Problematic internet shopping behavior a mediation analysis of online interpersonal relationships and data ownership awareness

Dissertation 異常網路購物行為研究：網路人際關係與資料所有權意識的中介分析 Problematic Internet Shopping Behavior: A Mediation Analysis of Online Interpersonal Relationships and Data Ownership Awareness 指導教授Advisor：廖世

INTRODUCTION

Background of the Study

Over the last few decades, the e-commerce industry has evolved rapidly with the popularization of the internet The internet is generating profound changes in how retailing is operated The burgeoning of digitization is generating opportunities for the extensive capture, storage, and analysis of business data, optimally used with big data analytics (BDA), which can release extraordinary business benefits The organization has increasingly implemented the internet and computing technologies to assemble massive amounts of consumer data and provide effective and efficient marketing strategies Companies were shown to consistently embrace technological artefacts to gather, analyze, and distribute personal information despite growing concern on whether individuals’ data is being compromised

The internet offers an opportunity to be a "retail therapy," but shopping overuse could be problematic There has been an upward trend in research exploring problematic buying-shopping behavior Psychological literature commonly refers to this maladaptive consumption as compulsive buying or addictive buying-shopping Many efforts have been made to characterize problematic shopping behavior, mainly focused on potential antecedents, possible comorbidity with other established disorders, or estimating the occurrence among the clinical and general population The crux of the issue is that psychiatrists have not agreed on defining or measuring it, and estimates of its prevalence vary widely Thus, an individual's inability to regulate their buying-shopping urges has attracted much research attention It has been listed in the appendix of the most recent version of the Diagnostic and Statistical Manual of Mental Disorder (American Psychiatric Association, 2013) However, it has not been classified as a separate mental illness because of inadequate scientific evidence to establish the diagnostic criteria The advent of internet technologies and the prevalence of BDA and technologies in e- commerce provide customers countless benefits when shopping online Popular e-commerce websites have argued that integrated features could stimulate unregulated buying/shopping (Larose & Eastin, 2002) However, the relationship between the application of big data analytics in e-commerce and problematic internet shopping behavior (PISB) has not been thoroughly examined Therefore, the present study intends to (i) establish a framework for studying problematic internet shopping behavior, taking into consideration the impact of benefits and risks entailed by BDA and data-intensive technologies; (ii) investigate the effect of big data advantages and disadvantages on consumers’ data ownership awareness; (iii) explore the mediating effect of online interpersonal relationships and data ownership awareness on the association between consumers’ perceived benefits and risks and problematic internet shopping behavior, and (iv) construct predictive models to predict problematic internet shopping behavior and compare the performance of relevant classifiers to determine the best classification algorithms based on the predictive performance comparison.

Statement of the Problem

Scholarly work on consumer behaviors has undergone a significant transformation in shopping habits resulting in a stronger disposition to internet shopping as a consequence of the expansion of the internet and the acceleration of the electronic commerce market The utilization of information and communication technologies in e-commerce has provided immense opportunities for online consumers to lure into the realm of shopping enjoyment

Shopping has long been and is becoming an integral part of our everyday activities that helps people express their self-identity and escape negative feelings However, such a revolution is accompanied by a coincidental development of some problematic behaviors related to buying and shopping Although consumers’ irrational purchasing has continually been investigated, significant gaps exist in the literature In addition, online shopping entails numerous benefits and risks to consumers, and such factors affect the consumer's attitude and behavior differently While the benefits have been reported to influence customer satisfaction and positive attitude toward online purchasing, the risk associated with internet shopping is suggested to impact online purchasing intention negatively

The Stimulus-Organism-Response model is an influential theoretical model used widely in retailing and marketing literature to understand the effect of environments on customer behavior The S-O-R model incorporates stimulus as the independent variable(s), the organism as the mediator, and response as the outcome variable (Vieira, 2013) However, studies examining big-data- related stimuli that affect customers’ response, mainly focusing on the dark side of consumer behavior, remains scarce Therefore, designing a framework for studying problematic online consumers’ behavior based on the theoretical ground that incorporates environmental factors can allow a better understanding of problematic internet shopping behavior The present research will design a framework for studying problematic online shopping and assess the relationship between the consumers’ perceptions regarding the pros and cons of online shopping and problematic online shopping behavior.

Objectives of the Study

Compulsive buying and shopping addiction are terms used interchangeably in behavioral psychology and consumer behavior literature It refers to the preoccupation, chronic, repetitive, impulsive, obsessive, and excessive buying-shopping that results in adverse consequences associated with psychological, financial, interpersonal, and legal problems Consumers are confronted by a dilemma over how to indulge themselves in shopping enjoyment while not stumbled down into the addiction The benefits and risks of online shopping play a role as stimulation affecting customer response Therefore, the objectives of the current study are to (1) establish a research framework based on Stimulus-Organism-Response for studying problematic internet shopping behavior; (2) to accentuate the effect of perceived benefits and risks of online shopping on consumers’ data ownership awareness and problematic internet shopping behavior; (3) explore the mediating effect of online interpersonal relationships and data ownership awareness on the association between consumers’ perceived benefits and risks and problematic online shopping behavior, and (4) construct a model to predict problematic internet shopping behavior based on classification algorithms and compare the predictive performance of relevant classifiers.

Contribution of the Study

The present study's findings contribute to the expansion of problematic internet shopping literature relevant to BDA and the development of the online retailing market The study offers a potential avenue of research in developing approaches for moderating and managing problematic internet shopping behavior and its possible intervention for treating this mental disorder Research findings highlight the influence of perceived benefits and risks of shopping online and its relationship with data ownership awareness and problematic online shopping behavior It means that the mechanisms in which online interpersonal relationships and data ownership awareness intervention the proposed association was scrutinized Additionally, machine learning models were constructed to predict problematic internet shoppers based on consumers' perception of the advantages and disadvantages of internet shopping and their relevant demographic characteristics The efficient machine learning classification algorithms are also determined to provide beneficial tools for businesses, e-commerce practitioners, and future researchers in promulgating policy and interventions for the incidence of problematic internet shopping behavior.

Operational Definitions

Buying and shopping: The words are often used interchangeably, but there is a substantial difference between the two While buying refers to locating and purchasing a product, shopping indicates the act of searching for inspiration and discovering a product that may not be of interest

Online shopping or internet shopping generally pertains to purchasing products and services through the internet This process includes but is not limited to looking for e-vendors and products, searching for product information, comparing prices, choosing payment options, and completing a transaction

Compulsive buying is “a chronic, repetitive purchasing that occurs as a response to negative events or feelings” (O'Guinn & Faber, 1989)

Impulsive buying: refers to the feeling of a persistent urge to make a sudden and immediate purchase without careful consideration of the cause and consequently complete the purchase Impulsive buying is a less evident manifestation of compulsive buying in which the former describes the opening stage and the latter represents the utmost side of the same behavior (Kyrios et al., 2020; Ridgway et al., 2008)

Shopping addiction is yet another derivative being used interchangeably with compulsive buying Shopping addicts are characterized by an obsession with shopping, pre-purchase stress, or anxiety, followed by a sense of relief after purchases There have been many controversies on whether shopping addiction is a legitimate mental disorder or a pastime that individuals operate to regulate their feelings or reveal their self-identity

Convergence validity refers to the extent to which indicators of a specific construct converge or share a high proportion of variance in common Convergent validity can be ascertained by calculating the average variance extracted value and composite reliability of each dimension of the respected construct

Discriminant validity is the degree to which a measure is genuinely distinct from another In other words, each item is suggested to correlate more substantially with other items of the same construct than those of other constructs.

