	New content
Environment and Planning B: Urban Analytics and City Science

Free access

Research article

First published online March 25, 2021

The geography of online shopping in China and its key drivers

Zhouying Song https://orcid.org/0000-0001-8771-3633 [email protected]View all authors and affiliations

https://doi.org/10.1177/23998083211002189

Abstract

Many studies shed light on online shopping adoption and its determinants, most of which were conducted in developed countries and focused on sociodemographic factors. However, the influence of spatial attributes remains unclear or underestimated, especially in emerging economies. China is the largest e-commerce market in terms of the number of e-shoppers and the volume of online retail sales, with a distinctive geographical pattern of e-shopping usage. This study developed a conceptual model of China’s e-shopping adoption, examined the spatial characteristics of the e-shopping usage across 1918 counties, and explored its key drivers. The findings indicate a digital divide of e-shopping in China, with counties in metropolitan areas having a relatively high online shopping index (SI) value. The counties’ online shopping levels in China decrease from the eastern coastal regions to the rural-mountainous regions in west China. This study also found that the key drivers of e-shopping adoption are the rate of urbanization, local delivery systems, and the internet connection, which indicates that geographical attributes, rather than socioeconomic problems, have to be solved for improving online shopping adoption.

Introduction

During the coronavirus pandemic, online shopping sales, especially online grocery sales, have skyrocketed in many countries and regions. For example, in the United States, overall online sales increased by 25% amidst the COVID-19 crisis, while online grocery shopping has jumped over 100%. Since 4 March 2020, the “New Trade Festival” launched by Alibaba, an international online shop, increased the international trade volume by 71% in just one week.

Indeed, online shopping has proliferated in the last decades. In 2019, global online e-commerce sales reached $3.53 trillion, with an annual growth of about 10% in the last five years (Statista, 2020). The proliferation of online shopping has significantly affected the retail economy, consumption landscape, logistics system, and transport geography (Cao et al., 2013; Farag et al., 2006; Hood et al., 2020; Monica, 2019; Weltevreden et al., 2008). For example, in 2019, retail e-commerce sales in the USA were $599.51 billion, accounting for 9.8% of the total retail sales; and the European e-commerce market grew to €621 billion, accounting for almost 8.0% of the total retail sales (Statista, 2020). Furthermore, the popularity of e-shopping has gradually affected shopping behaviors, daily lives, and the places where people live (Jaller and Pahwa, 2020; Lee et al., 2017; Ren and Kwan, 2009; Song et al., 2020).

In the last two decades, many studies have been conducted on e-shopping, most of which are about the relationships between e-shopping and in-store shopping, and why people adopt e-shopping (Cao, 2012; Farag et al., 2007; Weltevreden and Van Rietbergen, 2009; Zhou and Wang, 2014). However, there has been little research on the geography of e-shopping adoption (Farag et al., 2006; Kirby-Hawkins et al., 2019). Moreover, the impacts of spatial factors on e-shopping remain unclear or have been underestimated (Cao et al., 2013; Farag et al., 2006). Furthermore, the existing research on e-shopping has mainly been conducted in developed countries, such as the USA (Cao et al., 2013), the Netherlands (Farag et al., 2006; Weltevreden et al., 2008), the UK (Clarke et al., 2015; Kirby-Hawkins et al., 2019; Longley et al., 2008), and Belgium (Beckers et al., 2018). So far, little geographical research has been conducted concerning the spatial characteristics of e-shopping in emerging economies and their driving factors.

This study sought to fill these gaps. Based on online shopping data from the Alibaba group, with more than 600 million active e-shoppers in China, we try to identify the spatial distribution of e-shopping adoption across 1918 counties in China and explore its main drivers. The aim of this study is twofold. First, we analyze how the Chinese online shopping was spatially distributed at the county level in 2017. It extends the empirical research on e-shopping in emerging economies and, thus, answers calls for multiple-level studies (Cruz et al., 2018; Song et al., 2020). As the world’s largest online shopping market, there is still no nationwide study of the spatial patterns of e-shopping in China and its determinants. Second, we explain how these spatial patterns are influenced by sociodemographics, land use features, personal characteristics, the physical retail economy, and geographical factors. This paper constructs a conceptual model of e-shopping adoption, in which sociodemographical and geographical factors are both considered.

