‘What do you mean by that?’: a research towards the usage of linguistic language and the effect on perceived trustworthiness in OCRs.
By Robin Damen
S2003147
(this page was left blank intentionally).
‘What do you mean by that?’: a research towards the usage of linguistic language and the effect on perceived trustworthiness in OCRs.
By:
Robin Damen 18-06-2018
MSc Marketing, Management Master Thesis
University of Groningen Faculty of Economics and Business
Department of Marketing 9700 AV, Groningen
Supervisors:
First: Dr. J. A. Voerman Second: Dr. J. Berger
Student:
Robin Damen
Address: Admiraal de Ruyterlaan 26G 9726GT Groningen
Phone: +31650292211 Email: R.damen@student.rug.nl
Student number: S2003147
Acknowledgements
I would like to express my sincere gratitude to my supervisor, Dr. J. A. Voerman, for her
continuous and helpful support, motivation and enthusiasm. Next, I also want to thank Dr. J. Berger as the second reader of this thesis.
I want to thank all the respondents for their help in this research by participating in the survey.
Without them it would not have been possible to conduct this project.
I also want to take every group member for the support and advices during this project.
Finally, I want to thank my mother, Wilma ten Have, my father Eric Damen and my girlfriend Maaike Haan for supporting me during this project. Without their continuous support and believe this accomplishment would not have been possible.
Thank you.
Executive Summary
Through the increased online word-of-mouth communication many consumers use the internet to read online consumer reviews (OCRs). These OCRs are written by consumers for other consumers and are used to search information on products and services before making a purchase. However, sometimes it happens that readers of the OCRs cannot fully interpret the message conveyed by the writer and therefore this can affect the perceived trustworthiness of the OCR.
An important reason for this is the language that is used to write the OCR, which has received little attention in existing literature (Kronrod and Danziger, 2013). The focus in this research lies on the distinction between figurative and literal language since better understanding of this distinction will help to give insights to which kind of wording a consumer will respond better to when reading an OCR (Wu et al., 2017). Next, OCRs can influence the probability of sales and therefore it is important to gain insights in the use of language to convey a message in the most effective manner (Eaton, 2005). Furthermore, in recent studies have been argued that figurative language is used more when an OCR is written on a hedonic product and literal language when the OCR is written on a utilitarian product (Kronrod and Danziger, 2013). Therefore, it is empiricallly tested if this also affect the perceived trustworthiness of an OCR. Another factor in this research that can affect the trustworthiness through the use of language is the valence of the review, which can be either
positive or negative Guo and Zhou (2016). Negative information is tended to be considered as more trustworthy (Sen and Lerman, 2007). It is therefore empirically tested in this study which valence of an OCR is perceived as more trustworthy and if language has an impact on this. Furthermore, the covariates skepticism to OCRs, preference for language and sensitivity to negativity have been hypothesised to have an impact on perceived trustworthiness of an OCR.
Use of eWOM has increased in recent years. It has become more of a key factor in consumer decision making (Cheung & Lee, 2012). Therefore, the purpose of this study is to investigate what the effect of type of linguistic language, review valence and type of product is on perceived trustworthiness of OCRs. In addition, this study investigates if skepticism to OCRs, sensitivity to negativity and preference for figurative/literal language influence this effect.
To empirically test the hypotheses, a 2 x 2 x 2 between-subject experimental design is conducted
through the use of a survey. This survey was distributed online and participants were assigned to
one of the 8 possible scenarios that differed on at least one factor of type of language (figurative or
literal), review valance (positive or negative) or product category (hedonic or utilitarian). This study
was filled in by 196 respondents randomly chosen through the internet. After seeing the scenario, the respondent answered a manipulation check question on how literal or figurative the respondent perceived the text.
The analysis of the hypotheses started with two manipulation checks. The first was to check if the respondents really perceived the literal text as literal and the figurative as figurative. A significant difference was found. There was also a manipulation check for the review valence, to see if a respondent perceived a text as positive or negative. Again, this was confirmed by the results. The use of hedonic products and utilitarian products in this study were based on a study from Kronrod and Danziger (2013), who found evidence that these specific products were perceived by consumers as either hedonic or utilitarian.
After the manipulation checks an ANOVA analysis was conducted, which resulted in a significant effect of language, product category and interaction between product category and review valence on perceived trustworthiness. The regression analysis also showed significant effects on these variables through all the 5 models. In addition, the regression analysis also showed a significant negative effect of sensitivity to negativity on perceived trustworthiness of an OCR. This implies that a higher sensitivity to negativity leads to a lower perceieved trustworthiness of an OCR.
Findings in this study suggest that figurative language has a positive impact on perceived
trustworthiness of an OCR. Next, results indicate that OCRs on utilitarian products negatively affect perceived trustworthiness of OC s. Furthermore, results indicate that a negative OCRs written on utilitarian products positively increase perception of trustworthiness of OCRs and that sensitivity to negativity negatively affects trustworthiness of OCRs. The other variables and interactions did not show significant effects.
These findings are an important contribution to previous studies: it has found that these results are both supportive and contradicting with existing literature. It seems that there is a lot of disunity in existing literature in this area of research. This study therefore creates awareness on the fact that there is still little known in the field of eWOM, especially what influences perceived
trustworthiness of OCRs, and therefore can be used for further research in this area. Managers can use this study to use the right kind of OCRs on their websites, in order to achieve positive perceived trustworthiness of the OCR, in order to boost the sales (Eaton, 2005).
Furthermore, this study argues to have several limitations. This study is limited by a small
population, with people belonging to the same age group. In addition, mood and literacy may have impacted the results.
To summarize, this study provides a good basis for further investigation of perceived
trustworthiness in OCRs in the field of eWOM.
