Variable: RGE

Method: ARMA Maximum Likelihood (OPG - BHHH)
Date: 11/30/20 Time: 10:45
Sample: 2002M02 2013M04
Included observations: 135
Convergence achieved after 34 iterations
Coefficient covariance computed using outer product of gradients

Variable Coefficient Std. Error t-Statistic Prob.  

C -0.104907 0.850082 -0.123407 0.9020

MA(1) 0.107519 0.057638 1.865410 0.0644
MA(2) -0.031695 0.063423 -0.499745 0.6181
SIGMASQ 69.18273 6.603423 10.47680 0.0000

R-squared 0.014214    Mean dependent var -0.105447

Adjusted R-squared -0.008361    S.D. dependent var 8.408567
S.E. of regression 8.443647    Akaike info criterion 7.133994
Sum squared resid 9339.669    Schwarz criterion 7.220077
Log likelihood -477.5446    Hannan-Quinn criter. 7.168976
F-statistic 0.629625    Durbin-Watson stat 1.990264
Prob(F-statistic) 0.597146

Inverted MA Roots       .13          -.24

ARMA (1,0)
Dependent Variable: RGE
Method: ARMA Maximum Likelihood (OPG - BHHH)
Date: 11/30/20 Time: 07:34
Sample: 2002M02 2013M04
Included observations: 135
Convergence achieved after 24 iterations
Coefficient covariance computed using outer product of gradients

Variable Coefficient Std. Error t-Statistic Prob.  

C -0.104960 0.873445 -0.120167 0.9045

AR(1) 0.105366 0.055860 1.886263 0.0615
SIGMASQ 69.39202 6.526505 10.63234 0.0000

R-squared 0.011232    Mean dependent var -0.105447

Adjusted R-squared -0.003749    S.D. dependent var 8.408567
S.E. of regression 8.424316    Akaike info criterion 7.122176
Sum squared resid 9367.922    Schwarz criterion 7.186738
Log likelihood -477.7469    Hannan-Quinn criter. 7.148412
F-statistic 0.749723    Durbin-Watson stat 1.979902
Prob(F-statistic) 0.474499

Inverted AR Roots       .11

ARMA (2,0)

Dependent Variable: RGE

Method: ARMA Maximum Likelihood (OPG - BHHH)
Date: 11/30/20 Time: 10:46
Sample: 2002M02 2013M04
Included observations: 135
Convergence achieved after 23 iterations
Coefficient covariance computed using outer product of gradients

Variable Coefficient Std. Error t-Statistic Prob.  

C -0.103499 0.824575 -0.125519 0.9003

AR(1) 0.113392 0.057473 1.972958 0.0506
AR(2) -0.074314 0.060999 -1.218275 0.2253
SIGMASQ 68.99904 6.542740 10.54589 0.0000

R-squared 0.016831    Mean dependent var -0.105447

Adjusted R-squared -0.005684    S.D. dependent var 8.408567
S.E. of regression 8.432430    Akaike info criterion 7.131394
Sum squared resid 9314.871    Schwarz criterion 7.217476
Log likelihood -477.3691    Hannan-Quinn criter. 7.166375
F-statistic 0.747551    Durbin-Watson stat 1.999362
Prob(F-statistic) 0.525649

Inverted AR Roots  .06+.27i      .06-.27i

ARMA (1,1)

Dependent Variable: RGE

Method: ARMA Maximum Likelihood (OPG - BHHH)
Date: 11/30/20 Time: 07:21
Sample: 2002M02 2013M04
Included observations: 135
Convergence achieved after 30 iterations
Coefficient covariance computed using outer product of gradients

Variable Coefficient Std. Error t-Statistic Prob.  

C -0.104988 0.861818 -0.121821 0.9032

AR(1) -0.104623 0.564548 -0.185321 0.8533
MA(1) 0.219736 0.562706 0.390499 0.6968
SIGMASQ 69.24129 6.612957 10.47055 0.0000

R-squared 0.013380    Mean dependent var -0.105447

Adjusted R-squared -0.009215    S.D. dependent var 8.408567
S.E. of regression 8.447220    Akaike info criterion 7.134837
Sum squared resid 9347.574    Schwarz criterion 7.220919
Log likelihood -477.6015    Hannan-Quinn criter. 7.169819
F-statistic 0.592164    Durbin-Watson stat 2.002282
Prob(F-statistic) 0.621209

Inverted AR Roots      -.10

Inverted MA Roots      -.22
ARMA (1,2)
Dependent Variable: RGE
Method: ARMA Maximum Likelihood (OPG - BHHH)
Date: 11/30/20 Time: 10:40
Sample: 2002M02 2013M04
Included observations: 135
Convergence achieved after 34 iterations
Coefficient covariance computed using outer product of gradients

Variable Coefficient Std. Error t-Statistic Prob.  

C -0.104183 0.882043 -0.118115 0.9062

AR(1) 0.164678 1.753055 0.093937 0.9253
MA(1) -0.055488 1.752973 -0.031654 0.9748
MA(2) -0.054191 0.173507 -0.312329 0.7553
SIGMASQ 69.14842 6.807569 10.15758 0.0000

R-squared 0.014703    Mean dependent var -0.105447

Adjusted R-squared -0.015614    S.D. dependent var 8.408567
S.E. of regression 8.473959    Akaike info criterion 7.148321
Sum squared resid 9335.037    Schwarz criterion 7.255924
Log likelihood -477.5117    Hannan-Quinn criter. 7.192048
F-statistic 0.484972    Durbin-Watson stat 1.991409
Prob(F-statistic) 0.746740

Inverted AR Roots       .16

Inverted MA Roots       .26          -.21

ARMA (2,1)
Dependent Variable: RGE
Method: ARMA Maximum Likelihood (OPG - BHHH)
Date: 11/30/20 Time: 10:42
Sample: 2002M02 2013M04
Included observations: 135
Convergence achieved after 32 iterations
Coefficient covariance computed using outer product of gradients

Variable Coefficient Std. Error t-Statistic Prob.  

