Jurnal 4 Main

Journal of Banking and Finance 148 (2023) 106742
Contents lists available at ScienceDirect
Journal of Banking and Finance

journal homepage: www.elsevier.com/locate/jbf
Fintech and big tech credit: Drivers of the growth of digital lending ✩
Giulio Cornelli a,d, Jon Frost a,b, Leonardo Gambacorta a,c, P. Raghavendra Rau b,∗,
Robert Wardrop b, Tania Ziegler b
a
Bank for International Settlements, Switzerland
b
Cambridge Centre for Alternative Finance, University of Cambridge, United Kingdom
c
CEPR, United Kingdom
d
University of Zürich, Switzerland
a r t i c l e i n f o a b s t r a c t
Article history: Fintech and big tech companies are making rapid inroads into credit markets. We hand construct a global
Received 20 October 2020 database of fintech and big tech lending volumes for 79 countries over 2013–2018. Using a panel regres-
Accepted 1 December 2022
sion analysis, we find these new forms of digital lending are larger in countries with higher GDP per
Available online 5 December 2022
capita (albeit at a declining rate), where banking sector mark-ups are higher, and where banking regula-
JEL classification: tion is less stringent. We also find that these alternative forms of credit are more developed where the
E51 ease of doing business is greater, investor protection disclosure and the efficiency of the judicial system
G23 are more advanced, and where bond and equity markets are more developed. Overall, fintech and big
O31 tech credit seem to complement other forms of credit, rather than substitute for them.
© 2022 The Author(s). Published by Elsevier B.V.
Keywords:
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)
Fintech
Big tech
Credit
Data
Technology
Digital innovation
1. Introduction
Credit markets around the world are undergoing a transforma-

tion. While banks, credit unions, and other traditional lenders re-
✩
We thank the editor Thorsten Beck, an Associate editor, and two anonymous main the chief source of financing for companies and households
referees for very useful comments and suggestions. We also acknowledge comments in most economies (with capital markets playing an important role
and input from Raphael Auer, Tobias Berg, Marcel Bluhm, Stijn Claessens, Sebas-
in some cases), new intermediaries have recently emerged. In par-
tian Doerr, Boris Hofmann, Martin Hood, Pawee Jenweeranon, Ross Leckow, Loriana
Pelizzon, Jermy Prenio, Antoinette Schoar, Jose Maria Serena, René Stulz, Cheng-Yun ticular, digital lending models such as peer-to-peer (P2P) / market-
Tsang, and Joy Wann. We also thank participants at the Deutsche Bundesbank con- place lending and invoice trading have grown in many economies
ference “Banking and Payments in the Digital World”, a Zhejiang University Inter- in the past decade. These types of credit, facilitated by online plat-
national Business School webinar, a Vaduz Roundtable and a BIS research meeting. forms rather than traditional banks or lending companies, are re-
We thank Stephen Ambore, Masaki Bessho, Cyprian Brytan, Iuliia Burkova, Teresa
ferred to as “debt-based alternative finance” (Wardrop et al., 2015),
Caminero, Greg Chen, Anrich Daseman, Graeme Denny, Darren Flood, Sergio Gor-
jón Rivas, Aleksi Grym, Cheryl Ho, Tobias Irrcher, Arif Ismail, Chandan Kumar, Lyu “fintech lending” (Berg et al., 2021), or “fintech credit” (FSB and
Yuan, Nur Fazila Mat Salleh, Nicolas Même, Manoranjan Mishra, Aiaze Mitha, Irina CGFS, 2017). Moreover, in the past few years, large companies
Mnohoghitnei, Mu Changchun, Michelle O’Donnell Keating, Vichett Oung, Jisoo Park, whose primary business is technology (typically referred to as “big
Naphongthawat Phothikit, Melchor Plabasan, Bintang Prabowo, Ricky Satria, Martina
techs”) have entered credit markets, lending either directly or in
Sherman, Paul Shi, Joshua Slive, Ylva Søvik, Edward Tan, Rupert Taylor, Triyono, Vi-
cente de Villa, Chris Welch, Maarten Willemen and Melanie Wulff for help with partnership with financial institutions (BIS, 2019; Stulz, 2019). In
data for individual jurisdictions. We thank Tyler Aveni, Matías Fernandez, Gil Guan, this paper, we hand collect data on both fintech and big tech lend-
Daisy Mwanzia, Devyani Parameshwar and Huiya Yao for assistance with company- ing volumes for 79 countries over the period 2013–18. Using a
level data. Research assistance by Haiwei Cao and Yuuki Ikeda is gratefully acknowl- panel analysis, we then analyze the main economic and institu-
edged. The views are those of the authors and not necessarily of the Bank for In-
tional drivers of these alternative forms of credit.
ternational Settlements
∗
Corresponding author. Research on fintech and big tech credit is important, because
E-mail address: r.rau@jbs.cam.ac.uk (P.R. Rau). data on the overall size of these markets are notably scarce.
https://doi.org/10.1016/j.jbankfin.2022.106742
0378-4266/© 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)
G. Cornelli, J. Frost, L. Gambacorta et al. Journal of Banking and Finance 148 (2023) 106742
Rapidly expanding credit in an economy (as in a credit boom) In this paper, we construct a new global database on both fin-
has the potential to predict both financial crises and severe re- tech and big tech credit volumes, We use data on fintech credit
cessions (see Drehmann et al., 2010; Schularick and Taylor, 2012; available from the annual CCAF global survey, and hand collect
Kindleberger and Aliber, 2015). As fintech and big tech credit be- data on big tech credit after contacting central banks and big tech
come more economically significant forms of financing, it is im- firms, to complement the data available from public sources. We
portant to have greater understanding of the extent and drivers of then analyze the data using panel regressions to examine the main
such lending. economic and institutional factors that are driving the growth of
There are key differences between the two types of credit. Fin- these new forms of digital lending. We also compare the determi-
tech credit models were originally built around decentralised plat- nants of fintech and big tech lending.
forms where individual lenders choose borrowers or projects to To assess the overall drivers and differences between fintech
lend to in a market framework. Platforms help to solve prob- and big tech credit, we distinguish between macroeconomic factors
lems of asymmetric information both through their screening prac- and institutional characteristics. We hypothesize that the volume
tices, and by providing investors with information on loan risks of fintech and big tech credit should be higher where it is more
and other borrower characteristics. Over time, some platforms have attractive for new intermediaries to offer credit, and where there
moved to fund loans from institutional investors rather than only is an un(der)met demand for credit. This should result in higher
individuals, and many use increasingly sophisticated credit mod- volumes in countries with higher institutional quality.
els (see e.g. Jagtiani and Lemieux, 2019). Yet the core business of We estimate that, in 2018, fintech and big tech credit (together
fintech credit platforms remains financial services. “total alternative credit”) reached USD 694 billion globally. Big tech
Big tech firms, by contrast, have a range of business lines, (USD 397 billion) has shown particularly rapid growth in Asia
of which lending represents only one (often small) part, while (China, Japan, Korea, and Southeast Asia), and some countries in
their core business activity is typically of a non-financial nature. Africa and Latin America. In contrast, global fintech credit volumes
These firms have an existing user base, which facilitates the pro- (USD 297 billion) declined in 2018 due to market and regulatory
cess of onboarding borrowers. They can use large-scale micro-level developments in China. Outside China, fintech credit is still grow-
data on users, often obtained from non-financial activities, to miti- ing.
gate asymmetric information problems. While these large volumes We also find that such alternative forms of credit are more de-
of information allow big tech firms to effectively measure loan veloped in countries with higher GDP per capita (at a declining
quality and potentially reduce loan defaults, it is also possible rate), where banking sector mark-ups are higher and where bank-
that they could raise problems of price discrimination (Morse and ing regulation is less stringent.
Pence, 2020; Philippon, 2019), and concomitant issues for compe- Finally, we find key differences between the determinants of
tition and data privacy (Carstens, 2018; BIS, 2019; Petralia et al., fintech and big tech credit. Regulation and banking sector mark-
2019; Boissay et al., 2020). ups appear relatively more important for fintech credit. Fintech
Unfortunately, while there are well-developed systems for of- credit is more developed where there are fewer bank branches
ficial reporting of bank lending, such reporting for these new per capita. For both types of credit, we find that fintech and big
forms of digital lending is lacking. Recently, there have been tech credit are more developed where the ease of doing business
efforts to improve the data on non-bank credit to the private is greater, investor protection disclosure and the efficiency of the
sector (Dembiermont et al., 2013; FSB, 2020) and on fintech judicial system are more advanced, the bank credit-to-deposit ra-
(Serena, 2019; IFC, 2020). Central banks and public sector authori- tio is lower, and where bond and equity markets are deeper. Over-
ties use such data to monitor economic and financial conditions, to all, alternative credit seems to complement other forms of credit,
guide monetary policy decisions and to set macroprudential poli- rather than substitute for them.
cies, such as levels of countercyclical capital buffers. However, au- Our paper contributes to the literature in several ways. First,
thorities often rely on non-official sources for fintech and big tech it constructs a new global dataset on digital lending and makes
credit. Some individual fintech credit platforms voluntarily pub- it available for use by other researchers. Second, it uses panel re-
