Therefore, the objective of this research is to explore the relationship between trust, commitment, loyalty and retention, the four main areas identified in various previous studies on c
INTRODUCTION
BACKGROUND
Over the past decades, the Internet has transformed social life by enabling virtual communication and allowing people to buy almost anything without visiting stores E-commerce and online shopping have created new, more efficient distribution channels beyond traditional retail while saving customers time As a result, online shopping has grown exponentially, driven by rising e-business revenues and a surge in Internet-based transactions.
Vietnam's e-commerce sector benefits from government support, coordinated by the Ministry of Industry and Trade, which established the E-commerce Development Centre to monitor growth, analyze market trends, and craft strategies to advance e-commerce in Vietnam The centre continually tracks the Vietnam e-commerce landscape and publishes annual reports that outline current conditions, challenges, and recommended actions for stakeholders seeking to capitalize on opportunities in this dynamic market.
From 2006 to 2009, the share of Vietnamese enterprises applying e-business rose from 8% to 12%, according to the E-commerce Development Centre's 2009 report This uptick suggests more products will be supplied online and that Vietnamese consumers can search for and purchase goods without visiting supermarkets or stores However, traditional buying habits—such as the close, communicative relationship between buyers and sellers and the ease with which sellers can introduce new products through local convenience stores—can hinder the development of online business in Vietnam Additional barriers to online shopping cited in the media include electronic payment systems and the banking and network infrastructures that support them.
PURPOSES AND RESEARCH QUESTIONS
E-commerce in Vietnam has government and commercial enterprise attention but the number of customers shopping online is still limited People are not yet ready to trade online Moreover, with the limited number of current customers, what should companies do to keep them and develop close relationships with them? How can companies ensure customer product repurchase or recommend others people to use them?
Research in Vietnam has largely focused on e-commerce development from the perspective of the firms implementing it, while scant attention has been paid to how to attract and retain customers who shop online This study aims to examine the relationships among trust, commitment, loyalty, and retention—the four core areas identified in prior customer relationship marketing research—and to determine which factors drive customer loyalty and retention in Vietnamese online retail Building on these relationship patterns, the paper discusses how to encourage and sustain online shopping in Vietnam and offers practical recommendations for online retailers Two research questions arise to guide the inquiry.
1 What are the antecedents of trust and commitment?
2 How do trust and commitment influence customer loyalty and customer retention in online shopping in Vietnam?
METHODOLOGY
The objective of research is to test the hypothesis to find out the relationship between variables in online shopping in Vietnam lending to a deductive research approach and the research purpose is explanatory
The research will start with a literature review which reviews the theories related to the topic researched to investigate often studied areas in customer relationship in marketing areas and to develop hypotheses for this research According to the recommendations of the research method, a survey strategy with questionnaire is chosen to find the answers to the research questions
A data-collection questionnaire comprising 28 questions was developed by adapting items from multiple previous studies to measure both independent and dependent variables A pilot test was conducted to ensure that both participants and researchers understand the meaning of the questions.
Data collected for this study were analyzed with SPSS First, normality of the distributions was assessed to determine whether the data meet the assumption of normality Subsequently, inferential statistics were computed using correlation analyses and multiple regression to test the stated hypotheses The results indicate which hypotheses are supported or rejected and illuminate the relationships among the four main areas Finally, the findings are compared with current studies to assess whether these relationships are consistent with or differ from existing research.
DEFINITIONS
This study examines the relationships among customer trust, customer commitment, customer loyalty, and customer retention as articulated in theories of customer relationship management, and tests these relationships within the e-commerce and online shopping sector To support the analysis, the research defines and operationalizes the four core concepts—customer trust, customer commitment, customer loyalty, and customer retention—for clarity and comparability By focusing on a single business domain, the study seeks to understand how trust influences commitment and loyalty, and how these factors together drive retention in online retail environments.
Customer relationship management (CRM) is a strategic approach aimed at increasing shareholder value by developing strong relationships with key customers and segments It integrates relationship marketing strategies with information technology to build profitable, long-term connections with customers and other key stakeholders By leveraging data and insights, CRM enables a deeper understanding of customers and the co-creation of value with them Realizing this vision requires cross-functional integration of processes, people, operations, and marketing capabilities, all enabled through information, technology, and applications.
Turban and King (2003) defined E-commerce (EC) as “the process of buying, selling, or exchanging products, services, and information via computer networks, including the Internet”
Online shopping refers to customers’ intentions to shop on the Internet and describes the way people buy products from Internet-based companies with e-business operations It represents a business-to-consumer (B2C) segment within the broader e-commerce landscape, as defined by Mosuwe et al (2004).
