国际清算银行-金融科技和大型科技信贷：一个新的数据库.docx

BAN K FO R I NTERN ATI0 N ALSETTLEM ENTSBIS Working PapersNo 887Fintech and big tech credit: a new databaseby Giulio Cornelli, Jon Frost, Leonardo Gambacorta, Raghavendra Rau, Robert Wardrop and Tania ZieglerMonetary and Economic DepartmentSeptember 2020JEL classification: E51z G23z 031.Keywords: fintech, big tech, credit, data, technology, digital innovation.Graph 2Fintech credit platforms continue to exit in China, as stock and flow of loans fallNumber of fintech credit platforms in ChinaStock and flow of loans is falling as average tenors riseNumber of platformsNumber of entries/exits RMB bnMaturity in monthsSources: WDZJ ; authors' calculations.Supervisory Service and Bank of Korea. Fintech credit volumes reached about USD 2.3 billion in 2019, and the market is dominated by P2P/marketplace property lending.The stock of fintech and big tech credit is estimated at USD 15.2 billion as of end-2019.The UK, meanwhile, had estimated fintech credit volumes of USD 11.5 billion in 2019 (up from USD 9.3 billion in 2018), made up of a vibrant mix of P2P/marketplace business, consumer and property lending, and smaller volumes of balance sheet lending and invoice trading. After rapid growth in 2013 16, fintech lending volumes have been relatively steady in the United Kingdom in the past three years, perhaps reflecting greater maturity and saturation in the relevant market segments. For instance, Ziegler et al. (2020) estimate that fintech credit platforms accounted for up to 27.7% of equivalent bank credit to small and medium enterprises with annual turnover below GBP 2 million in 2018. This may have been encouraged by public policy; for instance, the government-owned British Business Bank invested over GBP 165 million over 2014-18 for lending through Funding Circle, a UK credit platform, and announced a commitment for a further GBP 150 million to support small business lending (British Business Bank, 2018). Big tech credit volumes are estimated to be much smaller, at an estimated USD 100 million in 2017 and 2018, primarily through Amazon's Seller Lending programme.Looking beyond the largest fintech and big tech credit markets, higher-frequency data from Brismo and WDZJ show that fintech credit volumes have continued to grow rapidly in the European Union, Australia and New Zealand, even as they have plateaued in the United States and United Kingdom and declined in China (Graph 3Z left-hand panel). In many emerging market and developing countries (not shown), fintech lenders are becoming economically significant lenders for specific segments, such as small and medium-sized enterprises (SMEs) (Cornell! et al., 2019; World Bank, 2020). Some fulfil so-called agency banking functions, by which they function as agents to expand the reach of banks, especially in Latin America and parts of Asia and Africa.Graph 3Fintech credit is growing in Europe, big tech credit is booming in AsiaBig tech credit is booming in Asia, the United States and Africa2Big tech credit is booming in Asia, the United States and Africa22018 2019Fintech lending volumes are diverging1Index, QI 2017 = 100CN = China, JP = Japan, KR = Korea, US = United States, KE = Kenya, ID = Indonesia.1 Data are based on five platforms for Australia and New Zealand, all platforms covered by WDZJ for China, 49 platforms for Europe, 34 for the United Kingdom and five for the United States. Volumes are reported in local currency. 2 Figures include estimates.ioSource: Brismo ; WDZJ ; companies* reports; authors' calculations.Big tech credit is achieving economically significant scale in China, Japan, Korea； parts of Southeast Asia, East Africa and (to a lesser extent) some countries in Latin America (Graph 3, right-hand panel). This is driven by the lending activities of e- commerce platforms like Mercado Libre, ride-hailing companies like Grab and Go-Jek, and telecommunication and mobile money providers like M-Pesa. In many cases, these lenders initially target a specific group of users (e.g. sellers on the e-commerce platform, or drivers) but then expand such credit offerings to more users over time.Interest rates, defaults and marginsInformation on interest rates, defaults and profit margins is not available for all countries in the sample, but available data can give some useful insights.The interest rates charged on fintech credit appear to be roughly in line with comparable bank loans. For borrowers, as of the latest readings by CCAFZ typical interest rates charged on the major US fintech credit platforms range between 9 and 28%. In the UKZ interest rates are between 6.5 and 24%. In China, interest rates have been more volatile in past years given the changes in the market and regulation (Gambacorta et al., 2019). Data on the interest rates charged by big tech companies are not available. Looking across the largest fintech