4. Integration and Cointegration Flashcards
What are the different types of nonstationarity?
- series with a deterministic trend
- series with structural breaks
- unit root behaviour
What is unit root behaviour?
Where we have stochastic trends. This can be stationary or non stationary
When do we have a random walk?
When p=1
When do shocks have a permenant impact and the variance is a function of time?
When we have a non stationary series
Order of integration
The number of times we need fo difference the variable to make it stationary
Why do we use a dickey fuller test?
Because the inferential stats (t ratio) arent normally distributed
What is a problem with Dickey Fuller test and how do we deal with it?
We may mistake serial correlation for unit root behaviour. We use the Augmented Dickey Fuller which includes lags of the dependent variable to take out serial correlation
Problems with time series unit root tests
- low power of tests in near unit root cases
- inference is sensitive to treatment of serially correlated errors and treatment of trends and means
- sensitivity to structural breaks
- non linearities
What is a spurious regression
Where two or more varisbles are associated but not causally related
What is cointegration?
A technique used to find a possible correlation between time series processes in the long run
What are the two conditions for cointegration?
- Yt and Xt are both non stationary and integrated of order 1
- There exists a linear combination of the two variables, Yt- ØXt that is stationary
What is the error correlation model?
It is simply a rearranged ARDL model but the LR multiplier effect is already contained
What are the steps to cointegration testing: Engle and Granger approach
- We run the regression
- We run a unit root test of these residuals
- Ho: if ¥=0 no cointegration
H1 if ¥<0 there is cointegration
What can we study from the cointegration test: engle and granger approach?
The SR relationship and the speed of convergence (half life)
