DYNA

2.1. Technology acceptance model

Since early 1970, numerous models have been proposed to understand and explain the factors that determine the acceptance of the use of information technologies. Some of them examine the relationship between the attitudes, perceptions, and beliefs of technology users and the level of use of the technology itself [21]. They include the following:

2.1.8. The Unified Theory of Acceptance and Use of Technology Model (UTAUT)

Due to the diversity of the technology acceptance models with similar theoretical foundations, initiatives to develop unifying models have emerged [34], including the UTAUT proposed by Venkatesh, Morris, Davis, and Davis [35]. The UTAUT was formulated from a conceptual and empirical synthesis of the above-mentioned models, providing a coherent theoretical perspective in the study of the adoption of online shopping [36].

The main concepts that group all of the adoption theories in the UTAUT are the following: performance expectation, expectation of effort, social influence, and enabling conditions; all of these influence the intention and use of technology behaviour [37] (Table 1).

Table 1: Theories unified in the UTAUT.

Source: Adapted from [69]

Since its original publication, the UTAUT has served as a guiding model and has been applied to the study of a variety of technologies in both organizational and non-organizational [37] contexts. According to [38], the UTAUT is a model that can be very useful in describing the adoption of e-commerce in its initial development because, in more mature developments, many concepts lose significance. In addition, there is a lack of studies that use this model in countries where e-commerce has these characteristics.

Therefore, the following model hypotheses are proposed (Fig. 1):

Figure 1: Proposed model from the UTAUT.

Source: The authors

2.2. Probing new moderating variables of purchase intention

The UTAUT model makes use of moderating variables such as gender, age, the mandatory use of computer equipment, and the previous experience of users [32]. However, the adoption of electronic commerce can also be affected by other factors such as socioeconomic status and educational level [47], especially in countries where the digital gap is high [48].

Socioeconomic status determines the purchasing power of buyers. It is important because it has been estimated that innovations traditionally break into society via subjects of high socioeconomic status [33], and although there are no studies that analyse the moderating effect of socioeconomic status on models of e-commerce adoption, it has been found that users of e-commerce tend to have higher income levels [47].

Educational level is another traditional marketing variable, just like socioeconomic status. Adoption studies [49-51] show greater levels of buyers with high levels of education [52-56]. In this same approach, other authors find that higher levels of education convey more information to shop online. Therefore, educational level may affect variables such as effort expectations and ease of use, and perceived risk versus electronic shopping may be reduced [47,57,58].

From the above, the following hypotheses are proposed,

3.1. Measuring tool

Because the goal is to compare two groups of Internet users differentiated in terms of being non-users and users of electronic commerce, as explained when we built the model, a unique analysis of the basic variables that affect the electronic purchase is presented to use the same evaluating tool for the two groups. To test the hypotheses of the proposed model, items acquired from previous literature were used, choosing the measurement scale adapted from the Spanish study for the general adoption of electronic commerce in Spain [42]. Similarly, to avoid comprehension problems with regard to the questions when using the language, a pre-test was performed using a group of 50 Internet users representing various cities in Colombia to evaluate each of the questions, without major changes in the proposed final tool (Appendix 1).

3.2. Sample and data collection

Because this is an exploratory study for the entire Colombian territory and given that the population in this country is mostly concentrated in large urban areas, it was decided to select Internet users in the major cities of the country, such as Bogota, Medellin, Cali, and Barranquilla, as the target population. Correspondingly, each city generated an area that geographically and culturally integrated other nearby cities. From this point, it was determined to create a quota sampling, taking into account the selection of proportional data from each area as a first condition. The subsequent guidelines were for gender to be equitable and to disseminate the survey to all socioeconomic statuses of the population to avoid any possible bias. (Tables 2, 3)

Table 2: Sample characteristics of Internet non-buyers.

¹ Adapted from the stratification used in Colombia according to public sources

Source: The authors.

