GVU's 4th WWW Survey Page
GVU's Fourth WWW User Survey

GVU's 4th WWW User Survey

[ Survey Home ] [ 4th Survey Home ] [ Graphs ] [ Bulleted Lists ] [ Datasets ]
This is the main page for the Graphic, Visualization, & Usability Center's (GVU) 4th WWW User Survey. GVU runs the Surveys as PUBLIC SERVICE and as such, ALL RESULTS ARE FREE (subject to certain terms and conditions). The 4th Survey was run from October 10, 1995 through November 10, 1995 and was endorsed by the World Wide Web Consortium (W3C) (which exists to develop common standards for the evolution of the Web) , NCSA's Software Development Group (SDG) (the folks who develop Mosaic and other Web technologies), and INRIA (the acting European host for the W3C in collaboration with CERN, where the Web originated). Over 23,000 unique responses were collected to eight sets of questionnaires, including: General Demographics, WWW Browser Usage, Authoring Information, Consumer Attitudes & Preferences and Web Service Providers. This page provides access to the following:

Executive Summary

It is amazing to both Colleen and me that GVU's WWW User Surveys continue to be such a huge success. The Fourth Survey, conducted from October 10 through November 10, 1995, received over 23,000 responses, which (if you go by the current estimates of the total number of WWW users as 18 million), means that the Survey was completed by nearly 1 out of every 1,000 Web users. Note that we reached similar participation rates for the Third Survey conducted in the spring. This makes GVU's Surveys the oldest (we pioneered the field of Web-based surveying) and the largest survey to date as measured in terms of the number of responses multiplied by the number of questions asked.

Some of the main findings are: The age profile has shifted slightly younger, with the average age around 31 years old. The integration of female users into the population seems to be accelerating, with a 10% increase since the Third Survey pushing the gender proportions to 30% female - 70% male. It seems that the Web is still in the early phases of commercial acceptance, with the primary utility being within enterprises. We see the likely migration path of the Web being from enterprise solutions, to business-to-business use, to consumer transactions. Our results indicate that users still do not spend much time shopping, but rather prefer to use the Web for entertainment and work/research. The Survey results support the notion that this results from a perceived weakness in Web security, as 60% of the users cite security concerns as the primary reason for not buying merchandise. Most users prefer to use the Web to gather product information rather than make purchases.

Other interesting findings include: On a scale of 1 (Never) - 5 (Sometimes) - 9 (Regularly), users report using their Web browser instead of watching TV (6.1). This ranks third against all other options, trailing behind replacing other interfaces to the Internet (6.8) (e.g. XFTP, MacGopher) and accessing Reference Information (6.4). The finding is even stronger among Weekend users (6.6), indicating that the Web is replacing traditional recreational/entertainment activities. We will be investigating this further in future Surveys. Perl is the most common CGI scripting language (46.7%), followed by C (12.5%), C++ (9.5%), and shell scripts (8.1%). Finally, the biggest problem reported by users of the Web is the speed it takes to download and display content (i.e., more bandwidth and cycles, please).

As part of GVU's commitment toward the growth of the Web and the Web community itself, we offer access to the collected results and datasets free of charge. It is important to note that there are certain restrictions on the use of the data. Nevertheless, we strongly believe that all Web users, regardless of ability to pay, should have access to the most comprehensive data gathered on Web users to date. Remember though, the data presented on the following pages is only a snapshot of the current Web user population - we do not make any claims about the representativeness of the data to the entire Web population.

Presentation of all the results is an arduous task (please forgive any typos and spelling errors, although we are indebted to Wendy Mattson (posh@sirius.com) for the wonderful job proofreading the online results). Not only is the process time-consuming, but finding a meaningful way for users to find the results of interest is not easy either. Toward this end, we've created over 200 graphs (See: graphs and tables) of the results and added our interpretation to each question asked in the Survey. These interpretations are also available in a separate, non-graphical format (See: bulleted lists of the findings). These bulleted lists provide an easy way for users to scan the results non-graphically first, and then inspect the graphs for only those questions of interest. There are a lot of interesting results.

