Rational Integrative Model of Online Consumer Decision Making

Padmini Patwardhan

Winthrop University

Jyotika Ramaprasad

Southern Illinois University


Based on traditional rational consumer theories about beliefs preceding intent to act and knowledge preceding behavior, this study proposed, and empirically tested, a hierarchical path model of decision making in the online environment, focusing on the Internet’s role in two decision stages: pre-purchase search and evaluation, and actual purchase. Both direct and indirect effects were posited in the sequential model using four Internet related variables: pre-purchase search beliefs, purchase beliefs, actual pre-purchase search, and actual purchase.

The empirical test was conducted among consumers in the United States and India with 291 respondents taking the online survey (186 for the United States, 105 for India). For both U.S. and Indian respondents, each conceptualized stage of online decision making was significantly impacted by the stages preceding it, either directly or indirectly. In terms of direct effects, an antecedent Internet belief variable (pre-purchase search beliefs) impacted a consequent belief variable (purchase beliefs), and an antecedent action variable (pre-purchase search) impacted a consequent action variable (purchase). Further, the consequent belief variable (purchase beliefs) impacted the immediately following antecedent action variable (pre-purchase search). In terms of indirect effects, all antecedent variables impacted consequent variables at each stage of the model.


The applicability of traditional models to the e-shopping process is still a new area of study, with some (e.g. Hoffman and Novak 1996) holding that traditional decision making patterns will prevail and others (e.g. Peterson, Balasubramanian, and Bronenberg 1997) suggesting that they are likely to change. Within the context of such discussion, the purpose of this study was to propose a rational model of online decision making and empirically examine the application of rational hierarchy theories to the Internet environment. The cross-cultural applicability of this study’s proposed model was also explored by surveying Internet users in countries at different stages of e-shopping adoption: the United States and India.

Many traditional consumer decision-making models are based on the assumption that the consumer is rational and adaptive (Moorthy, Ratchford, and Talukdar 1997). These models propose a series of mental and motor steps (Howard and Sheth 1969) that envisage decision making as progressing from problem/need recognition, through information search and evaluation and purchase decision, to post purchase behavior (Schmidt and Spreng 1996). The various models that have emerged differ slightly in the number of steps (Haubl and Trifts 2000) and the nomenclature used for the steps, but at their core they maintain a knowledge-attitude-behavior sequence (Lavidge and Steiner 1961). Such an approach assumes a rational, goal-oriented approach to decision making, as the consumer moves from awareness and knowledge (cognitive stage) to liking and preference (affective stage) to conviction and purchase (conative stage) (Aaker and Myers 1982).

Criticism of linear and staged rational models has led us to understand that the product, consumer, buying situation, and other factors may not only reverse or change the hierarchy (in any possible combination), but also result in abbreviated or expanded hierarchies (see e.g., Crozier and McLean 1997; Sheth 1974). Ray (1973) suggests that three combinations appear to represent a majority of the situations: 1) the rational hierarchy (cognitive, affective, conative), often called the learning hierarchy because of its behavioristic approach, 2) the dissonance attribution hierarchy (conative, affective, cognitive), and 3) the low-involvement hierarchy (cognitive, conative, affective). Miracle (1987) adds the dependency hierarchy (affective, cognitive, conative) for Japanese television commercials.

In addition, MacKenzie and Lutz’s (1989) theory of consumer preference formation also assumes a rational sequence, where beliefs inform attitudes and attitudes inform intent to act. Beliefs are information (true or otherwise, based on fact or opinion) that consumers have about an object (Duncan and Olshavsky 1982; Petty and Cacioppo 1981), while attitudes are evaluative (like/dislike) and are based on this information. Beliefs have received considerable attention in multi-attribute models of consumer preference formation.

Because the Internet appears to lend itself to such decision-making, a conceptual model was developed bringing these rational theories together (where knowledge precedes behavior and beliefs precede intent to act). Consumer use of the Internet is generally assumed to be more purposive and goal-directed, and therefore more “rational,” at least at this point in time. In addition, online shoppers are far more likely to operate under conditions of higher involvement presaging more active information seeking and processing (Petty and Cacioppo 1981, 1983). In the rational model, pre-purchase search is followed by purchase. The Internet collapses in one medium the search for information and the act of purchasing, making it a perfect venue for the operation of this study’s two-stage rational model. In online shopping contexts, consumers’ need for information is large because of the lack of real interaction with the product, because such information is easily available on the Internet, and because the Internet also allows direct purchase. The Internet’s vast capacity for information storage, search and retrieval, information customization, and interactive communication makes it an efficient medium for accessing, organizing, and communicating information (Peterson, Balasubramanian, and Bronnenberg 1997). Informational use of the Internet can significantly reduce pre-purchase anxiety among consumers (Ghose and Dou 1998) and pre-purchase sales information appears to be a major part of a web site’s value (Bruner 1997). Thus potential informational benefits include increased search efficiency, better product evaluation, and enhanced transaction convenience (Zeng and Reinartz 2003).

