A longitudinal study of e-commerce diversity in Europe

Based on a review of the literature comprising 174 articles on Operations Research and Logistics, Giménez and Lourenço [19] define E-SCM “as the impact that the Internet has on the integration of key business processes from end user through original suppliers that provide products, services, and information that add value for customers and other stakeholders”. They underline that, along with positive results such as reduced costs or increased supply chain process efficiency, the implementation of Internet technology brings about certain concerns associated with culture and technology.

Mahmood et al. [30] conducted a cross-country study exploring factors significantly positively impacting on online shopping behaviours. The scope of data used in the study included 37 countries with 42,238 interviewed individuals derived from the “TNS Interactive-Global eCommerce Report”. The results obtained led to the conclusion that online shopper behaviours are actually affected by interpersonal trust and e-shoppers’ economic situation. Results showed that an increase of 1 percent in the standard deviation in the trust index, for example, was associated with an increase of 1.86 percent in on-line shopping. For the economic situation, the rise in online shopping was even higher—at 2.69 percent. Interestingly, there was no significant relationship between technological practical understanding and behaviours connected with online shopping. It is also worth mentioning that the study drew attention to the influence of culture on online shopping as a factor playing an extraordinary role in e-business, which requires in-depth research. E-commerce has accordingly a complex and multifaceted structure. Many combinations of factors influence e-commerce diffusion. This includes considerations of its barriers to adaptation limited in analyses to particular countries and regions.

Other examples of e-commerce diversity include Couclelis [14] who examined one particular soft effect of information and communication technologies of special interest to geography and regional science, related to the modification of individual activity patterns and the modification of a person’s activity space associated therewith. She put forward a hypothesis on activity disintegration assuming that “it is not distance that is dead; it is activity that is disintegrating”. The hypothesis presumes a potential spatial and temporal disintegration of activity related to shopping in the context of substituting other activities such as, for example, professional work. The research leads to the conclusion that, thanks to ICTs, spatial and temporal boundaries of customer behaviours may change so radically that traditional e-shopper behaviour assumptions can no longer be applied. Hernández et al. [21] analysed perceptions leading customers to shop online in Spain. In the study, they assumed the Technology Acceptance Model (TAM) and included three factors, i.e. Internet acceptance, frequency of use and satisfaction. The results of their research lead to the conclusion that Internet acceptance does not affect the frequency of Internet use in the context of online shopping. It was also emphasised that the characteristics of a country, from both the economic and social perspectives, should also be analysed regarding e-shopper behaviours. In respect of the social perspective, cultural differences are of key importance, assuming the form of two cultural e-purchasing characteristics: authority distance and uncertainty avoidance in the case of Spain.

Kauffman and Kumar [24] conducted a study aimed at determining a relationship between ICT readiness (e-readiness) and trade flows as well as R&D. The results indicate a positive relationship among the studied factors, although the strength of the relationship depends on the country. The impact of ICTs on different parameters is simultaneous, e.g. wider Internet accessibility provides greater opportunities for e-commerce, but also for democracy development and health awareness. The results also confirm that the Internet use variable is significant solely for commerce and ICTs. It should be highlighted that the Internet facilitates communication and shopping, making them faster and, in many cases, more convenient. However, the results of research performed by Çikrıkci [12], applying the method of meta-analysis, confirmed the occurrence of a negative relationship between Internet use and well-being. The study assumed the components of well-being to be life satisfaction, psychological well-being and self-esteem, which may significantly vary among countries and regions depending on socioeconomic, demographic and ethnic structures.

Results of research carried out in the United Kingdom prove that age and income are crucial demographic discriminators of e-commerce usage, as is rural versus urban location, and distance from physical stores [13]. Conclusions drawn from their study indicate that, although e-commerce diffusion began in big cities such as London, it is currently not a trend limited solely to cities. There is evidence of growing e-commerce use in rural areas derives from increasing high-quality broadband Internet access.

Smith et al. [41] conducted a cross-cultural exploration of online shopping behaviours, comparing Norway, Germany and the USA, and juxtaposing common views on increasing similarities in cross-national customer demand preferences and shopping behaviours with the fact that they represent distinct cultures in terms of commonly examined frameworks. Their conceptual model was grounded in the technology acceptance model (TAM) [16]. The survey was administered to undergraduate business students from three different countries—Norway, Germany and the USA. The research results emphasise the pervasiveness of cultural influences upon consumer behaviour. Germans and Norwegians appear to base their use of online shopping entirely on perceptions that online shopping enables them to accomplish tasks that cannot be done as effectively in traditional retail settings. This constitutes a utilitarian approach to online shopping treated as facilitating the process of acquiring goods and services [11].

