Tóm tắt Luận án A measurement model of city - Based consumer patriotism: A case of research in Vietnam

ration, conflict, prejudice or discrimination in a process called social comparison (Tajfel, 1982). 3.1.2 Consumer ethnocentrism (CET) 7 Shimp and Sharma (1987) defined consumer ethnocentrism as a unique economic form of ethnocentrism that captures the beliefs of consumers about the appropriateness and morality of purchasing foreign products. Consumers having an ethnocentric attitude consider that the products produced in their own country are good and consider the purchasing of foreign products as wrong because they can adversely affect workers’ jobs, restrain the development of domestic businesses, and endanger the economy. 3.1.3 National identity and patriotism (PAT) In marketing literature, national identity has been discussed using the concept of patriotism (Kim t al., 2013, pp.77) because patriotism is one of the main sources of national identity (Petya, Marco, 2014). Patriotism is often perceived as a love of one’s country and devotion to it, but it can take many forms (Schatz, Staub, and Lavine, 1999). Vida and Reardon (2008) noted that the concept of patriotism was attached to culture, so it should be operationalised in market scenarios. Because of the broad perspectives on the concept of national identity and patriotism, up to now, in addition to a number of works measured patriotism in a particular way, four main points can be listed in the measurement. They are point of Adorno et al. (1950), Kosterman and Feshbach (1989), Keillor et al. (1996) and Karasawa (2002). The author noticed Keillor et al. (1996) based on SIT when developing the measure of patriotism, but the scale was designed quite complicated with four diemensions that one of which bias consumer ethnocentrism. So the author decided to reference simply this scale to measure Vietnamese patriotism. Therefore, the author continued to review marketing science studies that referenced the scale of Keillor et al. (1996) to find the most reasonable reference. Through the review, it could be seen that most these studies had the statements expressing main ideas such as: Love for the country; Pride in nationality; The meaning of being a citizen of that country; The connection with the country. These statements exhibit similar psychological characteristics of patriots in many different countries, including Kazakhstan (Rybinal et al., 2010). The author assumed that Kazakhstan and Vietnam have a number of geographical - historical and political similarities that can make people's psychology similar. So the application the Kazakh patriotism scale for Vietnamese case 8 could be a reasonable choice. As a result, the author decided to use the measurement scale of the construct of patriotism in the work of Rybina et al. (2010) for Vietnamese people. 3.1.4 Consumer cosmopolitanism and open-mindedness 3.1.4.1 Cosmopolitanism (COS) In the 20th century, sociology literature defined cosmopolitans as people who oriented themselves outside their community rather than being influenced solely by local traditions and values (Riefler và ctg., 2012). Cosmopolitanism is like a form of openness to different cultures, along with a desire to work to seek experiences from other cultures rather than from one's own culture (Hannerz, 1990; Roudometof, 2005; Thompson and Tambyah, 1999). In the preliminary study, the author used the CYMYC scale (Cannon et al., 1994) to measure consumer cosmopolitanism. 3.1.4.2 Open-mindedness (OPEN) After conducting preliminary study, in order to increase the monological validity of the construct of consumer cosmopolitanism, the author reviewed the literature to find out a newer measurement scale called C-COSMO (Riefler và ctg., 2012). This scale defines consumer cosmopolitanism as a three-dimensional, second-order construct capturing the extent to which a consumer (1) exhibits an open-mindedness towards foreign countries and cultures, (2) appreciates the diversity brought about by the availability of products from different national and cultural origins, and (3) is positively disposed towards consuming products from foreign countries (Riefler et al., 2012). According to Riefler et al. (2012), the first dimension named "open-mindedness" is an important characteristic of consumer cosmopolitanism. Riefler et al. (2012) established a scale for the construct of open-mindedness with statements revolving the consumer desire for cultural exchange and travelling to interact with people and other cultures. The author decided to choose open- mindedness to represent the consumer cosmopolitanism in the formal stydy (second quantitative study). 