GARCH Model (Volatility Forecasting)
Generalized Autoregressive Conditional Heteroskedasticity Model · Also known as: GARCH, GARCH(1,1), conditional volatility model, GARCH Modeli (Oynaklık Tahmini)
The Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model, introduced by Tim Bollerslev in 1986, models the time-varying conditional variance of a financial time series. It captures volatility clustering and the ARCH effect, and is the standard tool for estimating risk and volatility in return series.
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.
+29 more
When to use it
Use GARCH for stationary (or differenced, e.g. log-return) financial time series in which volatility clustering is present, ideally with at least 250 observations. Before fitting, confirm the series is stationary with an ADF test, verify an ARCH effect with the ARCH-LM test, and check the residuals for autocorrelation with the Ljung-Box test. If there is no ARCH effect a simpler model such as ARIMA suffices, and if volatility responds asymmetrically to good and bad news an extension such as EGARCH or GJR-GARCH is preferable.
Strengths & limitations
- Captures volatility clustering and the ARCH effect, the dominant feature of financial return series.
- The standard, widely accepted tool for conditional volatility and risk forecasting, including Value-at-Risk.
- The parsimonious GARCH(1,1) often fits return series well with only a handful of parameters.
- Needs a long series (at least about 250 observations); with fewer, maximum-likelihood estimation can fail to converge and parameters become unreliable.
- Symmetric GARCH treats positive and negative shocks alike, so it misses the leverage effect — for that, EGARCH or GJR-GARCH is needed.
- Requires a stationary input series and a confirmed ARCH effect; without volatility clustering the model is unnecessary.
Frequently asked
What is volatility clustering?
It is the empirical pattern in which large price changes tend to be followed by large changes and small changes by small changes. GARCH captures this by letting the conditional variance depend on recent squared shocks and past variance.
How do I know if GARCH is needed?
Run the ARCH-LM test on the residuals of a mean model. If it is significant there is an ARCH effect (volatility clustering) and GARCH is warranted; if it is not significant, a simpler model such as ARIMA is enough.
What does GARCH(1,1) mean?
The two numbers are the number of lagged squared-shock terms (the ARCH order q) and the number of lagged variance terms (the GARCH order p). GARCH(1,1) uses one of each — ω + α·ε²_{t-1} + β·σ²_{t-1} — and is often adequate for return series.
When should I use EGARCH instead?
When volatility reacts asymmetrically to good versus bad news (the leverage effect), symmetric GARCH understates risk after negative shocks. Asymmetric extensions such as EGARCH or GJR-GARCH model that asymmetry explicitly.
Sources
- Bollerslev, T. (1986). Generalized Autoregressive Conditional Heteroskedasticity. Journal of Econometrics, 31(3), 307–327. DOI: 10.1016/0304-4076(86)90063-1 ↗
How to cite this page
ScholarGate. (2026, June 1). Generalized Autoregressive Conditional Heteroskedasticity Model. ScholarGate. https://scholargate.app/en/econometrics/garch-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.
- ARIMAEconometrics↔ compare
- EGARCHEconometrics↔ compare
- Exponential SmoothingEconometrics↔ compare
- OLS RegressionEconometrics↔ compare
- Quantile RegressionEconometrics↔ compare