The relationship between service quality and customer loyalty, an empirical study on pharmasy customrrs in ho chi minh city

HOANG NGOC CAM TUTHE RELATIONSHIP BETWEEN SERVICE QUALITY AND CUSTOMER LOYALTY: AN EMPIRICAL STUDY ON PHARMACY CUSTOMERS IN HO CHI MINH CITY MASTER THESIS OF BUSINESS ADMINISTRATION HO C

HOANG NGOC CAM TU

THE RELATIONSHIP BETWEEN

SERVICE QUALITY AND CUSTOMER LOYALTY: AN EMPIRICAL STUDY ON PHARMACY CUSTOMERS IN HO CHI

MINH CITY

MASTER THESIS OF BUSINESS ADMINISTRATION

HO CHI MINH CITY – 2012

HOANG NGOC CAM TU

THE RELATIONSHIP BETWEEN

SERVICE QUALITY AND CUSTOMER LOYALTY: AN EMPIRICAL STUDY ON PHARMACY CUSTOMERS IN HO CHI

MINH CITY

Subject: Master of Business AdministrationCode: 60.34.01.02

MASTER THESIS OF BUSINESS ADMINISTRATION

For the completion of this thesis, I would like to extend my deep gratitude

to Dr Tran Ha Minh Quan, who gave me intensive support, valuable suggestion,guidance and encouragement during the time I conduct my thesis Through hisguiding I found the way to conduct thesis successfully and also the way to solveresearch problems

I also want to say thankful to my classmate at eMBA K19 for their sharingknowledge and honestly support during the time we studied and conducted thesis,their passion to finish thesis and encouragement gave me more strength to getover objection on the way to graduation

Last but not least, I am grateful to my husband and my family for theirwarm encouragement during the time I finished the thesis

Although I has tried the best to complete the thesis, but errors could not becomprehensively avoided Therefore, I am looking forward to receivingcomments from respectful lecturers and friends, so that I can improve more onthe quality of this thesis

Hoang Ngoc Cam Tu

Ho Chi Minh, 25 October 2012

I would like to commit that this thesis, “THE RELATIONSHIPBETWEEN SERVICE QUALITY AND CUSTOMER LOYALTY: ANEMPIRICAL STUDY ON PHARMACY CUSTOMERS IN HO CHI MINHCITY”, was accomplished based on my independent and serious studies andscientific researches The data was collected in reality and it has clear origins Inaddition to that, the data would be trust-worthily handled and it has never beenreleased in any menu

Hoang Ngoc Cam Tu

ACKNOWLEDGEMENT I COMMITMENT II CONTENTS III TABLES IV FIGURES V

CHAPTER I: INTRODUCTION 1

1.1 Problem Statement 1

1.2 Research Objective 2

1.3 Contribution 2

1.3.1 Academic Contribution 2

1.3.2 Business Contribution 3

1.4 Scope and Approach 3

1.5 Research Structure 4

2.1 Service Quality 6

2.2 Customer Loyalty 9

2.3 Relationship Between Service Quality and Loyalty 11

3.1Research Design 14

3.2 Questionnaire Development 16

3.2.1 Service quality scale 16

3.2.2 Customer loyalty scale 18

3.3 Questionnaire Translation 19

3.4Group Discussion 20

3.5 Target population 21

3.6 Sample size 21

3.7 Selecting the sample and collecting data 22

3.8 Method of data analysis 23

3.8.1 Data cleaning . 23

3.8.2 Reliability . 23

3.8.3 Confirmatory Factor Analysis (CFA) 24

3.8.3 Structural Equation Model (SEM) 26

3.8.5 Analysis of variance (one-way and two-way ANOVA) 27

4.1 Characteristic of qualified respondents 29

4.2 Normality analysis 31

4.3 Reliability of the measurements 33

4.3.1 Reliable Test of Service Quality (SERVPERF) Measurement 33

4.3.2 Reliable test of customer loyalty scale 36

4.4 Confirmatory factor analysis 36

4.5 HYPOTHESIS TESTING 42

4.6 Analysis of variance 45

4.6.1 The effect of gender on loyalty 45

4.6.2 The effect of age on loyalty 47

4.6.3 The effect of income on loyalty 48

5.1 Discussion on findings 50

5.2 Practical implications 51

5.3 Contribution of the current study 52

5.4 Limitations and Future research 52

REFERENCES VIII APPENDIX XVIII

TABLES

Table 2-1: Five Dimensions of Service quality

Table 2-2: Loyalty Scale

Table 3-1: Survey Items Used in The Study-Service Quality Scale

Table 3-2: Survey Items Used in The Study-Customer Loyalty Scale

Table 3-3: Assessing fit indices

Table 4-6: Composite reliability

Table 4-7: Standardized Regression Weights – Hypothesized Model

Table 4-8: Assessing Fit Indices – Modified Model

Table 4-9: Regression Weights: (Group number 1 - Default model)