For those who want a more compact way to browse the results from this and previous Surveys, we offer a modified version of our 4th International WWW Conference presentation in PDF and Postscript Form. These files contain a lot of new material and comparisons of characteristics across Surveys that are not yet available in HTML form.

For all questions, analysis between the following groups were performed: European vs US users, Weekend vs Weekday users, and Female vs Male users. These comparisons provider deeper insight into the characteristics of these user segments. Don't forget to view the Consumer Survey pages, developed and analyzed by the Hermes Team, as part of GVU's WWW User Surveys. Also, former Survey Team member Mimi Recker and John Greenwood have performed an analysis of the Oceania respondents. For those seeking more analysis of European countries see Erik Granered's Back to the top

High Level Summary and Trend Analysis

General Demographics

The state of the Web seems to be in as much of a state of flux as with previous Surveys. While certain characteristics of the Web users sampled in the Surveys has remained the same or changed slightly, other characteristics have changed dramatically. More than ever, the users in the Fourth Survey represent less and less the "technology developers/pioneers" of the First Survey (primarily young, computer-savvy users) and more of what we refer to as the "early adopters/seekers of technology." The adopters do not typically have access to the Web through work or school, but actively seek out local or major Internet access providers. As the Web continues to expand its horizon of users, we expect, and indeed find, that more and more users from diverse segments of the population participate in the Surveys. Much of the following content is derived from the bulleted lists and the graphs. Please refer to them for more complete coverage of the questions.

What's the average age?

Reversing the trend of the previous Surveys, the average age across all users in the Fourth Survey(32.7 years old) was down two years from the Third Survey (35.0 years old). This downward shift brings the average age of the Web users sampled closer to the average measured in the Second Survey (31.7 years old). Once again, the youngest users were European (average age 29.7), with the eldest being Weekend users (average age 33.4). The Fourth Survey shows a significant change in age distributions between genders: there are more female users between the ages of 16 and 20 years old (14.2% female vs. 11.5% male) and fewer above the age of 46 years old (13.82% female vs. 17.87% male). This trend toward younger females is consistent with the observed increase in female users in college and K-12 education occupations.

What's the gender ratio & how has this changed over time?

Overall, 29.3% of the users were female, 70.7% male The actual numbers are 6842 women and 16,506 men. Compared to the last Survey, women represent a 15% increase and men a 12% decrease (the last Survey had a "Rather not Say!" option). This is nearly double the growth rate of women users observed across the first three Surveys (5% increase per Survey). In the US, 32.5% of the users were female, 67.5% male. This represents a strong shift in the increased acceptance and use of the Web by women. The US continues to integrate female users into the Web user population faster than Europe. However, the percent of US users is still far from the 52% female 48% male composition of the entire US population (1995 Estimated US Census). As noted above, the increase in women users occurred largely in college students and K-12 educators. For Weekend users, 24.8% of the respondents were female with 75.2% male. These ratios reflect the traditionally male-dominated field of computer hobbyists.

What's the average and median income?

The estimated average income across all users is $63,000 (US dollars), which is slightly lower than the Third Survey, which had an estimated income of $69,000. The median income is in the range of $50,000 and $60,000, which is the same as the Third Survey. The estimated average US income is $64,700 which is lower than the Third Survey of $67,600, but still higher than the estimated average European income of $56,000 ($53,500 Third Survey). 18.3% of the women surveyed would "Rather not say!" their income, whereas 14.5% of the men chose this response. These are exactly the same proportions as last time. This non-response, nearly one in five, interferes with robust gender comparisons between incomes. Female users typically report making less money than their male counterparts, with the average income for females being $58,900. As one would expect of computer hobbyists, the Weekend users have slightly more income than Weekday users. Thus, we see that as more US users are joining the Web, fewer are ultra-elite computer-users, but the estimated median income is still quite above the national median -- $36,950 as reported by the 1993 US Census.

What about location, marital status, & occupations?

