Mapping Diversities and Tracing Trends of Cultural Homogeneity/Heterogeneity in Cyberspace

Authors

  • Elad Segev,

    1. School of Politics, International Relations, & Philosophy
      Keele University
    Search for more papers by this author
    • Elad Segev is a doctoral candidate at Keele University, a lecturer of Media and Communications at Aston University, and a former research student at Tel Aviv University in the Faculty of Management. His research deals with technology in general and the Internet in particular, and its social, political, and cultural implications.

      Webpage: http://www.eladsegev.com

      Address: School of Politics, International Relations and Philosophy (SPIRE), Keele University, Keele, Staffordshire ST5 5BG, UK

  • Niv Ahituv,

    1. The Faculty of Management
      Tel Aviv University
    Search for more papers by this author
    • Niv Ahituv is the Marko and Lucie Chaoul Chair for Research in Information Evaluation and the Academic Director of the Institute of Internet Studies of Tel Aviv University. His research deals with information economics, information system strategy and management, and social and business implications of the Internet.

      Webpage: http://recanati.tau.ac.il/Index.asp?ArticleID=76&CategoryID164

      Address: Faculty of Management, Tel Aviv University, Tel Aviv, 69978, Israel

  • Karine Barzilai-Nahon

    1. The Information School
      University of Washington
    Search for more papers by this author
    • Karine Barzilai-Nahon is Assistant Professor in the Information School at the University of Washington. Formerly she held senior positions in Research and Development in hi-tech industry. Her research deals with social and business aspects of the Internet and telecommunications.

      Webpage: http://projects.ischool.washington.edu/karineb/

      Address: The Information School, University of Washington, Mary Gates Hall, Seattle, WA 98195, USA


Abstract

Many Internet global commercial services were originally initiated in the United States and later exported to other countries, continents, and cultures. By analyzing content and form characteristics of two leading portal sites—MSN and Yahoo!—on 33 and 23 (respectively) of their local country homepages in a comparative and longitudinal study over a period of six months, this study addresses two research questions: What differences, if any, are found among homepages of the same parent site located in different cultures (or targeted at audiences in different cultures)? What trends, if any, do those differences show over time? Although MSN is a U.S. brand, the analysis of its homepages reveals increasing cultural heterogeneity and localization of content and form. In contrast, Yahoo!’s homepages are found to be more similar to each other and to the “parent” American portal. A metric was developed to measure the distance between various websites in terms of form and content, along with a structured procedure to analyze and cluster groups of websites. The results suggest that the diversity of local homepages in MSN and Yahoo! follows the geographic and ethnographic scatter of their countries.

Introduction

Many Internet global commercial services (news portals, ISPs, weather portals, and the like) were originally initiated in the United States and later exported to other countries, continents, and cultures. An interesting question is to what extent cultural variance is reflected through the Internet and whether one dominant culture, such as that of the U.S., enforces cultural homogeneity in cyberspace. The Internet is a unique cultural space in that it combines a variety of means and forms of communication (e.g., mutual interaction, broadcasting, self-searching for information, forum groups, human computer interaction) and different types of content (e.g., text, image, voice, video). This provides a wide range of opportunities, which in turn have significant social and cultural implications (DiMaggio, Hargittai, Neuman, & Robinson, 2001).

Cultural Diversity in Cyberspace

Some scholars have suggested that mass communication channels shaped by early media forms (e.g., written press, radio, television) might create a global uniform taste. They argued that commercial considerations and dependence on advertisements might produce homogeneous and popular programs or texts in order to attract the attention of maximally large audiences (Adorno, 1991; Bagdikian, 2004; Mattelart, 1991; Shils, 1963). Eventually, together with the advent of the Internet and the increasing availability and openness of information (Ahituv, 2001), some media channels would be likely to represent a certain universal homogeneity. This claim stands in contrast to the increasing customization of production (Caves, 2000; Shapiro & Varian, 1999; Sunstein, 2001) and the long-tail theory, which shows that profit follows the economics of abundance (Anderson, 2006).

Other researchers suggest that economic imperatives will keep the Internet from realizing its technical potential as a source of cultural abundance (Castells, 1996; Neuman, McKnight, & Solomon, 1998). They indicate that although entrance barriers to the network may be low, competitors can bias the attention of users and consequently confine it (Barzilai-Nahon, 2006). In fact, 80% of visits are to only 0.5% of websites (Waxman, 2000). Similarly, some scholars suggest that online information is not equally spread throughout the global network but rather is channeled by economic, technological, cultural, and social factors (Barabási, 2001). From this perspective, online information is controlled by dominant actors and is concentrated in “live-zones,” i.e., information-saturated centers that have a global cultural impact (Kellerman, 2002; Lash, 2002). Moreover, evidence has been reported that the English language dominates online content out of proportion to its share of native speaker Internet users (Global Reach, 2004).

The question is whether the Internet promotes individual and local expressions and therefore heterogeneous culture, or whether it serves as an instrument for dominant organizations (and nations) to increase markets and reinforce common, homogeneous dispositions. This study proposes quantitative tools to explore cultural heterogeneity through popular online artifacts. The idea of exploiting quantitative tools to study cultural differences is not new in the social sciences; however, the attempt to establish a metrics of distance between the content and form of different websites is quite new.

A recent study by Norris (2004) measured the bridging and bonding functions of various online groups in the context of ideological, political, and social homogeneity and heterogeneity. Her study used an ordinary least squares (OLS) regression model to predict these measures, while taking into account cultural attributes such as interests, lifestyle, religious group, and more.