In research into omnichannel (coopetition between various supply chains) customer behaviours, Juaneda-Ayensa, Mosquera and Murillo [23] applied the UTAUT 2 (Unified Theory of Acceptance and Use of Technology) model with two additional variables: innovativeness and perceived security, to test 628 Spanish customers of the Zara shop, referred to as customers 3.0., i.e. shopping online along with taking advantage of traditional forms of sales. Their results indicate that personal innovativeness is the strongest predictor of a shopping intention in the omnichannel context. At the same time, they proved that perceived security did not influence omnichannel shopping intention, which is key evidence of the possibility to extend e-commerce to include more conservative customers who perceive online shopping as a high-risk activity.

Shi, Chen and Sethi [40] conducted a study aimed at examining the possibility of forecasting the repetitiveness of online shopping and determining the heterogeneity of customers using general distribution. In their research, they assumed a typical Online-to-Offline (O2O) model, differing from the traditional Business-to-Consumer model (B2C) in the inclusion and coordination of deliveries by third party logistics. The proposed stochastic model of predicting future spending by e-customers considered solely historical purchase transactions, without knowing customers’ demographic data and preferences.

Onstein et al. [36] included the spatial layout of freight transport and storage systems used to move goods between production sites and consumption places in e-commerce research. They identified three important research streams: Supply Chain Management (SCM), Transportation and Geography, based on the Systematic Literature Review (SLR) method, distinguishing modelling methods as well as strengths and limitations of each research stream. The contribution of the study is the identification of relationships among seven groups of factors affecting the distribution model adoption. These groups of factors are: (1) demand level, (2) service level, (3) product characteristics, (4) logistics costs (transport costs, inventory costs and warehousing costs), (5) labour and land availability, (6) accessibility and (7) contextual factors.

Basing their research on a sample of 562 Swedish retail firms using multiple regression analysis, Tolstoy et al. [44] revealed that actions consisting of adapting sales channels have a positive impact on international sales in online channels. This enabled generating new insights into how the geographical export scope negatively moderates the effect that international online sales channel adoption has on international sales.

Cárdenas et al. [10] examined the issue of the e-commerce last mile in Belgium from the perspective of external costs of delivery. Their research results indicate that e-commerce consumption per capita is higher in rural areas, which means that individuals living outside urban areas buy more from online distribution channels. At the same time, in respect of the number of kilometres connected with e-commerce freight transport, the results obtained indicate that urbanised areas absorbed over a half of deliveries in the country, significantly contributing to considerable CO₂ emissions. The spatial distribution of deliveries unambiguously indicates that the majority of e-commerce deliveries in Belgium are directed to three towns: Brussels, Antwerp and Ghent located in the North-Eastern part of the country.

Wu and Lin [46] employed unstructured data analytics for retrieving valuable business models of e-commerce logistics comprising 2370 documents for the period 2010–2016 from professional websites dedicated to e-commerce, using web crawling using Python programming. The results of this topic mining identified eight main topics regarding e-commerce logistics, one of which refers to the spatial and temporal diversity of e-commerce in Europe: Europe e-commerce trends. This indicates that Finns prefer bank transfer payments and Danes—credit card payments, while online shoppers in the Netherlands, Spain and Germany consider free shipping and delivery within 3–5 days to be important requirements. The research also proves that cultural differences to customising appropriate technologies of e-commerce logistics could increase the likelihood of the acceptance of innovative technology.

Research presented and discussed in this section reveals how the diffusion of E-commerce and reducing the existing divergence are strongly influenced by the presence of the online commerce distribution chain, which allows the high utility of this form of buying goods and services by households. The distribution chain configuration directly affects the logistic service level, subject to different expectations in particular countries and regions.

Research into e-commerce spatial diversity has been dominated by approaches describing the phenomenon at the national level [4]. The approach of e-commerce spatial diversity analysis considering regions has been applied to a lesser extent. This also applies to the regional e-commerce approach, which is most commonly presented from the perspective of regional diversity within a single country [38]. As stated by Perrons [39], the new economy is characterised by globalisation whose inherent trait is the increasing use of information and communication technologies. Positive globalisation effects are compensated for by the widening digital divide between those who are active knowledge economy participants and those who are digitally excluded. Research results indicate that workers in the new economy can be organically connected and simultaneously part of the broadening digital divide concerning different spatial scales. By extending conclusions related to new media clusters researched by Perrons to include e-commerce assumes the dispersed occurrence of similar e-retail clusters.

