Journal Pre-proof

Can fintech improve the efficiency of commercial banks? —An analysis

based on big data

Yang Wang, Xiuping Sui

PII: S0275-5319(20)30946-6
DOI: https://doi.org/10.1016/j.ribaf.2020.101338
Reference: RIBAF 101338

To appear in: Research in International Business and Finance

Received Date: 17 January 2020

Revised Date: 13 September 2020
Accepted Date: 27 September 2020

Please cite this article as: { doi: https://doi.org/

This is a PDF file of an article that has undergone enhancements after acceptance, such as
the addition of a cover page and metadata, and formatting for readability, but it is not yet the
definitive version of record. This version will undergo additional copyediting, typesetting and
review before it is published in its final form, but we are providing this version to give early
visibility of the article. Please note that, during the production process, errors may be
discovered which could affect the content, and all legal disclaimers that apply to the journal
pertain.

© 2020 Published by Elsevier.

Can Fintech Improve the Efficiency of Commercial Banks?

—An Analysis Based on Big Data

Yang WANG, Xiuping Sui

Yang WANG,

of
School of Economics, Capital University of Economics and Business

ro
E-mail:[email protected]

-p
Xiuping Sui
re
School of international Economics，University of International Business and Economics

E-mail：[email protected]
lP
na

Highlights
ur

 .build more comprehensive fintech index

 .expend the calculation of TFP of commercial banks to input side based on DEA method
Jo

 .testify the relationship of finech and tip of commercial banks and find that fintech can boost
total factor productivity of commercial banks

 The improvement of fintech's total factor productivity of commercial banks varies according
to the degree the banks apply technology in china,

1. Introduction

1
With the rapid development of network technology and Internet, various traditional industries

innovations are booming in China. Since 2014, a new term has entered the public's vocabulary:

financial technology, or fintech(Pedersen,2015; Rysman, M. and S. Schuh ,2016). Broadly speaking,

fintech involves the application of a variety of advanced technologies to support the development of

the finance industry (Darolles, 2016). Related fields include big data, artificial intelligence (AI), cloud

computing, blockchain, and quantum computing. Fintech is a new type of computer algorithm that is

of
currently difficult to implement (Barocas and Selbst, 2016; Dhar, V, 2016). Fintech encompasses

both digital innovations and technology-enabled business model innovations in the financial

ro
sector(Abraham, F,2019). Such innovations can disrupt existing industry structures, blur industry

-p
boundaries, facilitate strategic disintermediation, change how existing firms create and deliver

products and services, provide new gateways for entrepreneurship, and democratise access to
re
financial services (Admati, 2013; Philippon, 2019). Fintech include areas of finance in which

technology is widely used, such as front-end consumer products, competition between new entrants
lP

and existing players, and even innovative block chain technology, and cryptocurrencies such as

bitcoin (Dranev, Y., et al，2019). Examples of innovations that are central to fintech today include
na

various applications of blockchain technologies, new digital advisory and trading systems, artificial

intelligence and machine learning, peer-to-peer lending, equity crowdfunding, and mobile payment
ur

systems (Philippon, 2015; Darolles,2016). The development of fintech has enhanced competitiveness

of commercial banks, as digital technologies have played a significant role in improving the efficiency
Jo

of services provided by banks and other financial institutions to small and micro enterprises, and to

private enterprises (Berg and Burg, 2019). Banks and other financial institutions are seeking to

minimise the costs of customer acquisition and risk control, reduce operating costs and improve

efficiency, and enhance the user experience for a wider range of consumers, leading to increasingly

strong demand for fintech applications (Allen and Gale, 2000;).

2
Intelligent decision-making, marketing, risk control, operations, and customer service enabled by

fintech can optimise an institution’s credit process and customer evaluation model（Aylin Aslan，

2020）, enable the quick lending of money, reduce the overall cost of corporate financing, and

enhance the economic efficiency of financial services (Bartlett, 2018). For example, after a firm

submits a loan application online, a fintech system can automatically decide whether to provide a

of
loan or not without manual intervention (Bazot, 2013). Banks can model and analyse user data

through customer profile regression, filter users according to nearly 10,000 rules, refer to Google's

ro
PageRank and analyse the weights of ordinary users on social networks (Dobbie, 2018). Furthermore,

-p
fintech in China has accumulated more than a billion anonymous samples encompassing six major

categories of core data, such as social and e-commerce data, and has verified and optimised the
re
customer identification model in practice(Bickenbach,2016). The impact and challenges of fintech

for commercial banks are primarily reflected in the impact of online payments (including third-party
lP

and mobile payments) and intermediary services such as payment and settlement (Chamley, 2012).

At the same time, the traditional asset and liability business of commercial banks has been challenged
na

by the trend toward financial disintermediation. Therefore, fintech has immediately affected both

commercial banks’ customer base and their market competition (Dhar, 2016). This study assesses the

potential impact of fintech on the banking industry by examining the relationship between the use of
ur

fintech and commercial banks’ total factor productivity (TFP). This paper is structured as follows:
Jo

Section 2 briefly reviews the related literature and presents hypotheses; Section 3 introduces the

model, data, and methodology for the study; Section 4 provides my empirical results and analysis;

and Section 5 summarises the main findings and presents the conclusion.

