The customers’ degree of involvement affects their purchase decisions (Laurent & Kapferer, 1985; Pieniak, Verbeke, Scholderer, Brunsø, & Olsen, 2008). To further understand the behaviour of consumers when making decisions, scholars have studied their involvement in various industries, such as banking (Sanchez-Franco, 2009), food (Pieniak et al., 2008), wine (Rahman & Reynolds, 2015) and tourism (Ferns & Walls, 2012). From the perspectives of different industries, these studies draw a full-scale map of the relationships between involvement and several factors, such as satisfaction, trust, commitment and positive mood. Their findings offer managerial implications for marketers who attempt to incorporate involvement into their marketing strategies.

Involvement, which exhibits relevance to personal values (Barki & Hartwick, 1994; Bloch & Richins, 1983), is an individual, internal interest of intensity, direction and persistence toward specific products (Andrews, Durvasula, & Akhter, 1990). Although many papers have classified involvement into various types, product and purchase involvement have been widely acknowledged as its two main categories (Belanche, Flavián, & Pérez-Rueda, 2017; Bloch & Richins, 1983; Lockshin, Spawton, & Macintosh, 1997; Vanwesenbeeck, Walrave, & Ponnet, 2016). Product involvement (or enduring involvement) is a feeling of interest, enthusiasm and excitement about specific product categories that consumers build over the long term (Houston & Rothschild, 1978), while purchase involvement (or situational involvement) is stimulated by a particular situation, such as a purchase occasion, and prompts pre-purchase internal or external information searches to reduce the perceived risks associated with the selection of products or services (Bloch & Richins, 1983; Houston & Rothschild, 1978).

Consumer trust in products or service providers is a major factor that fuels the rapid development of e-commerce (Gefen, 2000). Given their inability to confirm the quality of a product in person, consumers have to trust the entire e-commerce mechanism to determine the transactions that follow (Lee & Turban, 2001). Customer trust is a psychological state that allows consumers to accept their own vulnerability and take risks based on their positive expectations regarding the intentions and behaviours of the other party (Singh & Sirdeshmukh, 2000). Grabner–Kräuter and Kaluscha (2003) performed an integrative review of 11 empirical studies on trust in e-commerce and found that most of these studies have explored the consumers’ trust in online retailers or retail websites. Online retailers are stores which sell products or services via the Internet, while retail websites are online portals that allow online businesses to manage their website, marketing, sales and operations. Many authors highlighted brand trust as another important research object that can help customers reduce their uncertainty and simplify their choices of online shopping (Plassmann, Kenning, Deppe, Kugel, & Schwindt, 2008). Even though many e-merchants are unknown to consumers, they still engage in business online because e-service providers play an intermediary role in guaranteeing customers’ equity. Moreover, brands, public images conceived of as something to be marketed, can shape some consumers’ perceptions toward products. When product information is accessed from unfamiliar retailers, the brand name can increase the customer’s confidence in claims (Delgado–Ballester & Munuera-Alemán, 2001). Trust in online shopping must be conceptualised as a separate construct that functions differently in the e-commerce mechanism with an aim to reduce uncertainty.

The inherent nature of perceived risks somehow bridges the relationship between involvement and trust in the e-commerce context. Sanchez-Franco (2009) investigated the moderating effect of involvement on the relationship between trust and commitment and argued that when product involvement exists, the importance of trust is reduced. By contrast, high purchase involvement results from a high perceived risk, thereby enhancing the relevance of trust. People tend to make trust-related assumptions about others based on what they know (Riquelme, Román, Cuestas, & Iacobucci, 2019), but in an uncertain situation, people confirm their beliefs through experiential evidence (Ouyang, Gursoy, & Chen, 2019), such as their perceptions toward a website. After identifying their needs, those customers with a high purchase involvement engage in an information search to avoid uncertainty. They feel vulnerable in the unfamiliar online world because of their lack of confidence in their ability to purchase their desired product online (Sanchez-Franco, 2009). Therefore, those purchase-involved consumers who temporarily increase the relevance or their interest toward a product or service tend to seek information from formal or informal sources to evaluate their alternatives and ponder over their purchase decisions. These temporal risk-reduction activities may reinforce the customers’ trust in any trading party in e-commerce (Sanchez-Franco, 2009). As their degree of purchase involvement increases, these customers’ level of trust in e-retailers, brands being sold online and e-service providers increases. The following hypotheses are proposed:

On the other hand, familiarity with products can make consumers feel secure when shopping online because they know how to identify useful information and where to find bargains (Kim et al., 2008). Product involvement allows customers to develop expectations regarding the trading parties (Sanchez-Franco, 2009). Therefore, given their rich experience or knowledge about e-commerce, those consumers with a high product involvement build their faith in others’ benevolence, competence and integrity, which consequently develops their trust in online businesses (McKnight, Choudhury, & Kacmar, 2002). The following hypotheses are proposed:

