The Use of Alternative Data in Credit Risk Assessment: Opportunities, Risks, and Challenges 2024 2 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges © 2024 The World Bank Group 1818 H Street NW Washington, DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org All rights reserved. This volume is a product of the staff of the World Bank Group. The World Bank Group refers to the member institutions of the World Bank Group: The World Bank (International Bank for Reconstruction and Development); International Finance Corporation (IFC); and Multilateral Investment Guarantee Agency (MIGA), which are separate and distinct legal entities each organized under its respective Articles of Agreement. We encourage use for educational and non-commercial purposes. The findings, interpretations, and conclusions expressed in this volume do not necessarily reflect the views of the Directors or Executive Directors of the respective institutions of the World Bank Group or the governments they represent. The World Bank Group does not guarantee the accuracy of the data included in this work. Rights and Permissions The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. All queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2422; e-mail: pubrights@worldbank.org. The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 3 ABBREVIATIONS AND ACRONYMS AI Artificial intelligence AIA Artificial Intelligence Act AIDA Artificial intelligence and data analytics API Application programming interface B2B Business-to-business BCEAO Banque Centrale des États de l’Afrique de l’Ouest BIS Bank for International Settlements BNPL Buy now pay later BOT Bank of Thailand CARES Coronavirus Aid, Relief, and Economic Security Act (US) CBN Central Bank of Nigeria CCD Consumer Credit Directive (EU) CFPB Consumer Financial Protection Bureau (US) CGAP Consultative Group to Assist the Poor CICs Credit information companies CRA Credit reporting agency CRSPs Credit reporting service providers CV Credit Vision DLAs Digital lending apps DPIs Digital public infrastructures DRIP Data-rich, information-poor EBA Euro Banking Association ECOA Equal Credit Opportunity Act (US) EDPS European Data Protection Supervisor FAR Finance Against Remittances (Nepal) FCMB First City Monument Bank FCRA Fair Credit Reporting Act (US) FDI Foreign direct investment FDIC Federal Deposit Insurance Corporation (US) FEAT Fairness Ethics Accountability Transparency (Singapore) FICO Fair Isaac Corporation FINA Financial agency Fintech Financial technology FMCG Fast-moving consumer goods GAO Government Accountability Office (US) GDP Gross domestic product GDPR General Data Protection Regulation (EU) GPFI Global Partnership for Financial Inclusion GSMA Global System for Mobile Communications Association ICCR International Committee on Credit Reporting IDB Inter-American Development Bank IFC International Financial Corporation IoT Internet of Things KYC Know your customer LATAM Latin America LMICs Low-and Middle-Income Countries 4 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges ABBREVIATIONS AND ACRONYMS MAS Monetary Authority of Singapore MENA Middle East and North Africa MFIs Microfinance institutions MNO Mobile network operator MSE Micro and small enterprise MSME Micro, small, and medium enterprise MTO Money transfer operator NAL No-action letter NCLC National Consumer Law Center (US) NCR National Credit Regulator (South Africa) NCUA National Credit Union Administration (US) NFIR National Financial Information Registry OCC Office of the Comptroller of the Currency (US) ODA Official development assistance OJK Otoritas Jasa Keuangan (Indonesia) PBoC People’s Bank of China POS Point of sale PPP Paycheck Protection Program (US) PSD2 Payment Services Directive Two RBI Reserve Bank of India SIDS Small Island Developing States SME Small and medium-sized enterprises SNS Social Networking Services WEF World Economic Forum The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 5 ACKNOWLEDGMENTS This landscape study is a product of the International Committee on Credit Reporting (ICCR) and the World Bank Group (WBG). This paper builds on the ICCR Guidance Note, “The Use of Alternative Data to Enhance Credit Reporting to Enable Access to Digital Financial Services,” issued in 2018. The paper was prepared by Collen Masunda (Global Financial Inclusion Specialist & Secretariat of the ICCR) and Khadija Jabri (independent consultant) under the leadership and guidance of the ICCR Communications and Knowledge Management Working Group, chaired by Giovanna Cardellicchio (Alacred). The document benefited from a consultation process and the contributions of plenary members, representative organizations, and peer reviewers Gian Boeddu (Senior Financial Sector Specialist, World Bank), Maria Fernandez Vidal (Senior Financial Sector Specialist, CGAP), and Sephooko Ignatius Motelle (Senior Economist, IFC). The Committee also gratefully acknowledges executives from the following firms for their insights and perspectives: BCEAO, TransUnion (South Africa), Safaricom (DigiFarm), Experian, Laxmi Bank, Nowcast, Kabbage, Destacame, Begini, and LendMN. The ICCR would also like to thank the Chairman of the ICCR, Jean Pesme, and Secretariat members Luz Maria Salamina and Collen Masunda for guiding the process. Susan Boulanger provided editorial services. The layout and design of the report was prepared by Nitin Kapoor. 6 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 7 List of Tables / Boxes Table 1: Summary of Various Alternative Data Types for Credit Risk Assessment 11 Table 2: Alternative data use, opportunities, risks and challenges 23 Box 1: The Global State of Digitalization 15 Box 2: Integrating the Rental Payment Records into Credit Files 22 Box 3: Examples of Lenders and Credit Scoring Companies Leveraging Alternative Data Sources 25 Box 4: Laxmi Bank—Introducing the Concept of Finance Against Remittances (FAR) in Nepal 26 Box 5: Credit Reporting Agencies’ Use of Alternative Data 27 Box 6: Open banking and open finance examples 28 Box 7: Examples of Credit Building Products 29 Box 8: COVID-19 Credit Scoring Products 30 Box 9: Indonesia’s Regulatory Sandbox: OJKR No. 13/POJK.02/2018 on Digital Financial Innovation (Inovasi Keuangan Digital (IKD)) 36 Box 10: The Veritas Consortium of Singapore 36 Box 11: Global Open Banking Initiatives and Interventions 38 8 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges Table of Contents EXECUTIVE SUMMARY 10 1. INTRODUCTION 14 2. The use of alternative data in credit risk assessment 18 2.1 Rationale and definitions 18 2.2 The evolution of alternative data use in credit risk assessment 18 2.3 The impact of alternative data on credit access 19 2.4 Opportunities, challenges, and risks of using alternative data 22 3. Evolution of adoption and use of alternative data in credit risk assessment 24 3.1 Regional differences in types of alternative data use 24 3.2 Emerging business models and strategies 26 3.3 The impact of COVID-19 on the alternative data market 29 4. Regulatory and policy environment 32 4.1 Alternative data: Regulatory challenges and considerations 32 4.2 Regional regulatory responses 33 4.3 Open banking: Key implications and considerations 37 5. Navigating the path ahead: Recommendations and policy considerations 40 5.1 Recommendations 40 5.2 Policy considerations 41 References & Annotations 44 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 9 EXECUTIVE SUMMARY Digitalization accelerated by the pandemic has exponentially profiles for thin- or no-file customers. This study has identified increased the number and variability of alternative data emerging business models and innovative products that sources. The stay-at-home requirements during the pandemic collectively drive innovations in the credit scoring industry, as forced several business models to be digitized. This helped illustrated by the following examples: spur the digitization that had been ongoing pre-pandemic, largely driven by governments as countries sought to leverage • Credit reporting entities are integrating alternative data into the benefits of the digital economy. Advances in technological their databases: Building on the enabling environment, innovations are also facilitating the development of advanced credit bureaus have started integrating alternative data, credit underwriting models and the use of structured and such as utility, rental, and trade data. In addition, bureaus unstructured data to assess creditworthiness. This has enabled are introducing products that enable borrowers to share the financial industry to actively integrate alternative data to alternative data for credit scoring. gain a competitive edge. • Banks are partnering with alternative scoring providers: Banks are leveraging their robust infrastructure alongside This study examines the evolution of the use of alternative the cutting-edge analytics of alternative scoring companies data in creditworthiness assessment since the ICCR issued its to enhance their credit assessments and broaden their Policy Guidance Note in 2018. The report identifies the use cases service offerings. Alternative scoring companies can collect of alternative data in creditworthiness and the emerging models and aggregate alternative data from sources that banks across geographical regions. After assessing the evolution of would otherwise not easily access for compliance reasons. associated legal and regulatory frameworks, it concludes by • Alliances are being formed between credit reporting agencies offering recommendations and policy considerations. The and fintechs: These alliances blend traditional credit history findings of the study are detailed below. with the alternative data to which fintechs have access, thus creating enhanced models that extend creditworthiness Financial institutions are using alternative data across assessments to a larger pool of consumers. the credit value chain, from lead sourcing to account • Advancements are being made in open frameworks: Open management. Uses of alternative data include new product banking and finance and data frameworks integrate development, marketing, prescreening, underwriting, portfolio alternative data from banks, financial services, and other monitoring, and collections. Firms use alternative data not only data sources to improve inclusion of individuals, small for credit invisibles, but also to complement traditional credit borrowers, and gig workers. history. • Tech companies leverage alternative data from their own platforms: Tech companies, such as agtechs, leverage The use of alternative data in credit risk assessment provides varied data sets from their platforms, primarily to tailor opportunities but also presents risks and challenges. Some of credit assessments to their customer base. the key benefits include enhancing visibility and gaining a more • Incumbents with in-house built-in capabilities: These players comprehensive view of prospective borrowers, especially thin- often enter the market early, developing proprietary or no-file consumers, thus improving access to underserved solutions and refining them over time. Examples include populations. Alternative data also reduces credit losses and MNOs and e-commerce companies. improves approval rates, loan limits, borrower retention, • New credit building products are emerging: New tools are and expansion to new markets. Use of alternative data can being introduced to help individuals enhance their credit introduce biases, however, leading to discriminatory outcomes scores through innovative features like positive payment and perpetuating inequalities. Consequently, while alternative reporting. data can improve financial inclusion, it also risks reinforcing disparities for marginalized groups, including women and The specific alternative data adopted varies across regions, minorities. markets, institution types, and levels of digitization. Stakeholders use various forms of alternative data to build on Collaboration in leveraging alternative data has been the digital footprints created by businesses and individuals in increasing among incumbent and innovative players. To fend their everyday activities. They seek a balance in the adoption off competition, incumbents are building in-house capabilities of conventional and nonconventional alternative data for and collaborating with new entrants to effectively leverage the assessing creditworthiness. Table 1 categorizes these data power of alternative data. Furthermore, leveraging alternative sources into distinct domains, adding specific examples of how data is helping financial institutions to develop new products, these can inform credit evaluations. such as credit builders, which are designed to help build credit 10 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges EXECUTIVE SUMMARY Dimension Examples of Alternative Data Description Consumer Financial Telco data Billing history, call patterns, usage data, SIM card age, top-up habits Behavior Text messages Analysis of patterns, including frequency of financial transactions or bills-related messages Digital wallet Data from platforms, like PayPal, revealing payments and transfers Buy now, pay later (BNPL) Repayment history, credit use, and transaction frequency Remittances Indicates stable income, useful for verifying income for those with irregular earnings Credit/debit card and bank Financial transactions and patterns account records Consumer Lifestyle App use Offers insights into financial responsibility, lifestyle, and interests through installed and Habits apps and usage patterns Social media insights SNS/sentiment, social media; offers insights into lifestyle, interests, connections; privacy concerns noted Travel data Provides insights into income level, lifestyle, and spending habits through travel details Geolocation Uncovers routines and movements; potential biases from associations with socioeconomic statuses Biometric data Offers insights into behavior and lifestyle from wearable/IoT devices Business Operations Corporate data Business market share, competitive position, strategic moves and Analytics Accounting software data Revenue growth, revenue source diversification, dependency analytics Shipping data Operational scale and market penetration insights from transaction volume and frequency Supply chain data Supplier dependency and stability Patent data Provides insights into innovation capacity and future revenue potential Survey data Provides insights into business operational efficiency and market position Cloud platform data Business operational scale and efficiency Gig economy platforms Income stability, earning potential, and demand for skills Housing and Utility Rental history Assesses reliability in making consistent and timely payments Management Utility payment Payments of electricity, water, gas, and internet bills Economic Activity and E-commerce data (consumers) Offers insights into purchasing behavior and spending patterns Commerce E-commerce data (MSMEs) Analyzes sales performance, inventory management, customer satisfaction POS data Transactional data indicating cash flow and income stability Risk Management and Weather forecast Impacts on agriculture and construction projects affecting potential revenues Insurance Insurance Provides insights from payment of insurance premiums, claims history, and renewals Agriculture Bioclimatic Possible insights into climatic conditions and soil and crop types Satellite data Provides insights into crop health and agricultural practices for agriculture-focused MSMEs Weather forecast Impacts on agriculture and construction projects affecting potential revenues Market and Web traffic Brand recognition, customer interest and retention Environmental CCTV Customer density and foot traffic as business activity indicators Insights Regulatory Government data Legal and financial status from licenses, permits, and tax filings Compliance Table 1: Summary of Various Alternative Data Types for Credit Risk Assessment Source: Authors The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 11 EXECUTIVE SUMMARY The alternative data-based lending market is poised for results of experimentation have validated that the use of continuous growth. The expanding availability of data sources, alternative data is adequately covered under existing legal rising demand for credit access, and global initiatives to and regulatory frameworks. promote financial inclusion collectively indicate a steady rise in the demand and supply side of alternative data in lending. Most data protection and openness frameworks (including Additionally, integration of new technologies such as AI and open banking and open data standards) can accelerate advanced computing, already being employed, will further responsible use of alternative data. Open frameworks, reinforce the trajectory of growth. including open banking and open finance, facilitate access to banking and/or financial data proven to have high levels The COVID-19 pandemic accelerated the integration and use of predictiveness. This enriches the available data pool and of alternative data. The crisis impacted borrowers’ ability to provides lenders with a multidimensional understanding of a make payments, resulting in difficulties for lenders seeking to potential borrower’s profile. By providing legal and regulatory distinguish between false and real defaults. With the pandemic frameworks, policy makers introduce safeguards that can accelerating digital transformation, lenders actively sought enhance responsible sharing of data. out and evaluated various new alternative data sources to get a real-time understanding of borrowers. While initially Alternative data-based lending introduces complexities to alternative data was used for macroeconomic analysis to detect the existing credit reporting ecosystem. Questions about economic distress and recovery signals, this data was eventually where alternative credit scoring providers fit within current integrated into creditworthiness assessments and came to form regulatory boundaries and rules include concerns about a key component of national strategies to mitigate the economic potential governance evasion and regulatory gaps. Another impacts of COVID-19. issue is whether loans issued through digital platforms are reported to credit information companies; recent digital lending Notwithstanding the promising results from alternative data regulations in some countries enforce credit information use for credit underwriting, regulation has not as yet been sharing. Furthermore, the use of alternative data in credit standardized in several markets and regions. Responsibly scoring complicates the explanation of adverse action notices, integrating alternative data into credit risk assessment requires adding another layer of complexity to the process. regulatory clarity and consistency that provides guidance on issues such as data privacy, accuracy, consent, transparency, Recommendations and fairness of algorithmic models. Regulatory responses vary from hard and prescriptive measures to soft and incremental To unlock the full potential of alternative data to transform ones: creditworthiness assessment, several foundational elements and enabling policy frameworks are essential. To help • Prescriptive approaches: Two main prescriptive approaches stakeholders in developing these, this report offers the following incorporate alternative data elements: either within credit recommendations and policy considerations. reporting laws or within other credit-related regulations. Countries such as Brazil, China, Tajikistan, and Uganda, Recommendation 1: Implement robust legal and regulatory for example, included provisions relating to alternative frameworks. data collection within their credit-reporting laws, thus harmonizing all laws related to information sharing for While there is no one-size-fits-all approach, regulators must creditworthiness assessment. Other jurisdictions, such the consider enacting robust legal and regulatory frameworks European Union, India, and Thailand, provide guidance on that promote the sharing and use of alternative data with the the use of alternative data through credit-related directives necessary consumer protection and cyber security safeguards. such as the EU Consumer Credit Directive (CCD), India’s Account Aggregator Framework, and the Bank of Thailand Recommendation 2: Implement a regulatory blacklist for (BOT) Circular on Rules, Procedures, and Conditions for alternative data in credit scoring. the Undertaking of Digital Personal Loan Business. These regulations create a framework that encourages responsible In view of the potential risks associated with the use of use of alternative data in credit underwriting. alternative data, such as privacy and discrimination concerns, • Soft and incremental approaches: Soft and incremental policy makers should consider implementing a blacklist that approaches are characterized by their flexibility and delineates data elements prohibited from inclusion in credit gradual implementation, allowing regulators and industry scoring algorithms. participants to learn and understand the evolving technologies. The approaches in this bucket have taken Recommendation 3: Leverage regulatory innovation the form of no-action letters (e.g., the Consumer Financial platforms to promote experimentation and testing. Protection Bureau in the United States), pilots (the National Credit Regulator in South Africa), and regulatory In the absence of an existing legal framework that enables sandboxes (Otoritas Jasa Keuangan in