3
 2. Literature Review

The rise of fintech has had a major impact on the traditional business of commercial banks (Petralia

et al., 2019). In key areas such as residential mortgages, commercial banks have lost market share to

shadow banks and fintech lenders, which are subject to different regulations and enjoy technological

advantages (Buchak et al., 2018). Fintech lenders serve more creditworthy borrowers than shadow

banks but charge higher interest rates (14–16 basis points), which supports the idea that consumers

of
are willing to pay more for a better user experience and faster lending decisions. Another difference

ro
between fintech lenders and traditional lenders in the mortgage market is that the former process

applications 20% faster, without increasing loan risk (Fuster et al., 2019). Fintech lenders also respond

-p
more elastically to demand shocks and have a higher propensity to refinance, especially for borrowers

who are likely to benefit from it. In this way, fintech lenders have improved the efficiency of financial
re
intermediation in mortgage markets (Egan, 2016).
lP

The advent of fintech is often considered a promising avenue for reducing unequal access to credit.

Misalignment of incentives within finance firms can lead to biased lending decisions (Dobbie et al.,

2018). Fintech lenders may alleviate discrimination in mortgage markets; traditional lenders charge
na

minorities more for purchase and refinance mortgages, and fintech algorithms discriminate 40% less

than do face-to-face lenders (Bartlett et al., 2018). New financial technologies and data may offer
ur

superior capability for screening borrowers (Berg et al., 2019). The predictive power of data collected

by fintech, which is based on consumers’ digital footprints, equals or exceeds that of traditional credit
Jo

bureau scores when it comes to predicting consumer default.

Fintech companies are also competing in the market for wealth management. The United States is the

leading market for robo-advisors, and accounts for more than half of all investments managed by

robo-advisors in 2017 (Abraham et al., 2019). Nevertheless, the assets managed by robo-advisors are

4
still a small portion of the total assets under management, with average client wealth much less than

the average in the industry (The Economist, 2017). Because they save on fixed costs (such as the

salaries of financial advisors and maintenance of physical offices), robo-advisors can reduce

minimum investment requirements and lower fees (Abraham et al., 2019).

Finally, fintech has also intensively impacted the intermediary businesses of commercial banks, as

well as the incentives within organisations (Foà, 2015). Payment settlement has always been one of

of
the most basic and traditional intermediate businesses of commercial banks. According to information

ro
asymmetry theory, commercial banks, as financial intermediaries, help alleviate information

asymmetry to a certain extent (Kelly, 2016). Their information-based advantage and the resulting

-p
monopoly position have granted commercial banks long-term and unique advantages. Fintech, which

enables third-party and mobile payments, has reduced these advantages (Berger, A., R. Demsetz,
re
1999). Third-party and mobile payments have far lower costs than the services provided by banks.

Cloud computing support and other technologies can efficiently store and manage customer data,
lP

thereby more effectively alleviating information asymmetry, and can realise payment and settlement

more conveniently and efficiently than traditional methods (Baker,2015).

na

In conclusion, fintech influences the efficiency of commercial banks in several different ways. In this

study, I apply a data envelopment analysis (DEA) Malmquist non-parametric method (Fare, 1994) to
ur

evaluate the multi-input and multi-output effects on the banking industry and to calculate the TFP of

commercial banks, then analyse the impact of fintech on the efficiency of commercial banks.
Jo

Based on this, the hypotheses are:

1. Fintech promotes the TFP of commercial banks.

5
2. The positive impact of fintech on the TFP of commercial banks varies according to the degree to

which banks apply the technology.

3. Model and Data

Based on the above theoretical analysis, the development of fintech has resulted in increased

profitability, financial business innovations, and improved risk control for commercial banks. That is,

of
by using fintech, commercial banks can improve their traditional business models, reduce operating

costs, improve service efficiency, strengthen risk control capabilities, and directly create more

ro
attractive business models for customers, thereby improving their comprehensive competitiveness.

Because commercial banks of different sizes and types are affected differently by the development of

-p
fintech, econometric models should fully consider the impact of variables on the heterogeneity of

commercial banks on regression. Therefore, in developing my model, it was necessary to introduce

re
heterogeneity control variables. The competitiveness of commercial banks involves many factors
lP

which are difficult to quantify. Therefore, I used the TFP of commercial banks as a proxy for their

competitiveness. Because TFP has a certain viscous effect, it was necessary to use a lagging TFP.
na

3.1 Model

TFP𝑖𝑡 = 𝛼 + 𝛽𝐹𝑖𝑛𝑡𝑒𝑐ℎ𝑖𝑡 + 𝛾𝑋𝑖𝑡 + 𝛿TFP𝑖𝑡−1 + 𝜇𝑖𝑡 + 𝜀𝑖𝑡

ur

In the above model, the main explanatory variable, TFP𝑖𝑡 , indicates the change in the TFP of bank i

in year t. When it is greater than 1, the bank's TFP in that year improved over the previous year, and
Jo

when it is less than 1, this indicates a decrease compared with the previous year. When it is equal to

1, it is the same as in the previous year. 𝐹𝑖𝑛𝑡𝑒𝑐ℎ𝑖𝑡 indicates China's fintech index (i.e. China's fintech

development indicators) for t years. 𝑋𝑖𝑡 is the control variable, including macroeconomic indicators,

the maturity of the financial market, monetary policy indicators (i.e. the strength of monetary policy),

6
capital market development indicators, and some bank-level heterogeneity indicators, such as sector

concentration at the industry level, bank size, risk-bearing capacity, profitability, resource allocation

ability, innovation ability, and whether it is a listed bank. Because of the lagging period of TFP𝑖𝑡−1 , I

form a dynamic data panel model. To solve the endogenous problem, the baseline regression

discussed in this paper uses two methods: systematic generalised moment estimation (SYS-GMM)

and differential generalised moment estimation (DIF-GMM). On this basis, I use regression as a

of
robustness test.

ro
TFP. My analysis uses TFP as the proxy variable for the competitiveness of commercial banks. When

calculating the TFP of a commercial bank, two choices must be considered: the TFP calculation

-p
method and the input and output(Brei, M,2016).

