You are on page 1of 13

International Review of Financial Analysis 85 (2023) 102472

Contents lists available at ScienceDirect

International Review of Financial Analysis


journal homepage: www.elsevier.com/locate/irfa

Electronic payment, natural environment and household consumption:


Evidence from China household finance survey
Jiayi Li a, Sumei Luo b, Guangyou Zhou c, *
a
PBC School of Finance, Tsinghua University, Beijing 100084, China
b
School of Finance, Shanghai University of Finance and Economics, Shanghai 200433, China
c
School of Economics, Fudan University, Shanghai 200433, China

A R T I C L E I N F O A B S T R A C T

JEL codes: In this paper, whether and how electronic payment affects households’ consumption decisions are discussed.
D12 Data of China Household Finance Survey from 2011 to 2017 are adopted to carry out empirical tests. It is found
E20 that electronic payment would significantly induce households to consume more in discretionary goods. The
Q51
potential mechanism lies in that the electronic payment cuts down transaction costs of venturing out or carrying
Keywords: banknotes that may hinder consumption. Using the weather conditions near the household location, this paper
Electronic payment
finds that the electronic payment would increase a household’s discretionary consumption more significantly
Household consumption
Natural environment
when the rain is heavy or the sunshine is weak. But necessary consumption like medical care is unaffected. It is
Transaction costs found that electronic payment increases consumption more significantly in urban, well-educated and young
households.

1. Introduction 2022). Another important factor that fosters the development of elec­
tronic payment is the emergence of online shopping platforms. In China,
“Where is the QR codes? Can I use Alipay?” the biggest mobile payment platform, Alipay, and the biggest e-com­
merce platform, Taobao, are provided by the same commercial giant,
This is probably the most frequent sentence that one would hear in
Alibaba. In 2021, China’s online retail sales reached 13.09 billion yuan
Chinese shops in recent years. According to the ‘General Situation of the
(1.87 billion $), an increase of 14.1% over the previous year. Online
Payment System Operation’ in 2021 published by the People’s Bank of
shopping has become the main way of consumption for consumers. If
China, the business volume of electronic payment is growing rapidly,
they want to purchase things online, they have to use electronic pay­
among which mobile payment is the fastest growing, with the volume of
ment. Thanks to the well-developed express services in China, you can
transaction increasing by 16.90%. The users of some third-party pay­
just stay at home and buy what you want using the electronic payment
ment platforms like Alipay and WeChat Pay have increased dramati­
without taking the bother to venture out.
cally. It is estimated that the total users of Alipay have reached 711
Now that the electronic payment makes it easier to pay for com­
million by the end of 2020, which account for almost half of the Chinese
modities, it is natural to ask whether it would increase household
population.
consumption.
Why is electronic payment so prevalent in China?
Some previous studies have already provided some insights into this
Using electronic payment only takes a few seconds by scanning a QR
issue from different perspectives. One strand of literature attaches sig­
code or swiping a card, making the consumption process fast and easy.
nificance to the psychology of consumers. Ma, Wang, He, Tan, and
The only thing that you need to carry is a mobile phone. Using cash
Zhang (2021) study the brains of consumers and find that using cash
instead would incur various transaction costs. For example, some studies
would cost consumers more mental resources, so consumers would
suggest that carrying large bills is likely to suffer from counterfeiting,
consume more using mobile payment. Another strand of studies focuses
but this is not a problem when you use electronic payment (Bajaj &
on the liquidity constraints that consumers face. Brito and Hartley
Damodaran, 2022). Other studies also suggest that the long distance to
(1995) find that the electronic payment method like credit cards would
banks makes it difficult to withdraw cash to consume (Zhao, Wu, & Guo,

* Corresponding author.
E-mail address: zgy@fudan.edu.cn (G. Zhou).

https://doi.org/10.1016/j.irfa.2022.102472
Received 10 August 2022; Received in revised form 4 November 2022; Accepted 19 December 2022
Available online 22 December 2022
1057-5219/© 2022 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
J. Li et al. International Review of Financial Analysis 85 (2023) 102472

Heavy Weak
Rain Sunshine

Represents

Bad
Weather

Increases

Decreases Decreases
Electronic Transaction
Consumption
Payment Costs

Increases

Fig. 1. The logic of this paper.


Note: In this figure, we summarize the logic of this paper.

Table 1 Table 2
Definition and measurement methods of dependent variables. Definition and measurement methods of key independent variables.
Variables Measurement methods Variables Measurement methods

To measure the family’s expenditure on clothing To measure whether the electronic payment is the main consumption
consumption in that year. The question on the mode of the family. The question on the questionnaire is: Which of the
consumption_clothing
questionnaire is: How much did your family following payment methods do you and your family usually use when
epay
spend on buying clothes last year? shopping? The variable is defined as 0 for households that only choose
To measure the family’s expenditure on dining the cash option, i.e. do not use electronic payment, and 1 for households
out consumption. The question on the that choose other payment methods, i.e. use electronic payment.
questionnaire is: How much did you spend on The annual sunshine time in the area where the family is located in the
consumption_foods_outside
dining out last month? In the calculation process, sunshine current year. It is calculated by averaging the annual sunshine time
this result is multiplied by 12 as the total collected by all weather stations in the province.
consumption expenditure of dining out last year. The annual rainfall in the area where the family is located in the current
To measure the family’s expenditure on tourism rain year. It is calculated by averaging the annual rainfall collected by all
consumption. The question on the questionnaire weather stations in the province.
consumption_travel
is: How much did your family spend on traveling
and visiting relatives last year? Note: In this table, we show the definition of key independent variables that are
To measure the family’s expenditure on health used in our empirical tests.
care in that year. The question on the
consumption_health
questionnaire is: How much did your family
offer liquidity to the consumers and induce them to consume more. This
spend on health care last year?
To measure the family’s expenditure on
paper, however, takes a different perspective on this issue. This paper
transportation purchase in that year. The mainly focuses on the transaction costs that using electronic payment
consumption_transportation
question on the questionnaire is: How much did can save, which directly determine the costs to buy. This mechanism is
your family spend on transportation last year? important in the context of China in that the well-developed online
To measure the family’s expenditure on
shopping saves a lot of transaction costs by allowing people to stay at
education in that year. The question on the
consumption_education
questionnaire is: How much did your family home with a few clicks using electronic payment and they can get almost
spend on education and training last year? everything they want without venturing out.
To measure the family’s expenditure on utilities. This paper mainly empirically establishes the proposition that elec­
The question on the questionnaire is: How much
consumption_utility tronic payment would cut down transaction costs for consumers, thus
did your family spend on water, electricity, fuel,
property management, heating, etc. last year?
encouraging them to consume more. But an important concern is, how to
To measure the discretionary consumption measure the transaction costs that a consumer may face in the con­
consumption_discretionary
expenditure of the family in that year. It is equal sumption process?
to the sum of consumption_clothing, Some previous studies have already shed light on this issue. Most of
consumption_foods_outside and consumption_travel.
them focus on the geographic distance. Jack and Suri (2014) adopt the
To measure the necessary consumption
expenditure of the family in that year. It is equal context of Kenya, in which the geographic distance is a major concern.
consumption_necessary
to the sum of consumption_health and They argue that the mobile payment makes it easier for households to
consumption_transportation. transfer money and save transaction costs for them. Zhao et al. (2022)
To measure the necessary consumption also use the distance to banks to measure the transaction costs of using
expenditure of the family in that year. It is equal
consumption_necessary_robustness to the sum of consumption_health,
cash. But using geographic distance to measure transaction costs still has
consumption_transportation, consumption_utility some identification concerns. First and foremost, those studies use the
and consumption_education. distance of a household to the nearest bank to measure transaction costs.
Note: In this table, we show the definition of dependent variables that are used in But it is very likely that the investigated household has no account in the
our empirical tests. nearest bank. Second, the geographic distance is not completely

