We investigate the impact of fiscal expansions on firm investment through corporate lending by exploiting highly granular Colombian data during 2004-2015. For identification, we focus on firms that have multiple banking relationships and trace the entire loan history across lenders as they change their stock of government bonds. Further, we conduct a localized RDD approach and compare the lending behaviour of: (i) banks that barely met and missed the criteria of being a primary dealer, and (ii) barely winners and losers at government auctions. We argue that the lending strategy of banks within the vicinity of the cutoff point is very similar ex ante and can reveal crowding-out effects when confronting liquidity shortages. Given the richness of the data set, we further investigate whether the effects of high government borrowing is heterogeneous across firms. Our findings are grounded in a quantitative default model with financial and real sectors. Government spending limits the amount of available funds to firms and primary dealers then pass on these costs to local firms. Essentially, the heightened cost of credit lowers private investment.