Cointegration for Price levels Time Series

Question

I don't understand why is the difference between price levels is a stationary process while the time series of price levels themselves is a non-stationary process.

For example:

s_1,t = delta_1 + b_1*Z_t + e_1,t
s_2,t = delta_2 + b_2*Z_t + e_2,t

--both have unit roots

but, 
b_2*s_1,t - b_1*s_2,t = b_2*delta_1 - b_2*delta_2 + b_2*e_1,t - b_1*e_2,t

--because of co integration, we are allowed to take the differences.

*the underscores are subscripts

Answer 1

As far as I know they are assumed to be stationary.