Impacts of agri-food e-commerce on traditional wholesale industry: Evidence from China

Ruyi Yang ,Jifang Liu ,Shanshan Cao ,Wei Sun ,Fantao Kong

1 Institute of Agricultural Information,Chinese Academy of Agricultural Sciences,Beijing 100081,China

2 Institute of Special Animal and Plant Sciences,Chinese Academy of Agricultural Sciences,Changchun 130112,China

Abstract Rapidly expanding studies investigate the effects of e-commerce on company operations in the retail market.However,the interaction between agri-food e-commerce (AEC) and the traditional agri-food wholesale industry (AWI) has not received enough attention in the existing literature.Based on the provincial panel data from 2013 to 2020 in China,this paper examines the effect of AEC on AWI,comprising three dimensions: digitalization (DIGITAL),agrifood e-commerce infrastructure and supporting services (AECI),and agri-food e-commerce economy (AECE).First,AWI and AEC are measured using an entropy-based combination of indicators.The results indicate that for China as a whole,AWI has remained practically unchanged,whereas AEC exhibits a significant rising trend.Second,the findings of the fixed-effect regression reveal that DIGITAL and AECE tend to raise AWI,whereas AECI negatively affects AWI.Third,threshold regression results indicate that AECI tends to diminish AWI with three-stage inhibitory intensity,which manifests as a first increase and then a drop in the inhibition degree.These results suggest that with the introduction of e-commerce for agricultural product circulation,digital development will have catfish effects that tend to stimulate the vitality of the conventional wholesale industry and promote technical progress.Furthermore,the traditional wholesale industry benefits financially from e-commerce even while it diverts part of the traditional wholesale circulation for agricultural products.

Keywords: agri-food e-commerce,traditional wholesale industry,panel threshold model,dual-channel circulation

1.Introduction

In the era of Internet plus,e-commerce,as a new economic mechanism in the digital market (Bodini and Zanoli 2011),has aggressively entered the real economy to expand the scope of offline-to-online development of traditional industries,thus enabling dual-channel circulation (Lu and Liu 2015).The various benefits of e-commerce include streamlining the supply chain,enhancing internal and external communication,increasing revenues,reducing expenses,and improving the company’s brand (Kadam 2019;Tokaret al.2021).Based on its demonstrated impact on industrial markets (Zhu and Kraemer 2002;Hamadet al.2018),it is believed that e-commerce can stabilize the circulation chain of the agricultural sector.According to research (Pan and Li 2020),e-commerce has a significant impact on increasing the scale of agricultural trade.It can also enhance the profitability of agricultural markets by boosting sales and reducing search and transaction costs (Zhenget al.2019).Achieving dualchannel circulation is a critical step for maintaining a competitive advantage and realizing sustained economic growth,especially for China’s traditional agri-food wholesale markets,which encounter both opportunities and challenges from digital transformation.In February 2023,China unveiled its “No.1 central document” for the year,calling for efforts to strengthen the processing and circulation industry of agricultural products,which provides policy guidance for the digital transformation of the wholesale industry.

Since the production of Chinese agricultural products is characterized by small scale and large quantity (Luet al.2016),the main form of agri-food circulation in China is based on the “small production and big market” model involving approximately 200 million peasant households (Zou 2019),with 70% of agri-food requiring centralized distribution through wholesale markets.Additionally,parts of the wholesale markets have characteristics of sociality and commonweal for the public and consumers by ample supply of safe food (Ren and An 2010).Currently,a network of Chinese agri-food markets has been established,covering both urban and rural areas,controlled by intermediary enterprises and wholesale markets,and remaining the predominant distribution mechanism (Gan and Huang 2022).Moreover,the sales scale of the wholesale industry of agricultural products continues to expand.According to the data from the National Bureau of Statistics of China,the number of people employed in the wholesale industry of agricultural products in China reached 186,000,and the sales volume of commodities reached 352.4 billion USD in 2021,with the year-on-year growth of 6.29 and 32.58%,respectively.The influence of agri-food wholesale industry (AWI) extends beyond any direct interaction between farmers and consumers,or between farmers and restaurants (Hall and Gössling 2016).Nonetheless,there are still a number of major factors hindering the efficient circulation of the AWI,including a lack of standardization for the construction of trading facilities,outdated warehousing facilities and technology,insufficient investment in cold chain infrastructure,and a lack of public welfare function in wholesale markets (Wu and Qiu 2019).Several studies suggest that agri-food wholesale markets will continue to have significant development potential in China over the next 20 to 30 years,and their fundamental public welfare function will prevent them from being replaced by digital commerce.

The e-commerce industry of agricultural products has gradually grown as the country shows greater support for the digital economy,enabling the agricultural product circulation industry to share the dividends of the digital economy.In March 2015,the Chinese government introduced the “Internet Plus” action plan,which has since led to strong government support for agri-food e-commerce (AEC) (Liet al.2020).E-commerce giants,including Alibaba,have been actively contributing to the development of AEC,which has become a significant driving force for the rural economy due to its rapid growth and potential.The development of the Chinese AEC has gone through three stages (Liu and Walsh 2019): the beginning period (2005-2012),the exploration period (2012-2013),and the development period (2013-).The online retail sales of agricultural products reached 59.7 billion USD in 2021,with a year-on-year growth of 2.80%,accounting for about 16.89% of the annual wholesale sales of agricultural products in China (CIECC 2022).There were nearly 4,000 e-commerce platforms in China dedicated to agricultural transactions (Liu and Walsh 2019).Thus far,five mainstream models have been developed,including B2C (business to customer),C2C (customer to customer),O2O (online to offline),B2B (business to business),and CGP (community group purchasing).