C -0.103188 0.850639 -0.121306 0.9036

AR(1) 0.150713 0.866624 0.173909 0.8622
AR(2) -0.078613 0.090588 -0.867804 0.3871
MA(1) -0.037471 0.871012 -0.043020 0.9658
SIGMASQ 68.99470 6.832507 10.09801 0.0000

R-squared 0.016893    Mean dependent var -0.105447

Adjusted R-squared -0.013356    S.D. dependent var 8.408567
S.E. of regression 8.464534    Akaike info criterion 7.146147
Sum squared resid 9314.285    Schwarz criterion 7.253750
Log likelihood -477.3649    Hannan-Quinn criter. 7.189874
F-statistic 0.558463    Durbin-Watson stat 1.997636
Prob(F-statistic) 0.693207

Inverted AR Roots  .08-.27i      .08+.27i

Inverted MA Roots       .04
ARMA (2,2)
Dependent Variable: RGE
Method: ARMA Maximum Likelihood (OPG - BHHH)
Date: 11/30/20 Time: 10:44
Sample: 2002M02 2013M04
Included observations: 135
Convergence achieved after 29 iterations
Coefficient covariance computed using outer product of gradients

Variable Coefficient Std. Error t-Statistic Prob.  

C -0.112032 0.824234 -0.135923 0.8921

AR(1) -0.179746 0.111165 -1.616940 0.1083
AR(2) -0.841669 0.091376 -9.211090 0.0000
MA(1) 0.328173 0.122992 2.668257 0.0086
MA(2) 0.834065 0.114346 7.294221 0.0000
SIGMASQ 64.87709 5.927220 10.94562 0.0000

R-squared 0.075565    Mean dependent var -0.105447

Adjusted R-squared 0.039734    S.D. dependent var 8.408567
S.E. of regression 8.239820    Akaike info criterion 7.102204
Sum squared resid 8758.407    Schwarz criterion 7.231328
Log likelihood -473.3988    Hannan-Quinn criter. 7.154677
F-statistic 2.108942    Durbin-Watson stat 2.049190
Prob(F-statistic) 0.068391

Inverted AR Roots -.09+.91i     -.09-.91i

Inverted MA Roots -.16+.90i     -.16-.90i
Explanation
Summary of the estimation of the information criteria for each ARMA model order from (0,0) to (2,2)
  ARMA (0,0) ARMA (0,1) ARMA (0,2) ARMA (1,0) ARMA (2,0) ARMA (1,1)ARMA (1,2) ARMA (2,1) ARMA (2,2)
Number of
0 1 0 0 1 0 0 0 3
Sig. coefficient
Sigma2 (Volatility) 0.0000 69.2812 69.1827 69.3920 68.9990 69.2413 69.1484 68.9947 64.8771
2
Adj. R 0.0000 -0.0021 -0.0084 -0.0037 -0.0057 -0.0092 -0.0156 -0.0134 0.0397
Akaike Info C 7.1038 7.1206 7.1340 7.1222 7.1314 7.1348 7.1483 7.1461 7.1022
Schwarz C 7.1253 7.1852 7.2201 7.1867 7.2175 7.2209 7.2559 7.2538 7.2313
Hannan-Q C 7.11250 7.14684 7.16898 7.14841 7.16638 7.16982 7.19205 7.18987 7.15468
Source: Student’s compilation from e-view output

Decision
From the above, I chose ARMA (2,2) model because it fulfils the following conditions.
1. ARMA (2,2) model has the lowest Akaike info criterion, although its Schwarz criterion, and
Hannan-Quinn criterion are not the lowest.
2. ARMA (2,2) model is the model with the highest number of significant estimated
coefficients.
3. ARMA (2,2) model is the model with the least volatility (Sigma 2), this condition enhances
the predictive power of the model.
4. ARMA (2,2) model is the model with the highest adjusted R 2, the model is good because the
included variables have significant role they play in the model as shown by the value of the
adjusted R2.
Question 6.13 C
An estimation of ARMA (2,2) model as the most appropriate.
ARMA ( 2,2 ) : RGE=α + β 1 RGEt −1+ β2 RGEt −2+ δ 0 μ0 +δ 1 μt −1 + δ 2 μ t−2

ARMA (2,2)
Dependent Variable: RGE
Method: ARMA Maximum Likelihood (OPG - BHHH)
Sample: 2002M02 2013M04
Included observations: 135
Convergence achieved after 29 iterations
Coefficient covariance computed using outer product of gradients

Variable Coefficient Std. Error t-Statistic Prob.  