lish detailed data on their loan portfolios, but these are generally gressions rather than a simple cross-section analysis to assess the
not comparable across platforms and reporting is not standardised drivers of fintech credit (Claessens et al. (2018)) and big tech credit
across jurisdictions. (Frost et al., 2019). Third, it shows how specific institutional char-
The most comparable data on fintech credit volumes come acteristics could influence the development of these new forms
from the Cambridge center for Alternative Finance (CCAF), e.g. of digital finance, expanding the analysis beyond simple forms of
Rau (2020) and Ziegler et al. (2020). These data, based on surveys crowdfunding (Rau, 2020). In this way, the analysis helps to shed
of platforms around the world, provide annual flows of new lend- light on why some economies have experienced such dramatic
ing. They show that debt-based alternative finance (fintech credit) growth of such lending and the economic underpinnings of this
reached USD 297 billion in 2018. The economic relevance is high important transformation.
in some markets; for instance, the volume of P2P business lending The rest of the paper is organised as follows. Section 2 discusses
in the UK was estimated to be equivalent of 27.7% of all lending to the construction of our database. Section 3 explores key character-
UK small businesses in 2018. istics of fintech and big tech credit markets. Section 4 introduces
In contrast, data on big tech credit volumes are very patchy. our empirical approach on the drivers of fintech and big tech credit
Frost et al. (2019) assemble estimates of big tech credit for 2017, volumes over time. Section 5 presents our results, including the
and attempt to explain volumes in a cross-country setting. How- differences between fintech and big tech credit. Section 6 discusses
ever, big tech companies have expanded their credit offerings sub- institutional drivers of total alternative credit. Section 7 concludes
stantially since then, and lend in ever more markets to households with policy implications and avenues for future research. Some
and businesses, either directly or in partnership with financial in- methodological notes on data construction are set out in Appendix
stitutions. As we discuss below, we estimate a global volume of A and the results of robustness checks in Appendix B.
USD 397 billion in 2018. In China, such lending is equivalent to
2% of the stock of total private credit in the economy. Despite this 2. Database construction
growing importance, we are not aware of any other comparable
cross-country data sources on big tech credit. This section discusses the construction of our database. It de-
scribes the sources used, the scope and the necessary choices on
2
data aggregation and estimation, as well as external validation. Ad- active in their jurisdiction. Many of these firms lend to house-
ditional details are reported in Appendix A. holds, businesses, or both. They lend either directly or in partner-
The data on fintech credit come from the Global Alternative ship with financial institutions, using the data and networks from
Finance Database held at the CCAF. Data are collected from an their non-financial business lines. The big techs in question are
annual industry survey and web-scraping by CCAF and academic Airtel, Amazon, Alibaba/Ant Group, Au Jibun Bank, Baidu/Du Xiao-
partners (Wardrop et al., 2015; Ziegler et al., 2018; Zhang et al., man, BKash, Facebook (Meta), Fuse, Go-Jek, Google, Grab, JD.com,
2018, 2020). Firms are asked in an online questionnaire to re- Jumia, Kakao Bank, K-bank, LINE, Mercado Libre, Microsoft, MTN,
port annual alternative finance volumes, with 11 time-series re- MTS bank, Mynt, Ola Cabs, Orange, Ovo, Ozon, Rakuten, Samsung,
quired questions that serve to pinpoint exact transaction values, STC, Tencent/WeBank, Telenor, Tigo, Tokopedia, Toss, Uber, Voda-
the number of stakeholders etc. All loan-based business models fone M-Pesa, and Yandex.
are counted as fintech credit. This includes peer-to-peer (P2P) or Some big techs (particularly those that are publicly listed) file
marketplace lending to consumers, businesses or for property; bal- annual reports, which often give aggregate credit volumes (lending
ance sheet lending to consumers, businesses or for property; in- flow or, in some cases, year-end stock) for the company. Some big
voice trading, debt-based securities (debentures and bonds) and techs only lend in one country. Yet others lend in multiple mar-
mini-bonds (see Table A1 in Appendix A). Equity-based, donation- kets, and do not break down their lending flows by country. Firms
based, and reward-based crowdfunding are not included in fin- that are not publicly listed often do not even file annual reports.
tech credit. This excludes profit-sharing crowdfunding, community This lack of transparency is a key challenge, and improving disclo-
shares, pension-led funding and real estate crowdfunding, which sures by firms remains an important challenge for central banks
are counted in the broader category of alternative finance. and regulators seeking to better monitor these markets.
The data from CCAF are widely used by researchers and pol- In this paper, we attempt to fill these gaps as effectively as pos-
icymakers. Nonetheless, there are some differences in definition sible. In many cases, we refine our estimates based on information
and scope with other sources. The CCAF data do not include US provided by company and by central bank contacts in the jurisdic-
fintech mortgage lenders like Quicken Loans, Amerisave, Cashcall, tion in question. Where needed, we make a few assumptions:
Guaranteed Rate, Homeward Residential and Move Mortgage (see - In the few cases where only end-year stocks of outstanding
Buchak et al., 2018 and Fuster et al., 2018). These lenders are credit are available, we estimate the credit flow as the difference
very large, but generally originate loans for the US government- between these stocks, plus those loans assumed to have matured
sponsored enterprises (GSEs) like Fannie Mae and Freddie Mac, and over the year based on estimates of the average loan maturity
thus have a somewhat different model than lenders in other coun- (see Appendix A for the methodology used). Available numbers for
tries. Online lenders operated by commercial banks (e.g. Marcus firms with both stock and flow data suggest that this is a reason-
by Goldman Sachs) are explicitly excluded from the focus. How- able assumption.
ever, fintech lenders that work with commercial banks for origi- - Where the firm only reported its total lending over several
nation, or who sell (securitised) portfolios to banks, are included. countries, we distribute this lending volume proportionally to its
The CCAF definition aligns quite well with public sector estimates overall revenue (for all its business lines) in those countries. Where
of fintech credit or P2P lending.1 the revenue by country is not available, we distribute this lending
We hand collect data on big tech credit from contacts at cen- proportionally to the GDP of these countries. Where possible, we
tral banks and big tech firms, and from a variety of publicly avail- check that the growth in revenues was aligned with the growth in
able sources. Specifically, we sent e-mails to managers and staff- the number of unique users. While the distribution of users in fi-
level contacts in the central banks of all G20 countries (with whom nancial services business lines may differ from those of other busi-
the authors interact in the FSB Financial Innovation Network and ness lines, these checks broadly support our assumption.
other fora), and central banks and supervisory agencies in other We estimate lending volumes at the firm-specific level at an-
jurisdictions, in some cases based on e-mail introductions by oth- nual frequency, and aggregate these to country-year observation
ers. The e-mails contained a matrix to be filled in, with lending levels over 2013–18. We cross-check the firm-specific and aggre-
volumes for each big tech firm in the jurisdiction over 2013–20. gate estimates with central banks directly.
Big tech firms were asked to provide the same information for Because of the patchy nature of disclosure by big tech firms,
their own firm, and, in some cases, estimates of their own market and the assumptions needed to estimate fintech and big tech credit
share in big tech lending. These responses were collected under volumes, there will necessarily be some measurement error. In par-
agreements to keep company-specific figures confidential. In each ticular, it is possible that some big tech firms or some of their ac-
case, we also shared our country-level estimates by e-mail with tivities are not captured in our sample. Moreover, in those cases
the central banks after the estimation process was complete and where lending flows have been estimated, it may be that actual
asked if the estimates were reasonable. We received corrections lending flows differ from estimates based on end-year stocks, and
from some central banks and big tech firms, and non-objection that the distribution of firms’ activities across countries differs
comments from others. Appendix A provides further insights on from our estimation. Overall, these factors may mean that our
the process and respondents. database tends to underestimate actual alternative credit volumes
Big tech firms are defined as large companies whose primary (though volumes could be overestimated in specific country cases).
activity is technology, rather than financial services. These firms Based on the quality of available data, we restrict our analysis to
can be identified by their large user networks in areas such as e- the period 2013–18. We have higher confidence around the lending
commerce, social media, internet search and advertising, telecom- flow estimates than around the stocks. Flows are made available in
munications, etc. Globally, we are aware of 37 big techs that were the database accompanying this paper.
active in financial services at the end of 2018. This list was re-
fined with help from central bank contacts regarding big techs 3. Characteristics of fintech and big tech credit markets
This section discusses the characteristics of fintech and big tech

