Using U.S. quarterly data, we show that in response to a positive productivity shock: (i) firms’ creation increases (ii) firms’ destruction reduces at impact, then overshoots its long-run level, peaking almost four years later above its steady-state (iii) banks’ markup reduces. To address these three facts, we provide an NK-DSGE model where firm dynamics are endogenous, the banking sector is monopolistic competitive, and defaulting firms do not repay loans to banks. We show that the interaction between firms and banks is key to replicate the empirical evidence. Contrary to conventional wisdom, in the baseline model, the effects of the shock are dampened with respect to a model without banks.