1 s2.0 S0378426623001966 Main

Journal of Banking and Finance 156 (2023) 107005
Contents lists available at ScienceDirect
Journal of Banking and Finance

journal homepage: www.elsevier.com/locate/jbf
Digital finance and financial literacy: Evidence from Chinese households☆

Junhong Yang a, Yu Wu b, *, Bihong Huang c, 1
a
Management School, University of Sheffield, Conduit Road, Sheffield S10 1FL, UK
b
Survey and Research Center for China Household Finance, Southwestern University of Finance and Economics, Chengdu, Sichuan 610074, China
c
Monetary and Capital Markets Department, International Monetary Fund, 1900 Pennsylvania Avenue NW, Washington, DC 20431, USA
A R T I C L E I N F O A B S T R A C T
JEL classifications: We measure financial literacy and study its impact on household use of digital finance using the 2015 and 2017
D10 China Household Finance Survey. We find that financial literacy significantly boosts the use of digital finance,
G11 including mobile payments, online borrowing, and online financial products. This effect is more pronounced for
D91
online borrowing and online financial products than for mobile payments. This result suggests that the impact of
D83
financial literacy increases with the complexity of digital finance. Furthermore, financial literacy plays a more
Keywords:
important role in promoting the use of digital financial services among disadvantaged groups, such as families
Financial literacy
Digital finance
with low income and wealth, the elderly, and residents in rural areas, compared with their counterparts.
Household finance
CHFS
China
1. Introduction portfolio managers (robo-advisors), and advanced trading platforms,

