1

So I have a data set, it containss 4 variables (GDP, consumer price index, interest, house price index) and around 88 sample. Quarterly data from years 1995-2007.

Two of them look like this https://i.stack.imgur.com/GXyiP.jpg and I difference mortgage and HOX, mortgage to the first to I(1) and HOX needed to be at I(2) so they have a unit root (and therefore also are stationary time series, correct me if I'm wrong).

My question is: why do I need to do Johanses cointegration test? I found this online: https://i.stack.imgur.com/IF7Qr.jpg and https://i.stack.imgur.com/Pko4f.jpg .. :/

When I ran Johanses cointegration test, and I got this: https://i.stack.imgur.com/8S4Uu.jpg

The lags are chosen using the AIC criteria (It picked 5 as the best)

  1. What is going on?
  2. Do I really need the cointegration test? since I've already make them stationary/they have now unit root
  3. I don't get this at all.
  4. What can I do - if this is wrong?

ps. I'm using EViews.

Michael R. Chernick
  • 42,857
soet irl
  • 41
  • I think the OP means that the series are stationary after differencing. The unit roots correspond to the original characteristic polynomials. – Michael R. Chernick May 13 '18 at 18:56
  • Regarding (2), whether or not the series are cointegrated is a different question than whether their differences are stationary – Taylor May 17 '18 at 12:20

0 Answers0