What is an ARMA model and when is it appropriate?

Question

AcadiFi · Accepted Answer

An ARMA(p, q) model combines autoregressive and moving average components: x_t equals c plus phi_1 times x_{t-1} plus and so on plus phi_p times x_{t-p} plus epsilon_t plus theta_1 times epsilon_{t-1} plus and so on plus theta_q times epsilon_{t-q}.

ARMA is appropriate when a series shows both persistent feedback (warranting AR terms) and transient shock propagation (warranting MA terms). Many financial time series — GDP growth, interest rate changes, exchange rate returns — fit ARMA structures better than pure AR or MA.

Model selection follows the Box-Jenkins methodology: (1) ensure stationarity via differencing if needed, (2) examine ACF and PACF patterns to propose candidate orders, (3) estimate competing models, (4) select by information criteria like AIC or BIC, (5) check residuals for whiteness via Ljung-Box Q-statistic.

Example: researcher Kimiyo Tanaka-Walsh at Obsidian Credit Partners finds quarterly corporate default rates fit an ARMA(1,1): default_t equals 0.8 plus 0.7 times default_{t-1} plus epsilon_t plus 0.3 times epsilon_{t-1}. The AR(1) term captures persistence; the MA(1) captures quick reaction to credit shocks.

```mermaid
flowchart TD
    S[Stationary Series] --> A[Check ACF PACF]
    A --> B{Pattern?}
    B -->|Cutoff PACF| AR[AR p]
    B -->|Cutoff ACF| MA[MA q]
    B -->|Both decay| ARMA[ARMA p q]
```

If differencing is required to achieve stationarity, the result is an ARIMA(p, d, q) model, relevant for non-stationary series with trends.

What is an ARMA model and when is it appropriate?

Master Level II with our CFA Course

Related Questions

Practice Questions