论文信息

Jing Zhan / Junlin Jiang / Bin Liu

Frontiers in Sustainable Food Systems

论文解读

AI 辅助整理
方法
中国家庭金融调查2015-2019年面板数据,中介效应分析。
发现
数字乡村转型显著提升农户金融资产及收益,推动消费向发展与享受型转变。
机会点
可探究数字乡村金融资产积累机制在不同区域与群体间的差异及长期福祉效应。

论文摘要

已有研究主要将乡村数字化视为一种交易成本冲击,而较少关注另一个边际:数字乡村转型通过改善金融服务、市场和投资机会的可及性,可能使农村家庭得以积累金融资产,进而改善其经济福祉。本文利用中国家庭金融调查(CHFS)2015年、2017年和2019年三轮合并面板数据,实证分析数字乡村转型对农村家庭消费结构的影响以及金融资产的中介作用。研究发现,数字乡村转型显著提高了农村家庭的金融资产持有量及其产生的收入,且这种积累解释了其后向发展与享受型消费转变的部分原因。异质性分析进一步表明,较年轻、受教育程度较高、已婚、工资性收入较高、有贷款以及处于欠发达地区的农村家庭从数字乡村转型中获益更多。本研究的核心贡献在于将金融资产积累确立为数字乡村转型发挥作用的农村生计机制,从而为数字经济时代的农村消费研究提供了新的理论视角。

全文

Frontiers in Sustainable Food Systems

TYPE Original Research
PUBLISHED 16 November 2026
DOI 10.3389/fsufs.2026.1760055
OPEN ACCESS
EDITED BY
Justice Gameli Djokoto,
Dominion University College, Ghana
REVIEWED BY
Jingkuang Liu,
Guangzhou University, China
Yanxin Zhou,
Chongqing University, China
Tao Cen,
Ningbo University of Finance and
Economics, China
CORRESPONDENCE
Bin Liu
2229401012@st.gxu.edu.cn;
202601132@jyu.edu.cn
†PRESENTADDRESS
Bin Liu,
School of Economics and Management,
Jiaying University, Meizhou, China
RECEIVED 03 December 2025
REVISED 10 October 2026
ACCEPTED 20 October 2026
PUBLISHED 16 November 2026
CITATION
Zhan J, Jiang J and Liu B (2026) Digital
rural transformation, financial assets, and
rural household wellbeing.
Front. Sustain. Food Syst. 10:1760055.
doi: 10.3389/fsufs.2026.1760055
COPYRIGHT
© 2026 Zhan, Jiang and Liu. This is an
open-access article distributed under the
terms of theCreative Commons
Attribution License (CC BY).T h eu s e ,
distribution or reproduction in other
forums is permitted, provided the
original author(s) and the copyright
owner(s) are credited and that the
original publication in this journal is
cited, in accordance with accepted
academic practice. No use, distribution
or reproduction is permitted which does
not comply with these terms.
Digital rural transformation,
financial assets, and rural
household wellbeing
Jing Zhan1, Junlin Jiang1 and Bin Liu2
†
1School of Economics, Management and Law, University of South China, Hengyang, China,2School of
Business, Guangxi University, Nanning, China
Existing research treats rural digitalization mainly as a transaction-cost shock
and pays little attention to a second margin: by improving access to financial
services, markets and investment opportunities, digital rural transformation may
enableruralhouseholdstoaccumulatefinancialassetsandtherebyimprovetheir
economic well-being. Using pooled panel data from the 2015, 2017 and 2019
waves of the China Household Finance Survey (CHFS), we empirically analyze
the impact of digital rural transformation on rural households’ consumption
structure and the mediating role of financial assets. The findings reveal that
digital rural transformation significantly raises rural households’ financial asset
holdings and the income those assets generate, and that this accumulation
accounts for part of the subsequent shift toward development and enjoyment-
oriented consumption. Heterogeneity analysis further indicates that younger,
more educated, married rural households with higher wage income, loans, and
those in less developed regions benefit more from digital rural transformation.
The central contribution of this study is to establish financial asset accumulation
as a rural livelihood mechanism through which digital rural transformation
operates, thereby offering a new theoretical perspective for rural consumption
research in the digital economy era.
KEYWORDS
consumptionupgrading,digitalruraltransformation,financialassetaccumulation,rural
China, rural household wellbeing
1 Introduction
The digital rural initiative is poised to revitalize China’s sluggish rural household
consumption. In recent years, the disparity between urban and rural regions
in China has been gradually diminishing ( Suetal., 2015 ), and the consumption
capacity of rural households has increased accordingly ( Lagakos, 2020). However,
basic consumption still accounts for 70–80% of the total consumption of rural
households (Kanemotoetal., 2019 ; Sunetal., 2021 ), suggesting substantial potential for
optimization in rural households’ consumption patterns. Upgrading the consumption
structure of rural households is an important engine for balanced urban–rural
development and for sustained, domestically driven economic growth ( Yuetal.,
2021). Accordingly, in 2021, China’s Ministry of Commerce and twelve other
government agencies jointly issued the “Notice on Measures to Promote Key Bulk
Consumption While Unleashing Rural Consumption Potential,” implementing targeted
policies to address structural deficiencies in rural consumption and stimulate market
vitality. Currently, online consumption in rural areas is becoming increasingly active
Frontiersin SustainableFoodSystems 01 frontiersin.org

Zhan et al. 10.3389/fsufs.2026.1760055
(RäisänenandTuovinen, 2020 ). While physical infrastructure
(e.g., water, electricity, and logistics) continues to improve
(Xiaoetal., 2022 ), the soft environment—particularly digital
ruraltransformation—requiresfurtherenhancement( Huangetal.,
2023). In 2022, the Action Plan for the Development of Digital
Villages (2022-2025), issued by the Central Internet Information
Office and 10 other departments, points out that digital rural
transformation is a crucial initiative to promote the upgrading
of the rural consumption structure. Consequently, how digital
ruraltransformationstimulatestheruralhouseholds’consumption
structureupgradinghasattractedgrowingscholarlyattention.
The digital rural initiative encompasses the comprehensive
integrationofmoderninformationtechnologyintoruraleconomic
and social activities (Maliketal., 2022 ), which in turn drives
ruralmodernizationandtransformationfromwithin.Thisconcept
highlights the crucial function of information technology in
rural advancement and aims to advance the rural revitalization
strategy through digital tools (Shengetal., 2022). Rural household
consumption is a key indicator of rural prosperity (Xieetal., 2022)
and has drawn extensive attention in the digital economy era
(AnkrahTwumasietal., 2021 ; Bakkeretal., 2021 ; Kebede, 2022).
Scholars have mainly examined how internet usage (Houetal.,
2018), mobile payments (Zhaoetal., 2022 ), and digital finance
(Liuetal., 2021 ) influence rural household consumption behavior
and overall consumption. First, the widespread adoption of the
internethasremarkablyenhancedtheruralconsumerenvironment
(Khanetal., 2022 ). It has not only increased rural household
consumptionbutalsoimproveditsstructure,particularlyspending
on hedonic goods and services. Additionally, the impact of
the internet on rural household consumption varies across
different income groups (KhanalandMishra, 2016 ). Second, the
low cost, security, and convenience of mobile payments have
led rural households to prefer electronic payment methods
(Deichmannetal., 2016 ), thereby altering their consumption
habits. Third, digital finance has been found to provide more
consumption options for rural households (Yangetal., 2022). The
provisionofdiversefinancialserviceshasimprovedfinancialaccess
and raised non-agricultural employment among rural households,
ultimatelydrivinguptheirconsumption.
While existing research has established the positive effects
of digital technologies (e.g., Internet access, mobile payments,
digital finance) on rural consumption patterns, scholars have
yet to systematically examine how digital rural transformation
facilitates consumption structure upgrading and the mechanisms
through which it operates. We position our study within three
ongoing debates. The first concerns the outcome: prior work
has predominantly measured rural consumption in aggregate
terms, and thus speaks to whether digital technologies expand
consumption, but leaves open whether they restructure it. The
second concerns the mechanism: the received explanation runs
through lower transaction costs, relaxed liquidity constraints
and payment convenience, whereas the household-balance-sheet
channel — whether digitalization changes what rural households
hold, and hence what they can spend — has received little formal
attention. We treat this channel as a livelihood mechanism rather
than as a financing convenience: financial assets are a store of
value and a recurrent source of household income, and access
to the financial services, markets and investment opportunities
needed to build such a position is itself one of the things digital
rural transformation changes. The third concerns the distribution:
a substantial literature argues that digital dividends accrue
disproportionately to already-advantaged households, implying a
wideningdigitaldivide.
Thisstudyusesmicrodatafromthe2015,2017and2019waves
of the China Household Finance Survey (CHFS) microdata to
empirically examine how digital rural transformation influences
rural consumption structure upgrading and its underlying
mechanisms. Read against these debates, our contribution is
threefold. First, and most importantly, we formalize and test
a financial-asset channel that prior work has largely assumed
away, and we interpret it as a rural livelihood mechanism: digital
ruraltransformationimprovesruralhouseholds’accesstofinancial
services, markets and investment opportunities, and hence their
capacity to accumulate assets. Second, we shift the outcome from
the level to the composition of consumption, which is where a
changeinwhathouseholdsholdshouldshowup.Third,wenuance
the distributional debate, since our heterogeneity results show
larger effects for households in central and western provinces and
for those at the lower end of the consumption distribution — a
pattern consistent with a narrowing, rather than widening, rural
consumptiongap.
The remainder of the paper is organized as follows. Section
2 develops the theoretical framework and proposes testable
hypotheses. Section 3 details the empirical methodology and data.
Section 4 reports the core empirical findings. Section 5 concludes
withkeyinsightsandpolicyimplications.
2 Theoretical mechanisms and
research hypotheses
As digital rural transformation deepens, the time rural
households need to purchase goods and services falls sharply,
alleviating the problem of information asymmetry between buyers
andsellers( Heeksetal.,2021 ;Xieetal.,2020 ).Thischangeenables
ruralhouseholdstoobtainbetter-pricedgoodsandservicesatlower
marginal search costs, thus helping to optimize the consumption
structureandmaximizeconsumptionutility.Inthisprocess,digital
rural transformation influences rural household consumption
primarily through two aspects: First, it changes consumption
decisionsbyloweringtransactioncosts—thepriceeffect.Second,it
widensthesetofchoicesavailabletoruralhouseholds—themarket
scopeeffect.
2.1 Price effect
The price effect exhibits its complex dual mechanism of
action in shaping the consumption decisions of rural households.
GalorandTsiddon (1997) show that the returns to unobservable
skills of the labor force (e.g., teamwork skills, leadership,
and learning abilities) have increased in response to the new
technological push, which may lead to an increase in the price
Frontiersin SustainableFoodSystems 02 frontiersin.org

