CHAPTER 3 METHODOLOGY
3.4 Questionnaire Design
All our data will be collected by paper-based questionnaires. Most items will be answered on a five-point Likert scale ranging from 1 to 5, with 1 being “strongly disagree”
and 5 being “strongly agree.” Willingness to communicate to others is a self-report scale which is scored by the participants themselves. All the original items were in English and then translated into Traditional Chinese and then back into English by another native speaker who is also fluent in Chinese to ensure the content and meaning remained the same during the translation process. There are two different parts of this questionnaire design. According to Churchill, “once a construct is clearly defined, the next step is to generate a set of items that clearly capture the domain of the construct (Churchill, 1979).” A pre-test has to be conducted in order to be sure the wording and expressions are clear to our participants. The first step of our questionnaire is to have 30 college students fill out the questionnaire and give opinions and suggestions on the wording and content. After making appropriate adjustments, a revised version will be handed out to college students in Taiwan.
Our official questionnaire is divided into 5 sections, electronic word of mouth, trust, willingness to communicate, tie strength, and demographic variables. Electronic word of mouth activities will be measured by Churchill 1979’s word of mouth scale. Originally there
were 13 items to measure the construct, yet Harrison-Walker (2001) eliminated 7 items after performing a scale purification process which refined the scale into a “more reliable and meaningful scale in assessing word of mouth (Harrison-Walker, 2001).” Positive electronic word of mouth was measured by 3 items from Liljander and Strandvik (1997)’s loyalty scale which demonstrated high reliability statistics. On the other hand, the scale used to measure electronic negative word of mouth was originally used in Liu and McClure’s research on cross-cultural customer complaint behavior study. The scale was adapted to suit the
contextual characteristics of this particular study. In order to help our participants be clear of the difference between electronic and traditional word of mouth, examples of electronic platforms and the definition of traditional word of mouth was given at the beginning of section 1 and section 2. The original set of items was measured on a 7 point Likert type scale.
Yet to help improve the efficiency and increase the unity of our questionnaire, scales were modified from 7 point to 5 point.
Section 1: Demographic Variables
Table 3-1. Demographic variables Construct Scales of measurement Item
Gender Nominal scale 1. Male 2. Female
Age Ordinal scale 1. Under 18
2. 19~23 3. 24~28 4. 29~33
Construct Scales of measurement Item 5. 34~38 6. Above 39 University
Section 2: Electronic Word of Mouth
Table 3-2. Measuring an individual’s electronic word of mouth items Construct Dimension Scales of Measurement Items
Electronic
5 point Likert type scale 1. I mention the product which I use to others quite frequently.
2. I’ve told more people about certain products than I’ve told about most other products.
3. I seldom miss an opportunity to tell others about certain products.
4. When I tell others about a certain product/service, I tend to talk about it in great detail.
Positive
Section 3: Trust
Rotter’s Interpersonal trust scale is an extremely often used tool to measure the level of trust one has towards others. Developed in 1967 with 25 items measuring trust, and 15 filler items, this scale is “designed to measure a person’s generalized expectancy that the promises of another individual which can be relied on (Rotter et al., 1971)”
Table 3-3. Measuring an individual’s interpersonal trust items Construct Dimension Scales of Measurement Items
Trust Interpersonal trust
5 point Likert type scale
1. Hypocrisy is on the increase in our society.
2. This country has a dark future unless we can attract better people into politics.
3. Using the honor system of not having a teacher present during exams would probably result in increased cheating.
4. Taiwan will never be an
effective force in keeping world peace.
5. Most people would be horrified if they knew how much news the public hears and sees is distorted.
Construct Dimension Scales of Measurement Items
6. Even though we have reports in newspapers, radio, and T.V., it is hard to get objective accounts of public events.
7. If we really knew what was going on in international politics, the public would have reason to be more frightened than they now seem to be.
8. Many major national sports contests are fixed in one way or another.
9. Most people can be counted on to do what they say they will do.
10. In dealing with strangers one is better off to be cautious until they have provided evidence that they are trustworthy.