Hofstede (1997) suggested that national culture has significant implications for organizational behavior. He defined culture as “the collective mental programming” (p. 5) that occurs along various dimensions. Hofstede measured cultural differences in the workplace using survey data from IBM employees in over 50 countries. Ultimately, he designed a special index for each dimension and ranked the countries by their cultural differences in the workplace.

Content and Form as Cultural Artifacts

Various recent studies have implemented Hofstede’s (1997) framework to explore the design and content of websites in different countries (Callahan, 2005; Cyr & Trevor-Smith, 2004; Lo & Gong, 2005; Sheridan, 2001; Zhao, Massey, Murphy, & Liu, 2003). Their findings suggest that differences in website form and content reflect a variety of cultural attributes. Relatedly, a recent study by Würtz (2005) compared the form and content of transnational corporation websites in different countries. It revealed that the websites of high-context cultures (Hall & Hall, 1990) such as Japan, China, and Korea, in which communication is relatively indirect and rich in gestures, displayed more extensive use of graphic elements and indirect messages. By contrast, websites of low-context cultures, such as Germany, Norway, and the U.S., in which communication is relatively direct, were more static and displayed rather direct messages. In terms of content, it was found that websites of high-context cultures tended to emphasize family and social values, while websites of low-context cultures tended to emphasize freedom and personal lifestyle. Similar trends were observed in other studies (Leonardi, 2002; Marcus & Gould, 2000), indicating that different national websites place different emphases on collective and individual values.

While a growing number of studies have focused on website design and layout (e.g., animations, colors, graphical density, and navigation menus) as cultural markers (Badre, 2000; Becker, 2002; Callahan, 2005; Cyr & Trevor-Smith, 2004; Sun, 2001; Tsui & Paynter, 2004), fewer studies have been conducted on differences in content and their cultural implications. Most content analyses are based on Hofstede’s (1997) frameworks and thus are limited in focus.

Robbins and Stylianou (2003) looked at the information provided by the largest global corporation websites. They found that website content is significantly different across national cultures. Singh and Baack (2004), Singh, Furrer, and Ostinelli (2004), and Singh, Xhao, and Hu (2003) compared the content of American, Mexican, European, Indian, and Chinese websites, identifying significant differences in emphasis on family and social values, masculinity, and power. Additionally, the authors argued that online users prefer localized over standardized web content.

In contrast to cultural sensitivity and localization of content and form, Kang and Corbitt (2001) revealed trends of homogeneity and standardization in web design. They examined the differences in display colors, images, animation, and functionality of 10 U.S.-based IT companies and their Australian and Singaporean versions. Their study concluded that local homepages used a design similar to the parent U.S. website, taking only minor account of cultural issues.

Finally, a study by Schmied (2003) looked at websites of transnational corporations, and particularly websites of airline companies, finding certain trends of standardization in content and form, but at the same time identifying distinctive local cultural markers. In short, there are very few studies that indicate trends of homogeneity in web design and content, and even then their findings still suggest that there are local elements that cannot be ignored.

By exploring cultural diversity through differences in the content and form of national homepages, it is presumed that online culture may be reflected through content and form. Barber and Badre (1998) defined the term “culturability” as the relationship between culture and usability in the design of webpages. They argued that the interface design, interactivity and content reflect cultural sensitivity and understanding of the targeted audience. Their study indicated that there are design patterns (or what they also called “cultural markers”), which are highly prevalent within different cultures and different genres. Their study indicates, for example, that Brazilian websites tend to be more colorful and Lebanese websites tend to use light graphics and more text (Badre, 2000; Barber & Badre, 1998). Similarly, the present study uses the term “culture” somewhat loosely, as a means to distinguish among the different countries and their respective websites. There is no intention to explore the various aspects of this term but rather to enable a comparison of national websites as cultural products or “artifacts” (Appadurai, 1996) with different form and content attributes. Thus, the term “online culture” is used here to refer to national cultures as reflected on the Internet.

While most of the aforementioned studies compared content and form elements using Hofstede’s framework, in this study we develop an original metric to measure the diversity of content and form online. We choose to study the Internet portals MSN and Yahoo! mainly due to their significant rating and popularity in the observed countries (Nielsen/Netratings, 2004; see also Appendix C). We focus on the differences in their local homepages, assuming that the previous decisions of web designers and corporate strategies are incorporated in their content and form, and that they are constantly affected by the values and norms of their local audiences.

Van Couvering (2004) suggested that MSN and Yahoo! conform to the mass communication business model known as the “dual product” market. On the one hand, they produce cultural artifacts for individuals, and on the other hand, they produce advertisements. But here the business models of MSN and Yahoo! differ, in that MSN gets 29% of its revenue from advertisements, while Yahoo! gets 82% of its revenue from advertisements (Van Couvering, 2004). These different revenue strategies also suggest that MSN benefits more from direct subscription, and therefore compels its customers to pay more attention to local and customized cultural artifacts, than its competitor Yahoo!. Consequently, it was expected that the content and form of national homepages in MSN would be more diverse and localized than in Yahoo!. The results of this study strongly support that hypothesis.

Research Questions

This article addressed two primary research questions:

RQ1: What are the differences in content and form among local homepages of the same parent site?

RQ2: What trends, if any, do these differences display over a period of time?

The first question explores diversity and raises sub-questions such as: What are the similarities of each homepage to the parent U.S. site? Is there any relation between geographical or ethnographical diversity and online cultural diversity? The second question explores trends and raises sub-questions such as: Is it possible to trace trends in content and form of popular Internet portals? If so, are they progressing toward cultural homogeneity or toward heterogeneity?

Since both MSN and Yahoo! originated in the U.S., we expected to find cultural homogeneity in local homepages, reflecting the form and content of the parent U.S. site, and an increasing similarity in their form and content over time.