C -0.112032 0.824234 -0.135923 0.8921

AR(1) -0.179746 0.111165 -1.616940 0.1083
AR(2) -0.841669 0.091376 -9.211090 0.0000
MA(1) 0.328173 0.122992 2.668257 0.0086
MA(2) 0.834065 0.114346 7.294221 0.0000
SIGMASQ 64.87709 5.927220 10.94562 0.0000

R-squared 0.075565    Mean dependent var -0.105447

Adjusted R-squared 0.039734    S.D. dependent var 8.408567
S.E. of regression 8.239820    Akaike info criterion 7.102204
Sum squared resid 8758.407    Schwarz criterion 7.231328
Log likelihood -473.3988    Hannan-Quinn criter. 7.154677
F-statistic 2.108942    Durbin-Watson stat 2.049190
Prob(F-statistic) 0.068391

Inverted AR Roots -.09+.91i     -.09-.91i

Inverted MA Roots -.16+.90i     -.16-.90i

The estimated ARMA (2,2) model

ARMA ( 2,2 ) : RGE=−0.112−0.179 RGEt −1−0.842 RGE t −2 +0.328 μt −1+ 0.834 μ t−2

Diagnosis of ARMA (2,2) Model

The Correlogram of the residual of ARMA (2,2) above is flat since all the autocorrelation lies
within the standard error bounds implying there is not really anything left in the data to explain.
Question 7.1e
Use a method for selecting the lag length in the VAR optimally.
Vector Autoregression Estimates
Date: 12/02/20 Time: 10:32
Sample (adjusted): 1986M06 2013M04
Included observations: 323 after adjustments
Standard errors in ( ) & t-statistics in [ ]

RUSTB10Y RUSTB5Y RUSTB3Y RUSTB1Y RUSTB6M RUSTB3M

RUSTB10Y(-1)  0.085822  0.705618  0.642417 -0.451622 -0.973615 -1.829489

 (0.19499)  (0.27266)  (0.33024)  (0.31824)  (0.37057)  (0.68631)
[ 0.44013] [ 2.58790] [ 1.94528] [-1.41913] [-2.62736] [-2.66568]

RUSTB10Y(-2) -0.034995  0.374213  0.669140  0.482703  0.422992  0.342361

 (0.19614)  (0.27427)  (0.33219)  (0.32011)  (0.37275)  (0.69035)
[-0.17842] [ 1.36442] [ 2.01434] [ 1.50792] [ 1.13479] [ 0.49592]

RUSTB5Y(-1) -0.405824 -1.179333 -1.093968  0.375973  0.905808  1.076909

 (0.25718)  (0.35961)  (0.43556)  (0.41972)  (0.48874)  (0.90518)
[-1.57800] [-3.27947] [-2.51164] [ 0.89576] [ 1.85335] [ 1.18972]

RUSTB5Y(-2) -0.227184 -0.573117 -0.729356 -0.975426 -1.232114 -1.518490

 (0.25708)  (0.35948)  (0.43540)  (0.41957)  (0.48856)  (0.90484)
[-0.88371] [-1.59431] [-1.67515] [-2.32483] [-2.52194] [-1.67819]

RUSTB3Y(-1)  0.551370  0.968170  1.017604  0.501136  0.276696  0.923342

 (0.15060)  (0.21059)  (0.25506)  (0.24579)  (0.28621)  (0.53007)
[ 3.66110] [ 4.59746] [ 3.98962] [ 2.03887] [ 0.96677] [ 1.74192]

RUSTB3Y(-2)  0.134023  0.185790  0.161552  0.692519  0.889852  1.518590

 (0.15047)  (0.21040)  (0.25483)  (0.24557)  (0.28595)  (0.52959)
[ 0.89073] [ 0.88304] [ 0.63396] [ 2.82008] [ 3.11195] [ 2.86749]

RUSTB1Y(-1) -0.296018 -0.442149 -0.546789 -0.826826 -0.285963 -0.765635

 (0.08838)  (0.12358)  (0.14968)  (0.14424)  (0.16796)  (0.31107)
[-3.34936] [-3.57775] [-3.65298] [-5.73223] [-1.70257] [-2.46129]

RUSTB1Y(-2) -0.006291  0.002087 -0.090490 -0.777099 -0.676894 -1.470460

 (0.08877)  (0.12413)  (0.15034)  (0.14488)  (0.16870)  (0.31244)
[-0.07087] [ 0.01681] [-0.60190] [-5.36389] [-4.01245] [-4.70639]

RUSTB6M(-1)  0.097445  0.117228  0.059080  0.354092 -0.144142  0.179463

 (0.09093)  (0.12714)  (0.15400)  (0.14840)  (0.17280)  (0.32003)
[ 1.07168] [ 0.92202] [ 0.38365] [ 2.38611] [-0.83416] [ 0.56077]

RUSTB6M(-2) -0.041171 -0.074038  0.061380  0.685698  0.751955  1.659231

 (0.08985)  (0.12564)  (0.15218)  (0.14664)  (0.17076)  (0.31625)
[-0.45820] [-0.58928] [ 0.40335] [ 4.67595] [ 4.40367] [ 5.24657]

RUSTB3M(-1)  0.014256  0.005460  0.044947  0.082186  0.145475  0.000604

 (0.03320)  (0.04642)  (0.05623)  (0.05418)  (0.06309)  (0.11685)
[ 0.42941] [ 0.11762] [ 0.79938] [ 1.51679] [ 2.30572] [ 0.00517]
 RUSTB3M(-2)  0.042112  0.099056  0.105156 -0.016053 -0.047022 -0.317819
 (0.03367)  (0.04709)  (0.05703)  (0.05496)  (0.06400)  (0.11852)
[ 1.25054] [ 2.10364] [ 1.84380] [-0.29208] [-0.73477] [-2.68146]

C -0.411883 -0.684948 -0.903649 -0.958154 -0.857981 -0.765822

 (0.34157)  (0.47762)  (0.57848)  (0.55745)  (0.64912)  (1.20220)
[-1.20587] [-1.43410] [-1.56210] [-1.71880] [-1.32176] [-0.63702]