Morgan and Hunt (1994) defined Trust as “the perception of confidence in the exchange partner‟s reliability and integrity”
Morgan and Hunt (1994) define commitment as an enduring desire to maintain a valued relationship, representing the customer’s mental dedication to a company and the ongoing effort to sustain that bond In this framework, customer commitment underpins long-term loyalty by emphasizing mutual value, trust, and the sustained maintenance of the relationship between customer and firm.
Customer loyalty is a deeply rooted commitment to repurchasing or re-patronizing a preferred product or service consistently in the future, driving repeated purchases of the same brand or brand family This loyalty persists despite situational influences and marketing efforts that could prompt switching behavior, reinforcing long-term patronage with the chosen brand.
Gerpott (2001) defined Customer retention as “maintaining the business relationship established between a supplier and a customer”
LITERATURE REVIEW
CUSTOMER RELATIONSHIP MANAGEMENT
CRM is a strategic approach aimed at improving shareholder value through developing appropriate relationships with key customers and customer segments It unites the potential of relationship marketing strategies and information technology to create profitable, long-term relationships with customers and other key stakeholders CRM enables enhanced use of data and information to understand customers and cocreate value with them, requiring cross-functional integration of processes, people, operations, and marketing capabilities that is enabled through information, technology, and applications This definition is complete because it shows the constituent activities of CRM and how CRM is incorporated in companies.
CRM has evolved through three generations, as identified by Kumar and Reinartz (2006) The first-generation functional CRMs aimed to boost sales and improve service, capturing activities such as sales force automation and customer support The second generation introduced a customer-facing front-end approach, designed to fill gaps in ERP functionality and meet business needs through direct customer interaction During the 1990s, CRM sought to enable end-to-end customer communication—from pre-sale to post-sale—via channels such as telephone and the Internet, but achieving this full goal remained elusive in that era.
By the end of 2002, a strategic shift toward the third generation of CRM began as the company drew lessons from the unsuccessful implementation of its old CRM version It moved beyond the customer-facing front end of the second generation and also strengthened back-end systems, focusing on partners and suppliers, before integrating these with Internet technology As a result, CRM became not merely a technology solution but a core company strategy, playing an important role in driving revenue growth.
Uncertainties in online transactions have led researchers to identify trust as a critical factor in the successful development of e-commerce Trust underpins many social and economic interactions that involve uncertainty and dependence, shaping how individuals and businesses engage in digital markets In online commerce, trust mainly influences two key elements—security and privacy—making a precise definition of trust essential for understanding and improving online transactions.
Although e-commerce offers convenient connections between buyers and sellers, it also has limitations, including a lack of direct communication between buyers and sellers and between buyers and goods To overcome these barriers, suppliers must cultivate a trustworthy relationship with customers, which in turn boosts customer loyalty and encourages repeat purchases.
Teo and Liu (2007) regard consumer trust as a crucial aspect of e-commerce, and argue that understanding its antecedents and consequences is a primary concern: the antecedents reveal the relative importance of the factors that affect trust and inform the design of measures to foster it, while the consequences clarify trust's impact on online buying behavior.
In organizational trust literature, Mayer et al (1995) proposed a model describing the relationship between a trusting party and the party to be trusted In the e-commerce domain, Jarvenpaa et al (2000) examined whether customers’ perceptions of an Internet store’s reputation and size influence their trust in the store Overall, research shows that trust has meaningful consequences for consumers, shaping their attitudes, intentions, and behaviors.
Trust in marketing relationships is best understood through two foundational definitions Morgan and Hunt (1994) define trust as the perception of confidence in a partner's reliability and integrity, underscoring confidence and integrity as the core elements of trustworthy exchange In contrast, Mayer, Davis, and Schoorman (1995) define trust as a willingness to be vulnerable to another's actions based on the expectation that the other will perform a specified act important to the trustor, even when monitoring or controlling the other party is limited Taken together, these definitions show that confidence and reliability are the essential building blocks of trust in business relationships, with vulnerability and anticipated dependable actions shaping how trust operates in practice.
Although the original concept of trust provides a foundation, its definition does not capture all dimensions of trust because trust is a broad, multi-dimensional construct Therefore, trust research often requires classification based on the analysis and comparison of definitions, with different classifications emerging depending on the definitions used and the factors considered—attitudes, beliefs, behaviors, and tendencies—and on the referents of trust, whether trusting in a person, in a broader sense, or in a specific characteristic such as honesty.
