This thesis addresses the independence of bank supervisory agencies and its effects on bank soundness in industrial countries. It tackles the following underresearched questions: whether and, if so, to what extent delegation to independent bank supervisors strengthens banking systems; what forms delegation arrangements take: and \\ hich politico-institutional and banking market conditions affect the etTecti\ eness of delegation. The study isolates theoretically and empiricall) the role of agency independence as a source of credibility in banking supervision. The quantitative analysis builds on a new cross-sectional data set on legal bank supen isor independence from government and industry in OECD countries. The qualitative anal) sis opts for a two-country (the UK and Japan), two-period (1980-1998 and 1998-2006) case study framework. The Bank of England, the Ministry of Finance of Japan. and their respective successor bodies, the Financial Services Authority and the Financial Services Agency, are strongly differentiated according to their independence profile. The thesis offers qualified support for independence as a prerequisite for effective banking supervision, in line with the virtually untested policy prescription contained in the Basel Core Principles. Quantitative evidence suggests that higher bank supervisor independence from government enhances bank soundness, and does so the more prevailing is the rule of law. Delegation is more effective in bank-centred, less concentrated financial systems. Legal independence is rarely complete, suggesting that countries have a variety of models of partial delegation to choose from to suit their particular accountability needs. Qualitative evidence provides further support for the beneficial effects of delegation. It suggests that \vhile legal independence from industry is not a necessary condition for bank soundness, supervisory capture, at least above some critical thresholds. is c1earl) detrimental. The transmission mechanisms between independence and bank conditions are discllssed, including the possibility of two-way causality .