This study posits that sources of online product reviews can induce differential effects on 2 dimensions of source credibility, perceived expertise and perceived trustworthiness. Study 1 shows that experts are perceived as having more expert knowledge, but at the same time as having less trustworthiness than laypersons, and vice versa. These opposing credibility evaluations suppress the effect of online source identification on readers' attitudes toward online product reviews. Study 2 finds that these opposing credibility assessments only emerge when the expert status of the source is based on self-claims. When the expert status of the online source is based on peer ratings, the source is assessed as having both expertise and trustworthiness.
User-generated content moves advocacy away from traditional one-way mass communication in which a central sender addresses a mass audience. Using today's interactive media, people formerly known as “the audience” (Rosen, 2006) are increasingly dictating product information themselves. The opportunity for people to engage actively in the public information process with regard to products and services provides consumers with a rich and varied set of electronic word-of-mouth messages, often posted in the form of online product reviews.
User-generated product reviews are a persuasive source of information in shaping consumers' attitudes and their purchases (see, e.g., Bickart & Schindler, 2001; Senecal & Nantel, 2004). Consumers rely on online reviews to make decisions that range from which movie to watch to which insurance to buy. The persuasiveness of online reviews has often been attributed to their source credibility. Consumers regard product evaluations from fellow consumers, whose brand or product endorsements presumably have no persuasive intent, as more credible than product information from marketers (Bickart & Schindler, 2001; Brown, Broderick, & Lee, 2007).
However, not all reviewers are likely to be equally credible. Review sites allow anyone to post anything about any product (Winter, Krämer, Appel, & Schielke, 2010), whether one is qualified to assess a product critically or not. Reviewers thus include laypersons who claim little knowledge of the product under review, to self-proclaimed experts who claim to know about a product as part of their profession (Mackiewicz, 2010). Finally, rated experts are reviewers whose status has been vetted through the ratings of other users based on their past helpfulness.
This study examines the relative effects of laypersons and self-proclaimed experts, and compares these with the effects of rated experts in invoking different evaluations of perceived source expertise and perceived source trustworthiness, i.e., two dimensions of source credibility. We posit that, due to these differential evaluations, there may be an ironic effect in which a form of source identification has a boomerang effect on readers' assessment of source credibility, and subsequently, on readers' attitudes toward the review. Teasing apart these differential evaluations may facilitate a better understanding of how source identification affects readers' judgments in the online context.
Consumers are inclined to seek and use reviewers' personal identifying information as cues for assessing their credibility (Xie, Miao, Kuo & Lee, 2011), including information that categorizes the reviewer as an expert or layperson (Metzger, Flanagin, & Medders, 2010; Smith, Menon & Sivakumar, 2005). However, inferences about the expertise of a source on a review sites can often only be drawn from what the source discloses in the review (Brown et al., 2007). According to a study of digital camera reviews, Mackiewicz (2010) found that online reviewers often begin their reviews with expertise claims. With these claims, reviewers construct a persona of expertise from the outset (“In my line of work, wedding and special event photography, durability and adaptability are critical”), or the lack thereof (“I am not an expert on digital cameras or digital photography”). The presence of such claims in online reviews has also been observed in a number of other studies. For example, Otterbacher (2011, cf. Willemsen, Neijens, Bronner, & de Ridder, 2011) found that almost a quarter of online reviews makes reference to the reviewer's level of expertise. Do consumers readily use these claims as an indication of source credibility? Given the prominence of expertise claims in online reviews, this study aims to shed light on how expertise claims are interpreted in terms of perceived credibility.
Perceived Credibility of Review Sources
Source credibility is a multifaceted construct, and generally believed to consist of expertise and trustworthiness perceptions (see Ohanian, 1990 for a review). Whereas perceived expertise refers to the degree to which a source is considered to be capable of making valid assertions, perceived trustworthiness refers to the degree of confidence that a source is motivated to communicate valid assertions (Hovland, Janis, & Kelley, 1953; McCracken, 1989). The rich body of credibility research demonstrates that more credible sources produce more attitude change than less credible sources (see Pornpitakpan, 2004 for a review). Reliance on perceived source expertise versus trustworthiness as a criterion for credibility assessments does not seem to affect these results (Mills & Jellison, 1967; Rhine & Severance, 1970).