Research Systematic Discussion

This research includes five chapters: introduction, literature review, methodology, results, discussion, conclusion, and recommendations

To begin with, the author identifies the research problem at the beginning of the dissertation Next, the research topic was determined by considering the influence of perceived benefits and risks of online shopping on problematic internet shopping behavior and the mediating role of online interpersonal relationships and data ownership awareness Once the research topic is identified, the research objectives are stated

Subsequently, the literature review explores and synthesizes previous studies on big data and big data analytics in e-commerce and introduces the benefits and risks of online shopping, online interpersonal relationships, and data ownership awareness The rationale for the mediating roles of online interpersonal relationships and data ownership awareness is also justified

Afterwards, the research methodologies are discussed in the third chapter, where the research models are presented The hypotheses are proposed within this section, along with the sample size determination, data collection, data screening and pre-processing, and data analysis

Further, the results and discussion are presented in the fourth chapter, wherein three studies' results are presented, followed by interpretation, discussions, and sub-conclusion More specifically, Study I characterize the effect of perceived benefits and risks of shopping online on consumers’ data ownership awareness Study II examines the mediating roles of online interpersonal relationships and data ownership awareness on the relationship between perceived benefits and risks of shopping online and problematic internet shopping behavior Study III presents a comparative analysis of data mining classification algorithms to predict problematic internet shopping behavior Consequently, the conclusion chapter summarizes the main research findings, and the thesis concludes with implications and limitations Figure 1 depicts the whole process through which the thesis is undergone

Questionnaire design and Data collection

Analytical methods to be used

Study III (N 3 = TW + VN + MTurk)

The Organization of the Dissertation

The present research is comprised of five chapters: (1) introduction, (2) literature review, (3) methodology, (4) results and discussion, and (5) conclusion and research recommendation

The introduction chapter contains the study’s background, problem statement, research objective, contributions, operational definitions, and systematic research discussion

Chapter 2 delineates and analyzes the revolutionary and relevancy of the identified research problem The author focuses on the customers’ perception of the benefits and risks of online shopping by applying big data analytics in e- retailing operations In addition, the foundation for the cross-cultural comparison was elaborated with the justification for the relevancy of associated constructs used within this study, i.e., online interpersonal relationships and data ownership awareness

Chapter 3 justifies the research methods that are employed in this research The author explains hypotheses development, measurement design, sample size determination, data collection, and analysis techniques

Chapter 4 provides research findings for three studies described in the previous section Each study includes the analytical research results, interpretation of results, discussion, and sub-conclusion

Consequently, the final chapter concludes the thesis with concluding remarks, managerial and practical implications, limitations, and recommendations for future research Additional research outputs are provided in the Appendix section.

LITERATURE REVIEW

Internet Use and Online Consumer Behavior

There has been a phenomenal increase in global internet usage over the years As of April 2022, the number of internet users worldwide stood at 5.0 billion, meaning that almost two-thirds of the worldwide population is connected to the cyber world (WeAreSocial, 2022) Notable increases in total internet users have been observed worldwide, especially in Eastern Asia, Southern Asia, Southeast Asia, and Northern America (Statista, 2022) The internet has transformed how products are traded worldwide, and e-commerce has become an indispensable part of the global retail landscape Statista (2022) reports that global retail e-commerce sales constituted nearly 4.9 trillion U.S dollars in 2021 There appears to have been an upward trend in internet use globally due to the physical constraints driven by implementing measures to inhibit the coronavirus disease (e.g., lockdowns, quarantines, self-isolation)

Moreover, consumer behavior is subjected to two fundamental transformations: (i) converting consumers into digital purchasers necessitating the usage of internet technologies, and (ii) transforming brick-and-mortar retails into online marketplaces that are information-intensive based It is revealed that changes in consumer purchasing behavior significantly and positively influence the business model design (Tao et al., 2022) However, consumer behavior is difficult to understand due to the complexity and the involvement of various entities such as consumers, businesses, and products Several theoretical frameworks have been employed to investigate online consumer behavior or examine consumer willingness to adopt e-commerce technologies (Hwang & Jeong, 2014) These studies can be categorized into two groups, i.e., (i) factors affecting the acceptance and utilization of e- commerce and (ii) determinants that promote or inhibit consumers from making purchases In addition, the internet and social media have changed consumer behavior and changed how companies conduct their business Online marketing provides seamless opportunities to businesses through lower costs, improved brand awareness and increased sales.

Big Data Analytics in Electronic Commerce

The notion of big data concerns the abundance of data elaborated on the data acquisition, and traditional analysis tools cannot consummate data collection, handling, and examination in a short time (Wamba et al., 2015) Big data analytics (BDA) refers to the process of analyzing huge amounts of complex data to reveal hidden patterns, unexplored relationships, and other insights Big data analytics apply advanced analytic techniques against large data sets for advancing business and provides insights to improve business decisions Electronic commerce enterprises have been one of the pioneer groups adopting BDA due to their inbuilt necessity to overtake the rivalry (Akter & Wamba, 2016) The utilization of BDA allows e-vendors to explore customers’ profiles comprehensively For instance, transactional or business- related activity, click-stream, video, and voice are processed to gain consumer insights Big data analytics application in e-retailing has enabled the personalization of customers’ shopping experience, so long as the instance of the product recommender systems introduced by electronic marketplace leaders, such as eBay and Amazon (Chen et al., 2012) The utilization of BDA in enterprises could enhance their production efficiency and competitiveness, predict the behavior of consumers, and propose new business modes (Chen,

2014) Raguseo (2018) perceives four different benefits that can be achieved from adopting big data technologies: transactional, strategic, transformational, and informational Le and Liaw (2017) suggest that the advantages of applying big data analytics, including information search, recommendation systems, dynamic pricing, and customer service, have a positive effect, while privacy, security, shopping addiction, and group influence negatively affect customer intention and behavior Privacy and security issues and the scarcity of information system structure support are the most frequently recognized risks that firms must consider before using big data technologies.

E-commerce and Cross-cultural Context: Taiwan and Vietnam

Culture has been recognized as one of the most influential consumer behavior factors, shaping needs and affecting consumers’ behaviors, attitudes, and preferences (Mooij, 2019) Vietnam is considered a dynamic economy with a high economic growth rate and improving the living standards of the country’s population This Asia-Pacific nation has the highest internet penetration rates, which could lead to an undeniable e-commerce market development By July 2019, Vietnam had approximately 72 million internet users, accounting for almost three-fifths of the total population The country is ranked seventh globally, with around 65 million social media users (Kemp, 2020; Meng, 2020) In addition, there was a significant increase of over 10% in national retail growth during the 2013 to 2018 period (Deloitte, 2019) Although such favorable conditions are necessary for the advancement of online shopping, the e-commerce industry has not developed adequately Some reasons include inadequate infrastructure and legal documents to regulate and create fair competition and protect consumer information (Yang et al., 2018) Also, less is known about consumer behavior in Vietnam due to its complexity with the involvement of many factors (Le et al., 2022) Research on internet- related issues in Vietnam has primarily concentrated on the pros and cons of adopting the internet into their social lives and business activities (Lin et al., 2015; Zhang et al., 2017) Numerous studies have been implemented to investigate issues related to e-commerce and mobile commerce within the context of Vietnam, such as internet banking (Le et al., 2022; Liang, 2016; Lin et al., 2015), electronic payment, and e-invoice (Lin & Nguyen, 2011; Nguyen et al., 2020), mobile commerce (Han et al., 2016), and credit card usage (Nguyen & Cassidy, 2018) These studies suggest that most Vietnamese customers choose cash on delivery, leaving the e-payment methods to constitute a negligible proportion of total e-commerce payments The lack of customers’ trust in e-commerce and the insufficient online payment methods have resulted in less preference for e-payment adoption

Against this backdrop, Taiwan is a high developed country in terms of information technology (Pi & Sangruang, 2011) About 90% of the Taiwanese population uses the internet to assemble and exchange information for various purposes The e-commerce industry in Taiwan is highly developed According to Taiwan Internet Report 2020, almost 60% of internet users have experience shopping digitally Mobile payments have increased from 14.2% in 2017 to 25.8% in 2020 (Bergstrửm, 2022) In addition, Taiwan is ranked third in Asia regarding internet use, just behind South Korea and Japan (TWNIC, 2022) In Asia, the average rate of online shopping usage among Taiwanese users surpasses the global average but is not as much as in Indonesia, Malaysia, and Thailand According to Global Web Index, the percentage of internet users aged

16 to 64 researching products online before purchasing is 61.5% and 56.5% in Taiwan and Vietnam, respectively However, Taiwanese netizens still express concerns about online security, making it one of the most pivotal problems when using internet services

The two countries that have been selected to conduct research differ considerably regarding economic development and materialistic consumer values (Nguyen, 2019) Taiwan is currently ranked as a high-income industrialized country, whereas Vietnam, on the contrary, is a developing country For a collectivist culture like Vietnam, the social dimensions of shopping are particularly important Traditional face-to-face contact is imperative to build trust and enhance the buyer-seller interaction (Cutshall et al., 2022) In the third quarter of 2020, according to GlobalWebIndex.Com, 31.8% of Vietnamese internet users voice concerns over the extent to which companies use their data online, higher than that of Taiwanese internet users at

29.7% Besides, the two countries also differ in several cultural dimensions (Hofstede, 1991) For instance, the avoiding uncertainty score of the Taiwanese is higher than that of Vietnamese people The difference in internet use behavior and cultural dimensions would pave the way for seamless cross- cultural comparison between the two countries.