3.1.5 Consumer patriotism (COPAT) and its position in SIT 3.1.5.1 The definition of consumer patriotism Clift, Woll (2012a, pp.308-33) defined “economic patriotism as economic choices which seek to discriminate in favour of particular 9 social groups, firms or sectors understood by the decision-makers as insiders because of their territorial status. Economic patriotism entails a form of economic partiality: a desire to shape market outcomes to privilege the position of certain actors”. The author used the main idea of the definition of consumer patriotism from the empirical studies of Han (1988); MacGregor, Wilkinson (2012); Tsai, (2010) and improved the definition of the nature of social identity in consumer patriotism proposed by Clift, Woll (2012a) to get the definition of consumer patriotism: Consumer patriotism is the biased consumer choices made to support the domestic economy by buying domestic goods to help domestic businesses and workers, and the emotional reasons for that biased choice. 3.1.5.2 The relationship of consumer patriotism with other constructs of SIT In social identity theory, the origin of both consumer ethnocentrism and patriotism arises from the attachment of individuals to their in-group, which in this case is their mother country, but consumer cosmopolitanism focuses primarily on a person's relationship with outside groups (ie, foreign countries) (Z- Roth et al., 2015). In terms of direction, patriotism is concerned with bias that favors the in-group (Bizumic et al., 2009; Brewer, 1999), but consumer ethnocentrism and cosmopolitanism are a bias against out-groups (Z-Roth et al., 2015). A similar classification is also used for the construct of consumer patriotism. Consumer patriotism is the economically positive bias that is based on an individual's attachment to his/her in-group. Consumer patriotism is only directed toward the relationship with the in-group and not directled toward the out-group. From these ideas, the authors identified the potential relationships between the these constructs like that: (1) The relationship between patriotism and consumer patriotism: Z- Roth et al. (2015) proposed that the national identity of a person brings bias for his/her in-group. In the current study, consumer patriotism is an economic in-group bias so it is considered to be the result of national identity. The author can summarize the relationship as follows: people who are more patriotic will be more biased towards domestic economic support. In other words, the consumer patriotism will be positively influenced by their patriotism. 10 (2) Relationship between consumer patriotism and consumer ethnocentrism: Z-Roth et al. (2015) stated that consumer ethnocentrism is an out-group bias against foreign countries. However, consumer patriotism is the individual's positive bias for his/her in-group. Therefore, there is potential for these two factors to be positively correlated. (3) Relationship between consumer patriotism and consumer cosmopolitanism: Roudometof (2005, pp.122) argueed that the open-mindedness and variety seeking of consumer cosmopolitans make it difficult for them to take a "viewpoint of unconditional support for one's country". On the contrary, consumer patriotism is the positive in-group bias. Therefore, the author hypothesized that consumer cosmopolitanism would reduce the consumer patriotism. 3.2 Theory of evaluating the measurement model 3.2.1 Requirements for the measurement scale 3.2.1.1 Content validity Assessing the appropriateness of the measurement variables designed for a measurement scale of construct with the definition of that construct. 3.2.1.2 Unidimensionaltity The overall fit of the model provides the necessary and sufficient conditions to determine whether a set of observed variables is unidimensional (Steenkamp and Van Trijp, 1991). However, this is true only if there is no correlation between the residuals of the observed variables. 3.2.1.3 Reliability The researchers used the following criterion: item-total correlation> 0.5 (Hair et al., 2010); Cronbach's alpha coefficient> 0.7 (Hair et al., 2010); AVE> 0.5 (Bagozzi and Yi, 1988); CR> 0.6 (Bagozzi and Yi, 1988). 3.2.1.4. Validity Convergent validity: Hair et al. (2010) proposed that the condition for convergent validity of scales is FLs ≥ 0.5 or ideally, 0.7. Discriminant validity: It involves confirming whether the population correlation among constructs is truly different from a value of 1. The Fornell–Larcker criterion (1981) requires that the AVE of a construct should be higher than the highest squared 11 correlation between this construct and other constructs to ensure this construct has discriminant validity. Content validity: Steenkamp and Van Trijp (1991, p.294) say the content validity is assessed by statistical testing the relationship between construct and other constructs in the theoretical system. 