Measuring the inputs and outputs of banks is a difficult problem, because the characteristics of banks
re
are distinctive(O’Mahony,2009). Unlike, for example, companies in the manufacturing industry,
lP

which produce tangible goods, banks produce products that are both intangible (i.e. intermediate

services) and composite (i.e. composed of a range of products). Bank efficiency studies have used

many methods of measuring bank output, such as the number of deposit and loan accounts and the
na

revenue in each account(Bolton,2011). Because my analysis emphasises the basic nature of banks’

production processes, rather than stock changes, I consider the services provided to customers as bank
ur

outputs.
Jo

There are two main methods for measuring service products: production and intermediary

methods(Chamley, C.,2012). The production method considers banks to be companies that produce

different deposit and loan accounts. The number and type of transactions and vouchers are considered

the best measure of bank output(Favara, G,2009). However, it is generally difficult to obtain this data.

Therefore, in practice, the number of deposit and loan accounts alone is usually used to measure the

7
output of banks. The intermediary method considers banks to be financial intermediaries that transfer

funds between depositors and lenders. In the production method, a bank’s number of bank loans and

investments represent its output, while labour and deposits represent inputs. The intermediary method,

on the other hand, uses deposits as an input while taking into account operating and interest costs

(Goddard et al., 2001). Neither method is perfect, and they may complement each other (Berger and

Humphrey, 1997). Each of the above methods emphasises the functions of a certain aspect of a bank,

of
and they can be used to analyse efficiency on different levels(Levine, R,2015). Given that the

production method considers the operating costs of banks, it is most suitable for studying cost

ro
efficiency. As the intermediary method assesses the overall cost of banks, it is most suitable for

-p
analysing the economic differences between banks (Ferrier and Lovell, 1990). This method takes

interest costs into account and is useful for assessing bank efficiency and undertaking boundary
re
analyses. However, there are different theoretical interpretations of the way commercial banks

operate, as well as differences in the definitions of the input and output variables of banks.
lP

In the existing literature, there are five commonly used methods for defining bank inputs(Levine,

R. ,2005): the production method, the intermediary method, the asset method, the user cost method,
na

and the added value method(Glode, V.,2012). The main difference between these methods is the

existence of a rationality of the bank and a different understanding of the role of the bank, and the
ur

resulting choice of input(Greenwood, R,2013). As discussed above, according to the production

method, the output of banks is measured by the number of their deposit and loan accounts. A bank's
Jo

inputs are usually capital and fixed labour costs(Kelly, B.,2016). In contrast, the intermediary method,

as discussed above, considers that banks pool idle funds and distribute them to the parties who need

them, acting as an intermediary between fund suppliers and demanders.

8
The asset method also considers banks as intermediaries between cash suppliers and demanders, and

defines the output of a bank as assets on its balance sheet, mainly including loans(Kovner, A.,2014).

Under this method, deposits are viewed as liabilities rather than outputs. This method considers that

assets should be regarded as outputs when the opportunity cost for the bank is lower than its return

on the assets. Deposits are also considered to be liabilities when the opportunity cost for the bank is

higher than the value of the liability(Yermack, D, 2015).

of
The added value method considers that a bank's inputs include labour and physical capital to purchase

ro
funds, and its outputs are the activities that generate high added value, such as loans, demand deposits,

and time deposits(Kumar, S,2016).

-p
The first three methods mentioned above are currently the main methods used in the literature. The

same calculation method may use different input and output variables due to different data sources or
re
focus points. The production method requires an analysis of the volume of business done by a bank,
lP

and this information is difficult to obtain, so I do not use this method. However, whether the asset

method or intermediary method is used, production is the most important factor, and the final result

of production is the attainment of income and maximisation of profits. Therefore, my research

na

focusses on the production results of banks.

I define bank input variables as their labour force costs and registered capital. I consider a bank’s
ur

labour force to be the number of full-time employees in a given year, including managers, sales staff,
Jo

and other personnel at all levels within the head office and branches, and define capital as the owner

equity component of a bank's income statement, to include physical and financial capital.

As discussed above, at present, there is no unified view of the definition of bank output. Based on

existing research both in China and abroad, I combine the characteristics of China's banking industry

to define bank output variables as follows: loan amount, profit amount, and deposit amount. As

9
special financial enterprises, the business objective of banks is to maximise income and profit.

Therefore, I include loan and profit amounts as outputs. Due to the liquidity and security of some

deposits, it is difficult to judge whether a bank’s deposit amount should be attributed to output or

input. However, China's commercial banks do not charge any fees when handling deposits, and this

intermediary behaviour has a positive external effect on society. Therefore, I consider it reasonable

to also use deposits as an output.

of
As mentioned above, I use the DEA Malmquist non-parametric method (Fare, 1994) to evaluate the

ro
multiple inputs and outputs of the banking industry. This method calculates the total factor

productivity of a commercial bank and then analyses the impact of fintech on the efficiency of the

-p
bank.

Fintech. Fintech can be evaluated across different dimensions, including payment calculation,
re
resource allocation, risk management, and network channels (Asktitas, 2009). These dimensions can
lP

be combined with fintech performance and big data indices (Zimmermann, 2010).

I use factor analysis to comprehensively calculate China's fintech index (i.e. China's fintech
na

development indicators) based on the following five dimensions: big data（Vu Le Tran，2020）,

artificial intelligence （ Shay-KeeTan,2020 ） , distributed technology, the interconnectedness of

ur

technology （ Shaker Ahmed,2020 ） , and the security of technology(Sun Xiaolei,2020). The

Jo

application layer used another four dimensions, based on those suggested by Asktitas (2009) and

listed above.

