AR(1): Stationarity, Variance, and Whitening

Reported in interviews at BlackRock

Problem statement

Consider the AR(1) model

$$x_t = a\, x_{t-1} + e_t, \qquad e_t \sim N(0,1) \text{ i.i.d.}$$

(a) What condition on $a$ makes $x_t$ (weakly) stationary?

(b) What is the stationary variance of $x_t$? Write the variance relation in vector/matrix form for the whole vector $x = (x_1, \dots, x_T)^{\top}$.

(c) The $x_t$ are correlated with a non-trivial covariance matrix $\Sigma$. How do you standardize (whiten) the series so the transformed variables are uncorrelated with unit variance, and what does the model look like after that transform?

Show hint

What must happen to old shocks over time for the process to settle into a steady state?

Relevant roles: Data Scientist, Applied Scientist and Research Scientist.

Related free problems