For classification of location by major geographical location, 76.2% of the respondents were from the US, 10.2% from Canada and Mexico, and 8.4% from Europe. Compared to the Third Survey (80.6% from the US, 9.8% from Europe, and 5.8% from Canada and Mexico), this represents a shift toward less US dominance in Web users. Additionally, notable increases occurred in most of the other geographical areas such as Asia, Africa, Oceania, the Middle East, and Central and South America. As with the Third Survey, a higher percentage of women were from the US (84.6%) than men (72.7%), although the proportion of women to men was reversed for all other locations. For example, Europe had 10.6% male and 3.0% female users. This trend supports the hypothesis that the US is integrating women into its user population faster than other countries. It also suggests that as the Web increases its user population, a more balanced female/male ratio is expected.

Overall, slightly more users are married than single; 41.7% of the users reported being married, and 40.8% reported being single. The users who reported living with another (a new option for the Fourth Survey) was 8.8%; the number who reported being divorced was 5%. Europeans were twice as likely to report living with another person (16.5%) compared to the US (7.6%). Compared to the Third Survey, there are more single and less married Web users. European users were more likely to be single (50.3%) than married (28.1%). The converse occurs in the US, where users are more likely to be married than single, with 39.1% of US users being single and 44.2% being married.

Computer (29.1%) and Educational (30.9%) occupations still represent the majority of respondents, though Educational has taken the lead over Computer occupations on the Web. Professional (19.9%), Management (10.2%), and "other" occupations (9.8%) fill out the other categories. Europeans tend to be predominantly in Education (33.8%) and Computers (33.6%). Educational responses gained heavily since the Third Survey, 30.9% (Fourth) compared to 23.7% (Third). Again, this paints the picture that the Web is no longer dominated by computer-related users. More and more users are from the Educational field and the general public. Weekend users tend to be employed less in Computer-related jobs (25.5%) and more in Management positions (11.5%) than Weekday users (29.9% and 9.9% respectively). However, Educational Weekend users still dominate in both segments (30.9% Weekday and 30.9% Weekend). There was an increase of over 10% in the proportion of women in Education from the Third Survey (24.2% to 37.1%). As with the Third Survey, there are less women than men in Computers (20.1% women vs 32.8% men) and Management (9.1% women vs 10.7% men), but more women in the Educational (37.1% women vs 28.4% men) and "other" occupation category (12.9% women vs 8.6% men).

How willing are users to pay for access to Web sites?

One of the most stable characteristics of previous Surveys has been that one of five users stated outright that they would not pay for access to WWW sites. For the Fourth Survey, this segment of the population has increased to 31.8% from 22.6% in the Third Survey! This is indeed alarming for those who wish to apply a subscription business model to the Web. This may also very well reflect the perceived value of the material and resources currently available on the Web by its users. And this places the financial burden upon sites that wish to generate monies by advertising, rather than user subscriptions. Related, the Fourth Survey reports that over three-quarters of the sites do not allow advertising (See: Site Policy Toward Advertising). Most users reported that their willingness would depend on both the cost of access as well as the quality of the material provided (58.7%). As with the Third Survey, Europeans tend to be slightly more concerned about the quality of information if they had to pay for access (7.3% Europe vs 3.2% US).

WWW Usage & Preferences

How often do people use their Web browser?

The majority of users who participated in the Surveys (78.4%) report using their browsers daily. Users in Europe spend slightly less time using their browsers than user in the US, which is the reverse of the Third Survey. 65.8% of European users spend 6 hours or less per week compared to only 59.4% of US users. As one might expect, Weekend and Female users report using their browsers less frequently than Weekday and Male users. This reflects more casual and recreational use of the Web. As noted in the bulleted lists, there seems to be evidence that users who use computers for work are less likely to spend casual and recreational time computing. One hypothesis is that users are only willing to spend a fixed amount of time per week in front of a computer. More research into this issue is warranted before any conclusions can be drawn.

Why do people use their Web browsers?

The most common use of Web browsers is simply for browsing (79.0%) followed by entertainment (63.6%) and work (51.8%). The category with the least number of responses was "other uses" (10.8%) followed closely by shopping (11.1%). The lack of use of the Web for shopping was also found in the other Surveys, where 10.0% (Third) and 8.0% (Second) reported using the Web for this purpose. This profile of use has remained fairly consistent since the Second Survey, where browsing and entertainment dominated. Hence, while the Web is being used for academic and business purposes, the main uses are still recreational.

What do people do with their Web browsers and with what regularity?