Table 1Fintech dimensions

Payment dimension Online payment

mobile payment

10
 Third party payment
QR code payment
Network payment
Resource allocation dimension internet loan
Internet lending
Network investment
Online lending
P2P loan
Risk management dimension Internet insurance
Internet financing

of
Network financing
Online financing
Network insurance

ro
Network channel dimension Mobile banking
Online Banking Service
Internet Banking

-p
E-bank
E-bank
Big data dimension Big data
re
Big data
data mining
Big data analysis
lP

Big data application

AI dimension AI
Artificial intelligence
intelligent robot
na

natural language processing

machine learning
Distributed technology cloud computing
dimension
Cloud platform
ur

Digital currency
Bitcoin
Blockchain Technology
Jo

Internet technology dimension Internet of things

Vehicle interconnection
Mobile Internet
5G
Mobile communication
Security technology dimension Biometrics
fingerprint identification
Iris Recognition

11
 Face recognition
Voice recognition

Bank size. I calculated bank size based on the log value of a bank's total assets. Research has shown

that large banks take advantage of economies of scale and scope to reduce transaction costs and

improve operating efficiency. However, some studies have suggested that the expansion of banks

makes management more complicated and reduces efficiency. Considering that fintech impacts all

of
aspects of the banking business, commercial banks with a larger asset scale are more capable of

ro
introducing fintech and reforming and innovating traditional business models, and are thus more

likely to improve their operating efficiency.

-p
Capital adequacy ratio. Most recent empirical studies have used the capital adequacy ratio to
re
measure the robustness of banks. Considering that the capital adequacy ratio was not available for

most commercial banks in the sample, and that the calculation method for the ratio was also adjusted
lP

by the CBRC in 2004, I use the ratio of owner equity to total assets as the risk control agent variable

to calculate the robustness of each bank.

na

Bank return on assets (ROA). Banks retain capital to increase capital and improve their ability to

manage risks, and therefore the current profitability of banks has a positive impact on TFP. This study

adopts the same method as Rime (2001) and replaces the bank's current profitability with its total
ur

asset income replacement rate (ROA) as the profitability proxy variable.

Jo

Bank loan-to-deposit ratio (LDR). I used bank LDR as the bank resource allocation capability proxy

variable. To calculate the LDR, I divided a bank's total amount of loans by its total amount of deposits

for the same period. I drew these figures from bank balance sheets, on which loans are listed as assets,

while deposits are listed as liabilities. A bank’s LDR shows its ability to cover loan losses and

12
withdrawals by its customers. Investors monitor the LDR of banks to make sure they have adequate

liquidity to cover loans in the event of an economic downturn resulting in loan defaults.

Deposit ratio. In my model, the proxy variable for a commercial bank's innovation capability is equal

to its non-interest income divided by its total revenue ratio. An increase in a bank’s cash deposit ratio

leads to a decrease in its money multiplier. An increase in deposit rates induces depositors to deposit

more, thereby leading to a decrease in the cash-to-aggregate-deposit ratio.

of
Other macro-level control variables. I also use the GDP growth rate or monetary policy to control

ro
for the impact of macroeconomic shifts on TFP.

-p
3.2 Data

Based on data availability, I selected a data panel to include 113 domestic commercial banks from
re
2009 to 2018. However, TFP, the main explanatory variable for this study, has a certain sticky effect.

Generally, when performing a regression analysis, it is necessary to include a TFP lag. Given that I
lP

use the DEA Malmquist method to calculate TFP, I had to obtain input-output information for the

year prior to the beginning of the dataset (i.e. to calculate the change in TFP from 2009 to 2018, it
na

would be necessary to obtain data from 2008 to 2018). Thus, I used data from 2008 to 2017.

Table 2 Data Description

ur

Variables Obs mean Standard min max

TFP 1243 1003 0.124 0 2.547
Fintech index 1130 0.594 0.319 0 1
Jo

GDP% 1130 7.96 1.346 6.6 10.6

Extend of finance 1130 59.125 12.526 42.02 78.5
Assets scale 1130 7.599 1.624 4.417 12.532
Capital adequacy
1130 13.013 2.313 3.24 40.3
ratio
Deposit ratio 1130 69.011 182.405 13.81 61.81
Proportion of non-
1130 11.195 8.766 0 47.1
interest income
ROA 1130 5.093 23.206 -59 209

13
 listed 1130 0.212 0.409 0 1

I analyse the multicollinearity problem between variables by describing the correlation coefficient

matrix of the main variables. The correlation coefficient for each variable reported in the table below

was less than 0.7, indicating that there is no serious multicollinearity problem.

Table 3 multicollinearity matrix

of
maturity
GDP Capital
Fintech of asset deposit deposit
TFP growth adequacy ROA

ro
index financial scale ratio ratio
rate ratio
market
TFP 1.000
Fintech
0.043 1.000

-p
index
GDP
-0.007 -0.615 1.000
growth rate
maturity of
re
financial -0.053 -0.014 -0.060 1.000
market
asset scale 0.032 0.329 -0.304 0.004 1.000
lP

Capital
adequacy 0.052 -0.058 0.093 -0.072 -0.204 1.000
ratio
deposit ratio 0.024 0.006 -0.013 -0.034 -0.005 0.011 1.000
deposit ratio 0.067 0.167 -0.162 0.106 0.016 -0.001 -0.020 1.000
na

ROA 0.022 -0.049 0.049 -0.040 -0.182 0.078 -0.003 -0.019 1.000
ur
Jo

14
 4. Empirical testing

First, I conducted an empirical test of Hypothesis 1 using the dependent and independent variables

outlined in Section 3 for the benchmark regression. To make the results in this section more robust, I

used two regression methods: SYS-GMM and DFF-GMM.

