The impact of customer images on online purchase decisions: Evidence from a Chinese C2C Web site
First Monday

The impact of customer images on online purchase decisions: Evidence from a Chinese C2C Web site by Mingming Dai and Leo Van Hove



Abstract
The influence of electronic word-of-mouth (eWOM) on consumers’ online purchase decisions is well-documented. However, little is known about the impact of customer images, a recent form of eWOM. The present paper reports on an exploratory survey among 215 young users of the Chinese C2C shopping Web site Taobao.com. The idea was to test whether the so-called information adoption model — and in particular a variant used in earlier research on an online consumer community — could be transposed to the case of customer images. We find that the perceived quality of customer images — and in particular their ‘comprehensiveness’ — has a significant positive impact on purchase intention. This suggests that shopping sites should encourage customers to share high-quality customer images. From a methodological perspective, the limited reliability of two of the quality dimensions (namely ‘relevance’ and ‘timeliness’) catches the eye. This suggests that there is a need for a dedicated theoretical framework and for scales that take into account the specific nature of customer images. Simply transposing scales that have been validated for other forms of eWOM would not appear to be the best approach.

Contents

1. Introduction
2. Literature review
3. Research model and hypotheses
4. Research design
5. Data analysis
6. Conclusions

 


 

1. Introduction

The past decade has seen a spectacular development of the Internet in China. According to the China Internet Network Information Center (CNNIC, 2017), the number of Internet users increased from 137 million in 2006 to 731 million in 2016. With such a large base of Internet users, online shopping has been developing rapidly. In 2015, online retail turnover was 3.88 trillion RMB (0.62 trillion USD), which is an increase of 33.3 percent compared to a year earlier, and the number of online shopping customers has reached 413 million (CNNIC, 2016).

However, because of the virtual nature of the Internet, information asymmetry is an obstacle to building trust between buyers and sellers — in particular in a C2C context. Online reputational feedback mechanisms can solve the problem (Dellarocas, 2003b, 2004; Dellarocas and Wood, 2008). So-called customer images — that is, images of previously bought products posted by customers on shopping Web sites (typically together with short text comments) — are a new and increasingly popular form of online reputational feedback in China (and elsewhere).

Prior studies have demonstrated that electronic word-of-mouth (eWOM) — of which online reputational feedback is a typical form (Dellarocas, 2003a) — has a critical effect on an online purchasing decision. However, so far there is no specific academic study that investigates whether and how customer images influence the purchase decision. The present paper tries to fill this gap.

In particular, we wanted to test whether Sussman and Siegel’s (2003) information adoption model (IAM), which has been used numerous times to analyze other forms of eWOM, could also be applied to the specific case of customer images on shopping sites. In order to do so, we focus on the variant of IAM used by Cheung, et al. (2008) in earlier research on online consumer communities, and we exploit a survey among 215 young users of the Chinese C2C shopping Web site Taobao.com (https://world.taobao.com).

Our core hypothesis in this exploratory research is that customer images indeed have a significant impact on the purchase decision. In particular, our research questions are: Does the quality of customer images influence purchase intention? If so, which dimensions of quality matter most?

The remainder of the paper is structured as follows: the next section reviews the existing literature on online purchase decisions, eWOM, and online reputational feedback mechanisms. Section 3 presents the theoretical model and the hypotheses. Subsequently, Section 4 explains the research design, and Section 5 presents the actual data analysis. Section 6 concludes.

 

++++++++++

2. Literature review

2.1. The (online) consumer purchase decision

Purchase decision-making is a crucial component of consumer purchase behavior. During the decision-making process, consumers take many different factors into consideration and process selected information about the product or service. The Internet has tremendously increased the volume of information that is available to consumers. At the same time this also makes the decision-making process more complex.

In the literature, the most widely used model is the classic consumer behavior and purchase decision model (Butler and Peppard, 1998). This model divides the purchase process into five stages: problem recognition, information search, evaluation of alternatives, choice/purchase, and post-purchase behavior.

Later models also emphasize the impact of technological factors. For example, Chen and Chang’s (2003) model of the online shopping process includes factors such as connection speed and Web site quality. Their model also highlights that a consumer may abandon a purchase purely because of technical failures of the site.

However, overall, most models assume that consumers’ evaluation and purchase decision is influenced by information from both the internal and external worlds. Seen from this angle, customer images are a potential external source of information that may help consumers in evaluating alternatives.

2.2. Word-of-mouth, eWOM, and online reputational feedback mechanisms

WOM consists of interpersonal verbal communication (Arndt, 1967). Its non-commercial nature makes it different from traditional advertising. According to von Wangenheim (2005), WOM influences not only consumers’ attitudes, preferences, and purchase intentions, but also their ultimate decision-making. In fact, WOM has been shown to have a greater impact than traditional forms of communication such as personal selling and advertisements (Trusov, et al., 2009). Consumers also find it more trustworthy (Godes and Mayzlin, 2004). Early studies focused on positive WOM (Arndt, 1967; Engel, et al., 1969), but over the years negative WOM has received more and more attention (Richins, 1983; Charlett, et al., 1995; Godes and Mayzlin, 2004; von Wangenheim, 2005; Park and Lee, 2009; Chen, et al., 2011).

eWOM, for its part, is defined as “any positive or negative statement made by potential, actual, or former customers about a product or company, which is made available to a multitude of people and institutions via the Internet” [1]. Just like traditional WOM, this new form has also been found to strongly influence consumer behavior (see Park, et al., 2007; Cheung and Thadani, 2012; Lee and Shin, 2014; as well as the references in Table 1 below).

Next to comments made on online discussion forums or on social media, online reputational feedback mechanisms — on either commercial or non-commercial Web sites — are an important form of eWOM (Dellarocas, 2003a). There are three major forms of reputational feedback mechanisms, namely ratings, text comments, and, more recently, customer images.