Indonesia). The the use of alternative data, regulators should consider using use of innovation platforms enables experimentation, innovation platforms, such sandboxes, to test the predictiveness which can create demonstration effects that help promote of alternative data within the country context. In the absence regulatory reforms where necessary. In some markets, the of formal sandboxes, regulators might consider other policy 12 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges EXECUTIVE SUMMARY instruments or tools, such as the no action letter (NAL) or no Policy considerations objective letter and controlled pilots to promote responsible use of alternative data. In addition to the above recommendations, policy makers should consider implementing policies and interventions that Recommendation 4: Promote the adoption of consumer- contribute to the generation of alternative data and harness the permissioned, secure data-sharing protocols that enable potential of such data in creditworthiness assessment. Among effective sharing of alternative data at scale. others, policy consideration may include: Regulators should accelerate informed consumer-permissioned Policy Consideration 1: Incentivizing digitization of economic data-sharing frameworks that are frictionless to both data activities providers and consumers, such as open banking / data frameworks. To the extent possible, policy makers should implement financial and nonfinancial incentives to promote digitization. In addition, Recommendation 5: Advance gender equity in alternative it is also necessary to ensure equitable access to infrastructure, credit scoring through inclusive data practices. such as internet connectivity and digital devices. Policy makers and industry players can adopt several inclusive Policy Consideration 2: Digitizing government services and approaches to designing alternative scoring models. Financial making the data readily accessible institutions and credit bureaus might, for example, collect and use sex-disaggregated data—in an aggregated and anonymized Public databases such as company registries, vehicle registries, form to comply with existing regulations—to inform their models tax registries, collateral registries, and court records can be designs; they might also offer products tailored to better reflect sources of alternative data. To the extent possible, governments the unique financial behaviors and creditworthiness factors of should digitize their services and make their databases readily certain borrowers, e.g., women. The use of sex-disaggregated accessible for credit reporting services providers. alternative data is expected to promote access, as more women than men tend to be credit invisible or thin-file customers. Policy Consideration 3: Supporting infrastructure development Recommendation 6: Establish a comprehensive industry code governing the use of alternative data in credit scoring. Policy makers should consider policies that facilitate the development of robust digital public infrastructure—namely, In the absence of legal and regulatory guidance, industry identity, payments, and data exchange systems—as these are participants might consider self-regulation guided by a set of essential for generating and spurring alternative data use. comprehensive principles that ensure responsible and ethical use of alternative data in creditworthiness assessment. Policy Consideration 4: Supporting digital literacy and consumer awareness Recommendation 7: Adopt a risk-based approach for the collection and use of alternative data for creditworthiness Policy makers should promote literacy and awareness programs assessment by lenders. a means for promoting responsible use of alternative data. Regulators should ensure that lending institutions have Policy Consideration 5: Promoting cross-border collaboration adequate data policies that govern consent management; alternative data collection, processing, correction, and storage; Given increasing regional integration and migration (voluntary and reliance on third-party data providers. The policies should and involuntary), policy makers should implement measures that also cover the responsible application of machine learning and promote cross-border data flows to support creditworthiness AI in credit scoring, considering that credit scoring models might assessment of MSMEs and individuals across borders. incorporate these more advanced techniques. The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 13 1. INTRODUCTION Traditionally, access to credit is primarily based on the assessment of the credit repayment history of borrowers. This approach poses a significant access barrier for many credit applicants due to information asymmetry. The access constraint is especially acute for borrower segments with limited credit histories, such as young people, recent immigrants, women, and ethnic minorities, as well as small and medium enterprises (SMEs), as customers in these groups tend to have thin or no credit files. Economic agents operating in the informal sector are also disproportionately affected by access constraints, as they have low and/ or irregular incomes and much of their economic and financial activity is not recorded. Also, a significant part of the population in developing economies operates in the informal sector and cannot build personal credit histories because they may not have mainstream credit facilities in their own names. Thus, the absence of an adequate credit history impedes their ability to access credit from formal credit providers, compelling them to turn to informal lenders who offer less than favorable lending terms, including charging exorbitant fees. Despite their lack of credit history, these potential borrowers create digital footprints when they use cloud-based services, browse the internet, use their mobile phones, engage in social media, use e-commerce platforms, ship packages, or manage their receivables, payables, and recordkeeping online. These digital footprints can be leveraged to assess these potential borrowers’ creditworthiness. Additionally, these potential borrowers have other recurrent obligations—including mobile phone, utility, and rental payments—which can also be used to assess their creditworthiness. As digitized data and technological advancements proliferate, credit reporting service providers are increasingly incorporating such additional, structured, and unstructured data into their credit assessments.1 14 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges INTRODUCTION Digitalization has proven to be an asset during the pandemic, catalyzing widespread change and facilitating access to various social and economic activities. Its benefits, however, were disproportionally felt across countries and segments, with the most marginalized communities, unsurprisingly, benefitting the least. Even the most developed countries have much room for growth. A 2022 global survey by PYMTS, in collaboration with Stripe, looked at 11 developed countries, constituting nearly half of the world’s GDP, and found that digital engagement has only reached 27 percent of its full potential, with 87 percent of studied consumers connected to the internet but only 19 percent highly engaged in digital activities. Moreover, consumers are nearly 1.5 times more likely to be engaged in activities purpose-built for digital environments than in purely transactional ones, such as banking or making retail purchases.2 The Global Findex 2021, for instance, examined the issue of digitalization and its impact across payments. Looking at private sector wages in Pakistan, for example, it found that such digitalization would result in a 13 percentage point increase in account ownership and would bring around 20 million unbanked adults into the formal financial system. The survey also finds that COVID-19 accelerated the digitalization of utility payments in some developing economies, especially across countries in Latin America and the Caribbean, where a high number of adults made a utility payment from an account for the first time. The study also revealed an important insight into the financial literacy and receptivity to digitalization of unbanked adults in developing economies. When asked if they would feel comfortable using a financial account without assistance, respondents indicated a significant level of insecurity regarding such accounts, with at least 64 percent of unbanked adults stating that they could not use an account without help.3 Digital payment ecosystems are the invisible infrastructure that powers modern commerce, trade, and people’s lives,4 but a joint study by the World Bank and World Economic Forum (WEF) estimates that the global value of MSME retail transactions as of 2015 was $34 trillion, of which $19 trillion were paper based.5 The identified barriers to acceptance of digital payments by MSMEs were consistent with those encountered in any digital solution implementation: inadequate digital infrastructure and an inconsistent regulatory environment. While the pandemic expedited acceptance of digital payments, further research is needed to quantify the impact on the digitalization of business payments. A 2019 survey on the national digital strategies of 17 African countries found that only one had an actual digital economy strategy.6 The selected sample of countries were diverse across factors such as size of the economy, wealth, infrastructural challenges, and digital development. These statistics serve to underline the need for revisiting national digital transformation strategies, with a focus on the digital economy and associated needs. Cyprus’s National Digital Strategy 2020–2025, for instance, has among its key strategic objectives achieving a vibrant, sustainable, resilient digital economy through several specific stages, such as fostering development of basic and lifelong skills relevant to daily and domestic activities, helping consumers connect with digital government services, teaching consumers how to make digital transactions, and supporting the digitalization of local businesses and business sectors by providing upskilling, tools, support, and guidance. 7 Box 1: The Global State of Digitalization Credit reporting systems increasingly incorporate alternative data when assessing creditworthiness.8 Likewise, credit providers leverage alternative data generated on their platforms or sourced from credit bureaus or data aggregators to enhance their credit risk assessments.9 Therefore, alternative data is creating opportunities to increase access to credit by expanding the information available to lenders and credit reporting service providers about borrowers, particularly informationally opaque but creditworthy borrowers. Over the past decade, significant progress has been made in using utility, telecoms, rental, and cashflow data to improve the assessment of the creditworthiness of thin- or no-file customers and to promote their access to credit. Alternative data is also being used to enhance the credit files of borrowers who fail to meet lenders’ criteria or are accessing credit at higher interest rates. The COVID-19 pandemic increased the demand for alternative data as the credit industry had to find new ways to assess creditworthiness and augment traditional credit history. During the pandemic, even traditionally good borrowers were unable to meet their obligations as they fell due, leading to false negatives. Credit bureaus and lenders adopted real-time alternative data to distinguish between good corporate borrowers that were failing to meet their obligations due to the pandemic and zombie firms that were unlikely to recover from the crisis. Improvements in consumers’ credit scores and enhanced access to financial products and services for the underserved have been attributed to the integration of alternative data into credit risk assessment.10 Various types and combinations of alternative data have been shown to enhance the predictive capabilities of models using only traditional data by 5 to 20 percent.11 Opening data access to many parties also reduces incumbent firms’ information advantage, thus increasing competition and broadening the range of products and services available to borrowers. Although considerable progress and success have been achieved in the use of alternative data in credit risk assessment, concerns about its use have also been raised. In addition, large numbers of people—as many as four in ten globally—continue to have thin or no credit files. In the United States, an estimated 53 million potential customers have no credit files with the big credit reporting service providers.12 The high numbers of ethnic minorities and protected classes lacking files brings to the fore a key question about whether the adoption of alternative data promotes credit inclusion. In addition to its potential for bias and discrimination, the use of alternative data in credit assessment has also been under scrutiny for its significant risks and challenges to data privacy and The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 15 INTRODUCTION potential cybersecurity issues. Questions have also emerged over whether credit reporting service providers’ use of alternative data increases their oligopolistic power.13 Those who believe that it does argue for decentralized mechanisms for data-sharing, including bilateral arrangements. Such arrangements, however, are generally considered expensive and difficult to scale. In 2018, the ICCR released a policy guidance note14 recommending the adoption and integration of alternative data to promote access to credit for marginalized segments. The guide proffers 21 recommendations across six themes: 1. Adoption of alternative data for credit reporting 2. Improving availability and accuracy of alternative data 3. Expanding credit information sharing 4. Enabling responsible cross-border data exchanges, if relevant and applicable 5. Balancing integrity, innovation, and competition 6. Data privacy and consumer protection Objectives and context This study seeks to assess how the market has evolved since the ICCR Policy Guide was issued in 2018. It does so by assessing the use of alternative data in the credit risk assessment of borrowers, including underserved individuals and MSMEs. With this goal in mind, this paper reviews market dynamics through regional lenses; identifies the key opportunities, challenges, and risks of adopting alternative data; and proposes recommendations on how alternative data can be mainstreamed responsibly. This study is designed to answer the following questions: • How has alternative data definition or scope evolved over time? • What factors have laid the groundwork for alternative data to emerge? • What opportunities and risks does alternative data present? • What is the added value of using alternative data along with traditional data? • How is alternative data currently being used in the credit market? Who are the key players? What are the current and emerging business models or applications? • Has the use of alternative data accelerated financial inclusion of the underserved and unserved segments? • How did the policy and regulatory framework adapt to the emergence of alternative credit scoring? • What are the best practices in the governance of alternative data use in the context of credit risk assessment? • What are the prerequisites necessary to mainstream alternative data in creditworthiness assessments responsibly and sustainably? Methodology This study is based on extensive desk review and semi-structured interviews with 41 stakeholders, including regulators, lenders (conventional banks to fintechs), credit bureaus, and industry bodies from Europe, Asia, North America, Latin America, Africa, and Oceania. The interviews focused on achieving a more comprehensive understanding of the use of alternative data across different regions, including the types of alternative data used, the drivers for using it, the impact of its use, and the regulatory framework that governs its use across various jurisdictions. Structure of the report Following this introduction, the rest of this report is structured as follows: • Section 2 presents and defines the alternative data ecosystem broadly, including potential risks, challenges, opportunities, and efficacy in advancing financial inclusion. • Section 3 discusses alternative data case studies and key business models from around the world, with a focus on emerging economies. • Section 4 highlights key emerging approaches to regulating the use of alternative data in the credit market and examines subsequent implications and gaps from the legal and regulatory angle. • Section 5 presents policy considerations and recommendations. 16 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges Introduction The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 17 2. The use of alternative data in credit risk assessment 2.1 Rationale and definitions The range of alternative data available is influenced by factors including an individual’s digital literacy and engagement and, In view of the ongoing challenges faced by prospective for MSMEs, the level of digitalization and the nature of their borrowers in providing adequate credible financial data for business. For instance, e-commerce MSMEs are likely to generate credit assessment, banks and other lenders have continued to alternative data differently than do construction companies rely on collateral or guarantees. Consequently, many borrowers or other labor-intensive businesses. This makes compiling have been unable to access credit, and the credit industry has a definitive list of alternative data sources challenging, as it responded by developing innovative strategies and tools to requires ongoing updates to reflect new insights from various evaluate the creditworthiness of these borrowers. Notably, the data types. integration of alternative data has emerged as a significant advancement in credit scoring practices. 2.2 The evolution of alternative data use in credit risk assessment No single official definition describes the full spectrum of specific forms of data that fall under the term “alternative Against the backdrop of the global financial crisis of 2008, the data.” Most definitions are descriptive rather than definitive. ensuing erosion of public trust in formal financial institutions Generally, alternative data is defined as encompassing a catalyzed the implementation of new regulatory reforms. vast range of nontraditional data sources. The US Consumer Coinciding with this period, dissemination and adoption of Financial Protection Bureau (CFPB) defines alternative data digital technologies increased significantly. Leveraging these as “information that is not detailed in the traditional credit concurrent developments, several prominent players in the report provided by the three main nationwide credit reporting fintech landscape emerged, particularly in the lending and agencies, namely Equifax, Experian, and TransUnion.”15 alternative credit scoring markets. In the context of credit reporting, the International Committee Historically, innovation tends to originate in developed on Credit Reporting (ICCR) views “alternative data” as serving countries, where favorable conditions—such as supportive merely to describe nonconventional methods for collecting and regulatory environments, robust institutions, advanced analyzing data to support assessment of creditworthiness.16 infrastructure, and a literate user base—foster its adoption. The Additionally, ICCR notes that alternative data often takes the landscape surrounding the adoption of alternative data presents form of readily available digital information collected through a more nuanced picture, however. On the one hand, integration technological and electronic platforms. ICCR also notes, of alternative data into assessments of creditworthiness is however, that what currently constitutes alternative data may increasingly prevalent in emerging markets, driven by factors evolve over time as more work is done in the area. Moreover, the such as widespread smartphone adoption, increased mobile definition of alternative data may be country specific, meaning data usage, improved infrastructure, accelerated digitalization, that what is considered alternative data in one market may very and the imperative to address financial inclusion gaps in society. well be considered traditional by another market’s standards.17 On the other hand, the fintech boom, characterized by digital Alternative data exists in two forms. First, alternative data can payment as its initial focus, has shifted significantly toward either be financial (transactional) in nature, such as utility digital lending. A study by the Bank for International Settlements payments, suppliers’ payments, or purchases on e-commerce (BIS) highlights the rapid growth in fintech credit activity18 sites, or nonfinancial, such as internet browsing history and between 2013 and 2016, with global fintech credit extended patterns, social media activity, and online traffic patterns. soaring from $11 billion in 2013 to an estimated $284 billion in Second, the data format can be either structured or unstructured. 