Zhan et al. 10.3389/fsufs.2026.1760055
of commodities. However, the advantage of these unobservable
skills diminishes with the widespread use of Internet technology,
which reduces the level of the commodity premium. The
spread of the Internet and the slower growth in labor costs
have lowered the barriers to entry for firms into the Internet
market,whilepushingtheproductionpossibilitiesfrontieroutward
(Bhagwatietal., 2004 ). The change in the market entry threshold
attracts more enterprises to join the Internet platform, creating
a competitive effect, i.e., the openness of the Internet platform
breaks the geographical monopoly, introduces online and offline
competition, and improves the substitutability of products, thus
directlyreducingthemarketprice( Fanetal.,2018 ;Ratchfordetal.,
2022). The rightward shift of the production possibilities frontier
implies an increase in the quantity of goods on the market, and in
aneffectivelycompetitivemarketenvironment,thevendors’pricing
space for their own products and services is further squeezed
(KauffmanandWang, 2001 ). Under the condition of keeping
marginal cost basically stable and market demand unchanged in
the short term, an increase in production is often accompanied by
a decrease in price, which makes the Internet platform tend to pull
downtheactualsellingpriceofproducts( Cavallo,2017).
Second, digital rural transformation helps to reduce the
transaction costs of rural households’ consumption. Transaction
costs mainly include information search costs, contract signing
costs and fulfillment costs (ChoiandKrause, 2006; Geyskensetal.,
2006; Teece, 2014). Rural households can use digital platforms to
directly and efficiently obtain information on goods and services,
andonlinesearchisfarmoreefficientthantraditionalofflinesearch.
Digital platforms leverage advanced technologies like artificial
intelligence and big data to precisely identify rural households’
needs and enhance transaction matching efficiency. According to
Bakos(1997),theequilibriumpriceinthemarketis P∗ = √sf (P∗,
s, andf stand for the equilibrium price, search cost, and utility loss
cost, respectively) when search cost exists. Changes in search costs
affecttheequilibriumpriceinthesamedirection.Inthetraditional
transaction model, due to the existence of multiple links such as
acquisition and distribution, intermediaries often take advantage
ofinformationtoincreasethesellingprice( ReschandKock,2021 ).
Digital rural transformation enhances the information advantage
of rural households in the market by providing transparent
product information through Internet information technology
and directly connecting buyers and sellers. This avoids additional
intermediary costs and helps break down information barriers
and reduce information asymmetry (Razzaqetal., 2023 ). Also,
it improves the efficiency of commodity information collection,
thus reducing the marginal search cost ( GoldfarbandTucker,
2019). Ultimately, the equilibrium price of market commodities
is reduced, enabling rural households to purchase commodities
at more favorable prices, which promotes the upgrading of
rural households’ consumption structure. Furthermore, digital
rural platforms facilitate streamlined information exchange
between rural households and producers through multi-channel
communicationsystems( TsunodaandZennyo,2021 ).Thisenables
full disclosure of key information such as price, style and quality
of products. The information gap between the two parties to a
transaction is narrowed, reducing contracting costs and increasing
thesuccessrateofthetransaction.
FIGURE 1
Market structure under the long-tail theory.
2.2 Market scope effect
Digital rural transformation transcends traditional geographic
constraints in commodity transactions, thereby more effectively
meeting rural households’ consumption demands. In the
traditional business model, vendors are restricted by geographical
location (Eidseetal., 2016 ), and their market share is limited by
information costs, making it difficult to expand market coverage
(Blankespooretal., 2014 ). However, digital rural transformation,
by connecting people, devices and platforms, blurs geographic
boundaries(Hampshireetal.,2021 ),andthiseffectismoreobvious
in online transactions. Consequently, vendors have achieved
substantial expansion in both market penetration and service
coverage. They have realized the synchronized development
of online and offline, facilitated cross-city, cross-province, and
even cross-border transactions by rural households, met the
consumption needs of less developed regions, and gradually
achieved undifferentiated market coverage. On the other hand,
digital rural transformation optimizes the efficiency of resource
allocation and effectively responds to the challenges faced by
the traditional long-tail market ( CoelhoandMendes, 2019 ).
Under the familiar 80/20 regularity, the mainstream market
accounts for about 20 percent of products, while the remaining 80
percent constitutes the long-tail market of heterogeneous products
(IancuandTrichakis, 2014 ). As shown in Figure1, the Head
segment represents the mainstream market and the Tail segment
represents the long-tail market. In the long-tail market, the degree
of personalization is high and the matching success rate is low, so
the marginal search cost is high. As digital rural transformation
proceeds,ruralhouseholdsgaineasieraccesstovariousnichegoods
and services (Daumetal., 2021 ). These goods and services satisfy
theconsumerdemandtriggeredbythelong-taileffectofdiversified
and personalized demand. This attracts manufacturers to focus on
and actively develop niche markets, and mitigates market failures.
Only when the basic needs of rural households are satisfied will
they pursue differentiated and personalized needs. This shift is a
reflectionoftheprocessofupgradingtheconsumptionstructure.
Frontiersin SustainableFoodSystems 03 frontiersin.org