The empirical test of the conceptual model was conducted for online consumers in both the United States and India, presenting an opportunity to test the cross-cultural validity of the proposed model among consumers at different stages of Internet and e-shopping adoption. Specifically at the time of the study, overall Internet penetration in the two countries stood at 62% of the total U.S. population and 16% of the Indian population, while e-commerce penetration stood at 32% of American Internet users and about four percent of Indian Internet users (TNS Interactive-Global eCommerce Report 2002). The U.S. Department of Commerce reported online retail transactions by American consumers in the year 2001 at U.S. $33.7 billion (Pastore 2002). Comparable figures for India stood at about U.S. $100 million (Rs. 450 crores) in 1999-2000, of which retail and business-to-consumer transactions constituted only about U.S. $11 million (Rs. 50 crores) (Times of India 2000).

The rationale underlying the choice of India was multi-fold. First, while decision models are routinely tested using American consumers, theory-driven studies of Indian consumers are practically non-existent. Second, economic liberalization since the 1990s has seen a shift from socialism to capitalism in India which, in turn, is fostering a culture of consumer spending similar to that of more developed societies. Third, India’s largest export to the United States is its pool of computer technology professionals and it may not be wrong to assume that Internet users in India are reasonably familiar with various online activities, whether they shop online or not. Finally, there is some evidence to suggest that Indians have begun leapfrogging the digital divide and Internet users in that country are using online resources to inform their purchase decisions (Patwardhan 2003).

Conceptual Path Model of Rational Decision Making

Several barriers to actual online purchase have been reported, including consumer concerns and transaction anxiety (Korgaonkar and Wolin 1999), payment security and distribution issues, perceptions of informational rather than transactional value of the Internet (Zeng and Reinartz 2003), lack of face-to-face contact in online shopping (Times of India 2000), as well as product factors like need for physical experience (e.g., perfumes) and complexity of purchase (e.g., real estate). However, it is equally evident that e-shopping is becoming an increasingly global phenomenon with over 15% of global Internet users transacting online, and 18% intending to do so in the next six months (TNS Interactive-Global eCommerce Report 2002). In many consumer surveys, convenience, interactivity, and shopping ease are often mentioned as primary reasons for the growth of consumers’ e-shopping activities (Ghose and Dou 1998; GVU survey 1998; Korgaonkar and Wolin 1999). While some consider only the final act of buying a product online as defining the transactional nature of e-shopping, others include an entire range of activities from purchase-related information searches, use of price and brand comparison tools, use of online coupons and discounts, and interaction with marketing personnel via web sites, to actual purchase and post-purchase use of the Internet. This expansive definition of online purchase activities can be broken down into three basic, staged components of consumer decision making–pre-purchase search and evaluation, actual use/purchase, and post-purchase interaction.

This study modeled two of the three decision stages within the e-shopping context: pre-purchase search and evaluation, and actual purchase. Using the rational hierarchy of effects model (Lavidge and Steiner 1961) where knowledge precedes behavior, it proposed the following relationship as the first conceptual step in model building:

Online Pre-Purchase search and evaluation

Based on the consumer preference formation model, each of the two stages above was further conceptualized as a pair of belief-action variables. Since beliefs have theoretically been considered to precede use (Duncan and Olshavsky 1982; MacKenzie and Lutz 1989), each belief variable was posited to precede its related action. Thus beliefs about the Internet’s potential for pre-purchase search and evaluation (pre-purchase search beliefs) and actual pre-purchase search and evaluation (pre-purchase search) were the cognition (knowledge) stage variables, and beliefs about the Internet’s potential for purchase (purchase beliefs) and actual purchase (purchase) were behavior stage variables. The model was expanded to propose the following relationships:

Pre-Purchase search beliefs

The final step in the conceptual model building was integrating these traditional theories to propose the following rational path model of relationships between the variables. In the model, both belief variables precede action variables, and both search related variables precede action related variables.

Figure 1. Path Model of Rational Decision Making for Online Purchases

Path Model of Rational Decision Making for Online Purchases

Hypotheses and Research Questions

The hypotheses (direct effects) and research questions (indirect effects) paralleled the paths outlined in the diagram. The primary focus of this study was to test, using path analysis, our model’s applicability to online purchase decision-making. The study put the model to test in two countries for stringency and to assess the model’s cross-cultural validity. As a result, the study’s hypotheses and research questions do not focus on differences between countries, but on relationships, direct or indirect, between variables in both countries.