2016.19 This growth has been unevenly distributed, however, Structured data lends itself to clear definitions, formats, and with China, the United Kingdom, and the United States leading searchability, whereas unstructured data is data in its native the market, while Latin America and the Caribbean region format and cannot be analyzed using conventional data tools have been catching up in recent years. According to a study20 and methods. by the Inter-American Development Bank (IDB), the number of fintech platforms surged by 112 percent from 2018 to 2021, 18 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges The use of alternative data in credit risk assessment with Argentina, Brazil, Colombia, Mexico, and Peru leading this MSMEs represent over 95 percent of firms around the world growth trajectory. Although payment and remittance platforms and play a critical role in employment creation in low-and continue to dominate the fintech market, lending is projected to middle-income countries.24 Thus, given their role as important grow at a faster rate in the coming years, eventually taking the contributors to economic output and employment in emerging lead in the region. markets and developing economies, MSMEs’ access to credit is vital for maintaining their business operations, funding Technological advancements also play a key role in this business expansion, and ensuring survival during crises. Their expansion. On the supply side, factors such as the proliferation of lack of access to credit services can stunt the development big data, advanced analytics tools (including machine learning), of the economy.25 MSMEs tend to face sharper access to sophisticated computing solutions, and accessible cloud storage credit constraints than do larger firms, however. In emerging incentivize lenders to incorporate alternative data into credit economies, for example, the access to finance gap for MSMEs is risk assessments. On the demand side, advancements in digital estimated to be approximately $5 trillion—1.3 times the current infrastructure facilitate the widespread adoption of digital tools, level of MSME lending.26 The figure would be even higher if activities, and services, resulting in a wealth of valuable data informal businesses were included. Unlike large firms, MSMEs that can either substitute for or complement traditional cash are informationally opaque, making it challenging for lenders flow information. For instance, MSMEs increasingly embrace to assess their capacity and willingness to honor their credit digital tools that facilitate enhanced data collection and analysis obligations. of their credit risk, including digital wallets, electronic invoices and receipts, payment gateways, and e-commerce platforms for As mentioned earlier, the conventional approach to credit sales. scoring can exclude MSMEs that lack sufficient reliable financial data to support a credit risk assessment. Alternative data can Various industry players are gradually endorsing and adopting address this challenge by offering lenders a better understanding the use of alternative data. In the United States, for example, of the creditworthiness of MSMEs and can be used in tandem Nova Credit conducted a study21 in 2024 to assess the state of with conventional credit data to generate important information alternative data use in lending. The company surveyed 125 about the creditworthiness of underserved borrowers, including decision-makers at banks, credit unions, fintechs, and other MSMEs. For example, Fundfina, a fintech in India, partners lending institutions. The study revealed that 43 percent of with suppliers and agent networks for fast-moving consumer lenders were currently using alternative data to gain insights goods (FMCG) to access transactional data to provide loans for into their customers. The main types of data used included micro and small enterprises (MSEs). The company estimates income, employment, rent, bank transactions, cross border that 80 percent of its clientele lacks a formal credit history, transactions, telco and utility statements, among others. More which presents a challenge to effective underwriting. To assess than nine in ten respondents indicated that the incorporation creditworthiness, Fundfina therefore uses a range of alternative of alternative data resulted in higher approvals, credit limits, data including characteristics of the enterprise, characteristics of and use timelines. The top four reasons for including alternative its partner, and transactional data. To evaluate the effectiveness data are reducing fraud risk, growing the customer base, of this approach, CGAP (Consultative Group to Assist the Poor) expanding to new market segments, and reducing loss rates. conducted an analysis27 of over 5,000 loans issued by Fundfina Young consumers (73 percent) and new-to-country consumers to MSEs under various partnerships. The analysis uses the (41 percent) were considered the top two segments benefiting repayment records of Fundfina’s customers as a proxy for their from inclusion of alternative data. credit history. Two sample sets were used for the study: a full sample comprising all clients, among which 60 percent are 2.3 The impact of alternative data on credit access repeat customers with a “credit history,” and a smaller sample consisting only of repeat clients with a credit history. The CGAP 2.3.1. Impact on MSMEs, including the gig economy analysis found that a model built solely using transactional data had predictive power comparable to one relying solely on According to Beck et al. (2007), financial development plays a credit history. Combining both data sources, however, offered positive role in reducing inequality and poverty22, and financial better prediction accuracy. In the repeat customer sample, inclusion can catalyze the distributional effects of credit in the predictive power remains similar when using either credit economy. For this reason, Barajas et al. (2020, 14) perceive history or transactional data alone, whereas combining both financial inclusion as “an additional dimension of financial strengthened the model. In the full sample, transactional data development.” They argue that, even for two countries with outperforms credit history data; since some borrowers in this identical financial depth, if their patterns of credit allocation sample had no credit history data, the study highlighted the differ such that in one country credit is concentrated in a few potential of transactional data as an effective tool for financial large firms while in the other country MSMEs also benefit from inclusion, particularly for borrowers new to credit. credit access, the differential in access to finance by firm size will lead to differing distributional outcomes. In essence, financial The gig economy has the potential to offer a lifeline for millions inclusion provides the mechanism for sharing the benefits of worldwide. Gig economy workers, however, regardless of their financial development in the economy by facilitating access occupation or sector, face the same lending needs as many other to finance for underserved segments such as low-income workers seeking to both accelerate and maintain the initial stages households and micro, small, and medium-sized enterprises of their businesses. Ride-hailing apps or delivery drivers, for (MSMEs)23. example, need loans to purchase and maintain the vehicles they The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 19 The use of alternative data in credit risk assessment use for work. Furthermore, being self-employed, gig workers’ problem, lenders need critical information about the borrower, often-fluctuating monthly incomes mean they stand less chance including credit repayment history and financial statements. than do standard salaried individuals of demonstrating the Lenders can use self-collected information and/or leverage credit income stability necessary to qualify for a loan. Realizing that information from credit reporting service providers, who upon the gig economy is the next frontier to monitor when it comes to request can provide a borrower’s credit score, which serves as a lending, innovators saw an opportunity to leverage data such as numerical representation of a lender’s trust and confidence in a payment patterns, driving records, brand loyalty, and ridership borrower’s capacity to meet debt payments on time. Evaluating patterns to tap into the alternative lending market. credit applications involves examining evidence that provides insights into the applicant’s financial behavior and financial KarmaLife, for example, is a fintech start-up operating in responsibility. Applicants are typically assessed based on two India that leverages partner platform data to tailor financial primary criteria: their financial capability to repay a loan and solutions and loans to gig platform workers and blue-collar their willingness to do so. workers, including drivers and food delivery couriers. These solutions address the growing demand within the mobility Although consumers and MSMEs may have different financial sector for higher-ticket, installment-linked loans, which are obligations, they tend to share some common factors often used for vehicle purchases or repairs that in turn enhance that contribute to their creditworthiness. For individuals, the earning potential of workers in ride-hailing jobs. To assess establishing a track record of making timely payments, such as the value of platform data for credit scoring, CGAP analyzed28 credit card bills, loan installments, or rent payments, is essential 15,000 KarmaLife-facilitated loans to drivers of the truck and for building a positive credit history. Similarly, having consistent bike delivery platform Porter, over 95 percent of whom had a and predictable income streams, whether from employment, file in the credit bureau. The analysis indicates that platform investments, or other sources, instils confidence in lenders data performs comparably to credit bureau data in predicting a regarding the borrower’s ability to repay debts. Going beyond driver’s creditworthiness. Additionally, the analysis reveals that these, the use of alternative data helps lenders expand access to integrating platform data significantly enhances the accuracy of credit for underserved individuals using a range of data sources credit models using just bureau data. The Porter data showed such as telco data and rent data. that higher earnings, excellent driving ratings, and increased login hours all correlated with lower repayment risk. Conversely, A FICO study in 2015 revealed that use of alternative data account suspensions and negative balances were indicative of alongside traditional data can accurately score more than higher repayment risk 50 percent of previously unscorable credit applicants.33 The study specifies that telecommunication payment data, for 2.3.2. Impact on households instance, can provide high quality and predictive data. A later study34 by FICO in 2023 revealed that adding alternative data to For households, credit can serve as a means of accessing traditional information increases predictive value on margin to essential improvements in quality of life, such as housing, risk models based on traditional data. The study broke down education, or transportation, or as a safety net to mitigate the increase in predictive value35 added to models according adverse effects of unexpected financial setbacks. In 2021, to customers relationships with lenders. The study showed however, the World Bank reported that 1.4 billion adults remain that more predictive value could be realized for the category of unbanked. The report found that in high-income countries the borrowers new to the lending institution and those with weak most prominent source of credit for households was formal relationships with the lender. For new customers, predictive credit from a financial institution, a credit card, or a mobile value enhancements include 5 to 10 percent from rental and money account. Although the use of formal credit is lower in utility data, 0 to 15 percent from social profile data, 10 to 20 developing countries, it has increasingly become an important percent from social network data, and 0 to 10 percent from source of credit for individuals living in developing countries.29 clickstream information. For customers with weak relationships For example, over the last decade, the share of adults who use with their lenders, the predictive value gained from the inclusion formal credit increased from 16 percent of adults in 2014 and of transactional data ranges from 0 to 5 percent, rental and utility 2017 to 23 percent in 2021.30 While this trend is promising, it is from 5 to 10 percent, social profile data from 0 to 15 percent, important that lenders and credit reporting service providers social network data from 10 to 20 percent, clickstream data continue to explore innovative ways of deepening financial from 0 to 10 percent, and text data from 0 to 5 percent. Lastly, inclusion. for customers with strong relationships to lenders, predictive value increases by 5 to 10 percent with transactional data, 0 to Agarwal et al. (2018) show that in Rwanda credit information 5 percent with rental and utility data, 0 to 5 percent with social sharing (CIS) effectively augmented the readiness of lenders in profile data, 5 to 15 percent with social network data, 0 to 5 improving household access to credit from commercial banks at percent with clickstream data, and 0 to 5 percent with text data. favorable terms.31 Similarly, de Janvry, McIntosh and Sadoulet (2019) noted an improvement in credit as one microfinance Similarly, a study by the US Government Accountability Office lender in Guatemala began using credit reporting services.32 (GAO) noted the significance of including positive rental payment history in automated credit underwriting. It indicated Lenders deploy different methods to assess the creditworthiness that 17 percent of a sample of mortgage loan applications of potential borrowers. Regardless of the credit assessment previously shown to be ineligible could have been found eligible method used, however, to address the information asymmetry had the applicants’ rental payment histories been considered.36 20 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges The use of alternative data in credit risk assessment The study added that while use of alternative data shows societal expectations, combined with limited time to seek encouraging results in improving credit scores and extending employment or engage in digital activities, hinder women’s access to credit, its effects are most impactful for those on the financial inclusion. This situation affects women in two main lower end of the spectrum, notably consumers with lower credit ways: first, their caregiving responsibilities reduce the time scores, who are disproportionately minorities. available for seeking traditional employment, leading many to opt for flexible jobs or intermittent sources of income streams. A study37 conducted by Equifax in the United Kingdom compared Second, when women own smartphones, their multiple the predictive power of traditional UK bureau data, such as domestic duties often leave them with less time for online credit repayment history and credit balances, to the predictive activity, thus limiting their digital footprint. power of bank account and transactional data, such as income and salary, rent, and council tax. While the predictive power of Similarly, marginalized and low-income households, facing the credit bureau score surpassed that of the transaction data challenges with supply and demand economics and rising score, combining both in the credit risk assessment resulted construction costs, may encounter difficulties in meeting rent in higher predictive power, outperforming the traditional CRA payments—a significant expense. Renters often find themselves score by up to 10 percent on a full sample and up to 25 percent burdened with costly payments that consume a large portion on thin-file borrowers. Results demonstrate that 50 percent of of their income. The utility payment history, another factor the population experiences a score increase, 20 percent scores impacting consumer scores, adds further complexity. For the same, and 30 percent experiences a score decrease. These instance, in certain regions where these are public utilities, the findings add to the growing body of research demonstrating culture on repayment has been shown to be poor. In markets the potential value of integrating alternative data, such as where these utilities are privately owned, fluctuations in prices transactional data, with traditional credit data. due to exogenous factors can challenge both consumers and providers. To understand the credit underserved and unserved markets, TransUnion conducted a study38 using depersonalized credit These examples shed light on the varying effects of alternative data from 572 million credit consumers in Canada, Colombia, data on women and other marginalized groups. It is essential Hong Kong, India, South Africa, and the United States. The study for innovators to consider these dynamics and to develop assessed pre-pandemic and pandemic credit participation, market solutions that address the challenges. Such solutions activity, originations, scores, balances, and performance. To may involve inclusive credit scoring models that incorporate a complete the data, the company conducted an online global wider range of alternative data sources and financial products survey of 11,128 credit consumers from a range of developed designed to cater to the unique needs of and limitations faced and developing credit markets. Among other findings, the study by women and other marginalized communities. found that for the United States, the inclusion of rental payment tradelines in the credit file resulted in 9 percent of the formerly credit invisible becoming scorable, with an average credit score of 631. A separate analysis for Colombia found that inclusion of alternative credit data could raise the country’s financial inclusion to 97 percent, higher than the government target of 85 percent. Although the potential of alternative data use to enhance the credit scores of existing credit holders has been acknowledged, uncertainties persist regarding whether these benefits effectively extend to minorities and marginalized segments, including women and rural communities. For example, women often encounter discrimination in the credit scoring landscape, and the inclusion of non-gender-neutral data could exacerbate this disparity. Instances abound where lenders heavily factor social media presence, web, and app activity into their credit scoring systems, potentially excluding individuals and small businesses with limited online engagement due to personal choice or time constraints. Moreover, the extensive use of mobile phone data in credit scoring reveals disparities. As evidenced by a recent GSMA (Global System for Mobile Communications Association) report,39 women are statistically less likely to own smartphones and to use mobile internet, potentially limiting the digital footprints of this demographic. Additionally, prevailing gender norms dictate that women often bear more caregiving responsibilities than do men,40 impacting women’s income levels and employment opportunities. These The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 21 The use of alternative data in credit risk assessment New York has consistently been on the list of the most expensive US cities in which to rent a home. According to the latest available official figures,41 the average rent-to-income ratio in New York is 39.4 percent, meaning that households spend over one-third of their income on rent, most paid duly and on time. Yet, such households are denied one critical benefit: unlike mortgage payments, their rent payment history for the most part is not factored into their credit score. A study42 published in 2017 by New York City Comptroller Scott M. Stringer revealed that 44 percent of New Yorkers’ credit scores are a mix of subprime (26 percent), nonprime (13 percent), and unscorable (5 percent)—with some sharp disparities across the city’s neighborhoods.43 In zip codes where the median credit score is below 630, 78 percent of people rent their homes—a strong indication of an individual’s financial responsibility and ability to pay rent. The study, conducted in consultation with all three major credit bureaus in the United States, used consumer files from Experian’s database that already contained rent information. A sample of files of consumers whose rent is under $2,000 was scored with and without rent payment history to project and compare how the inclusion of rent payment would impact credit scores. The study found the following: • 28.7 percent of renters obtained a credit score for the first time. • 76 percent of New York City tenants had their credit scores increase, including significant increases of 11 points or more for an estimated 19 percent of participating renters. • Credit scores for New York’s low-income and minority communities whose credit score averages 630 or lower were lifted. It is worth noting that the residents of zip codes with a low average score of 630 had at least 90 percent Black and Hispanic residents. These figures demonstrate that rent payment data could help lift the scores of renters in general, while minorities in particular were likely to benefit greatly. Similarly, TransUnion (2021)44 noted that the inclusion of rental payments in credit files boosted the average credit score of consumers by 60 points. The study found that the largest score increases were among the unscorable and subprime consumer cohorts with least access to favorable terms for financial goods and services. For example, 9 percent of consumers went from unscorable to scorable with an average credit score of 631. The study also found that 12 percent of consumers shifted to a higher score tier, with subprime consumers being the most likely to be classified as near prime. It revealed that 60 percent of credit scores increased as early as the first month of rent payment reporting. Box 2: Integrating the Rental Payment Records into Credit Files 2.4 Opportunities, challenges, and risks of using alternative data The ability to leverage alternative data is closely linked to the quality of digital infrastructure and level of digital literacy. While digitalization has expanded avenues for financial inclusion and innovation, it has also exposed disparities in access and exacerbated existing barriers to credit. Individuals from marginalized backgrounds, for example, may encounter difficulties due to limited digital literacy, further constraining their ability to navigate digital financial services and to access credit. Similarly, MSMEs must adjust to the transition to digital payment ecosystems, particularly those in markets that rely heavily on cash transactions. Therefore, the shift toward digital finance, encompassing alternative data-informed lending, requires significant investments in digital infrastructure and digital literacy to mainstream such practices sustainably and inclusively. The integration of alternative data in credit risk assessment brings forth several opportunities, challenges, and risks across different dimensions. The table below outlines the key impacts of alternative data use on each aspect, including opportunities, challenges, and risks. 