Zhan et al. 10.3389/fsufs.2026.1760055
2.3 A theoretical model of digital rural
transformation, financial assets and
consumption structure
Relative to Kongsamutetal. (2001) , the model introduces
three features that are specific to the digital rural setting. First,
the developmental and enjoyment consumption that a household
provides for itself is made a decreasing function of digital rural
transformation,amargintheoriginalframeworkdoesnotcontain.
Second, the relative price of developmental goods is endogenized
asadecreasingfunctionofdigitalruraltransformation,whichgives
the price effect of Section 2.1a formal counterpart. Third, financial
asset holdings are made an increasing function of digital rural
transformation and enter the intertemporal budget constraint; this
is what generates the asset channel and separates our mechanism
fromthetransaction-costchannelemphasizedinearlierwork.
In this model, we incorporate digital rural transformation
and financial assets to explore the intrinsic mechanisms by which
digital rural transformation promotes consumption upgrading.
In order to simplify the analysis while maintaining the general
applicability of the study, we divide consumer goods into two
main categories: survival and developmental enjoyment. Survival-
type consumer goods are necessary to maintain a basic life, and
they satisfy people’s most basic survival needs (JensenandMiller,
2008; Yangetal., 2022 ). Consumer goods for development
and enjoyment provide comfort, pleasure, and life-enhancing
fulfillment (AlbaandWilliams, 2013 ). The former represents a
lower level of consumption needs, while the latter belongs to a
higherlevelofconsumption.Whentheproportionofdevelopment
and enjoyment-type consumption increases, it indicates that the
consumptionlevelofruralhouseholdsisrising.
Assuming that there are many homogenized rural households
in the economy, each capable of perpetuating itself, and that
the preferences of the representative households are divisible in
time and have a logarithmic form of the utility function, the
instantaneous utility function and the additive discounted utility
function of the representative households can be expressed as
follows,respectively:
u
(
cst,cdt
)
= ϕln
(
cst − α)+ (1− ϕ)ln (cdt +St ) (1)
U =
∞∑
t=0
ρu
(
cst,cdt
)
(2)
In Equations1 and 2, cst and cdt denote the consumption of
subsistence and developmental enjoyment goods purchased by
the household in period t. ϕ is the weight of the household’s
preference,anditsvaluerangesfrom(0,1). αdenotestheminimum
consumption of subsistence goods required by the household in
order to maintain the basic life. St = St(Wt,DIGt) denotes the
developmental enjoyment consumption in periodt provided by
members of the household,1 which will be influenced by the
1 St represents the household’s developmental enjoyment consumption
which is the residual excluding market purchases(cdt ) and consists mainly of
leisureandrecreationthatrequiresthefamilytoparticipatetogether,suchas
socializationofintra-householdmembers,playingchessorcards,andsoon.
overall wealth Wt and the digital rural transformation DIGt
together. Wt includes the household’s wage income and financial
assets, and an increase in Wt boosts the demand of farm
households for developmental enjoyment consumption provided
by the market and suppresses the demand for developmental
enjoyment consumption provided by intra-household members.
The enhancement ofDIGt is conducive to the exposure of rural
households to the personalized goods and services provided by the
market, thus further promoting consumption upgrading. It can be
seen thatS′t (Wt) < 0,S′t (DIGt) < 0. ρ is the discount factor, the
value range is between (0,1). The budget constraint of the farmer
householdcanbeexpressedas:
Wt = wtlt +FINt = cst +ptcdt +FINt+1 (3)
FINt = FINt (rt,DIGt),FIN
′
t (DIGt) > 0 (4)
In Equations3, 4, wt denotes the wage level per unit of labor
time in the periodt.lt denotes the input labor time in the periodt.
FINt denotes the value of financial assets held by rural households
in period t, which is related to the interest rate levelrt and the
level of digital rural transformationDIGt in the current period.
Assuming that the rt is constant, the higher DIGt, the more
conducive to increase the financial market participation of rural
households, so thatFIN
′
t(DIGt) > 0. It should be stressed that
FIN
′
t(DIGt) > 0 enters the model as a reduced-form restriction
rather than as a result derived from preferences. It summarizes the
fact that digital rural transformation lowers the transaction and
information costs of participating in financial markets — search
costs,accountopening,andthecostofacquiringinformationabout
financialproducts—whichisthesamemechanismdocumentedin
Section 2.1. We adopt this restriction because it is what allows the
model to be solved in closed form, and we treat it as a hypothesis
to be tested rather than as an assumption to be maintained:
Section 4.4 estimates the corresponding empirical relationship
directly.Importantly,therestrictionconcernstheaccesstofinancial
markets, not the returns earned on the resulting asset position.
The model is therefore agnostic about whether that position
generates positive realized returns for every household, which is
precisely what the empirical mediation in Section 4.4 is designed
toestablish.
For the sake of simplicity, the market price of subsistence
consumer goods cst is normalized to 1, and the market price
of developmental enjoyment consumer goods pt is pt > 1.
As mentioned above, when the search cost exists, the market
equilibrium price of the product is p∗t = √sf (s, f are the
search cost and the utility loss cost, respectively), and the search
cost s is positively related to the market equilibrium price p∗t .
The higher the level of digital rural is, the lower the search cost
is, i.e., p′t(DIGt) < 0. Based on this, the maximization of the
consumption utility of the farmer household can be expressed
asfollows:
MaxU =
∑∞
t=0 ρu
(
cst,cdt
)
(5)
s.t. wtlt +FINt = cst +ptcdt +FINt+1 (6)
Frontiersin SustainableFoodSystems 04 frontiersin.org

Zhan et al. 10.3389/fsufs.2026.1760055
Employing the Lagrange function method and the asset lag
formula substitution solution, the equilibrium condition for the
maximization of consumption utility by the farmer household can
beobtainedas:
[ ϕ
1− ϕ
]
×
[
cdt +St(
cst − α)
]
×pt = FINt (7)
TakeSt(Wt,DIGt), pt(DIGt), FINt(rt,DIGt) intoEquation7,i t
canbeobtained:
[ ϕ
1− ϕ
]
×
[
cdt +St (Wt,DIGt)(
cst − α)
]
= FINt (rt,DIGt)pt (DIGt) (8)
Assuming that the wage ratewt and the interest ratert are
constant, the developmental enjoyment consumption expenditure
shareinperiod t isexpressedas:
cdt
(cst +cdt ) = 1[ ϕ1−ϕ × pt(DIGt)FINt(rt,DIGt) × (1+ St(DIGt)cdt )+ αcst +1]
= r (DIGt) (9)
Equation9 defines the expenditure share implicitly, because
themarket-purchasedcomponentofdevelopmentalandenjoyment
consumption appears on both sides of the expression. Holding
the wage rate and the interest rate constant and differentiating
implicitly with respect to digital rural transformation decomposes
the total effect into three parts: one operating through the relative
price,onethroughassetholdings,andonethroughhouseholdself-
provision. Equation8 determines the sign of each part. The share
is decreasing in the relative price, increasing in asset holdings, and
decreasing in self-provision; combined with the three restrictions
that the relative price falls, asset holdings rise, and self-provision
falls as digital rural transformation advances, all three parts are
positive. The three margins therefore reinforce rather than offset
one another, which establishes H1. The same decomposition
delivers H3 as a separate claim, because the asset term can be
isolated while the other two margins are left intact — the exercise
thatSection4.4performsempirically.
Building upon the preceding analytical findings, digital rural
transformation can causep′t(DIGt) < 0 through the price effect,
andcause S′t(DIGt) < 0throughthemarketscopeeffect.
Awordisneededonwhythemarketscopeeffectisrepresented
as a decline in household self-provision. Self-provision is the
developmental and enjoyment consumption that a household
producesforitself—home-producedgoods,informalservices,and
locallyobtainedsubstitutes.Whatlimitsitisnotwillingnessbutthe
range of goods and services that is physically and informationally
within reach: when the market boundary is narrow, a rural
household must meet many of its higher-order needs through
its own production. Digital rural transformation widens that
boundary by making niche, differentiated, and personalized goods
discoverable and deliverable, so that a larger share of these needs
is met by market purchases and a smaller share by self-provision.
Themarketscopeeffectthereforeentersthemodelasasubstitution
of market provision for household provision at a higher level of
variety,notasachangeinpreferences.
Meanwhile, digital rural transformation will increase the
participation of rural households in the financial market, so that
FIN
′
t(DIGt) > 0. In summary, it can be concluded thatr(DIGt) >
0, i.e. Digital rural transformation increases developmental
consumption expenditure shares, thereby driving structural
upgradinginruralhouseholdconsumption.Consequently,weposit
thefollowinghypothesis:
H1. Digital rural transformation boosts rural households’
consumptionstructureupgrading.
H2. Digital rural transformation promotes the holding of
financial assets by rural households, and heterogeneity in holdings
mayexistacrossfinancialassettypes.
H3. Digital rural transformation promotes rural
households’ consumption structure upgrading through financial
assetaccumulation.
3 Econometric modeling and data
description
3.1 Model setting
Toinvestigatetheeffectofdigitalruraltransformationonrural
households’consumptionstructureupgrading,weadoptatwo-way
fixedeffectsmodelbasedonpaneldata:
consumpipt = β1 + β2digpt + β3Controlipt + δi + μt + εipt (10)
In Equation10, we focus on several key variables. The
explained variable consumpipt denotes the proportion of the
ith farmer household’s developmental enjoyment consumption
in total consumption in province p,y e a rt. This proportion
reflects the level of rural households’ consumption structure
upgrading. The core explanatory variable isdigpt, which measures
the level of development of digital rural transformation in
the same province and year. The control variables Controlipt
encompass three dimensions: (1) the household head
characteristics, (2) household socioeconomic attributes, and
(3) provincial-level contextual factors. The model incorporates
δi as household fixed effects and μt as year fixed effects, and
εipt is a random error term. Our analysis focuses primarily
on coefficient β2, which quantifies the magnitude of digital
rural transformation’s effect on rural households’ consumption
structureupgrading.
3.2 Description of main variables
3.2.1 Explained variable: rural households’
consumption structure upgrading
We measure consumption upgrading (consump) as the share
of developmental and enjoyment-oriented expenditure in total
household consumption. Three approaches are commonly used
Frontiersin SustainableFoodSystems 05 frontiersin.org

