JCMC 9 (4) July 2004
Collab-U  CMC Play  E-Commerce Symposium Net Law  InfoSpaces Usenet
  NetStudy VEs VOs O-Journ HigherEd Conversation Cyberspace Web Commerce
Vol. 6 No. 1  Vol. 6 No. 2  Vol. 6 No. 3  Vol. 6 No. 4  Vol. 7 No. 1  Vol. 7 No. 2  Vol. 7 No. 3
 Vol. 7 No. 4  Vol. 8 No. 1  Vol. 8 No. 2   Vol. 8 No. 3  Vol. 8 No. 4   Vol. 9 No. 1   Vol. 9 No. 2   Vol. 9 No. 3

Cross-National Differences in Website Appeal: A Framework for Assessment

Brian F. Blake
Kimberly Neuendorf
Cleveland State University


Abstract

Based on existing theory and research, a framework for the examination of national differences in Website feature appeal is developed. The framework is applied to data collected in five nations (U.S.A., Canada, Austria, Iran, Taiwan). The results confirm the importance, when assessing website appeal cross-nationally, of considering the type of user evaluation (i.e., "elevation" or sheer level of demand for features as a group, "differentiation" or the degree to which features are distinct in the extent to which they are preferred, and "priority" or the relative preference for a feature over others). The approach also highlights the importance of distinguishing between "individual" vs. "societal" level mechanisms when gauging inter-nation differences.


Introduction

As reported by e-commerce monitoring organizations1, nations differ in the number of people shopping online, the demographic profiles of shoppers, and the types of products purchased. Beneath these surface differences, do nations and cultures differ in the features that attract shoppers to one e-commerce website in preference to another? The answer is pivotal to social science theories explaining cross-cultural variations in an increasingly important arena of consumer behavior. The answer is relevant to marketing science concepts of integrated global marketing strategies. Still further, the answer is important to website developers and other communication/marketing professionals hoping to generate sales by tailoring websites to fit the demands of the targeted national markets (e.g., Lynch, Kent, & Srinivasan, 2001).

Given that nations/cultures differ in media perceptions (e.g., Rice & D'Ambra, 1998; Ross, 2001), it comes as no surprise that nations/cultures have been found to differ in the role played by website features in attracting shoppers. Illustratively, Lynch et al., (2001) noted that nations differ in the role played by trust, site quality, and elicited affect in producing purchase intent and site loyalty. Jarvenpaa and Tractinsky (1999) determined that cultures vary in consumer expectations of what makes a web merchant trustworthy and in the consequences of those judgments of trustworthiness. As the latter note, however, their results are not entirely consistent with the conclusions of some previous work (Quelch & Klein, 1996).

Objectives

Unfortunately, beyond work such as the above, there has been limited empirical and theoretical analysis of the nature of national differences in the drawing power of specific website features. More fundamentally, there has not emerged a common conceptual framework for quantitatively assessing such differences. Our purposes in this report are: (a) to suggest a framework or model for assessing the nature of cross-national distinctions in reactions to website features, (b) to demonstrate its application in an illustrative sampling of online shoppers in five nations, (c) based on these data, to propose particular evaluative dimensions to operationalize this framework in future cross-national studies of website appeal, and (d) to suggest the wide-ranging implications of the framework for behavioral scientists and for practitioners. Data are based on small convenience samples of online shoppers in each nation. While the limited sampling precludes generalizing results from any sample to the larger nation populations,2 the data are quite suitable to demonstrate the framework's applications and implications.

Site Features

         Typological Perspectives

Several investigators have suggested that specific features (e.g., download speed, reputation of merchant, variety of products available) have an impact upon website appeal because they perform particular underlying roles or functions. Notably, Srinivasan, Anderson, and Donnavolu (2002) isolated eight factors or functions of site features that impact customer loyalty to a retail site: 1) customization, 2) interactivity, 3) "cultivation" (i.e., provision of information and incentives to extend customer purchasing over time), 4) "care" (operationally, features that keep customers informed of the availability of preferred products and of the status of their orders, or features that minimize disruption in service), 5) "community" (i.e., provision of a structure to facilitate the exchange of user opinions and information about offered products/services), 6) "choice" (variety of products), 7) convenience, and 8) "character" (i.e., text/graphics/slogans, etc., projecting an image or personality of the web merchant). All but convenience were found to enhance customer loyalty. Palmer (2002) suggested that various specific features are important to the success of a website because they contribute either to a site's usability or to its media richness (i.e., its ability to communicate information).

Other investigations have considered the perceptual or evaluative dimensions a consumer uses in assessing the appeal of a website. Chen and Wells (1999) suggested that users evaluate a website along the three dimensions of how entertaining, how informative, and how organized it is. Yoo and Donthu (2001) noted that users evaluate the quality of a site along four dimensions: ease of use, aesthetic design, processing speed, and security. From the perspective of consumer motives, Keeney (1999) derived 25 categories of online consumer shopping objectives. Objectives were categorized as means-oriented (e.g., maximize product information) or fundamentally ends oriented (e.g., maximize product quality). Parsons (2002) provided a taxonomy of online shopper motives, differentiating among the functional (e.g., convenience), the personal non-functional (e.g., diversion from daily routine), and the social non-functional (e.g., communication with like-minded others) motives.

Other typologies have been suggested, such as Eighmey's (1997) assessment of the perceived benefits delivered by commercial websites, Liu and Arnett's (2000) exploration of factors associated with the success of commercial websites, and the assessment of preferences for online and traditional retail formats by Keen, Wetzels, de Ruyter, and Feinberg (2001).

         Specific Features

A variety of particular functions, user responses, or specific site features have been proposed to impact site appeal. A partial listing of these is contained in Table 1.


Features
Key Citations
   
Operational Features:  
Navigability Neilsen, 2000; Palmer, 2002; Radosevich, 1997
Interactivity, including customization Coyle & Thorson, 2001; Ghose & Dou, 1998; Palmer, 2002; Schneiderman, 1997; Selz & Shubert, 1997
Time delay/download speed Yoo & Danthu, 2001
Flow Goldsmith et al., 2001; Nel, van Niekerk, Berhon, & Davies, 1999; Novak, Hoffman, & Yung, 2000
Responsiveness Palmer, 2002
Communication utility Li, Kuo, & Russell, 1999
Organization Palmer, 2002
Convenience Bhatnagar, Misra, & Rao, 2000; Swaminathan et al., 1999; Syzmanski & Hise, 2000
   
Safety Features:  
Riskiness Bhatnagar et al., 2000; Van den Poel, & Leunis, 1999
Security Swaminathan, Lepkowska-White, & Rao, 1999; Szymanski & Hise, 2000
   
Content Features:  
Complexity Bucy, Lang, Potter, & Grabe, 1999
Quality of content Jarvenpaa & Todd, 1997
Vividness Coyle & Thorson, 2001
Entertainment (fun) value Chen, Wigand, & Nilan, 2000; Eighmey, 1997; Goldsmith, Bridges, & Freiden, 2001; Koufaris, 2002
   
Social Features:  
Approval by referent others, like family/friends Shim, Eastlick, Lotz, & Warrington, 2001
   
Product-related Features:  
Price Lynch et al., 2001; Swaminthan et al., 1999
Recognizability and/or desirability of brand Balabanis & Reynolds, 2001
Choice (i.e., range of products offered) Srinivasan, Anderson, & Donnavolu, 2002

Table 1. Specific websites features.


