A research on determinants influencing the acceptance of electronic word of mouth of consumers in Ho Chi Minh city

acceptance of eWOM information in Ho Chi Minh City; define research concepts and measurement variables. Qualitative research Step 1 results in a research model and draft scales. Qualitative Research Step 2 aims to assess the factors influencing the acceptance of eWOM information were identified in Step 1; Adjust research concepts, edit / add measurement variables (in the field generate new variables). Qualitative research Step 2 was conducted by interviewing academic experts (major in Business Administration, Linguistics, Psychology); Focus group discussions with industry experts (website administrator, consumer representative, expert in online marketing field). 3.2.1.1 Objective: The expert interview helps to adjust the model in the optimal direction and adjust the scale to suit reality of Vietnam. The group discussion helps to redefine the interview questions that are easy to understand, simple and appropriate for the context needed research or not (the context of searching for pre-purchase evaluation information) 3.2.1.2 Subject and expert interviewing method: Subjects to interview academic experts: lecturers in linguistics, business administration, psychology of the University of Foreign Languages and Informatics of HCM City (HUFLIT), University of Economics and Finance (UEF), University of Technology (HUTECH), University of Education of HCM City. Method of organizing interviews: two weeks before the interview, the interviewees are suggested their abilities If they agree, the author will formally send an invitation with the basic contents of the interview, in which note the content of the interview is background theory assessment, model correction (for lecturers of Business Administration) and editing metrics (for Linguistics and Business Administration and Psychology faculty). The interview is implement each object individually. 3.2.1.3 Subject and method of group discussion: Group discussion subjects: industry experts such as website administrators, managers of companies with positions online business capabilities, experts in the field of online marketing, consumers have consulted commodity reviews before buying. Method of organizing interviews: a week before the interview, the subjects discussed are suggested abilities, if they agree the author will formally send an invitation with the main content of the discussion, in which note the issue main point of discussion is to edit the observed variables and the subjects do not need to prepare their responses in advance but only need to discuss and express thoughts on the issues raised during the discussion. 3.2.1.4 Collection and processing data: At the end of each interview, collected information will be recorded on A4 paper according to the exact text has been answered. The expert interview questions as well as the group discussion questions are semi-structured questions or open- ended questions that encourage and direct the respondent to answer the question in terms of thinking and language terms (term) of the individual interviewee (discussion). 3.2.1.5 Data analysis: Information collected from interviews and focus group discussions will be aggregated to serve as a basis for drawing most substantial and substantial conclusions about the issues raised in the interviews and seminars to evaluate and correct the shortcomings of the model and scale of this study. 3.2.2 Design of Quantitative research The quantitative study aims to quantify the impact of these factors on the information acceptability variable eWOM andintermediate effects. Shows the statistical values, the reliability of the scale and the suitability of the model's data research, confirming the reliability of qualitative research. Quantitative research is conducted in 2 steps: 11 Step 1. With data collected from about 50, the thesis will conduct the construction of the main scale by evaluating the reliability of the preliminary scale (Cronbach's Alpha). Likert scale with 5 levels is recommended to measure observed variables. The SPSS application is used to perform these assessments. The quantitative results in Step 1 are the formal scale of the factors that influencing the acceptance of eWOM information in Ho Chi Minh City. Step 2. On the basis of the formal scale built in Step 1, we use the data collected from the questionnaire online of about 500 questionnaires to test the formal scale by assessing the reliability of the main scale formula (Cronbach's Alpha), test convergence by exploratory factor analysis (using EFA method), test measurement model using Confirmation Factor Analysis (CFA). Next is to test the structure model by the method of structural equation modeling (SEM). All these tasks are performed by SPSS – AMOS application. The quantitative study in Step 2 resulted in quantifying the effects of the factors influencing the acceptance of eWOM information in Ho Chi Minh City. 3.2.2.1 Evaluation of scale reliability The purpose of assessing the reliability of the scale is to reject observed variables and the scale are not suitable. If the coefficient Corrected Item- Total Correlation correlation of an observed variable is less than 0.3, the observed variable will be eliminated, when Cronbach's Alpha coefficient is less than 0.6, that scale will be disqualified (Nunnally 1975). 