Consider the AR(1) model with parameter a = −1. As yt is just the negative of its predecessor yt−1 plus an error term, the process will tend to oscillate back and forth around zero. Use recursive substitution to show that Y0, Y2, Y4,…, is a random walk with i.i.d. N(0, 2????2) innovations.

Then simulate and plot the process for a large value of T to see how the variance appears to change through time.