The remainder of this paper is structured as follows: The upcoming section summarizes the existing research and introduces a conceptual framework based on established theory from the e-commerce literature, which will be used to discuss the results. A further section explains the data and methods in detail, which is followed by a section that examines the characteristics of the geographical pattern of online shopping in China. The penultimate section details the empirical estimation and discusses major influential factors. The final section concludes the study and places our results into the policy debate about e-commerce.

Theory and hypotheses

The study on e-shopping is part of a much broader field of research: social and policy implications of the internet (Ren and Kwan, 2009), in which an important question is the inequality of internet use. Many studies have focused on digital inequality and its driving factors (Van Dijk, 2006) and revealed that there was a huge inequality in internet access, use, and outcomes, which were named as the first-, second-, and third-order digital divides, respectively (Cruz et al., 2018; Dewan and Riggins, 2005; Song et al., 2020). Online shopping, more often categorized as e-commerce (Zhou and Wang, 2014), has been discussed as the second-order digital divide.

Many studies shed light on the determinants of e-shopping, most of which have been conducted by disciplines outside geography (Farag et al., 2007; Lee, 2002; Sim and Koi, 2002; Vrechopoulos et al., 2001). Some geographers have analyzed e-shopping from a spatial perspective (Cao et al., 2013; Farag et al., 2006). However, the influence of spatial attributes remains unclear or underestimated, which is critical in measuring the influence of e-shopping on the future consumption landscape, retail economy, and regional development. Based on the existing literature, factors capable of affecting e-shopping can be divided into two sets: sociodemographic factors and geographical factors. Sociodemographics refer to the demographic and socioeconomic attributes of e-shoppers, such as gender, age, education, income, and so forth; geographical attributes concern transport accessibility, the local delivery system, and the local environment. In this study, we explore both the impacts of sociodemographic and geographical factors. Based on the factors identified in the literature, 11 testable hypotheses (H1-H9B) are discussed in this section. The theoretical research model developed to test the relationships between the factors and e-shopping adoption is presented in Figure 1.

Sociodemographics

Many studies have emphasized the impacts of age and gender on e-shopping adoption (Fuchs, 2009; Song et al., 2020). For example, it has been argued that the relation between age and e-shopping adoption is piecewise linear, while older people are less likely to buy online (Beilock and Dimitrova, 2003; Casas et al., 2001). Some studies found that there is a non-linear inverse correlation between age and e-shopping, while people under the age of 40 tend to shop online more often (Clarke et al., 2015; Farag et al., 2007; Vrechopoulos et al., 2001). As the largest country in the world, China’s “young” online shoppers (under the age of 40) have been given a high priority. Moreover, it is widely known that there is an e-shopping divide between genders (Fuchs, 2009; Raijas, 2002; Song et al., 2020). Some research argued that men tend to have more positive attitudes towards online shopping (Lee et al., 2017). Others found that women tend to shop online more frequently than men (Hasan, 2010; Lee et al., 2017). Based on the aforementioned findings and the demographic situation in China, we therefore posit the following:

H1:There is a positive relationship between the youth population ratio (YON) and online shopping.

H2: Gender (GEN) would be associated with online shopping adoption.

The literature review revealed that individuals with a higher education are more inclined to shop online (Casas et al., 2001; Sim and Koi, 2002). However, a growing number of people with basic education has begun to use the internet and shop online in China, against the background of cheap smartphones and the popularity of personal computers (Song et al., 2020; Yang et al., 2013). As a result, the education level of individuals fosters internet skills, which would stimulate them to use the internet and shop online (Song et al., 2014). Thus, the following hypothesis is formulated:

H3: Adult literacy rate (ALR) would increase online shopping usage.

The literature indicates, perhaps not surprisingly, that the economy has significant impacts on e-shopping (Farag et al., 2007; Pick and Azari, 2011). Empirical research of different countries revealed that the local income is one of the most critical determinants of online shopping (Casas et al., 2001; Farag et al., 2006; Sim and Koi, 2002). For example, Clarke et al. (2015) indicated that the wealthiest households are 10 times more inclined to shop online than lower income households. In China, higher income individuals are better able to afford the costs of internet use (Song et al., 2014). Consequently, our next hypothesis is:

H4: Residential income (RIN) is expected to increase the ratio of online shopping.