22 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges The use of alternative data in credit risk assessment Aspect Opportunities Risks and challenges • Advancement of financial • Widening the digital divide: limited access to technology or digital inclusion through wider credit literacy among certain segments of the population may exacerbate availability. disparities in accessing financial services and opportunities. • Expansion of credit access for Financial inclusion individuals and MSMEs with limited credit history. • Potential for improved credit scores. • Expedited credit risk assessment processes. • Enhanced accuracy in credit evaluations. • Higher approval rates for credit applicants. • Deeper insights into borrower behavior through diverse data inputs. • Access to a wider range of data • Inaccurate or outdated data can lead to flawed risk assessments. points for comprehensive risk • Inability to accurately link the alternative data to right data subjects. assessment. • Market volatility and economic uncertainty may affect the predictive • Deeper insights into borrower power of alternative data. Data availability, behavior through diverse data • It will take time to understand the stability and reliability of new relevance, and inputs. types of alternative data, increasing the importance of scorecard usefulness management. • Difficulty accessing relevant data for individuals with limited digital footprints. • Evaluating data from the informal sector characterized by volatility. • Difficulty to validate certain data sets. • Promoting financial • Mutability of certain alternative data sets. responsibility and good credit • Applicants, unaware of how their data is used, may feel constrained, Borrower behavior culture across the market. fearing negative credit score impacts. • Stimulating increased digitization among MSMEs to enhance credit accessibility. • Fostering fair lending practices. • Some types of alternative data may serve as proxies for sensitive • Addressing ethical issues attributes and may introduce hidden biases in credit scoring models, Bias and fairness surrounding the use of potentially leading to discriminatory outcomes or inequitable alternative data. treatment of certain demographic groups. • Potential accountability and • Some algorithms’ opacity may obscure errors and biases, complicating Algorithmic transparency in algorithmic understanding and raising accountability and trust concerns. decision-making, credit scoring. transparency / model • Opportunity to mitigate bias in interpretability credit decisions. • Re-evaluation and • Lack of unified regulation for alternative data in credit scoring. modernization of existing • Compliance with a complex array of regulations, including data privacy, regulatory frameworks. fair lending, and credit reporting standards. Regulatory • Constructive discourse on • Difficulty meeting explainability requirements for alternative credit frameworks issues such as data privacy, scoring processes. accessibility, and lending • Unregulated entities and regulatory arbitrage. accountability. • Scalability for larger scale • Integration into existing credit risk assessment frameworks can be Integration and adoption. technically challenging and resource intensive. scalability • Enhanced operational efficiency. Collaboration with • Access to vast amount of data • Development of standards for secure and seamless data flow from third- party providers through collaboration with generation to final use. third-party data providers. • Strengthening security and data • Addressing additional security risks, particularly concerning the Cybersecurity protection and enhancing data storage and transmission of sensitive information. governance. Table 2: Alternative data use, opportunities, risks, and challenges Source: Authors The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 23 3. Evolution of adoption and use of alternative data in credit risk assessment 3.1 Regional differences in types of alternative data use While the use of alternative data is expanding globally, disparities in adoption are seen across regions. In Europe, initiatives such as open banking (which allows third-party service providers to access consumer banking, transactions, and other financial data through application programming interfaces (APIs)) and open finance (which goes beyond open banking to allow consumers to access and share their data from a broad suite of financial products, including mortgages, investments, insurance, and pensions) are both contributing to innovation in credit risk assessment. Similarly, countries across Latin America, driven by a growing fintech sector, are increasingly exploring the use of alternative data to improve financial inclusion and expand access to credit for underserved populations. In North America, the three largest credit bureaus in the country—Equifax, Experian, and TransUnion— have developed products that integrate alternative data in their credit reporting processes. In contrast, regions such as the Middle East and North Africa (MENA), some Central Asian countries, and Small Island Developing States (SIDS) prioritize the enhancement of their credit reporting frameworks. In Asia, platforms such Alibaba and Flipkart have played pivotal roles in driving the adoption of alternative data in credit scoring. These companies, with their extensive e-commerce ecosystems, analyze vast amounts of data generated on their platforms, such as customers’ behavior and payment histories, to underwrite credit to millions of users (both MSMEs and individuals). In Sub-Saharan Africa, the increasing adoption of mobile technology has led to greater social and economic connectivity, with mobile devices serving as tools not only for communication and interaction but also for driving financial inclusion. The region has experienced significant growth in mobile money adoption, accounting for nearly half of the world’s active mobile money services, registered accounts, and transaction volume in 2021, as reported by GSMA.45 Safaricom, a leading provider of mobile money services, has capitalized on this trend for over a decade through services like M-Shwari and Digifarm. The influence of Mobile network operators (MNOs), led by Safaricom, remains strong in the alternative credit market across Sub-Saharan Africa, owing to factors such as widespread network coverage, a large customer base, and efficient service delivery. Transactional and third-party verified data types remain prevalent across various regions, including telco data, utility payments, app usage, mobile money transactions, and e-commerce participation. Social media data, however, remains underutilized by established providers, notwithstanding its potential application by smaller market players, such as micro-lenders. Stakeholders interviewed unanimously agree that, despite the attention, the significance of social media data in credit assessment is limited due to its susceptibility to manipulation. Moreover, the use of social media data for credit risk assessment faces additional challenges due to platform policies, such as those enforced by Meta (formerly Facebook), which explicitly prohibit use of platform data for credit eligibility determination. The extent of compliance among alternative lenders and credit scoring providers remains uncertain, however. The following is a sample of key players in the MSME and consumer lending sectors in various countries, each employing a distinctive approach to credit assessment and using various types of alternative data. MSMEs Kabbage, a US-based fintech established in 2008 and acquired by American Express in 2020, specializes in small business lending. It primarily operates online, offering lines of credit to small businesses. The company assesses creditworthiness by analyzing a diverse set of data points sourced from business checking accounts, accounting software, online payment platforms like PayPal and Stripe, shipping data from UPS, and transactions on e-commerce marketplaces such as Amazon, eBay, and Etsy. Additionally, a personal credit check supplements the assessment. WeBank, a digital-only bank headquartered in China, offers a variety of financial products and services, including small business loans. It leverages data from Tencent’s ecosystem, which encompasses WeChat, a popular messaging and social media application widely used in the country, to assess credit risk and provide customized financial solutions for MSMEs. Tencent, a Chinese conglomerate, operates several digital platforms and services in addition to WeChat. 24 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges Evolution of adoption and use of alternative data in credit risk assessment Mujeres WOW, operating in Ecuador, the fintech company specializes in lending to women-owned businesses. Its credit scoring model incorporates various data points, including demographic information, social media activity, SME expenses, and government data related to SME such as registration. Additionally, it uses a review-based referral system through which other women provide feedback on the borrower’s social reputation, based on their relationships in different capacities as friends, clients, colleagues, and business partners. Kifiya is an Ethiopia-based alternative credit scoring and digital lending infrastructure provider. The company offers solutions that enable financial institutions to extend uncollateralized loans to underserved segments with a focus on MSMEs. Through its Qena digital lending platform, the company offers an alternative credit scoring service that leverages various sources of data including SMS data, surveys, and finance and accounting data. Consumers FinScore, is an alternative scoring company primarily based in the Philippines. It uses telco data in credit risk assessment, capitalizing on the widespread use of mobile phones in the country. FinScore’s Telco Credit Score system incorporates 400 telco data variables, encompassing the consumer’s bill payment history, call patterns, usage history, SIM card age, and top-up patterns. Findo, an Argentine company, offers a range of alternative credit scoring products. Their solutions leverage various alternative data such as the applicant’s living capital, social capital, inferred income, and other data obtainable from cell phones. The company uses over 10,000 real-time data points obtained from cell phones. LENDMN is a Mongolia-based digital lender that provides noncollateralized microloans to consumers in urban and rural areas. The company’s credit scoring platform uses a combination of credit and behavioral scoring methods to assess the creditworthiness of loan applicants. Data sources include banking transaction history, in-app activities, and repayment tendencies. Simbrella is an Azerbaijan-based company specializing in Mobile Fintech services. The company offers various instant financial services platforms to banks, MNOs, and mobile wallets by leveraging mobile wallet data and telco data to extend loans to individuals, including those without a credit history. Box 3: Examples of Lenders and Credit Scoring Companies Leveraging Alternative Data Sources To expand the range of alternative data sources, industry participants are exploring additional avenues. For example, satellite imagery has emerged as a promising tool for assessing the creditworthiness of farmers. This method provides insights into factors such as land use and agricultural productivity, offering valuable indicators of assessing agricultural facilities. The use of satellite imagery for credit assessment presents challenges, however, including the need to address issues such as land ownership determination as well as the substantial capital and expertise required to implement such methods. Remittances are another emerging data source. Global remittances increased to $794 billion in 2022, of which $626 billion went to low- and middle-income countries.46 Remittances are the largest source of external finance in LMICs, exceeding foreign direct investment (FDI) and official development assistance (ODA) in some markets. The size of total remittances is estimated to be up to 50 percent higher, as people often use informal methods that are not captured.47 The pandemic and the war in Ukraine have shown that remittances are largely resilient to macroeconomic shocks. Financial institutions in Latin America and parts of Asia already leverage data on remittances for credit underwriting to remittance beneficiaries. The regional disparities in remittances as an alternative data source in creditworthiness assessments partly reflect the wealth distribution, migration patterns, and the level of organization of the migration corridors. For example, regions with organized corridors, such as Asia, where there are government-to-government work arrangements and where lenders have serving points in host countries, tend to be more advanced in the use of remittances data. The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 25 Evolution of adoption and use of alternative data in credit risk assessment This finance against remittances project was based on a collaborative effort involving Laxmi Bank, a commercial bank in Nepal; AlFardan Exchange, a United Arab Emirates–licensed money transfer operator (MTO); New Street Tech, a blockchain enterprise based in India and the Middle East; and the United Nations Capital Development Fund (UNCDF) for investment and technical support. The United Arab Emirates is a common destination for Nepali migrants seeking sustainable opportunities to support their families in Nepal. Pathways to financial services through which these families can receive financial support remain scarce, however. This is due mainly to the information asymmetry between financial institutions in Nepal and the country in which the remitter resides, limiting access to financial services such as loan products. The consortium of partners sought to address this problem by launching FAR (Finance Against Remittances) in the Nepal–United Arab Emirates corridor. This system collateralizes the remittances to extend loans to migrants’ families in Nepal. The project started as a pilot in 2019, but it was disrupted during the COVID-19 pandemic. The partnership between Laxmi Bank and UKaid Skills for Employment, a Nepalese program, has further accelerated the financing against remittances effort. Building on its experience in financial innovations with UNCDF, Laxmi Bank further modified and expanded technology- based financial products to better reach and service the target migrant Nepalis. Laxmi Bank also drew on its prior experience to re-engineer its value proposition by integrating an application programing interface (API) platform to make loans against remittances more accessible and affordable. It increased its focus on migrant households in the Madhesh and Lumbini provinces, targeting and expanding its base of migrants working in and traveling to corridors in Malaysia, Qatar, Saudi Arabia, and United Arab Emirates. The revamped API technology eliminated bottlenecks and hiccups faced earlier in the project, allowing migrant users to make loan requests and upload their information and documents from their destination countries. This information was then further validated through the beneficiaries residing in Nepal. Despite the progress made in recent years, access to information and formal finance is still limited among migrants and migrant households in the target areas. Laxmi Bank has been able to leverage its digital solutions network, as well as to make arrangements with money transfer operators (MTOs) and employers in destination countries, to redesign and refine its products and services to better reach the formerly unreached sector and communities. This has not only made loans accessible; it has also enabled productive use of remittances by making the credit access and disbursement process easier and by lowering the cost of migration through affordable pre-migration loans that give the target group increased access to financing. In addition, through its microfinance partner Laxmi Laghubitta, Laxmi Bank has extended to its customers tailored financial products and skills development training to improve their financial literacy, increase their capacity to generate income, and develop their entrepreneurship abilities. The FAR model for finance against remittances represents an emerging trend in the lending market. By leveraging blockchain technology and considering remittances as potential collateral, it offers a solution to address data asymmetry challenges and expand access to credit for underserved populations. Box 4: Laxmi Bank—Introducing the Concept of Finance Against Remittances (FAR) in Nepal 3.2 Emerging business models and strategies An extensive study conducted by the Global Partnership for Financial Inclusion (GPFI) on the transformative potential of alternative data in SME financing identified several key institutional models using alternative data for SME lending decisions. These include SME marketplace lenders, tech, e-commerce, and payment giants, supply chain platforms, mobile data–based lending models, and digital bank models.48 This section expands on some of the key findings of that study by providing an updated outlook on emerging trends in the alternative credit market serving both consumers and MSMEs. As alternative credit scoring markets rapidly expand, driven by reduced digital entry barriers for market players, demand is ongoing for innovative solutions. This growth has fostered economies of scale, underscoring the importance of partnerships in which players leverage their strengths. By examining the credit market through both a global and a regional lens, various strategies, business models and products, delineated below, emerge. 3.2.1. Credit reporting entities integrating alternative data into their databases In view of the proliferation of digitized data and technological advancements, credit reporting service providers (CRSPs), such as credit bureaus, are increasingly turning to additional, structured, and unstructured data to score consumers and businesses with limited credit bureau information and provide additional insights to help lenders make decisions. The bureaus have benefited from supportive legal frameworks that enable them to collect alternative data such as utility, rent, and trade data. As a result, most of the bureaus have developed alternative data products. 