Table of contents
1.Introduction………..………..……….9
1.1 Perceived trustworthiness………..………..….…10
1.2 Language in OCRs……….………..10
1.3 Differences in conversational norms between advertisements and user-generated content………...……….………....…11
1.4 Focus on OCRs written by consumers……….…………12
1.5 Product categories……….…...12
1.6 Review valence in OCRS……….…12
1.7 Implications for companies………..13
2. Literature review……….…14
2.1 Linguistic language of OCRs……….…..14
2.1.1 Figurative language……….……..14
2.1.2 Literal language……….15
2.1.3 Differences between figurative and literal language……….…15
2.2 Review valence……….………...16
2.2.1 Negative valence……….…..16
2.2.2 Positive valence..……….…..17
2.3 Product category………...18
2.3.1 Negative valence……….…..19
2.3.2 Positive valence……….………19
2.4 Covariate characteristics………...19
2.4.1 Skepticism towards OCRs….………19
2.4.2 Sensitivity of consumers on negative information………20
2.4.3 Preference for Literal or Figurative language by consumers………21
2.5 Conceptual model……….….……...22
3. Research design………23
3.1 Type of research and method……….…..23
3.2 Set up of experimental design……….….24
3.3 Operationalization of the research………..……..26
3.3.1 Dependent variable………..…..28
3.3.2 Manipulation check questions………...28
3.3.3 Covariates………..29
3.3.4 Control variable ……….………...31
3.3.5 Factor validity of the scales……….………..32
3.4 Population and sample design……….……….32
3.5 Manipulation check results………...32
3.5.1 Linguistic language…….……….……….33
3.5.2 Review valence……….……….33
3.6 Plan of analysis……….………33
3.6.1 ANOVA analysis……….…….……….35
3.6.2 Regression analysis estimations………36
3.6.3 Multicollineairty………36
4. Results……….………...36
4.1 ANOVA analysis results………..40
4.2 Lineair regression results……….43
4.3 Implications for the hypothesizes………43
5. Conclusion and Recommendation…….……….44
5.1 Conclusion…….……….……….46
5.2 Limitations and future research…….………..46
5.3 Recommendation…….……….………...47
5.3.1 Managerial implications…….………..43
5.3.2 Academic relevance……….47
6. References…….……….………..48
6.1 Articles…….……….………..48
6.2 Books…….………..55
Appendix A – Pre-test survey on linguistic language…….………..57
Appendix B – Pre-test review valence…….……….63
Appendix C – Survey…….……….………..65
Appendix D – Means per scenario…….………...91
Appendix E – Descriptives of pre-test on linguistic language………..92
Appendix F – T-test paired samples on pre-test linguistic language………93
Appendix G – Factor analysis………..…….………94
Appendix H – T-test of two indepedent samples manipulation check on type of language……….99
Appendix I – T-test of paired samples on review valence…….………..100
Appendix J – ANOVA analysis…….……….……….101
Appendix K – Linear regression models 1-5…….………..103
1.0 Introduction
In recent years, the Internet has increased word of mouth (WOM) communication dramatically, particularly in the form of online consumer reviews (OCRs) on e-commerce websites (Guo & Zhou, 2016). Many consumers search the Internet before they make purchases (Smith et al. 2005), not only searching for product information provided by the producer or vendor, but also for reviews written by other consumers, so-called OCRs (Utz et al., 2012). OCRs are regarded as a form of user generated content since it is made by consumers for consumers, and as a type of electronic word-of- mouth (eWOM) through which users of the Internet can informally and non-commercially interact with each other, exchanging positive and negative consumer experiences, by usually focusing on the quality of the product (Utz et al., 2012; Hu et al., 2008).
Companies have incorporated user-generated content like OCRs, on their e-commerce websites, to use the advantages offered by this incorporation to create trust around a product. Early literature have identified more than 80 variables as factors that can influence the intention to purchase a certain product, and through the decision to purchase (Hazari et al., 2016). However, when looking at user-generated content it has been found that this contributes to trust, which is seen as a key characteristic for shopping (Hazari et al., 2016). Cheung et al. (2009) state that OCRs with high perceived trustworthiness have a positive impact on the degree to which users adopt information.
However, there is a difference between evaluating trustworthiness of offline and online reviews since online reviews can be written by a wide array of strangers worldwide. For example, in the context of online consumer reviews, consumers do not have complete information about reviewers’
actions, thoughts, or motives (Racherla & Friske, 2012). Therefore, it is important to do further research towards what makes an online review seem trustworthy to consumers (Cheung et al., 2012).
In a research towards eWOM adoption, Fang (2014) suggest that perceived credibility of eWOM
belongs to one of the most crucial drivers of eWOM adoption. In the same research, Fang (2014)
composed a framework in which perceived credibility of eWOM could be influenced by source
expertise, task attraction, argument strength and recommendation rating. Fang (2014) argues that
among these four factors, argument strength has the strongest positive influence on perceived
credibility of eWOM reviews. In Fang (2014) argument strength concentrates on the persuasive
strength of received eWOM reviews. Cheung et al. (2009) have identified the direct impact of
argument strength on receivers’ attitudes, specifically in online environments. An important way of
sharing information is through the use of language and language style (Kronrod & Danziger, 2013).
Next to the possible effect of usage of language within an OCR on perceived trustworthiness, existing literature has argued that review valence also has an effect on perceived trustworthiness of OCRs (Guo and Zhou, 2016), and is therefore also taken into account in this research as a factor that can influence perceived trustworthiness.
Furthermore, Kronrod and Danziger (2013) argue that online consumer reviews written on hedonic products contain more figurative language than online consumer reviews which are written on utilitarian products. Thus, it is expected that when the product is considered hedonic, consumers will use figurative language more to describe the product than when the product is utilitarian.
However, it is unknown if this difference in usage of language is because of the fact that consumers have a certain level in trust associated with a writing style of the review (Kronrod and Danziger, 2013) or because of an other reason. Although a great deal of research has demonstrated that consumer opinions shape behavior, there has been less work examining how such opinions are conveyed (Packard & Berger, 2017).
1.1 Perceived trustworthiness
Mayer et al. (1995) have defined trust as the ‘willingness of a party’ to be vulnerable to the actions of another party based on the expectation that the other will perform a particular action important to the trustier, irrespective of the ability to monitor or control that other party’. Trust and uncertainty form two ends of the same continuum in the sense that the higher the uncertainty, the lower the trust and vice versa (Racherla & Friske, 2012). Further, the notion of “willingness” in the above
definition reflects a trustor’s perception that the other party is competent, open, and reliable and that there is a complementarity of intention and expectation between the two parties (Racherla & Friske.
2012). The level of trustworthiness is related in Grewal & Baker (1994) to how individuals perceive and respond to the information provided to them. In this research this reflects to the use of language which consumers use to describe a certain phenomenon. Few studies have attempted to investigate how consumers assess credibility and trustworthiness in e-WOM (Cantallops & Salvi, 2014; Yoo &
Gretzel, 2009). Filieri (2016) tried to fill this gap by doing research towards causes for trustworthy vs. untrustworthy reviews and found different factors that are the cause of a certain level of
trustworthiness of a review.