4.1 Baseline regression

of
The following table shows the benchmark regression results. The regression results using DIF-GMM

estimation are reported in columns (1)-(3). The fintech index is significantly positive at 1%, indicating

ro
that the development of fintech has increased the TFP of commercial banks. The regression results of

-p
the SYS-GMM are reported in columns (4)–(6), showing that the fintech index is still significantly

positive, and the coefficient of the variable's estimation is always significant when the control variable
re
is included, thus confirming the basic hypotheses of this article.

Table 4 Baseline regression after winsor

lP

DFF-GMM SYS-GMM
（1） （2） （3） （4） （5） （6）
3.780*** 3.408*** 3.505*** 3.588*** 3.400*** 3.478***
na

Fintech
(0.665) (0.670) (0.683) (0.648) (0.660) (0.664)
index
- - - - - -
TFP
0.441*** 0.444*** 0.442*** 0.298*** 0.301*** 0.301***
ur

(0.042) (0.043) (0.042) (0.042) (0.042) (0.042)

- - - - - -
GDPgrowth
1.457*** 1.352*** 1.388*** 1.425*** 1.366*** 1.395***
Jo

rate (0.248) (0.248) (0.250) (0.247) (0.251) (0.251)

- - - - - -
maturity of
0.018*** 0.016*** 0.017*** 0.018*** 0.017*** 0.017***
financial (0.003) (0.003) (0.003) (0.003) (0.003) (0.003)
market
-0.009 -0.007 0.002 0.006**
asset scale
(0.012) (0.012) (0.002) (0.003)

15
 0.005* 0.006** 0.003 0.003*
Capital
(0.003) (0.003) (0.002) (0.002)
adequacy
ratio
0.002** 0.002** 0.000 0.001
deposit
(0.001) (0.001) (0.000) (0.000)
ratio
0.000 0.000 0.001* 0.001*
deposit
(0.001) (0.001) (0.000) (0.000)
ratio

of
0.001** 0.001** 0.000 0.000
ROA
(0.000) (0.000) (0.000) (0.000)
-0.023 -0.020*
listed
(0.024) (0.012)

ro
constant
678 678 678 791 791 791
obs

-p
Note: The values in parentheses are the T statistics after correction of heteroscedasticity: *, **, ***
represent the significance levels of 10%, 5%, and 1%, respectively.
re
lP

As shown in columns (5) and (6), the asset scale is significantly positive, indicating that the larger

the asset scale, the higher the TFP of a commercial bank. As discussed, large-scale commercial banks

can generate economies of scale and are thus more capable of introducing fintech to improve their
na

operating efficiency.
ur

The estimated coefficient of capital adequacy ratio is significantly positive, indicating that the higher

the risk-bearing capacity of a commercial bank, the higher the TFP. This is consistent with the
Jo

findings of existing research. For example, John et al. (2008) noted that risk-bearing capacity was

related directly to corporate sales growth and investment in innovation. Appropriate risk-bearing

capacity can strengthen the ability of commercial banks to launch new businesses, thereby improving

TFP.

16
The estimated deposit ratio coefficient is significantly positive, indicating that commercial banks with

a stronger ability to allocate resources have a higher TFP. The deposit ratio reflects the ability of a

commercial bank to allocate funds. After the central bank relaxes the deposit ratio requirements for a

commercial bank, the bank can put more funds into their asset business, increasing its TFP.

As shown in columns (5) and (6), the deposit ratio is significantly positive, indicating that the more

innovative the commercial bank, the higher its TFP. This is logical; highly innovative commercial

of
banks are better able to accept technological changes to traditional business models.

ro
The estimated coefficient of GDP growth rate is significantly negative, indicating an inverse

relationship between the speed of economic development and the TFP of commercial banks. This

-p
may be because when the economy is developing rapidly, commercial banks can easily acquire a

large number of high-quality deposit and loan customers, and stable interest rate spreads reduce the
re
willingness of banks to innovate and restructure their businesses, leading to a decline in TFP.
lP

The estimated coefficient of capital market deepening is significantly negative, indicating that the

development of the capital market has a negative impact on the TFP of commercial banks. As a
na

country's capital market becomes mature, more funds flow from commercial banks to the capital

market, making it more difficult for commercial banks to absorb savings, and lowering their TFP.
ur

The estimated coefficients of ROA and listed status are not significant, indicating that in the domestic

banking system, these factors have only a small impact on the TFP of commercial banks.
Jo

4.2 Heterogeneous impact on different types of commercial banks

Based on benchmark regression, I next divided the sample of 113 banks into three sub-samples to

study the impact of fintech on the TFP of different types of commercial banks. The three sub-samples

were national banks (a total of 18, including 6 state-owned large banks and 12 joint-stock banks),

17
urban commercial banks (a total of 72), and rural commercial banks (a total of 23). The banks were

classified into these categories on the following basis:

1) National banks have branches across the whole country;

2) Urban commercial banks are local commercial banks that can only open branches in specific

regions, and generally provide financial services to urban enterprises and residents; and

3) Rural commercial banks are local commercial banks that can only open branches in specific

of
regions and generally provide financial services to rural enterprises and farmers.

