Fourier TGARCH Model
Fourier Threshold Generalized Autoregressive Conditional Heteroscedasticity Model · Also known as: Fourier TGARCH, Fourier Threshold GARCH, Fourier GJR-GARCH, smooth structural break TGARCH
The Fourier TGARCH model extends the Threshold GARCH framework by embedding Fourier trigonometric terms in the conditional variance equation to capture smooth, gradual structural breaks in volatility dynamics. It jointly models asymmetric leverage effects — where negative shocks amplify volatility more than positive shocks of the same magnitude — and time-varying intercept shifts caused by unobserved structural change.
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 Fourier TGARCH when you have a financial or macroeconomic return series that exhibits both asymmetric volatility clustering (leverage effects) and evidence of gradual, smooth structural change in the volatility level — for example, long daily or weekly financial return series spanning multiple economic cycles, policy regimes, or crisis episodes. It is preferable to plain TGARCH when breakpoint tests suggest time-varying intercepts but no single sharp date can be identified. Do not use it when the series is very short (fewer than 200 observations), as the additional Fourier parameters reduce estimation precision, or when a single well-identified structural break exists — in that case a dummy-variable TGARCH is simpler and better identified.
Strengths & limitations
- Simultaneously models asymmetric leverage effects and smooth structural breaks in volatility without requiring a pre-specified break date.
- Fourier approximation is parsimonious: a single low-frequency component (K=1 or K=2) often suffices to capture gradual regime shifts.
- Consistent with the Enders-Lee Fourier principle, which is robust to both gradual and abrupt breaks.
- QML estimation retains consistency and asymptotic normality under mild distributional assumptions.
- Provides a formal test for leverage effects (significance of gamma) conditional on the structural break correction.
- Requires long time series (at least 200 observations) to identify both the Fourier components and TGARCH parameters reliably.
- Selection of the optimal Fourier frequency K adds a model-selection step, typically via information criteria (AIC/BIC).
- Does not accommodate multiple volatility regimes with distinct dynamics in each regime — a Markov-switching GARCH may be more appropriate for that.
- The smooth-break assumption may be violated if the structural change is genuinely abrupt.
Frequently asked
How do I choose the number of Fourier frequencies K?
Estimate the model for K = 1, 2, and 3, and select the specification with the lowest AIC or BIC. In most applications K = 1 is sufficient. Avoid K greater than 3 to prevent overfitting.
What is the difference between Fourier TGARCH and Fourier EGARCH?
Both incorporate Fourier terms for smooth breaks, but they differ in how asymmetry is parameterised. TGARCH uses a piecewise indicator on past shocks and operates on the conditional variance directly, while EGARCH operates on the log-variance and uses a sign function. TGARCH is somewhat easier to interpret, while EGARCH automatically satisfies positivity.
Does Fourier TGARCH assume a specific distribution for errors?
No. It is typically estimated by quasi-maximum likelihood (QML) assuming conditional normality, which yields consistent estimates even if the true distribution has fat tails. For better efficiency with fat-tailed data, a Student-t or GED likelihood can be used instead.
How do I test whether the Fourier terms are necessary?
Run a standard TGARCH as the restricted model and the Fourier TGARCH as the unrestricted model, then perform a likelihood ratio test. Alternatively, compare AIC and BIC across the two specifications.
Can I apply Fourier TGARCH in panel data settings?
Individual-series Fourier TGARCH estimation is standard. Panel extensions (Panel Fourier TGARCH) exist in recent literature but are computationally demanding and require specialised software. For most applications, cross-sectional averaging of series-level estimates or DCC-GARCH is preferable.
Sources
- Zakoian, J.-M. (1994). Threshold heteroskedastic models. Journal of Economic Dynamics and Control, 18(5), 931-955. DOI: 10.1016/0165-1889(94)90039-6 ↗
- 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 ↗
How to cite this page
ScholarGate. (2026, June 3). Fourier Threshold Generalized Autoregressive Conditional Heteroscedasticity Model. ScholarGate. https://scholargate.app/en/econometrics/fourier-tgarch
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.
- DCC-GARCH modelEconometrics↔ compare
- EGARCH modelEconometrics↔ compare
- Fourier EGARCHEconometrics↔ compare
- Fourier GARCH ModelEconometrics↔ compare
- TGARCH modelEconometrics↔ compare