BIS Working Papers No 841 On fintech and financial inclusion by Thomas Philippon Monetary and Economic Department February 2020 JEL classification: E2, G2, N2. Keywords: fintech, discrimination, robo advising, credit scoring, big data, machine learningBIS Working Papers are written by members of the Monetary and Economic Department of the Bank for International Settlements, and from time to time by other economists, and are published by the Bank. The papers are on subjects of topical interest and are technical in character. The views expressed in them are those of their authors and not necessarily the views of the BIS. This publication is available on the BIS website (bis). Bank for International Settlements 2020. All rights reserved. Brief excerpts may be reproduced or translated provided the source is stated. ISSN 1020-0959 (print) ISSN 1682-7678 (online)On Fintech and Financial Inclusion Thomas Philippon September 2019 Abstract The cost of nancial intermediation has declined in recent years thanks to technology and increased competition in some parts ofthe nanceindustry. I documentthis fact andI analyze two features of newnancial technologies that have stirred controversy: returns to scale and the use of big data and machine learning. I argue that the nature of xed versus variable costs in robo-advising is likely to democratize access to nancial services. Big data is likely to reduce the impact of negative prejudice in the credit market but it could reduce the eectiveness of existing policies aimed at protecting minorities. JEL: E2, G2, N2 Stern School of Business, New York University; NBER and CEPR. This paper was prepared for the 2019 BIS Annual Research Conference. I am grateful to my discussants Manju Puri and David Dorn, to Hyun Shin, Marina Niessner, and participants at the 2019 BIS Annual Research Conference. I thank Marcos Sonnervig for outstanding research assistance. 1Fintech covers digital innovations and technology-enabled business model innovations in the nancial sector. Such innovations can disrupt existing industry structures and blur industry boundaries, facilitate strategic disin- termediation, revolutionize how existing rms create and deliver products and services, provide new gateways for entrepreneurship, and democratize access to nancial services. On the other hand, they create signicant privacy, regulatory and law-enforcement challenges and they could increase the scope for some forms of discrimination. Examples of innovations that are central to Fintech today include various application of blockchain technologies, new digital advisory and trading systems, articial intelligence and machine learning, peer-to-peer lending, equity crowdfunding and mobile payment systems. In this paper I oer some preliminary evidence and theoretical analysis about the impact of technological progress in the nance industry. The rst question is whether there has been any material change in nancial intermediation in recent years. To shed some light on this question, I update the work of Philippon (2015) with post-crisis U.S. data. The puzzle emphasizedinpreviousworkwasthat theunitcostofnancialintermediationhadremainedstubbornlycloseto200 basis points for more than a century, despite advances and large investments in computers and communication technologies. The post-crisis data suggests that this puzzle might be diminishing. I nd that the unit cost of nancial intermediation has declined over the past 10 years. I then study two issues that are at the heart of the Fintech debate: access to nance and discrimination. If we accept the fact that Fintech brings eciency gains to nancial intermediation, the next question is: how will these gains be shared? Will Fintech democratize access to nancial services or will it increase inequality? I highlight two forces that will shape the answer to these questions. The rst force is increasing returns to scale brought by technology. I argue that the nature of xed versus variable costs has changed in a way that is likely to improve access to nancial services. It may not, however, reduce inequality among all groups. The second force is the use of big data and machine learning (BDML for short). I argue that this technology is likely to reduce unwarranted human biases against minorities, but it will probably decrease the eectiveness of existing regulations. The tentative conclusion is that Fintech can bring widely-shared welfare benets but changes in existing policies and regulations are necessary to achieve its full potential. Recent literature Philippon (2016)discussesthe literatureup to2016soIwill mentionheresomerecentpapers. Focusing on residential mortgages, Buchak et al. (2018) study the growth in the market share of shadow bank and Fintechlenders,arguingthatitcanbeexplainedbydierencesinregulationandtechnologicaladvantages. Theynd that Fintech lenders serve more creditworthy borrowers (relative to shadow banks) but charge higher interest rates (14-16basispoints),whichisconsistentwiththeideathatconsumerarewillingtopayforbetteruserexperienceand quick decisions. Fuster et al. (2019) study the dierences between Fintech and traditional lenders in the mortgage market and nd that the former is quicker in processing applications (20% faster), without increasing loan risk. They also provide evidence that Fintech lenders adjust supply more elastically to demand shocks and increase the 2propensity to renance, especially among borrowers that are likely to benet from it. Their results suggest that Fintech rms have improved the eciency of nancial intermediation in mortgage markets. The advent of Fintech is often seen as a promising avenue for reducing inequality in access to credit. Bartlett et al.