Direct effects are that part of the total effect that is not transmitted through intervening variables (Alwin and Hauser 1975), while indirect effects are that part of the total effect that is transmitted by mediating variables. Hypotheses were proposed for direct effects because these may be, and were, tested using a series of standard regression analyses following Pedhazur (1982) and Alwin and Hauser (1975). Research questions were posited for indirect effects because these may be examined only descriptively (Goldsmith, Lafferty, and Newell 2000) due to lack of a statistical test of significance.

Our first hypothesis proposed a relationship between the study’s two belief variables: pre-purchase search beliefs and purchase beliefs. While linear hierarchies do not specifically address belief-belief relationships, research on attitudes suggests that an antecedent attitude (e.g. attitude toward the ad) often impacts a consequent one (e.g. attitude toward the brand). Following this line of thinking, we hypothesized the two belief variables in our study to be directly related. Since we assumed a linear, staged sequence of rational decision making, the direction of the relationship was determined to be from pre-purchase to purchase. Our first hypothesis proposed that:

H1a: Pre-purchase search beliefs positively and directly influence purchase beliefs.

The next four hypotheses proposed paths between beliefs and actions. Following Duncan and Olshavsky’s (1982) argument that, at a minimum, beliefs and actions will be consistent, both belief variables were seen as directly related to both actions in the hierarchy.

H1b: Pre-purchase search beliefs positively and directly influence pre-purchase search.

H1c: Pre-purchase search beliefs positively and directly influence purchase.

H2a: Purchase beliefs positively and directly influence pre-purchase search.

H2b: Purchase beliefs positively and directly influence purchase.

Our final hypothesis dealt with a behavior-behavior (action-action) relationship: pre-purchase search and purchase. A rationale similar to H1a (linear, rational decision-making) was used to propose a path from pre-purchase action to purchase action:

H3: Pre-purchase search positively and directly influences purchase.

Apart from forcing researchers to explicate theory with regard to the sequence of relationship(s) among variables rather than simply testing linear relationships, what distinguishes path analysis from a multiple regression analysis is its ability to represent and calculate indirect effects (Alwin and Hauser 1975), thus ensuring that the influence of a causal variable is not underestimated. In the tradition of Pedhazur (1982) and Alwin and Hauser (1975), indirect effects were examined through the following research questions:

RQ1a: To what extent do pre-purchase search beliefs positively and indirectly influence pre-purchase search through an effect on purchase beliefs?

RQ1b: To what extent do pre-purchase search beliefs positively and indirectly influence purchase through an effect on purchase beliefs as well as an effect on pre-purchase search?

RQ2: To what extent do purchase beliefs positively and indirectly influence purchase through an effect on pre-purchase search?


This research employed a questionnaire which was administered via a web site to a sample of U.S. and Indian consumers.

Population and Sample

The survey population was U.S. and Indian consumers with email/Internet access because the focus of the study was online purchase behavior. Hence consumers without Internet access were consciously excluded. The sampling process did not take into consideration the populations of the two countries largely because Internet penetration and use were vastly different in terms of the percentage of the population that had access. Previous researchers surveying online populations have made use of both convenience and purposive samples. While thousands of user databases are commercially available from online and offline sources for payment, no comprehensive national lists of online consumers are freely available, one of the major problems in generating probability samples of Internet users. Following Sheehan and Hoy (1999), with some departures, this study deployed a geographic parameter based on stratification of the country to obtain a sample of consumers with Internet access. To ensure randomness, all U.S. states were first assigned to different geographic zones based on their general location: North, South, East, West, and Central. Two states were randomly picked from each zone, providing a set of 10 states. Names of the states were then entered, one at a time, into the Yahoo People Search engine. At the time of sampling, each state-wise search generated 75 email addresses, although this search method is no longer available on Yahoo. While geographic stratification and random selection of states from the strata were used, it is unknown whether the 75 email addresses were generated randomly for each state. Further, the population from which these names were generated is also unknown. The emails were then manually entered into a directory created by the researchers, providing a list of 750 email addresses. Due to a low initial response, five more U.S. states (one from each zone) were subsequently sampled without replacement to generate an additional 350 email addresses.

A similar procedure was followed for India, although only one state from each of five geographic regions (India has half the number of states compared to the United States) was randomly selected to generate an Indian sample of 375. Second round sampling for India (similar to the U.S. process) had to be abandoned. Yahoo failed to return a sufficient number of email addresses from other randomly selected Indian states, sometimes returning none. Thus a pool of less than fifty email addresses was returned in the second round. We next attempted to resample from the Indian states selected in the first round. The same email addresses were returned, rendering the re-sampling useless. Emails to Yahoo seeking clarification yielded no response. Admittedly, this sampling method had limitations.

An invitation to participate in the survey was delivered via email to respondents, with a link to the survey website. The email explained the purpose and nature of the study, the time required to complete the survey, and the researchers’ affiliations. An opt-out option and anonymity assurances were also provided.