3.2.2.2 Exploratory Factor Analysis (EFA) Exploratory factor analysis (EFA) method is used to test convergent values and discriminant values of the scale. According to Burns & Burns (2008), the indicators to consider when analyzing the EFA method include: - KMO index (Kaiser-Meyer-Olkin): used to check the appropriateness of the results of EFA analysis. If the index KMO between 0.5 and 1, the analysis results are consistent with the available data. The Bartlett test evaluates the correlation between observed variables. If this test is significant (Sig <0.05), the observed variables are correlated in overall. - Factor loadings: use to evaluate the correlation between observed variables and a factor, if the coefficient is greater than 0.5 (Hair, 2006). Factor load factor is used to evaluate the actual significance of the results of EFA analysis. Load factor must be at least 0.3; If greater than 0.4 is significant and greater than or equal to 0.5 is of practical significance. Difference of weight <0.3 is an acceptable value, but consider needing a content value before removing. - A scale is statistically significant if the total variance extracted > 50%. The extraction method is used in the thesis is "Principal Axis Factoring", together with the rotation "Promax". - Requires Eigenvalue coefficient greater than 1. Therefore, only factors with Eigenvalue coefficient greater than 1 can be retained. If this coefficient is less than 1, it will not have a good effect of explaining factor variability. 3.2.2.3 Validation of Measurement Model – Confirmatory Factor Analysis - CFA The testing of measurement model of a latent variable is done by Confirmatory Factor Analysis (CFA). The CFA method allows the evaluation and modification of a measurement model measure of an underlying variable, consisting of three ratings of Unidimensionality, Validity, and Reliability of the measurement model. 3.2.2.4 Validation of Structural Model – Structural Equation Modeling - SEM If the test measurement model meets the requirements (including unidirectionality, validity and reliability), step analysis is followed by evaluating the Structured Model using path analysis techniques in SEM. Analysis method of SEM uses many other statistical analysis techniques multivariate regression, factor analysis and correlation analysis allow to evaluate the model with complex relationships. SEM analysis estimates the factor load of latent variables in the model, and the causal relationships among potential variables. SEM structural model 12 evaluation method allows to consider causal structures (path analysis) and feces analyzing the influence of exogenous variables on endogenous variables and the influence of endogenous variables plays an intermediate role (mediator) up the result variable. 3.2.2.5 Validation of Overall Suitability –Bootstrap The Bootstrap method allows replaceable iterative sampling from an existing data sample (called original data) with the purpose of considering the suitability of the existing data with the population. The estimation process uses the number of samples vast N times, usually N = 1000. If the difference between the mean is obtained using Bootstrap and the smaller the mean of the model calculated with the original data sample, the more likely it is to model consistent with the larger overall data. 3.2.3 Method of data collection 3.2.3.1 Preliminary research sample The preliminary study sample is selected according to the convenient method using the Questionnaire and Interview the survey subjects directly. Sample collection: Samples were collected by hand-out and direct collection: number of questionnaires handed out was 80, number of questionnaires collected 56. After analyzing and evaluating, there are 6 unsatisfactory tables that must be removed because of missing information or only recording 1 level reviews for all questions. Therefore, through this method, 50 valid samples were obtained. 