2
J. Li et al. International Review of Financial Analysis 85 (2023) 102472

Table 3 also includes sunshine as a measure of environmental condition. When


Definition and measurement methods of control variables. the rain is heavy, the sunshine would be weak (Hirshleifer & Shumway,
Variables Measurement methods 2003).
This paper adopts the data of China Household Finance Survey from
To measure the family’s disposable income after tax. We add up the
salary income of each individual in the household survey. The 2011 to 2017 to carry out empirical analysis. It is found that using
income
question on the questionnaire is: How much after-tax money wages electronic payment would significantly increase a household’s discre­
did you receive last year? tionary consumption like clothing and traveling. But it would not in­
To measure the family’s average age. It is measured by dividing the fluence a household’s necessary consumption like medical care and
age
total age of each family member by the number of family members.
To measure the proportion of men in the family. It is measured by is
transportation. When the rain is heavy or the sunshine is weak, the
male measured by dividing the number of men in the family by the electronic payment can save the transaction costs more and increase
number of family members. consumption more significantly. It is also found that the electronic
To measure the education level of the family. It is calculated by payment would increase the consumption of urban, well-educated and
dividing the sum of the education level of each family member by the
young households, probably due to their proficiency in using electronic
number of family members. The education level in the questionnaire
is divided into: 1. No schooling, 2. Primary school, 3. Junior high devices. Our regression results remain robust in various robustness
education checks.
school, 4. High school, 5. Technical secondary school/ Vocational
high school, 6. Junior college/ Higher vocational school, 7. This paper contributes to the literature in several ways. First,
Undergraduate, 8. Master and 9. Doctor. The higher the value, the although some previous studies like Liu, Pan, and Yin (2020) and Zhao
higher the education level of the family.
To measure the marital status of the family. The specific setting in
et al. (2022) have tested the relationship between mobile payment and
the questionnaire is: Single person, Married person, Cohabitation, consumption, they use a cross-sectional data. This paper, however, uses
Marital separation, Divorce and Widowed. a panel data that contains observations of several years for a household.
Define Marital separation, Divorce and Widowed as experiencing This is important in that it allows us to control fixed effect at household
divorce emotional frustration with a value of 1, while the other categories
level to control for the unobservable characteristics that could generate
are defined as not experiencing emotional frustration with a value of
0. The marital status of the whole family is measured by summing up endogeneity concerns. Second, unlike studies that adopt a behavioral
the individual status and then dividing it by the number of family approach on the consumers’ mental account (Thaler, 1985) and infor­
members. mation salience (Sexton, 2015), this paper highlights the transaction
To measure the risk aversion degree of the family. The question on costs that a consumer has to bear during the purchase, which directly
the questionnaire is: If you have an asset, which investment project
determine the real price of the goods. Empirically, this paper uses the
would you like to choose: 1. Project with high risk and high return, 2.
Project with slightly higher risk and slightly higher return, 3. Project weather condition, a fully exogenous and random variable, to proxy the
risk_aversion
with average risk and average return, 4. Project with slightly lower transaction costs. Third, this paper also supplements the previous papers
risk and slightly lower return, 5. Not willing to take any risk. The (Liébana-Cabanillas, García-Maroto, Muñoz-Leiva, & Ramos-de-Luna,
higher the number selected in the multiple-choice question, the
2020) finding that characteristics like age would influence a person’s
higher the risk aversion degree of the family.
To measure the degree of unhappiness in the family. It is measured acceptance of mobile payment. We further find that characteristics like
by the question asked in the questionnaire: do you feel happy now? education would influence their use of electronic payment, thus influ­
unhappiness
1. Very happy, 2. Happy, 3. Neutral, 4. Unhappy, 5. Very unhappy. encing their consumption decisions.
The higher the value of this variable, the unhappier the family. The rest of this paper is organized as follows: In Section 2, the related
The number of houses, measured by the total number of properties
house_number studies are reviewed and hypotheses are put forward. In Section 3, the
owned by the family.
The number of cars, measured by the total number of cars owned by empirical methods and models are presented. In Section 4, the empirical
car_number
the family. results are shown and discussed. In Section 5, robustness checks are
ln_gdp
The logarithm of the GDP of the province in which the family lived in carried out to ensure the robustness of previous analysis. In Section 6, we
that year.
conclude and point out directions for further research.
cpi The CPI index of the province in which the family lived in that year.

Note: In this table, we show the definition of control variables that are used in 2. Literature and hypotheses
our empirical tests.
This part first reviews the studies that investigate the relationship
exogenous and still has potential endogeneity concerns (Dai, Rau, between electronic payment and consumption. And then it turns to re­
Stouraitis, & Tan, 2020). For example, the households that live near to a view the potential economic mechanisms that could explain the rela­
lot of banks may dwell in the center of the city. Those people are more tionship between electronic payment and consumption. This paper
likely to have higher intelligence that may help them to make rational focuses on the mechanism of transaction costs and justify the use of
consumption choices (Ahn & Nam, 2022). The unobservable variables weather conditions to proxy this mechanism. Finally, this part summa­
are correlated with the dependent variable, which would cause endo­ rizes some studies focus on the heterogeneity of electronic payment use.
geneity concerns. This paper, however, adopts a new proxy for the Many previous studies have tested the relationship between elec­
transaction costs, that is, the environmental condition. Some previous tronic payment and consumption, but the methods and conclusions are
studies have already found that the weather would influence consumers’ not consistent. Some studies mainly take on a macroeconomic
choices (Govind, Garg, & Sun, 2014; Kang & Reiner, 2022), especially perspective. Moody’s recent research shows that the increase in the use
during the offline shopping (Chintagunta, Chu, & Cebollada, 2012). of electronic payments has contributed to consumption and economic
Agnew and Thornes (1995) find that the traffic flow around the mall is growth (Zandi, Singh, & Irving, 2013). Zhang, Zhang, Liu, De Renzis,
significantly reduced when the rain is heavy. Liu, Sockin, and Xiong and Schmiedel (2019) use macro-data and find that the high penetration
(2021) find that heavy rain would greatly increase the online food or­ rate of electronic payment would increase GDP significantly in a coun­
derings. This paper adopts rain and sunshine to proxy the transaction try. Tee and Ong (2016) and Oyelami, Adebiyi, and Adekunle (2020)
costs that a consumer has to face during the consumption process. First, find similar evidence that electronic payment would contribute to the
the heavy rain would prevent consumers from venturing out and shift consumption growth. Some empirical works also adopt micro-data to
them to online shopping, which is only possible with the appearance of test this issue. Jack and Suri (2014) use the household data from Kenya
electronic payment. Second, the heavy rain would increase the cost of and find that the mobile payment will reduce non-users’ consumption by
carrying cash or withdrawing cash from banks for consumers, but the 7%. However, the consumption of users will not be affected. Hou,
electronic payment avoids this trouble and makes the consumption Hsueh, and Zhang (2021) find that households using electronic payment
process convenient. To ensure the robustness of our findings, this paper consume 20.63% more than those using other payment methods. Some

3
J. Li et al. International Review of Financial Analysis 85 (2023) 102472

Table 4
Descriptive statistics.
(1) (2) (3) (4) (5) (6)

Variables N mean sd min max skewness

consumption_clothing 76,590.00 2623.11 5837.65 0.00 500,000.00 21.80


consumption_foods_outside 76,294.00 3449.65 12,735.58 0.00 1,200,000.00 51.58
consumption_travel 76,590.00 1993.97 9301.34 0.00 1,000,000.00 35.27
consumption_health 76,590.00 5203.43 28,250.73 0.00 3,333,333.00 38.80
consumption_transportation 76,590.00 163.11 3211.94 0.00 520,500.00 96.60
consumption_education 76,590.00 4022.93 14,255.03 0.00 1,000,000.00 20.78
consumption_utility 74,842.00 291.58 576.29 0.00 50,000.00 25.43
consumption_discretionary 75,377.00 8735.21 21,936.63 0.00 1,712,000.00 27.44
consumption_necessary 76,590.00 5366.54 28,449.89 0.00 3,333,333.00 38.12
consumption_necessay_robustness 74,842.00 9593.29 31,370.28 0.00 3,333,673.00 29.39
epay 76,582.00 0.04 0.20 0.00 1.00 4.47
sunshine 73,561.00 2006.03 406.13 1135.63 2938.74 − 0.02
rain 76,590.00 990.69 450.04 250.32 2144.22 0.54
male 76,589.00 0.50 0.20 0.00 1.00 − 0.18
divorce 76,590.00 0.09 0.24 0.00 1.00 2.93
risk_aversion 76,590.00 4.10 1.28 1.00 9.00 − 0.80
income 76,590.00 28,164.78 54,063.74 0.00 3,100,000.00 10.88
unhappiness 76,527.00 2.25 0.86 1.00 9.00 0.54
house_number 76,545.00 0.99 0.66 0.00 27.00 2.24
car_number 76,567.00 0.22 0.49 0.00 12.00 2.57
age 76,587.00 44.81 16.19 10.60 117.00 0.56
education 76,590.00 3.11 1.46 0.00 9.00 0.79
ln_gdp 76,590.00 10.92 0.78 8.23 12.28 − 0.64
cpi 76,590.00 104.00 1.56 100.90 106.30 − 0.25

Note: In this table, we show the descriptive statistics for our variables. ‘N’ represents the number of observations. ‘mean’ represents the sample average. ‘sd’ represents
the standard error. ‘min’ represents the minimum value. ‘max’ represents the maximum value. ‘skewness’ represents the skewness of the sample. We keep two decimal
places for all numbers.

