Exploring Factors Affecting Consumers’ Adoption of Shopping via Mobile Applications in Turkey

The aim of this study is to identify the factors behind consumers’ adoption of shopping via mobile applications and to develop a new model that explains this situation. The related literature was examined for this purpose. Delphi technique was preferred to determine factors in the study. Data was collected through questionnaires. Exploratory Factor Analysis (EFA) was conducted with SPSS. A research model based on an integration of various theoretical fields was developed. As a result of EFA, ten new dimensions emerged in the study. And then, in order to statistically analyze the measurement and structural models, this study used Smart PLS for Structural Equation Modeling (SEM) technique. After path analysis with Smart PLS, a new conceptual model was developed to explain adoption of shopping via mobile applications by consumers in Turkey. Structures such as Personalization, Word of Mouth Communication and Perceived Mobility used in the model developed within the scope of this research, but rarely used in this field of studies, were verified to be determinants of shopping behavior via mobile applications in Turkey. The model developed within the study is both valid and reliable in terms of its structure and all relations established within the scope of the model are significant.

model, the social influence process which consists of subjective norm, voluntariness and image components and the cognitive instrumental process consisting of job relevance, output quality, result demonstrability and perceived ease of use significantly affect user's acceptance (Venkatesh & Davis, 2000). In the study, marketing experts reached on a consensus about that image has an impact on the consumers' adoption of shopping via mobile applications. The image is described as the perception that use of an innovation increases one's status within the social system (Moore & Benbasat, 1991).
Another study within the field developed an integrated model at the user level (TAM 3), based on the suggestion that there is a lack of research on the role of interventions in information technologies which are considered helpful for managerial decision making mechanisms. The study also suggests new concepts as determinants of the perceived usefulness and the perceived ease of use concepts, namely individual differences, system characteristics, social influence and facilitating conditions. In the study, marketing experts reached on a consensus about that facilitating conditions have an impact on the consumers' adoption of shopping via mobile applications. Facilitating conditions refer to organizational support which makes use of information technologies possible (Venkatesh & Bala, 2008). According to the model (TAM3), determinants of the perceived usefulness are the perceived ease of use, subjective norms, image, job relevance, output quality and result demonstrability concepts. In addition, determinants of the perceived ease of use concept mentioned by the model include computer self-efficacy, perception of external control, computer anxiety, computer playfulness, perceived enjoyment, and objective usability. In the study, marketing experts reached on a consensus about that perceived enjoyment has an impact on the consumers' adoption of shopping via mobile applications. Perceived enjoyment refers the degree to which an activity of using a system is perceived to be enjoyable (Venkatesh, 2000).

Theory of Planned Behavior
Theory extends Theory of Reasoned Action by taking perceived behavioral control in consideration, which refers to users' perception about internal and external restrictions that may affect the behavior (Ajzen, 1991). Perceived behavioral control means availability of opportunities, sources and purposes (time, money, skills, cooperation with others) for an individual to enable him or her to display a specific behavior and to ensure success of this behavior (Ajzen, 1991). In the study, marketing experts reached on a consensus about that perceived behavioral control has an impact on the consumers' adoption of shopping via mobile applications.

Innovation and Diffusion Theory
The theory explains the main component in terms of diffusion of new ideas and thoughts as; diffusion of (1) an innovation (2) through communication using certain channels, (3) in a specific time and (4) among members of the social system (Rogers, 1995). According to this theory, the degree to which members of the social system perceive characteristics of an innovation determines the rate of adoption. An innovation has five characteristics: (1) Relative advantage, (2) compatibility, (3) complexity, (4) trialability and (5) observability (Rogers, 1995). First three of these factors were identified as in relation with the decision to adopt (Tornatzky & Klein, 1982) and use of information technologies (Moore & Benbasat, 1993).
Relative advantage indicates the degree to which the innovation is more useful than the replaced situation (Roger, 1995). The benefits in this regard consist of economic benefits, image development, ease of use and satisfaction. These benefits share similarities with the usefulness factor included in the TAM (Taylor & Todd, 1995). Complexity refers to the level of difficulty in learning, understanding and applying an innovation (Rogers & Shoemaker, 1971). This concept shows similarity with the ease of use component of the TAM (Taylor & Todd, 1995). Compatibility, on the other hand, explains the degree to which the innovation complies with values, previous experience and current needs of the adopter (Roger, 1995). In the study, marketing experts reached on a consensus about that compatibility has an impact on the consumers' adoption of shopping via mobile applications.