Table 4-10: T-test of gender toward loyalty (BI)

Table 4-11: ANOVA Test Of Age Toward Loyalty

Table 4-12: ANOVA Test Of Education Toward Loyalty

Table 4-13: ANOVA test of income toward loyalty

FIGURES

Figure 2-1: The conceptual Model

Figure 3-1: The research process

Figure 3-2: The translation process

Figure 4-1: CFA model

Figure 4-2: SEM result (standardized estimate)

CHAPTER I: INTRODUCTION

1.1 Problem Statement

From 2001 to 2008, Vietnamese pharmaceutical market has increasedsignificantly with average about 19.9% per year, especially in 2008 the growthwas up to 25.5% compared to 2007 (Tan et al., 2009) In recent years, with thehigher average income, Vietnamese people tend to spend more on healthcareservice While average expense for medicine of Vietnamese people in 1998 wasjust about 5.5 USD, in 2008 the average expense for medicine was 16.5 USDand up to 19.77 USD in 2009 (Tan et al., 2009) The growth in pharmaceuticalmarket is attributed by increasing trend in self-medication, inline with the greaterawareness of health issues

As end-stage in medicine distribution (along with hospitals), pharmacies hasbecome the most important factor to ensure the good consuming medicineproduct to most of consumer, the best place for self-medication Especially inrural areas, where patient rather visit pharmacies than hospital for their minorhealth problem, pharmacies are not only places to buy drug but also places wherepatient can get fast and effective medicine for their simple disease Atpharmacies, patient can be advised how to use medicine safety and effectively

In Vietnam, when patient got health problem, most of them will go topharmacies nearby to take some medicine; this situation make the role ofpharmacies become more and more important

Adapted to Ministry of Health regulation in 2007, that all pharmacies in HoChi Minh City should be improved to GPP (Good Pharmacy Practices) standard

to increase service to consumer, pharmacies has been put in a new competitionfor increasing infrastructure and service quality More and more new GPPpharmacies appeared in HCM city in year 2011 with many formalism ofincreasing service: pharmacist or doctor available, economic price, gift for loyalcustomer…etc Pharmacies are doing their best to increase service quality toattract loyal customer to maximize profit

Aim at understanding how to improve loyalty pharmacy customer, thisresearch is conducted to clarify relationship between service quality andcustomer loyalty

1.2 Research Objective

To examine if any dimension of service quality, that are tangibility,reliability, responsiveness, assurance and empathy effect on loyalty inpharmaceutical market

To investigate if there is any difference on mean of loyalty ofpharmaceutical consumer between demographic factors such as income, age, andgender

1.3 Contribution

1.3.1 Academic Contribution

Although there have been many studies on bank, supermarket and otherservices, the study on pharmacy service is still spare This study is one of fewstudies conducted on pharmacy service This paper contributes to the researchfield of service quality and loyalty in two aspects: First, this paper examine onpharmacy service, which is not many research have been conduct so far Second,this paper explores a model to demonstrate that if pharmacy loyalty is aconsequence of pharmacy service quality?

1.3.2 Business Contribution

For pharmacies: This paper examines if and how factors of service qualityaffect customer loyalty so that pharmacies can chose which factor of servicequality should be improved to retain customers and maximize profit

For retail chain of pharmacies: The understanding of relationship betweenfactors of service quality and customer loyalty in pharmacy market can helpretail pharmacy companies building suitable marketing strategy

1.4 Scope and Approach

This study is conducted in HCM city, the biggest city in Vietnam HCM city

is also the city with highest number of pharmacies nation- wide, this city is alsothe 1st city to have new GPP pharmacy and pharmacy chain With highcompetition, pharmacies in HCM city also focus on increasing service qualitythan in other areas

Data were collected by direct interview to pharmacies customer in HCM city.Respondents who are collected randomly are diversified at any career, age,income, men or women and all of them are customers of different pharmacies inHCM city.

