We evaluate the mechanisms behind relationship lending and its macroeconomic con-sequences. Using merged credit registry and firm data for Chile, we find that closer bank relationships give firms easier and cheaper access to credit and that more productive firms select into relationships. We build and calibrate a dynamic model where firms choose their relationship status along with investment and borrowing. Borrowing in relationships allows for greater monitoring, provides implicit guarantees to other creditors and substitutes for physical collateral. Our model can account for about 40percent of the difference in observed outcomes of firms with bank relationships. Counterfactual experiments indicate that eliminating the benefits of relationship contracts to all firms results in an output loss of about 20 percent.