1.2 Language in OCRs
Although user generated content has garnered much recent interest in the marketing and consumer
behaviour research community, the language used in the user-generated content has received little interest (Kronrod & Danziger, 2013). According to Packard & Berger (2017), prior research tended to consider all positive word of mouth as having the same impact but did not take into account the words used to express this positive word of mouth (Chen & Lurie, 2013; De Angelis et al. 2012). In the context of consumer social interactions, researchers have begun to examine the consequences of language-related phenomena such as abstract versus concrete word use (Schellekens et al., 2010), explained actions (Moore, 2015), figurative language (Kronrod & Danziger, 2013), emotional words (Berger & Milkman, 2012), linguistic mimicry (Moore & Mcferran, 2016), and boasting (Packard & Berger, 2017). However, if the focus lies on OCRs a direct comparison of figurative language vs. literal language would help understand to what kind of wording of the review a
consumer will respond better (Wu et al., 2017). In Kronrod & Danziger (2013), figurative language and literal language are concepts of the style of linguistic language used. Linguistic style refers to the way how the message is conveyed, so how the review is written by the reviewer (Ireland &
Pennebaker, 2010). Linguistic language is also the term used in this research as the term for the usage of language in OCRs, with the possibility of being either figurative or literal.
1.3 Differences in conversational norms between advertisements and user-generated content According to Kronrod & Danziger (2013) research towards the use of language has mostly been done in the context of advertisements. However, people generally trust consumer reviews more than advertising (Edwards, Li & Lee, 2002) Since the focus in this research will be on OCRs it is
important to distinguish the differences between research done in advertisements and user- generated content. The reason for this is that some differences between advertisements and user- generated content can cause a difference in trust (Edwards, Li & Lee, 2002). Advertising differs from user-generated content in several aspects related to conversational norms (Kronrod &
Danziger, 2013). First, advertising is considered to be generally positive, or elicit positive attitudes towards a certain product or phenomenon, while user-generated content contains positive as well as negative opinions (Sen & Lerman, 2007). Second, advertisements are being perceived as biased and persuasive attempts, while user-generated content is generally perceived as an objective sharing of opinions (Sen & Lerman, 2007). Third, advertisements consist of professionally preplanned
methods with the goal to achieve mass communication, while user-generated content is spontaneous
and natural; often generated by authors that distribute messages which are usually not censored
(Kronrod & Danziger, 2013). Because of these differences, Kronrod & Danziger (2013) suggest
readers to have a different set of conversational norms regarding advertising and user-generated
content. It is plausible that text which adheres to conversational norms influences not only product
attitudes but also attitudes towards the perceived trustworthiness of reviews (Kronrod & Danziger,
2013).
1.4 Focus on OCRs written by consumers
Within the context of online consumer reviews a distinction can be made between recommendations made by customers and recommendations made by experts (Huang & Chen, 2006). In this research the focus will be on the reviews written by consumers. The reason for this is that Huang & Chen (2006) contrasted the effects of expert recommendations and customer recommendations and found that customer recommendations had more influence on other consumers when reading OCRs (Huang & Chen, 2006). Recommendations by consumers are usually made by giving an overall rating of the product (e.g. 1-5 stars), but many consumers regard the added description of the concrete experiences with the product given in the text as more informative than a simple rating (Utz et al. 2012). In this research the focus will lie on the description of experiences with a product since this can be done in, among others, either figurative or literal language. Existing literature have linked usage of figurative and literal wording in writing OCRs on hedonic and utilitarian products (Babin et al., 1994; Kronrod & Danziger, 2013). Because of this link between language and products, the focus will lie on hedonic and utilitarian products in this research. A more detailed description of the link between language and products will be given in the section below.
1.5 Product categories
Babin et al. (1994) have argued that hedonic perceptions of products are related to emotional and aroused consumption and therefore can be linked towards figurative language (Kronrod &
Danziger, 2013). Research towards the emotional intensity and usage of figurative language in interpersonal communication have argued a positive correlation between the two (Bryant and Gibbs, 2002; Gibbs et al., 2002; Zemanova, 2007). Kronrod & Danziger (2013) conclude that people express themselves more in figurative language when talking about issues of an emotional character in an OCR. An example for this are OCRs on experiences with products. This is in
contrast to utilitarian aspects, which are considered rational and are therefore considered to be more typical to be conveyed in a literal manner in OCRs (Kronrod & Danziger, 2013).
1.6 Review valence in OCRs
As with other communications, online reviews differ in their efficacy. A reason for this is given in Guo & Zhou (2016). They argue that review valence is a key characteristic of reviews. Review valence refers to the evaluation direction of an OCR, which can be positive or negative
(Purnawirawan et al., 2012). The valence of a review is typically shown through a numeric five-star
format assigned by the reviewer to the product, ranging from one star (very negative) to five stars
(very positive) (Yin et al., 2016). Average reviews of products are often displayed by prominent review sites and can therefore have a high impact on product sales (Chevalier & Mayzlin, 2006;
Duan et al., 2008). Furthermore, according to Hazari et al. (2016) review valence may affect consumer behavior since it also has the potential to impact the ability of consumers to form
opinions about products and to change attitudes towards products. This may also have an effect on the propensity to make purchase decisions (Hazari et al., 2016).
1.6 Implications for companies
According to Chen & Xie (2008) OCRs can serve as “sales assistant” to help consumers identify the products that best match their idiosyncratic needs. This dependence of consumers on OCR’s to make purchase decisions has benefitted online retailers in the form of increased consumer loyalty and increased product sales (Schindler & Bickart, 2012). Not only does the dependence of online user-generated content to make purchase decisions is increasing for consumers, it it also becoming vitally important for online retailers to understand exactly how consumers make use of such content. In special this counts for online retailers that boost product sales through OCR’s (Guo &
Zhou, 2016). This current study helps explain why some OCRs are perceived to be more
trustworthy than other OCRs. Online retailers can use this study to create strategies to encourage more useful OCRs that aid consumers in making decisions and judgments on products.
Because of the gaps in literature as mentioned above between linguistic language and perceived trustworthiness, the problem that this research tends to answer is:
What is the effect of linguistic language, review valence and type of product on the perceived trustworthiness of OCRs by consumers?
To help answer this main question, the following sub questions are formulated:
- What is the influence of type of language on the perceived trustworthiness of an OCR?
- What is the influence of type of product on the perceived trustworthiness of an OCR?
- What is the influence of type of valence on the perceived trustworthiness of an OCR?
- Which covariates have an influence on the perceived trustworthiness of an OCR?
In the next section there will be a review of used literature in this research and the hypothesis of this
research. First there will be further elaborated on linguistic language. After that the focus will lie on
the perceived trustworthiness experienced by consumers, followed by review valence and the two product categories used in this research.