To provide an overarching definition for this study, we adopt Morgan and Hunt's (1994) characterization of trust: “the perception of confidence in the exchange partner’s reliability and integrity.” This definition anchors our analysis of trust in inter-organizational relationships by emphasizing both dependable behavior and ethical conduct as the core components that build trust.
To answer the research question "What are the antecedents of trust?", this study synthesizes prior findings that identify two foundational antecedents: e-retailer reputation and trust, as explored by Bennett & Gabriel (2001) and Josang et al (2007); and privacy concern and trust, as examined by Eastlick (2006) The discussion that follows examines these antecedents and analyzes the relationships among them, outlining how retailer reputation and privacy concerns shape consumer trust in online shopping environments and how these factors interact to influence trust formation across e-commerce contexts.
Various factors influence the decision to participate in e-commerce, with retailer reputation emerging as a central consideration According to Bennett and Gabriel (2001), e-retailer reputation is essentially the same as brand reputation, encompassing the retailer’s name, term, symbol, sign, or design that distinguishes its goods and services from those of competitors offering similar items A retailer’s reputation comprises not just visual imagery but also an outsider’s subjective judgment of the organization’s past performance and overall quality, shaping consumer trust and online purchasing decisions.
Brand reputation reduces customers' perceived risk in online trading, as demonstrated by Van and Leunis (1999) E-retailer reputation is a key factor shaping customers' decisions to participate in e-commerce, influencing their willingness to engage in online shopping Moreover, Bennett and Gabriel (2000) and Josang et al (2007) show a positive relationship between reputation and trust, indicating that stronger reputational signals enhance customer trust These findings provide theoretical evidence for building and testing hypotheses about how reputation drives trust and participation in online markets.
Reputation is a sensitive, valuable asset for any company, organization, or even a category of individuals, and it’s far easier to lose than to earn It is strategic and fragile, easily tarnished if not carefully protected Therefore, a supplier with a strong reputation remains highly conscious of protecting it against negative effects, investing in trust and credibility to sustain long-term success.
Traditional marketing literature identifies reputation as a positive factor that boosts buyer and seller confidence In online shopping, Teo and Liu (2007) show that the perceived reputation of a vendor is significantly related to consumers’ trust in the vendor, providing theoretical evidence for the relationship between e-retailer reputation and customer trust, as described in Hypothesis H1a.
METHOD OF STATISCAL ANALYSIS
Data analysis in this study begins with a correlation analysis to examine the relationships among the study variables The results of this correlation analysis reveal how variables relate to one another, and the strength and direction of these relationships are quantified through the standardization of covariance, specifically using Pearson’s correlation coefficient (r).
The correlation coefficient has to lie between -1 and +1
A coefficient of +1 indicates a perfectly positive relationship; a coefficient of -1 indicates a perfectly negative relationship, while a coefficient of 0 indicates no linear relationship at all
The correlation coefficient is a commonly used measure of the size of an effect: values of ±0.1 present a small effect, ±0.3 a medium effect and ±0.5 a large effect
Multiple regression is a statistical technique used to analyze the relationship between a dependent variable and several independent variables In this study, all regression analyses model the dependent variable as a function of multiple independent variables, so multiple regression analysis was employed The major procedures for analyzing multiple regression in this study are presented in the following sections.
In social science, there are three major regression methods: standard (forced-entry) regression, sequential (hierarchical) regression, and statistical (stepwise) regression Standard regression enters all predictors into the model simultaneously, hierarchical regression enters predictors in a researcher-determined sequence, while stepwise regression decides the entry order based on a purely mathematical criterion The main goal of this study is to test mediating relationships among variables Predictors were entered one by one under the researcher’s control rather than all at once, so hierarchical regression was used to test the hypotheses.
Analysis of regression result in this study was based on major statistics such as sums of squares (R 2 , adjusted R 2 , and R 2 Change) and regression coefficients (Bi & òi)
R-squared (R^2) measures how much of the variation in the dependent variable is explained by the independent variables in a regression model, effectively showing the percentage of outcome variation the model accounts for In other words, it represents the portion of the variation that the model explains R^2 is calculated as 1 minus the residual sum of squares (SSE) divided by the total sum of squares (SST), or equivalently as the regression sum of squares (SSR) divided by SST The p-value of the F-ratio is used to assess the significance of the overall model and the observed R^2; R^2 is considered significant if this p-value is less than 0.05 at alpha = 0.05.