Table 3: Sample characteristics of Internet buyers

¹ Adapted from the stratification used in Colombia according to public sources

Source: The authors.

The mechanism of data collection was performed through an electronic questionnaire validated in other studies for this type of research [40,42]. To facilitate the responses, the methodology that was followed entailed the use of a questionnaire with multiple items per construct; to provide their answers, the respondents used a Likert scale ranging from 1 (= “I strongly disagree”) to 7 (= “I strongly agree”) to effectively measure variables that are not directly visible [59].

The period for data collection occurred from November 2015 to May 2016. To achieve the objectives of the fieldwork, a national team with coordinators in each of the central cities was created to manage the distribution of the questionnaire. In addition, among the people who voluntarily responded to the survey, an incentive was offered, raffling a laptop that was given to the winner of the drawing held among all of the participants in June 2015 in Bogota and audited by the Gran Colombian Polytechnic University Institution (Institución Universitaria Politécnico Gran Colombiano). A total of 1,836 forms were obtained, of which 1,245 corresponded to online shoppers; 309 had consistency problems in the responses, and therefore, a final sample of 936 valid questionnaires for this group was obtained. Regarding the non-buyers, 875 forms were collected, and by eliminating 284 containing erroneous or inconsistent data, a final sample of 591 surveys for this group was obtained.

4.1. Validation of the measurement model

The validation process of the measuring instrument was performed from the exploratory analysis. The first step was to establish the convergent and discriminant validity of the constructs and the reliability of each item. The convergent validity of each construct was acceptable because they all had loads higher than 0.505 [62]. The individual reliability of each item was measured by the correlation between their loads and each variable. Table 4 verifies that the loads for each indicator were significant and validated.

Table 4: Indicator’s loads for both groups.

Source: The authors.

To calculate the internal measurement consistency of all indicators in relation to their corresponding variables, Dillon-Goldstein's test, known as the compound reliability coefficient, was applied; all values were higher than the acceptable minimum of 0.70 [63]. In addition, the Cronbach's alpha test was applied, obtaining for both groups values ​​above 0.7, the minimum value allowed for confirmatory studies [59].

Finally, the convergent validity was analysed by once again taking into account the variance. In other words, there had to be a similar variance between the indicators and their construct, which must be greater than 0.50 of the variability explained by the indicators [64], being greater in the two groups (Table 5). The discriminant validity was verified by comparing the value of the average variance extracted (AVE) of each variable with the squared correlation of each construct’s variable. The values ​​obtained from the square root of the AVE were higher than those in the constructs, and therefore, it can be considered that each variable is more strongly related to their own items than to those of other variables, which is acceptable for both groups [64] (Tables 6 and 7).

Table 5: Convergent validity of the indicators.

Source: The authors.

Table 6: Discriminant validity of the indicators - Fornell & Larcker Test.

Source: The authors.

Table 7: Discriminant validity of the indicators - Henseler-Ringle test

Source: The authors.

4.2. Validation of the structural model

To continue with the structural model, another sampling was made through the bootstrapping technique. Using 1,000 sub-samples from the data of the study, a significant contrast of the model parameters was performed, all of the above, on the basis that the model has fulfilled its predictive ability by obtaining R² values greater than 0.10 [62]. The hypotheses have been verified as follows: for the group of e-commerce users, the performance expectations variable was validated as having the highest incidence (H1a: B = 0.457), followed by effort expectations (H2a: B = 0.071), social influence (H3a: B = 0.088), and facilitating conditions (H4a: B = 0.261); although it was positive, the perceived risk construct was the only construct that was not significant (H5a: B = 0.031). For the non-online shoppers group, it was found that the variable with the highest incidence on future electronic purchase intention is performance expectations (H1b: B = 0.597), followed by social influence (H3b: B = 0.160), and, finally, with low but valid incidence, effort expectations (H2b: B = 0.078). Conversely, for the non-users of e-commerce, the variables of facilitating conditions (H4b: B = -0.001) and perceived risk (H5b: B= -0.008) were not validated (Table 8) (Figs. 2 and 3).