2. Literature review
In this chapter, the dependent variable perceived trustworthiness, and the independent variables linguistic language, review valence and product category will be reviewed based on existing literature, as are the covariates skepticism towards OCRs, sensitivity to negativity and preference for figurative/literal language. The findings in existing literature on these subjects are used to help formulate the hypotheses. Based on this, a conceptual model is created and will be explained in the last paragraph of this chapter.
2.1 Linguistic language of OCRs
Linguistic language is considered by Niederhoffer & Pennebaker (2002) as the observable output of an individual’s innate mental processes. Similarity in linguistic style can facilitate the vocal users’
efficiency in review information processing and content comprehension (Pickering & Garrod, 2004). As mentioned in chapter 1, linguistic language can be divided in figurative and literal language (Kronrod & Danziger, 2013), which will be further elaborated on in the section below.
First a more detailed description on figurative language will be given followed by a detailed description of literal language. In the third section the differences between the two categories of language will be given and this section will end with a hypothesis on the main effect between the use of linguistic language on perceived trustworthiness of OCRs.
2.1.1 Figurative language
In this research figurative language is defined as ‘the use of words and expressions employing their
indirect meaning, to convey an additional connotation beyond that of their lexical sense (Fogelin,
1988). Figurative language is not uncommon or exclusively poetic (Roberts & Kreuz, 1994). It is a
ubiquitous part of spoken and written discourse (Pollio et al., 1990). Unfortunately, figurative
language is not always clear or precise through which the possibility rises that the message
conveyed by the sender is interpreted different than intended (Roberts & Kreuz, 1994). For
example, if one metaphorically states, “My aunt is an elephant.” one could be referring to girth,
length of nose, or fondness for peanuts (Roberts & Kreuz, 1994). In some languages, such as
English, the figurative idiomatic meaning of the expression climbing the wall is to be extremely
nervous or upset (Kronrod & Danziger, 2013). In other languages, such as Russian or Hebrew, the
expression has a different figurative meaning, which is to be very bored or have nothing to do
(Kronrod & Danziger, 2013). Additional instances of figurative language include metaphor (the
Ferrari of vacuum cleaners), word play (Don’t leave without a good buy), idiomatic expressions (My car’s a lemon), hyperbole (the service person was a cell phone professor), or imitating sounds (this teacher is wrrrufff, meaning a tough teacher, or ouch!, meaning I was offended)(Kronrod &
Danziger, 2013). If people who participate in a discourse cooperate by expressing themselves as clearly, concisely, and completely as possible, as Grice (1975, 1978) hypothesised, then potentially ambiguous figurative language must accomplish certain communicative goals better than literal language (Gerrig & Gibbs, 1988; Glucksberg, 1989; Kreuz, Long & Church, 1991). According to the research done by Kreuz & Roberts (1994) this means that the benefits of using figuration must outweigh the potential costs of being misunderstood. However, the question rises whether this also influence the trustworthiness of the online consumer review.
2.1.2 Literal language
As words and phrases can have different meanings in figurative language, this is not the case when the words and phrases are communicated in literal language. The literal meaning of words and phrases is their plain dictionary meaning (Kronrod & Danziger, 2013). For example, the literal meaning of climbing the wall, means the action of traveling up or down along a vertical surface (Kronrod & Danziger, 2013). In this example, the words exactly mean what they intend to describe, as literal language is defined by Fogelin (1988) who defines literal language as ‘words meaning exactly what it says to describe a certain phenomenon’.
However, because of all these different possibilities to express and interpret figurative language compared to literal language there are also differences between figurative and literal language, which are considered as key characteristics in this research because of the way of how these can be used to have the most effect when conveying a message towards other consumers of a specific product.
2.1.3 Differences between Figurative and Literal language
In recent literature, linguistic research suggests that the advantage of figurative language (vs. literal
language) is conditioned on the communication context. A key difference between figurative
language and literal language according to Gibbs (2008); Kronrod & Danziger (2013); Shen (2002)
is the context-defined norm of communication (Wu et al. (2017). Figurative language is perceived
as more emotional and effect-rich compared with literal language. Therefore, it is regarded as more
appropriate to communicate emotional experiences rather than regional and functional experiences
(Gibbs, 2008; Ireland & Pennebaker, 2010; Shen, 2002) in Wu et al. (2017). In the online review
context, where a majority of the posts are communicated from one individual to a large group of
unknown other customers, literal language sets the conversational norm and should be regarded as a more typical type of language than figurative language (Wu et al., 2017). Furthermore, Hoffman &
Kemper (1987) state that there is no difference in time between figurative and literal language to understand an expression in the text, when sufficient context is provided. According to Inhoff, Lima
& Carroll (1984) in Giora (1997) argue that contexts that are longer than 3 sentences were found equally easy to process when metaphoric and literal language interpretations were compared.
Kemper (1981) found that the longer the paragraph, the easier it was to interpret it figuratively.
This could imply that, in the context of online consumer reviews, the longer the review, the easier it is to interpret figurative language for consumers.
As mentioned before, figurative language is used more in order to describe reviews on hedonic products and linguistic language on reviews with utilitarian products (Kronrod & Danziger, 2013).
Based on researches towards advertising, McQuarrie & Mick (1999; 2003); McQuarrie & Philips (2005); Philips & McQuarrie (2009) have argued that messages using figurative language elicit more positive attitudes toward the advertisement and the product, compared with advertising that does not employ figurative language. Utz et al. (2012) did a research towards the positivity and negativity of online consumer reviews. Utz et al. (2012) had a hypothesis set in which they expected that the more positive a review was, the higher the perceived trustworthiness. From the results from the same research could be drawn that the perceived trustworthiness was indeed higher for positive reviews than for negative reviews. Because of the results drawn from the researches of Utz et al.
(2012); McQuarrie & Mick (1999; 2003); McQuarrie & Philips (2005); Philips & McQuarrie (2009) the first hypothesis is:
H1: Online consumer reviews consisting of figurative language will be perceived as more trustworthy than online consumer reviews that consist of literal language by consumers.
2.2 Review valence
As mentioned in chapter 1; the valence of the OCRs in this research is either positive or negative. In this section there will be further elaborated on both sides of the review valence.
2.2.1 Negative valence
A negative OCR is considered to have more value to the vocal users than a positive OCR (Sen &
Lerman, 2007; Yang & Mai, 2010). An important explanation for the greater impact of negative
OCRs compared to positive OCRs is that those are more likely to be attributed to product
performance while positive OCRs are more likely to be attributed to non-product causes.
