Estimating elasticities to settlement offers in a market for unsecured credit

How to cite

Gupta, Apoorv; Mezzanotti, Filippo; Zinman, Jonathan; Ghosh, Pulak. 2024. "Estimating elasticities to settlement offers in a market for unsecured credit." AEA RCT Registry. AEARCTR-0013807. https://doi.org/10.1257/rct.13807-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Apoorv Gupta* , Filippo Mezzanotti , Jonathan Zinman , Pulak Ghosh

* Primary investigator

Registry ID
AEARCTR-0013807
DOI
https://doi.org/10.1257/rct.13807-1.0
First registered
Last updated
Status
In development
Trial dates
2024-06-20 – 2025-07-01
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/13807

Abstract

We plan to identify effects of randomly assigned settlement discount offers made to delinquent consumer borrowers by a large digital fintech lender in India. The lender will randomly assign late borrowers in its portfolio to offers to settle their outstanding loan amount (inclusive of principal, interest, and other charges) with a single lump sum payment. Offers will vary in the amount of discount (i.e., haircut) on the contractually obligated amount. Estimates will help shed light on the importance of borrowers’ strategic behavior and perceived costs of defaulting.

Intervention

The experiment will include consumers (“borrowers”) who have borrowed from the lender and are 90 days to 180 days late on their repayment (“delinquent borrowers”) at the time of random assignment. The lender provides loans for maturity of 3 months, 6 months, 9 months, 12 months, and 18 months.

The company plans to experiment with randomized settlement offers on about 3000 delinquent borrowers on a monthly basis. As detailed below, settlement discount offers will range from zero to 40 %, subject to various constraints, and the lender will send each delinquent borrower at least one communication offering the borrower the option to settle her account for a specified “all in” lump-sum cash amount.

The lender will communicate offers, per its standard practices, with direct email and text messages that include instructions to contact the company for more information. Offer communications will be harmonized—in frequency, timing, and content—with the routine communications sent to the control group, to the fullest extent possible.

Experimental design

For each borrower in the sample, the lender will send a settlement offer (the “effective discount” below) based on random assignment (the “assigned discount” below) subject to the lender’s three standard operational constraints on the discount offer’s composition and maximum allowable amount:
1)on unpaid interest and late fees, offer up to a 100% haircut (i.e., complete debt forgiveness)
2)on unpaid principal, offer a maximum of:
25% haircut for 91-120 days late borrowers
40% haircut for 121-150 days late borrowers
50% haircut on loan principal for 151-180 days late borrowers

Each month, each borrower in the sample to one of five discount values: zero (control group), 10%, 20%, 30%, 40%. The randomization will be stratified across six groups. These six groups are created by dividing the sample across two variables: (a) 3 bins reflecting whether the borrower was late by 91-120 days, 121-150 days or 151-180 days during the experiment month; and (b) whether the borrower never made a non-zero payment after loan origination (“never payer”) as defined in the month they first become part of the experiment sample. However, the effective discount that a borrower will receive could be lower than the randomized discount, because of the constraints described above. To be precise, the effective discount is defined as min(Assigned Discountit, Maximum Allowable Discountit), where i indexes borrowers and t the month of randomization.

We will use the conditionally assigned discount for any intent-to-treat (ITT) analysis. However, the effective discount will be useful for descriptive purposes, for measuring compliance (in econometric terms) with the random assignment, and possibly for estimating ITT heterogeneous treatment effects and/or IV estimates.

We will analyze the take-up rate of the settlement offers using the following specification:

y_it = a + b1 * T_it1 + b2 * T_it2 + b3 * T_it3 + b4 * T_it4 + m * X_it + s_it + u_it

where y_it is one of the second-stage outcomes defined above, and T_itg is takes the value of 1 if borrower i in randomization month t belonged to the treatment arm g={1,2,3,4}. Control group is the omitted category. s_it represents strata x experiment month fixed effect, where strata is one of the 6 groups defined by combination of the three days late bin and whether the borrower was a “never payer”. X_it is a vector of indicators reflecting the borrower’s randomly assigned discount(s) in any prior month(s).

Design details

Randomization method
Randomization will be done in office by a computer.
Randomization unit
The unit of randomization is the loan account (borrower) – month.
Clusters
NA
Observations
We expect to have about 3000 late borrowers per month. However, estimating the effective sample over time is hard at this point. The company has agreed to run the experiment for about 6 months between June 2024 to December 2024.
Arms
The sample will be split equally across the 4 treatment arms and one control arm (i.e. 20% in each of the treatment and control arms).
Primary outcomes
(i) Effective discount received by the borrowers
Secondary outcomes
(i) 1(any payment): whether the borrower made any payment to the lender in the week following the offer. (ii) 1(account settled): whether the borrower paid the lender to settle the account in the week following the offer. (iii) total repayment to the lender made by the borrower in the week following the offer. We will look at both the levels of the variable, and the version where this variable is scaled by the total amount owed. We will also examine longer horizons (one month and potentially more), to the best extent possible.

Cited by

No citation data available for this registration yet.

Keywords