Payday Loans, Consumption Shocks, and Discounting

Payday loans are high-interest rate, short-term loans which mature on borrowers' paydays, have terms equal to the duration of one pay cycle, and are collateralized with post-dated personal checks. Roughly nine million American households borrowed on payday loans in 2002, typically paying annualized interest rates above seven thousand percent. With a unique new dataset of nearly two million payday loan applications, we examine two hypotheses for why consumers use these extremely expensive financial instruments. First, consumers may experience shocks to consumption needs like expenses for health care or car repairs. These shocks could raise the marginal utility of consumption enough to account for borrowing at very high interest rates. Second, consumers may have strong preferences for immediate consumption over consumption in the future. To test these hypotheses formally we build a dynamic structural model of consumption and borrowing behavior. In our benchmark specification sophisticated agents have access to liquid assets and face income uncertainty as is standard in this literature. Consumers also face borrowing constraints except for the opportunity to borrow on institutionally realistic payday loans. As the key innovation in our benchmark model, we incorporate stochastic consumption shocks. We use a numerical backward induction algorithm to solve for the stationary Markov Perfect Equilibrium of the model, and with these MPE strategies we simulate the behavior of a population of consumers. Moments calculated from the simulated and empirical populations permit us to estimate the free parameters of the structural model using a two-stage Method of Simulated Moments (MSM) procedure (Gourinchas and Parker 2002, Laibson, Repetto, and Tobacman 2004). We find that consumption shocks alone cannot account qualitatively or quantitatively for observed payday borrowing behavior, so we examine two extensions of the benchmark model. When we permit overoptimism about future consumption shocks and higher discount rates in the short run than in the long run, the model's fit improves and we statistically reject the benchmark model's assumptions. Since the median income of payday borrowers in our sample is about $20,000 per year, our results contribute important insights about the circumstances and choices of low-income decision-makers