Stimulus-Organism-Response Model

The current research adopted the Stimulus-Organism-Response (S-O-R) model as the research framework It was introduced by Mehranbian and Russell

(1974) to explain the consumer decision-making process The S-O-R paradigm postulates that environmental stimuli (S) can promote consumers’ cognitive and affective reactions (O), which sequentially propel their behavioral responses (R) Accordingly, external stimuli such as e-commerce platforms’ features affect consumers' psychological states In the second component, the organism (O) models the consumers’ internal processes that respond to stimuli, including mood, cognitive and emotional perceptions The term response depicts the consequence or the reactions consumers have against internal processes Scholars have widely adopted this framework to uncover consumers’ internet buying-shopping behavior, with much research focusing on impulse buying behavior per se (Chang et al., 2011) The S-O-R framework puts forward a coherent basis to investigate the impact of various e-commerce related-stimuli on consumer perception and, in turn, consumer impulse buying behavior The current study considers the benefits and risks of shopping online brought by integrated and driven by big data technologies into e-commerce platforms (S) Their impact on online interpersonal relationships and data ownership awareness is considered as the consumer’s emotional state (O) The latter, successively, influences the consumer and results in problematic internet shopping behavior (R) Embracing the S-O-R model in understanding consumer behaviors helps distinguish environmental stimuli and consumers’ internal and external behaviors.

Perceived Benefits of Online Shopping

Information search is an initial step of the decision-making process wherein individuals intensely retrieve and incorporate information from a different source (Chiu et al., 2019), with the overall approach being to diminish risk and uncertainties (Grabner-Krọuter & Kaluscha, 2003) Big data's emergence has altered how online consumers acquire product information If consumers are interested in specific goods or services, they are more likely to compare until they feel they have adequate information to make a rational decision The existence of search engines embodied by big data analytics can assist consumers in identifying more significant amounts of relevant information than could be retrieved in a brick-and-mortar context Information relating to product attributes, the authenticity of the e-vendors, the reliability of previous customer reviews, or comparing product prices across e-commerce websites before making a purchase decision can be easily obtained from different sources on the internet The notion of information overload is an overwhelming problem for consumers during decision-making (Yan et al.,

2016) The tremendous amount of information available through the internet has yet generated value without adequate mechanisms for which information is specified, retrieved, and organized An e-commerce website with big data analytics tools can inspect and refine a wealth of customers' information Generally, big data relates to insights and provides the relevant product information to the right consumer for a reasonable price through the right marketing channel to make right-time decisions As such, the ability to retrieve products or services information online and acquire constant updates about new product launch can become a crucial incentive for consumers to plump for the internet in place of physical storefront shopping (Peterson & Merino, 2003)

Information overload has become a critical consumer problem during decision-making (Yan et al., 2016) Recommendation systems incorporate high-tech software and data mining algorithms to provide suggestions for items mostly matching a particular user's interest, helping consumers filter irrelevant information and find items of interest among massive databases These techniques are mainly intended for individuals with bad or incompetent personal experience in evaluating the potentially enormous alternatives offered by a website (Resnick & Varian, 1997) The overall objective of recommendation systems is to reduce information overload and provide more fruitful information retrieval They have become an essential strategic tool for organizations in the digital marketplaces According to Chen (2018), the e- commerce recommendation system based on big data analysis can increase business operation capability, identify potential customers, and enhance customized offerings Likewise, Schafer et al (2001) argue that recommendation systems help boost conversion rates by increasing cross- selling opportunities and raising customer loyalty The systems use consumers’ purchasing data and user-specific preferences to produce a list of recommended items (Hu et al., 2019) Recommendation systems benefit users greatly from a more purified list of items provided by processing a large amount of data These are powerful tools for internet users to confront information overload and allow them to make rational decisions Hence, recommendation systems benefit both consumers and firms The recommendation systems significantly minimize the requisites for collecting, screening, and assessing product’s information These systems allow consumers set up the buying plan that is more in consonance with their preferences at a more reasonable price

Price has been a critical driver affecting consumer decision-making, and how price is set influences business performance Consumers often compare an item’s worth with the prices of similar items or the same items that came from other stores Dynamic pricing offers goods at different prices, which vary according to the customer’s demand This price-setting tactic has steadily become common as long as online marketing predominance increases Developed initially by airlines (Chiang, 2007), dynamic pricing is now widely adopted across a myriad of industries like retail (Chen, 2009), transportation (Chiang, 2007; Puller & Taylor, 2012), mobile communication, hospitality (Vives & Jacob, 2020) and sharing economy (Gibbs et al., 2018), to name a few Big data analytics assist e-retailers in using a dynamic pricing strategy to attract more customers Shiller (2014) predicts that if Netflix had used pricing based merely on consumer demographics, it could have increased its profit by 0.8 percent, but by 12.2 percent if it had utilized personalized prices based on web-browsing data Retail firms leverage digital technology advancement and make it possible to analyze enormous data about potential customers, enabling companies to personalize prices according to a specific algorithm Price discrimination algorithms can help firms recognize customers through cookies, previous purchases, and click-streams to assign them into price-sensitive or price-insensitive categories and charge the latter with higher prices Therefore, two customers could purchase the same product at different prices, and those with an increased willingness to pay will probably be worse off While dynamic pricing is naturally good for the firm's profitability, it often accompanies with the negative effect of price volatility resulting from short-term ups and downs in supply and demand conditions (Yan et al., 2020) In addition, many individuals consider personalized pricing as tricky or manipulated (Haws & Bearden, 2006; Hindermann, 2018; Zuiderveen Borgesius & Poort, 2017) Therefore, consumers' reaction to dynamic pricing tactics will significantly affect their satisfaction with purchases and ensuing behavioral intentions

Customer services aim to satisfy customers' needs, desires, and aspirations regardless of prior, during, or after the transaction Therefore, ascertaining superior customer service is essential to keeping the customers satisfied Big data delivers many customer insights that businesses can utilize to captivate customers Harnessing and using the collected data intelligently can drive better customer service across the organization Big data allows firms to track negative remarks and respond to customer problems efficiently Research exploration has emphasized big data for improved customer satisfaction, notably the personalization and customization of services Many companies have begun to establish a competent customer service system through which machine learning algorithms are deployed to stimulate people’s reactions and achieve the effect of user-firm communication An online customer service system can facilitate efficient and effective searching, buying, and dispatching goods (Zeithaml et al., 2002), thereby enriching the traditional customer-firm interaction Social media can be a supplementary customer service and communication channel to provide insight into customers’ desires, concerns, and behaviors (He et al., 2013) However, although digital media have been increasingly utilized for instant response time, many customers prefer using human services rather than a computer (Mero, 2018) Thus, hybrid approaches incorporating the reciprocal capacity of human and artificial intelligence manifest terrific opportunities for deployment in digital customer service (Graef et al., 2021) Social media is becoming an efficient communication channel in which firms can collect and respond to customer inquiries quickly, thus enhancing the firm's reputation (Guo et al., 2020).

Perceived Risks of Online Shopping

Ko et al (2004) define perceived risk as “ the potential for loss in pursuing the desired outcome from online shopping.” When associated with online transactions, perceived risk includes performance, financial, convenience, and psychological dimensions (Forsythe & Shi, 2003) Empirical evidence suggests risk perception is more pronounced when consumers shop online than at a retail storefront (Tan, 1999), indicating that the lower the perceived risk, the higher the propensity for internet purchasing

Privacy has always been an important research topic for many scholars in marketing and e-commerce According to Betzing et al (2020), privacy pertains to knowing what kind of data will be utilized, how the data is acquired, how it is handled, with whom it will be shared, and how long it will be retained With the prevalence of the internet and online shopping, the issue of privacy has become even more essential In online marketing, privacy is associated with disclosing personal data on an e-commerce website and the threat of such information being unveiled (Bart, 2005; Limbu et al., 2011) The privacy incident may contain the illegitimate revealing of personal data, unsolicited contact from the e-retailer, or confidential tracking of shopping behavior (Miyazaki & Fernandez, 2001) Hence, privacy concerns individuals' control over their data disclosure and subsequent use (Román, 2007) Privacy concerns stem from cybercitizens requesting personal information in exchange for products and services The collection and storage of a large amount of data have aggravated consumer privacy concerns The convenience and relevance of personalization carry the risk of privacy violation Scholars have tried to delineate the nature of the decision-making process wherein the decision is based on risks-benefits calculation, known as privacy calculus theory (Barth &

De Jong, 2017) In sum, privacy and security have been continuously identified as obstacles to e-commerce The author follows Román (2007) to classify security and privacy into two distinct variables for increased clarity and validity