meld technology, regulation, user behavior, and global market dy
Digital finance has shown tremendous potential in reaching previ namics. These developments contribute to the increasing complexity of
ously excluded and underserved populations by offering further tailored digital finance, which requires households to have adequate financial
financial services and products. The existing literature showed that ad literacy and knowledge. Moreover, the liberalization of financial mar
vancements in digital finance create new and affordable investment kets among competing financial technology (fintech) companies has
opportunities, leading to good welfare benefits and great financial in placed a significant responsibility on individuals to acquire financial
clusion (Cocco et al., 2005; Demir et al., 2020; Demirguc-Kunt et al., knowledge. Financial knowledge allows individuals to make informed
2018; Shen et al., 2020). Furthermore, digital finance fortifies financial economic and financial choices, avoid financial errors or missteps, and
system resilience and bolsters financial stability by rivaling traditional ensure their financial well-being. A lack of financial knowledge2 may
finance businesses (Jack and Suri 2014; Buchak et al., 2018; Ozili 2018). lead to the exclusion of certain individuals from digital financial services
Yet, emerging financial innovations, such as mobile banking, online and products, consequently depriving them of the relevant technological
financial management products, peer-to-peer (P2P) lending, automated empowerment. This study aims to analyze the role of financial literacy in
☆
We thank Geert Bekaert, the editor, and two anonymous referees for their invaluable suggestions and constructive comments. The work was supported by the
Asian Development Bank Institute. We also appreciate the helpful comments received during seminars at the School of Oriental and African Studies (London) and the
University of Sheffield. Junhong Yang is grateful for the support from the Centre for Research into Accounting and Finance in Context (CRAFiC) at the University of
Sheffield. Yu Wu is grateful for the support received from multiple sources: the Humanities and Social Sciences Program of the Ministry of Education (Grant No.
21JHQ060), the National Natural Science Foundation of China (Grant No. 71903160), the Fundamental Research Funds for the Central Universities (Grant No.
JBK2304055, JBK230113), and the Higher Education Discipline Innovation Project (111 Project) (Grant No. B16040).
* Corresponding author.
E-mail addresses: [email protected] (J. Yang), [email protected] (Y. Wu), [email protected] (B. Huang).
1
This research was conducted when Dr. Bihong Huang served as an Economist at the Asian Development Bank Institute. The views expressed in this paper are
those of the author(s) and do not necessarily represent the views of the IMF, its Executive Board, or IMF management.
2
Substantial evidence indicates the prevalence of financial illiteracy, with many individuals lacking even basic financial knowledge (Lusardi and Mitchell 2007;
van Rooij et al. 2011). For example, in the 2016 OECD/INFE Survey of Adult Financial Literacy Competencies (OECD 2016) covering 30 countries and economies,
including 17 OECD countries, only 56% of adults achieved the minimum target score for financial knowledge (five out of seven).
https://doi.org/10.1016/j.jbankfin.2023.107005
Received 14 February 2021; Accepted 8 September 2023
Available online 13 September 2023
0378-4266/© 2023 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-
nc/4.0/).
J. Yang et al. Journal of Banking and Finance 156 (2023) 107005
explaining households’ access decisions to digital finance. identities. The IV estimation results confirm the positive effect of
Financial literacy—“a combination of awareness, knowledge, skill, financial literacy on the use of digital finance. Moreover, we employ
attitude and behavior necessary to make sound financial decisions and various alternative measures of financial literacy to mitigate potential
ultimately achieve individual financial well-being” (Atkinson and Messy measurement errors. We account for cognitive abilities and peers’
2012)—is identified as a critical factor in diverse aspects of individual financial literacy to disentangle the impact of knowledge from innate
financial decision-making. Examples include opening bank accounts talents and peer influences on individuals’ decisions to use digital
(Cole et al., 2011), investing in stocks (van Rooij et al. 2011), purchasing finance. We also rule out the potential influence of households that have
life insurance (Wang et al., 2021), planning for retirement (Lusardi and voluntarily opted out of using digital finance.
Mitchell 2017), accumulating wealth (Jappelli and Padula 2013), Our study also explores the mechanisms through which financial
diversifying portfolios (Gaudecker 2015), and making good financial knowledge facilitates the use of digital finance. Our findings suggest that
decisions (Disney and Gathergood 2013; Guisoand Viviano 2014). financial literacy increases individuals’ access to online financial infor
Despite the burgeoning body of research accentuating financial lit mation, helping them make informed decisions about using digital
eracy, its relationship with the uptake of digital finance remains under- finance. In addition, households possessing high levels of financial lit
examined. Predominantly, the extant studies mainly focused on the role eracy are likely to trust in new things and have a high tolerance for
that financial literacy plays in contexts such as Internet banking financial risks. Hence, they are receptive to digital finance.
(Andreou and Anyfantaki, 2021). However, a gap in understanding the This research contributes to the existing literature in several ways.
influence of financial literacy on household engagement with digital We build on Andreou and Anyfantaki (2021), who focused on the role of
financial services of varying intricacy exists. Furthermore, the impact of financial literacy in Internet banking in Cyprus. We broaden this
this effect on different demographic, socio-economic, and geographic perspective by evaluating the influence of financial literacy on house
dimensions remains to be elucidated. hold adoption of various digital financial services, including mobile
In this context, this study employs the China Household Finance payments, online borrowing, and online financial products. The finding
Survey (CHFS), which interviews approximately 40,000 households that the impacts of financial literacy increase with digital finance
across the country. The objectives of our study are to measure the use of complexity has important policy implications. Given the growing com
different types of digital finance by Chinese households, estimate the plexities in fintech offerings, consumers should be urgently equipped
financial literacy level of individuals with different demographic char with the requisite knowledge and competencies, to ensure optimal
acteristics, and assess the extent to which changes in financial literacy choices in various digital finance for enhanced financial well-being.
affect the use of digital finance. The survey results reveal that the usage Our study distinguishes from the extant literature by exploring the
of digital finance is still low and mainly dominated by mobile payment. mechanisms that underpin the relationship between financial knowl
Overall, 29.1% of the households use digital finance, whereas the edge and digital finance adoption. The role of financial literacy in
adoption of mobile payments, online financial products, and online affecting households’ decision-making and improving their financial
borrowing among the respondents is 28.1%, 6.6%, and 4.4%, respec outcome has been recognized. However, knowledge of its working
tively. In addition, the level of financial knowledge proficiency, mechanism is still limited. Most studies in this area focused on estab
measured by the answers to three questions involving economic con lishing causality, with little effort devoted to investigating the pathways
cepts of interest rates, inflation, and financial risk, among Chinese of the effects. We carefully unpack the channels through which financial
households is fairly low. Notably, the weighted percentages of re literacy changes consumers’ adoption of digital finance. In particular,
spondents who did not understand and could not answer these three we examine how financial literacy assists consumers in lowering the
questions are 48.6% (interest rate), 45.9% (inflation), and 39.0% costs of obtaining information, improves their trust in digital technol
(financial risk). The level of financial literacy is significantly low among ogy, and increases their receptivity to financial risks. Understanding the
the elderly; male households with low levels of education, personal mechanisms of consumer and societal acceptance of digital finance is
wealth, and income; and residents living in rural and economically critical not only to fill a gap in the academic literature but also to inform
underdeveloped areas. public policies on designing additional effective financial education
The empirical analyses show that household financial literacy initiatives in the digital age.
significantly enhances the use of digital finance. Other things being This study also sheds new light on the financial inclusion literature.
equal, for every one standard deviation increase in financial literacy, the Cole et al. (2011) show that financial literacy education in emerging
probability of using digital finance increases by 2.87 percentage points. market countries only increases the likelihood of opening a bank ac
Notably, the impacts of financial literacy are more pronounced for on count for households with little education or financial literacy. Based on
line borrowing and online financial products than for mobile payments. a representative survey of 143 countries, Grohmann et al. (2018) find
This result indicates that the importance of financial literacy is amplified that financial literacy has a positive impact on financial inclusion. We
as the complexity of digital finance rises. We further implement the extend the literature by examining digital financial inclusion. Despite
heterogeneous analyses and find that the effects of financial literacy the promise of digital finance to provide affordable, accessible, and
significantly vary across different population groups. Specifically, adaptable financial services, its usage remains low in many emerging
financial literacy plays a more important role in encouraging the use of economies, such as China. Our research shows that a lack of financial
digital financial services among disadvantaged groups, such as families literacy significantly curtails digital finance adoption. Although digital
with low levels of wealth and income, the elderly, and rural residents, finance serves as an essential instrument for advancing financial inclu
compared with their counterparts. Financial literacy is particularly sion, households are required to cope with increasingly complex finan
important for them to overcome barriers to digital financial inclusion. cial decisions.
In the empirical analysis, we control for demographic, socio-
economic, and geographic factors in the regressions. However, unob 2. Institutional background
servable factors that affect the respondents’ financial literacy and their
use of digital finance simultaneously might still exist. Another potential Despite China’s remarkable progress in financial inclusion, the
concern is reverse causality. Households who are actively engaged in country still has the world’s largest unbanked population (Demi
digital finance are likely to show a heightened interest in improving rguc-Kunt et al., 2018). Approximately one-fifth of all adults (255
their financial knowledge, given their increased stake in the process. We million) do not have an account with a formal financial institution, ac
employ several instrumental variables (IVs) to address potential endo counting for 13% of the world’s unbanked population. Digital finance
geneity concerns, including the education levels of the respondents’ could synergistically complement traditional financial intermediaries,
parents, neighborhood financial literacy, and the respondents’ social particularly by extending financial services to areas that are often
2
underserved or neglected by inefficient Chinese financial institutions 3. Data, key variables, and summary statistics
(Demirguc-Kunt et al., 2018).
Over the past decade, digital finance has developed rapidly in China 3.1. Data
as an alternative finance model. This trajectory can be traced back to
December 2004, with the introduction of Alipay, a mobile payment The data used in this research come from the 2015 and 2017 CHFS,
solution developed by the e-commerce platform—Taobao. Mobile pay which contain very detailed demographic and financial information at
ment rapidly gained popularity in China owing to its greater conve the household and individual levels, including assets, wealth, and in
nience (wider coverage without intermediaries), better features (easier come. The survey is designed to be representative of Chinese house
to use and lower costs), and higher quality of service (faster and safer holds, which uses a multi-stage stratified random sampling process with
transactions) than other traditional modes of payment. Remarkably, probability proportionate to size. The survey covered 83 cities in 29
each of the two leading mobile payment service providers in China, provinces, municipalities, and autonomous regions, including a total of
Alipay and WeChat Pay, now has approximately 900 million users. 37,340 households in 2015 and 40,011 households in 2017. We exclude
In addition to mobile payment, FinTech Unicorns, such as Ant observations in the 1% tails for the key variables, such as household
Finance, Tencent, Baidu, and JD Digits, offer a range of online financial wealth and income, householder age, and number of household mem
products and services to their customers by utilizing digital technology bers, to minimize the potential influence of outliers. To alleviate
and big data analysis (Huang 2020). For example, Ant Finance launched simultaneity bias, our dependent variable—the use of digital finan
Yu’ebao in 2013. Yu’ebao enables Alipay customers to transfer money ce—comes from the 2017 wave of the survey, whereas our explanatory
between their payment account and money market account and earn an variables—including financial literacy—are from the 2015 wave. The
interest rate on their residual funds, which is much higher than con final sample contains 19,788 observations.3
ventional bank deposit rates.
The investment opportunities offered by online financial products 3.2. Measures of digital finance
hold considerable advantages over traditional bank offerings. On the
one hand, online financial products have significantly reduced the bar The CHFS has several survey questions designed to measure whether
riers that households face when investing in the financial market. For respondents have used any form of digital finance, that is, mobile pay
instance, one can invest in a money market fund with as low as one ments, online financial products, and online borrowing. Specifically,
Chinese yuan. Online financial products are also appealing owing to mobile payment refers to a payment made through portable electronic
their convenience. Consumers can invest seamlessly, anywhere and devices (tablets or mobile phones), such as Alipay (Alibaba), WeChat
anytime by using their mobile devices without having to physically visit Pay (Tencent), and Apple Pay (Apple). Online financial products are
a bank branch. On the other hand, the inherently low overhead costs of Internet-based wealth management products, such as Yu’ebao (Ali
digital platforms enable these online financial products to offer reduced baba), Licaitong (Tencent), JD Xiaojinku (JD.com), and Baifa (Baidu).
fees and competitive interest rates. This cost-efficiency, in turn, provides Online borrowing includes online consumer finance (e.g., Huabei and
investors with access to a variety of investment products, facilitating JD IOU), loans provided by Internet banks (Webank, Mybank, etc.),
portfolio diversification and the potential for superior returns. There online cash lending (Ali microfinance, Ant Borrow, etc.), and P2P
fore, the customer response to the online financial products provided by lending.
FinTech Unicorns was astonishing. Within a mere six months post- Therefore, we construct three binary variables to reflect the use of
launch, Yu’ebao garnered an impressive 43 million subscribers. In Q1 these three types of digital financial services, namely, Mobile_Payments,
2017, Yu’ebao, with assets of CNY 1.1 trillion (USD 117 billion), over Online_FPs, and Online_Borrowing. They take a value of 1 if the respon
took JP Morgan’s US government money market fund and became the dent has used the service and 0 if otherwise. We collectively denote the
largest money market fund in the world (Chui 2021). use of digital finance with a binary variable (Digital_Finance), which
Recent advances in fintech have also facilitated the development of takes a value of 1 if the respondent has used any of these services and 0 if
online borrowing as an alternative source of funding, particularly otherwise.
benefiting low-income households and small and medium-sized enter
prises (SMEs). They often face challenges in accessing traditional bank 3.3. Measures of financial literacy
credit owing to the lack of historical data, insufficient collateral, or
absence of government guarantees. To bridge this gap, digital banks, Financial literacy is a form of human capital that includes under
such as WeBank, MyBank, and XWBank, have developed proprietary standing financial concepts and knowledge needed to make important
credit scoring models using big data analytics and innovative solutions financial decisions. In our study, financial literacy refers specifically to
(Gambacorta et al., 2019). Moreover, online consumer lending services financial knowledge as a form of human capital, rather than the broad
provided by fintech giants, such as Ant Group’s Huabei and Jiebei, have term “financial capability,” which includes financial knowledge, finan
grown rapidly and become important drivers of consumer spending. cial behavior, and financial self-efficacy (Lusardi and Mitchell 2014;
Between June 2020 and June 2021, nearly 500 million users borrowed Feng et al., 2019).
money from Huabei and Jiebei. In the first half of 2020, the revenue of In the CHFS, similar to Lusardi and Mitchell (2011, 2014), Lusardi
these lending services reached RMB 28.3 billion, accounting for 39% of and de Bassa Scheresberg (2013), and Grohmann et al. (2018), three
Ant Group’s revenue before the introduction of new regulations at the questions are asked to assess respondents’ financial literacy. These three
end of 2020 (Yu and McMorrow 2021). Furthermore, in 2015, the questions deal with the concept of interest rates and compound interest
Chinese witnessed a boom in P2P lending, enabling thousands of plat (Quiz_Interest), the effects of inflation (Quiz_Inflation), and the financial
forms and hundreds of billions of dollars in loans, easing financial risk of stocks and bonds (Quiz_Risk). Three types of answers are recorded
constraints faced by small businesses and consumers. As one of the for each question: (1) the respondent understood the question and
booming fintech industries, China’s P2P online lending credit scale answered correctly (Correct); (2) the respondent understood the ques
reached RMB 3.9 trillion in 2017, equivalent to USD 700 billion tion but did not give the correct answer (Incorrect); (3) the respondent
(Nemoto et al., 2019). However, years of unregulated expansion have did not understand the question and could not answer (Do_Not_Know).
given rise to numerous cases of fraud, ultimately leading to the closure
of these platforms in the Chinese financial market.
3
The 2017 wave only interviewed approximately 1000 new respondents
about financial literacy, and data on digital finance are only available from the
2017 wave.
3
Appendix A1 contains the exact wording of the three questions that (employed/entrepreneur), and political status (communist party mem
measure financial literacy. ber); and geographic factors—rural or urban area, local economic and
Table 1 reports the descriptive statistics for the responses to these financial development, and Internet access. Specifically, we include the
three questions. We used sampling weights to ensure that our statistics age of the household head (Age) and its square (Age2) to control for
are representative of the population. The weighted percentages of cor lifecycle factors; education level (Education), marital status (Married),
rect responses to the Quiz_Interest, Quiz_Inflation, and Quiz_Risk questions health status (Unhealth_Ratio), family size (Size_Household), and political
are 28.5%, 16.2%, and 51.4%, respectively.4 However, the weighted status (Party) to reflect the individual’s/household’s background; and
percentages of respondents who did not understand and could not home ownership (Homeowner), employment status (Employed or Entre
answer the question are 48.6% (Quiz_Interest), 45.9% (Quiz_Inflation), preneur), total wealth (Wealth), and total disposable income (Income) to
and 39.0% (Quiz_Risk). This result suggests that the respondents sur control for the household financial situation.
veyed have limited financial/economic knowledge. The prevalence of Furthermore, we control for householders’ cognitive abilities by
financial illiteracy in China poses a significant obstacle to the use of including two additional dummy variables, Ability_High1 and Abil
digital finance in households. ity_High2. Ability_High1 is equal to 1 if the respondent could answer the
We follow Lusardi and Mitchell (2008), van Rooij et al. (2011), and survey questions with little assistance from the interviewer and 0 if
Feng et al. (2019) in constructing two dummy variables to differentiate otherwise. Then, Ability_High2 is equal to 1 if the respondent was overall
between the “incorrect” and “do not know” answers. Quiz_Correct is able to understand the questions and 0 if otherwise. Respondents who
equal to 1 if the respondent answered the question correctly and 0 if could answer the questions by themselves and understand them easily
otherwise, and Quiz_DK is equal to 1 if the respondent chose the “do not should have high cognitive abilities. The perceived cognitive ability of
know” answer and 0 if otherwise. We compute financial literacy scores the respondent is based on the subjective assessment of the respondent
for each respondent using the six indicators associated with these three by the interviewer and the objective quality of the respondent’s re
financial literacy questions. Specifically, we identify the presence of a sponses to the questions.
single factor using factor analysis as only one component exhibits ei Last, we include a dummy variable for whether the household lives in
genvalues greater than one. Next, we conduct an iterated principal a rural area (Rural), the number of bank branches in the community
factor analysis to compute our composite index of financial literacy where the household resides (No_Branches), broadband access rates
(Literacy_Index). An advantage of factor analysis is that it considers the (Broadband_Access), and the gross regional product per capita (GRP_PC)
difference between “incorrect” and “do not know” answers but does not in the city where the household resides. The objective is to control for
assume that a respondent who provided an “incorrect” answer had the potential impacts of local economic and financial development on
greater financial knowledge than one who responded “do not know.” the use of digital finance. We also include province dummies to control
Appendix A2 and Table A2 give details of the factor analysis. for uneven development across different provinces. As shown in Table 2,
Alternatively, we use rating scales (Literacy_Score1 and Literacy_ the summary statistics for these variables are roughly in line with Song
Score2) to measure the degree of respondents’ financial knowledge. et al. (2020) and Wang et al. (2021). See Appendix B for detailed defi
Literacy_Score1 is a variable with a three-point scale. Specifically, if the nitions of all the variables.
respondent answered the question correctly, then she/he receives 1
point and 0 if otherwise. With three financial literacy questions, the 3.5. Stylized facts
value of Literacy_Score1 ranges from 0 to 3, with 0 representing the
lowest level of financial literacy and 3 the highest. Likewise, Literacy_ Table 2 presents descriptive statistics for the key variables used in
Score2 is a variable with a six-point scale. Specifically, the respondent this research. On average, 29.1% of the households in our sample use
receives two points if she/he answered the question correctly (Correct); digital finance. Among the respondents, 28.1% used mobile payments,
one point if she/he understood the question but gave an incorrect 6.6% used online financial products, and 4.4% used online borrowing.
answer (Incorrect); and zero points if she/he did not understand the The mean values of Literacy_Index, Literacy_Score1 and Literacy_Score2
question (Do_not_Know). The value of Literacy_Score2 therefore ranges are 0.014, 0.892, and 2.484, respectively, suggesting that, on average,
from 0 to 6, with 0 reflecting the lowest level of financial literacy and 6 households possess a very low level of financial knowledge.
the highest.5 We further examine differences in financial literacy across popula
tion groups. The descriptive statistics presented in Table 3 reveal sig
nificant gaps among regions. For example, on average, urban
3.4. Demographic, socio-economic, and geographic factors households have a financial literacy index (Literacy_Index) of 0.315,
which is significantly higher than that of rural households (− 0.364).
In addition to the financial literacy index, our empirical specification These results are in line with those of Cui et al. (2019), who find that
considers other determinants of household financial behavior, which people living in urban areas have better financial capabilities than those
have been widely recognized by previous studies (Haliassos and Bertaut in rural areas. The financial literacy of households in the economically
1995; van Rooij et al. 2011, 2012; Lusardi and Mitchell 2014; Gau well-developed eastern regions (0.065) is significantly higher than that
decker 2015). We control for demographic factors—age, gender, marital of households in the western (0.043) and central (− 0.062) regions with
status, health status, and household size; socio-economic fac lower per capita income. Moreover, women are more financially literate
tors—financial situation, personal income, home ownership, occupation than men, younger people more than mature people, and married people
more than unmarried people. These results are different from the find
ings of Atkinson and Messy (2012), who report that overall female re
4
Our data on the three financial literacy questions (Quiz_Interest, Quiz_Infla spondents have lower financial knowledge than male counterparts using
tion, and Quiz_Risk) are consistent with Feng et al. (2019), who also used the a sample covering 13 countries across four continents. According to
2015 wave of the CHFS data to analyze the impact of financial literacy on Lusardi and Mitchell (2007), financial literacy is notably deficient
household debt and assets. They show that 30.2%, 17.6%, and 54.6% of re
among young individuals. Additionally, Agarwal et al. (2009) find that
spondents correctly answered questions about interest rates, inflation, and risk,
younger and older adults make more financial errors than middle-aged
respectively.
5
The Literacy_Score2 variable entails an assumption that a respondent who adults. In China, older individuals tend to have lower levels of finan
provides an “incorrect” answer has greater financial knowledge than the one cial knowledge than younger counterparts probably because of their
who provides a “do not know” answer, which might not be entirely fair. long experience of living in a centrally planned economy. Individuals
Therefore, we only use Literacy_Score2 in robustness tests as it may over- who grow up with limited formal financial literacy education often find
interpret “do not know” answers. themselves compelled to acquire financial knowledge through their own
4
Table 1
Responses to the three questions about financial literacy.
Quiz_Interest Quiz_Inflation Quiz_Risk
Number Percentage Number Percentage Number Percentage
Correct 10,412 28.5% 5901 16.2% 17,911 51.4%