R-squared  0.140091  0.126039  0.138594  0.285892  0.284893  0.223851

Adj. R-squared  0.106804  0.092208  0.105250  0.258249  0.257212  0.193806
Sum sq. resids  11411.04  22311.65  32730.97  30394.44  41211.94  141361.4
S.E. equation  6.067110  8.483694  10.27540  9.901847  11.53004  21.35426
F-statistic  4.208608  3.725583  4.156413  10.34235  10.29180  7.450651
Log likelihood -1034.014 -1142.304 -1204.193 -1192.232 -1241.403 -1440.467
Akaike AIC  6.483057  7.153584  7.536797  7.462735  7.767203  8.999795
Schwarz SC  6.635099  7.305626  7.688839  7.614777  7.919245  9.151837
Mean dependent -0.429193 -0.696565 -0.921830 -1.191054 -1.259520 -1.351297
S.D. dependent  6.419609  8.904139  10.86295  11.49707  13.37821  23.78290

Determinant resid covariance (dof adj.)  4.41E+08

Determinant resid covariance  3.45E+08
Log likelihood -5924.611
Akaike information criterion  37.16787
Schwarz criterion  38.08012
Number of coefficients  78
 Question 7.1e continuation

The lag length selection in the VAR

VAR Lag Order Selection Criteria
Endogenous variables: RUSTB10Y RUSTB5Y RUSTB3Y RUSTB1Y RUSTB6M
RUSTB3M 
Exogenous variables: C 
Date: 12/02/20 Time: 10:34
Sample: 1986M03 2013M04
Included observations: 315

 Lag LogL LR FPE AIC SC HQ

0 -6039.923 NA   1.89e+09  38.38681  38.45829  38.41537

1 -5908.325  257.3457  1.03e+09  37.77984  38.28019  37.97975
2 -5794.724  217.8268  6.29e+08  37.28714   38.21634*  37.65839
3 -5731.272  119.2492  5.29e+08  37.11284  38.47091  37.65544
4 -5592.708  255.1327  2.76e+08  36.46164  38.24858  37.17559
5 -5495.525  175.2392  1.88e+08  36.07317  38.28898  36.95847
6 -5430.157  115.3799  1.56e+08  35.88671  38.53138  36.94335
7 -5340.157  155.4275  1.11e+08  35.54386  38.61739  36.77185
8 -5277.311  106.1410  94245799  35.37340  38.87580  36.77274
9 -5173.026  172.1521  61483309  34.93985  38.87112  36.51054
10 -5087.265   138.3074*   45188814*   34.62390*  38.98403   36.36594*

 * indicates lag order selected by the criterion

 LR: sequential modified LR test statistic (each test at 5% level)
 FPE: Final prediction error
 AIC: Akaike information criterion
 SC: Schwarz information criterion
 HQ: Hannan-Quinn information criterion

Decision
From the VAR Lag Order Selection Criteria result above, I selected a tenth order as optimal since
both the Akaike’s and the Hannan-Quinn criteria select the tenth order as optimal.
Question 7.1e continuation
Determine whether certain maturities lead or lag others, by conducting Granger
causality tests.

Granger causality test

VAR Granger Causality/Block Exogeneity Wald Tests

Date: 12/02/20 Time: 10:50
Sample: 1986M03 2013M04
Included observations: 323

Dependent variable: RUSTB10Y

Excluded Chi-sq df Prob.

RUSTB5Y  2.714096 2  0.2574

RUSTB3Y  13.57440 2  0.0011
RUSTB1Y  12.63513 2  0.0018
RUSTB6M  1.928596 2  0.3813
RUSTB3M  1.621148 2  0.4446

All  34.90158 10  0.0001

Dependent variable: RUSTB5Y

Excluded Chi-sq df Prob.

RUSTB10Y  7.071421 2  0.0291

RUSTB3Y  21.21574 2  0.0000
RUSTB1Y  14.67836 2  0.0006
RUSTB6M  1.786698 2  0.4093
RUSTB3M  4.469166 2  0.1070

All  33.56618 10  0.0002

Dependent variable: RUSTB3Y

Excluded Chi-sq df Prob.

RUSTB10Y  6.024864 2  0.0492

RUSTB5Y  7.374172 2  0.0250
RUSTB1Y  13.88803 2  0.0010
RUSTB6M  0.230416 2  0.8912
RUSTB3M  3.671195 2  0.1595

All  43.02686 10  0.0000

Dependent variable: RUSTB1Y

Excluded Chi-sq df Prob.

RUSTB10Y  6.139318 2  0.0464

RUSTB5Y  7.920462 2  0.0191
RUSTB3Y  10.78485 2  0.0046
RUSTB6M  22.53680 2  0.0000
RUSTB3M  2.584172 2  0.2747

All  123.6365 10  0.0000

Dependent variable: RUSTB6M

Excluded Chi-sq df Prob.

RUSTB10Y  10.99280 2  0.0041

RUSTB5Y  13.31151 2  0.0013
RUSTB3Y  10.00085 2  0.0067
RUSTB1Y  16.19178 2  0.0003
RUSTB3M  6.534476 2  0.0381

All  97.01278 10  0.0000

Dependent variable: RUSTB3M

Excluded Chi-sq df Prob.