In line with traditional credibility research, research on electronic word of mouth has found source credibility to have a profound effect on consumers' judgment and choice (Brown et al., 2007; Lim, Sia, Lee, & Benbasat, 2006). However, the contribution of source identification—experts or laypersons—to a source's perceived credibility remains unclear (Vermeulen & Seegers, 2009). In face-to-face settings expert endorsers have repeatedly been found to induce more source credibility than laypersons (DeBono & Harnish, 1988; Pallak, Murroni, & Koch, 1983). In online settings, however, although some studies have found expert endorsers to induce more credibility than laypersons, others have found laypersons to induce more credibility than experts (e.g., Schindler & Bickart, 2005; Smith, Menon & Sivakumar, 2005; Wang, 2005). A third group of studies finds no differences in the perceived source credibility of experts and laypersons (see Willemsen et al., 2011 for a review).
The ambiguous results with regard to the role of online source identification may be due to their differential effects on perceived expertise versus trustworthiness: A source may be perceived to have great expertise but little trustworthiness, and vice versa (Pornpitakpan, 2004). These differential effects may cancel each other out when expertise and trustworthiness are assessed as part of one and the same credibility judgment, as many studies involve. The approach of the present study is therefore to establish the effect of different review sources separately on perceived source expertise and trustworthiness.
Perceived Expertise and Perceived Trustworthiness
Source expertise is based on the accumulation of skills, competences, or knowledge through experience (Ohanian, 1990). If a reviewer claims to have topic mastery because of (formal) training or a hobby relevant to the product under review, consumers may be likely to perceive this reviewer to be an expert. The self-proclaimed expert status of a source serves as a cue signalling that the source provides valid information due to his/her authority on the subject. As humans are “cognitive misers” (Taylor & Fiske, 1978), who do rely on no more cues than necessary to form an impression, this cue prompts people to attribute expertise to a source, regardless of the quality of the information he or she offers. Metzger et al. (2010) found that discerning expertise was critical for consumers in determining a source's competence and authority on a subject.
Different results may be expected for source trustworthiness. Source trustworthiness is determined by attributions about the motives of a source to share particular information. Thus, consumers who judge a review source on trustworthiness, will base their judgment on the causal inferences they make regarding the reviewer's motivation to endorse a product (McCracken, 1989). Consumers thereby attribute product endorsement to either the product's actual attributes or to factors unrelated to the product, such as dispositional characteristics of the reviewer (Sen & Lerman, 2007). Following the discounting principle of attribution theory (Kelley, 1973), consumers will discredit an endorsement when they attribute it to nonproduct-related factors, that is, when the consumer suspects that the endorsement is not based on product performance but, instead, a reviewer's intent to persuade.
On this basis, a reviewer who claims to be an expert is likely to be perceived as less trustworthy than a reviewer who claims to be a layperson (Huang & Chen, 2006; Senecal & Nantel, 2004). Self-proclaimed expertise creates the impression that the reviewer wants to present him/herself favorably, potentially with the intention to persuade others. Consumers are often skeptical about the truthfulness of such favorable self-presentations (Walther et al., 2009) and the presence of persuasive intent (Doh & Hwang, 2009; Metzger et al., 2010; Schindler & Bickart, 2005). These concerns are legitimate since some marketers stimulate expert consumers (given their knowledge of the product category) to endorse their products by providing an incentive for posting online reviews (Huang & Chen, 2006). Marketing messages can be easily disguised as unbiased reviews given the lack of social cues in online environments. This makes reviews prone to manipulation, and consumers wary of their authenticity.
Authenticity concerns stemming from fear of other's manipulations may be less likely when the reviewer claims to be a layperson. A claimed lack of expert knowledge signals sincerity and serves as a validation cue that the review does not come from someone who (falsely) claims to have expert knowledge to impress online consumers, and influence their judgment or behavior (Mackiewicz, 2010; Schindler & Bickart, 2005). Moreover, consumers are inclined to trust people whom they perceive to be homophilous to themselves (McCroskey, Richmond, & Daly, 1975; Wang, Walther, Pingree & Hawkins, 2008). As everyday consumers have generally no expert knowledge of the product that they may read about, they are more likely to identify with, and form trusting attitudes towards a layperson than an expert (Huang & Chen, 2006; Metzger et al., 2010).