3.2.3.2 Formal research sample The formal research sample is selected by a convenient method using Google Forms for posting Web interview questionnaires, and thanks to relationships to go viral on Facebook pages. Sample collection and sample size determination: Number of samples collected: The number of questionnaires collected was 565. After analysis, there were 43 questionnaires that failed and were removed for filling out information are not consistent. Therefore, through this method, 522 valid questionnaires are obtained to be used in the thesis. Sample size determination: The number of questionnaires collected is guaranteed to be larger than n = 15m, where m = 28 is the number of qualified observational variables performing factor analysis methods (Comrey, 1973). In the thesis, the number of collected questionnaires is 522>15 * 28 = 420 ensures enough conditions to perform analysis according to factor analysis method. 3.2.3.3 Questionnaire Questionnaire structure - The questionnaire is divided into 3 parts: Part 1 is for exclusion questions, if the respondent does not use the Internet to find information before making a purchasethe following sections will not be continued. Part 2 is survey questions about the influence of information, the social influence, and the effects of skepticism. Total, there are 28 sentences, each sentence corresponding to an observed variable. Part 3 is part of personal information such as gender, age, occupation ... 3.3 QUALITATIVE RESEARCH RESULTS 3.3.1 Results of expert interviews The results of expert interviews include 3 parts, Assessing the conformity of the background theories used in research of the thesis, explore the scale of research concepts and evaluate the research model. Outline details of expert interviews and interview results are presented in Appendix 4 - Outline and results of expert interviews. 13 3.3.1.3 Research Model H10: eWOM information quality has a positive impact on perception of the reliability of information eWOM (H10+). H11: Sources reliability have a positive impact on perceiving the reliability of information eWOM (H11 +). Combined with the results discussed above the Scale, the research model of the thesis after in-depth discussion with the experts is presented in Figure 3.2 3.3.2 Results of group discussion The thesis uses the results of interviewing experts on the scale to summarize it into a preliminary and progressive scale discussion in order to further edit the content related to the preliminary scale for measurement variables to be achieved ease of understanding, adjusting obscure words and adjusting and supplementing content to suit the survey needs reality. 3.3.3 Research Model and Formal Scale Research hypotheses, revised research models, preliminary scales are synthesized based on alternative recommendations expert discussion and group discussion sessions were then tested for quality with sample numbers N = 50, the test results show that the preliminary scale is qualified and is used as the formal scale. 3.3.3.1 Research hypothesis – Modified research model Source: Author compiled according to expert opinion Figure 3.2 – Research model after interviewing experts 3.3.3.2 Formal Scale Through the quantitative analysis in Appendix 6 – Details of scale analysis with N=50 shows that preliminary measurement can be used as the formal scale to test the model. 14 CHAPTER 4. RESEARCH RESULTS AND DISCUSSION 4.1 QUANTITATIVE RESEARCH RESULTS 4.1.1 Description of research sample Data was collected using the Google Form software to post questionnaires on the Web, and thanks to relationships to go viral on Facebook pages. Number of questionnaires collected is 565. After analysis and check, there are 43 questionnaires rejected due to the inconsistent evaluation information. Hence through this collection method 522 valid questionnaires are used in the thesis. 4.1.2 Validation of the scale 4.1.2.1 Cronbach’s Alpha reliability test In essence, the higher the Cronbach's Alpha coefficient, the higher the correlation of the observed variables with the observed variables the higher the other in the same group. A scale with Cronbach's Alpha value > 0.6 is often chosen, because the scale is valid Cronbach's Alpha number > 0.6 is eligible to continue to perform the exploratory method of EFA. In addition, observed variables with variable-total correlation coefficient < 0.3 do not have much correlation with other variables and should be removed from the scale. 4.1.2.2 Test the scale by Exploratory Factor Analysis – EFA method KMO index (Kaiser-Meyer-Olkin): used to check the appropriateness of the results of EFA analysis. KMO Index =.859 is between 0.5 and 1, so the analysis result is consistent with the available data. The Bartlett test evaluates the degree correlation between observed variables. Because this test is significant (Sig. = .000 <0.05), the observed variables are correlated related to each other in the whole. Doing EFA analysis for all factors showed that: The variables observed in each factor were grouped into 1 group. There is no variable belonging to two factors, ensuring distinct validity. The factor load factors are all larger 0.5, ensure the convergence value. 