Experimental Design Details
The assignment of consumers to treatment legs is done prior to the consumer becoming delinquent. We assign all the customers at the bank (1m+) to 8 bins, a {0,1} bin with respect each contract feature: rate, term, relief. We do this by drawing three random numbers for each contract feature, and assigning a customer to group 1 if the customer is above a specific threshold. This threshold equals 0.5 for rate and maturity, 0.66 for relief. This leads to 8 groups of participants.
When randomizing, we stratify the participants into non-overlapping and exhaustive bins with respect to variables S. The variable set S includes the number of late pay days; outstanding balance D and city of residence, latter included in order to capture local economic shocks. Finally we check for balance between H and L treatments, for each treatment leg, by comparing the difference in means, and the regression coefficients of assignment to treatment on observables.
The participants in the experiment are 25,000 consumers that fall into delinquent status as of or after June 2017. These customers have failed to make a payment on at least one debt contract and are being monitored by the bank. They have been previously nudged via text messages, and are contacted via phone calls to remind them of their payments.
At the point when the customer is late, we anticipate a restructuring and assign a new rate, and maturity for the potentially restructured contract.
- Interest rates. The restructuring features an interest rate reduction to R' < R. Consumers in the four high rate treatment legs get a 60bps APR rate reduction, and the typical contract in the low rate treatment group gets a 480bps APR rate reduction. However, the new minimum rate is bounded below by an experimental minimum set by the financial institution.
- Maturity. The restructured contract features a maturity extension offer to T'>T. When assigning customers to maturity offer bins, we first group to grids of width 12 with respect to the remaining maturity of the old contract. We then offer a maturity extension of 150% to those in the low maturity treatment, and a 200% to those in the high maturity treatment. The offer is bounded above by the maximum maturity set by the regulatory authority, 48 or 72 months depending on when the old contract is originated.
- We select a subset of customers for potential temporary payment relief. For these customers, we offer to postpone the payment of the principal in the first quarter, and the total payments for this group in the first quarter then consists only of interest payment on the principal, DR'. The amortizing payments for these consumers start in month 4.
After a new contract rate and maturity has been assigned, if the consumer still has not made a payment after 60 days, the financial institution contacts the consumer through an in-house call-center for a potential restructuring. The call-center representative asks the customer about the nature of their financial distress, and asks if the customer is interested in restructuring the contract. At the restructuring screen, the call-center employee sees the outstanding balance, the new interest rate and a maturity offer on her screen. The employee and customer discuss how much the customer could pay each month, where the employee offers maturity T'.
If they agree on the contract term, before they finalize the restructuring, the customers in the relief group are offered 3 months of principle payment relief. The customer can then accept or reject the relief offer. If the customer takes the relief offer, payments in the first three months equal $D R$, interest on the principal, and payments after month $t=3$ are given by the mortgage formula, with $T-3$ equal payments, as the payment relief is offered only after the customer agrees on an interest rate and maturity. After the payment schedule is determined, the contract is forwarded to processing.