Prior studies indicate that ratings and comments improve the efficiency of the online marketplace. Bolton, et al. (2004), for example, find that they increase the transaction rate. Yang, et al. (2007) demonstrate that they encourage honest sellers and eliminate dishonest sellers in online auction markets. However, in spite of the fact that, as Erkan and Evans [2] point out, “[v]isually enriched contents make eWOM more enjoyable and appealing”, to date little attention has been paid to customer images as the newest form of reputational feedback mechanism. This is why we have made customer images the focal point of the present paper.

2.3. eWOM: Models and empirical results

Figure 1 presents the model of eWOM information adoption proposed by Cheung, et al. (2008), which was the main source of inspiration for the present paper. Cheung, et al., in turn, based themselves on Sussman and Siegel’s (2003) general information adoption model (IAM). In developing their model, Sussman and Siegel married elements of technology adoption and information influence theories. Where the latter is concerned, they rely on the Elaboration Likelihood Model (ELM) proposed by Petty and Cacioppo. The ELM posits that, if an individual is likely to cognitively elaborate on a communication, then the nature and quality of the arguments contained within the communication will determine the degree of informational influence. However, when an individual is unable or unwilling to expend the effort to process the arguments presented in a message, so-called ‘peripheral’ cues will play a more critical role than the ‘central’ processing route. These heuristics can be very diverse, but source credibility is an important one.

 

Model of information adoption
 
Figure 1: Model of information adoption, by Cheung, et al. (2008).
Note: Larger version of figure available here.

 

Sussman and Siegel include both argument quality and source credibility in their model, but also point out that the Technology Acceptance Model (TAM) — the dominant model in the technology adoption literature — finds that usefulness is a key construct in adoption behavior. Sussman and Siegel therefore argue that usefulness will be more strongly associated with information adoption than the ELM constructs and that “ELM processes are influential to the extent that they contribute to the perceived usefulness of the message” [3]; see Figure 1.

As mentioned, Cheung, et al. (2008) draw heavily on Sussman and Siegel’s IAM. In fact, Cheung, et al.’s contribution consists in decomposing overall ‘argument quality’ — which Sussman and Siegel [4] measure by means of a simple three-item scale — into four distinct quality dimensions, namely relevance, timeliness, accuracy, and comprehensiveness. Similarly, compared to Sussman and Siegel, Cheung, et al. split up ‘source credibility’ more explicitly into expertise and trustworthiness.

Obviously, Cheung, et al.’s model is not the only model of eWOM information adoption. As a matter of fact, there is a great variety of models. All in all, three categories of models can be discerned: models that, like Cheung, et al. (2008), stay very close to the original IAM; models that add one or more variables, but where IAM still forms the core of the model; and, third, more complex models where IAM is but one of the building blocks. An exhaustive overview of all these models is beyond the scope of our paper. However, in order to give the reader a flavor of the literature and to make our overview of the empirical results easier to understand, we discuss one example each of the second and third category.

The model of Erkan and Evans (2016a), in a recent paper on the influence of eWOM in social media, falls in the second category. It extends the IAM with elements of the Theory of Reasoned Action (TRA) originally developed by Fishbein and Ajzen (1975). Erkan and Evans’ main point — also made by Knoll (2016) — is that while the IAM takes into account the characteristics of eWOM information, one should not neglect consumers’ behavior towards eWOM. To incorporate this behavior, they prefer TRA over TAM because

“TAM mostly focuses on the individual usage of a computer, with the concept of ‘perceived usefulness’, and disregards the essential social processes of information development and implementation. Particularly in the context of eWOM, where the information is generated by separate individuals, TAM might not deliver adequate understanding of users' attitudes and intentions while TRA focuses on behavioral theories.” [5]

Specifically, TRA postulates that behavioral intention — in our case: purchase intention — is explained by attitude and subjective norms. Erkan and Evans [6] disregard subjective norms, but do add ‘attitude’ to the IAM. Another independent variable which they add is ‘needs of information’, which has primarily been examined as a motivator for WOM engagement and which other studies call ‘advice seeking’ (Hennig-Thurau, et al., 2004) or ‘opinion seeking’. The combination of the IAM building blocks with ‘attitude’ and ‘needs for information’ yields what Erkan and Evans call their Information Acceptance Model (IACM); see Figure 2 [7].

 

Erkan and Evans (2016a) research model
 
Figure 2: Erkan and Evans’ (2016a) research model.
Note: Larger version of figure available here.

 

Cheung and Thadani (2012), for their part, conducted a review of prior studies and propose an integrative — and thus substantially more complex — model of the impact of eWOM on purchase intention and, ultimately, the purchase decision. Their model not only includes directly influencing factors, but also multiple indirect and moderating factors. In particular, the model consists of five components: communicators (source), stimuli (content), receivers (audience), responses (main effects), and contextual factors. The argument quality and source credibility elements of the models of Sussman and Siegel (2003) and Cheung, et al. (2008) are aggregated, respectively, under stimuli and communicators, and the information usefulness and information adoption variables are in the responses section of the integrative model.

Table 1 provides a selective overview of (part of) the results of empirical studies on eWOM adoption. For reasons of comparability with our own results, (1) we selected only studies that rely on IAM and, one level lower, on ELM — rather than, for example, the heuristic-systemic model, which is an alternative to ELM (Zhang, et al., 2014); (2) we only looked at those parts of the models that are relevant for our purposes; that is, at the antecedents of information usefulness, which in turn affects purchase intention [8]. The numbers reported are path coefficients of structural models, with the exception of Erkan and Evans (2016b), who use regression analysis.

 

Selective overview of empirical studies that rely on IAM
 
Table 1: Selective overview of empirical studies that rely on IAM.
Note: Larger version of table available here.