Table 8: Summary of the structural validity of the model.

Source: The authors.

Figure 2: Model for e-commerce non-users

Source: The authors.

Figure 3: UTAUT model for e-commerce users

Source: The authors.

4.3 Analysis of the moderating effect of socioeconomic status

The moderating effect test for each of the model relations was applied, taking the socioeconomic status variable as the moderator. The results were that hypotheses H6g (B = 0.040) and H6h (B = 0.056) are significant for buyers, given that socioeconomic status exerts a positive moderating effect on the relationship between the facilitating conditions and purchase intention for both buyers and non-buyers. The remaining hypotheses are rejected (Table 9).

Table 9: Validation of moderating effects.

Source: The authors.

4.4 Moderating effect of educational level

In the case of the moderating effects of educational level on non-online shoppers, hypothesis H7d (B = 0.097) is validated; that is, educational level has an effect that increases the positive influence of effort expectations on electronic purchase intention. The same occurs with hypothesis H7f (B = 0.071) because a higher educational level generates a greater social influence, affecting electronic purchase intention.

In the case of buyers, the following hypotheses were validated: H7g (B = 0.065), i.e., the higher the educational level is, the greater the incidence of facilitating conditions in electronic purchase intention; H7E (B = 0.032), i.e., educational level’s positive effect on the relationship between social influence and electronic purchase intention is validated; H7c (B = 0.022), i.e., educational level exerts a positive effect on the relationship between effort expectations and electronic purchase intention; Finally, H7a (B= -0.057), i.e., educational level exerts a negative effect on the significance of the relationship between performance expectations and electronic purchase intention (Table 9).

5.1 Theoretical contribution

This article aims to perform a comparative analysis between electronic buyers and non-buyers in Colombia with regard to the factors that affect electronic purchase intention. Similarly, through an exploratory analysis, it whether there is a moderating effect of the variables of socioeconomic status and educational level for both groups is examined. The results confirm that there are differences in the adoption of electronic purchase intentions between the two groups.

As a first theoretical contribution, the basic variables proposed in the UTAUT model concerning the adoption of electronic purchase intention [37,41] by electronic buyers have been validated, with a high incidence of performance expectations in both groups. This finding confirms one of the most important relationships in the literature for electronic consumers in other countries [9,40,65,66].

In addition, effort expectations are another factor that influences the intended use of e-commerce in Colombia for the group of buyers and that corresponds to previous studies that require the buyer to have motives that are related to the positive utility generated by the use of this technology [8,11,16,65,67,68]. Conversely, in the case of non-buyers, this variable has a very low validation because they may not perceive a high utility in using this commercial cannel, which would be a major reason not to use it. This result can also relate to the DIT [33], in which late adopters dislike using something new because they feel that adopting it will not be necessary.

Following this validation order for the variables in the UTAUT model, another important predictor of the intention to use e-commerce is social influence, which has been positively validated for both groups, with a significant difference between them. First, regarding buyers, this relationship has a low validation, indicating that, possibly, relying on the statistics of the sample of experience in the use of e-commerce (approximately 70% of the online shopping for more than one year) (Table 3) may explain these results. According to [69], experience shows that buyers no longer take into account the opinions of others to make a utilitarian use of e-commerce. Additionally, for non-buyers, the experience is zero, which can therefore validate this relationship to a greater degree, given the same results in previous studies in which the adopters were beginners [14,39].

One variable linked to the tools facilitating the use of this technology is facilitating conditions, which have been validated for the group of buyers, confirming the proposal of the UTAUT model regarding electronic buyers’ need to have technical conditions that enable them to make purchases effectively; however, this variable was not validated for non-buyers.