credit markets, it is apparent that the returns paid to investors in fintech credit platforms have been relatively high in the past five years, but are trending downward globally (Graph 4, left-hand panel). For big tech companies, the return on loans could also include the benefit obtained from supporting companies' core business lines (e-commerce, social media, advertising etc.), user loyalty to the platform's overall services, and user data.Graph 4Big tech firms are highly profitable, while fintech platforms have often struggledReturns on fintech credit platforms have trended down Big tech firms are more profitable than fintech platformsNet profit margin:Fintech2 Big tech3Per cent1 Average interest rate. 2 Simple average of Black Knight Financial Services, Elevate, Enova International, Fellow Finance, Funding Circle, LendingClub, Lendingtree, Nelnet, OnDeck and Synchrony. 3 Simple average of Alibaba, Amazon, Apple, Baidu/Du Xiaoman, Facebook Google, JD , Kakao, LINE, Microsoft, MTS bank, Orange, Rakuten, Samsung, Tencent, Uber, Vodacom, Vodafone and Yandex.Source: Brismo ; Refin it iv Eikon; WDZJ ; authors* calculations.Defaults at fintech credit pla甘orms have picked up in the past few years. More granular default data show that certain loans segments like US consumer lending have seen a worsening of credit quality in the past three years; high-frequency data on the impact of the Covid-19 pandemic on credit quality are not yet available. At the same time, available empirical evidence suggests that some fintech and big tech lenders, through the use of alternative data and machine learning, have been able to achieve lower default rates than with traditional data and models, and even to achieve superior performance after a downturn in the credit cycle (Gambacorta et al, 2019). It is an open question how credit models will perform in the current downturn.For fintech and big tech platforms to continue to grow, intermediation needs to be profitable for the providers (a separate consideration than the return provided to investors). In this light, the profit margins of big tech firms (that benefit, however, from a more diversified bundle of activities) are relatively high. Margins are particularly high when compared with those of fintech credit platforms, which have often struggled to achieve profitability (Graph 4, right-hand panel) and have relied on new investor funding for expansion. For big tech firms, this often relates to the high profit margins in core businesses lines. In some cases, there are questions about whether big tech platforms wish to engage directly in lending at allz since it is less profitable than these other activities (FSBZ 2019). This, as well as regulations, may be a factor behind the use of partnership models, where the big tech distributes financial products but a financial institution retains such lending on its balance sheet.3. Drivers of credit volumes across economies: a panelanalysisIn this section, we seek to explain fintech and big tech volumes in different economies over time. This is a novelty with respect to earlier studies (Claessens et al.，2018; Frost et aLz 2019) that analyse such volumes in the cross section. Leveraging on the new database and following Rau (2020), we extend the analysis using a panel approach. We look at the drivers of fintech and big tech separately, and then take a deeper look at a range of specific country characteristics that are most salient in the cross-section dimension, for their sum (total alternative credit).We hypothesise that fintech and big tech credit per capita can be broadly related to demand side and supply side drivers. On the demand side, we expect that more developed economies (with higher GDP per capita) will have a higher demand for credit from firms and households, but that this relationship may show a decreasing trend for very high levels of development (see Claessens et al., 2018; Frost et al” 2019; Bazarbash and Beaton, 2020). Similarly, we expect that fintech and big tech credit will be higher when incumbent banking services are more expensive (higher banking sector mark-ups), for instance because of less competition, and where there is a larger un(der)met demand for financial services, as proxied by fewer bank branches per capita. On the supply side, we expect that more stringent banking regulation (a proxy for the overall stance of financial regulation) will create barriers to the entry for fintech and big tech firms. A number of additional institutional characteristics, such as the ease of doing business, investor protection and disclosure, the judicial system and characteristics of the incumbent banking system will be discussed later. We will try also to answer the openquestion of whether fintech and big tech credit complement or substitute for bank credit and other forms of finance.11Our baseline regression takes the form:in(cccccccfltfe) = ® + 帆州,I + 662y猿I + "叫,1 + 制解用6 i +腆%T +。