Research & Policy Briefs From the World Bank Chile Center, Malaysia Hub, and Seoul Center No. 26, Nov 2019 Using Big Data to Expand Financial Services: Benefits and Risks Facundo Abraham Sergio L. Schmukler José Tessada Big data is transforming financial services around the world. Advances in data analytics and computational power are allowing firms to exploit data in an easier, faster, and more reliable manner, and at a larger scale. By using big data, financial firms and new entrants from other sectors are able to provide more and better financial services. Governments are also exploring ways to use big data collected by the financial sector more systematically to get a better picture of the financial system as a whole and the overall economy. Despite its benefits, the wider use of big data has raised concerns related to consumer privacy, data security, discrimination, data accuracy, and competition. Hence, policy makers have started to regulate and monitor the use of big data by financial institutions and to think about how to use big data for the benefit of all. Big Data and the Financial Sector have immense potential to help financial institutions better understand the environment in which they operate and, as a result, The finance industry has always been driven by information. create new financial products and reach new segments of the Providing finance is the act of offering money in exchange for a population. But systematically using this huge body of data for promise that it will be returned at a later date. Because financial market analysis is very cumbersome. With millions of clients and institutions have imperfect knowledge about their customers, they transactions per day, financial institutions have to process and need to acquire substantial information to assess customers’ analyze an enormous amount of data. repayment capacity as accurately as possible. Therefore, collecting data on prospective borrowers is central to the provision of New methods to analyze data and higher computational power finance. These data typically include “hard data” (such as credit in the context of the so-called big data revolution are now enabling history, income, employment, education level, tax records, and financial sector participants to collect and analyze their data more financial statements) and “soft data” (such as opinions from loan easily and quickly than ever. For example, to map the use of officers, internal discussions, and economic prospects). Financial different financial services with the characteristics of customers institutions need information not only about their borrowers, but that use them (such as their income, residence, or educational about their lenders. Financial institutions lend and invest using level), financial institutions can employ association rules. They can money collected from third parties, which requires knowing their also use text mining tools to search for specific keywords in large, creditors’ investment preferences and risk tolerance (including complex documents such as internal reports and firms’ financial their financial goals, investment horizons, income, and future statements. expenditures). The big data revolution has also allowed financial institutions to Once clients are using financial services, financial institutions augment their own private data with public data and proprietary collect data from that use. For example, financial institutions data sold by other providers. For instance, financial institutions can document all transactions clients make to keep track of their mine data from social media profiles or public records to running balances. The amount of transactional data collected by supplement their information on clients. Similarly, they can track financial institutions has grown exponentially as economic activity news, tweets, blog posts, and other online publications to monitor is becoming less cash-based and transactions are increasingly market sentiment and predict market upturns and downturns. conducted through financial sector providers. Several financial activities such as personal and business banking, asset and wealth management, and insurance are already using big Financial institutions are also required to compile data to data (figure 1). submit to regulators, supervisors, and other institutions, such as credit bureaus. Financial institutions are unlike any other This brief discusses how big data is transforming the provision businesses in two ways. First, because they intermediate other of financial services. In particular, it discusses how financial peoples’ money, any losses end up being borne not only by the intermediaries use big data both to adapt services they offer to financiers but also possibly by the original creditors. Second, existing clients and to incorporate new clients into the financial financial institutions can pose systemic risk. A crisis in one system. The brief does not cover how big data is affecting other institution can lead to instability in the entire industry and the important aspects of the financial sector, such as algorithmic economy. For these reasons, the financial industry is tightly trading, the prevention of fraud, or artificial intelligence chatbots. regulated and supervised. As a consequence, financial institutions need to periodically prepare comprehensive data about their Potential Benefits of Big Data activities and associated risks and submit this information to the authorities. Traditionally, to know and understand the needs of their customers, financial institutions use