(2018)studythisissue, analyzingthe roleofFintechlendersinalleviatingdiscriminationinmortgagemarkets. They nd that all lenders, including Fintech, charge minorities more for purchase and renance mortgagesbut that Fintech algorithms discriminate 40% less than face-to-face lenders. Regarding the use of new technologies in credit markets,Bergetal.(2019)analysetheinformationcontentofthe“digitalfootprint” (aneasilyaccessibleinformation for any rm conducting business in the digital sphere) for predicting consumer default. With data from a German e-commerce, they nd that it equals or exceeds the predictive power of traditional credit bureau scores. Their results suggest that new technologies and new data might bring a superior ability for screening borrowers. FinTechs are also competing in the market for wealth management. The United States is the leading market for robo-advisors. In 2017, it accounted for more than half of all investments in robo-advisors (Abraham et al., 2019). Nevertheless, the amount of assets managed by robo-advisors is still a small portion of total assets under management,withaverageclientwealthmuchsmallerthantheaverageintheindustry(Economist,2017). Abraham et al. (2019) argues that because they save on xed costs (such as salaries of nancial advisors or maintenance of physical oces), robo-advisors can reduce minimum investment requirements and charge lower fees. 1 (In)eciency of the Existing System The main nding in Philippon (2015) is that the unit cost of nancial intermediation in the U.S. has remained around 200 basis points for the past 130 years. Improvements in information technologies have not been passed through to the end users of nancial services. This section oers an update of this work. 1.1 Financial Expenses and Intermediated Assets To organize the discussion I use a simple model economy consisting of households, a non-nancial business sector, and a nancial intermediation sector. The details of the model are in the Appendix. The income share of nance, shown in Figure 1, is dened as 1 y f t y t = Value Added of Finance Industry GDP . 1 Philippon (2015) discusses various issues of measurement. Conceptually, the best measure is value added, which is the sum of prots and wages. Whenever possible, I therefore use the GDP share of the nance industry, i.e.,the nominal value added of the nance industry divided by the nominal GDP of the U.S. economy. One issue, however, is that before 1945 prots are not always properly measured and value added is not available. As an alternative measure I then use the labor compensation share of the nance industry, i.e., the compensation of all employees of the nance industry divided by the compensation of all employees in the U.S. economy. Philippon (2015) also explains the robustness of the main ndings to large changes in government spending (because of wars), the rise of services (nance as a share of services displays a similar pattern to the one presented here), globalization (netting out imports and exports of nancial services). 3The model assumes that nancial services are produced under constant returns to scale. The income of the nance industry y f t is then given by y f t = c,t b c,t + m,t m t + k,t k t , (1) whereb c,t isconsumercredit,m t areassetsprovidingliquidityservices,andk t isthevalueofintermediatedcorporate assets. The parameters i,t s are the unit cost of intermediation, pinned down by the intermediation technology. The model therefore says that the income of the nance industry is proportional to the quantity of intermediated assets, properlydened. The model predictsno income eect, i.e., notendency forthe nance incomeshareto grow with per-capitaGDP. This doesnot mean that the nance incomeshareshould be constant, since the ratioofassets to GDP can change. But it says that the income share does not grow mechanically with total factor productivity. This is consistent with the historical evidence. 