Technology Readiness Index
Technology Readiness Index was developed by Parasuraman in 2000. The aim of this index is to identify preventive and encouraging factors in terms of individuals' tendency to use new technologies. The Technology Readiness Index consists of four dimensions. These four dimensions are optimism, which refers to a positive approach toward technologies; innovativeness, which indicates the tendency to be a leader in a specific technology and thought; discomfort, which is perceived lack of control over technology; and finally insecurity, distrust of technology and skepticism about its ability to work properly (Parasuraman, 2000).
consumers' adoption of shopping via mobile applications.
This section of the study covers original structures which are not mentioned in aforementioned models but encountered within the literature. It is observed that the Value Based Adoption Model, which explains consumers' adoption of mobile internet in terms of value maximization, uses the concept of the perceived value (Kim, Chan, & Gupta, 2007). Within the scope of this model, the perceived value refers to the exchange between advantages obtained by the consumers such as relevant features and benefits and the resources used such as price, opportunity cost, time and necessary effort. Therefore the concept covers the perceived monetary value and the perceived cost as well (Zhang, Zhu, & Liu, 2012;Wei, Marthandan, Chong, Ooi, & Arumugam, 2009;Faziharudean & Li-Ly, 2011;Kim, 2012). The level of preexisting knowledge about the innovation itself or a similar product/service, or existing knowledge, reduces the innovation's degree of complexity (Bauer, Reichardt, Barnes, & Neumann, 2005;Yang, 2005). Perceived mobility refers to the fact that mobile services enable widespread and instant communication. "Positive Word of Mouth Communication" focuses on individuals' willingness to recommend products and services and encourage use of them as a result of loyalty, rather than their own future use and purchasing behavior (Assarut & Eiamkanchanalai, 2015;Turel, Serenko, & Bontis, 2010). Perceived Informativeness refers to his or her perceiving products and services of mobile applications as a significant source of knowledge (Bauer, Reichardt, Barnes, & Neumann, 2005;Okazaki, 2007). Perceived risk consists of three key concepts (Lu, Liu, Yu, & Wang, 2008). These are protection during the data transmission process and of data storages, privacy concerns, success in achieving desired outcomes. In conclusion, personalization refers to the customized offerings related to products and services sent by a mobile application (Xu, 2006).
Marketing experts reached on a consensus about seven structures mentioned above in this section. On the other hand, structures like need for uniqueness, mobile affinity, media effect, trialability were excluded from the study.