Conceptual definitions of the four major variables were as follows. Pre-purchase search beliefs were beliefs about the Internet’s potential forgathering and evaluating reliable and comprehensive product and purchase related information with ease and speed. Purchase beliefs were beliefs about the Internet’s potential for product purchase with security, convenience, ease, and enjoyment. Actual pre-purchase search was use of the Internet for retrieving and evaluating product, price, service and other purchase related information, and actual purchase was use of the Internet for making product purchases and related actions such as using online discounts.

The questionnaire design went through several qualitative and one quantitative iteration based on the results of a pilot test. Apart from questions on demographics (age, gender, education, income) and Internet use (years of Internet use, daily hours online), the questionnaire included scales for the two belief and two behavior variables (see Appendix) developed by the authors using the following procedure. A pool of face valid scale items was first generated from literature and through discussion. Scale items were administered in a pre-test to a convenience sample of Internet users from a university population that included both American and Indian students (n = 80). Factor analysis (varimax rotation) and reliability testing was conducted to purify the scales. Single factor solutions with eigenvalues above one emerged for all four variables supporting construct unidimensionality. Cronbach’s alphas were: pre-purchase search beliefs (.85), purchase beliefs (.90), pre-purchase search (.93) and purchase (.91). Thus all scale items were retained for the final questionnaire. During final survey administration, post -test reliabilities were as follows, presented by country: pre-purchase search beliefs U.S. (.85) and India (.84), purchase beliefs U.S. (.90) and India (.91), pre-purchase U.S. (.77) and India (.86), and purchase U.S. (.91) and India (.92).


Altogether, 1,500 email messages were sent out to the U.S. (750 first mailing, 350 second mailing) and Indian (375) sample. A total of 509 email messages were returned as undeliverable (422 for the U.S. sample, 97 for the India sample), and five respondents sent email messages declining participation, reducing the generated sample size to 986. A total of 291 respondents took the survey (186 for the U.S., 105 for India) on the web site, for a response rate of 19% for the U.S. and 28% for India. However, because the survey site was subsequently listed on an Indian listserv by an interested respondent, and access to the web site was not controlled, the response rate must be interpreted with caution because the unsolicited listing prevented estimation of accurate response rate from the original sample.

Description of Sample

Indian Internet users in the survey were somewhat older (average age 39 years) than their U.S. counterparts (average age 37 years) (Table 1). Exactly half (50%) of U.S. respondents, and a little over half (56.4%) of Indian respondents, were in the 25-40 age group. Interestingly, those below 25 years constituted the smallest segment of Internet users among U.S. respondents (14.4%), while the above 40 age group (10%) was the smallest segment among Indian respondents.

While better balance among male (55.4%) and female (41.9%) users was observed for the U.S. sample, male users (66.7%) outnumbered female users (29.5%) for the Indian sample. In terms of education, more than half (51.1%) of U.S. respondents were college graduates, 25% were in college, and another 18% had completed high school. Among Indian respondents, about half (51.4%) were college graduates, 34% were in college, and another four percent had completed high school (Table 1).

In terms of Internet experience and use, on average, U.S. respondents had used the Internet for about five to six years (m = 5.63, s.d. = 2.99), while Indian users had used it for about four years (m = 3.86, s.d. = 2.01). Interestingly, average daily Internet use suggested that Indian users were online for longer daily durations (m = 3.44 hours, s.d. = 3.34) than U.S. respondents (m = 2.91 hours, s.d. = 2.32) (Table 1).

Table 1. Respondent Profile and Mean Scores* by Country

Respondent Profile and Mean Scores* by Country

Both U.S. (m = 2.3, s.d. = .56) and Indian (m = 2.2, s.d. = .74) Internet users had fairly positive and strong beliefs about the Internet’s potential for pre-purchase search and evaluation (lower value = more positive belief on a scale of 1 to 5) (Table 1). However, their beliefs about its potential for product purchase were more or less neutral: U.S. users (m = 3.0, s.d. = .53), Indian users (m = 3.1, s.d. = .62). Both U.S. (m = 2.8, s.d. = .84) and Indian (m = 2.9, s.d. = 1.0) users reported slightly greater than average use of the Internet for pre-purchase search evaluation. However, when it came to actual online purchase, U.S. users reported buying online “sometimes” (m = 3.1, s.d. = 1), while Indian users seldom shopped online (m = 3.6, s.d. = 1.2) (Table 1).

Path Analysis Results

Direct Effects

The strength of the path from pre-purchase search beliefs to purchase beliefs for U.S. consumers (ß = .618, p = .00) suggested that the former had a direct and significant influence on the latter (Figure 2 and Table 2). Such a direct effect (ß = .608, p = .00) was observed for Indian consumers as well (Figure 2 and Table 3). Hypothesis 1a was, therefore, supported. Regression analysis also showed that pre-purchase search beliefs significantly predicted 38% of the variance in purchase beliefs (F = 113.8, p = .00) for U.S. consumers and 37% (F = 60.45, p = .00) for Indian consumers.

