DCC-GARCH (Dynamic Conditional Correlation)
Dynamic Conditional Correlation GARCH · Also known as: dynamic conditional correlation, Engle DCC, multivariate GARCH, DCC-GARCH — Dinamik Koşullu Korelasyon
DCC-GARCH is Engle's (2002) multivariate volatility model that lets the correlations between several assets change over time. A separate univariate GARCH model is fitted to each series, and then the dynamic correlation matrix is estimated in a second, separate step.
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 DCC-GARCH for several continuous financial return series observed over time when you need correlations that vary across periods rather than a single static value, and you have a reasonably long sample (the model needs at least about 100 observations, and roughly 250 or more for reliable parameter estimates). It assumes each series follows its own GARCH(p,q) process, that the standardized residuals are conditionally independent, and that the DCC parameters satisfy a + b < 1. For very high dimensions (N > 50) the cDCC or BEKK variants are preferable, and the series should be stationary, so difference them first if needed.
Strengths & limitations
- Captures correlations that change over time instead of forcing a single static correlation between assets.
- Two-step estimation separates volatilities from correlations, keeping the model tractable for many assets at once.
- Built on familiar univariate GARCH components, so it integrates naturally into volatility and risk workflows.
- Needs a long, stationary sample; with short series (n < 250) the DCC parameters cannot be estimated reliably and a simpler GARCH is preferable.
- Scales poorly to very high dimensions (N > 50), where cDCC or BEKK is recommended instead.
- Non-stationary series produce misleading results and must be differenced to stationarity first.
Frequently asked
How is DCC-GARCH different from a single GARCH model?
A univariate GARCH tracks the changing volatility of one series. DCC-GARCH extends this to several series at once and adds a second stage that models how the correlations between them change over time, decomposing the covariance matrix into volatilities and a dynamic correlation matrix.
Why is the model estimated in two steps?
Separating the problem keeps it tractable. The first step fits an individual GARCH to each series to capture its own volatility; the second step uses the resulting standardized residuals to estimate the correlation dynamics. This avoids estimating a huge multivariate likelihood all at once.
What does the constraint a + b < 1 mean?
The DCC parameters a and b govern how strongly past shocks and past correlations carry forward. Requiring a + b < 1 keeps the correlation process stationary and mean-reverting toward the long-run correlation; violating it makes the dynamics explosive and the estimates unreliable.
What if I have many assets or a short sample?
For very high dimensions (N > 50) the cDCC or BEKK variants are recommended over the standard DCC. For short series (n < 250) the parameters cannot be estimated reliably, so a simpler single-series GARCH such as EGARCH is the better choice.
Sources
- Engle, R. (2002). Dynamic Conditional Correlation: A Simple Class of Multivariate GARCH Models. Journal of Business & Economic Statistics, 20(3), 339-350. DOI: 10.1198/073500102288618487 ↗
- Aielli, G. P. (2013). Dynamic Conditional Correlation: On Properties and Estimation. Journal of Business & Economic Statistics, 31(3), 282-299. DOI: 10.1080/07350015.2013.771027 ↗
How to cite this page
ScholarGate. (2026, June 1). Dynamic Conditional Correlation GARCH. ScholarGate. https://scholargate.app/en/finance/dcc-garch
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
- Copula ModelsFinance↔ compare
- EGARCHEconometrics↔ compare
- Extreme Value TheoryFinance↔ compare
- Value at RiskFinance↔ compare