The Wold Theorem states that any (weakly) stationary process $(x_t)_{t=-\infty}^{+\infty}$ with zero mean can be decomposed into $$ x_t=\sum_{j=0}^{\infty} b_j\epsilon_{t-j}\ + \eta_t , $$ where the first summand is the stochastic and the second term a determinstic part.

Let $x_t$ be any (weakly stationary) ARMA(p,q)-process and consider I would like to find the Wold representation of that process. Is it true that the Wold decomposition always coincides with the Moving Avergage MA($\infty$)-representation of that ARMA(p,q)-model, such that $\eta_t=0$ and $b_j$ being the MA weights ?


The Wold decomposition does not say what you state. It says that any weakly stationary $(x_t)_{t=-\infty}^{\infty}$, there exists a white noise process $\{\epsilon_t\}_{t=-\infty}^{+\infty}$ such that $(x_t)_{t=-\infty}^{\infty}$ has two-sided MA representation $$ x_t=\sum_{-\infty < j < \infty} b_j\epsilon_{t-j}. $$

What you're asking is whether the two-sided MA representation is unique.

No. Existence does not imply uniqueness, i.e. there is no "the Wold decomposition." Given a two-sided MA representation $$ x_t=\sum_{-\infty < j < \infty} b_j\epsilon_{t-j}, $$ it is easy to find another MA representation---a different white noise process $\{\epsilon'_t\}_{t=-\infty}^{+\infty}$ and a different sequence $\{b'_t\}_{t=-\infty}^{+\infty}$ such that $$ x_t=\sum_{-\infty < j < \infty} b'_j\epsilon'_{t-j}. $$

So it does not make sense to speak of "the Wold decomposition."

Given a stationary ARMA $(x_t)_{t=-\infty}^{\infty}$, you can write down an MA(∞) representation $$ x_t=\sum_{-\infty < j < \infty} \psi_j\epsilon_{t-j}. $$ But it does not make sense to ask whether "...the Wold decomposition...coincides with the MA(∞)-representation".


Yes, conditional on ARMA(p,q) being the true model, what you said is correct.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.