The discussion above implies that consumers may be left in a quandary when assessing the credibility of reviewers. While a self-proclaimed expert may exhibit more expertise than a layperson, the latter may be more trustworthy. This means that the effect of source identification on attitude formation may not be as straightforward as suggested in the traditional credibility literature. The effect of a source who claims to be an expert (vs. layperson) is likely to induce two mechanisms that work in opposite directions by increasing one dimension of source credibility (perceived expertise) while simultaneously decreasing the other dimension of credibility (perceived trustworthiness). Since perceived expertise and trustworthiness are both positively related to attitude change (Gilly, Graham, Wolfinbarger, & Yale, 1998; Harmon & Coney, 1982; Lascu, Bearden, & Rose, 1995), source identification is likely to produce two opposing indirect effects on attitude towards the review—i.e., one positive indirect effect through perceived expertise and one negative indirect effect through perceived trustworthiness. More precisely, a self-proclaimed expert (vs. layperson) will increase perceived expertise, which then induces a more positive attitude towards the review, but simultaneously lower perceived trustworthy, which induces a less positive attitude towards the review. When operating together, these opposing indirect effects are likely to cancel each other out, thereby producing a nonsignificant effect of source identification on attitude towards the online review. Such a situation, also referred to as one of suppression (MacKinnon, Krull, & Lockwood, 2000; Preacher & Hayes, 2008), would explain the in consistent findings in the literature regarding the role of source identification.
Indeed, the literature suggests evidence of a suppression situation. In a series of experiments on herding in online product choices, Huang and Chen (2006) showed that an expert source was perceived as having significantly more expert knowledge than a layperson, but as being less trustworthy. Although these results suggest that source identification invokes different mechanisms through perceived expertise and trustworthiness, it is yet to be established that suppression is an outcome of these differential mechanisms en route to influencing message attitude. Hence, it is hypothesized:
H1. A review source who claims to be an expert is perceived as having (a) more expertise, but (b) less trustworthiness than a review source who claims to be layperson.
H2. Perceived expertise (a) and perceived trustworthiness (b) are positively related to attitude towards the review.
H3. Review source identification (self-proclaimed expert vs. layperson) induces (a) a positive indirect effect on review attitude through perceived expertise, and (b) a negative indirect effect through perceived trustworthiness, which (c) in tandem, will suppress the effect of source identification on attitude towards the review.
Participants in a student subject pool were invited via e-mail to participate in the experiment, which was administered online to allow maximum realism. A total of N = 265 participated and were randomly assigned to the experimental conditions. The sample consisted of 31% men and 69% women, with a mean age of 22.9 years (SD = 3.19).
Stimuli and Independent Measures
Participants were asked to evaluate a product review posted on a fictitious review site which endorsed either a television set or a hotel accommodation. Reviews were identical except for the their source. In the layperson condition, the source stated in the textual body of the review that he had no special knowledge of the product class. In the self-proclaimed expert condition, the source stated that he had a profession that was related to the product (i.e., lecturer in hotel management or technician for the hotel and television conditions, respectively). After exposure to the review, participants completed a survey which included questions about the review, the source of the review, and review involvement.
To measure review attitude, participants reported their evaluations of the review on four bipolar scales ranging from 1 to 5 (e.g., bad-good, unuseful-useful), based on prior research (LaTour & Rotfeld, 1997), α = .87. The two dimensions of source credibility were adopted from scales developed by Ohanian (1990) using 5-point bipolar items, four measuring perceived trustworthiness (e.g., dishonest-honest, insincere-sincere), α = .91; and four measuring perceived expertise (e.g., inexperienced-experienced, unknowledgeable-knowledgeable) α = .94. Finally, perceived review quality, message tone, participants' self-reported expertise and involvement were included in the survey as potential controls. Specifically, we asked participants to rate the review in terms of perceived quality (1 = low quality; 10 = high quality), and message tone (1 = negative; 5 = positive). Participants' expertise was measured by asking them to rate their knowledge with regard to the product category (1 = no expert knowledge; 5 = expert knowledge).
The scale items from source trustworthiness, source expertise, and review attitude were subjected to confirmatory factor analysis (CFA) to assess the measurement model. Constructs were modeled as three correlated first-order constructs. Indices signaled a satisfactory model fit: χ2(51) = 118.10, p < .001, CFI = .98, TLI = .97, RMSEA = .058 (90% CI: .04, .07). Item reliabilities were evaluated by examining the standardized loading of each measure. All items yielded high standardized loadings on their intended constructs (λ > .70), and hence were retained for analysis.
