Auto-stabilizing debt repayment schedules for common shocks: A large-scale field experiment

The intervention is a temporary loan modification triggered by an identified negative common shock, coupled with an ex-ante commitment to make the modification. The modification reduces the minimum required loan installment payment by approximately 75% for 3 months, with the loan term extending to accommodate the lower principal amount paid in the modification months (“the stabilization period”). The loan installment will revert to the account’s original pre-treatment amount, approximately, after the three-month stabilizing period. Borrowers can pay more than the minimum at any time, without a prepayment penalty, so the intervention is designed to give borrowers flexibility to weather an exogenous, temporary shock.

The experiment focuses on a set of small, privately-owned businesses, providing mass transport services, that have a vehicle loan with our partner lender (“partner”). The typical borrower is a closely-held small business with one owner-operated minibus taxi or a small fleet. Borrowers already have GPS devices installed in their loaned vehicles due to credit contract covenants.

The unit of randomization is an operating area for minibus taxi services: a route or cluster of overlapping/contiguous routes. The partner decided to run the experiment on 118 route-clusters (with 3,024 loan accounts): 34 route-cluster (with 942 loan accounts assigned to them at baseline) in Metro Area 1 and 84 route-clusters in Metro Area 2 (with 2,082 loan accounts assigned to them at baseline). Identifying routes, and mapping borrowers to where they operate, are technically challenging tasks, and we have developed algorithms to do so.

For each metro area, the experiment randomly assigns half of the route-clusters with at least 5 loan accounts from our partner lender to the treatment group, with remaining half to the control group. We stratify by one variable (for each metro area separately): an indicator for whether a route-cluster has more than the median number of loans provided by our lender.

The partner sends communications to treatment-group accounts before any triggering shock, then again if a shock is identified and the stabilizing modification will be triggered, and then again during the temporary modification period. Borrower take-up is opt-out: customers on route-clusters assigned to treatment arm will receive the auto-stabilizing modification, in the event of an identified negative shock, unless they opt-out beforehand. Text messages summarize the stabilizer and link to an account-customized pdf providing additional details and a summary visualization of how the new product feature works, from the borrower’s perspective. Per standard practice, communications also provide contact info for specially-trained customer service representatives who can answer questions and handle opt-out requests. Treatment-group accounts on routes that have not (yet) been triggered will also receive a monthly communication similar to what they received at program inception and that will serve as a reminder that their contract features the commitment of loan modification from the lender in the event of a trigger in the future. Treatment group also receives all business-as-usual communications (of which there are many).

Control group accounts receive only business-as-usual communications.

2026-04-01

2027-09-30

We will study the following outcomes during three periods that span the duration of our experiment: (a) pre-trigger period (i.e. the period before an account’s route is triggered), (b) stabilization period (i.e. the period of three months of lower installment following the triggered month), (c) post-stabilization period (i.e. the period after the loan account reverts back to its original installment). We also plan to use an alternative definition of the stabilization period based on the duration of the shock instead of three months.

(1) Loan performance: First, we plan to examine outcomes that capture the repayment activity by borrowers. This includes (i) how much borrowers pay month by month; (ii) total arrears; (iii) whether the borrower is current on the loan; (iv) scaled arrears (=arrears/ required monthly payment); (v) a summary index of above measures.

(2) Effort Outcomes: Second, we plan to examine the outcomes of firm’s effort measure during the intervention period. For instance, using the lender’s proprietary data on borrowers’ driving, we can observe the distance driven, the time on job, and the number of days worked.

(3) Repayment and borrowings from other debt sources: We will analyze spillovers on other forms of borrowings reported in the credit bureau data. This includes utilization of outside credit lines, any adverse flag, amount of debt overdue, late payments, new account openings, the share of borrowings overdue, total borrowings, and defaults on those debts.

We plan to do this at the loan-account level (for the sample of eligible accounts at baseline) as well as at the level of route-clusters.

Mapping financed vehicles to routes: At baseline, each loan account is assigned to an operating area (route-cluster aka “route” below) based on proprietary algorithm that utilizes detailed driving behavior from installed telematics devices in the vehicles. We use information from last six months of driving to assign the most-frequented route for each vehicle.

Random assignment: We then identify experiment-eligible routes (those with >=5 loans financed by our lender at baseline) and assign each to Treatment or Control (50-50 probability) at baseline. An account’s assignment to a given route, and thus to T or C, will remain fixed during the duration of the experiment.


Measurement of stabilizer-triggering events: Our goal is to identify route-months where there is a substantial decrease in total activity on the route. We do this by first assigning all financed vehicles to a primary route this month (updating this dynamically each month, in contrast to our random assignment), and then calculate the median of the total distance driven by vehicles assigned to a route during the month. We then define a route to be triggered if the percentage difference between this measure in the last month and its mean over the past twelve months is below a threshold chosen by the lender ex-ante. Borrowers do not know the exact threshold: it is communicated to them as a “major disruption in your operating area”. The partner initially agreed on a 15% threshold (i.e., triggered routes are those where driving declines by more than 15%) but decided to switch to a more generous 10% threshold after the first month of implementation.

Eligibility to receive a loan modification: Receiving a modification is conditional on clearing eligibility criteria determined by the lender. In particular, an account on a treated triggered route is ineligible to receive a modification during that month if the account is under default status (90+ days delinquent). To be eligible to receive modification the account has to be less than 90 days delinquent on their loan. If not, borrowers are provided information on the minimum payments they would need to make to be eligible. As of now, each account is eligible of one loan modification during the experiment duration of 12 months, but this might be modified during the experiment duration due to the recent fuel cost shock and/or other considerations.

For the 12-month lookback period and each current month, for classifying the route as triggered or not, we only include route-months with at least 3 vehicles, since requiring more than two limits the chance of misclassifying a route as shocked on the basis of too few observations.

Not available

Randomization done in office by a computer.

The unit of randomization is at the route level. All eligible loan accounts (i.e. loaned vehicles) assigned to the same route receive the same offer. Loan account’s treatment assignment remains fixed throughout the duration of the experiment.

Yes

118 routes

3,024 loan accounts

1,483 loan accounts; 59 routes (control)
1,541 loan accounts; 59 routes (treatment)