The study is conducted in 2 main phases: pilot study and main study Thepurpose of the pilot study is to check the contents and to examine themeasurement scales Then the main study is conducted to confirm themeasurement scales, to test the hypotheses and to confirm the research model

1.5 Research Structure

The study is reported in 5 chapters:

Chapter 1: Introduction will provide the research background, researchobjectives, as well as research scope and approach

Chapter 2: Literature review provides theoretical and empiricalbackgroundbehind the hypothesized research model

Chapter 3: Methodology details the methodologies the author use to conductthisresearch

Chapter 4: Analysis is conducted based on the collected data to testthehypotheses and answer the research questions

Chapter 5: The author discusses the results; provide practical implication,andpossible directions for future research

CHAPTER II: LITERATURE REVIEW

In this chapter, the author review about previous investigations onrelationship between service quality and customer loyalty to develop thehypothesized model The service quality, which has been investigated anddefined long time ago by many researchers, may be captured with perceptionmodel (Cronin and Taylor, 1994) or captured with disconfirmation model(Parasuraman et al., 1994) However, despite these attempts and its perceivedimportance in marketing theory and practice (Jones and Sasser, 1995; Reichfeld

&Sasser, 1990), customer loyalty still “presents an enigma to researchers”(Oliver, 1996) Implicitly, service loyalty is often incorporated as an ultimate

dependent variable in service quality models (Fornell et al., 1996), yet the

precise nature of the service quality-service loyalty connection has remainedfuzzy as the result of limited conceptualizations and contradictory empirical

results (Boulding et al., 1993;Cronin & Taylor, 1992) Most studies that have

examined the service quality-service loyalty relationship focused on one specificindustry only,limiting the general ability of results (e.g Crosby and Stephens,

1987; Kelleyet al., 1993; Rust and Zahorik, 1993) Some results show that if

customers satisfy with service quality, they tend to repeat purchase with thatservice (Chaudhuri, 1994) In this research the author examine the relationshipbetween service quality and customer loyalty in pharmacy service

2.1 Service Quality

Service quality is difficult to evaluate clearly because services are intangibles,heterogeneous, cannot be stored, and production and consumption aresimultaneous (Athnassopoulos et al., 2001) Researchers are interested in findingthe identification of service quality but they have not agreed how to measure ityet Cronin and Taylor (1992) suggest that perceptions of service quality moreclosely match customer evaluation of the service provided However,Parasuraman et al., (1994) argue that disconfirmation is valid and further, itallows service providers to identify gaps in the service provided In thedisconfirmation model, service quality is defined as the comparison of serviceexpectations with actual performance perceptions (Parasuraman et al., 1991).They also called SERVQUAL model The core idea is that service quality is afunction of the difference scores or gaps between expectations and perception.Therefore, service quality is a multidimensional concept They find fivedimension of service quality are reliability (ability to deliver the promisedservice dependably and accurately), responsiveness (willingness to helpcustomers and provide prompt service), assurance (ability to inspire trust andconfidence), empathy (customers are individuals) and tangibles (elements thatrepresent the service physically)

SERVQUAL is useful scale to measure service quality by making suitableadjustment for industry and country contextual effect SERVQUAL has beenproven the validity and reliability through many ranges of service fields: Tyre

Shop (Carman, 1990); discount and department store (Finn and Lamb, 1991;Teas, 1993), medical services (Brown & Swartz, 1989), hospitals (Babakus

&Mangold, 1992), higher education (Boulding et al., 1993)

However, SERVQUAL is criticized for theoretical and operational issues.Brown et al., (1993) have questioned the validity and the reliability of thedifferences between expectations and performance Tea (1993) questioned thevalidity of perception-expectation gap with conceptual and operational problem

in the definition of the expectation Cronin and Taylor (1992) have suggestedthat perceptions scores alone offer a better indication of service quality, andapplication of SERVQUAL is not possible in new services, but only for existingones

The perceived performance model deviates from SERVQUAL model in thatexpectations play a less significant role in satisfaction formation Because theexpectation comes mainly from past experience of using service provided, butwith new services which consumers have not consume yet they lack ofexperiences to compare or expect about the service they would get Increasingly,Mittal and Lassar (1996) are simply measuring perceptions (SERVQUAL-P) asindicators of service quality, ignoring expectations completely Then, Cronin andTaylor (1992, 1996) have compared difference scores with perception toconclude that perceptions are a better predictor of service quality thandisconfirmation They suggested that the performance based measurementapproach (SERVPERF) is more in conformance with the existing attitudeliterature and is superior to the perception-expectation gap approach

