ScholarGate
Get started
Regression modelEconometricsEconometrics / time seriesModel

Nonlinear TGARCH Model

Also known as: NL-TGARCH, Nonlinear Threshold GARCH, Asymmetric TGARCH, GJR-GARCH variant

OriginatorJean-Michel Zakoian; related work by Glosten, Jagannathan & RunkleYear1993–1994Sources2Related methods4

The Nonlinear TGARCH (Threshold GARCH) model extends the standard GARCH framework by allowing positive and negative shocks of equal magnitude to exert different effects on future volatility. It models conditional volatility in terms of the absolute value of lagged residuals split by a sign threshold, capturing the well-documented leverage effect in financial return series.

Key highlights

  • Explicitly captures the leverage effect, producing more accurate volatility forecasts for equity and foreign-exchange markets.
  • Modelling in terms of absolute rather than squared residuals reduces the influence of extreme outliers compared with standard GARCH.
  • Straightforward interpretation: gamma directly quantifies the asymmetry in the volatility response.
  • Nests the symmetric TGARCH as a special case (gamma = 0), making nested hypothesis testing easy.
  • Compatible with a wide range of error distributions (normal, Student-t, skewed-t) to accommodate fat tails.

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 Nonlinear TGARCH when modelling daily or high-frequency financial returns (equity indices, exchange rates, commodity prices) where you suspect an asymmetric volatility response to positive versus negative news. It is preferable to symmetric GARCH whenever the leverage coefficient gamma is significant. Avoid it for low-frequency macroeconomic series (monthly GDP growth) where volatility clustering is weak, for short samples (fewer than ~300 observations) that make likelihood-based estimation unreliable, or when the research question concerns the conditional mean rather than the conditional variance.

Strengths & limitations

Strengths
  • Explicitly captures the leverage effect, producing more accurate volatility forecasts for equity and foreign-exchange markets.
  • Modelling in terms of absolute rather than squared residuals reduces the influence of extreme outliers compared with standard GARCH.
  • Straightforward interpretation: gamma directly quantifies the asymmetry in the volatility response.
  • Nests the symmetric TGARCH as a special case (gamma = 0), making nested hypothesis testing easy.
  • Compatible with a wide range of error distributions (normal, Student-t, skewed-t) to accommodate fat tails.
Limitations
  • Requires a reasonably large sample (at least 300–500 observations) for reliable maximum-likelihood estimation.
  • Stationarity requires parameter constraints (alpha + gamma/2 + beta < 1 in squared-volatility terms) that must be imposed or checked post-estimation.
  • The model does not capture long-memory in volatility; Fractionally Integrated GARCH (FIGARCH) or HYGARCH may be better for very persistent processes.
  • Selecting the conditional distribution (normal vs. Student-t vs. skewed-t) adds a model-choice step that affects inference.

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

What is the difference between TGARCH and GJR-GARCH?

Both capture asymmetric volatility, but TGARCH models conditional standard deviation using absolute residuals, while GJR-GARCH models conditional variance using squared residuals. The threshold mechanism is similar, but the resulting dynamics and parameter interpretations differ.

How do I test whether the leverage effect is significant?

Estimate the model and inspect the t-statistic (or p-value) on the threshold coefficient gamma. A significantly positive gamma confirms asymmetric volatility. You can also compare AIC/BIC of the Nonlinear TGARCH with a symmetric GARCH to assess model fit improvement.

What sample size is needed for reliable TGARCH estimation?

A minimum of roughly 300 daily observations is typically recommended; 500 or more is preferred. With fewer observations, the likelihood surface can be flat around gamma, making the asymmetry estimate unreliable.

Can TGARCH be combined with an ARMA mean equation?

Yes. A standard approach is to fit an ARMA(p,q)-TGARCH(1,1) specification, where the ARMA part handles any autocorrelation in returns and the TGARCH part models the residual volatility dynamics.

Which software packages support Nonlinear TGARCH estimation?

The model is available in R (rugarch package), Python (arch package), EViews, and Stata (mgarch suite). All implement maximum-likelihood estimation with multiple distributional options.

Sources

  1. 1.
    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
  2. 2.
    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

You have read it. What now?

Ask about this methodSimilar methods

Cite this page

ScholarGate. (2026, June 3). Nonlinear TGARCH model. ScholarGate. https://scholargate.app/econometrics/nonlinear-tgarch-model

Nonlinear TGARCH Model — Nonlinear Threshold GARCH Model