Geographical factors

Anderson et al. (2003) established two hypotheses for the relationship between e-commerce and transportation. On the one hand, the innovation hypothesis states that urban residents with high transport accessibility are more likely to shop online (Weltevreden, 2007; Farag et al., 2006). On the other hand, the efficiency hypothesis states that individuals who live in low accessibility areas prefer e-shopping due to the spatiotemporal compression function of the internet (Weltevreden and Van Rietbergen, 2009). To date, some studies have offered evidence for these hypotheses, while they generated mixed results on the impacts of transport accessibility on e-shopping (Rotem-Mindali and Weltevreden, 2013). In China, there is an increasing difference in traffic accessibility between counties, which may impact the local e-shopping adoption. Therefore, the following hypothesis is proposed:

H5: Transport accessibility (TAC) has significant impacts on online shopping.

With the development of online shopping, urban logistics systems have transformed in the past two decades (Sakai et al., 2020). Previous research has focused on characterizing and modeling logistics facility systems, and home delivery and end-delivery movements of e-commerce (Gonzalez-Feliu et al., 2012; Punakivi and Saranen, 2001; Taniguchi and Kakimoto, 2003). Recent studies are more focused on e-commerce delivery systems, such as the deployment of pickup point networks, reception point delivery systems, proximity reception points, and site location distribution (Morganti et al., 2014). It is argued that in the near future, the deployment of drop-off and collection schemes will significantly affect local logistics systems and their competitiveness (Morganti et al., 2014). Therefore, for e-commerce delivery systems, we posit,

H6: The number of drop-off and collection points in local delivery system (DCP) is expected to increase the ratio of online shopping.

The residence location, such as whether or not located in an urban area (Zhen et al., 2018), has an important role in e-shopping usage. There are urban–rural differences in internet use and online behavior worldwide (Song et al., 2020), and researchers found that urban residents tend to shop online more likely (Sener and Reeder, 2012). For example, Farag et al. (2006) found that in the Netherlands, people living in urban areas are more inclined to shop online than people in less urbanized areas. In China, large metropolitan areas tend to have a higher level of internet adoption, while rural-mountainous regions have a relatively low internet usage (Fong, 2009; Song et al., 2020). Therefore, the relationship between urbanization and online shopping adoption is hypothesized as follows.

H7: The ratio of urbanization (URB) would enhance e-shopping adoption.

Many studies have focused on the relationship between online shopping and physical shopping. For example, Mokhtarian (2004) analyzed the transportation impacts of e-commerce and stated that e-shopping could replace, generate, or modify shopping trips, which may occur simultaneously. Farag et al. (2007) found that according to shopping trip frequencies, online shopping and physical shopping tend to complement or generate each other. Weltevreden and Van Rietbergen (2009) distinguished four impacts of e-shopping on in-store shopping, i.e., substitution, complementarity, modification, and neutrality. Moreover, recent research suggests that it should be focused more on the complex relationships between e-shopping and in-store shopping, as these relationships (such as substitution, generation, complementarity, modification, and neutrality) may co-occur (Cao, 2012; Zhai et al., 2019; Zhen et al., 2018; Zhou and Wang, 2014). Given the fact that the relationship between e-shopping and physical shopping is mixed, this study aims at measuring the impacts of physical retail shopping in two dimensions, retail volume and retail accessibility. Thus, the following hypotheses are proposed:

H8A: The retail sales per capita (RET) will be associated with e-shopping adoption.

H8B: Physical retail store numbers (PRT) is expected to influence online shopping.

Internet factors

A better internet connection could improve the adoption of e-shopping (Liao and Cheung, 2001; Swinyard and Smith, 2003). The internet can provide boundless shopping opportunities for every e-shopper, even for the internet users in non-urban regions, where retail opportunities are rather scarce. With the gradual popularization of smartphones in the last two decades, mobile phones and personal computers have been the main equipment to access the internet. According to CCNIC (2020), China is the largest nation in terms of the number of internet and mobile phone users, with approximately 904 million internet users and 897 million mobile internet users in March 2020. Consequently, we put forward the following hypotheses:

H9A: Internet penetration (INT) has a positive impact on online shopping.

H9B: Mobile cellular penetration (MOB) is expected to improve online shopping adoption.