Table 5
Electronic payment and different types of consumption.
(1) (2) (3) (4) (5)

Variables consumption_clothing consumption_foods_outside consumption_travel consumption_health consumption_transportation

epay 1892.87*** 1521.30*** 2746.04*** − 2044.00 − 48.82


(99.49) (230.78) (164.51) (1585.59) (183.12)
male − 860.90*** 1689.69*** − 609.76*** − 291.45 91.66
(97.86) (227.17) (161.81) (506.16) (58.48)
divorce − 141.42 158.29 − 205.83 − 668.70 21.44
(86.56) (200.78) (143.14) (447.74) (51.73)
risk_aversion − 220.24*** − 512.77*** − 225.47*** 98.06 − 8.59
(15.99) (37.12) (26.43) (82.69) (9.55)
income 0.01*** 0.00 0.01*** − 0.01*** 0.00
(0.00) (0.00) (0.00) (0.00) (0.00)
unhappiness − 213.75*** − 146.94*** − 164.64*** 565.91*** − 10.34
(23.26) (54.00) (38.46) (120.31) (13.90)
house_number 517.20*** 724.97*** 879.00*** 478.34*** − 26.99
(31.52) (73.18) (52.11) (163.01) (18.83)
car_number 2166.88*** 3454.89*** 2160.78*** 424.05* 221.41***
(45.22) (105.14) (74.77) (233.88) (27.02)
age − 36.29*** − 37.08*** 16.97*** 96.94*** − 2.97***
(1.41) (3.27) (2.33) (7.29) (0.84)
education 428.14*** 379.91*** 699.38*** 155.34** − 17.37*
(14.87) (34.52) (24.59) (76.90) (8.88)
ln_gdp 13.41 269.68 936.91* 2105.93 99.44
(301.47) (699.33) (498.50) (1559.32) (180.15)
cpi 21.29 779.67*** − 523.29*** − 2461.62*** 9.39
(88.91) (206.60) (147.02) (459.89) (53.13)
Constant 982.98 − 80,124.48*** 42,987.41*** 231,567.81*** − 1720.95
(9055.65) (21,038.54) (14,973.91) (46,838.68) (5411.42)
Province FE Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Observations 76,460 76,169 76,460 76,460 76,460
R-squared 0.16 0.05 0.09 0.03 0.00

Note: In this table, we test whether using electronic payment would influence different types of consumption. All variables are defined in Table 1, Table 2 and Table 3.
The ‘Province FE’ represents province fixed effect and ‘Year FE’ represents year fixed effect. ‘Observations’ represent observations in the sample and ‘R-squared’
represents the goodness of fit. Standard error in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. We keep two decimal places for all numbers.

4
J. Li et al. International Review of Financial Analysis 85 (2023) 102472

Table 6 offer consumers interest-free loans for consumption (Ekici & Dunn,
Electronic payment and two types of comprehensive consumption. 2010). Mobile payments like Alipay usually offer interest-free loans for
(1) (2) consumption too. For Alipay, it has a service called Ant Credit Pay,
which allows consumers to borrow and consume without interest as long
Variables consumption_discretionary consumption_necessary
as they can repay the debt next month. The extra credit looses a con­
epay 7587.49*** − 2093.27 sumer’s budget constraint and increases his consumption (Bechlioulis &
(379.73) (1595.93)
male 444.00 − 199.79
Brissimis, 2019; Zeldes, 1989). But Gross and Souleles (2002) argue that
(372.04) (509.44) although consumer credit can relax liquidity constraints in the short
divorce 9.33 − 647.26 term, it will bring debt increases in the long run, so the role of consumer
(328.48) (450.64) credit is limited. Some behavioral scholars also argue that consumers are
risk_aversion − 1138.10*** 89.47
likely to overspend with their credit cards due to over optimism (Yang,
(60.90) (83.22)
income 0.03*** − 0.01*** Markoczy, & Qi, 2007). In this way, the total consumption may still
(0.00) (0.00) increase.
unhappiness − 553.41*** 555.57*** The last strand of literature focuses on the transaction costs that
(88.44) (121.09) using electronic payment may save, which is closely related to this
house_number 2337.13*** 451.35***
(120.01) (164.07)
paper. Schierz, Schilke, and Wirtz (2010) argue that convenience is the
car_number 8589.16*** 645.46*** major factor that affects consumers’ use of mobile payment. This paper
(173.04) (235.40) highlights this intrinsic characteristic of electronic payment, that is,
age − 63.97*** 93.98*** convenience. The convenience is measured by the transaction costs in
(5.36) (7.34)
the economic sense (Bennion & Nickerson, 2011). The electronic pay­
education 1690.62*** 137.97*
(56.65) (77.40) ment, however, could cut down these transaction costs. Zhao et al.
ln_gdp 1567.76 2205.36 (2022) finds that some urban households in China may not have
(1143.94) (1569.44) convenient access to banks. But the electronic payment breaks the
cpi 216.49 − 2452.23*** physical limitations and allows for remote payment, saving the costs of
(337.75) (462.88)
Constant − 33,030.59 229,846.86***
traveling long distance. The electronic payment also cuts down the
(34,381.50) (47,142.69) transaction costs of consumption due to its transparency (Dalton,
Province FE Yes Yes Pamuk, Ramrattan, van Soest, & Uras, 2019). A consumer can check his
Year FE Yes Yes account balance at any time using his mobile phone in China, thus
Observations 75,270 76,460
increasing their incentives to use mobile payment. The transaction costs
R-squared 0.15 0.04
directly determine the real price of commodities. The real price is lower
Note: In this table, we test whether using electronic payment would influence when the transaction costs decrease. Kim and Li (2009) argue that the
two different types of comprehensive consumption. All variables are defined in satisfaction and loyalty of customers would be decreased by the high
Table 1, Table 2 and Table 3. The ‘Province FE’ represents province fixed effect
transaction costs. It is found that high transaction costs would hamper
and ‘Year FE’ represents year fixed effect. ‘Observations’ represent observations
consumers’ consumption level as well as their consumption smoothing
in the sample and ‘R-squared’ represents the goodness of fit. Standard error in
(Bonaparte, Cooper, & Zhu, 2012).
parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. We keep two decimal places for
all numbers. According to the above analysis, the electronic payment can save
transaction costs, provide more credit and induce more impulse buying,
thus increasing household consumption. We propose the first
studies focus on specific event like COVID-19 to test whether the elec­
hypothesis:
tronic payment is useful to boost consumption during the quarantine,
especially in China. Liu et al. (2020) empirically show that mobile Hypothesis 1. Using electronic payment is positively related to a
payments boost urban household consumption during the COVID-19 household’s consumption level.
pandemic, while rural households are not affected.
As is mentioned above, this paper mainly focuses on the economic
However, simply testing the relationship between electronic pay­
mechanism that electronic payment can save transaction costs during
ment and consumption is not convincing enough. It is important to
the consumption process. But another important question is, how to
address and test the underlying economic mechanism. Why does elec­
measure the transaction costs that a consumer face.
tronic payment influence consumption patterns?
Some previous studies like Zhao et al. (2022) and Jack and Suri
One strand of literature focuses on the behavioral bias of the con­
(2014) use the distance to the nearest bank as a measure of transaction
sumers and argues that the use of electronic payment would amplify its
costs. But this is not precise in that the investigated household may not
magnitude. As early as Laibson (1997), scholars have been modelling
have an account in the nearest bank at all. A more serious concern is that
consumers’ irrational behaviors. The Laibson (1997) put forward a
the location of a household is not purely exogenous. Some unobservable
benchmark model of hyperbolic discounting that may induce consumers
characteristics may influence a household’s location and consumption
to become more ‘hand-to-mouth’. They may not be able to resist the
together, thus creating endogeneity problems.
temptation to consume instantly. In a world where the consumers are
This paper uses an exogenous variable, that is, the environmental
exposed to various temptation online, it becomes more difficult for them
condition (Cortés, Duchin, & Sosyura, 2016) to measure the transaction
to resist consuming excessively (Liu, Wang, & Zhao, 2021). Some studies
costs. Some previous studies have already shown that bad weather
show that the mobile payment would induce consumers to overspend,
condition would hamper consumption by increasing the transaction
especially when they financial knowledge is poor (Ahn & Nam, 2022). A
costs of venturing out. Some studies focus on the impact of rainfall.
possible explanation could be that consumers perceive less pain when
Agnew and Thornes (1995) use shopping mall traffic data and find that
they use virtual accounts and the information salience of using elec­
shopping mall traffic would decrease significantly when the rain is
tronic devices is not as significant as using cash (Huebner, Fleisch, & Ilic,
heavy, and consumers prefer avoiding going out in stormy weather
2020; Manshad & Brannon, 2021).
conditions. Joo, Kang, and Moon (2014) find that rain would signifi­
Another strand of literature attaches importance to the extra credit
cantly prevent a consumer from visiting a theme park. Aggarwal (2021)
that electronic payment would provide and argues that electronic pay­
find that the large precipitation would significantly reduce consump­
ment can relax consumers’ liquidity constraints and increase their
tion. Some studies focus on the impact of light. Summers and Hebert
consumption. Some electronic payment method like credit cards would
(2001) design an experiment by installing supplemental lighting and

5
J. Li et al. International Review of Financial Analysis 85 (2023) 102472

Table 7
Sunshine mechanism.
(1) (2) (3) (4) (5)