Finally, the perceived risk variable is not validated for the two groups. For buyers, although the perception of risk in online shopping is positive, it is not significant, which can be attributed to the perception of risk in business to consumer (B2C) purchases, i.e., recognized websites and the security in banking system-related means of payment, leading to the rejection of this variable as a determinant of electronic purchase intention. In the case of non-buyers, this relationship is negative and very low; although this group is clearly distrustful of online shopping, the levels do not allow for verification.

To close this first analysis, the UTAUT model for the purchase intention of two groups of Internet users in Colombia (buyers and non-buyers) is verified. For buyers, the results consistently validate the relationships between the predictor variables and the dependent variable, noting that the R² is greater than 0.590 (Table 8), which reinforces that the model has measured a large proportion of the variables that a person takes into account to generate Internet purchase intention in general and according to previous studies of e-procurement in other countries [8, 40, 41, 66, 70]. For non-buyers, this study is one of the first to describe their intention to adopt e-commerce, equally validating the model as a good predictor of these relationships, with an R² of 0.588 (Table 8).

Regarding the exploration of the possible moderating effects of the socioeconomic status and educational level variables, this study has validated that socioeconomic status has a positive moderating effect on the relationship between the facilitating conditions and purchase intention variables for both groups (Figs. 4, 5). The above can be related to the studies of [33], who find that a higher economic level facilitates the access to and enjoyment of new technology, given that people with lower incomes are more affected by the cost of the goods that are necessary to connect to the Internet and, therefore, facilitating conditions are not be considered to be an influential and positive factor [69]. Hence, this finding may be related to the fact that, in the non-buyers, the facilitating conditions variable is not significant because 49% of buyers have lower income levels (Table 3).

Figure 4: Moderating effects of socioeconomic status on purchase intention for electronic buyers.

Source: The authors.

Figure 5: Moderating effects of socioeconomic status on purchase intention for electronic non-buyers.

Source: The authors.

In addition, it has been found that educational level has a moderating effect on both groups (Figs. 6, 7), although this effect has been stronger for buyers in the following relationships: between social influence and purchase intention, effort expectations and purchase intention, and facilitating conditions and purchase intention; there is a negative moderating effect on the relationship between performance expectations and purchase intention. The results regarding educational level and its moderating effect regarding social influence may be because, when adopters gain more knowledge, they tend to share information related to their experiences among each other [33]. Social influence has been validated as influential in purchase intention, and buyers have a high degree of educational levels, with 81% of them having completed the first, second, and third cycles of higher education (Table 3), which supports the exploratory results for this group. The next moderating effect of educational level on the relationship between effort expectations and purchase intention has been proposed by [33], given that the higher the knowledge acquired is, the easier it is to handle technology. This may also explain the positive moderation in facilitating conditions, given that, for people with higher education levels, it is easier to use e-commerce [15,42].

Figure 6: Moderating effects of educational level on purchase intention for electronic buyers.

Source: The authors.

Figure 7: Moderating effects of educational level on purchase intention for electronic non-buyers.

Source: The authors.

Finally, the negative moderating effect of educational level has been validated for this group with regard to the relationship between performance expectations and purchase intention. This may occur because a higher acquired knowledge tends to generate a more advanced use of e-commerce and, therefore, high experience, which may decrease the expectations regarding the benefits of its use for this group of buyers and become negative.

Finally, in the case of non-buyers, it has been shown that there is a positive moderating effect of educational level on the relationship between effort expectations and purchase intention, which may be logical for this group of e-commerce non-adopters. Given that, in the scale proposed by Rogers, people with higher levels of educational attainment are placed as innovative adopters because they are more knowledgeable, they tend to evaluate new technologies positively and as having high value [33,69]. Furthermore, the moderating effect exerted by the education of non-buyers on the relationship between social influence and electronic purchase intention is valid. These results can be equally related to the buyers, given that people who have more information and knowledge tend to share it among the other members of the community [33].

