Structural Time Series Model (Basic Structural Model)
Also known as: BSM, basic structural model, unobserved components model, Yapısal Zaman Serisi Modeli (BSM)
The Structural Time Series Model, in its Basic Structural Model (BSM) form, is Andrew Harvey's state-space approach that decomposes a series into separate stochastic trend, seasonal, cyclical, and irregular components. Developed in Harvey's 1990 treatment, it is prized for interpretability and component decomposition where ARIMA only delivers a black-box fit.
Key highlights
- Decomposes the series into separate, directly interpretable trend, seasonal, cyclical, and irregular components.
- Often more interpretable than ARIMA because each component carries an explicit meaning rather than abstract autoregressive and moving-average terms.
- The state-space and Kalman-filter formulation handles missing observations and time-varying structure naturally.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use the BSM for a continuous time series (at least about 30 observations) when you want an interpretable decomposition into trend, seasonal, and cyclical parts rather than just a forecast. It assumes the series can be written in state-space form, that the Kalman filter is applicable, and that the components are meaningful and interpretable. It is a strong choice when you need to read off the underlying trend or seasonal pattern, handle missing observations, or explain a forecast to non-specialists; it is less necessary when a plain ARIMA fit suffices and interpretability is not a concern.
Strengths & limitations
- Decomposes the series into separate, directly interpretable trend, seasonal, cyclical, and irregular components.
- Often more interpretable than ARIMA because each component carries an explicit meaning rather than abstract autoregressive and moving-average terms.
- The state-space and Kalman-filter formulation handles missing observations and time-varying structure naturally.
- Requires casting the problem in state-space form and choosing which components (trend, seasonal, cycle) to include.
- Maximum-likelihood estimation of the component variances can be sensitive to starting values and the optimisation routine.
- Needs a reasonable run of data (at least about 30 observations) and well-defined, interpretable components to be reliable.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
How does the BSM differ from ARIMA?
ARIMA describes a series through abstract autoregressive and moving-average terms after differencing, which fits well but is hard to interpret. The BSM instead models explicit trend, seasonal, cyclical, and irregular components, so you can read each part directly. For decomposition and interpretability the BSM is generally preferred.
What role does the Kalman filter play?
The model is written in state-space form, and the Kalman filter recursively estimates the unobserved components as each observation arrives, updating its prediction by a correction proportional to the prediction error. It also produces the likelihood used to estimate the component variances.
What does the 'local linear trend' mean?
It is a trend with both a level and a slope that are each allowed to drift stochastically over time, so the trend can change its direction gradually rather than being fixed. It is the standard trend specification in the Basic Structural Model.
How much data do I need?
At least about 30 observations, and enough span to identify any seasonal and cyclical components you include; otherwise their variances can become unstable or collapse to zero.
Sources
- 1.Harvey, A. C. (1990). Forecasting, Structural Time Series Models and the Kalman Filter. Cambridge University Press.ISBN 978-0521405737
- 2.Harvey, A. C. & Shephard, N. (1993). Structural Time Series Models. In G. S. Maddala, C. R. Rao & H. D. Vinod (Eds.), Handbook of Statistics, Vol. 11 (pp. 261-302). Elsevier.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 1). Structural Time Series Model. ScholarGate. https://scholargate.app/econometrics/structural-time-series