Many of the banks that failed in the years 1985–1990 borrowed from the Federal Reserve for extended periods in their last year. This article tests hypotheses about the determinants of borrowings by banks that failed in these years. Results are consistent with the hypothesis that borrowings were greatest among the banks with the greatest liquidity needs in their last year. They do not support the hypothesis that the Fed favored member banks in its allocation of credit to troubled banks. The results indicate significant variation in lending practices across Federal Reserve districts, and there is weaker evidence of variation in lending practices across time. The rate of bank failure in the second half of the 1980s and early 1990s was high relative to failure rates in earlier decades. Many of the failed banks borrowed from the Federal Reserve for extended periods in their last year. Of the sample of failed banks in this study, 58% borrowed at some time in their last year, and 48% borrowed in their last three months. In most cases, the Federal Reserve would have been aware of the financial problems of these banks when lending to them, based on the supervisory ratings of the condition of the banks. Congress acted in 1991 to restrict Federal Reserve lending to undercapitalized banks, in the Federal Deposit Insurance Corporation Improvement Act of 1991 (FDICIA). These restrictions were based on the view that Federal Reserve lending to undercapitalized banks increased the losses of the FDIC in bank failure cases. There was a lot of variation among the borrowers in terms of the length of time they borrowed and average borrowings relative to their total deposits. This variation makes it possible to test hypotheses about the borrowings of these banks near the time of their failure. One hypothesis is that the Federal Reserve made credit available to the troubled banks with the greatest liquidity needs. Banks with liquidity needs have exhausted most of their liquid assets, must draw down reserves to pay depositors who are withdrawing funds, and cannot raise funds in the private sector. Fed lending to the troubled banks with the greatest liquidity needs would have given supervisors time to determine which banks to close and the methods for resolving the failed bank cases. 1 1 While the Federal Reserve describes its discount window policy as extending credit only to solvent banks, Sprague (1986) gives an insider's account of how the Federal Reserve worked with the other bank supervisory agencies to extend the lives of major banks in serious financial trouble through discount window loans. Another hypothesis is that variation among the banks in the patterns of their borrowings reflected preferences of the Fed to aid some banks rather than others, such as banks that were members of the Federal Reserve System. Yet another hypothesis is that the variation in patterns of borrowings reflected differences in Fed practices across districts and across time in lending to troubled banks. Tests of these hypotheses do not indicate whether the practice of Federal Reserve lending to troubled banks was good policy. 2 2 For criticisms of Federal Reserve lending to troubled banks, see Schwartz (1992), Todd (1988, 1991, 1992), and U.S. House of Representatives (1991a, 1991b, 1991c). These tests, however, may shed light on the factors that motivated the Fed to lend to troubled banks.