1
There are some exceptions. For instance, the 2018 lending volume from CCAF credit markets, lending institutions and borrowers in the countries
for Finland is slightly larger than the P2P lending volume provided by Bank of Fin-
land, and 172% of the volume of P2P lending in New Zealand provided by the Re-
where it is most developed.
serve Bank of New Zealand. Compared to the Korean P2P Lending Association, the Our database shows that fintech credit flows reached USD 297
CCAF numbers are roughly one-half. billion in 2018, while big tech credit flows surged to USD 397
3
Fig. 1. Global fintech and big tech credit over time and across countries.
billion.2 This represents a dramatic increase since 2013, when forms, which had been numerous and fast-growing in the period
volumes were only USD 9.9 billion and 10.6 billion, respectively through 2017, actually contracted their lending in 2018 and 2019,
(Fig. 1, left-hand panel). This is still small relative to global out- as a series of defaults and platform failures took their toll on the
put and global credit markets. Notably, fintech and big tech lend- sector. From a peak of 3600 fintech credit platforms in Novem-
ing flows (together “total alternative credit”) are much larger rela- ber 2015, only 343 were still in operation in December 2019, with
tive to GDP and the stock of lending in certain economies, such as steady exits but no new platform entries since September 2018
China, Kenya and Korea (Fig. 1, right-hand panel). In other major (Fig. 2, left-hand panel). New lending was only at 17% of its peak
markets like the United States, Japan and the UK, fintech and big (July 2017) level, while the stock of loans was at 46% of its (May
tech lending flows are large in absolute terms but are equal to less 2018) peak (Fig. 2, right-hand panel). Of the lending that has still
than 1% of the stock of total credit. happened, loan tenors have steadily risen, reaching 15.7 months, at
The largest market for both fintech credit and big tech credit an average interest rate of 9.5%. Because of the (previously) large
in absolute terms is China.3 Big tech companies like Alibaba’s Ant size of the Chinese fintech credit market, this has a large impact
Group, Tencent’s WeBank, Baidu’s Du Xiaoman and e-commerce on global fintech credit volumes.
platform JD.com lent USD 363 billion in 2018 (2.6% of GDP), ac- Kenya stands out as the second-largest market for big tech
cording to data provided to us by the People’s Bank of China. credit relative to the size of its economy – with lending volume
This covered a wide range of borrower types, from small busi- of USD 1.0 billion (1.1% of GDP) in 2018. This relates to both con-
nesses on Alibaba’s Taobao e-commerce platform (Ant Group) sumer and business lending through mobile money providers such
to smartphone-based consumer loans (WeBank) to rural student as Vodafone M-Pesa that began in payments and have since begun
loans (Du Xiaoman). Ant Group and WeBank, in particular, were to offer further financial serices.4 Credit is offered through mo-
able to make use of the extensive payments data from their mobile bile interfaces using the existing network and user data from the
payment services to price credit (see Frost et al., 2019). In some mobile network, and is generally at short tenors. Fintech credit is
cases, this lending was on the balance sheet of the big tech or its much smaller, at USD 32 million.
financial services subsidiary, with funding from wholesale markets. In Korea, big tech credit (lending flow of USD 8.0 billion or 0.5%
In other cases, lending was in partnership with banks, through an of GDP in 2018) is provided by two major firms – KakaoBank and
originate-to-distribute model. Meanwhile, fintech credit (P2P) plat- KBank – both of which are full subsidiaries of big tech companies,
namely messaging platform Kakao and telecommunication com-
pany KT, respectively. Both were launched in 2017 with a virtual
2
Throughout the paper, fintech and big tech credit volumes refer to the flow banking license, which offers a tailored regulatory framework.5
of new lending over a calendar year. This differs from the standard methods for
reporting bank credit, which are end-year stocks of loans outstanding. A quirk of
4
the fintech and big tech lending market is that firms are more likely to report the Available information provided to the authors by Vodafone M-Pesa shows that
accumulative loan flow over a year or since the inception of their business, rather such lending expanded significantly in 2019 and 2020.
than a current outstanding loan book. For more details, see Appendix A. 5
As of end-2018, these were the only two firms with virtual banking licenses in
3
The list of top 30 countries for total flow of alternative credit in the period Korea. Outside our sample period, fintech payments firm Toss received the country’s
2013-2018 is presented in Table A2 of Appendix A. third virtual banking license.
4
Fig. 2. Fintech credit platforms in China.
Their activities are fundamentally different than those of incum- In Japan, big tech lenders are quite active (USD 19.0 billion of
bent banks, as they have no branches, fully-digital interfaces and lending in 2018, or 0.4% of GDP), while fintech credit is small. In
a strong operational link with the other business lines of the big particular, Rakuten has offered a suite of financial products since
tech group. They offer financial services that are integrated into 2013, including payments, transaction lending, credit card issuing
their parent company’s “eco-system”. These firms offer loans with and acquiring, mortgages and insurance. Meanwhile, social media
an average maturity of roughly one year to users of their respective company LINE offers consumer lending (through a joint venture
networks. They benefit from user data and existing customer rela- with Mizuho Bank and a credit card company). Telecommunica-
tionships from the big techs’ non-financial business lines. Informa- tion firm NTT DoCoMo provides customer credit-scoring services
tion on their lending is provided publicly by the Financial Super- (upon contractual agreement with banks and customers) and Ama-
visory Service and Bank of Korea. Separately, Korea has a growing zon lends through its seller lending program.8 Fintech credit (esti-
fintech credit sector, dominated by P2P/marketplace property lend- mated at USD 920 million in 2018) in Japan is primarily through
ing. Fintech credit reached an estimated USD 550 million in 2018. P2P/marketplace business and property lending.9
The United States has a large market for fintech credit in abso- The UK had estimated fintech credit volumes of USD 11.5 bil-
lute terms, but big tech credit volumes are relatively small – cer- lion in 2019 (up from USD 9.3 billion in 2018), made up of a mix
tainly compared with the economy’s deep credit markets. Fintech of P2P/marketplace business, consumer and property lending, and
credit reached USD 57.7 billion in 2018 (0.3% of GDP). This was smaller volumes of balance sheet lending and invoice trading. Af-
made up primarily of P2P/marketplace consumer lending, with inter rapid growth in 2013–16, fintech lending volumes have been
vestment coming predominantly from institutional investors rather relatively steady in the United Kingdom in the past three years,
than individual lenders. It came in large part from platforms like perhaps reflecting greater maturity and saturation in the relevant
Lending Club, SoFi, Prosper and OnDeck. These platforms often market segments. For instance, Ziegler et al. (2020) estimate that
partner with financial institutions, originating loans that are sold fintech credit platforms accounted for up to 27.7% of equivalent
on to banks and other institutional investors.6 While the United bank credit to small and medium enterprises with annual turnover
States is home to many of the largest big tech companies in the below GBP 2 million in 2018. This may have been encouraged by
world, only Amazon engaged in any significant lending in 2018, to public policy; for instance, the government-owned British Busi-
the tune of roughly USD 1 billion, according to public reporting.7 ness Bank invested over GBP 165 million over 2014–18 for lend-
ing through Funding Circle, a UK credit platform, and announced a
commitment for a further GBP 150 million to support small busi-
ness lending (British Business Bank, 2018). Big tech credit volumes
6
Another large group of lenders, not captured in our fintech credit data, are are estimated to be much smaller, at an estimated USD 100 mil-
fintech mortgage lenders such as Quicken Loans, Amerisave, Cashcall, Guaranteed lion in 2017 and 2018, primarily through Amazon’s Seller Lending
Rate, Homeward Residential and Move Mortgage, which often originate loans for program.
the government-sponsored enterprises (GSEs) like Fannie Mae and Freddie Mac. See Looking beyond the largest fintech and big tech credit mar-
Buchak et al. (2018) and Fuster et al. (2018).
7
kets, higher-frequency data from Brismo and WDZJ show that fin-
As of 2019, Apple launched its Apple Card in cooperation with Goldman Sachs,
which had outstanding balances of USD 7 billion by year-end. Since this is out-
side our sample period, Apple is not included in our sample, though it is a poten-
8
tial big tech entrant in future years. Plans by Google to offer a checking account Outside the period of analysis, the second largest telecommunications provider,
product, in conjunction with Citi, Stanford Federal Credit Union and several other KDDI, has a joint venture with MUFG Bank called Au Jibun Bank. On 1 April 2019,
banks, were announced but then abandoned. Uber Money began to offer payment when it became a consolidated subsidiary of KDDI. The announcement emphasized
and wallet products to its drivers, but executives noted that lending was not yet on that the bank would be able to benefit from the big data and user network of KDDI
the roadmap (Shevlin, 2019). A cooperation between Amazon and Goldman Sachs – consistent with the characteristics of several other big tech lenders.
9
on small business lending may be significant in the future but was not yet in oper- The stock of fintech and big tech credit is estimated to have reached USD 34.3
ation in 2018. billion at the end of 2019.
5
Fig. 3. Geographical comparison of fintech and big tech credit development.
tech credit volumes have continued to grow rapidly in the Eu- more developed economies (with higher GDP per capita) will have
ropean Union, Australia and New Zealand, even as they have a higher demand for credit from firms and households, but that
plateaued in the United States and United Kingdom and declined in this relationship may show a decreasing trend for very high lev-
China (Fig. 3). In many emerging market and developing countries els of development (see Claessens et al., 2018; Frost et al., 2019;
(not shown), fintech lenders are becoming economically significant Bazarbash and Beaton, 2020). Similarly, we expect that fintech and
lenders for specific segments, such as small and medium-sized en- big tech credit will be higher when incumbent banking services
terprises (SMEs) (Cornelli et al., 2019; World Bank, 2020). Some are more expensive (higher banking sector mark-ups), for instance,
fulfill so-called agency banking functions, by which they function because of higher market power in the banking sector, and where
as agents to expand the reach of banks, especially in Latin America there is a larger un(der)met demand for financial services, as prox-
and parts of Asia and Africa. ied by fewer bank branches per capita.
Big tech credit is achieving economically significant scale in On the institutional side, we expect that a more stringent bank-
China, Kenya, Korea, parts of Southeast Asia, East Africa and (to a ing regulation (a proxy for the overall stance of financial regu-
lesser extent) some countries in Latin America (Fig. 3, right-hand lation) may create barriers to the entry for fintech and big tech
panel). This is driven by the lending activities of e-commerce plat- firms. Several additional institutional characteristics, such as the
forms like Mercado Libre, ride-hailing companies like Grab and Go- ease of doing business, investor protection and disclosure, the ju-
Jek, and telecommunication and mobile money providers like M- dicial system and characteristics of the incumbent banking system
Pesa. In many cases, these lenders initially target a specific group will be discussed later, and refer to the quality of domestic institu-
of users (e.g. sellers on the e-commerce platform, or drivers) but tions to support innovative credit activities. We also examine the
then expand their credit offerings to more users over time. open question of whether fintech and big tech credit complement
or substitute for bank credit and other forms of finance.10
Our baseline regression takes the form:
4. Drivers of credit volumes: a panel approach
In our core empirical analysis, we consider the determinants of