Data and methods

Data

The online shopping data at the county level came from the Alibaba group and are not publicly available. Alibaba is China’s largest e-commerce retailer, including the Taobao and Tmall platforms. In 2019, with 654 million active users, the Alibaba online marketplace reported a total sale of over 5.73 billion RMB, accounting for 58.87% of the total online retail sales in China. The online shopping index (SI) is built by Aliresearch, which can intuitively reflect the development level of online shopping in the counties. Its value ranges from 0 to 100, with a larger value indicating a higher development level of online shopping. The calculated formula is as follows

SI = 0.6 \times α + 0.4 \times β

(1)

where SI represents the online shopping index; α represents the online shopping density index, which is the ratio of the number of online shopping consumers to the population；β represents the online shopping consumption level index, which is the ratio of the amount of online consumption to the number of online shopping consumers.

This study collected the data of factors from different sources (Table 1). Data for the 1918 counties in mainland China (not including Hong Kong, Macao, and Taiwan) were collected from the China County Statistical Yearbook and the Statistical Report on Internet Development in China (China Internet Network Information Center (CNNIC), 2020). Retail sales data were collected from China's regional economic and social development database. The data of physical retail store were collected from the POI (point of information) data websites of fourteen shopping service places through web crawler; logistics outlet data were obtained from the crawler data of the “express 100” website. Transport accessibility was measured by the density of the transportation infrastructure in each county, which included many layers of grid data, referring to the density of railways, highways, and water transport channels. Finally, transportation infrastructure data were collected from the database of the Institute of Geographical Sciences and Natural Resources Research.

Table 1. Definition of independent variables.

Category	Variable	Code	Source	Year
Sociodemographic	Proportion of youth population	YON	CCSY	2017
Sociodemographic	Gender (Female/male)	GEN	CCSY	2017
Sociodemographic	Adult literacy rate	ALR	CCSY	2017
Sociodemographic	Residential income	RIN	CCSY	2017
Geographical	Transport accessibility	TAC	IGSNRR	2018
Geographical	Drop-off and collection points in local delivery systems	DCP	EWEB	2018
Geographical	The ratio of urbanization	URB	CCSY	2017
Geographical	The retail sales per capita	RET	CRESD	2018
Geographical	Physical retail store numbers	PRT	POIW	2018
Internet	The internet penetration	INT	CSRIN	2017
Internet	Mobile cellular penetration	MOB	CSRIN	2017

CCSY: China County Statistical Yearbook; IGSNRR: Institute of Geographical Sciences and Natural Resources Research; EWEB: Homepage of “express 100” website; CRESD: China's regional economic and social development database; the POI data websites of fourteen shopping service places; POIW: The POI data websites of fourteen shopping service places; CSRID: Statistical Report on Internet Development in China.

Methods

We first mapped the dependent e-shopping variable using the geographical information system (GIS) software (Figure 2) with cartogram techniques to provide useful visual clues for spatial patterns of the dependent variable. Then, we applied a geographically weighted regression (GWR) model to analyze the determinants of the SI value based on fourteen independent variables.

Figure 2. Cartogram map of the spatial distribution of e-shopping across 1918 counties in China, 2017.

The GWR model is an improvement of the traditional regression model, which can reflect the effects adjacent points have on the characteristic value of a certain unit (Song et al., 2020). In the GWR model, we used SI as the dependent variable and eleven factors as input variables: the youth population ratio (YON), gender (GEN), adult literacy rate (ALR), residential income (RIN), transport accessibility (TAC), the number of drop-off and collection points in local delivery system (DCP), the ratio of urbanization (URB), the retail sales per capita (RET), physical retail store numbers (PRT), internet penetration (INT), and mobile cellular penetration (MOB). Generally, the regression model is

y_{i} = a_{0} + \sum_{j = 1}^{n} a_{j} x_{i j} + b_{i} i = 1, 2, \dots m; j = 1, 2, \dots n

(2)

where

x_{i j}

represents the jth observation value in county C_i; y_i represents the SI for the explained variable;

b_{i}

is a random error term; and

a_{0}

is a constant.

The GWR model uses each point’s distance weight for a local linear regression and uses the distribution of output parameters to determine the spatial heterogeneity of each unit. As a result, GWR can obtain the regression coefficient of each factor without geographical bias.

