The Price of Asking: Social Frictions in Relational Lending and FinTech Adoption

How to cite

XU, Yuchen; Sun, Yuchen; Naso, Pedro; Wang, Mengye. 2026. "The Price of Asking: Social Frictions in Relational Lending and FinTech Adoption." AEA RCT Registry. AEARCTR-0018242. https://doi.org/10.1257/rct.18242-1.0

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

Yuchen XU* , Yuchen Sun , Pedro Naso , Mengye Wang

* Primary investigator

Registry ID
AEARCTR-0018242
DOI
https://doi.org/10.1257/rct.18242-1.0
First registered
Last updated
Status
On going
Trial dates
2025-12-13 – 2026-08-31
Countries
Burundi (Muramvya)
JEL codes
C93; D91; G21
Registry record
http://www.socialscienceregistry.org/trials/18242

Abstract

Access to credit is a central determinant of household welfare in low-income settings. Relational lending has been a main source of credit, but it is not socially costless. Borrowing from a friend or neighbor requires asking, negotiating, and potentially damaging the relationship if repayment becomes difficult. This paper uses a lab-in-the-field experiment in rural Burundi to study two such frictions: the mental cost of asking and reputational concern. We recruit real borrower-lender pairs and allow borrowers to finance repeated investment opportunities either through their peer lender or through a simulated FinTech application with fixed, anonymous terms. By exogenously varying the FinApp interest rate, we identify how borrowers’ credit-channel choices respond to the price of anonymous borrowing and recover the implied interest-rate premium associated with avoiding social frictions. We further vary network tie strength and project default risk to isolate the two underlying mechanisms. The design also allows us to study whether anonymous credit crowds out or complements relational lending and whether it improves borrower welfare.

Intervention

We conduct a lab-in-the-field experiment in rural Burundi to study the social frictions embedded in relational lending. We recruit real borrower-lender pairs drawn from existing social networks and embed them in a five-round loan game with real-value incentives. Borrowers face repeated risky agricultural investment opportunities and may finance investment either through their peer lender (via WhatsApp negotiation) or through a simulated anonymous FinTech application (FinApp) with fixed, non-negotiable terms and no interpersonal interaction. We vary three treatment dimensions: (1) the FinApp interest rate (3%, 5%, 8%, 10%, or 15%), which serves as the main price shifter for anonymous credit; (2) the strength of the borrower’s network tie to the peer lender (close tie versus less-close tie), which shifts the mental cost of asking; and (3) whether the assigned investment project places the borrower in an environment with default risk (PD) or not (PN), which shifts reputational concern. Extension arms include a relational-credit-only arm, a FinApp-only arm, a risk-matched no-default-risk project (PN1), and a no-renegotiation arm.

Experimental design

The primary experiment employs a 5×2×2 factorial design crossing FinApp interest rate (3%, 5%, 8%, 10%, 15%), network tie strength (close tie versus less-close tie), and investment project risk (default risk versus no default risk), yielding 20 cells. Borrower-lender pairs are recruited from real social networks identified in a baseline survey. In each session, borrowers and lenders are seated in separate rooms and communicate via WhatsApp for loan negotiations. Borrowers choose each round whether to borrow from the peer lender, from FinApp, or not at all. Four extension arms (relational-credit-only, FinApp-only, a risk-matched no-default-risk project, and a no-renegotiation arm) are implemented at the two benchmark FinApp rates of 5% and 10%. Final token balances determine real compensation.

Design details

Randomization method
Borrowers were recruited from a baseline survey conducted using a random-walk procedure in randomly selected collines. From the eligible baseline survey sample, borrowers were randomly selected for invitation to the experiment. Each invited borrower was randomly assigned to a tie-strength category, which determined the two eligible nominees from the borrower’s pre-listed borrowing contacts. The borrower then invited one of these two nominees to participate as the peer lender. Eligible borrower-lender pairs were then randomly assigned by computer to the remaining treatment arms using pre-generated assignment lists, with randomization stratified by borrower gender and age.
Randomization unit
Borrower-lender pair.
Clusters
640 borrower-lender pairs.
Observations
2,560 borrower-round observations for the primary round-level outcomes (640 borrower-lender pairs × 4 decision rounds), plus 640 borrower-level observations for the final-balance welfare outcome.
Arms
Primary 5×2×2 factorial (20 cells): 22 pairs per cell, 440 pairs total. This corresponds to 88 pairs at each of the five FinApp-rate arms (3%, 5%, 8%, 10%, 15%), split evenly across 2 tie-strength × 2 project conditions. Extension arms (implemented at 5% and 10% only): 10 pairs per cell, 200 pairs total, comprising 40 pairs in the relational-credit-only arm (4 cells), 40 pairs in the FinApp-only arm (4 cells), 40 pairs in the PN1 arm (4 cells), and 80 pairs in the no-renegotiation arm (8 cells).
Primary outcomes
(1) Negotiation initiation: a binary indicator equal to one if the borrower initiates a WhatsApp loan request to the peer lender in a given round. (2) FinApp take-up: a binary indicator equal to one if the borrower borrows from FinApp in a given round. (3) Final token balance: the borrower’s total token balance at the end of the five-round game (main welfare outcome).

Cited by

No citation data available for this registration yet.

Keywords