Adjusted R-squared measures the proportion of variance in the dependent variable explained by the independent variables, while penalizing the addition of extra predictors to reflect model complexity When the model is derived from the population rather than a sample, this statistic provides a more reliable assessment of fit by accounting for the number of predictors and the source of the data In practice, adjusted R-squared tends to be close to R-squared when the model is well-specified and the sample size is adequate, making it a robust indicator of the explanatory power of the predictors.
R-squared change, or R^2 Change, is used in regression analysis to assess the incremental explanatory power of a block of one or more independent variables It measures the difference in R^2 before and after adding those predictors, indicating how much additional variance in the dependent variable is explained The change in R^2 is considered statistically significant if the p-value for the R^2 Change (the F-test for the change in R^2) is less than 0.05 when alpha is set at 0.05.
Regression diagnostics evaluate whether the model fits the observed data well or is unduly affected by a small number of cases in the sample This assessment hinges on detecting outliers and influential observations that can distort parameter estimates and conclusions By identifying these influential points, analysts improve model validity and ensure more reliable inference from the data.
An outlier is a data point that differs substantially from the main trend, and such observations can bias the estimated regression coefficients, risking a biased model Detecting outliers helps minimize this bias by examining the residuals—the differences between observed values and those predicted by the model If a model fits the data well, residuals are small; if the fit is poor, residuals are large Standardized residuals are a popular tool for detecting these influential observations and identifying data points that warrant further investigation.
According to Field (2005), one general rule for residuals:
(1) “ standard residuals with an absolute value greater than 3.29 are cause for concern because in an average sample a value high like this is unlikely to happen by chance”;
(2) “ if more than 1% of a sample has standardized residuals with an absolute value greater than 2.58 there is evidence that the level of error within our model is unacceptable”; and
(3) “ if more than 5% of cases have standardized residuals with an absolute value greater than 1.96 then there is also evidence that the model is a poor representation of the actual data”
Residuals are used to detect outliers by examining the errors in the model, while influential cases are observations that exert undue influence on the estimated parameters In regression analysis and statistical modeling, the most commonly used statistics to identify influential observations include Cook's distance, leverage, Mahalanobis distance, DFBeta, and the covariance ratio (CVR).
Cook’s distance is a statistic that considers the effect of a single case on the model as a whole The values of Cook’s distance greater than 1 is may be cause for concern
Mahalanobis distance relates to leverage values and measures how far each observation lies from the mean vector of the predictor variables In multivariate analysis, it helps identify observations that sit far from the center of the predictor space With a sample size of N = 500 and five predictors, Mahalanobis distances above 25 are a cause for concern and may indicate influential cases or outliers that warrant further investigation.
In smaller samplers (N = 100) and with fewer predictors (namely three) values greater than 15 are problematic, and in very small sample (N0) with two predictors values greater than 11 should be examined
DFBETAs quantify how much each observation changes a parameter estimate when that observation is removed, with DFBeta defined as the difference between the parameter estimated using all cases and the parameter estimated when one case is excluded DFBETAs are calculated for every case and for each parameter in the model, enabling the identification of observations that exert a large influence on the regression coefficients By inspecting these standardized DFBETAs, you can spot influential observations that affect the model, and a case with the absolute value of the standardized DFBeta greater than 2 is typically considered to have substantial influence on a coefficient.
Regression analysis yields a model that fits the observed sample, but it may not generalize to the broader population unless key assumptions are satisfied, including independent residuals, normally distributed residuals, homoscedastic residuals, linear relationships between predictors and the outcome, and the absence of multicollinearity among predictors.
(1) Independent residuals (Durbin-Watson test)
One key assumption of regression is that the residuals for any two observations are uncorrelated, and this can be tested with the Durbin-Watson test, which detects serial correlation in the residuals from a regression model The Durbin-Watson statistic ranges from 0 to 4, with a value of 2 indicating no autocorrelation among residuals; values above 2 indicate negative correlation between adjacent residuals, while values below 2 indicate positive correlation The size of the Durbin-Watson statistic depends on the number of predictors in the model and the total number of observations.
As a very conservative rule of thumb, values less than 1 or greater than 3 are definitely cause for concern; however, values closer to 2 may still be problematic depending on sample and mode l
This is assumed that the residuals in the model are random, normally distributed variables with a mean of 0 This assumption indicates that the differences between the model and the observed data are most frequently zero or close to zero, and those differences much greater than zero happen only occasionally This assumption is tested by two graphical methods: histogram and normal probability The histogram plots of the residuals should be similar to the normal curve with the same mean and standard deviation as the data If the residuals are normally distributed, all points of normal probability plot should lie on the normal distribution line
METHODOLOGY
RESEARCH STRATEGY
Deductive and inductive reasoning are foundational approaches in research, discussed across many studies Understanding their methodological differences helps researchers decide whether a deductive or inductive approach should be used to answer the research questions The deductive approach derives hypotheses from existing theories and tests them through empirical data, while the inductive approach collects observations to generate new theories By contrasting these methods, researchers can select the approach that best aligns with their aims, data availability, and the desired balance between theory testing and theory development, ultimately improving the rigor and relevance of the study.