ln (Creditit ) = α + β1 yi,t−1 + β2 y2i,t−1 + γ LIi,t−1 + δ RSi,t−1
fintech and big tech volumes in different economies over time. This + μBNi,t−1 + σ Xi,t−1 + ϑ Dk + εi (1)
is a novelty with respect to earlier studies (Claessens et al., 2018;
Frost et al., 2019) that analyze such volumes in the cross section. where Creditit is the volume of fintech or big tech credit per capita
Leveraging on the new database and following Rau (2020), we ex- in economy i at time t, or total alternative credit. The three credit
tend 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 the most signif-
10
icant determinants of total alternative credit volume (the sum of For Germany, De Roure et al. (2016) find that P2P lending substitutes bank
fintech and big tech credit). credit 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
We hypothesize that fintech and big tech credit per capita can States, Tang (2019) finds that P2P lending is a substitute for bank lending in terms
be broadly related to: (i) macroeconomic factors and (ii) institu- of serving infra-marginal bank borrowers, but that it complements bank lending
tional characteristics. On the macroeconomic side, we expect that with respect to small loans.
6
Table 1
Descriptive statistics.
Variable Observations Mean Standard deviation Min Max
GDP per capita (in thousands of USD) 453 21.53 18.21 0.67 87.76
Lerner index1 453 0.30 0.15 –0.05 1.00
Bank branches per 100,000 adults 453 17.65 14.03 1.43 83.75
Normalised index of bank regulatory stringency2 453 0.72 0.10 0.38 0.96
Score-Starting a business (overall) 425 82.55 11.46 23.04 99.96
Score-Time (days) 425 81.96 17.33 0.00 100.00
Score-Paid-in Minimum capital (% of income per capita) 425 94.85 15.45 0.00 100.00
Score-Cost (% of income per capita) 425 84.92 26.95 0.00 100.00
Extent of disclosure index (0–10) 425 64.44 23.80 0.00 100.00
Trial and judgment (days) 425 407.85 203.51 90.00 1095.00
Enforcement of judgment (days) 425 177.98 110.59 26.00 597.00
Enforcement fees (% of claim) 425 5.37 5.23 0.00 23.30
Bank credit to bank deposits (%) 212 105.39 80.48 27.73 702.09
Bank regulatory capital to risk-weighted assets (%) 197 17.20 3.81 10.59 35.65
Provisions to non-performing loans (%) 187 64.29 37.26 0.00 232.06
Loans from non-resident banks to GDP (%) 194 27.53 28.33 1.24 158.53
Proportion of firms with a transactions account (%) 276 85.48 15.18 18 100
Corporate bond average maturity (years) 135 10.26 5.75 3.54 34.09
Corporate bond issuance volume to GDP (%) 137 2.14 1.81 0.05 13.83
Total factoring volume to GDP (%) 145 5.16 4.90 0.07 16.29
Global leasing volume to GDP (%) 78 1.32 0.95 0.01 4.81
Stock market total value traded to GDP (%) 167 50.64 89.08 0.00 562.92
Stock market turnover ratio (%) 161 53.61 67.37 0.84 556.91
Ln(Total alternative credit per capita (in USD)3 ) 453 0.93 1.43 –1.97 5.11
Ln(Big tech credit per capita (in USD)) 453 0.09 0.99 –3.57 4.55
Ln(Fintech credit per capita (in USD)) 453 –1.14 2.76 –7.20 4.81
The table shows the descriptive statistics for the variables used in the regressions. Ln = natural logarithm. The dependent variables are winsorized at the 1% and 99% level.
1
The Lerner index of banking sector mark-ups in economy i reflects market power by incumbent banks. World Bank data. For 2015–17, data are estimated based on Igan et
el (2020).
2
The index is normalised between 0 (no regulation) and 1 (max regulation). The index is calculated from a survey conducted by the World Bank in given years, and
therefore data are not available over the whole sample period, but proceed in steps. See https://datacatalog.worldbank.org/.
3
Defined as the sum of big tech and fintech credit.
Sources: IMF, World Economic Outlook, World Bank; Cambridge center for Alternative Finance and research partners; company statements; authors’ estimates/.
aggregate volumes serve as our left-hand side variables, each with Bank branches and mobile subscriptions are measured relative to
the same regressors.11 the adult population.14 Dk is a vector of geographical area fixed ef-
Several regressors on the right-side equation above are lagged fects and εi,t is an error term.15
by one year to mitigate endogeneity issues. yi,t−1 is the GDP per Table 1 reports descriptive statistics for our sample of 79 coun-
capita in economy i at year t-1, and the variable y2i,t−1 is its tries over the period 2013–18. Given the lower coverage of many
quadratic term, to address the non-linear relationship between variables for 2019, we have excluded this year from regressions.
credit development and income levels. LIi,t−1 is the Lerner in- Data come from a variety of sources, including the IMF’s World
dex12 of banking sector mark-ups in economy i, reflecting mar- Economic Outlook and the World Bank’s Global Financial Develop-
ket power by incumbent banks; a higher value may reflect a less ment Database (GFDD) and Findex.
competitive banking sector. RSi,t−1 is an index of regulatory strin-
gency for the banking sector of economy i, as constructed by Barba 5. Economic drivers of fintech and big tech credit
Barba Navaretti et al. (2017) from World Bank data.13 BNi,t is the
density of the bank branch network in country i relative to the Our panel regression results show that fintech and big tech
adult population (which may capture both the reach of the bank- lending volumes are significantly associated with several key
ing sector and its relative cost base). Xi,t is a vector of control economic and institutional drivers. The first three columns of
variables including the growth in GDP and total credit; a real Table 2 report estimations for total alternative credit (big tech plus
short-term interest rate; a dummy for whether a country had suf- fintech credit; column 1), big tech credit (column 2), and fintech
fered a financial crisis since 2006, as defined by Laeven and Va- credit (column 3) respectively. Column 4 reports formal tests for
lencia (2018); mobile phone subscriptions (given the mobile-based differences between the coefficients for big tech credit and fintech
nature of many platforms); and a dummy for advanced economies. credit.
The first key driver is overall economic development. As in
Claessens et al. (2018) and Frost et al. (2019), we find that to-
11
Frost et al. (2019) refer to the sum of fintech and big tech credit as “total fintech tal alternative credit activity, as well as its two component parts
credit”. Here, to prevent confusion, we refer to the sum of fintech and big tech (big tech and fintech credit), is positively associated with GDP per
credit as “total alternative credit”. We also regress bank credit per capita using the
capita, but at a declining rate. Since GDP per capita is likely to be
same specification to check for significant differences.
12
The Lerner Index of banking sector mark-ups has been updated over the pe-
a proxy for many aspects of a country’s stage of development, this
riod 2015–17 using information on the alternative cyclical measure devolped by confirms a positive relationship between a country’s overall devel-
Igan et al. (2020). Appendix A provides details. A higher value indicates higher mar-
gins and profitability among traditional banks, and thus less competition.
13 14
The regulatory stringency variable is constructed as an index (normalised be- When observation for bank branches, mobiles and credit growth were not yet
tween 0 and 1) based on the World Bank Bank Regulation and Supervision Survey. available, we extrapolated the figures using the cross-country average growth rate
The index takes a value between 0 (least stringent) and 1 (most stringent) based or the cross-country average change.
15
on 22 questions (2011 survey) or 23 questions (2019 survey) about bank capital The inclusion of some (barely) time-invariant country-specific regressors (see
requirements, disclosure, the legal powers of supervisory agencies etc. next section) prevents us from using a complete set of country dummies.
7
Table 2
Drivers of fintech and big tech credit volumes.
All variables are expressed in current USD, except where indicated