Consumer confusion is a mental state with conscious and unconscious elements in which a customer struggles when making purchase decisions (Moon, Costello, & Koo, 2016). Consumer confusion can occur before or after the purchase (Mitchell, Walsh, & Yami, 2005) and can be viewed as the failure to develop an accurate interpretation of the various facets of a product or service during the information processing procedure (Turnbull, Leek, & Ying, 2000). Three types of confusion, namely, similarity, overload and ambiguity have been identified from the literature (Turnbull et al., 2000; Walsh et al., 2007). Similarity, confusion refers to the consumers’ propensity to think that different products in a product category are similar with the brand names, attributes or quality (Walsh et al., 2007). Overload confusion takes place when individuals are bombarded with an abundance of information (Mitchell et al., 2005). Ambiguity confusion occurs when individuals are forced to re-evaluate and revise current beliefs or assumptions about the product or purchasing environment (Mitchell et al., 2005). These three types of confusion not only take place when shopping in physical stores but also when shopping online (Mitchell et al., 2005). Browsing an excessive number of similar websites or reading excessively similar product information from multiple sources may engender similarity confusion; introducing newly updated online payment technologies or assorted product combinations at varying prices can trigger ambiguity confusion; and providing a vast amount of information may cause overload confusion (Walsh et al., 2007).

Consumer confusion has long been considered an important topic in the consumer behaviour literature (Mitchell et al., 2005; Walsh & Mitchell, 2010; Walsh et al., 2007). Understanding consumer confusion can help businesses improve their marketing strategies and establish sustainable and profitable relationships with their target consumers (Turnbull et al., 2000). Most papers on consumer confusion have investigated its antecedents and consequences (Cheng, Lu, & Gursoy, 2015; Tjiptono, Arli, & Bucic, 2014; Walsh & Mitchell, 2010; Walsh et al., 2007; Wobker, Eberhardt, & Kenning, 2015). Among those consequences, marketplace trust has been verified to be significantly decreased by consumer’s confusion (Cheng et al., 2015; Walsh & Mitchell, 2010; Walsh et al., 2007).

However, only a few papers have discussed the moderating effects of consumer confusion. In the complex online shopping environment, this study examines whether consumer confusion can alter the strength or direction of the relationships between distinct involvement types and specific trust objects (Baron & Kenny, 1986). Product-involved consumers who adopt a long-term uncertainty-reduction strategy are less susceptible to the influence of a confusing context. Conversely, purchase-involved consumers often spend time and effort in seeking or processing information to evaluate their alternatives during the purchase process (Sanchez-Franco, 2009); this type of situational involvement is easily influenced by product attributes (e.g., product complexity and similarity) and situational variables (e.g., asymmetry in the availability of information related to decision-making) (Houston & Rothschild, 1978). Therefore, a complex and unfamiliar shopping environment places the consumers’ purchase decisions in a highly confusing situation. If these consumers feel confused when shopping online, then the effect of their purchase involvement on their trust may be reduced. On the basis of the above discussion, we formed the following hypotheses:

This study developed a two-section questionnaire based on the literature review. The first section consists of three constructs and eight dimensions: consumer confusion (i.e., similarity, overload and ambiguity), product involvement (i.e., product and purchase) and trust (i.e., in e-vendor, retail website and brand). The consumer confusion scale was adapted from Walsh et al. (2007), while product involvement and purchase involvement were measured by using scales from Sanchez-Franco (2009) and Lockshin et al. (1997). The respondents’ trust was measured by using three dimensions, namely, trust in e-vendors, trust in retail websites and trust in brands, with scales adapted from Bhattacherjee (2002), Gefen (2000), Delgado–Ballester and Munuera-Alemán (2001) and Jarvenpaa, Tractinsky and Vitale (2000) (see Table 1-A1, Appendix 1 for details). The second section gathers the demographic information of the respondents, including their gender, age, education, marital status, years of online shopping and monthly personal income. All items in the constructs were measured by using a seven-point Likert-type scale.

The English measurement items were translated into Chinese and then translated back into English by a bilingual expert to ensure translation accuracy. The survey instrument was then pilot-tested among 35 randomly selected college students to ensure its validity and reliability. In China, most college students are active Internet shoppers (Huang, Zhou, Liao, Mo, & Wang, 2017), which justifies the pilot-test sampling. Based on the results of reliability analysis and item analysis, some measure items were deleted from the original scales to reach better scale consistency.