Incorrect 8358 22.9% 13,806 37.9% 3364 9.7%
Do_Not_Know 17,726 48.6% 16,728 45.9% 13,589 39.0%
Aggregate 36,496 100.0% 36,435 100.0% 34,864 100.0%
Notes: This table reports the number of respondents and the weighted percentage of households providing correct, incorrect and Do_Not_Know answers to each of the
financial literacy questions. The data are from the 2015 CHFS. The three questions concern the following concepts: numeracy and capacity to perform calculations
relating to interest rates such as compound interest (Quiz_Interest); understanding of inflation (Quiz_Inflation); and understanding of financial risk (Quiz_Risk). We use
sampling weights to ensure that the statistics are representative of the population. Appendix B provides detailed definitions of all the variables.
Table 2
Summary statistics of the key variables.
N Mean SD. Min Max
Digital finance 19,788 0.2908 0.4542 0 1

Dependent Variables Digital_Payment 19,788 0.2811 0.4495 0 1
Online_FP 19,788 0.0661 0.2485 0 1
Online_Borrowing 19,788 0.0437 0.2044 0 1
Independent Variables Literacy_Index 19,788 0.0144 1.0704 − 1.7631 1.4782
Literacy_Score1 19,788 0.8928 0.8989 0 3
Literacy_Score2 19,788 2.4838 1.9702 0 6
Financial_Class 19,788 0.0565 0.2309 0 1
Log(Wealth) 19,788 12.3487 1.6879 7.0246 16.2186
Control Variables Log(Income) 19,788 10.5335 1.2956 6.2166 13.3897
Age 19,788 53.9158 13.0647 25 87
Age^2/100 19,788 30.7759 14.5627 6.2500 75.6900
Male 19,788 0.8055 0.3958 0 1
Single Female 19,788 0.0579 0.2336 0 1
Married 19,788 0.8065 0.3950 0 1
Education 19,788 8.8501 4.0116 0 22
Size_Household 19,788 4.1512 1.8096 1 10
Unhealthy_Ratio 19,788 0.1187 0.2237 0 1
Employed 19,788 0.4401 0.4964 0 1
Homeowner 19,788 0.8978 0.3029 0 1
Entrepreneurship 19,788 0.1581 0.3648 0 1
Party 19,788 0.1550 0.3619 0 1
Ability_High1 19,788 0.5553 0.4969 0 1
Ability_High2 19,788 0.5964 0.4906 0 1
Rural 19,788 0.4431 0.4968 0 1
No_Branches 19,788 1.0784 1.8184 0 25
Broadband_Access 19,788 0.4199 0.2766 0.0428 1
Log(GRP_PC) 19,788 10.6740 0.5389 9.5597 11.9150
Notes: This table reports summary statistics of the variables. We use sampling weights to ensure that the statistics are representative of the population. Detailed
definitions of all the variables are provided in Appendix B.
life experiences. A correlation exists between education/wealth/income 4. Financial literacy and digital finance
and the level of financial literacy. People with higher education, wealth,
and income are likely to have higher financial literacy than those with 4.1. Baseline results
lower levels of education, personal wealth, and income.
Table 4 compares the use of digital finance among various popula We first estimate the impact of financial literacy on the use of digital
tion groups. Similar to financial literacy, the use of digital finance also finance using a probit model. Table 5 reports the regression results with
varies by region. Specifically, access to digital finance is more prevalent cluster-robust standard errors.6
among households in the eastern regions, where 32.7% have access, Column 1 shows that financial literacy significantly increases the
compared with 24.2% in the central regions and 30.3% in the western probability of using digital finance, with a marginal effect of 2.69%.
regions. The disparity between rural and urban areas is remarkable. Of Given that the standard deviation of the financial literacy index is 1.070,
the urban residents, 42.4% have used digital financial services, whereas a one standard deviation increase in financial literacy is associated with
the proportion is only 12.4% among rural residents. Additionally, a 2.87 percentage point (=2.69% × 1.070) increase in the probability of
younger, married, and female householders are more likely to use digital
finance than their older, unmarried, and male counterparts. The use of
digital finance increases with wealth, income, and education levels. 6
Importantly, the percentage of households using digital finance in Although pooled probit models cannot consider unobserved heterogeneity,
they provide consistent estimates of the relevant parameters. In our study, we
creases with the financial literacy index, suggesting that households
employ clustered standard errors at the community level to ensure appropriate
with greater financial knowledge are more likely to use digital finance.
inference. The main advantages of pooled estimations are that exogenous re
gressors are not strictly necessary, and predetermined variables can be
accommodated (Wooldridge 2010; Ding et al. 2021). As a result, pooled probit
models can be more robust than random-effects probit models, which assume
strict exogeneity.
5
Table 3
Financial literacy by demographics.
Literacy_Index Literacy_Score1 Literacy_Score2
Nation 0.0144 0.8928 2.4838

Urban 0.3153 1.1310 3.0855
Rural − 0.3638 0.5934 1.7275
Eastern 0.0646 0.9391 2.5653
Central − 0.0616 0.8424 2.3994
Western 0.0433 0.8900 2.4721
Female 0.1679 0.9665 2.6580
Male − 0.0227 0.8750 2.4417
Young_Grp 0.3743 1.2038 3.2538
Middle_Grp 0.0572 0.9212 2.5755
Old_Grp − 0.3032 0.6290 1.8041
Married 0.0319 0.9083 2.5299
Unmarried − 0.1113 0.7810 2.1520
No Schooling − 0.6315 0.3343 1.0237
Primary_Education − 0.1342 0.7471 2.1205
Secondary_Education 0.3150 1.1271 3.1520
Higher_Education 0.7265 1.6279 4.2421
Wealth_Grp1 − 0.3669 0.5718 1.6676
Wealth_Grp2 0.0111 0.8705 2.4606
Wealth_Grp3 0.4263 1.2617 3.3837
Income_Grp1 − 0.4133 0.5569 1.6331
Income_Grp2 0.0547 0.8916 2.4934
Income_Grp3 0.4053 1.2328 3.3319
Notes: This table reports the means of the literacy measures by demographics. We use sampling weights to ensure that the statistics are representative of the population.
For the age groups, Young_Grp refers to household heads younger than 44 years old; Middle_Grp refers to household heads aged between 44 and 55; Old_Grp refers to
household heads older than 60. For the education groups, No_Schooling, Primary_Education, Secondary_Education and Higher_Education mean 0, up to 9, 9–12 and more
than 12 years of education, respectively. For income/wealth groups, we group the corresponding variables based on the terciles of the distribution of the values.
Detailed definitions of all the variables are provided in Appendix B.
using digital finance. Considering that the mean value of the use of likely to use digital finance. A one percent increase in household wealth
digital finance is 0.291, this is a sizable effect. and income respectively lead to a 0.04 percentage point and a 0.05
Focusing on the other regressors, we observe significant and positive percentage point rise in the probability of using digital finance.
marginal effects associated with household wealth and income. This The financial literacy index we use is widely used to measure
result suggests that wealthier and higher-income households are more household financial knowledge (Lusardi and Mitchell 2011, 2014;
Table 4
Digital finance usage by subgroups.
Digital_Finance Mobile_Payments Online_FPs Online_Borrowing
Nation 29.1% 28.1% 6.6% 4.4%

Urban 42.4% 41.0% 10.3% 6.8%
Rural 12.4% 11.9% 2.0% 1.3%
Eastern 32.7% 31.4% 9.2% 5.2%
Central 24.2% 23.4% 4.6% 3.0%
Western 30.3% 29.5% 5.3% 5.0%
Female 36.4% 35.8% 9.3% 5.8%
Male 27.3% 26.2% 6.0% 4.0%
Young_Grp 54.8% 53.3% 13.5% 9.4%
Middle_Grp 28.9% 27.9% 6.2% 4.0%
Old_Grp 10.4% 10.0% 2.1% 1.2%
Married 30.0% 29.0% 6.8% 4.4%
Unmarried 22.6% 22.0% 5.6% 4.1%
No_Schooling 7.6% 7.1% 1.3% 1.1%
Primary_Education 20.8% 20.1% 3.5% 2.5%
Secondary_Education 39.7% 38.7% 8.6% 6.2%
Higher_Education 68.1% 65.2% 22.4% 13.0%
Wealth_Grp1 8.4% 8.2% 1.1% 1.1%
Wealth_Grp2 26.2% 25.3% 4.6% 3.6%
Wealth_Grp3 54.4% 52.5% 14.8% 8.8%
Income_Grp1 9.3% 8.9% 1.2% 1.2%
Income_Grp2 25.5% 24.9% 3.6% 2.7%
Income_Grp3 52.6% 50.6% 15.1% 9.2%
literacy_Grp1 11.1% 10.7% 1.5% 1.0%
literacy_Grp2 43.9% 42.4% 11.7% 7.4%
literacy_Grp3 40.4% 39.1% 8.9% 6.3%
Notes: This table reports the weighted percentages of financial literacy measures by demographics. We use sampling weights to ensure that the statistics are repre
sentative of the population. For the age groups, Young_Grp refers to household heads younger than 44 years old; Middle_Grp refers to household heads aged between 44
and 55; Old_Grp refers to household heads older than 60. For the education groups, No_Schooling, Primary_Education, Secondary_Education and Higher_Education mean 0,
up to 9, 9–12 and more than 12 years of education, respectively. For income/wealth/literacy groups, we group the corresponding variables based on the terciles of the
distribution of the values. Detailed definitions of all the variables are provided in Appendix B.
6
Table 5
Baseline results of probit model regressions.
(1) (2) (3) (4)
Literacy_Index 0.0269*** 0.0264*** 0.0090*** 0.0080***