Methodology

In order to capture the content and form of different local homepages, an observation item list was constructed. The item list is an electronic form that was used to report content and form in the different local homepages of MSN and Yahoo!. One month of observations (May 2003) served as a pilot sample to fine-tune the item list. Subsequently, samplings of content and form were taken over a period of six months (June-November 2003) from the 33 MSN local homepages and 23 Yahoo! local homepages (see Appendix A for the full list of countries). Since six months is a relatively short period, we focus our analysis on the diversity of local homepages, and suggest implications concerning their trends mainly to encourage further investigation in this domain.

The content and form elements of the item list that served as a basis for comparison across local homepages followed two main principles: consistency and remarkableness. Consistency refers to the form and content elements that appear in all or most local homepages. Thus, it is useful to compare the number of pictures in different local homepages only if it is a consistent element, that is, it appears in all or most local homepages. It is less useful to compare elements that are not consistent, such as the number of banners with sound, which hardly ever appear.

Remarkableness refers to the form and content elements that are most distinguishable and that are referred to as “cultural markers” by Badre (2000). Some form elements, such as banners and active (animated) banners, were identified in previous studies as remarkable form elements (Cyr & Trevor-Smith, 2004; Lo & Gong, 2005; Würtz, 2005; Zhao et al., 2003); these may further reflect high-context cultures (Hall & Hall, 1990). Similarly, number of links, number of menus, number of headlines, search box, and page layout have been identified as remarkable form elements with distinctive cultural attributes (Barber & Badre, 1998; Callahan, 2005; Cyr & Trevor-Smith, 2004).

Based on previous studies of “cultural markers” and our ongoing observations of MSN and Yahoo!, we identified 10 form elements for examination: number of frames, number of banners, number of active banners, number of background colors, number of photos, number of main links (headline links), number of links in general, size of search box, number of menus, and number of pop up windows. Each one of these elements was further operationalized in detail; thus, for example, the number of pictures is the number of any graphics above 40 pixels. Although one researcher conducted the coding, the detailed operationalization of each element helped to maintain the consistency of the coding process.

Measuring content quantitatively is often a less straightforward process. Here we focus on topical classification of seven categories: news, economics, entertainment, sports, shopping, technology, and general information. These topics were chosen due to their high frequency of appearance in all local homepages. Classifying content by topics was a natural choice also because both MSN and Yahoo! use topical classification to differentiate graphically and conceptually among different kinds of content.

Excluding rare exceptions, most content could be sorted into those seven categories. Each category had structured and detailed rules regarding the inclusion of content in it. For example, links to purchase products and services would be categorized under the “shopping” category. A particular challenge was to define the rules of inclusion and exclusion for the category “news,” which can be relevant for many different topics, such as economics or sports. In order to maintain the consistency of the classification process, it was decided to count as news only content that deals with politics and society in general.

All local homepages were translated by means of online dictionaries into English. Since local homepages in MSN and Yahoo! consist of very short sentences and are often already graphically divided into topics, the classification of content into the seven broad categories was straightforward. Additionally, a research student was employed to validate the online translation of the Japanese, Chinese, and Korean homepages.

Apart from the general number of links from each category, content was further subdivided by its location and function. Thus content variables could be, for example, the number of news, economics, or sports headlines, or the number of entertainment or shopping banners (see Appendix B for the full item list).

To conclude, 44 different attributes of content and form were analyzed monthly over a period of six months in 33 countries for MSN and 23 countries for Yahoo!. In order to explore the diversity of content and form, three indices were developed (General Diversity Index, Difference Diversity Index, and National Similarity Index) and a structured statistical analysis was employed (factor and cluster analyses). Both indices and statistical analyses examined the diversity of local homepages of the same parent site and its trends over time; thus they complemented and supported each other.

Three Diversity Indices

The first index, the General Diversity Index (GDI), was designed to explore the diversity of the content and form of local homepages. It is based on calculating the coefficient of variation of the form and content variables through three steps:

  • 1Compute the average and standard deviation across all homepages for each independent variable.
  • 2Compute the scattering index of the variable (standard deviation/average), also known as the coefficient of variation.
  • 3The first index, the GDI, is derived from the average scattering index of the variables of form and content separately.

Table 1 demonstrates how to calculate the GDI index for a partial sample taken from actual observations of MSN in June 2003.

Table 1.  First index (GDI)—Sample from MSN, June 2003
CountryBannersActive BannersBackground ColorsPictures 
U.S.0046 
Arabia2248 
Italy0058 
Netherlands21315 
Norway21414 
Spain5039 
U.K.3146 
Brazil2248 
Canada (English)0046 
Mexico62312 
India3354 
Korea11410 
Singapore4247 
Australia1035 
Taiwan4240 
Average2.331.133.877.87 
Standard Deviation1.840.990.643.85 
Scattering Index0.790.870.170.490.58

The second index, the Difference Diversity Index (DDI), was designed to explore the diversity of the absolute differences between local homepages and the parent U.S. homepage. Similar to the GDI, it is based on calculating the coefficient of variation through the following steps:

  • 1Compute the absolute differences between each local homepage and the U.S. homepage for each observation.
  • 2Compute the average and the standard deviation among all homepages for each independent variable.
  • 3Compute the scattering index of the variable (standard deviation/average), also known as the coefficient of variation.
  • 4The second index, the DDI, is derived from the average scattering index of the variables of form and content separately.

A trend of decline in the second index does not indicate that the local homepages become similar to the U.S. homepage. It only indicates that the absolute differences become more homogenous. The full meaning of the second index will be further discussed in the analysis of the results.

The first and the second indices engender a dimensionless number, which is meaningful only if compared over a period of time to track diversity trends. Since the calculation of indices is based on the coefficient of variation, it can be implemented easily in other fields to explore the scattering level and its trends.

