The second Taiwanese participant also had a below average knowledge about cars but was surprisingly familiar with VW as a brand (compound SK scale was 3.17). She showed an
products made in Germany possible (5.00). She was neither affiliated with VW nor did she believe that the scandal was product-related. She considered the brand as a stereotypical brand for Germany. The brand VW reminded her of a multinational company, sophisticated engineering, the hippie era, their slogan “Das Auto” and her time as an exchange student in Germany. She also used the adjective eco-friendly to describe VW. She was very well-informed about the scandal and the means VW used to deceive environmental agencies.
Her attitude towards the brand didn’t change significantly after the scandal which she partly attributed to her time in Germany where she gathered some knowledge about the German car industry and about what differentiates German cars from US or Japanese ones. Furthermore, she thinks that it would be wrong to blame the whole company for a crime that was committed by some high-level executives. In her opinion, this will definitely have some negative implications for the brands in Taiwan, but people who appreciate high production standards, high quality and well-designed products would still stick to VW. Her statement indicated a tendency to bolster her attitude towards VW. By stressing some of VW’s qualities such as superior manufacturing or well-designed products, she uses some of the brand most positive characteristics to boost her own attitudes.
Furthermore, she also admits that she is not an avid reader of Taiwanese news since “most of them seem to be too sensational for my taste”. However, she prefers to read critical articles, even if her standpoint is very different from the one of the author.
Summarizing the results from the interviews, there are four important results. Whereas the first two confirmed the authors assumptions, finding three and four unveiled interactions between resistance strategies and existing / new variables potentially useful for the successing quantitative analysis.
1.! Participants, regardless of sex or nationality, maintained a rather favorable attitude towards VW after the scandal
2.! CA was the most commonly-used strategy to resist attitude change and was not influence by the participant’s SK or their attitude towards products made in Germany 3.! Social-validation was used by most participants including those with high-knowledge
about the industry and the scandal itself
4.! The two German participants with the highest knowledge on both VW and their scandal were suspected to yield high levels of social vigilantism. Forming counter arguments was significantly influence by a feeling of superior knowledge on the technical specifications of the scandal.
3.3! Quantitative Analysis
The main purpose of the survey is the application of the findings obtained from interviews. Due to the larger scale, assumptions based on prior research and on the results of conducted interviews, the hypotheses can be validated or falsified. Interactions are initially tested in an interview setting and results are interpreted to improve hypotheses or form new assumptions about the interdependencies of the variables observed.
3.3.1! Design
Considering the hypotheses formed from the literature review in this paper on brand transgressions, prototype brands, country images and resistance to persuasion, the author of this study takes the view that knowledge and country attitude largely interact with the application of certain strategies. As stated by Verlegh and Steenkamp (1999) country image can be seen as an external cue moderating the consumer’s product perceptions and product evaluations.
Therefore, the main design purpose of the questionnaire was the accurate measurement of SK
and country image to measure the impact of these variables on the likelihood of using a certain resistance to persuasion strategy.
3.3.2! Variables
Originally, the author of this study planned to use SK and attitudes towards “Made in Germany”
as independent variables to measure the dependent variable, level of attitude changes towards VW. After the review of the literature on brand transgressions and resistance to persuasion, a scale measuring controlling for any ethnocentric effect was added to the design. Furthermore, the analysis of the interviews revealed that social vigilantism might have a significant impact on the likelihood of employing counter-argumentative resistance strategies. Using the variable attitude change as a dependent variable was discarded by the author due to a strong possibility of contamination.
The questionnaire is measuring SK and attitude towards products “Made in Germany” as independent variables and is control for effects fueled by EC or social vigilantism. The dependent variable will take into account the most significant factors of the overall resistance scale serving as a close substitute approximation for attitude change.
3.3.3! Participants
Out of 260 participants 173 completed the questionnaire attesting the survey has a reasonable sample size. The composition of the sample reflects the population surprisingly well. There is an almost equal number of male and female participants (cf. Tab. 4), and with 45,78 %, almost half of the participants were Taiwanese (cf. Tab. 5). Approximately 40 % of all participants were between 25 and 34 years old, followed by participants between 18 – 24 years. Other age groups were less common and accounted for roughly 40 % of the sample (cf. Tab. 6).
