Motivation
Stationary autoregressive moving average (ARMA) models play an important role in financial forecasting (we referred to it here and here). My experience is that online explanations of stationarity often are either very long or implicit about certain steps. Therefore, this post presents a fairly short but hopefully complete explanation of when an ARMA model is stationary.
Stationary ARMA models
Consider the following basic linear filter model:
where is the value of a time series at time , where are weights and are random terms, at different times, drawn from a probability distribution with a fixed mean and variance (we don’t bother with the level of the time series, which would equal a constant on the right-hand-side of the expression). A times series is called stationary whenever .
Now an ARMA model i just a special case of the linear filter model. To see this, consider the following ARMA model:
where and are weights of the the AR and MA parts of the model, respectively, and is the back-shift operator (i.e. ). For this particular ARMA model depends on the two most recent values of the series, , as well as three random terms, . By re-arranging the expression we get:
Now we can compare this with the linear filter model (and using the back-shift operator for the linear filter model as well):
which is equal to:
If we cancel on both sides and compare the weights in front of , respectively, we find that:
(for )
(for )
(for )
(for )
…
(for )
For with we have a second-order linear homogeneous recurrence relation (with initial conditions given by finite values ). What we want to do is to find an closed expression for . We follow the standard approach when solving recurrence relations, by setting , where :
We divide by and get:
.
This is a simple quadratic equation. If we solve for we either get two distinct (independent) real roots, two complex roots or one repeated root. If we have two distinct roots we add them for the following general solution (this is a standard technique when solving recurrance relations):
For one repeated root we instead have the general solution:
.
We could use the initial conditions to determine the values of the constants . But for stationarity it is sufficiant that the absolute values of the roots of the quadratic equation is less than one. This is because, when the roots are distinct, we have that:
.
This follows from the triangle inequality and because the sum of infinite geometric series are finite when .
The cases will be similar if we have complex roots or one repeated root. Our method for determining stationarity also holds for other ARMA models as well (and of course for all AR and MA models since they equal ARMA models with some weights set to zero).