Security refers to the website's ability to protect the customers' purchasing-related data from any unauthorized use of information disclosure throughout the electronic transaction (Guo et al., 2012) In an e-commerce context, security implies consumers’ perceptions about the extent to which purchasing data and financial information are protected for confidentiality, integrity, and availability From a big data perspective, the problem is to assure that individuals (i.e., data subjects) have complete control over their personal information and to refrain from data misuse while maintaining its utility Security and privacy are significant factors of online marketing Security is usually considered critical in the relationship marketing pattern for a successful business relationship (Hirschheim & Klein, 1994) Prior studies reveal that consumers are reluctant to negotiate with e-vendors if they are aware that their personal information may be leaked out and stolen by hackers Trust, therefore, presents an essential role in dealing with perceptions of risk and insecurity Thus, the case in which e-vendors carelessly or purposefully giving out their customer data to other third parties may provoke ethical concerns among internet users, resulting in negatively perceived ethical behavior (Limbu et al.,

2011) Internet users regularly regard privacy and security as the two most ethical severe apprehension (Román & Cuestas, 2008)

A reference group comprises individuals or groups influencing our opinions, beliefs, attitudes, and behaviors Retailers regard reference groups as crucial, affecting consumers’ information interpretation and purchase intentions Reference groups may influence determining what types of merchandise consumers want to put money into and which brand of product they choose Extant literature on consumer behavior has acknowledged that reference groups influence consumers' purchasing behavior According to Eastin and LaRose (2005), social support – the resources provided by another person, is essential for group influence as it generates supportive systems among individuals Empirical evidence reinforces that peer communication significantly impacts shopping orientations, attitudes toward advertising (De Gregorio & Sung, 2010), and consumer decision-making (Sheikha et al., 2019) Likewise, social relationships, including social media communication, can significantly affect an individual’s buying behavior (Liang et al., 2011) Most consumers tend to exchange pre-purchase information with virtual companions and emphasize online compadres’ recommendations Moreover, online shoppers rely considerably on the information shared by previous purchasers, termed electronic word-of-mouth (e-WOM), and they trust the e-WOM over and above conventional advertisements (Amblee & Bui, 2011; Cheung & Thadani, 2012) Consumers often consult product reviews before buying the product on e-commerce platforms A customer’s review represents customers' thoughts, ideas, beliefs, and experiences, which strongly influence consumers’ purchase decisions on e-commerce platforms (Akar & Topỗu, 2011; Hajli,

2015) When the reviews of a particular product are positive, the consumer will be very likely to buy it Yao et al (2009) reported that positive reviews elevate purchase intentions while negative ones may discourage buying intent

Furthermore, affluent literature accentuates the effect of celebrities and influencers on individuals’ purchase behavior (Croes & Bartels, 2021; Zafar et al., 2021) Influencer marketing has been recognized as a component of seeding marketing strategies, becoming a critical constituent in the new marketing mix of fast-paced consumables enterprises (Dost et al., 2019) Some people passively accept the consequences of goods worth buying because a celebrity or an influencer has influenced them The main reasons to trust and follow an influencer on social media are physical appearance, credibility, number of followers, cultural values, and trustworthiness (Croes & Bartels, 2021)

Deception is the intentional transmission of messages to a recipient to promote a misleading belief or conclusion (Caspi & Gorsky, 2006) Deceptive practices occur when an e-vendor gives the consumer a false or potentially misleading sense of trust that differs from what the consumer may reasonably expect with decent comprehension Academic literature primarily looks at deception from marketers (objective deception) and consumer perspectives (perceived deception) (Held & Germelmann, 2018) Accordingly, perceived deception is the consumers’ perception of the responsibility of a marketer for trying to set untrue beliefs with any marketing communication (Held & Germelmann, 2018) While deceptive practice is also present in brick-and- mortar stores, it is much more striking in a digital market environment The rapid growth of e-commerce has put forward a fertile ground for cyber-fraud and deception (Freestone & Mitchell, 2004; Grazioli & Jarvenpaa, 2003) owing to the absence of in-person communication between consumers and sellers (Román, 2007) A considerable body of research has examined the consequences of deception Recent research has scrutinized the consumers' perception of deception and the associated effects In general, the perception of deception always leads to a decrease in brand attitude and a more negative evaluation of the marketer or even advertising (Shanahan & Hopkins, 2007), thereby worsening purchase intention (Xie et al., 2015) This process is particularly evident when consumers have negative prior experiences with the advertiser (Darke & Ritchie, 2007) Darke and Ritchie (2007) and Riquelme and Román (2014) contend that the impression of deceptive advertising strategies and their activation of negative stereotypes reduces the persuasive power of advertising and may lead to ineffective marketing communication The increasing concerns over deceptive marketing and other ethical behavior in online retailing can disrupt and impede the expansion of online retail and scare consumers away internet purchases

Shopping has long been and will continue to become integral to our everyday pursuit However, excessive shopping can deteriorate emotional functioning, interpersonal relationships, and economic circumstances for a small percentage of the population Nakken (1996) describes behaviors as occurring on a continuum, which can progress from patterns and habits to compulsions and addictions Correspondingly, Larose and Eastin (2002) suggest that impulsive, compulsive, and addictive shopping lie along a spectrum and portray different levels of impaired self-regulation Evidence support that it is easier for consumers to be uncontrollable during online shopping than in retail stores Marketers have consistently provided consumers with a greater chance to buy excessively and impulsively (Sun et al., 2012) by providing an immense assortment of products selection and enhancing the ease of shopping operation (Strack et al., 2006)

Impulse buying occurs when an environmental stimulus stimulates one to purchase, while compulsive buying is induced by inner states where shopping and buying become a relief While impulsive buying is reported to influence compulsive buying (Darrat et al., 2016), they are considered nonstandard purchasing behavior and are frequently described in the same spectrum

Figure 2 Problematic buying-shopping continuum

Use resulting in more positive than harmful effect

Recreational, negligible health or social effects

Compulsion Disorder Initial adverse consequences

Chronic, dependence Poorly controlled Severe consequences

Furthermore, several marketing tactics induce and foster problematic buying-shopping behavior Gilbert and Jackaria (2002) reported a high correlation between coupons, price discounts, samples, and buy one get one free versus compulsive buying Marketing cues influence the path towards and away from maladaptive consumption and potentially addictive behaviors over time (Martin et al., 2013) Besides, individuals with the potential buying- shopping disorder appear to be more sensitive to marketing tactics (Chauchard et al., 2020) Thus, sensitivity to marketing was put forward as a determinant of buying-shopping disorder (Duroy et al., 2014)

In the psychological literature, problematic buying-shopping behavior has been increasingly researched, typically referred to as compulsive buying, compulsive buying shopping disorder, and shopping addiction, among the other derivatives The classification of problematic buying-shopping has been the subject of considerable controversies Some scholars propose that problematic buying-shopping is a disorder on the impulse control and obsessive-compulsive spectrum (Christenson, 1994; Ridgway et al., 2008), while others contend that it should be considered behavioral addiction (Rose & Dhandayudham, 2014) According to Shaffer (1996), it is not material goods but the interplay between addicted individual and their addiction subject that matters Thus, readily available the subject of addiction is a critical antecedent for the maintenance and development of addiction Also, in considering shopping as an addiction, it is imperative to distinguish between normal and excessive behaviors and identify thresholds at which normal behaviors can become problematic and eventually be classified as an addiction (Harris, 2000) Numerous terms have been used interchangeably to signalize this particular behavior (Maraz et al.,

2016) This status-quo calls for consistent terminology that would specify this consumption behavior and minimize confusion

The growing accessibility to technology and the widespread internet favored the maintenance and development of problematic online buying- shopping (Niedermoser et al., 2021) Well-known e-commerce sites with appealing characteristics can reduce self-regulation (Larose & Eastin, 2002) From an individual perspective, impotent self-control is a significant cause of online shopping addiction Online shopping addiction indicates an individual’s incapability to regulate online shopping (Jiang et al., 2017) Addictive shoppers are tempted to buy things they do not need but fail to resist the urge or control themselves in the virtual storefront After experiencing online shopping, consumers with low self-control are more inclined to indulge in internet purchasing, resulting in higher susceptibility to shopping addiction (Zhang et al., 2019) Online compulsive buying – a subtype of internet addiction, is highly related to problematic internet use (Suresh & Biswas, 2020); Trotzke et al

