EGARCH
Allow different volatility responses to positive and negative news
EGARCH models the logarithm of conditional variance, ensuring positive variance without the same non-negativity restrictions used in linear GARCH. Signed and magnitude terms allow equally sized positive and negative innovations to have different effects. Parameter interpretation depends on the exact equation and error distribution, so the sign of an asymmetry coefficient must be read within the stated convention.
Use when variance responds differently to negative and positive innovations and a log-variance specification is defensible. Compare forecast performance with symmetric GARCH and threshold alternatives rather than selecting by in-sample fit alone.
Strengths
- Log variance guarantees a positive fitted variance
- Allows shock magnitude and sign to contribute separately
Limitations
- Log-scale persistence and moment conditions need care
- An asymmetric statistical response does not identify its economic cause
Know the boundary
The asymmetry often called a leverage effect is an estimated response pattern, not proof that changing financial leverage caused it.