Time-Varying Parameter TGARCH Model
Time-Varying Parameter Threshold Generalized Autoregressive Conditional Heteroscedasticity Model · Also known as: TVP-TGARCH, time-varying TGARCH, threshold GARCH with time-varying parameters, TVP Threshold GARCH
The TVP-TGARCH model extends Threshold GARCH by allowing its volatility parameters to evolve over time via a state-space representation. It captures both the leverage effect — that negative return shocks increase volatility more than positive ones — and structural change in that asymmetry, making it well-suited for long financial time series subject to regime shifts.
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 TVP-TGARCH when you have a long financial return series (hundreds to thousands of observations) and suspect that the leverage effect or overall volatility dynamics have shifted over the sample — for example spanning multiple business cycles, crises, or regulatory regimes. It is valuable for accurate long-horizon volatility forecasting and risk management when structural stability is doubtful. Avoid it with short samples (fewer than ~300 observations) where the time-varying parameters are poorly identified, or when a standard TGARCH or GJR-GARCH with regime dummies suffices and parsimony is preferred.
Strengths & limitations
- Captures the leverage effect of TGARCH while simultaneously allowing those asymmetry parameters to change over time.
- Detects gradual or abrupt structural change in volatility dynamics without requiring pre-specified break dates.
- State-space formulation provides a principled probabilistic framework with well-developed Kalman-filter estimation.
- Produces a full time-path of estimated parameters, enabling analysis of how financial risk sensitivity has evolved.
- Outperforms constant-parameter models in out-of-sample volatility forecasting over long horizons.
- Requires large samples for the time-varying parameters to be well identified; short panels yield unstable estimates.
- Computationally demanding compared with standard GARCH family models, especially with many parameters allowed to vary.
- Model selection (which parameters to allow to vary, choice of transition covariance Q) involves non-trivial specification decisions.
- Filtering uncertainty in the Kalman smoother is rarely fully propagated into forecast confidence intervals in practice.
Frequently asked
How is TVP-TGARCH different from a Markov-switching GARCH model?
Markov-switching GARCH assumes a finite number of discrete regimes with abrupt transitions governed by an unobserved Markov chain. TVP-TGARCH instead allows parameters to drift continuously over time via a random-walk state equation, which better captures gradual structural change rather than sudden jumps.
Which parameters are typically allowed to vary over time?
In practice, researchers often allow the asymmetry coefficients α⁺ and α⁻ and sometimes the persistence parameter β to vary, while holding the intercept ω fixed or also time-varying. The choice depends on the sample length and economic hypothesis being tested.
What estimation method is used?
The standard approach is quasi-maximum likelihood via the Kalman filter: the state-space form allows the prediction-error decomposition to evaluate the likelihood, which is then maximised numerically. Bayesian approaches using particle filters or MCMC are also found in the literature.
How large a sample is needed?
A rule of thumb is at least 300 observations, and ideally 500 or more, so that the time-varying parameters have enough data to be identified at each period. With fewer observations the estimates of the transition variance Q become unreliable.
Can TVP-TGARCH be extended to multivariate settings?
Yes, multivariate extensions (TVP-DCC-TGARCH, TVP-BEKK with asymmetry) exist in the literature, though they are significantly more computationally demanding and the identification challenges multiply with the number of assets.
Sources
- Zakoïan, J.-M. (1994). Threshold heteroskedastic models. Journal of Economic Dynamics and Control, 18(5), 931–955. DOI: 10.1016/0165-1889(94)90039-6 ↗
- Glosten, L. R., Jagannathan, R., & Runkle, D. E. (1993). On the relation between the expected value and the volatility of the nominal excess return on stocks. Journal of Finance, 48(5), 1779–1801. DOI: 10.1111/j.1540-6261.1993.tb05128.x ↗
How to cite this page
ScholarGate. (2026, June 3). Time-Varying Parameter Threshold Generalized Autoregressive Conditional Heteroscedasticity Model. ScholarGate. https://scholargate.app/en/econometrics/time-varying-parameter-tgarch-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.
- EGARCH modelEconometrics↔ compare
- GARCH ModelEconometrics↔ compare
- State Space ModelEconometrics↔ compare
- TGARCH modelEconometrics↔ compare