In the agri-food trade industry,where e-commerce giants desire to participate,B2B and CGP prefer to optimize their operations in order to bolster their position in the agricultural supply chain.The agri-food B2B model involves the use of e-commerce to enhance business prospects and efficiency with suppliers and distributors (Gregoryet al.2019).These companies provide supply chain services by collecting procurement data to determine fresh food production volume,making direct purchases from production bases or farms,and delivering the produce to the terminal.Songxiaocai,a B2B company that translates to “fresh produce delivery” in Chinese,traded 200,000 tons of fresh produce annually in 2017 and operated in 45 cities across North,Central,and East China,achieving remarkable success.The CGP model experienced rapid growth due to the COVID-19 outbreak,which led to an increase in home consumption.This model involves consumers purchasing goods on platforms like WeChat or mini-programs and collecting them from offline physical stores located in residential areas (Liuet al.2021),making the process of collecting goods convenient.WeChat,the most popular social media platform in China,is used by almost 90% of the country’s community-supported agriculture farms (Zhenget al.2019).Similarly,the “community store+platform shopping” strategy is prevalent in medium and large communities in nearly all Chinese cities.

The B2B and CGP models,in comparison to general agricultural e-commerce,possess two unique characteristics that give them a role similar to that of the wholesale sector.Firstly,these models facilitate large sales volume per order,with B2B acting as a large online wholesaler that connects producers selling in large volumes to distributors and consumers who buy in large volumes.Secondly,these models have selfbuilt infrastructure and supporting services,with most B2B firms having their own cold storage and cold chain logistics that maximize existing industrial parks and idle factories.In the CGP operation system,offline stores like community service stations perform warehouse roles,while community chiefs or technology centers provide logistics and transport functions,thereby reducing distribution costs.

Inevitably,the rapid expansion of AEC impacts AWI,particularly B2B and CGP,which serves the same distribution function in the agri-food supply chain.Effective online and offline integration,from the perspective of e-commerce,can enhance channel performance and accelerate the rapid evolution of distribution integration,leading to the rapid development of vertical fresh e-commerce (Yu 2021).From the standpoint of the wholesale industry,sharing digital information and e-commerce facilities can enhance the operating efficiency of wholesale markets.However,e-commerce platforms have attracted numerous agrifood producers and brokers,as well as diverted some sub terminal distributors,resulting in the agglomeration diseconomies of conventional wholesale markets.This may lead to a trend from agglomeration to diffusion.

AEC has rapidly expanded,while AWI has progressed slowly but constantly.In many countries,the digital economy and the real economy are converging rapidly,and the digital transformation of AWI in China is also an unavoidable trend.Based on the existing literature,this study poses two questions: (1) Will AEC impact the growth of AWI? (2) Will the two achieve coordinated growth? If so,how? To address these questions,this study examines the development level of AWI and AEC by entropy weight method (EWM) using China’s province panel data from 2013 to 2020.It clarifies the relationship between AEC and AWI by a panel fixed effect regression model and a panel threshold model,which has theoretical and practical value for fostering the coordinated and healthy development of the two,as well as the digital transformation of the conventional wholesale industry in China’s regions.

The following are the three primary contributions.First,this research contributes to the increasing literature on the effects of AEC on AWI.Although AEC reveals its positive impacts from an economic and social standpoint,as demonstrated by studies on agribusiness (Negrão 2018;Lin Jet al.2020),rural areas (Tang and Zhu 2020;Jing and Jie 2021;Karine 2021),farmers (Ashokkumaret al.2019;Maet al.2020),and consumers (Liu 2018;Jiet al.2020) as proving subjects,most related studies have focused on the e-commerce retail industry (Roselliet al.2016;Zenget al.2019).Little is known about the impact mechanism of AEC on the conventional wholesale sector.This study addresses this information gap by investigating the impact of AEC on AWI,including operation efficiency,profit efficiency,circulation efficiency,development capacity.Second,most previous research has used proxy factors or questionnaire data to measure the level of AEC development,which lacks impartiality and comprehensiveness.This study addresses this issue by developing a novel and comprehensive evaluation index system that ingeniously incorporates three dimensions of AEC (digitalization,agri-food e-commerce infrastructure and supporting services,and agri-food e-commerce economy).This compensates for the problem of insufficient correlation between variables and ensures the validity of the results.Third,the present research primarily employs linear models to analyze the impact of e-commerce,overlooking the potential non-linear interaction between factors.This study is one of the first to investigate the non-linear relationship between AEC and AWI using a panel threshold model,which enables a more precise examination of the effect’s stage difference.

The rest of the paper is organized as follows.Section 2 presents a theoretical analysis and research hypothesis.Section 3 describes the data and methods used in the study.Section 4 presents the results and discussion.The conclusion and implications are presented in Section 5.

2.Theoretical analysis and research hypothesis

E-commerce of agricultural products is an online circulation channel that has emerged from the development of digitalization.It enables the online trading of agricultural products and generates certain economic benefits.It functions as a typical bilateral market,providing an online information matching and trading space between users and suppliers.This will inevitably have an impact on the traditional wholesale industry for offline sales of main businesses.We will discuss the potential impact of e-commerce’s digital development foundation,circulation channel development,and economic benefits on the operation efficiency of the wholesale industry from a theoretical perspective.

2.1.Effects of the e-commerce’s digital development foundation on the agri-food wholesale industry

With the deepening integration of digital technology in the commodity circulation industry,the digital wave has become the main driving force to promote the modernization of the traditional wholesale industry for agricultural products.The rapid spread and advancement of digital technology has led to its widespread adoption in daily life,resulting in reduced trade costs such as search,transportation,and communication (Wu and Yang 2019).Digital communication has overcome geographical space constraints and promoted operational efficiency,expanding the scope of commodity trading in agricultural wholesale markets and achieving a largerscale economy.The wholesale industry of agricultural products can upgrade technology applications by using digital technology to process the circulation of information and optimize the circulation process (Wei 2021).The high permeability of big data can break traditional circulation boundaries and establish digital connections between all links of the supply chain,integrating massive circulation information data within and between the wholesale industry using digital technology to optimize resource allocation (Hoffman and Novak 2018).The agricultural wholesale industry can create a new collaborative network using Internet thinking,establishing an intelligent connection between production and consumer ends,and carrying out digital and online transactions to promote business development.Based on the above theoretical analysis,this study proposes the following hypothesis.Hypothesis 1: E-commerce’s digital development foundation positively contributes to the operation efficiency of the agri-food conventional wholesale industry.