The third index, the National Similarity Index (NSI), is different from the first two in the sense that it does not describe the level of diversity among all local homepages of the same parent site, but rather indicates the level of similarity of each local homepage to the parent U.S. homepage (MSN or Yahoo!). The steps to calculate the third index are as follows:

  • 1Compute the ratio between each local homepage and the U.S. homepage for each observation.
  • 2If the ratio is greater than 1, use the reciprocal value (absolute ratio), so that it will be possible to intersect the ratio’s average per each local homepage.
  • 3The third index, the NSI, is derived from the average of those ratios among all independent variables for each local homepage.

Table 2 demonstrates how to calculate the third index for a partial sample taken from actual observations of MSN, June 2003.

Table 2.  Third index (NSI)—Sample from MSN, June 2003
CountryBackground Color Absolute RatioPicture Absolute RatioMenu Absolute RatioSearch Box Size Absolute RatioAverage
U.S.1.001.001.001.001.00
Arabia1.000.750.880.740.84
Italy0.800.750.910.800.82
Netherlands0.750.400.880.880.73
Norway1.000.420.850.970.81
Spain0.750.670.950.740.78
U.K.1.001.000.940.800.94
Brazil1.000.750.910.880.89
Canada (English)1.001.000.910.810.93
Mexico0.750.500.680.760.67
India0.800.670.880.820.79
Korea1.000.601.000.720.83
Singapore1.000.860.950.760.89
Australia0.750.830.730.460.69
Taiwan1.000.000.880.740.66
 Average0.80

The third index indicates the percentage of similarity for each local homepage to the U.S. homepage. The NSI value ranges from 0 to 1, where a value of 0 indicates a large difference from the U.S. homepage, while a value of 1 designates the highest possible similarity to the U.S. homepage. The NSI applies to both form and content attributes.

Factor and Cluster Analyses

In addition to the three indices, we use statistical analysis to validate and support the new metrics of indices and to examine the scatter of local homepages in detail. This was performed through a sequence of three stages:

  • 1Factor analysis was used to identify correlated variables and replace them with fewer components that explain the content and form.
  • 2Cluster analysis was used in order to group local homepages according to similar characteristics. The homepages in each cluster share common factors of content or form. The distance between clusters can be measured either by the least squares method, or by the Ward method based on variation analysis. The Ward (1963) method was found more adequate for this study, as will be later argued.
  • 3The common factors in each cluster were traced. Since the clustering process was based on the new components that the factor analysis generated, we could indicate the common factors of homepages in the same cluster.

All stages of the diversity analysis were carried out each month for six months on each site for the content and form variables separately, thus enabling us also to construct a trend component to the study. The observation item list contained 44 variables (10 of form, 34 of content) that were sampled over 332 homepages (33 MSN + 23 Yahoo! × 6 months). Consequently, 14,608 observations were collected.

Results

Analysis of the Three Indices

General Diversity Index (GDI)

Figure 1 does not indicate a clear trend in the general diversity of local homepages. However, interestingly, form and content have similar trends in MSN, and in October, the homepages were relatively more similar to each other. If October is regarded as an exception, it is possible to observe a slight increase in the general diversity level both in form and content. The sampling covered a relatively short period, however; thus a further statistical investigation is required to support this trend.

Figure 1.

MSN GDI trends

Similarly, it is difficult to trace general trends in the diversity of local homepages in Yahoo! over the observation period. Figure 2 indicates that form and content have different trends in Yahoo!. The best-fit line of the GDI in the content only implies a slight trend of decline. Although six months may be a relatively short period to draw a conclusion about the GDI trends in MSN and Yahoo!, those slight trends will gain further support by the other indices and the statistical analysis and therefore are worth mentioning.

Figure 2.

Yahoo! GDI trends

Difference Diversity Index (DDI)

The second index describes the level of homogeneity of the absolute differences between the sampled homepages and the main U.S. homepage.

Figure 3 portrays the trends of the DDI in MSN, indicating a general trend of decline mainly in content. This suggests that the absolute differences became similar to each other. In Yahoo! it was difficult to trace a general trend for the DDI, and again content and form had opposite trends. The main power of the DDI is shown when it is compared with the first index (GDI), as demonstrated in the following example.

Figure 3.

MSN DDI trends

In this example it is assumed that in month A there were 10 pictures in each homepage (including the U.S. homepage), except for one local homepage that had only five pictures in its homepage. Assuming that in the next month another local homepage had five pictures, the first index increased, while the second index decreased. This indicates that some homepages become similar to one another, but different from the U.S. homepage. This phenomenon is called in this research “the alternative block,” since homepages that become similar to one another suggest alternatives to the form and content of the U.S. homepage.

Comparing the trends of the first and the second indices therefore indicates that the alternative block in MSN is getting stronger over the period of six months. In other words, localization forces appear to be stronger than globalization forces. This phenomenon will receive further support from the cluster analysis.

In Yahoo! the trends of the first and the second indices are similar during five out of the six months both in form and content variables, and therefore there is no reason to discuss the alternative block and its trends.

National Similarity Index (NSI)

The third index indicates the degree of proximity (similarity) of local homepages to the U.S. homepage in terms of form and content. Its value ranges from 0 (least similarity) to 1 (greatest similarity).

Figure 4 portrays the trends of the third index over a period of six months in MSN. It indicates that there was a general trend of decline in form and content, which means that local homepages become more different from the U.S. homepage in terms of form and content over a period of six months. These findings are consistent with the trends of the first index that implied an increase of heterogeneity and with the growth of the alternative block that the second index revealed.

Figure 4.