The Framework

Overview of Proposed Framework

The objective of the framework or model is to identify the nature of cross-national differences in the appeal of online shopping site features. As clarified by Kollman (2001), the features are defined at a concrete or specific level (e.g., download speed) rather than more abstractly (e.g., interactivity, flow). It was anticipated that this specificity, first, would make the framework more actionable for practitioners without loss of value to researchers interested in theory development, and second, make it possible for study respondents to grasp more clearly the issue they were asked to evaluate.

The model posits that there are three "Types" of national differences in feature appeal, termed here "Elevation," "Differentiation," and "Priority." Next, national differences in regard to a particular type, such as Elevation, can occur at two levels, Individual and Societal. Further, the framework assumes that the assessment of national differences in feature appeal is conducted in a hierarchical fashion, going from the most to the least general for the Types and from the more specific to the more abstract for the Levels. In each case, the examination begins with the most fundamental indicator(s) and builds from one level to the next, incorporating a consideration of the last level. For Types, this building is analogous to the "moments" in descriptive, univariate statistics (Blalock, 1979), with the first representing the single most important summative indicator (i.e., central tendency), the next adding to that an indication of dispersion, and so on. For Levels, the building process moves from person-specific characteristics outward to encompass societal-level characteristics.

Framework Components

Figure 1 sketches the key components of the model.


 
Type of Differences
Mechanism Level   Elevation     Differentiation     Priority  
Individual Level
(e.g., Demographics, Attitudes) 		     
Societal Level
(e.g., Cultural Values, National Infrastructure ) 		     

Assumptions
      Feature Groupings
      Hierarchical Assessment

Figure 1. The framework for examining national differences in website feature appeal.


Let us illustrate the model in the hypothetical case of three site features (product variety, short delivery time, discounted prices), two shoppers (one in the US and one in mainland China), and an appeal index running from 0 (low attraction to a feature) to 10 (highest appeal). Suppose that the features' attractiveness in each shopper's eyes was:

Product Variety
Short Delivery Time
Discounted Prices
Chinese
3
0
7
American
8
7
6

Figure 2. An example.


Type

First consider the Type (i.e., the nature or form) of the cross-national differences.

     Elevation

Elevation is the overall demand for, or responsiveness to, website features in general. In our hypothetical case, the overall demand for website features is greater to the USA shopper than to his or her counterpart in mainland China. A hypothetical explanation could be the fact that, comparatively, Americans are more exposed to online shopping and to the availability of a wide variety of high quality shopping sites (Economist, 2000) and, so, have learned the value of a wide range of features. Chinese online shoppers, by contrast, have fewer sites accessible to the citizenry (e.g., few websites are in Chinese characters) and so have less opportunity to have learned what the numerous features can do for them.

In our example, the typical feature has 3.7 points less appeal to the Chinese than to the American (i.e., the American's mean is 7, the Chinese shopper's is 3.3). So to account for the national differences in the appeal of a single feature, say product variety, some of the 5-point differential can be due to Elevation.

     Differentiation

Differentiation is the variability in the appeal of the various features in the eye of each individual shopper. In our hypothetical scenario there is minimal distinction among features as seen by the USA shopper (average absolute deviation of .7), but more as viewed by the shopper in China (with an average absolute deviation of 2.3). In this hypothetical case, the cross-national difference in Differentiation can be due to the same dynamic as influenced elevation, i.e., the greater inexperience of the Chinese shopper. The latter may have learned the value of some features (low price) but does not as yet care as much about delivery time or, to a degree, product variety.

     Priority

Priority is the relative appeal of a feature compared to the appeals of the other features in the eyes of a particular shopper. In the illustrative scenario, short delivery time to the American and discounted prices to the Chinese shopper have the same absolute level of appeal, i.e., 7. But the value of 7 renders that feature most important for the China shopper but only mid-range for the USA shopper. Hypothetically, perhaps with a limited number of websites tailored for the Chinese market in Chinese characters, shoppers have learned to appreciate the value of some obvious features (low prices), but not as yet the worth of some less obvious ones (short delivery time). In the USA, though, the less obvious are by this time more obvious; shoppers want and expect short delivery time for many products. So to understand cross-national differences we have to assess the relative Priority of one feature over another.

Assumption: Feature Groupings

The above reasoning implies that to gauge preferences for given site features we must consider people's responses to an adequate sample of features. By definition, Elevation is reflected in the central tendency (here, the mean) of the preferences for the group of features. Differentiation is the variability in preferences within that group of features. Priority is the relative standing of a particular feature in the preference hierarchy in the group of features.

A corollary is that the feature sample should include a broad, diversified mix of features. The framework achieves diversity by selecting features to represent, first, the various categories in a theoretical taxonomy of the determinants of a website's drawing power (Torkzadeh & Dhillon, 2002) and, second, the characteristics of an innovation determining its likelihood of adoption in the theoretical model of Rogers (1995).

A second corollary is that when differences among specific features are considered (Priority), those differences should be viewed in terms not just of single features like product delivery time, but in terms of clusters of features grouped on the dimensions underlying the evaluations of the features. The dimensional clusters may be more interpretable by behavioral scientists and, for practitioners, more actionable than are the single features. As an instance of the latter, suppose it were found in a certain nation that shoppers' demands for short delivery time go hand in hand with their demands for low or no shipping charges. A practitioner, then, would learn that an e-tailer cannot provide short delivery time by building in a hefty charge for overnight shipping.

Mechanism Levels

In light of the great variability in knowledge, habits, and other predispositions within any nation or culture, it is a gross oversimplification to describe differences found between, say, an American and a Chinese sample as a reflection of differences between American and Chinese consumers per se (e.g., Lynch & Beck, 2001; Weber & Hsee, 1998).

         Individual Level

Individual Level mechanisms are linked to demographics, attitudinal, personality, experiential, and other such characteristics of persons that vary within and between nations. Differences among nations in appeal of a feature may reflect solely the fact that, first, the appeal of that feature is a function of particular individual characteristics and, second, the nations in question differ in their concentrations of residents with those key characteristics. For example, suppose Nation X has a younger population than Nation Y. If preference for a specific website feature is invariably greater among younger users, we will find that a representative sample of Nation X users prefers that feature more than does a sample of Nation Y users. In this case the two nations do differ in their preferences, but the difference does not directly reflect4 the unique social or cultural structures of Nation X or Nation Y (Lynch & Beck, 2001; Weber & Hsee, 1998).

Individual Level mechanisms may involve demographics, although several studies suggest that Internet use is becoming less and less related to a user's demographic profile (e.g., Goldsmith, 2000; Sultan, 2002). This follows the reasoning and findings of innovation adoption researchers who have identified a weakened role of demographic factors for those innovations that are more mature (Atkin, Jeffres, & Neuendorf, 1998; Neuendorf, Atkin, & Jeffres, 2002) and those that are more continuous (i.e., similar to other innovations; Atkin & LaRose, 1994; LaRose & Atkin, 1992). On the other hand, while a decrease in the impact of demogaphics on online behavior may be true in the United States or in other developed countries, this conclusion is not necessarily valid on a global basis (e.g., Taylor Nelson Sofres Interactive, 2000).