A second important explanation in existing literature is given in Vaish et al. (2008) in a research on the impact of negative OCR’s. Vaish et al. (2008) stress the importance of risk alertness in the environment. This importance derives from the fact that humans are more alert to risks because of the fact that risks have been more critical to our survival. Negative OCR’s are therefore more impactful because they stress the importance of product attributes and reflect on risks associated with the use of the product itself.
2.2.2 Positive valence
In the context of positive reviews there can be argued that positive reviews elaborate on multiple aspects of a product. A positive review stresses for example not only the quality of the product, but also the design, the ease of use or the understandability of the instruction manual (Utz et al. 2012).
An important reason for this is given by Friestad & Wright (1994) and Kelley (1973) in Guo &
Zhou (2016). In the same article they suggest that consumers often make inferences about why product experiences are shared and use these inferences to judge the value of the information that has been provided by the sender, the reviewer. The implication rises that an OCR that is considered more attributed to non-product causes will be perceived as not credible (Lee & Youn, 2009; Sen &
Lerman, 2007).
Therefore, this research argues that negative OCR’s are to be found more credible by consumers compared to positive OCR’s because of the fact that OCR’s that are considered negative by consumers describe more product attributes (Sen & Lerman, 2007) and are found to be more trustworthy. From this, the following hypotheses rises:
H2: Negative valence in OCRs is considered to be more trustworthy than positive valence in OCR’s by consumers.
Next to this hypothesised effect there is also the effect of linguistic language on trustworthiness, with review valence as the moderating effect. Pennebaker & King (1999) state that the usage of language can be seen as the way people demonstrate individual differences in self-expression, providing insights in the way they perceive the world. Vocal uses of words can generate the
underlying motivation of the reason why the review is being generated (Pickering & Garrod, 2004).
When the review valence is negative, literal language will be used more compared to figurative
language, which will be used more in the context of a positive review (Sen & Lerman, 2007).
OCRs with a negative valence are perceived as more trustworthy then OCRs with a positive valence (Sen & Lerman, 2007). The reason for this is that OCRs with a negative valence tend to focus more on product attributes than positive OCRs. If this is reflected back on linguistic language than literature describes that literal language is used to describe product attributes in OCRs. Therefore, this research hypothesises:
H3: The positive effect of negative valence on perceived trustworthiness in OCRs is stronger for literal language than for figurative language
2.3 Product Category
As mentioned before, the product category will be distinguished in two types, hedonic and
utilitarian. A suggestion for this distinction has been made in Kronrod & Danziger (2013) where the two contexts are commonly associated with emotional (as in figurative language) and rational attitudes (as in literal language) (Alba & Williams, 2013; Chaudhuri & Ligas, 2006; Dhar &
Wertenbroch, 2000). Reasons for this are given in Botti & McGill (2011), who suggest that
utilitarian consumption is less affectively rich than hedonic consumption. In early literature, Babin et al. (1994) have found distinct characteristics of hedonic perceptions of products, which are related to emotional consumption, whereas the perception of utilitarian products are related to rational and task-related aspects, by focusing on product attributes (Sen & Lerman, 2007).
Examples of hedonic products are Toys, Wines, Jewellery, CDs & DVDs and Beauty & Cosmetics (Kushwaha & Shankar, 2013). Shopping for these products will, in general, elicit a positive mood (Chaudhuri & Holbrook, 2001). On the other hand, examples of utilitarian products are sports equipment, electronics, pet items, office supplies and musical instruments (Kushawa & Shankar, 2013). Sen & Lerman (2007) concluded that OCRs that focus on the description of product attributes are found to be more trustworthy. In a research done by Lee & Youn (2009) the
description of product attributes in OCRs have been linked to OCRs on utilitarian products. Since utilitarian products are acquired for their product attributes to complete certain tasks (Sen &
Lerman, 2007), this research hypothesizes the following effect:
H4: OCRs on Utilitarian products will be perceived as more trustworthy than OCRs on Hedonic products.
Furthermore, in existing literature, Kronrod, Grinstein & Wathieu (2012) computed a study towards the usage of assertive and non-assertive messages of promotion for hedonic and utilitarian
consumption. From this study the conclusion could be derived that more assertive messages can
cause greater compliance in the context of hedonic consumption. Also, in the same study was stated that hedonic consumption contexts are more likely to generate a positive mood. This implies that the consumers expect assertive language. This is interesting since in this research figurative language is expected to elicit a positive mood by the usage of it in the OCRs and therefore
suggesting that figurative language is more useful for hedonic consumption since this is found to be more positive (McQuarrie & Mick (1999; 2003); McQuarrie and Philips (2005); Philips &
McQuarrie (2009). Furthermore, Kronrod & Danziger (2013) hypothesize a strong tie between figurative language and hedonic consumption, and between literal language and utilitarian
consumption. In an experiment towards the choice between hedonic and utilitarian options, Kronrod
& Danziger (2013) have argued that in their study figurative language foster choice of hedonic products whereas reading literal language fosters choice of utilitarian products. Therefore, this research hypothesizes that:
H5a: The positive effect of usage of figurative language in OCR’s on hedonic products will be perceived more trustworthy than the usage of literal language in OCR’s on hedonic products.
H5b: The positive effect of usage of literal language will in OCR’s on utilitarian products will be perceived more trustworthy than the usage of figurative language in OCR’s on utilitarian products.
As mentioned before, consumers tend to trust OCRs with a negative valence more than consumers with a positive valence (Sen & Lerman, 2007). The reason for this is that negative OCRs tend to focus more on product attributes than positive OCRs. If this is reflected back on a specific product category than literature describes that product attributes are more important in the search for utilitarian product than hedonic products. There this research hypothesizes:
H6: The positive effect of negative valence on perceived trustworthiness is stronger for utilitarian OCRs.
2.4 Covariate characteristics
In this research there are three different covariates that can influence the perceived trustworthiness of OCRs. In this section the variables skepticism towards OCRs, sensitivity towards negativity and preference for figurative/literal language will be described and hypothesized since existing literature argues an effect on perceived trustworthiness caused by these covariates.