Ln(total alternative credit per capita) Ln(big tech credit per capita4 ) Ln(fintech credit per capita5 ) Difference │b-a│
(a) (b) H0 : │b-a│<0
Variables of core interest

GDP per capita1 0.123∗ ∗ ∗ 0.069∗ ∗ ∗ 0.171∗ ∗ ∗ 0.102∗ ∗ ∗
(0.022) (0.020) (0.038) (0.043)
GDP per capita^2 –0.002∗ ∗ ∗ –0.001∗ ∗ ∗ –0.002∗ ∗ ∗ 0.001∗ ∗
(0.000) (0.000) (0.001) (0.001)
Lerner index2 1.438∗ ∗ ∗ 0.867∗ ∗ 2.436∗ ∗ ∗ 1.569∗ ∗
(0.401) (0.365) (0.732) (0.818)
Bank branches per 100,000 –0.017∗ ∗ ∗ 0.005 –0.028∗ ∗ ∗ 0.033∗ ∗ ∗
adult population (0.005) (0.005) (0.009) (0.010)
Normalised regulation –4.665∗ ∗ ∗ –1.735∗ ∗ ∗ –8.427∗ ∗ ∗ 6.692∗ ∗ ∗
index3 (0.560) (0.544) (1.068) (1.199)
Other controls
GDP growth 0.034∗ 0.005 0.046 0.041
(0.019) (0.016) (0.039) (0.042)
Crisis dummy 0.221 0.106 –0.206 0.099
(0.234) (0.217) (0.326) (0.392)
Total bank credit growth to the –0.000 0.004 –0.002 0.001
private non-financial sector (0.005) (0.005) (0.008) (0.010)
Mobile phone –0.001 –0.001 0.002 0.001
subscriptions (0.002) (0.002) (0.004) (0.004)
Dummy advanced 0.681∗ ∗ –0.194 2.088∗ ∗ ∗ 1.894∗ ∗ ∗
economy (0.311) (0.339) (0.481) (0.589)
Real interest rate 0.021∗ ∗ ∗ 0.006∗ 0.035∗ ∗ ∗ 0.029∗ ∗
(0.007) (0.003) (0.012) (0.013)
Geographic area fixed effects6 Yes Yes Yes
No. of observations 453 453 453
R2 0.469 0.112 0.516
The table shows the results of a panel regression of the natural logarithm of total alternative credit per capita, (column 1), fintech credit per capita (column 2) and big tech
credit per capita (column 3) on the independent variables reported in the row headings. The last column reports the results of test for differences between the coefficients
for big tech and fintech credit. Estimation period 2013–18. Robust standard errors in parentheses. ∗ ∗ ∗ /∗ ∗ /∗ denotes results significant at the 1/5/10% level. Ln = natural
logarithm. The dependent variables are winsorized at the 1% and 99% level.
1
GDP per capita (in USD thousands).
2
Lerner index of banking sector mark-ups in economy i, reflecting market power by incumbent banks.
3
The index is normalised between 0 (no regulation) and 1 (max regulation).
4
Big tech credit is zero in 47 countries. To allow the computation of the log of the ratio (not defined for zero), big tech credit has been rescaled summing an arbitrary
constant (the minimum value).
5
Fintech credit is defined as credit activity facilitated by electronic platforms that are not operated by commercial banks or big tech firms.
6
The sample is divided into five geographical areas: Africa, Asia Pacific, Europe, Latin America, Middle East and North America.
Sources: CCAF; IMF, World Economic Outlook; World Bank; authors’ calculations.
opment and innovative credit activities. The negative coefficient on fintech credit serves clients in underbanked areas and that it is
GDP per capita squared suggests that the link becomes less impor- therefore complementary to traditional bank credit. This is also
tant at higher levels of development. The difference between the consistent with the use of agency banking. Big tech credit, while
estimated coefficient for fintech and big tech credit for GDP per also relying on digital distribution channels rather than physical
capita is statistically significant, meaning that the relationship is branches, does not appear to be correlated with the number of
stronger for fintech than big tech credit. As we show later, these bank branches relative to the adult population, all else equal. This
differences are also detected in the cross-sectional analysis.16 result could depend upon the global nature of big techs business
Second, fintech and big tech credit are more developed, other models that has the potential to reduce the link with domestic
things being equal, in those jurisdictions with higher banking bank distribution conditions.
mark-ups. Both forms of credit are likely to be more flexible than Fourth, the coefficient of the stringency of banking regula-
bank credit, and that the ease and speed of receiving the loan de- tion is negative for both forms of credit: more stringent bank-
cision is higher – an element that could be particularly important ing regulation is significantly linked to less big tech and fin-
for underbanked clients. It may also be that high margins make en- tech credit activity. Banking regulation explains around 10% of
try more attractive for the fintech and big tech firms, themselves. the variability of total alternative credit per capita in the baseline
Bank mark-ups explain around 5% of the variability of total alter- model, and contributes to more than one fifth of the R2 . This re-
native credit per capita (the overall R2 is 46.9%). The elasticity of sult, similar to that found by Barba Barba Navaretti et al., 2017,
big tech credit to the Lerner index is lower than for fintech credit. Claessens et al. (2018) and Frost et al. (2019), could have several
As a third driver, the density of the bank branch network is possible explanations. This could suggest that regulation of alterna-
negatively correlated with the development of fintech credit, but tive forms of credit in general is more liberal in jurisdictions where
not of big tech credit. The difference between the two coefficients banking regulation is more liberal. Conversely, it may be more dif-
is statistically significant. This is consistent with the view that ficult to launch new lending activities in countries with relatively
strict prudential and bank licensing regimes. The elasticity of big
tech credit to the regulatory index is significantly lower than for
16
These differences are not due to a different average size of fintech and big tech
fintech credit. The quantitative effects of regulation are also eco-
loans. Both forms of credit are typically granted to households and small enterprises nomically significant. Specifically, a country with an index that is
and are of modest average loan size. See Gambacorta et al. (2020) for the case of one standard deviation lower in the cross section (looser regula-
China.
8
Table 3
Drivers of total alternative credit volumes (region and year fixed effects).

Ln(total alternative credit)
Baseline model with Panel with Pooled System GMM

regional country cross-section with regional
fixed effects fixed effects regression fixed effects
(i) (ii) (iii) (iv)
GDP per capita1 0.123∗ ∗ ∗ 0.502∗ ∗ ∗ 0.139∗ ∗ ∗ 0.046∗ ∗ ∗

(0.022) (0.109) (0.032) (0.010)
GDP per capita^2 –0.002∗ ∗ ∗ –0.002∗ ∗ –0.002∗ ∗ ∗ –0.001∗ ∗ ∗
(0.000) (0.001) (0.000) (0.000)
Lerner index2 1.438∗ ∗ ∗ 0.646 1.223 0.558∗ ∗ ∗
(0.401) (0.903) (0.777) (0.190)
Bank branches per 100,000 –0.017∗ ∗ ∗ –0.091∗ ∗ ∗ –0.014 –0.004∗ ∗
Normalised regulation –4.665∗ ∗ ∗ –0.743 –4.428∗ ∗ ∗ –1.263∗ ∗ ∗
index3 (0.560) (0.911) (1.491) (0.412)
Lagged dependent variable 0.934∗ ∗ ∗
(0.038)
Other controls4 Yes Yes Yes Yes
Geographic area fixed effects5 Yes No Yes Yes
Country fixed effects No Yes No No
No. of observations 453 449 79 453
R2 0.469 0.824 0.578
Serial correlation test6 0.370
Hansen test7 0.085
The table shows the results of four different regression models: our baseline panel with regional fixed effects (column (i)), a panel with country fixed effects (columns (ii)),
a pooled cross-section regression (column (iii)), a system GMM (column (iv)). The dependent variable is the natural logarithm of total alternative credit per capita. It is
winsorized at the 1% and 99% level. The system GMM is instrumented with one to three lags of the dependent variable and one lag of the independent variables. Estimation
period 2013–18. Robust standard errors in parentheses. ∗ ∗ ∗ /∗ ∗ /∗ denotes results significant at the 1/5/10% level.
1
GDP per capita, in USD thousands.
2
3
4
Other controls include: GDP growth; a crisis dummy that takes a value of 1 if the country experienced a financial crisis since 2006 and 0 elsewhere; total bank credit
growth to the private non-financial sector; mobile phones per 100 persons; and country-specific real interest rates.
5
6
Reports p-values for the null hypothesis that the errors in the first difference regression exhibit no second order serial correlation.
7
Reports p-values for the null hypothesis that the instruments used are not correlated with the residuals.
Sources: CCAF, authors’ calculations.
tion) has a ratio of total alternative credit per capita that is 0.5 nificance of the coefficients are stable. Table B1–B3 in Appendix B
percentage points higher.17 Six other variables are included as con- report the results for models (ii)-(iv) on fintech and big tech credit
trols.18 Overall, our estimations can explain 51.6% of the variation separately
of fintech credit but only 11.2% of the variation of big tech credit. Model (i) includes regional fixed effects. Most of the coun-
This may relate to the smaller number of countries and years in try controls have a limited variability during the estimation pe-
which big tech credit is present, and the fact that big tech firms riod, preventing us from using country fixed effects in the base-
operate in different countries. This could reduce the capacity of line model. Despite this fact, model (ii) includes country fixed
domestic controls to capture the global nature of big tech business effects. As expected, the normalised regulation index retains the
models. same sign but becomes insignificant, because this control variable
Our results are qualitatively very similar when looking at alter- does not change very much over time in each country and it is
native estimation methods. Table 3 reports the results for total al- captured in this model by the country-specific fixed effects. Model
ternative credit comparing: (i) the baseline model (1) with regional (iii) is a simple cross section, obtained by averaging all values over
fixed effects; (ii) a panel with country fixed effects; (iii) a pooled 2013–18. This model confirms that our results are also robust to
cross-section regression; and (iv) a system generalised method of analysing only the “within” variability across countries. To account
moments (GMM) with regional fixed effects. The sign and the sig- for possible reverse causality issues, we also use a dynamic GMM
panel (see e.g. Arellano and Bond (1991) and Blundell and Bond
(1998)). In model (iv) the inclusion of the lagged dependent vari-
17
Unfortunately, limited information does not allow us to properly investigate the able and the use of instruments do not qualitatively change the re-
effects of regulation that is specific to fintech and big tech credit. Surveys from sults. Both the sign and size of the coefficients of interest are very
Rowan et al. (2019) and Ehrentraud et al. (2020) provide relevant insights on reg- similar to those obtained for model (i), (ii) and (iii).
ulatory frameworks in 2019. However, this is later than the sample period for our
regressions, which ends in 2018 and the surveys do not report systematically when
The results are also robust to the inclusion of a complete set of
these regulations were introduced. Regulations that were only enacted shortly be- time dummies that, however, tend to capture the common global
fore the survey would not be expected to influence fintech and big tech credit vol- trend in the evolution of these forms of credit (see Table B4 in
umes over 2013-2018. Rau (2020) estimates the year that dedicated frameworks for Appendix B). The results are also robust to using the log of bank
debt and equity crowdfunding were enacted and does find a significantly positive
credit to the private non-financial sector per capita, rather than
link with actual volumes one year later. This remains an important area for further
investigation. credit growth, as a control (Table B5 in Appendix B).19
18
Among the additional controls, the real interest rate and advanced economy
dummy have positive and significant coefficients, and GDP growth has a positive
19
and significant coefficient for total alternative credit. The coefficients for the crisis Our results are also robust to a number of additional checks. For example, our
dummy, mobile phones per 100 persons and credit growth are insignificant. results are very similar to using the log of GDP per capita rather than GDP per
9
Table 4
Drivers of total alternative credit – Ease of doing business indicators.