Convergent validity was assessed using Composite Reliability (CR) and the average variance extracted (AVE) as recommended by Fornell and Larcker (1981). The CR values ranged from .82 to .89, all exceeding the cut-off value of .70. AVE values ranged from .77 to .86 and thus surpassed the cut-off value of .50. Hence, the measurement model was considered acceptable. Discriminant validity among constructs was examined by comparing the square root of the average variance extracted (AVE) with the correlation between two latent constructs. Discriminant validity is satisfied when the square root of the AVE exceeds this correlation. This criterion was met for all pairs of constructs in the model, including expertise and trustworthiness, indicating that source expertise and trustworthiness are two separate constructs. This was also confirmed by the weak correlation between source expertise and trustworthiness (r = .14).
Manipulation and confound check
The manipulation check indicated a successful manipulation of review source identification. Respondents assigned to the layperson condition (vs. self-proclaimed expert condition) more often mentioned that they read a review from a layperson than respondents assigned to the self-proclaimed expert condition and vice versa, χ2Layperson (1) = 98.48, p < .001; χ2Self-proclaimed expert(1) = 79.26, p < .001. The conditions showed no differences in respect to review quality, tone-of-voice, self-reported expertise, participant expertise, gender or age (all Fs < 1.19).
To test whether source identification has differential effects on the two dimensions of source credibility— perceived expertise and trustworthiness—a mixed analysis of variance was conducted with source identification and product type as between-factors, and trustworthiness and expertise as repeated measures. As expected, the results revealed a significant interaction between source identification and its effect on perceived expertise versus trustworthiness, Wilk's λ = .71, F (1, 263) = 106.41, p < .001, multivariate η2 = .29, indicating that source identification indeed has different effects on perceived expertise and trustworthiness. As shown in Table 1, follow-up contrast tests showed that a self-proclaimed expert scored significantly greater on perceived expertise than a layperson, which provides support for H1a. A reverse pattern was found for perceived trustworthiness. In line with H1b, a self-proclaimed expert was considered to be less trustworthy than a layperson. No significant differences were found between the perceived expertise and trustworthiness of review sources among the different product categories (F < 1). Therefore, the product conditions were collapsed in subsequent hypotheses tests.
Table 1. Effects of Source identification on Perceived Source Expertise and Trustworthiness
Note. Mean scores on a 5-point scale (with standard deviations). Scores with different superscripts in the same column denote significant differences, p < .05, in post hoc tests with Bonferroni adjustments.
Attitude Towards the Review
Hypothesis 2 posits that perceived source expertise and trustworthiness are positively related to attitude towards the review. Regression analyses was performed with attitude towards the review as outcome variable, and perceived expertise and perceived trustworthiness as predictor variables. In line with our hypothesis, the results revealed that perceived expertise (β = .34, p < .001) and trustworthiness (β = .43, p < .001) were both positively related to attitude towards the review. The two variables together explained 35% in the variance of attitude towards the review, F (2, 264) = 69.28, p < .001.
The results presented above showed that a review source (rated expert vs. layperson) has differential effects on what are assumed to be the two dimensions of credibility—i.e., perceived source expertise and trustworthiness—and that both dimensions have a positive effect on attitude towards the review. To assess whether these differential effects suppress the relationship between source identification and attitude towards the review, a formal test of suppression was performed as recommended by MacKinnon et al. (2000; cf. Cheung & Lau, 2008; Preacher & Hayes, 2008; Shrout & Bolger, 2002). According to these authors, suppression is closely related to mediation and can be tested using the same methods, i.e., by analyzing the total, direct, and indirect effects between a set of variables. Within the mediation model presented in Figure 1, suppression would be present when: (1) the direct effect and indirect effects of source identification on review attitude through perceived expertise or trustworthiness have opposite signs, which produce a total effect close to zero; and (2) controlling for the effects of perceived expertise and trustworthiness increases the regression coefficient of source identification such that the direct effect becomes larger in magnitude than the total effect. To assess these criteria, bootstrapping analyses were conducted with 5,000 resamples and a bias corrected and accelerated 95% confidence interval (CI), using Preacher and Hayes' (2008) INDIRECT macro for SPSS. Generally, when the bias-corrected CI does not include zero, the effect can be said to differ significantly from zero (Preacher & Hayes, 2008).
In line with H3a, bootstrapping analyses revealed a significant positive indirect effect of source identification (dummy-coded) on attitude towards the review through perceived expertise, b = 0.31, 95% CI [.21, .43]. A self-proclaimed expert (vs. layperson) had a positive effect on perceived expertise (b = 0.72, p < .001), and perceived expertise, in turn, had a positive effect on review attitude (b = 0.43, p < .001). As predicted by H3b, the results also showed a significant negative indirect effect of source identification on attitude towards the review through perceived trustworthiness, b = -0.19, 95% CI [-.28, -.11]. A self-proclaimed expert source (vs. layperson) had a negative effect on perceived trustworthiness (b = -0.41, p < .001), which in turn, had a positive effect on attitude towards the review (b = 0.45, p < .001). Thus, source identification (self-proclaimed expert vs. layperson) produces two indirect effects on attitude towards the review, i.e., one positive indirect effect through perceived expertise and one negative indirect effect through perceived trustworthiness.