Spatial characteristics of online shopping in China

As shown in Figure 2, there were enormous spatial differences of the SI values among the counties in China. The SI values in Yiwu, Kunshan, Bixian, and Shishi in Eastern China were all higher than 23, nearly four times of the national average. However, in Jiuzhi in Qinghai Autonomous Region, Dongxiang in Gansu Province, Nanmulin and Andren in Tibet Autonomous Region, and Tulufan in Xinjiang Autonomous Region, the SI values were less than 1, only one-fifth of the national average. In terms of the spatial distribution, most counties with high SI levels agglomerated in the eastern coastal areas of China, while the counties with relatively low SI values were mainly located in the densely populated areas in Southwestern China and the thinly populated areas in Northeastern China. There was a prominent strip of high SI value counties extending from the east coast region to the northwest region. However, we find a few high SI value counties scattered in the inland of China, such as Xinzheng in Henan Province, Qinghe in Hebei Province, Xilinhot in Inner Mongolia Autonomous Region, and Houma in Shanxi Province (Figure 2).

More specifically, we identified six distinctive geographical clusters of counties, based on the SI value. The highest level cluster contained 26 counties, the ratio of the cluster with the lowest value being 6.69:1. Among them, 22 counties were concentrated in the eastern coastal region of China, including Zhejiang, Jiangsu, Fujian, and Guangdong Provinces. For example, in 2018, among the top 20 e-shopping counties in China, 13 were located in the Yangtze River Delta, 4 were located in the West Coast Economic Zone of the Taiwan Straits, and 3 were located in the Pearl River Delta. The high-level cluster contained 167 counties, with SI values between 8.01 and 13.95. There were mainly two kinds of counties in this cluster. The first kind was situated on the periphery of the highest counties, which were also concentrated in the southeast coastal provinces, such as Zhejiang, Jiangsu, and Fujian Province. The second kind of county in this cluster were counties around the provincial capitals in north China and the southwest provinces. The medium-high level cluster contained 402 counties, with SI values between 5.16 and 8.01. We identified three kinds of counties in this cluster. Some counties were located in the west of large metropolitan areas in eastern China, such as Anhui, Jiangxi, and Hunan Province. Some counties were concentrated in Central and North China. Furthermore, some counties scattered in the northeast border area, northwest area, and the southern edge of the Qinghai Tibet Plateau. The medium-level cluster contained 753 counties, with SI values between 3.28 and 5.16. The spatial distribution of these counties was relatively scattered, mainly distributed in the central, northeast, and southwest regions of China, especially in the southern margin of the Qinghai Tibet Plateau. The medium-low & low-level clusters contained 570 counties, with SI values lower than 3.28. These counties were mainly distributed in northwest, southwest, and northeast China. There were still some counties with an SI value lower than 1, such as Dongxiang in Gansu Province, Jiuzhi in Qinghai Autonomous Region, Nanmulin and Andren in Tibet Autonomous Region, and Tulufan in Xinjiang Autonomous Region.

Overall, the online shopping levels in China displayed a spatial pattern that decreased progressively from the core zone of the eastern coastal regions to the rural-mountainous regions in west China. Moreover, we found that an effect of urban agglomeration was apparent (Song et al., 2020), as the SI values were higher in metropolitan areas, such as in the Yangtze River Delta.

Results and discussion

Due to the data limitation, we only analyzed a sample of 1862 counties with GWR. Table 2 shows that the rate of urbanization (URB), the local delivery system (DCP), internet penetration (INT), retail sales per capita (RET), and residential income (RIN) were the factors most associated with the SI. Some of these findings were supported by previous online shopping and e-commerce literature, but others revealed new information.

Table 2. Regression results of the dependent variables for 1812 counties, China, 2017.

Hypothesis	Variable	Code	Coefficient	VIF(Variance inflation factor)
H1	Proportion of youth population	YON	0.206	2.450
H2	Gender (Female/male)	GEN	−0.014**	1.405
H3A	Adult literacy rate	ALR	0.069*	1.248
H3B	Secondary gross enrollment ratio	SER	0.050	1.690
H3C	Tertiary gross enrollment ratio	TER	0.135**	1.582
H4	Residential income	RIN	0.300***	1.011
H5	Transportation accessibility	TAC	0.064**	1.185
H6	Local delivery system	DCP	0.378***	2.552
H7	Rate of urbanization	URB	0.492*	2.194
H8A	Retail sales per capita	RET	0.306	2.784
H8B	Physical retail store numbers	PRT	0.052*	3.234
H9A	Internet penetration	INT	0.324*	5.499
H9B	Mobile cellular penetration	MOB	−0.027	5.004

YON: youth population ratio; GEN: gender; ALR: adult literacy rate; SER: gross secondary education; TER: gross tertiary education; RIN: residential income; TAC: transport accessibility; DCP: the number of drop-off and collection points in local delivery system; URB: the ratio of urbanization; RET: the retail sales per capita; PRT: physical retail store numbers; INT: internet penetration; MOB: mobile cellular penetration.