Variables consumption_clothing consumption_foods_outside consumption_travel consumption_health consumption_transportation

epay 3195.94*** 6395.60*** 4653.44*** − 745.74 257.34


(506.00) (1218.24) (866.63) (2650.37) (316.85)
sunshine_epay − 0.71*** − 2.47*** − 0.96** − 0.45 − 0.20
(0.25) (0.61) (0.43) (1.32) (0.16)
sunshine 0.14 − 1.73*** 2.16*** 3.75*** 0.02
(0.23) (0.56) (0.40) (1.22) (0.15)
male − 904.70*** 1666.18*** − 597.49*** − 58.75 95.50
(97.43) (234.84) (166.87) (510.32) (61.01)
divorce − 165.37* 133.52 − 215.08 − 705.03 22.96
(86.50) (208.31) (148.15) (453.06) (54.16)
risk_aversion − 211.97*** − 510.65*** − 221.45*** 74.50 − 8.57
(15.83) (38.16) (27.11) (82.90) (9.91)
income 0.01*** 0.00 0.01*** − 0.01** 0.00
(0.00) (0.00) (0.00) (0.00) (0.00)
unhappiness − 202.09*** − 125.53** − 156.07*** 532.08*** − 10.71
(23.12) (55.74) (39.60) (121.10) (14.48)
house_number 518.33*** 747.06*** 887.67*** 491.92*** − 27.03
(31.46) (75.83) (53.88) (164.76) (19.70)
car_number 2132.24*** 3359.48*** 2137.12*** 413.38* 232.79***
(45.04) (108.74) (77.13) (235.90) (28.20)
age − 36.55*** − 37.09*** 16.95*** 93.15*** − 3.06***
(1.40) (3.37) (2.40) (7.33) (0.88)
education 431.79*** 381.65*** 697.45*** 127.59* − 18.17**
(14.79) (35.64) (25.32) (77.44) (9.26)
ln_gdp − 25.93 − 573.13 1833.96*** 3490.92** 140.14
(318.68) (767.77) (545.81) (1669.23) (199.56)
cpi 50.85 706.54*** − 508.84*** − 1651.28*** − 31.26
(94.85) (228.96) (162.46) (496.83) (59.40)
Constant − 1960.77 − 59,935.01*** 27,278.30* 124,871.23** 2019.47
(9605.95) (23,181.41) (16,452.24) (50,314.85) (6015.19)
Province FE Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Observations 73,434 73,159 73,434 73,434 73,434
R-squared 0.16 0.05 0.09 0.03 0.00

Note: In this table, we test the sunlight mechanism of electronic payment to different types of consumption. All variables are defined in Table 1, Table 2 and Table 3.
The ‘Province FE’ represents province fixed effect and ‘Year FE’ represents year fixed effect. ‘Observations’ represent observations in the sample and ‘R-squared’
represents the goodness of fit. Standard error in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. We keep two decimal places for all numbers.

find that strong light would induce consumers to consume more. It instructions on using electronic payment would influence a person’s
seems that the costs of venturing out in days with heavy rain and weak frequency of using it. Alemu, Bandyopadhyay, and Negash (2015) find
sunshine would be high. But the electronic payment makes it possible for that elderly people are less likely to accept electronic payment. Based on
consumers to stay at home and consume online, which is not affected too the above facts, we propose the last hypothesis:
much by bad weather conditions. Using mobile payment also avoids the
Hypothesis 3. Electronic payment increases consumption more
trouble of traveling a long distance to a bank in bad weather conditions.
significantly in urban, well-educated and young households.
Riley (2018) finds that under extreme impacts such as natural disasters,
mobile payment users can effectively prevent the decline in consump­ Finally, we summarize the logic of this paper in Fig. 1. The basic
tion demand. Chintagunta et al. (2012) propose that when shopping relationship is that electronic payment can increase consumption. But
offline, consumers need to bear transaction costs. They use empirical why it can increase consumption? Because it can save transaction costs
data to estimate the relative transaction costs of online shopping and for consumers. How to measure the transaction costs that consumers
offline shopping and find that the transaction costs are greater in bad face? The bad weather exogenously generates high transaction costs for
weather. them. How to define the bad weather? This paper adopts rainfall and
According to the above analysis, heavy rainfall or weak sunshine sunshine, which are negatively related to each other. Finally, this paper
incurs higher transaction costs to consumers, thus hampering con­ also tests the relationship between electronic payment and consumption
sumption. But electronic payment could alleviate this problem because in different subsamples.
it cuts down transaction costs more significantly in bad weather condi­
tions. We propose the second hypothesis: 3. Empirical strategy
Hypothesis 2. Electronic payment increases consumption more
3.1. Data and variables
significantly when the rain is heavy or the sunshine is weak.
Finally, we are curious about whether electronic payment would This paper adopts the data from China Household Financial Survey
influence households with different characteristics differently. Some Database for empirical analysis. This dataset is widely used in empirical
previous studies have already pointed out that different groups of people research nowadays (Zhao et al., 2022). Due to the fact that our key in­
would manifest different acceptance of electronic payment. Research by dependent variable, electronic payment is not included in the ques­
Anderson, Strand, and Collins (2018) shows that consumers may face tionnaire in 2015, this paper only uses the data in 2011, 2013 and 2017
barriers to the use of electronic payments, especially among rural for our research. Having excluded the missing values, a total sample of
household consumers. Kumaga (2011) finds that a proper education or 76,582 observed values of 52,908 households in three years is obtained.

6
J. Li et al. International Review of Financial Analysis 85 (2023) 102472

Table 8
Sunshine and rain mechanism.
(1) (2) (3) (4)

Variables consumption_discretionary consumption_necessary consumption_discretionary consumption_necessary

epay 16,246.57*** − 488.40 1672.76* − 5053.21


(1983.66) (2668.81) (921.77) (5161.02)
sunshine_epay − 4.43*** − 0.64
(0.99) (1.33)
sunshine 0.81 3.77***
(0.91) (1.23)
rain_epay 5.58*** 1.57
(0.79) (1.58)
rain 0.20 − 2.79***
(0.73) (1.01)
male 361.21 36.75 460.27 33.30
(381.38) (513.87) (371.93) (513.92)
divorce − 44.86 − 682.07 15.57 − 686.40
(337.98) (456.22) (328.38) (456.21)
risk_aversion − 1124.96*** 65.94 − 1138.91*** 67.79
(62.10) (83.48) (60.89) (83.48)
income 0.03*** − 0.01** 0.03*** − 0.01**
(0.00) (0.00) (0.00) (0.00)
unhappiness − 515.94*** 521.37*** − 551.94*** 526.04***
(90.53) (121.94) (88.41) (121.91)
house_number 2393.92*** 464.89*** 2342.47*** 472.40***
(123.33) (165.91) (120.02) (166.00)
car_number 8385.18*** 646.17*** 8589.18*** 640.61***
(177.47) (237.54) (173.01) (237.62)
age − 64.68*** 90.09*** − 64.73*** 90.29***
(5.48) (7.38) (5.36) (7.38)
education 1691.82*** 109.41 1685.74*** 115.13
(58.01) (77.98) (56.64) (77.97)
ln_gdp 1740.04 3631.06** 1170.67 3503.23**
(1245.68) (1680.84) (1213.88) (1686.56)
cpi 128.50 − 1682.55*** 217.25 − 1775.22***
(371.21) (500.28) (339.24) (501.80)
Constant − 27,541.35 126,890.71** − 28,929.42 148,267.42***
(37,562.43) (50,664.83) (34,376.52) (50,061.23)
Province FE Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Observations 72,277 73,434 75,270 73,434
R-squared 0.15 0.03 0.15 0.03

Note: In this table, we test the sunlight and rain mechanism of electronic payment to two different types of comprehensive consumption. All variables are defined in
Table 1, Table 2 and Table 3. The ‘Province FE’ represents province fixed effect and ‘Year FE’ represents year fixed effect. ‘Observations’ represent observations in the
sample and ‘R-squared’ represents the goodness of fit. Standard error in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. We keep two decimal places for all numbers.

