Credit Bureau 3.0

SME-1

MSMEs in ASEAN are significantly underfunded

Micro, small, and medium-sized enterprises (MSMEs) play a key role in contributing to economic growth and creating employment, yet they face disproportionally large challenges in financing. The Asia-Pacific Trade Facilitation Report 2019 of the Asian Development Bank stated that “[M]SMEs are most affected as they tend to have higher rejection rates than larger firms. Banks have higher transaction and information costs when dealing with smaller companies. Lending institutions’ reluctance to lend to MSMEs or offer them competitive interest rates stems from the relatively costly information acquisition for small loans.

Blind financial inclusion encourages capital misallocation

Governments have rolled out financial assistance programmes for MSMEs in response, with additional stimulus post-COVID. However, not all MSMEs deserve or should receive financial assistance. When a business idea is unsound, and operation is unprofitable, blind financial inclusion through these programmes not only misallocates capital but also deepens the unnecessary losses that the owner-entrepreneur incurs. By enhancing the quality of credit assessments, lenders’ competition can naturally achieve fairer financing of MSMEs.

SME-1
Picture 4

Existing models lack sufficient information to accurately assess credit risk

A good quality credit model enables lending at a competitive interest rate reflective of the credit risk faced by individual entities. Many MSME lending decisions still heavily rely on qualitative and subjective assessments, such as scorecards based on discretionary weights. It is also estimated that half of the MSMEs in ASEAN are not covered by existing credit infrastructure such as credit bureaus.

An evidence-based model, calibrated on observed default data, will reduce informational asymmetry between lenders and borrowers and enhance the quality of credit assessments. However, a good quality model needs to be trained by a large number of data instances. Given credit events are a relatively rare occurrence, most financial institutions are unable to construct a good quality in-house model based on their limited lending experience.

Picture 4

Mr Heng Swee Keat, Deputy Prime Minister and Coordinating Minister for Economic Policies, introduces the SME Credit Analytics Consortium during the official launch of AIDF on September 3, 2021

Ms Chuchi Fonacier, Deputy Governor of Bangko Sentral Ng Philipinas, shared an insightful keynote speech about the landscape of the MSME in Southeast Asia and the Philippines during The Official Launch of the SME Credit Analytics Consortium on 17 December 2021.

SME-3

Credit Bureau 3.0 – iCASS facilitates the construction of credit models using a privacy-protected distributed dataset, bringing forth the next generation of credit bureaus

The past generations of credit bureaus relied on member institutions directly sharing underlying non-payment data to a central authority. Recent cutting-edge scientific developments in machine learning technology have extended the realm of possibilities beyond this requirement.

Our lab, in collaboration with CriAT, has developed iCASS, a new "federated learning" system that can robustly calibrate credit models using credit data from multiple lenders (banks, finance companies and P2P platforms) without compromising data privacy, ushering in a new generation of technology-enabled decentralized credit bureaus.

"Federated learning" enables model calibration using only highly aggregated functional values derived from each member's data, removing the need for information on clients and client non-payment events to leave the consortium member's local data site. This enables the next generation of credit bureaus to draw insights from multiple institutions while respecting the privacy of each member, advancing the capabilities of traditional credit bureaus.

Utilising alternative data to enhance traditional models – leveraging on over 10 years of leadership in scientific credit research

With our extensive credit-risk modelling experience, we aim to create a benchmark model based on common features/variables available to all members. Members who wish to utilise unique features/variables such as alternative data will be able to access enhanced models tailored for them. These models will allow members to have better assessments of credit risk while fully utilising the unique advantage that their alternative data possess. As such, the consortium then caters for coopetition between member institutions through a coordinated but decentralised operation that can leverage both conventional and alternative data for better credit risk assessments.

AIDF will also contribute its NUS-CRI database to the Consortium’s modelling efforts. The database comprises, among other things, 30 years of comprehensive data on exchange-listed SMEs in the ASEAN region.

SME-3

iCASS Technology Demonstration

Technical Papers

White Paper

Contact Us

Please reach out to us for more information regarding the SME Credit Analytics Consortium