Glossary
GARCH
A model in which the variance depends on the latest squared return and the previous variance, and drifts back to a long-run level.
In GARCH(1,1), var_t = omega + alpha x r_(t-1)^2 + beta x var_(t-1). The long-run variance is omega / (1 - alpha - beta) and the sum alpha + beta is the persistence. Engle (1982) introduced ARCH; Bollerslev (1986) generalised it.
It captures volatility clustering and gives forecasts that revert to the long-run level.
Example: with alpha 0.08 and beta 0.90, persistence is 0.98 and a variance shock has a half-life of ln(0.5) / ln(0.98) = 34.3 days.