26 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges Evolution of adoption and use of alternative data in credit risk assessment Experian Launched in 2019, Experian Boost enables customers to boost their scores by adding real-time alternative data, such as positive billing histories on utility, telecom, and Netflix use, directly into their credit file. The tool allows customers the option to add or remove data at any time. The customized version of the tool has been launched in emerging markets including Brazil and South Africa. As of March 2024, more than 15 million consumers in the United States have connected to Experian Boost.49 TransUnion TransUnion developed its suite of Credit Vision (CV) scoring products for consumers, formal businesses, informal businesses, and microfinance clients that leverage alternative data to enable firms to serve thin-file and credit-invisible clients. The products leverage data such as mobile use, commercial interests, asset ownership, employment, geolocation, trade credit, and rent and utility payment, among others. The company found that CV products increased K-S value50 by between 17 percent and 126 percent for subprime and super prime clients, respectively,51 while combining Credit Vision Link with traditional credit data scoring solutions improved risk predictability of thin-file consumers by up to 25 percent. Equifax In 2016, Equifax partnered with FICO and LexisNexis Risk Solutions, a data and technology provider, to create FICO Score XD, which leverages alternative financial data such as public records and phone, cable, and utility payment histories. The results of its pilot with many of the largest US lenders found that as many 50 percent of previously unscorable clients received scores of at least 620. Over time, the bureau has added more data, such as specialty finance data (which includes short term installment loans, rent-to-own information, and other nontraditional lending attributes) and telco and utility data (which includes telecommunications, pay TV, home security and utility payment history). Other credit bureaus such as CRIF, Creditinfo, and Metropol have also developed alternative data products leveraging other sets of data, including bioclimatic data. Box 5: Credit Reporting Agencies’ Use of Alternative Data 3.2.2. Collaboration between banks and alternative scoring companies Banks bring established infrastructure and customer bases, while alternative scoring companies offer innovative data analytics and risk assessment methods. A collaboration between Tonik Bank, a digital bank in the Philippines, and FinScore showcases such synergies. Initiated in 2021, this partnership involves integrating FinScore’s AI-powered Telco Data Credit Scoring module into Tonik’s digital platform. This initiative aims to enhance loan accessibility for Filipino borrowers by using alternative data, capitalizing on the nation’s high smartphone penetration rate. The partnership between Nigeria’s First City Monument Bank (FCMB) and Simbrella, an Azerbaijan-based fintech providing credit scoring solutions using mobile wallet telco data, offers another example. Their joint initiative, the FastCash product, launched in Nigeria in 2018 and makes available an instant loan solution requiring no collateral. 3.2.3. Partnership between credit reporting agencies and fintechs Credit reporting agencies typically have extensive payment-related data, including utilities and telecommunications records, to which fintechs often lack access. Conversely, fintechs leverage alternative data from nontraditional sources that credit bureaus typically cannot access. This discrepancy in data availability highlights the potential for collaboration, or even acquisition, between the two entities. By merging their respective datasets and analytical capabilities, these organizations can enhance credit scoring models to better evaluate the creditworthiness of both existing and new credit holders. For example, in 2019 CRIF High Mark partnered with CreditVidya, an alternative credit scoring company based in India, to boost credit scoring by incorporating additional digital footprints. In 2021, Experian and FinScore joined forces to provide telco data credit scoring to the unbanked and underbanked in the Philippines. Through this exchange of expertise, the two companies developed Experian PowerScore, leveraging FinScore’s capabilities in the field of alternative data.52 Additionally, in May 2022, Experian acquired a majority stake in MOVA, a Brazilian fintech specializing in data-driven credit assessments for SMEs, enhancing Experian’s B2B solutions and facilitating SME credit access.53 3.2.4. Open banking, open finance, and the gig economy Open banking and open finance models capitalize on the vast financial data from platforms that facilitate transactions for both merchants and gig workers. These platforms access transactional data via open banking, informing revenue streams and business performance for merchants. Credit scoring firms use this data to tailor credit solutions, such as cash advances, specifically for merchants. On the other hand, gig workers who generate transactional data on digital platforms offer valuable insights into their income and payment reliability. Credit scoring firms use this data to develop alternative scoring models, which assess gig workers’ The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 27 Evolution of adoption and use of alternative data in credit risk assessment creditworthiness based on their earnings history and transaction patterns. This approach facilitates their access to products like personal loans or credit lines (see Box 6 for more examples). The samples below illustrate how the principles of open banking and open finance, coupled with the use of alternative data, shape tailored financial solutions for various market segments, including individual consumers, MSMEs, and gig economy workers. Consumers In February 2021, Monzo Bank—a UK-based digital bank—partnered with Experian for Open Banking credit scoring, enabling customers to connect their accounts to Experian Boost, which uplifts Experian credit scores by factoring in subscription payments and utility bills. Similarly, in March 2022, Santander and Experian collaborated to use Experian Boost to offer a competitive credit card to Experian customers that would help improve their credit scores. Experian Boost was officially introduced in the United States in 2019 and in the United Kingdom in 2020. Plaid, a US-based fintech, enables consumers to share specific financial information from their banks with apps, services, and companies. The company helps to share account and routing numbers, balances, transaction histories, personal plans and credit cards, and investment holdings from 11,000 different financial institutions to thousands of services with the goal of helping consumers to mobile bank, save, invest, build businesses, make payments, and lend. Plaid acts as an intermediary between a consumer’s financial accounts and the financial tool or service they seek to access. The company also provides consumers with the option to send positive information to financial institutions, but it is not a data furnisher that automatically provides information to credit bureaus. In 2019, Experian partnered with FICO and Finicity, an open banking platform provider, to create an UltraFICO Score, which factors in alternative data such as how long accounts have been open, the frequency and recency of bank account transactions, evidence of cash on hand, and history of positive account balances. Currently, consumers of participating financial institutions can grant permission for Finicity to access data from their checking, savings, and money market accounts and for prospective lenders to review their data in the decision-making process. MSMEs In March 2022, CreditVidya announced a partnership with Flipkart Wholesale, a digital B2B marketplace.54 Under this alliance, CreditVidya provides an embedded buy now and pay later (BNPL) credit solution to Flipkart wholesale merchants, facilitating access to a capital pool of US$100 million through various financial institutions. This initiative is particularly beneficial for MSMEs looking to expand their inventory. In 2022, Tillful, a US-based small business credit platform fintech that helps businesses gain access to finance, announced a partnership with Experian, with the aim of building credit profiles for SMEs new to credit. Through the partnership, Tillful users can connect their business accounts with Experian, which will use their real-time cash flow data to build a credit report. Gig economy workers In 2021, Uber introduced the “Drive to your Mibanco benefits” program in Peru, allowing Uber drivers to digitally request loans.55 This collaboration between Uber and Mibanco, the microfinance arm of Credicorp and Peru’s largest financial services holding company, leverages open finance to extend loans to drivers. Drivers simply log into Mibanco using their Uber credentials, provide necessary information and the desired loan amount, and a Mibanco advisor evaluates the request. Similarly, Abaco Latam, a fintech of Colombian-Spanish origin, explored this frontier. The company extends credit products to gig economy workers by analyzing their digital wallets and use of gig economy apps via open APIs. In addition to this analysis, a psychometric test is used to further assess gig workers’ creditworthiness. Box 6: Open banking and open finance examples 3.2.5. AgriData credit model Agritech companies are reshaping credit access for farmers, particularly in regions where traditional banks have been hesitant to lend due to data scarcity and the inherent volatility of agriculture. These startups harness various data types, including crop yields, weather patterns, soil health, historical farm performance, financial transactions, and satellite imagery, to evaluate the creditworthiness of farmers. 3.2.6. Incumbents with in-house built-in capabilities These players often enter the market early, developing proprietary solutions and refining them over time. For instance, MNO Safaricom leveraged M-Pesa data to expand into the credit market, offering products like M-Shwari and DigiFarm. Similarly, in the United States, the fintech Kabbage incorporates diverse alternative data into its credit scoring algorithms and acquired several data analytics firms over time to enhance its services for SMEs. E-commerce platforms like Alibaba, and Mercado Libre have developed internal capabilities to leverage alternative data to lend to users on its ecosystem. Ant Group, through its Alipay platform, uses transactional data (generated on Alibaba e-commerce platforms) and its Sesame Credit scoring system to provide credit scoring and lending services to millions of users. On the other hand, Mercado Libre is also lending to small businesses through Mercado Pago, leveraging the vast amounts of data generated on its platforms. 28 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges Evolution of adoption and use of alternative data in credit risk assessment 3.2.7. Credit building products The emergence of credit building products marks another important shift in the credit landscape, offering consumers new avenues to enhance their credit scores. At the forefront of this paradigm shift are credit reporting and scoring platforms (CRSPs). These products often include features such as positive payment reporting, which factors timely payments for utilities and other recurring expenses into individuals’ credit scores, providing a more holistic representation of their creditworthiness. eCredable, a financial services company based in the United States, offers a service called eCredable Lift, which allows consumers and small business owners to improve their credit scores. Through this service, consumers can add relevant information to their credit files, resulting in enhanced credit scores. These updated scores are reported monthly directly to TransUnion, provided that the consumer’s account remains linked to eCredable. eCredable has partnerships with over 2,000 utility companies, allowing consumers to include up to 24 months of complete payment history from these providers. Unlike traditional credit reporting practices that only include missed payments, eCredable Lift incorporates both positive and negative payment information. Experian Go, launched in 2022, is a tool designed to help consumers with limited or no credit history initiate, manage, and improve their credit profiles. The service is accessible through Experian’s free mobile app, where users can enroll in a complimentary Experian membership and verify their identities. Users are then provided with tailored recommendations to kickstart their credit journey and establish a FICO score by adding credit accounts, also known as tradelines, to their Experian credit report. Also, users can include their on-time bill payments directly into their credit history using Experian Boost. UltraFICO Score, which began as a pilot program in 2019 through collaboration among FICO, Experian, and Finicity, a financial data aggregator, aims to broaden access to credit for more consumers. This initiative connected the consumer’s checking, savings, or money market accounts, which can be overlooked by traditional credit assessment methods. These accounts are then examined to evaluate the consumer’s financial management practices. FICO integrates this data with the Experian credit file to generate an UltraFICO Score.56 This scoring model can be particularly beneficial for individuals navigating financial challenges, offering them an opportunity to demonstrate responsible financial behavior and gradually rebuild their creditworthiness. Box 7: Examples of Credit Building Products 3.3 The impact of COVID-19 on the alternative data market The COVID-19 pandemic affected the ability of traditionally good borrowers to meet their obligations as they fell due, leading to a decline in scores and thus interfering with their ability to access additional credit. MSMEs were among those most negatively affected by the pandemic. As is often the case in times of crisis, lenders took a more conservative approach in lending to MSMEs during the pandemic—typically by limiting lending, prioritizing or relocating lending to more creditworthy borrowers, or asking borrowers for more reassurances in the form of credit assessment information. The COVID-19 pandemic placed significant strain on MSMEs due to intermittent lockdowns and containment measures, resulting in both temporary and permanent business closures. This downturn led to adverse outcomes such as job losses and reduced turnovers. While some MSMEs adapted to the growing digitalization trend by transitioning from offline to online operations, it remains uncertain whether this growth will be sustainable in the long term, particularly in the competitive e-commerce sector. MSMEs unable to transition to e-commerce, either due to the nature of their businesses or because they operate in the informal sector, faced numerous challenges, including limited access to financial support and credit, making survival their primary concern. To mitigate the impact of the crisis, regulatory authorities implemented prudential measures, including public and private moratoria. In response, CRSPs implemented credit reporting measures, such as using technical reporting codes to distinguish borrowers affected by the pandemic. The industry also accelerated the adoption of alternative data such as transactional data, utility and rental data, e-commerce data, and subscriptions (such as cable, Amazon, and Netflix) to complement traditional data. By incorporating alternative data, the bureaus and lenders could get a more complete, real-time view of their customers, thus making it easier to distinguish false negatives from real negatives, averting the potential credit rationing that typically follows a crisis. In South Africa, for example, the rate of “false bads” increased from 1.5 percent of credit active population pre-COVID to 8 percent in October 2020.57 Real-time alternative data also helped distinguish between good corporate borrowers that were failing to meet their obligations due to the pandemic and zombie firms that would not recover from the crisis. Lenders and CRSPs also actively sought new sources of alternative data capable of capturing borrowers’ evolving economic and social behaviors, while ensuring reliability for integration into their predictive models through data such as geolocation, satellite images, foot traffic, flights, and freight movement. Additionally, various technology firms began generating mobility trend reports. For instance, Apple and Google introduced mobility indexes based on mobile phone data. Foursquare, a location data company, developed a visitation index reflecting activity levels at businesses and recreational facilities across the United States during the pandemic. Similarly, OpenTable tracked global restaurant reservations to gauge consumer financial health. The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 29 Evolution of adoption and use of alternative data in credit risk assessment The pandemic inadvertently brought forth an unprecedented wave of innovation in credit risk assessment. Credit reporting service providers introduced next-generation credit decisioning models, and alternative lenders such as Kabbage leveraged their experience in digital SME lending to participate in the Paycheck Protection Program (PPP). The Creditinfo COVID Impact score combines alternative data including short- and long-term industry impact outlook and a company’s bureau credit score. The score helped identify the companies hit hardest by the COVID crisis and that were thus likely to have solvency problems in the near future. Credit providers used information and scores to design and execute a set of focused actions. The product was launched in Baltic markets, namely Iceland, Latvia, and Estonia, at the onset of the pandemic and has now been deployed across several markets.58 The Fina COVID score, introduced in April 2020 by Financial Agency (Fina), to assess the vulnerability of businesses in Croatia during the pandemic and to help process applications for government support and liquidity loans. The scoring system assesses nine elements of risk to determine a business’s liquidity: business type, dominant activity, Fina rating, business subjects with frozen accounts, reduction in the number of employees, expected decline of business revenues compared to the same period in 2019 (prior to the pandemic), projected liquidity, failure to pay employees’ salaries (eliminating criteria), and whether the business is on the list of tax debtors.59 In the United States, FICO introduced a Resilience Index in the context of the CARES Act,60 when payment deferrals extended by banks during the COVID-19 pandemic were not reflected in consumers’ credit scores. The natural consequence of this was that lenders could not accurately determine the creditworthiness of borrowers, causing creditors to tighten standards and amend their lending rules to limit risk in a period characterized by huge uncertainty. To help creditors assess the creditworthiness of borrowers during the rapidly changing conditions of the pandemic, FICO introduced the Resilience Index, an innovative metric designed to provide insights into consumers’ resilience during an economically volatile time. Used to complement the industry standard FICO Score, the Resilience Index rates consumer resilience on a scale from 1 to 99, with the 1 to 44 range considered the most prepared and able to weather an economic shift.61 Highly resilient consumers tend to satisfy the following criteria: more experience managing credit, lower total revolving balances, fewer active accounts, and fewer credit inquiries in the last year preceding the pandemic. No clear or available statistics are available on the adoption of the Resilience Index during the pandemic. Box 8: COVID-19 Credit Scoring Products 30 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges Overindebtedness The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 31 4. Regulatory and policy environment 4.1 Alternative data: Regulatory challenges and track the borrower’s movements and exert pressure on them to considerations repay the debt, potentially leading to harassment or coercion. 