eg., consider the following time series process:
yt = μ + Φyt
+ ut
For doing the unit root test analysis we consider the case
Φ<1
Φ=1
We ignore the case for Φ>1
This is because assuming the scenario of Φ>1 would constitute of what we say an "explosive" process. This would mean that shock from previous times will be having an increasingly greater influence as time progresses.
For example if an incident happens on a stock today, this will have more impact tommorow, even more so a week later and so on. In reality however we observe that this is not the case.
Hence this case does not describe any data of financial time series and can be ignored.