Review of Study Findings

Studies on Web Use

The Internet has evolved to become a technology that serves multiple needs. Users can access various types of services (such as news, e-banking, information search, etc.). Studies that evaluated use behavior focused on actual use or willingness to use these services. Some studies assessed use of the Internet in general, without contextualizing use for a specific service. The predictors of the use behavior can be segmented into user characteristics, user perceptions, and the social context of the user ( Table 3 ).

User Characteristics

Two dominant aspects within user characteristics that have been subjected to empirical analysis are demographic variables and psychographic variables. The demographic variables investigated by studies as predictors of Internet use included race, gender, generation, and culture. The findings supported the notion that the white population used the Internet more than minorities, males were marginally heavier users than females, and subjects younger than 19 years of age displayed a much higher usage behavior (Kraut et al., 1999). Culture (subjects in the U.S. and Hong Kong) not only impacted the use behavior but also influenced the underlying purpose of the use (Chau et al., 2002). The subjects in the U.S. were found to be more oriented toward using the Internet for commerce and entertainment, while subjects in Hong Kong primarily used the Internet for hobbies and social communication. In terms of psychographics, researchers have found that personal innovativeness, playfulness, and computer skill were distal determinants of use, achieving their effects through ease of use and usefulness (Agarwal & Karahanna, 2000; Agarwal & Prasad, 1998; Kraut et al., 1999; Moon & Kim, 2001).

User Perceptions

User perceptions were widely used as the main variables of interest in a variety of studies. User perceptions regarding lack of data security, instability of the system, information content and accuracy, responsiveness, download delay, navigation, interactivity, system design quality, ease of use, and usefulness were found to be significant predictors of use behavior (Agarwal & Venkatesh, 2002; Han & Noh, 2000; Liao & Cheung, 2002; Liu & Arnett, 2000; Moon & Kim, 2001; Palmer, 2002). In addition, it was found that the difference between expectation and perceived performance regarding web information quality and service quality significantly explained web customer satisfaction (McKinney et al., 2002). Factors such as control, curiosity , heightened enjoyment, focused immersion and temporal dissociation collectively proposed as cognitive absorption were also found to influence perceptions such as ease of use and usefulness, which subsequently impacted use (Agarwal & Karahanna, 2000).

Social Context

A limited number of studies have investigated the impact of social context on web use behavior. Use of the Internet by other family members, external influence (articles, reviews, and promotion of the web site), and interpersonal influence (relatives and colleagues) were identified as significant predictors of web use (Agarwal & Venkatesh, 2002; Kruat et al., 1999; Parthasarathy & Bhattacherjee, 1998).

Studies on Online Purchase

The studies within this category focused on identifying factors that impacted the intention to purchase or the actual purchase behavior. The variables used as predictors of purchase behavior are categorized into consumer characteristics, consumer perceptions, technology attributes, and social context ( Table 3 ).

Consumer Characteristics

Studies found that the higher a person's income, education, and age, the more likely he or she was to buy online (Bellman et al., 1999; Liao & Cheung, 2001). Gender was found to significantly impact perceptions toward shopping through the Web. Women view shopping as a social activity and were found to be less technology oriented compared to men (Slyke et al., 2002). However, researchers have cautioned that demographic variables alone explain a very low percentage of variance in the purchase decision (Bellman et al., 1999). An interesting result that emerged was that consumers that are more likely to buy online have a "wired lifestyle." Such consumers have used the Internet for a long time, received a large number of emails everyday, believed the Internet improves productivity at work, and used the Internet for most of their other activities such as reading news and searching for information (Bellman et al., 1999). Other consumer characteristics, such as personal innovativeness, discretionary time, search for product information, web skill, Internet self-efficacy, email use, and prior web use were also found to be predictors of willingness to purchase (Agarwal & Prasad, 1998; Liao & Cheung, 2001; Limayem et al., 2000; Ramasawami et al., 2001). The impact of those variables on intention to purchase may be mediated through factors such as ease of use, shopping enjoyment, and perceived control (Koufaris, 2002; Limayem et al., 2000).

Consumer Perceptions

Consumer perceptions constituted an important category that influenced purchase related behavior. However, it was also one of the categories that showed a high level of diversity in terms of study variables. Perceived consequences and perceived risk were found to predict purchase behavior (Grazioli & Jarvenpaa, 2000; Liao & Cheung, 2001; Limayem et al., 2000). Perceived control and involvement with the product were also found to significantly impact shopping behavior. Consumer perceptions about different types of quality attributes of the web site and the vendor were also evaluated. Perceptual variables from the Technology Acceptance Model (TAM) (Davis, 1989) and Service Quality (SERVQUAL) (Parasuraman et al., 1988) were examined. The TAM variables of perceived usefulness and ease of use were found to be distinguishing factors between bidders and non-bidders in an online auction context (Stafford & Stern, 2002). The SERVQUAL construct consists of the five sub-dimensions of tangibles, reliability, responsiveness, assurance, and empathy (Pitt et al., 1995), and were often used in a disaggregated fashion resulting in mixed findings. Vendor quality was found to influence willingness to shop online (Liao & Cheung, 2001), and information or content quality was also a predictor of purchase behavior (Jarvenpaa & Todd, 1997; Ranganathan & Ganapathy, 2002).