Figure 2 Results of Path Analysis of the Rational Decision Making Model for Online Purchases

Results of Path Analysis of the Rational Decision Making Model for Online Purchases

The path from pre-purchase search beliefs to pre-purchase search was significant for both countries’ consumers (U.S.: ß = .466, p = .00; India: ß = .315, p = .00) (Figure 2 and Tables 2 and 3). That is, pre-purchase search beliefs positively and directly predicted pre-purchase search for both groups of consumers. Hypothesis 1b was, therefore, supported.

The effect of pre-purchase search beliefs on purchase was not significant for either U.S. or Indian consumers. This suggests that, for both groups, beliefs about the Internet’s information potential did not significantly predict their online purchasing. Hypothesis 1c was, therefore, not supported.

The paths from purchase beliefs to a) pre-purchase search, and b) purchase were significant for both countries’ consumers (U.S.: ß = .216, p = .00; ß = .407, p = .00; India: ß = .335, p = .00; ß = .400, p = .00) (Figure 2 and Tables 2 and 3). That is, purchase beliefs positively and directly predicted pre-purchase search as well as actual purchase in both groups of consumers. Hypotheses 2a and 2b were, therefore, supported.

A significant direct, positive path was also found from pre-purchase search to purchase for both U.S. (ß = .478, p = .00) and Indian (ß = .348, p = .00) consumers (Figure 2 and Tables 2 and 3). This suggests that, for both groups, pre-purchase search had a direct, positive impact on purchasing. Hypothesis 3 was, therefore, supported.

Indirect Effects

For RQ1a, an indirect path from pre-purchase search beliefs to pre-purchase search was found through purchase beliefs for both U.S. (ß =.135) and Indian (ß =.204) consumers (Tables 2 and 3). However, for both groups, the direct effect was stronger than the indirect effect.

For RQ1b, indirect paths from pre-purchase search beliefs to purchase were found for both U.S. and Indian consumers through i) purchase beliefs (U.S.: ß = .252; India: ß =.243), ii) pre-purchase search (U.S.: ß =.223; India: ß =.110), and iii) the two combined (U.S.: ß = .064; India: ß = .071) (Tables 2 and 3). For both groups, therefore, pre-purchase search beliefs exerted an indirect effect on purchase. Thus while pre-purchase search beliefs did not directly predict purchase, they did exert an indirect influence on purchase through mediating variables.

For RQ2, purchase beliefs had an indirect influence on purchase through pre-purchase search for both U.S. (ß = .103) and Indian (ß =.117) consumers (Tables 2 and 3). However, for both groups, the direct effect was greater than the indirect effect.

Table 2. Interpretation of Significant Effects in Path Model of Rational Decision Making for Online Purchases (U.S. Consumers)

Interpretation of Significant Effects in Path Model of Rational Decision Making for Online Purchases (U.S. Consumers)

Table 3. Interpretation of Significant Effects in Path Model of Rational Decision Making for Online Purchases (Indian Consumers)

Interpretation of Significant Effects in Path Model of Rational Decision Making for Online Purchases (Indian Consumers)

Results of the Revised Path Model

A judgmental criterion of meaningfulness (e.g., a regression coefficient < .05) is used by researchers to delete paths even if coefficients are significant, particularly when it is suspected that with a large sample size substantively meaningless regression coefficients may be found to be significant (Pedhazur 1982). In this study, only the non-significant path (pre-purchase search beliefs to purchase) was eliminated for theory trimming because all significant coefficients appeared to be relatively healthy.

Because only the last stage of the model (the regression of purchase on the three other variables) was affected by the non-significant path, only the last stage of the path analysis was re-run excluding pre-purchase search beliefs as a direct predictor of purchase. The revised path diagram and results are presented in Figure 3. All remaining paths continued to be significant though ß weights went down slightly.

Figure 3. Revised Path Model of Rational Decision Making for Online Purchases

Revised Path Model of Rational Decision Making for Online Purchases


Based on the sequential predictive relationships empirically observed through this study, it may be concluded that the mental and motor steps (Howard and Sheth 1969) of online purchase decision-making may be explained using hierarchical models. The fact that the findings were identical across two rather different cultures at different stages of Internet adoption provides further reinforcement to the conclusions. The lack of one direct path does not change the model’s basic hierarchies, and the rerun of the model without the non-significant path did not change results much, further confirming the conclusions.

For both countries, with one exception, all direct paths, i.e., effects, in the sequential, rational consumer decision-making model were statistically significant. Thus, pre-purchase search beliefs led to purchase beliefs, which led to pre-purchase search, which, in turn, led to purchase. In addition, pre-purchase search beliefs led to pre-purchase search, and purchase beliefs led to purchase. Only pre-purchase search did not directly lead to purchase.

