Liquidity matters when measuring bank output

We develop a new method for calculating bank output that addresses the flaws of the current approach of the System of National Accounts. We implement a simple model-free method that removes the "pure" credit risk premium from the production of banks while keeping the liquidity provision as part of the total bank output. Using both local projections and autoregressive distributed lag models, we show that our method produces bank output estimates that are consistent with the evolution of the economic activity and that remain always positive including during periods of financial stress. This method satisfies the four conditions set by the Inter-Secretariat Working Group on National Accounts. Furthermore, our method reveals that the banking output of the eurozone is overestimated by approximately 40 percent over the period 2003-2017.