Most recent articles have supported Cronin and Taylor's (1992) based paradigm over Parasuraman et al (1985) disconfirmation-based paradigm

performance-(Oliver 1993; and Babakus and Mangold 1992) One of the most telling reviews

is provided by Boulding et al (1993) They state "our results are incompatiblewith both the one dimensional view of expectations and the gap formation forservice quality Instead, we find that service quality is directly influenced only

by perceptions"

Table 2 -1: Five Dimensions of Service

quality Service quality Definition

Responsiveness Willingness to help customer and provide prompt service

Assurance Employees’ knowledge and courtesy and their ability to inspire

trust and confidence

Empathy Caring, individualized attention given to customers

Adapted from Parasuraman et al., 1991

The author choose the SERVPERF scale primarily instead ofSERVQUAL because of many criticisms of SERVQUAL (Cronin & Taylor1992; Teas, 1993; Churchill et al., 1993), and the failure of empirical studies toreplicate SERVQUAL's initial success (Carman, 1990; Babakus & Boller 1992)

Although there are many studies on diversified service such assupermarket, fast food, banking, retail store…but there are not many studiesconducted on pharmacy service except Taylor and Cronin (1994) measure

Appearance of physical facilities, equipment, personnel and written materials

service quality of professional healthcare service One study reported inpharmacy industry is only about the pharmacist service, not on the pharmacy as astore (Schommer and Wiederholt, 1994).On the other hand; Bastos & Gallego(2008) have found that SERVPERF are suitable to measure service quality inpharmacy service In this study, the author expects that there is a replicatedsuccess to apply SERVPERF in pharmacy service like previous different service,because SERVPERF has been proved to be suitable to measure service quality inthese above services.

2.2 Customer Loyalty

Customer loyalty contributes a lot on the success for a service brand.Marketing research showed that defensive marketing referring to retainingexisting customer is superior than offensive marketing referring to capturing newcustomers For examples, Reichheld & Sasser (1990) found that loweringcustomer defections can have a strong impact on company benefits and marketshare The reason is because fee for marketing of current brands is less than of anew one Chaudhuri (1999) claimed that customer loyalty toward a brand orservice expresses their intention to use and repeat purchase As a consequence,the higher loyalty toward a brand or service, the higher profit the companies gain.Several studies have examined the association between service quality and morespecific behavioral intentions such as positive word-of-mouth, willingness torecommend, repurchase intention For example, Parasuraman et al (1991) find apositive and significant relationship between customer perceptions of servicequality and their willingness to recommend the company Boulding et al (1993)

find a positive correlation between service quality, and repurchase intentions andwillingness to recommend.

Bastos & Gallego (2008) stated that customer loyalty expressed anintended behavior related to the product or service Loyalty also defined asobserved behavior (Liljander & Strandvik, 1995) or actual behavior that drivesthe performance of an industry Repeat purchasing and purchasing sequence aremeasures of actual behavior Loyalty is also an attitude, expressed for example,

in the willingness to recommend a service provider to other consumers (Selnes,1993) Loyalty is also cognitive, that could be operationalized as a product orservice that comes first to mind when making a purchase decision or the product

or service that is the first choice among alternatives (Ostrowski et al., 1993), orprice tolerance (Anderson, 1996; Fornell et al., 1996)

For the construction of service loyalty the constructs are more diversified.Meanwhile the investigations of Bloemer et al (1998), Bloemer (2002) andZeithaml et al (1996) are the best contributions to the research Themeasurements of loyalty are as follow

Table 2-2: Loyalty Scale

Number Items wording

of items

1 Say positive things about XYZ to other people

2 Recommend XYZ to someone who seek for your advices

3 Encourages friends and relative to do business with XYZ

4 Consider XYZ your first choice to buy

5 Do more business with XYZ in the next few years

Source: Zeithalm et al., 1996.

2.3 Relationship Between Service Quality and Loyalty

The relationship between service quality and customer preference loyaltyhas been examined in different research Cronin and Taylor (1992) focused

solely on repurchase intentions, whereas Boulding et al (1993) focused on the

elements of repurchasing and willingness to recommend However, while Croninand Taylor did not find significant positive effect of service quality onrepurchase intentions (in contrast to the significant positive relation between

satisfaction and repurchase intention), Boulding et al (1993) found positive

relationships between service quality and repurchase intentions and willingness

to recommend

Furthermore, there are a number of factors that limit a comprehensiveunderstanding of service quality and its impact on customer loyalty in serviceand prevent the generalization of research findings First, the construct ofcustomer loyalty remain limited, as it ignores the full range of conceivable

loyalty action that may follow the evaluation of the service (Zeithalm et al.,1996) Second, the link between the individual dimensions of service quality andcustomer loyalty received relatively little research attention.