Due to the lack of physical boundaries in Europe and the ease of handling e-customers from any region, shopping patterns intertwine. This is also associated with a new approach to the location of logistics facilities designed to provide services to the European e-retail sector from the point of view of the availability of regions, not countries, as in the past. This also applies to the reconfiguration of supply chains and their adaptation to the requirements of e-customers in e-retail sales channels [48].

The growing importance of the regional approach to e-retail, supply networks and e-customer behaviour leads to a new way of presenting, analysing and assessing phenomena that were previously treated as exclusively national. Billón et al. [6, 7] in research on the spatial distribution of the Internet in EU firms, they showed the influence of the spatial effect on ICT adoption by firms in the enlarged EU. The obtained research results in the form of the determination of clusters of regions with similar characteristics are significant from the point of view of cohesion policy and the development of the information society in Europe. Similar results [8] were obtained in the spatial analysis of the distribution of ICT use in European regions. Research has confirmed sharp disparities in the level of Internet use in households at a regional level in Europe. Nevertheless, based on these results, the existence of a division between the north–south and east–west regions cannot be unequivocally confirmed.

Lutz [29] performed research applying the Exploratory Spatial Data Analysis of the Digital Single Market in EU countries, excluding Turkey and Iceland. It proved the occurrence of, both globally and locally, strong spatial autocorrelation and the north–south polarisation pattern. In the years 2011–2016, the digital divide remained very stable and lasting, which was noticeable in the case of Internet use and e-commerce. Spatial patterns of the e-commerce index and its rise indicate saturation in growth through low clusters around Belgium and Scandinavia, but not in the surrounding areas of London.

Lucendo-Monedero, Ruiz-Rodríguez and González-Relaño [28] performed research into inequalities in the digital development of households and individuals in Europe. The results indicate that the basis for development and reducing existing inequalities is broadband Internet access, which is nationally and regionally diversified. Mainstream research on the diversity of e-commerce and e-retail involves conducting comparisons at the national level and, to a lesser extent, in-depth analyses of regional diversity at the level of individual countries. At the same time, there is a scarcity of research focused on the regional presentation of e-commerce diffusion, which would assume that the existence of administrative divisions is not relevant to the development of e-retail. This approach is based on the assumption that there are clusters of regions similar to each other from the point of view of online shopping development, which are also associated with the level of broadband Internet access and the frequency of its use.

Zhang et al. [49] conducted an analysis of spatial patterns and factors affecting e-commerce in China. Results of their research prove that e-commerce development is associated with the phenomenon of the formation of spatial and temporal clusters. Fundamental social productivity and residents’ standard of living ought to be indicated among the principal factors positively impacting on this phenomenon. At the same time, the calculated Moran’s I statistics show that the geographical agglomeration of e-commerce in China is gradually weakening [15]. Another conclusion for China is the fact that e-commerce development differences between regions began to shrink.

The occurrence of spatial inequalities in access to the Internet was also confirmed in research by Sujarwoto and Tampubolon [42], who tested two hypotheses: normalisation and stratification in developing countries on the example of Indonesia in the years 2010–2012. Their results indicate that, although a continuous and significant increase in telecommunication technologies’ diffusion in the country occurred in the final 10 years of the study period, digital media development was spatially unequal. This confirms the stratification theory referring to the occurrence of divergence and the deepening of the divide in Internet use in the cross-section of socioeconomic groups, but also in other cross-sections, such as urban–rural or urban-suburban areas. The study provided no evidence of the reduction or closing of the divide according to the normalisation thesis in the study period.

In a questionnaire survey carried out in France, Motte-Baumvol, Belton-Chevallier and Dablanc [34] put forward two research hypotheses to be tested. The first one—the efficiency hypothesis—posited that people with low accessibility tend to buy more online. The second—the innovation-diffusion hypothesis—presumed that urban residents are more open to innovations and technology. The authors believe that, due to the increasing pace of spatial Internet coverage growth, the innovation-diffusion hypothesis should reduce spatial differences in online shopping in highly urbanised areas and areas with low population density, e.g. small towns and rural areas. They conclude that the Internet can be perceived as an additional space for shopping, and be included by households in their daily activities. They also assert that the propensity for online shopping is higher for suburban households in particular with those households being the main online shopping group. Furthermore, they note that online shopping considerably increases spatial accessibility with no need to travel to collect parcels.