2 Measuringintermediatedassetsiscomplicatedbecausetheseassetsareheterogenous. Asfarascorporatenance is concerned, the model is fundamentally a user cost model. Improvements in corporate nance (a decrease in k ) lower the user cost of capital and increase the capital stock, which, from a theoretical perspective, should include all intangible investments and should be measured at market value. A signicant part of the growth of the nance industry over the past 30 years is linked to household credit. The model provides a simple way to model household nance. The model also incorporates liquidity services provided by specic liabilities (deposits, checking accounts, some form of repurchase agreements) issued by nancial intermediaries. One can always write the RHS of (1) as c,t b c,t + m,t c,t m t + k,t c,t k t . Philippon (2015) nds that the ratios m,t c,t and k,t c,t are close to one. 3 As a result one can dene intermediated assets as q t b c,t +m t +k t . (2) The principle is to measure the instruments on the balance sheets of non-nancial users, households and non- nancial rms. This is the correct way to do the accounting, rather than looking at the balance sheet of nancial intermediaries. After aggregating the various types of credit, equity issuances and liquid assets into one measure, I obtain the quantity of nancial assets intermediated by the nancial sector for the non-nancial sector, displayed in Figure 1. 1.2 Unit Cost and Quality Adjustments I can then divide the income of the nance industry by the quantity of intermediated assets to obtain a measure of the unit cost t y f t q t . (3) 2 Thefactthatthenance shareofGDPisthesamein1925andin1980makesisalreadyclearthatthere isnomechanicalrelationship between GDP per capita and the nance income share. Similarly, Bickenbach et al. (2009) show that the income share of nance has remained remarkably constant inGermany overthe past30years. Moreprecisely, usingKLEMSforEurope (seeOMahonyand Timmer (2009) one can see that the nance share in Germany was 4.3% in 1980, 4.68% in 1990, 4.19% in 2000, and 4.47% in 2006. 3 This is true most of the time, but not when quality adjustments are too large. Philippon (2015) provides calibrated quality adjustments for the U.S. nancial system. 4Figure 1: Finance Income and Intermediated Assets 1 2 3 4 Intermediated Assets/GDP .02 .04 .06 .08 Share of GDP 1880 1900 1920 1940 1960 1980 2000 2020 year. Share of GDP Intermediated Assets/GDP Notes: Both series are expressed as a share of GDP. Finance Income is the domestic income of the nance and insurance industries, i.e., aggregate income minus net exports. Intermediated Assets include debt and equity issued by non nancial rms, household debt, and various assets providing liquidity services. Data range for Intermediated Assets is 1886 - 2012. See Philippon (2015) for historical sources and details about the underlying data. Figure 2 shows that this unit cost is around 2% and relatively stable over time. In other words, I estimate that it costs two cents per yearto create and maintain one dollar of intermediated nancial asset. Equivalently, the annual rate of return of savers is on average 2 percentage points below the funding cost of borrowers. The updated series are similar to the ones in the original paper. The unit costs for other countries are estimated by Bazot (2013) who nds convergence to US levels. Figure 2: Unit Cost of Financial Intermediation 0 .005 .01 .015 .02 .025 .03 1880 1900 1920 1940 1960 1980 2000 2020 time 2012 Data New Data Unit Cost Notes: The raw measure is the ratio of nance income to intermediated assets, displayed in Figure 1. The 2012 data is from Philippon (2015), while the new data was accessed in May 2016. Data range is 1886 - 2015. The raw measure of Figure 2, however, does not take into account changes in the characteristics of borrowers. These changes require quality adjustments to the raw measure of intermediated assets. For instance, corporate 5nanceinvolvesissuingcommercialpaperforbluechip companiesaswellasraisingequityforhigh-technologystart- ups. The monitoring requirements per dollar intermediated are clearly dierent in these two activities. Similarly, with household nance, it is more expensive to lend to poor households than to wealthy ones, and relatively poor households have gained access to credit in recent years. 4 Measurement problems arise when the mix of high- and low-quality borrowers changes over time. Following Philippon (2015), I then perform a quality adjustment to the intermediated assets series. Figure 3 shows the quality adjusted unit cost series. It is lower than the unadjusted series by construction since quality adjusted assets are (weakly) larger than raw intermediated assets. The gap between the two series grows when there is entry of new rms, and/or when there is credit expansion at the extensive margin (i.e., new borrowers). Even with the adjusted series, however, we do not see a signicant decrease in the unit cost of intermediation over time. Figure 3: Unit Cost and Quality Adjustment 0 .005 .01 .015 .02 .025 .03 1880 1900 1920 1940 1960 1980 2000 2020 time Raw Quality Adjusted Unit Cost, with Quality Adjustment Notes: The quality adjusted measure takes into account changes in rms and households characteristics. Data range is 1886 - 2015. As I ha