Zhan et al. 10.3389/fsufs.2026.1760055
to operationalize household consumption upgrading: the Engel
coefficient, the expenditure share of development and enjoyment-
oriented items, and composite indices constructed from several
categories. We follow the second. It has been adopted in
recent work on Chinese rural households — most directly by
Yuetal. (2021) , who study rural consumption structure with
the same CHFS data and classify consumption into subsistence
vs. development and enjoyment items, drawing on the need-
hierarchy logic formalized by JensenandMiller (2008) and
AlbaandWilliams (2013) . We prefer it to the Engel coefficient
for two reasons. First, the Engel coefficient is sensitive to relative
food prices and to the imputation of home-produced food,
both of which vary systematically across provinces and over
time in rural China, whereas the development/enjoyment share
is far less exposed to these measurement problems. Second,
and more importantly for our purposes, the expenditure share
is the theoretical object of interest in the non-homothetic
framework of Section 2.3: Equation9 is an expression for
exactly this share, so using it keeps the measurement and the
modelaligned.
3.2.2 Core explanatory variable: digital rural
transformation
There is no uniform standard on how to measure the
development level of digital rural transformation. Some scholars
use a single indicator, while others construct a multi-dimensional
indicator system in order to obtain a more comprehensive
assessment; a multi-dimensional system is appropriate here
because digital rural transformation is not a single technology
but a bundle of complementary components that must be
present together for digital services to reach a rural household.
Following Ghobakhloo (2020) and Hatukaetal. (2021) ,w e
evaluate digital rural transformation along four dimensions,
which are distinguished not by sector but by the function
each performs in the chain that connects digital technology
to rural household outcomes: (1) digital infrastructure and
human-capital inputs, (2) digital industry and markets, (3)
digital life and household adoption, and (4) digital service
coverageandlast-mileaccessibility.Thespecificassessmentindexes
are shown in Table1. The digital rural transformation index
(dig) is computed using the entropy method ( LiuandHuang,
2025).
3.2.3 Control variables
To mitigate potential omitted-variable bias, we include a
comprehensivesetofcontrolvariablesthatmayaffectconsumption
upgradingatthreedimensions.
Individual level of household head: referring to the life-
cycle hypothesis, we include the age of the household head
(age) and its square term (age2) as control variables. We then
also control for the household head’s educational attainment
(educ), measured in years of schooling, to account for its
well-documented influence on household income. Furthermore,
we control for Communist Party membership ( parmen)a sa
TABLE 1 Digitalrural transformationevaluationindex system.
Categories Indicator Attributes
Digitalinfrastructure
andhuman-capital
inputs
Percentageofruralbroadband
accesshouseholds
+
Mobilephoneownershipper
100ruralhouseholds
+
Computersper100rural
households
+
Expenditureoneducation +
Digitalindustryand
markets
PercentageofTaobaovillages
amongalladministrative
villages
+
Ruraldigitalfinancial
inclusionindex
+
Shareofe-commercesalesin
tertiaryvalueadded
+
Digitallifeand
householdadoption
Ratiooftransportand
communicationcoststototal
consumptionexpenditureof
agriculturalhouseholds
+
Percentageofconsumer
spendingondigitalproducts
purchasedbyfarmer
households
+
Digitalservicecoverage
andlast-mile
accessibility
Lengthofruraldeliveryroutes +
Averagepopulationservedby
ruralpostalshops
–
binary variable, given its potential influence on consumption
behavior and preferences, with communist party member as 1
and non-communist party member as 0. Marital status may
also affect the consumption habit, so marital status (marr)i s
set as a control variable, with married as 1 and unmarried
as 0. Current physical condition may affect consumption
demand, so physical condition ( phycon) is set as a control
variable, with very good physical condition as 3, fair as 2, and
badas1.
Household level: the structure of household members,
especially the proportion of children and the elderly, has a
significant impact on consumption patterns, so we have calculated
the household dependency ratio (depend) as a control variable.
The household’s wage income, which is a determinant of the
consumption level, is included as a control variable after taking its
logarithmicform( lninc).
Provinciallevel:toaccountforprovincial-levelheterogeneityin
economic development, we control for the logarithm of provincial
GDP per capita (lnpgdp). Table2 presents the descriptive statistics
forallkeyvariables.
Frontiersin SustainableFoodSystems 06 frontiersin.org

Zhan et al. 10.3389/fsufs.2026.1760055
TABLE 2 Descriptivestatistics.
Variables Variable meaning Obs. Mean Std Min Max
consump Ruralhouseholds’consumptionstructureupgrading 20,391 0.539 0.227 0 1
dig Digitalruraltransformationindex 20,391 0.355 0.114 0.154 0.652
age A geofheadofhousehold 20,391 64.71 12.16 12 103
age2 Squareofageofheadofhousehold 20,391 43.36 15.74 1.440 106.1
educ Educationalattainment 20,391 4.867 5.064 0 19
parmen Membershipofthecommunistpartyornot 20,391 0.0960 0.295 0 1
marr Marriedornot 20,391 0.888 0.316 0 1
phycon Physicalcondition 20,391 2.105 0.778 1 3
depend Householddependencyratio 20,391 0.535 0.366 0 1
lninc Logarithmofhouseholdwageincome 20,391 1.941 2.878 0 13.65
lnpgdp LogarithmofprovincialGDPpercapita 20,391 10.90 0.361 10.17 12.01
3.3 Data sources
The micro data are mainly derived from the China Household
Finance Survey (CHFS), which contains data on personal
information of household heads, household consumption and
income.Themacroeconomicindicatorsareprimarilysourcedfrom
two authoritative datasets: (1) the China Statistical Yearbook, and
(2) the Digital Financial Inclusion Index compiled by Peking
University’sDigitalFinanceResearchCenter.Wehaveselecteddata
from 2015, 2017 and 2019 for integration, and after a rigorous
samplescreeninganddatacleaningprocess,wefinallyobtain6,797
householdswithatotalof20,391validobservations.
4 Empirical findings and discussion
4.1 Benchmark regression results
Table3 reports the baseline regression results examining
the impact of digital rural transformation on rural household
consumption structure upgrading, presenting four progressively
controlled specifications: (1) no control variables, (2) household
head characteristics only, (3) additional household-level controls,
and (4) full controls including provincial-level factors, with all
models incorporating both year and household fixed effects. The
regression results indicate that the coefficientβ2 for digital rural
transformation is consistently positive and statistically significant
at the 1% level across all model specifications, demonstrating
that digital rural initiatives significantly increase the share of
developmental and enjoyment-oriented consumption. Therefore,
H1isvalidated.
The Column (4) results inTable3 demonstrate that age, age
squared, marriage, and years of education do not have a significant
effect on rural households’ consumption structure upgrading.
However, the negative and statistically significant coefficient for
Communist Party membership (p < 0.05) suggests that party
member households allocate a smaller share of expenditures
TABLE 3 Benchmark regression results.
Variables consump
(1) (2) (3) (4)
dig 0.780∗∗∗ 0.763∗∗∗ 0.738∗∗∗ 0.550∗∗∗
(0.038) (0.038) (0.040) (0.063)
age −0.000 −0.000 −0.000
(0.002) (0.002) (0.002)
age2 −0.001 −0.000 −0.000
(0.002) (0.002) (0.002)
educ 0.001∗ 0.001∗ 0.001
(0.001) (0.001) (0.001)
parmen −0.019∗∗ −0.019∗∗ −0.020∗∗
(0.009) (0.009) (0.009)
marr 0.009 0.007 0.007
(0.010) (0.010) (0.010)
phycon −0.018∗∗∗ −0.018∗∗∗ −0.018∗∗∗
(0.003) (0.003) (0.003)
depend −0.035∗∗∗ −0.039∗∗∗
(0.012) (0.012)
lninc 0.008∗∗∗ 0.008∗∗∗
(0.001) (0.001)
lnpgdp 0.072∗∗∗
(0.017)
Constant 0.262∗∗∗ 0.326∗∗∗ 0.269∗∗∗ −0.441∗∗
(0.013) (0.067) (0.067) (0.182)
YearFE YES YES YES YES
HouseholdFE YES YES YES YES
Obs. 20,391 20,391 20,391 20,391
R2 0.033 0.037 0.043 0.044
∗, ∗∗, ∗∗∗ represent 10%, 5% and 1% significance levels, respectively. Robust standard errors
areinparentheses.Thesamemeaningsapplytothetablebelowifnototherwisespecified.
Frontiersin SustainableFoodSystems 07 frontiersin.org