Another area of these Individual Level mechanisms is attitudinal, i.e., chronic consumer orientations that are characteristic of individuals. There are substantial differences within and between nations in these orientations since they reflect personality factors, unique experiences with products, etc. Consider innovativeness, a highly consequential dimension of cross-national differences (Steenkamp, Hofstede, & Wedel, 1999). Within a given national market, innovative buyers are more ready to adopt an innovation like online shopping than are other consumers (cf. Rogers, 1995). Goldsmith (2000, 2001) and Citrin, Sprott, Silverman, and Stem (2000) have shown that, within a nation or culture, domain-specific innovativeness predicts the frequency of shopping online. Blake, Neuendorf, and Valdiserri (2003) have shown that domain-specific innovativeness can explain the variety of products shopped for online, and can do so even when one's demographic characteristics, experience with the Internet, and the prevalence of online shopping in one's social milieu are controlled. Cross-national differences in online shopping behavior and in the appeal of various types of websites, then, may possibly reflect differences among nations in the distribution of individual innovativeness orientations found there (Goldsmith, d' Hauteville, & Flynn, 1998).

To conclude that an observed difference among nations is due to an Individual Level process or mechanism, we must show that the determinant factor (e.g., age) has a comparable impact from one nation to another. For example, Blake, Neuendorf, Valdiserri, and Hughes (2004) found that consumers' innovativeness predicted differences among individuals in the overall frequency of online shopping, and did so within an independent sample drawn from each of the Asian, European, and North American nations studied. Operationally, the specific form of an Individual Level process can be seen in a predictive coefficient (e.g., a beta coefficient regressing an index of online shopping frequency on the innovativeness scale) estimated from a pooled multinational sample in which each nation has an equal weight or contribution.

But a demographic or an attitudinal variable may not have a comparable effect from one nation to another. For example, in the Blake, Neuendorf, Valdiserri. and Hughes (2004) study, innovativeness was predictive of the range of unusual products purchased, but only in some of the nations investigated. In the same study, gender was a significant predictor of online shopping proclivities, but only in some nations. To explain these nation by innovativeness or nation by gender interactions, it is necessary to consider Societal Level differences among the nations.5

One potentially important Individual Level mechanism involves the shopper's degree of Internet usage. Nations differ considerably in how extensively residents use the Internet. Since the frequency of one's use of the Internet can influence the manner in which one shops online (e.g., Blake, Neuendorf, & Valdiserri, 2003; Citrin et al., 2000; Goldsmith, 2001), differences among nations in the desire for website features may be traceable to Internet experience per se. Given the possible gravity of this factor, our demonstration analysis paid particular attention to Internet experience. It was anticipated that, compared to light users, heavy users of the Internet would be more familiar with the value of a wide range of site features and, so, would be more demanding of commercial websites. Hence, it was hypothesized that across nations, as an Individual Level phenomenon, heavy users would show higher Elevation than would light users. Due to the absence of any compelling conceptual reasons, hypotheses were not offered for the impact of usage frequency upon Differentiation or Priority.

         Societal Level

Societal Level processes involve the operation of values, social structures, etc., that are specific to a national population. Consider, first, cultural values; these are more general, abstract orientations about the importance placed upon various conditions, e.g., personal freedom, financial success. The values of interest are those held by broad sectors of a nation or culture, and so serve to differentiate among nations or among national cultures. For example, Schwartz (1992) has suggested a set of basic values that can typify a national culture and explain a wide variety of behavioral differences among those nations. Additionally, he has developed an individual-level scale for the assessment of values that taps what some investigators believe to be a universal system of ten value dimensions (Schwartz, 1994).

The Societal Level includes a potpourri of a nation's historic events (e.g., credit card scandals, an epidemic of "dot-com busts"), traditional economic patterns, social movements, etc. Of the myriad factors that can be considered we propose focusing upon those aspects of a nation that may be particularly important to the readiness of consumers to use the Internet or to shop online. One point of departure in developing a list of such variables has been provided by Saeed (1998) who suggested that impediments to Internet use and online shopping are both structural (a nation's information technology infrastructure) and functional (marketing programs and process issues). The former includes: extent of PC ownership, prevalence of computer literacy, cost to consumers to access the Internet, language barriers, "culture" (preferences for personal involvement in purchasing and for dealing directly with people, uncertainty avoidance), government regulations. The functional level includes the fit of the nation's consumer culture to the marketing programs employed, information management techniques available, and extent of customer discontent. Examples are the comparative absence of credit cards in Iran and the long tradition of cash (rather than credit) purchases in mainland China (cf. Economist, 2000).

Operationally, there are many potentially important aspects of a nation's social or cultural structure to estimate separately in an empirical study of website appeal. Hence, in the absence of clearly specified societal factors in an empirical analysis, the Societal Level of determinants may be treated as a residual category present in the observed differences among national samples after pertinent Individual Level processes have been accounted for. In cross-cultural studies this research design is often used, i.e., differences observed among national/cultural groups after demographic and specific Individual Level variables have been controlled are considered to be attributable to nationality/cultural forces. Although widely used, the weaknesses of this design mitigate against its use in investigations in which theoretically pertinent Societal Level mechanisms can be identified.

Assumption: Hierarchical Assessment

The model proposes that the impact on a given Type should be assessed in sequence from the most to the least general. Thus, we view the order of the effects or impacts as: upon the appeal of website features in general (Elevation), on the degree to which features are treated as comparable or distinct (Differentiation), and then upon the comparative appeal of some specific features/feature dimensions over others (Priority). It would follow that, empirically, the impact upon Differentiation can be assessed after the impact upon Elevation has been removed, and the effect on Priority can be gauged properly only after the influence upon Elevation and Differentiation have been eliminated. In regard to the Levels we would account for nation differences in terms of the simplest, most concrete explanations, turning to more complex and more abstract explanations only if the simpler ones are unsuccessful. Thus, we would proceed from the Individual to the Societal Levels.

This Demonstration

The analysis focuses, first, upon a perspective on inter-nation differences not addressed in previous investigations, the distinction among the three Types. Second, we concentrate more upon the Societal than upon the Individual Level. Demographic variables represent Individual Level factors; Societal Level effects are seen as differences among nation samples once Individual Level demographic effects are removed. Further, given the foci of this demonstration, given the many possible relationships between the demographic variables and the three Types, and given the need for brevity, let us simply mention when Individual Level demographic effects are present rather than detail the precise nature of those effects.


Method

Site Features Assessed

As noted earlier, to provide a wide variety of features that consumers might value, the works of Torkzadeh and Dhillon (2002) and Rogers (cf. 1995) were used to suggest the features/functions to include in the set. Based on the earlier analyses by Keeney (1999), Torkzadeh and Dhillon proposed that consumer values serve as yardsticks against which shoppers evaluate a website. Values are either "means objectives" (i.e., instrumental in achieving more basic priorities) or "fundamental objectives" (i.e., the basic goals/practices). Their factor analytic study of ratings of site features identified four fundamental and five means factors. The former were: 1) shopping convenience, 2) Internet ecology, 3) customer relations, 4) product value. The latter were: 5) product choice, 6) online payment, 7) vendor trust, 8) shopping travel, and 9) shopping errors. Two of the domains may pertain to consumer evaluations of online shopping per se, rather than to evaluations of one website relative to other websites. These two, Internet ecology (No. 3) and shopping travel (No. 8), were therefore dropped from consideration.

Next, online shopping is often classified as an instance of the adoption of an innovation (e.g., Citrin et al., 2000; Goldsmith, 2000; 2001). Hence, features that express underlying factors that impact the adoption of innovations might as well attract browsers or buyers to a website. Rogers (1995) has suggested that five characteristics of an innovation can attract or repel adopters: a) comparative advantage over competitors, b) compatibility with one's social environment, c) complexity/simplicity of product use, d) trialability/divisibility which allows the shopper to assess an innovation's value without risking a major commitment, and e) observability of adoption by others in the social environment.