2.2.Non-linear effects of the e-commerce’s circulation channel development on the agri-food wholesale industry

New business forms enabled by digital technology are embedded in the real economy,and online distribution channels like B2B and CGY are gaining clout,posing a threat to the circulation and operational performance of the traditional agricultural wholesale industry.On the one hand,the disintermediation effect of agri-food e-commerce is a hindrance.Agricultural products require little added value from distributors,making e-commerce disintermediation a phenomenon not only related to online transactions,but also symbolizing the removal or weakening of intermediaries in the supply chain (Mills and Camek 2004).The e-commerce circulation channel may divert some of the original business volume of the agricultural wholesale market,leading to a decrease in the number of wholesalers and demonstrating that market agglomeration is not economically viable.On the other hand,the channel barrier effect of e-commerce has somewhat hindered the potential of the agricultural wholesale industry to build its online channel due to disparities in core business logic.E-commerce,as a typical bilateral market,aims to increase user penetration and capture more market share for a higher enterprise valuation (Lin Xet al.2020).As a result,it frequently employs the cross-price subsidy operation strategy and reduces price rigidity (Hillen and Fedoseeva 2021) to increase customer stickiness and raise enterprise valuation.This user-scale-based e-commerce profit model makes it more challenging for traditional wholesale marketplaces to adapt to e-commerce.After all,the conventional wholesale industry’s primary focus is on ensuring adequate supply and price stability.the price subsidy or low price-rigidity strategy is not a beneficial move to the wholesale industry,which is accustomed to making tiny profits and seeing big sales at a stable price.

Furthermore,the restraining effect of the e-commerce circulation channel on the operation of the traditional wholesale industry could be dynamic.In the early stage of e-commerce development,most e-commerce platforms are still exploring profit methods and boosting economic benefits,impacting the circulation of the wholesale industry to some extent with faster and more convenient online trading platforms.As e-commerce rapidly diffuses,more e-commerce platforms join in succession and compete for market share through various exclusive price subsidy strategies,exacerbating the channel inhibition effect on traditional industries.Over time,as a stable competition mechanism and user scale are formed in the e-commerce field,and the level of digital technology absorption and application in the wholesale market of agricultural products improves,some distributors in the wholesale market may spontaneously integrate into the e-commerce channel to realize the dual circulation,gradually weakening the channel diversion and barrier effect caused by e-commerce.Based on this,the following hypothesis is proposed.Hypothesis 2: E-commerce’s circulation channel development has non-linear effects on the operation efficiency of the agri-food conventional wholesale industry.

2.3.Effects of the e-economy on the agri-food wholesale industry

In the process of promoting the technological progress of the wholesale industry through digitalization,the agri-food e-commerce channel may have a certain degree of negative inhibition effect,but the application of e-commerce has opened up sales for agricultural products (Jin 2018) in the following two ways.Specialized e-commerce marketing capabilities directly increase a firm’s distribution and communication efficiency (Gregoryet al.2019),which positively impacts the trade extension of agricultural products.China’s agricultural production has the characteristics of geographical dispersion and smallholder economy.The wholesale market remains the distribution center for agricultural products under the “wide circulation and large market” circulation pattern.Therefore,a sizable portion of agricultural products sold through e-commerce still originate from the wholesale market,providing e-commerce economic benefits to the wholesale business.Additionally,the intangible advantages of e-commerce would trickle down to the traditional wholesale industry through individual wholesale market dealers frequently and impulsively using the e-commerce trading platform.Based on this,the following hypothesis is proposed.Hypothesis 3: The e-commerce economy benefits the traditional wholesale industry for agricultural products.

3.Data and methods

3.1.Data source

This study is based on the panel data of 30 provinces,municipalities and autonomous regions (hereinafter referred to as provinces) in China from 2013 to 2020.Given the integrity of the data,the relevant data for Xizang,Taiwan,Hong Kong,and Macao are excluded from this research sample.China Statistical Yearbook(NBSC 2014-2021a) andChina Trade and External Economic Statistical Yearbook(NBSC 2014-2021b) are the primary sources for relevant statistics in this study.For a simple imputation of basic data missing in a single year,the average value approach was adopted.To minimize the impact of pricing factors,we used the 2013 as the base year and the wholesale value-added deflator to deflate sales,net profit,assets,and other data.The GDP deflator of the base year was used to deflate e-commerce sales.

3.2.Measurement variables

Dependent variablePrevious research has identified scale (Zhuet al.2018),productivity (See and Ma 2018),profitability (Zubairet al.2021),and development capacity (Milliganet al.2017) as explanatory variables for the performance and efficiency of an industry.Based on the existing literature,a four-dimensional assessment index system was developed to evaluate the development level of AWI (Table 1): operational productivity (OE),profit effectiveness (PE),circulatory effectiveness (CE),and developmental potential (DC).Among them,purchasesale rate,management cost rate,and asset-liability ratio are included as negative indicators,meaning the lower the index value,the more favorable the evaluation outcome.

Table 1 Evaluation index system of agri-food wholesale industry (AWI)