personalized relationships of Financial institutions collect vast and detailed data on their clients with officers at a branch. However, this mode can have clients’ preferences, behavior, characteristics, and risks. These data several limitations, such as when clients are new or when officers Affiliation: Development Research Group, the World Bank; Pontificia Universidad Católica de Chile (PUC & Finance UC). E-mail addresses: fabraham@worldbank.org, sschmukler@worldbank.org, jtessada@gmail.com Acknowledgement: We thank Anushka Baloian for able research assistance. We received very useful comments from Ana Maria Aviles, Youjin Choi, Bob Cull, Sameer Goyal, Norman Loayza, and Rekha Reddy, and helpful edits from Nancy Morrison. The World Bank Chile Research and Development Center, the FCI group at the Malaysia Global Knowledge and Research Hub, the FCI group’s Seoul Center for Financial Sector Development partnership with the Korean Ministry of Economy and Finance, the Knowledge for Change Program (KCP), and the Research Support Budget (RSB) provided financial support for this brief. Objective and disclaimer: Research & Policy Briefs synthesize existing research and data to shed light on a useful and interesting question for policy debate. Research & Policy Briefs carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions are entirely those of the authors. They do not necessarily represent the views of the World Bank Group, its Executive Directors, or the governments they represent. Global Knowledge & Research Hub in Malaysia Using Big Data to Expand Financial Services: Benefits and Risks Other financial institutions are taking advantage of new data to Figure 1. Investments in Big Data by Financial Sector Industry, build alternative credit scores for low-income individuals. Examples 2018 include fintech companies like Lenddo and Tala, with presence in Lending & Africa, Asia, and Latin America. These companies typically offer an Financing, 8% app that individuals can download on their phone and use to apply Investment for loans. When they submit a loan application, the app scans the Banking & Personal & digital footprint of a mobile phone (contacts, social network Capital Markets, Business profile, geographical patterns) and instantaneously decides 11% Banking, 29% whether to extend the loan. Many of these alternative credit providers are located in East Asia, such as Ayannah (the Credit Cards & Philippines); CredoLab (Indonesia, Malaysia, and Singapore, among Payment other countries); and Kakao Bank (Korea) (Fintech News 2017; The Processing, 12% Economist 2019). Automated credit scoring not only can help expand credit, but also saves time and might reduce some of the human biases that can emerge in the evaluation process. Asset & Wealth Management, Big data can also promote access to finance for small and 13% Insurance medium enterprises (SMEs). Financial institutions have difficulties Services, 28% assessing the risk of SMEs because they may not have been in business long and public information about them is limited. Large A wide range of industries within the financial sector use big data. Source: SNS Telecom & IT (2018). online retailers, such as Alibaba, Amazon, eBay, and Mercado Libre, are using big data to gather and analyze information on firms participating in their platforms (such as sales and customer ratings) are relocated to another branch. Large banks, with millions of and then providing loans (The Economist 2015). Alibaba offers clients, can find it particularly difficult to maintain personalized credit to SMEs through its own digital bank called MyBank, which relations with clients. compiles data on monthly sales on Alibaba and then preapproves firms for loans. When a firm operating in Alibaba requires a loan, it Nowadays, financial institutions can leverage big data to offer simply completes an online application that is processed in only personalized services to millions of customers. For example, three minutes. Loans are proportionate to the volume of sales by Banorte, a Mexican bank with more than 13 million clients, is the firm, are unsecured, and carry lower rates than traditional studying clients’ banking behavior to know which products best banks. As of 2018, MyBank serviced roughly half of all SMEs in suit each client, create a personalized experience when customers China, 6 million SMEs (Chataing and Kushnir 2018). use Banorte’s app, and enhance customers’ direct interactions with bank personnel. Barclays has used customer sentiment in social Big data is also helping underserved groups by enhancing media to detect and correct shortcomings in its app. Garanti, a financial networks. In several emerging economies, mobile money leading Turkish bank, combines purchase data with GPS location still relies on networks of agents that allow users to deposit and data to identify customers’ favorite stores and notify them when withdraw cash. If poorly managed, agents in these networks can they are close to a preferred store that has a special offer (Forbes run out of money or not be located close enough to their 2015). customers. Zoona, a financial services provider in Zambia, uses simulations, data from field staff, and Google Maps to determine Financial institutions are also using big data to conduct the optimal location of agents as well as potential demand targeted advertising and promote cross-selling of their products. (Mastercard Foundation and IFC 2017). DBS Bank, a Singaporean Instead of launching “one-size-fits-all” marketing campaigns, big bank, uses data