*Significant at 0.05, **Significant at 0.01, ***Significant at 0.005.

Sociodemographic factors

Youth population

The regression results show that demographic factors were no longer the most important determinants (Table 2). The proportion of youth population (YON) had a relatively positive effect on the counties’ SI value, with a correlation coefficient of 0.206. Because of the technical and operational barriers in the process of online shopping, such as goods search and electronic payment, older people are less likely to be e-shoppers (Beilock and Dimitrova, 2003; Casas et al., 2001; Fuchs, 2009 ). According to CNNIC (2020), in March 2020, the proportion of internet users under the age of 40 in China was 65.6%. Today, accessing the internet has become a regular activity for most young people (Clarke et al., 2015). The Statista report shows that 62% of the 16- to 24-year-old people shop online at least fortnight, compared to only 29% of those aged over 65 (Statista, 2020). In China, the youth population has stimulated the development of e-shopping in the last two decades. Age is still a related discriminator in terms of e-shopping, with people under 40 being more inclined to buy online (Clarke et al., 2015; Farag et al., 2007; Vrechopoulos et al., 2001).

Gender

The correlation coefficient between gender and the SI was –0.014, which indicates that gender was not a notable determinant of e-shopping adoption. This result is in line with those of Clarke et al. (2015) and Lee et al. (2017), who reported a minimal to no gender effect. According to CNNIC (2020), in March 2020, the ratio of male and female internet users in China was 51.9:48.1. Although some new information technologies can still pose barriers to women’s participation (Selwyn, 2004; Song et al., 2020), women are increasingly involved in e-shopping.

Education factors

Although all education factors had positive effects on the SI value in the counties, the impacts were small and not significant. The existing literature shows that more highly educated people have more e-shopping experience than less educated people (Casas et al., 2001; Farag et al., 2007; Pick et al., 2013). Our GWR results were partly supported by previous studies, which reveal that in China, gross tertiary education (TER) has a notable impact on the e-shopping gap, while gross secondary education (SER) and adult literacy rate (ALR) had much lower impacts. However, with the rapid development of smart phones and the popularity of e-commerce, more and more people with basic literacy can shop online in China (Song et al., 2020; Yang et al., 2013). As a result, education has become a less considerable factor.

Income factors

Corresponding to previous studies, the GWR results show that the income level had significant positive impacts on online shopping in China. Households with a higher income are more inclined to buy online, and their single purchase amount is much higher (Clarke et al., 2015; Farag et al., 2006; Sim and Koi, 2002). According to interview records with e-retailers in the Alibaba group, the highest income households had the strongest e-retail preferences. Although there could be some interaction with the age variable (Clarke et al., 2015), it could be confirmed that the income variable (RIN) was among the most important drivers of the e-shopping development in China’s counties.

Geographical factors

Transport accessibility

The results showed that transport accessibility (TAC) did not have a significant impact on the counties’ SI value, which was entirely unexpected. This may be partially related to the concept and method we applied to measure transport accessibility, which has been defined and operationalized in many different ways (Dewan and Riggins, 2005; Nassir et al., 2016). We applied the most common variable from the existing literature, like travel time, which was estimated based on simplifying assumptions considering road density, speed limits, and route frequencies (Dewan and Riggins, 2005; Mavoa et al., 2012; Nassir et al., 2016). On the other hand, most of the time, physical retail sales were weaker than online shopping sales in counties with poor transport accessibility. The entry cost of physical retail sales is much higher in counties with low transport accessibility, which would promote the local residents’ choice of online shopping. Furthermore, in recent years, there has been a boost in online service consumption in China, with its growth rate being 1.46 times higher than that of physical consumption. Moreover, online service consumption has little demand for regional accessibility.