Ln(total alternative credit per capita)
GDP per capita1 0.108∗ ∗ ∗ 0.112∗ ∗ ∗ 0.124∗ ∗ ∗ 0.112∗ ∗ ∗ 0.107∗ ∗ ∗

(0.023) (0.023) (0.023) (0.024) (0.024)
GDP per capita^2 –0.001∗ ∗ ∗ –0.002∗ ∗ ∗ –0.002∗ ∗ ∗ –0.001∗ ∗ ∗ –0.001∗ ∗ ∗
(0.000) (0.000) (0.000) (0.000) (0.000)
Lerner index2 1.475∗ ∗ ∗ 1.520∗ ∗ ∗ 1.639∗ ∗ ∗ 1.518∗ ∗ ∗ 1.470∗ ∗ ∗
(0.418) (0.421) (0.416) (0.418) (0.419)
Bank branches per 100,000 –0.019∗ ∗ ∗ –0.020∗ ∗ ∗ –0.019∗ ∗ ∗ –0.019∗ ∗ ∗ –0.019∗ ∗ ∗
adult population (0.005) (0.005) (0.005) (0.005) (0.005)
Normalised regulation index3 –4.570∗ ∗ ∗ –4.604∗ ∗ ∗ –4.816∗ ∗ ∗ –4.684∗ ∗ ∗ –4.600∗ ∗ ∗
(0.605) (0.607) (0.599) (0.600) (0.607)
Score starting a business (overall) 0.010∗ ∗
(0.005)
Score – Time (days) 0.006∗ ∗
(0.003)
Score – Paid-in Minimum capital (% of 0.004∗ ∗
income per capita) (0.002)
Score – Cost (% of income per capita) 0.005∗
(0.003)
First principal component4 0.080∗ ∗
(0.038)
Other controls5 Yes Yes Yes Yes Yes
Geographic area fixed effects6 Yes Yes Yes Yes Yes
No. of observations 425 425 425 425 425
R2 0.459 0.459 0.466 0.457 0.459
The table shows the results of a panel regression of the natural logarithm of total alternative credit per capita
on the independent variables reported in the row headings. Estimation period 2013–18. Robust standard errors in
parentheses. ∗ ∗ ∗ /∗ ∗ /∗ denotes results significant at the 1/5/10% level. Ln = natural logarithm. The dependent variable
is winsorized at the 1% and 99% level.
1
2
3
4
The loadings are 0.63, 0.48, 0.48, and 0.38 respectively.
5
Other controls include: GDP growth; a crisis dummy that takes a value of 1 if the country experienced a financial
crisis since 2006 and 0 elsewhere; total bank credit growth to the private non-financial sector; mobile phones per
100 persons; a dummy that takes a value of 1 for advanced economies and 0 elsewhere; and country-specific real
interest rates.
6
The sample is divided into five geographical areas: Africa, Asia Pacific, Europe, Latin America, Middle East and
North America.
A key question is the extent to which fintech and big tech bank branches is, as expected, opposite. Finally, total alternative
credit differ from bank credit. Table B6 in Appendix B reports re- credit appears to be more sensitive to the normalised regulation
sults when the log of the stock of bank credit to the private sector index than bank credit.
is considered as a dependent variable. We consider both a sim- As an additional check we consider the impact of explicit
ple cross section (model (a)) and a panel (model (b)). The nega- fintech regulation. Following Rau (2020), we include in model (1)
tive correlation with the Lerner index of banking sector mark-ups a dummy variable that takes a value of 1 if an explicit regulation
suggests that, other things being equal, bank credit is more devel- of fintech credit (“crowdfunding debt models”) was in place in
oped in those jurisdictions with a more competitive banking sec- a given country and year, and 0 elsewhere. During our sample
tor. This underscores how fintech and big tech credit complement period, 21 countries introduced explicit regulation for fintech
bank credit. As expected, the density of the bank branch network is credit. The results in Table B7 in Appendix B indicate a positive
positively correlated with the development of bank credit. Finally, correlation between explicit regulation and fintech credit. Due
more stringent banking regulation is significantly associated with to potential endogeneity issues, we do not claim a clear causal
lower bank credit activity. The final column of Table B6 compares relationship from the introduction of a fintech regulation to
the coefficients for the panel model on bank credit (model b) with greater fintech credit volumes; indeed, regulatory initiatives could
those obtained for the panel model on total alternative credit. The take place as a specific reaction to the development of fintech
elasticity of bank credit to GDP per capita is significantly higher credit markets. By clarifying the institutional environment, the
than that of total alternative credit. The elasticities of these two regulations might reduce the risk that new activities started by
forms of credit with respect to the Lerner index of banking sector the fintech firms are eventually prohibited. Ran et al. (2022) also
mark-ups have opposite signs, because fintech and big tech credit document a positive relation between regulatory clarity and fin-
appear to develop where bank credit competition is lower. The ef- tech volume using a staggered difference-in-difference analysis
fect detected on the two forms of credit due to the presence of and a propensity matching approach to establish at least partial
causality. For the purposes of this study, what is important is that
the main results of the study remain unaffected when controlling
capita and its square term, or when simply excluding the GDP per capita squared
for explicit regulation of fintech credit.
term. However, in both cases we observe a significant reduction in the R2 of the As a final test, we also check for differences between big tech,
regression for the log of big tech credit per capita (from 0.112 to 0.075) and a drop fintech and bank credit using the ratios between the different
of the significance of some of the coefficients (not reported for the sake of brevity credit types as dependent variables in equation (1). As we do not
but available from the authors upon request).
10
Table 5

Ln(total alternative credit per capita)
GDP per capita1 0.109∗ ∗ ∗ 0.108∗ ∗ ∗ 0.110∗ ∗ ∗ 0.113∗ ∗ ∗ 0.095∗ ∗ ∗

(0.024) (0.023) (0.025) (0.023) (0.025)
GDP per capita^2 –0.001∗ ∗ ∗ –0.001∗ ∗ ∗ –0.002∗ ∗ ∗ –0.002∗ ∗ ∗ –0.001∗ ∗ ∗
(0.000) (0.000) (0.000) (0.000) (0.000)
Lerner index2 1.428∗ ∗ ∗ 1.391∗ ∗ ∗ 1.647∗ ∗ ∗ 1.525∗ ∗ ∗ 1.467∗ ∗ ∗
(0.418) (0.410) (0.413) (0.423) (0.406)
Bank branches per 100,000 –0.019∗ ∗ ∗ –0.017∗ ∗ ∗ –0.017∗ ∗ ∗ –0.020∗ ∗ ∗ –0.016∗ ∗ ∗
adult population (0.005) (0.005) (0.005) (0.005) (0.005)
Normalised regulation index3 –4.541∗ ∗ ∗ –4.495∗ ∗ ∗ –4.678∗ ∗ ∗ –4.544∗ ∗ ∗ –4.529∗ ∗ ∗
(0.605) (0.603) (0.610) (0.607) (0.609)
Extent of disclosure index (0–10) 0.004∗
(0.002)
Trial and judgment (days) –0.001∗ ∗ ∗
(0.000)
Enforcement of judgment –0.001∗ ∗
(days) (0.001)
Enforcement fees (% of claim) –0.018∗
(0.010)
First principal component4 –0.168∗ ∗ ∗
(0.043)
Other controls5 Yes Yes Yes Yes Yes
Geographic area fixed effects6 Yes Yes Yes Yes Yes
No. of observations 425 425 425 425 425
R2 0.460 0.466 0.462 0.459 0.471
The table shows the results of a panel regression of the natural logarithm of total alternative credit per capita
on the independent variables reported in the row headings. Estimation period 2013–18. Robust standard errors in
parentheses. ∗ ∗ ∗ /∗ ∗ /∗ denotes results significant at the 1/5/10% level. Ln = natural logarithm. The dependent variable
is winsorized at the 1% and 99% level.
1
2
3
4
The loadings are –0.46, 0.63, 0.61, and 0.14 respectively.
5
Other controls include: GDP growth; a crisis dummy that takes a value of 1 if the country experienced a financial
crisis since 2006 and 0 elsewhere; total bank credit growth to the private non-financial sector; mobile phones per
100 persons; a dummy that takes a value of 1 for advanced economies and 0 elsewhere; and country-specific real
interest rates.
6
The sample is divided into five geographical areas: Africa, Asia Pacific, Europe, Latin America, Middle East and
North America.
have information on annual bank credit flows, we approximate Bank, 2019). Descriptive statistics for these variables are included
them using the annual changes in stocks. The results are reported in Table 1.20
in Table B8 in Appendix B. Both big tech and fintech credit re- First, we evaluate whether potential barriers to entry, such as
act more to GDP per capita changes than bank credit but at a restrictions on starting a new business, could affect the develop-
decreasing rate. In case of an increase in competition (a drop in ment of total alternative credit. Easier procedures to start a busi-
the Lerner index), big tech credit tends to grow more than fintech ness may allow new firms – such as those that sell products on
credit, while the latter grows less than bank credit. An increase in e-commerce platforms and use fintech or big tech credit – to en-
the number of bank branches reduces the growth of fintech credit ter the market, thus increasing the demand for credit. Easier pro-
with respect to bank credit, while no significant effect is detected cedures may also allow fintech and big tech intermediaries to
for big tech credit. Finally, a tightening in regulation pushes big emerge, or foreign firms to enter these markets, thus increasing
tech credit to develop more than fintech credit, but there is no sig- the fintech and big tech credit supply.
nificant effect when the two forms of credit are compared to bank In Table 4, we add to the baseline specification for total alter-
lending. native credit a number of indicators that measure how easy is to
open a new business. Each indicator ranges from 0 (difficult to
open a business) to 10 (maximum ease). In particular, we consider
6. The role of institutional characteristics one at a time (to avoid collinearity problems) the following mea-
sures: (i) the overall score for the ease to start business; (ii) a spe-
In this section, we include additional country characteristics in cific score based on the median duration to complete the proce-
the baseline model (1). This represents a novelty with respect to dure for creation of a firm; (iii) a specific score based on the mini-
the literature so far, which focuses primarily on cross-sectional mum capital required for an entrepreneur to start up and formally
analysis and a very limited number of explanatory variables. While operate a business; and (iv) a specific score based on overall costs
we continue to use the full range of country-year observations, we officially required for an entrepreneur to start up and formally op-
focus on the impact of specific indicators that could be highly cor- erate an industrial or commercial business. In 2013 and 2014, these
related with one another. To avoid multicollinearity problems, we
therefore include these relevant country-specific characteristic one
at a time. In particular, we include both variables proxying for bar- 20
We note that institutional factors and social arrangements may influence not
riers to entry, as expressed by the ease of doing business variables only the volume of alternative credit, but also the form that such credit takes
and investor disclosure and efficiency of the judicial system (World (Wardrop, 2020). However, this analysis is beyond the focus of this paper.
11
Table 6