Since the population this study focused on the Internet shoppers, we applied the online survey as our sampling method. The online survey, which was run by a professional opinion pollster, randomly selected participants from online users in Guangdong Province for 3 months. According to the Alibaba Group (China Business Intelligence Network, 2018), e-retailing sales in Guangdong Province accounted for 25% of national sales and exceeded those in other parts of China in 2017, thereby enhancing the representativeness of the collected data. Those respondents who had never shopped online were excluded from the interviews. A total of 598 participants completed the questionnaire, among which 28 were excluded because they were inconsistent or had too many missing data, thereby leaving 570 effective samples for the analysis. The sample included 263 men (46.1%) and 307 women (53.9%). Respondents aged below 33 years and with a bachelor’s degree or above accounted for 82.4% and 67.5% of the sample respectively. Almost all respondents (92.8%) had more than 2 years of online shopping experience, and most of them (64.9%) were earning below 5000 RMB every month.

This article followed all ethical standards for carrying out research without direct contact with human or animal subjects.

Maximum likelihood estimation was conducted to estimate the confirmatory and structural equation models. The multi-dimensional factor of consumer confusion was tested before evaluating the fitness of the entire structural model. Confirmatory factor analysis was performed first to test the model fit of three second-order components (i.e., similarity, overload and ambiguity confusion) and one first-order construct (i.e., consumer confusion). As shown in Tables 1 and 2, all factor loadings (i.e., standardised coefficients) are above 0.6. The Cronbach’s α and CR of the second-order components all exceed 0.7, while their average variance extracted (AVE) meet the suggested level of 0.5, thereby confirming the convergent validity and internal consistency of the measurement scales (Fornell & Larcker, 1981; Hair, Black, Babin, & Anderson, 2010). Discriminant validity was also achieved because the inter-factor correlations in the corresponding rows and columns were less than the square roots of AVE on the main diagonal (Fornell & Larcker, 1981). The fit indices showed a favourable overall fit for the second-order model (Hair et al., 2010). Consequently, the items concerning the three types of consumer confusion were averaged dimensionally to obtain three indicators of the main construct.

The proposed model was tested after performing second-order CFA on consumer confusion. An item-parcelling strategy was applied across the entire model to obtain highly reliable indicators, to estimate fewer parameters and to simplify the model interpretation (Hau & Marsh, 2004). Confirmatory factor analysis was also conducted to examine whether the collected data fit the hypothesised measurement model (i.e., consumer confusion, product involvement, purchase involvement, trust in e-vendor, trust in retail website and trust in brand). As shown in Tables 3 and 4, a favourable model fit is achieved because the χ²/d.f. ratio is less than 3, the values of NFI, GFI, CFI and TLI are all above 0.90 and the RMSEA is below 0.08 (Hair et al., 2010). The convergent validity and internal consistency of the measurement scales were also confirmed because almost all values for the factor loadings, Cronbach’s α, and CR exceeded 0.7 while their AVEs met the suggested level of 0.5. Similarly, the test for discriminant validity presented a satisfactory outcome because the square root AVE estimates of a single construct were greater than the inter-construct correlations.

SEM regression paths were constructed to test the proposed hypotheses in the conceptual model as shown in Figure 1. Purchase involvement is negatively related to trust in e-vendors (β = −1.187, p < 0.001), retail websites (β = −1.238, p < 0.001) and brands (β = −1.332, p < 0.001), thereby not supporting H1a, H1b and H1c. Meanwhile, consumers with a high purchase involvement exhibit low trust in e-commerce. Hypothesis 2, which predicts a positive relationship between product involvement and trust in e-business, was also verified by the results concerning the e-vendors (β = 3.034, p < 0.001), trust in retail websites (β = 2.920, p < 0.05) and trust in brands (β = 3.109, p < 0.001). Therefore, H2a, H2b and H2c are all supported. In other words, as the consumers’ product involvement increases, they tend to trust retailers, retail websites and brands while shopping online.

A quadratic latent variable with a single indicant was applied to test the moderating effect of consumer confusion. Although all possible pairwise products of the main effect were recommended as indicators of the latent product to examine the interaction (Kenny & Judd, 1984), unfortunately, the product terms violate the multivariate normality of the maximum likelihood assumption (Cortina, Chen, & Dunlap, 2001; Ping, 1995). Given this problem, many authors have proposed a single indicator of the latent product (Jöreskog & Yang, 1996; Mathieu, Tannenbaum, & Salas, 1992; Ping, 1995). The moderating effects of consumer confusion was tested by using a single indicant specification, and the results are also shown in Figure 1. Consumer confusion demonstrates a significant and negative moderating effect on the negative relationship between purchase involvement and trust in e-vendors (β = −0.153, p < 0.01). By contrast, the moderating effects of consumer confusion on the relationships between purchase involvement and trust in retail websites (β = −0.047, ns) and trust in brands (β = −0.032, ns) fail to reach a significant level. Therefore, H3a is supported, while H3b and H3c are rejected. The interacting effect of purchase involvement and confusion on trust in e-vendors is illustrated in Figure 2. The plot of interaction shows that trust in e-vendors decreases along with purchase involvement; however, this slope descends more sharply for higher confusion. This finding indicates that high purchase-involved and confused consumers tend to have lower trust in e-vendors.