Both e-retail and e-commerce diffusion are associated with an increase in Internet access. Changes in patterns of use as well as online customer experience in e-retail and technological know-how, which leads to an increase in e-trust and consumer confidence [1, 31, 32]. The spread of e-retail is associated with e-trust and is much more developed in countries and regions with high GDP and which are geographically, politically and economically stable [9]. They provided empirical evidence of this and, with the use of linear regression analysis, calculated that the correlation coefficient between IDI (Information and Communication Technologies Development Index) and GDP (Gross Domestic Product) is very high and amounts to 0.921. Although this reductionist approach provides knowledge about general dependencies between ICT diffusion and GDP, it fails to explain the causes of regional diversity, and does not take into account other factors (e.g. socioeconomic, demographic or cultural) affecting e-retail.

The absorption of ICT and the increase in broadband Internet access by European societies demonstrate a varied pace of change over time, and it is not imperative to affect Internet use to make online purchases. Approaching e-commerce with a focus on regional diversity is not clearly confirmed in previous studies of e-commerce. Some study of online shopping has focussed on the north–south scheme of division of Europe. Therefore, classifying clusters of e-commerce perception and use from the point of view of e-logistics and e-customer service helps fill a current research gap. It is vital to know the dynamics of changes in e-retailing in European regions in the years 2010–2019, especially from the standpoint of the convergence process.

The analyses of e-retailing in Europe were performed based on statistical data available at the European Statistical Office (the source of data was Regional Digital Economy and Society). In order to show the most complete picture of e-retail in Europe, in addition to the European Union countries (Austria, Belgium, Bulgaria, Croatia, Cyprus, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden and the United Kingdom), the analyses also included Iceland, Montenegro, Northern Macedonia, Norway, Switzerland and Turkey—i.e. 34 European countries. The main goal of the research was to analyse the diversity of e-retail in Europe at the regional level. Therefore, the most appropriate way to conduct the analysis seemed to be at the NUTS 2 level, which represents the basic regions for the application of regional policies. Nevertheless, due to the lack of statistical data on e-commerce at the NUTS 2 level for some countries (the largest data gaps occurred in the case of France, Germany, Poland, Turkey and the United Kingdom), the analyses were carried out for NUTS 1—major socio-economic regions—i.e. 117 spatial units.

The analysed variables describing e-commerce were:

Statistical data were collected for the years 2010–2019. Unfortunately, there were some small gaps in the database, especially in the case of France, where information for NUTS 1 was available until 2014. Due to the fact that an important element of the research was to analyse e-retail changes over time, in order not to reduce the sample scope (data for most spatial units have been available since 2010), the missing data were supplemented on the basis of trends.

Descriptive statistics for the variables are presented in Table 1.

In order to characterise the spatial structure of e-retail in European regions, basic methods of Exploratory Spatial Data Analysis, such as global and local spatial autocorrelation statistics, were used. Spatial autocorrelation can be defined as the coincidence of value similarity with locational similarity. Positive spatial autocorrelation occurs when high or low values of a random variable tend to cluster in space. In turn, negative spatial autocorrelation occurs when geographical areas tend to be surrounded by neighbours with dissimilar values [3].

Moran’s I statistic is the most commonly used indicator of global spatial autocorrelation. It was introduced by Moran [33], then adapted for the purpose of spatial analyses and popularised by Cliff and Ord [37]. In essence, it is a cross-product statistic between a variable and its spatial lag, with the variable expressed in deviations from its mean.where: n—number of observations; xi, xj—values of variable x in locations i and j;—mean value of xi observations; wij—elements of spatial weights matrix W; z—a vector which takes the form: \({\mathbf{z}} = \left[ \begin{gathered} z_{1} \hfill \\ z_{2} \hfill \\ \cdots \hfill \\ z_{n} \hfill \\ \end{gathered} \right]\), where the form: \(z_{i} = x_{i} - \overline{x}.\)

Values for Moran’s I statistic fall in the interval. Values above zero indicate the presence of positive spatial autocorrelation and below zero—the occurrence of negative spatial autocorrelation. In the event that there is no correlation between adjacent values, the expected value is close to zero. In order to verify hypotheses concerning spatial autocorrelation (H0: observed values of the variable are randomly distributed, hence there is no spatial autocorrelation between specific locations, H1: there is spatial autocorrelation), randomisation tests are performed.

In order to identify spatial relationships of a given variable in a specific location, local indicators of spatial association (LISA) developed by Anselin [2] are used. One of them is a local Moran’s I statistic. A characteristic feature of these measures is that the sum for all locations is proportional to the value of the global autocorrelation measure ΣiLi = γM. LISA provide a more detailed insight into the structure of spatial distribution of the studied variable by indicating, among others, statistically significant clusters of spatial units characterised by similar values of the analysed variable.