Valuing volatility spillovers

George Milunovich, Susan Thorp

Research output: Contribution to journalArticle


We measure the reduction in realized portfolio risk that can be achieved by allowing for volatility spillover in forecasts of equity covariance. The conditional second moment matrix of equity returns for pairs of major European equity markets is estimated via two asymmetric dynamic conditional correlation models (A-DCC): the unrestricted model includes volatility spillover effects and the restricted model does not. Data are daily returns on the London, Frankfurt and Paris equity market price indices synchronized at London 16:00 time. Covariance forecasts from the restricted and unrestricted models are combined with assumed expected returns to compute efficient three-asset portfolios (two equity indices and the risk-free asset). The impact of expected return choice on out-of-sample portfolio efficiency is minimized via the polar co-ordinates method of Engel and Colacito (2004), which allows expected equity returns to span all relatives. Out-of-sample realized portfolio returns and variances from efficient portfolios are computed and tested. Allowing for volatility spillover effects produces small, statistically significant reductions in portfolio risk. Portfolio standard deviations for the unrestricted model are at most one per cent smaller than standard deviations for restricted models. Significant risk reductions persist across daily, weekly, and monthly rebalancing horizons. Tests for second degree stochastic dominance indicate that realized returns from portfolios based on the volatility spillover model would be preferred by risk averse agents.
Original languageEnglish
Pages (from-to)1-31
Number of pages31
JournalMacquarie economics research papers
Publication statusPublished - 2005


  • portfolio risk
  • volatility spillover
  • equity return
  • asymmetric dynamic conditional correlation models (A-DCC)
  • expected equity return


Dive into the research topics of 'Valuing volatility spillovers'. Together they form a unique fingerprint.

Cite this