I examine the causal effect of bank failures during the Great Depression using the quasi-experimental setup of Richardson and Troost (2009). The experiment is based on Mississippi being divided into two Federal Reserve districts, which followed different policies for liquidity provision. This translated into variation in bank failures across the state. Employing a plant-level sample from the Census of Manufactures, I find that banking failures had a negative effect on revenue stemming from a fall in physical output. I find no effect on employment at the plant-level and a large decline at the county-level. (JEL E32, E44, G21, G33, N12, N22, N92) B anking crises are often associated with declines in the aggregate economy, but how much do bank failures exacerbate these downturns? And at what cost can policy mitigate these deleterious effects? These questions are starkest in the context of the Great Depression where an unprecedented decline in output was paired with an unprecedented collapse in the banking sector and serious questions about policy choices. The problem in addressing this question during the Depression, or at any other point in time, is one of endogeneity. Did bank failures drive business failures or the reverse, with business failures leading to bank insolvencies? Furthermore, in order to formulate an appropriate policy response, if bank failures do lead to declines in economic activity, what is the mechanism?While Keynes and other authors writing around the time of the Depression believed the arrow of causality ran from banks to firms, this was downplayed by future generations of economists who saw the banking crisis during the Depression as a sideshow. The seminal work by Friedman and Schwartz (1971) reinvigorated the old banking-led hypothesis and simultaneously indicted policymakers for their