We study a variant of the one-sector neoclassical growth model of Diamond in which capital investment must be credit financed, and an adverse selection problem appears in loan markets. The result is that the unfettered operation of credit markets leads to a one-dimensional indeterminacy of equilibrium. Many equilibria display economic fluctuations which do not vanish asymptotically; such equilibria are characterized by transitions between a Walrasian regime in which the adverse selection problem does not matter, and a regime of credit rationing in which it does. Moreover, for some configurations of parameters, all equilibria display such transitions for two reasons. One, the banking system imposes ceilings on credit when the economy expands and floors when it contracts because the quality of public information about the applicant pool of potential borrowers is negatively correlated with the demand for credit. Two, depositors believe that returns on bank deposits will be low (or high): these beliefs lead them to transfer savings out of (into) the banking system and into less (more) productive uses. The associated disintermediation (or its opposite) causes banks to contract (expand) credit. The result is a set of equilibrium interest rates on loans that validate depositors' original beliefs. We investigate the existence of perfect foresight equilibria displaying periodic (possibly asymmetric) cycles that consist of m periods of expansion followed by n periods of contraction, and propose an algorithm that detects all such cycles.