4.1.1. Privacy concerns, data quality, accuracy, and • Data accuracy and quality consumer consent The General Principle on Credit Reporting GP 1 underscores Integrating alternative data into credit risk assessment prompts the importance of relevant, accurate, timely, and sufficient many considerations, including data privacy, accuracy, and data in credit reporting systems. Similarly, both European consumer consent, as well as transparency and fairness in the Banking Authority (EBA) guidelines on loan origination and use of algorithmic models. Consultations conducted for this monitoring and the Consumer Credit Directive (CCD) underline report, supplemented by comprehensive desk reviews, revealed the necessity of having adequate data to assess borrower risk a global absence of specific regulations governing alternative effectively. These guidelines emphasize the need to consider credit scoring and its associated complexities. These findings various factors influencing repayment capacity and credit highlight critical ethical and legal challenges surrounding the risk, including the consumer’s income, expenses, and other use of alternative data in credit risk assessment. relevant financial information. Currently, no unified standards apply to determining the suitability of alternative data in credit • Data privacy assessment, and lenders are tasked with experimenting with different data types and determining their usefulness based on Data privacy laws, such as the European Union’s General their own criteria. Data Protection Regulation (GDPR), have given data subjects sovereignty over their own data and have introduced important • Consent principles regarding purpose, limitation, relevance, adequacy, proportionality, and transparency of data processing. Similarly, The concept of consent lies at the heart of alternative data credit several African countries have strengthened their existing data scoring models, where borrowers are typically required to grant privacy regulations by providing individuals with greater control permission to collect and use their personal data. Navigating over their data and implementing stricter rules on its usage. the intricacies of consent in the context of alternative data use While such regulatory frameworks aim to safeguard individuals’ presents notable challenges and potential regulatory loopholes, rights, navigating potential loopholes remains a challenge given however. While borrowers may seem to provide consent for the diverse nature of alternative data sources that can be used data access, the complexity of modern data-sharing practices, and the potential obstacles to ensuring full compliance. coupled with the opacity of how alternative data is used for credit scoring, raises questions about consumers’ full understanding Data privacy implications arise particularly significantly in the of such consent. context of alternative credit scoring. The nature of data used, such as mobile phones use, text messages, and geolocation, A key challenge revolves around the clarity and comprehensibility escalate the data privacy and security implications for subjects. of consent mechanisms employed by alternative credit scoring Unlike traditional credit data, which is typically governed by strict platforms. These consent agreements can be incomprehensible regulations and obtained with explicit consent, alternative data to the average consumer, leaving borrowers vulnerable to collection often involves data points that individuals may not consent to terms they may not have fully understood. This leads consciously provide and that may include private and sensitive to a gap between the intended purpose of consent—enabling information about the data subjects themselves and their social borrowers to make informed decisions about their data—and connections. If not adequately protected, this information could the practical reality, wherein borrowers may unintentionally be vulnerable to breaches or exploitation, potentially leading to agree to data use terms without fully understanding the identity theft, discrimination, or other forms of harm. Moreover, implications. In other cases, borrowers may feel compelled to some data types could be leveraged to harass borrowers consent to data collection and use to access loans even if they into repaying. For instance, if a borrower falls behind on loan don’t understand the privacy implications. This phenomenon of payments, a lender could potentially use geolocation data to coerced or uninformed consent may undermine the principles 32 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges Regulatory and policy environment of autonomy and agency in data-sharing. Aronowitz, Golding, and Choi suggest, a possible remedy is to rein in risk-based pricing—which they note is a form of price In addition to consent from the primary service user (called here discrimination. Limiting or eliminating risk-based pricing is a subject A), understanding data privacy warrants discussion of the controversial solution, to say the least, but Aronowitz, Golding, consent of any silent parties (subject B). Subject B’s data may be and Choi argue that risk-based pricing is not necessary for embedded across a range of subject A’s data: in a phone book, in safe lending and a profitable mortgage industry. This example message exchanges, or even in bank statements. While subject A raises two issues. First, traditional data may contain biases has provided a service with consent to access and process his or that can have unintended and often expensive consequences if her data, subject B has not. Subject B is not using the service, nor borrowers’ alternative data inadvertently triggers hidden biases, does she or he have any contractual agreement with the service with possible detrimental socioeconomic effects. Second, a provider. Furthermore, subject B may be unaware that the data risk-based pricing system layered on decades of institutional is being accessed at all. As a result, subject B, as a silent party, discrimination may further exacerbate the inequalities that has not given consent for her or his data to be used, but it has marginalized communities suffer. become, by its nature, unavoidable bycatch in alternative data models. The growing scope of alternative data use by several The findings of the study underscore how systemic biases, such market players—some of which may overlook this situation, as those revealed in mortgage pricing disparities, intersect intentionally or not—clearly could lead to potential data privacy with lending practices, potentially perpetuating economic violations. disparities. Vulnerability factors such as socioeconomic status, employment history, and access to traditional financial services For countries under the (combined) provisions of Directive can all influence a borrower’s perceived creditworthiness. 2002/58/EC and the GDPR, use of silent party data in subject A’s When alternative data is used in credit scoring alongside credit scoring can take place only after explicit authorization traditional data, these vulnerability factors may be amplified from the silent party affected or after anonymization of the or misrepresented, leading to inaccurate assessments of data. In other parts of the world, some newly introduced digital creditworthiness, consequently resulting in less favorable terms, credit regulations delineate a clear boundary protecting silent such as higher interest rates or stricter lending conditions. party data, such as India’s digital lending regulation, which prohibits access to sources where such data could be found. This Addressing potential discrimination in automated decision- necessarily includes mobile phone resources, such as contact making processes remains a significant challenge. Efforts to lists, call logs, etc., thus systematically addressing the issue of tackle discrimination using practical means are ongoing, with silent party consent. Overall, regulatory frameworks should the application of parity principles in machine learning being address the nuanced dynamics of consent in the digital age to one widely supported approach. For example, “fairness through ensure that borrowers can retain meaningful control over their unawareness” and impact parity concepts have been extensively personal information while accessing financial services. discussed, yet they possess inherent limitations, particularly inadvertent correlation of seemingly neutral features with 4.1.2. Mitigating discriminatory risks in alternative data protected attributes, thereby challenging their perceived credit scoring fairness. Furthermore, comparing algorithmic risk assessments with actual repayment behaviors, especially among vulnerable An issue inherent to using alternative data in credit risk demographic groups, offers another avenue for addressing assessment is discrimination. For example, the US Equal discriminatory outcomes. Credit Opportunity Act (ECOA) explicitly prohibits lenders from discriminating against individuals based on certain protected 4.2 Regional regulatory responses characteristics, such as race, religion, or marital status. Alternative data sources may, however, inadvertently include Regulatory responses to adoption of alternative data have varied information that correlates with these protected characteristics. across markets, with some adopting a prescriptive approach For instance, details about an individual’s social connections while others opt for a soft and incremental approach. or educational background could indirectly suggest their socioeconomic status. Consequently, relying on alternative data 4.2.1. Examples of prescriptive approaches that serve as proxies for protected characteristics may result in credit scoring outcomes that inadvertently perpetuate biases, Several countries, including Brazil, Tajikistan, and Uganda, negating the intended purpose of existing laws and regulations as well as Banque Centrale des États de l’Afrique de l’Ouest aimed at preventing such discriminatory practices. (BCEAO), have sought to codify the integration of alternative data by incorporating it into their credit reporting laws and A recent study62 examines what Black mortgagors in the US pay regulations. Most of these markets have aimed to incorporate as compared to their white counterparts; significant inequities alternative data sources such as trade credit and telco, rental, were found, with Black populations paying more. This stems not and utility histories into the data submitted to credit bureaus. only from the effect of a long history of redlining and of fewer Other countries have chosen a different approach by issuing opportunities to accumulate generational wealth, but also digital lending regulations that indirectly govern the use of historical bias across other economic and social pillars, such alternative data in credit scoring. The examples below provide a as employment, that put Black Americans at a disadvantage sample of key regulatory responses across various jurisdictions. and lead institutions to view them as riskier borrowers. As The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 33 Regulatory and policy environment People’s Bank of China are increasingly utilizing alternative data in their scoring models). In 2021, the People’s Bank of China (PBoC) issued its “Measures for the Administration of Credit Reporting Services,” which The regulations and guidelines now in effect in Europe, along came into force in January 2022. Under the new measures, with the GDPR, present the most comprehensive and significant “credit information” refers to the basic information, lending regulatory effort to date. information, and other relevant information lawfully collected to identify and assess the creditworthiness of enterprises and The Reserve Bank of India individuals to facilitate financial and other activities, as well as the analyses and evaluations made based on that information. While India has not enacted any specific reforms directly Considering the wide application of alternative data in the credit targeting integrating alternative data, several reforms reporting industry, the new measures expand the definition implemented by the Reserve Bank of India (RBI) address of data that can be collected to include alternative data. issues associated with the use of alternative data in credit The regulations, however, impose a requirement that credit risk assessment. For example, in 2021, the RBI introduced reporting agencies conduct the checks necessary to verify the the Account Aggregator (AA) framework, a consent-based source, quality, and accuracy of data obtained from different system that streamlined the accessibility of financial data for information providers.63 individuals and small businesses. The AA framework gives data subjects control over how their data is shared and used and The European Commission enables the use of alternative data in credit underwriting. The following year, the Reserve Bank of India issued Guidelines on The European Commission issued its new Consumer Credit Digital Lending66 addressing lending activities through online Directive (CCD) in 2023 to adapt the consumer credit regime to platforms and mobile apps. The guidelines establish rules for consider the new forms of credit available online and the new regulated entities (REs) and lending service providers (LSPs), risks that they present. The CCD64 acknowledges and addresses mandating compliance in functions like underwriting support the newfound risks and challenges associated with the use of and loan recovery. Notably, the regulations explicitly govern alternative data in creditworthiness assessments. One key clause the use of alternative data. For instance, digital lending apps that touches on alternative creditworthiness is Article (3): “The (DLAs) are required to collect data based solely on necessity, to assessment of creditworthiness shall be carried out on the basis maintain a transparent audit trail, and to obtain explicit prior of relevant and accurate information on the consumer’s income consent from borrowers. Furthermore, access to certain mobile and expenses and other financial and economic circumstances phone resources, such as media, contact lists, and call logs, is which is necessary and proportionate to the nature, duration, prohibited except for specific instances like on-boarding and value and risks of the credit for the consumer. That information know your customer (KYC) requirements. Additionally, REs must may include evidence of income or other sources of repayment, ensure that lending activities through DLAs, including the buy information on financial assets and liabilities, or information on now pay later (BNPL) mode, are reported to credit information other financial commitments.” The Directive excludes the use of companies (CICs). social networks. In 2023, plans for setting up the National Financial Information Prior to the finalization of the Directive, the European Data Registry (NFIR) were announced. The NFIR, which the Reserve Protection Supervisor (EDPS) issued an opinion paper65 Bank of India (RBI) is planning to establish, will collate credit emphasizing the complementary relationship between and ancillary information that relates to the determination of consumer and data protection. In it EDPS suggested the an entity’s creditworthiness. This includes alternative data such following, among other recommendations: as paid taxes and electricity consumption patterns. The registry will enable lenders to have a comprehensive “360-degree” • Clearly define the categories and sources of personal perspective on the creditworthiness of potential borrowers by data that should and should not be used for the purpose accessing an expanded pool of data. of creditworthiness assessment. In addition to the data sources already prohibited by the European legislation, Bank of Thailand such as social media platforms and health data (including cancer diagnoses), the EDPS recommended expanding the The Bank of Thailand (BOT) issued a circular in 2020 on rules, prohibited list to include search query and online browsing procedures, and conditions for undertaking a digital personal data (which are often used/considered as alternative data loan business.67 This directive, aimed at potential business sources). operators seeking digital loan business licenses, emphasized • In line with the principle of data minimization, collect only the importance of integrating alternative data into credit risk adequate and necessary data for the purpose of credit risk assessment processes. Specifically, the circular encouraged assessment, and establish a clear relationship between the the inclusion of alternative data sources such as utility and data collected and the borrower’s ability to repay the loan mobile phone bill payments, as well as e-commerce earning and (aligning with the responsible use of alternative data in spending behaviors. credit assessment). • Regulate the requirements, role, and responsibilities of credit databases or third-party credit score providers (which 34 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges Regulatory and policy environment 4.2.2. Examples of soft and incremental regulatory International Finance Corporation (IFC) and under the South approaches Africa Credit Reporting and Financial Inclusion Project Phase 1, launched a pilot to test the predictiveness of alternative data. • No-action Letter The project supported pilots led by private sector players (credit bureaus, loan marching platforms, and lenders) to develop In 2017, the Consumer Financial Protection Bureau (CFPB) alternative data-based scores to promote the creditworthiness issued a no-action letter (NAL) to Upstart Network—an AI of underserved segments. As of April 2024, the project had lending platform that uses nontraditional data as part of its produced diagnostic research on alternative data in the country, credit scoring.68 Unlike a formal regulatory action, the issuance and five alternative data-based scoring products were developed of the NAL was not intended as a supervisory measure or and tested with banks, MFIs, and MNOs, among other users.70 enforcement action. Instead, its primary aim was to facilitate the CFPB’s understanding of the implications of using alternative In addition, the Inter Fintech Working Group, a gathering of data in credit risk assessments, supported by both quantitative financial regulators in South Africa, admitted into its sandbox an and qualitative evidence. Under the terms of the arrangement, institution that wanted to test the predictive value of rental data Upstart committed to implementing a “model risk management for credit assessment. The test was briefly paused to comply and compliance plan,” whereby the company would routinely with regulatory requirements but has since resumed.71 furnish the bureau with pertinent lending and compliance data, thus assisting in risk mitigation for consumers. This • Regulatory sandboxes arrangement has persisted quarterly since its initiation in 2017 and was renewed upon its expiration in 2020. The issuance of Regulatory sandboxes serve as a critical component in the the NAL serves as an example of how regulatory agencies seek evolving landscape of alternative data use in credit scoring to understand the potential benefits and risks associated with regulation. These frameworks provide a structured environment alternative data in credit scoring. for testing innovative credit scoring models, ensuring that they comply with existing regulatory standards while also In 2019, US financial regulators (the Federal Reserve System, accommodating the unique challenges posed by alternative the Consumer Financial Protection Bureau, the Federal Deposit data. By allowing experimentation within a controlled setting, Insurance Corporation, the National Credit Union Administration, sandboxes enable regulators to better understand the and the Office of the Comptroller of the Currency) issued an implications of integrating alternative data into credit scoring interagency statement on the use of alternative data in credit processes and to tailor regulations accordingly. In Indonesia, underwriting that indicated their awareness that financial for instance, Regulation OJKR No. 13/POJK.02/2018 provides a institutions were using it, or contemplating its use, in credit framework for digital financial innovation. Within this framework, underwriting.69 The statement highlighted the benefits and the Financial Services Authority (Otoritas Jasa Keuangan (OJK)) consumer protection implications of using alternative data and has established a regulatory sandbox that allows companies encouraged reference to applicable laws. The agencies also to operate under relaxed regulatory requirements for up to 18 encouraged responsible use of alternative data. months. In 2020, CredoLab, a Singapore-based fintech company specializing in alternative credit scoring solutions leveraging • Collaboration between regulators and the private sector smartphone data, officially registered as a provider of digital financial technology after meeting the necessary criteria. The OCC, with the goal of expanding access to credit and capital, launched Project REACh to help identify and reduce barriers to full and fair participation in the nation’s banking system and in the economy. In view of the huge number of credit invisibles, a workstream on alternative credit assessment was created that promoted partnerships to facilitate entry into mainstream financial services for credit invisibles and economically disadvantaged individuals, communities, and small businesses. The workstream has been able to convene financial institutions, technology firms, and community organizations in pursuit of two goals: (1) exploring, developing, and analyzing credit assessment methods that can consistently predict a borrower’s ability to repay, and (2) to collaborate on an alternative credit assessment method that integrates traditional credit bureau data and existing permissioned deposit account data shared across participating financial institutions to extend credit to people who previously lacked access to mainstream credit. • Legislation-led pilot program Following the ICCR Policy Guide on Alternative Data, the National Credit Regulator (NCR) in South Africa, with the support of the The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 35 Regulatory and policy environment Fintech in Indonesia is regulated by two entities: the Bank of Indonesia, which represents the central bank and is responsible for the regulation of payment-based innovations, and the Financial Services Authority (Otoritas Jasa Keuangan (OJK)), which is responsible for fintech activities related to lending and all other aspects of financial technology. On August 15, 2018, OJK issued Regulation 13, Digital Financial Innovation (Inovasi Keuangan Digital (IKD)),72 to regulate information technology–based services in the financial sector. The term “digital financial innovation” encompasses numerous facets of business, including a business process, a business model, and financial instruments that provide new and additional value in the financial services sector by involving a digital ecosystem. Such added value often derives from multiple financial services, including collection and distribution of funds, transaction settlement, capital gathering, and more. Furthermore, OJK put forth some additional specific criteria for IKD companies or institutions wishing to participate in the sandbox. Among the main criteria was that their business model should be based on ICT innovations and focused on the financial sector—with the added conditions of being collaborative and promoting financial literacy and inclusion; showing signs of being widely used and useful; offering the possibility of a smooth integration into existing financial systems; and complying with consumer data privacy and protection regulations. The IKD candidates should not manage a portfolio on their own but should provide a platform to facilitate financial services. Upon registration into the regulatory sandbox, the IKD participant is assessed over one year, with a potential extension of six additional months. If the IKD company is assessed and evaluated as “recommended,” its registration application