(0.0026) (0.0026) (0.0018) (0.0015)
Ln(Wealth) 0.0420*** 0.0413*** 0.0127*** 0.0059***
(0.0026) (0.0026) (0.0017) (0.0012)
Ln(Income) 0.0504*** 0.0486*** 0.0236*** 0.0110***
(0.0029) (0.0029) (0.0024) (0.0018)
Age − 0.0107*** − 0.0103*** − 0.0025*** − 0.0035***
(0.0016) (0.0016) (0.0009) (0.0007)
Age^2/100 0.0036** 0.0034** 0.0005 0.0020***
(0.0015) (0.0015) (0.0009) (0.0006)
Male − 0.0194*** − 0.0203*** − 0.0048 0.0008
(0.0071) (0.0071) (0.0043) (0.0035)
Single_Female 0.0430*** 0.0430*** 0.0135 0.0076
(0.0149) (0.0151) (0.0085) (0.0071)
Married − 0.0009 − 0.0014 0.0050 − 0.0029
(0.0081) (0.0081) (0.0049) (0.0037)
Education 0.0057*** 0.0051*** 0.0034*** 0.0013***
(0.0009) (0.0009) (0.0006) (0.0005)
Size_Household 0.0106*** 0.0109*** 0.0004 0.0019**
(0.0017) (0.0017) (0.0011) (0.0008)
Unhealthy_Ratio − 0.0837*** − 0.0874*** − 0.0323*** − 0.0206**
(0.0154) (0.0154) (0.0122) (0.0100)
Employed 0.0204*** 0.0220*** − 0.0002 0.0076**
(0.0060) (0.0061) (0.0040) (0.0032)
Homeowner − 0.0695*** − 0.0663*** − 0.0212*** − 0.0188***
(0.0092) (0.0092) (0.0050) (0.0039)
Entrepreneurship 0.0576*** 0.0544*** 0.0071* 0.0119***
(0.0072) (0.0072) (0.0040) (0.0031)
Party − 0.0212*** − 0.0164** − 0.0097** − 0.0169***
(0.0074) (0.0073) (0.0044) (0.0037)
Ability_High1 0.0270*** 0.0264*** 0.0085** 0.0037
(0.0064) (0.0064) (0.0040) (0.0033)
Ability_High2 0.0336*** 0.0323*** 0.0086** 0.0067*
(0.0066) (0.0065) (0.0043) (0.0037)
Rural − 0.0708*** − 0.0697*** − 0.0185*** − 0.0172***
(0.0075) (0.0075) (0.0048) (0.0038)
No_Branches 0.0032** 0.0021 0.0014 0.0001
(0.0016) (0.0016) (0.0009) (0.0007)
Broadband_Access 0.0075 0.0172 − 0.0080 − 0.0015
(0.0238) (0.0242) (0.0137) (0.0101)
Ln(GRP_PC) 0.0222** 0.0228** 0.0035 0.0006
(0.0103) (0.0105) (0.0062) (0.0046)
(△Y/△X) / E(Y) 9.25% 9.39% 13.62% 18.31%
Province Fixed Effects Yes Yes Yes Yes
N 19,788 19,788 19,788 19,788
pseudo R2 0.3352 0.3278 0.2513 0.2125
Notes: The table reports marginal effects and standard errors (in parentheses). All the specifications were estimated using the probit estimator. Test statistics and
standard errors of all the variables in the regressions are clustered at the community level. (△Y/△X)/E(Y) is the marginal effect associated with Literacy_Index
evaluated at the mean values of the different forms of digital finance. See Appendix B for detailed definitions of all the variables. ***, ** and * indicate significance at
the 1%, 5% and 10% levels, respectively.
Lusardi and de Bassa Scheresberg 2013; Grohmann et al., 2018). How digital finance but does not necessarily reflect financial knowledge or
ever, it may be a concern that this index, serving as a proxy for per cognitive ability. The significant and positive marginal effects of edu
sonality traits, could also capture cognitive ability. We therefore cation indicate its potential role in overcoming barriers to the use of
consider this concern in Table 5 by including additional controls (Abil digital finance owing to ignorance and misperceptions (Haliassos and
ity_High1 and Ability_High2) for cognitive ability. We find that these Bertaut 1995).
proxies for cognitive ability are often significant and have the expected We also find that employment and entrepreneurship positively
signs that cognitive ability has a positive effect on the use of digital impact the use of digital finance. Nonetheless, households that own their
finance. Notably, if our financial literacy index only reflects cognitive homes are less inclined to use digital finance, which may indicate a
ability, then we would no longer see any significant effect of the preference for investing in property relative to other types of financial
financial literacy index on the use of digital finance in our regressions.7 assets. We observe that the likelihood of using digital finance is signif
In addition, we find that the use of digital finance significantly increases icantly lower among men than women (particularly among those who
with household education. Education has a positive impact on the use of are single), communist party members than non-party members, and
those living with unhealthy family members and smaller family sizes
than those with healthy family members and larger family sizes.
Following Finke et al. (2017), we test the influence of age and find that
7
Our results are robust when we estimate our models without accounting for age is negatively and significantly related to the use of digital finance,
cognitive ability, which shows the significantly positive marginal effects asso whereas the square of age is significantly positively related. This finding
ciated with the financial literacy index on the use of digital finance. For brevity, suggests a U-shaped relationship between age and the use of digital
these results are not reported, but they are available upon request.
7
finance. However, when looking at the magnitudes of the single and online financial products (column 3), and online borrowing (column 4).
squared terms, we find that the turning point of the U-shaped relation In addition, financial literacy has a higher impact on online borrowing
ship between age and the use of digital finance (i.e., the quadratic graph and online financial products than on mobile payments when evaluated
goes from downward to upward sloping) is greater than 100.8 Therefore, at the mean values of these three forms of digital finance. For example,
age overall has a negative impact on the use of digital finance. The as shown in Panel A, the impacts of financial literacy for online financial
negative marginal effects of Rural and the positive marginal effects of products and online borrowing are 25.11% and 18.99%, respectively,
GRP_PC and No_Branches show that the use of digital finance is higher in which are greater than that for mobile payments (11.85%). These results
urban areas and regions with higher income per capita and financial confirm the importance of financial literacy in the adoption of highly
outreach than in rural areas and regions with lower income per capita complicated financial products.
and financial outreach. As for the other explanatory variables, the estimates are qualitatively
Columns 2–4 of Table 5 show the relationships between financial similar to those reported in Table 5, but we do not report them for
literacy and the use of mobile payments, online financial products, and brevity. In summary, these results suggest that our main findings are
online borrowing. Specifically, even after accounting for demographic, robust to the use of different financial literacy indicators.
socio-economic, and geographic factors, financial literacy has a signifi
cantly positive impact on the use of all of them. The corresponding 4.2.2. IV estimation
marginal effects on the use of mobile payments, online financial prod Previous studies often considered financial literacy to be endogenous
ucts, and online borrowing are 2.64% (column 2), 0.9% (column 3), and (van Rooij et al. 2011). Although our baseline specification controls for
0.8% (column 4), respectively. The adoption rates of mobile payments, different demographic, socio-economic, and geographic factors, unob
online financial products, and online borrowing vary significantly with servable factors may drive our results. For instance, households with
mean values of 0.281, 0.066, and 0.044, respectively. Due to these high levels of financial knowledge may exhibit great optimism, poten
disparities, we cannot directly compare the effects of financial literacy tially making them inclined to engage with innovative digital finance
on these three forms of digital finance by analyzing their marginal ef solutions. Additionally, digital financial tools can potentially elevate
fects. Instead, we use the mean values of these three forms of digital users’ financial literacy by offering accessible, personalized, and
finance as reference levels to assess the impact of financial literacy engaging educational content. Such digital tools may also encourage
evaluated at these mean values.9 Specifically, a one-unit increase in the users to seek relevant financial information or acquire knowledge
financial literacy index enhances the probability of using mobile pay through experience, thereby helping them understand financial con
ments, online financial products, and online borrowing by 9.39%, cepts and adopt best practices. Unobservable factors and simultaneity
13.62%, and 18.31%, respectively, when evaluated at these mean may result in a spurious relationship between financial literacy and the
values. These figures reveal that financial literacy plays a more impor use of digital finance.
tant role in promoting the use of online borrowing and online financial We employ an IV approach to infer the causal impact of financial
products than mobile payments. The reason may be that financial lit knowledge on the use of digital finance to overcome these endogeneity
eracy has a greater influence on the adoption of more complex digital concerns. Specifically, following van Rooij et al. (2011), we first in
finance products, such as online borrowing and online financial prod strument respondents’ financial literacy by the education levels of their
ucts, than mobile payments, which are accessible, user-friendly, and less parents, which are measured as years of schooling. The intuition is that
risky. The signs and significance of the other control variables are parents’ education is unlikely to change as a result of the financial
generally similar to those in our baseline models (column 1). behavior of their children but is likely to be correlated with their chil
dren’s financial knowledge. Second, inspired by Bucher-Koenen and
Lusardi (2011), we use neighborhood financial literacy as another in
4.2. Robustness tests
strument for household financial knowledge. In particular, we divide
local neighborhoods into three groups according to their total wealth
We conduct a series of robustness tests to check the validity of our
and calculate the mean financial literacy value for each group (i.e.,
results. These tests use alternative measures of financial literacy, employ
Wealth_Grp1, Wealth_Grp2, and Wealth_Grp3). The rationale is that an
an IV approach, and consider the influence of peer effects and voluntary
individual’s financial knowledge is likely to be linked with that of
self-exclusion.
households belonging to the same community and similar wealth groups
although this linkage would not directly affect a household’s adoption of
4.2.1. Different measures of financial literacy
digital finance. Third, we use the respondent’s social identity as an in
We first verify whether our results are robust to different proxies for
strument for financial literacy. Specifically, we use the respondent’s
financial literacy. Panels A and B of Table 6 present the estimates based
place of birth, which is recorded on his or her household registration
on Literacy_Score1 and Literacy_Score2, respectively. In line with our
book or “hukou bu,”10 dummies for the decade in which the respondent
main results, financial literacy is significantly and positively associated
was born (e.g., the 1970s, 1980s) and interaction terms between them to
with the use of digital finance regardless of the different measures used.
reflect the person’s social identity (Afridi et al., 2015). In theory,
Panel C further measures financial knowledge based on whether the
householders’ places and dates of birth are reasonably exogenous (and
respondent has taken any finance–related classes. Specifically, Finan
are out of their control). However, they strongly influence individuals’
cial_Class, a dummy variable, is equal to 1 if the respondent has attended
social identities, which might explain cross-sectional differences in in
an economics and/or finance class in the past and 0 if otherwise. The
dividuals’ financial literacy.
finding further confirms that financial literacy increases the use of dig
We then re-estimate our baseline models using the instrumental
ital finance.
approach. Panels A, B, and C of Table 7 present the IV probit estimates
Similar results are found for the use of mobile payments (column 2),
when the financial literacy index is instrumented by the education levels
of the respondent’s parents, neighborhood financial literacy, and the
8
respondent’s social identity, respectively. Column 1 in Panels A and B
For example, in column 1, given that the magnitudes of the marginal effects
reports the coefficients from our first-stage regression of financial
associated with the single term (Age) and the squared term (Age2/100) are
− 0.0107 and 0.0036, respectively, the turning point of age in the quadratic
graph is 148.6=0.0107/(2×0.0036/100).
9 10
In other words, we aim to find how the marginal effects compare relatively The Chinese hukou system, also known as the household registration sys
to the vastly different means. Notably, standardizing binary outcome variables tem, was introduced in 1958. Every citizen must be registered at birth with this
may render the results meaningless. basic demographic information.
8
Table 6
Probit model regressions: alternative measures of financial literacy.
(1) (2) (3) (4)
Panel A Digital_Finance Mobile_Payments Online_FPs Online_Borrowing
Literacy_Score1 0.0349*** 0.0333*** 0.0166*** 0.0083***

(0.0029) (0.0029) (0.0018) (0.0014)
(△Y/△X) / E(Y) 12.00% 11.85% 25.11% 18.99%
Other Control Variables Yes Yes Yes Yes
N 19,788 19,788 19,788 19,788
pseudo R2 0.3364 0.3286 0.2573 0.2129
Panel B Digital_Finance Mobile_Payments Online_FPs Online_Borrowing
Literacy_Score2 0.0215*** 0.0209*** 0.0091*** 0.0043***

(0.0015) (0.0015) (0.0010) (0.0008)
(△Y/△X) / E(Y) 7.39% 7.44% 13.77% 9.84%
N 19,788 19,788 19,788 19,788
pseudo R2 0.3396 0.3320 0.2575 0.2125
Panel C Digital_Finance Mobile_Payments Online_FPs Online_Borrowing
Financial_Class 0.1953*** 0.1772*** 0.0728*** 0.0309***