Ln(total alternative credit per capita) Ln(big tech credit per capita6 ) Ln(fintech credit per capita7 ) Difference │b-a│
(a) (b) H0 : │b-a│<0
GDP per capita1 0.080∗∗∗ 0.059∗∗∗ 0.098∗∗ 0.039

(0.026) (0.021) (0.046) (0.050)
GDP per capita^2 –0.001∗∗∗ –0.001∗∗∗ –0.001∗∗ 0.000
(0.000) (0.000) (0.001) (0.001)
Lerner index2 1.415∗∗∗ 0.912∗∗ 2.032∗∗∗ 1.120∗
(0.406) (0.374) (0.710) (0.802)
Bank branches per 100,000 –0.016∗∗∗ 0.003 –0.020∗∗ 0.023∗ ∗ ∗
Normalised regulation –4.528∗∗∗ –1.629∗∗∗ –7.535∗∗∗ 5.906∗ ∗ ∗
index3 (0.608) (0.560) (1.165) (1.293)
Ease of doing business4 0.097∗∗ –0.041 0.107∗ 0.148∗ ∗
(0.040) (0.033) (0.064) (0.083)
Investor protection and –0.176∗∗∗ –0.083∗ –0.280∗∗∗ 0.197∗ ∗
judicial system5 (0.044) (0.042) (0.098) (0.107)
Other controls Yes Yes Yes
Geographic area fixed effects8 Yes Yes Yes
No. of observations 425 425 425
R2 0.475 0.112 0.518
The table shows the results of a panel regression of the natural logarithm of total alternative credit per capita, (column 1), fintech credit per capita (column 2) and
big tech credit per capita (column 3) on the independent variables reported in the row headings. The last column reports the results of test for differences between
the coefficients for big tech and fintech credit. Estimation period 2013–18. Robust standard errors in parentheses. ∗ ∗ ∗ /∗ ∗ /∗ denotes results significant at the 1/5/10%
level. Ln = natural logarithm. The dependent variables are winsorized at the 1% and 99% level.
1
GDP per capita (in USD thousands).
2
3
4
Corresponds to the first principal components of a PCA over the overall score, the time, the minimum capital and of the cost of starting a business. The loadings
are 0.63, 0.48, 0.48, and 0.38 respectively.
5
Corresponds to the first principal component of a PCA over the extent of disclosure index, the number of days for trial and judgement, the number of days for
enforcement of judgement, and the enforcement fees. The loadings are –0.46, 0.63, 0.61, and 0.14 respectively.
6
Big tech credit is zero in 47 countries. To allow the computation of the log of the ratio (not defined for zero), big tech credit has been rescaled summing an
arbitrary constant (the minimum value).
7
Fintech credit is defined as credit activity facilitated by electronic platforms that are not operated by commercial banks or big tech firms.
8
indicators are available only for 65 of the 79 countries in the sam- system and the strength of insolvency resolution. All indicators are
ple, and therefore, the number of observations drops from 453 to taken from the World Bank Ease of Doing Business database.
425. Countries with more disclosure and stronger judicial systems
The results indicate that total alternative credit is positively cor- have more developed alternative forms of credit. The latter is
related with all the indicators that measure the ease of starting a higher where investors are protected through laws that allow for
new business. To get an indication of the quantitative effect, we higher disclosure of ownership and financial information. The sum
can observe countries in different quartiles of the overall score dis- of fintech and big tech credit per capita is also larger in countries
tribution. For example, the difference in total alternative credit per with a lower average number of days to complete a trial/judgment;
capita between those countries in the first quartile of the distri- a lower number of days to enforce the law; and lower judicial en-
bution (where it is relatively difficult to create a new business) forcement fees. The results are also confirmed in the last column of
and countries in the last quartile (where it is relatively easier to Table 5 where we include the first principal component of the four
create a new business) is between 1 and 16%, depending on the indicators on investor protection and judicial system. This principal
specification. The results are also confirmed in the last column of component captures 53% of the variance of these indicators.21
Table 4 where we include as a catch-all variable, the first principal Table 6 includes, jointly in our baseline specifications for to-
component of the four indicators, on the ease of doing business. tal alternative credit, fintech and big tech credit, the first principal
This principal component captures 57% of the variance of these in- components for the ease of doing business and investor protection
dicators. / judicial system. The last column reports the results of tests for
In Table 5, we analyze how the development of total alterna-
tive credit per capita depends on investor protection disclosure
and efficiency of the judicial system. Higher investor protection 21
We also examine how total alternative credit is influenced by: (i) the impact of
may make it easier to set up a new lending platform and to find specific characteristics of the banking sector, and (ii) bond and equity market de-
investors. Superior contract enforcement frameworks may limit velopment. Unfortunately, the data coverage of these two measures is quite limited.
credit risk and thus make lending more attractive. Again, to avoid Results are reported in Tables B9 and B10 in Appendix B. We find that total alter-
native credit is less developed in countries where the banking system supplies a
multicollinearity problems, we add one indicator at a time. The larger amount of credit relative to their deposit capacity (loan-to-deposit ratio). By
first column includes an indicator for the business extent of dis- contrast, total alternative credit is more developed where banks have a higher level
closure (the extent to which investors are protected through dis- of capital. Interestingly, alternative credit is also higher in countries with a higher
closure of ownership and financial information). The index ranges level of bank provision to non-performing loans. This could indicate a role for fin-
tech and big tech credit in economies where the banking system is more cautious
from 0 to 10, with higher values indicating more disclosure. The
or is more constrained by credit losses. Finally, we find that total alternative credit
other indicators taken into account the efficiency of the judicial is positively correlated with indicators of development of the bond and equity mar-
ket, and with other forms of non-bank credit: factoring and leasing. More details
are provided in the working paper version of this study (Cornelli et al., 2020).
12
differences between the coefficients for big tech and fintech credit. teeing financial stability should not have to “fly blind” or rely ex-
All the main results remain unchanged. The two institutional char- clusively on non-official data sources; they should have access to
acteristics analyzed – namely (i) the ease of opening a new busi- timely and accurate information about fintech and big tech credit
ness and (ii) investor protection disclosure and efficiency of the ju- in their own economy and economies around the world. With the
dicial system – jointly have a significant effect on the alternative database accompanying this paper, we contribute toward this goal.
forms of credit. However, the last column indicates that the effect
on fintech credit is significantly higher than that on big tech credit. CReDiT author statement
While fintech and big tech credit often serve smaller (e.g. SME)
corporate borrowers, and individuals, volumes do correlate with Giulio Cornelli: Conceptualization, Methodology, Analysis, Re-
market financing for larger firms. Further tests (not reported) show viewing; Jon Frost: Conceptualization, Methodology, Writing, Re-
that total alternative credit shows a strong positive association viewing and Editing; Leonardo Gambacorta: Conceptualization,
with venture capital, private equity and merger and acquisition ac- Analysis, Reviewing and Editing; P. Raghavendra Rau: Conceptu-
tivity and the number of such deals, from PitchBook Data. Thus, al- alization, Writing, Reviewing and Editing; Robert Wardrop: Con-
ternative credit seems to complement these other forms of finance, ceptualization; Tania Ziegler: Data curation
not to substitute for them. This is consistent with other work on
how alternative finance and capital market financing can reach un- Declaration of Competing Interest
derserved borrowers, particularly SMEs (World Bank, 2020).
None.
Conclusion
Data Availability
We document the recent growth of fintech credit, as provided
by non-bank online platforms, and big tech credit, as provided by Data will be made available on request.
large companies whose primary business is technology, sometimes
in partnership with traditional financial institutions. Based on data Supplementary materials
collected from the CCAF surveys, public sources and contacts with
firms and central banks, we show that both forms of credit have Supplementary material associated with this article can be
risen dramatically since 2013, but that since 2018, big tech credit found, in the online version, at doi:10.1016/j.jbankfin.2022.106742.
has overtaken fintech credit in total size. Based on preliminary
data, the shift to big tech credit was likely to have been even more References
pronounced in 2019.
We assess the economic and institutional factors driving the Barba Navaretti, G., G. Calzolari, J.M. Mansilla-Fernandez and A.F. Pozzolo (2017):
growth and adoption of fintech and big tech credit. We find that “FinTech and banks: friends or foes?”, European economy: banks, regulation,
and the real sector, 2, December.
fintech and big tech credit are more developed in countries with Bank for International Settlements (BIS) (2019): “Big Tech in finance: opportunities
higher GDP per capita, but at a declining rate. We also show that and risks”, BIS Annual Economic Report Ch. III, June.
these alternative forms of credit appear to be larger where banking Bazarbash, M. and K. Beaton (2020): “Filling the gap: digital credit and financial
inclusion”, IMF Working Papers, no 20/150.
mark-ups are higher and where banking regulation is less strin- Boissay, F., Ehlers, T., Gambacorta, L., Shin, H.S., 2020. Big techs in finance: on the
gent. Regulation and, to a lesser extent, mark-ups are particularly new nexus between data privacy and competition. In: Rau, R, Wardrop, R, Zin-
important for fintech credit. Fintech credit is also more prevalent gales, L (Eds.), The Handbook of Technological Finance. Palgrave Macmillan, Lon-
don.
where there are fewer bank branches per capita. We also find that Buchak, G., Matvos, G., Piskorski, T., Seru, A., 2018. Fintech, regulatory arbitrage, and
fintech and big tech credit are more developed where ease of do- the rise of shadow banks. J. Financ. Econ. 130 (3), 453–483.
ing business is greater, investor protection disclosure and the effi- Claessens, S., Frost, J., Turner, G., Zhu, F., 2018. Fintech credit markets around the
world: size, drivers and policy issues. BIS Quart. Rev. September.
ciency of the judicial system are more advanced, the bank credit- Cornelli, G., Frost, J., Gambacorta, L., Rau, R., Wardrop, R., Ziegler, T., 2020. Fintech
to-deposit ratio is lower, and where bond and equity markets are and big tech credit: a new database. BIS Work. Pap. 887.
more developed. Overall, these alternative forms of credit seem to De Roure, C., Pelizzon, L., Tasca, P., 2016. How Does P2P Lending Fit into the Con-
sumer Credit Market? Deutsche Bundesbank Discussion Papers, no 30/2016.
complement more traditional credit markets, rather than substitute
De Roure, C., Pelizzon, L., Thakor, A., 2018. P2P Lenders Versus Banks: Cream Skim-
for them. ming or Bottom Fishing? SAFE Working Papers, no 206.
It is of course difficult to forecast how large fintech and big tech Dembiermont, C., Drehmann, M., Muksakunratana, S., 2013. How much does the
credit will become in the future. Given accommodative drivers in private sector really borrow – a new database for total credit to the private
non-financial sector. BIS Quart. Rev..
some economies, fintech and big tech credit are growing to sizes Drehmann, M., Borio, C., Gambacorta, L., Jiménez, G., Trucharte, C., 2010. Counter-
that could be relevant to financial stability. In specific markets, for cyclical capital buffers: exploring options. BIS Work. Pap. 317.
instance, in small businesses lending in the United Kingdom and Ehrentraud, J., García Ocampo, D., Garzoni, L., Piccolo, M., 2020. Policy Responses to
Fintech: a Cross-Country Overview FSI Insights, no 23, January.
China, and consumer lending in Kenya, fintech and big tech lending Financial Stability Board (FSB) and Committee on the Global Financial System
have significant market shares. Some big tech players (particularly (CGFS), 2017. FinTech Credit: Market Structure, Business Models and Financial
in China) have likely reached a level of systemic importance. More- Stability Implications May.
FSB (2020): “Global monitoring report on non-bank financial intermediation”, Jan-
over, as credit has grown rapidly, there is the potential for individ- uary.
ual borrowers to become overindebted, and – as in past periods of Frost, J., Gambacorta, L., Huang, Y., Shin, H.S., Zbinden, P., 2019. BigTech and the
rapid credit growth – even for risks to financial stability. Whether changing structure of financial intermediation. Econ. Policy 34 (100), 761–799.
Fuster, A., Plosser, M., Schnabel, P., Vickery, J., 2018. The Role of Technology in Mort-
this growth represents the natural diffusion of a promising new
gage Lending Federal Reserve Bank of New York Staff Reports, no 836, February.
type of intermediation or a credit bubble remains an open ques- Gambacorta, L., Y.P. Huang, Z.H. Li, H. Qiu and S. Chen (2020): “Data vs collateral”,
tion; it may be possible to assess this only after the economy has BIS Working Papers, no 881.
Igan, D., Martinez Peria, M.S., Pierri, N., Presbitero, A., 2020. When They Go Low, We
undergone stressed conditions. In this light, the economic shock
Go high? Bank Market Power and Interest Rates IMF Working Papersforthcom-
from the Covid-19 pandemic represents an important test for fin- ing.
tech and big tech credit. Irving Fisher Committee on Central Bank Statistics (IFC), 2020. Central Banks and
Overall, there is a need to ensure that authorities and re- Fintech Data Issues February.
Jagtiani, J., Lemieux, C., 2019. The roles of alternative data and machine learning
searchers have the proper data to monitor and study fintech and in fintech lending: evidence from the LendingClub consumer platform. Financ.
big tech credit platforms. Authorities with a mandate for guaran- Manag. 48 (4), 1009–1029.
13
Kindleberger, C., Aliber, A., 2015. Manias, panics, and crashes: a History of Financial Shevlin, R., 2019. Uber’s Fintech strategy: a Conversation With Peter Hazlehurst,
Crises, 7th edition Palgrave Macmillan, London. Head of Uber Money. Forbes 4 November.
Laeven, L., Valencia, F., 2018. Systemic Banking Crises Revisited. International Mon- Stulz, R., 2019. FinTech, BigTech, and the Future of Banks. J. Appl. Corp. Finance 31
etary Fund, Washington, D.C. IMF Working Papers, WP/18/206. (4), 86–97.
Morse, A., Pence, K., 2020. Technological Innovation and Discrimination in House- Tang, H., 2019. Peer-to-peer lenders versus banks: substitutes or complements? Rev.
hold Finance NBER Working Papers, no 26739. Financ. Stud. 32 (5), 1900–1938.
Petralia, K., Philippon, T., Rice, T., Véron, N., 2019. Banking Disrupted? Financial Wardrop, R., Zhang, B., Rau, R., Gray, M., 2015. Moving Mainstream: the European
Intermediation in an Era of Transformational Technology Geneva Report 22, alternative Finance Benchmarking Report. Cambridge Centre for Alternative Fi-
September. nance.
Philippon, T., 2019. On Fintech and Financial Inclusion NBER Working Papers, no Wardrop, R., 2020. The Role of Trust in the Divergent Development of Alternative
26330. Lending Channels in the UK and Germany. University of Cambridge.
Ran, Z., Rau, P.R., Ziegler, T., 2022. Sometimes, Always, Never: The Impact of Regula- World Bank (2019): “Doing Business 2020: comparing business regulation in 190
tory Clarity on Digital Capital Raising. University of Cambridge working paper. economies”, October.
Rau, P.R., 2020. Law, Trust, and the Development of Crowdfunding. University of World Bank (2020): “Capital markets and SMEs in emerging markets and developing
Cambridge working paper. economies: can they go the distance?”, January.
Rowan, P., Miller, M., Schizas, E., Zhang, B., Carvajal, A., Blandin, A., Garvey, K., Zhang, B., Ziegler, T., Mammadova, L., Johanson, D., Gray, M., Yerolemou, N., 2018.
Ziegler, T., Rau, R., Randall, D., Hu, A., Umer, Z., Cloud, K., Mammadova, L., The Fifth UK Alternative Finance Industry Report. Cambridge Centre for Alter-
Kim, J., Yerolemou, N., 2019. Regulating Alternative Finance – Results from a native Finance November.
Global Regulator Survey. Cambridge Centre for Alternative Finance. Ziegler, T., Shneor, R., Wenzlaff, K., Odorović, A., Johanson, D., Hao, R., Ryll, L., 2018.
Schularick, M., Taylor, A.M., 2012. Credit booms gone bust: monetary policy, leverage Shifting Paradigms: The Fourth European alternative Finance Benchmarking Re-
cycles, and financial crises, 1870-2008. Am. Econ. Rev. 102 (2), 1029–1061. port. Cambridge Centre for Alternative Finance March.
Serena, J.M., 2019. IFC survey on fintech data: statistical initiatives to close data Ziegler, T., R. Shneor, K. Wenzlaff, B. Wang, J. Kim, A. Odorovic, F. Ferri de Camargo
gaps. Presentation to Irving Fisher Committee on Central Bank Statistics, August. Paes, K. Suresh, B. Zhang, D. Johanson, C. Lopez, L. Mammadova, N. Adams, D.
Luo (2020): “The global alternative finance market benchmarking report”, April.
14