Most of the participants were either employees (38.92 %) or students (34.13 %), a significant smaller number was self-employed (14.37 %) or employed on the executive level (7.78 %).
Two participants were unemployed and 3.59 % of participants were retired (cf. Tab. 7). 90 % of all participants have an academic education. One third of all participants is holding a graduate degree, roughly one third has an undergraduate degree educational level, followed by 23 % holding a high school diploma (cf. Tab. 8).
On average, it took one participant 12:30 minutes to complete the survey. The median for questionnaire processing time was with 8:13 minutes significantly shorter, since only 173 out of 260 participants completed the whole questionnaire. 78 % of participants provided a valid email address and were eligible to take part in the lottery. 46 % of participants answered in German language, 44 % in Chinese, and 11 % answered in English language. This number is surprising in so far that the author of the study specifically stated that only Taiwanese and German citizens are eligible to take part in the survey. A possible explanation for the unexpected large number of English language preference over German or Chinese can be seen in a higher than average amount of bilingual participants or a higher than average command of English language among the participants.
3.3.4! Procedure and Measurement
The survey was launched as an online questionnaire on the platform unipark.de which is the academic application of questback, the world’s leading online survey software. The main advantages of this software is the large amount of statistical online tools as well as the possibility to create questionnaires in multiple languages. The survey was available online for participation for eight days. The URL linking participants to the questionnaire was distributed via the social media platforms Facebook, Twitter as well as business platforms such as LinkedIn and the German business network Xing. Due to a large number of items and an average response time of more than 10 minutes, participation was incentivized by a lottery draw. Each Taiwanese participant who finished the questionnaire was eligible to be part of a raffle with the chance of
winning a monetary reward of NTD 2000. In the same manner, German participants who completed the questionnaire and were eligible to a 50 Euro price draw.
The first set of questions were designed to measure the participant’s SK about VW and automobiles in general. This part was positioned in the very beginning of the questionnaire.
This way, the positioning of the SK scale should attract participants. Scale items were uncritical, required low cognitive effort and covered a highly familiar. Moreover, the first set of questions gave participants a sense for the direction of the questionnaire’s topic in case they missed to read the description provided. SK was measured by using a combination of the brand familiarity scale by Diamantopoulos, Smith and Grime and Flynn (2005) and Goldsmith’s subjective brand knowledge scale (1999). The overall scale consists of 13 items and should be capable of measuring each participant’s SK level about the transgressing brand VW and its products. Each item is measured on a 5-point Likert scale consisting of negative and positive statements to avoid answer tendencies. The mean value of all 13 items will be used for further statistical analysis.
The second scale in the questionnaire was measuring the attitude participants had towards Germany as a CoO. The variable country image was measured by a slightly modified version of of Roth and Romeo’s country image scale (1992). The scale is examining product characteristics and conceptualizes country image as summary construct5. Roth and Romeo use a two-dimensional approach focusing on the product and product-marketing aspects of a country. As stated in chapter 2.2.2, VW can be seen as a prototype brand for Germany.
Therefore, a product dimension scale is likelier to measure the construct country image accurately than a country characteristics scale. Eventually, it’s easier for participants to rate
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
5 The usage of a summary construct as a measurement of country image is appropriate for this study, since participants are expected to have a prior product knowledge about Germany as a CoO whereas a country characteristics concept would only make sense when consumers have little knowledge about the products of a certain country (Lala et al., 2009). For instance, Taiwanese consumers might not be familiar with products manufactured in Austria but are perfectly familiar with certain products manufactured in Germany such as BMW, Mercedes Benz or Volkswagen. Furthermore, there is disagreement among researchers about the amount and type of dimensions.
product features such as attractiveness or quality since these items are perceived as less abstract in this study than country characteristics such as political structure or working conditions.
Therefore, the employed scale consists of four 5-point sematic-differential items which are significantly lowering the cognitive requirements of participants compared to a 7-dimensional scale with 21 items used by Lala et al. (2009).
The participant’s level of EC was measured by an adaption of the scale used by Siamagka and Balabanis (2015). To achieve a feasible measurement of EC for both Taiwanese and German participants, words indicating a certain nationality were replaced by neutral words such as
“domestic”, “my country” or “home country”. In addition, the original scale was shortened from 17 to 10 items with one dimension being represented by two items instead of three or more to not overstretch the patience of the participants. Every item of the EC scale was measured by a 5-point Likert scale.
