Causality in Variance Test
Test for Causality in Variance · Also known as: Volatility spillover test
The causality-in-variance test detects whether shocks to one variable cause changes in the conditional variance (volatility) of another variable, distinct from mean-level causality. Introduced by Cheung and Ng (1996), it identifies volatility spillovers and contagion effects—crucial for risk management and understanding financial market interdependencies. This approach has become standard in studying shock transmission across asset classes and geographies.
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 causality-in-variance test when studying volatility spillovers—e.g., testing whether stock-market shocks increase currency volatility, whether bank failures increase credit-spread volatility, or whether oil-price shocks increase manufacturing-output uncertainty. It is essential for risk modeling and systemic financial risk assessment.
Strengths & limitations
- Captures variance causality distinct from mean effects
- Identifies contagion and shock transmission via volatility
- Straightforward implementation using standard GARCH software
- Economically meaningful for risk management and financial stability
- Requires parametric modeling of volatility (GARCH or similar); model mis-specification biases tests
- Small-sample properties can be poor; requires large T for reliable inference
- Multiple testing (testing many pairs of variables) requires multiple-comparison adjustments
- Conditional vs unconditional variance causality distinction can be subtle in interpretation
Frequently asked
Is causality in variance the same as Granger causality in squared residuals?
Conceptually similar, but causality-in-variance formally tests conditional variance relationships using GARCH models, whereas Granger causality is mean-based. The variance test is more powerful for detecting volatility spillovers.
Should I use univariate or multivariate GARCH?
Multivariate GARCH (e.g., DCC, BEKK, Copulas) is more flexible but requires larger samples. Start with univariate GARCH if sample is limited or variables are many.
How do I correct for multiple testing?
Use Bonferroni or false-discovery-rate (FDR) adjustments if testing many variable pairs. Alternatively, focus on economically motivated pairs to reduce testing.
What if GARCH models don't fit well?
Try alternative volatility models (exponential GARCH, stochastic volatility). If issues persist, results may be unreliable; report robustness across specifications.
Sources
- Cheung, Y. W., & Ng, L. K. (1996). A causality-in-variance test and its application to financial market prices. Journal of Econometrics, 72(1-2), 33-61. DOI: 10.1016/0304-4076(94)01714-X ↗
- Hafner, C. M., & Herwartz, H. (2006). Testing for causality in variance using multivariate GARCH models. Journal of Econometrics, 135(1-2), 129-153. link ↗
How to cite this page
ScholarGate. (2026, June 3). Test for Causality in Variance. ScholarGate. https://scholargate.app/en/econometrics/causality-in-variance-test
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.
- Component GARCHEconometrics↔ compare
- DCC-MIDASEconometrics↔ compare
- GARCH-MIDASEconometrics↔ compare