This paper examines the information content of bank loan agreements. We differentiate borrowers according to financial analysts' percentage earnings forecast errors and most recent forecast revisions. The empirical results suggest that banks rely on other indicators as initial screening devices to determine where to best deploy their evaluation and monitoring efforts. If these other indicators are reliable and signal-improving prospects, banks do little further investigation. However, if the indicators are noisy and signal-declining prospects, banks have incentives to expend resources to investigate the borrowers, resulting in the production of valuable information.