From these points of departure 20 features were developed to represent the consumer values from Torkzadeh and Dhillon (1-9) and the innovation characteristics in the Rogers perspective (a-e). The specific operational definitions employed were selected to be consistent with prior empirical demonstrations of factor impact on website appeal, as listed earlier (in Table 1).6

Wide selection and variety of products (5) Easy to find the product looked for (1)
Good price incentives (coupons, frequent shopper programs, etc.) (4a) Fast response time from customer service (3)
Customer feedback (i.e., site provides a place to learn about other customers' evaluations of the product) (d) Return policy is easy to understand and use (3)
Reputation and credibility of the company on the web (7) Credit card safety (c)
Order process is easy to use (1) No tax (4,a)
No language barrier (1,b) Good place to find a bargain (4,a)
Download speed of page (1) Low or no charge for shipping and handling (a,4,a)
Delivery time is short (a) Product information including FAQ's (d)
(My) friends and family have been happy shopping there (b) Friends and family will like to know (my) opinion of the site (b)
Website is new and different (c) Hear about it on radio, TV or in newspapers (e)

Table 2. Site features considered


Questionnaire

Within a longer multipurpose survey were the twenty features, for which respondents were asked: "How much would the following encourage you to shop (seek information, make a purchase, etc.) at a particular website? For each item, indicate if it would attract you: not at all (1), a little (2), moderately (3), a lot (4)."

Respondents were also asked : "On average, how many hours per week, if any, do you use the Internet?" Responses were: "0," "1-5," "6-10," "11-15," "16-20," "21 or more." They also reported their gender, age, education, and employment status.

The questionnaire was translated into Farsi, Mandarin Chinese, and German. Multiple back translations were conducted to establish comparability between the English and the non-English versions. The original translator was a native born citizen of the nation in question, speaking English as a second language, as were all but one of the back translators. A minimum of two back translations was required before the non-English version was judged comparable to the English one.

Data Collection

An online survey was not used, for it was anticipated that light users of the Internet might be somewhat intimidated by an online survey, particularly in nations in which relatively few use the Internet. An online survey, then, might result in a nation sample seriously biased toward more sophisticated users. Instead, a self-administered print questionnaire booklet was employed.

A convenience sample of adults completed the self administered questionnaire. A cover letter explained that the study, performed by University researchers, was meant to explore the factors underlying Internet usage and shopping. The anonymous and voluntary nature of participation was noted. Questionnaires were distributed by the investigators' graduate students and colleagues to co-workers, families, friends of family members, etc. The objective was to obtain as diverse a sample as possible in demographics, in Internet use, and in online shopping. It was stipulated that respondents were to be native to the nation in question (no temporary visitors, etc). When completed, questionnaires were placed in a sealed envelope and returned to a central location for transmittal to the investigators' university.

Questionnaires were distributed in Toronto, Canada, in Tehran, Iran, in Taipei, Taiwan, in Vienna, Austria, and in the Greater Cleveland, Ohio, region of the USA. Lynch and Beck (1999) observed that a broad sample of nations is needed to assess cross-national differences in the appeal of websites. They found large differences among geographic regions of the world, but small differences among the contiguous nations comprising a given region. Such a broad selection was used in the present demonstration.


Results & Discussion

Sample

A total of 670 respondents qualified as having ever used the World Wide Web and also provided the relevant demographic information. Table 3 presents the demographic and Internet usage characteristics of each nation sample.

The sampling procedure was successful in obtaining samples of the type needed to illustrate the framework, i.e., one diverse in demographics and usage within each nation sample. Further, the differences among the nation samples were as one would expect; Internet users tend more often to be male, younger, and better educated in nations outside the USA, while in the USA users tend to be more balanced in gender, age, and education (e.g., Taylor Nelson Sofres Interactive, 2000).

Characteristics
 USA
n = 235		 Austria
n = 121		 Canada
n = 85		 Iran
n = 94		 Taiwan
n = 135		 Total
n = 670
Gender Male 44.7 62.8 52.9 61.7 54.1 53.3
Female 55.3 37.2 47.1 38.3 45.9 46.7
Age 18/less 1.3 1.7 3.5 6.4 2.3 2.6
19-24 19.7 41.3 47.1 34.0 39.8 33.2
25-30 17.2 22.3 18.8 14.9 21.1 18.8
31-40 22.7 19.0 16.5 17.0 21.1 20.1
41-50 25.8 9.9 10.6 23.4 12.8 18.0
51/over 13.3 5.8 3.5 4.3 3.0 7.4
Education Some High School 2.1 .9 1.2 5.4 1.5 2.1
High School Grad. 14.1 14.9 20.2 21.5 6.8 14.6
Tech. School 3.4 31.6 11.9 0.0 10.5 10.3
Some College 30.3 34.2 29.8 4.3 20.3 25.2
College Grad. 23.9 10.5 29.8 34.4 52.6 29.6
Grad. School 26.1 7.9 7.1 34.4 8.3 18.1
Employment Full-Time 75.0 53.8 51.2 10.6 53.8 54.8
Part-Time 9.5 9.4 6.0 44.7 2.3 12.6
Temp. Unemployed .9 1.7 2.4 4.3 3.0 2.1
Self-Employed 5.2 .9 1.2 4.3 2.3 3.2
Student 6.5 32.5 36.9 30.9 37.9 24.7
Housewife 2.6 0 2.4 4.3 .8 2.0
Retired .4 1.7 0 1.1 0 .6
Internet Use (hours per week) 0 4.7 4.1 0 6.4 1.5 3.6
1-5 51.3 34.1 45.3 36.2 45.5 44.1
6-10 19.1 28.5 26.7 30.9 27.6 25.1
11-15 13.6 15.4 16.3 10.6 9.7 13.1
16-20 2.1 7.3 10.5 7.4 6.7 5.8
21+ 9.3 10.6 1.2 8.5 9.0 8.3

Table 3. Percent of sample with each demographic and Internet usage characteristic.


The Framework

         Individual Level

As discussed earlier, Individual Level demographic effects were estimated by pooling the five nation samples into a single composite sample group.7 Since the nation samples differed greatly in size, simply pooling all respondents would allow the larger samples to unduly influence the analysis. Accordingly, the samples were weighted so as to equate the n's. The weighting coefficients were: USA .61, Austria 1.47, Canada 1.85, Iran 1.67, Taiwan .94.

The demographics were converted to dummy variables, with the lowest level of age (18 and under) and education (less than high school) serving as the excluded group. The "other employment" status (including student, homemaker, retired, etc.) was also excluded. Females were scored "1," males "0." The 13 dummy predictors - five age, five education, two employment, and one gender level - were entered as a single step in the OLS regression. The analysis was based upon 490 cases providing a complete set of responses.

         Elevation

A person's Elevation score was indexed as the mean of that person's 20 feature ratings on the attractiveness scale. The Individual Level demographic multiple regression was then conducted with the pooled nation set. Demographic effects were significant (R = .268, F = 2.84, df = 13/476, p = .001), although the effect was modest (R2 = .072, adjusted R2 = .047). Across nations, then, a person's Elevation score was influenced by that person's demographic profile. As discussed previously, in light of this report's focus on Societal Level differences and of the need for brevity, let us simply note that Individual Level effects appear.