This question provided a scale for users to specify the regularity of information accessed, with 1 corresponding to Never and 9 corresponding to Regularity. The use of the Web to access Reference Information (6.4) remains the primary use besides replacing the traditional Internet interfaces to FTP and Gopher (6.8). Other popular uses were accessing using the Web instead of watching TV (6.1), accessing Electronic News and Magazines (5.8), Product Information (5.5), and Research (5.1). These findings are quite similar to the Third Survey, with almost every category showing increased access. The one area that remains at the bottom of the list is Shopping (2.6 for both the Third and the Fourth Surveys). Thus, while users regularly use the Web to seek Product Information, the users sampled do not report regularly using the Web for Shopping.

Why do people save documents on the Web?

This was a new addition to the Survey, which attempts to gain insight into why people save content from the Web. (Users were allowed to select more than one answer). The most widely cited reason was to use the information in the document offline (59.7%), followed by reading the document offline (50.9%), distribute to others not online (45.2%), and archiving the content (30.9%). These findings correlate fundamentally to the types and regularity of information being accessed, where people are most likely to read and use the content offline when it is reference, news, or product information. Saving documents in fear that the item would no longer be available was only reported by 18.8% of the users.

What are the main problems with using the Web?

The most widely cited problem was that it takes too long to view/download pages (69.1%). This problem is supported by the finding that 60.3% of the users report using either 14.4 or 28.8 kbs modems (see next summary). The other problem areas identified by users were: not being able to find a page that they know is out there (34.5%), not being able to organize the pages & information they gather (25.8%) and not being able to find a page once visited (23.7%). Additionally, 14.3% of the users reported that not being able to visualize where they have been and where they can go was a problem. Surprisingly, users did not report that being able to determine where they are (the classical "lost in hypertext" problem) as being very problematic (6.5%). Only 3.2% reported that their Web browser was either poorly designed or did not work well as being a problem. Thus, some of the traditional problems of hypertext do not seem to be as pressing for the Web users surveyed.

How fast are people's connection to the Internet?

While 14.5% of the users were unsure of their connection speed, 33.7% of the users report that they have 14.4 kbs connections, and 26.6% report speeds of 28.8 kbs. Many more users have 28.8 kbs than in the Third Survey (14.6% more) but the percentage of users with speeds less than or equal to 28.8 kbs is the same across Surveys (approximately 61%). In general, users in Europe seem to have faster connections than users in the US. Weekend users tend to have slower connections than Weekday users with 74.6% of Weekend users having connections under 28.8 kbs compared to 57.6% of Weekday users. This firmly backs the finding that users find speed to be the biggest problem in using the Web. It also has clear implications for Web page designers - the bandwidth is not there for many of the users for large or a large number of graphics.

Information Providers/HTML Authors

How easy was it for people to learn HTML?

Just over half of the respondants who have learned HTML, learned the basics in under 3 hours, with 79.4% learning in under 6 hours. In the Second Survey, 62% of the users learned HTML in under 3 hours and in the Third Survey, the number dropped to 55.2%--the trend is clearly toward users taking longer to learn HTML. Yet, taking into account the shifts in the demographics toward less experienced computer users, this increase is extremely small. The ease with which the Web enables many people to provide content remains one of its primary strengths.

How did users learn HTML?

Online documentation was consulted by 83.9% of users in learning HTML. The next two most popular sources, books and friends, were consulted by only 44.5% and 29.0% of users, respectively. The use of books has risen noticeably since the Third Survey, from 29.2% to 44.5% overall. The rise was even more pronounced for Europeans (up 15.9%) and females (up 17.0%). This is mainly due to the numerous HTML authoring books that are now readily available in most bookstores. It also is a sign that users who are not familiar with the Web can learn HTML, since using online documentation requires that the user is familiar with the Web and how to use it (which may not always be the case). As one would expect, given that more Weekend users are the sole users of their machines, fewer Weekend users consulted friends (24.5%) and gurus (17.3%) than did Weekday users (30.3% friends, 21.0% gurus).

What do users typically author pages on?