For both countries, this time without exception, all indirect paths, i.e., effects, were present. Pre-purchase search beliefs led to purchase through purchase beliefs and pre-purchase search. Also, purchase beliefs led to purchase through pre-purchase search. Finally, pre-purchase search beliefs led to pre-purchase search through purchase beliefs.

While the direct path from pre-purchase search beliefs to purchase was not significant, the indirect path was. In sum, each stage of the online purchase decision-making process as conceptualized in this study’s path model was significantly impacted by the stages preceding it, either directly or indirectly, for both the U.S. and Indian respondents.

Some general conclusions can be drawn from the above analysis. It was argued that the rational model might be particularly well-suited to Internet-based consumer purchases because of consumers’ need for information in an online buying situation where the product is not physically available for inspection, and because of the tremendous potential of the Internet for information in terms of amount of availability of information and ease of retrieval of this information. Respondents confirmed the latter by rating the medium quite high on potential for information search and around midpoint for actual information search. (Such a drop at each stage of decision making models is typically expected. Also, access related constraints could be responsible for the lower rating for actual retrieval as compared with potential for retrieval).

It was also argued that the rational model might be well-suited to online consumer purchases because of the Internet’s potential for direct purchase by consumers. In fact, it may be that the Internet, more than any other direct marketing medium, has the greatest potential for actual purchase because of ease and ability to create a virtual shopping experience. However, while our respondents were generally positive about the Internet’s purchase utility, some restraint and caution was evident at the actual purchase stage. Considering the concerns regarding online security and privacy, as well as practical issues regarding transacting online in India (e.g. less prevalent credit card usage and online payment gateways), the limited purchase action is not surprising.

It appears that the basic rational online decision model proposed in this study (defined in terms of traditional knowledge-behavior and beliefs-action hierarchies) is a viable one, regardless of the stage of Internet and e-shopping adoption. In addition, when effects are compared, some very interesting conclusions emerge to further our knowledge of rational hierarchies. First, each antecedent belief better predicted the action most closely related to itself. Thus, in the Internet environment, pre-purchase search beliefs positively influenced pre-purchase search and purchase beliefs influenced purchase. Second, an antecedent belief predicted a consequent belief better than it predicted a consequent action. In fact, the strongest effect in the model was of one belief variable on another. In both countries, for total (direct and indirect) effects, pre-purchase search beliefs had the largest effect on purchase beliefs, the second largest effect on pre-purchase search, and the lowest on purchase. This is an interesting finding.

Some of the limitations of this study are as follows. Response rate for the study was small. Based on the fact that studies with low response rates are the norm for online surveys of general Internet users, this study’s response rate may be deemed acceptable. Comparison of respondent characteristics with Internet population demographics for both the U.S. (GVU Survey 1998) and India (NASSCOM 2000) suggested that the samples were fairly representative, with the single exception of age for Indian respondents. Indian respondents in the sample were somewhat older than the general Internet user population in India. Still, because a true random sample could not be used (a practical impossibility when surveying general Internet populations), the results from this study’s sample must be interpreted with caution.

Purchase was measured as an interval, not ratio, level measure to capture a range of purchase-related actions. Despite differences in accessibility, adoption, and e-commerce availability levels, a ratio level measure may be a more valid measure of online purchasing activity.

While quite strongly supported by theories of consumer decision making processes and by the theoretical basis of decision support systems research (Haubl and Trifts 2000; Pereira 1999), the path model used in this study did not represent all stages of a rational model. Particularly, the exclusion of the attitudinal stage in both the knowledge-behavior and beliefs-action sequences is a limitation. Inclusion of this stage may have provided greater explanation of variance in the model. As already pointed out, the sequence adopted by consumers is not necessarily always rational; that is, information search may not always precede purchase and beliefs may not always precede action.

Further, the theoretical underpinning of this study–the rational hierarchy–is generally applied to a specific purchase (find out about a product, develop an attitude, and then make a purchase); this study does not do that. By not referencing a specific purchase, questions about the links between the stages may arise. For example, it may be argued that consumers may make use of several information channels that may or may not include the Internet (for example, call an 1-800 number or talk to a dealer) even if their purchase is made online; or consumers may collect information from the Internet even if their purchase is made offline. The exclusion of the interaction of offline searches and offline purchasing with online search and purchase is therefore a limitation.

The path model proposed in this study is a simple linear one. More complex models may be proposed and tested in future research, adding mediating stages (e.g. attitudinal) or including moderating variables (e.g. consumer expertise) that may affect the strength of the conceptualized relationships. In addition, the online shopping environment has evolved considerably (both positively and negatively) since this study was conducted. The development of new online shopping resources and tools (e.g. independent consumer feedback and advice pages, consumer blogs, sophisticated information search tools) are likely to impact decision making as are new concerns about identity theft and hacking.