Independent variablesThis research concentrates primarily on AEC with wholesale equivalent functions.Due to disparities in resource endowments and economic foundations,the development level of AEC measured by proxy variables (such as agri-food e-commerce sales and express delivery volume) can only reflect the realization of agricultural e-commerce from the perspective of a single output,and there may be issues such as insufficient correlation between variables.To better evaluate AEC,we further decompose it into three components (Table 2): (1) The growth of the AEC relies heavily on the digitalization (DIGITAL) of its IT infrastructure,which facilitates digital transactions and cooperation.The evaluation index system includes seven sub-indicatorsof DIGITAL,as per the research of Liuet al.(2020).(2) the indicator of agri-food e-commerce infrastructure and supporting services (AECI) is crucial for distinguishing AEC circulation from traditional wholesale circulation and represents the variety of online and offline channels for selling agricultural products.Considering AEC’s function as an intermediary between producers,distributors,and consumers,five sub-indicators from both the production and consumption viewpoints are included.On the production side,the introduction of Taobao Village1The term “Taobao Village” was initially used in April 2007 by AliResearch,a member of the Alibaba Group,to describe a village of several e-businesses that rely on Taobao as their primary trade platform,creating an Internet business cluster with agglomeration effects.The following three characteristics must exist in a Taobao village: First,businesses must be located in rural areas where the administrative village is used as a unit of measurement;second,annual e-commerce transactions coming from the village must reach 10 million CNY;third,there must be more than 100 active online shops or more than 10% of the village’s households must be represented by active online shops.has led logistics firms to establish numerous new distribution hubs and collecting locations in rural areas,promoting the expansion of e-commerce (Zenget al.2017).Therefore,the number of Taobao villages is utilized to indicate thedevelopment of digital agriculture bases,which is also a crucial indicator selected by many researchers (e.g.,Weiet al.2020) as the agent variable of e-commerce.Inclusive finance digitalization serves as the basis for farmers to access e-commerce to conduct online sales,and thus,the degree of inclusive finance digitalization (based on the digital inclusive finance index evaluated by Guoet al.2020) is used.In terms of consumption,with the rise of e-commerce sales and the passive impact of the COVID-19 pandemic,consumers have gradually formed a food consumption preference for “online purchase and offline delivery”,and more new businesses with the community as the core have emerged,such as WeChat group purchase,forming a unique e-commerce circulation and transaction mode.Therefore,we consider the new food purchasing methods of “online purchase with offline delivery or self-picking.” The number of postal service outlets and community service stations reflect the number of offline receiving base stations and self-picking base stations,respectively.In addition,the urban Engel coefficient serves as a negative indicator to determine the potential capacity of urban consumers to purchase agrifood online.(3) As the output component of AEC,the agri-food e-commerce economy (AECE) comprises the e-commerce foundation and sales scale of enterprises that sell agricultural products online.This study uses the proportion of firms engaged in e-commerce trading activities to all firms to assess the level of e-commerce development and the number of agri-food wholesale firms with official WeChat accounts to estimate the extent of online promotion and marketing.Additionally,based on available data,the value of online retail sales of physical items is used to determine the scope of agri-food sales online,and the value of online sales is used to estimate the future expansion of agri-food e-commerce.

Table 2 Evaluation index system of agri-food e-commerce (AEC)

Control variablesIn addition to the above-mentioned primary explanatory variables,numerous other variables have been found to directly influence the development of AWI.Consequently,Table 3 presents the control variables chosen for this study based on previous research.Table 4 presents the descriptive statistical results of each variable.

Table 3 Control variables

Table 4 Descriptive statistical results

3.3.Model selection

Entropy weight methodEWM is an objective weighting approach that overcomes information overlap among multiple index variables,reduces the subjectivity of index weights,and minimizes interference of human factors.This method ensures objectivity and high credibility of calculated index weights.According to the variation degree of each index,the entropy weight of each index was calculated by the information entropy,and then the weight of each index was modified through the entropy weight.Finally,more objective index weights have been obtained (Bouraimaet al.2021).The EWM was used in this study to evaluate two primary indicators: (1) the level of development of AEC and (2) the level of development of AWI.The EWM weighing procedure is as follows:Step 1: data standardization

where,Xij,minXij,maxXij,andYijare the original value,minimum value,maximum value,and standardized value of theith object under thejth index,respectively.Among them,i=1,2,…,m;j=1,2,…,n.The same below.

Step 2:Pijis used to represent the index value of theith object under thejth index,and the expression is:

Step 3: solve the entropy value under thejth index:

Step 4: calculate the coefficient of variation under thejth index:

Step 5: calculate the weight value of thejth index:

Step 6:Zisis used to represent the index score of thesth criteria layer under theith object,and the expression is:

among them,qis the total number of indexes contained in the criteria layer.

Step 7: the comprehensive evaluation value (e.g.,of the core indicator in each province from 2013 to 2020) is:

where,prepresents the total number of criteria layers.

Panel regression modelThe panel regression model may simultaneously reflect the change rule,attributes,and influence of variables in a two-dimensional section and temporal space.We use panel regression to evaluate,on a preliminary basis,the relationship between the level of development of AEC and the level of development of AWI.This is the baseline regression model:

where,idenotes the province,tdenotes the year,μiis the individual effects,AWI denotes the development level of the agri-food traditional wholesale industry,DIGITALdenotes the development level of digitalization,AECIdenotes the development level of agri-food e-commerce infrastructure and supporting services,AECEdenotes the development level of agri-food e-commerce economy,Hdenotes a series of control variables,βis the variable coefficient,μiis the individual effects,δtis the time effects,andεitis a random error term.

Panel threshold modelNevertheless,there may also be a non-linear influence link between AEC and the conventional wholesale industry.The panel threshold regression model based on individual fixed effects suggested by Hansen (1999) is capable of reliably estimating the threshold value and testing the importance of the endogenous threshold effect,reducing the bias resulting from subjective arbitrariness.Consequently,a panel threshold model is constructed for an empirical test with AECI as the threshold variable:

where,AWIitrepresents the development level of the agri-food traditional wholesale industry of provinceiin yeart;AECIitis the threshold variable,representing the development level of agri-food e-commerce infrastructure and supporting services of provinceiin yeart;αirepresents the unobserved individual fixed effects;γis the specific threshold value;and I (·) is an indicator function whose value depends on the relationship between the threshold variable (AECI) and the threshold value (γ).If there is a single threshold,the impact of AECI on AWI can be divided into two stages by a unique threshold valueγ1,which can replace the real case of AECI.Its economic meaning is attaining a specific level of AECI development will most effectively encourage or limit the expansion of the AWI,or its direction of action will remain the same while its degree of influence will alter.If there are multiple thresholds,γ2is introduced to make the empirical calculation coefficients display “three phases,” i.e.,there are two values as the turning point,so dividing the influence of the development level of an AECI on AWI into three stages.In each step,the degree of effect varies,and the direction of influence may not be consistent.

4.Results and discussion

4.1.Evaluation results of AWI and AEC

The entropy weight method was initially employed to evaluate the AWI and AEC,examining their temporal evolution and spatial differentiation characteristics.The regional division method proposed by the 7th Five-Year Plan of China was used to divide the provinces into three regions: eastern,central,and western regions.

From 2013 to 2020,the AWI at the national level remained stable,with a small fluctuation (no more than 0.01) and a steady yearly average value of 0.3454.In contrast,The AEC exhibited a consistent increasing trend,with the average annual value increasing from 0.1445 in 2013 to 0.3684 in 2020,a roughly 2.5-fold increase.In 2019,the two values intersected,and in 2020,the AEC held a 0.01-point advantage over the AWI (Fig.1).