画ii,gi +册叫攵+网(1)where CCCCCCCCfffe is the volume of fintech or big tech credit per capita in economy i at time t, or total alternative credit Thus, we consider three credit aggregates as left-hand side variables, each with the same regressors.For Germany, De Roure et al. (2016) find that P2P lending substitutes the banking sector for high-risk consumer loans. De Roure et al. (2018) present a theoretical model and further evidence in favour of such "bottom fishing”. For the United States, Tang (2019) finds that P2P lending is a substitute for bank lending in terms of serving infra-marginal bank borrowers, but that it complements bank lending with respect to small loans. Frost et al. (2019) refer to the sum of fintech and big tech credit as “total fintech credit”. Here, to prevent confusion, we refer to the sum of fintech and big tech credit as "total alternative credit0. We also regress bank credit per capita using the same specification to check for significant differences.The right-hand side includes a number of regressors that are lagged by one year to mitigate endogeneity issues. is the GDP per capita in economy i at year t-1, and the variable 恍1 is its quadratic term, to address the non-linear relationship between credit development and income levels. is the Lerner indexThe Lerner Index of banking sector mark-ups has been updated over the period 2015-17 using information on the alternative cyclical measure devolped by Igan et al. (2020). See Annex A. A higher value indicates higher margins and profitability among traditional banks, and thus less competition. of banking sector mark-ups in economy i, reflecting market power by incumbent banks; a higher value may reflect a less competitive banking sector. Whji-i is an index of regulatory stringency for the banking sector of economy i, as constructed by Barba Navaretti et al. (2017) from World Bank data.The regulatoiy stringency variable is constructed as an index (normalised between 0 and 1) based on the World Bank's Bank Regulation and Supervision Survey. The index takes a value between 0 (least stringent) and 1 (most stringent) based on 22 questions (2011 survey) or 23 questions (2019 survey) about bank capital requirements, disclosure, the legal powers of supervisory agencies etc. BBS8 is the density of the bank branch network in country i compared with the adult population (which may capture both the reach of the banking sector and its relative cost base).瑞弘 is a vector of control variables that includes: growth in GDP and total credit; a real short-term interest rate; a dummy for whether a country had suffered a financial crisis since 2006, as defined by Laeven and Valencia (2018); mobile phone subscriptions (given the mobile-based nature of many platforms); and a dummy for advanced economies. Bank branches and mobile subscriptions are measured relative to the adult population.When observation for bank branches, mobiles and credit growth were not yet available, we extrapolated the figures using the cross-country average growth rate or the cross-country average change. is a vector of geographical area fixed effects and %坨 is an error term.The inclusion of some (barely) time-invariant country-specific regressors (see next section) prevents us from using a complete set of country dummies.Table 1 reports descriptive statistics for our sample of 79 countries over the period 2013-18. Given the lower coverage of many variables for 2019, we have excluded this year from regressions. Data come from a variety of sources, including the IMF's World Economic Outlook and the World Bank's Global Financial Development Database (GFDD) and Findex.VariableObservationsMeanStandard deviationMinMaxGDP per capita (in thousands of USD)45321.5318.210.6787.76Lerner index14530.300.15-0.051.00Bank branches per 100,000 adults45317.6514.031.4383.75Normalised index of bank regulatory stringency24530.720.100.380.96Score-Starting a business (overall)42582.5511.4623.0499.96Score-Time (days)42581.9617.330.00100.00Score-Paid-in Minimum capital (% of income per capita)42594.8515.450.00100.00Score-Cost (% of income per capita)42584.9226.950.00100.00Extent of disclosure index (0-10)42564.4423.800.00100.00Trial and judgment (days)425407.85203.5190.001095.00Enforcement of judgment (days)425177.98110.5926.00597.00Enforcement fees (% of claim)4255.375.230.0023.30Bank credit to bank deposits (%)212105.3980.4827.73702.09Bank regulatory capital to risk-weighted assets (%)19717.203.8110.5935.65Provisions to non-performing loans (%)18764.2937.260.00232.06Loans from non-resident banks to GDP (%)19427.5328.331.24158.53Proportion of firms with a transactions account (%)27685.4815.1818100Corporate bond average maturity (years)13510.265.753.5434.09Corporate bond issuance volume to GDP (%)1372.141.810.0513.83Total factori