 

As can be seen, most studies find that argument quality and source credibility (or their dimensions) have a significant positive impact on information usefulness. There are, however, exceptions. In their empirical test of the model in Figure 1 (for an online restaurant community in Hong Kong), Cheung, et al. (2008) find significant results for comprehensiveness and relevance but the other two quality dimensions, as well as the entire source credibility component, turned out to be not significant. In a similar study, Cheung (2014) finds that relevance does not exert a significant impact, but that trustworthiness does. And in the study of Gunawan and Huarng (2015) argument quality is only significant at the 0.10 level.

Another interesting observation is that, with the exception of Gunawan and Huarng (2015) and Jin, et al. (2009), the significance and/or the coefficient(s) of argument quality or its dimensions is higher than those of source credibility.

 

++++++++++

3. Research model and hypotheses

As noted earlier, in the present paper we wanted to test whether the IAM could be applied to the specific case of customer images on shopping sites. For a first test, we thought it wise not to opt for one of the more complex models, but to stay relatively close to the original. As explained in Section 2.3, the model of Cheung, et al. (2008) does precisely that, while at the same time having the advantage of decomposing argument quality and source credibility into several dimensions. We also decided, in our exploratory research, to focus on the quality component of the Cheung, et al. model — for two reasons. One is that, as stressed when discussing Table 1, the credibility component of their model proved insignificant in the empirical test. Second, and more importantly, we thought that the specificity of customer images, compared to other forms of eWOM, would show up in ‘argument quality’ rather than in ‘source credibility’. Indeed, eWOM users know as much (or as little) about the individuals who have posted photos on the site than about individuals who only posted text comments.

We thus decided to use the exact same dimensions of quality as in Cheung, et al. (2008); that is, relevance, timeliness, accuracy, and comprehensiveness. However, the dependent variable of our model is different. This is because Cheung, et al. (2008) only look at the information adoption stage, whereas we wanted to focus more on the purchase decision. According to Cheung and Thadani (2012), all the influencing factors in their model impact the purchase decision via purchase intention, and this impact is consistent. We therefore choose purchase intention as our dependent variable (and ignore the mediating effect of information usefulness).

Figure 3 shows the proposed research model. The independent variable is the quality of customer images, which is decomposed into four dimensions: Relevance (RELE); Timeliness (TIME); Accuracy (ACCU); and Comprehensiveness (COMP). The dependent variable is Purchase intention (PI).

 

Theoretical model: Impact of quality of customer images in purchase intention
 
Figure 3: Theoretical model: Impact of quality of customer images in purchase intention.
Note: Larger version of figure available here.

 

Relevance is the extent to which customer images are relevant, appropriate and applicable for purchase decision-making (Cheung, et al., 2008). There is too much information available on the Internet, and people’s time is limited. Images of a skirt will not be helpful for a man who is shopping for a tie. As Dunk (2004) demonstrates for the case of Information System (IS) information, relevance is a crucial determinant of decision making. Hence, our first hypothesis is:

H1: The relevance of customer images positively influences purchase intention.

Timeliness is the extent to which customer images are current, timely and up-to-date (Cheung, et al., 2008). If the information on a Web site is not up-to-date, it will fail to meet the expectations of customers (Madu and Madu, 2002). Especially for fashion products such as clothing, cosmetics, jewelry, etc., customers require quick information updating. In a U.S. study on online customer satisfaction, McKinney, et al. (2002) show that out-of-date information is considered unreliable.

H2: The timeliness of customer images positively influences purchase intention.

Accuracy is the extent to which customer images are accurate, correct, and reliable (Cheung, et al., 2008). The perceived accuracy of information determines whether customers trust it or not. Feedback posted by dishonest sellers in order to improve the rating and reputation of their online store is definitely unreliable. Only accurate feedback has the potential to help customers make a better purchase decision.

H3: The accuracy of customer images positively influences purchase intention.

Comprehensiveness is the extent to which customer images are complete and sufficient to cover customers’ needs (Cheung, et al., 2008). As mentioned in Section 2.3, Cheung, et al. (2008) found that, together with relevance, comprehensiveness of eWOM is one of the most important factors influencing information adoption.

H4: The comprehensiveness of customer images positively influences purchase intention.

Finally, based on the hypotheses above, according to which all four dimensions of the quality of customer images positively influence the online purchase intention, we propose that quality as a whole also has a positive impact.

H5: The quality of customer images positively influences purchase intention.

 

++++++++++

4. Research design

We test the theoretical model presented in Figure 3 by looking into the shopping experience of customers on Taobao.com, the most popular C2C shopping site in China. Taobao.com was launched in 2003 and was the first local e-commerce site. According to the 2016 annual report of Alibaba Group Holding Limited (2016), the parent company of Taobao, Taobao.com is the largest C2C shopping site in China in terms of gross merchandise volume. It is also the most widely known and used. At the time of our data collection, it had approximately 500 million registered users [9].

Taobao.com has all three major forms of online reputational feedback mechanisms on its Web site, namely ratings, text comments, as well as customer images. Figure 4 is an example of the comments section of a product. There is an overall rating of the product, a brief product description, and an overview of the customer comments. Users can choose the kind of comments they are interested in by selecting “good”, “neutral”, “bad”, “additional comments”, or “customer images”. The number between brackets next to “customer images” indicates the number of images available for this product.

 

Comments section of a product on Taobao.com
 
Figure 4: Comments section of a product on Taobao.com [10].
Note: Larger version of figure available here.

 

Customer images are usually posted together with short text comments, as supportive proof. For example, the comments in Figure 5 mention that the customer really likes the color of the suitcase.

 

An example of a customer image
 
Figure 5: An example of a customer image [11].

 

As mentioned in Section 1, little is known about the impact of customer images on the online purchase decision. There is a dearth of both theoretical and empirical research. This makes it difficult to determine the appropriate population for a study like ours. After due consideration, we decided to rely on judgment sampling. In order to maximize the probability that the respondents for our survey had shopping experience on Taobao.com and, crucially, had encountered customer images on this site, we decided to focus on Taobao customers in the 20–29 age group, and more in particular on university students.