on ATM withdrawals and deposits to forecast ATM data allows financial institutions to use a wide range of individual cash demands and optimize its ATM network (Fitzgerald 2014). characteristics to segment clients into different groups and provide tailored offers according to their preferences. Some institutions are Some financial institutions, such as the BBVA Data & Analytics even using big data to anticipate customers’ future needs. For Center and the JP Morgan Chase Institute, are using their unique example, a credit card company in the Republic of Korea uses proprietary data to produce new knowledge that could be used to transaction records of doctor visits and product purchases to better understand consumers and provide better and new services. determine whether a user is expecting a child and, as a result, For example, one study examines income and consumption provide tailored recommendations for credit cards and discounts patterns of 2.5 million U.S. account holders (and 135 related million on child care services (Korea Joongang Daily 2019). transactions) to draw conclusions about households’ earnings and spending volatility (JP Morgan Chase Institute 2015). Financial institutions can also use big data to expand their customer base. One of the main reasons for denying credit is lack Risks in the Use of Big Data of or inadequate credit history, which prevents financial While big data has the potential to transform the way financial institutions from properly assessing prospective customers’ services are provided, successfully using big data presents some creditworthiness. Big data can mitigate this issue by enabling important challenges that, if not addressed, could have adverse financial institutions to profile prospective borrowers based on the effects on consumers and society. These concerns are not exclusive repayment of services other than loans, such as past payments of to the financial sector (Forbes 2017; Carrière-Swallow and Haksar rent, utilities, or mobile phone plans. This information is not new; 2019). financial institutions already collected payment data from online payment of bills. But new technologies are making it easier to Using big data in financial institutions involves constructing organize these data and report them to credit bureaus and large data sets that include private information (such as income, registries. In some instances, new fintech companies, such as details about assets held, credit card numbers, and spending Destacame in Chile and Mexico, are the ones centralizing and habits) submitted by individuals, willingly or de facto. Financial verifying payment data. They then provide these data to financial institutions can augment their own data by hiring third parties to institutions to assess prospective borrowers and extend loans. collect consumer data on their behalf. Financial institutions could 2 Research & Policy Brief No.26 be tempted to sell their proprietary data to interested third parties, two-thirds of respondents claimed that their consumer such as marketing and retail companies, that are interested in information was only 0–50 percent correct (Lucker, Hogan, and understanding spending habits of different demographic and social Bischoff 2017). Different factors can explain these errors, including groups (Asrow and Xu 2018). In this context, data privacy is an the cost of maintaining up-to-date information and individuals important concern. Financial institutions need to be transparent consciously providing wrong information to protect their privacy. and promptly disclose what information they have on consumers and what are they doing with it. In some circumstances, consumers While reducing information asymmetries between consumers could be granted the power to limit financial institutions from and financial institutions, big data could potentially create selling specific information. information asymmetries across financial institutions. Large financial institutions start with a big base of consumers. With more Once data are collected, financial institutions need to data, large financial institutions can conduct better analyses and safeguard the information. In the wrong hands, it could be used to improve their services more easily than small institutions, widening commit identity theft, scams, and extortions. Even as financial their data advantage and reinforcing their dominant position. This institutions have increased spending on IT security, the number of feedback loop, which can lead to higher entry costs and data breaches has grown. Around one-quarter of financial service monopolistic positions, can be amplified by the existence of companies reported a data breach in 2018—and half of them said economies of scale in big data (OECD 2016). Big data could also that it was not the first time (Thales 2018). Financial institutions widen asymmetries between users of financial services. For need to take steps to enhance data security by, for example, instance, large firms tend to be older and have more available data strengthening data access monitoring, data masking, and data than small firms. By making relatively better forecasts, big data encryption. could reduce the risk premiums for large firms relatively more than for smaller firms, widening their relative cost of capital and Cloud computing also poses benefits and risks. Cloud impairing competition (Begenau, Farboodi, and Veldkamp 2018). computing refers to storing, managing, and processing data in the Concentration might also arise from other financial sector activities