3.1 Conceptual research framework
Design scenario with the same service and the different level of participation and self-efficacy.
Choose 10 participants for each scenario and make sure the efficiency of scenarios and scales.
Sample size would be 2(participation level:
high and low) X 2(self-efficacy: high and low) X2(outcome: successful and
unsuccessful). Participants of each cell are composed of nearly half male and half female.
Randomly assign each participant to a cell.
Collect 240 samples. Each cell has 30 samples.
17
3.2 Designing scenario
The scenario used in this study was: an experience of English cram school. In that scenario, the service failure is defined as “You didn’t get a good grade and didn’t pass the
test. “ The reasons why we choose the English cram school are as follow. First of all, Yen
et al. (2004) showed that participation consists of information sharing, responsible
behavior and personal interaction. Using school for the scenario can fit that definition and
differentiate the level of participation easily. Second, self-efficacy has often been
investigated in educational studies, and it is easy to measure the level of self-efficacy.
Third, the experience of going to a cram school is familiar to people in Taiwan, and thus
participants could imagine such scenario easily. Although the scenario has some
deficiencies, we believed it was still appropriate for this research.
3.3 Experiment Design
A 2(participation level: high and low) x2(self-efficacy: high and low) x2(outcome:
successful and unsuccessful) between-subject factorial design matrix was used to test our
hypotheses: the impacts of attribution with distinct participation level, self-efficacy and
outcome. In this study, two levels of participation (high and low) were matched with two
levels of self-efficacy (high and low). Two conditions of outcome (successful and
unsuccessful) were viewed as the moderating variables that influenced the attribution. In
18
total, there are eight scenarios in this study. Thirty participants were asked to participate in
each scenario independently, which meant the sample size would be 240 (30x8=240).
Participants were exposed to a written scenario describing a service in an English
cram school. Half of them have service failure and others have not, such scenarios are
attached as Appendix I. Participants were told that it was a study about consumer behavior
and were given a questionnaire. The questionnaire contained three major parts. The first
part included three examples in English and listed questions about (SE) self-efficacy, such
examples were attached as Appendix II. In second part, participants were asked to read the
scenarios carefully and imagine being in the scenarios. And it listed some questions about
(AtF) attributions to firm and (AtS) attribution to self. The third part contained questions
about (LoC) locus of control and demographic information, and such questionnaire is
attached as Appendix III.
3.4 Measurement
3.4.1 Participation
There are two participation levels in this study: high and low. The definition of participation level in this study is “the tendency of attendance and the effort of learning “.
With higher level of participation, participants always attended classes and made preview,
review and interaction with teacher. As with low level of participation, participants always
19
be absented and did no preview, review or interaction with teacher. There were two items
constructed in this study. They are, “I have high level of participation for the class.” and
“I didn’t have the sense of participation.” All of the items are verified on 7 points scales
represented from “extremely agree” to “extremely disagree.”
3.4.2 Self-efficacy
There are two self-efficacy levels in this study: high and low. The definition of
self-efficacy level in this study is “the confidence of English test and comprehension of class”. With higher level of self-efficacy, participants would have confidence in passing
the English test and believe that they can learn well in the class. As with lower level of
self-efficacy, participants did neither have confidence on passing the English test nor have
ability to learn. Participants would be asked to do all the questions and recorded the
number of correct answers. According to Mallin & Mayo (2006), there were some items
available for this study. They can be adapted in two parts. One of them measuring the confidence of passing the test, and the items were “I am confident to pass the English
test”, “I can answer most of the questions with the same difficulty as the examples” and
“ I believed I can get good grades in this English test.” Another category measured the
comprehension of class, and the items were “I am able to learn English”, “I can
comprehend the materials in class” and “I am more efficient than others in the class.” All
of the items are 7-points Likert-scale.
20
3.4.3 Attribution to Firm
In research by Maxham III and Netemeyer (2002), the level of attribution to firm
could be measured in three items: They were “Overall, the cram school should take that responsibility”, “the cram school make that outcome” and “For the outcome, I would
blame/praise the cram school.” All of the items are verified on 7-points Likert-scale.
3.4.4 Attribution to Self
In Coe et al (1982), attribution to self can be measured in three items. All of the
items were verified on 7-points Likert-scale. They were “ For the outcome, that was my fault/credit”, “I should take responsibility” and “I would blame/praise myself.”
3.5 Data Collection
The data was gathered from 80 participants in the pilot study and 240 participants in
the main study. Data were collected via two major channels. First, the questionnaires were
delivered to students in the classes at National Chiao Tung University (NCTU) and
National Cheng Chi University (NCCU). The second channel was using the Internet to
distribute and collect questionnaires online. Eight questionnaires with distinct scenarios
were mixed and were given randomly to subjects. All participants were told about the
purposes of this study, and were asked to complete the questionnaires carefully.
21
3.6 Manipulation Check Items
The reliability for the scenario descriptions can be measured by two-item with
Likert-scale. They are “The story reflects what might happen in the real world” and “I had
no difficulty imagining myself in the situation.” The mean is 5.19 (above 4) meaning that
the scenario is realistic. And one manipulation check was conducted to test whether the
participation was identified a high or low. Another manipulation check is conducted to test
whether the self-efficacy can be grouped into high or low. The scale items are mentioned
in 3.4.1 and 3.4.2 above. In addition, the results of the manipulations are reported in
chapter four.
3.7 Pilot Study
A pilot study was conducted to test the reliability of the questionnaire. It could
discover the limitations and misunderstandings in the experiment. Then, the limitations or
misunderstandings could be modified in the main study. After two failed and subsequent
modifications, the third trial of the pilot study was successful.
The pretest was made through giving 80 participants the experimental questionnaires,
and telling them the research was about consumer behavior. There were 41 male and 39
female participants. Forty-one of the 80 participants were students.
22
The reliability of scale for attribute to firm is 0.829 (Table 1) and the reliability of scale for attribute to individual is 0.901 (Table 1). Both were higher than 0.7, and hence
there is a significant difference between high and low participation group (p<0.00). The
difference between the self-efficacy groups is significant as well (p<0.00).
Table 1 Reliability Statistics of Attribution to firm and Attribution to self
Construct Cronbach's Alpha N of Items
Attribution to firm .829 3
Attribution to self .901 3
23