11. Fear of social disgrace or punishment rather than conscience prevents most people from breaking the law.
Construct Dimension Scales of Measurement Items
12. Parents usually can be relied upon to keep their promises.
13. The judiciary is a place where we can all get unbiased treatment.
14. It is safe to believe that in spite of what people say most people are primarily interested in their own welfare.
15. The future seems very promising.
16. Most elected public officials are really sincere in their campaign promises.
17. Most experts can be relied upon to tell the truth about the limits of their knowledge.
18. Most parents can be relied upon to carry out their threats of punishment.
19. In these competitive times one has to be alert or someone is likely to take advantage of you.
Construct Dimension Scales of Measurement Items
20. Most idealists are sincere and usually practice what they preach.
21. Most salesmen are honest in describing their products.
22. Most students in school would not cheat even if they were sure
of getting away with it.
23. Most repairmen will not overcharge even if they think you are ignorant of their specialty.
24. A large share of accident claims filed against insurance
companies are phony.
25. Most people answer public opinion polls honestly.
Section 4: Willingness to Communicate
In 1985, McCroskey proposed a scale to measure willingness to communicate which is a personality-based trait like scale that is consistent across between different receivers and different communication contexts. McCroskey has proved that the level of willingness of a particular context (small group) is correlated with this individual’s willingness in a different context (with individuals and large meetings). In addition, the willingness to communicate
with a certain type of audience (for example, friends) is also correlated with the willingness to talk with other different types of people (strangers and acquaintances). Yet this does not mean that a person will be equally willing to communicate in all different contexts and receivers.
What this means is that they will be correlated. This scale has 20 items, with 8 items being filler items that will not count nor influence the final results of our scale. The willingness to communicate scale includes four communication contexts and three different receivers. The scale can measure the overall willingness to communicate represents an individual’s general personality orientation to communicate with others as well as 7 different sub-scores.
Participants will fill out the percentage of how much they are willing to communicate in each different scenario.
Table 3-4. Measuring an individual’s willingness to communicate items Construct Dimension Scales of
Measurement
1. Talk with an acquaintance while standing in line
Construct Dimension Scales of Filler items 1. Talk with a service station
attendant 5. Talk with a waiter/waitress in a
restaurant
6. Talk with a secretary
Construct Dimension Scales of
Filler items 7. Talk with a garbage collector 8. Talk with a spouse
(girlfriend/boyfriend)
Section 5: tie strength between individual and electronic word of mouth receiver
In section 4, we study the tie strength of each participants and the electronic word of mouth receiver whom he has interacted with most recently. Hansen’s two item scale measuring the closeness of a working relationship and frequency of contact developed on 1999 is used in our study to measure tie strength. According to Levin’s pre-test, instructions were given to “choose 7 for these two questions if you have never had prior contact with this person (Levin, 2002).” Again, to help simplify the procedure of filling out this questionnaire, this set of items were modified from a 7 point Likert scale to a 5 point scale.
Table 3-5. Measuring an individual’s tie strength with another person in eWOM items Construct Dimension Scales of measurement Items
Tie Strength
5 point Likert scale 1. How close was your relationship with him/her?
2. How often did you communicate with him/her?
3.5 Questionnaire Pretest
Two pretests were done before the official questionnaire was handed out to participants.
Cronbach Alpha of 0.7. The reliability of a scale means the consistency and stability of a scale.
In 1951, Cronbach proposed that a method that is now known as “Cronbach Alpha” that could measure the reliability. The standard of how much the Alpha has to be for the instrument to be accepted is different for each area of research. In most studies, 0.6 is enough for the
instrument to be declared acceptable. In the first pretest executed, many participants
responded that the items of the trust scale were hard to understand and confusing. Therefore there was no surprise when the internal consistency of this scale received a fairly low score of 0.593. Item 8 had a negative item to total correlation, and was removed since the scale could achieve a 0.657 Cronbach Alpha if this was done. The original electronic word of mouth scale consisted of 2 dimensions which are word of mouth activity and positive word of mouth.
Electronic word of mouth activity reached a Cronbach Alpha of 0.622, with the last item having a negative correlation with the entire scale. If the item were to be deleted, the Cronbach Alpha would rise to 0.886. Therefore, the fourth item was deleted in our second pretest. Along with the replacement of a new positive word of mouth scale that consisted of 3 items, 2 more items measuring negative word of mouth were added to strengthen the
contribution of our study. As mentioned, the positive word of mouth scale was replaced by a new one due to the extremely low Alpha of 0.197 which could not be fixed by removing items or revision of translation and wording. Although electronic word of mouth activity also
achieved a low internal consistency score, yet participants also responded that the wording was difficult and hard to understand. For the same reason, this scale was given another chance believing that it would operate better in our second pretest after careful revision. Willingness
to communicate achieved a high Alpha of 0.803 as well as tie strength’s 0.910.