(0.0380) (0.0367) (0.0142) (0.0115)
(△Y/△X) / E(Y) 67.16% 63.04% 110.14% 70.71%
N 19,788 19,788 19,788 19,788
pseudo R2 0.3318 0.3243 0.2517 0.2093
literacy on the education level of the respondent’s parents and neigh associated with an increase in the probability of using digital finance,
borhood financial literacy, respectively. The coefficients on both IVs are ranging from 2.28 to 7.30 percentage points. Moreover, the Kleibergen-
significantly positive at the 1% level, which satisfies the relevance Paap rk LM and the Anderson tests for instrument validity have signif
condition. Individuals whose parents are well-educated tend to have icant p-values (i.e., p-values less than 0.05), rejecting the null hypothesis
great financial knowledge. Additionally, a respondent’s understanding that the equation is under-identified. The tests suggest that our in
of financial matters is closely related to his or her neighborhood. These struments are adequate for identifying equations. In short, our main
results indicate possible channels through which financial knowledge results are robust in accounting for the potential endogeneity of finan
may be acquired. Panel C does not include the results of the first-stage cial literacy: people with higher financial literacy are more likely to use
regression owing to an excessive number of instruments, which consist digital finance than those with less financial literacy.
of dummies for householders’ places and decades of birth and their in
teractions. However, regardless of the instruments used, the first-stage 4.2.3. Peer effects
F-statistics are greater than the rule of thumb of 10, suggesting that Previous research showed that the experiences of peers can influence
our instruments are valid and have high explanatory power (Stock and respondents’ portfolio choices (Hong et al., 2004; Brown et al., 2008).
Yogo 2005). In addition, Wald tests for the exogeneity of the instru Our IV results also suggest that interactions with individuals with close
mented variables reject the null hypothesis of no endogeneity. ties, such as family and neighbors, are a potential channel for acquiring
Columns 2–5 of Table 7 present the second-stage regression results of financial knowledge. In this case, information about using digital
digital finance, mobile payments, online financial products, and online finance could spread through peer groups via word of mouth. This
borrowing, respectively. The results again show that financial literacy subsection investigates whether our estimates hold after controlling for
has a positive and statistically significant impact regardless of the in peer effects of using digital finance. To this end, we construct a new
struments employed, confirming the importance of financial literacy in financial literacy index in Table 8 by subtracting peer financial literacy
promoting the use of digital finance. In particular, a 10 percentage point from the respondent’s financial literacy index. Specifically, we first
increase in financial literacy enhances the probability of using digital construct 24 subgroups based on the interaction between age (six age
finance by a range of 1.71 (Panel B) to 5.71 (Panel C) percentage points groups: 18–30, 31–40, 41–50, 51–60, 61–70, and >70) and education
(column 2). We also calculate the marginal effects of financial literacy, (four education groups as shown in Table 3) in a given province. Then,
evaluated at the mean values of these three types of digital finance, to we calculate the average value of the financial literacy index as a proxy
compare the impact of financial literacy on the three forms of digital for peers’ financial literacy. After considering peer group effects, the
finance. In line with the baseline estimation results, financial literacy marginal effects of financial literacy remain significantly positive. In
has a greater influence on online borrowing and online financial prod addition, the marginal effects of financial literacy relative to their mean
ucts than mobile payments, except for the results in Panel B. values are more pronounced for online borrowing (16.93%) and online
In unreported results, we re-estimate the empirical models using the financial products (12.86%) than mobile payments (8.64%).
two-stage least-squares IV approach. The coefficients on Literacy_Index
are still significant and positive for the different forms of digital finance, 4.2.4. Voluntary self-exclusion
that is, mobile payments, online financial products, and online Empirical analyses to date have assumed that the exclusion of re
borrowing. In particular, a 10% increase in financial literacy is spondents from digital finance is involuntary, resulting from a lack of
9
Table 7
Robustness check: instrumental variable probit model regressions.
(1) (2) (3) (4) (5)
Panel A First Stage Digital_Finance Mobile_Payments Online_FPs Online_Borrowing
Literacy_Index 0.4440*** 0.4290*** 0.1232*** 0.1078***

(0.0408) (0.0408) (0.0252) (0.0210)
Parents’ Education 0.0183***
(0.0020)
(△Y/△X) / E(Y) 152.68% 152.61% 186.38% 246.68%
Other Control Variables Yes Yes Yes Yes Yes
Province Fixed Effects Yes Yes Yes Yes Yes
P-value Exogeneity 0.000*** 0.000*** 0.000*** 0.000***
First Stage F-stat 81.180
N 15,862 15,862 15,862 15,862 15,862
Panel B First Stage Digital_Finance Mobile_Payments Online_FPs Online_Borrowing
Literacy_Index 0.1714*** 0.1713*** 0.0377* 0.0336**

(0.0311) (0.0319) (0.0208) (0.0158)
Neighbourhood Financial Literacy 0.1635***
(0.0198)
(△Y/△X) / E(Y) 58.94% 60.94% 57.03% 76.89%
First Stage F-stat 69.280 131.42
N 19,364 19,364 19,364 19,364 19,364
Panel C First Stage Digital_Finance Mobile_Payments Online_FPs Online_Borrowing
Literacy_Index 0.5711*** 0.5649*** 0.2071*** 0.1425***

(0.0163) (0.0163) (0.0117) (0.0100)
Hukou_Decade_Dummy
(△Y/△X) / E(Y) 196.39% 200.96% 313.31% 326.09%
First Stage F-stat 41.500
N 19,788 19,788 19,788 19,788 19,788
Notes: The table reports marginal effects and standard errors (in parentheses). We estimate all the specifications using the instrumental variable (IV) probit estimator.
Test statistics and standard errors of all the variables in the regressions are clustered at the community level. The p-value exogeneity test is the Wald test of exogeneity
of the instruments. The F-statistics of the first stage regression of Literacy_Index are also reported. (△Y/△X)/E(Y) is the marginal effect associated with Literacy_Index
evaluated at the mean values of the different forms of digital finance. See Appendix B for detailed definitions of all variables. ***, ** and * indicate significance at the
1%, 5% and 10% levels, respectively.
financial knowledge. However, some households may have access to interest in digital finance are indeed less likely to use digital finance.12 In
digital finance services, but they may opt not to use them as they Panel B, we further removed households that voluntarily excluded
perceive no need for them, resulting in voluntary exclusion from digital themselves from digital finance to alleviate self-selection bias, and the
finance.11 We consider respondents to be voluntarily self-excluded from results remain qualitatively unchanged.
using digital finance if they answered “no need or no interest” to the
follow-up question “Why don’t you use digital finance?” Specifically, we 5. Further tests
construct a dummy variable (Self-Exclusion) which is equal to 1 if the
respondent answered “no need or no interest” and 0 if otherwise. In 5.1. Financial literacy working mechanisms
Panel A of Table 9, we include the Self-Exclusion variable as a control
variable in our regressions. For brevity, we only report the probit esti Thus far, our findings suggest that increasing financial literacy can
mates of the financial literacy and self-exclusion variables. We find that facilitate the use of digital finance. The literature suggests that the cost
the marginal effect of financial literacy on the use of digital finance of acquiring information, lack of trust in new technology, and risk
remains significantly positive. Moreover, the marginal effects of finan aversion could be potential obstacles to using digital finance.
cial literacy concerning their mean values are more prominent for online First, financial literacy equips individuals with basic financial con
borrowing and online financial products than for mobile payments. cepts, such as interest rates, asset allocation, diversification, and
Notably, Self-Exclusion has a significantly negative impact on the use of financial risk. This level of proficiency can reduce the costs of obtaining
digital finance, confirming that households who perceived no need or information related to the use of digital finance. For example, financial
literacy can assist households in collecting and processing economic
information from diverse sources, including the Internet. Financial lit
11
eracy, in turn, can aid households in making informed decisions about
We do not exclude from the sample households those who voluntarily self-
digital financial services and products, which can be complex and
exclude from using digital finance because households, which do not need or
confusing (Wang et al., 2021; Niu et al., 2020).
are not interested in digital finance, may lack sufficient financial knowledge,
and thus, they feel that digital finance is not necessary. To the best of our Second, financial literacy can bolster consumers’ trust and confi
knowledge, this study is the first to consider voluntary self-exclusion when dence in using digital financial services (Hansen 2012; Malady 2016).
examining the use of digital finance. This issue has been largely ignored in
previous research. We keep the voluntary self-exclusion sample in our main
regressions to make easy comparisons with prior research.
12
The self-exclusion variable is omitted in column 3 because of collinearity.
10
Table 8
Probit model regressions: taking peer effects into consideration.
(1) (2) (3) (4)
Literacy_Index 0.0249*** 0.0243*** 0.0085*** 0.0074***

(0.0026) (0.0026) (0.0018) (0.0015)
Ln(Wealth) 0.0422*** 0.0416*** 0.0128*** 0.0059***
(0.0026) (0.0026) (0.0017) (0.0012)
Ln(Income) 0.0506*** 0.0488*** 0.0236*** 0.0111***
(0.0029) (0.0030) (0.0024) (0.0018)
Age − 0.0112*** − 0.0108*** − 0.0026*** − 0.0035***
(0.0016) (0.0016) (0.0009) (0.0007)
Age^2/100 0.0038** 0.0036** 0.0005 0.0021***
(0.0015) (0.0015) (0.0009) (0.0007)
Male − 0.0202*** − 0.0211*** − 0.0051 0.0006
(0.0071) (0.0071) (0.0043) (0.0035)
Single_Female 0.0431*** 0.0432*** 0.0134 0.0076
(0.0149) (0.0151) (0.0085) (0.0071)
Married − 0.0010 − 0.0014 0.0050 − 0.0029
(0.0081) (0.0081) (0.0049) (0.0037)
Education 0.0076*** 0.0069*** 0.0041*** 0.0019***
(0.0009) (0.0009) (0.0006) (0.0005)
Size_Household 0.0105*** 0.0108*** 0.0004 0.0018**
(0.0017) (0.0017) (0.0011) (0.0008)
Unhealthy_Ratio − 0.0845*** − 0.0883*** − 0.0325*** − 0.0209**
(0.0154) (0.0154) (0.0122) (0.0100)
Employed 0.0213*** 0.0229*** 0.0001 0.0078**
(0.0060) (0.0061) (0.0040) (0.0032)
Homeowner − 0.0695*** − 0.0663*** − 0.0212*** − 0.0188***
(0.0092) (0.0092) (0.0050) (0.0039)
Entrepreneurship 0.0574*** 0.0541*** 0.0070* 0.0118***
(0.0072) (0.0072) (0.0040) (0.0031)
Party − 0.0204*** − 0.0156** − 0.0097** − 0.0168***
(0.0074) (0.0073) (0.0044) (0.0037)
Ability_High1 0.0272*** 0.0267*** 0.0085** 0.0038
(0.0064) (0.0064) (0.0040) (0.0033)
Ability_High2 0.0341*** 0.0329*** 0.0088** 0.0068*
(0.0066) (0.0065) (0.0043) (0.0037)
Rural − 0.0717*** − 0.0706*** − 0.0188*** − 0.0175***
(0.0075) (0.0075) (0.0048) (0.0038)
No_Branches 0.0034** 0.0022 0.0014* 0.0002
(0.0016) (0.0016) (0.0009) (0.0007)
Broadband_Access 0.0075 0.0171 − 0.0082 − 0.0017
(0.0239) (0.0243) (0.0138) (0.0102)
Ln(GRP_PC) 0.0227** 0.0232** 0.0037 0.0007
(0.0104) (0.0106) (0.0062) (0.0047)
(△Y/△X) / E(Y) 8.56% 8.64% 12.86% 16.93%
N 19,788 19,788 19,788 19,788
pseudo R2 0.3345 0.3271 0.2510 0.2119
evaluated at the mean values of the different forms of digital finance. See Appendix B for detailed definitions of all variables. ***, ** and * indicate significance at the
1%, 5% and 10% levels, respectively.
Financial knowledge enhances households’ awareness of new financial choices can lead to overly cautious decision-making. Financial literacy
services, thereby enabling individuals to evaluate the costs, fees, and may enhance users’ risk awareness, help them identify and manage
risks associated with various digital financial services. Moreover, digital financial risks, navigate digital platforms, and conduct secure
financial knowledge helps individuals recognize the benefits of digital transactions. Furthermore, financial knowledge can help consumers
financial services, such as convenience, functionality, and accessibility, protect themselves from fraud and other risks associated with digital
which collectively contribute to building trust. By contrast, individuals financial services and safeguard their personal and financial
with low financial literacy may hesitate to use digital financial services information.
owing to a lack of trust. A survey conducted in Cyprus in 2018 found that In this section, we examine various channels through which financial
financially illiterate consumers are less likely to trust Internet banking literacy may affect the use of digital finance. Specifically, we first
and exhibit lower confidence in their digital and financial skills examine the extent to which financial literacy can enhance households’
(Andreou and Anyfantaki 2021). access to financial information online owing to a reduction in the costs
Third, financial literacy can reduce risk aversion and increase associated with information acquisition. Second, we test whether
tolerance to digital finance–related risks. New technologies often face financial literacy increases households’ trust in the use of new tech
difficulty in appealing to risk-averse individuals (Han et al., 2019). nologies, thereby facilitating their adoption of digital finance. Third, we
According to choice bracketing theory (Read et al.1999; Dohmen et al., test whether financial literacy increases households’ tolerance for
2010), people tend to behave in a risk-averse manner if they fail to financial risks, which may also lead to greater use of digital finance.
incorporate future considerations in their decision-making or to Table 10 presents the estimation results of Watch_Fin_Online, Trust,
perceive the broad context of their choices. Narrow bracketing of and Risk_Averse serving as mediator variables for the impact of financial
11
Table 9
Probit model regressions: taking voluntary self-exclusion into consideration.
(1) (2) (3) (4)
Panel A Digital_Finance Mobile_Payments Online_FPs Online_Borrowing
Literacy_Index 0.0270*** 0.0265*** 0.0095*** 0.0080***