2.4.1 Skepticism towards OCRs
Previous research of the skepticism towards OCRs have concluded that consumer reviews are seen as less biased than advertisements (Herr et al., 1991; Hoch & Young-Won, 1986). Online shoppers trust OCRs even more than store reputation and assurance seals (Utz el al., 2012). Senecal & Nantel (2004) have computed a research towards the influence of online product recommendations on consumers’ online choices. They argue that there are three determinants that can influence the impact of recommendations on consumers’ online product choices. The first determinant is the nature of the website on which the recommendation is proposed. Alba et al. (1997) and Bakos (1997) argue that more independent websites such as non-commercial linked third-party websites are preferred by consumers when searching for OCRs. These websites provide more objective information and alternative products or services to choose from and therefore will be perceived as more useful by consumers (Alba et al. 1997). Senegal and Nantel (2002) conclude that consumers would follow product recommendations in a greater proportion when the OCR was read on a more independent website. Next, in the same research was also concluded that a recommender system is more influential than a recommendation source, like other consumers providing the information.
The reason for this influence of websites and sources is according to Kelman (1961) based on trustworthiness. In this article trustworthiness is linked to credibility and expertise. It has been found that trustworthiness and source expertise are positively correlated towards attitudes towards a brand or product and the behavioral intentions of consumers (Gilly et al., 1998).
Based on this, consumers would show more trust in OCRs written on more independent websites and based on recommendation systems. However, in this research websites are not a factor on which the relations between the constructs are based. As for the recommendation systems, the reviews are written by consumers and are not based on such a system. Therefore, consumers can show less trust in the OCR at the start of reading it since the recommendation system is missing in this research.
Based on this and the factors used in this research the following hypothesis rises:
H7: The higher the skepticism towards OCRs the lower the perceived trustworthiness of OCRs.
2.4.2 Sensitivity of consumers on negative information
According to Sen & Lerman (2007) research in other areas of consumer behavior has concluded
that negative information, distributed by consumers towards other consumers, has more value to the
consumer that receives the negative information than when one receives positive information. This
counts for both decision-making tasks as well as judgement tasks (Feldman, 1966). Reasons for this
are given in Feldman (1966); Zajonc (1968) and Kanouse & Hanson (1972). They state that the social environment where one lives in exist of more positive cues compared to negative cues.
Negative cues are seen as counter normative. Because of the latter, the negative cues that do appear have the tendency to attract attention and are more heavily attributed to the stimulus object than positive cues are (Kanouse & Hanson, 1972) in Sen & Lerman (2007). A reason for this is a higher sensitivity to negativity (Masland et al., 2015). When a person is more sensitive to negativity this person will value negative information more than positive information. This means for OCRs that, when a consumer is more sensitive to negativity, this consumer will tend to value negative
information more than positive information and therefore will trust OCRs with a negative valence more than OCRs with a positive valence (Sen & Lerman, 2007; Masland et al., 2015).
Based on this, the following hypotheses rises:
H8: The positive effect of a negative OCR on perceived trustworthiness is higher if consumers are more sensitive to negativity.
2.4.3 Preference for Literal or Figurative language by consumers
In Kronrod & Danziger (2013), have argued been argued that there is a positive link between emotional intensity and the preference for figurative language in OCRs. Emotional intensity, in this research, is defined by Babin et al., (1994) as emotional and aroused consumption, and is depended of how emotional stable and extraverted a consumer is (Gosling et al., 2003). If a consumer is more emotional stable and more extraverted, this consumer will have a higher preference for figurative language in OCRs since this consumer is more receptive for figurative language. On the other hand, OCRs that are more ‘emotional intense’ are more often written in figurative language compared to literal language (Babin et al., 1994). These OCRs focus more on experiences with a product instead of focussing on product attributes. Consumers with a higher emotional intensity are more affected by these figurative messages.
From this the following hypothesis rises:
H9: A high score on emotional intensity positively affects the relation between figurative language
and perceived trustworthiness.
2.5 conceptual model
This figure shows that the use of figurative and literal language in an OCR influences the perceived trustworthiness of OCRs. Next, it shows that the valence of an OCR and the product category the OCR is about can influence the perceived trustworthiness of the OCR. Furthermore, in this model the covariates sensitivity to negativity, preference for figurative/literal language and skepticism towards OCR are hypothesised to influence the dependent variable perceived trustworthiness of OCR’s. The conceptual model is seen in figure 2.5 below.
Figure 2.5: Conceptual model
Preference for fig/lit language (covariate)
Sensitivity to negativity (covariate)
Review Valence (Positive vs Negative)
Skepticism towards OCR
(covariate)
H2
Perceived Trustworthiness of OCR’s H3
Linguistic language of OCR’s (Literal vs Figurative)
H1
H5a + H5b Product Category (Hedonic vs
Utilitarian)
H4
H8
H7 H9
H6
3.0 Research design
In this section a detailed description will be given on the type of research and methods used to conduct this research and the set-up of experimental design, followed by the operationalization and population and sample design. Next, manipulation checks will be described followed by the plan of analysis
3.1 type of research and method
Present studies suggest that there is a positive relationship between the use of figurative language in an OCR and the perceived trustworthiness of an OCR by consumers. A second suggestion made by other studies is that an OCR with a negative valence is perceived more trustworthy than an OCR with a positive valence. Furthermore, OCRs written in figurative language have a higher positive effect on OCRs written on hedonic products while OCRs written in literal language have a higher positive effect on OCRs written on utilitarian products. It is also assumed that OCRs written on utilitarian products will be more perceived as trustworthy than OCRs on hedonic products. In addition, 3 covariates are investigated—Skepticism on OCRs, Preference for figurative/literal language and sensitivity to negativity. According to Malhotra (2009) this type of research is therefore a causal experiment.
In the present study, a 2 (figurative language vs literal language used in writing OCR) x 2 (negative valance vs. positive valence of an OCR) x 2 (hedonic vs. utilitarian product category) a between- subject experimental design is used to test the hypotheses. The questionnaire was randomly distributed online. A number of 296 respondents answered the questionnaire of which 196 where usable. Of these respondents, 89 are male and 107 females. Each participant was assigned to one of the 8 scenarios randomly and saw a stimulus: a scenario that is distributed as an OCR. Each
scenario consisted of a use of language (figurative or literal), valence (positive or negative) and product category (hedonic or utilitarian) and were different from each other based on at least one of the 3 factors. A more detailed overview can be seen in table 3.1.1 below.
Scenario Linguistic language Valence Product category
1 figurative positive hedonic
2 literal positive hedonic
3 figurative negative hedonic
4 literal negative utilitarian
5 figurative negative utilitarian
6 literal negative hedonic
7 literal positive utilitarian
8 figurative positive utilitarian
Table 3.1.1: Description of the different scenarios
The different scenarios were tested on the dependent variable perceived trustworthiness to test which scenario was perceived as the most trustworthy OCR.