Deductive research is characterized by explaining causal relationships between variables, requiring quantitative measurement of all variables or concepts, and pursuing generalization through sufficiently large sample sizes By contrast, inductive research collects data first and develops theory from analysis of that data, focusing on how people interpret events and meanings As a result, inductive studies rely on qualitative data and are less concerned with broad generalization.
This study adopts a quantitative research approach, centering on numerical observations and aiming to generalize findings through formalized analysis of selected data in which statistical indicators play a central role The survey method, implemented as a structured questionnaire, directly tests whether the collected data can effectively answer the research questions.
This study applies a quantitative methodology using a cross-sectional survey to examine the phenomenon at a specific point in time A longitudinal or experimental design was not used because it would require substantial time and resources Since the population is large, a representative sample was drawn to test the theoretical model, and data were collected only once over a short period from diverse contexts within the population.
Survey methodology comprises four approaches: self-administered questionnaires, interviews, structured record reviews, and structured observations This study seeks to collect data on individuals’ attitudes at a single point in time under limited resources Among these methods, the self-administered questionnaire offers clear advantages for identifying and describing respondents’ attitudes and variability across phenomena Interviews demand substantial time and resources, particularly with larger samples, while structured record reviews and structured observations—focusing on visual or recorded data—are not well suited for measuring attitudes Given that self-administered questionnaires are a staple in social science research and meet the survey’s requirements, the study adopts a self-administered questionnaire completed by respondents as the most suitable approach.
DATA COLLECTION METHOD
This study examines how trust and commitment influence customer loyalty and retention in online shopping in Vietnam, drawing on respondents who have experience with online purchases In Vietnam, online shopping is commonly seen as a straightforward process—selecting goods via a website and paying either online or with cash on delivery—reflecting a distinctive feature of e-commerce development in the country, driven by consumers' limited cash resources and relatively low levels of information technology adoption.
Because the population of online shoppers in Vietnam is large and there is no fixed total, using a representative sample helps estimate the characteristics of the whole population Analyzing data from the selected sample allows the findings to be generalized to the entire population to answer the research questions In addition, sampling saves time and helps meet strict project deadlines, making it a practical approach for online consumer research in Vietnam.
The research uses probability samples which is most popular in the survey- based research strategy Because the research questions is concerned with online customer
This study uses multiple regression to test hypotheses and derive overall results Following Green's (1991) rule of 50+8k for sample size, where k is the number of predictors, the calculation with nine predictors (e-retailer reputation, privacy concerns, alternative attractiveness, switching cost, customer satisfaction, trust, commitment, customer loyalty, and customer retention) yields a minimum required sample of 50 + 9×8 = 122 With an anticipated response rate of 10%, the total number of questionnaires to distribute should be 1,220.
Sample size is important for population representativeness and reliability This study uses a sample of 1,000 online shoppers in Vietnam, mainly Internet users including university students (notably at the National Economics University in Hanoi) and working professionals who shop online and pay by credit card or cash, with an expected response rate of about 10–15% To ensure representativeness, after estimating the relationships among studied variables with multiple regression, the sample will be further evaluated using Cook's distance, Mahalanobis distance, and DFBeta to detect influential cases or bias If these diagnostics show no substantial bias, the results can be generalized to the broader population.
To fit time and resource constraints, this study uses a self-administered questionnaire with a sample size of 1,000 respondents who complete the survey themselves To achieve the expected response rate of 10–15%, the delivery and collection questionnaire method is employed: the questionnaire is handed to participants and collected after they finish, providing a practical and efficient data collection process.
The content of the questions focuses on customers‟ assessments expressed by
Respondents rate statements on a scale ranging from extremely disagree to extremely agree about their last shopping experience (questions 1–14, which assess the determinants of trust and commitment) and about online shopping in general (questions 15–28, which assess the determinants of customer loyalty and customer retention and examine the relationships among the four core constructs—trust, commitment, customer loyalty, and customer retention).
This questionnaire was designed to gather information from Vietnamese online shoppers to identify the factors that influence online purchasing decisions and to test the hypothesized relationships among nine variables—e-retailer reputation, privacy concerns, alternative attractiveness, switching costs, customer satisfaction, customer trust, customer commitment, customer loyalty, and customer retention Each variable is measured by several questions in the questionnaire, with items drawn from prior studies to ensure clarity and understandability for participants.