This pattern of indirect effects demonstrate the presence of two mechanisms that work in opposite directions, which in tandem, could undermine the relationship between source identification and attitude towards the review. This is confirmed after inspection of the direct and total effect. Controlling for the effects of perceived expertise and trustworthiness, a self-proclaimed expert, in contrast to a layperson, was found to have a negative direct effect on attitude towards the review (b = -0.17, p < .05). As such, the results meet the first criterion of suppression, i.e., an opposing direct effect and indirect effect: The direct effect of source identification on attitude towards the review is consistent in sign with its indirect effects through perceived trustworthiness, but opposite in sign with its indirect effect through perceived expertise. The results also support the second criterion of suppression. Accounting for perceived expertise and trustworthiness increases the regression coefficient of source identification such that the direct effect becomes larger in magnitude than the total effect (b = -0.17, p < .05 vs. b = -0.05, p = .61). Thus, perceived trustworthiness and expertise show indirect effects that cancel out as demonstrated by a total effect closer to zero than the direct effect (Preacher & Hayes, 2008). Hence, H3c is supported.
Study 1 examined the relative effects of self-proclaimed experts and laypersons on perceived trustworthiness and perceived expertise—i.e., two dimensions of source credibility— and their subsequent effects on consumers' attitude towards an online product review. In line with hypotheses, the results indicate that a self-proclaimed expert, in comparison to a layperson, is perceived as having more expertise, but at the same time, less trustworthiness (H1). Also, perceived expertise and trustworthiness were both positively related to review attitude (H2). The result is the presence of two competing mechanisms: A layperson (vs. self-proclaimed expert source) had a positive indirect effect on review attitude through perceived trustworthiness, and a negative indirect effect through perceived expertise. The co-occurrence of these mechanisms were found to suppress the relationship between source identification and review attitude (supporting H3). Thus, self-proclaimed experts and laypersons evoke different evaluations in terms of perceived source expertise and perceived source trustworthiness, and due to these differential evaluations, there is an ironic effect of source identification on readers' assessment of source credibility, and subsequently, readers' attitudes towards online reviews.
The expert status of a source can be expressed not only by means of self-claims, but also by means of peer ratings (Winter et al., 2010). In the context of review sites, peer ratings allow website users to provide feedback about the performance of reviewers by rating whether they found the review helpful (Resnick, Zeckhauser, Friedman, & Kuwabara, 2000). Based on such ratings, review sites such as Amazon or Epinions publicly recognize reviewers through so-called badges. Badges are credentials such as “Top Reviewer” or “Advisor” granted to reviewers who have proven to be a helpful source of information in the consumer decision-making process. Badges are usually presented in a sidebar next to the name of the reviewer, and allow demarcation between laypersons and experts (hereafter: rated experts).
Similar to those who claim to be experts, rated experts are likely to accrue perceptions of greater expertise. Unlike self-proclaimed experts, their status as experts is warranted by others. According to the warranting theory, people attach credence to personal information when it is perceived to be immune to manipulation by the person to whom the information refers (see for review Walther, Van Der Heide, Hamel, & Shulman, 2009). Since rated experts are not able to manipulate the peer ratings on which their status is based, they are likely to be perceived as having expert knowledge more so than self-proclaimed experts.
Also, unlike proclaimed experts, rated experts are likely to be perceived as very trustworthy. Their identification by others reduces readers' suspicion about the possibility that reviewers are driven by persuasive intent. Websites only publicly recognize the expert status of reviewers when peers have consistently vouched for reviewers based on good review performance across different products of the same product category. When a reviewer has provided valid information on different products, it demonstrates that he/she has no intentions to promote one in particular. Good performance as evidenced by peer ratings is thus likely to translate into attributions about good intentions, and, subsequently, trustworthiness. Evidence consistent with this logic appears in e-commerce literature that repeatedly has shown peer rating systems to build trust (e.g., Ba & Pavlou, 2002; Bolton, Katok, & Ockenfels, 2004). Hence, rated experts may be perceived as equally trustworthy, or even more trustworthy, than a layperson.