5.2 Practical and business contribution

This study contributes to describing the adoption of electronic purchase intention in a developing country, where studies related to this context are scarce [69]. The electronic consumer in Colombia considers e-commerce to be very valuable, which is reflected in their performance expectations, the strongest predictor of the intention to engage in electronic shopping. E-commerce is perceived as something useful that offers great benefits when used, which is a great opportunity for companies to market their products and services through this channel. In addition, effort expectations validate the characteristics of the electronic consumer with some experience to make this process easy.

Linked to this variable, the positive significance of facilitating conditions describes an overview of the access to the requirements for the use of this commercial channel, in which it can be assumed that the Colombian electronic consumer has all the necessary conditions to connect to the Internet, has valid means for making payments, and possesses the knowledge to perform the process. According to [33,48], all of this is directly affected by socioeconomic status, which is a factor that requires special attention given the current level of development of e-commerce in this country because consumers have high purchasing power and a profile that allow them to pioneer the use of new technologies.

This information is valuable for companies that market their products online in Colombia because they must take into account that the electronic consumer of this country has some interesting features for products and services focused on the medium-high segments. Simultaneously, however, this is not a good sign, given that few people in this country use this commercial channel, with only 7% of the total population accessing Internet shopping [71]. For the non-buyers group, a large part of the population has a high perception of risk regarding the use of the Internet to make purchases, low positive utility in using this channel, difficulty of use, and lack of conditions needed to make purchases online, which are factors that are associated with the high digital gap [71] and which require a government policy to allow for more equitable and effective access and use of e-commerce in Colombia. Businesses are required to take into account that there is a high degree of mistrust in the use of this channel by those who still do not dare to use it, and they should disseminate strategies and policies for the promotion of electronic shopping.

5.3 Limitations

The main limitation of this study was that, even when a sample collection method by quota of all major Colombian cities was completed, a data spread across all age groups was not obtained. Although new communication and Internet technologies are used mostly by young people, a sample percentage of adults and older age people that would allow an analysis of this group within electronic buyers and non-buyers was not obtained.

Another aspect considered to be a possible limitation is that, when conducting a nationwide study that is exploratory in nature in terms of understanding the dynamics of the adoption of e-commerce from a general perspective, products and services that have a different characteristic affecting the purchasing decision are ignored, and therefore, there could be a different adoption for each type of good or service marketed over the Internet in Colombia. However, how Colombians perceive the primary use of this commercial channel, i.e., as a tool, following the initial investigation of the technological adoption models has been identified.

5.4 Suggestions for future research

It is recommended for future research on the adoption of e-commerce in other countries to conduct cross studies between buyers and non-buyers so that they can provide successful solutions in the development of this channel for the entire population, especially in countries where the development of e-commerce is strongly conditioned by a digital gap [69, 48]. In addition, the use of the unified theory of technology adoption with the UTAUT model makes it possible to test other variables that were not taken into account in this study and that can provide a further description of the virtual consumer’s behaviour; these may include the personality trait to innovations, the trait of buying low prices, and hedonic motivations, among other new factors that have been recently incorporated into technological adoption theories. It may also be interesting to conduct preliminary studies on groups of products and/or services purchased preferentially through this channel and to apply this model to find possible differences. Finally, this study can be replicated in the near future to compare the behavioural changes in the adoption of e-commerce in this country through a longitudinal analysis.

Regarding the moderating effects discovered in an exploratory manner, it is necessary to examine whether they are similarly generated in other countries with the same level of e-commerce adoption or whether, instead, they have only been found in this study because it is possible that, in countries with a high development of e-commerce, the moderating effect of socioeconomic status may not be significant because the use of e-commerce is fairly extensive. The opposite occurs with the moderating effect of educational leel because the results of this study have found that it is possible that the higher the educational level is, the higher the effects on the adoption of e-commerce in general. In addition, by having some relationships more greatly affected, future studies should test these hypotheses regarding their moderating effect with respect to social influence and performance expectations.