4.1.3 Test results of the measurement model by Pooled CFA method Measurement model test was performed by Pooled CFA confirmation factor analysis (Control of all potential variables in the model). General assessment of CFA confirmatory analysis: The unidirection is satisfactory because the load coefficients of the observed variables for each concept are greater than 0.6. The validation indicators are satisfactory, specifically: The measurement variables are statistically significant and AVE values> 0.5, so the scale achieves convergent value. The values related to the suitability such as RMSEA = 0.014 0.9, GFI = 0.956 and CMIN / df =1,106 are all satisfactory. The square root of AVE> the correlations between the two concepts is therefore satisfactory for the degree of distinction. The test indicators of reliability are satisfactory. Through the above results of factor analysis, CFA proves that the scale of research concepts is uniform model is consistent with existing data. The research model does not change, and is a formal model of the thesis. 4.1.4 Test results of structural model by Structural Equation Modeling – SEM method CFA analysis allows to test measurement models, but fails to evaluate the causal effects of these concepts. Therefore, the next step is to run the SEM model. 4.1.4.4 Hypothesis testing results SEM analysis results show that the proposed hypotheses are statistically significant (with p-value <0.05), therefore these hypotheses is accepted (Table 4.23). (Estimates are normalized values). 15 Table 4.23 – Results of hypothesis testing Hypo- thesis Influencing Factors Affected Factors Estimate p- value Results H1 Information Quality Perception of the usefulness .226 .000 Accept H2 Source Reliability Perception of the usefulness .155 .006 Accept H3 Perception of the usefulness eWOM information acceptance .380 .000 Accept H4 Perception of the ease of use Perception of the usefulness .403 .000 Accept H5 Perception of the usefulness eWOM information acceptance .494 .000 Accept H6a Information Rating Perception of the ease of use .238 .000 Accept H6b Information Rating Perception of eWOM reliability .201 .000 Accept H7 Perception of eWOM reliability eWOM information acceptance .290 .000 Accept H8 Perception of Identity Perception of eWOM reliability -.109 .000 Accept H9 Perception of Motivation Perception of eWOM reliability -.135 .010 Accept H10 Information Quality Perception of eWOM reliability .224 .000 Accept H11 Source Reliability Perception of eWOM reliability .136 .023 Accept Source: SPSS-AMOS analysis results 4.1.4.5 Bootstrap result Bootstrap testing allows to evaluate the suitability of the data with the population. In this study the repeatability level of the existing data set is N = 1000. The estimated results from 1000 random samples generated by the program are obtained the mean and standard deviation. The results show that the CR values are all less than 1.96, so we can conclude that the estimates in the model can be indicates that the sample can be representative of the population meaning that the study data is consistent with the marketing data. 4.1.5 Evaluation of the competitive model by SEM method The thesis proposes to test a competitive model to evaluate the suitability with existing data of the research model. If the competitive model has a higher relevance, then the model chosen will be a competitive model, otherwise, the current research model is considered to be the most appropriate for the available data. The competition model is built by adding a causal link between the information rating variable and the usefulness of information variable (Figure 4.4). Hence, a new hypothesis is proposed: Hypothesis H12: Ranking information eWOM has a positive impact on Perceived usefulness of informationeWOM. 4.1.5.4 Hypothesis testing results of the competitive model The testing of competitive model by SPSS-AMOS program shows that hypothesis H12 is proposed unsatisfactory. Means the relationship with the information rating and the usefulness of the information eWOM does not exist. Therefore, the competition model does not explain better than the current research model. Specifically, Table 4.26 gives found that the estimation of CHAPNHANEWOM dependent variable explanation of the competition model (.694) is same with explanation of current research model (.693). Current research model with this comparison is considered model best suited to the data collected. 