Structural Break EGARCH Model
Exponential GARCH Model with Structural Breaks · Also known as: SB-EGARCH, EGARCH with regime shifts, break-adjusted EGARCH, structural change EGARCH
Structural Break EGARCH combines Nelson's Exponential GARCH framework with explicit allowance for one or more structural breaks in the volatility process. By letting the intercept and persistence parameters of the log-variance equation shift at detected break dates, the model avoids the spurious long-memory and inflated persistence that standard EGARCH suffers when the data contain regime changes.
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 Structural Break EGARCH when: (1) you model financial or macroeconomic volatility over long samples that span clearly identifiable regime changes (crises, policy reforms, exchange-rate pegs); (2) a standard EGARCH shows implausibly high persistence (β close to 1), which is a classic sign of unmodelled breaks inflating persistence; (3) formal break tests on squared returns or GARCH residuals reject parameter stability. Do not use it when the sample is short (fewer than ~200 observations) — break detection has low power and the regime sub-samples will be too small for reliable ML estimation; prefer a rolling or time-varying EGARCH instead. Also avoid it when no structural breaks are theoretically or statistically supported — unnecessary break terms consume degrees of freedom without adding explanatory power.
Strengths & limitations
- Removes the upward bias in volatility persistence caused by unmodelled structural changes, yielding more accurate impulse-response analysis.
- Retains the EGARCH leverage asymmetry (bad news raises volatility more than good news) within each regime.
- Log-variance specification guarantees non-negative conditional variance without imposing parameter sign restrictions.
- Enables regime-specific unconditional variance levels, useful for studying how crises or policy changes shift the long-run risk environment.
- Compatible with fat-tailed error distributions (Student-t, GED) that better capture extreme observations in financial series.
- Break dates are rarely known with certainty; endogenous detection adds a pre-testing layer that complicates inference and can overfit to noise.
- Small sub-sample sizes between breaks reduce the precision of ML estimates and can make standard errors unreliable.
- The number of parameters grows quickly with additional breaks and regimes, increasing the risk of overfitting in moderate samples.
- Joint estimation of break dates and GARCH parameters is computationally intensive and may not converge to the global optimum.
- Ignoring breaks in the mean equation (while only modelling variance breaks) can still leave residual structure that distorts the volatility estimates.
Frequently asked
Why does standard EGARCH overestimate volatility persistence when breaks are present?
When the unconditional variance shifts upward at a break date, a model without break terms must interpolate between the pre- and post-break levels by assigning high persistence (β close to 1) to the conditional variance. This makes shocks appear to die out slowly, when in reality they die out quickly within each regime and the elevated level is simply a permanent shift.
How do I choose the number of break points?
Start with formal tests: apply the Bai-Perron sequential test to squared residuals from an initial EGARCH fit to determine the number of structural breaks. Compare models with 0, 1, 2, … breaks using the Bayesian Information Criterion (BIC), which penalises extra breaks and guards against overfitting.
Can I let all EGARCH parameters (not just the intercept) shift at breaks?
Yes, but this dramatically increases the number of parameters. A common compromise is to allow only the intercept ω to shift (partial break model) while constraining α, β, and γ to be constant across regimes. Test whether the additional flexibility of full parameter shifts is warranted with a likelihood ratio test.
What is the minimum sample size recommended for reliable estimation?
As a rule of thumb, each regime sub-sample should contain at least 100–200 daily observations for reliable ML convergence and precise parameter estimates. Very short sub-samples produce large standard errors and may cause numerical instability in optimisation.
How does structural break EGARCH differ from a Markov-switching GARCH?
Both handle regime changes in volatility, but Markov-switching GARCH treats the regime as a latent variable that can recur according to transition probabilities, making it suited to cyclical regime shifts. Structural break EGARCH assumes permanent, one-directional shifts at known or estimated dates, making it simpler and more interpretable for single-event policy or crisis studies.
Sources
- Nelson, D. B. (1991). Conditional heteroskedasticity in asset returns: A new approach. Econometrica, 59(2), 347–370. DOI: 10.2307/2938260 ↗
- Lamoureux, C. G., & Lastrapes, W. D. (1990). Persistence in variance, structural change, and the GARCH model. Journal of Business and Economic Statistics, 8(2), 225–234. DOI: 10.1080/07350015.1990.10509794 ↗
How to cite this page
ScholarGate. (2026, June 3). Exponential GARCH Model with Structural Breaks. ScholarGate. https://scholargate.app/en/econometrics/structural-break-egarch
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.
- ARCH modelEconometrics↔ compare
- DCC-GARCH modelEconometrics↔ compare
- EGARCH modelEconometrics↔ compare
- TGARCH modelEconometrics↔ compare
- Zivot-Andrews Structural Break TestEconometrics↔ compare