The main hallmarks of the global financial crisis were too-big-to-fail institutions taking on too much risk with other people’s money while gains were privatised and losses socialised. It is shown that banks need little capital in calm periods, but in a crisis they need too much – there is no reasonable ex-ante capital rule for large systemically important financial institutions that will make them safe. The bank regulators paradox is that large complex and interconnected banks need very little capital in the good times, but they can never have enough in an extreme crisis. Separation is required to deal with this problem, which derives mainly from counterparty risk. The study suggests banks should be considered for separation into a ring-fenced non-operating holding company (NOHC) structure with ring-fencing when they pass a key allowable threshold for the gross market value (GMV) of derivatives, a case which is reinforced if the bank has high wholesale funding and low levels of liquid trading assets. The pricing of derivatives and repos would become more commensurate with the risks if the NOHC proposal were to be pursued as a unifying strategy for the different national approaches. Most of the objections to this structure are summarised and rebutted. Other national proposals for separation in Switzerland, the Volcker rule, the Vickers rule, and the Liikanen proposal are argued to be inferior to the ring-fenced NOHC proposal, on the grounds that empirical evidence about what matters for a safe business model is not taken properly into account. JEL classification: G01, G15, G18, G20, G21, G24, G28 Keywords: Financial crisis, derivatives, bank business models, distance-to-default, structural bank separation, banking reform, GSIFI banks