A three-step procedure was then used to calculate Societal Level effects when these impacts of Individual Level demographics were controlled. First, a residual Elevation score was computed for each respondent. This was done by subtracting from one's Elevation score the Elevation score for that person as predicted from his/her demographic profile. Next, a 5 (Nation) X 2 (Internet Usage) between subject ANOVA was computed with the residual scores as the dependent variable. The Usage variable was a comparison of light (5 or fewer hours per week) and heavy (over 5 hours per week) users. Then the Duncan test was used post hoc to isolate the components of the significant main effect of Nation.

Table 4 contains the mean residual scores and cell sizes. The ANOVA, based on 443 persons providing complete responses, yielded a significant main effect of Nation, F(4/433) = 13.851, p < .001, and a main effect of Internet Usage, F(1/433) = 13.458, p < .001. The Nation X Internet Usage interaction was not significant, F(4/433) = 1.319, p = .262.

Nation
Light Usage
Heavy Usage
Total
USA .172 (n = 86) .551( n = 78) .352 (n = 164)
Austria -.332 (n = 18) -.085 (n = 46) -.155 (n = 64)
Canada .268 (n = 19) .407 (n = 35) .358 (n = 54)
Iran -.845 (n = 21) .016 (n = 39) -.285 (n = 60)
Taiwan -.435 (n = 39) -.199 (n = 62) -.290 (n = 101)
Total -.114 (n = 183)  .160 (n = 260)  .047 (n = 443)

Table 4. Mean residual elevation scores and sample sizes in each nation and Internet usage group.


As hypothesized, heavy users wanted more in their website features than did light users. Further, the Duncan post hoc tests of differences among Nations revealed two clusters of means. The Elevation did not differ between the USA and Canada (p > .950), but both were higher (p < .05) than the other three countries. The latter three countries did not differ among themselves (p > .400).

In summary, Elevation depended somewhat upon the Internet user's Individual Level demographic profile. In addition, users in the USA and in Canada had stronger demands (higher Elevation) for website features than did users in Austria, Iran and Taiwan. In all nation samples, heavy users were more demanding than were light users (Individual Level).

         Differentiation

Each person's Differentiation score was the variance of that person's 20 website feature ratings. The sequence of analyses was the same as for the Elevation scores.

The regression of the Differentiation scores on Individual Level demographic variables was significant (R = .230, F = 2.050, df = 13/476, p = .016). The effect was slight (R2 = .053, adjusted R2 = .027) once again.

The ANOVA of the impacts of Nation and Usage upon Differentiation residual scores (See Table 5 for cell means) showed a significant main effect of Nation, F(4/433) = 9.660, p < .001. Internet Usage, F(1/433) = 0, p = .98, and the interaction, F(4/433) = 1.233, p = .296, were not significant. A Duncan test revealed three clusters of nations. Taiwan, the lowest, was significantly lower (p < .05) than the other nations. Next, three nations were not different from each other (p = .094) - Canada, USA, and Iran. The highest cluster was Canada, Iran, and Austria; again they were not significantly different from each other (p = .053). Basically, then, Taiwan showed the least Differentiation, Canada/Iran/Austria the most.

Nation
Light
Usage
Heavy
Usage
Total
USA -.018 -.225 -.116
Austria .210 .453 .385
Canada .016 .106 .074
Iran .340 .052 .153
Taiwan -.613 -.445 -.509
Total -.078 -.071 -.074

Table 5. Means of residual differentiation scores.


In summary, Individual Level demographics had a significant but minimal effect on Differentiation. More importantly for this demonstration, national differences did appear with the Taiwanese showing the least and the Austrians/Canadians/Iranians the most Differentiation.

         Priority

Each individual's appeal ratings were converted to z scores based upon that person's own mean and the standard deviation of his/her 20 feature ratings. The higher a given z score, then, the more appealing that person found that feature relative to other features. These Priority scores then served as the dependent variable in the regression for the Individual Level demographics within the composite nation sample. It was found (see Table 6) that the demographics were predictive of 14 of the 20 features. Across nations, then, the Internet users' demographic profile was related to the features of a website they find comparatively appealing. In general, the amount of variance explained by the demographics was modest, but these results are consistent with the notion that when nations differ in the demographic profile of their Internet users, some of the observed differences between nations in the Priority placed on various site features can be due to these Individual Level demographic differences (e.g., Lynch & Beck, 1999; Weber & Hsee, 1998).

Feature
R
R2
Adjusted
R2
F
df
p
 
1) Wide Selection
.245
.060
.033
2.198
13/449
.009
2) Easy to Find
.222
.049
.021
1.779
13/448
.044
3) Price Incentives
.172
.030
.002
1.059
13/449
.394
4) Fast Customer Service
.224
.050
.023
1.818
13/446
.038
5) Customer Feedback
.232
.054
.026
1.962
13/448
.022
6) Easy Returns
.178
.032
.003
1.124
13/449
.336
7) Company Reputation
.277
.077
.050
2.865
13/447
.001
8) Credit Card Safety
.357
.128
.102
5.015
13/446
<.001
9) Easy to Order
.259
.067
.040
2.477
13/448
.003
10) No Tax
.218
.048
.020
1.720
13/446
.054
11) No Language Barrier
.171
.029
.001
1.026
13/445
.425
12) Good Bargains
.196
.038
.010
1.367
13/446
.172
13) Download Speed
.231
.053
.026
1.916
13/441
.027
14)Low S & H Charges
.318
.101
.075
3.836
13/443
<.001
15)Quick Delivery
.215
.046
.018
1.651
13/443
.069
16) Product Information
.297
.088
.062
3.296
13/442
<.001
17)Friends/Family Shopped
.297
.088
.061
3.302
13/444
<.001
18) Family/Friends Want Opinions
.283
.080
.053
2.957
13/441
<.001
19) New & Different
.325
.106
.079
4.031
13/445
<.001
20) Radio/TV/Press Exposure
.257
.066
.039
2.417
13/445
.004

Table 6. Multiple regression predicting from demographics to priority scores.


The next step was to create residual Priority scores by subtracting the Priority score (in z units) predicted from one's demographic profile from one's obtained Priority score. These residual scores, then, reflected a person's Priority for a feature once the effect of one's Individual Level demographic profile has been subtracted. Next, since the adjustment for demographics rendered the z scores no longer strictly ipsative, it was possible to factor analyze the residual Priority scores to find underlying dimensions or groups of associated features. Again, the nation samples were weighted to allow the smaller and the larger samples to have a more equal contribution to the results of the factor analysis. A varimax rotation of principal components indicated that seven factors underlay the residual Priority scores, accounting for 57.7% of the variance (see Table 7). Inspection of the key loadings led us to define the factors as follows:

Factor 1: "Hot and Trendy." Higher scorers are more strongly drawn to a site that is new and different, that has TV/radio/press exposure, and about which friends/family want to know one's opinions. Higher scorers are less drawn to a site with: low shipping and handling charge, credit card safety, no tax, and easy ordering. It seems that higher scorers want a site that is new and hot with a lot of "buzz" (both local word of mouth and media). They are willing to give up some convenience and even security to get it.

Factor 2: "Product Choice." Higher scorers show more demand for a site on which it is easy to find a desired product and which offers a wide selection of products. These high scorers are less concerned that their friends and family have been happy shopping there. High scorers appear, then, to want a site with good product choice, and find it especially appealing when it is not the "same ol', same ol'" favored by friends and family.