Delivery system

As shown in Table 2, drop-off and collection points in local logistics systems (DCP) are one of the key determinants of the development of the counties’ online shopping. This is supported by previous studies, which indicated that the efficiency and effectiveness of logistics systems were critical for e-commerce (Cho et al., 2008; Ramanathan, 2010). According to CNNIC (2020), in 2019, 80.15% (i.e., 8.52 trillion RMB) of the total online retail sales in China were physical goods, which need delivery service. This enormous market has stimulated the boost of advanced logistics services in China (He et al., 2019). In 2019, there were a total of 63.52 billion parcels delivered by express companies in China, among which more than 70% were generated by e-commerce (China National Statistic Bureau, 2020). Besides third-party logistics, more and more e-retailers begin to build self-run logistics, such as Amazon in the USA, and Tmall.com (under Alibaba), JD.com, and Suning.com in China. However, e-retailers are still being challenged to offer their customers the ever-more responsive last-mile delivery services, which should be affordable and reliable (Dutta et al., 2020; Giuffrida et al., 2017). The spatial distribution of local drop-off and collection points is the most important part in last-mile logistics, which is one of the “most inefficient and most expensive” part of the e-commerce delivery system (Dutta et al., 2020; Fernie et al., 2010; He et al., 2019). Consequently, with the development of e-commerce, advanced drop-off and collection points in local logistics services play an increasing role in online shopping adoption.

Urban factors

The regression results show that the ratio of urbanization (URB) was the most crucial determinant of e-shopping usage in China, which offered evidence for the innovation-diffusion hypothesis (Anderson et al., 2003). Some research suggests that e-shopping is a predominantly urban phenomenon because it often starts in the centers of innovation (Clarke et al., 2015; Farag et al., 2006). In contrast, some studies argued that there was no significant or even a negative relationship between e-shopping and urban areas (Clarke et al., 2015; Krizek et al., 2005). It is an interesting debate as to whether we should expect urban to be higher than rural e-shopping adoption. Due to the urban–rural differences in information infrastructure and internet use in China, rural consumers may face more significant accessibility problems in shopping online and home delivery. According to CNNIC (2020), in March 2020, there were 649 million urban internet users in China, accounting for 71.8% of the total internet users; while the number of rural internet users was 255 million, accounting for 28.2% of the total internet users. However, the rise and diffusion of broadband technology in rural areas has been much faster than that in urban areas in the past five years. In March 2020, the digital divide of urban–rural internet penetration has reduced by 5.9%, as compared to the end of 2018 (CNNIC, 2020). Thus, another consideration here is that the location of residence could become less important in the future.

Physical shopping

We found that the relationship between online shopping and physical shopping is “cross-mode.” On the one hand, in-store shopping tends to complement or generate online shopping, with the correlation coefficient of retail sales per capita being 0.306. Shoppers seem to switch equally between physical shopping and e-shopping, or physical shoppers become almost exclusively e-shoppers (Clarke et al., 2015). In China, shoppers may make extensive use of both modes and often search online before purchasing at physical stores, which has been indicated in some existing literature (Farag et al., 2006; Ferrell, 2004; Longley et al., 2008). On the other hand, in-store shopping modifies e-shopping or has no effect at all, with the correlation coefficient of physical retail store numbers to the SI being 0.052. Shoppers at a physical store may not be interested in e-commerce or modify their behavior to include infrequent e-shopping. The absence of an obvious relationship between e-shopping likelihood and shopping center accessibility is also supported by previous studies (Krizek et al., 2005; Ren and Kwan, 2009).

Internet factors

The internet penetration variable (INT) had significant positive effects on the SI scores among counties in China, while mobile-cellular penetration (MOB) showed only a slight negative influence on the SI score. Internet accessibility is the basis for online buying, which refers to internet access, internet skill, and internet use. However, there is still a spatial gap in internet access and internet use in China (Fong, 2009; Pick et al., 2013; Zhu and Chen, 2016). According to CNNIC (2020), there is still an enormous digital divide in the internet access among the counties in China. In March 2020, there were 449 million fixed Internet broadband access users, in which the number of users with an access rate of 100 Mbps and above was 384 million, accounting for 85.4% of the total users; the number of users with an access rate of 1000 Mbps and above reached 870 thousand. With the spatial gap in internet penetration in China, the internet variable is still one of the main determinants of e-shopping adoption.

