Fourier Structural Vector Autoregression (Fourier SVAR) Model
Fourier Structural Vector Autoregression Model · Also known as: Fourier SVAR, Fourier structural VAR, Fourier-approximation SVAR, frequency-domain SVAR
The Fourier SVAR model integrates Fourier series approximations into the structural VAR framework, allowing the model to capture smooth, gradual structural breaks and time-varying dynamics in multivariate time series without requiring a priori knowledge of break dates. It recovers structural shocks and their propagation effects while remaining robust to low-frequency parameter drift.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use the Fourier SVAR when you have multivariate macroeconomic or financial time series and suspect smooth, gradual structural change — for example, evolving monetary policy transmission, gradual financial integration, or slow institutional drift — but do not know the exact break dates. It is well-suited when standard SVAR residual diagnostics suggest parameter instability yet discrete break tests find no single dominant break. Avoid it for short samples (T < 80) where Fourier components consume too many degrees of freedom, for data with sharp discrete breaks (prefer Markov-switching SVAR), or when the structural identification scheme is contested and uncertainty is large.
Strengths & limitations
- Accommodates smooth, gradual structural breaks without requiring knowledge of break dates.
- Retains full structural interpretation: orthogonal shocks, IRFs, and FEVD are all recoverable.
- More parsimonious than time-varying-parameter VAR for slowly evolving dynamics.
- Compatible with standard identification strategies (Cholesky, sign restrictions, long-run restrictions).
- Reduces misspecification bias in structural inference caused by unmodelled parameter drift.
- Requires a sufficiently long time series (T >= 80) to estimate Fourier components reliably alongside VAR parameters.
- Choice of the number of frequencies n introduces researcher discretion; overfitting is possible if n is chosen carelessly.
- Cannot capture abrupt, discrete regime changes — a Markov-switching SVAR is more appropriate in those cases.
- Identification of structural shocks still requires external theory; the Fourier extension does not resolve fundamental identification uncertainty.
Frequently asked
How many Fourier frequencies should I include?
Start with a small number (n = 1 or 2) and use an information criterion such as BIC to choose n. Adding too many frequencies relative to sample size wastes degrees of freedom and can produce spurious smooth patterns.
How is the Fourier SVAR different from a time-varying-parameter VAR?
A TVP-VAR allows all coefficients to change every period via a state-space model, which is very flexible but demands large samples and heavy computation. The Fourier SVAR parameterises change through a fixed set of trigonometric terms, making it more parsimonious and suitable for moderate-length samples with slow, smooth drift.
Can I still use standard identification schemes with a Fourier SVAR?
Yes. The Fourier augmentation modifies the deterministic or slowly varying parts of the model; structural identification — Cholesky ordering, sign restrictions, or long-run restrictions — is applied to the structural matrix A0 as in a standard SVAR.
What sample size is sufficient for a Fourier SVAR?
A rough guideline is at least 80 observations per equation. The Fourier terms add parameters, so small samples (T < 60) leave too few degrees of freedom for reliable estimation of both the VAR dynamics and the Fourier components.
When should I choose a Markov-switching SVAR instead?
If theory or preliminary tests suggest a sharp, discrete regime change (e.g., a sudden policy break, a financial crisis with a clear onset date), a Markov-switching SVAR is more appropriate. The Fourier SVAR is designed for gradual, smooth transitions.
Sources
- Enders, W., & Lee, J. (2012). A unit root test using a Fourier series to approximate smooth breaks. Oxford Bulletin of Economics and Statistics, 74(4), 574-599. DOI: 10.1111/j.1468-0084.2011.00662.x ↗
- Bernal, O., & Gnabo, J. Y. (2023). Fourier-based structural VAR models with time-varying parameters. Journal of Applied Econometrics, 38(3), 321-345. link ↗
How to cite this page
ScholarGate. (2026, June 3). Fourier Structural Vector Autoregression Model. ScholarGate. https://scholargate.app/en/econometrics/fourier-svar-model
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Bayesian VAR modelEconometrics↔ compare
- Fourier VAR modelEconometrics↔ compare
- VAR ModelEconometrics↔ compare