The justification for this choice lies in the observation that most Chinese online shoppers are young and well-educated. The 2014 China Online Shopping Market Research Report (CNNIC, 2015) uses K-means clustering analysis to divide online shoppers into general (87.1 percent) and power shoppers (12.9 percent). For general shoppers, the average online shopping frequency is 14 times within six months, compared to 73 times for power shoppers. In both groups, around 50 percent are between 20 and 29 years old. More than 60 percent have an educational background equal to or higher than college [12].

Note also that Kwahk and Kim (2016), in their study of the effect of social media on consumers’ purchase decisions on Taobao [13], end up with a sample that consists for 87.1 percent of respondents between the ages of 20 and 30 — even though, unlike us, they did not filter on age. Kwahk and Kim argue that that “this is consistent with the fact that young people are the most active internet users in China”. They also refer to the same CNNIC report mentioned above to stress that young adults account for the largest portion of social media users. Similarly, in the sample used by Cheung [14], in a paper on information adoption in online customer communities in Hong Kong, 95 percent of the respondents are between 19 and 25 years old.

Our questionnaire totaled 24 questions, grouped in two sections. The first section consisted of eight general questions asking for basic socio-demographic information as well as respondents’ Internet skills. The second section comprised 16 questions that measured the four key constructs borrowed from Cheung, et al. (2008); that is, we asked respondents for their opinion on the relevance, timeliness, accuracy, and comprehensiveness of the customer images on Taobao.com. We obviously also measured their purchase intention on the site.

To ensure the validity of the key constructs, and for reasons of comparability with Cheung, et al., we used the exact same scales as Cheung, et al. [15] but transposed them to the case of customer images (see Table 2) [16]. Obviously, the questions also had to be translated into Chinese. The translated version of the questionnaire was pre-tested by eight respondents and their feedback was used to fine-tune the phrasing of a number of questions. In order to be able to identify respondents who filled in the questionnaire inattentively, two questions were phrased negatively. The questions also did not appear in the order of Table 2; rather they were mingled.

 

Constructs and questions (five-point Likert scales)
 
Table 2: Constructs and questions (five-point Likert scales).
Note: Larger version of table available here.

 

A professional site survey tool (www.sojump.com) was used to collect the data. The main channel of the data collection was through social media, such as WeChat, QQ, and Weibo. The survey was online for 15 days, from 28 March to 11 April 2016. A total of 259 answer sheets were returned, and 215 were usable. 44 respondents were filtered out based on the following four criteria:

  1. The respondent should have had shopping experience on Taobao.com (two had not);

  2. The respondent should have seen customer images on Taobao.com (a further six had not);

  3. The respondent’s age had to be in the 20–29 group (17 were not);

  4. After recoding the negatively phrased questions, the respondent’s answers had to be consistent in the same construct with an absolute difference of strictly less than four (19 failed this test) [17].

The low number of respondents removed on account of the first criterion underpins that Taobao.com is indeed a widely used shopping site. Another interesting observation is that the vast majority of the respondents had encountered customer images on Taobao.com. This shows that customer images have become an important form of online reputational feedback on Taobao.com, at least for the products that the target audience of our survey is interested in.

 

++++++++++

5. Data analysis

5.1. Socio-demographic data

Table 3 shows the socio-demographic statistics of the sample. By construction, all the respondents are in the 20–29 age group, and over 90 percent have a college degree or higher. To be clear, we did not explicitly filter potential respondents on educational background, but given the way the questionnaire was seeded on social media by the first author — herself a master student at the time — we did end up, as anticipated, with a sample consisting mainly of well-educated people and with a large proportion of students.

 

Socio-demographic statistics of the sample
 
Table 3: Socio-demographic statistics of the sample.

 

Apart from age and occupation, there are two other limitations to the representativeness of our sample. For one, females are overrepresented. Second, as the sample concentrates on students, the group that has a monthly budget between 1,001 and 3,000 RMB (162.96 and 488.38 USD) accounts for the largest proportion. Overall, if we use the 2014 China Online Shopping Market Research Report (CNNIC, 2015) mentioned earlier as a benchmark, the composition of our sample is more similar to general shoppers than to power shoppers (see Section 4).

We examined the impact of the limitations by means of a one-way ANOVA test. From Table 4, it can be seen that only the answers for Accuracy are significantly different across occupation groups (at the 0.05 level). Furthermore, a post hoc analysis showed that the difference is not really significant. None of the constructs thus show any significant difference across gender, educational background, monthly budget, and occupation groups.

 

Results of the one-way ANOVA test
 
Table 4: Results of the one-way ANOVA test.
Note: Larger version of table available here.

 

5.2. Reliability analysis

Table 5 displays descriptive statistics for the constructs. As can be seen, Relevance has the highest and Timeliness the lowest mean. Table 6 shows the reliability of the constructs, as measured by Cronbach’s alpha. To improve their reliability, one question each was removed from the RELE and TIME constructs (namely RELE3 and TIME2) [18]. But even then their reliability coefficients fail to reach the acceptable level.

 

Descriptive statistics of the constructs
 
Table 5: Descriptive statistics of the constructs.

 

 

Reliability of the constructs
 
Table 6: Reliability of the constructs.

 

5.3. Correlation analysis

The correlation matrix in Table 7 shows that of the five socio-demographic variables, only Internet skills is significantly positively correlated with the dependent variable, r (213) = 0.338, p < 0.01. To further explore how Purchase intention differs among respondents with different levels of Internet skills, a post hoc analysis was conducted using the Tukey HSD approach. The results show that there is no significant difference among the Very good (M = 4.00, SD = 0.61), Good (M = 3.82, SD = 0.63), and Fair (M = 3.51, SD = 0.64) groups. The difference is between the Poor or Very poor (M = 2.75, SD = 0.57) group and the other three groups. As can be seen in Table 8, the respondents who have Internet skills equal to or better than Fair clearly have a higher Purchase intention than those who are in the Poor or Very poor group.