RUSTB10Y  8.967244 2  0.0113

RUSTB5Y  5.737660 2  0.0568
RUSTB3Y  10.11868 2  0.0063
RUSTB1Y  22.87633 2  0.0000
RUSTB6M  29.31055 2  0.0000

All  84.82398 10  0.0000

Explanation of Granger Causality result

There is a causality from US three-year treasury bill (RUSTB3Y) and US one-year treasury bill
(RUSTB1Y) to US ten-year treasury bill (RUSTB10Y) and also from US ten-year treasury bill
(RUSTB10Y) to both the US three-year treasury bill (RUSTB3Y) and US one-year treasury bill
(RUSTB1Y).
There is a causality from US ten-year treasury bill (RUSTB10Y), US three-year treasury bill
(RUSTB3Y), and US one-year treasury bill (RUSTB1Y) to US five-year treasury bill
(RUSTB5Y) and also from US five-year treasury bill (RUSTB5Y) to the US three-year treasury
bill (RUSTB3Y) and the US one-year treasury bill (RUSTB1Y).
There is a causality from the US three-year treasury bill (RUSTB3Y) and the US six-month
treasury bill (RUSTB6M) to the US one-year treasury bill (RUSTB1Y) and also from the US
one-year treasury bill (RUSTB1Y) to the US six-month treasury bill (RUSTB6M) and the US
three-year treasury bill (RUSTB3Y).
There is a causality from the US ten-year treasury bill (RUSTB10Y), the US five-year treasury
bill (RUSTB5Y), the US three-year treasury bill (RUSTB3Y) and the US three-month treasury
bill (RUSTB3M) to the US six-month treasury bill (RUSTB6M).

There is a causality from the US ten-year treasury bill (RUSTB10Y), the US three-year treasury
bill (RUSTB3Y), and the US one-year treasury bill (RUSTB1Y) to the US three-month treasury
bill (RUSTB3M).
Hence, there is bi-directional causality between the US three-year treasury bill
(RUSTB3Y) and the US ten-year treasury bill (RUSTB10Y), the US one-year treasury bill
(RUSTB1Y) and the US ten-year treasury bill (RUSTB10Y), the US three-year treasury bill
(RUSTB3Y) and the US five-year treasury bill (RUSTB5Y), the US one-year treasury bill
(RUSTB1Y) and the US five-year treasury bill (RUSTB5Y), t he US three-year treasury bill
(RUSTB3Y) and the US one-year treasury bill (RUSTB1Y), and between the US six-month
treasury bill (RUSTB6M) and the US one-year treasury bill (RUSTB1Y). This mean causality
runs in both directions between these variables as stated above.
But no causality between the US five-year treasury bill (RUSTB5Y) and the US three-
month treasury bill (RUSTB3M) in any direction.

Summary of the Granger causality test

Causality from RUSTB3Y and RUSTB1Y to RUSTB10Y
Causality from RUSTB10Y, RUSTB3Y and RUSTB1Y to RUSTB5Y
Causality from RUSTB10Y, RUSTB5Y and RUSTB1Y to RUSTB3Y
Causality from RUSTB10Y, RUSTB5Y, RUSTB3Y and RUSTB6M to RUSTB1Y
Causality from RUSTB10Y, RUSTB5Y, RUSTB3Y, RUSTB1Y and RUSTB3M to RUSTB6M
Causality from RUSTB10Y, RUSTB3Y, RUSTB1Y and RUSTB6M to RUSTB3M
Causality goes in both directions RUSTB3Y AND RUSTB10Y
Causality goes in both directions RUSTB1Y AND RUSTB10Y
Causality goes in both directions RUSTB3Y AND 5Y
Causality goes in both directions RUSTB3Y AND RUSTB1Y
Causality goes in both directions RUSTB6M AND RUSTB1Y
Causality goes in both directions RUSTB3M AND RUSTB6M
No causality in any direction between RUSTB5Y and RUSTB3M
Question 8.8
Determine whether the exchange rate series (in their raw levels forms) are non-stationary. If
that is the case, test for cointegration between them using the Engle–Granger approach. Would
you have expected the series to cointegrate? Why or why not?

UNIT ROOT TEST

Augmented Dickey-Fuller Test Equation
EUR
@LEVEL

Null Hypothesis: EUR has a unit root

Exogenous: Constant
Lag Length: 2 (Automatic - based on SIC, maxlag=12)

t-Statistic   Prob.*

Augmented Dickey-Fuller test statistic -2.001725  0.2863

Test critical values: 1% level -3.434624
5% level -2.863315
10% level -2.567763

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

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(EUR)
Method: Least Squares
Date: 12/03/20 Time: 10:02
Sample (adjusted): 6/09/2009 6/06/2013
Included observations: 1459 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.  

EUR(-1) -0.004109 0.002053 -2.001725 0.0455

D(EUR(-1)) 0.332404 0.026071 12.74989 0.0000
D(EUR(-2)) -0.088199 0.026053 -3.385435 0.0007
C 0.003081 0.001528 2.015854 0.0440

R-squared 0.102217    Mean dependent var 3.31E-05

Adjusted R-squared 0.100365    S.D. dependent var 0.003201
S.E. of regression 0.003036    Akaike info criterion -8.753887
Sum squared resid 0.013410    Schwarz criterion -8.739396
Log likelihood 6389.961    Hannan-Quinn criter. -8.748481
F-statistic 55.21936    Durbin-Watson stat 1.997158
Prob(F-statistic) 0.000000
@1ST DIFF

Null Hypothesis: D(EUR) has a unit root

Exogenous: Constant
Lag Length: 1 (Automatic - based on SIC, maxlag=12)

t-Statistic   Prob.*

Augmented Dickey-Fuller test statistic -24.68641  0.0000

Test critical values: 1% level -3.434624
5% level -2.863315
10% level -2.567763

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

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(EUR,2)
Method: Least Squares
Date: 12/03/20 Time: 10:04
Sample (adjusted): 6/09/2009 6/06/2013
Included observations: 1459 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.  