The two most common types of pages authored were work related (75.0%) and personal (73.2%). The most common type of hyperlink used in pages is to other WWW documents with 93.6% of users including them in their documents. The next most common are images and "other" links with 85.8% and 64.9% of users including them, respectively. Links to other kinds of servers (FTP, Gopher, etc.) were down compared to the Third Survey, indicating that the other Internet Information Services may very well become legacy services. Just the same, the the number of interactive links is being authored is rising - a good sign for entertainment seeking users/browsers.

What programming languages are people using for CGI scripts?

This is a new question to the Fourth Survey. The idea here is to try to gain a better understanding in who is using what languages for CGI programming. The majority of people who develop CGI applications use Perl (46.7%), followed by C (12.5%), C++ (9.5%), and shell scripts (8.1%). Users in Europe were less likely to use Perl (43.7% Europe vs 47.8% US), and more likely to use either C (14.8% Europe vs 11.9% US), shell scripts (12.5% Europe vs 7.5% US), and OLE (8.0% Europe vs 0.3% US). There were no significant differences between Weekend and Weekday users, though Weekend users reported using Apple Script more than their Weeday counterparts (8.9% Weekend vs 6.0% Weekday). Women CGI programmers reported using Perl (50.3% women vs 46.1% men) and "other" languages (12.7% women vs 7.8% men). Men reported using more C than women (13.2% vs 8.1%) and shell scripts (8.6% vs 5.2%). There were no significant differences between gender for the other languages.

What about servers?

The most common server connection speed is 10 Mb/sec (38.9%), followed by 1 Mb/sec with 12.9% and 56 Kb/sec with 11.1% of the responses. Thus, it would appear that there is usually ample bandwidth from most servers to the Internet (excluding the million hit/day club). The majority of Webmasters replied that they "Don't Allow Ads" (78.9%). For those who do allow ads, the largest percentage (11.7%) charge under $50 per week. Only 2.1% charge over $510 per week. In general, more servers are allowing ads across all price ranges when compared to the Third Survey. There were also modest gains in the percent of Webmasters that operate mirrors (88.9%) and proxies (16.7%) from the Third Survey, though the majority still does not.

Web Service Providers (WSPss)

What types of services are being offered?

Overall, European WSPs provide more services than their US counterparts, which may be due to the smaller number of WSPs in Europe compared to the US. Nearly the same distribution of services was observed as with the Third Survey. Companies provide page design (78.1%), Internet/Web consulting (68.9%), other types of services (59.8%), disk space (59.0%), Internet/Web marketing advice (55.3%), CGI scripting (57.4%), traffic analysis of page accesses (51.5%) and finally shell accounts (45.2%).

How long have providers been in business?

Over half of the providers have been in business over 10 months (56.7%). Between the Third Survey in April and the Fourth Survey in October, 33.8.% of the providers surveyed went online. The start-up rate for Web Service Provider companies varies between 8.1% (under 1 month) and 13.3% (2 to 4 months) for the preceding 10-month period.

What factors affect the prices WSPs charge?

In order to better understand the pricing of services by Web Service Providers, this question allowed respondents to choose from more than one option (thus the totals will exceed 100%). The most common factor affecting pricing was the size of the project (77.5%), followed by the complexity of the project (74.4%), duration of maintenance (60.5%), other factors (59.9%), number of services used (59.4%) and the size of the customer's company (25.7%). European WSPss used more factors in determining their pricing, which is reflected in their preference for customized page design and CGI scripting services.

Back to the top

Limitations of the Results

Highly distributed, heterogeneous, electronic Surveying is a new field, especially with respect to the Web. Our adaptive WWW based surveying techniques are pioneering and as such, require conservative interpretation of collected data due to the absence of time-tested validation and correction metrics. Basically, our Survey suffers two problems: sampling and self-selection. Essentially, when people decide to participate in a survey, they select themselves. This decision may reflect some systematic selecting principle (or judgment) that effects the collected data. Almost all surveys suffer from self-selection problems. That is, when a potential respondent hangs up on a telephone based surveyor, self-selection has occurred. Likewise, when a potential respondent does not send back a direct mail survey, self-selection has occurred.