(0.0026) (0.0026) (0.0019) (0.0015)
Self-Exclusion − 0.0496*** − 0.0477*** − 0.0275*
(0.0184) (0.0184) (0.0146)
(△Y/△X) / E(Y) 9.28% 9.43% 14.37% 18.31%
N 19,788 19,788 19,040 19,788
pseudo R2 0.3356 0.3281 0.2466 0.2132
Panel B Digital_Finance Mobile_Payments Online_FPs Online_Borrowing
Literacy_Index 0.0280*** 0.0275*** 0.0095*** 0.0084***

(0.0027) (0.0027) (0.0019) (0.0015)
(△Y/△X) / E(Y) 9.63% 9.78% 14.37% 19.22%
N 19,040 19,040 19,040 19,040
pseudo R2 0.3337 0.3262 0.2466 0.2086
literacy on the use of digital finance. Specifically, in columns 1–3 of In Panels C, D, and E of Table 10, we repeat all the exercises as in
Panel A, we regress Watch_Fin_Online, Trust, and Risk_Averse on financial Panel B for the three forms of digital finance, namely, mobile payments,
literacy while controlling for all the demographic, socio-economic, and online financial products, and online borrowing, respectively. After
geographic factors, respectively. See Appendix B for detailed definitions adding these moderators into our regressions, the marginal effects
of these three variables. As expected, the coefficients on Literacy_Index associated with financial literacy remain positive and significant at the
for Watch_Fin_Online and Trust are significantly positive, whereas the 1% level, although their magnitudes are smaller than those in column 1
coefficient for Risk_Averse is significantly negative. These results suggest (with an absence of moderator variables). For example, the figures in
that households with greater financial literacy tend to access more column 5 of Panels C (Mobile_Payments), D (Online_FPs), and E (Onli
financial news and information online, have greater trust in new things ne_Borrowing) show that the marginal effects associated with financial
and people, and are less risk-averse than those with less financial literacy fall by 15.2%, 18.9%, and 12.5%, respectively, compared with
literacy. the marginal effects in column 1. The results suggest that Watch_
Furthermore, we examine the extent to which financial literacy af Fin_Online, Trust, and Risk_averse partially mediate the effects of financial
fects the use of digital finance by controlling for these three moderator literacy on the use of these digital financial products.
variables. Panel B of Table 10 shows the results. In columns 2–4, we In short, the mechanism analysis suggests that Watch_Fin_Online,
include each of these three moderator variables individually. Our Trust, and Risk_Averse are suitable mediator variables. In other words,
analysis reveals that the marginal effects of Watch_Fin_Online and Trust financial literacy plays a vital role in promoting the use of digital finance
are significantly positive, whereas the marginal effect of Risk_Averse is by enhancing access to online information, boosting digital trust, and
significantly negative. The findings suggest that promoting online access increasing risk tolerance for digital finance.
to financial information can increase the use of digital finance. More
over, increased trust and risk tolerance associated with new technology 5.2. Heterogeneous effects of financial literacy
will also lead to increased use of digital finance. We find that the mar
ginal effects of financial literacy are still significant after accounting for 5.2.1. Variation by wealth and income
these moderators in our regressions. However, their magnitudes are Digital finance use is influenced by household wealth and income.
smaller than those in column 1 (without moderator variables). For This subsection further investigates how the effect of financial literacy
example, the marginal effects associated with financial literacy in col on the use of digital finance varies across households with different
umns 2–4 drop by approximately 11.9%, 1.1%, and 3.7%, respectively, wealth or income levels. We classify households into three groups based
compared with the marginal effect in column 1 (2.69%). These results on their total wealth and income in Panels A and B of Table 11. Spe
suggest that Watch_Fin_Online, Trust, and Risk_Averse partially mediate cifically, in column 1 (3), we consider a household to have a low (high)
the effect of financial literacy on the use of digital finance. Sobel tests level of wealth or income if its total wealth or income lies in the bottom
(Sobel 1982) reject the null hypothesis of no mediation effects.13 In (top) third of the distribution of the variable in the data sample. The
column 5, we include all three moderators in the regression and find that remaining households form the medium level of wealth or income
the marginal effects of all three moderators have signs consistent with group. In all the wealth groups, the marginal effects associated with
our expectations and are significant at the 1% level. The marginal effect financial literacy are always positive and significant. Considering the
associated with financial literacy remains positive and significant, significant discrepancies in the use of digital finance across groups, we
whereas its magnitude decreases by approximately 15.6%, that is, from also estimate the marginal effects of financial literacy, which were
2.69% to 2.27%. evaluated at their respective mean values. Financial literacy has the
greatest effects on the use of digital finance for families with the lowest
levels of wealth (16.17%) and income (18.28%), followed by households
13
Sobel (1982) proposed the Sobel test. This test examines whether a medi with medium levels of wealth (11.40%) and income (11.95%), and the
ator carries the influence of a given independent variable to a dependent least for those with the highest levels of wealth (6.7%) and income
variable. (5.63%). However, low-wealth/income households are less likely to
12
Table 10
Further mechanism tests.
(1) (2) (3) (4) (5)
Panel A Watch_Fin_Online Trust Risk Averse
Literacy_Index 0.0161*** 0.0086*** − 0.0148***

(0.0024) (0.0028) (0.0036)
Other Control Variables Yes Yes Yes
Province Fixed Effects Yes Yes Yes
N 19,788 19,788 19,788
Pseudo R2 0.2186 0.0181 0.0317
Panel B Digital Finance
Literacy_Index 0.0269*** 0.0237*** 0.0266*** 0.0259*** 0.0227***

(0.0026) (0.0025) (0.0026) (0.0026) (0.0025)
Watch_Fin_Online 0.1556*** 0.1514***
(0.0068) (0.0069)
Trust 0.0299*** 0.0239***
(0.0064) (0.0063)
Risk Averse − 0.0463*** − 0.0350***
(0.0053) (0.0052)
N 19,788 19,788 19,788 19,788 19,788
Pseudo R2 0.3352 0.3588 0.3362 0.3385 0.3616
Sobel Test Z(p) 5.460(0.000) 2.457(0.014) 8.852(0.000)
Panel C Mobile_Payments
Literacy_Index 0.0264*** 0.0232*** 0.0262*** 0.0255*** 0.0224***

(0.0026) (0.0025) (0.0026) (0.0026) (0.0025)
Watch_Fin_Online 0.1516*** 0.1477***
(0.0067) (0.0067)
Trust 0.0245*** 0.0189***
(0.0065) (0.0064)
Risk Averse − 0.0439*** − 0.0327***
(0.0053) (0.0052)
N 19,788 19,788 19,788 19,788 19,788
Pseudo R2 0.3278 0.3511 0.3284 0.3308 0.3533
Sobel Test Z(p) 5.459(0.000) 2.378(0.174) 8.750(0.000)
Panel D Online_FPs
Literacy_Index 0.0090*** 0.0078*** 0.0088*** 0.0086*** 0.0073***

(0.0018) (0.0018) (0.0018) (0.0018) (0.0018)
Watch_Fin_Online 0.0555*** 0.0534***
(0.0036) (0.0036)
Trust 0.0175*** 0.0149***
(0.0035) (0.0034)
Risk Averse − 0.0209*** − 0.0147***
(0.0032) (0.0031)
N 19,788 19,788 19,788 19,788 19,788
Pseudo R2 0.2513 0.2788 0.2538 0.2557 0.2832
Sobel Test Z(p) 5.422(0.000) 2.506(0.012) 7.951(0.000)
Panel E Online_Borrowing
Literacy_Index 0.0080*** 0.0073*** 0.0079*** 0.0078*** 0.0070***

(0.0015) (0.0015) (0.0015) (0.0015) (0.0015)
Watch_Fin_Online 0.0350*** 0.0336***
(0.0030) (0.0030)
Trust 0.0083*** 0.0067**
(0.0031) (0.0031)
Risk Averse − 0.0141*** − 0.0103***
(0.0026) (0.0026)
N 19,788 19,788 19,788 19,788 19,788
Pseudo R2 0.2125 0.2376 0.2138 0.2172 0.2411
Sobel Test Z(p) 5.333(0.000) 2.135(0.033) 6.595(0.000)
standard errors of all the variables in the regressions are clustered at the community level. Under the null hypothesis of no mediation effect, the Sobel tests are used to
test whether a mediator carries the influence of a given independent variable to a dependent variable. See Appendix B for detailed definitions of all the variables. ***,
** and * indicate significance at the 1%, 5% and 10% levels, respectively.
13
Table 11 Table 12
Probit model regressions: considering the heterogeneous effects of financial Probit model regressions: considering the heterogeneous effects of financial
literacy based on demographic and socio-economic factors. literacy based on geographic factors.
(1) (2) (3) (1) (2) (3)
Panel A Low_Wealth Mid_Wealth High_Wealth Panel A Rural Urban
Literacy_Index 0.0136*** 0.0299*** 0.0365*** Literacy_Index 0.0168*** 0.0341***

(0.0028) (0.0046) (0.0058) (0.0031) (0.0038)
(△Y/△X) / E(Y) 16.17% 11.40% 6.70% (△Y/△X) / E(Y) 13.57% 8.05%
Other Control Variables Yes Yes Yes Other Control Variables Yes Yes
Province Fixed Effects Yes Yes Yes Province Fixed Effects Yes Yes
N 6596 6596 6596 N 7887 11,901
pseudo R2 0.2925 0.2221 0.2390 Pseudo R2 0.2489 0.3001
Panel B Low_Income Mid_Income High_Income Panel B Western Central Coastal
Literacy_Index 0.0170*** 0.0305*** 0.0296*** Literacy_Index 0.0286*** 0.0214*** 0.0299***