Internet through third-party servers rather than in a firm’s own beyond the provision of finance, such as in algorithmic trading. servers. It reduces costs and increases flexibility. Despite these benefits, financial institutions might prefer to keep data in-house, Policy Discussion arguing that storing it online would make it more vulnerable to hacking. However, hacking of in-house servers is just as likely Policy makers around the world have started to discuss ways to through an Internet connection or other means. Because cloud meet the new challenges that accompany big data and to regulate computing services are specialized in data storage, they might even its use (EBA 2016; U.S. Treasury 2018). Several countries have have more expertise in dealing with data security issues than a issued regulations to enhance consumers’ rights over their own financial institution. data and have increased penalties for data misuse (Greig 2019). Policy makers have also started to discuss whether big data can be Another challenge of relying on big data is that it can a source of market power and whether steps to promote a more sometimes lead to biases if used without the appropriate care efficient distribution of data are needed (OECD 2016). Some (so-called algorithm bias) (FTC 2016). In the process of analyzing governments have gone even further, regulating the use of big data through algorithms, human bias can be introduced in automatic algorithms (Bloomberg Law 2019). multiple stages: from choosing the variables in the model and their relative importance to how the sample is collected and results Beyond regulating big data, policy makers could be interested interpreted. Furthermore, models can unintentionally introduce in collecting data from financial institutions. In the hands of biases even when they are not programmed to do so. For example, financial institutions, data are used to identify new business involuntary bias can occur when models are calibrated using opportunities that might result in better financial services. But if historical data that was generated under conditions of the government could collect, centralize, and asses data dispersed discrimination (such as loan applications where applicants were among financial institutions, it could use that information for policy discriminated based on race) (Barocas and Selbst 2016). making, such as identifying and mitigating gaps in financial markets. A recent survey (Central Banking 2018) shows that about Financial institutions might not only use big data to offer 60 percent of central banks are using big data for multiple products that consumers might demand, but could also tailor purposes, including forecasting, stress testing, and combating prices to extract the maximum profit from each consumer. For money laundering (figure 2). Uses include analyzing credit card example, algorithm-based lending can result in higher interest spending data to gain real-time information on current economic rates being charged to minorities after determining that these activity. Compared to traditional retail survey data that are groups rarely do cost comparisons and live in areas where the available only with considerable lags and are not collected for all supply of financial services, and thus competition, is limited locations, big data could prove useful to study localized and (Bartlett et al. 2019). Some financial institutions even use big data short-lived shocks (Aladangady et al. 2019). to identify vulnerable groups in need of quick cash and offer them financially risky products, such as payday loans (Business Insider Moreover, partially disclosing some of these data could 2013). increase data access, level the playing field, and encourage competition. Korea is an example of a country that is moving More data do not necessarily lead to better analyses. Data sets forward to open data for digital development and competition in can be so large that is hard to know where to start and what to look the financial sector. In 2020, Korea is expected to launch CreDB, an for; the relevant information might be hidden underneath a mound open data system that will grant fintech firms, financial companies, of useless data (O’Donnell 2017). Analytical tools used in big data and the academia access to de-identified credit data. Korea is also can be good at finding correlations between variables but not so creating a platform for data exchange and a new agency to ensure good at distinguishing correlation from actual causation. the safe use and intermediation of big data (FSC 2019). Traditional techniques to infer causation, such as randomized control trials and natural experiments, may also be needed (Athey However, obtaining the needed information presents its own 2017). Furthermore, big data is not always completely accurate. challenges. Financial institutions might be reluctant to disclose For example, a survey in the United States found that more than information that they use to gain a comparative edge. Forcing 3 Using Big Data to Expand Financial Services: Benefits and Risks Figure 2. Use of Big Data by Central Banks Nowcasting 55% Forecasting 55% Stress testing 42% Anti-money laundering 32% Fraud detection/prevention 30% Cyber security 26% Other 13% 0% 10% 20% 30% 40% 50% 60% % of survey respondents Central banks are using big data in a variety of ways to enhance their operations. Source: Central Banking (2018). financial institutions to disclose information could also have the export credit is helping exporters. Research institutions, such as unintended