Вы находитесь на странице: 1из 9

We present a preliminary analysis of one of the variables we are going to analyze, the Greek

GDP (1970-2005).
Firstly, we show you some descriptive statistics and the line graph with the evolution of the
series

2.50E+11

2.00E+11

1.50E+11

1.00E+11

5.00E+10

0.00E+00
1970 1975 1980 1985 1990 1995 2000 2005

GRCGDP

20
Series: GRCGDP
Sample 1970 2009
16 Observations 40

Mean 7.44e+10
12 Median 4.01e+10
Maximum 2.41e+11
Minimum 1.11e+09
8 Std. Dev. 7.85e+10
Skewness 0.807969
Kurtosis 2.296607
4
Jarque-Bera 5.176695
Probability 0.075144
0
0.00000 1.0e+11 2.0e+11
Then, here there are the correlograms, to see wether the series is stationary or not.
The correlogram of the series is:

Date: 05/05/10
Time: 19:12
Sample: 1970 2009
Included
observations: 40

Autocorrelation Partial Correlation AC   PAC  Q-Stat  Prob

      . |*******|       . |*******| 1 0.926 0.926 36.961 0.000


      . |****** |       .*| . | 2 0.844 -0.097 68.469 0.000
      . |****** |       . | . | 3 0.762 -0.041 94.839 0.000
      . |***** |       . | . | 4 0.683 -0.028 116.59 0.000
      . |***** |       . | . | 5 0.606 -0.032 134.20 0.000
      . |**** |       . | . | 6 0.528 -0.056 147.97 0.000
      . |*** |       . | . | 7 0.453 -0.033 158.40 0.000
      . |*** |       . | . | 8 0.382 -0.021 166.05 0.000
      . |** |       . | . | 9 0.312 -0.047 171.33 0.000
      . |** |       . | . | 10 0.244 -0.043 174.65 0.000
      . |*. |       . | . | 11 0.177 -0.042 176.48 0.000
      . |*. |       . | . | 12 0.112 -0.054 177.22 0.000
      . | . |       . | . | 13 0.048 -0.043 177.37 0.000
      . | . |       . | . | 14 -0.012 -0.037 177.38 0.000
      .*| . |       . | . | 15 -0.069 -0.042 177.70 0.000
      .*| . |       . | . | 16 -0.122 -0.036 178.74 0.000
      .*| . |       . | . | 17 -0.170 -0.033 180.86 0.000
      **| . |       . | . | 18 -0.215 -0.040 184.40 0.000
      **| . |       . | . | 19 -0.255 -0.028 189.61 0.000
      **| . |       . | . | 20 -0.288 -0.014 196.56 0.000

It seems that coefficients decrease rapidly, like in an AR(1), showing possible stationariety.
Here there are the correlograms of the first and second difference.

Date: 05/05/10
Time: 19:12
Sample: 1970 2009
Included
observations: 39

Autocorrelation Partial Correlation AC   PAC  Q-Stat  Prob

     . |****** |      . |****** | 1 0.832 0.832 29.107 0.000


     . |***** |      . | . | 2 0.711 0.062 50.940 0.000
     . |***** |      . |** | 3 0.685 0.257 71.774 0.000
     . |***** |      . | . | 4 0.643 0.014 90.645 0.000
     . |**** |      . *| . | 5 0.535 -0.163 104.12 0.000
     . |*** |      . | . | 6 0.455 -0.027 114.13 0.000
     . |*** |      . | . | 7 0.399 -0.048 122.08 0.000
     . |** |      . *| . | 8 0.327 -0.061 127.58 0.000
     . |** |      . |*. | 9 0.290 0.111 132.05 0.000
     . |** |      . *| . | 10 0.241 -0.060 135.25 0.000
     . |*. |      . *| . | 11 0.168 -0.080 136.87 0.000
     . |*. |      . | . | 12 0.122 0.010 137.75 0.000
     . | . |      . *| . | 13 0.052 -0.188 137.92 0.000
     . | . |      . *| . | 14 -0.037 -0.117 138.01 0.000
     . *| . |      . | . | 15 -0.084 0.039 138.48 0.000
     . *| . |      . *| . | 16 -0.138 -0.121 139.80 0.000

