One of the remarkable consequences of the recent financial crisis has been the shift in emphasis, in economic analysis as well as policy practice, from the microprudential to the macro-prudential approach to banking regulation and supervision (Hanson et al. (2011)). The micro-prudential approach focuses on controlling the risk of individual banks, taking the rest of the system as given. Since banks are funded largely by deposits, an instrument protected in most countries by a collectively funded safety net, the micro-prudential regulator/supervisor needs to control that bank managers do not take advantage of such protection to assume an unduly high level of risk (moral hazard). However, in the presence of risk externalities, i.e. when risk-taking behavior of individual institutions affects other institutions and the system as a whole, another source of social inefficiency arises, requiring the prudential regulator/supervisor to take a macro-prudential approach. Banks need to be controlled not only in relation to their propensity to undertake high risk due to the protection they enjoy on their liability side, but also in relation to the risk they transmit to other banks with which they are connected by a reciprocal web of exposures. To guide policy it becomes necessary, in this context, to measure the systemic importance of individual banks, i.e. the degree to which they propagate systemic risk by exerting influence on the rest of the system. A large body of literature has developed to define and measure systemic risk and to analyze its implication for the conduct of prudential regulation and supervision (for a comprehensive overview see Bisias et al. (2012)). Broadly speaking, there are two classes of indicators of systemic importance proposed in the recent literature. The first and more widely used class of indicators relies, directly or indirectly, on asset prices distributions and correlations across institutions, assets and time. The second approach, with a rapidly rising literature is that of network analysis Our contribution here consists in proposing an alternative method for measuring the transmission of risk within banking systems, which logically belongs to the second class mentioned in the preceding paragraph, namely that based on balance sheet measures of interconnection. Our approach is new and old at the same time. It is new because, as such, it has never been used in the analysis of banking. It is old because it draws on a well-established and time-honored strand of literature, that of input-output analysis.Taking the balance sheet of the banking system as a point of departure, we derive expressions that closely resemble the traditional Leontief (1941) input-output model. The input-output model makes evident the fact that the production of any sector has two distinct effects on the remaining sectors: on the one hand by increasing production it will demand more inputs from other sectors ("upstream"), on the other it will be able to provide more output to the sectors that depend its production as in...