However, recent research gives some evidence that service quality has animpact on customer satisfaction, which in turn leads to later behavior toward theservice firm (Cronin &Taylor, 1992; Anderson & Sullivan, 1993) Furthermore,Bloemer &Ruyter (1998) find a positive relationship between service quality andservice loyalty in five different service industries Similarly, in another studyconducted by Wong et al (1999), a positive relationship was found betweenservice quality and customer loyalty

Indeed, most of the studies linking service quality and behavioralintentions have been conducted in one specific service setting For instance, Rustand Zahorik (1993) relate service quality perceptions to consumer loyalty inbanking, Crosby and Stephens (1987) investigate loyalty in the insuranceindustry and with regard to retailing, customer patronage was investigated in

relation to service encounter failures by Kelley et al (1993).However, there is

very limited research to examine the direct relationship between service qualityand customer loyalty in pharmacy service, although relationship between servicequality and customer satisfaction has been examined in some studies (Hayashi et

al (2005), Kamei et al (2001)) Bastos & Gallego (2008) do not find out asignificant direct relationship between service quality and customer loyalty inpharmaceutical market

Although prior research has not comprehensively examined the impact ofservice quality on service loyalty dimensions, the author expect, in accordance

with Zeithaml et al (1996), positive relationships between service quality and

customer loyalty This leads to the following hypotheses:

Hypothesis 1, 2, 3, 4, 5: Tangibility, reliability, assurance, responsiveness and empathy have direct positive effect on Loyalty

Figure 2-1: The conceptual Model

CHAPTER III: RESEARCH METHODOLOGY

The purpose of this chapter is to address the methods used in this study.Items that will be addressed include research design, questionnaire development,questionnaire translation, pilot study, target population, sample frame, samplesize, sample selection, sample characteristics, data collection and methods ofdata analysis

3.1Research Design

The study contained 2 main phases: Pilot study and main study

Pilot study was conducted by qualitative method to explore, adjust items

to measure research concepts Based on the literature, the 1st draft questionnairewas translated from English to Vietnamese and back by two professionaltranslators and was use to discuss in focus group One group of 10 pharmacycustomers (5 men and 5 women, who are from 25 to 32 years old) was invited to

a focus group in HCM city The purpose of the focus group was to check ifmeaning and content of the questionnaire are understood exactly and, toadjust/add more or delete any items which were not suitable in Viet Nam market.With result from focus group, 1st draft questionnaire was adjusted to finalquestionnaire for the main study

Main study was conducted on 310 pharmacy customers in HCM city bydirect interview The purposes were to test the measurement scales, thehypotheses and the research model 40 of 310 responses were excluded if therewere unanswered questions of the whole questionnaire or illogical answer 270other responses were test reliability, CFA and SEM.

Figure 3-1: The research process

LITERATURE REVIEW

(Develop research model and hypotheses)

GROUP DISCUSSION

10 attendants (To check measurement scale wording

and adjust the scale)

THE MAIN STUDY USING THE

Test Reliability CFA, SEM

CFA: Confirmatory Factor Analysis

SEM: Structural Equation Model

3.2 Questionnaire Development

The survey contained 2 sections including service quality, and customerloyalty All items were assessed on Likert-type scales on which 1 represents

“strongly disagree” and 7 represents “strongly agree”

3.2.1 Service quality scale

Service quality scale was based on SERVPERF scale (Cronin and Taylor,1992) which was used in previous study of many services This instrumentmeasured 5 dimensions of service quality:

Tangibility: This dimension captured tangible facilities of pharmacyincluded costume and appearance of the pharmacy employees, pharmacyinstallation and equipment in view of service

Reliability: This dimension captured pharmacy ability to conduct service

on time and same as their commitment at the first time

Responsiveness: This dimension captured pharmacy’s employee capability

to be willing to serve customer punctually

Assurance: This dimension captured pharmacy’s employee professionalknowledge, polite and warm service to customer

Empathy: This dimension captured that pharmacy’s employee take carecustomer individually

Table 3-1: Service Quality Scale

This pharmacy has modern equipment

The installations of this pharmacy are visually agreeable

The employees of this pharmacy have an agreeable aspect

The equipment of this pharmacy are in view of the service

This pharmacy does what it promises

When I have a problem, this pharmacy demonstrates interest in it

resolution

This pharmacy does the service well at first time

This pharmacy does the service in the promised time

In this pharmacy does not commit errors

This pharmacy has a quickly attendance

The employees of this pharmacy inform you conveniently

The employees of this pharmacy aren’t ever occupied to answer to yourquestions

The employees of this pharmacy are always prepared to help you

You feel secure when you buy this pharmacy

The employees of this pharmacy are always pleasant

The employees of this pharmacy have sufficient knowledge’s to answer to

your questions.