Date: 05/05/10
Time: 19:12
Sample: 1970 2009
Included
observations: 38

Autocorrelation Partial Correlation AC   PAC  Q-Stat  Prob

     . | . |      . | . | 1 0.027 0.027 0.0298 0.863


     .**| . |      .**| . | 2 -0.283 -0.284 3.4177 0.181
     . | . |      . | . | 3 0.038 0.060 3.4790 0.323
     . |*. |      . | . | 4 0.141 0.061 4.3634 0.359
     . | . |      . | . | 5 -0.026 -0.009 4.3956 0.494
     . | . |      . | . | 6 -0.050 0.005 4.5166 0.607
     . | . |      . *| . | 7 -0.037 -0.059 4.5829 0.711
     . *| . |      . *| . | 8 -0.122 -0.151 5.3413 0.721
     . | . |      . | . | 9 0.040 0.037 5.4273 0.796
     . | . |      . | . | 10 0.050 -0.019 5.5619 0.851
     . | . |      . | . | 11 -0.048 -0.010 5.6899 0.893
     . |*. |      . |*. | 12 0.097 0.149 6.2390 0.904
     . |*. |      . | . | 13 0.066 0.023 6.5017 0.926
     . *| . |      . *| . | 14 -0.133 -0.100 7.6245 0.908
     . | . |      . | . | 15 0.028 0.059 7.6773 0.936
     . | . |      . *| . | 16 0.058 -0.063 7.9064 0.952
After looking at the correlograms, it seems that the GDP series is stationary.

Then, we have done the Augmented Dickey Fuller test to look more analytically at the possible
existence of a unit root. We have done the test with and without trend and intercept. Only
without trend and intercept we find that there is no unit root.

ADF tests: Without intercept and trend

Null Hypothesis:
GRCGDP has a unit
root
Exogenous: None
Lag Length: 3
(Automatic based on
AIC, MAXLAG=3)

t-Statistic   Prob.*

Augmented Dickey-
Fuller test statistic -1.804679  0.0680
Test critical values: 1% level -2.630762
5% level -1.950394
10% level -1.611202

*MacKinnon (1996)
one-sided p-values.
Augmented Dickey-
Fuller Test Equation
Dependent Variable:
D(GRCGDP)
Method: Least Squares
Date: 05/05/10 Time:
19:23
Sample (adjusted):
1974 2009
Included observations:
36 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.  

GRCGDP(-1) -0.029362 0.016270 -1.804679 0.0805


D(GRCGDP(-1)) 1.020553 0.226193 4.511872 0.0001
D(GRCGDP(-2)) -0.400187 0.287184 -1.393487 0.1731
D(GRCGDP(-3)) 0.762160 0.263871 2.888381 0.0069

    Mean
dependent
R-squared 0.764083 var 6.63E+09
    S.D.
dependent
Adjusted R-squared 0.741966 var 4.87E+09
    Akaike
S.E. of regression 2.47E+09 info criterion 46.19848
    Schwarz
Sum squared resid 1.95E+20 criterion 46.37443
    Durbin-
Log likelihood -827.5727 Watson stat 1.681752
With intercept, without trend:

Null Hypothesis:
GRCGDP has a unit
root
Exogenous: Constant
Lag Length: 3
(Automatic based on
AIC, MAXLAG=3)

t-Statistic   Prob.*

Augmented Dickey-
Fuller test statistic -1.255189  0.6394
Test critical values: 1% level -3.626784
5% level -2.945842
10% level -2.611531

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

Augmented Dickey-
Fuller Test Equation
Dependent Variable:
D(GRCGDP)
Method: Least Squares
Date: 05/05/10 Time:
19:26
Sample (adjusted):
1974 2009
Included observations:
36 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.  

GRCGDP(-1) -0.022647 0.018043 -1.255189 0.2188


D(GRCGDP(-1)) 0.943990 0.243303 3.879890 0.0005
D(GRCGDP(-2)) -0.437987 0.291460 -1.502732 0.1430
D(GRCGDP(-3)) 0.718808 0.269436 2.667824 0.0120
C 6.78E+08 7.75E+08 0.875033 0.3883

    Mean
dependent
R-squared 0.769770 var 6.63E+09
    S.D.
dependent
Adjusted R-squared 0.740063 var 4.87E+09
    Akaike
S.E. of regression 2.48E+09 info criterion 46.22964
    Schwarz
Sum squared resid 1.91E+20 criterion 46.44957
Log likelihood -827.1335     F-statistic 25.91195
    Prob(F-
Durbin-Watson stat 1.668449 statistic) 0.000000
With trend and intercept:

Null Hypothesis:
GRCGDP has a unit
root
Exogenous: Constant,
Linear Trend
Lag Length: 3
(Automatic based on
AIC, MAXLAG=3)

t-Statistic   Prob.*

Augmented Dickey-
Fuller test statistic -1.951980  0.6069
Test critical values: 1% level -4.234972
5% level -3.540328
10% level -3.202445

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

Augmented Dickey-
Fuller Test Equation
Dependent Variable:
D(GRCGDP)
Method: Least Squares
Date: 05/05/10 Time:
19:26
Sample (adjusted):
1974 2009
Included observations:
36 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.  

GRCGDP(-1) -0.035708 0.018293 -1.951980 0.0603


D(GRCGDP(-1)) 0.648620 0.272129 2.383500 0.0237
D(GRCGDP(-2)) -0.579272 0.285582 -2.028388 0.0515
D(GRCGDP(-3)) 0.466836 0.283903 1.644347 0.1105
C -2.70E+09 1.80E+09 -1.504780 0.1428
@TREND(1970) 4.03E+08 1.95E+08 2.063564 0.0478

    Mean
dependent
R-squared 0.798387 var 6.63E+09
    S.D.
dependent
Adjusted R-squared 0.764785 var 4.87E+09
    Akaike
S.E. of regression 2.36E+09 info criterion 46.15246
    Schwarz
Sum squared resid 1.67E+20 criterion 46.41638
Log likelihood -824.7443     F-statistic 23.76002
    Prob(F-
Durbin-Watson stat 1.626867 statistic) 0.000000
Without trend and intercept, we found that at 10%, there is no unit root.

Then, we have tried for another unit root test.


If we do the KPSS test:

Null Hypothesis:
GRCGDP is stationary
Exogenous: Constant
Bandwidth: 5 (Newey-
West using Bartlett
kernel)

LM-Stat.

Kwiatkowski-Phillips-
Schmidt-Shin test
statistic  0.718806
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)  6.01E+21
HAC corrected
variance (Bartlett
kernel)  3.06E+22

KPSS Test Equation


Dependent Variable:
GRCGDP
Method: Least Squares
Date: 05/05/10 Time:
19:30
Sample: 1970 2009
Included observations:
40

Variable Coefficient Std. Error t-Statistic Prob.  

C 7.44E+10 1.24E+10 5.992553 0.0000

    Mean
dependent
R-squared -0.000000 var 7.44E+10
    S.D.
dependent
Adjusted R-squared -0.000000 var 7.85E+10
S.E. of regression 7.85E+10     Akaike 53.03539
info criterion
    Schwarz
Sum squared resid 2.40E+23 criterion 53.07761
    Durbin-
Log likelihood -1059.708 Watson stat 0.010038

With trend, we cannot reject the stationariety of the series.

With trend and intercept, we neither can reject the stationariety. So conclusions seem
contradictory.

Null Hypothesis:
GRCGDP is stationary
Exogenous: Constant,
Linear Trend
Bandwidth: 5 (Newey-
West using Bartlett
kernel)

LM-Stat.

Kwiatkowski-Phillips-
Schmidt-Shin test
statistic  0.207471
Asymptotic critical
values*: 1% level  0.216000
5% level  0.146000
10% level  0.119000

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

Residual variance (no


correction)  6.96E+20
HAC corrected
variance (Bartlett
kernel)  3.29E+21

KPSS Test Equation


Dependent Variable:
GRCGDP
Method: Least Squares
Date: 05/05/10 Time:
19:32
Sample: 1970 2009
Included observations:
40

Variable Coefficient Std. Error t-Statistic Prob.  

C -4.87E+10 8.40E+09 -5.801947 0.0000


@TREND(1970) 6.31E+09 3.71E+08 17.03061 0.0000

    Mean
dependent
R-squared 0.884161 var 7.44E+10
    S.D.
dependent
Adjusted R-squared 0.881113 var 7.85E+10
    Akaike
S.E. of regression 2.71E+10 info criterion 50.92984
    Schwarz
Sum squared resid 2.78E+22 criterion 51.01428
Log likelihood -1016.597     F-statistic 290.0416
    Prob(F-
Durbin-Watson stat 0.033851 statistic) 0.000000

Вам также может понравиться