must be submitted within six months or its approved status will be revoked. OJK enforces stringent penalties on any IKD company that breaches its obligations to uphold the confidentiality, integrity, and availability of the personal data, transaction data, and financial data entrusted to and managed by the IKD company. Box 9: Indonesia’s Regulatory Sandbox: OJKR No. 13/POJK.02/2018 on Digital Financial Innovation (Inovasi Keuangan Digital (IKD)) 4.2.3. Intersections of AI governance and alternative data in credit scoring regulation The developments in AI governance and regulation in recent years, including initiatives such as the launch of national AI strategies, the introduction of regulatory frameworks like the AI Act in Europe, and the enactment of AI regulations in China, are closely tied to the use of alternative data in credit scoring. The connection between this context and the use of alternative data in credit scoring lies in the broader regulatory framework governing artificial intelligence (AI), which often encompasses the use of alternative data in credit assessment processes. For instance, the emphasis on transparency and accountability in AI algorithms, as seen in China’s AI regulation requiring accessible explanations for AI algorithms, directly intersects with the responsible use of alternative data in credit risk assessment processes. Similarly, initiatives like the Fairness, Ethics, Accountability, and Transparency (FEAT) principles in Singapore, aimed at guiding the responsible use of AI in the finance sector, likely span considerations related to the incorporation of alternative data into credit scoring models. The EU’s Artificial Intelligence Act (AIA) established a direct link between AI and use of alternative data in credit scoring by classifying both AI-based credit scoring and AI-supported assessment of creditworthiness as a “high-risk application.”73 Furthermore, the AI Act complements existing EU law by introducing requirements aimed at minimizing the risk of algorithmic discrimination. This includes ensuring the quality and fairness of the data sets used, which is particularly relevant in the context of alternative data sources that may introduce biases or inaccuracies into credit scoring models. Additionally, AIA emphasizes transparency and human oversight in AI systems, recognizing that certain AI algorithms, especially those using alternative data, can be complex and opaque. By mandating transparency measures, the AIA aims to ensure that regulators, users, and consumers can understand and interpret the output of AI-based credit scoring models adequately, consequently promoting accountability and trust in automated credit scoring. Launched in late 2018, the Veritas initiative is an integral part of Singapore’s National AI strategy. This endeavor, led by the Monetary Authority of Singapore (MAS), seeks to enable financial institutions to evaluate their artificial intelligence and data analytics (AIDA) driven solutions against the FEAT Principles (launched earlier that year)—by providing them with the necessary practical tools and a collaborative environment for sharing best practices. It is a multiphase, collaborative project with the financial industry, and having evolved into a consortium, Veritas had 27 participants in its second phase, including members such as Amazon, Microsoft, Google, Goldman Sachs, Accenture, Union Bank of the Philippines, and HSBC.74 In its first phase, the convened consortium focused on developing fairness metrics and an assessment methodology for two banking use cases: (1) credit risk scoring, and (2) customer (individuals and/or firms) marketing. The second phase focused on developing assessment methodology for insurance predictive underwriting, customer marketing, and insurance fraud detection, as well as transparency assessment methodology for credit risk scoring and customer marketing. At the end of each phase, several white papers were published documenting the findings and bringing forward perspectives for the financial industry to consider when putting AIDA concepts and systems into practice. The Veritas initiative presents one of the very few efforts undertaken by a regulator to design a framework to operationalize AI ethical principles for the responsible use of artificial intelligence across data analytics specifically in the financial sector. Box 10: The Veritas Consortium of Singapore 36 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges Regulatory and policy environment 4.3 Open banking: Key implications and considerations In the context of open banking, competition extends beyond traditional banks and fintech companies to include competition among traditional players. This dynamic is particularly evident in markets where banks are at the forefront of digital innovation. As consumers explore options such as switching or refinancing loans, traditional banks are prompted to enhance their digital infrastructure to remain competitive. Consequently, open banking acts as a catalyst for ongoing innovation within the financial sector, encouraging traditional players to embrace technological advancements and enhance their services to meet evolving consumer needs in a competitive landscape. Open banking gives lenders and other stakeholders the opportunity to access detailed data from conventional and alternative sources. While its adoption is experiencing rapid growth, albeit unevenly, in regions like Europe and in other data-mature markets such as Singapore, the true potential for leveraging alternative credit scoring exists within the framework of open banking itself, as highlighted by several interviewees. Despite offering significant advantages, however, the open banking framework also raises several concerns. The extensive sharing of third-party data introduces multiple potential points of failure at which customer data may be compromised, stolen, or misused. Consequently, open banking presents a range of challenges for both system participants and regulators in defining liability and accountability for data breaches, as well as in determining who is responsible for the associated costs of customer remedies. Regulators face the challenge of addressing “information asymmetry” among participants in the open banking system. The principle of reciprocity is particularly significant, as its failure to be addressed may result in an imbalance that allows one party to gain advantages over the other. This imbalance has been characterized as an “unfair and regulatory-driven competitive disadvantage.”75 Indeed, the pool of participants and the rules governing data-sharing may create a nuanced power balance where traditional financial institutions, which are often data-rich but information-poor (DRIP), are required to share customers’ financial data with data-rich big tech companies; this dynamic could potentially allow big tech firms to dominate key segments of financial services. Credit bureaus swiftly acknowledged the potential benefits of open banking. For instance, as part of its Open Banking strategy, Experian partnered in 2020 with Invers, a leading bank data aggregator in the Benelux region. Together, Experian and Invers developed a technology platform to optimize credit granting decisions by leveraging applicants’ transactional banking history.76 Similarly, in 2021, Equifax acquired AccountScore, a transaction data analytics company, to strengthen its own Open Banking strategy. The acquisition aimed to enhance Equifax’s consumer and commercial product offerings by combining traditional credit bureau information with bank transaction data facilitated by AccountScore.77 The increasing integration of digital tools and automation into daily life has generated diverse footprints reflecting financial activities ranging from mortgages and taxes to pensions, insurance, and for MSMEs, accounting, logistics, and trade data. By collectively aggregating and analyzing these footprints, lenders gain a comprehensive understanding of borrowers’ financial well-being, facilitating informed credit decision-making. This concept, known as open finance, holds significant promise. A senior executive at a leading global information services company suggests that the true potential of open finance lies in the evolution toward open data. As alternative data types continue to proliferate and diversify, he observes, “alternative data essentially embodies the principles of open data.” After the launch of the European Union’s Payment Services Directive Two (PSD2) in 2018, several countries around the world aligned behind the open banking movement; below are examples of initiatives taken by various countries to establish their own open banking regimes: • North America: In July 2021, President Biden signed an Executive Order titled Promoting Competition in the American Economy, encouraging the Consumer Financial Protection Bureau (CFPB) to consider crafting rules under section 1033 of the Dodd-Frank Act to support open banking with the goal of allowing consumers to easily switch financial institutions and use new and innovative financial products.78 In October 2023, the Consumer Financial Protection Bureau (CFPB) proposed the Personal Financial Data rule aimed at accelerating the shift toward open banking. The rule is expected to be finalized by fall 2024.79 • Australia: The country introduced consumer data rights legislation using a regime similar to open banking but more broadly designed to empower individuals to control their own data. The multiphase implementation started with all major banks and the energy sector, with the telecommunications sector set to follow. • Latin America: The Central Bank of Brazil approved an open banking project in 2019 and officially launched it in 2021. The four-phase model allows consumers to regain control of their data and includes a reciprocity principle under which all institutions receiving data must also share data. As of December 2023, there were 42 million active consents connecting accounts to open finance.80 In Chile, the Financial Portability Law, passed in 2020, took the first steps toward open banking by allowing individuals and companies to freely change their financial service providers. In August 2021, the Chilean Ministry of Finance published its Guidelines for Developing an Open Finance Framework in Chile, with a Focus on Competition and Financial Inclusion.81 Mexico has been in the vanguard of the movement in Latin America, introducing a regulatory regime for open banking in its 2018 Fintech Law. Implementation of this regime remains to be completed, with further regulations anticipated to be finalized in 2024 and potentially extending into 2025. The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 37 Regulatory and policy environment • APAC: In 2020, Singapore launched the Singapore Financial Data Exchange (SGFinDex) as the first public-private open banking collaboration in the world to enable individuals to plan their finances by retrieving their financial data from participating banks as well as from several government agencies and entities.82 In India, the Account Aggregator (AA) framework provides the country’s entry into open banking. First announced in 2021, the framework enables data flows and manages consent for financial data-sharing among a group of entities regulated by the Reserve Bank of India (RBI), subject to explicit permission from the customer. • Africa: As data privacy laws are progressively rolled out in Africa, the open banking movement is set to gain momentum or at the very least to occupy the center of the finance industry debate in the coming years. In Kenya, the Central Bank’s 2021–2025 Vision and Strategy document set the agenda for the future of the country’s digital payments ecosystems, explicitly mentioning the use of open banking as a strategic objective.83 In March 2023, the Central Bank of Nigeria (CBN) issued operational guidelines detailing the processes and standards for Nigeria’s open banking ecosystem.84 In South Africa, while there is no open banking regime, Nedbank became the first African bank to open its API platform and marketplace to fintechs that meet its technical standards for open banking.85 Box 11: Global Open Banking Initiatives and Interventions 38 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges Regulatory and policy environment The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 39 5. Navigating the path ahead: Recommendations and policy considerations 5.1 Recommendations Notably, certain credit bureaus and alternative credit scoring entities currently develop comprehensive country-specific Building upon the recommendations from ICCR’s Guidance profiles of acceptable data, prioritizing sources that have Note, “Use of Alternative Data to Enhance Credit Reporting to demonstrated high predictive validity. For instance, telecom Enable Access to Digital Financial Services by Individuals and (telco) and utilities data are often considered exemplary SMEs operating in the Informal Economy” (ICCR 2018), this alternative data sources due to their third-party verification, research further supplements those findings while introducing effectiveness in reducing synthetic identity fraud, and ability to a comprehensive framework intended to guide the responsible provide evidence of consistent and punctual payments. use of alternative data in credit risk assessment. As the landscape evolves, particularly in the post-pandemic era, the integration Recommendation 3: Leverage regulatory innovation of alternative data sources has become increasingly prevalent. platforms to promote experimentation and testing. It becomes essential, therefore, to establish principles that prioritize consumer protection, fairness, and transparency. In the absence of an existing legal framework that enables the use of alternative data, regulators should consider Recommendation 1: Implement robust legal and regulatory using innovation platforms such as sandboxes to test the frameworks. predictiveness of alternative data within the country context. The use of innovation platforms will provide an avenue to foster Regulators should consider enacting robust legal and regulatory responsible testing of alternative data use and prove its ability frameworks that promote the sharing and use of alternative to solve some existing challenges, such as promoting access for data with the necessary consumer protection and cybersecurity the unserved and underserved communities. safeguards. The legal and regulatory reforms will provide legal clarity that will enable the leveraging of alternative data. Most of the countries that have established sandboxes have Although there is no one-size-fits-all approach to legislation, incorporated mandates geared toward promoting financial some key practices include: inclusion. The use of alternative data can be one of the instruments that can promote credit inclusion for unserved and 1. Integration into the credit reporting legislation and/or underserved segments of the population. The alternative data regulations. tests should help in surfacing data that is relevant in the country 2. Leveraging the open banking / finance legislation. but also in addressing potential risks and challenges such as 3. Integration into digital lending or consumer lending data accuracy, biases and discrimination, and predictiveness. directives. Regulators without formal sandboxes can also consider other Recommendation 2: Implement a regulatory blacklist for policy instruments or tools, such as the no-action letter (NAL) alternative data in credit scoring. or the no-objective letter and controlled pilots, to promote responsible use of alternative data. Such initiatives can help Considering the variability in data quality, stability, accessibility, regulators understand innovations, provide guidance to the predictive power, nuances, and associated risks across different whole market, and shape regulatory reforms using evidence- sources, the accuracy of algorithmic credit scoring relies heavily based insights. on the data fed into the algorithm. To mitigate these risks in a structured manner, regulatory bodies should consider Recommendation 4: Promote the adoption of informed implementing a formalized “data blacklist” for alternative data consumer-permissioned secure data-sharing protocols that usage. This blacklist would delineate data elements prohibited enable effective sharing of alternative data at scale. from inclusion in credit scoring algorithms. The concept of a blacklist is not new in credit reporting, as traditional credit Some alternative data is difficult to mainstream through credit data also includes protected classes. For example, the US Equal reporting services providers because of its nature, structure, and Credit Opportunity Act (ECOA) and Fair Credit Reporting Act form. Regulators must therefore accelerate informed consumer- (FCRA) both prohibit collection of gender and race data. permissioned data-sharing frameworks such as open banking / 40 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges Navigating the path ahead: Recommendations and policy considerations data frameworks. These frameworks should provide solutions • Security, Robustness, and Resilience: Prioritize high that are frictionless to data providers and consumers while standards of security to safeguard customer data, especially ensuring that consumers can maintain ownership and control considering the sensitive nature of alternative data. of their data. • Lawfulness: Ensure compliance with existing laws and adherence to professional standards and ethical guidelines, Regulators have introduced legal and regulatory frameworks including those specific to the collection and use of in several markets, confirming the general awareness of the alternative data. importance of openness. In other markets, the industry has • Fitness for Purpose: Aim to address and improve existing led in innovation, resulting in the prominence of consumer- challenges in credit scoring by leveraging alternative data permissioned data-sharing solutions. The ability of such sources effectively and responsibly. initiatives to scale depends on regulatory guidance on their use. • Sustainability and Well-Being: Promote sustainable development principles across social, economic, and Recommendation 5: Advance gender equity in alternative environmental dimensions, considering the long-term credit scoring through inclusive data practices. implications of alternative data use. • Diversity and Inclusion: Foster inclusivity by involving Women may face disadvantages in the use of alternative data diverse data sets and teams in model development to in credit scoring. While alternative data can assist in addressing optimize accessibility and reflect the diversity of borrowers the challenge of providing access to credit to women in certain in credit models, including those whose credit profiles may regions or socioeconomic groups who may be underrepresented benefit from alternative data. in traditional credit data, the use of alternative data sources can also inadvertently perpetuate societal biases. Lack of Recommendation 7: Adopt a risk-based approach to the gender-specific data points in alternative sources may lead to a collection and use of alternative data for creditworthiness misrepresentation of women’s credit risk, potentially resulting by lenders. in less favorable credit terms or denial of credit opportunities. To address these challenges, policy makers and industry Regulators should ensure that lending institutions have players can adopt several inclusive approaches when designing adequate data policies that govern the consent management, alternative scoring models. Financial institutions and credit collection, processing, correction, storage of alternative data, bureaus might, for example, collect and use sex-disaggregated and reliance on third-party data providers. The policies should data—in an aggregated and anonymized form to comply with also cover the responsible application of machine learning and existing regulations—to inform their models designs; they might AI in credit scoring, given that most credit scoring models will rely also offer products tailored to better reflect the unique financial on advanced AI techniques and machine learning algorithms. behaviors and creditworthiness factors of certain borrowers, e.g., women. The use of sex-disaggregated alternative data is expected to promote access, as more women than men tend to 5.2 Policy considerations be credit invisible or thin-file customers. In addition to recommendations, policy makers should consider Recommendation 6: Establish a comprehensive industry implementing policies and interventions that contribute to the code governing the use of alternative data in credit scoring. generation of alternative data and the harnessing of its potential in creditworthiness assessment. In the absence of legal and regulatory guidance, industry participants should consider self-regulation guided by a set of Policy Consideration 1: Incentivizing digitization of economic comprehensive principles to ensure responsible and ethical activities use of alternative data in credit scoring. These principles should prioritize consumer protection and industry integrity The study revealed that more digitized economies far to address the diverse risks associated with alternative data outperformed less digitized economies in innovative use of implementation. Responsible automated credit scoring, alternative data. To fully capitalize on the benefits of alternative incorporating