Factor 3: "Back up Service" pertains to how much one wants fast customer service, an easy return policy, and extensive product information/FAQ's. Lower factor scorers find this more appealing.

Factor 4: "Language Over Quick Delivery" When ordering directly from a foreign manufacturer a product that is not readily available locally, one may encounter a language barrier but might receive the order more quickly than would be possible from a local source. Ordering that hard-to-get product from a domestic manufacturer may avoid the language problem, but may require a more complex ordering/delivery process. Factor 4 is composed of two items; higher scorers find the absence of a language barrier more appealing, but are less demanding of quick delivery time.

Factor 5: "Good Prices." Good price incentives and good bargains strongly draw those with a low score on this factor.

Factor 6: "Company Credibility." A company's reputation is a stronger draw to those with a low factor score. It can be hazardous to label a single-item factor, but we do so because: 1) it is axiomatic that company reputation is an important decision factor in choosing to purchase at a company's website (e.g., Balabanis & Reynolds, 2001); 2) the factor does correlate strongly with an item that seems to tap credibility, i.e., the referral by friends and family who shop there.

Factor 7: "Customer Ratings." High scorers want more customer feedback and are willing to sacrifice quick download time to get this feedback.

Feature
Factor 1
Factor 2
Factor 3
Factor 4
Factor 5
Factor 6
Factor 7
Communality
19. New & Different
.656
-.212
.343
.052
.158
.208
.083
.670
20. Radio/TV /Press Exposure
.619
-.138
.180
.169
.289
.167
.109
.587
14. Low S & H Charges
-.541
-.359
-.219
-.259
.036
.426
.053
.722
8. Credit Card Safety
-.531
.214
.288
-.279
.263
-.260
.027
.626
10. No Tax
-.527
-.061
.255
.046
-.098
.254
.228
.475
18. Family/ Friends Want Opinion
.459
-.415
.120
-.275
-.197
.291
.160
.622
9. Easy to Order
-.414
.121
.353
-.067
.199
-.003
.210
.399
2. Easy to Find
.308
.551
.098
-.165
-.057
.307
-.219
.581
17. Friends/ Family Shopped
.226
-.502
.097
-.273
-.301
-.433
.011
.665
1. Wide Selection
.393
.444
.039
-.219
-.021
.274
.077
.482
4. Fast Customer Service
.032
.369
-.516
.135
-.223
-.132
.152
.512
6. Easy Returns
-.217
.185
-.413
.227
.200
-.164
-.309
.466
16. Product Information
.360
-.130
-.406
-.037
.178
-.059
-.288
.431
11. No Language Barrier
-.262
-.134
.106
.568
.086
-.147
.220
.498
15. Quick Delivery
.380
-.322
-.355
-.467
-.051
.237
.033
.653
3. Price Incentives
-.081
.470
.051
-.031
-.598
.018
.027
.590
12. Good Bargains
-.133
-.131
.256
.394
-.581
-.034
-.147
.616
7. Company Reputation
-.232
.193
.304
-.194
.229
-.497
-.389
.671
5. Customer Feedback
.089
.239
-.406
.187
.230
-.161
.539
.634
13. Download Speed
-.212
-.391
-.055
.418
.057
.225
-.459
.640
Eigenvalue
2.780
1.977
1.574
1.422
1.297
1.277
1.124
% Total Variance
9.044
9.035
8.557
8.328
8.105
7.859
6.776

Table 7. Loadings and variance explained for each rotated factor.


A person's score on each of the seven factors was that respondent's Priority score. These scores previously had been adjusted for Individual Level demographic impacts and thus differences among national samples are attributable to Societal Level nation differences. Accordingly, the seven Priority scores were entered as dependent variables into a 5 (Nation) X 2 (Usage) between subject multiple analysis of variance (MANOVA). The cell sizes were the same as used in the prior ANOVAs.

Box's test of equality of covariance matrices was not significant (M = 288.297, F = 1.022, df = 252/33228.942, p = .391), permitting the assumption of equal covariance matrices across groups. The MANOVA results (see Table 8) indicated that the main effect of Nations was significant, but Usage was significant neither as a main nor as an interaction effect.

Effect
Value
F
df
p
Nation:
Pillai's Trace
.372
5.486
28/1500
.000
Wilks' Lambda
.661
5.834
23/1342.687
.000
Usage:
Pillai's Trace
.008
.422
7/372
.889
Wilks' Lambda
.992
.422
7/372
.889
Nation x Usage:
Pillai's Trace
.105
1.438
28/1500
.065
Wilks' Lambda
.899
1.443
28/1342.687
.064

Table 8. MANOVA of nation and usage effects on the seven priority scores.


Since the MANOVA indicated significance, a 5 (Nations) x 2 (Usage) ANOVA was computed separately for each Priority score. When the main effect of Nation was significant, a Duncan post hoc test was performed to find which particular nations differed from others on that Priority.

         Priority No. 1: Hot & Trendy

The ANOVA indicated a significant main effect of Nation, F(4/378) = 6.864, p < .001. The Usage main effect, F(1/378) = .320, p = .572, and the interaction, F(4/378) = .085, p = .987, were not significant. The means for the significant Nation main effect were:

USA
.238
Canada
-.061
Iran
-.253
Austria
-.328
Taiwan
.432
Total
.094

Table 9. Cell means on priority no. 1: Hot and Trendy.


The Duncan test again sought groups of nations which did not differ from each other in Priority, but differed from other nations. The strongest appeal of Hot & Trendy features was to Taiwan and the USA who did not differ from each other (p = .257), but were higher than other groups (p < .05). The appeal was least to Canada, Iran, and Austria who did not differ, p = .14, from each other, but were lower than the other groups.

          Priority No. 2: Product Choice

The ANOVA indicated that there was a Nation main effect, F(4/378) = 5.355, p < .001; the nations were different in the mean demand for the Product Choice features. The Usage main effect, F(1/378) = .287, p = .592, and the interaction, F(4/378) = .866, p = .484, were not significant. The Nation mean Priority scores were:

USA
.154
Canada
.192
Iran
-.459
Austria
.259
Taiwan
-.192
Total
.018

Table 10. Cell means on priority no. 2: Product Choice.


In the Duncan test a higher attraction to Product Choice was found in Austria, USA, and Canada; the three did not differ from each other (p = .566). Taiwan and Iran had a lower preference than the former countries (p < .05), but did not differ from each other (p = .116).

          Priority No. 3: Back Up Services

Recall that lower scores were indicative of a stronger desire for Back Up Services. The mean Nation Priorities were:

USA
.287
Canada
.232
Iran
-.402
Austria
-.136
Taiwan
-.011
Total
.061

Table 11. Cell means on priority no. 3: Backup Services .


The nations were not equivalent in their preferences for Back Up Services, as shown in the ANOVA. The main effect of Nation was significant, F(4/378) = 6.344, p < .001. The Usage main effect, F(1/378) = .384, p = .536, and the interaction term, F(4/378) = 1.259, p = .286, were not significant.

In the Duncan test, USA, Canada, and Taiwan were equivalent (p= .175) and lowest in their attraction to Back Up Services. Austrians and Iranians, conversely, displayed higher (p < .05) regard for these features. The latter two did differ, p = .113.