Research Design and Methodology
The literature review within the scope of this study identified a good number of factors that affect consumers' adoption of shopping via mobile applications and these factors were analyzed by ten marketing experts using the Delphi Technique. technologies reviewed within the literature and modified to be compatible with the study. The present study includes: four items measuring Perceived Usefulness (Davis, Bagozzi, & Warshaw, 1989;Ahn, Ryu, & Han, 2004;Wu & Wang, 2005); five items measuring Perceived Ease of Use (Davis, Bagozzi, & Warshaw, 1989;Nysveen, Pedersen, & Thorbjørnsen, 2005); four items measuring Perceived Value (Sirdeshmukh, Singh, & Sabol, 2002;Kim, Chan, & Gupta, 2007); four items measuring Perceived Risk (Bauer, Reichardt, Barnes, & Neumann, 2005;Wu & Wang, 2005); four items measuring Perceived Enjoyment (Nysveen, Pedersen, & Thorbjørnsen, 2005); five items measuring Innovativeness (Goldsmith & Hofacker, 1991;Parasuraman, 2000); three items measuring Existing Knowledge (Flynn & Goldsmith, 1999;Bauer, Reichardt, Barnes, & Neumann, 2005); four items measuring Word of Mouth Communication (Lin, Sher, & Shih, 2005;Turel, Serenko, & Bontis, 2010); three items measuring Perceived Mobility (Hong, Thong, Moon, & Tam, 2008); three items measuring Personalization (Mittal & Walfried, 1996;Chellappa & Sin, 2005;Xu, 2006); three items measuring Perceived Informativeness (Bauer, Reichardt, Barnes, & Neumann, 2005); two items measuring Subjective Norm (Bhattacherjee, 2000;Nysveen, Pedersen, & Thorbjørnsen, 2005); two items measuring Image (Moore & Benbasat, 1991). Moreover, Attitude and Behavioral Intention concepts mentioned in the final model of the study were measured with three items for each (Shimp & Kavas, 1984;Bauer, Reichardt, Barnes, & Neumann, 2005;Yang & Zhou, 2011). In conclusion, the conceptual structure of Behavior was examined by addressing two questions (Kim, 2012). In order to efficiently identify the factors behind consumers' adoption of shopping via mobile applications, the sample of this study was chosen from people who do shopping via mobile applications. It means that the study uses a nonrandom method which is called as the judgmental sampling. The data was collected through questionnaires conducted between April-June, 2017, from 450 consumers who live in Istanbul, Turkey and use mobile applications for shopping but only 438 of them were included in the analysis process. 64 variables to be analyzed were scaled by a 5point Likert type scale and the respondents were asked to answer these questions by choosing points from (1) strongly disagree to (5) strongly agree.
An EFA with SPSS was conducted in order to identify dimensions of the data collected from consumers. The dimensions obtained as a result of the EFA were included in the research model and the relations within the literature (Figure 1) Contrary to covariance based AMOS and LISREL path modeling techniques (CB-SEM), SMART PLS is a regression based technique which rests upon the path analysis (Chinomona & Sandada, 2013). A study conducted with SMART PLS and AMOS to explore the structural equation modeling compatible with confirmatory factor analysis revealed that values of PLS-SEM are more efficient than those of CBSEM, even when the data is same (Afthanorhan & Afthanorhan, 2013). Smart PLS software has sufficient capacity to process complicated prediction models on small and medium scaled samples. After EFA, due to the reasons mentioned above, this study used Smart PLS software in order to statistically analyze the measurement and structural models.