ro
Table 5 DFF-GMM subsample regression results

National City Rural

Fintech
（1）
3.179***
（2）
2.800**
（3）
4.558***
-p（4）
4.180***
（5）
1.608*
（6）
1.933**
re
(0.808) (1.243) (0.977) (1.055) (0.937) (0.896)
- - - - - -
TFP
0.283*** 0.313*** 0.432*** 0.437*** 0.431*** 0.435***
(0.074) (0.066) (0.063) (0.064) (0.048) (0.046)
lP

- - - - - -
GDPgrowth
1.181*** 1.135** 1.751*** 1.622*** 0.716** 0.805**
rate (0.282) (0.441) (0.370) (0.386) (0.362) (0.358)
maturity of
na

- - - - -
-0.009*
financial 0.015*** 0.013** 0.022*** 0.020*** 0.010**
(0.005)
(0.004) (0.006) (0.005) (0.005) (0.005)
market
-
ur

asset scale -0.182* -0.067

0.016**
(0.095) (0.063)
(0.007)
Capital
Jo

-0.011 0.010*** 0.007*

adequacy (0.013) (0.004) (0.004)
ratio
0.004** 0.001*** 0.001
deposit ratio
(0.001) (0.000) (0.001)
0.000 0.000 0.001
deposit ratio
(0.004) (0.001) (0.001)

18
 -
ROA 0.031 0.001**
0.159**
(0.031) (0.000)
(0.071)
obs 108 108 432 432 138 138
Note: The values in parentheses are the T statistics after correction of heteroscedasticity: *, **, ***
represent the significance levels of 10%, 5%, and 1%, respectively.

of
Tables 5 and 6 report the regression results using DIF-GMM and SYS-GMM, respectively. As can

be seen from the tables, fintech has the most positive impact on the TFP of urban commercial banks,

ro
followed by national banks, and finally by rural commercial banks. A reasonable explanation is that

because of the restrictions on opening a new branch, urban commercial banks are at a disadvantage

-p
in terms of acquiring offline customers compared with national commercial banks.
re
Table 6 SYS-GMM subsample regression results

National City Rural

lP

（1） （2） （3） （4） （5） （6）

3.047** 2.628** 4.374**
Fintech 3.923*** 1.448* 1.730**
* * *
(0.995) (0.784) (0.792)
(0.840) (0.851) (0.957)
na

- - - - -
TFP -
0.236** 0.258** 4.167** 0.303** 0.326**
0.301***
* * * * *
(0.058)
(0.077) (0.070) (0.887) (0.051) (0.042)
ur

- - -
GDPgrowt - - -
1.140** 1.023** 1.728**
1.578*** 0.679** 0.753**
h rate * * *
(0.382) (0.316) (0.324)
Jo

(0.289) (0.318) (0.373)

maturity of - - -
- - -
0.014** 0.013** 0.021**
financial 0.019*** 0.008** 0.010**
* * *
(0.005) (0.004) (0.004)
market (0.004) (0.004) (0.005)
0.020** -0.004 -0.012
asset scale
(0.009) (0.005) (0.008)

19
 Capital
-0.004 0.007*** 0.006*
adequacy (0.006) (0.003) (0.003)
ratio
0.001**
deposit 0.001** 0.001***
*
(0.001) (0.000)
ratio (0.000)
deposit 0.001 0.000 0.001
(0.001) (0.000) (0.001)
ratio

of
-
ROA
0.079** 0.004 0.000
* (0.014) (0.000)

ro
(0.030)
9.243** 13.887** 7.591**
constant
* * *
(2.504) (3.068) (2.576)
obs 126 126 504

-p
504 161 161
Note: The values in parentheses are the T statistics after correction of heteroscedasticity: *, **, ***
re
represent the significance levels of 10%, 5%, and 1%, respectively.

Therefore, they are more eager to use fintech to offset this disadvantage and cultivate their
lP

competitive advantage. Rural commercial banks have the same needs as urban commercial banks, but

are generally small in scale and cannot afford the upfront costs of fintech applications. Moreover,
na

compared with national commercial banks, the level of organisation and strength of human capital of

rural commercial banks are relatively low. A lack of understanding of fintech and its applications in
ur

rural commercial banks has resulted in a low rate of adoption of new technologies and new business

models.
Jo

4.3 Robustness test

Based on the above analysis, I next used lagged TFP as a robustness test. The results are reported in

the table below. The results in columns (1)-(6) are identical to those in the table above. The sign of

the estimated coefficient of the control variable was consistent across the columns of the robustness

20
test. To some extent, this shows that the selected control variables can control the impact of various

factors on bank performance with relative stability.

Table 7 DFF-GMM Robustness Regression Results

DFF-GMM DFF-GMM DFF-GMM DFF-GMM

(1) (2) (3) (4)

- - - -

of
LagTFP 0.405*** 0.296*** 0.420*** 0.436***
(0.052) (0.028) (0.068) (0.056)
3.155*** 3.877*** 3.777*** 1.153**

ro
Fintech
(0.449) (0.860) (0.706) (0.558)
- - - -
GDPgrowth rate 1.390*** 1.680*** 1.633*** 0.550***

-p
(0.164) (0.306) (0.248) (0.209)
- - - -
maturity of financial market 0.017*** 0.021*** 0.020*** 0.007***
re
(0.002) (0.004) (0.003) (0.003)
-
asset scale 0.005 0.230*** 0.004 -0.004
lP

(0.008) (0.081) (0.043) (0.005)

0.002 -0.013 0.004 -0.001
Capital adequacy ratio
(0.002) (0.009) (0.002) (0.002)
0.000*** -0.000 0.001*** 0.001**
deposit ratio
na

(0.000) (0.001) (0.000) (0.001)

-
deposit ratio 0.001** 0.003 -0.001* -0.001
(0.000) (0.002) (0.000) (0.001)
ur

-0.000 -0.169* 0.024 0.000

ROA
(0.000) (0.093) (0.017) (0.000)
obs 678 108 432 138
Jo

Note: The values in parentheses are the T statistics after correction of heteroscedasticity: *, **, ***
represent the significance levels of 10%, 5%, and 1%, respectively.