D(EUR(-1)) -0.759799 0.030778 -24.68641 0.0000

D(EUR(-1),2) 0.091056 0.026040 3.496739 0.0005
C 2.57E-05 7.96E-05 0.323418 0.7464

R-squared 0.353610    Mean dependent var -4.11E-07

Adjusted R-squared 0.352723    S.D. dependent var 0.003777
S.E. of regression 0.003039    Akaike info criterion -8.752508
Sum squared resid 0.013447    Schwarz criterion -8.741640
Log likelihood 6387.954    Hannan-Quinn criter. -8.748454
F-statistic 398.2558    Durbin-Watson stat 1.997562
Prob(F-statistic) 0.000000

GBP @ LEVEL

Null Hypothesis: GBP has a unit root

Exogenous: Constant
Lag Length: 2 (Automatic - based on SIC, maxlag=12)

t-Statistic   Prob.*

Augmented Dickey-Fuller test statistic -2.589767  0.0953

Test critical values: 1% level -3.434624
5% level -2.863315
10% level -2.567763

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

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(GBP)
Method: Least Squares
Date: 12/03/20 Time: 10:05
Sample (adjusted): 6/09/2009 6/06/2013
Included observations: 1459 after adjustments
 Variable Coefficient Std. Error t-Statistic Prob.  

GBP(-1) -0.007908 0.003054 -2.589767 0.0097

D(GBP(-1)) 0.312045 0.025904 12.04632 0.0000
D(GBP(-2)) -0.138504 0.025960 -5.335311 0.0000
C 0.005016 0.001932 2.596372 0.0095

R-squared 0.096564    Mean dependent var 1.75E-05

Adjusted R-squared 0.094701    S.D. dependent var 0.002322
S.E. of regression 0.002209    Akaike info criterion -9.389401
Sum squared resid 0.007103    Schwarz criterion -9.374911
Log likelihood 6853.568    Hannan-Quinn criter. -9.383995
F-statistic 51.83922    Durbin-Watson stat 1.996447
Prob(F-statistic) 0.000000

GBP @ 1ST DIFF

Null Hypothesis: D(GBP) has a unit root
Exogenous: Constant
Lag Length: 1 (Automatic - based on SIC, maxlag=12)

t-Statistic   Prob.*

Augmented Dickey-Fuller test statistic -26.64198  0.0000

Test critical values: 1% level -3.434624
5% level -2.863315
10% level -2.567763

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

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(GBP,2)
Method: Least Squares
Date: 12/03/20 Time: 10:08
Sample (adjusted): 6/09/2009 6/06/2013
Included observations: 1459 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.  

D(GBP(-1)) -0.834694 0.031330 -26.64198 0.0000

D(GBP(-1),2) 0.144095 0.025921 5.559079 0.0000
C 1.50E-05 5.80E-05 0.258877 0.7958

R-squared 0.377868    Mean dependent var -1.23E-06

Adjusted R-squared 0.377013    S.D. dependent var 0.002805
S.E. of regression 0.002214    Akaike info criterion -9.386173
Sum squared resid 0.007136    Schwarz criterion -9.375305
Log likelihood 6850.213    Hannan-Quinn criter. -9.382119
F-statistic 442.1690    Durbin-Watson stat 1.997708
Prob(F-statistic) 0.000000
JPY @ LEVEL

Null Hypothesis: JPY has a unit root

Exogenous: Constant
Lag Length: 1 (Automatic - based on SIC, maxlag=12)

t-Statistic   Prob.*

Augmented Dickey-Fuller test statistic -1.337897  0.6137

Test critical values: 1% level -3.434621
5% level -2.863313
10% level -2.567763

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

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(JPY)
Method: Least Squares
Date: 12/03/20 Time: 10:09
Sample (adjusted): 6/08/2009 6/06/2013
Included observations: 1460 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.  

JPY(-1) -0.001889 0.001412 -1.337897 0.1811

D(JPY(-1)) 0.255410 0.025208 10.13209 0.0000
C 0.161201 0.120714 1.335396 0.1820

R-squared 0.066367    Mean dependent var 0.000661

Adjusted R-squared 0.065085    S.D. dependent var 0.372010
S.E. of regression 0.359700    Akaike info criterion 0.794961
Sum squared resid 188.5128    Schwarz criterion 0.805823
Log likelihood -577.3216    Hannan-Quinn criter. 0.799013
F-statistic 51.78481    Durbin-Watson stat 1.963717
Prob(F-statistic) 0.000000

JPY @ 1ST DIFF

Null Hypothesis: D(JPY) has a unit root

Exogenous: Constant
Lag Length: 0 (Automatic - based on SIC, maxlag=12)

t-Statistic   Prob.*

Augmented Dickey-Fuller test statistic -29.60186  0.0000

Test critical values: 1% level -3.434621
5% level -2.863313
10% level -2.567763

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

Kwiatkowski-Philips-Schmidt-Shin (KPSS)
EUR at Level
Null Hypothesis: EUR is stationary
Exogenous: Constant
Bandwidth: 31 (Newey-West automatic) using Bartlett kernel

LM-Stat.

Kwiatkowski-Phillips-Schmidt-Shin test statistic  1.427329

Asymptotic critical values*: 1% level  0.739000
5% level  0.463000
10% level  0.347000

*Kwiatkowski-Phillips-Schmidt-Shin (1992, Table 1) 

Residual variance (no correction)  0.001505

HAC corrected variance (Bartlett kernel)  0.045341

KPSS Test Equation

Dependent Variable: EUR
Method: Least Squares
Date: 12/06/20 Time: 19:43
Sample: 6/06/2009 6/06/2013
Included observations: 1462

Variable Coefficient Std. Error t-Statistic Prob.  

C 0.743577 0.001015 732.6459 0.0000

R-squared 0.000000    Mean dependent var 0.743577

Adjusted R-squared -0.000000    S.D. dependent var 0.038807
S.E. of regression 0.038807    Akaike info criterion -3.659769
Sum squared resid 2.200197    Schwarz criterion -3.656152
Log likelihood 2676.291    Hannan-Quinn criter. -3.658420
Durbin-Watson stat 0.006820

EUR @ 1st Diff

Null Hypothesis: D(EUR) is stationary
Exogenous: Constant
Bandwidth: 2 (Newey-West automatic) using Bartlett kernel

LM-Stat.