(2015) Adamczyk (2021) reveals that female online shoppers with positive attitudes and more intensively do online shopping are incredibly susceptible to compulsive buying He confirms that “…online shoppers show a stronger susceptibility to compulsive buying than offline shoppers,” and it is improbable for consumers to regulate their cravings incredibly when they are utterly engrossed with internet purchasing than shopping on physical storefronts (Augsburger et al., 2020) The enticement of new products lures consumers to buy compulsively However, existing literature has not determined whether problematic internet shopping is merely the virtual equivalent of conventional compulsive buying shopping disorder or could be hypothesized as a distinct entity or a subtype of internet addiction (Müller et al., 2022) Previous studies revealed several unfavorable consequences associated with problematic buying-shopping behavior such as bankruptcy, psychological distress, and jeopardized social relationships

Interpersonal relationships refer to personal connections in the public life of an individual Thanks to the popularity of the internet, social networking sites have evolved rapidly and become an essential source of social interaction

Data Ownership Awareness

The ownership of data is considered to be “the possession of complete control over the data and its rights including, but not limited to access, creation, generation, modification, analysis, use, sell, or deletion of the data, in addition to the right to grant rights over the data to others” (Asswad & Marx Gómez,

2021) The notion of data ownership has been identified as one common issue that is confronted in using big data Commercial firms typically decide how consumer data are collected, the types, and how consumers’ data will be utilized, resulting in customers being granted varying degrees of control over their data Line et al (2020) highlight that the transparency in the data management process and the uniformity of knowledge between firms and consumers regarding the governance of personal data will determine the different levels of data ownership More specifically, an inclusive and transparent mechanism in which users can clarify how their data is collected and handled by the firms defines a high degree of data ownership In contrast, a knowledge asymmetry between customers and firms regarding the data types and their resulting manipulation characterizes low levels of data ownership According to the theory of the adoption of innovation (Rogers, 2003), awareness is the first step in five-step decision-making, all-embracing awareness, interest, evaluation, trial, and adoption In the first step, an individual is exposed to an idea or innovation but lacks information about that idea or invention and is less likely to seek more information about the idea or innovation Before proceeding further, people must first be aware of the ideas and determine whether they are worth adopting Data ownership is multifaceted, including legal, ethical, and technological considerations Thus, the author affirms that data ownership awareness is the extent to which an individual is conscious of the existence of associated legitimate rights and control measures they are entitled to and exert toward their digital data The author also contends that the attitude of consumers will be shaped once they are fully aware of data ownership

Correspondingly, the dearth of transparency regarding data collection introduces complexity when trying to figure out what happens to consumers’ data—i.e., how it is stored, processed, or disseminated to other parties This status quo underscores the necessity to obtain control over the manipulation and accessibility of consumer data Existing literature from a juridical perspective suggests that data ownership is predominantly associated with the privacy of the individual (Fadler & Legner, 2021) Likewise, privacy is one aspect of data ownership (Jagadish et al., 2014), wherein the former, within the firm–consumers relationship, reflects “the extent to which a consumer is aware of and can control the collection, storage, and use of personal information by a firm” (Beke et al., 2018) More specifically, the involvement of different entities who have authorized access to such data, often without prior consent from the data subject (Mendelson & Mendelson, 2017), accrues more complexity to the concept of data ownership (Galvin & DeMuro, 2020).

METHODOLOGY

Conceptual Models and Hypotheses Development

3.1.1 Study I - Impact of Perceived Benefits and Risks of Online Shopping on Data Ownership Awareness

Study I aims to provide insights into the issue of data ownership awareness based on consumers' perceptions regarding the benefits and risks of shopping online Specifically, the current research examines whether the consumers' perception of the advantages and disadvantages of shopping online could affect their consciousness of data ownership This study will analyze the data collected from 630 online consumers in Taiwan and Vietnam To comprehend data ownership awareness, the author first assigned the first-order constructs in a fundamental independent-dependent research model to provide insights into the structure and the mechanism in which each component of these variables prevails over the other

Distinct from shopping in a physical storefront, more significant efforts must be dedicated to online consumers before making a final purchase decision These efforts include but are not limited to seeking out product details, researching previous customer reviews, comparing prices across the e-retailer website, and asking for customer services Perceived benefits of e-commerce websites employing big data analytics have been reported to positively affect consumer responses (Le & Liaw, 2017) Users whom voice privacy concerns fail to act accordingly but freely share their information in return for improved services and personalization It means they pay less attention to tackling data misuse, and their behavior does not align with their concerns Extant literature has documented the difference between the individual’s declared privacy apprehension and actual behavior and named it the “privacy paradox.” The privacy paradox is frequently interpreted as a trade-off between risks and benefits Due to this trade-off, the willingness to reveal individual information varies In essence, the privacy calculus paradigm posits that consumers rely on the risk-benefit judgment to determine “with whom, how, and to what extent personal data is shared” (Li et al., 2010) Only a small proportion of people take the necessary measures to secure their privacy To remedy the information disclosure, individuals strive to maximize utility and minimize risk using rational calculus for better decision-making Those who expressed more significant concern about privacy share less personal data and hold a more pessimistic attitude toward information disclosure (Dienlin et al., 2021)

There have been many possible reasons individuals divulge personal data despite elevated privacy concerns The concept of privacy calculus was introduced to represent a rational decision where individuals reveal personal information after weighing benefits in opposition to risks (Lee et al., 2013) Another probable explanation for online information disclosure is the scarcity of risk awareness and lack of knowledge of possible consequences of online self-disclosure due to digital literacy (Hoffmann et al., 2016) In addition, privacy concerns and information disclosure decisions are influenced by context (Waldman, 2018) Based on comparative judgment, an individual is more likely to disclose if she comprehends that others are willing to do so (Acquisti et al., 2012) Another school of thought perceives that disclosure can be emotionally manipulated A favorable attitude toward a website is associated with a higher inclination to divulge information (Li et al., 2008) To put it straightforwardly, internet users may neglect their reported privacy concerns and tend to overshare their data when interacting with an entertaining website Taken together, in accordance with previous discussion, we put forward the subsequent hypotheses:

H 1-1 : Information search has a significant effect on data ownership awareness

H 1-2 : Recommendation systems have a significant effect on data ownership awareness

H 1-3 : Dynamic pricing has a significant effect on data ownership awareness

H 1-4 : Customer services have a significant effect on data ownership awareness

H 1-5 : Privacy has a significant influence on data ownership awareness

H 1-6 : Security has a significant influence on data ownership awareness

H 1-7 : Group influence has a significant effect on data ownership awareness

H 1-8 : Deception has a significant influence on data ownership awareness

The proposed research model used to test the hypotheses mentioned above is presented in Figure 3

Figure 3 Research model for Study I (lower-order construct)

Furthermore, using more abstract levels of constructs is consistently becoming prominent for different theoretical and empirical justifications (Jarvis, 2003) Polites et al (2012) emphasize the significance of evaluating the conceptual relationships between the lower-order constructs and their indicators and between first-order and second-order constructs Likewise, the author appreciates that multiple related latent variables can also converge at a higher level predicting a common trait According to Hair et al (2010), including a higher order in an SEM analysis consumes less degree of freedom, resulting in a more parsimonious model and a better model fit Consequently, the author proposes a second-order path model to observe a general predisposition of a group of respective first-order constructs as presented in Figure 4

Figure 4 Research model for Study I (higher-order construct)

Perceived benefits of online shopping (PBOS)

Perceived risks of online shopping (PROS)

Perceived benefit is the consumers’ belief about the extent to which they become superior from the online transaction with a particular online purchase or the consumers’ subjective perception of gain from shopping online (Forsythe et al., 2006) Perceived benefit is an essential factor influencing consumers’ decisions when shopping online Some researchers have discussed several benefits that online shopping provides consumers that are not available in traditional shopping channels For example, online shopping provides consumers with more information and opportunities to compare products and prices, with more fabulous product selection, convenience, and ease of finding desired products online In the online shopping environment, businesses embrace massive consumer data to deliver improved information search results through efficient recommendation systems to compile a personalized pricing list and provide better customer services The author hypothesizes that the more values consumers perceive, the more likely they will make a purchase decision Also, benefits from online shopping under the prevalence of data-driven technologies may play a catalyst, facilitating the consciousness of data ownership Thus, the author postulates that:

H 2-1 : Perceived benefits of shopping online significantly influence consumers' data ownership awareness

Correspondingly, the advantages of e-shopping also present potential risks that the consumers must notice Perceived risks are consumers' personal expectations of suffering a loss in pursuing the desired outcome Consumers perceive risk because they face uncertainty and potentially undesirable consequences due to purchases Thus, the more risk they perceive, the less likely they will purchase Several authors have observed that the perceived risk in e-commerce has a negative effect on shopping behavior on the internet, attitude toward usage behavior, and intention to adopt e-commerce (Katta & Patro, 2020; Pi & Sangruang, 2011) Many scholars have divided perceived risk into several parts and measured the risk based on different survey samples, demonstrating that perceived risk exists in online shopping As explained earlier, four risk groups can be perceived by online consumers, which are associated with the embrace of big data analytics and big data technologies in e-commerce, i.e., privacy, security, group influence, and deception The perceived risk decreases the willingness of consumers to buy goods over the internet (San Martín & Camarero, 2009; Tan, 1999) Consequently, the author hypothesizes that individuals with higher risk perceptions are less likely to disclose personal information, significantly influencing their data ownership awareness Accordingly, the following hypothesis is formulated:

H 2-2 : Perceived risks of online shopping significantly influence consumers’ data ownership awareness

Apart from exploring the hypothesized relationships, the author investigates the cross-cultural differences among Vietnamese and Taiwanese online consumers for specific hypotheses The justification mentioned earlier and findings from the literature review enables us to speculate compelling insights when exploring different perceptions of shopping online and its relations with data ownership awareness

3.1.2 Study II - The Mediating Role of Online Interpersonal Relationships and Data Ownership Awareness

A growing body of literature has focused on the effects of perceived benefits and risks of shopping online on consumer attitude and purchasing behavior No study has ever considered the relationship with problematic internet shopping According to Le and Liaw (2017), the advantage of applying BDA in e-commerce positively influence customers’ intention and behavior; the inverse relationship is found for its negative components Consistent with the research rationale presented previously and aligned with Le and Liaw

(2017), the author hypothesizes that consumers who perceive more significant benefits from internet shopping are more likely to purchase online Also, benefits from online shopping under the prevalence of data-driven technologies might be an accelerator, facilitating the occurrence of problematic internet shopping Likewise, individuals with higher risk perceptions may find online shopping too risky and thus become less likely to purchase online As problematic buying-shopping behavior only occurs at the continuum end of the consumption operation, with a critical antecedent that consumers devote a considerable amount of time to online retailing stores Thus, the following hypotheses are proposed:

H 3-1 : Perceived benefits have a positive and significant effect on the problematic internet shopping

H 3-2 : Perceived risks have a negative and significant effect on the problematic internet shopping

Moreover, the degree to which OIR and DOA mediate the relationship between PBOS, PROS, and PIS remains unexplored However, several anecdotal pieces of evidence have supported this association Firstly, efforts have been dedicated to scrutinizing the mediating role of interpersonal relations in the association between psychological factors and internet addictive behaviors (Baek et al., 2016; Fioravanti et al., 2012; Ryu et al., 2018) For example, interpersonal relationships were a protective factor against smartphone addiction and partially mediated the relationship between stress and problematic smartphone use (Baek et al., 2016; Kwon, 2021) Secondly, the association between privacy and social capital in a social networking site context was reported (Stutzman et al., 2012) Scholars have demonstrated the close relationship between privacy and disclosures, emphasizing the complex privacy-disclosure trade-off in people’s intention to disclose information (Krasnova et al., 2010) Self-disclosure is identified as a precondition for the establishment of interpersonal relationships Kwak et al (2014) show that self- disclosure behavior seemingly and positively alters the scale and intimacy of interpersonal relationships Likewise, the linkage between self-disclosure and social support was documented (Trepte et al., 2018), highlighting the direct compensation of social support for sharing and disclosing information Lastly, the existing scholarly research reveals the essential role of parents and peers in shaping individuals’ engagement or avoiding risky behavior by balancing risks and benefits (Resnick et al., 1997) For instance, online social support positively relates to internet addiction (Lin et al., 2018; Wang & Wang, 2013) Thus far, no research has scrutinized the proposed relationships among PBOS, PROS, OIR, DOA, and PIS

Consequently, the author contends that examining the mediation effect of OIR and DOA can better elucidate the complex intervening mechanism of how consumers’ perceptions are associated with PIS To understand the mechanisms guiding the relationship between PBOS, PROS, and PIS, the author proposes multiple mediation models, including parallel and sequential mediation, to explore how PBOS and PROS influence PISB (Hayes, 2018) Multiple mediation analysis allows the researcher to examine a higher level of complexity in a process, thus broadening the understanding of the phenomenon (Mathieu et al., 2008) Parallel mediation infers that all constructs mediate the relationship between independent and dependent variables comparably, allowing a simultaneous test of each mechanism while accounting for their shared association

3.1.2.1 The Mediating Role of Online Interpersonal Relationships

Research on social influence in consumerism is critical, particularly among adolescents who are highly susceptible to such effects Adolescents are consistently more vulnerable to reference group influence than adults and can be considered an “at-risk” cohort For adolescents, besides their parents, peers are also essential for developing personal relationships Roberts et al (2008) examine how susceptibility to peer and parental influence impacts adolescent materialism and compulsive buying Their findings suggest that peer normative influence has a relatively more substantial effect than the informative parental role on materialism and compulsive buying In addition, scholars have elucidated internet gaming disorder in terms of interpersonal difficulties As such, adolescents with internet gaming disorder have higher conflicts with their parents and receive little attention from them (Lee, 2005) In addition, the quality of peer relationships was inversely correlated with problematic internet gaming (Yang, 2007) Low-quality interpersonal relationships can expose individuals, especially adolescents, to an increased risk of developing problematic behaviors such as problematic internet use (Milani et al., 2009) and suicidal ideation (Sewall et al., 2020) Likewise, the author contemplates that poor interpersonal relationship quality may be associated with an increased risk of problematic internet shopping Thus, the following hypotheses are proposed:

H 4-1 : Online interpersonal relationships mediate the relationship between perceived benefits and problematic internet shopping

H 4-2 : Online interpersonal relationships mediate the relationship between perceived risks and problematic internet shopping

3.1.2.2 The Mediating Role of Data Ownership Awareness

In the present study, data ownership awareness refers to the appreciation from which users consent to and understand their legitimate rights and control measures they can exert toward their personal data Prior studies reported that perceived privacy risk significantly and negatively impacts the amount of information disclosed (Krasnova et al., 2010) Internet users regulate how much information they disclose based on the threats of privacy violations Kim et al

(2019) examine factors affecting the willingness to provide privacy information based on the privacy calculus and degree of personalization perspective Their findings suggest that people do not pay much attention to perceived privacy risks when providing privacy information for a better-personalized service Besides, the increase in online purchasing activities has made consumers more willing to disclose their personal data (Gao et al., 2020) This tendency has put forward higher business requirements to ensure consumers' information Hence, the research puts forward the following hypotheses:

H 5-1 : Data ownership awareness mediates the relationship between perceived benefits of online shopping and problematic internet shopping

H 5-2 : Data ownership awareness mediates the relationship between perceived risks of online shopping and problematic internet shopping

Nevertheless, it is unclear whether the mediating effects of OIR and DOA (if any) are equally important or is one more important than the other Therefore, the author counts on the serial mediation hypothesis to isolate the effect of the proposed mediators Serial mediation casts aside the assumption of no causality between mediators and describes how the distinct mediator variables of a proposed model are connected in a particular way along a chain (Hayes, 2012, p 14) According to Tan et al (2016), serial mediation may reveal mediators' unique effects and compare the indirect effects' strengths (Hayes, 2018) Thus, the author explicitly tests a model in which PBOS and PROS are related to OIR and DOA, which are, in turn, associated with PIS Thus, the following hypotheses were proposed:

H 6-1 : Online interpersonal relationships and data ownership awareness sequentially mediate the relationship between perceived benefits and problematic internet shopping (PBOS – OIR – DOA – PIS)

H 6-2 : Online interpersonal relationships and data ownership awareness sequentially mediate the relationship between perceived risks and problematic internet shopping (PROS – OIR - DOA – PIS)

Data Collection

The data used in this study was collected online using a survey questionnaire The questionnaire includes a range of questions or statements requiring respondents to answer or indicate the extent to which they agree or disagree with the given statement Online consumers residing in Taiwan and Vietnam who had shopped online over the last twelve months were invited to participate in the study A holdout sample was also collected via Amazon Mechanical Turk (MTurk) – a crowdsourcing marketplace, to make up a sufficiently large dataset for the analysis MTurk is an online data collection platform that has been shown to produce data comparable in quality and reliability to those provided by student and community samples (Buhrmester et al., 2011) Respondents were asked to rate how they agreed or disagreed with statements on perceived benefits, perceived risk of online shopping, online interpersonal relationships, data ownership awareness, and problematic internet shopping behavior The survey also requires respondents to provide general information such as gender, gender, education level, marital status, and online shopping budget for classification and comparison purposes