(0.0026) (0.0047) (0.0059) (0.0054) (0.0042) (0.0041)
(△Y/△X) / E(Y) 18.28% 11.95% 5.63% (△Y/△X) / E(Y) 9.44% 8.83% 9.16%
Other Control Variables Yes Yes Yes Other control variables Yes Yes Yes
Province Fixed Effects Yes Yes Yes Province fixed effects Yes Yes Yes
N 6596 6596 6596 N 4364 6705 8719
pseudo R2 0.3054 0.2200 0.2302 Pseudo R2 0.3428 0.3219 0.3384
Panel C Young_Age Middle_Age Old_Age Notes: The table reports marginal effects and standard errors (in parentheses).
Literacy_Index 0.0168*** 0.0341*** 0.0168*** All the specifications were estimated using the probit estimator. Test statistics
(0.0031) (0.0038) (0.0031) and standard errors of all the variables in the regressions are clustered at the
(△Y/△X) / E(Y) 3.06% 11.78% 16.18% community level. (△Y/△X)/E(Y) is the marginal effect associated with Liter
Other Control Variables Yes Yes Yes acy_Index evaluated at the mean values of the different forms of digital finance.
Province Fixed Effects Yes Yes Yes See Appendix B for detailed definitions of all the variables. ***, ** and * indicate
N 3847 7728 8213 significance at the 1%, 5% and 10% levels, respectively.
pseudo R2 0.3041 0.2394 0.2704
Panel D Male Female less trust in adopting new technologies, and are more risk-averse. In
Literacy_Index 0.0257*** 0.0336*** Panel D, when considering the mean values of digital finance use, the
(0.0028) (0.0062) marginal effects of financial literacy are similar for women (9.22%) and
(△Y/△X) / E(Y) 9.41% 9.22%
men (9.41%), and both are positive and significant.
Other Control Variables Yes Yes
Province Fixed Effects Yes Yes
N 15,849 3939 5.2.3. Variation by region
pseudo R2 0.3284 0.3484 China is a vast nation with significant rural and urban disparities
Notes: The table reports marginal effects and standard errors (in parentheses). (Huang et al., 2008). Table 4 shows that the use of digital finance is
All the specifications were estimated using the probit estimator. Test statistics much higher among urban than rural families. This subsection takes a
and standard errors of all the variables in the regressions are clustered at the close look at the possible rural–urban differences in the relationship
community level. (△Y/△X)/E(Y) is the marginal effect associated with Liter between financial literacy and the use of digital finance. To this end, we
acy_Index evaluated at the mean values of the different forms of digital finance. divide the whole sample into rural and urban regions. Panel A in
See Appendix B for detailed definitions of all the variables. ***, ** and * indicate Table 12 shows that financial literacy has significant and positive mar
significance at the 1%, 5% and 10% levels, respectively. ginal effects on rural and urban families. However, when considering the
mean values of digital finance use, the marginal effects of financial lit
access digital finance than high-wealth/income ones (Table 4). eracy are higher for rural families (13.57%) than urban families
Although low-wealth/income families tend to have higher costs of (8.05%). This finding implies that promoting financial knowledge
accessing information and lower levels of trust and tolerance regarding among rural households would lead to greater adoption of digital
financial risks, financial literacy can assist these families in overcoming financial services in rural areas than in urban areas. In Panel B, we
these obstacles. Therefore, promoting financial literacy among poor divide the sample into three groups according to household places of
households may be an effective means to increase the adoption and use residence: coastal, central, and western regions. We find that in all the
of digital financial services, which in turn could help bridge the digital specifications, the marginal effects associated with financial literacy are
divide between the rich and the poor. always positive and significant. When considering the mean values of
digital finance, the marginal effects are slightly higher in western re
5.2.2. Variation by age and gender gions (9.44%, column 1) than in coastal (9.16%, column 3) and central
Panels C and D of Table 11 further explore whether the impact of (8.83%, column 2) regions.
financial knowledge on the use of digital finance varies with age and
gender, respectively. In Panel C, we split the sample into three groups by 6. Conclusions
age: young (18–44 years old), middle-aged (45–59 years old), and old-
aged (older than 59 years). We find that the marginal effects associ This study uses CHFS data to investigate the impact of financial lit
ated with financial literacy are all positive and significant in the three eracy on the use of digital finance in China, which has the largest
different age groups. However, when considering the mean values of unbanked population and yet the world’s largest fintech market
digital finance use, the impacts of financial literacy are the highest for (Guariglia and Yang 2016; Demirguc-Kunt et al., 2018). We find that
old-aged people (16.18%), followed by middle-aged (11.78%) and financial literacy plays a significant role in promoting the use of digital
young adults (3.06%). This result suggests the vital role of financial financial services despite the low level of financial knowledge among
knowledge in encouraging the old-aged to embrace digital finance. This Chinese households. Overall, a one standard deviation increase in
case may be related to the role financial literacy plays in mitigating financial literacy raises the probability of using digital finance by 2.87
barriers faced by the elderly. Compared with young adults, older in percentage points. Using the mean values of three forms of digital
dividuals often encounter higher costs in acquiring information, have finance as reference levels, we find that a one-unit increase in the
14
financial literacy index increases the probability of using mobile pay spreads in emerging markets, such as China, developing effective
ments, online financial products, and online borrowing by 9.39%, financial education initiatives and helping customers navigate unfa
13.62%, and 18.31%, respectively. These results show that the impact of miliar and potentially complex digital financial products are vital to
financial literacy is greater for more sophisticated financial products developing responsible digital financial practices, thereby achieving
such as online borrowing and online financial products than for mobile equitable financial inclusion in the digital age. In particular, the au
payments. thorities should develop and implement national strategies to promote
Furthermore, we explore different mechanisms by which financial digital financial education and specifically tailored programs that target
literacy promotes the use of digital finance. Our findings confirm the vulnerable and disadvantaged groups, including the less educated, the
critical role of financial literacy in enhancing access to online financial elderly, the rural, and the poor.
information, promoting digital trust, and reducing risk aversion, leading
to increased use of digital finance. Last, a heterogeneity analysis in CRediT authorship contribution statement
dicates that financial literacy plays a more important role for disad
vantaged groups, such as families with low levels of wealth and income, Junhong Yang: Conceptualization, Methodology, Investigation,
the elderly, and rural residents, than their counterparts, illustrating the Writing – original draft, Writing – review & editing, Funding acquisition,
promising role of financial literacy in closing the digital divide. Project administration, Visualization, Supervision. Yu Wu: Conceptu
Digital finance in emerging markets has great potential to address alization, Methodology, Resources, Formal analysis, Data curation,
the financial needs of underserved population segments, including the Project administration, Supervision. Bihong Huang: Conceptualization,
unbanked, startups, and micro, small and medium-sized enterprises Project administration, Resources, Writing – review & editing.
(MSMEs), by providing them with solutions for essential aspects of their
lives and businesses, such as payments and credit. However, great Data availability
financial knowledge is necessary to understand the risks and benefits of
various digital financial services and products. As financial illiteracy The authors do not have permission to share data.
Appendix A
A1. Questions in the CHFS related to financial literacy
The answers to the following three questions were used to calculate our composite index of financial literacy (Literacy_Index). The three questions
concern the following concepts: numeracy and capacity to perform calculations related to interest rates, such as compound interest (Quiz_Interest);
understanding of inflation (Quiz_Inflation); and understanding of financial risk (Quiz_Risk).
Quiz_Interest (H3105): Suppose you have 100 yuan and a bank’s interest rate is 4% per year. If you deposit this money in the bank for 1 year, how
much (the principal and interest earned) will you have in 1 year’s time?
1. Less than 104 yuan

2. Equal to 104 yuan
3. More than 104 yuan
4. Do not know
Quiz_Inflation (H3106): Suppose the annual interest rate of your bank is 5% and the inflation rate is 3%. If you deposit 100 yuan in the bank, after
1 year, how much will you be able to buy with the money?
1. More than you could buy a year ago

2. The same as you could buy a year ago
3. Less than you could buy a year ago
4. Do not know
Quiz_Risk (H3111): Which investment do you think is riskier in general when you buy a stock or equity funds?
1. Stocks
2. Equity funds
3. Never heard of “stocks”
4. Never heard of “equity funds”
5. Never heard of either
A2. Constructing the financial literacy index using factor analysis
We constructed the Literacy Index using a factor analysis of six components of the answers to the three questions regarding the respondents’
financial sophistication. Specifically, we constructed a dummy variable (Quiz_Correct) which is equal to 1 if the respondent gave the correct answer to
the question and 0 if otherwise. In addition, if a respondent provided the “do not know” answer, then the dummy variable (Quiz_DK) is equal to 1 and
0 if otherwise.
For the factor analysis, one component has eigenvalues above 1, suggesting that the study should use one factor. The Kaiser–Meyer–Olkin (KMO)
measure of sampling adequacy takes values between 0 and 1, with small values indicating that overall the variables have little in common to warrant a
factor analysis, and values above 0.5 are satisfactory for a factor analysis. The higher the KMO index, the more efficient the factorization. We observe
that all the KMO values in our table are above 0.6, which satisfies the minimum requirement for sample adequacy (0.5).
15
Table A2
Constructing a financial literacy index using factor analysis.
Variables Loading KMO test Eigenvalue Proportion Cumulative
Quiz_Interest_DK 0.5781 0.7417 2.7624 0.6660 0.6660

