英语论文网

The ADF and PP test has been criticized for having low power, where the root of stationary process is very close to the non-stationary boundary. KPSS test is run on the variables, with opposite null and alteative hypothesis, that checks for stationarity directly.

From the KPSS test, for all the variables, unit root at first differenceis found. The alternative hypothesis has been accepted for all the variables, and can be concluded that, all the series are stationary at first difference, as non-stationarity is rejected.

Finally, from all the three tests in looking for unit root, all the series of data are stationary, and the are fit to run the cointegration on the data, as the possibility of spurious regression has been removed and the series will provide desired and valued results.

5.2. Cointegration in Multivariate Sytstem – Johansen Approach

As the variables contain a unit root, than the components or the variables are integrated at order 1, expressed as I(1). A set of variables will be cointegrated only if the linear combination of these variables are stationary. If not stationary, the variables might still move together, due to some influence that is bound on the variables, in the long run. A cointegration is a long run phenomenon, since it is possible that, the variables in a cointegration may deviate from their mean in the short run, but will return into the association in the long run. This association is the desirable property that needs to be estimated. If the variables of interest are non-stationary and not-cointegrated, than there would be no long-run relationship that will bind the variables together. In absence of a constant mean, there will not be any mean reversion,and will not be any association among the variables (Brooks, 2008).

As mentioned before, due to some constraints and limitations in Engle-Granger 2 steop method, Johansen-Jesulius full rank cointegration method has been applied. The null hypothesis in this system implies of no-cointegration, where the alternative hypothesis implies the existence of cointegrating equation. The result is organizaed in the following table for cointegartion rank test based on λ_trace test. Here, r denotes rank (number) of cointegration.

From both the trace test and maximum eigenvalue test, the null hypthesis of no cointergration is rejected once. Therefore, the rank of cointegration vector is 1 and there exists one cointegrating equation. The implication of this decision is that the model is integrated at order I(1). In addition, there exists significant long run association among the variables.

5.3. Vector Error correction procedure

“ One definition of the long run that is employed in econometrics implies that the variables have converged upon some long term values and are no longer changing,…..hence all the difference terms will be zero……the class of model that can overcome this problem is error correction models” (Brooks, 本论文由英语论文网提供整理，提供论文代写，英语论文代写，代写论文，代写英语论文，代写留学生论文，代写英文论文，留学生论文代写相关核心关键词搜索。