Time Series • Multivariate Time Series Models and Forecasting
Time Series / Multivariate Stationary Process

Multivariate Stationary Process

Notes 6 Multivariate Time Series Models and Forecasting

Unit 3 used stationarity for one series: a reasonably stable typical level, a reasonably stable amount of variation, and a dependence structure that does not keep changing in a systematic way. A multivariate stationary process extends that idea to several series at once. Not only each series, but also the relationships among the series (including lagged cross-links), stay reasonably stable over time.

Notes

Multivariate Stationary Process

Definition

Unit 3 used stationarity for one series: a reasonably stable typical level, a reasonably stable amount of variation, and a dependence structure that does not keep changing in a systematic way. A multivariate stationary process extends that idea to several series at once. Not only each series, but also the relationships among the series (including lagged cross-links), stay reasonably stable over time.

Example

Two related economic series, such as a regional price index and a wage index, recorded monthly. If both wander upward without bound and the gap between them keeps changing regime, the pair is not behaving as a stationary multivariate process. If, after a simple transformation or differencing, the pair fluctuates around stable means and a stable co-movement, modelling is on safer ground.

What Should Stay Stable

  • Mean behaviour: each series does not keep drifting in level (after any justified differencing or trend removal).
  • Variance / covariance behaviour: the typical size of each series, and how strongly they move together at the same time, does not keep jumping to a new regime.
  • Lagged relationships: the way series A now relates to series B last month stays comparable through the sample.

Covariance at the same time is “do they move together this month?” Cross-covariance at a lag is “does today's A move with last month's B?” Both are association measures. They are not causation. Advanced matrix proofs of these quantities are not required at this level.

Univariate vs Multivariate Stationarity

Univariate stationarity Multivariate stationarity
Focus One series: level, variation, own lags The whole vector: own properties plus cross-series links
Can fail even if… The single series wanders or changes volatility Each series looks calmer but the relationship between them keeps shifting

Why It Matters for Modelling

Joint models and simultaneous forecasts assume that the pattern you estimated will still be relevant next month. If the means, spreads or cross-links are still evolving, the fitted joint structure can be a poor guide to the future.

Non-stationary series ↓ May require transformation / differencing ↓ More stable structure ↓ Then model the multivariate process

As in Unit 3, not every series needs differencing. Check plots and whether associations look stable. Do not difference automatically “because the data are multivariate.”

Exam-Oriented Key Points

  1. Multivariate stationarity = stable individual behaviour and stable cross-variable relationships.
  2. Cross-covariance describes lagged association between two series; it is not causation.
  3. Non-stationary pairs may need transformation or differencing before a joint model.
  4. Do not assume that separate univariate stationarity automatically gives multivariate stationarity.