However, based on the principles of perceived homophily, a rated expert may also be perceived as being less trustworthy than a layperson (Hovland et al., 1953; McCroskey et al., 1975). Based on this principle, consumers are likely to ascribe less similarity to a reviewer that has expert knowledge of a product than a reviewer that, like themselves, lack such expertise. As perceived similarity is positively related to trust, a layperson instead of a rated expert may induce more trust (Metzger et al., 2010; Huang & Chen, 2006; Smith et al., 2005). Based on the above, the following research questions arise:
RQ1a. How do rated experts, self-proclaimed experts, and laypersons differ with respect to perceived expertise?
RQ1b. How do rated experts, self-proclaimed experts, and laypersons differ with respect to perceived trustworthiness?
RQ2. Can reviewers' differences in perceived trustworthiness be attributed to differences in (a) suspicion of persuasive intent, or (b) perceived similarity?
Participants and design
Ninety-six graduate students voluntarily participated in the experiment, and were randomly assigned to one of three conditions (layperson vs. self-proclaimed expert vs. rated expert). The average age of the participants was 29.3 years, and 53% were females.
Procedure and stimuli
The procedure and stimuli were similar to Study 1, except for the type of product (digital camera) and the source identification manipulations. Respondents read a review endorsing a digital camera. Reviews were identical across conditions except for their source. The layperson source stated in the first sentence of the review that he had bought the camera for taking pictures during a holiday trip. The self-proclaimed expert source stated that he had bought the camera for his work as an event photographer. The rated expert source was similar to the other conditions with the exception that it did not contain any expertise claims. Instead, this condition showed a badge with a label designating “Rated as Top Reviewer” in a sidebar next to the reviewer's name.
The questionnaire included the same items as in Study 1 to measure perceived trustworthiness (α = .90) and perceived expertise (α = .91. In addition, based on prior homophily research (McCroskey et al., 1975), subjects indicated, their agreement with four statements (e.g., “The reviewer has a similar background as I have”) on 5-point scales, α = .70. To measure suspicion of persuasive intent, subjects indicated their agreement with six items from Obermiller and Spangenberg's (1998) skepticism scale that was adapted for the purpose of the study (e.g., “The reviewer is driven by ulterior motives”), α = .78.
To determine whether there was a perceptible difference in the treatment levels of source identification, participants indicated (1) the extent to which the source of the review was presented to be an expert, and (2) the extent to which the review community rated the source of the review to be an expert (1—Not at all an expert, 5—Very much an expert). An ANOVA yielded a significant main effect for source identification on the first manipulation check: expertise presentation, F(2, 93) = 11.47, p < .001, η2 = .20. According to post hoc contrast tests, participants more strongly agreed that the sources in both expert conditions (Mself-proclaimed expert = 3.59, Mrated expert = 3.81) were presented as experts than the source in the layperson condition (Mlayperson = 2.53), t(93) = 3.73, p <.001. An ANOVA on the second manipulation check, expertise rating, also indicated significant differences across sources, F (2, 93) = 7.78, p = .001, η2 = .14. As intended by the study design, participants agreed that the review community rated the source in the rated expert condition as significantly more expert (Mrated expert = 3.69, SD=1.09) than the source in the self-proclaimed expert condition or layperson condition (Mself-proclaimed expert = 3.20, SD = 1.00; Mlayperson = 2.60, SD = 1.16), t(93) = 3.34, p = .001. Thus, these manipulation checks suggest that source identification was manipulated successfully.
Perceived source expertise and perceived source trustworthiness
RQ1 asked how rated experts, self-proclaimed experts, and laypersons differ with respect to perceived expertise and perceived trustworthiness. A mixed analysis of variance was conducted, using source identification (Layperson vs. Self-Proclaimed Expert vs. Rated Expert) as the between-subjects factor, and perceived expertise and trustworthiness as repeated measures. A significant interaction emerged between source identification, and its effect on perceived expertise versus trustworthiness, Wilk's λ = .84, F(2, 93) = 9.13, p < .001, multivariate η2 = .164, indicating that source identification differed in perceived expertise and trustworthiness. Post hoc contrasts were conducted using Bonferroni corrections (See Table 1). Results demonstrated that a layperson was perceived as having significantly less expert knowledge than the self-proclaimed expert, or the rated expert source. No significant differences were found between the latter two sources. With regard to perceived trustworthiness, a layperson was perceived as significantly more trustworthy than the self-proclaimed expert source, whereas the rated expert source was not significantly different from either of the two.