All constructs items were operationalized on a seven-point Likert scale ranging from “strongly disagree” to “strongly agree.” The author employs construct items that were validated and then tailored to fit this study's specific context The authors constructed the initial questionnaire in English and translated it into Vietnamese and Chinese Two doctoral candidates and one professor at the business administration department who is bilingual in Chinese and English were then consulted to back-translate the questionnaire into English to ascertain that the items' meaning and wordings were not compromised Pilot tests were conducted among twenty volunteers in Vietnam and Taiwan, providing evidence for further content validity amendments

The questionnaires’ cover page explains the objectives of the research and its academic contribution, the confidential declaration, and the anonymous nature following the principles of the Declaration of Helsinki Also, the author incorporated a screening question at the beginning of the questionnaire to exclude those who had not participated in online purchasing activities over the last twelve months (i.e., ineligible to participate in the study) The final version of the questionnaire (see Appendix A-C) was circulated to social network channels from June to August 2020 to obtain the desired number of respondents The author hosted our online questionnaires on Survey Cake - a professional survey platform The survey management system allows us to detect duplicate responses by inspecting each answer's IP address which provides more evidence for excluding invalid responses

Additionally, the screening data procedure revealed that 147 responses were eliminated due to the inability to pass embedded attention check questions, leaving a usable sample of 1,039 Participation in the research was entirely voluntary, and all participants gave written informed consent to participate The survey commenced with questions regarding demographic information, followed by a battery of standardized questionnaires measuring key constructs of our research No monetary compensation was offered, except for those recruited from MTurk, who received $1.50 for participation

According to Hair et al (2017), the power analysis is recommended to determine the required sample size for a specific research model Therefore, a priori and posthoc power analysis utilizing G*power analysis software to ensure the adequacy of the sample size This research follows the recommendation of (Dattalo, 2008) to set up the input of the power analysis (𝛼

= 0.05 and 𝛽 = 0.95 for error Type I and II, effect size = 0.15, and the number of predictors based on a specific research design Using suggested minimum values by Cohen (1988), a priori G*Power calculation specified that a sample size of 195 would be required In addition, the post-hoc G*Power estimate for a minimum R 2 of 0.10, a sample size of 1,039, and four predictors indicated that the statistical power achieved using the study’s sample size of 1,039 was 0.99, which surpassed Cohen’s (1988) recommended benchmark, thus satisfying the adequacy of our sample size.

Measurements

Much of the measurement instruments utilized in the present study were adapted from the literature and subjected to rational adjustment to meet the specific requirements of this research Perceived benefits of shopping online indicate the subjective evaluation of the usefulness or advantages of internet purchasing based on four dimensions: information search, recommendation system, dynamic pricing, and customer services (Le & Liaw, 2017) These four dimensions are critical for the success of organizations within the big data framework, where commercial firms rely increasingly on BDA to improve their business performance Likewise, the author adapted some indicators from Le and Liaw (2017) and Román (2007) to construct a scale measuring the perceived risk of internet shopping This 16-item scale comprises four privacy, four security, four group influence, and four deception items and is operationalized by a seven-point Likert scale, ranging from 1 (strongly disagree) and 7 (strongly agree) Data ownership awareness indicates how consumers are conscientious of the rights and practices they can exert toward purchasing data Based on a holistic literature review and expert consultancy, the author designs an 8-item scale to measure consumers’ data ownership awareness The abovementioned question items are measured using a seven-point Likert scale

(1 = strongly disagree) to (7 = strongly agree) Furthermore, the scale to measure consumers’ online interpersonal relationships (OIR) was adapted from the Inventory of Interpersonal Problems 32 (Barkham et al., 1996), which tapped into the difficulties individuals confront in their interpersonal relationships The author attempted to simplify and exclude all irrelevant items and reworded all remaining items that refer to online acquaintances ascribed to virtual connection to fit the purpose of the current study That way, online interpersonal relationships correspond to cyberspace consumers' challenges and difficulties A seven-point Likert scale was operationalized to measure online interpersonal relationships, ranging from 1 (strongly disagree) to 7 (strongly agree)

(Agag et al., 2016; Le & Liaw, 2017; Román,

Data ownership awareness 01 08 Expert consultation & self-collaboration Online interpersonal relationships 01 08 (Barkham et al., 1996)

Problematic internet shopping 03 12 Zhao et al (2017) and self-collaboration

Note: IS = information search, RS = recommender systems, DP = dynamic pricing, CS = customer services, PR = privacy, SC = security, GI = group influence, DC = deception

Problematic internet shopping was measured using the Online Shopping Addiction Scale (Zhao et al., 2017) with judicious modification to match the context of this research The original 18-item scale includes six subscales based on addiction's core components: salience, mood modification, tolerance, withdrawal, conflict, and relapse (Griffiths, 2005) A seven-point Likert scale administers the scale from “strongly disagree” to “strongly agree.” The higher the total score, the more severe the consumer experiences the symptoms of problematic internet shopping The original scale was subjected to psychometric properties evaluation resulting in a minor revision of a 12-item scale (Duong & Liaw, 2022b) The scale was highly reliable, as evidenced by Cronbach’s alpha value of 0.928.

Data Analysis

3.4.1 Data Cleaning and Data Pre-processing

The collected data was submitted to straight-lining patterns, unengaged responses, and missing value assessment to minimize bias and variability of the answers before proceeding further Provided that multivariate outliers may lead to biased findings, the author diagnosed multivariate outlier incidence using the Mahalanobis distance, D 2 , measure A case with a probability of a D 2 value less than 0.001 is considered for removal Thirty-seven observations were detected, and 1,039 sets of responses were retained The data was then assessed if it is normally distributed using skewness and kurtosis by SmartPLS 3.2.8 software The recommended thresholds proposed by West et al (1995) and Kline (2016) were opted for examining the skew and kurtosis values The analysis revealed that the collected data were non-normally distributed In addition, the multivariate normality test results show that Mardia’s multivariate coefficient p-value is less than 0.05, further confirming the multivariate non-normality of the data Therefore, Smart-PLS – i.e., a non-parametric analysis tool, is deemed a suitable statistical software for the present analysis Besides, given that the data were self-reported, the potentiality of the common method bias (CMB) problem was examined The author employed ex-ante and ex-post procedures to remedy the CMB concern (Conway & Lance, 2010) Maintaining the respondents’ anonymity is recommended to lessen the evaluation apprehension and diminish method bias Therefore, the survey was designed to be anonymous, and respondents were informed that there were no precise or imprecise responses, thereby giving their honest answers This remedy aims to minimize potential CMB ex-ante by reducing social desirability issues Second, the author counterbalanced the order of dependent and independent variables, as recommended by Chang et al (2010) and Podsakoff et al (2003) Third, the results from Harman’s one-factor test reveal that the first component that emerged from the unrotated factor solution explains 19.11% of the total variance, which was far below the 50% threshold proposed by Wong et al

(2015), indicating the absence of CMB in this study Provided that reassuring Harman’s one-factor test does not thoroughly negate the risk of CMB, the test employed by Pavlou et al (2007) was implemented with the examination of the correlation matrix In no case did the author find any pair of constructs with their correlation exceeding 0.85, further supporting that CMB was not of significant apprehension Consequently, the collinearity assessment was executed by assessing the inner collinearity assessment function (VIF) Our results revealed that VIF values were far below the benchmark of 3.3 (Hair et al., 2010; Kock, 2015), indicating the absence of collinearity concern

As not all scales used in this study were established and validated through prior literature, we decided to conduct exploratory factor analysis because of the varying results from previous publications concerning the dimensionality of studied constructs Exploratory factor analysis using the principal axis factoring approach with the Varimax rotation method was used to reduce the measurement items by recognizing the latent components The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy, internal consistency (Cronbach's 𝛼), and Bartlett's test of sphericity were evaluated These standard tests were employed to specify the suitability of the measurement item scales for exploratory factor analysis (Kaiser, 1974) More specifically, a set of variables is deemed suitable for factor analysis if its KMO value surpasses 0.5 and the Bartlett test score attains statistical significance The KMO for the present study was 0.934, and Bartlett’s test of sphericity was statistically significant (𝜒 2 = 25285.305, p < 0.001); thus, our sample was considered suitable for the exploratory factor analysis We inspected the rotated factor loading matrix to identify measurement items with factor loading exceeding 0.5 The Varimax rotated solution returned ten dimensions for this study, accounting for 69.58% of the total variance The Cronbach’s alpha for each factor surpassed 0.70, except for IS (0.698) and RS (0.697), which were at the marginal threshold, signifying a high level of reliability A minor revision was amended by examining the factor loadings and cross-loading items Consequently, some items with insignificant loadings were excluded from the final scale, including four items of dynamic pricing, one item of information search (IS1), one item of recommendation system (RS4), one item of privacy (PR2), and four items from data ownership awareness (DO2, DO3, DO7, and DO8) A final scale comprising 29 items was used for Study I