Fig.1 Time series evolution of the agri-food wholesale industry (AWI) and agri-food e-commerce (AEC).

In terms of spatial characteristics,the AWI exhibited greater spatial balance than the AEC in 2020.However,the AWI of 18 provinces was still lower than its total value (0.3540),whereas the AEC of 18 provinces exceeded that of the AWI (Fig.2).

Fig.2 Values of the agri-food wholesale industry (AWI) and agri-food e-commerce (AEC) in 2020.

Evolution characteristics of the AWIFig.3-A displays the progression characteristics of the AWI.The average annual AWI of the eastern,western,and central regions was categorized into high,medium,and low grades using every 0.01 as a time series separating point.From 2013 to 2020,the average annual AWI of the eastern region was 0.3584,which was higher than the average value of 0.3454.In contrast,the average annual AWI of the western and central regions was 0.3425 and 0.3338,respectively,both of which were lower than the average value.The peak of the AWI in the eastern and central regions occurred in 2014.From 2014 to 2015,the AWI decreased rapidly at the national level.Since 2015,the AWI of the eastern region has continued to decline after recovering,while the AWI of the western region has continued to improve.The AWI of the central region slowly recovered,reaching 0.3649 by 2020,which was higher than the corresponding levels of 0.3535 and 0.3447 for the eastern and western regions.The lack of discernible regional differences in the AWI can be primarily attributed to China’s longstanding practice of “wide circulation and large market”.The slight regional disparity in AWI was mainly due to variations in market size and the concentration of production and marketing.The eastern region consists of major agri-food sales areas with large populations,such as Beijing and Shanghai,as well as regions with favorable geographical and climatic conditions for agricultural production,such as Shandong and Hebei.These characteristics indicate the presence of substantial agri-food wholesale markets with high concentration and circulation efficiency.The western region,mainly comprising Yunnan,Guizhou,and Sichuan,is continuously improving the circulation efficiency of its agri-food wholesale markets by creating clusters of agricultural product-producing areas,thanks to its suitable climate and abundant resources.

Fig.3 Evolution characteristics of the agri-food wholesale industry (AWI) and agri-food e-commerce (AEC).

Evolution characteristics of the AECFig.3-B shows the characteristics of the AEC’s evolution.From 2013 to 2020,the overall AEC increased from 0.1445 to 0.3684,with an average annual growth rate of 15.01%.In 2014-2015,the AEC’s average annual growth rate was 43.39%,as it rose from 0.1770 to 0.2538.The eastern area has the highest average yearly AEC value (0.3216),followed by the western region (0.2499),while the central regions have the lowest value (0.2356).The rate of increase,however,was inverted: the AEC of western regions almost tripled from 0.1165 in 2013 to 0.3507 in 2020,while the AEC of the central regions increased 1.8 times from 0.1175 to 0.3307 between 2013 and 2020.The AEC of the eastern regions multiplied by 1.2 from 0.1920 in 2013 to 0.4172 in 2020.The regional disparities in AEC are primarily due to variances in consumer scale and infrastructure development.The relatively high level of urbanization in the eastern region makes consumer demand the primary driver of e-commerce growth.For instance,Beijing’s AEC (with an average yearly value of 0.4591) is significantly higher than that of other provinces,and it has remained high and stable due to its unique status as a political center and a megacity with more than 21 million permanent residents.The convenient e-commerce distribution channel for agricultural products,with the community as the consuming core,has a large audience and makes it easy to create consumer preferences.Often,the development of e-commerce auxiliary facilities follows the same trajectory as economic growth.Jiangsu,Zhejiang,and Shanghai,for instance,have relatively high economic levels,making it easy to generate a high-level development trend through the establishment of agricultural e-commerce infrastructure.The need for production and distribution is another critical factor in the expansion of e-commerce.The AEC of the western region experienced the highest increase,mostly owing to the fact that the region is a major agrifood-producing area with abundant natural resources.These provinces,including Xinjiang,Gansu,Yunnan,Guizhou,Guangxi,and Henan,have a low level of AEC development,with average yearly values close to 0.2.This could be attributed to their remote location and low economic status.There is still considerable room for advancement.

4.2.Panel fixed effect regression

This study started by constructing a panel regression model to investigate the overall relationship between AWI and AEC.To avoid false correlation caused by the unsteady indices,we conducted the unit root testing and cointegration test using two methods: the Levin Lin-Chu unit root test and the Pedroni test.The test results showed that all variables were stable,and there was a cointegration relationship between the variables,indicating that the variables used in this study met the data stability requirements of regression models.To improve the accuracy of the estimation results,the coefficient of variance expansion (VIF) was used.The results show that the mean value of VIF between each explanatory variable was 2.63.Thus,the variable selection is effective without the possibility of multicollinearity between variables.In addition,an ordinary least squares method (OLS) and a random effect method (RE) were also used for comparative analysis and robustness consideration.The Hausman test results indicated that theP-value is 0.0121,which is significant at the 5% level.Therefore,it was appropriate to adopt the FE as the core model.The estimation results for the three regression models are presented in column (3) of Table 5.

Table 5 Estimation results of panel regression models

Estimation resultsThe three aspects of AEC have different effects on AWI.The results show that DIGITAL has a positive effect on AWI,with an estimated coefficient of 0.4826 and significance at the 1% level.AECI has a negative effect on AWI,with an estimated coefficient of -0.7890 and significance at the 1% level,indicating a significant impact of channel diversion.AECE has a positive effect on AWI,with an estimated coefficient of 0.4457 and significance at the 5% level,suggesting that an increase in AECE will encourage the growth of AWI.The results of the OLS in column (1) and the RE in column (2) are consistent with the FE estimation results,with a significance level of at least 5% and only slight differences in impact intensity.So far,Hypothesis 1 and Hypothesis 3 have been verified.

Evidently,digital development,as the embodiment of enhanced productivity,is a key foundation for the development of agri-food e-commerce and one of the most influential factors in promoting the growth of AWI.Furthermore,even though e-commerce diverted some of the conventional wholesale channel flow,the online economic advantage has still contributed to the growth of AWI.In other words,many conventional wholesale businesses have modernized and now operate online.