Table 3-6. Reliablity statistics of first pretest
Construct Dimension Cronbach α Item(s) removed Cronbach α after item removed
eWOM eWOM activity 0.622 Item 4 0.886
Positive eWOM 0.197 Scale Removed --
Tie Strength -- 0.910 -- --
Trust -- 0.593 Item 8 0.657
Willingness to Communicate
-- 0.803 -- --
In our second pretest, the questionnaire made several revisions. First, demographic variables were moved from Part 1 to Part 5. This was because most scholars believe that demographic questions might be more sensitive and lead to a low return rate of questionnaires if they were placed at the beginning of the survey. Second, electronic word of mouth was rated with a total of 8 items including newly added 2 item scale on negative electronic word of mouth. Third, 1 item was removed from our trust scale in hope of improving the internal consistency. Furthermore, more directions were added to each section to help our participants understand each part of the questionnaire and how to respond to the questions better. A total of 15 questionnaires were used to run a statistical analysis in our second pretest.
Table 3-7. Reliability statistics of second pretest
Construct Dimension Cronbach α Item removed Cronbach α after item
The results for our second pretest all passed the bar of 0.7 internal consistencies except for “trust.” Four items that were negatively correlated with the scale were removed, which resulted in a Cronbach’s α 0.713 if so. The final official questionnaire had a total of 53 questions, including demographic variables.
In our study, we will use several types of statistical analysis method to explain the collected data and help understand our research framework. These methods include:
1. Regression Analysis: Regression analysis is used to explore the relationship between dependent and independent variables. In our study, we will use regression analysis to analyze the relationships hypothesized in our study.
2. Cronbach’s α Coefficient: Cronbach’s α is used to test the reliability of the data
we have retrieved. It will test whether or not the data is internally consistent. Therefore, the higher the constructs achieve on this score, the higher the constructs are internally related and reliable. According to Roberts and Wortzel, the alpha coefficient between 0.7 and 0.98 reflects high reliability. Therefore 0.7 will be the standard to examine whether out constructs are internally consistent.
3. Pearson Coefficient: Pearson Coefficient Correlation is used to examine the relationship between the variables in our study. Whether they are positively or negatively correlated, and if there is a significant relationship and the strength of that relationship is discussed using Pearson analysis.
CHAPTER 4 RESEARCH RESULTS AND DATA ANALYSIS
With the help of acquaintances who are students in different colleges in Taipei, questionnaires were distributed in classrooms and a total of 171 valid questionnaires were retrieved. Using SPSS version 17 software, the procedure of statistical analysis will be conducted to examine our study and test the hypotheses.
4.1 Reliability Analysis
Our study conducted a reliability analysis for the 4 different variables included in the study. The results were fairly good, with each constructs’ Cronbach’s α above 0.7. This means that the items in the questionnaire are highly correlated and also consistent. Electronic word of mouth activity reached a Cronbach α of 0.783, positive electronic word of mouth 0.775, negative electronic word of mouth 0.811. Our trust scale has a Cronbach αof 0.756, scoring much higher than the 2 pretests conducted. Furthermore, the 12 items of willingness to communicate had an internal consistency of 0.810, while tie strength had the highest internal consistency of all, 0.921.
Table 4-1 Cronbach Alpha Analysis Results
Construct Cronbach α Number of Items
Trust 0.765 20
WTC 0.810 12
Tie Strength 0.921 2
eWOM Activity 0.783 3
Positive eWOM 0.775 3
Negative eWOM 0.811 2
4.2 Regression Analysis
Table 4-2 Regression Analysis Results
Independent Variable Dependent Variable P-Value R-Square
Trust WTC 0.048 0.023
WTC eWOM Activity 0.001 0.059
WTC Positive eWOM 0.003 0.051
WTC Negative eWOM 0.271 0.007
Trust eWOM Activity 0.547 0.002
Trust Positive eWOM 0.677 0.001
Trust Negative eWOM 0.461 0.003
Through table 4-2, we can see that the R-square is relatively low. This means indicates that the research framework can only explain a very low percentage of the dependent variables. However, these results do not mean that the model should be rejected since there might be other possibilities that influence the result. The regression analysis also points out that four of our hypotheses are accepted with a P-Value that is significant. Trust has a significant effect on willingness to communicate, whereas willingness to communicate has a significant relationship with electronic word of mouth activity and positive electronic word of mouth.
