Friday, June 13, 2008

The impact of credit constraints

How important are credit constraints for households? This is not obvious to determine empirically due to a host of issues: one observes only successful applications, data centers typically on consumption, or it is about loans, it is based on opinion surveys that Economists are rightfully uneasy about. In a paper just published in the International Economic Review, Attanasio, Goldberg and Kyriazidou found a nifty way to address better the question.

Specifically, they exploit data in the Consumer Expenditure Survey that includes auto loans contracts. Credit constrained households are limited by how much they can borrow, and are little affected by interest rates. Those that are not constrained are able to optimize how much to borrow and are responding more strongly to interest rates. Think about it: the choices of the first is really determined by the kink in the budget constraint, and by its slope for the latter. The data they use contains large heterogeneity in interest rates and maturities, the latter being highly correlated with the size of the loan.

Results? Increasing maturity by one year increases loan demand by about 90%, and the demand elasticity with respect to the interest rate is undistinguishable from zero. These results are very strong and indicative of strong aggregate credit constraints. Age does not matter, surprisingly, but income does, obviously.

What does this mean? We know the permanent income hypothesis does not hold strictly due to credit constraints. A model with such constraints, however, should not pile them on the young, but on the poor.

1 comment:

Vilfredo said...

I wonder what these results imply with respect to the Ricardian Equivalence. We know that some households are credit constrained, and thus initiatives like the recent fiscal stimulus will have some impact. But it is the size of this impact that is questioned. I do not know how to interpret the results of this paper to evaluate how strong the deviation from the Ricardian Equivalence is.