Zhan et al. 10.3389/fsufs.2026.1760055
TABLE 4 Regressionresults after replacing variables.
Variables consump Total development and enjoyment consumption
(1) (2) (3) (4) (5) (6) (7) (8)
L.dig 0.731∗∗∗ 0.716∗∗∗ 0.695∗∗∗ 0.550∗∗∗
(0.034) (0.034) (0.036) (0.056)
dig 5.459∗∗∗ 5.319∗∗∗ 5.430∗∗∗ 3.974∗∗∗
(0.179) (0.179) (0.189) (0.299)
Constant 0.298∗∗∗ 0.365∗∗∗ 0.310∗∗∗ −0.307∗ 7.618∗∗∗ 7.417∗∗∗ 6.855∗∗∗ 1.359
(0.011) (0.066) (0.067) (0.182) (0.064) (0.323) (0.313) (0.872)
Householdheadcontrol NO YES YES YES NO YES YES YES
Householdcontrol NO NO YES YES NO NO YES YES
ProvinceFE NO NO NO YES NO NO NO YES
YearFE YES YES YES YES YES YES YES YES
HouseholdFE YES YES YES YES YES YES YES YES
Obs. 20,391 20,391 20,391 20,391 20,391 20,391 20,391 20,391
R2 0.035 0.039 0.045 0.046 0.074 0.088 0.124 0.128
to developmental and enjoyment consumption compared to
non-member households. This pattern may reflect distinctive
consumption preferences or value orientations among party-
affiliatedruralhouseholds.Inaddition,thehealthstatuscoefficient
is negative and statistically significant at the 1% level, suggesting
that less healthy rural households exhibit higher proportions of
developmental and enjoyment consumption. This pattern may
reflect increased healthcare expenditures required to maintain
basic wellbeing, potentially crowding out other consumption
categories.Meanwhile,thecoefficientofthehouseholddependency
ratio is also negative and significant at the 1% level, indicating
that as the proportion of elderly people and children in the
household increases, the share of rural households’ demand
for subsistence goods rises accordingly, which reflects the
fact that farming households are more pragmatic in their
consumption structure. Finally, the results show statistically
significant positive coefficients (p < 0.01) for both household
wage income and provincial economic development level. This
suggests that economically advantaged households in more
developed provinces demonstrate greater propensity for higher-
level consumption, thereby facilitating consumption structure
optimizationandupgrading.
4.2 Robustness testing
4.2.1 Replacement of core explanatory variable
and explained variable
To assess robustness, we employ alternative variable
specifications: (1) a one-period lagged digital rural transformation
index (L.dig) as the core explanatory variable, and (2) total
developmental and enjoyment consumption expenditures as
the consumption upgrading measure. As shown inTable4,t h e
regression results confirm our baseline findings, with the positive
effect of digital rural transformation remaining statistically
significant. These results provide further empirical support for
Hypothesis1.
4.2.2 Removing extreme values and municipalities
Considering that municipalities may be subject to policy
favoritism, this may affect the robustness of the regression results.
Therefore, we decide to exclude municipalities from the sample
and re-run the regression analysis. Additionally, to exclude the
possible interference of outliers, we trim the explained variable at
the 1st and 99th percentiles.Table5 presents the regression results
after these adjustments. The estimated coefficients for digital rural
transformation (dig) demonstrate consistent statistical significance
and magnitude across model specifications, including robustness
checksthatexcludebothoutliersandmunicipal-levelobservations.
4.2.3 Panel quantile model
In the baseline regression analysis, the conditional expectation
canonlyreflectthecentralizedtrendoftheconditionaldistribution
and cannot exhaustively describe the effect of the explanatory
variablesontheconditionaldistributionoftheexplainedvariables.
Additionally, traditional regression methods based on minimizing
the sumof squared residuals are susceptibleto the phenomenon of
extremevaluesandsignificantinterferencefromheteroskedasticity.
Therefore, we have constructed a panel quantile regression model
to capture more comprehensively the impact of explanatory
variablesonthedistributionoftheexplainedvariables.
consumpipt = θ1 + θ2digpt + θ3Controlipt + σipt (11)
Frontiersin SustainableFoodSystems 08 frontiersin.org

Zhan et al. 10.3389/fsufs.2026.1760055
TABLE 5 Regressionresults after removing outliersand municipalities.
Variables consump
(1) (2) (3) (4)
dig 0.753∗∗∗ 0.735∗∗∗ 0.708∗∗∗ 0.508∗∗∗
(0.038) (0.038) (0.041) (0.064)
Constant 0.268∗∗∗ 0.324∗∗∗ 0.267∗∗∗ −0.510∗∗∗
(0.014) (0.068) (0.069) (0.191)
Householdhead
control
NO YES YES YES
Householdcontrol NO NO YES YES
ProvinceFE NO NO NO YES
YearFE YES YES YES YES
HouseholdFE YES YES YES YES
Obs. 18,792 18,792 18,792 18,792
R2 0.033 0.037 0.043 0.044
In Equation11, θ1, θ2, andθ3 represent estimable parameters,
while σipt denotes the stochastic error term. Referring to
BudigandHodges (2010) , we estimate the explained variables
sequentially from the 10% quantile to the 90% quantile. Consistent
with the previous section, Controlipt contains exactly the same
variables as the benchmark regression. The results inTable6 show
that the digital rural transformation index is significant at all
quantiles. Specifically, the maximum impact is reached at the 10%
quantile, which is 0.672. As the quantile increases, the impact on
consumption structure upgrading gradually diminishes, implying
that the digital rural transformation exerts a more pronounced
impact on the upgrading of rural households’ consumption
structure at low consumption levels, while the impact on the
residents at high consumption levels is smaller, which may be
attributed to the fact that the latter’s consumption level is already
high, and the marginal effect of the digital rural transformation is
relatively low. The results of the panel quantile regression further
validate the robustness of the benchmark regression, and H1 is
reconfirmed. They also carry a distributional implication: because
the estimated effect is largest at the bottom of the consumption
distribution and declines monotonically toward the top, digital
rural transformation raises the consumption share of the least-
advantaged rural households by more than it raises that of
householdsalreadyatthetop.
4.3 Endogeneity test
There are two possible endogeneity problems with the
benchmark regression model. First, at the household level, farm
households with more advanced consumption structures may also
be better placed to adopt digital services, so that the direction of
causality is not identified. Second, at the provincial level, rural
households’ consumption structure upgrading is not only affected
by digital rural transformation, but may also be disturbed by other
macro-factors,whichmayleadtoinaccuratebenchmarkregression
estimates. In order to solve the endogeneity problem, suitable
instrumentalvariablesneedtobefound.
Given that the digital rural transformation index is measured
at the provincial level, we follow Lietal. (2022) in employing
“the spherical distance from provincial capitals to Hangzhou”
(hereinafter called “spherical distance”) as our instrumental
variable (dis_tel) for digital rural transformation. First, as the
carrier of digital rural transformation, the popularity of the
Internet is affected by geographical location, and it is more
difficult to popularize the Internet in areas far away from
Hangzhou. Second, “spherical distance” is related to digital rural
transformation, but does not directly affect the consumption of
rural households. In addition, referring toLiuetal. (2023) ,t h e
“post and telecommunication data of provincial capital cities in
1984” (hereinafter called “post and telecommunication data”) is
used as another instrumental variable (post_net) for digital rural
transformation. On one hand, it is because the laying of fixed
telephone lines is mainly done by post and telecommunications,
which directly affects the construction of Internet infrastructure.
Meanwhile,thehigherthenumberofpostandtelecommunication,
the higher the local communication demand. On the other hand,
thedataonpostandtelecommunicationsarehistoricaldata,which
have little impact on rural households’ consumption structure
upgrading. These two instrumental variables satisfy exogeneity
and correlation, and meet the basic conditions for instrumental
variableselection.
Since both “spherical distance” and “postal data” are
cross-sectional data, these variables cannot serve as valid
instrumental variables in panel data analyses. Therefore, referring
to NunnandQian (2014), the interaction term between “spherical
distance” and the total telecom business of each province in the
previous year is set as the first instrumental variable (dis_tel).
The interaction term between “postal data” and the number of
rural Internet access users in each province in the previous year
is used as the second instrumental variable (post_net). Table7
reports the results of the second-stage regressions. Results using
dis_tel as the instrumental variable are presented in columns (1)
and (2), while columns (3) and (4) report estimates employing
post_net as the alternative instrument. Column (5) is the result of
using bothdis_tel and post_net as instrumental variables. All these
results show that digital rural transformation significantly boosts
rural households’ consumption structure upgrading. Meanwhile,
the F statistic of all regression models is greater than the critical
value at the 10% level of the weak identification test, indicating
that the models do not suffer from weak instrumental variables
and non-identifiability. It further supports the rationality of
instrumentalvariablesselection.
4.4 Mechanism analysis
Different types of financial assets have different risk and return
characteristics. Therefore, we categorize financial assets into risk-
free and risky types. Risk-free financial assets mainly include low-
risk and low-yield investments such as demand deposits, time
deposits and gold. Risky financial assets, in contrast, include
high-risk, high-yield investments such as stocks, funds, bonds
Frontiersin SustainableFoodSystems 09 frontiersin.org