Jurnal 4 Main

Uploaded by

Copyright:

Available Formats

Jurnal 4 Main

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Jurnal 4 Main

Uploaded by

Copyright:

Available Formats

Journal of Banking and Finance 148 (2023) 106742

Contents lists available at ScienceDirect

Journal of Banking and Finance

Credit markets around the world are undergoing a transforma-

This section discusses the characteristics of ﬁntech and big tech

Fig. 2. Fintech credit platforms in China.

Fig. 3. Geographical comparison of ﬁntech and big tech credit development.

In our core empirical analysis, we consider the determinants of

Variable Observations Mean Standard deviation Min Max

All variables are expressed in current USD, except where indicated

Variables of core interest

All variables are expressed in current USD, except where indicated

Baseline model with Panel with Pooled System GMM

GDP per capita1 0.123∗ ∗ ∗ 0.502∗ ∗ ∗ 0.139∗ ∗ ∗ 0.046∗ ∗ ∗

All variables are expressed in current USD, except where indicated

GDP per capita1 0.108∗ ∗ ∗ 0.112∗ ∗ ∗ 0.124∗ ∗ ∗ 0.112∗ ∗ ∗ 0.107∗ ∗ ∗

All variables are expressed in current USD, except where indicated

GDP per capita1 0.109∗ ∗ ∗ 0.108∗ ∗ ∗ 0.110∗ ∗ ∗ 0.113∗ ∗ ∗ 0.095∗ ∗ ∗

All variables are expressed in current USD, except where indicated

GDP per capita1 0.080∗∗∗ 0.059∗∗∗ 0.098∗∗ 0.039

You might also like