MSN—Trends of the NSI

Figure 5 shows that in Yahoo!, unlike in MSN, the NSI increases in content only. There are no clear trends in form in the period of the observation, although it might be suggested that the content of local homepages in Yahoo! becomes more similar to the main U.S. homepage. These findings conform to the trend of homogeneity in content that the GDI previously suggested.

Figure 5.

Yahoo!—Trends of the NSI

Another interesting result provided by the third index is the level of similarity of each homepage to the U.S. homepage. Table 4 summarizes this proximity for MSN, June 2003, where values close to 1 indicate great similarity and values close to 0 indicate great difference from the U.S. homepage.

Table 4.  NSI in MSN, June 2003
Ranking Form Diversity (MSN)Ranking Content Diversity (MSN)
1U.S.0.819India1U.S.0.466Belgium (French)
0.937Canada0.810Hong Kong0.707Netherlands0.461Norway
0.914Germany0.805Spain0.701U.K.0.45Korea
0.908Brazil0.759Taiwan0.701Japan0.427Sweden
0.888Malaysia0.746Australia0.673Arabia0.423Spain
0.882France0.736Sweden0.669Germany0.403Hong Kong S.A.R.
0.874U.K.0.726Israel0.641Canada0.361Taiwan
0.873Italy0.683Denmark0.606France0.354Austria
0.866Latin America0.668Czech Republic0.598Italy0.352Czech Republic
0.865Korea0.659New Zealand0.593Brazil0.337Singapore
0.861Belgium0.642Mexico0.554New Zealand0.28India
0.86Norway0.519Switzerland0.554Israel0.279Denmark
0.852Arabia0.499China0.536Latin America0.249Switzerland (German)
0.846Japan0.495Finland0.519Malaysia0.187Finland
0.842Netherlands0.467South Africa0.498Australia0.118South Africa
0.840Singapore0.463Austria0.492Mexico0China

Interestingly, Table 4 (June 2003) and the results for the next five months (July-December 2003) display the same patterns over a period of time for form and content attributes. Local homepages in MSN that were similar to the U.S. homepage in form (e.g., the U.K., Canada, Germany, Belgium, and France) were often similar in content as well. Local homepages in MSN that were different from the U.S. homepage in form (e.g. China, Slovakia, South Africa, Finland, Denmark, The Czech Republic, Austria, and Switzerland) were often different in content as well. A further analysis will reveal the reasons behind those trends.

In Yahoo! it was difficult to find a pattern over time. There was no significant consistency in homepage ranking order for form or content.

Results of the Statistical Analysis

Factor Analysis

In addition to the three indices, statistical tools were exploited to further analyze the factors behind the diversity of the homepages’ form and content. By means of principal component analysis the correlated variables were extracted and a linear combination that could explain them was found.

Table 5 displays the percentage of explained variation accounted for by the new components. Each of the four new components represents a group of correlated variables. The first component represents the “number of links,” the “number of main links,” and the “number of menus.” Those highly correlated variables together comprise the form of the content. Similarly, the second new component represents the “size of the search box,” the “number of background colors,” and the “number of frames,” which together comprise the page layout or the pattern. The third component represents the “number of pop-up windows” and the “number of active banners,” which together comprise the activity. Finally, the fourth component represents the “number of photos” and “number of banners” that comprise colorfulness. Apparently, each month the same form variables were consistently correlated and grouped into the same new components that comprise four aspects of the form: form of content, patterns, activity, and colorfulness. Those components will be used later for the cluster analysis.

Table 5.  New form components—MSN, June 2003
  1. Note: Extraction Method: Principal Component Analysis.

  2. Rotation Method: Varimax with Kaiser Normalization.

  3. a. Rotation converged in 6 iterations.

inline image

In Yahoo! the factor analysis of the form shows similar patterns. The only difference is that local homepages in Yahoo! are more similar to one another. The variable “number of background colors,” for instance, always gets the value of 4, and there were hardly any pop-up windows. This is why the factor analysis process in Yahoo! generates only three new components that describe three aspects of the form (the colorfulness and activity variables were reduced into one factor). Those three new components appeared consistently in all six months with no variation.

The factor analysis for content was less consistent. The correlation matrix showed some unreasonable relations among the 34 independent variables. The unreasonable relations appeared mostly within variables that usually get a zero value and therefore create “noise” (e.g., the “number of sports banners”). After filtering those variables, the correlation matrix of the 20 content variables that were left made much more sense.

Table 6 presents the new groups of correlated variables after filtering the variables that create “noise.” The variables in each group are reasonably correlated, since they deal mostly with the same subject. Interestingly, the variables that deal with economics and shopping were highly correlated in almost every month and therefore were replaced by a single new component. This may indicate that MSN content editors in many countries have similar attitudes and preferences as regards those particular topics. Entertainment does not attain representation as an independent component, but rather it is reflected in some components in combination with other issues.

Table 6.  New content components—MSN, June 2003
  1. Notes: Extraction Method: Principal Component Analysis.

  2. Rotation Method: Varimax with Kaiser Normalization.

inline image

Unlike the factor analysis of form, the factor analysis of content in both MSN and Yahoo! creates slightly different groups of variables for each month. Variables of the same subject were almost always correlated and therefore were replaced by a single new component.

Cluster Analysis and Tracing the Factors

The cluster analysis is the second stage in the diversity analysis process. It uses the factor analysis components as input and groups together local homepages with similar patterns of form and content.

Figure 6 delineates the results of a cluster analysis of form in MSN, June 2003. The horizontal axis shows the distance between each cluster using the Ward method (Ward, 1963). This method was found to be most adequate, since it creates a small number of clusters with relatively more homepages. We decided to focus on clusters whose distance was between 5 and 10 units from each other, entailing six main clusters. In this way, we could trace the common factors for all homepages within the same cluster and observe the trends of those six clusters each month.

