autocorrelation equal to unity. Under the null of no integration, the DW statistic value will not be significantly from zero. Therefore, a Cointegrating Regression Durbin-Watson (CRDW) statistic different from zero implies cointegration.
iii. Slight problems with DW test for cointegration
This test suffers from two major problems, it is extremely sensitive to the assumption of being a true random walk and, the critical values of the test statistic are not consistent as the number of regressors increases over the sample size. The practical use of this test is therefore extremely limited. 7
3. Data
We select monthly data of adjusted close share prices of SSE and S&P 500 from 2001.12 to 2015.04, this is because China joint the World Trade Organization (WTO) on 2001.11.12 from which China began to globalize with the world’s financial system. (Notes: all data are in natural logarithms) Also, the reason why we do not use daily data is because of the difference between Chinese and US stock markets open dates, they are always mismatched due to different national holidays as well as time lag. Therefore it is difficult to judge which dates should be paired.
4. Results
a. Unit Root Test
The first stage of the analysis is to determine whether the time series data possess a unit root by using Dickey-Fuller test.
Unit root test for SSE Index
X Variable β coefficient: -0.00863
Test statistic: -1.24184
Significance Level
10%
5%
1%
C.V for constant but no trend C.V for constant and trend
-2.57
-2.86
-3.43
-3.12
-3.41
-3.96
Table 2: Critical Values for Dickey Fuller test (Fuller, 1976, p373)
7 Sjo, B. (2008) Testing for Unit Roots and Cointegration .
11
Made with FlippingBook - professional solution for displaying marketing and sales documents online