The study uses a 28-item questionnaire divided into two sections: Section 1 comprises 14 items probing the most recent online shopping experience to identify factors that influence two mediated variables—trust and commitment—while Section 2 comprises 14 items capturing customers’ perceptions of e-retailers across four dimensions—trust, commitment, loyalty, and retention Together, Sections 1 and 2 address two objectives: (1) to map relationships among the variables and identify the key drivers of customer loyalty and retention in online shopping within the Vietnamese market; and (2) to examine the impact of these drivers on loyalty and retention by testing the hypotheses outlined in Table 3.1 (Independent variables) and Table 3.2 (Dependent variables).
H1a E-retailer reputation positively affects customer trust
Questions 1-3 measure e-retailer reputation, questions 15-19 measure customer trust
H1b Privacy concerns negatively affect customer trust
Questions 4-6 measure privacy concern, questions 15-19 measure customer trust
H2a Alternative attractiveness negatively affects customer commitment
Questions 7-9 measure alternative attractiveness, questions 20-22 measure customer commitment
H2b Switching cost positively affects customer commitment
Questions 10-12 measure switching cost, Questions 20-22 measure customer commitment
H2c Customer satisfaction positively affects customer commitment
Questions 13,14 measure customer satisfaction, Questions 20-22 measure customer commitment
H3a Increasing customer trust leads to higher customer loyalty
Questions 15-19 measure customer trust, Questions 23-25 measure customer loyalty
H3b Increasing customer trust leads to higher customer retention
Questions 15-19 measure customer trust, Questions 26-28 measure customer retention
H4a Increasing customer commitment leads to higher customer loyalty
Questions 20-22 measure customer commitment, Questions 23-25 measure customer loyalty
H4b Increasing customer commitment leads to higher customer retention
Questions 20-22 measure customer commitment, Questions 26-28 measures customer retention
The questionnaire uses a Likert-scale of seven levels from Extremely Disagree (1) to Extremely Agree (7) to measure different variables in the model of the research and their relationships
Grounded in prior research, this study investigates the relationships among trust, commitment, loyalty, and retention within e-commerce, with a focus on online shopping It examines how trust and commitment interact to shape customer bonds, and how loyalty translates into higher retention in digital marketplaces By analyzing two or three focal areas—trust and commitment, as well as loyalty and retention—the research clarifies the drivers of customer engagement in both e-commerce generally and online shopping specifically The findings highlight that stronger trust and commitment foster greater loyalty, which in turn leads to improved retention across online retail platforms These insights inform practical strategies for building trust, cultivating commitment, sustaining loyalty, and boosting retention in online shopping environments.
Probability sampling encompasses techniques such as simple random sampling, systematic sampling, and stratified random sampling, with a random sample giving every member of the population an equal chance of selection In practice, accessing customers' online shopping information from companies is often difficult, which makes it hard to identify the full population or to select participants randomly Consequently, the study uses convenience sampling by choosing participants based on their availability Questionnaires were distributed in person to friends who are over eighteen and have online shopping experience in Vietnam, and they were asked to forward the survey to others Before handing over the questionnaire, an initial screening question—“Have you ever shopped online?”—was asked to confirm participants had online shopping experience.
To safeguard ethics and clarity in customer assessment surveys, the questionnaire was reviewed by experienced experts who confirmed there were no ethical issues and that the questions were clearly understood by participants Although the items were drawn from or adapted from prior research, a pilot test was conducted to ensure the instrument would function well with the study population Consequently, the pilot was evaluated by additional experts specializing in online shopping in Vietnam.
DATA ANALYSIS
Variables measured on a Likert scale are discrete, since responses occupy a fixed set of values Common Likert scales are 5-point or 7-point scales, with options such as 1 to 5 or 1 to 7 This discreteness means the actual values observed are limited to these scale points, which can be a disadvantage when the underlying construct warrants greater granularity and nuance beyond the defined options.
3, 4, 5, 6, or 7 and cannot be 3.15 or 5.18
As previously stated in the Literature Review chapter, and in relations to the questions in the questionnaire, the independent variables are determined as following table 3.1
(*): all questions displayed in the questionnaire are attached in the Appendix 3.3.3 Dependent Variables
The independent variables are listed and explained in the table 3.2
(*): all questions displayed in the questionnaires are attached in the Appendix
Data from 219 printed-survey responses were entered into SPSS as the primary data file for this study Normality was first assessed using descriptive statistics (means, minimums, maximums) The analysis then examined the relationships among variables to address the research questions about how trust and commitment influence customer loyalty and customer retention Results were derived using standard SPSS procedures to quantify associations and effects among the constructs.
