"Alternative Credit Assessment Requires Technological Advancement and Policy Support"

There are opinions that, in order for alternative credit assessment?which measures credit ratings using non-financial information such as consumption and rent payment records instead of traditional financial data?to be more widely adopted, it is necessary not only to improve technological sophistication but also to have policy support from financial authorities.

According to the financial sector on July 22, Jung Yoon-young, a research fellow at Hana Institute of Finance, recently stated in "Alternative Credit Assessment: The Reasons Behind Its Slow Adoption" that "a multidimensional approach is needed for the stable establishment of alternative credit assessment."

Traditionally, when evaluating the credit status of individuals and companies, credit rating agencies assessed the borrower's employment information, income data, and repayment history, evaluating their ability to repay debts within a set period. Financial institutions then combined these assessments with their own credit evaluations during the loan screening process. However, this method has limitations, such as difficulties in evaluating borrowers with insufficient financial transaction records and challenges in reflecting changes in credit status in real time. As a result, alternative credit assessment has emerged, proposing that analyzing data such as utility payment records, mobile usage, and social networking service (SNS) activity can supplement traditional credit assessment. Alternative credit assessment refers to the method of measuring or assigning credit ratings by applying new analytical technologies, such as machine learning or deep learning, to non-financial information that was not previously used.

Currently, in Korea, the adoption of evaluation models is mainly led by fintech (finance + technology) companies. These models utilize telecommunications data and lifestyle information, and there is growing interest in adoption among internet banks and secondary financial institutions. Notable examples include Equal, which uses data from the three major telecom companies (SKT, KT, LG Uplus), as well as Naver Pay Score and KakaoBank Score.

원본보기 아이콘

However, Jung explained that only a few financial institutions actually use alternative credit assessment models in their official credit screening processes. Even when such models are adopted, rather than realizing their benefits, institutions face practical issues due to model stability and regulatory concerns. He added, "During the process of expanding to mid- to low-credit borrowers, institutions have encountered practical constraints such as rising delinquency rates."

The institute identified technical issues as the first reason for the slow spread of alternative credit assessment. Data quality and model opacity are fundamental problems. Unstructured data can be manipulated, leading to reliability issues, and if some data are missing, risks may be excessively interpreted, resulting in discriminatory outcomes. Alternatively, if the model is trained on biased data, it can produce unfavorable results for certain groups. For example, in the past, Apple Card granted lower credit limits to female applicants than to male applicants, and in U.S. fintech lending, Black and Latino borrowers with the same credit ratings as White borrowers were charged higher interest rates. When using complex machine learning or artificial intelligence (AI) algorithms, the "black box" problem can arise, making it difficult to clearly explain the decision-making process.

Jung also pointed out that, in addition, if there is a lack of clear legal and institutional foundations for the use of new types of data, practical utilization may be difficult. For example, in the European Union (EU), the expansion of the General Data Protection Regulation (GDPR) has led to stricter legal scrutiny of practices that use automation and AI in credit scoring.

Nevertheless, Jung stated that, given the potential of alternative credit assessment models, it is necessary to ensure their stable establishment. He emphasized that, beyond simply introducing new technologies or advancing models, it is important to strengthen data utilization capabilities and develop models that enhance the explainability and transparency of algorithmic models to improve predictive accuracy. Jung added that financial authorities should create an environment where alternative credit assessment can become more active, stating, "Policy measures should be established to eliminate policy uncertainty, such as through regulatory sandboxes."

© The Asia Business Daily(www.asiae.co.kr). All rights reserved.