consequence of discouraging them from investing in http://socialobservatory.worldbank.org/categories/democratizing-data academic organizations and think tanks, and international technologies to collect and analyze data. Consumers might also http://socialobservatory.worldbank.org/categories/democratizing-data organizations could help policy makers in these efforts to organize raise concerns https://govinsider.asia/ about the government gaining access to their smart-gov/inside-jakartas-new-smart-city-hq/ and analyze data. personal data. https://govinsider.asia/ smart-gov/inside-jakartas-new-smart-city-hq/ https://www.brookings.edu/blog/future-development/2019/06/07/can-technology-improve-service-delivery/ Overall, countries would benefit greatly from developing an If the government manages to get big data from the financial https://www.brookings.edu/blog/future-development/2019/06/07/can-technology-improve-service-delivery/ information system that allows financial and economic information sector, the next challenge would be to integrate these data with collected by different agencies to be matched and used more other data that the government collects, such as data from the real systematically. While building this system might be challenging and sector, to obtain a more complete picture. For example, policy take time, the developmental benefits might well be worth the makers might obtain data from the financial sector about firms that costs. A more thorough understanding of the financial sector can use export credit. Combining these data with other data (such as prove of immense value in the quest for building a more exports by firm) would enable governments to examine whether developed, competitive, and stable financial sector. References Aladangady, A., S. Aron-Dine, W. Dunn, L. Feiveson, P. Lengermannm, and C. Sahm, Fitzgerald, M. 2014. “DBS Bank Pumps Up the Volume on its Technology.” Interview 2019. “From Transactions Data to Economic Statistics: Constructing Real-time, with David Gledhill, DBS Bank. MIT Sloan Management Review, January 6. High-frequency, Geographic Measures of Consumer Spending.” NBER Working Paper 26253, National Bureau of Economic Research. Forbes. 2015. “Turkey’s Garanti Bank Captures Elusive Millennials.” June 15. Asrow, Kaitlin, and Joanne Xu. 2018. The 2018 California Consumer Privacy Act: ---------. 2017. “3 Massive Big Data Problems Everyone Should Know About.” June 15. Understanding Its Implications and Ambiguities. Fintech Edge Special Report. FSC (Financial Services Commission). 2019. “FSC Announces Plans to Establish Financial Federal Reserve Board of San Francisco Big Data Infrastructure.” June 3. Athey, S. 2017. “Beyond Prediction: Using Big Data for Policy Problems.” Science 355 FTC (U.S. Federal Trade Commission). 2016. Big Data: A Tool for Inclusion or Exclusion? (6324): 483–85. Understanding the Issues. Barocas, S., and A. D. Selbst. 2016. “Big Data’s Disparate Impact.” California Law Review Greig, J. 2019. “How More Countries Plan to Pass Stringent Privacy Laws in 2019.” 104 (3): 671–732. TechRepublic, June 25. Bartlett, R., A. Morse, R. Stanton, and N. Wallace. 2019. “Consumer-lending Discrimination JP Morgan Chase Institute. 2015. Weathering Volatility: Big Data on the Financial Ups in the FinTech Era.” NBER Working Paper 25943, National Bureau of Economic Research. and Downs of U.S. Individuals. Begenau, J., M. Farboodi, and L. Veldkamp. 2018. “Big Data in Finance and the Growth Korea Joongang Daily. 2019. “Big Data Means Big Money.” May 20. of Large Firms.” Journal of Monetary Economics 97 (C): 71–87. Lucker, J., S. K. Hogan, and T. Bischoff. 2017. “Predictably Inaccurate: The Prevalence Bloomberg Law. 2019. “AI Regulations Aim at Eliminating Bias.” May 31. and Perils of Bad Big Data.” Deloitte Review 21 (July): 8–25. Business Insider. 2013. “How Marketers Use Big Data to Prey on the Poor.” December 19. Mastercard Foundation and IFC (International Finance Corporation). 2017. Handbook: Carrière-Swallow, Y., and V. Haksar. 2019. “The Economics and Implications of Data: An Data Analytics and Digital Financial Services. Partnership for Financial Inclusion. Integrated Perspective.” Departmental Paper No. 19/16, International Monetary Fund. O’Donnell, B. 2017. “We All Thought Having More Data Was Better. We Were Wrong.” Central Banking. 2018. Big Data in Central Banks Focus Report 2018. Vox, April 12. Chataing, M-S., and K. Kushnir. 2018. “How Is Ant Financial Closing the SME Finance OECD (Organisation for Economic Co-operation and Development). 2016. “Big Data: Gap in China.” SME Finance Forum Blog, June 7. Bringing Competition Policy to the Digital Era.” Background Note by the EBA (European Banking Authority). 2016. “On the Use of Big Data by Financial Secretariat. DAF/COMP 14. Institutions.” Joint Committee Discussion Paper JC/2016/86. SNS Telecom & IT. 2018. Big Data in the Financial Services Industry: Economist, The. 2015. “Cool, Man.” May 7. 2018–2030—Opportunities, Challenges, Strategies & Forecasts. Economist, The. 2019. “South Korea is Trying to Make Banking Fun.” May 2. Thales. 2018. 2018 Thales Data Threat Report. Fintech News. 2017. “Alternative Credit Scoring Providers in Southeast Asia You Have to http://documents.worldbank.org/curated/en/795161508437504068/Corporate-debt-maturity-in-developing-countries-sources-of-long-and-short-termism U.S. Department of the Treasury. 2018. A Financial System that Creates Economic Know.” June 9. Opportunities Nonbank Financials, Fintech, and Innovation. Global Knowledge & Research Hub in Malaysia