Based on the mentioned hypotheses, a multivariate regression model is then built, as stated in following equation:
5 LOYALTY = ò 1 ER + ò 2 PC + ò 3 AA + ò 4 SC + ò 5 CS
6 LOYALTY = ò 1 ER + ò 2 PC + ò 3 AA + ò 4 SC + ò 5 CS + ò 6 TR + ò 7 CO
7 RETENTION = ò 1 ER + ò 2 PC + ò 3 AA + ò 4 SC + ò 5 CS
8 RETENTION = ò 1 ER + ò 2 PC + ò 3 AA + ò 4 SC + ò 5 CS + ò 6 TR + ò 7 CO
Where ER: E-retailer reputation; PC: Privacy concerns; AA: Alternative attractiveness; SC: Switching costs; CS: Customer satisfaction; TR: Trust; CO: Commitment; CL: Customer loyalty and CR: Customer retention.
FINDINGS AND CONCLUSION
DESCRIPTIVE STATISTICS
At the outset of the data analysis, we assessed the data distribution using descriptive statistics, with the scale scores summarized in Table 4.1 The minimum and maximum values ranged from 1 to 7, indicating no constraint on variability The means clustered around 4.0, with a maximum of 4.4581 and a minimum of 3.1689 The standard deviations hovered near 1.2, ranging from 0.94741 to 1.51073 Moreover, the absolute values of skewness and kurtosis were below the conventional thresholds (3 for skewness and 5 for kurtosis), indicating that the variables are normally distributed.
Table 4.1 Descriptive Statistics of Scales
Scales N Min Max Mean Std Skewness Kurtosis
CORRELATIONS
Table 4.2 presents the correlations matrix for the study variables, showing that all relationships lie within an acceptable range from zero to 0.8, which indicates meaningful associations among the variables Specifically, e-retailer reputation positively correlates with trust, meaning higher reputation increases customer trust in e-retailers This increased trust, in turn, drives greater customer loyalty, illustrating the reputation–trust–loyalty linkage revealed by the analysis.
ER PC AA SC CS TR CO CL CR
* Correlation is significant at the 0.05 level (2-tailed)
** Correlation is significant at the 0.01 level (2-tailed).
HYPOTHESIS TESTING
This study uses multivariate regression in SPSS to explore the relationships among trust, commitment, customer loyalty, customer retention, and their determinants The first section tests the relationship between trust and its determinants, while the next section analyzes the correlation between commitment and its variables The subsequent sections then present the direct and indirect effects of factors on customer retention and on customer loyalty, offering a comprehensive view of how these constructs interrelate and influence retention and loyalty outcomes.
This section examines the effect of e-retailer reputation and privacy concerns on trust Hypothesis 1a states that e-retailer reputation positively affects customer trust, while Hypothesis 1b states that privacy concerns negatively affect customer trust Table 4.3 presents the regression results that test these relationships, showing the estimated effects of e-retailer reputation and privacy concerns on customer trust and their statistical significance In short, the results describe the direction and strength of the relationships and reveal how reputation acts as a trust-building factor and privacy concerns as a potential barrier to trust in online retail.
Table 4.3 The Determinants of Trust
95% Confidence Interval for B Collinearity Statistics
Durbin-Watson 1.807 a Dependent Variable: Trust (TR)
Hypothesis H1a posits that e-retailer reputation positively affects customer trust, and this relationship is supported by a significant positive link between ER and TR Hypothesis H1b posits that privacy concern negatively affects customer trust, and this relationship is supported by a significant negative association between PC and TR Overall, ER and PC both show significant correlations with trust, underscoring the influential roles of retailer reputation and privacy concerns in shaping customer trust.
TR at p= 000 & and 000 with coefficient B = 156 & (-.152) respectively
After running a multiple regression analysis, the relationships among variables are identified, but regression diagnostics using Cook’s distance, Mahalanobis distances, and DFBetas are necessary to ensure influential cases do not bias the results The general rule for regression diagnostics is outlined, with more detail noting that a standard residual greater than 3.29 signals cause for concern; if more than 1% of cases have standard residuals above 2.58, the model’s error level is unacceptable; and if more than 5% of cases have standard residuals above 1.96, the model poorly represents the population.