Kwiatkowski-Phillips-Schmidt-Shin test statistic  0.055039

Asymptotic critical values*: 1% level  0.739000
5% level  0.463000
10% level  0.347000

*Kwiatkowski-Phillips-Schmidt-Shin (1992, Table 1) 

Residual variance (no correction)  1.03E-05

HAC corrected variance (Bartlett kernel)  1.45E-05
KPSS Test Equation
Dependent Variable: D(EUR)
Method: Least Squares
Date: 12/06/20 Time: 19:44
Sample (adjusted): 6/07/2009 6/06/2013
Included observations: 1461 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.  

C 3.87E-05 8.39E-05 0.461107 0.6448

R-squared 0.000000    Mean dependent var 3.87E-05

Adjusted R-squared 0.000000    S.D. dependent var 0.003206
S.E. of regression 0.003206    Akaike info criterion -8.647093
Sum squared resid 0.015004    Schwarz criterion -8.643474
Log likelihood 6317.701    Hannan-Quinn criter. -8.645743
Durbin-Watson stat 1.391039

GBP at Level
Null Hypothesis: GBP is stationary
Exogenous: Constant
Bandwidth: 30 (Newey-West automatic) using Bartlett kernel

LM-Stat.

Kwiatkowski-Phillips-Schmidt-Shin test statistic  0.314274

Asymptotic critical values*: 1% level  0.739000
5% level  0.463000
10% level  0.347000

*Kwiatkowski-Phillips-Schmidt-Shin (1992, Table 1) 

Residual variance (no correction)  0.000362

HAC corrected variance (Bartlett kernel)  0.009938

KPSS Test Equation

Dependent Variable: GBP
Method: Least Squares
Date: 12/06/20 Time: 19:47
Sample: 6/06/2009 6/06/2013
Included observations: 1462

Variable Coefficient Std. Error t-Statistic Prob.  

C 0.632376 0.000498 1269.666 0.0000

R-squared 0.000000    Mean dependent var 0.632376

Adjusted R-squared 0.000000    S.D. dependent var 0.019044
S.E. of regression 0.019044    Akaike info criterion -5.083440
Sum squared resid 0.529869    Schwarz criterion -5.079824
Log likelihood 3716.995    Hannan-Quinn criter. -5.082091
Durbin-Watson stat 0.014863
GBP at 1st Diff

Null Hypothesis: D(GBP) is stationary

Exogenous: Constant
Bandwidth: 5 (Newey-West automatic) using Bartlett kernel

LM-Stat.

Kwiatkowski-Phillips-Schmidt-Shin test statistic  0.038220

Asymptotic critical values*: 1% level  0.739000
5% level  0.463000
10% level  0.347000

*Kwiatkowski-Phillips-Schmidt-Shin (1992, Table 1) 

Residual variance (no correction)  5.39E-06

HAC corrected variance (Bartlett kernel)  7.08E-06

KPSS Test Equation

Dependent Variable: D(GBP)
Method: Least Squares
Date: 12/06/20 Time: 19:48
Sample (adjusted): 6/07/2009 6/06/2013
Included observations: 1461 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.  

C 2.01E-05 6.08E-05 0.330063 0.7414

R-squared 0.000000    Mean dependent var 2.01E-05

Adjusted R-squared 0.000000    S.D. dependent var 0.002322
S.E. of regression 0.002322    Akaike info criterion -9.291705
Sum squared resid 0.007875    Schwarz criterion -9.288086
Log likelihood 6788.590    Hannan-Quinn criter. -9.290355
Durbin-Watson stat 1.458033

JPY at Level
Null Hypothesis: JPY is stationary
Exogenous: Constant
Bandwidth: 31 (Newey-West automatic) using Bartlett kernel

LM-Stat.

Kwiatkowski-Phillips-Schmidt-Shin test statistic  1.320787

Asymptotic critical values*: 1% level  0.739000
5% level  0.463000
10% level  0.347000

*Kwiatkowski-Phillips-Schmidt-Shin (1992, Table 1) 

Residual variance (no correction)  44.69918

HAC corrected variance (Bartlett kernel)  1352.377
KPSS Test Equation
Dependent Variable: JPY
Method: Least Squares
Date: 12/06/20 Time: 19:49
Sample: 6/06/2009 6/06/2013
Included observations: 1462

Variable Coefficient Std. Error t-Statistic Prob.  

C 85.25612 0.174914 487.4172 0.0000

R-squared 0.000000    Mean dependent var 85.25612

Adjusted R-squared 0.000000    S.D. dependent var 6.688032
S.E. of regression 6.688032    Akaike info criterion 6.639200
Sum squared resid 65350.20    Schwarz criterion 6.642817
Log likelihood -4852.255    Hannan-Quinn criter. 6.640549
Durbin-Watson stat 0.003122

JPY at 1st Diff

Null Hypothesis: D(JPY) is stationary
Exogenous: Constant
Bandwidth: 7 (Newey-West automatic) using Bartlett kernel

LM-Stat.

Kwiatkowski-Phillips-Schmidt-Shin test statistic  0.797834

Asymptotic critical values*: 1% level  0.739000
5% level  0.463000
10% level  0.347000

*Kwiatkowski-Phillips-Schmidt-Shin (1992, Table 1) 

Residual variance (no correction)  0.139626

HAC corrected variance (Bartlett kernel)  0.191106

KPSS Test Equation

Dependent Variable: D(JPY)
Method: Least Squares
Date: 12/06/20 Time: 19:50
Sample (adjusted): 6/07/2009 6/06/2013
Included observations: 1461 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.  

