The Financial trailblazing of Digital Finance in Rural China: The Stepping Stone Effect

How to cite

Zhang, Longyao. 2023. "The Financial trailblazing of Digital Finance in Rural China: The Stepping Stone Effect." AEA RCT Registry. AEARCTR-0011539. https://doi.org/10.1257/rct.11539-1.0

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

Longyao Zhang*

* Primary investigator

Registry ID
AEARCTR-0011539
DOI
https://doi.org/10.1257/rct.11539-1.0
First registered
Last updated
Status
On going
Trial dates
2020-09-01 – 2023-08-31
Countries
China ()
JEL codes
G14, G20, G24
Registry record
http://www.socialscienceregistry.org/trials/11539

Abstract

This study draws from the approach of Karlan et al. (2011), using an algorithm (a computer program) to seek out a group of "marginal reject lenders" among customers applying for digital finance for the first time. Subsequently, a Randomized Controlled Trial (RCT) is conducted on this group of "Marginal Reject Lenders", randomly approving digital finance applications to assess the marginal effect of digital finance.

The reason for focusing on "marginal reject lenders" is that directly carrying out an RCT targeting all loan applicants would be costly. Moreover, the RCT may reject some applicants who could have been approved in regular business procedures, thus harming the welfare of these applicants.

To address this, this study aims to find among those rejected by normal business rules, a group of loan applicants who exhibit similar characteristics, have comparable credit qualifications, but due to minor issues, fall slightly below the loan approval threshold (the "marginal reject lenders" group). By implementing an additional randomized digital finance approval intervention (either approval or continued rejection), we form corresponding digital finance treatment and control groups. By comparing the relative changes in welfare performance indicators between the two groups, we attribute these changes to the marginal contribution of digital finance.

Intervention

This study, in conjunction with the sample financial institution, develops a credit scoring algorithm to identify "marginal reject lenders". The algorithm uses credit-related business indicators, combined with the requested loan amount, term, and repayment method. It trains and formulates this algorithm based on the default or non-default history of the financial institution's previous customers. This algorithm is then used to credit score first-time digital credit applicants.

Labels of "major reject lenders" and "marginal reject lenders" are assigned based on this algorithm to those who apply for digital credit for the first time but have their applications rejected. "Major reject lenders" are loan applicants whose qualifications are far below the loan threshold due to historical defaults, fraud, poor credit records, and other factors. "Marginal reject lenders", on the other hand, are a group of loan applicants whose credit qualifications are slightly below the loan threshold.

In the group of "marginal reject lenders", digital credit loan applications are randomly approved, constituting the treatment group; the remaining digital credit loan applications continue to be rejected, forming the control group.

Experimental design

This study, in conjunction with the sample financial institution, develops a credit scoring algorithm to identify "marginal reject lenders". The algorithm uses credit-related business indicators, combined with the requested loan amount, term, and repayment method. It trains and formulates this algorithm based on the default or non-default history of the financial institution's previous customers. This algorithm is then used to credit score first-time digital credit applicants.

Labels of "major reject lenders" and "marginal reject lenders" are assigned based on this algorithm to those who apply for digital credit for the first time but have their applications rejected. "Major reject lenders" are loan applicants whose qualifications are far below the loan threshold due to historical defaults, fraud, poor credit records, and other factors. "Marginal reject lenders", on the other hand, are a group of loan applicants whose credit qualifications are slightly below the loan threshold.

In the group of "marginal reject lenders", digital credit loan applications are randomly approved, constituting the treatment group; the remaining digital credit loan applications continue to be rejected, forming the control group.

Design details

Randomization method
The "marginal reject lenders" group is identified through a computer-developed algorithm. Furthermore, the division of the "marginal reject lenders" into the treatment group and control group is also achieved through random assignment by a computer.
Randomization unit
Individual
Clusters
4000 "marginal reject" digital credit applicants
Observations
4000 "marginal reject" digital credit applicants
Arms
680 applicants in treatment group and 3320 applicants in control group
Primary outcomes
Farmers' access to formal credit; Farmers' access to formal credit; Borrowing costs for farmers
Secondary outcomes
Farmers' financial literacy

Cited by

No citation data available for this registration yet.

Keywords