3.2.2 Exploratory factor analysis technique for preliminary assessment of measurement scale Assessing the discriminant validity: The number of factors extracted by EFA must be consistent with the original hypothesis of the number of factors, combined with the condition that the observed variables download exactly the latent concept that it is designed for measuring (Nguyen Dinh Tho, 2011). Assessing the measurement model by EFA requires that the cumulative percentage of variance is higher than the threshold of 50% (Gerbing and Anderson, 1988). Hair et al. (2010, pp.117) noted that Factor loadings FL greater than 0.5 to meet the requirement of convergent validity for the construct scales. 3.2.3 Covariance based structural equation modeling 3.2.3.1 Basic steps of SEM Step 1. Specify the model. Step 2. Evaluate model identifcation Step 3. Select the measures and collect, prepare, and screen the data. Step 4. Estimate the model: a. Evaluate model ft (if poor, skip to step 5). b. Interpret parameter estimates. c. Consider equivalent or near-equivalent models (skip to step 6). Step 5. Respecify the model (return to step 4). Step 6. Report the results. 3.2.3.2 Path model (PA) A path model (PA) is a structural model for observed variables, and a structural model represents hypotheses about effect priority. In the PA model, the variables analyzed are observed variables but not the latent variables are measured through a set of observed variables. Calculate the PA degrees of freedom: dfM = Number of observations - Number of free model parameters 12 Types of PA: There are two basic types of path models: recursive and non-recursive. There are two basic requirements for the identification of any kind of structural equation model, not only PA: (1) dfM ≥ 0, and (2) every unobserved (latent) variable must be assigned a scale (metric). The sample size of PA: the relationship between sample size and model parameters is 20:1. Bollen (1989) said that this relationship must be at least a 5:1 ratio. Estimation procedure: the most common procedure is the MLE. The goodness-of-fit of PA: χ2M =(n-1)FML to test the hypothesis H0: The model is correct; χ 2 M /dfM < 2 or 3 or even 5 (Bollen, 1989); 0,06 0,9 show the goodness-of-fit of model (Hair et al., 2010). 3.2.3.3 The measurement model of latent variables and confirmatory factor analysis model (CFA) Identification of CFA models: Necessary Requirements: (1) the number of free parameters is less than or equal to the number of observations, and (2) every latent variable, which includes the measurement errors and factors, must have a scale. Sufficient Requirements: If a standard CFA model with a single factor has at least three indicators, the model is identified. If a standard model with two or more factors has at least two indicators per factor, the model is identified. Checking multivariate normal distribution of data: Researchers can test multivariate normal distribution of data with two tests: Royston's Multivariate Normality Test and Henze-Zirkler's Multivariate Normality Test. 3.2.4.4 SR model combines measurement and structural models The SR structure regression model is a combination of CFA measurement model and PA model, they are the most general form. 3.2.4 The procedure of mediator effect test 3.2.4.1 The definition of mediator variable. The mediator variable is defined as the variable that explains the relationship between a cause variable and a result variable (Baron and Kenny, 1986). 3.2.4.2 The four-step procedure of mediator effect test of Baron and Kenny (1986) 13 Step 1 - determine the relationship between the cause and result variables. Step 2 - determine the relationship between the cause variable and the mediator variable Step 3 - determine the relationship between the mediator variable and the result variable Step 4 - Test the role of mediator variables through the following 3 equations: First, the result is regressed on the cause variable Second, the mediator is regressed on the cause variable Third, the result variable is regressed on both the cause and the mediator variables After that, based on the values of regression coefficients and standard errors of the estimated regression coefficients, a Z score value is calculated to decide whether to accept or reject the null hypothesis H0 :The considering variable has no mediate role. 3.2.4.3 The mediator effect test with BC boostrap Based on the limit of the ci% confidence interval of the MLE estimate in Sem, the researcher can conclude the statistical significance of the mediator variable. CHAPTER 4 RESEARCH METHODOLOGY 4.1 Qualitative research to identify research hypotheses and models 4.1.1Topic of qualitative research To determine the following 04 causal relationships: Patriotism/consumer patriotism ; consumer patriotism/ Consumer ethnocentrism; consumer cosmopolitanism/ consumer patriotism; consumer cosmopolitanism/ Consumer ethnocentrism. 