3.4.3 Partial Least Squares Structural Equation Modeling

The partial least square structural equation modelling (PLS-SEM) was used to evaluate the research model This prediction-oriented structural equation modelling technique allows researchers to analyze latent constructs, specifying direct and indirect relationships PLS-SEM allows consideration of the construct's reliability and validity simultaneously PLS-SEM is a robust analytical technique due to the “minimal demands on measurement scales, sample size, and residual distributions” (Chin, 2010) The current study utilized PLS-SEM through the SmartPLS software 3.2.8 to analyze data This method is chosen because it can handle complex models, making it a primary priority for consumer behavior research

The model was assessed following the suggested two-step approach to the structural equation model, including the assessment of the measurement model and the structural model (Anderson & Gerbing, 1988) While the measurement model quantifies the relationship between indicators and their underlying latent variables, the structural model represents the causal relationship among latent constructs (Byrne, 2016) The examination of the measurement model concentrates on its reliability, internal consistency, convergent validity, and discriminant validity (Hair et al., 2017) Convergent validity is assessed first by inspecting whether each indicator's loading is 0.6 or more (Bagozzi & Yi, 1988) The Cronbach's alphas and the composite reliability for each construct should exceed the benchmark value of 0.7 Besides, convergent validity is assessed by examining the loadings and the average variance extracted (AVE) values The minimum threshold for AVE is 0.5 An accurate discriminant validity assessment is crucial because it ensures that each construct is empirically unique and captures a phenomenon not represented by other constructs in a statistical model To this, discriminant validity is established based on the heterotrait-monotrait ratio of correlations (HTMT), a said procedure superior to the commonly considered Fornell-Larcker (1981) criterion and assessments of cross-loadings (Henseler et al., 2015) Model estimation was executed with R 2 , Q 2 , effect size (f 2 ), importance-performance map analysis (IPMA), and PLSpredict (Hair et al., 2019; Shmueli et al., 2016)

The current study employs partial least square structural equation modelling to assess significant differences in outer weights, outer loadings, and path coefficients The PLS-SEM approach was opted for to perform the multigroup analysis because of its ease and robustness (Henseler, 2012), whereby a bootstrap option provided path coefficients of the differences and the p-values of the differences The author performed a multigroup analysis to determine if significant differences existed between Taiwanese consumers (322) and Vietnamese consumers (308) The sample sizes were deemed sufficient to obtain a 5% significance level and statistical power of 80% (Hair et al., 2017)

It is imperative to confirm the existence of measurement invariance across constructs before comparing model estimates between groups (Sarstedt et al.,

2011) Thus, the measurement invariance of composite models’ analysis (MICOM) procedure was executed to verify whether the model estimators' differences did not originate from different interpretations in the composite variables MICOM requires the fulfilment of the following criterion: (1) the configurational invariance, (2) the compositional invariance, and (3) the equality of composite means and variances (Henseler et al., 2016b) For configural invariance, the indicators, data treatment, and algorithm settings were inspected to see whether they were identical for the two subsamples Next, using 5,000 permutation tests, we examine if compositional invariance is present statistically The last step evaluates the equality of the composite mean values and variances by examining whether or not the original differences between the mean and variance falls between the 2.5% and 97.5% boundaries Once the measurement invariance is confirmed, the group differences between two identical models can be unraveled by applying Henseler’s multigroup analysis in SmartPLS 3.2.8 This approach does not rely on distributional assumptions but is based on bootstrapping combined with a rank-sum test to determine the existence of multigroup differences According to Matthews

(2017), the method benefits global researchers conducting cross-cultural studies The multigroup analysis will be carried out for Study I and Study II to examine whether significant differences exist between Taiwanese and Vietnamese consumers

Mediation refers to a mechanism in which a third variable intervenes the relationship between two relevant constructs That is to say, a change in the exogenous construct results in a difference in the mediator, causing a change in the endogenous construct in the PLS path model According to Hair et al

(2019), PLS-SEM is a superior approach for mediation analysis, provided that it examines the overall theoretical model in the estimation process and rules out measurement errors in the analysis, thus consequently reducing bias If the confidence intervals do not include zero (p < 0.05), it can be concluded that the mediated effects are statistically significant (Preacher & Hayes, 2008) The single-mediator model is shown in Figure 6, where the variables X, M, and Y are in rectangles, and the arrows represent relations among variables Noted that a represents the association of X to M, b denotes the relation of M to Y adjusted for X, and c’ is the relation of X to Y, adjusted for M The symbols e 2 and e 3 represent residuals in the M and Y variables, respectively

Figure 6 A conceptual diagram of the mediation model

The indirect effect of X on Y through M can be quantified as the product of a and b (i.e., ab) The total effect of X on Y is quantified with the unstandardized regression weight c The total effect of X on Y can be expressed as the sum of the direct and indirect effects: c = c’ + ab Equivalently, c’ is the difference between the total effect of X on Y and the indirect effect of X on Y through M – i.e., c’ = c – ab

Bootstrapping is a nonparametric technique that repeatedly samples the data set and estimates PLS path relationships for each subsample According to Preacher and Hayes (2008), by repeating this process thousands of times, an empirical approximation of the sampling distribution of ab is built and used to construct confidence intervals for the indirect effect Bootstrapping has been acknowledged as one of the more accurate methods for examining mediating effects (Zhao et al., 2010) The single mediation model can be extended to a

Dependent variable (Y) c model with multiple mediators, of which such mediators are presented in parallel or sequential manners Integrating various mediators in a proposed model allows us to investigate a higher level of complexity sequentially and broadens the knowledge horizon of a particular phenomenon Mediation analysis will be carried out in Study II to examine the mediating role of online interpersonal relationships and data ownership awareness on the relationship between perceived benefits, perceived risks of shopping online and problematic internet shopping behavior

3.4.6 Big Data Analysis - Data Mining

Data mining is a significant area of data sciences that seeks to uncover hidden patterns and trends in data, often massive databases The standard methods of data analysis using data mining methods include classification, regression analysis, clustering, association rules, characteristics, deviation analysis, and web page mining, to name a few Classification is probably the most commonly utilized data mining approach to extract rigorous insights from massive amounts of data Classification and prediction tasks are the most widely used type of data mining, aiming to classify the given data records into one of the many possible cases already known Classification algorithms are supervised learning methods that uncover unexplored correlations or associations between the targeted and the input variables (Maimon & Rokach,

2005) However, selecting a classifier to utilize in a data mining task is difficult Dogan and Tanrikulu (2013) suggest that the nature of the data is a determining factor influencing the provision of the best solution for a specific classification task Hence, it is reasonable for each algorithm to build several models, pick the one with the highest performance, and select the best implementation

Figure 7 The process of data mining

RESULTS AND DISCUSSION

Descriptive Statistics

Observations with missing values were excluded during data collation

A total of 1,039 complete cases were used for further statistical analysis, of which 308 (29.64%) were collected in Vietnam, 322 (30.99%) usable responses were gathered from Taiwan, and 409 usable responses were obtained from MTurk (39.36%) The final sample includes 575 females (55.35%) and 464 males (44.65%), of which 535 (51.49%) live with their partners On average, respondents in our sample have 12.86 years of internet experience (SD = 6.41) They spend almost 6.54 hours on internet usage (SD = 3.79) but only spare approximately 1.81 hours per day for online shopping activities About one- third of the respondents were under 25, followed by 23.87 per cent of those older than 35 Regarding education, 696 (66.98%) have a university degree, and 243 (23.38%) have obtained post-graduate certification Differences in sociodemographic characteristics between samples are reported in Table 3

1.69 (1.24) 1.22 (2.05) 2.35 (2.19) Note: Numbers in parentheses of [5-7] are standard deviations

A more significant proportion of females (77.92%), relative to males (22.08%), were in the Vietnam group, compared to the Taiwan group (60.25% female, 39.75% male) and MTurk (34.47% female, 65.53% male) In addition, a larger percentage of consumers in the MTurk group reported living with a partner or being in a relationship with others (66.51%) than those in the

Vietnam group (45.13%) and Taiwan cohort (38.51%), respectively Within the Vietnam group, the proportion of those who aged less than 25 (62.01%) was higher than the corresponding percentages in the Taiwan group (13.35%) and MTurk (6.60%) Lastly, the portion of individuals who have yet completed higher education compared to graduates and post-graduate degree holders’ school in the Vietnam group (15.26%) was higher than the percentage of the high school cohort in Taiwan (8.39%) and MTurk (6.36%)

In our sample, males have higher online shopping usage (p