A short paragraph followed summarizing the very basic characteristics of the VW scandal along with a 5 point-scaled statement designed to evaluate the participant’s knowledge about the scandal. Participants with no knowledge about the scandal were automatically redirected to the final page of the questionnaire6. Prior attitudes towards VW before its scandal were measured afterwards, followed by a subjective evaluation of the individual’s perceived attitude change7. After that, participants were asked to state their email address to enroll for the prize draw and were eventually asked about their current attitude towards VW. All items were measured using
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
6 Participants who haven’t heard about the VW scandal would conatminate the sample and distort the change in attitude after VW’s trangression.
7 To ensure that the assumption of relatively stable attitude after brand transgressions is correct, attitudes and potential attitude changes towards VW are measured in the questionnaire in two different ways. Two items are measuring the participant’s attitudes towards VW before and after the scandal. An additional question aims at the subjective attitude change perceived by the participants. However, attitude will not be part of the actual analysis and is not considered as a dependent variable by the author since the variable is prone to contamination.
a 5-point Likert scale.
To control for a potential impact of social vigilantism on certain resistance strategies, a 5-point-scaled version of Saucier and Webster’s Social Vigilantism Scale (cf. 2010, p. 22 f.) was employed to compare then mean values between Taiwanese and German participants and to compute the overall impact of social vigilantism on self-assertions and high-cognition strategies.
Since Magnusson’s et al. (2015) latest study examining the extent of brand transgressions on country images and related brands couldn’t find any effect between the transgression extent and attitude formation, the extent of brand transgression (wide / deep) is regarded as covariate and won’t be part of the analysis.
Resistance strategies are measured by a 28-item scale which is based on Zuwerink Jacks and Cameron’s 14-item resistance to persuasion scale (2003, p. 151, cf. Tab. 2). The wording of each item was adjusted to the VW scandal making every statement a reasonably-sounding way of reacting to a counter-attitudinal message about VW. For each resistance category, two items were added to increase the validity of the scale. Every item added was designed closely to the original items and some of the additional items were already employed in the qualitative part of the study. The only differentiation criteria for the 4 items in every resistance strategy category was the source of the counter-attitudinal information combined with a slightly different sounding wording. For instance, resistance by CA was measured by participant’s evaluation of the likelihood of forming arguments after watching a TV report, reading a newspaper or having a conversation with a friend about the latest VW scandal. Besides being the most vital part of the questionnaire, 28 similar-worded statements demanded a lot from remaining participants and increased the dropout rate significantly.
4! Results and Discussion
In this section, the quality of the data obtained will be evaluated and the results of the study are presented. The first part of chapter 4 deals with descriptive statistics which are necessary to identify answer tendencies, dropout rates and mean values of the scales employed. The second paragraph is devoted to inform the reader about the underlying procedures used to edit the data obtained for further statistical testing. Scales adopted from reference studies are expected to yield valid results and will only be tested for reliability whereas the resistance scale’s validity and reliability is tested extensively. The third part contains the verification / falsification of the hypotheses employed along with the interpretation of the results obtained.
4.1! Analysis of Descriptive Statistics
Table 12 shows the results of the first set of questions about the SK of participants about cars.
The first scale was completed by 192 participants and the mean values range from 2.45 to 4.11.
There is a significant gender difference for SK as well as for brand familiarity. As expected, male participants have a higher knowledge about cars and are more familiar with the brand Volkswagen than their female counterparts. The mean difference for the compound knowledge scale (SK + brand familiarity) is .69176 and significant on the .000 level. Levene’s test for equal variances is with p = .316 not significant indicating that equal variances can be assumed.
Therefore, the gender of a participants might already be a strong predictor for the likelihood of resisting attitude change in this particular VW case. However, the actual impact of knowledge on resistance to persuasion will be analyzed in chapter 4.3.3. Not a single participant has a direct affiliation with VW. One participant used to work for VW and four participants used to work for a VW subsidiary. A potential contamination of results due to an (formerly) active professional relationship with VW Internet survey is highly unlikely. More than 85 percent of
The table of the SK scale (cf. Tab. 12) shows that most items circle around the mean value of the scale. However, arithmetic means of the scale items range from 2.45 to 4.11. Both items, ”I have heard of the most car brands that are around” and “Among my circle of friends, I’m one of the experts on cars” yield rather extreme values and might decrease the validity of the scale.