The specific definition and measurement method of each variable are as happiness (Gertler & Gruber, 2002; Blundell, Pistaferri, & Saporta-
follows: Eksten, 2016;Dominko & Verbič, 2022). So, we include them as our
Table 1 shows the definitions and measurement methods of the control variables. See Table 3 for specific definitions.
dependent variables. In this paper, clothing consumption, dining out
consumption and tourism consumption are selected as discretionary 3.2. Models
consumption, and they are tested respectively. In contrast, health care
expenditure and transportation expenditure are selected as necessary This paper uses fixed effect model to test the impact of electronic
consumption. This paper also changes the definition of necessary con­ payment on households’ consumption. Controlsi, t represents the control

sumption to include the utility consumption (such as electricity con­ variables, and their definitions are given in Table 3. province repre­

sumption) and education consumption to ensure the robustness. sents the province fixed effect. year represents the year fixed effect.
Table 2 shows the definitions and measurement methods of the key sunshinei, t ⋅ epayi, t represents the interaction term of sunshine and
independent variables in this paper. This paper includes sunshine and electronic payment. raini, t ⋅ epayi, t represents the interaction term of
rain as the key instruments to measure the transaction costs that a rain and the electronic payment. The variables above are used to test the
household faces in its consumption decision. At the same time, two underlying mechanisms of electronic payment on households’
interaction terms, sunshine⋅epay, rain⋅epay, are also added to the model consumption.
to test the underlying mechanism of electronic payment to consumption The benchmark regression model examines electronic payment’s
under different environmental scenarios and different transaction costs. effects on clothing consumption, dinning out consumption, tourism
Finally, this paper selects variables that may have an impact on consumption, medical care consumption and transportation
household consumption as control variables to avoid the omitted vari­ consumption:
able concerns. Existing studies have shown that the main factors ∑
affecting household consumption are income and savings (Lise & Seitz, consumptioni,t =β0 + β1 epayi,t + β2 Controlsi,t + β3

province
(1)
2011; Meyer & Sullivan, 2013; Aguiar & Bils, 2015; Blundell, Pistaferri, + β4 year + εi,t
& Saporta-Eksten, 2018), household assets (Bonaparte et al., 2012),
short-term credit (Ludvigson, 1999), macroeconomic policy (Bachmann In Eq. (1), consumptioni, t represents the five different types of con­
& Sims, 2012), and aging, labor supply, life cycle working conditions, sumption including consumption_clothing, consumption_foods_outside,
consumption_travel, consumption_health and consumption_transportation in

7
J. Li et al. International Review of Financial Analysis 85 (2023) 102472

Table 9
Heterogeneity tests.
(1) (2) (3)

Variables consumption_discretionary consumption_discretionary consumption_discretionary

epay 7979.20*** 2951.81*** 9735.82***


(395.38) (1039.84) (1074.88)
rural_epay − 5764.65***
(1351.05)
rural − 1815.07***
(177.81)
education_epay 1062.07***
(221.79)
education 1480.42*** 1628.99*** 1689.42***
(59.76) (58.09) (56.65)
age_epay − 59.30**
(27.76)
age − 63.40*** − 64.22*** − 62.30***
(5.36) (5.36) (5.42)
male 731.54** 457.05 439.85
(372.69) (371.99) (372.03)
divorce − 88.90 22.49 18.51
(328.36) (328.45) (328.50)
risk_aversion − 1132.04*** − 1138.25*** − 1137.20***
(60.85) (60.89) (60.90)
income 0.03*** 0.03*** 0.03***
(0.00) (0.00) (0.00)
unhappiness − 541.06*** − 557.18*** − 552.25***
(88.38) (88.43) (88.44)
house_number 2404.36*** 2346.71*** 2347.14***
(120.07) (120.01) (120.10)
car_number 8481.55*** 8609.47*** 8588.29***
(173.17) (173.06) (173.03)
ln_gdp 1356.23 1576.42 1565.13
(1143.13) (1143.78) (1143.91)
cpi 203.58 217.43 210.77
(337.46) (337.70) (337.75)
Constant − 28,376.68 − 33,029.20 − 32,493.97
(34,355.31) (34,376.49) (34,381.60)
Province FE Yes Yes Yes
Year FE Yes Yes Yes
Observations 75,270 75,270 75,270
R-squared 0.15 0.15 0.15

Note: In this table, we test the heterogeneity effects of electronic payment on discretionary consumption. All variables are defined in Table 1, Table 2 and Table 3. The
‘Province FE’ represents province fixed effect and ‘Year FE’ represents year fixed effect. ‘Observations’ represent observations in the sample and ‘R-squared’ rep­
resents the goodness of fit. Standard error in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. We keep two decimal places for all numbers.

Table 1. ∑ ∑
Second, this paper tests the electronic payment’s effect on two + β3 sunshinei,t + β4 Controlsi,t + β5 province + β6 year + εi,t (4)
comprehensive measures of consumption. One is the relationship be­
tween electronic payment and household discretionary consumption, consumption necessaryi,t = β0 + β1 epayi,t + β2 sunshinei,t ⋅epayi,t
the other one is the relationship between electronic payment and ∑ ∑
household necessary consumption. + β3 sunshinei,t + β4 Controlsi,t + β5 province + β6 year + εi,t (5)

consumption discretionaryi,t =β0 + β1 epayi,t + β2 Controlsi,t + β3 province Rain mechanism models:

+ β4 year + εi,t consumption discretionaryi,t = β0 + β1 epayi,t + β2 raini,t ⋅epayi,t
(2) ∑ ∑
∑ + β3 raini,t + β4 Controlsi,t + β5 province + β6 year + εi,t (6)
consumption necessaryi,t =β0 + β1 epayi,t + β2 Controlsi,t + β3 province
∑ consumption necessaryi,t = β0 + β1 epayi,t + β2 raini,t ⋅epayi,t
+ β4 year + εi,t
(3) ∑ ∑
+ β3 raini,t + β4 Controlsi,t + β5 province + β6 year + εi,t (7)
Third, this paper tests the underlying mechanism between electronic
payment and consumption. In Eqs. (4)–(7), sunshine and rain are added,
4. Empirical results
as well as the interaction items of them and electronic payment. In this
way, the underlying mechanism between electronic payment and con­
4.1. Descriptive statistics
sumption under different environmental scenarios and different trans­
action costs is tested.
Table 4 reports the descriptive statistics of the variables used in this
Sunshine mechanism models:
paper. It can be seen from the table that all variables are within
consumption discretionaryi,t = β0 + β1 epayi,t + β2 sunshinei,t ⋅epayi,t reasonable value ranges. At the same time, it can be seen from the
descriptive statistics that sample households spend an average of 2623
yuan (375 $) on clothes, 3450 yuan (493 $) on dining out and 5203 yuan

8
J. Li et al. International Review of Financial Analysis 85 (2023) 102472

Table 10
Household fixed effect model to test different types of consumption.
(1) (2) (3) (4) (5)

Variables consumption_clothing consumption_foods_outside consumption_travel consumption_health consumption_transportation

epay 342.41** 1096.22*** 565.85** 1106.25 − 80.43


(142.97) (282.31) (238.16) (4412.84) (144.40)
male − 519.13*** 1239.70*** − 47.44 850.12 − 80.19
(192.17) (379.69) (320.12) (1140.16) (194.10)
divorce − 134.55 − 315.08 − 196.94 2634.13*** 38.94
(139.43) (275.12) (232.26) (829.74) (140.83)
risk_aversion − 100.65*** − 125.75** − 92.25* − 14.66 − 8.73
(31.88) (63.21) (53.11) (186.46) (32.20)
income 0.01*** 0.00*** 0.01*** − 0.01*** 0.00
(0.00) (0.00) (0.00) (0.00) (0.00)
unhappiness − 94.66*** − 224.21*** − 58.20 437.47** − 8.62
(28.84) (56.99) (48.05) (170.87) (29.13)
house_number 159.64*** 339.80*** 75.87 − 7.63 − 167.45***
(40.12) (79.26) (66.83) (238.02) (40.52)
car_number 647.31*** 161.69 414.43*** − 442.34 441.59***
(66.20) (131.27) (110.27) (393.22) (66.86)
age − 31.39*** − 24.93*** − 0.62 − 72.42*** − 1.25
(2.97) (5.86) (4.95) (17.53) (3.00)
education 162.37*** 214.13*** 264.24*** − 688.02*** 5.67
(34.71) (68.59) (57.82) (205.85) (35.06)
ln_gdp 167.00 740.66 775.78* 1758.21 186.21
(252.71) (497.80) (420.96) (1507.20) (255.24)
cpi − 85.08 449.30*** − 172.00 − 1202.10*** − 23.40
(72.28) (142.87) (120.40) (449.19) (73.00)
Constant 10,613.13 − 51,216.17*** 10,353.45 113,162.31*** 749.80
(7315.77) (14,458.94) (12,186.65) (44,804.00) (7389.06)
Household FE Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Observations 42,519 42,224 42,519 40,274 42,519
R-squared 0.74 0.56 0.65 0.47 0.49

Note: In this table, we test the electronic payment on different types of consumption by controlling household fixed effect. All variables are defined in Table 1, Table 2
and Table 3. The ‘Household FE’ represents household fixed effect and ‘Year FE’ represents year fixed effect. ‘Observations’ represent observations in the sample and ‘R-
squared’ represents the goodness of fit. Standard error in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. We keep two decimal places for all numbers.