Figure 6.

Form cluster analysis—MSN, June 2003

Figure 6 and Table 7 show that the U.S. homepage was located in the biggest cluster (no. 1) and that MSN Israel was located in a separate cluster (no. 6). This is because the form of MSN Israel is significantly different from the form of other local homepages in MSN. Homepages that have relatively less content and therefore a simple form, like MSN Austria and Finland, are located together in the same cluster (no. 3).

Table 7.  MSN Form clusters and common factors over time
June 2003July 2003August 2003September 2003October 2003November 2003
Patterns, Form of ContentPatterns, Form of ContentPatternsPatterns, Form of ContentPatterns, Colorfulness, Form of ContentForm of Content, Patterns,
U.S.U.S.U.S.U.S.U.S.U.S.
ArabiaArabiaItalyBelgiumBelgiumBelgium
BelgiumBelgiumSpainFranceFranceFrance
Czech ReFranceSwedenItalyNetherlaUnited K
FranceGermanyUnited KUnited KNorwayBrazil
ItalyItalyLatin AmBrazilSpainCanada
NetherlandsNetherlandsChinaCanadaSwedenLatin
NorwayNorwayJapanLatin AmUnited KIndia
SpainSpainMalaysiaJapanBrazilJapan
United KSwedenSingaporeKoreaCanadaKorea
BrazilUnited KTaiwanTaiwanLatin AmSingapore
CanadaBrazil JapanTaiwan
Latin AmCanada Singapore 
IndiaLatin Am Taiwan 
JapanJapan 
KoreaIndia 
MalaysiaKorea 
SingaporeMalaysia 
 Taiwan 
Poor FormPoor FormPoor FormPoor FormPoor FormPoor Form
AustriaAustriaAustriaAustriaAustriaAustria
FinlandFinlandCzech ReCzech ReCzech ReCzech Re.
SwitzerlandSwitzerlandFinlandFinlandFinlandFinland
South AfricaSouth AfricaSlovakiaSlovakiaSlovakiaSlovakia
 SwitzerlandSwitzerlandSwitzerlandSwitzerland
 South AfricaSouth AfricaSouth AfricaLatin Am.
 China
 Malaysia
 South Africa
Patterns, activePatterns, activeColorfulnessPatternsPatterns, Colorfulness, activeColorfulness, Patterns
DenmarkCzech ReNetherlandsDenmarkMexicoDenmark
SwedenDenmarkNorwayNetherlaChinaItaly
MexicoMexico Spain Netherlands
ChinaChina Sweden Norway
 Singapor China Spain
 India Sweden
 Mexico
Colorfulness, activeActive, PatternsActive, PatternsActiveActive, PatternsActive
GermanyHong KongArabiaArabiaArabiaArabia
Hong Kong DenmarkGermanyDenmarkGermany
Taiwan GermanyNorwayGermanyHong Kong
 MexicoMexicoItaly 
 Hong KongHong KongHong Kong 
 IndiaSingaporeIndia 
 Korea Korea 
 Malaysia 
Colorfulness, active, Form of ContentColorfulness, active, Form of ContentColorfulness, active, Form of ContentColorfulness, Form of ContentColorfulness, active, Form of ContentColorfulness, active
IsraelIsraelIsraelIsraelIsraelIsrael
Form of ContentForm of ContentForm of ContentForm of ContentForm of ContentForm of Content
AustraliaAustraliaBelgiumAustraliaustraliaAustralia
New ZealandNew ZealandFranceNew ZealandNew ZealandNew Zealand
 Brazil 
 Canada 
 Australia 
 New Zeal 

Figure 7 displays screenshots of four representative local homepages in MSN. It clearly indicates that MSN U.S. and MSN Canada, which are also located in the same cluster (no. 1), have very similar page layouts as well as a number of background colors and photos. By contrast, MSN Israel has a very different design and a greater number of photos. Finally, MSN Finland has a relatively simple layout and less content.

Figure 7.

Screenshots of MSN’s local homepages, July 2003

Table 7 shows how the composition of each form cluster varies over time in MSN. It shows that the biggest cluster contains MSN U.S., together with other countries that appear consistently over the observation period such as the homepages of the U.K., Brazil, Canada, Italy, Japan, and France. This is consistent with the analysis of the third index (NSI). The common factor that makes them belong to the same cluster is the similarity in the pattern of the form (number of frames and size of the search box). Table 7 indicates that this cluster becomes smaller over a period of six months (from 19 to 12 homepages), supporting previous findings. The heterogeneity of local homepages in MSN grows, and they become more different from the U.S. homepage over time.

Another cluster consists of MSN Austria, the Czech Republic, Finland, Slovakia, Switzerland, and South Africa. The common factor for this cluster is the limited content and therefore the simple form. This cluster grows over the period of time (from 4 to 9 homepages). It therefore supports the findings of the second index (DDI), which indicate the growth of an “alternative block.” The cluster analysis reveals exactly what this “alternative block” is, and what the variables that make it distinctive are.

Using the same method, a cluster analysis was employed to analyze the content diversity of homepages based on the content components in MSN. It was found that similar to the form clusters, the biggest content cluster contains MSN U.S., together with the local homepages of the U.K., Brazil, Canada, Japan, and France. This finding is consistent with the analysis of the third index (NSI), in which those countries ranked first as the most similar to MSN U.S. in both content and form. The common factor that makes them belong to the same cluster is the relatively frequent appearance of economics and shopping-related content. Similar to the main form cluster, the main content cluster shrinks over a period of six months (from 14 to 5 homepages), supporting previous findings. The heterogeneity in content grows as well, and local homepages in MSN become more different from the U.S. homepage over time.