Table 4.4 reports 11 cases (5%) beyond the ±2 threshold, two cases (0.91%) beyond ±2.5, and no cases beyond ±3, indicating that regression diagnostics are warranted Cook’s distance, Mahalanobis distances, and DFBetas were calculated For the cases examined, Cook’s distance values are well below 1, suggesting no undue influence on the regression, and all have Mahalanobis distances below the established threshold.
15 and the absolute values of all DFBetas are far below the threshold of 2 indicating that no case influences the regression parameters
Another important thing need to consider after running multiple-regression is to ensure the result from the sample can be generalized to the whole population
Durbin-Watson statistic for the regression is 1.807 (Table 4.3), a value within the 1–3 range that indicates the residuals are independent and the assumption of independent residuals is met in this sample.
Table 4.3 shows that all Variance Inflation Factor (VIF) values are well below the conventional cutoff of 10, and tolerance values exceed 0.2, indicating that multicollinearity does not distort the regression model and that the coefficient estimates are reliable for the regression analysis.
Regression results in Table 4.3 indicate that all significant coefficients have tight confidence intervals and do not cross zero, signifying that the estimated relationships are representative of the true population parameters As a result, the model demonstrates robustness and is generalizable to the entire population.
This section investigates the relationship between customer commitment and its determinants by testing three hypotheses: alternative attractiveness negatively affects customer commitment (Hypothesis 2a), switching costs positively affect customer commitment (Hypothesis 2b), and customer satisfaction positively affects customer commitment (Hypothesis 2c) Regression analysis is used to assess these relationships, and the results are reported in Table 4.5, detailing the direction and strength of the effects of these variables on customer commitment.
Table 4.5 The Determinants of Commitment
Upper Bound Tolerance VIF (Constant) 1.098 2.091 038 063 2.134
Durbin-Watson 1.461 a Dependent Variable: Commitment (CO)
The output in Table 4.5 shows that there is a significant relationship between
The analysis reveals that alternative attractiveness (AA) is negatively related to customer commitment (CO), with p = 020 and B = -0.180, supporting H2a In contrast, switching cost (SC) and customer satisfaction (CS) both show positive relationships with CO, with the SC–CO and CS–CO correlations reaching statistical significance (p = 000 and p = 000, respectively), leading to the acceptance of H2b and H2c Put differently, higher switching costs and higher customer satisfaction are associated with greater customer commitment, while higher alternative attractiveness is associated with lower customer commitment.
Table 4.6 shows that 10 cases (4.56%) fall within ±2, and 4 cases (1.8%) fall within ±2.5, with no cases beyond ±3, indicating that regression diagnostics are warranted The Cook's distance values for these cases are well below 1, suggesting that none of these observations has undue influence on the regression results Additionally, all cases have Mahalanobis distances below the established threshold, indicating a lack of influential multivariate outliers.
15 Finally, the absolute values of all DFBetas which are far below the threshold of 2 indicate that there are no case influences the regression parameters
The value of Durbin-Watson in the regression is 1.461 (see Table 4.5)
This value is in the range between 1 and 3, close to 2, so the assumption independent residuals are thus met or no cause for concern in this sample
All variance inflation factors (VIFs) are well below 10 and tolerance values exceed 0.2, indicating low multicollinearity among predictors The significant coefficients have tight confidence intervals that do not cross zero, implying that the estimated effects are robust and representative of the true population values Therefore, the regression model can be generalized to the entire population.
4.3.3.1 The direct determinants of Customer Loyalty
This section tests the influence of customer trust and commitment on the customer loyalty (Hypothesis 3a: Increasing trust leads to higher loyalty and Hypothesis 4a: Increasing commitment leads to higher loyalty) Table
3.7 shows the output of regression predicting the relationships between these variables Both hypotheses H3a and H4a are accepted because the relationships between trust (TR) and customer loyalty (CL) as well as between customer commitment (CO) and customer loyalty (CL) are significant (.002 and 000 < 05 respectively)
Table 4.7 The Direct Determinants of Customer Loyalty
Upper Bound Tolerance VIF (Constant) 1.196 3.797 000 575 1.817
Durbin-Watson 1.999 a Dependent Variable: Customer Loyalty (CL)
Table 4.8 shows 12 cases (5.47%) out of ± 2, 6 cases (2.7) out of ± 2.5 and more than ±3 Thus regression diagnostics is required in this case As Cook‟s distances values of these cases are significantly lower than1, none of these cases thus have an undue influence on the regression Moreover, all of these cases have the value of Mahalanobis distance lower than the threshold of 15 Finally, the absolute values of all DFBetas far below the threshold of 2 indicate that no case influences regression parameters