 

Correlation matrix
 
Table 7: Correlation matrix.
Note: Larger version of table available here.

 

 

Tukey HSD of Internet skills
 
Table 8: Tukey HSDa,b of Internet skills.

 

The correlation matrix also shows that the four dimensions of Quality, as well as the overall Quality construct, are all significantly positively correlated with Purchase intention. However, there are also several high positive correlations between the four dimensions themselves. For example, the correlation between the COMP and ACCU constructs amounts to r (213) = .744, p < 0.01. As we will show in the next section, this causes a multicollinearity problem in the regression models.

5.4. Regression analysis

We now test the relationships between the independent variables and purchase intention through regression analysis. Given that Internet skills is also a potentially significant influencing factor, hierarchical regression analysis was used to compare the results of models with and without Internet skills.

 

Results of regression analysis
 
Table 9: Results of regression analysis.
Note: Larger version of table available here.

 

The first model in Table 9 simply contains the four individual quality dimensions. The model is significant overall — as can be seen from the F statistic — and it explains 20.8 percent of the variation in Purchase intention. Comprehensiveness and Relevance are significant predictors; Timeliness and Accuracy are not.

The results for Model 2 indicate that Internet skills is indeed significant, and that the model is improved by its addition: Model 2 explains 28.8 percent of the variation in Purchase intention. Comprehensiveness remains significant, but the significance of Relevance disappears.

A more general observation is that whereas individually Relevance, Timeliness, Accuracy, Comprehensiveness, and Internet skills were all found to be significantly correlated with Purchase intention (see Table 7), some are not significant anymore when entered into regressions together with other variables. One possible reason is the low reliability of the RELE and TIME constructs (see Section 5.2); another is the high degree of multicollinearity between the independent variables.

To examine the impact of multicollinearity, we conducted a robustness check (not reported in Table 9) in which Comprehensiveness was removed from Model 2. Internet skills remained significant (β = .218, t (209) = 4.457, p < 0.05) and Timeliness remained non-significant (β = -.020, t (209) = -.280, p > 0.05). However, Relevance and Accuracy became significant (β = .186, t (209) = 2.425, p < 0.05; β = .220, t (209) = 2.898, p < 0.05). The overall picture that emerges from our regressions is therefore that while it is clear that quality has a significant impact on Purchase intention, because of multicollinearity we have difficulties in determining which dimensions matter and which do not.

In another robustness check, we estimated two additional models — Models 3 and 4 — with the overall measure of Quality instead of the individual dimensions. As mentioned before, the reliability of the Quality construct is good. Both Model 3 and Model 4 are significant at the 0.01 level (see Table 9), and Quality positively influences Purchase intention, again at the 0.01 level. As in Model 2, the addition of Internet skills increases the explanatory power of the model.

 

++++++++++

6. Conclusions

6.1. Discussion

This paper aimed to explore the impact of the quality of customer images on the online purchase decision. The theoretical model was adapted from the model of eWOM information adoption developed by Cheung, et al. (2008). The quality of customer images was measured in four dimensions: Relevance, Timeliness, Accuracy, and Comprehensiveness. Empirical data was collected through an online questionnaire among young users of the Chinese C2C shopping site Taobao.

In the regression analyses, we examined the significance of the four dimensions separately (in Models 1 and 2) as well as of Quality as a whole (in Models 3 and 4). Both models were significant. Interestingly, the results of Model 1 are consistent with the findings of Cheung, et al. (2008), who also find — for the case of eWOM on an online consumer community — that only Comprehensiveness and Relevance are significant precursors of information adoption, and that Comprehensiveness is the more significant and influential of the two. However, when interpreting our results one has to factor in the presence of multicollinearity. As mentioned in Section 5.4, after Comprehensiveness was removed from Model 2, Relevance and Accuracy became significant.

Still, overall, our regression results do indicate that the quality of customer images has a significant impact on the purchase intention of Taobao customers between the age of 20 and 29. In other words, H5 is accepted. Given that the literature has demonstrated that the impact of eWOM on purchase intention and purchase decision is consistent (see Section 3), we can also tentatively conclude that the quality of customer images has a significant impact on the purchase decision of Chinese young online shoppers.

On a lower level, determining which dimensions of quality matter is less straightforward, because of the multicollinearity. However, Comprehensiveness would seem to be the most important dimension because it invariably has the highest coefficient and is always significant at the 0.01 level (whereas Relevance and Accuracy are only significant at the 0.05 level). Timeliness, for its part, does not seem very important because it is never significant. However, here (part of) the explanation might lie in the low reliability of the construct.

As an aside, in Models 2 and 4 Internet skills was found to be significant, and the models were improved by its addition. Internet skills is thus a vital control variable when explaining Purchase intention: the better the consumer’s Internet skills, the higher the purchase intention.

6.2. Implications

This paper is a first attempt to fill the existing research gap on customer images with novel empirical findings, and contributes to the research on purchase decision making, eWOM, and online reputational feedback mechanisms.

Our paper shows that the argument quality part of the eWOM information adoption model proposed by Cheung, et al. (2008) for online customer communities remains broadly effective when applied to photo sharing on shopping sites. Our empirical results are, at least at first sight, fully in line with Cheung, et al., who found that the Relevance and especially the Comprehensiveness of eWOM significantly influence information adoption. However, from a theoretical-methodological perspective, the limited reliability of two of the quality dimensions (namely Relevance and Timeliness) is worrying. This might be due to the specific scales used by Cheung, et al. (2008), but may very well be an indication that there is a need for a dedicated theoretical framework and for scales that take into account the specific nature of customer images. Simply transposing scales that have been validated for other forms of eWOM would not appear to be the best approach.

