Why Subprime Auto Loans Default

The average purchaser finances around 90 percent of the price of the automobile, with the average loan size being around $11,000. Repayment is highly uncertain: more than half of the loans default, and the majority of these default within the first year of repayment.

Access to credit markets is generally considered a hallmark of developed economies. In the United States, most households appear to have substantial ability to borrow. As of September 2007, U.S. households had a total of more than $2.4 trillion in non-mortgage debt. Still, economists often point to limited borrowing opportunities, or liquidity constraints, to explain certain findings about consumption behavior, labor supply, and the demand for credit.

In Liquidity Constraints and Imperfect Information in Subprime Lending (NBER Working Paper No. 13067), authors William Adams, Liran Einav, and Jonathan Levin use unique data from a large U.S. auto sales company to study credit market conditions for precisely the population that is most likely to have a difficult time borrowing: those with low incomes and poor credit histories. These consumers, who typically cannot qualify for regular bank loans, comprise the so-called sub-prime market. The authors combine proprietary data on loan applications, transactions, and repayment records from 2001 to 2004 to provide a snapshot of the market, to analyze consumer borrowing behavior, and to document the informational problems facing sub-prime lenders.

The authors use the data to document two important facts about the market. The first is that this population of consumers appears highly sensitive to cash-on-hand, or liquidity-constrained. The second is that imperfect information substantially constrains lenders in extending credit to this population.

The loan applicants in this dataset fall toward the bottom of both the income distribution and the distribution of credit scores. The U.S. median household income is on the order of $30,000 dollars; less than half of the company's loan applicants have a Fair Isaac (FICO) score above 500, whereas the national median is over 700. These kinds of low credit scores indicate either a sparse or checkered credit record. Nearly a third of the loan applicants have neither a checking nor a savings account.

The company's transaction records indicate a high demand for borrowing. The average purchaser finances around 90 percent of the price of the automobile, with the average loan size being around $11,000. Repayment is highly uncertain: more than half of the loans default, and the majority of these default within the first year of repayment. Interest rates reflect the high probability of default: a typical loan in the authors' dataset has an annual interest rate on the order of 25-30 percent.

The evidence on liquidity constraints comes in two forms. First, the authors document a striking degree of seasonality in purchasing: demand is almost 50 percent higher in February, when consumers receive tax rebate checks, than in other months. This seasonal spike in demand correlates closely with eligibility for the earned income tax credit. Second, the authors estimate that consumers' purchasing decisions are much more sensitive to immediate down payment requirements than to changes in the price of the car, which can be financed. Without liquidity constraints, only an inordinately high degree of impatience would explain these differing sensitivities.

The authors then use the data on borrowing and repayment behavior to estimate the informational problems facing lenders. They estimate that, all else equal, extending a given buyer an additional $1000 in credit increases the default rate on the loan by around 15 percent. This kind of sensitivity of repayment to loan size is the driving force in moral hazard models of credit imperfections. At the same time, a buyer who chooses to finance an extra $1000 of her purchase (that is, who self-selects into a larger loan) has an even greater default rate, around 24 percent higher than a buyer who opts to pay the $1000 dollars upfront. In other words, the decision to finance more heavily reveals additional adverse information about the likelihood of default, as in standard models of adverse selection.

The authors do not provide a welfare analysis or specific evidence on the growth in sub-prime lending that has occurred over the last decade. The last part of their paper finds that modern credit scoring techniques can go a significant distance toward mitigating adverse selection problems in the credit market, which suggests that innovations in this area may be related to the rise in sub-prime lending. Such credit scoring is less likely to mitigate moral hazard problems in repayment, thereby still restricting credit to sub-prime borrowers.

-- Les Picker