Zhan et al. 10.3389/fsufs.2026.1760055
TABLE 6 Quantileregression results.
Variables consump
10% 20% 30% 40% 50% 60% 70% 80% 90%
dig 0.672∗∗∗ 0.655∗∗∗ 0.622∗∗∗ 0.590∗∗∗ 0.552∗∗∗ 0.510∗∗∗ 0.477∗∗∗ 0.440∗∗∗ 0.426∗∗∗
(0.110) (0.100) (0.082) (0.069) (0.061) (0.066) (0.079) (0.098) (0.107)
Obs. 20,391 20,391 20,391 20,391 20,391 20,391 20,391 20,391 20,391
TABLE 7 Endogeneitytest.
Variables IV=dis_tel IV=post_net IV=dis_telpost_net
(1) (2) (3) (4) (5)
dig 0.904∗∗∗ 0.942∗∗∗ 1.080∗∗∗ 1.740∗∗∗ 1.587∗∗∗
(0.066) (0.237) (0.048) (0.167) (0.161)
Controls NO YES NO YES YES
YearFE YES YES YES YES YES
HouseholdFE YES YES YES YES YES
Obs. 20,391 20,391 20,391 20,391 20,391
Fstatistic 5660.275 909.395 1.7e+04 2025.600 1103.573
R2 0.033 0.041 0.028 0.016 0.022
and futures. To measure the financial asset position of rural
households and the income it generates, we use the property
income of rural households ( prop_inc) as a measure, which
covers the gains and losses of all financial assets. Meanwhile, we
use a dummy variable to indicate whether the farm household
holds financial assets or not. If holding is 1 and not holding
is 0. Risky financial assets are denoted byrisk_fin and risk-free
financialassetsaredenotedby rifre_fin.Buildinguponthebaseline
specification, we further use a mediation effect model to test
the impact of digital rural transformation on property income of
farmhouseholds.
prop_incipt = β1 + β2digpt + β3Controlipt + δi + μi + εipt (12)
consumpipt = β1 + β2digpt + β3prop_incipt + β4Controlipt

  • δi + μi + εipt (13)
    In Equations12 and 13, prop_incipt is the property income
    of the farm household. We use Probit model to test whether
    the rural households hold financial assets. Table8 reveals
    that digital rural transformation significantly increases the
    probability of holding financial assets, and the effect on risky
    financial assets is more significant (column (1) and (2)). Column
    (3) further demonstrates that digital rural transformation
    significantly enhances farm households’ property income.
    In summary, H2 is validated. Column (4) demonstrates
    that both digital rural transformation and rural households’
    property income significantly enhance consumption structure
    upgrading. Read together with columns (1)–(3), this indicates
    that digital rural transformation improves rural household
    wellbeing by first enlarging the household’s financial asset
    TABLE 8 Mechanism analysis results.
    Variables risk_fin rifre_fin prop_inc consump
    (1) (2) (3) (4)
    dig 1.094∗∗∗ 0.325∗∗∗ 25.374∗∗∗ 0.458∗∗∗
    (0.203) (0.091) (0.718) (0.065)
    prop_inc 0.004∗∗∗
    (0.001)
    Constant −4.749∗∗∗ −2.025∗∗∗ −23.607∗∗∗ −0.356∗
    (0.815) (0.358) (2.127) (0.182)
    Controls YES YES YES YES
    YearFE YES YES YES YES
    HouseholdFE YES YES YES YES
    Obs. 20,391 20,391 20,391 20,391
    R2 0.279 0.046
    position and then raising the income that position generates. H3
    isverified.
    4.5 Heterogeneity analysis
    To explore more precisely how digital rural transformation
    affects rural households’ consumption structure upgrading,
    we conduct a heterogeneity analysis based on the personal
    characteristics of the household head, the characteristics of the
    family,andthegeographiclocationofthehousehold.
    Frontiersin SustainableFoodSystems 10 frontiersin.org

Zhan et al. 10.3389/fsufs.2026.1760055
TABLE 9 Heterogeneityanalysis of household head characteristics.
Variables Age >60 Age ≤60 Years of education>6 Years of education≤6 Married Unmarried
(1) (2) (3) (4) (5) (6)
dig 0.467∗∗∗ 0.812∗∗∗ 0.772∗∗∗ 0.277∗∗∗ 0.613∗∗∗ 0.592∗∗
(0.085) (0.098) (0.088) (0.103) (0.066) (0.281)
Constant −1.801∗∗ 0.074 −0.403 −0.855∗∗∗ −0.324 −0.858
(0.714) (0.421) (0.285) (0.302) (0.198) (0.775)
Controls YES YES YES YES YES YES
YearFE YES YES YES YES YES YES
HouseholdFE YES YES YES YES YES YES
Obs. 10,101 10,290 10,055 10,336 18,104 2,287
R2 0.042 0.062 0.067 0.032 0.048 0.040
4.5.1 Household head characteristics
In this section, we analyze heterogeneity across farm
households with respect to their age, years of education and
marriage. Based on descriptive statistics, age is categorized into
two groups: above 60 and 60 and below. Years of education
is categorized into two groups: above 6 years and 6 years and
below. Marriage is categorized into two groups: married and
unmarried.Table9 presentsheterogeneityanalysisresultsbasedon
householdheadcharacteristics.Incolumns(1)and(2),digitalrural
transformation shows statistically significant positive effects across
bothagecohorts,withstrongerestimatedimpactsobservedamong
households headed by individuals below 60 years old, suggesting
that digital rural transformation exerts a more pronounced impact
onconsumptionupgradingamonghouseholdswithyoungerheads.
In columns (3) and (4), digital rural transformation contributes
more strongly to consumption upgrading among rural households
with more than 6 years of education. In columns (5) and (6),
digital rural transformation demonstrates marginally stronger
effects on consumption structure upgrading among married rural
householdscomparedtounmarriedhouseholds.
4.5.2 Family characteristics
We analyze the heterogeneity of farm households with
respect to their indebtedness and wage income. First, households
with loans may be more conservative in consumption due to
loan repayment pressure. Second, we analyze whether digital
rural transformation can simultaneously promote consumption
upgrading among farm households with different income levels.
Farm households are categorized into two groups, high-income
and low-income, according to the average wage income. Columns
(1) and (2) of Table10 show that digital rural transformation
promotes consumption upgrading among farm households
regardless of whether they have loans or not. However, the
coefficients are larger for households with loans, probably
because their original consumption level is lower and digital rural
transformation has stimulated consumption demand. Columns
TABLE 10 Heterogeneityanalysis ofhousehold characteristics.
Variables With
loan
Without
loan
High-
income
Low-
income
(1) (2) (3) (4)
dig 0.894∗∗∗ 0.532∗∗∗ 0.995∗∗∗ 0.399∗∗∗
(0.156) (0.080) (0.137) (0.083)
Constant 0.519 −0.530∗∗ 0.034 −0.459∗
(0.422) (0.240) (0.419) (0.244)
Controls YES YES YES YES
YearFE YES YES YES YES
HouseholdFE YES YES YES YES
Obs. 5,447 14,944 6,322 14,069
R2 0.052 0.044 0.075 0.032
(3) and (4) reveal that the promotion effect of digital rural
transformation on high-income farm households is significantly
higher than that of low-income farm households, indicating
that digital rural transformation demonstrates greater efficacy
in promoting consumption upgrading among high-income
farmhouseholds.
4.5.3 Geographic location
Rural households in the eastern region exhibit more advanced
consumption structures compared to their central and western
counterparts, attributable to superior regional economic
development. By analyzing the heterogeneity of geographic
locations, this study aims to investigate whether digital rural
transformationcontributestotheconvergenceoftheconsumption
structure of farm households in different regions. Following the
official regional delineation established by China’s National Bureau
of Statistics in 2021, we classify ten provincial-level administrative
units as the eastern region. These include Beijing, Tianjin, Hebei,
Frontiersin SustainableFoodSystems 11 frontiersin.org