4.3 Pearson Coefficient Analysis
Table 4-3. Pearson Coefficient Analysis of Variables eWOM
**. Correlation is significant at the 0.01 level (2-tailed).
*. Correlation is significant at the 0.05 level (2-tailed).
Using Pearson Correlation Coefficient to analyze the correlation between our variables, we can explain the relationships between them and examine whether or not our hypotheses are accepted. Table 4-3 gives a clear view of the correlations between our constructs.
Electronic Word of Mouth Activity has a positive relationship of 0.425 with Positive Electronic Word of Mouth, 0.348 with Negative Electronic Word of Mouth, and 0.243 coefficients with willingness to communicate. This means that participants who engage in electronic word of mouth activity more often share positive feedback and advice instead of spreading out negative opinions. On the other hand, electronic word of mouth activity does have a highly positive significant relationship with willingness to communicate. Positive electronic word of mouth has a positive relationship with negative word of mouth, yet not as
highly correlated as it is with electronic word of mouth activity. This might be individuals who tend to spread the good about products either keep the bad shopping experience to themselves or don’t have unhappy purchase experiences. Positive word of mouth also has a positive relationship with tie strength and willingness to communicate. This means that the stronger the relationship between two individuals, the more they are likely to share their experiences of good product purchases with each other. Also, the higher of communication willingness, the more an individual will share positive word of mouth with. To our surprise, negative electronic word of mouth is the only dimension of Electronic Word of Mouth that does not have a significant relationship with willingness to communicate. We assume a logical explanation might be that perhaps the people tend to keep quiet about negative experiences. Therefore, even if one is willing to communicate and trusts another individual, this does not represent that they will share their negative experience with you. Yet this result is contrary to the past literature, since customers are known to spread negative opinions further and wider compared to their positive experiences.
Tie strength has a positive and significant relationship with both positive and negative electronic word of mouth, yet not with electronic word of mouth activity. An explanation might be people who have stronger tie strength spend less time engaging in the same
cyberspace taking about products, while they do give more straightforward comments to each other about products they like or dislike. Although the results of the data analysis does not support hypotheses 2a, 2b, and 2c, yet trust does have a significant relationship of 0.152 correlation with willingness to communicate. This shows that there needs to be a mediator to
bridge the relationship between trust and willingness to communicate and also that the main argument of our study is accepted, the higher of interpersonal trust an individual has towards, the more he is willing to communicate and talk with others.
CHAPTER 5 CONCLUSION AND DISCUSSION
Our primary goal is to study whether or not willingness to communicate is a mediator between trust and electronic word of mouth. The results show that our hypothesis is accepted.
In Chapter 5, we will discuss our research results and also provide suggestions on future research in this area.
5.1 Conclusion
Figure 5.1 Correlations and Significance Relationship proven to exist
The purpose and aim of our research focused on the interaction of electronic word of mouth, trust, willingness to communicate and tie strength. After a thorough procedure of testing our hypotheses, several of the hypotheses established by literature review were proven to be accepted. In other words, there was a significant relationship found between several variables. First and most importantly, we find a significant positive relationship between trust and willingness to communicate. This means that willingness to communicate is indeed a mediator that acts like a bridge between trust and electronic word of mouth activity, and positive electronic word of mouth. However, it does not have the same effect on negative
0.152*
0.243**
Trust WTC
eWOM Activity
Positive eWOM Negative eWOM 0.226**
Tie Strength
0.212**
0.181*
word of mouth. We search for a reasonable explanation for this research result and find that there might be two possibilities. There is a famous Chinese saying called: “Praise the good, and hide the evil” which means that there is a tendency in people that wants to portray the best of themselves in front of others. Giving positive comments and praising products might give such a desired impression and help build a positive personal image. On the contrary, consumers might take into account the more negative self image or character that might be seen as a reflection if given complaints and negative reviews. The second explanation is the
word of mouth. We search for a reasonable explanation for this research result and find that there might be two possibilities. There is a famous Chinese saying called: “Praise the good, and hide the evil” which means that there is a tendency in people that wants to portray the best of themselves in front of others. Giving positive comments and praising products might give such a desired impression and help build a positive personal image. On the contrary, consumers might take into account the more negative self image or character that might be seen as a reflection if given complaints and negative reviews. The second explanation is the