Survey Study

In the survey section, a total 280 questionnaires were submitted and a total of 241 questionnaires returned. Nevertheless, respondents marking “Other” as their mother tongue were excluded due to thesis purpose being narrowed to comparison between Filipinos and Taiwanese. Thus, after exclusion of 15 questionnaires, the sample consisted of 226 respondents, 91 (40.3%) males, 135 (59.7%) females. The local

43

language being used is Taiwanese (114, 50.4%) and Tagalog (112, 49.6%), respectively. The age groups of respondents are drawn in the figure 3.

Figure 3. The average age of participants

Respondents submitted information concerning the highest education they have reached. In this case, the sample consisted predominantly of people obtaining university/polytechnic degree (54 %).

Figure 4. The education level of participants

44

3.2.2.2 Instruments

As recommended in the theoretical part, service quality perception in the call center background was most regularly deliberated by two tools, SERVQUAL and

RECOVSTAT (Jaishwal, 2008). Statements contained in the questionnaire reflect categories of SERVQUAL. As stated in previous chapter. SERVQUAL, SERVQUAL contains of five service dimensions (tangibles, reliability, responsiveness, assurance, and empathy). Tangibles item was left out because of its irrelevance for the thesis. A 7-point Likert-type scale (completely disagree – completely agree) was chosen because it gives respondent enough variety for expressing opinion. Table 2 presents SERVQUAL categories and example statements in the questionnaire.

Table 2. SERVQUAL categories reflected in the questionnaire

SERVQUAL Example question

RELIABILITY My overall impression from the service is good

RESPONSSIVENESS I am satisfied with the way my complaint was handled ASSURANCE I believe that Customer Care person did her best to

help me

EMPATHY In my opinion Customer Care person was polite

The purpose of thesis was to determine whether service provided by non-local speaker is perceived differently compared to the one conducted by local speaker. Therefore, two questionnaire´s version were developed differing in the one detail:

(1) In scenario respondents were supposed to read prior to question´s answering (non-local speaker) was added following sentence “You can clearly hear that she [Customer Care employee] is not a local speaker of your language because she has a foreign accent”

45

After reading a short scenario respondents were asked to express their opinion on the seven-item Likert-type scale (1= completely disagree, 7= completely agree). Below are the questions from the questionnaires. Negatively worded questions were reversed prior to statistical analysis. Statements marked with (r) are the reverse coded items.

Example questionnaires may be found in Appendix 4. (Local speaker version) and 5.

(Non-local speaker version). The statements were preliminary divided into following 4 categories.

(1) Perceived service quality (4 items)

I am satisfied with the way my complaint was handled

The solution Customer Care person offered was customer-friendly I will purchase from this company also in the future

My overall impression from the service is good

(2) Opinion on the AGENT´s competence and behavior (5 items) I think that Customer Care person was competent

I am satisfied with behavior of Customer Care person towards me I wonder why call center employed exactly this person (r)

I felt annoyed by need to repeat my sentences (r)

There was a high probability of misunderstanding during this call (r)

(3) Opinion on the AGENT´s helpfulness and politeness (3 items) In my opinion Customer Care person was polite

I believe that Customer Care person did her best to help me

I believe that other Customer Care person would serve me better (r)

46

(4) Opinion on the AGENT´s trustworthiness (2 items)

I wonder why Customer Care person complained about poor lines when I heard her well (r)

I think that poor phone connection was just an excuse for her not really understanding me (r)

One statement was not preliminary placed into any category because it did not particularly suit to any cluster. Statement “I think that call center is located outside of country” was intended to test whether customers perceive higher probability of

service being allocated abroad when interacting with a non-local speaker.

3.2.2.3 Procedures

Questionnaire, originally developed in English, was translated into English and Chinese. Both versions were reviewed by local speakers to ensure that content is grammatically correct and wording is as similar as possible. Both Taiwanese and Filipino respondents were told that the purpose of thesis is to determine call center function influence on the company´s image.

Chinese version was obtained via networking. Asking friends and acquaintances to fill and distribute the questionnaires could have affected the final results (57 valid answers) [respondents marking “Other” as their mother tongue were omitted from the study). Unfortunately, no company was interested to participate in questionnaires´

distribution and later benefit from the results.

The English sample answers were obtained in two ways. 56 valid answers were obtained via networking. Asking friends and acquaintances to fill and distribute the

47

questionnaires could have affected the final results. Unfortunately, no company was interested to participate in questionnaires´ distribution and later benefit from the results.

Prior to questions answering, respondents were briefly explained the nature of research and its purpose (master thesis). After reading a short scenario, respondents answered a set of questions directly related to the text. Firstly, they were asked to evaluate the company performance and satisfaction with the complaint handling.

Finally, after filling the questionnaire, respondents were kindly asked for information regarding their sex, age, education and mother tongue. Since the thesis purpose is comparison between Taiwanese and Filipinos, only questionnaires where respondent marked Chinese or English as a mother tongue were included.

3.2.2.4 Analyses

As Hair et al. (2010) point out MANOVA is suitable for assessing group differences across multiple dependent variables simultaneously. In this study MANOVA was used for hypotheses testing as well as for assessing differences in cross-national comparison in case of more variables involved. Additionally, t-test was conducted in case of single dependent variable. Hair et al. (2010) define t-test as a “test to assess the statistical significance of the difference between two sample means for a single dependent variable.” When comparing only two means, t-test for independent samples and ANOVA gives identical results (www.experiment-resources.com). Thus, t-test was performed when possible.

Prior to the analysis, gained data was examined for not violating the basic assumptions. Firstly, checking for missing data was performed. As Pallant (2005)

48

points out it is hardly the case that complete data would be obtained from all respondents. In this study, the check showed that 11 values were, indeed, missing.

Nevertheless, the number did not exceed three responses per item and mostly one answer per question was reported. Table 10 lists the missing answers per item.

However, since the sample was rather large (N=217), the amount of data missing did not threaten the analysis course. Participants with missing responses were not excluded from the analyses.

Next, the normality of the data was assessed. This process was done in order to test basic assumptions requested by method used for hypotheses´ testing. Normality of the data may be assessed by multiple ways. In this study it was screened by skewness and kurtosis scores which are described below. Additionally, the PASW analysis was run (By using “Explore” option). For data to be considered as normal, the Kolmogorov – Smirnov statistics should be non-significant (over .05). In this study all items were significance of .00 which is considered as normality assumption violation (Pallant, 2005). However, as Pallant (2005) states this is fairly ordinary in bigger samples.

Skewness and kurtosis were checked next. In the perfect case both values would be 0 but as Pallant (2005) states out this is rather strange in social sciences. Out of 15 items checked for skewness, 8 items scored positive values and 7 items reached negative skewness. These values indicate whether the mean deviations for items are going to be positive or negative. Regarding kurtosis, all items were assigned negative values suggesting the flat distribution.

49

According to Pallant (2005) this fact indicates that many cases fall into extreme values. Nevertheless, with larger samples (N≥200) skewness and kurtosis values are not of high difference (Tabachnick and Fidell, 2000 in Pallant, 2005). Therefore, it may be supposed that even though data is not distributed normally, it is not an obstacle for this study. Moreover, data suitability for the particular method was monitored prior to concrete analysis (MANOVA and t-test).

50