4.2 DISCUSSION OF RESEARCH RESULTS 4.2.1 Discussion of the explanatory level of the research model of the thesis compared with the original model (Information Adoption Model - IAM) Compared with the original models of Sussman & Siegal (2003); Cheung, Lee & Rabjohn (2008); Cheung (2014), level the explanation of the research model of the thesis for CHAPNHANEWOM dependent variable increases 16 significantly. One partly because the research model has added a number of independent variables. Specifically those are the XEPHANG variables, DANHTINH and CN_DONGCO. When running SEM analysis, the original model of Sussman & Siegal (2003) with present data collected, the explanation level of the original model is 42.5% - Details of the results of running the SEM analysis of the original model is shown in Appendix 11 - Original Model – IAM with running result details in SPPS - AMOS. The reason why the level of explanation of the research model in the current context (69.3%) is higher than before is due to consumers have changed the way they perceive eWOM information. How to find and absorb information of Consumers' eWOMs have changed, they no longer fully trust eWOM information as they once did, and there is too much information eWOM to handle, many times needing to rely on public (social) comments to support the process make their own buying decision. Hence a newer model that reflects these changes is required, here is viewed the new point of the thesis. 4.2.2 Discussing the model building’s approach of the thesis compared with the model building's approach of other related studies Experimental research model of Shen, Cheung & Lee (2013) has a research related approach of the thesis by using the expanded approach of Information Adoption Model - IAM of Sussman & Siegal (2003) (Figure 2.14). However, the research subjects and research objectives of Shen, Cheung & Lee (2013) are different with the dissertation's research subjects and objectives. Since then, research models and scales are also built differently. Specifically, the research objective of Shen, Cheung & Lee (2013) is to evaluate the effects of trustworthiness and the usefulness of Wikipedia (the open dictionary page) to accepting information provided by Wikipedia. Another research that has a related approach to the thesis research is that of Shen, Zhang &Zhao (2016) (Figure 2.19). The research of Shen, Zhang & Zhao (2016) expands the Information Adoption Model - IAM by adding two intermediate variables, Discounting own Information and Immitating Others. 4.2.3 Discussion of direct effects on the eWOM information acceptance All 3 intermediate variables CN_HUUDUNG, CN_DESUDUNG and CN_DOTINCAY have a direct impact on CHAPNHANEWOM variable at different levels. The intermediate variable CN_DESUDUNG has the highest impact level on the acceptance of eWOM information, which shows that consumers often do not really learn the content of the eWOM, they rely solely on community reviews and use that rating information as a basis just accept eWOM information. To the knowledge of the author, the impact of perception of the usefulness of eWOM information to the acceptance of eWOM information has not had any studies related to eWOM, and therefore can be considered as a new point of the thesis. Especially when consumers are overwhelmed with information, choosing to go with the crowd is potentially decisive to buy is sometimes better than making their own decisions because too much information eWOM will make it difficult for them to choose correct advice. This reduces the time spent searching and researching - the consumer just needs to go after what is already highly unanimous community. However, the crowd is not always right and thus limits the effectiveness of the purchase process. In addition, the intermediate variable CN_DOTINCAY also has a quite large impact, showing another cause of the acceptance of eWOM information is the consumer's belief in the objective truthfulness of eWOM information. The impact of perception of eWOM information reliability on the acceptance of eWOM information is consistent with research results of Cheung et al (2009); Fan et al (2013). However, the difference from the thesis is research by Cheung et al. (2009) evaluated the effect of perception of eWOM information reliability on acceptance knowledge, do not the acceptance of eWOM information. Fan et al. (2013) also differ in ways approach compared to the research of the thesis because only researching information effects (Information quality, source reliability, quantity of information, ...) to perceive the reliability of information; while the thesis's approach is evaluate the impact of informational factors 17 (Information quality, Sources reliability), normative factors (information rating and the factor of skepticism (perception of motivation) to perception of eWOM information reliability. This is considered a new point of the thesis. In addition, the intermediate variable CN_HUUDUNG also has a relatively high impact weight, showing the usefulness of eWOM information plays an important role in consumer’s acceptance of eWOM information. Impact of perception of the usefulness of eWOM information to the acceptance of eWOM information consistent with the research of Sussman & Siegal (2003); Cheung, Lee & Rabjohn (2008); Shen, Cheung & Lee (2013); Tseng & Wang (2016). However of course, research by Sussman & Siegal (20