Suspicion of Persuasive Intent and Perceived Similarity
A MANOVA assessed source identification as independent variable and suspicion of persuasive intent and perceived similarity as dependent variables. Results showed a significant effect of source identification on suspicion of persuasive intent, F(2, 93) = 4.95, p < .01, η2 = .10, but not perceived similarity, F (2, 93) = 1.58, p = .21. Post hoc tests with Bonferroni corrections showed that a self-proclaimed expert scored higher on suspicion of persuasive intent than a layperson or a rated expert, with no significant differences among the latter two conditions (see Table 2).
Table 2. Effects of Source identification on Suspicion of Persuasive Intent and Perceived Similarity
Suspicion of Persuasive Intent
Note. Mean scores on a 5-point scale (with standard deviations). Scores with different superscripts in the same column denote significant differences, p < .01, in post hoc tests with Bonferroni adjustments.
2.89 a (0.59)
3.31 b (0.70)
2.89 a (0.57)
To test further whether source differences with regard to trustworthiness could be explained by perceived suspicion of persuasive intent (RQ2), bootstrapping analyses were conducted with 5,000 resamples and a bias corrected and accelerated 95% confidence interval (CI), using Preacher and Hayes' (2008) INDIRECT macro for SPSS. Bootstrapping analyses revealed a significant negative indirect effect of source identification (dummy-coded: self-proclaimed expert vs. rest) on perceived source trustworthiness through suspicion of persuasive intent, b = -0.13, 95% CI [-.29, -.04]. A self-proclaimed expert (vs. layperson) had a positive effect on suspicion of persuasive intent (b = 0.42, p < .01), and suspicion of persuasive intent, in turn, had a negative effect on perceived trustworthiness (b = -0.32, p < .01). Moreover, the results indicated that the effect of source identification on perceived source trustworthiness (b = -0.33, p <.02) became nonsignificant when controlling for the effect of suspicion of persuasive intent, while the effect of suspicion of persuasive intent remained significant (b = -.020, p = .16). This provides support for full mediation, meaning that suspicion of persuasive intent explains why a self-proclaimed expert is perceived as less trustworthy than a rated expert or a layperson.
The present study examined three types of sources of online reviews in terms of source credibility: laypersons, self-proclaimed experts and rated experts. Study 1 showed an ironic effect where source identification induced opposing evaluations on the two dimensions of source credibility: perceived source expertise and perceived source trustworthiness. In the perception of readers, self-proclaimed expert reviewers exhibit more expertise but at the same time less trustworthiness than laypersons. As both perceived expertise and trustworthiness are positively related to attitude towards the review, the ironic effect of source identification produced two competing mechanisms en route to attitude formation. A self-proclaimed expert (vs. layperson) was found to have a positive indirect effect on attitude towards the review through perceived expertise and a negative indirect effect through perceived trustworthiness. In tandem, these opposing indirect effects suppress the relationship between source identification and readers' attitude towards review. Although opposing effects of (self-proclaimed) expert and layperson sources on perceived trustworthiness and perceived expertise have been found in previous research on online reviews (Huang & Chen, 2006), our study was the first to demonstrate such a suppression situation.
Study 2 showed that the presence of an ironic effect depended on the way an expert source was demarcated from a layperson. An online source was found to induce opposing credibility evaluations only when the expert status of the source was based on self-claims. When the expert status of the source was based on peer ratings, the twin judgments of source credibility were in line with each other. Replicating the results of study 1, a self-proclaimed expert is considered to have more expert knowledge than a layperson, but also less trustworthiness. A rated expert, however, is perceived to have as much expert knowledge as a self-proclaimed expert, and as much trustworthiness as a layperson. Expertise conferred by ratings seems capable of closing the gap between the two dimensions of source credibility.
The finding that rated experts are evaluated equally trustworthy as laypersons, suggests that, in the context of online review sites, it is not perceived similarity that drives trustworthiness. Prior research demonstrates that consumers are likely to trust people who they perceive to be homophilous, that is, people who like themselves do not have expert knowledge of the product under review (McCroskey et al., 1975; Huang & Chen, 2006). In contrast to what can be expected based on this prior work, rated experts did not differ from laypersons in perceived trustworthiness. Both sources, however, were evaluated more trustworthy than self-proclaimed experts. This seems to indicate that reviewers' self-presentations as experts make consumers suspicious of reviewers' motivations to share reviews, and doubt their trustworthiness. This was supported by mediation analyses. Suspicion of persuasive intent, and not perceived similarity, was found to explain the effects of review source identification on perceived trustworthiness. Reviewers who present themselves favorably in terms of expertise (i.e., self-proclaimed experts) make themselves suspect of the intention to persuade other consumers, which negatively affects perceived trustworthiness. This was not the case with reviewers whose expert status has been vetted through peer ratings (i.e., rated experts). A record of good conduct as evaluated by the review community thus serves as a validation cue that the reviewer does not intent to disguise persuasive messages as unbiased reviews, and can be trusted as a source.