In the early stage of e-commerce development, the local mobile cellular penetration is a major driving factor in China (Yang et al., 2013). However, in the last decade, with the popularity of mobile phones, the influence of mobile cellular penetration (MOB) on e-shopping has gradually declined. This may be partially explained by the fact that the decline in handset prices and the rise of China-made mobile phones have almost eliminated the digital divide of mobile phone usage in China. However, we should notice that mobile phones do not refer to smartphones, as some mobile phones used in rural China are cheap and without support for internet access. According to CNNIC (2020), at the end of 2019, there were 1.6 billion mobile phone users in China, with a mobile phone penetration of 114.14%. In March 2020, 99.3% of the internet users in China used mobile phones to access the internet. As the mobile phone has become a daily necessity for most inhabitants in China, the mobile phone variable is no longer a notable driver for e-shopping usage.

Conclusions

China is one of the world’s most significant countries in terms of e-shoppers and online retail sales (Song et al., 2020). However, there is a large digital divide of e-shopping in China, which could put enormous pressure on traditional retailers and have significant ramifications for economic geography. This is also a broader topic for geographers and planners because online shopping has already affected the retail economy, consumption landscape, land use patterns, and regional environment (Miguel and Anmol, 2020; Song et al., 2020; Yang, 2017). In this study, we validated the online shopping gap in China and explored its major determinants.

The findings demonstrate a generational gap of e-shopping in China, which has been discussed in the existing literature. At the county level, we also found that counties in the eastern coastal areas had relatively high SI values, which spread from the eastern coastal areas to the western regions, and from the metropolitan areas to the rural-mountainous regions.

The GWR results show that spatial attributes had significant effects on the online shopping landscape in China. The key drivers for the e-shopping gap were found to be the rate of urbanization (URB), the local delivery system (DCP), and internet penetration (INT), while retail sales per capita (RET) and the residential income (RIN) had significant positive effects on the counties’ SI value, and the gender (GEN) had negative effects on the SI. Previous studies indicated that age and higher education are essential discriminators in terms of e-shopping, but our results indicate that age and higher education are no longer the most important determinants. We also found that spatial attributes, such as drop-off and collection points in the local logistics system (DCP), the ratio of urbanization (URB), and retail sales per capita (RET) had significant positive effects on e-shopping. Furthermore, the influence of gender and basic education on the e-shopping divide tended to be very small and insignificant.

As contrary to some e-shopping studies, our results indicate that geographical factors are more important for the e-shopping development than sociodemographics. According to the results, the local governments should pay more attention to improve the internet connection, such as internet penetration, mobile cellular penetration, and even 5G in the future. The local governments should also optimize the drop-off and collection points in local logistics systems, which could help to ensure a timely delivery of online shopping goods. Strengthening long-standing investments in the physical retail infrastructure would be another positive policy, that would replace, generate, or modify e-shopping.

We should note that a weakness of the study is that there may be differences between the Alibaba county online shopping index (SI) and the real county online shopping level. Although Alibaba is China’s largest e-commerce retailer, there are still about 9.2% of e-shoppers who do not use e-retailers of the Alibaba group. More and more e-retailers, such as Jingdong, Pingduoduo, Suning, and Amazon, have provided a diversified choice for e-shoppers in China. On the other hand, the impact of spatial attributes on e-shopping can vary across product types (Zhen et al., 2018), such as physical goods and digital goods. However, we did not have access to online shopping data on different product types, which should be discussed in the future via interviews of e-shoppers.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was primarily funded by Program of the Natural Science Foundation of China (No. 41871120 and No. 41371006).

ORCID iD

Zhouying Song https://orcid.org/0000-0001-8771-3633

References

Anderson WP, Chatterjee L, Lakshmanan TR (2003) E-commerce, transportation, and economic geography. Growth and Change 34(4): 415–432.

Abstract

Introduction

Theory and hypotheses

Sociodemographics

Geographical factors

Internet factors

Data and methods

Data

Methods

Spatial characteristics of online shopping in China

Results and discussion

Sociodemographic factors

Youth population

Gender

Education factors

Income factors

Geographical factors

Transport accessibility

Delivery system

Urban factors

Physical shopping

Internet factors

Conclusions

Declaration of conflicting interests

Funding

ORCID iD

References

Biographies

Cite article

Cite article

Download to reference manager

Share

Share this article

Share with email

Share on social media

Share access to this article

Information

Published In

Keywords

Rights and permissions

Authors

Affiliations

Notes

Metrics

Journals metrics

Article usage*

Articles citing this one

Figures & Media

Tables

View options

PDF/ePub

Get access

Access options

Sign in

Also from Sage

Article usage^*