(743 $) on health care. The average gender of the surveyed population is rigidity to some extent, and it is not easily affected by impulsive be­
close to 0.5, which is consistent with the basic demographic balance. haviors or transaction costs.
The average age is about 45 years old and the average income is about The values and significance of other variables in Table 5 are also in
30,000 yuan (4286 $). At the same time, the average level of education line with economic intuition. For example, the increase in the proportion
is about 3, that is, the level of junior high school graduates, indicating of men in a family has a significant negative effect on reducing the
that the level of education is still not optimistic. It seems that the con­ expenditure on clothing, suggesting that women are more likely to buy
sumption variables show a strong skewness, mainly because they are clothes than men. On the other hand, the higher the risk aversion co­
absolute positive values. This paper takes a logarithm of them in the efficient of a family is, the more likely the family will save to prevent
robustness checks to avoid concerns in statistical inference. unexpected risk events. Therefore, the consumption amount of clothing,
dining out and tourism will be significantly reduced. In addition, the
4.2. Benchmark regressions subjective unhappiness of a family also affects its consumption behav­
iors. For families caught in serious mental sufferings, their consumption
Before formal analysis, Hausman test is conducted on regressions amount of purchasing clothing, dining out and traveling is significantly
after setting fixed effect model and random effect model respectively, reduced, but their health care expenditure is significantly increased.
and the Chi-square statistic is 1699.96 while the critical P-value is 1.96. This suggests that an unhappy family is more likely to suffer from health
The results show that the random effect model should be strongly issues. For the age variable, the higher the average age of a family is, the
rejected, indicating that the fixed effect model should be used. less clothing it chooses to buy, and its consumption of dining out is also
Table 5 reports the influence of electronic payment on different types significantly reduced. But its tourism expenditure is significantly
of consumption. As is shown in the table, using electronic payment increased, reflecting the consumption choice for retirement leisure. At
would increase a household’s consumption on clothing, dinning out and the same time, its health expenditure is significantly increased, which is
tourism significantly at the level of 1%. The economic significance is in line with the conventional logic. Finally, the number of houses and
also large, for example, using electronic payment would increase cars owned by the family reflects the accumulated wealth of it. Higher
clothing consumption by 1893 yuan (270 $). This could be explained wealth will lead to a significant increase in all kinds of household con­
potentially by two facts: First, using electronic payment would cut down sumption, which is also consistent with economic intuition. Therefore,
the consumers’ costs of visiting the shops or carrying banknotes. The use the test results verify Hypothesis 1, that is, electronic payment would
of electronic payment during travel is more convenient and avoids the significantly increase household consumption.
transaction costs and the robbery risk of carrying cash, which leads to Next, this paper classifies all kinds of consumption into discretionary
the increase of various types of consumption. Second, using electronic consumption and necessary consumption and investigates the impact of
payment would induce more impulse buying and increase consumption. electronic payment on the two types of comprehensive consumption. As
On the other hand, for the necessary expenditure such as health care can be seen from the results in Table 6, electronic payment significantly
expenditure and transportation expenditure, the electronic payment improves discretionary consumption but has no obvious effect on
shows no significant effect, indicating that their consumption has necessary consumption. This suggests that the electronic payment is

9
J. Li et al. International Review of Financial Analysis 85 (2023) 102472

Table 11
Household fixed effect model to test two types of comprehensive consumption.
(1) (2) (3) (4) (5) (6)

Variables consumption_discretionary consumption_necessary consumption_discretionary consumption_necessary consumption_discretionary consumption_necessary

epay 1988.78*** 1281.21 7366.64*** 2107.97 4172.90*** − 6535.33


(480.52) (4482.00) (2595.75) (4488.85) (1170.45) (8491.92)
sunshine_epay − 2.61** − 2.09
(1.28) (2.21)
sunshine − 0.36 3.60***
(0.66) (1.15)
rain_epay 6.16*** 3.09
(1.07) (2.72)
rain 0.27 − 1.80*
(0.51) (0.92)
male 542.90 785.12 405.58 785.42 587.79 817.44
(641.46) (1158.00) (664.62) (1157.97) (641.08) (1157.95)
divorce − 793.24* 2694.11*** − 868.96* 2714.00*** − 802.59* 2675.62***
(464.44) (842.71) (482.59) (842.54) (464.13) (842.71)
risk_aversion − 381.81*** − 23.27 − 390.82*** − 23.96 − 381.51*** − 30.37
(106.38) (189.51) (108.66) (189.42) (106.32) (189.50)
income 0.02*** − 0.01*** 0.02*** − 0.01*** 0.02*** − 0.01***
(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
unhappiness − 394.93*** 429.30** − 372.46*** 421.50** − 392.50*** 428.38**
(96.43) (173.60) (99.79) (173.57) (96.36) (173.61)
house_number 623.01*** − 185.30 650.36*** − 170.37 611.70*** − 158.20
(134.01) (241.80) (138.87) (241.75) (134.09) (242.20)
car_number 2119.95*** 28.21 2218.71*** − 7.39 2144.95*** 31.08
(222.62) (399.40) (230.96) (399.48) (222.51) (399.61)
age − 62.15*** − 73.32*** − 60.44*** − 73.08*** − 62.64*** − 73.26***
(9.90) (17.81) (10.20) (17.81) (9.89) (17.81)
education 612.87*** − 679.51*** 604.44*** − 674.86*** 599.31*** − 684.02***
(116.18) (209.11) (120.30) (209.04) (116.12) (209.14)
ln_gdp 1822.58** 2037.01 1800.05* 3736.32** 1431.26 3014.85*
(841.94) (1531.41) (930.88) (1623.63) (891.68) (1616.00)
cpi 36.23 − 1276.11*** − 178.21 − 1341.27*** 42.47 − 1363.14***
(241.06) (456.42) (261.96) (456.63) (242.61) (458.81)
Constant − 14,645.34 118,075.90*** 8433.14 99,142.91** − 11,278.76 118,282.23***
(24,392.30) (45,506.14) (26,309.56) (45,895.85) (24,384.61) (45,505.62)
Household FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Observations 41,569 42,519 39,357 40,274 41,569 40,274
R-squared 0.71 0.47 0.70 0.47 0.71 0.47

Note: In this table, we test the electronic payment on two different types of comprehensive consumption by controlling household fixed effect. All variables are defined
in Table 1, Table 2 and Table 3. The ‘Household FE’ represents household fixed effect and ‘Year FE’ represents year fixed effect. ‘Observations’ represent observations in
the sample and ‘R-squared’ represents the goodness of fit. Standard error in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. We keep two decimal places for all
numbers.

more likely to induce impulse consumption of the discretionary goods hamper consumption.
rather than necessary goods. At the same time, the coefficients of other Table 7 reports the regression results using the fixed effect model. It
variables are also reasonable. For example, for families with higher can be seen from the table that, consistent with the previous results, the
levels of unhappiness, their discretionary consumption amount de­ use of electronic payment will greatly increase the expenditure on
creases significantly, while the amount of necessary consumption in­ clothing, dining out and tourism. While it does not significantly influ­
creases significantly, which is consistent with above analysis. For ence the consumption of health and transportation. Furthermore, the
families with higher risk aversion coefficients, their discretionary con­ interaction term of sunshine and electronic payment is significantly
sumption decreases significantly, but the necessary consumption is not negative at the significance level of 1%. This means that electronic
obviously influenced. Meanwhile, for families with higher average age, payment would increase consumption more in rainy days than sunny
their discretionary consumption decreases significantly, while the days. As has been analyzed before, this is because the transaction costs
necessary consumption for health care and transportation increases on rainy day are higher than sunny days. Using electronic payment
significantly. This reflects the fact that elderly people need more medical would cut transaction costs more significantly in rainy days. On the
assistance. contrary, the stronger the sunshine is, the lower the transaction costs of
using cash payment. This in turn weakens the electronic payment’s
ability to cut down transaction costs to increase consumption. It shows
4.3. Mechanism analysis that reducing transaction costs is an important economic mechanism for
electronic payment to influence consumption.
We have claimed that the electronic payment would increase con­ This paper further tests the mechanism of electronic payment on
sumption in that it can reduce transaction costs for the consumers. But it discretionary consumption and necessary consumption, which are
is hard to find a variable that could measure the transaction costs that a defined as two types of comprehensive consumption. The test results are
consumer may face in a consumption experience. Instead, we focus on shown in Table 8. Consistent with the above conclusions, the use of
the natural environment that gives us a completely exogenous measure. electronic payment has a significant effect on the increase of discre­
It is quite intuitive that the transaction costs in rainy days with less tionary consumption but has no significant effect on the necessary
sunshine are more likely to be high. The costs of venturing out in a consumption. This paper further adds the amount of rainfall to identify
stormy day to withdraw cash from the bank are disturbing and may

10
J. Li et al. International Review of Financial Analysis 85 (2023) 102472

Table 12
Alternative definition of necessary consumption.
(1) (2) (3)

Variables consumption_education consumption_utility consumption_necessay_robustness

epay 310.02 83.59 741.68


(425.15) (86.52) (581.26)
male − 777.52 23.97 − 1610.01***
(571.47) (22.10) (568.88)
divorce 156.59 − 5.77 − 633.37
(414.62) (16.12) (504.25)
risk_aversion 12.15 − 4.64 − 18.19
(94.82) (3.63) (92.50)
income 0.01*** 0.00*** 0.02***
(0.00) (0.00) (0.00)
unhappiness 107.78 − 2.12 512.41***
(85.77) (3.31) (134.54)
house_number 141.31 8.47* 1562.40***
(119.31) (4.59) (182.52)
car_number 32.41 62.52*** 2784.83***
(196.85) (7.60) (262.46)
age − 100.53*** − 1.38*** 3.20
(8.83) (0.34) (8.18)
education 1574.72*** − 1.15 943.85***
(103.22) (3.98) (86.41)
ln_gdp − 415.14 − 69.64** 2533.45
(751.47) (28.89) (1739.35)
cpi − 189.26 3.88 − 2384.66***
(214.93) (8.27) (512.86)
Constant 27,285.64 647.57 223,868.43***
(21,754.89) (836.40) (52,188.93)
Household FE Yes Yes Yes
Year FE Yes Yes Yes
Observations 42,519 41,420 74,740
R-squared 0.63 0.59 0.04