          Priority No. 4: Language Over Quick Delivery

The ANOVA revealed another main effect of Nation, F(4/378) = 2.883, p < .05. Again, Usage, F(1/378) = .571, p = .450, and the Nation x Usage interaction, F(4/378) = 2.130, p = .076, did not reach significance. The Nation means were:

USA
-.015
Canada
.111
Iran
-.181
Austria
.315
Taiwan
-.264
Total
-.027

Table 12. Cell means on priority no. 4: Language over Quick Delivery.


The Duncan test showed that Austrian and Canadian respondents were particularly desirous (p < .05) of avoiding language barriers at the expense of delivery time. Conversely, Taiwanese and Iranians opted relatively more (p < .05) for delivery speed and less for avoidance of language barriers. The U.S.A. respondents were mid-range with respect to the factor, not differing significantly from any of the other countries.

          Priority No. 5: Good Prices

Still again, the ANOVA determined that the main effect of Nation was significant, F(4/378) = 10.328, p < .001. Usage, F(1/378) = .255, p = .614, and Nation x Usage, F(4/378) = .855, p = .491, were not. The Nation means were:

USA
.217
Canada
.644
Iran
-.315
Austria
-.307
Taiwan
-.284
Total
.012

Table 13. Cell means on priority no. 5: Good Prices.


It should be recalled that lower scores indicate stronger interest in good prices. Duncan post hoc tests revealed three groups of nations. Canada had the least regard for pricing (p < .05); the USA was midrange, revealing fairly slight interest; the other three nations were equivalently (p=.859) desirous of sites with good prices.

          Priority No. 6: Company Credibility

As will be recalled, the lower the score, the more was the appeal of the company's reputation. In the ANOVA the Nation factor was significant, F(4/378) = 3.317, p < .05. Usage, F(1/378) = 1.213, p = .271, and the interaction, F(4/378) = 1.555, p= .186, were not significant. The Nation means were:

USA
.125
Canada
.304
Iran
-.347
Austria
.026
Taiwan
-.124
Total
.016

Table 14. Cell means on priority no. 6: Company Credibility.


As shown by the Duncan test, company reputation was most important in Iran (p < .05), more important in Taiwan (p< .05), and less important to respondents from Canada, USA, and Austria.

          Priority No. 7: Customer Ratings

Table 15 displays the mean ratings of the heavy and light users in each country.

Nation
Heavy Usage
Light Usage
USA
-.286
.710
Austria
.023
-.452
Canada
-.284
.332
Iran
.215
-.132
Taiwan
.282
.026
Total
-.024
.017

Table 15. Cell means on priority no. 7: Customer Ratings.


The ANOVA indicated a significant interaction, F(4/378) = 3.366, p = .01. Neither the Nation, F(4/378) = 1.329, p = .258, nor the Usage, F(1/378) = .031, p = .854, main effects was significant. The differences among nations varied for heavy and light Internet users. We must be cautious in interpreting this interaction, however, since the earlier MANOVA revealed that the Nation main effect, but not the interaction, was significant.

Among Heavy Users customer feedback is appreciated more in Taiwan and Iran, but comparatively less in the USA and Canada. But among Light Users the U.S.A. is most desirous, Austria the least.

Summary of Societal Level Differences

Nations samples showed differences of all three Types in their attraction to the website features.

         USA

The U.S.A. respondents were comparatively high in Elevation and midrange in Differentiation. They were found to be higher in Priority 1 and 2; low in 3 and 6, and medium on 4 and 5. On Priority 7 they were especially high among Light Users, but low among Heavy Users. In other words, Internet users in the U.S.A. want a wide variety of site features, more than do users in the three overseas nations. They are especially drawn to features that are Hot & Trendy, that provide for wide Product Choice and (among Light Users) Customer Ratings. The Priority they place on these features is stronger than is found in most of the other nations covered. On the other hand, relative to other features, they are less attracted to Backup Services, to Customer Ratings (among Heavy Users) and to Company Credibility; the relatively low priority they place on these features is lower than found in some other nations studied.

         Austria

Austrian respondents are relatively low in Elevation and high in Differentiation. They are low in Priorities 1, 6 and 7 (among Light Users), but high in the others. That is, compared to the other nations studied, they demand fairly little vis-à-vis site features, differentiating widely in their preferences between the features they strongly desire and those to which they are indifferent. Compared to the priorities of other nations, they are fairly unresponsive to Hot & Trendy features, to Company Credibility and to Customer Ratings (among Light Users). They do, however, find the other four types of features (Good Prices, Product Choice, Language Over Quick Delivery, Back Up Services) particularly appealing.

         Canada

Canadian respondents are high in Elevation, midrange in Differentiation, high in Priorities 2, 4 and 7 (among Light Users); and they are low in Priorities 1, 3, 5 and 6. Thus, Canadians, like those in the U.S.A., want a broad range of site features. Compared to the degree of priority accorded to features elsewhere, they are quite unconcerned with Hot & Trendy and Company Credibility, Good Prices, Back Up Services, and Customer Ratings (among Heavy Users). On the other hand, relative to the degree of priority for these types of features found elsewhere, they are strongly drawn to Product Choice, to the absence of language barriers, and (among Light Users) Customer Ratings.

         Iran

Iranian respondents are fairly low in Elevation and midrange in Differentiation. They scored relatively high in Priorities 3, 5 and 6, but low in the other types of features. That is, Iranian users are fairly undemanding in regard to many website features. Compared to other nationals, the comparative emphasis they place on Back Up Services, Good Prices, Customer Ratings (among Heavy Users), and Company Credibility is high. Conversely, on a comparative basis, they are less drawn to Hot & Trendy features, Product Choice, and absence of language barriers.

         Taiwan

Taiwanese respondents are low in Elevation and in Differentiation. Their Priorities are high for Nos. 1, 5 and 6, low for the others. This pattern means that the Chinese in Taiwan see few site features as imperative, and their preferences do not vary greatly from one feature to another. While the priority they place on having Hot & Trendy features, Company Credibility and Good Prices is greater than found in other nations, the comparative importance they give to Product Choice, Language Over Quick Delivery, and to Back Up Services is low.

Implications of the Framework for Behavioral Scientists

The model's relevance to the development of social science theory and marketing science lies in its helping to isolate the components of cross-national differences in website appeal. What is the nature of the differences in website appeal - i.e., are there differences in Elevation? Differentiation? Priority? Are these differences due to the distribution in the respective nations of demographic or other characteristics operating at the Individual Level? Or are the differences due to broader social or cultural factors? We cannot devise viable theories of what underlies a phenomenon until we have identified the nature of the phenomenon to be explained!

An important application is to help an investigator avoid overlooking or misidentifying meaningful differences existing among national populations. Such problems can readily occur when an investigator simply compares various nations in the levels of preference existing therein for a small sample of website features. Illustratively, suppose a survey of preferences in nation A and in nation B for the feature "no shipping/handling charges" uses an index analogous to the 0-10 appeal scale discussed in an earlier illustration, and finds that the feature's average preference in both Nation A and Nation B is 6. Can the conclusion be drawn that the two nations do not differ meaningfully in their demand for that feature? No. It is possible that nation A's shoppers are high (say, 8) in Elevation and that the handling charges feature has a low Priority. Nation B's shoppers can be lower in Elevation (say, averaging 4) and that feature can have a high Priority. In such a case, the feature could be a compelling attraction to shoppers in nation B, but an uninfluential factor in site selection in nation A. This important conclusion would be overlooked if the investigator looked simply at the feature's preference in isolation rather than consider Elevation and Priority.

