Diebold and Yilmaz (2009) measure volatility spillovers using externally constructed volatility measures within a VAR framework. However, treating these noisy estimates as observed regressors introduces a generated-regressor problem. Furthermore, external proxies are susceptible to outliers and often unavailable for lower-frequency data, limiting the VAR approach. This paper proposes an alternative spillover index computed directly within a multivariate ECCC-GARCH framework, eliminating the need for external proxies. By modeling volatility dynamics internally, our approach provides a more robust and methodologically consistent measure. An application to G5 stock markets demonstrates that the proposed index is less sensitive to outliers and avoids the artificial spikes prevalent in proxy-based VAR models.