Reginal heterogeneityChina possesses an expansive territory with notable disparities in economic development,industrial structure,and resource endowment among its various regions,particularly the eastern,central,and western regions.Being at the forefront of China’s economic development,the eastern region surpasses the central and western regions in terms of economic development,technological innovation,and e-commerce advancement.Hence,we aim to examine whether the influence of AEC in the eastern region on AWI surpasses that in the central and western regions.Simultaneously,as agricultural production undergoes gradual commercialization,specialization,and largescale development,China has established a pattern of “big market and extensive circulation” for agricultural products.The primary agricultural production areas and sales regions share the responsibility of ensuring a stable supply of agricultural products.This is primarily accomplished by establishing a connection between production and sales across different regions.However,the nationwide development of e-commerce introduces potential disruptions at both ends of the circulation.Therefore,we aim to investigate whether the development of AEC in the primary agricultural production areas has a distinct influence on AWI compared to the sales regions.

Building upon the aforementioned hypothesis,regional dummy variable D1 was incorporated to assess the disparities in impact between the eastern region and the combined central and western regions.The eastern region was assigned a value of 1,while the central and western regions were assigned a value of 0.D2 investigated the disparities in the impact of the main agricultural production area and main sale areas2The main sale areas of agricultural products include Beijing,Tianjin,Hebei,Liaoning,Heilongjiang,Shanghai,Jiangsu,Zhejiang,Anhui,Hunan,Guangdong,Guangxi,and Chongqing.The main production areas of agricultural products include Shanxi,Inner Mongolia,Jilin,Fujian,Jiangxi,Shandong,Henan,Hubei,Sichuan,Guizhou,Hainan,Yunnan,Shaanxi,Gansu,Qinghai,Ningxia and Xinjiang.on AEC and AWI in the agricultural sector.The main agricultural production area was assigned a value of 1,while the sales area was assigned a value of 0.Interaction terms between regional dummy variables and the three primary explanatory variables of AEC were incorporated into the model to assess regional heterogeneity in the impact of AEC on AWI.The non-zero coefficient value of the interaction term indicates the presence of regional heterogeneity in the influence.Moreover,if the estimated coefficient value aligns with the direction of the explanatory variable,it suggests that the impact of DIGITAL/AECI/AECE on AWI in this region exceeds that of other regions.The estimation results of FE are shown in Table 6.

Table 6 Regression results of regional differences

From the perspective of the three regions,DIGITAL still has a substantial positive impact on AWI in the three regions,but the interaction term is significantly negative,suggesting that the influence of DIGITAL on AWI is relatively weaker in the eastern region compared to the central and western regions.AECI still has a significant inhibitory effect on AWI,but the interaction term is significantly negative,indicating that the inhibitory effect of AECI on AWI is more pronounced in the eastern region.AECE still has a significant positive impact on AWI,and the interaction term is significantly positive,indicating that the promoting effect of AECE on AWI is particularly notable in the eastern region.These findings suggest that the conventional agricultural wholesale markets in the central and western regions are more influenced by the knowledge spillover effect of digital technology development,having the enhanced digital connectivity facilitated by e-commerce development.The wholesalemarkets in the eastern region experience greater economic benefits from e-commerce due to the availability of supportive e-commerce facilities.

Moreover,DIGITAL still has a substantial positive impact on AWI in the main production area and sales area,and the interaction term is significantly positive,indicating that the DIGITAL impact on AWI is relatively greater in the main production area.It can be inferred that the development of digitalization has demonstrated a more positive promotional effect on the conventional wholesale markets in the main production area.Additionally,the impact of e-commerce development on both production and sales ends does not significantly differ within the large market and circulation pattern;inclusive economic benefits have exhibited similar effects in both the main production and sales areas.

4.3.Panel threshold regression

Combined with theoretical analysis and practical conditions,the impact of various levels of AECI on AWI may exhibit non-linear threshold characteristics,which can be examined using a panel threshold model.The Bootstrap method was used to test the AECI level as the threshold variable before conducting the threshold regression.The results,presented in Table 7,show that both the single and double thresholds of AECI haveF-values andP-values that pass the significance test of 5%,indicating the existence of two threshold values.The confidence intervals for the two threshold values (AECI=0.0206 and AECI=0.0236) are relatively narrow,respectively [0.0201,0.0210] and [0.0198,0.0245],so the recognition effect of the threshold value is relatively accurate.

Table 7 Test results of threshold effect (agri-food e-commerce infrastructure and support (AECI) as the threshold variable)

Fig.4 shows the likelihood ratio function,which explains the derivation of the threshold value and confidence interval.The intersection of the LR trend line and the dashed line in the figure represents the confidence interval for the threshold variable.The calculated confidence interval for the first threshold is [0.02201,0.0210].The first threshold value is 0.0206,as the LR value is closest to zero when the threshold variable (AECI) is at this point.Below the dashed line,there is a second point where the LR trend line is closest to zero,indicating a three-stage inhibitory effect of AECI on AWI.This is determined by the estimation findings of the double threshold model (Table 8),using the two threshold values of the threshold variable (AECI) as thedividing point.

Fig.4 Threshold values and confidence intervals.

Table 8 Estimation results of the double threshold model

When AECI is below 0.0206,the estimated coefficient is -1.0661,which is statistically significant at the 10% level.The estimated coefficient is -3.9950 when AECI is between 0.0206 and 0.0236,and it is statistically significant at the 1% level.AECI continues to reduce AWI.When AECI exceeds 0.0236,the estimated coefficient is -0.8384,which is statistically significant at the 1% level.At this stage,AECI still restricts the AWI level,but to a much lesser extent.Hypothesis 2 is supported.AECI exhibits a channel-diversion effect on AWI during the study period.When AECI is relatively low,increasing it will moderately reduce AWI,and as AECI increases,the inhibitory effect will first increase before diminishing.However,when AECI levels are relatively high,inhibition is significantly reduced.

