The benefits of banking relationships: evidence from Uganda’s banking crisis. Mimeographed document

Four Ugandan banks, including three domestic banks, were closed between September 1998 and May 1999 because of imprudent banking practices. This paper uses a unique bank-firm matched data set to estimate the effect of losing a banking relationship on firm performance. Employing a fixed effects difference-in-differences estimation that controls for unobservable firm heterogeneity, I find that firms that lost a banking relationship declined by 10-15 % relative to unaffected firms over the three years following the crisis. This result is robust to reverse causation; for a fairly well identified subset of firms, I find no evidence of firm decline causing banking failure. I investigate two potential explanations of this result: the information and looting views of relationship lending. I find evidence for both views for mutually exclusive sub-samples. Insider firms experience the sharpest decline in employment relative to all affected firms. The insider effect persists even after controlling for the level of outstanding debt to failed banks. I interpret this as evidence of the looting view. In the set of potential non-looting firms, I find that older affected firms have larger declines in employment relative to younger affected firms. In addition, affected firms that do not produce hard information have the largest declines in employment relative to other affected firms. Moreover, affected firms that produce hard information have the same growth rate as unaffected firms. Finally, affected firms are more likely to report being credit constrained in the post crisis period. I interpret this as evidence for the information view