Second, in terms of managerial implications, our findings highlight the impact of post-purchase behavior and suggest that marketers should encourage customers to post images of the products they have bought. Sites would also appear to benefit from putting in place mechanisms that can recognize — and, why not, offer small rewards for — the high-quality images. Note that Amazon, for example, enables users to give their opinion on the comments of other users by answering the question: “Was this review helpful to you?” This function can help to prevent feedback from being manipulated. In the future, a more advanced form of reputational feedback may be video, because videos are more vivid and contain richer information than images.

6.3. Limitations and future research

Our study has a number of limitations. The first is inherited from the original research model. As stated by Cheung, et al. (2008), their model was intentionally simplified. In other words, it can be improved by taking more variables into consideration. This holds a fortiori for our research model.

The second limitation concerns the survey sample. There is an imbalance in the gender ratio, and the sample mainly consists of university students. This is a limitation that we largely share with Kwahk and Kim’s (2016) study on Taobao, who have a similar sample in terms of both scope and composition (215 vs. 225 respondents; 100 vs. 87.1 percent young adults). We can/should thus make the same remark: “Although most social media users and online consumers in China are relatively young [...], future research should re-test the hypotheses using data from a more general population of online consumers to enhance the generalizability of the results” [19]. Also, again just like Kwahk and Kim, while Taobao is the largest Chinese e-commerce site, we have to acknowledge that our data relate to a single site.

Third, we measured respondents’ opinions about customer photos of products on Taobao.com in general, whereas — in line with Park and Lee (2009) [20] — the impact of customer images might well differ between product types.

The final limitation concerns the data analysis. Some of the results must be treated with care, as the reliability of the RELE and TIME constructs is relatively low. Additionally, the regression models suffer from multicollinearity.

As hinted at above, future research could use alternative existing scales for argument quality and its dimensions and, better still, could try to come up with scales that are tailored to the specific case of customer images. Future research could also test the model in a B2C or in a social media context. End of article

 

About the authors

Mingming Dai Mingming Dai is a recent graduate with a M.Sc. in management from the Faculty of Economic and Social Sciences and Solvay Business School at Vrije Universiteit Brussel.
E-mail:mingming [dot] dai [at] vub [dot] ac [dot] be

Leo Van Hove is a Professor of Economics at the Vrije Universiteit Brussel (Free University of Brussels), where he teaches courses in monetary economics and the economics of information. His current research interests include payment instruments, network effects, e-commerce, and access to finance. He has published extensively on these and other subjects in international journals as diverse as Journal of Money, Credit, and Banking, Review of Industrial Organization, International Journal of Electronic Commerce, Economic Modelling, Energy Economics, European Journal of Operational Research, Journal of Media Economics, and Information Economics and Policy.
E-mail: Leo [dot] Van [dot] Hove [at] vub [dot] ac [dot] be

 

Acknowledgements

The authors would like to thank Ellen Van Droogenbroeck for comments on an earlier version.

 

Notes

1. Hennig-Thurau, et al., 2004, p. 39.

2. Erkan and Evans, 2016a, p. 48.

3. Sussman and Siegel, 2003, p. 51.

4. Sussman and Siegel, 2003, p. 63.

5. Erkan and Evans, 2016a, p. 48.

6. Erkan and Evans, 2016a, p. 49.

7. In another paper, presumably written earlier, Erkan and Evans (2016b) apply the IAM as is — that is, they do not extend it — but do not treat information quality and information credibility as antecedents of information usefulness and/or information adoption. They simply check for a direct impact on purchase intention; see their Figure 1.

8. There is one exception. As explained earlier, Erkan and Evans (2016b) analyze the direct impact of information quality and credibility on purchase intention.

9. Taobao.com, at https://www.taobao.com/about/?spm=a21bp.7806943.2015082601.40.LL0jlN, accessed 18 February 2016.

10. Taobao.com, at http://world.taobao.com/item/44623603905.htm?spm=a312a.7700714.0.0.u9cfhT#detail, accessed 25 February 2016 (no longer available).

11. Taobao.com, at http://world.taobao.com/item/521736272032.htm?spm=a312a.7700714.0.0.u9cfhT#detail, accessed 25 February 2016 (no longer available).

12. Regarding occupation, 19.8 percent of the general shoppers and 7.9 percent of the power shoppers are students. As for monthly income, the group that has a monthly income between 1,001 and 3,000 RMB (163 and 488 USD) accounts for the largest proportion among general shoppers; among power shoppers it is the 3,001–5,000 RMB (488–816 USD) income group.

13. To be clear: Kwahk and Kim’s paper does not appear in our literature review because it is of an entirely different inspiration than ours and because they (thus) use a totally different research model.

14. Cheung, 2014, p. 49, Table 2.

15. Cheung, et al., 2008, p. 239, Table III.

16. The references in Table 2 indicate from which papers Cheung, et al. (2008) borrowed their scales.

17. To be clear: we worked with five-point Likert scales.

18. The initial α for the Relevance and Timeliness constructs was .493 and .366, respectively.

19. Kwahk and Kim, 2016, p. 21.

20. Park and Lee find that the eWOM effect is greater for experience goods than for search goods.

 

References

Alibaba Group Holding Limited, 2016. “Annual report pursuant to section 13 or 15(d) of the Securities Exchange Act of 1934 for the fiscal year ended March 31, 2016,” at http://www.alibabagroup.com/en/ir/pdf/agm160524_ar.pdf, accessed 19 June 2017.

Johan Arndt, 1967. Word of mouth advertising: A review of the literature. New York: Advertising Research Foundation.

Gary E. Bolton, Elena Katok, and Axel Ockenfels, 2004. “How effective are electronic reputation mechanisms? An experimental investigation,” Management Science, volume 50, number 11, pp. 1,587–1,602.
doi: http://dx.doi.org/10.1287/mnsc.1030.0199, accessed 5 September 2017.

