We assemble data on the structure of bank supervision, distinguishing supervision by the central bank from supervision by a nonbank governmental agency and independent from dependent governmental supervisors. Using observations for 140 countries from 1998 through 2010, we find that supervisory responsibility tends to be assigned to the central bank in low-income countries where that institution is one of few public-sector agencies with the requisite administrative capacity. It is more likely to be undertaken by a non-independent agency of the government in countries ranked high in terms of government efficiency and regulatory quality. We show that the choice of institutional arrangement makes a difference for outcomes. Countries with independent supervisors other than the central bank have fewer nonperforming loans as a share of GDP even after controlling for inflation, per capita income, and country and/or year fixed effects. Their banks are required to hold less capital against assets, presumably because they have less need to protect against loan losses. Savers in such countries enjoy higher deposit rates. There is some evidence, albeit more tentative, that countries with these arrangements are less prone to systemic banking crises. In 1997-8 the British government transferred responsibility for supervising commercial banks from the Bank of England to a newly-created independent supervisory agency, the Financial Services Authority (FSA). In 2010-11 it then took steps to transfer that authority back, creating a Financial Policy Committee within the Bank and establishing a Prudential Regulatory Authority as a Bank of England subsidiary while refocusing the FSA on consumer protection (Ferran 2011). This reversal reflected distinctive aspects of the UK’s experience, notably the failure of supervisors to head off problems in a prominent building society, Northern Rock. It reflected their failure to prevent the first bank run in the United Kingdom in more than a century, an outcome widely if not unanimously attributed to imperfect coordination and inadequate information sharing between the FSA and the Bank of England. More generally, this vacillation was indicative of uncertainty about the appropriate locus of responsibility for bank supervision – an uncertainty was by no means limited to the UK – and of a dearth of evidence about what works best.