Note: In this table, we change the definition of our necessary consumption to include education consumption and utility consumption. All variables are defined in
Table 1, Table 2 and Table 3. The ‘Household FE’ represents household fixed effect and ‘Year FE’ represents year fixed effect. ‘Observations’ represent obser­
vations in the sample and ‘R-squared’ represents the goodness of fit. Standard error in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. We keep two decimal
places for all numbers.

the transaction costs that the consumers may face. It can be seen that the acceptance of electronic payment would be more prevalent in urban
interaction term between sunshine and electronic payment is signifi­ areas. It can cut down transaction costs more significantly in cities than
cantly negative in discretionary consumption, while the interaction term in countryside, creating larger urban consumption promotion.
between rain and electronic payment is significantly positive in discre­ Second, this paper tests whether electronic payment would affect
tionary consumption. This indicates that weaker sunshine or heavier educated people in a different way. It is estimated that on average, the
rainfall would incur larger transaction costs. Therefore, the electronic effect of electronic payment on consumption would increase by 1062
payment becomes convenient, and it incentivizes households to yuan (152 $) if the average education level of a household increases by
consume more. When it comes to necessary consumption, the results are one. That means, the well-educated group would increase their con­
not significant. This is also intuitive in that the weather would not sumption more when they use electronic payment. This is due to the fact
prevent necessary consumption like medical care. This further shows that they have lower barriers to the efficient use of electronic payment,
that electronic payment can effectively reduce transaction costs and so the electronic payment would cut down the transaction costs more
increase consumption. This indicates that reducing transaction costs is significantly for them.
an important transmission channel for electronic payment to promote Third, this paper compares the effect of electronic payment on con­
household consumption, thus verifying Hypothesis 2. sumption in different age groups. It is found that one year increase in age
would weaken the effect by 59 yuan (8 $) on average. Younger people
4.4. Heterogeneity are more likely to increase their discretionary consumption if they use
electronic payment. This could be explained by the fact that young
This paper further tests the heterogeneity in different subsamples to people are more exposed to impulse buying but elderly people are more
find whether electronic payment would increase consumption differ­ likely to consume with prudence. Another reason is that young people
ently in different groups. This paper tests the urban & rural heteroge­ are more proficient in using electronic devices, so they would use
neity, education level heterogeneity and age heterogeneity. The results electronic payment more frequently. The higher frequency of using
are shown in Table 9. electronic payment, the more likely that it would save transaction costs.
First, this paper divides the full sample into two groups. One group So, there is a more pronounced effect in young people group than old
contains all the households dwelling in the urban areas, while the other people group.
group in the rural areas. This paper denotes the variable rural to be one if In conclusion, electronic payment is more likely to increase the
the household locates in the rural areas and zero otherwise. From the consumption of those urban, well-educated and young households, thus
first column of Table 9, it can be seen that the electronic payment in­ verifying Hypothesis 3.
creases urban households’ discretionary consumption by 7979 yuan
(1140 $), which is significant at the level of 1%. However, it is estimated
that this impact would only be 2215 yuan (316 $) in rural households.
This could be explained by the fact that the access to the internet and the

11
J. Li et al. International Review of Financial Analysis 85 (2023) 102472

Table 13 The regression results are shown in Table 12. It seems that the electronic
Log regressions. payment would not influence the consumption of education or elec­
(1) (2) tricity significantly. The newly defined necessary consumption is also
virtually unaffected by the electronic payment. This further supports the
Variables ln_consumption_discretionary ln_consumption_necessary
previous findings.
epay 0.28*** 0.02 As is mentioned before, the consumption variables show great
(0.02) (0.05)
male 0.09*** − 0.43***
skewness mainly because it takes value only larger than zero. To address
(0.02) (0.05) this problem, this paper takes a logarithm of those consumption vari­
divorce − 0.37*** − 0.34*** ables to alleviate the problem of skewness on statistical inference (Arjas
(0.02) (0.04) & Haara, 1988). As is shown in Table 13, the electronic payment still
risk_aversion − 0.11*** − 0.01
significantly increases discretionary consumption at the level of 1% but
(0.00) (0.01)
income 0.00*** 0.00*** it does not impact necessary consumption very much. This supports our
(0.00) (0.00) previous findings.
unhappiness − 0.11*** 0.07***
(0.01) (0.01) 6. Conclusions
house_number 0.13*** 0.18***
(0.01) (0.02)
car_number 0.51*** − 0.26*** This paper adopts the data of China Household Finance Survey from
(0.01) (0.02) 2011 to 2017 to empirically test the relationship between electronic
age − 0.02*** 0.02*** payment and households’ consumption. It is found that the use of
(0.00) (0.00)
electronic payment would significantly increase a household’s clothing,
education 0.24*** 0.07***
(0.00) (0.01) dinning out and tourism consumption. But it does not affect necessary
ln_gdp 0.43*** − 0.73*** consumption like medical care, transportation, education and utilities.
(0.07) (0.22) The underlying economic mechanism of transaction costs is tested
cpi 0.05** − 0.15** empirically. When the rain is heavy or sunshine is weak, the transaction
(0.02) (0.06)
Constant − 0.70 30.16***
costs are high. Using electronic payment in bad weather would reduce
(2.17) (6.40) transaction costs more significantly, thus increasing consumption more
Province FE Yes Yes significantly. The heterogeneity tests show that electronic payment
Year FE Yes Yes would foster consumption more significantly in urban, well-educated
Observations 69,864 40,876
and young households.
R-squared 0.33 0.27
This paper also implies some policy suggestions. First, we provide
Note: In this table, we take a log of the consumption level to avoid its skewness empirical evidence that using electronic payment would foster con­
concern. All variables are defined in Table 1, Table 2 and Table 3. The ‘Province sumption by cutting transaction costs. This is particularly important in
FE’ represents province fixed effect and ‘Year FE’ represents year fixed effect.
the context of COVID-19. In China, epidemic isolation is a frequent
‘Observations’ represent observations in the sample and ‘R-squared’ represents
occurrence, which puts the transaction costs of venturing out to extreme.
the goodness of fit. Standard error in parentheses. *** p < 0.01, ** p < 0.05, * p
< 0.1. We keep two decimal places for all numbers.
The electronic payment could solve this by allowing consumers to place
orders online at home, which alleviates the hazardous impact of trans­
action costs. This insight is also useful for other developing countries.
5. Robustness checks
Second, we find that the acceptance of electronic payment is not opti­
mistic in rural areas. This further suggests government to foster the
5.1. Household fixed effect
penetration of internet to rural areas. By allowing those remote areas to
use electronic payment, the government can further reduce their
There may be unobservable factors at the household level that affect
transaction costs of traveling a long distance for consumption. This is
both electronic payment and consumption, which would cause endo­
meaningful in increasing consumption and eliminating poverty in rural
geneity concerns. This paper further controls the fixed effect at the
areas.
family level to repeat the benchmark regressions. The results are shown
This paper also points out some future research directions. Due to the
in Table 10 and Table 11. As is shown in the table, the electronic pay­
data availability, this paper doesn’t-test other mechanisms like liquidity
ment still significantly increases a household’s consumption on clothing,
constraints, which require sophisticated measurement of consumers’
dinning out and tourism. It is estimated that the electronic payment
personal bank accounts. More micro-level data could be used to test this
would increase a household’s consumption on dinning out by 1096 yuan
problem in the future research.
(157 $), which is economically significant. Having controlled household
level fixed effect, the interactions between sunshine, rain and electronic
Author statement
payment are still significant. In the weather of weak sunshine and
torrential rain, the electronic payment cuts down transaction costs more
We declare that we do not have any commercial or associative in­
and increases consumption more significantly. Those results suggest that
terest that represents a conflict of interest in connection with the work
our previous conclusions are robust.
submitted.

5.2. Alternative measures Acknowledgements

To address the potential concern that this paper may not include This paper is supported by the Major Program of the National Social
some other consumption like education, water and electricity. This Science Foundation of China (Grant No. 21&ZD117).
paper includes two more consumption in the definition of necessary
consumption. The education expenditure is depicted by the variable References
consumption_education, and it is defined in Table 1. The utility expendi­
ture like water, electricity etc. is depicted by the variable con­ Aggarwal, R. (2021). Impacts of climate shocks on household consumption and
inequality in India. Environment and Development Economics, 26(5–6), 488–511.
sumption_utility, and it is defined in Table 1 too. Those two sources of Agnew, M. D., & Thornes, J. E. (1995). The weather sensitivity of the UK food retail and
expenditure are also added to our definition of necessary consumption. distribution industry. Meteorological Applications, 2(2), 137–147.