In addition,the study samples are divided into three intervals based on the double threshold value: low transfer interval (AECI≤0.0206),medium transfer interval (0.0206＜AECI≤0.0236),and high transfer interval (AECI＞0.0236),as shown in Table 9.At various times and places,AECI inhibits AWI to varying degrees.In 2013,Tianjin,Anhui,Jiangxi,Guangdong,Hainan,Chongqing,Sichuan,Guizhou,and Gansu were in the low transfer interval,while Hubei,Ningxia,and Qinghai were in the medium transfer interval.AECI strongly inhibited AWI in 9 provinces and had an even stronger effect in 3 provinces.Since 2014,this type of influence has significantly decreased.In 2014,it had a large inhibitive effect on AWI in only 3 provinces (Guangxi,Hainan,and Sichuan),and since 2015,this effect has diminished dramatically.Moreover,30 provinces joined the high transfer interval,with no regional variation.It is evident that the initial short-term negative restriction of AECI on AWI only occurred in a few provinces,and all of them quickly eliminated this situation.The consistency between the panel threshold model and the panel fixed effect regression model reaffirms the reliability and robustness of the conclusions.

Table 9 Interval division and province distribution of agri-food e-commerce infrastructure and support (AECI) in 30 provinces of China from 2013 to 2020

In summary,the panel fixed effect model and threshold model provide an interesting finding that the three dimensions of AEC have diverse effects on the direction and magnitude of AWI.Firstly,raising DIGITAL and AECEwill enhance AWI,while increasing AECI will have the opposite effect.The following are the reasons: Firstly,digital technology can enhance digital transfer,save time and costs,and ensure a consistent market supply (Leeet al.2000;Gaoet al.2021).Increasing DIGITAL tends to improve AWI,demonstrating that digital technology is not only a prerequisite for the development of AEC but also a factor that promotes the circulation of conventional wholesale channels.Secondly,increasing AECI tends to decrease AWI.This suggests that agricultural products that were previously distributed through the traditional wholesale channel have been significantly diverted through the e-commerce mode of “online purchase with offline receiving or self-picking”,which creates a barrier in the circulation channel and somewhat hinders the growth of AWI.

Unavoidably,the expansion of online transaction channels has impacted the traditional wholesale model (Gan and Huang 2022).This finding is consistent with Alvarezet al.’s (2020) research,which indicates that e-sales significantly affect the wholesale and retail trade sectors.On the one hand,AEC fosters the expansion of AWI through digital technology diffusion.On the other hand,it diverts conventional wholesale circulation flows.Despite a certain degree of diversion of traditional circulation flows by e-commerce,the economic expansion of AEC indicates a major trend toward promoting wholesale industry development.Consequently,a substantial share of agricultural products marketed online originate from wholesalers and brokers in wholesale markets.Thus,we may conclude that AWI remains the primary battlefield in China’s agricultural goods circulation.Parts of wholesale businesses have already been upgraded to function as e-commerce,transforming the traditional ecosystem into an entrepreneurial one (Songet al.2022).Furthermore,since a portion of agricultural products are diverted by some third-party trade platforms (such as Alibaba and Minyu e-commerce company,Yanget al.2020),the economic benefits they produce ultimately encourage the growth of the formal wholesale sector,indicating that AEC and the formal wholesale sector have a mutually beneficial relationship.

Furthermore,AECI has a three-stage (strong-strongerweak) inhibitory effect on AWI,confirming the conclusion of the panel regression model from the previous section.When the Chinese AEC was in its exploratory phase,more smallholders began selling their agricultural products directly to consumers online through online shops on third-party trade platforms (Zenget al.2017).The rise of online circulation of agricultural products as a new outlet impeded the improvement of traditional wholesale circulation to some extent,and many third-party trade platforms with storage and transportation advantages have entered the market.A number of conventional wholesale businesses are currently active in both online and offline business,such as the generalization of e-transaction services in wholesale markets (Sui and Geng 2021) and the emergence of agricultural-exclusive e-commerce platforms (Jiang 2015),demonstrating an inhibition-removed effect.The wholesale industry has maintained its central position in the system of circulation by combining online and offline transactions to promote the circulation of agricultural products.

4.4.Endogeneity and robustness test

EndogeneityTo address endogeneity caused by measurement errors,authoritative sources were used to obtain data,and the deflation index was adopted to minimize the influence of price factors on the results.Additionally,the entropy weight method was used to calculate core variables,thereby reducing the impact of data quality on the estimates.

To address the issue of omitted variables,a set of control variables was included in the estimation process,and the fixed effect model was employed as the primary model,which can account for individual heterogeneity.However,the control variables selected,although as far as possible the important control variables into the model,but also unable to achieve comprehensive selection.System GMM (SYS-GMM) can address endogeneity problems caused by variable measurement errors and omitted variables(Bondet al.2001).Therefore,SYSGMM was used with the lagged term of AWI as the instrumental variable.The corresponding test results are given in column (3) of Table 10.The results show that the perturbation term of the SYS-GMM exhibits first-order autocorrelation,while the second-order model does not exhibit autocorrelation,indicating that variable selection is reasonable.To ensure the stability of the results,the OLS method with fixed years and the OLS method with fixed individuals and years were used to verify the rationality of the SYS-GMM estimation results.The regression results are shown in columns (1) and (2) of Table 10,respectively.The coefficient of the lagged term of AWI in SYS-GMM estimation in column (3) is 0.3075,which is between the results given by the two models in columns (1) and (2).This finding meets Bondet al.’s (2001) criteria for evaluating the effectiveness of GMM estimation,thus supporting the reasonableness of the SYS-GMM results and further verifying the stability of the results of FE.

Table 10 Endogenous test results (OLS and SYS-GMM)

To address the endogeneity problems caused by twoway causality,DIGITAL,AECI,and AECE were used as the core explanatory variables in the regression,with each having a one-period lag.The reasoning is that the current period’s improvement of AWI has little effect on the lagged terms of DIGITAL,AECI,and AECE.If the corresponding relationship between AWI and the one-period lagged value of terms of DIGITAL,AECI,and AECE still holds in the current period,it can be inferred that the three variables are the main cause of the two-way causality,rather than AWI.The results are given in columns (1)-(3) of Table 11,which shows that the conclusions are consistent with the findings from the previous basic regression.