Results
When the data associated with participants are examined (Table 2), it is seen that 64% of those using mobile applications for shopping are in 26-36 age group, 42% are female and 58% were male, while 80% of all participants have a monthly income over the value of minimum wage applied in Turkey.
Data of the 438 participants were subjected to an EFA with SPSS software, within the framework of the study. It is often mentioned as a general rule that factor analysis should be applied to more than 300 samples (Çokluk, Şekercioğlu, & Büyüköztürk, 2012). The obtained data was analyzed through principal axis factoring method and the Promax rotation technique. Variables with sampling sufficiency values less than 0.4, single variables under specific factors, variables which are under multiple factors and at the same time have factor load differences less than 0.1 were excluded from the study and 10 different dimensions were obtained as a result of the factor analysis, each of them having eigenvalues over 1. Internal consistency of factors was measured by using the Cronbach's Alpha (α) value. The rate of variance explained as a result of the EFA is 69%. Internal consistency and explained variance rates of the factors, KMO values associated with the data set and significance level are given in Table 1.
Following the determination of factors that have influence over consumers' adoption of shopping via mobile applications by EFA, the research model given in Figure 1 was developed in order to test relations of these factors with the structures of Attitude, Behavioral Intention and Behavior. The relations within the scope of this study were tested with Smart PLS software using bootstrapping resampling method. Relations within the model were tested at a significance level of 0.01. The path coefficients and T values in the research model are located in Table 3. The conceptual model developed through the analyses conducted with Smart PLS software and used to explain consumers' adoption of shopping via mobile applications is given in Figure 2.
When the external loads of latent variables determined by the analysis are examined, it is seen that all variables belonging to the structure of behavior are over 0.6 except for one (B2:0,409). Some experts of psychometrics (Churchill, 1979) suggest that reflective indicators with standardized outer loads less than 0.4 should be excluded from the model. The lowest behavioral variable was not excluded from the model as its load was over the limit. Composite Reliability (CR) value is expected to be over 0.7 within the first phase of the study and over 0.8 or 0.9 during the following phases (Nunnally and Bernstein, 1994). When Table 1 is examined, it is seen that all CR values belonging to latent variables, except for Behavior (0.695), are over the value of 0.7. These results show that all latent variables used in this model are reliable.
In order to evaluate the validity, two subtypes should be analyzed. These are convergent validity and discriminant validity. It is recommended to use the Average Variance Extracted (AVE) for evaluation of the convergent validity (Fornell & Larcker, 1981). The AVE value should be over 0.5. Table-1 shows that threshold coefficients of AVE values belonging to the latent variable are higher than 0.5. Therefore convergent validity of the latent variable within this model was verified. It is also stated that as square roots of the AVE values belonging to latent variables can also be used for the discriminant validity, the square root values should be higher than the correlation value among latent variables (Fornell & Larcker, 1981). The Fornell Larcker Criterion section, which is given in the report output of Smart PLS software, is also summarized in Table 4. The values written with boldface letters inside the Table indicate square roots of AVE scores belonging to latent variables. Table-3 verifies the discriminant validity of the measurement used in the study. The study uses the Variance Inflation Factor (VIF) to test the indicator variables that form the formative latent variables in terms of multicollinearity. As a rule, a VIF value over 10 indicates existence of harmful collinearity (Henseler & Ringle, 2009). None of the indicators and latent variables used in this study has VIF values over 5. Therefore, there is no problem with multicollinearity between indicator variables and latent variables used in the model.
The analysis conducted with Smart PLS software measures the path coefficients associated with the structures used in the model. Significance of these coefficients is evaluated by using the bootstrapping method. This method provides "t values" for each path prediction. All paths indicating relations among latent variables of the model developed by the study have significance values of 0.01 (Figure 2    When the model developed in this study is examined within the context of relations among conceptual structures analyzed by previous studies in the literature, it is identified that Innovativeness (Pagani, 2004;Gao, Rohm, Sultan, & Pagani, 2013), Perceived Informativeness (Bauer, Reichardt, Barnes, & Neumann, 2005;Okazaki, 2007) and Personalization (Chellappa & Sin, 2005;Xu, 2006) (Davis, Bagozzi, & Warshaw, 1989;Yang, 2005;Yang, 2005;Bhatti, 2007).

Conclusions
The present study has some significant findings from Turkish culture about consumers' adoption of shopping via mobile applications. Structures such as Personalization, Word of Mouth Communication and Perceived Mobility used in the model developed within the scope of this research, but rarely used in this field of studies, were verified to be determinants of shopping behavior via mobile applications. Considering today's marketing approach to customize products and services in line with individual wishes and needs, it seems highly natural that personalization may be a determining factor for adoption of mobile applications. Therefore users of mobile shopping applications who have positive attitudes would be more willing to share their personal information to benefit customized offers. In the information age we are in, consumers want to bring together all the information they can obtain about products and services before making purchase decision for choosing the best option. In this regard, consumers are highly affected by comments and evaluations of those who have previously experienced a specific product, service or application. High sensitivity toward bandwidth, data transmission rates and geographical locations are among limiting factors for mobile applications' ability to perform functions. The concept of mobility, which expresses the perception of "everytime and everywhere", shows the expectations of consumers from products and services in a specific place and time interval. That's why to overcome these limiting factors would enhance the perception of mobility. Considering the developed model and factors affecting consumer adoption of shopping via mobile applications such as personalization, word of mouth communication and perceived mobility can give firms to gain a sustainable competitive advantage in a digital world. The results of the study are considered to beneficial for mobile application developers who are independent individuals and professional organizations like brands.

Limitations and Future Research
Some possible semantic and verbal mistakes in translation of variables given in survey forms from English to Turkish are considered among limitations of the study. The developing but not settled structure of the market associated with mobile technologies in Turkey is another limitation. On the other hand; as the research model was developed in consideration of relations within the literature, Attitude, Intention and Behavior concepts were not included in the exploratory factor analysis.
Considering the relatively low level of mobile shopping in Turkey, where numbers of smart phone users and mobile broadband subscribers are rather high, further studies to identify the determinants and explain the behavior in question with larger samples from different regions are needed in purpose of ensuring diffusion of mobile shopping in Turkey.