Page 1 of 1

ADF tests and cointegration inquiry

Posted: Mon Nov 18, 2013 3:33 am
by iliasvalasop
I want to run this estimation :ls loggdp c loggdp(-1) logdt(-1) , where loggdp is calculated in excel as well as logdt which is the log of Debt/GDP. in this estimation I am asked to investigate whether there is a unit root(Dickey- Fuller test) and cointegration. I wanted to ask you how to do this. Until now I found the residuals and then I made the latter tests but I dont think that my findings are correct. thank you in advance for your assistance. :)

Re: ADF tests and cointegration inquiry

Posted: Mon Nov 18, 2013 7:18 am
by CharlieEVIEWS
User Guide II (EViews Eight) - p.471 for Unit Root tests and p.849 for cointegration testing. If you have any specific questions I'd be happy to help!

Re: ADF tests and cointegration inquiry

Posted: Tue Dec 03, 2013 3:34 am
by iliasvalasop
• yt = β0 + β1 yt-1 + β2 dt-1 + ε3t ,where yt = log(GDP) and dt= Debt /GDP
i want to test whether this has a unit root or not. all numbers are known except the disturbance term ofcourse .
i ran it with the ls command and then i cannot make the ADF test. I know we have to check when 1- β1L is 0 and i know the proceedures explained in the tutorials but the outcome isn't what i want.i dont wish to run an ADF test in log(gdp) and its lags soley and separately.
Am I using the wrong test or am I doing it wrong in the first place? What is the exact procedure that i must follow?
Apologies for any inconvenience and thank you again for your time.

Re: ADF tests and cointegration inquiry

Posted: Fri Dec 06, 2013 7:27 am
by CharlieEVIEWS
I am confused as to the question: You cannot run ADF tests on an equation, only on a series. Do you mean you want to run an ADF test on the residuals (some kind of Engle Granger 2 step approach to cointegration?), or do you want to run the test on the variables in your equation?

Re: ADF tests and cointegration inquiry

Posted: Fri Dec 06, 2013 1:01 pm
by iliasvalasop
there is a test that refers to common unit root. I examined all the variables about unit roots and their differences but the question is whether i can come to a conclusion with this:

Null Hypothesis: Unit root (common unit root process)
Series: LOGGDP, LOGGDP(-1), DEBT_GDP(-1)
Sample: 1948 2014
Exogenous variables: None
Automatic selection of maximum lags
Automatic lag length selection based on SIC: 0 to 3
Newey-West automatic bandwidth selection and Bartlett kernel
Total number of observations: 168
Cross-sections included: 3

Method Statistic Prob.**
Levin, Lin & Chu t* 1.09171 0.8625

** Probabilities are computed assuming asymptotic normality

Intermediate results on GROUP01

2nd Stage Variance HAC of Max Band-
Series Coefficient of Reg Dep. Lag Lag width Obs
LOGGDP 0.00141 0.0006 0.0115 3 10 6.0 63
LOGGDP(-1) 0.00077 0.0006 0.0117 3 10 6.0 62
DEBT_GDP(-1) 0.04755 0.0046 0.0123 0 9 4.0 43

Coefficient t-Stat SE Reg mu* sig* Obs
Pooled 0.00202 1.106 1.040 0.001 1.012 168

Thank you again.

P.S. Do I have homoskedasticity as an outcome from the following White test?

Heteroskedasticity Test: White

F-statistic 1.348132 Prob. F(5,38) 0.2654
Obs*R-squared 6.629071 Prob. Chi-Square(5) 0.2497
Scaled explained SS 7.754722 Prob. Chi-Square(5 0.1703

Test Equation:
Dependent Variable: RESID^2
Method: Least Squares
Date: 12/04/13 Time: 12:51
Sample: 1971 2014
Included observations: 44

Variable Coefficient Std. Error t-Statistic Prob.

C 0.003738 0.008168 0.457620 0.6498
LOGGDP(-1)^2 0.004030 0.005343 0.754221 0.4554
LOGGDP(-1)*DEBT_GDP(-1) -0.008353 0.008991 -0.929021 0.3587
LOGGDP(-1) -0.007409 0.013397 -0.553049 0.5835
DEBT_GDP(-1)^2 0.004253 0.003221 1.320443 0.1946
DEBT_GDP(-1) 0.007545 0.012261 0.615422 0.5419

R-squared 0.150661 Mean dependent var 0.000517
Adjusted R-squared 0.038906 S.D. dependent var 0.000859
S.E. of regression 0.000842 Akaike info criterion -11.19575
Sum squared resid 2.69E-05 Schwarz criterion -10.95245
Log likelihood 252.3065 Hannan-Quinn criter. -11.10553
F-statistic 1.348132 Durbin-Watson stat 1.678735
Prob(F-statistic) 0.265444

Re: ADF tests and cointegration inquiry

Posted: Fri Dec 06, 2013 1:33 pm
by CharlieEVIEWS
For the Levin-Lin-Chu Test (LLC) test:

H0: each time series contains a unit root
H1: each time series is stationary

Look at your p-values and make your conclusion accordingly.