Similarly, cluster analysis was used to explore the diversity of form in Yahoo!’s local homepages over time. It was found that the biggest cluster contains Yahoo! U.S., together with the homepages of Asia, Brazil, India, Mexico, and Canada. The common factor of this cluster is the similarity in the patterns of the content (number of links and menus). There is no clear trend for this cluster to become smaller or bigger over time. These findings are consistent with the third index analysis.

Interestingly, the composition of countries in each cluster is consistent over time. One cluster consists of Yahoo! Catalan only, which provides little content and therefore has a simple form; another cluster consists of the homepages of Scandinavian countries: Denmark, Sweden, and Norway, all of which have relatively colorful and active homepages. Another cluster consists of the homepages of European countries: France, Germany, and Italy, all of which have a relatively high number of links in their homepages; another cluster consists of the Asian countries homepages: Japan, Hong Kong, and Taiwan, each of which has a relatively small number of links in their homepages; and finally there is a cluster for Yahoo! China, which has a relatively colorful, active, and content-full homepage.

The composition of factors was further explored in each content cluster over time in Yahoo!. It was found that the cluster that contains Yahoo! U.S. has no stable composition and therefore no stable common factor. The cluster that contains the U.S. homepage grows over the period of six months (from 2 to 13 homepages). This supports the results of the first and third indices analyses. The homogeneity in content increases and the homepages become more similar to the U.S. homepage over time.

Discussion: Towards A Dictated Heterogeneity?

This article presented new methods, indices, and metrics to explore cultural diversity online and implemented them to measure the similarities among homepages of MSN and Yahoo! in a large number of countries over time. The tools and methodology proved useful in measuring content and form proximities of affiliated websites. Since previous comparative studies in the field were limited to a small number of countries and often used Hofstede’s (1997) framework to analyze content and form, we hope to encourage further use of these new analytical tools and approaches to explore online trends of heterogeneity and localization on a large scale.

The study posed two main research questions, the first about cultural diversity online and the second about trends toward homogeneity or heterogeneity over time. Despite the relatively short observation period, the findings clearly implied that MSN tended to increase the customization of form and content in its local homepages. Additionally, it was found that local homepages in MSN tended to differ from the U.S. homepage over time in terms of both content and form.

Conversely, Yahoo!’s analysis identified a trend toward homogeneity of content. The resemblance of the local homepages to the U.S. homepage grew in term of content. However, in terms of form, the cluster analysis indicated the distinctive differentiation of local homepages by their region: America, Europe, Scandinavia, and the Far East. The relatively high activity and colorfulness of Asian homepages in both MSN and Yahoo! support previous studies, which found similar trends in websites of transnational corporations and e-businesses as a reflection of high-context cultures (Becker, 2002; Cyr, Bonanni, Bowes, & Ilsever, 2005; Würtz, 2005).

In general, the hypothesis that assumed trends toward homogeneity was rejected. MSN, an American brand and popular content provider, displayed heterogeneous homepages and trends toward localization. These findings may therefore support similar studies that found cultural heterogeneity and localization of web design and usability (Badre, 2000; Becker, 2002; Cyr & Trevor-Smith, 2004; Lo & Gong, 2005; Marcus & Gould, 2000; Sheridan, 2001; Singh & Baack, 2004; Singh, Xhao, et al., 2003; Singh, Furrer, et al., 2004; Sun, 2001; Tsui & Paynter, 2004; Würtz, 2005; Zhao et al., 2003). Moreover, our study found a growing diversity of content in MSN local homepages, suggesting a new perspective for further investigation in the field.

These trends may be also explained as a result of the business model of MSN that focuses on direct subscription and production of customized cultural artifacts (Van Couvering, 2004). Heterogeneity and localization trends do not imply less control by the parent site over the local sites. On the contrary, these may be analyzed as a stronger leverage to control users’ preferences. Through addressing local content, big portals and content providers strengthen their grip on users, attracting attention that non-American users might not otherwise pay to more global content and form.

In contrast to MSN, the relative homogeneity in the content of Yahoo!’s local homepages and their similarity to the U.S. parent site imply that Yahoo! takes less account of cultural issues. A few studies (Kang & Corbitt, 2001; Schmied, 2003) may support this trend; they propose a certain level of standardization in websites of transnational (and particularly U.S.-based) corporations. These findings are consistent with the different business model of Yahoo!, which is based mainly on advertisement (Van Couvering, 2004) rather than on the provision of customized information products.

To conclude, the increasing diversity of local homepages in MSN and the regional division of form clusters in Yahoo! clearly indicate trends of localization and cultural diversity in cyberspace and a gradual departure from their U.S. sources. Consequently, we may expect further customization of local homepages, especially in MSN, in the Far East and Latin America, as MSN adds new ethnic cultures to its customer base. This is probably why during the observation period MSN launched another homepage entitled “Latin.” This homepage is culture and Spanish-language oriented, rather than geography oriented.

It may be argued that the customization process of content and form proceeds in phases, starting with geographical and national distinctions and progressing towards a focus on ethnographic differences. A reason for that could be that it is technically easier for a corporation to conduct a geographical customization than an ethnographic and more specific one.

Future research into the increasing customization of local homepages, especially in popular and influential U.S.-based companies such as MSN and Yahoo!, may further contribute to the understanding of cultural globalization (Appadurai, 1996; Basch, Schiller, & Blanc, 1994; Foster, 1991; Friedman, 1990). In this sense, our findings support previous studies that highlighted the centrality of customization and cultural diversity (Anderson, 2006; Caves, 2000; Shapiro & Varian, 1999; Sunstein, 2001). Additionally, our findings suggest that future studies of this diversity should increasingly focus on ethnographic as well as geographical differences.