4.1.2 Qualitative research procedure The author had referred to the Grounded theory method of Corbin and Strauss (1990) in combination with the guidance of Charmaz (2006) to implement the following steps in the qualitative research procedure of the thesis: 4.1.2.1 Choosing the expert The first expert work as a professional researcher at a social science research organization cum a sociology lecturer. The second expert works as an economics lecturer cum an consultant at the Association of Vietnam companies produce high quality product. 14 4.1.2.2 Compiling the qualitative research results In the working process with experts, the rules of causal relationhips of constructs were not specified by the author. The author only presented the definition of the constructs and research contexts. Then, the experts would decide the relationship between them. After two interview phases, the research results show the following relationship rules: Patriotism/consumer patriotism: positive relationship; consumer patriotism/ Consumer ethnocentrism: positive relationship; consumer cosmopolitanism/ consumer patriotism: no relationship; consumer cosmopolitanism/ Consumer ethnocentrism: negative relationship. 4.1.3 Research hypotheses and research model 4.1.3.1 Research hypotheses H1: Patriotism increases the consumer ethnocentrism attitude. H2: Patriotism increases the consumer patriotim attitude. H3: Consumer patriotism increases the consumer ethnocentrism attitude. H4: Consumer cosmopolitanism increases the consumer ethnocentrism attitude. Although in the qualitative research both experts concluded that there was no causal relationship between consumer cosmopolitanism and consumer patriotism of consumers. The author still proposed the hypothesis H5 based on theoretical arguement has established about this relationship. The author hope that the results of empirical research about this relationship on Vietnamese urban consumer context could provide more information, enriching the theoretical basis of SIT about the role of consumer cosmopolitanism. H5: Consumer cosmopolitanism decreases the consumer patriotism attitude. 4.1.3.2 Research model Based on 05 research hypotheses above, the conceptual framework was formulated. 15 Figure 1 The conceptual framework 4.2 Quatitative research procedure 4.2 .1 Building/completing the scale of construct 4.2.1.1 Building the scale of construct in the first study The author wrote down the statements of the construct of consumer patriotism based on the definition of this construct following the combined approach of Hinkin (1995). The statements of the construct of consumer ethnocentrism in previous works in Vietnam were listed by the author. The author referenced the scale of consumer cosmopolitanism construct from Yoon et al. (1996). The statements of the construct of patriotism were adopted from the Rybina et al. (2010). The author then worked with two expert to assess the content validity of construct measurement scale, and to edit the statements. As a result, the author had formed 21 statements. They were the basis for the author to develop the official questionnaire for the first study. 4.2.1.2 Completing the scale of construct in the second study The author summarized the results of the first study and used this result to discuss in a focus group with 12 researchers at the Southern Institute of Social Sciences of Vietnam. The construct of consumer cosmopolitanism in the second study was measured through the open-minded construct of the C-COSMO scale of Riefler et al., (2012). The participants of focus group agreeded with the content validity of the meaurenment scale of constructs in the study. The results of the discussion formed the outline of the statements. From that outline, the author developed the draft questionnaire of the second study. After using this draft questionnaire to interview few customers, the author adjusted it into a formal questionnaire used for second study. 16 4.3 Data collection 4.3.1 Data collection method The author used face-to-face survey with consumers. 4.2.2 Sampling 4.2.2.1 Sampling in the first study The author applied the purposive sampling technique. Subjects were students of evening classes of some universities in Ho Chi Minh City which were econimical major. As a result, the author had 230 fullfil questionnaires. 