Therefore, it is important to run Cronbach’s alpha to guarantee a high internal consistency. The overall arithmetic mean value of the SK scale is 3.11. The brand familiarity scale has a similar arithmetic mean value of 3.06. Values range from 2.36 to 3.54 with the item “unfamiliar – familiar” yielding the lowest average value and “novice buyer – expert buyer” representing the upper end of the range. Consequently, a successively conducted Cronbach’s alpha is compulsory to out rule any distortion in the reliability of the scale.
The total arithmetic mean value for the participant’s attitudes towards products “Made in Germany” is 3.01. The scale has an agreeable range from 2.83 to 3.17 (cf. Tab. 17) indicating a good reliability. However, the author expected the individual perception of “Made in Germany” to be significantly higher. These findings are either accounted for by a wrong assumption overestimating the reputation of German products or a contamination of the scale by the VW transgression. Further interpretations of the impact of the independent variable made in Germany on attitudes can only be made with reservations. The EC scale yields a total arithmetic mean value of 3.17. The mean value range of individual items is still reasonable (2.83 – 3.51).
For the resistance scale, it is important to consider the results of the sub-scales measuring the different resistance strategies. Table 1 shows the distribution of range and mean values for the different resistance strategies. Every sub-scale is measuring one specific strategy and consists of 4 items. All items measuring SD yield similar mean values (cf. Tab. 20) and have an agreeable range. With an overall mean value of 2.74, participant’s SD scores are notably lower then their scores for AoC (cf. Tab. 19). The sub-scale measuring NA yields even lower total mean values of 2.12 indicating that participants in general are less likely to show negative
feelings when resisting persuasion (cf. Tab. 21). Similar results were obtained for the scale measuring content AV (cf. Tab. 22) with a total average mean value of 2.17. All items for CA yield similar mean values (cf. Tab. 23) and the range of its mean values is only 0.16. The sub-scale for SV has a total mean value of 2.86 and a rather problematic range of 0.63 which might indicate a rather poor reliability or validity of this construct (cf. Tab. 24). AB has an high average mean value of 3.33 and relative high range of 0.51 (cf. Tab. 25).
Table 1: Overview Resistance Sub-Scales (Item Range and Mean Values)
lowest average value highest average value range total arithmetic mean
AoC 3.24 3.73 0.49 3.46
SD 2.63 2.90 0.27 2.74
NA 1.94 2.42 0.52 2.12
AV 2.03 2.34 0.31 2.17
CA 3.17 3.33 0.16 3.25
SV 2.52 3.15 0.63 2.86
AB 3.12 3.63 0.51 3.33
All results considered, it becomes clear that there is a strong tendency for individuals to make use of AoC, CA or AB rather than low-cognition strategies such as AV, SV or SD. The lowest results were obtained for negative feelings, a finding which was already observed by Zuwerink-Jacks and Cameron (2003).
For social vigilantism (cf. Tab. 26), the overall scale shows a mean value of 3.07 and mean values ranging from 2.10 to 3.82, making a dropout of extreme items probable.
4.2! Data Prerequisites and Data Preparation
With a total of 83 variables, the questionnaire collected a tremendous amount of data that needs to be prepared for further analyses. For all scales employed in the survey, negative worded scale items have been recoded and compound variables have been computed to express the value of a scale with one single variable. This procedure was done for items of the scales SK about cars, brand familiarity, attitude towards “Made in Germany”, EC and social vigilantism. Items of both scales SK about cars and brand familiarity were used to form a compound SK scale. The
With a total of 83 variables, the questionnaire collected a tremendous amount of data that needs to be prepared for further analyses. For all scales employed in the survey, negative worded scale items have been recoded and compound variables have been computed to express the value of a scale with one single variable. This procedure was done for items of the scales SK about cars, brand familiarity, attitude towards “Made in Germany”, EC and social vigilantism. Items of both scales SK about cars and brand familiarity were used to form a compound SK scale. The