Table 3-1: Survey items used in the study-service quality scale (continue) Empathy

EM1 This pharmacy priority is the customer

EM2 This pharmacy has personnel attendance

EM3 This pharmacy has a convenient horary

EM4 The employees of this pharmacy do personnel attendance

EM5 The employees of this pharmacy answer to your more specific needs

Adapted scale from SERVPERF (Cronin and Taylor, 1992)

3.2.2 Customer loyalty scale

In this paper the scope of this study the author adapted loyalty dimension from behavioral intention scale of Zeithalm et al., 1996

Table 3-2: Customer Loyalty Scale

BI1 I pretend to continue to be customer to this pharmacy

BI2 I consider this pharmacy as my first choice in pharmaceutical servicesBI3 I have only positive things to transmit from this pharmacy

BI4 I recommend this pharmacy to someone that needs my advice

BI5 I stimulate my friends and familiars to buy in this pharmacy

Adapted scale from Zeithalm et al., 1996.

3.3 Questionnaire Translation

The process of translation was as follow:

Figure 3-2: The translation process

Translate Original LanguageQuestionnaire(OLQ1)

To Target Language Questionnaire (TLQ)

Translate Target Language Questionnaire (TLQ) to

Original Language Questionnaire (OLQ2)

OLQ1 was compared to OLQ2

Conclusion

Translator 1

Translator 2

Translator 1&2

The author

Source: Harness & Schoua-Glusberg (1998); Brislin (1980)

Because initial questionnaire was in English, to apply in Vietnam it should betranslated from English to Vietnamese and back in line with cross culturaltranslation procedures (Harkness & Schoua-Glusberg, 1998) The term backtranslation preferred to the translation of translation back in to the source oflanguage The purpose of back translation was to access the quality of the

translation In this study, the author had a committee included two translatorswho made independent translations of the same questionnaire asrecommendation of Brislin (1980) Then in a reconciliation meeting, the twotranslators compared the translations, reconciled discrepancies and agreed on afinal version which taped the best of the independent translations.

In this study OLQ1 and OLQ2 were close meaning and the author decided touse the original questionnaire Target language questionnaire (TLQ) was used asfirst draft of questionnaire for the pilot test

3.4Group Discussion

The 1st draft of questionnaire was used in group discussion The groupdiscussion included 10 respondents (5 men and 5 women) who were from 25 to

32 years old and studying in Ho Chi Minh University of Economy The purpose

of the focus group is to check if meaning and content of the questionnaire areunderstood exactly and, to adjust/add more or delete any items which are notsuitable in Viet Nam market The comments were mostly regarding wording andrepetition of the items

All modification was for wording to make the translation clearer andcloser to Vietnamese For example: changed “nó” to “họ” for more polite word,changed “thể hiện sự quan tâm chân thành trong việc giải quyết” to ”Nhà thuốcgiải quyết một cách chân thành”, changed “bắt mắt” to “đẹp mắt”, changed

“hứa” to “cam kết”…

Finally, the 1st draft of questionnaire was adjusted to final questionnaireused for main study

3.5 Target population

The respondents were collected by asking customers of differentpharmacies in Ho Chi Minh City to give their answers based on the pharmacythat they usually or like to buy (at least 2 times of buying at this pharmacy in theprevious month) The respondents were selected by random selection method.Data were collected by direct interview to pharmacies customer in HCM city.Respondents who are collected randomly are diversified at any career, income,men or women All of them used to be customer of pharmacies in HCM city

3.6 Sample size

The sample size is variable and depends on the estimation method to beused According to Hair et al (1998), sample size needed for Maximum Likelyhood (ML) method should be from at least 100 to 150 However, Hoelter (1993)claimed that the sample size must be at least 200

To conduct Confirmatory Factor Analysis (CFA), the sample size should

be large because this method is based on large-sample theory (Raykov &Widaman, 1995) Bollen (1989) suggested that sample size should be at least 5m(m is the number of parameters to be estimated)