Implications for Practitioners

The model has action implications for a practitioner tailoring a website to fit the requirements of a given national market (e.g., Lynch, Kent, & Srinivasan, 2001). These propositions are consistent with the conceptual framework and data, but, naturally, require verification in professional practice. Consider first the Type issues. Let us turn to the Levels:
What if the differences between the U.S.A. national market and another national market are mainly due to Individual Level attitudes (i.e., online shoppers with the "right" attitudinal profile are more numerous in one nation rather than another)? If these attitudes are pertinent to the purchase of particular product classes, then users with the appropriate attitudes could be targeted by practitioners. Illustratively, since online shoppers with high domain specific innovativeness are more prone to shop online (e.g., Blake et al., 2003; Citrin et al., 2000; Goldsmith, 2000; 2001) practitioners can tailor the features of a website to fit the most likely buyers, the innovators (Blake, Neuendorf, & Valdiserri, 2004; Citrin et al., 2000; Goldsmith, 2001), whatever their nation of residence. Further, since in this instance the drawing power of a website is responsive to Individual Level attitudes rather than to more Societal Level influences, the practitioner can beta test the website features here in the U.S.A. rather than do so in the other nation.


Conclusions

This demonstration analysis strongly supports the value of the proposed framework. The framework's three Types were shown to be separate forms of national differences in website feature appeal; each has its own important implications for theoretical work and for practice. Similarly, the difference between Individual and Societal Level mechanisms (as distinct from person-specific and aggregate-level variables) has meaningful implications for practitioners devising commercial websites as well as for behavioral scientists attempting to chart the nature of cross-national variation in website appeal.

Further, two taxonomies developed for the framework merit attention. First is the 20-item feature classification operationalizing the concepts of Torkzadeh and Dhillon (2002) and Rogers (cf. 1995). It was found to be a workable analytic platform. Future research, however, might enhance its value by considering ways to incorporate additional issues listed in Table 1, especially interactivity, entertainment value (fun), and flow. The second taxonomy is the seven dimensions underlying feature evaluations emerging from the pooled nation samples. Can they be replicated in a larger nation sample? Can they provide a point of departure for attempts to devise a measure suitable for gauging website effectiveness, one that is internationally applicable?

More research with the framework is needed, especially with larger representative samples of respondents and with more nations, before we can identify the precise nature of these inter-nation differences. Further, conceptual and empirical efforts should specify the key Individual and Societal Level mechanisms impacting website appeal. We believe that, given the results obtained here and given the importance of the theoretical and practical implications noted above, the framework merits attention in future investigations.


Acknowledgments

We thank the following for their efforts in collecting the data: Fariba Arab, Suzanne Grilli, Allison Wright-Frazier, Ruben Jurik, Chia Chi Liu, and Colin Valdiserri. Copies of the survey items and their Chinese/Farsi/German translations as well as the particulars of the regression analyses of the demographics are available from the authors.


Footnotes

1. See, for example, AC Nielsen at www.acnielsen.com , Taylor Nelson Sofres at www.tnsofres.com , Forrester at Forrester.com; E-commerce Times at www.ecommercetime.com/perl, and Graphic, Visualization, & Usability Center's (GVU) WWW User Surveys at www.cc.gatech.educ/gvu/user_surveys ).

2. Specifying characteristics of the populations (online shoppers in a given nation) is challenging. Not only are these populations rapidly evolving, but in some nations (e.g., Iran) population figures are unavailable or suspect. Still further, as we noted elsewhere, attempting to define the online shopper population by its demographic profile is becoming less and less viable.

3. Our conceptualization of Societal Level mechanisms includes both cultural values and national infrastructure factors. Generally, culture is the deposit of knowledge, habits, traditions, customs, norms, beliefs, and values that are accumulated and passed down over the course of generations (Padilla, 1980; Prosser, 1978). Cultures operate within nations; laws and practices established by a given nation will moderate the manifestations of culture (Maldonado & Tansuhaj, 1999; Penaloza, 1989).

4. There may be indirect impact of the society/culture in that the distribution of a demographic can be due to societal level events. For example, the dearth of middle-aged males in France in the decades after World War I and in Iran at the turn of the 21st century was due to extensive military fatalities.

5. This example makes it clear that in our framework "individual level" pertains to mechanisms or processes rather than to variables per se. Here, gender's impact reflected societal forces operating within a specific nation and thus was not explainable as a person characteristic divorced from the particular societal context in that nation.

6. Several points should be noted about this listing. First, the theoretical categories were used to provide broad coverage; the listing is not exhaustive or mutually exclusive. Second, it is not assumed that features representing the same dimension in the Torkzadeh and Dhillon (2002) or in the Rogers perspectives will necessarily correlate highly or form a common dimension. For example, within the Torkzadeh and Dhillon (2002) "product value" category or within Rogers' (1995) "comparative advantage" would fall "high quality goods" and "low prices." However, in the eyes of shoppers often high quality and low prices do not co-exist in the same product (e.g., Dodds, Monroe, & Grewal, 1991). Third, the two previous typologies focus upon "functional" rather than "specific" features, e.g., "provides good value" rather than "provides discounts." A practitioner can determine how to structure a website knowing the concrete features demanded by the target market. Knowing the functions desired, however, can leave ambiguity about how to provide it. For example, is "good value" best achieved by low prices? By larger quantity? By better quality? Fourth, the list of 20 features does not necessarily include a variety of factors suggested to influence a site's appeal (see Table 1) because: 1) often the factors are functional rather than concrete in nature, 2) the factors may represent concepts with which a survey respondent might be unfamiliar, and so cannot reliably evaluate, and 3) they are not strictly necessary if the principal goal of the 20 features list is to provide broad coverage.

7. By using a composite data matrix, the analysis estimated Individual Level demographic effects, i.e., effects common to the nations as a composite group. Such an analysis does not necessarily capture Societal Level demographic processes. That is, if residual scores were calculated on a dependent variable (e.g., respondents' Elevation scores) by subtracting the score predicted by the demographics from the observed score, within a specific nation sample the residual scores might still be predicted by demographic variables. These predictions would be different from one nation sample to another and thus would be seen as due to Societal Level demographic effects.


References

Atkin, D. J., Jeffres, L. W., & Neuendorf, K. A. (1998). Understanding Internet adoption as telecommunications behavior. Journal of Broadcasting and Electronic Media, 42, 475-490.

Atkin, D., & LaRose, R. (1994). An analysis of the information services adoption literature. In J. Hanson (Ed.), Advances in telematics (Vol. 2, pp. 91-110). New York: Ablex.

Balabanis, G., & Reynolds, N. L. (2001). Consumer attitudes toward multi-channel retailers' websites: The role of environment, brand attitude, Internet knowledge, and visit duration. Journal of Business Strategies, 18(2), 105-131.

Bhatnagar, A., Misra, S., & Rao, H. R. (2000). On risk, convenience, and Internet shopping behavior. Communications of the ACM, 43(11), 98-105.

Blake, B. F., Neuendorf, K. A., & Valdiserri, C. M. (2003). Innovativeness and variety of Internet shopping. Internet Research: Electronic Networking Applications and Policy, 13(3), 156-169.

Blake, B. F., Neuendorf, K., & Valdiserri, C. (2004, in press). Tailoring new websites to appeal to those most likely to shop online. Technovation.

Blake, B. F., Neuendorf, K. A., Valdiserri, C. M., & Hughes, J. (2004). The online shopping profile in t