Theory testing and development is this paper’s primary contribution. Its proposition and empirical validation of a hierarchical theoretical model as applied to online consumer decision making extends the scope of rational models to the Internet. This study is also cross-cultural providing validity to both the method and the findings. Also, by calculating and presenting indirect effects, it exploits one of the major advantages of path analysis, the ability to separate out the relative direct and indirect contribution of each independent variable in explaining the respective dependent variable, and thereby also assessing the total effect. Specifically, the contribution of pre-purchase search beliefs to explaining purchase indirectly would have remained undiscovered if only direct paths had been examined, and the analysis of effect sizes would have been misleading if total effects had not been gauged. The study also uncovered empirical evidence of connections between beliefs indicating that, in the rational decision making process, beliefs at one stage may have a positive direct effect on beliefs at a subsequent stage. This is a finding of some interest and merits further consideration. Finally, from a measurement perspective, the new scales created to operationalize the four variables used in the study demonstrated fairly high reliability and may be employed in future research.


Aaker, David and John G. Myers (1982), Advertising Management (2nd ed.), Englewood Cliffs, N.J.: Prentice-Hall.

Alwin, Duane F. and Robert M. Hauser (1975), “The Decomposition of Effects in Path Analysis,” American Sociological Review, 40 (February), 37-47.

Bruner, Gordon (1997), “Cyberspace: The Marketing Frontier,” Book Review in Journal of Marketing, 61 (1), 112-113.

Crozier, David A. and Fiona McLean (1997), “Consumer Decision-Making in the Purchase of Estate Agency Services,” Service Industries Journal, 17 (2), 278-293.

Duncan, Calvin P. and Richard W. Olshavsky (1982), “External Search: The Role of Consumer Beliefs,” Journal of Marketing Research, 19 (February), 32-43.

Ghose, Sanjoy and Wenyu Dou (1998), “Interactive Functions and Their Impacts on the Appeal of Internet Presence Sites,” Journal of Advertising Research, (March-April), 29-43.

Goldsmith, Ronald E., Barbara A. Lafferty, and Stephen J. Newell (2000), “The Impact of Corporate Credibility and Celebrity Credibility on Consumer Reaction to Advertisements and Brands,” Journal of Advertising, 29(3), 43-54.

“GVU 10th WWW User Survey (1998),” Report <http://www.cc.gatech.edu/gvu/user_surveys/survey-1998-10/tenthreport.html#ex>
(accessed on 1/2002)

Haubl, Gerald and Valerie Trifts (2000), “Consumer Decision Making in Online Shopping Environments: The Effect of Interactive Decision Aids,” Marketing Science, 19 (1, Winter), 4-21.

Hoffman, Donna L. and Thomas P. Novak (1996), “Marketing in Hypermedia Computer-Mediated Environments: Conceptual Foundations,” Journal of Marketing, 60 (July), 50-68.

Howard, John A. and Jagdish N. Sheth (1969), The Theory of Buyer Behavior, New York: John Wiley.

Korgaonkar, Pradeep K. and Lori D. Wolin (1999), “A Multivariate Analysis of Web Usage,” Journal of Advertising Research, (March-April), 53-68.

Lavidge, Robert and Gary Steiner (1961), “A Model for Predictive Measurements of Advertising Effectiveness,” Journal of Marketing, 25 (6), 59-62.

MacKenzie, Scott B. and Richard J. Lutz (1989), “An Empirical Examination of the Structural Antecedents of Attitude Toward the Ad in an Advertising Pretesting Context,” Journal of Marketing, 53 (April), 48-65.

Miracle, Gordon E. (1987), “Feel-Do-Learn: An Alternative Sequence Underlying Japanese Consumer Response to Television Commercials,” in The Proceedings of the 1987 Conference of the American Academy of Advertising, F. Feaseley, ed., Columbia, SC: University of South Carolina.

Moorthy, Srinivas, Brian T. Ratchford, and Debabrata Talukdar (1997), “Consumer Information Search Revisited: Theory and Empirical Analysis,” Journal of Consumer Research, 23 (March), 263-277.

NASSCOM (2000), “India’s Internet: Ready for Explosive Growth,” < http://isp-planet.com/research/india_stats.html> (accessed on 2/2002).

Pastore, Michael (2002, February 20), “U.S. E-Commerce Spikes in Q4 2001,” <http://cyberatlas.internet.com/markets/retailing/article/0,,6061_977751,00.html#table> (accessed on 7/10/2002).

Patwardhan, Padmini (2003), Internet Dependency Relations and Relationship with Exposure, Involvement, and Satisfaction with Internet Activities: A Cross National Survey of U.S. and Indian Internet Users. Unpublished Dissertation. Carbondale: Southern Illinois University.