Zhan et al. 10.3389/fsufs.2026.1760055
TABLE 11 Heterogeneityanalysis of geographiclocation.
Variables Eastern region Central and Western
region
(1) (2)
dig 0.427∗∗∗ 0.715∗∗∗
(0.104) (0.081)
Constant −0.514 −0.273
(0.401) (0.208)
Controls YES YES
YearFE YES YES
HouseholdFE YES YES
Obs. 6,555 13,836
R2 0.038 0.050
Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, and
Hainan. The remaining provinces (cities) are defined as the central
and western regions. Table11 presents the regression results,
showing that digital rural transformation significantly enhances
rural consumption structure upgrading (p < 0.01). The stronger
effectsobservedincentralandwesternregionscomparedtoeastern
areasindicateitsroleinreducingregionalconsumptiondisparities.
5 Conclusions and implications
Digital rural transformation reshapes rural livelihoods as
well as rural markets. By improving rural households’ access
to financial services, markets and investment opportunities, it
enables them to accumulate financial assets; we read the resulting
change in the composition of household consumption as the
observable dimension of rural household wellbeing. This study
amalgamates the 2015, 2017 and 2019 China Household Finance
Survey data with provincial level data. By applying panel fixed-
effects models, instrumental variables methods, Probit estimations
and mediation effect models, it investigates how digital rural
transformation affects the consumption structure upgrading of
rural households. The key findings are as follows. First, digital
rural transformation has a remarkable positive impact on the
proportion of development and enjoyment-oriented consumption
of rural households. This conclusion remains stable after a
battery of robustness checks. Moreover, as this share grows,
the boosting effect of digital rural transformation on farm
household consumption upgrading is mitigated. Second, and most
importantly for the interpretation advanced here, digital rural
transformation improves rural households’ access to financial
markets and significantly increases both their financial asset
holdings and the income those assets generate. The mediation
analysis shows that this accumulation is one of the channels
through which digital rural transformation raises rural household
wellbeing. Third, the effects are unevenly distributed, and the
pattern runs against the usual digital-divide expectation: the gains
are larger in the central and western provinces than in the eastern
region, and larger at the bottom of the consumption distribution
than at the top, so that less-developed regions and households
with initially low consumption shares benefit disproportionately.
Regarding household head characteristics, rural households with
heads under 60, with over 6 years of education, or who are
married also benefit more from digital rural initiatives. In terms
of household nature, loan - holding rural households experience
a more pronounced positive effect on consumption upgrading
due to digital rural transformation, and this effect varies across
different income level households. Geographically, central and
western regions’ rural households gain greater upgrading benefits
from digital rural transformation, which aids in reducing the
consumption disparity between eastern and central and western
ruralhouseholds.
The following policy recommendations are proposed. First, it
is necessary to boost investment in Internet infrastructure so as to
gradually elevate Internet penetration in rural areas. Meanwhile,
the application of 5G technology and artificial intelligence in rural
regions should be encouraged. This approach will not only help
lay a solid hardware foundation for the consumption upgrading
of rural households, but also enable rural areas to share the
benefits of digital development. Second, because financial asset
accumulation is the livelihood mechanism identified here, policy
should target the access margin rather than consumption itself.
Digital rural transformation lowers the barriers rural households
face to participate in the financial market; the financial system
shouldthereforebefurtheroptimizedtopromotefinancialservices
to the grassroots and rural areas. Meanwhile, the quality of
financial services should be upgraded, and financial intelligent
services, such as remote and door-to-door services, should be
innovated. This not only connects financial markets with rural
households, but also advances the deep integration of digital rural
transformation and financial products. Finally, a differentiated
digital rural transformation strategy should be formulated by
comprehensively considering the personal traits, family conditions
and regional characteristics of rural households. For married
rural households who are young and highly educated, providing
more digital technology training and financial literacy will
help them make full use of the opportunities of digital rural
transformation. More flexible financial service programs should
be provided for households with loans. It should also increase
investment in the central and western regions and strengthen
infrastructure development to help narrow regional development
differences and promote overall rural consumption upgrading and
sustainabledevelopment.
This study acknowledges certain limitations, particularly
regarding data constraints that currently restrict our digital rural
transformation index to provincial-level measurement. Future
research should prioritize developing prefecture-level indicators
to enhance measurement precision. In addition, this study
mainly focuses on digital rural transformation to promote rural
households’ consumption structure upgrading by influencing their
own behaviors, while other external pathways such as health
services and social welfare are less explored. Future research will
further enrich and improve the theoretical framework of digital
rural transformation influencing rural households’ consumption
Frontiersin SustainableFoodSystems 12 frontiersin.org

Zhan et al. 10.3389/fsufs.2026.1760055
structureupgradingfromotherperspectives,aimingtoobtainmore
comprehensiveresearchresults.
Data availability statement
The raw data supporting the conclusions of this article will be
madeavailablebytheauthors,withoutunduereservation.
Author contributions
JZ: Conceptualization, Funding acquisition, Resources,
Writing – review & editing, Project administration. JJ: Data
curation, Writing – original draft, Formal analysis, Methodology,
Investigation, Conceptualization. BL: Supervision, Visualization,
Writing–review&editing,Validation.
Funding
The author(s) declared that financial support was received for
this work and/or its publication. This research was funded by the
National Social Science Fund of China (NSSFC), grant number
26BJY200.
Conflict of interest
The author(s) declared that this work was conducted in the
absence of any commercial or financial relationships that could be
construedasapotentialconflictofinterest.
Generative AI statement
The author(s) declared that Generative AI was not used in the
creationofthismanuscript.
Any alternative text (alt text) provided alongside figures in
this article has been generated by Frontiers with the support of
artificial intelligence and reasonable efforts have been made to
ensureaccuracy,includingreviewbytheauthorswhereverpossible.
Ifyouidentifyanyissues,pleasecontactus.
Publisher’s note
All claims expressed in this article are solely those of the
authors and do not necessarily represent those of their affiliated
organizations, or those of the publisher, the editors and the
reviewers. Any product that may be evaluated in this article, or
claim that may be made by its manufacturer, is not guaranteed or
endorsedbythepublisher.
References
Alba, J. W., and Williams, E. F. (2013). Pleasure principles: a review of research on
hedonicconsumption. J.Consum.Psychol .23,2–18.doi:10.1016/j.jcps.2012.07.003
AnkrahTwumasi,M.,Jiang,Y.,Asante,D.,Addai,B.,Akuamoah-Boateng,S.,Fosu,P.,
et al. (2021). Internet use and farm households food and nutrition security nexus: the
caseofruralGhana. Technol.Soc.65:101592.doi:10.1016/j.techsoc.2021.101592
Bakker, L., Sok, J., van der Werf, W., and Bianchi, F. J. J. A. (2021). Kicking the
habit: what makes and breaks farmers’ intentions to reduce pesticide use?Ecol. Econ.
180:106868.doi:10.1016/j.ecolecon.2020.106868
Bakos, J. Y. (1997). Reducing buyer search costs: implications for electronic
marketplaces.Manag. Sci.43,1676–1692.doi:10.1287/mnsc.43.12.1676
Bhagwati, J., Panagariya, A., and Srinivasan, T. N. (2004). The muddles over
outsourcing.J.Econ.Perspect .18,93–114.doi:10.1257/0895330042632753
Blankespoor, E., Miller, G. S., and White, H. D. (2014). The role of dissemination in
market liquidity: evidence from firms’ use of TwitterTM.Account. Rev. 89, 79–112.
doi:10.2308/accr-50576
Budig, M. J., and Hodges, M. J. (2010). Differences in disadvantage: variation in the
motherhood penalty across white women’s earnings distribution.Am. Sociol. Rev. 75,
705–728.doi:10.1177/0003122410381593
Cavallo, A. (2017). Are online and offline prices similar? Evidence from large multi-
channelretailers. Am.Econ.Rev .107,283–303.doi:10.1257/aer.20160542
Choi,T.Y.,andKrause,D.R.(2006).Thesupplybaseanditscomplexity:implications
for transaction costs, risks, responsiveness, and innovation. J. Oper. Manag. 24,
637–652.doi:10.1016/j.jom.2005.07.002
Coelho, M. P., and Mendes, J. Z. (2019). Digital music and the “death of the long tail.”
J.Bus.Res .101,454–460.doi:10.1016/j.jbusres.2019.01.015
Daum, T., Villalba, R., Anidi, O., Mayienga, S. M., Gupta, S., Birner, R., et al. (2021).
Uberfortractors?OpportunitiesandchallengesofdigitaltoolsfortractorhireinIndia
andNigeria. WorldDev.144:105480.doi:10.1016/j.worlddev.2021.105480
Deichmann, U., Goyal, A., and Mishra, D. (2016). Will digital technologies transform
agricultureindevelopingcountries? Agric. Econ.47,21–33.doi:10.1111/agec.12300
Eidse, N., Turner, S., and Oswin, N. (2016). Contesting street spaces in a socialist city:
itinerantvending-scapesandtheeverydaypoliticsofmobilityinHanoi,Vietnam. Ann.
Am.Assoc.Geogr .106,340–349.doi:10.1080/00045608.2015.1117936
Fan, J., Tang, L., Zhu, W., and Zou, B. (2018). The alibaba effect: spatial consumption
inequality and the welfare gains from e-commerce. J. Int. Econ. 114, 203–220.
doi:10.1016/j.jinteco.2018.07.002
Galor, O., and Tsiddon, D. (1997). Technological progress, mobility, and economic
growth.Am.Econ.Rev .87,363–382.
Geyskens, I., Steenkamp, J-. B. E. M., and Kumar, N. (2006). Make, buy, or
ally: a transaction cost theory meta-analysis. Acad. Manage. J . 49, 519–543.
doi:10.5465/amj.2006.21794670
Ghobakhloo,M.(2020).Industry4.0,digitization,andopportunitiesforsustainability.
J.Clean.Prod .252:119869.doi:10.1016/j.jclepro.2019.119869
Goldfarb, A., and Tucker, C. (2019). Digital economics. J. Econ. Lit. 57, 3–43.
doi:10.1257/jel.20171452
Hampshire,K.,Mwase-Vuma,T.,Alemu,K.,Abane,A.,Munthali,A.,Awoke,T.,etal.
(2021). Informal mhealth at scale in Africa: opportunities and challenges.World Dev.
140:105257.doi:10.1016/j.worlddev.2020.105257
Hatuka, T., Zur, H., and Mendoza, J. A. (2021). The urban digital lifestyle: an
analyticalframeworkforplacingdigitalpracticesinaspatialcontextandfordeveloping
applicablepolicy. Cities111:102978.doi:10.1016/j.cities.2020.102978
Heeks,R.,Gomez-Morantes,J.E.,Graham,M.,Howson,K.,Mungai,P.,Nicholson,B.,
etal.(2021).Digitalplatformsandinstitutionalvoidsindevelopingcountries:thecase
ofride-hailingmarkets. WorldDev.145:105528.doi:10.1016/j.worlddev.2021.105528
Hou,J.,Huo,X.,andYin,R.(2018).Doescomputerusagechangefarmers’production
and consumption? Evidence from China. China Agric. Econ. Rev. 11, 387–410.
doi:10.1108/CAER-09-2016-0149
Huang, A., Liu, B., Zhang, A., and Zhan, J. (2023). Coordinated development of rural
habitatinChina:astudyofmeasurement,spatio-temporalevolutionandconvergence.
J.Clean.Prod .398:136651.doi:10.1016/j.jclepro.2023.136651
Frontiersin SustainableFoodSystems 13 frontiersin.org