Taken together, the results of these studies show that online review sources provide two routes to perceived source expertise and perceived source trustworthiness, i.e., the two dimensions of source credibility. The route to perceived expertise stems from self-proclaimed experts and rated experts and the route to perceived trustworthiness stems from laypersons and rated experts. However, only one type of source—rated experts—connects both routes such that it scores high on overall credibility.
Theoretically, the findings highlight the “authenticity dilemma” (Metzger et al., 2010) inherent to online media where impressions are formed in an environment that provides extremely few social cues. In forming impressions based on the cues that nevertheless remain, cues that sources disclose about themselves and their occupations are not as compelling as cues provided by other users, as credibility assessments move from “a model of single authority based on hierarchy…to a model of multiple authorities based on…networks of peers” (Metzger et al., 2010, p. 415). This is also in line with warranting theory (see Walther et al., 2009): Compared to online information based on sources' self-presentations, people give more weight to information that is immune to a source's manipulation, as peer ratings provide. In the context of online product reviews, if the expertise of a source is conferred by others, a reader has more faith in the source's identity and motivations (i.e., lack of persuasive intent) for reviews.
Methodologically, the results support previous calls in the literature (see Pornpitakpan, 2004) to assess the isolated effect of perceived source expertise and trustworthiness. This study finds that perceived source expertise and perceived source trustworthiness operate separately and produce differential effects on attitude formation. This suggests that scholars should assess these dimensions of credibility orthogonally to gain a better understanding of the effects of source evaluations in computer-mediated communication. Assessing perceived expertise and trustworthiness as part of one and the same credibility scale may obscure the complexity of source evaluations in online environments.
The results also show that consumers seem to integrate perceptions about the expertise and trustworthiness of a source into evaluations about the source's review itself. This mitigates societal concerns that consumers adopt content without discerning credible from noncredible online sources (Hu & Sundar, 2010; Metzger, 2007). However, the present results do not make clear whether consumers are generally disposed to judge online review sources as credible or not. The present study compared the conferral of credibility in terms of self-claims vs. peer ratings. It did not include a control condition in which no identification cue was present. More research is needed to learn how consumers evaluate reviewers' credibility when neither the reviewer nor the community offer expertise assessments.
Another consideration is that reviews were studied in just one context: a site on which reviews come from unknown sources with whom consumers have no prior relationship. The findings may not describe other platforms where recommendations are exchanged between acquainted contacts (e.g., social network sites, special interest communities, etc.). Research is needed to demonstrate how credibility perceptions differ depending on the platforms and relationships in which electronic word of mouth is exchanged. Finally, future research needs to compare source credibility across online versus offline contexts, in order to learn better how credibility differs between them and what might explain the differences.
This research was in part supported by a Top Talent grant from the Dutch National Science Foundation (NWO). The authors wish to thank Anne Koning and Sander Hengeveld for their assistance in the data collection and to the editor for his comments on the manuscript.
Lotte M. Willemsen, M.Sc. (email@example.com) is a Ph.D. candidate in the Amsterdam School of Communication Research (ASCoR) at the University of Amsterdam. Her research focuses on social media, electronic word of mouth, online persuasion, and consumer behavior.
Address: Kloveniersburgwal 48, 1012 CX Amsterdam, the Netherlands.
Peter C. Neijens, Ph.D. (http://www.pneijens.socsci.uva.nl) holds the Chair of Persuasive Communication and also serves as Dean of the Graduate School of Communication at the University of Amsterdam. His research focuses on media and advertising, public opinion, and political communication.
Address: Kloveniersburgwal 48, 1012 CX Amsterdam, the Netherlands.
Fred Bronner, Ph.D. (firstname.lastname@example.org) is professor emeritus of media and advertising research in the Amsterdam School of Communication Research (ASCoR) at the University of Amsterdam. His research interests include multimedia synergy, family decision making, electronic word of mouth, and economizing behavior of consumers.
Address: Kloveniersburgwal 48, 1012 CX Amsterdam, the Netherlands.