alternative data sources, should adhere to the credit scoring, it’s important to spur digital transformation of following principles: MSMEs and individuals. The ability of prospective borrowers to consistently produce a steady stream of digital footprints • Ethics: Uphold fundamental human rights and avoid is important for creating pools of alternative data that can be discrimination based on protected characteristics, ensuring leveraged for credit underwriting. The pandemic-induced individual autonomy is preserved. acceleration of digital adoption enabled digitization of • Accountability: Hold participants accountable for internally significant parts of the economy, especially the financial and externally sourced technologies, with measures in place services. Notwithstanding this progress, several industries and to address errors and disruptions affecting consumers. sectors remain underdigitized, partly due to cost and business • Transparency and Trust: Foster trust through clear and cultural practices. communication about the technologies used, their implementation methods, and their impact on outcomes, By promoting awareness of the commercial benefits of digital emphasizing informed consent for the use of alternative transformation and its potential to enhance credit access, data. MSMEs can be incentivized to embrace digitalization. This The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 41 Navigating the path ahead: Recommendations and policy considerations transition not only positions MSMEs to generate a wealth of Policy Consideration 4: Supporting Digital Literacy and alternative data but also improves their business prospects. Consumer Awareness To the extent possible, policy makers can also promote financial and nonfinancial incentives to promote digitization. Policy makers should promote literacy and awareness as means Ensuring equitable access to infrastructure (including internet of promoting responsible use of alternative data. As the level of connectivity and digital devices, particularly in underserved digitization increases, it brings with it new risks, such as cyber communities), can also spur digitization. and privacy risks, that can impact consumers and ultimately affect confidence. In addition, literacy is also important to Policy Consideration 2: Digitizing government services and increase awareness of and trust in digital platforms and to help making the data readily accessible reduce the digital divide. To foster financial inclusion for both individuals and MSMEs, it is important to prioritize investments Digitization of government services has the potential to unlock in financial literacy. alternative data usage. Government services often reach broad sectors of the economy, hence attaining higher coverage, yet Policy Consideration 5: Promoting cross-border collaboration they are often manual or not fully digitized. Even in markets with significant levels of digitization, most public databases are not To increase regional integration and voluntary/involuntary readily available to the private sector. migration, policy makers must implement measures that promote cross-border data flows. To further promote cross- Permitting responsible access to public databases such as border movement, policy makers should collaborate with company registries, vehicle registries, tax registries, collateral regional peers to standardize and harmonize regulations, data registries, and court records can significantly augment the data, dictionaries, attributes, and transfer protocols, among other help increase visibility, and enhance scoring of marginalized initiatives. This can help MSMEs and individuals access credit in segments. Opening and use of government data sources can markets in which they might not have a traditional credit history. also enhance market confidence. Through collaboration and engagements with peers, challenges associated with regulatory practices could be discovered and Policy Consideration 3: Supporting infrastructure unified practices could be broadly enacted. development Policy makers should implement policies that enable the development of robust digital public infrastructure (DPIs). The three foundational elements of DPIs—identity, payments, and data exchange—are essential to generate and spur alternative data usage. At the center of information sharing and exchange is the ability to uniquely identify a data subject. This is more relevant with alternative data, where it is sometimes difficult to pinpoint the owner of the digital trail. For example, a digital trail of utility and mobile phone transaction records might be attributable to the registered owner, but the owner might not necessarily be the user. Data from payment platforms has emerged as one of the most reliable forms of alternative data for enhancing credit underwriting. As a result, policy makers should consider implementing policies and offering financial and nonfinancial incentives to promote the use of digital payment platforms. Beyond the core data exchange interventions, policy makers should consider enabling other ecosystem components, such as APIs and e-consent infrastructure, that can enable cost- effective, frictionless, and secure transmission of alternative data. APIs can enable transmission of large volumes of real- time data exchange within markets and across borders. API frameworks can provide guidelines, including the necessary safeguards, that can facilitate the development and wider adoption of APIs by the market. To reduce the level of effort and cost associated with connecting to various data sources, policy makers should also strive to ensure that these APIs are interoperable. Responsible scaling of alternative use relies on an effective consent management framework than can assist data subjects to understand how their alternative data are shared. 42 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges Navigating the Path Ahead The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 43 References & Annotations 1. ICCR (2018), “Use of Alternative Data to Enhance Credit Reporting to Enable Access to Digital Financial Services by Individuals and SMEs Operating in the Informal Economy,” guidance note, https://www.gpfi.org/sites/gpfi/files/documents/Use_of_Alter- native_Data_to_Enhance_Credit_Reporting_to_Enable_Access_to_Digital_Financial_Services_ICCR.pdf, 14–15. 2. PYMNTS (2022), “New 11-Country Study Shows Digital Transformation Has Reached Only 27% of Full Potential,”https://www. pymnts.com/connectedeconomy/2022/new-11-country-study-shows-digital-transformation-has-reached-only-27-of-full-po- tential/. 3. Findex (2021), Global Findex Database, https://www.worldbank.org/en/publication/globalfindex/Report, 42. 4. World Economic Forum (2022), “Accelerating Digital Payments in Latin America and the Caribbean,” https://www3.weforum. org/docs/WEF_Accelerating_Digital_Payments_in_Latin_America_and_the_Caribbean_2022.pdf. 5. World Bank (2016), “Innovation in Electronic Payment Adoption: The Case of Small Retailers,” https://documents1.worldbank. org/curated/en/765851467037506667/pdf/106633-WP-PUBLIC-Innovative-Solutions-Accelerate-Adoption-Electronic-Pay- ments-Merchants-report-2016.pdf. 6. V. V. Korovkin (2019), “National Digital Economy Strategies: A Survey of Africa,” Orfonline.org, https://www.orfonline.org/ wp-content/uploads/2019/07/ORF_IssueBrief_303_DigitalEcon-Africa.pdf. 7. Gov.cy (2020), Digital Cyprus 2025, https://www.dmrid.gov.cy/dmrid/research.nsf/all/927EA351714F99EDC22587CE- 0028C090/$file/Digital%20Strategy%202020-2025.pdf?openelement. 8. ICCR (2018), 14–15. 9. Sahamati (2021), “Empowering Alternate Data Collection For Digital Lending with Account Aggregator Framework- Introduc- tion,” blog, https://sahamati.org.in/empowering-alternate-data-collection-for-digital-lending-with-account-aggregator-frame- work-introduction/. 10. Oliver Wyman (2017), “Alternative Data and the Unbanked,” https://icic.org/wp-content/uploads/2022/11/Oliver_Wyman_Al- ternative_Data.pdf. 11. FICO blog (2017), “Using Alternative Data in Credit Risk Modelling,” https://www.fico.com/blogs/using-alternative-data-cred- it-risk-modelling. 12. Urban Institute (2021), Utility, Telecommunications, and Rental Data in Underwriting Credit, https://www.urban.org/sites/de- fault/files/2022-06/utility-telecommunications-and-rental-data-in-underwriting-credit_0.pdf. 13. NCLC (2022), “No Silver Bullet: Using Alternative Data for Financial Inclusion and Racial Justice,” https://www.nclc.org/wp-con- tent/uploads/2022/08/IB_Alt_Data_Is_No_Silver_Bullet.pdf. 14. ICCR (2018). 15. CFPB (2019), “Interagency Statement on the Use of Alternative Data in Credit Underwriting,” https://files.consumerfinance. gov/f/documents/cfpb_interagency-statement_alternative-data.pdf. 16. ICCR (2017), “Policy Brief: Credit Reporting Systems Contribution to Financial Inclusion” (March 2017). 17. ICCR (2018), 13. 18. The study defines fintech credit as “credit facilitated by electronic platforms that are not operated by commercial banks.” 19. Bank for International Settlements (2018), “Fintech Credit Markets Around the World: Size, Drivers, and Policy Issues,” https:// www.bis.org/publ/qtrpdf/r_qt1809e.pdf. 20. Inter-American Development Bank (2022), Fintech in Latin America and the Caribbean: A Consolidated Ecosystem for Recov- ery, https://publications.iadb.org/publications/english/document/Fintech-in-Latin-America-and-the-Caribbean-A-Consolidat- ed-Ecosystem-for-Recovery.pdf. 21. Nova Credit (2024), “The State of Alternative Data in Lending,” https://go.novacredit.com/state-of-alt-data-in-lending-report. 22. Levine, Ross & Beck, Thorsten & Demirguc-Kunt, Asli. (2007), “Finance, Inequality and the Poor. Journal of Economic Growth”. 12. 27-49. 10.1007/s10887-007-9010-6. 23. Barajas, Adolfo & Beck, Thorsten & Belhaj, Mohammed & Ben Naceur, Sami. (2020). Financial Inclusion: What Have We Learned So Far? What Do We Have to Learn?. IMF Working Papers. 20. 10.5089/9781513553009.001. 24. M. Ayyagari, A. Demirgüç-Kunt and V. Maksimovic (2011), “Small vs. young firms across the world: Contribution to employment, job creation, and growth, World Bank Policy Research Working Paper No. 5631, 2011. 25. IMF (2021), “Is Digital Financial Inclusion Unlocking Growth?,” https://www.elibrary.imf.org/view/jour- nals/001/2021/167/001.2021.issue-167-en.xml?rskey=nxGX3j&result=103. 26. SME Finance Forum, https://www.smefinanceforum.org/data-sites/msme-finance-gap. 27. CGAP (2024), Leveraging Transactional Data for Micro and Small Enterprise Lending, https://www.cgap.org/research/publica- tion/leveraging-transactional-data-for-micro-and-small-enterprise-lending. 28. Ibid. 44 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges Bibliography 29. Findex (2021),33. 30. Ibid., 89. 31. Agarwal, T. K., C. Minoiu, A. F. Presbitero, and A. F. Silva (2018), “Financial Inclusion Under the Microscope”, IMF Working Paper 18/208, https://www.elibrary.imf.org/display/book/9798400200038/CH007.xml. 32. de Janvry, A., C. McIntosh, E. Sadoulet. (2010) “The Supply- and Demand-Side Impacts of Credit Market Information”. Journal of Development Economics 93(2): 173–188. https://www.sciencedirect.com/science/article/pii/S0304387809000984 33. FICO (2015), “Can Alternative Data Score More Consumers?,” FICO blog, https://www.fico.com/blogs/can-alternative-da- ta-score-more-consumers. 34. FICO (2023), “Using Alternative Data in Credit Risk Modelling,” https://www.fico.com/blogs/using-alternative-data-cred- it-risk-modelling. 35. Here, the percentage increase refers to the increase in the value of the Gini index. 36. GAO (2021), “Mortgage Lending: Use of Alternative Data Is Limited but Has Potential Benefits,” https://www.gao.gov/assets/ gao-22-104380.pdf. 37. Study presented at ACCIS Webinar, Informative Session on Alternative Data, October 26, 2021. 38. TransUnion (2022), “Empowering Credit Inclusion: A Deeper Perspective on Credit Underserved and Unserved Customers,” https://www.transunion.com/lp/empowering-credit-inclusion-new-to-credit. 39. GSMA (2021), Connected Women: The Mobile Gender Gap Report 2021, https://www.gsma.com/r/wp-content/uploads/2021/06/ The-Mobile-Gender-Gap-Report-2021.pdf. 40. N. Sharma, S. Chakrabarti, and S. Grover (2016), “Gender Differences in Caregiving Among Family—Caregivers of People with Mental Illnesses,” World Journal of Psychiatry 6 (1): 7–17, March 22, 2016, doi: 10.5498/wjp.v6.i1.7, PMID: 27014594, PMCID: PMC4804270. 41. Scott M. Stringer, Office of the New York City Comptroller (2014), “The Growing Gap: New York City’s Housing Affordability Chal- lenge,” https://comptroller.nyc.gov/wp-content/uploads/documents/Growing_Gap.pdf. 42. Scott M. Stringer, New York City Comptroller (2017), “Making Rent Count: How NYC Tenants can Lift Credit Scores and Save Mon- ey,” https://comptroller.nyc.gov/reports/making-rent-count/rent-and-credit-report/#_edn1. 43. For a description of the metric used to make these assessments, see Experian, “VantageScore 3.0: Improved Predictive Pow- er Raises the Standard of Performance,” https://www.experian.com/assets/consumer-information/product-sheets/vantages- core-3.pdf. In brief, VantageScore 3.0 ranges are as follows: Super prime: 781–850 Prime: 661–780 Near prime: 601–660 Subprime: 500–600 Deep subprime: 300–499 44. TransUnion (2021), “Alternative Data Such as Rent Payment Reporting Bridges the Gap for Unscorable Consumers and In- creases Financial Inclusion Opportunities,” https://newsroom.transunion.com/alternative-data-such-as-rent-payment-report- ing-bridges-the-gap-for-unscorable-consumers-and-increases-financial-inclusion-opportunities/. 45. GSMA (2022), State of the Industry Report on Mobile Money, https ://www.gsma.com/sotir/wp-content/uploads/2022/03/ GSMA_State_of_the_Industry_2022_English.pdf. 46. World Bank Group (2022), “Migration and Development Brief 37,” https://www.knomad.org/publication/migration-and-devel- opment-brief-37 47. IMF (n.d.), “What are remittances?,” https://www.imf.org/external/pubs/ft/fandd/basics/pdf/ratha-remittances.pdf. 48. Global Partnership for Financial Inclusion (2017), Alternative Data Transforming SME Finance, https://www.gpfi.org/sites/gpfi/ files/documents/GPFI%20Report%20Alternative%20Data%20Transforming%20SME%20Finance.pdf. 49. Experian (2024), news release, https://www.experianplc.com/content/dam/marketing/global/plc/en/assets/documents/re- sults-and-presentations/2024/Experianf-full-year-FY24-results-announcement.pdf. 50. Kolmogorov-Smirnov (K-S) value is a common metric used to assess the performance of a credit scoring model. Specifically, it helps evaluate how well the model differentiates between good and bad credit risk profiles. 51. TransUnion Credit Vision https://www.transunion.co.za/content/dam/transunion/za/business/collateral/infographic/creditvi- sion-link-infographic.pdf 52. Experian (2021), “Experian Teams Up with FinScore to Uplift Financial Inclusion Through Alternative Data Scoring in the Phil- ippines,” https://www.experianplc.com/media/latest-news/2021/experian-teams-up-with-finscore-to-uplift-financial-inclu- sion-through-alternative-data-scoring-in-the-philippines/. 53. Experian (2022), “Experian Agrees to Acquire Majority Stake in Mova,” press release, https://www.experianplc.com/media/lat- est-news/2022/experian-agrees-to-acquire-majority-stake-in-mova/. 54. The Economic Times (2022), “CreditVidya and Flipkart Wholesale Collaborate to Offer ‘Buy Now Pay Later’ Credit to Merchants,” https://economictimes.indiatimes.com/tech/technology/creditvidya-and-flipkart-wholesale-collaborate-to-offer-buy-now- pay-later-credit-to-merchants/articleshow/90411581.cms. 55. Uber (2021), “Préstamos exclusivos y beneficios con MiBanco,” blog, https://www.uber.com/es-PE/blog/prestamos-exclu- sivos-cuenta-de-ahorros-y-beneficios-con-mibanco/. 56. Experian (2020), “What Is UltraFICO and How Do I Use It?,” blog, https://www.experian.com/blogs/ask-experian/what-is-ultraf- The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 45 Bibliography ico/. 57. Experian (2021), “The Impact of COVID-19 on Business,” The State of Business Credit in South Africa, webinar held January 20, 2021. 58. Creditinfo (2020), “Players in Baltic Markets (Latvia, Estonia) and Iceland Measure Risk Better, Benefitting from COVID-19 Impact Score Developed by Creditinfo,” blog, https://chronicle.creditinfo.com/2020/09/15/players-in-baltic-markets-latvia-estonia- and-iceland-measure-risk-better-benefitting-from-covid-19-impact-score-developed-by-creditinfo/. 59. Fina (2020), “Methodology for Calculating the COVID Score,” in Croatian, https://www.fina.hr/-/metodologija-za-izra- cun-covid-score-a. 60. CFPB (2020), “Consumer Reporting and the CARES Act,” blog, https://www.consumerfinance.gov/about-us/blog/consumer-re- porting-and-cares-act/. 61. FICO (2020), “FICO Resilience Index Now Available for Lenders to Pilot,” blog, https://www.fico.com/blogs/fico-resilience-in- dex-now-available-lenders-pilot. 62. M. Aronowitz, E. Golding, and J. Choi (2020), “The Unequal Costs of Black Homeownership 1,” https://gcfp.mit.edu/wp-content/ uploads/2020/10/Mortgage-Cost-for-Black-Homeowners-10.1.pdf. 63. People’s Bank of China (2021), Order No. 4 [2021] of the People’s Bank of China (Measures for the Administration of Credit Re- porting Services); http://www.pbc.gov.cn/en/3688253/3689009/4180845/4393542/index.html. 64. European Commission (2023), Directive (EU) 2023/2225 of the European Parliament and of the Council, https://eur-lex.europa. eu/legal-content/EN/TXT/PDF/?uri=CELEX:32023L2225. 65. European Data Protection Supervisor (EDPS; 2021), Opinion 11/2021 on the Proposal for a Directive on Consumer Credits, https://edps.europa.eu/system/files/2021-08/opinion_consumercredit-final_en.pdf. 66. Reserve Bank of India (2022), “Digital Lending Guidelines,” https://www.rbi.org.in/scripts/NotificationUser.aspx?Id=12382&- Mode=0#AN1. 67. Bank of Thailand (2020), “The BOT Issued the Circular for the Undertaking of Digital Personal Loan Business,” press release, https://www.bot.or.th/English/PressandSpeeches/Press/2020/Pages/n6463.aspx. 68. Consumer Financial Protection Bureau (2020), Application of Upstart Network, Inc. for CFPB No-Action Letter, https://files.consumerfinance.gov/f/documents/cfpb_upstart-network-inc_no-action-letter-application_2020-11.pdf. 69. Interagency Statement on the Use of Alternative Data in Credit Underwriting, news release, https://www.occ.gov/news-issuanc- es/news-releases/2019/nr-ia-2019-142a.pdf. 70. https://disclosures.ifc.org/project-detail/AS/605940/south-africa-credit-reporting-financial-inclusion-project-phase-1. 71. Intergovernmental Fintech Working Group (IFWG; 2022), First Regulatory Sandbox Report, https://www.resbank.co.za/content/ dam/sarb/publications/media-releases/2022/fintech-iwfg-sandbox-report/IFWG%20First%20Regulatory%20Sandbox%20Re- port%20October%202022.pdf. 72. OJK (2018), Digital Finance Innovation in the Financial Services Sector, summary of OJK regulation, https://www.ojk.go.id/iru/ BE/uploads/regulation/files/file_35a2beef-e4c8-4ca8-8f42-4cfc0f180246-07092022161630.pdf. 73. https://artificialintelligenceact.eu/. 74. Monetary Authority of Singapore (MAS; 2021), Veritas Initiative, https://www.mas.gov.sg/schemes-and-initiatives/veritas. 75. Deloitte (n.d.), “Open Banking Around the World: Towards a Cross-Industry Data Sharing Ecosystem,” https://www2.deloitte. com/tw/en/pages/financial-services/articles/open-banking-around-the-world.html. 76. Experian (2020), “Experian Partners with Invers to Launch Its New Open Banking Solution,” blog, https://www.experian.nl/ over-experian/2020/11/24/experian-partners-with-invers-to-launch-its-new-open-banking-solution/. 77. Equifax (2021), “Equifax Acquires Open Banking and Transaction Data Analytics Company AccountScore,” blog, https://www. equifax.co.uk/about-equifax/press-releases/en_gb/-/blog/equifax-acquires-open-banking-and-transaction-data-analyt - ics-company-accountscore/. 78. The White House (2021), Executive Order on Promoting Competition in the American Economy, https://www.whitehouse.gov/ briefing-room/presidential-actions/2021/07/09/executive-order-on-promoting-competition-in-the-american-economy/. 79. CFPB (2023), “Prepared Remarks of CFPB Director Rohit Chopra on the Proposed Personal Financial Data Rights Rule,” https:// www.consumerfinance.gov/about-us/newsroom/prepared-remarks-of-cfpb-director-rohit-chopra-on-the-proposed-personal- financial-data-rights-rule/. 80. https://dashboard.openfinancebrasil.org.br. 81. Chilean Ministry of Finance (2021), “Guidelines for the Development of an Open Finance Framework in Chile, with a Focus on Com- petition and Financial Inclusion,” https://www.hacienda.cl/noticias-y-eventos/noticias/ministerio-de-hacienda-anuncia-el-in- greso-del-proyecto-de-ley-de-innovacion/lineamientos-para-el-desarrollo-de-un-marco-de-finanzas-abiertas-en-chile-con. 82. Monetary Authority of Singapore, https://www.mas.gov.sg/development/fintech/sgfindex. 83. Central Bank of Kenya (CBK; 2020), Kenya National Payments System Vision and Strategy 2021–2025, https://www.centralbank. go.ke/wp-content/uploads/2020/12/CBK-NPS-Vision-and-Strategy.pdf. 84. Central Bank of Nigeria (CBN) (2023), Central Bank of Nigeria Operational Guidelines for Open Banking in Nigeria, https://www. cbn.gov.ng/Out/2023/CCD/Operational%20Guidelines%20for%20Open%20Banking%20in%20Nigeria.pdf. 85. Nedbank, https://apim.nedbank.co.za/nedbank-global-banking-award. 46 The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges Bibliography The Use of Alternative Data in Credit Risk Assessment: The Opportunities, Risks, and Challenges 47