Technology Attributes

Factors included in this category related to the actual functionalities and attributes of the web site rather than the perceptions of the attributes. Paper-based catalogs were found to generate higher levels of consumer involvement as compared to web-based catalogs (Griffith et al., 2001). No difference was found in money spent or number of products purchased among different interface designs, including catalog interface designs, bundle-based interface designs, and virtual reality-based stores (Westland & Au, 1998). Other attributes of the technology, such as comparative shopping, assurance mechanisms, web page loading speed, value added search mechanisms, shopping carts, feedback mechanisms, and chat channels, were found to significantly influence intentions to shop and actual purchase behavior (Grazioli & Jarvenpaa, 2000; Koufaris, 2002; Liang & Lai, 2002; Limayem et al., 2000).

Social Context

Studies in psychology and sociology have presented a wealth of knowledge about how individuals are influenced by the social structures in which they live. Limayem et al. (2000) found that media and family influences significantly affected intentions to purchase while friends ' influence did not make a difference.

Studies on Post-Purchase

The primary dependent variables within this category were satisfaction with purchase, channel preference, switching, attrition, and re-visitation. These variables are grouped into consumer characteristics, consumer perceptions, technology attributes, and vendor and channel characteristics ( Table 3 ).

Consumer Characteristics

Chen & Hitt (2002) was the only study that investigated the role of user characteristics in determining two types of post-purchase behavior (switching and attrition). The study found that age and education impacted attrition negatively, and that females showed a higher propensity to become inactive users. However, they concluded that demographics overall did not explain much variance.

Consumer Perceptions

In the context of consumer perceptions, researchers found that perceptions regarding data security, inconvenient use, stability of the system, satisfaction with previous purchase, usefulness, ease of use, web site quality, time saving, empathy, assurance, and shopping enjoyment were significant predictors of channel satisfaction, intention to revisit, switching, and attrition (Chen & Hitt, 2002; Devaraj et al., 2002; Han & Noh, 2000; Koufaris, 2002; Koufaris et al., 2002).

Technology Attributes

The studies evaluating the role of attributes of technology on post-purchase behavior have yet to identify a significant predictor. No difference was found in web satisfaction when the download time of the web page was varied between 0 and 15 seconds (Otto et al., 2000). Chen & Hitt (2002) found no significant relationship between personalization enabled through the web site and switching behavior.

Vendor and Channel Characteristics

In the context of vendor characteristics, it was found that the breadth of offerings was negatively related to switching behavior, while a greater minimum deposit required to join an online broker reduced attrition rate (Chen & Hitt, 2002). In terms of channel characteristics, price differentials between online and offline channels were found to be a significant predictor of channel satisfaction and subsequent channel preference (Devaraj et al., 2002).

Quantitative Analysis of the Theoretic Models and Study Variable Relationships

The dominant theoretical model used in online consumer research was the Technology Acceptance Model (16% of the studies) (Davis, 1989), followed by the Theory of Planned Behavior (12%) (Ajzen, 1991), and Innovation Diffusion Theory (7%) (Rogers, 1983, 1995). Other theoretic models or paradigms included Transaction Cost Economics (5%) (Williamson, 1979, 1985), Flow Theory (5%) (Csikszentmihalyi, 1988), SERVQUAL (5%) (Parasuraman et al., 1988), and Involvement Theory (5%) (Reeves & Nass, 1996). These theories were used independently or in combination with each other. An effort to examine the predictive power of different theoretical models proved to be extremely difficult because studies combined variables from different theories and used different dependent variables, thus making the comparison task problematic . An exception is the study conducted by Devaraj et al. (2002), which compared three alternative models and found that TAM explained the most variance in electronic commerce channel satisfaction (76%), followed by Transaction Cost Analysis (72%) and SERVQUAL (56%).

A meta-analysis of the interrelationships among the study variables was conducted by aggregating the correlation coefficients reported by individual studies. Since path coefficients are influenced by other variables present in the model, methods that rely on correlations are deemed more desirable (Hunter & Schmidt, 1990). Our review of the 42 studies identified only 17 studies with the correlation table reported in the paper. A subsequent review further showed that there were only eight relationships examined more than once across studies. Table 4 summarizes the analysis results of these common relationships. A weighted average of the correlation coefficients, instead of the simple average across studies, was computed for each relationship to correct for sampling error, as recommended by Hunter & Schmidt (1990).