Recurring-Payment Sensitivity in Household Borrowing

I develop a test for payment sensitivity that exploits differences in predicted bunching at kinks and notches generated by mortgage insurance requirements. I find that borrowing is substantially more responsive to nominal recurring payments than to the net present value of total costs. To rationalize the result, outside borrowing costs would have to be implausibly high, exceeding 40 percent a year. Payment sensitivity is the most likely explanation for observed borrowing choices, as alternatives require implausible nonmortgage borrowing costs or household preferences. I develop a dynamic consumption-savings model and show that underlying preferences can generate the observed payment sensitivity only if borrowers initially have a high marginal utility of cash-on-hand that coincidentally and sharply falls by more than 50 percent in a narrow time window after loan origination. Payment sensitivity has important implications for regulation and policy. Lenders can manipulate loan features and shroud increases in total costs from payment-sensitive borrowers, even while keeping fixed or even decreasing recurring payments. This type of shrouding could enable excessive borrowing and attenuate the transmission of monetary policy.