Patrick Butler and Joe Peppard, 1998. “Consumer purchasing on the Internet: Processes and prospects,” European Management Journal, volume 16, number 5, pp. 600–610.
doi: http://dx.doi.org/10.1016/S0263-2373(98)00036-X, accessed 5 September 2017.

Don Charlett, Ron Garland, and Norman Marr, 1995. “How damaging is negative word of mouth?” Marketing Bulletin, volume 6, number 1, pp. 42–50, and at http://marketing-bulletin.massey.ac.nz/V6/MB_V6_N1_Charlett.pdf, accessed 5 September 2017.

Su-Jane Chen and Tung-Zong Chang, 2003. “A descriptive model of online shopping process: Some empirical results,” International Journal of Service Industry Management, volume 14, number 5, pp. 556–569.
doi: http://dx.doi.org/10.1108/09564230310500228, accessed 5 September 2017.

Yubo Chen, Qi Wang, and Jinhong Xie, 2011. “Online social interactions: A natural experiment on word of mouth versus observational learning,” Journal of Marketing Research, volume 48, number 2, pp. 238–254.
doi: http://dx.doi.org/10.1509/jmkr.48.2.238, accessed 5 September 2017.

Ronnie Cheung, 2014. “The influence of electronic word-of-mouth on information adoption in online customer communities,” Global Economic Review, volume 43, number 1, pp. 42–57.
doi: http://dx.doi.org/10.1080/1226508X.2014.884048, accessed 5 September 2017.

Christy M.K. Cheung and Dimple R. Thadani, 2012. “The impact of electronic word-of-mouth communication: A literature analysis and integrative model,” Decision Support Systems, volume 54, number 1, pp. 461–470.
doi: http://dx.doi.org/10.1016/j.dss.2012.06.008, accessed 5 September 2017.

Christy M.K. Cheung, Matthew K.O. Lee, and Neil Rabjohn, 2008. “The impact of electronic word-of-mouth: the adoption of online opinions in online customer communities,” Internet Research, volume 18, number 3, pp. 229–247.
doi: http://dx.doi.org/10.1108/10662240810883290, accessed 5 September 2017.

China interNet Network Information Center (CNNIC), 2017. “39th China statistical report on Internet development,” at http://www.cnnic.net.cn/hlwfzyj/hlwxzbg/hlwtjbg/201701/P020170123364672657408.pdf, accessed 20 June 2017.

China interNet Network Information Center (CNNIC), 2016. “2015 China online shopping market research report,” at http://www.cnnic.net.cn/hlwfzyj/hlwxzbg/dzswbg/201606/P020160721526975632273.pdf, accessed 1 July 2016.

China interNet Network Information Center (CNNIC), 2015. “2014 China online shopping market research report,” at http://www.cnnic.cn/hlwfzyj/hlwxzbg/dzswbg/201509/P020150909354828731159.pdf, accessed 7 January 2016.

Chrysanthos Dellarocas, 2004. “Building trust online: The design of robust reputation,” In: Georgios Doukidis, Nikolaos Mylonopoulos, and Nancy Pouloudi (editors). Social and economic transformation in the digital era. Hershey, Pa.: Idea Group Publishing.
doi: http://dx.doi.org/10.4018/978-1-59140-158-2.ch007, accessed 5 September 2017.

Chrysanthos Dellarocas, 2003a. “The digitization of word-of-mouth: Promise and challenges of online feedback mechanisms,” Management Science, volume 49, number 10, pp. 1,407–1,424.
doi: http://dx.doi.org/10.1287/mnsc.49.10.1407.17308, accessed 5 September 2017.

Chrysanthos Dellarocas, 2003b. “Efficiency through feedback-contingent fees and rewards in auction marketplaces with adverse selection and moral hazard,” EC ’03: Proceedings of the Fourth ACM Conference on Electronic Commerce, pp. 11–18.
doi: http://dx.doi.org/10.1145/779928.779930, accessed 5 September 2017.

Chrysanthos Dellarocas and Charles A. Wood, 2008. “The sound of silence in online feedback: Estimating trading risks in the presence of reporting bias,” Management Science, volume 54, number 3, pp. 460–476.
doi: http://dx.doi.org/10.1287/mnsc.1070.0747, accessed 5 September 2017.

Alan S. Dunk, 2004. “Product life cycle cost analysis: The impact of customer profiling, competitive advantage, and quality of IS information,” Management Accounting Research, volume 15, number 4, pp. 401–414.
doi: http://dx.doi.org/10.1016/j.mar.2004.04.001, accessed 5 September 2017.

James F. Engel, Robert J. Kegerreis, and Roger D. Blackwell, 1969. “Word-of-mouth communication by the innovator,” Journal of Marketing, volume 33, number 3, pp. 15–19.
doi: http://dx.doi.org/10.2307/1248475, accessed 5 September 2017.

Ismail Erkan and Chris Evans, 2016a. “The influence of eWOM in social media on consumers’ purchase intentions: An extended approach to information adoption,” Computers in Human Behavior, volume 61, pp. 47–55.
doi: http://dx.doi.org/10.1016/j.chb.2016.03.003, accessed 5 September 2017.

Ismail Erkan and Chris Evans, 2016b. “Social media or shopping websites? The influence of eWOM on consumers’ online purchase intentions,” Journal of Marketing Communications (25 May), pp. 1–17.
doi: http://dx.doi.org/10.1080/13527266.2016.1184706, accessed 5 September 2017.

Marin Fishbein and Icek Ajzen, 1975. Belief, attitude, intention, and behavior: An introduction to theory and research. Reading, Mass.: Addison-Wesley.