Researchers agreed that the larger the sample size is, the better themeasurement is (Tho, 2011) but not yet agree how large is enough In this paper,there are 27parameters to be estimated, the sample size should be at least 5 x27

=135 However, aim at collecting larger sample for a better result, the authorcollected 310 answer sheets

3.7 Selecting the sample and collecting data

According to Tho (2011), there are two types of sampling: Probabilitysampling and non-probability sampling Probability sampling only can beconducted if the author knows probability of each elements contributing tomodel Probability sampling method is allowed making statistical inferencesfrom sample to the defined population and provided greater external validity ofthe findings Non-probability sampling is a method that the author samplewithout any probability rules In this study, it’s difficult to use probabilitysampling with a complete of full details (income, sex, education, age) Besides,without any information about the probability of the research model is also alimitation to use probability sampling

In this study, samples were collected by snowball sampling method.Snowball sampling is defined as building a sample through referrals (Jean &Marry, 1997) Jean & Marry (1997) claimed that confidence of relationshipswhich were developed over is period of time could increase reliability of thedata

Main field study was conducted from 4-September to 25-September by theauthor and her supporters The respondents were purchaser at pharmaceuticalstores or who had ever purchased at pharmaceutical product at anypharmaceutical store in Ho Chi Minh City 310 responses were collected andbeing cleaned before using for analysis

3.8 Method of data analysis

3.8.1 Data cleaning

The data was cleaned by excluding any answer sheet that had missinganswer After cleaning data, there were remained 270 responses used for dataanalysis

3.8.2 Reliability

The author used Cronbach’s Alpha to test the reliability of the scale (atleast 3 items in each scale) If Cronbach’s Alpha > 60, that scale can be acceptedabout reliability (Nunnally & Berstein, 1994) A reliable scale is a scale withCronbach’s Alpha varying in range of [.70-.80] (Tho, 2011)

The author also use function Scale if Deleted Items of SPSS to considerwhether delete one items or not The author would base on the contentrelationship between item and whole scale and change of Cronbach’s Alpha ifdeleted that items to decide to delete or not

Finally, the author also checked item-total correlation by SPSS According

to Nunnally & Bernstein (1994), if item-total correlation of one item is > 0.3,that item data is qualified

After checking reliability, some items were deleted and the rest of itemswould be used to check exploratory factor analysis (CFA)

3.8.3 Confirmatory Factor Analysis (CFA)

Before conducting structural equation model (SEM), the author conductedconfirmatory factor analysis (CFA) to test the model fit or whether the data fitthe hypothesized in advance CFA is theory-or hypothesis driven With CFA it ispossible to place substantively meaningful constraints on the factor model.Researchers can specify the number of factors or set the effect of one latentvariable on observed variables to particular values CFA allows researchers totest hypotheses about a particular factor structure (e.g., factor loading betweenthe first factor and first observed variable is zero) Unlike Exploratory factoranalysis (EFA), CFA produces many goodness-of-fit measures to evaluate themodel but do not calculate factor scores

A large class of omnibus tests exists for assessing how well the modelmatches the observed data Chi-Square (χ2) is a classic goodness-of-fit measure

to determine overall model fit The null hypothesis is that the implied orpredicted covariance matrix Σ is equivalent to the observed sample covariancematrix S, Σ = S A large χ2 and rejection of the null hypothesis means that modelestimates do not sufficiently reproduce sample covariance; the model does not fitthe data well In contrast, a small χ2 and failure to reject the null hypothesis is asign of a good model fit However, the χ2 test may also be invalid whendistributional assumption are violated, leading to the rejection of good models orthe retention of bad ones χ2 is based on a very stringent hypothesis of Σ = S(Brown 2006:81)

Due to these drawbacks of χ2 test many alternative fit statistics have beendeveloped, though each has its own advantages and disadvantages They are the

χ2/df , the Comparative Fit Index (CFI), the Root Mean Square Error ofApproximation (RMSEA), the significance of parameter estimates, and theamount of squared multiple correlation (Sila and Ebrahimpour, 2005), Tucker-Lewis index (TLI), and Goodness of Fix Index (GFI)

CFI: this index compares the proposed model with a null model assumingthat there are no relationships between the measures CFI ranges from 0(for apoor fit) to 1 (for a good fit) A CFI value that is greater than 0.90 indicates anacceptable fit to the data (Bentler, 1992)

TLI: TLI is another index for comparative fit that “includes a penaltyfunction for adding freely estimated parameters” (Brown 2006:85) TLI can beinterpreted in a similar fashion as CFI, but it can have a value outside of therange of 0 to 1 (p.86)