21
Table 8 SYS-GMM robustness regression results

SYS-GMM SYS-GMM SYS-GMM SYS-GMM

（5） (6) (7) (8)

- - - -
Lag TFP 0.278*** 0.251*** 0.288*** 0.267***
(0.048) (0.021) (0.062) (0.046)

of
2.754*** 2.817*** 3.375*** 0.690
Fintech
(0.442) (0.708) (0.624) (0.465)

ro
- - -
-1.525***
GDPgrowth rate 1.268*** 1.335*** 0.391**
(0.231)
(0.165) (0.264) (0.183)

maturity of financial market

-
0.015***
-

-p
0.016***
-
0.018***
-
0.005**
re
(0.002) (0.004) (0.003) (0.002)

0.003 -0.005 -0.003 -0.005

asset scale
lP

(0.002) (0.008) (0.003) (0.005)

0.001 -0.001 0.002 -0.003

Capital adequacy ratio
(0.001) (0.005) (0.002) (0.002)
na

-
0.000*** 0.001** 0.000***
deposit ratio 0.001**
(0.000) (0.001) (0.000)
(0.000)
ur

0.000 0.002*** -0.000 -0.000

deposit ratio
(0.000) (0.001) (0.000) (0.000)
Jo

0.000* 0.023 0.006 0.000

ROA
(0.000) (0.028) (0.010) (0.000)

11.508*** 11.993*** 13.616*** 4.524***

constant
(1.338) (2.131) (1.862) (1.452)

22
 obs 791 126 504 161

5. Conclusion

of
Based on the above analysis, the development of fintech has increased profitability, led to innovations,

and improved risk control for commercial banks. By adopting fintech, commercial banks can improve

ro
their traditional business models, reduce operating costs, improve service efficiency, strengthen risk

control capabilities, and create more attractive business models for customers, thereby improving

-p
their comprehensive competitiveness. Different sizes and types of commercial banks are affected
re
differently by the development of fintech. The adoption of fintech is likely to decrease the costs of

financial intermediation, but also to create new regulatory issues. In this paper, I have highlighted
lP

two relevant forces that will shape the impact of fintech on inequality. In the case of robo-advisors, I

have argued that the new model of fixed costs is likely to improve participation by lower-income

households. However, this may not lower inequality across all groups. In relation to the credit market,
na

alternate data sources are likely to reduce non-statistical discrimination. The most basic prerequisite

for a commercial bank to achieving in-depth integration with fintech and improving its
ur

professionalism, responsiveness, and inclusivity is to have the required hardware and software

infrastructure in place. Hardware requirements include information network facilities, high-

Jo

performance computers and cloud servers, and large-capacity storage, whereas the required software

capacity includes powerful data-mining and calculation, distributed storage, batch processing, and

advanced artificial intelligence.

23
Reference

1. Abraham, F., S. L. Schmukler, and J. Tessada (2019). Robo-advisors: Investing through

machines. World Bank Policy Research Working Paper (134881).

2. Admati, A. R. and M. Hellwig (2013). The Bankers’ New Clothes. Princeton University Press.

3. Aylin Aslan. Ahmet Sensoy (2020). Intraday efficiency-frequency nexus in the

of
cryptocurrency markets. Finance Research Letters.07(35).101298

ro
4. Baker, M. and J. Wurgler (2015). Do strict capital requirements raise the cost of capital? bank

regulation, capital structure, and the low risk anomaly. American Economic Review Papers and

-p
Proceedings.

5. Barocas, S. and A. Selbst (2016). Big data’s disparate impact. California Law Review 104,
re
671–732.
lP

6. Bartlett, R., A. Morse, R. Stanton, and N. Wallace (2018). Consumer-lending discrimination

in the era of Fintech. Working paper.

na

7. Bazot, G. (2013). Financial consumption and the cost of finance: Measuring financial

efficiency in europe (1950-2007). Working Paper Paris School of Economics.

ur

8. Bazot, G. (2013). Financial consumption and the cost of finance: Measuring financial

efficiency in europe (1950-2007). Working Paper Paris School of Economics.

Jo

9. Berg, T., V. Burg, A. Gombović, and M. Puri (2019). On the rise of Fintechs – credit scoring

using digital footprints.Working paper.

24
10. Berger, A., R. Demsetz, and P. E. Strahan (1999). The consolidation of the financial services

industry: Causes, consequences, and implications for the future. Journal of Banking and Finance 23,

135–194.

11. Bergstresser, D., J. Chalmers, and P. Tufano (2009). Assessing the costs and benefits of

brokers in the mutual fund industry. The Review of Financial Studies 22(10), 4129–4156.

of
12. Bickenbach, F., E. Bode, D. Dohse, A. Hanley, and R. Schweickert (2009, October).

Adjustment after the crisis: Will the financial sector shrink? Kiel Policy Brief.

ro
13. Bickenbach, F., E. Bode, D. Dohse, A. Hanley, and R. Schweickert (2009, October).

-p
Adjustment after the crisis:Will the financial sector shrink? Kiel Policy Brief.

