Volatility Spillovers and Network Connectedness among Global Systemically Important Banks

Resumo

We map the transmission of risk across the world’s Global Systemically Important Banks (G-SIBs) by measuring volatility connectedness, the natural carrier of contagion, rather than return comovement. Using Garman–Klass realized volatility built from daily ranges for 19 USD-traded G-SIBs over 2007–2026, we estimate the Diebold–Yilmaz connectedness network under three specifications – a generalized VAR, a time-varying-parameter VAR (TVP-VAR) and a LASSO/Elastic-Net VAR – and at two sampling frequencies, weekly and daily. Total connectedness is high and stable, at about 91% (weekly) and 87% (daily) of the typical bank’s forecast-error variance. The network exhibits a robust hub structure: the net transmitters of risk are the Anglo-American, Swiss and Dutch banks (UBS, JPMorgan, Barclays, ING, Citigroup, Morgan Stanley, HSBC), while the net receivers are the three Japanese G-SIBs (Mizuho, Mitsubishi UFJ, Sumitomo Mitsui) together with Deutsche Bank, Santander and Toronto-Dominion. This structure is invariant to the estimator, the VAR lag, the forecast horizon and the sampling frequency, ruling out a non-synchronous-trading artefact. A currency-robustness check shows that the US-transmitter and Japanese-absorber poles survive removal of the dollar overlay, while the net positions of the cross-listed European banks are sensitive to denomination. We trace the robust structure to the banks’ funding models, revisit and qualify the common “US-dominance” finding, and offer a market-based, contagion-oriented complement to the regulatory G-SIB framework.