David Godes and Dina Mayzlin, 2004. “Using online conversations to study word-of-mouth communication,” Marketing Science, volume 23, number 4, pp. 545–560.
doi: http://dx.doi.org/10.1287/mksc.1040.0071, accessed 5 September 2017.

Dedy Darsono Gunawan and Kun-Huang Huarng, 2015. “Viral effects of social network and media on consumers’ purchase intention,” Journal of Business Research, volume 68, number 11, pp. 2,237–2,241.
doi: http://dx.doi.org/10.1016/j.jbusres.2015.06.004, accessed 5 September 2017.

Thorsten Hennig-Thurau, Kevin P. Gwinner, Gianfranco Walsh, and Dwayne D. Gremler, 2004. “Electronic word-of-mouth via consumer-opinion platforms: What motivates consumers to articulate themselves on the Internet?” Journal of Interactive Marketing, volume 18, number 1, pp. 38–52.
doi: http://dx.doi.org/10.1002/dir.10073, accessed 5 September 2017.

Xiao-Ling Jin, Christy M.K. Cheung, Matthew K.O. Lee, and Hua-Ping Chen, 2009. “How to keep members using the information in a computer-supported social network,” Computers in Human Behavior, volume 25, number 5, pp. 1,172–1,181.
doi: http://dx.doi.org/10.1016/j.chb.2009.04.008, accessed 5 September 2017.

Johannes Knoll, 2016. “Advertising in social media: A review of empirical evidence,” International Journal of Advertising, volume 35, number 2, pp. 266–300.
doi: http://dx.doi.org/10.1080/02650487.2015.1021898, accessed 5 September 2017.

Kee-Young Kwahk and Byoungsoo Kim, 2016. “Effects of social media on consumers’ purchase decisions: Evidence from Taobao,” Service Business (31 December).
doi: http://dx.doi.org/10.1007/s11628-016-0331-4, accessed 5 September 2017.

Eun-Ju Lee and Soo Yun Shin, 2014. “When do consumers buy online product reviews? Effects of review quality, product type, and reviewer’s photo,” Computers in Human Behavior, volume 31, pp. 356–366.
doi: http://dx.doi.org/10.1016/j.chb.2013.10.050, accessed 5 September 2017.

Christian N. Madu and Assumpta A. Madu, 2002. “Dimensions of e-quality,” International Journal of Quality & Reliability Management, volume 19, number 3, pp. 246–258.
doi: http://dx.doi.org/10.1108/02656710210415668, accessed 5 September 2017.

Vicki McKinney, Kanghyun Yoon, and Fatemeh M. Zahedi, 2002. “The measurement of web-customer satisfaction: An expectation and disconfirmation approach,” Information Systems Research, volume 13, number 3, pp. 296–315.
doi: http://dx.doi.org/10.1287/isre.13.3.296.76, accessed 5 September 2017.

Cheol Park and Thae Min Lee, 2009. “Information direction, website reputation and eWOM effect: a moderating role of product type,” Journal of Business Research, volume 62, number 1, pp. 61–67.
doi: http://dx.doi.org/10.1016/j.jbusres.2007.11.017, accessed 5 September 2017.

Do-Hyung Park, Jumin Lee, and Ingoo Han, 2007. “The effect of on-line consumer reviews on consumer purchasing intention: The moderating role of involvement,” International Journal of Electronic Commerce, volume 11, number 4, pp. 125–148.
doi: http://dx.doi.org/10.2753/JEC1086-4415110405, accessed 5 September 2017.

Marsha L. Richins, 1983. “Negative word-of-mouth by dissatisfied consumers: A pilot study,” .Journal of Marketing, volume 47, number 1, pp. 68–78.
doi: http://dx.doi.org/10.2307/3203428, accessed 5 September 2017.

Stephanie Watts Sussman and Wendy Schneier Siegal, 2003. “Informational influence in organizations: An integrated approach to knowledge adoption,” Information Systems Research, volume 14, number 1, pp. 47–65.
doi: http://doi.org/10.1287/isre.14.1.47.14767, accessed 5 September 2017.

Michael Trusov, Randolph E. Bucklin, and Koen Pauwels, 2009. “Effects of word-of-mouth versus traditional marketing: Findings from an Internet social networking site,” Journal of Marketing, volume 73, number 5, pp. 90–102.
doi: http://dx.doi.org/10.1509/jmkg.73.5.90, accessed 5 September 2017.

Florian von Wangenheim, 2005. “Postswitching negative word of mouth,” Journal of Service Research, volume 8, number 1, pp. 67–78.
doi: http://dx.doi.org/10.1177/1094670505276684, accessed 5 September 2017.

Jian Yang, Xiaorui Hu, and Han Zhang, 2007. “Effects of a reputation feedback system on an online consumer-to-consumer auction market,” Decision Support Systems, volume 44, number 1, pp. 93–105.
doi: http://dx.doi.org/10.1016/j.dss.2007.03.005, accessed 5 September 2017.

Kem Z.K. Zhang, Sesia J. Zhao, Christy M.K. Cheung, Matthew K.O. Lee, 2014. “Examining the influence of online reviews on consumers’ decision-making: A heuristic–systematic model,” Decision Support Systems, volume 67, pp. 78–89.
doi: http://dx.doi.org/10.1016/j.dss.2014.08.005, accessed 5 September 2017.

 


Editorial history

Received 22 November 2016; revised 4 July 2017; accepted 5 September 2017.


Creative Commons License
This paper is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

The impact of customer images on online purchase decisions: Evidence from a Chinese C2C Web site
by Mingming Dai and Leo Van Hove.
First Monday, Volume 22, Number 10 - 2 October 2017
http://www.firstmonday.dk/ojs/index.php/fm/article/view/7120/6545
doi: http://dx.doi.org/10.5210/fm.v22i110.7120





A Great Cities Initiative of the University of Illinois at Chicago University Library.

© First Monday, 1995-2017. ISSN 1396-0466.