Table 3.1.2: Description of the means per scenario on the perceived trustworthiness of the OCR
As can be seen from table 3.1.2 the participants that answered scenario 2, the positive OCR on a hedonic product written in literal language as the most trustworthy OCR in this research. A more detailed table is seen in Appendix D.
3.2 Set up of experimental design
Linguistic language
A pre-test was conducted with the goal to test when consumers find an OCR predominantly literal or figurative. A pre-test with six different texts was conducted online of which three of these texts where conveyed on a literal manner and three of them conveyed on a figurative manner based on the requirements of both literal and figurative texts as stated in section 2.1. The texts A, B and E were conveyed in a figurative context and texts C, D and F were conveyed in a literal context.
Respondents had the choice to rate the different texts as literal or figurative based on a 7-point likert scale varying from 1 = strongly literal, 2 = literal, 3 = more or less literal, 4 = neutral, 5 = more or less figurative, 6 = figurative to 7= strongly figurative. This means that the more a text’s mean is close to 7, the more it will be perceived as figurative. As with the literal texts it is the opposite, the closer the mean is to 1 the more it will be perceived as literal. The texts used to research this are seen in Appendix A. The results of the survey are seen in table 3.2.1 below. A more detailed table on the descriptive statistics of the pre-test is seen in Appendix E.
Scenario 1
Scenario 2
Scenario 3
Scenario 4
Scenario 5
Scenario 6
Scenario 7
Scenario 8
Mean 4,59 5,21 4,24 5,04 4,43 5,19 4,44 3,61
Table 3.2.1 - Descriptive statistics of the pre-test on linguistic language.
A total of 27 respondents have responded to the survey. As seen in table 3.2.1 above Text B have the highest mean, which is 5,81. From this it can be concluded that this text was perceived by the respondents as the strongest figurative text in the survey. The strongest literal text in this survey as perceived by the respondents is text D with a mean of 2.00. A paired sample t-test between the figurative text B and the literal text D was conducted in order to find out if the two texts were significantly different from each other. The results are seen in table 3.2.2 below.
Mean Std. Deviation Std. Error Mean Sig. (2-tailed)
Pair 1 Text A - Text C 3,222 1,928 0,371 0,000
Pair 2 Text A - Text D 3,519 1,805 0,347 0,000
Pair 3 Text A - Text F 3,333 2,353 0,453 0,000
Pair 4 Text B - Text C 3,519 1,968 0,379 0,000
Pair 5 Text B - Text D 3,815 1,902 0,366 0,000
Pair 6 Text B - Text F 3,630 2,356 0,453 0,000
Pair 7 Text E - Text C 3,333 1,641 0,316 0,000
Pair 8 Text E - Text D 3,630 1,497 0,288 0,000
Pair 9 Text E - Text F 3,444 1,826 0,351 0,000
Table 3.2.2: T-test of paired samples on literal text and figurative text
As can be seen in the results all the figurative and literal texts are significantly different from each other based on a significance level of p<0,01. This indicates that the respondent did perceive the figurative texts as figurative and the literal texts as literal. Based on these results and the results from table 3.2.1 there is chosen to continue with text B as a figurative text and text D as the literal text. The reason for this is that text B has the highest mean and text D the lowest mean, indicating that these where perceived as most figurative text and most literal text in this research. A more detailed table is seen in Appendix F.
Text A Text B Text C Text D Text E Text F
N 27 27 27 27 27 27
mean 5,52 5,81 2,30 2,00 5,63 2,19
std. dev 1,451 1,520 0,869 0,784 1,182 1,297
Review valence
The valence of OCRs in the survey will be either negative or positive. An additional survey is conducted to check if the manipulation check really worked. In this survey the respondent will see two different texts, of which one has a positive valence and the other one has a negative valence.
Each text has one question through which the respondent can indicate how positive/negative the text was perceived. The questions were based on a 7-point Likert scale which ranged from (1) very positive towards (7) very negative.
Product category
In the research of Kronrod & Danziger (2013) the authors gave a classification of products that either were found to have high loadings with the hedonic product category or with the utilitarian product category. According to this article the following products were considered hedonic: Live concert CDs, Leisure and cooking books, Travel guides, DVD about water parks in the world and Digital-frame key holder. The same article argued that Language learning CDs, Textbook on accountancy, Financial guides, DVD about global warming and Cordless mouse for the computer belonged to the utilitarian product category. The reason for the choice of this categorisation of products is the fact that this categorisation was tested by Kronrod & Danziger (2013) in practice to find out if respondents actually perceived these products as so. The results were positive. Also, Kronrod & Danziger (2013) provide the products in pairs. This means that the hedonic product, for example Travel guide, has a utilitarian alternative; financial guide in this research. Therefore, the products do not differ from each other much; both guides but with a hedonic and utilitarian goal.
Hedonic Utilitarian
Live concert CD Language learning CD
Leisure and cooking books Textbook on accountancy
Travel guides Financial guide
DVD about water parks in the world DVD about global warming
Digital-frame key holder Cordless mouse for the computer
Table 3.2.3 - Categorization of products in either the hedonic category or the utilitarian category
3.3 Operationalization of the research
In this section the focus will lie on the operationalization and preparation of the dependent variable
the covariates, the manipulation check and the control variables. In table 3.3 below a detailed
overview can be seen which focuses on the dependent variable, covariates and manipulation check.
• An OCR is a reliable source of information about the quality and performance of products
• An OCR is truth well told
• In general, an OCR presents a true picture of the product being reviewed
• I feel I’ve been accurately informed after viewing most OCRs
• Most OCRs provide consumers with essential information.
Sensitivity to negativity
Berna et al. (2011) • You have recently taken an important exam. Your results arrive with an unexpected letter of explanation about your grade.
• You are about to move with your partner into a new home. You think about living here.
• You are lost in a part of a big city you don't know well. You ask someone on the streets for directions
7-point likert scale. Response anchors range from 1 (very unpleasant) towards 7 (very pleasant).
Based on these 4 scenarios a score can be derived which measures how sensitive one is to negative information based on sensitivity to negativity.
• EV: 1,397
• CA: 0,346
Variable Source Items Scale Eigenvalue (EV) and
Cronbach’s Alpha (CA)
Perceived trustworthiness in OCRs
Andrews et al., (2001)
• I find this review (trustworthy)
• I find this review (believeable)
• I find this review (credible)
7-point likert scale.