Pedhazur, Elazar (1982), Multiple Regression in Behavioral Research, New York: Holt, Rinehart and Winston.

Pereira, Rex E. (1999), “Factors Influencing Consumer Perceptions of Web-based Decision Support Systems,” Logistics Information Management, 12 (1/2), 157-181.

Peterson, Robert, Sridhar Balasubramanian, and Bart J Bronnenberg (1997), “Exploring the Implications of the Internet for Consumer Marketing,” Journal of the Academy of Marketing Science, 25 (4), 329-346.

Petty, Richard E. and John T Cacioppo (1981), Attitudes and Persuasion: Classic and Contemporary Approaches, Dubuque, IA: William C. Brown.

— and — (1983), “Central and Peripheral Routes to Persuasion: Application to Advertising,” in L. Percy and A. Woodside, eds., Advertising and Consumer Psychology, Lexington, MA: Lexington Books, 3-23.

Ray, Michael L. (1973), “Marketing Communication and the Hierarch of Effects,” in P. Clarke, ed., New Models for Communication Research, Beverly Hills, CA: Sage, 147-176.

Schmidt, Jeffrey B. and Richard A. Spreng (1996), “A Proposed Model of External Consumer Information Search,” Journal of the Academy of Marketing Science, 24 (3), 246-256.

Sheehan, Kim B. and Maria G. Hoy (1999), “Using Email to Survey Internet Users in the United States: Methodology and Assessment,” Journal of Computer Mediated Communication, 4 (3), <http://www.ascusc.org/jcmc/vol4/issue3/sheehan.html> (accessed on 12/2001).

Sheth, Jagdish N. (1974), Models of Buying Behavior: Conceptual, Quantitative, and Empirical. New York: Harper and Row.

Times of India (2000, November 30), “The New E-volution,” <http://www.indya.com/aboutus/tim3011.html > (accessed on 2/2002).

“TNS Interactive-Global eCommerce Report” (2002, June), <http://www.tnsofres.com/ger2002/keyfindings/index.cfm> (accessed on 7/2002).

Zeng, Ming and Werner Reinartz (2003), “Beyond Online Search: The Road to Profitability,” California Management Review, 00081256, Winter, 45 (2).

About the Authors

Padmini Patwardhan, assistant professor, Department of Mass Communication, Winthrop University, has teaching and research interests in Internet-related issues, international advertising/public relations/communication, and media dependency effects. She has published in Gazette, Journal of Communication Management, Journal of Interactive Advertising, American Journalism, and Journal of Website Promotion.

Jyotika Ramaprasad is Associate Professor in the School of Journalism, SIUC. She has teaching and research interests in international communication and national development and her travels in pursuit of these interests include Asia and Africa. Her work has been published in (among others) Gazette, Journalism Quarterly, Journal of Advertising, and Journal of Advertising Research.


Authors’ names are arranged alphabetically by last name. A version of this manuscript won the top research paper award in the Advertising Division at the AEJMC national convention, Kansas City, MO, August 2003.

Appendix: Online Consumer Decision Making Scales

Pre-purchase Search Beliefs
(Five point scale from Strongly Agree to Strongly Disagree)

The Internet is a valuable source of product information
Internet-based information helps in making wise purchase decisions
The Internet helps people become better consumers
The Internet provides quick and easy access to product information
Searching for information online is an inexpensive way to learn about products
The Internet is a useful tool to make brand comparisons
Web-based information about products is unreliable
Searching for purchase related information online is a waste of time
Web advertising is a valuable source of product information
Web sites provide all the information consumers need to buy a product

Purchase Beliefs
(Five point scale from Strongly Agree to Strongly Disagree)

Placing an order on the Web is too complicated
Buying on the Web is more impersonal than shopping in a store
Buying on the Internet takes all the trouble out of shopping
The Internet is a one-stop shop for everything
It’s difficult to judge product quality while shopping online
Transaction security is a major concern in online shopping
Buying on the Web gives consumers more flexibility to shop when they want, where they want
The Internet is not the best place to buy all products
Buying on the Web is faster than going to a store
Shopping on the Internet is a convenient way to buy products
Shopping on the Web is an enjoyable experience
Consumers should be cautious about buying products on the Internet
It’s cheaper to buy products on the Web

Pre-Purchase Search
(Five point scale from Very Frequently to Never)

I visit Web sites to check out the best deals
I search for detailed information about the brand or product category
I compare several brands online before making a decision
I check out relevant Web ads to get more information about a product
I check out Web sites for sales and service information
I look for product information that is specific to my requirements
I check out company information online for products I would like to buy
I look for online discounts and bargains

(Five point scale from Very Frequently to Never)

I shop on the Internet
I buy many different products on the Internet
I make use of online discounts on goods and services
I follow up on good deals on the Internet
I buy a product online even if other buying options are available