Cointegration 211
and the Netherlands attain statistical significance (at the 5% level) and are
about the same size. This shows that when the economies of France and
the Netherlands deviate from the common stochastic trend, they adjust. In
France about 15% and in Netherlands about 17% of the last period devia-
tion is corrected during this period.
Key Points
■ (^) Many of the variables of interest to finance professionals are non-
stationary.
■ (^) The relationships among nonstationary variables can be analyzed if
they share a common stochastic trend. A way of capturing this common
stochastic trend is the application of cointegration.
■ (^) Cointegration analysis can reveal interesting long-run relationships
between variables.
■ (^) When the variables have stochastic trends, there is a spurious regression
problem and, as a result, the ordinary least squares estimation method
may provide misleading results.
■ (^) Cointegration analysis has been used in testing market price efficiency
and international stock market linkages.
■ (^) It is possible that cointegrating variables may deviate in the short run
from their relationship, but the error-correction model shows how these
variables adjust to the long-run equilibrium.
■ (^) There are two important methods of testing for cointegration: the
Engle-Granger cointegration test and the Johansen-Juselius test.
■ (^) The most often used method to test cointegration between two vari-
ables is the Engle-Granger cointegration test.
■ (^) The Johansen-Juselius test is employed to test cointegration among
multiple variables.