12
J. Li et al. International Review of Financial Analysis 85 (2023) 102472

Aguiar, M., & Bils, M. (2015). Has consumption inequality mirrored income inequality? Jack, W., & Suri, T. (2014). Risk sharing and transactions costs: Evidence from Kenya’s
American Economic Review, 105(9), 2725–2756. mobile money revolution. American Economic Review, 104(1), 183–223.
Ahn, S. Y., & Nam, Y. (2022). Does mobile payment use lead to overspending? The Joo, H. H., Kang, H. G., & Moon, J. J. (2014). The effect of rain on the decision to visit a
moderating role of financial knowledge. Computers in Human Behavior, 134, theme park. Asia Pacific Journal of Tourism Research, 19(1), 61–85.
107–319. Kang, J., & Reiner, D. M. (2022). What is the effect of weather on household electricity
Alemu, T., Bandyopadhyay, T., & Negash, S. (2015). Electronic payment adoption in the consumption? Empirical evidence from Ireland. Energy Economics, 111, 106–123.
banking sector of low-income countries. International Journal of Information Systems Kim, Y. G., & Li, G. (2009). Customer satisfaction with and loyalty towards online travel
in the Service Sector, 7(4), 27–47. products: A transaction cost economics perspective. Tourism Economics, 15(4),
Anderson, D. M., Strand, A., & Collins, J. M. (2018). The impact of electronic payments 825–846.
for vulnerable consumers: Evidence from social security. Journal of Consumer Affairs, Kumaga, D. (2011). The challenges of implementing electronic payment systems: The case of
52(1), 35–60. Ghana’s E-zwich payment system. Working Paper.
Arjas, E., & Haara, P. (1988). A note on the asymptotic normality in the cox regression Laibson, D. (1997). Golden eggs and hyperbolic discounting. Quarterly Journal of
model. The Annals of Statistics, 16(3), 1133–1140. Economics, 112(2), 443–478.
Bachmann, R., & Sims, E. R. (2012). Confidence and the transmission of government Liébana-Cabanillas, F., García-Maroto, I., Muñoz-Leiva, F., & Ramos-de-Luna, I. (2020).
spending shocks. Journal of Monetary Economics, 59(3), 235–249. Mobile payment adoption in the age of digital transformation: The case of apple pay.
Bajaj, A., & Damodaran, N. (2022). Consumer payment choice and the heterogeneous Sustainability, 12(13), 5443.
impact of India’s demonetization. Journal of Economic Dynamics and Control, 137, Lise, J., & Seitz, S. (2011). Consumption inequality and intra-household allocations.
104–329. Review of Economic Studies, 78(1), 328–355.
Bechlioulis, A. P., & Brissimis, S. N. (2019). Consumer debt non-payment and the Liu, D., Wang, W., & Zhao, Y. (2021). Effect of weather on online food ordering.
borrowing constraint: Implications for consumer behavior. Journal of Banking & Kybernetes, 51(1), 165–209.
Finance, 101, 161–172. Liu, J., Sockin, M., & Xiong, W. (2021). Data privacy and consumer vulnerability. Working
Bennion, E. A., & Nickerson, D. W. (2011). The cost of convenience: An experiment Paper.
showing e-mail outreach decreases voter registration. Political Research Quarterly, 64 Liu, T., Pan, B., & Yin, Z. (2020). Pandemic, mobile payment, and household
(4), 858–869. consumption: Micro-evidence from China. Emerging Markets Finance and Trade, 56
Blundell, R., Pistaferri, L., & Saporta-Eksten, I. (2016). Consumption inequality and (10), 2378–2389.
family labor supply. American Economic Review, 106(2), 387–435. Ludvigson, S. (1999). Consumption and credit: A model of time-varying liquidity
Blundell, R., Pistaferri, L., & Saporta-Eksten, I. (2018). Children, time allocation, and constraints. Review of Economics and Statistics, 81(3), 434–447.
consumption insurance. Journal of Political Economy, 126(1), 73–115. Ma, Q., Wang, M., He, Y., Tan, Y., & Zhang, L. (2021). Does mobile payment change
Bonaparte, Y., Cooper, R., & Zhu, G. (2012). Consumption smoothing and portfolio consumers’ perception during payment process? An ERP study. Neuroscience Letters,
rebalancing: The effects of adjustment costs. Journal of Monetary Economics, 59(8), 762, 136–138.
751–768. Manshad, M. S., & Brannon, D. (2021). Haptic-payment: Exploring vibration feedback as
Brito, D. L., & Hartley, P. R. (1995). Consumer rationality and credit cards. Journal of a means of reducing overspending in mobile payment. Journal of Business Research,
Political Economy, 103(2), 400–433. 122, 88–96.
Chintagunta, P. K., Chu, J., & Cebollada, J. (2012). Quantifying transaction costs in Meyer, B. D., & Sullivan, J. X. (2013). Consumption and income inequality and the great
online/off-line grocery channel choice. Marketing Science, 31(1), 96–114. recession. American Economic Review, 103(3), 178–183.
Cortés, K., Duchin, R., & Sosyura, D. (2016). Clouded judgment: The role of sentiment in Oyelami, L. O., Adebiyi, S. O., & Adekunle, B. S. (2020). Electronic payment adoption
credit origination. Journal of Financial Economics, 121(2), 392–413. and consumers’ spending growth: Empirical evidence from Nigeria. Future Business
Dai, Y., Rau, P. R., Stouraitis, A., & Tan, W. (2020). An ill wind? Terrorist attacks and Journal, 6(1), 1–14.
CEO compensation. Journal of Financial Economics, 135(2), 379–398. Riley, E. (2018). Mobile money and risk sharing against village shocks. Journal of
Dalton, P. S., Pamuk, H., Ramrattan, R., van Soest, D., & Uras, B. (2019). Transparency Development Economics, 135, 43–58.
and financial inclusion: Experimental evidence from mobile money. Working Paper. Schierz, P. G., Schilke, O., & Wirtz, B. W. (2010). Understanding consumer acceptance of
Dominko, M., & Verbič, M. (2022). The effect of subjective well-being on consumption mobile payment services: An empirical analysis. Electronic Commerce Research and
behavior. Journal of Consumer Affairs, 56(2), 876–898. Applications, 9(3), 209–216.
Ekici, T., & Dunn, L. (2010). Credit card debt and consumption: Evidence from Sexton, S. (2015). Automatic bill payment and salience effects: Evidence from electricity
household-level data. Applied Economics, 42(4), 455–462. consumption. Review of Economics and Statistics, 97(2), 229–241.
Gertler, P., & Gruber, J. (2002). Insuring consumption against illness. American Economic Summers, T. A., & Hebert, P. R. (2001). Shedding some light on store atmospherics:
Review, 92(1), 51–70. Influence of illumination on consumer behavior. Journal of Business Research, 54(2),
Govind, R., Garg, N., & Sun, W. (2014). Geographically varying effects of weather on 145–150.
tobacco consumption: Implications for health marketing initiatives. Health Marketing Tee, H. H., & Ong, H. B. (2016). Cashless payment and economic growth. Financial
Quarterly, 31(1), 46–64. Innovation, 2(1), 1–9.
Gross, D. B., & Souleles, N. S. (2002). Do liquidity constraints and interest rates matter Thaler, R. (1985). Mental accounting and consumer choice. Marketing Science, 4(3),
for consumer behavior? Evidence from credit card data. Quarterly Journal of 199–214.
Economics, 117(1), 149–185. Yang, S., Markoczy, L., & Qi, M. (2007). Unrealistic optimism in consumer credit card
Hirshleifer, D., & Shumway, T. (2003). Good day sunshine: Stock returns and the adoption. Journal of Economic Psychology, 28(2), 170–185.
weather. The Journal of Finance, 58(3), 1009–1032. Zandi, M., Singh, V., & Irving, J. (2013). The impact of electronic payments on economic
Hou, L., Hsueh, S. C., & Zhang, S. (2021). Digital payments and households’ growth. Moody’s Analytics: Economic and Consumer Credit Analytics, 217(2).
consumption: A mental accounting interpretation. Emerging Markets Finance and Zeldes, S. P. (1989). Consumption and liquidity constraints: An empirical investigation.
Trade, 57(7), 2079–2093. Journal of Political Economy, 97(2), 305–346.
Huebner, J., Fleisch, E., & Ilic, A. (2020). Assisting mental accounting using Zhang, Y., Zhang, G., Liu, L., De Renzis, T., & Schmiedel, H. (2019). Retail payments and
smartphones: Increasing the salience of credit card transactions helps consumer the real economy. Journal of Financial Stability, 44, 100–690.
reduce their spending. Computers in Human Behavior, 113, 106–504. Zhao, C., Wu, Y., & Guo, J. (2022). Mobile payment and Chinese rural household
consumption. China Economic Review, 71, 101–719.

13

You might also like