Quiz_Interest_Correct 0.8117 0.7075 0.9062 0.2185 0.8844
Quiz_Inflation_DK 0.3618 0.7462 0.3543 0.0854 0.9698
Quiz_Inflation_Correct 0.7324 0.7432 0.1250 0.0301 1.000
Quiz_Risk_DK 0.7333 0.6371 0.0002 0.0001 1.000
Quiz_Risk_Correct 0.7513 0.6428 0.000 0.000 1.000
Aggregate 0.6914
Appendix B. Definitions of all the variables
Variable name Definition
Ln(Wealth) Natural logarithm of total household wealth: Household total wealth includes financial and non-financial assets. The former are the total value of stocks, bonds,
funds, financial products, derivatives, foreign exchange assets, and gold owned by households. The latter are the total value of the household’s commercial,
agricultural and production projects, houses, land, vehicles, and other assets
Ln(Income) Natural logarithm of total household disposable income: Total household disposable income includes household wages, production, investment, and transfer
income
Age Age of the householder
Age^2 Square of the age of the householder divided by 100
Male Gender of the householder (one for male, zero for female)
Single_Female Dummy variable equal to 1 if the householder is a single woman and 0 if otherwise
Married Marital status of the householder (1 for married/cohabiting, 0 if otherwise)
Education Years of education of the householder
Size_Household Number of household members
Unhealth_Ratio Ratio of the number of unhealthy members to the household size
Employed Dummy variable equal to 1 if the householder is employed and 0 if otherwise
Homeowner Dummy variable equal to 1 if the household owns a house (the household head is a homeowner) and 0 if otherwise
Entrepreneurship Dummy variable equal to 1 if the household has its own business and 0 if otherwise
Party Dummy variable equal to 1 if the householder is a communist party member and 0 if otherwise
Ability_High1 Dummy variable equal to 1 if the respondent did not rely much on interpretation by the interviewer when they answered the questions in the survey and 0 if
otherwise
Ability_High2 Dummy variable equal to 1 if the respondent was able to understand the questions in the survey and 0 if otherwise
Watch_Fin_Online Dummy variable equal to 1 if the respondent watches finance/business news online and 0 if otherwise
Trust Dummy variable equal to 1 if the respondent trusts strangers and 0 if otherwise
Risk_Averse Dummy variable equal to 1 if the respondent is risk-averse and 0 if otherwise
No_Branches Number of bank branches in the community
Broadband_Access Percentage of households that have broadband access in the city
Rural Dummy variable equal to 1 if the household resides in a rural area and 0 if otherwise
Regions Coastal regions: Liaoning, Tianjin, Beijing, Hebei, Shandong, Jiangsu, Shanghai, Zhejiang, Fujian, Guangdong, Hainan
Central regions: Heilongjiang, Jilin, Shanxi, Henan, Anhui, Hubei, Jiangxi, Hunan
Western regions: Inner Mongolia, Guangxi, Chongqing, Sichuan, Guizhou, Yunnan, Tibet Autonomous Region, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang
Ln(GRP_PC) Logarithm of the city-level per capita gross regional product (GRP)
References Demir, A., Pesqué-Cela, V., Altunbas, Y., Murinde, V., 2020. Fintech, financial inclusion
and income inequality: a quantile regression approach. Eur. J. Finance 28 (1),
86–107.
Afridi, F., Li, S.X., Ren, Y., 2015. Social identity and inequality: the impact of China’s
Demirguc-Kunt, A., Klapper, L., Singer, D., Ansar, S., Hess, J., 2018. Global FinTech
hukou system. J. Public Econ. 123 (C), 17–29.
Database 2017: Measuring financial inclusion and the FinTech revolution. World
Agarwal, S., Driscoll, J.C., Gabaix, X., Laibson, D., 2009. The age of reason: financial
Bank, Washington, DC.
decisions over the life cycle and implications for regulation. Brook. Pap. Econ. Act. 2,
Ding, S., Guariglia, A., Knight, J., Yang, J., 2021. Negative investment in China:
51–101.
financing constraints and restructuring versus growth. Econ. Dev. Cult. Change 69
Andreou, P.C., Anyfantaki, S., 2021. Financial literacy and its influence on internet
(4), 1411–1449.
banking behavior. Eur. Manag. J. 39 (5), 658–674.
Disney, R., Gathergood, J., 2013. Financial literacy and consumer credit portfolios.
Atkinson, A., Messy, F., 2012. Measuring Financial literacy: Results of the OECD/
J. Bank. Finance 37 (7), 2246–2254.
international Network On Financial Education (INFE) Pilot study, OECD Working
Dohmen, T., Falk, A., Huffman, D., Sunde, U., 2010. Are risk aversion and impatience
Papers On Finance, Insurance and Private Pensions, No. 15.
related to cognitive ability? Am. Econ. Rev. 100 (3), 1238–1260.
Brown, J.R., Ivković, Z., Smith, P.A., Weisbenner, S., 2008. Neighbors matter: causal
Feng, X., Lu, B., Song, X., Ma, S., 2019. Financial literacy and household finances: a
community effects and stock market participation. J. Finance 63 (3), 1509–1531.
Bayesian two-part latent variable modeling approach. J. Empir. Finance 51,
Buchak, G., Matvos, G., Piskorski, T., Seru, A., 2018. Fintech, regulatory arbitrage, and
119–137.
the rise of shadow banks. J. Financ. Econ. 130 (3), 453–483.
Finke, M.S., Howe, J.S., Huston, S.J., 2017. Old age and the decline in financial literacy.
Bucher-Koenen, T., Lusardi, A., 2011. Financial literacy and retirement planning in
Manage. Sci. 63 (1), 213–230.
Germany. J. Pension Econ. Finance 10 (4), 565–584.
Gambacorta, L., Huang, Y., Qiu, H., Wang, J., 2019. How Do Machine Learning and Big
Cocco, J.F., Gomes, F.J., Maenhout, P.J., 2005. Consumption and portfolio choice over
Data Affect Credit scoring? Evidence from a Chinese fintech firm. BIS Working Paper
the life cycle. Rev. Financ. Stud. 18 (2), 491–533.
No. 834. Bank of International Settlements, Basel.
Cole, S., Sampson, T., Zia, B., 2011. Prices or knowledge? What drives demand for
Gaudecker, H.M.V., 2015. How does household portfolio diversification vary with
financial services in emerging markets? J. Finance 66 (6), 1933–1967.
financial literacy and financial advice? J. Finance 70 (2), 489–507.
Chui, M., 2021. BIS Working Papers No 947.
Grohmann, A., Klühs, T., Menkhoff, L., 2018. Does financial literacy improve financial
Cui, X., Xiao, J.J., Yi, J., 2019. Employment type, residential status and consumer
inclusion? Cross country evidence. World Develop. 111, 84–96.
financial capability: evidence from China Household Finance Survey. Singapore
Guariglia, A., Yang, J., 2016. A balancing act: managing financial constraints and agency
Econ. Rev. 64 (01), 57–81.
costs to minimize investment inefficiency in the Chinese market. J. Corp. Finance 36,
111–130.
16
Guiso, L., Viviano, E., 2014. How much can financial literacy help?1. Rev. Financ. 19 (4), Nemoto, N., Storey, D.J., Huang, B., 2019. Optimal Regulation of P2P Lending For Small
1347–1382. and Medium-Sized Enterprises. ADBI Working Paper, 912.
Haliassos, M., Bertaut, C.C., 1995. Why do so few hold stocks? Econ. J. 105 (432), Niu, G., Zhou, Y., Gan, H., 2020. Financial literacy and retirement preparation in China.
1110–1129. Pacif.-Basin Finance J. 59, 101262.
Han, L., Xiao, J.J., Su, Z., 2019. Financing knowledge, risk attitude and P2P borrowing in Ozili, P.K., 2018. Impact of digital finance on financial inclusion and stability. Borsa
China. Int. J. Consum. Stud. 43 (2), 166–177. Istanbul Rev. 18 (4), 329–340.
Hansen, T., 2012. Understanding trust in financial services: the influence of financial Read, D., Loewenstein, G., Rabin, M., Keren, G., Laibson, D., 1999. Choice bracketing. In:
healthiness, knowledge, and satisfaction. J. Serv. Res. 15 (3), 280–295. Fischhoff, B., Manski, C.F. (Eds.), Elicitation of Preferences, pp. 171–202.
Hong, H., Kubik, J.D., Stein, J.C., 2004. Social interaction and stock-market Shen, Y., Hueng, C.J., Hu, W., 2020. Using digital technology to improve financial
participation. J. Finance 59 (1), 137–163. inclusion in China. Appl. Econ. Lett. 27 (1), 30–34.
Huang, J., Zhang, Q., Rozelle, S., 2008. Economic growth, the nature of growth and Sobel, M.E., 1982. Asymptotic confidence intervals for indirect effects in structural
poverty reduction in rural China. China Econ. J. 1 (1), 107–122. equation models. Sociol. Methodol. 13, 290–312.
Huang, Y., 2020. Fintech development in the People’s Republic of China and its Song, Q., Li, J., Wu, Y., Yin, Z., 2020. Accessibility of financial services and household
macroeconomic implications. ADBI Working Paper Ser. 1169. consumption in China: evidence from micro data. N. Am. J. Econ. Finance 53,
Jack, W., Suri, T., 2014. Risk sharing and transactions costs: evidence from Kenya’s 101213.
mobile money revolution. Am. Econ. Rev. 104 (1), 183–223. Stock, J.H., Yogo, M., 2005. Identification and Inference For Econometric models: Essays
Jappelli, T., Padula, M., 2013. Investment in financial literacy and saving decisions. in Honor of Thomas Rothenberg. In Andrews, D.W.K., Stock, J.H. (Eds.). Cambridge
J. Bank. Finance 37 (8), 2779–2792. University Press, Cambridge, England, pp. 80–108.
Lusardi, A., de Bassa Scheresberg, C., 2013. National Bureau of Economic Research, van Rooij, M.C.J., Lusardi, A., Alessie, R., 2011. Financial literacy and stock market
Working Paper, 18969. participation. J. Financ. Econ. 101 (2), 449–472.
Lusardi, A., Mitchelli, O.S., 2007. Financial literacy and retirement preparedness: van Rooij, M.C.J., Lusardi, A., Alessie, R.J.M., 2012. Financial literacy, retirement
evidence and implications for financial education. Bus. Econ. 42 (1), 35–44. planning and household wealth. Econ. J. 122 (560), 449–478.
Lusardi, A., Mitchell, O.S., 2008. Planning and financial literacy: how do women fare? Wang, H., Zhang, D., Guariglia, A., Fan, G.Z., 2021. Growing out of the growing pain:
Am. Econ. Rev. 98 (2), 413–417. financial literacy and life insurance demand in China. Pacif.-Basin Finance J. 66,
Lusardi, A., Mitchell, O.S., 2011. Financial literacy and retirement planning in the United 101459.
States. J. Pens. Econ. Finance 10 (4), 509–525. Wooldridge, J.M., 2010. Econometric Analysis of Cross Section and Panel Data. MIT
Lusardi, A., Mitchell, O.S., 2014. The economic importance of financial literacy: theory press.
and evidence. J. Econ. Lit. 52 (1), 5–44. Yu, S., McMorrow, R., 2021. Beijing to break up Ant’s Alipay and force creation of
Lusardi, A., Mitchell, O.S., 2017. How ordinary consumers make complex economic separate loans app. Financ. Time. 13 September. Retrieved from < https://www.ft.
decisions: financial literacy and retirement readiness. Q. J. Finance 7 (03), 1750008. com/content/01b7c7ca-71ad-4baa-bddf-a4d5e65c5d79 >[Accessed 13 March
Malady, L., 2016. Consumer protection issues for digital financial services in emerging 2022].
markets. Bank. Finance Law Rev. 31 (2), 389–401.
17

1 s2.0 S0378426623001966 Main

Uploaded by

Copyright:

Available Formats

1 s2.0 S0378426623001966 Main

Uploaded by

Document Information

Original Title

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

1 s2.0 S0378426623001966 Main

Uploaded by

Copyright:

Available Formats

Journal of Banking and Finance 156 (2023) 107005

Contents lists available at ScienceDirect

Journal of Banking and Finance

Digital finance and financial literacy: Evidence from Chinese households☆

1. Introduction portfolio managers (robo-advisors), and advanced trading platforms,

Number Percentage Number Percentage Number Percentage

Correct 10,412 28.5% 5901 16.2% 17,911 51.4%

Digital finance 19,788 0.2908 0.4542 0 1

Nation 0.0144 0.8928 2.4838

Nation 29.1% 28.1% 6.6% 4.4%

Literacy_Index 0.0269*** 0.0264*** 0.0090*** 0.0080***

Panel A Digital_Finance Mobile_Payments Online_FPs Online_Borrowing

Literacy_Score1 0.0349*** 0.0333*** 0.0166*** 0.0083***

Panel B Digital_Finance Mobile_Payments Online_FPs Online_Borrowing

Literacy_Score2 0.0215*** 0.0209*** 0.0091*** 0.0043***

Panel C Digital_Finance Mobile_Payments Online_FPs Online_Borrowing

Financial_Class 0.1953*** 0.1772*** 0.0728*** 0.0309***

Panel A First Stage Digital_Finance Mobile_Payments Online_FPs Online_Borrowing

Literacy_Index 0.4440*** 0.4290*** 0.1232*** 0.1078***

Panel B First Stage Digital_Finance Mobile_Payments Online_FPs Online_Borrowing

Literacy_Index 0.1714*** 0.1713*** 0.0377* 0.0336**

Panel C First Stage Digital_Finance Mobile_Payments Online_FPs Online_Borrowing

Literacy_Index 0.5711*** 0.5649*** 0.2071*** 0.1425***

Literacy_Index 0.0249*** 0.0243*** 0.0085*** 0.0074***

Panel A Digital_Finance Mobile_Payments Online_FPs Online_Borrowing

Literacy_Index 0.0270*** 0.0265*** 0.0095*** 0.0080***

Panel B Digital_Finance Mobile_Payments Online_FPs Online_Borrowing

Literacy_Index 0.0280*** 0.0275*** 0.0095*** 0.0084***

Panel A Watch_Fin_Online Trust Risk Averse

Literacy_Index 0.0161*** 0.0086*** − 0.0148***

Panel B Digital Finance

Literacy_Index 0.0269*** 0.0237*** 0.0266*** 0.0259*** 0.0227***

Literacy_Index 0.0264*** 0.0232*** 0.0262*** 0.0255*** 0.0224***

Literacy_Index 0.0090*** 0.0078*** 0.0088*** 0.0086*** 0.0073***

Literacy_Index 0.0080*** 0.0073*** 0.0079*** 0.0078*** 0.0070***

Panel A Low_Wealth Mid_Wealth High_Wealth Panel A Rural Urban

Literacy_Index 0.0136*** 0.0299*** 0.0365*** Literacy_Index 0.0168*** 0.0341***

Panel B Low_Income Mid_Income High_Income Panel B Western Central Coastal

Literacy_Index 0.0170*** 0.0305*** 0.0296*** Literacy_Index 0.0286*** 0.0214*** 0.0299***

A1. Questions in the CHFS related to financial literacy

1. Less than 104 yuan

1. More than you could buy a year ago

A2. Constructing the financial literacy index using factor analysis

Variables Loading KMO test Eigenvalue Proportion Cumulative

Quiz_Interest_DK 0.5781 0.7417 2.7624 0.6660 0.6660

Appendix B. Definitions of all the variables

Variable name Definition

You might also like

Literacy_Index 0.0269* 0.0264* 0.0090* 0.0080*

Literacy_Score1 0.0349* 0.0333* 0.0166* 0.0083*

Literacy_Score2 0.0215* 0.0209* 0.0091* 0.0043*

Financial_Class 0.1953* 0.1772* 0.0728* 0.0309*

Literacy_Index 0.4440* 0.4290* 0.1232* 0.1078*

Literacy_Index 0.1714* 0.1713* 0.0377* 0.0336**

Literacy_Index 0.5711* 0.5649* 0.2071* 0.1425*

Literacy_Index 0.0249* 0.0243* 0.0085* 0.0074*

Literacy_Index 0.0270* 0.0265* 0.0095* 0.0080*

Literacy_Index 0.0280* 0.0275* 0.0095* 0.0084*

Literacy_Index 0.0161* 0.0086* − 0.0148***

Literacy_Index 0.0269* 0.0237* 0.0266* 0.0259* 0.0227***

Literacy_Index 0.0264* 0.0232* 0.0262* 0.0255* 0.0224***

Literacy_Index 0.0090* 0.0078* 0.0088* 0.0086* 0.0073***

Literacy_Index 0.0080* 0.0073* 0.0079* 0.0078* 0.0070***

Literacy_Index 0.0136* 0.0299* 0.0365* Literacy_Index 0.0168* 0.0341***

Literacy_Index 0.0170* 0.0305* 0.0296* Literacy_Index 0.0286* 0.0214* 0.0299*