Table 11 Endogenous test results (FE and 2SLS)

At last,to ensure the unbiasedness of the panel threshold regression estimator,we used the instrumental variable method to test it.Following Mogstadet al.’s (2021) study,the one-period and two-period lagged AECI were selected as the multiple instruments for AECI in the current period of the 2SLS model.The results are shown in column (4) of Table 11.The direction and significance of the influence coefficient of AECI remain the same as the previous estimated results,and the instrumental variable test supports the rationality of the results,indicating that the threshold regression estimationis unbiased.Thus,the research hypothesis 2 of this study has been further verified.

Robustness testTo test the robustness of the results,this study substituted the core explained variable with alternative specifications.Since the digital technology spillover effect and channel diversion effect may mainly affect the circulation process of the agricultural wholesale industry,circulation efficiency (CE) was used as an alternative variable in the OLS,RE,and FE models.The estimated results are presented in Table 12.

Table 12 Results of robustness test

For all models,the regression results are robust.The directions of influence of DIGITAL,AECI,and AECE are consistent across all models and significant at a level of at least 5%.Specifically,the estimated coefficient of DIGITAL is positive at the significance level of 1%,the estimated coefficient of AECI is negative at the significance level of 10%,and the estimated coefficient of AECE is positive at the significance level of 10%,whichare consistent with the previously estimated results.These findings further verify the robustness of the estimation.

5.Conclusion and implications

Many studies has explored how e-commerce influences circulation efficiency and corporate operations,but little attention has been given to the traditional agricultural wholesale sector.This study contributes to this emerging research topic by examining the effects of the three dimensions of the AEC (including DIGITAL,AECI,and AECE) on AWI in China.This helps to better understand the role of e-commerce in the digital transformation of conventional wholesale industry for agricultural products and provides a practical foundation for the integrated development of e-commerce and wholesale industry.

Based on China’s province panel data from 2013 to 2020,this study first evaluates the temporal evolution and regional differentiation characteristics of the entropybased AWI and AEC values.Furthermore,the impacts of DIGITAL,AECI,and AECE on AWI are determined using a panel fixed effect regression model.Additionally,this study investigates the threshold influence of AECI on AWI by a panel threshold model.First,the data indicate that AWI was essentially steady,with an annual average of 0.3454 and no significant regional disparities.AEC had a consistent upward trend,increasing approximately 2.5 times from 0.1445 in 2013 to 0.3684 in 2020.Furthermore,there is a significant regional imbalance in AEC,with the eastern region having a much higher value than the western and central regions.Second,the effects of the three dimensions of AEC on AWI have distinct directions and magnitudes.Increases in DIGITAL and AECE tend to increase AWI,while AECI tends to decrease AWI.Third,there is a significant double-threshold effect of AECI in the effect of AEC on AWI.This effect is reflected in the three-stage (strong-stronger-weak) inhibitory effect of AECI on AWI,confirming the results of the panel fixed effect regression model.The study samples are divided into three intervals based on the two threshold values: Low transfer interval (AECI≤0.0206),medium transfer interval (0.0206＜AECI≤0.0236),and high transfer interval (AECI＞0.0236),the significant negative inhibition of AECI on AWI occurred in the initial short term in 12 provinces,but they immediately eliminated this situation.In 2015,30 provinces entered the overall high transfer interval.Comparatively to the east,the western agrifood wholesale industry benefitted more from digital development.Similarly,DIGITAL has a greater facilitating role in the main production areas than in the main sales areas.Only in the eastern regions and the main sales areas does AECE play a significant role in developing the wholesale industry.The west and the main production areas have not experienced any economic benefits from the e-commerce of agricultural products.

Although e-commerce of agricultural products diverts some traditional wholesale circulation,the traditional wholesale sector still benefits economically from e-commerce.Specifically,“e-commerce dividends” such as digitalization,e-commerce infrastructure,and supporting service building have fostered the digital upgrading of traditional wholesale markets.To a certain extent,the e-commerce economy has also bolstered their development.This indicates that China’s agricultural commodity distribution relies heavily on traditional wholesale markets.Most agricultural products sold online come from traditional wholesale markets,where online and offline dual-channel circulation is common.Therefore,in the competition between e-commerce and the wholesale industry of agricultural products,e-commerce competes for market share through innovation rather than in the market itself (Katz 2021).As a robust sector of the real economy,the wholesale agricultural commodities market is adjusting its capital to adapt to the impact of e-commerce,and the dual-channel supply chain of the agri-food wholesale industry has ample room for future improvement.

In light of this,the government should modernize the operations of conventional wholesale markets to inject new growth momentum into the wholesale industry.With the advent of e-commerce for agricultural products,the traditional wholesale markets will have significantly less marketing duty.The government can utilize B2B e-commerce and community group purchasing to take on the sales function of the wholesale market,allowing for a greater focus on the distribution function of large wholesale markets and increasing the demand for the cold chain storage and product quality inspection functions in these markets.Secondly,the government should promote the dual-channel chain strategy of conventional wholesale markets to strengthen their integration and symbiosis with e-commerce.To align with the digital economy trend,the government could utilize internet technology to establish a digital platform for contract farming,integrating production and marketing and enabling online and offline connectivity of the wholesale industry.Simultaneously,the government should capitalize on the e-commerce dividends,using the convenience of e-commerce infrastructure,supporting services,and professional cold storage facilities.To address the deficiencies of inadequate function and low utilization rate,the government could relocate certain small wet markets to other purposes and focus more on the dual-channel circulation of large agricultural wholesale markets.Thirdly,the government should promote balanced regional growth of agricultural e-commerce and unleash the national benefits of e-commerce in line with the agri-food circulation pattern of “large circulation and large market.” Currently,China’s “south-to-north transportation of vegetables” and “westto-east transportation of fruits” are both essential modes of circulation for balancing supply and demand.To align the Chinese distribution pattern for agricultural products with the level of e-commerce,local governments in the western and central regions can improve infrastructure,expand agricultural and logistics parks,and implement talent introduction policies to expedite industrial agglomeration and unlock e-commerce dividends in agricultural product circulation.

Acknowledgements

This study was supported by the Leading Talent Support Program for Agricultural Talents of the Chinese Academy of Agricultural Sciences (TCS2022020) and the General program of National Natural Science Foundation of China (1573263).

Declaration of competing interest

The authors declare that they have no conflict of interest.