C 0.001648 0.009779 0.168552 0.8662

R-squared 0.000000    Mean dependent var 0.001648

Adjusted R-squared 0.000000    S.D. dependent var 0.373794
S.E. of regression 0.373794    Akaike info criterion 0.870460
Sum squared resid 203.9941    Schwarz criterion 0.874079
Log likelihood -634.8713    Hannan-Quinn criter. 0.871810
Durbin-Watson stat 1.481335
Decision on the Unit-root test using Augmented Dickey-Fuller test and KPSS test equation.

Augmented Dickey-Fuller Test

Test at Levels Test at 1st difference Inference
Variables
ADF statistic t-Statistic Prob.* ADF statistic t-Statistic Prob.*
EUR -2.001725 -2.863315 0.2863 -24.68641 -2.863315 0.0000 I(1)
GBP -2.589767 -2.863315 0.0953 -26.64198 -2.863315 0.0000 I(1)
JPY -1.337897 -2.863313 0.6137 -29.60186 -2.863313 0.0000 I(1)
Source: Student’s calculations from E-views software

Kwiatkowski-Philips-Schmidt-Shin (KPSS ) Test Equation

Test at Levels Test at 1st difference
Variables Inference
KPSS statistic LM-Statistic KPSS statistic LM-Statistic
EUR 0.463000 1.427329 0.463000 0.055039 I(1)
GBP 0.463000 0.314274 0.463000 0.038220 I(0)
JPY 0.463000 1.320787 0.463000 0.797834 I(1)
Source: Student’s calculations from E-views software

Decision Rule for rejecting Null Hypothesis

 Augmented Dickey-Fuller Test
If t-stat is more negative than critical value => the null hypothesis of unit root can be rejected

 Kwiatkowski-Philips-Schmidt-Shin (KPSS) Test Equation

If test statistic exceeds the critical value, the null hypothesis of a stationary series should be
rejected.

Decision of Unit Root Test at Level

KPSS H0: Null Hypothesis: is stationary
ADF H0: Null Hypothesis: has a unit root
ADF KPSS Conclusion
EUR Do not reject H0 Reject H0 Non-stationary
GBP Do not reject H0 Do not reject H0 Inconclusive
JPY Do not reject H0 Reject H0 Non-stationary

Decision of Unit Root Test at 1st Diff

KPSS H0: Null Hypothesis: is stationary
ADF H0: Null Hypothesis: has a unit root
ADF KPSS Conclusion
EUR Reject H0 Do not reject H0 Stationary
GBP Reject H0 Do not reject H0 Stationary
JPY Reject H0 Reject H0 Inconclusive
To test for cointegration between them using the Engle–Granger approach
Using log of the variables i.e.
LEUR=> LOG(EUR)
LGBP=>LOG(GBP)
LJPY=>LOG(JPY)

Augmented Dickey-Fuller Unit Root Test on Regression Error term

Null Hypothesis: STATRESIDS has a unit root
Exogenous: Constant
Lag Length: 2 (Automatic - based on SIC, maxlag=12)

t-Statistic   Prob.*

Augmented Dickey-Fuller test statistic -3.363186  0.0125

Test critical values: 1% level -3.434624
5% level -2.863315
10% level -2.567763

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

Augmented Dickey-Fuller Test Equation
Dependent Variable: D(STATRESIDS)
Method: Least Squares
Date: 12/06/20 Time: 20:44
Sample (adjusted): 6/09/2009 6/06/2013
Included observations: 1459 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.  

STATRESIDS(-1) -0.012724 0.003783 -3.363186 0.0008

D(STATRESIDS(-1)) 0.344802 0.025747 13.39196 0.0000
D(STATRESIDS(-2)) -0.164078 0.025862 -6.344413 0.0000
C 6.41E-06 7.20E-05 0.089054 0.9291

R-squared 0.118801    Mean dependent var 7.78E-06

Adjusted R-squared 0.116984    S.D. dependent var 0.002926
S.E. of regression 0.002749    Akaike info criterion -8.952188
Sum squared resid 0.010998    Schwarz criterion -8.937697
Log likelihood 6534.621    Hannan-Quinn criter. -8.946782
F-statistic 65.38636    Durbin-Watson stat 1.993708
Prob(F-statistic) 0.000000

Date: 12/03/20 Time: 10:14

Series: EUR GBP JPY 
Sample: 6/06/2009 6/06/2013
Included observations: 1462
Null hypothesis: Series are not cointegrated
Cointegrating equation deterministics: C 
Automatic lags specification based on Schwarz criterion (maxlag=23)

Dependent tau-statistic Prob.* z-statistic Prob.*

EUR -2.639250  0.4109 -14.20147  0.3681
GBP -3.329879  0.1310 -22.15785  0.1076
JPY -2.014683  0.7276 -8.259640  0.7196

*MacKinnon (1996) p-values.

Intermediate Results:
EUR GBP JPY
Rho - 1 -0.007738 -0.012519 -0.004483
Rho S.E.  0.002932  0.003760  0.002225
Residual variance  8.07E-06  3.06E-06  0.256400
Long-run residual variance  1.28E-05  4.51E-06  0.408805
Number of lags  2  2  2
Number of observations  1459  1459  1459
Number of stochastic trends**  3  3  3

**Number of stochastic trends in asymptotic distribution

Decision
The test statistic (-3.363186) is more negative than the critical values (at the 5% and 10% level),
therefore the null hypothesis of a unit root in the test regression residuals can be rejected.
Conclusion: The series are cointegrated.