Response anchors were
trustworthy/credible/believable and range from 1 (very
untrustworthy/unbelievable/uncredible) towards 7 (very
trustworthy/believable/credible).
• EV: 2,488
• CA: 0,897
Skepticism
Obermiller and
Spangenberg (1998); Hardesty et al., (2002)
• We can depend on getting the truth in most OCRs
• OCR’s aim is to inform the consumer
• I believe an OCR is informative
• An OCR is generally truthful
7-point Likert scale. Response anchors were strongly agree (1) towards strongly disagree (7).
A higher score would suggest greater skepticism.
• EV: 5,505
• CA: 0,919
when they pull out something from their pocket.
• You give a speech at your friend's wedding. When you have finished, you observe the audience's reaction.
Preference for figurative/literal language
Gosling et al., (2003).
I see myself as:
• Extraverted, enthusiastic
• Reserved, quiet.
• Anxious, easily upset
• - Calm, emotionally stable
7-point likert scale.
Response anchors were Strongly disagree (1) towards strongly agree (7)
• EV: 1,916
• CA: 0,663
Table 3.3: Operationalization scheme
3.3.1 Dependent variable
To measure the perceived trustworthiness of consumers in the OCRs that they have read in the survey seven Likert items were operationalized to reflect the perceived trust of the consumers in the OCRs. These items were adopted from studies of Lin (2006), Mortensen (2009) and Smith et al.
2005) and were modified to make them fit more for the purpose of this study. Respondents had the choice to rate the OCRs on the degree of trust through a seven-point scale. This 7-point Likert scale varied from 1 = “strongly agree” to 7 = “strongly disagree”.
3.3.2 Manipulation check questions
For the experimental variable ‘linguistic language’ a manipulation check question was being asked after reading the scenario and answering the questions on trustworthiness as mentioned in the section above, respondents had to rate the text on a 7-point Likert scale from ‘very literal’ towards
‘very figurative’. Through this question it was possible to check whether the respondents who received a literal (figurative) text really perceived this text as literal (figurative).
3.3.3 Covariates
The respondent rated each covariate on the items as seen in table 3.3. These different items measured how skeptic the respondent is towards OCRs, how sensitive the respondent is towards negativity in order to check whether one is more sensitive towards negative information and if the respondent prefers figurative or literal language in OCRs. Skepticism towards OCRs is measured through 9 items, sensitivity for negativity through 4 items and preference for figurative or literal language is also measured through 4 different items. Bachorowski & Braaten (1994) have
developed a scale to measure the emotional intensity of an individual. This scale is measured based
on 30 different factors and is measured through the 5 personality traits: Openness to experience, Neuroticism, Extraversion, Conscientiousness, and Agreeableness. These five dimensions are better known as the Big-Five framework. However, according to Bachorowski & Braaten (1994)
emotional intensity has the strongest association with the dimensions Extraversion and Neuroticism.
Therefore, the measurement of emotional intensity in this research is based on factors of these two dimensions. Gosling et al. (2003) provides a measurement scale on the personality traits. The scales that belong to these two dimensions will be used, all of these items were rated on a 7-point Likert scale. Skepticism was rated on a reversed scale since a higher score would suggest greater
skepticism (Obermiller & Spangenberg, 1998; Hardest et al., 2002). Furthermore, respondents that have a higher emotional intensity are more receptive towards figurative language and therefore will see figurative OCRs as more trustworthy compared to literal. Therefore, the items ‘I see myself as:
Reserved, Quiet’ and ‘I see myself as: Anxious, easily upset’ where also rated on a reversed scale since a higher score suggests great emotional intensity (Gosling et al. 2003).
3.3.4 Control variable
At the end of the questionnaire some demographic information is collected. This is done by the items ‘age’ and ‘gender’. It is possible that these variables have an effect on the perceived trustworthiness of OCRs.
3.3.5 Factor validity of the scales
A factor analysis is used to analyze the unidimensionality of the proposed scales. The factor
analysis was used on each variable separately first. In order for a factor analysis to be appropriate to use the Kaiser-Meyer-Olkin (KMO) needs to have a higher value than 0.5. Next to this, Bartlett’s test of sphericity needs to be significant, the eigenvalues have to be > 1, communalities above 0.4.
The factor analysis was used on the dependent variable ‘perceived trustworthiness of OCRs’, and the covariates ‘Skepticism towards OCRs’, ‘Preference for literal/figurative language’ and
‘Sensitivity to negativity’. After the factor analysis the internal consistency of the variables was measured through Cronbach’s Alpha.
The factor analysis on the depended variable indicated that all three items can be used in the regression. All the conditions which make a factor analysis appropriate are high enough and
Cronbach’s alpha shows that the items are internal consistent enough to continue, which is the case when a Cronbach’s alpha is higher than 0.6 (Malhotra, 2009). It was not possible to achieve a higher Cronbach’s alpha by deleting one of the items. This is also the case for the variable
‘Skepticism towards OCRs’. However, when the factor analysis was computed on the variable
‘Sensitivity to negativity’ item 1 and item 2 had a communality <.4 which makes them too low to continue. A factor analysis was distributed on item 3 and 4. KMO went down towards 0,500, which is still high enough. Bartlett’s test of Sphericity went down towards 0,01 and the commonalities of both items rose towards 0,613. Cronbach’s alpha was checked on item 3 and 4 and an alpha of 0,370 was derived, which is still too low. Since sensitivity to negativity is important in this research there will be chosen to continue with 1 item. In this study, item 3 showed a higher correlation with the other scenarios than item 4 did. This can be implied as that item 3 has more in common with the other scenarios that are used to measure sensitivity to negativity. Berna et al. (2011) stated that a higher score on the items results in a higher sensitivity to negativity. Furthermore, this study hypothesizes that there is an effect of sensitivity to negativity on perceived trustworthiness of OCRs. In order to test this in the most reliable way the item that reflects this best needs to be used.
Therefore, there is chosen to continue with item 3.
Furthermore, the factor analysis on ‘Preference for figurative/literal language’ also met the conditions to continue with a factor analysis. The factor analysis computed 2 factors. Factor 1 consisted of item 1, 2 and 3 and factor 2 of item 4. However, by deleting item 4 the Cronbach’s Alpha increased from 0,631 towards 0,663. Furthermore, by deleting this item all the other 3 items loaded on 1 factor. By deleting this item, the KMO measure also increased towards 0,601.
After that, a factor analysis with all items was performed to ensure that each variable load on the right factor. The component matrix in table 3.3.5 below shows which item loads on which factor.
Item Factor