Zhan et al. 10.3389/fsufs.2026.1760055
Iancu,D.A.,andTrichakis,N.(2014).Paretoefficiencyinrobustoptimization. Manag.
Sci.60,130–147.doi:10.1287/mnsc.2013.1753
Jensen, R. T., and Miller, N. H. (2008). Giffen behavior and subsistence consumption.
Am.Econ.Rev .98,1553–1577.doi:10.1257/aer.98.4.1553
Kanemoto, K., Moran, D., Shigetomi, Y., Reynolds, C., and Kondo, Y. (2019). Meat
consumption does not explain differences in household food carbon footprints in
Japan.One Earth1,464–471.doi:10.1016/j.oneear.2019.12.004
Kauffman, R. J., and Wang, B. (2001). New buyers’ arrival under dynamic pricing
market microstructure: the case of group-buying discounts on the internet.J. Manag.
Inf.Syst.18,157–188.doi:10.1080/07421222.2001.11045687
Kebede, H. A. (2022). Market integration and separability of production
and consumption decisions in farm households. J. Dev. Econ . 158:102939.
doi:10.1016/j.jdeveco.2022.102939
Khan, N., Ray, R. L., Zhang, S., Osabuohien, E., and Ihtisham, M. (2022).
Influence of mobile phone and internet technology on income of rural farmers:
evidence from Khyber Pakhtunkhwa Province, Pakistan. Technol. Soc. 68:101866.
doi:10.1016/j.techsoc.2022.101866
Khanal, A. R., and Mishra, A. K. (2016). Financial performance of small farm
business households: the role of internet. China Agric. Econ. Rev. 8, 553–571.
doi:10.1108/CAER-12-2014-0147
Kongsamut, P., Rebelo, S., and Xie, D. (2001). Beyond balanced growth.Rev. Econ.
Stud.68,869–882.doi:10.1111/1467-937X.00193
Lagakos,D.(2020).Urban-ruralgapsinthedevelopingworld:doesinternalmigration
offeropportunities? J.Econ.Perspect .34,174–192.doi:10.1257/jep.34.3.174
Li, G., Zhang, R., Feng, S., and Wang, Y. (2022). Digital finance and sustainable
development:evidencefromenvironmentalinequalityinChina. Bus.StrategyEnviron .
31,3574–3594.doi:10.1002/bse.3105
Liu, B., and Huang, A. (2025). Digital economy and tourism growth: the role of
online attention allocation and transportation infrastructure.Tour. Rev. 81, 931–949.
doi:10.1108/TR-07-2024-0565
Liu, B., Zhan, J., and Zhang, A. (2023). Empowering rural human
settlement: digital economy’s path to progress. J. Clean. Prod . 427:39243.
doi:10.1016/j.jclepro.2023.139243
Liu, Y., Luan, L., Wu, W., Zhang, Z., and Hsu, Y. (2021). Can digital financial
inclusion promote China’s economic growth? Int. Rev. Financ. Anal. 78:101889.
doi:10.1016/j.irfa.2021.101889
Malik,P.K.,Singh,R.,Gehlot,A.,Akram,S.V.,andKumarDas,P.(2022).Village4.0:
digitalization of village with smart internet of things technologies.Comput. Ind. Eng.
165:107938.doi:10.1016/j.cie.2022.107938
Nunn, N., and Qian, N. (2014). US food aid and civil conflict.Am. Econ. Rev. 104,
1630–1666.doi:10.1257/aer.104.6.1630
Räisänen,J.,andTuovinen,T.(2020).Digitalinnovationsinruralmicro-enterprises. J.
RuralStud.73,56–67.doi:10.1016/j.jrurstud.2019.09.010
Ratchford, B., Soysal, G., Zentner, A., and Gauri, D. K. (2022). Online and offline
retailing: what we know and directions for future research.J. Retail. 98, 152–177.
doi:10.1016/j.jretai.2022.02.007
Razzaq, A., Sharif, A., Ozturk, I., and Skare, M. (2023). Asymmetric influence of
digitalfinance,andrenewableenergytechnologyinnovationongreengrowthinChina.
Renew.Energy 202,310–319.doi:10.1016/j.renene.2022.11.082
Resch, C., and Kock, A. (2021). The influence of information depth and
information breadth on brokers’ idea newness in online maker communities.Res.
Policy, Treading New Ground in Household Sector Innovation Research: Scope,
Emergence, Business Implications, and Diffusion50:104142. doi: 10.1016/j.respol.2020.
104142
Sheng, Y., Zhao, Y., Zhang, Q., Dong, W., and Huang, J. (2022). Boosting rural labor
off-farm employment through urban expansion in China.World Dev. 151:105727.
doi:10.1016/j.worlddev.2021.105727
Su,C-.W.,Liu,T-.Y.,Chang,H-.L.,andJiang,X-.Z.(2015).Isurbanizationnarrowing
the urban-rural income gap? A cross-regional study of China.Habitat Int. 48, 79–86.
doi:10.1016/j.habitatint.2015.03.002
Sun, M., Xu, X., Hu, Y., Ren, Y., Zhang, L., Wang, Y., et al. (2021). What differentiates
food-related environmental footprints of rural Chinese households?Resour. Conserv.
Recycl.166:105347.doi:10.1016/j.resconrec.2020.105347
Teece, D. J. (2014). A dynamic capabilities-based entrepreneurial theory of the
multinationalenterprise. J.Int.Bus.Stud .45,8–37.doi:10.1057/jibs.2013.54
Tsunoda, Y., and Zennyo, Y. (2021). Platform information transparency and effects
on third-party suppliers and offline retailers. Prod. Oper. Manag. 30, 4219–4235.
doi:10.1111/poms.13518
Xiao, H., Zheng, X., and Xie, L. (2022). Promoting pro-poor growth through
infrastructure investment: evidence from the targeted poverty alleviation
program in China. China Econ. Rev . 71:101729. doi: 10.1016/j.chieco.2021.
101729
Xie, Q., Yi, F., and Tian, X. (2022). Disparate changes of living standard in
China: perspective from Engel’s coefficient. China Agric. Econ. Rev. 15, 481–505.
doi:10.1108/CAER-03-2022-0046
Xie, T., Wei, Y., Chen, W., and Huang, H. (2020). Parallel evolution and response
decision method for public sentiment based on system dynamics.Eur. J. Oper. Res.
287,1131–1148.doi:10.1016/j.ejor.2020.05.025
Yang, B., Ma, F., Deng, W., and Pi, Y. (2022). Digital inclusive finance and rural
household subsistence consumption in China. Econ. Anal. Policy 76, 627–642.
doi:10.1016/j.eap.2022.09.007
Yu, C., Jia, N., Li, W., and Wu, R. (2021). Digital inclusive finance and rural
consumption structure - evidence from Peking University digital inclusive financial
index and China household finance survey.China Agric. Econ. Rev. 14, 165–183.
doi:10.1108/CAER-10-2020-0255
Zhao, C., Wu, Y., and Guo, J. (2022). Mobile payment and Chinese rural household
consumption.ChinaEcon.Rev .71:101719.doi:10.1016/j.chieco.2021.101719
Frontiersin SustainableFoodSystems 14 frontiersin.org

论文原文与出版信息

Frontiers in Sustainable Food Systems
查看论文原文

继续探索