4.2.2.2 Sampling in the second study The author applied the representative sampling technique. There were 300 households in Hanoi and Ho Chi Minh City, respectively, were selected for the sample. The author used the sampling frame of Vietnam Population Change and Family Planning Survey 2017 of General Statistics Office of Vietnam to take the list of 600 households. Based on this address list, the interviewer approached and interviewed a suitable consumer in each household. CHAPTER 5 RESULTS 5.1 Results of the first study 5.1.1 Sample profile Subjects were university students of evening classes so they were quite similar in age and education level. The sample comprised 32% men. 5.1.2 Evaluating the measurement scale 5.1.2.1 Factor analysis procedure After the EFA and Cronbach Alpha procedures, the author concluded that the measurement scales of the four latent constructs had achieved reliability and had not violated unidimensionality require. The convergent validity of CET scales was a bit weak, but the rest three scales were satisfied. After that, the author conducted the CFA procedure following 06 steps of Kline (2011) and used the results of CFA to evaluate the measurement models of latent constructs for unidimensionality, reliability, convergent validity, discriminant validity. All these requirement were satisfied. The monological validity of the measurement models of a constructs was assessed by testing the research hypotheses using mediator effect test. 17 5.1.2.2 Mediator effect test (1) Baron and Kenny’s (1986) mediator effect test The results of testing the mediate role of COPAT in the relationship PAT/CET showed a full mediate role of COPAT. The author concluded that the hypotheses H2 and H3 were supported. The hypothesis H1 should be rejected if based on p-value > 0.05, but actually this situation was the consequence of multicollinearity in regression model. The results of testing the mediate role of COPAT in the relationship COS/CET showed that COPAT was not mediator variable. The hypotheses H4 and H3 (again) were supported. But the hypothesis H5 was rejected. (2) Mediator effect test of BC boostrap The author continued to analyze the SR model based on the results of the CFA model. The author eliminated the unsignificant causal relationship of COS/COPAT, so the research hypothese H5 was rejected. Then the author re-specified the SR model and re- estimated. The goodness of fit of re-specification SR model was also satisfied. Based on the significance test of the path coefficient of the re-specification SR model, the author concluded that the hypotheses H1, H2, H3 and H4 were supported. The author used the re-specification SR model to conduct the BC boostrap test on the mediate role of COPAT. Based on the 95% confidence intervel, the author concluded that COPAT was not mediator variable in the relationship between COS/CET but was full mediator variable in the relationship between PAT and CET. 5.1.2.3 Conclusion of the measurement models of latent constructs after the first study From the results of two times of mediator effect test, the author generally concluded of the research hypotheses as follows: H1; H2; H3; H4 were accepted, but H5 was rejected. Finally, the author concluded that the measuement models of constructs were satisfied the requirements avergely, because the sampling was not representative; some relationships had not been resolved clearly, the content validity was not completely convincing, the construct of consumer ethnocentrism had not reached totally the convergence validity. 18 5.2 Results of the second study 5.2.1 Sample profile The sample comprised 25,3% men and 74,7% women. A total of 45,2% of respondents had undergraduate education, 11,5% of respondents had graduate-level education or above, and 26,3 % of respondents had completed high school. The other respondents had a lower education level. A total of 15,8% of the sample were 24 years of age or younger, nearly 43,8% were aged from 25 to 40, and approximately 38% of respondents were 41 or over. 5.2.2 Evaluating the measurement scale 5.2.2.1 Factor analysis procedure After the EFA and Cronbach’s Alpha procedures, the author concluded that the measurement scales of the four latent constructs had achieved reliability and unidimensionality requirements. All of the measurement scales met the convergent and discriminant validity. Starting with 20 observed variables, after the preliminary evaluation procedure, the author got 19 observed variables for CFA analysis. The author performed the CFA procedure following 06 steps: Step 1. Specify the model. Step 2. Evaluate model identifcation Step 3. Select the measures and collect, prepare, and screen the data. Step 4. Estimate the model: The values of indices indicated the goodness of fit of the CFA model. In particular, NC = 3,5 <5; RMSEA = 0,065 was greater than 0,05 and lower than 0,07; TLI = 0,941 and CFI = 0,949 were greater than 0,9 and very close 1. All the factor loadings were statistically significant. Figure 2 CFA model in the second study Step 5. Because there was not any eliminated observed variable in step 4, so the author should not do respecify the CFA model. 19 Step 6. Report the CFA results: The author concluded that the measurement models of constructs achieved the requirements