RMSEA: this index used to assess residuals and adjusts parsimony in themodel Its value must be equal to or less than 0.08 for and adequate model fit(Hu and Bentler, 1999)

GFI: GFI introduced by Jöreskog and Sörbom (1984) Like TLI and CFI,GFI is less than or equal to 1 A value of 1 indicates a perfect fit

There is no single evaluation rule on which everyone agree However,researchers agree that when RMSEA values are below 0.8 and CFI, TLI rangefrom 0.9 to 1, CMIN/DF range from 2 to 5 indicate a good model fit

Table 3-3: Assessing fit indices

Goodness- of fit Indices Code Desired rang of values

for a good fit

Absolute Fit Indices

p > 05

Chi-Square/Degrees of Freedom ratio χ2 /df < 5

Root Mean Square Error of Approximation RMSEA ≤ 080

Incremental Fit Indices

Source: Albright & Park, 2009; Brown (2006); Jöreskog and Sörbom (1984)

3.8.3 Structural Equation Model (SEM)

In this paper, the author using the structural equation model (SEM)technique to test the hypothesis SEM was first seen in paper in 1960s(Anderson,J.C & Gerbing D,W., 1988) SEM has more advantages than othersmethod in validity, reliability and complex models (Werner, C & Schermelleh-Angel, K., 2009)

Validity: Theories in the social sciences usually analysis variables thatcannot be observed (constructs), but that can only be predicted throughobservable variables (latent variables, or indicator variables) To operationalizethese constructs, often many different variables come into consideration, andnone of them may provide an optimal operational in its own Structural equationmodeling allows the author to use several observable variables per constructsimultaneously, which leads to more valid conclusions on the construct level

Using other methods of analysis would often result in less clear conclusions,and/ or would need separate analyses.

Reliability: Date in the social sciences frequently contain a non-negligibleamount of measurement error Structural equation modeling can takemeasurement error into consideration by explicitly including measurement errorvariable that correspond to the measurement error portions of observed variables.Therefore, conclusions about relationships between constructs are not biased bymeasurement error, and are equivalent to relationships between variables ofperfect reliability

Complex models: Theories in the social sciences usually refer complexpatterns of relationships between a multitude of variables, conditions or groups.Structural equation modeling allows to model and test complex patterns ofrelationships, including multitude of hypotheses at the same time as a whole(including mean structures and group comparison) Using other methods onanalysis, this would usually need several separate analyses

The estimation method in this study is maximum likelihood, which hasbeen a predominated estimate method Under the assumption of a multivariatenormal distribution of the observed variables, maximum likelihood estimatorshave the desirable asymptotic, or large-sample, properties of being unbiased,consistent, and efficient (Anderson,J.C & Gerbing D,W., 1988)

3.8.5 Analysis of variance (one-way and two-way ANOVA)

Independent sample T-test was used to examine the equality betweengender (male and female) in customer loyalty One-way ANOVA was used to

test the equality among age, income and education group because these variablehave more than two groups Tho, 2011 cited that to test the equality of more thangroup, we use One-way ANOVA After conduct T-test and One-way ANOVA, ifthere was any different of equality in customer loyalty between group the authorwould continue conduct Two-way ANOVA to further test.

CHAPTER 4: DATA ANALYSIS

This chapter presents sample characteristics, statistical analysis, test ofhypotheses, answers the research questions and other finding from this study.Data analysis included descriptive statistics, reliable test of measurement,confirmatory factor analysis (CFA), and regression analysis through SPSS 16.0and AMOS 18.0 software

4.1 Characteristic of qualified respondents

Among 270 qualified respondents, 157 were female (58.1%) theremaining were male Even though the author did not consider age as a criteria tochoose respondent, their age ranges from 18 to 60; that may be due to all

respondent had full time job They mostly graduated employees Their Income mostly under 10 VND million/month (207/270 respondents, 77.2%) (Table 4-2).

Characteristic of qualified respondent were described as follow (table 4-2)

Table 4-2: Socio-demographic Characteristics of The Qualified respondents Variables Frequency Percent % Valid % Cumulative %

4.2 Normality analysis

Table 4-3: Assessment of Normality

Statistic Statistic Statistic Std Error Statistic Std Error

Table 4-3: Assessment of Normality (continue)

Statistic Statistic Statistic Std Error Statistic Std Error

Skewness is a measure of symmetry, or more precisely, the lack of

symmetry A distribution, or data set, is symmetric if it looks the same to the left