The other issue is sampling. There are essentially two types of sampling: random and non-probabilistic. Random selection is intended to ensure equal representation among populations. To accomplish this, steps need to be taken to get respondents in a random manner, e.g., drawing numbers out of a hat. Our Survey uses non-probabilistic sampling, which does not use randomization techniques to get respondents. This reduces the ability of the gathered data to generalize to the entire user population.

Since the Web does not have a broadcast mechanism (yet) we used the following diverse mediums to attract respondents:

One could argue that this diversified exposure minimizes any systematic effect introduced via the sampling method. We tend to agree, though have taken steps to further explore this issue. Specifically, we designed the Third and Fourth Surveys to enable us to determine how the respondents found out about the survey. This allows us to group respondents accordingly and look for significant differences between these user populations.

For the Third Survey, we reported that there were no significant differences between the response profiles of women and men for the following categories: remembering to take the survey, other Web pages, the newspaper, other sources, and listserve announcements. There were differences found for: finding out via friends, magazines, Usenet news, and the www-surveying mailing list. Given the low effectiveness of all but other Web pages and Usenet news announcements, most of these differences lead to nominal effects. Thus, the surveys do not appear to suffer critically from sampling biases with respect to gender. If a segment of the Web user population were excluded, statistically, we'd expect to find similar response distribution for women and men.

In the Fourth Survey, we note that there are more differences between groups, notably between male and female users. Yet despite these differences in the method of participation, the Fourth Survey's ratios for gender and other core demographic characteristics like income, marriage, etc., are almost exactly those reported by North American based random sampling surveys. This decreases the reliance we formerly placed upon method of entry as an indicator of bias and increases our confidence in the overall robustness of the Fourth Survey's results. While the WWW User Surveys do attract heavier users than random phone based survey, it does not appear that frequency of use is a differentiating characteristic within the population, as one might expect. It is also important to keep in mind that up until the Fall of 1995, no random number dialing survey had been publicly released to compare the WWW User Survey results to. Indeed, to date, no international random sampling survey has taken place, so the biases and corrective metrics necessary are still undeterminable.

Despite the evidence to support the Fourth Survey results, we remain unconvinced that the Survey's sampling methodology is optimal and welcome suggestions and further comments on this subject.

Back to the top

Technical Information

Statistical Inferences

All analyzes were performed using Splus version 3.3 for Unix. Tests for significant interactions among variables were performed using the classical chi-squared for independence of categorical data, with significance being determined at p <= 0.01 level. Test for differences between stratified samples was performed using a two-sided alternative for the Wilcox rank sum statistic. Since all tests included N > 49, the normal approximation was used. In the event of ties, the Lehmann approximation was used. Significance was determined at the p <= 0.01 and confirmed by checking that Z was either <= -2.58 or Z => 2.85.


The Surveys were load balanced using a dedicated Sun Sparc 20 and two Sparc 5's. All HTML pages were generated on the fly via our Survey software and query engine (written in PERL). For more information about how the Surveys actually work, see: the write-up in the paper on the Second Survey Results. For inquiries about the availability of the survey code, contact: www-survey@cc.gatech.edu.

Back to the top

In Appreciation

We owe the loan of the fabulous artwork to
Melissa House & Allyana Ziolko (please contact Allyana: allyana@cc.gatech.edu for permission to use the artwork) and technical support to Michael Mealling (OIT), Dan Forsyth (CoC), Dave Leonard (CoC), Randy Carpenter (GVU), & Kipp Jones (CoC). Of course, the resources necessary for the Surveys would not be possible without support from the GVU administrative staff and Dr. James Foley, GVU's Director.

Special thanks go to Greg Calhoun, Emil Sarpa, & John Dutra of Sun Microsystems, whose generously provided the machines which ran the Surveys. Additional thanks go to Wendy Mattson for the fantastic help with proofreading the questionnaires and online results. Back to the top

[ Survey Home ] [ 4th Survey Home ] [ Graphs ] [ Bulleted Lists ] [ Datasets ]
Copyright 1995
Georgia Tech Research Corporation
Atlanta, Georgia 30332-0415
Usage Restrictions
For more information or to submit comments:
send e-mail to www-survey@cc.gatech.edu.

GVU's WWW Surveying Team
Graphics, Visualization, & Usability Center
College of Computing
Georgia Institute of Technology
Atlanta, GA 30332-0280