Acknowledgments

The research for this article was supported by the faculty of management, Tel Aviv University. We would like to thank the administration staff, and especially Mrs. Orit Aviv-Prihed, who helped us to carry out this research. We would like to thank Dr. Glenn McGregor and Dr. Yaojun Li of Birmingham University for their contribution to designing the methodology of this research. A special thanks to Professor Robert E.F. Smith for his comments on an earlier version of this article.

Note

  • 1

    Information search plays a prominent role in websites such as MSN and Yahoo!, and therefore the size of the search box was considered a notable element of form (see also Cyr & Trevor-Smith [2004], who compared the search layout of various national websites). The size of the search box was measured in pixels and ranged between 140px, which is relatively small (3.7cm), and 620px, which is relatively large (16.3cm).

Appendices

Appendix A

Table 8.  Full list of local homepages in MSN and Yahoo!
MSN Yahoo! 
U.S.SwedenU.S.Italy
Canada (English)ItalyU.K. & IrelandKorea
JapanMalaysiaAustralia & NZCanada
BrazilKoreaFranceSpain
IsraelTaiwanBrazilCatalan
Latin AmericaIndiaGermanyChina
FranceMexicoArgentina 
Belgium (French)Hong Kong S.A.R.Norway 
GermanyDenmarkHong Kong 
SpainCzech RepublicDenmark 
United KingdomSouth AfricaTaiwan 
SingaporeSwitzerland (German)Mexico 
AustraliaAustriaSweden 
New ZealandFinlandAsia 
NorwaySlovakiaJapan 
ArabiaChinaSingapore 
NetherlandsSwedenIndia 

Appendix B

Table 9.  Full list of content and form elements
Content Form
  1. Note: *Within the top popular search queries

NewsLocal SportFrames
Main (headline) NewsSport BannerBanners
Local NewsKeyword Sport*Active
News BannerShoppingBanners
Keyword News*Main (headline) ShoppingBackground Colors
EconomicsShopping BannerPhotos
Main (headline) EconomicsKeyword Shopping*Main Links
Local EconomicsTechnologyLinks
Economics BannerMain (headline) TechnologySearch Box Size
Keyword Economics*Technology BannerMenus
EntertainmentKeyword Technology* 
Main (headline) EntertainmentGeneral Information 
Local EntertainmentMain General Info. 
Entertainment BannerGeneral Info. Banner 
Keyword Entertainment*Keyword General Info.* 
SportStudy English 
Main (headline) SportKeyword Study English* 

Appendix C

Table 10.  Top 10 parent companies worldwide (Home Panel February 2004)
Parent NameUniqueAudience (000)Reach %Time Per Person
  1. Source: Nielsen//NetRatings, retrieved May 13, 2004 (http://www.nielsen-netratings.com).

  2. Note: The data about the popularity of MSN and Yahoo! were retrieved near the observation period. For more recent statistics, see http://www.nielsen-netratings.com.

United States
Microsoft93,67266.2501:32:48
Time Warner82,53058.3703:59:38
Yahoo!79,60356.3001:57:25
France
Microsoft9,19066.6802:09:19
Wanadoo8,62462.5800:50:20
Google6,59147.8200:23:13
Iliad - Free5,99543.5000:27:30
Time Warner4,63533.6302:50:51
Yahoo!4,33331.4400:42:34
Hong Kong
Yahoo!2,06876.6101:41:05
Microsoft1,96472.7501:34:02
Spain
Microsoft6,46077.5404:30:54
Google5,23162.7900:27:14
Terra Lycos3,16137.9400:23:26
Wanadoo3,14237.7100:23:09
Yahoo!2,93535.2300:38:01
Sweden
Microsoft3,42374.2701:50:05
Aftonbladet Hierta1,82739.6400:48:11
Eniro1,55333.7000:15:42
TeliaSonera1,40030.3700:20:14
Bonnierförlagen1,31528.5400:24:02
Google1,30928.4100:15:38
Förenings1,21326.3200:47:08
Sparbanken
Yahoo!1,09323.7100:12:19
United Kingdom
Microsoft14,65269.9202:05:05
Google9,76246.5900:18:31
Yahoo!7,89937.7001:01:50
Switzerland
Microsoft2,16371.9001:24:03
Google1,63054.1800:20:37
Bluewin1,50149.8700:39:06
Yahoo!79426.3700:49:27

About the Authors

  1. Elad Segev is a doctoral candidate at Keele University, a lecturer of Media and Communications at Aston University, and a former research student at Tel Aviv University in the Faculty of Management. His research deals with technology in general and the Internet in particular, and its social, political, and cultural implications.Webpage: http://www.eladsegev.comAddress: School of Politics, International Relations and Philosophy (SPIRE), Keele University, Keele, Staffordshire ST5 5BG, UK

  2. Niv Ahituv is the Marko and Lucie Chaoul Chair for Research in Information Evaluation and the Academic Director of the Institute of Internet Studies of Tel Aviv University. His research deals with information economics, information system strategy and management, and social and business implications of the Internet.Webpage: http://recanati.tau.ac.il/Index.asp?ArticleID=76&CategoryID164Address: Faculty of Management, Tel Aviv University, Tel Aviv, 69978, Israel

  3. Karine Barzilai-Nahon is Assistant Professor in the Information School at the University of Washington. Formerly she held senior positions in Research and Development in hi-tech industry. Her research deals with social and business aspects of the Internet and telecommunications.Webpage: http://projects.ischool.washington.edu/karineb/Address: The Information School, University of Washington, Mary Gates Hall, Seattle, WA 98195, USA

Ancillary