14. Bodenhorn, H. (2000). A History of Banking in Antebellum America: Financial Markets and
re
Economic Development in an Era of Nation Building. New York: Cambridge University Press.
lP

15. Bolton, P., T. Santos, and J. Scheinkman (2011). Cream skimming in financial markets.

Working Paper, Columbia University.

na

16. Brei, M. and L. Gambacorta (2016). Are bank capital ratios pro-cyclical? new evidence and

perspectives. Economic Policy 31 (86), 357–403.

ur

17. Buchak, G., G. Matvos, T. Piskorski, and A. Seru (2018). Fintech, regulatory arbitrage, and

the rise of shadow banks. Journal of Financial Economics 130(3), 453 – 483.
Jo

18. Chamley, C., L. J. Kotlikoff, and H. Polemarchakis (2012). Limited-purpose banking–

moving from "trust me"to "show me" banking. American Economic Review 102(3), 113–19.

19. Darolles, S. (2016, April). The rise of Fintechs and their regulation. Financial Stability

Review (20).

25
20. Dell’Ariccia, G., D. Igan, L. Laeven, and H. Tong (2016). Credit booms and macrofinancial

stability. Economic Policy 31 (86), 299–355.

21. DeYoung, R., D. Evanoff, and P. Molyneux (2009). Mergers and acquisitions of financial

institutions: A review of the post-2000 literature. Journal of Financial Services Research 36, 87–110.

22. Dhar, V. (2016). When to trust robots with decisions, and when not to. Harvard Business

of
Review.

23. Dobbie, W., A. Liberman, D. Paravisini, and V. Pathania (2018). Measuring bias in consumer

ro
lending.

-p
24. Favara, G. (2009). An empirical reassessment of the relationship between finance and growth.

25. Fuster, A., M. Plosser, P. Schnabl, and J. Vickery (2019). The role of technology in mortgage
re
lending. The Review of Financial Studies 32(5), 1854–1899.
lP

26. Garleanu, N. and L. H. Pedersen (2018). Efficiently inefficient markets for assets and asset

management. The Journal of Finance 73(4), 1663–1712.

na

27. Glode, V., R. C. Green, and R. Lowery (2012). Financial expertise as an arms race. Journal

of Finance.
ur

28. Greenwood, R. and D. Scharfstein (2013). The growth of modern finance. Journal of

Economic Perspectives 27(2), 3–28.

Jo

29. Kelly, B., H. Lustig, and S. V. Nieuwerburgh (2016). Too-systemic-to-fail: What option

markets imply about sector-wide government guarantees. American Economic Review.

30. Klaus Grobys. Shaker Ahmed, Niranjan Sapkota(2020). Technical trading rules in the

cryptocurrency market Finance Research Letters 01(32),101396

26
31. Elie Bouri Brian Lucey. David Roubaud. The volatility surprise of leading cryptocurrencies:

Transitory and permanent linkages. Finance Research Letters.03(33).101188.

32. Kovner, A., J. Vickery, and L. Zhou (2014, December). Do big banks have lower operating

costs? FRBNY Economic.

33. Kumar, S. (2016). Relaunching innovation: Lessons from silion valley. Banking Perspective

of
4(1), 19–23.

34. Levine, R. (2005). Finance and growth: Theory and evidence. In P. Aghion and S. N. Durlauf

ro
(Eds.), Handbook of Economic Growth, Volume 1A, pp. 865–934.

-p
35. Lixin Sun (2020). Financial networks and systemic risk in China‘s banking system. Finance

Research Letters. 05(34).101236

re
36. Mayer, C. and K. Pence (2008). Subprime mortgages: What, where, and to whom? Staff
lP

Paper Federal Reserve Board.

37. Moore, K. B. and M. G. Palumbo (2010, June). The finances of American households in the
na

past three recessions: Evidence from the survey of consumer finances. Staff Paper Federal Reserve

Board.
ur

38. O’Mahony, M. and M. P. Timmer (2009). Output, input and productivity measures at the

industry level: The euklems database. The Economic Journal 119(538), F374–F403.
Jo

39. Pagnotta, E. and T. Philippon (2018, May). Competing on speed. Econometrica 86.

40. Pedersen, L. H. (2015). Efficiently Inefficient: How Smart Money Invests and Market Prices

Are Determined.Princeton University Press.

27
41. Petralia, K., T. Philippon, T. Rice, and N. Véron (2019). Banking disrupted? financial

intermediation in an era of transformational technology. Technical Report 22, Geneva Reports on the

World Economy, ICMB and CEPR.

42. Philippon, T. (2015). Has us finance industry become less efficient? on the theory and

measurement of financial intermediation. The American Economic Review 105(4), 1408–38.

of
43. Philippon, T. (2016). The Fintech opportunity. NBER Working Paper.

44. Philippon, T. and A. Reshef (2012). Wages and human capital in the u.s. financial industry:

ro
1909-2006. Quarterly Journal of Economics.

-p
45. Shaker Ahmed. Klaus Grobys. Niranjan Sapkota (2020). Profitability of technical trading

rules among cryptocurrencies with privacy function. Finance Research Letters.07(35).101495

re
46. Shay-KeeTan.Jennifer So-Kuen Chan. Kok-Haur Ng(2020).On the speculative nature of
lP

cryptocurrencies: A study on Garman and Klass volatility measure. Finance Research

Letters.01(32).101075
na

47. Sun Xiaolei. Liu Mingxi.Sima Zeqian(2020). A novel cryptocurrency price trend forecasting

model based on Light GBM. Finance Research Letters.01(32).101084

ur

48. Vu Le Tran. Thomas Leiviska (2020). Efficiency in the markets of crypto-currencies.

Finance Research Letters.07(35) 101382

Jo

49. Dranev, Y., et al. (2019). "The impact of fintech M&A on stock returns." Research in

International Business and Finance 48: 353-364.

28