Gender and the Internet: Causes of Variation in Access, Level, and

Gender and the Internet:
Causes of Variation in Access, Level,
and Scope of Use n
Ira M. Wasserman, Eastern Michigan University
Marie Richmond-Abbott, Eastern Michigan University
Objective. The article examines differences in the use of the Internet by gender, with
a consideration of access to the web, use of communication facilities related to email
and chat rooms, frequency of use, and types of websites used. The study considers
the impact of socioeconomic status and social, geographic, racial, and ethnic variables for explaining variations in the use of the web by men and women, and how
these factors are mediated by knowledge of how to use the web. Methods. The
study employs data collected by the General Social Survey (GSS) in 2000, and
relates access, communication levels, frequency of use, and types of sites used to
gender and other relevant variables. The relevant variables are analyzed by multivariate analysis. Results. Access to the web was independent of gender, but was
related to education, race, income, age, and marital status. Women were less likely
than men to chat on the web, but were slightly more likely to use email, and they
utilized different types of sites than men. Conclusions. Women access the web as
frequently as men, but they communicate on the Internet differently than men, are
online less than men, and utilize different types of websites than men. Knowledge
related to web use is an important independent variable that influences Internet use
by men and women.
Prior to the 1990s, the scientific and military communities developed a
form of the Internet for their own purpose (ARPANET), but it was only in
the 1990s, with the development of hyper-text language, that the Internet
became a mass tool that could be employed by economic, political, religious,
and social groups that developed their own websites to communicate with
the larger public (Norris, 2001:27). The use of this new technology in an
effective manner required a home computer, and/or access to a computer at
work, or in a public setting (e.g., a public library or a school), funds (private
or public) to purchase monthly access to a service provider, and knowledge
of the computer in relation to web use. One would expect that the lower
socioeconomic class and culturally deprived minorities (e.g., racial and
ethnic minorities, rural residents) would be limited in their use of this new
n
Direct correspondence to Ira M. Wasserman, Department of Sociology, Eastern Michigan University, Ypsilanti, MI 48197 [email protected]. The authors will
share coding procedures for purposes of replication.
SOCIAL SCIENCE QUARTERLY, Volume 86, Number 1, March 2005
r2005 by the Southwestern Social Science Association
Gender and the Internet
253
technology (Wilhelm, 2000; Martin, 2003). These social groups would be
less likely to have a home computer and/or be able to pay the monthly fee
for access to such a service provider (Wilson, 2000).
One status group whose level of web use has been problematic has been
women. In the early 1990s women had less experience with computers than
men. They were more likely than men to use computers at work (Kaplan,
1994), but this use was for routine office activities, such as word processing
and spreadsheet work. Women saw men as being better able to comprehend
the Internet (Newton, 2001), possessed less self-efficacy toward the computer, and had high levels of computer anxiety (Durndell and Haag, 2002).
All these technophobic factors led to a gender gap in Internet use in the
1990s (Dhalokia, Dhalokia, and Pedersen, 1994).
The accelerated growth of this new technology in the 1990s narrowed the
gap between men and women with regard to Internet access, and by 2000
this access gap had almost disappeared (Nie and Erbring, 2000; Norris,
2001:82–84; U.S. Department of Commerce, 2000). Since 1995, numerous
survey studies (Chilsolm, 1996; Clemente, 1998; Flagg, 1999; Cummings
and Krout, 2002) have found that the new users of the Internet tend to be
women. Proportionately more women are attending medical and law schools
as well as entering science and engineering professions. Women have also
increased their participation in office administrative activities and have become knowledgeable about word processing, spreadsheets, and email communication (Morahan-Martin, 1998), although women in lower-status
occupational positions are less likely to have access to the web. Women are
also more likely than men to purchase products for home use (e.g., kitchen
products, decorative products, books), often on the web (Hilts, 1997; Raymond, 2000). There is also evidence that women are more likely than men
to employ the web to maintain social contacts (Howard, Rainee, and Jones,
2001; Jackson et al., 2001). Email is the most common form of Internet
social activity, and women tend to employ it slightly more than men (Nie
and Erbring, 2000). Chat-room communication is a more specialized form
of social communication, since it stresses anonymous communication.
All the changes just mentioned have influenced Internet use by gender and
recent studies (Bimber, 2000; Ono and Zavodny, 2003) find no gender
variation in access to the web, although these studies do find that women use
fewer websites, and also employ the Internet less frequently, than men. It
would be useful to extend this analysis of the gender gap with regard to the
Internet by considering possible causes for the variation in the scope and
frequency of website use.
Theoretical Perspective on the Gender Gap on the Internet
In considering the gender gap on the Internet, it is necessary to define
three aspects of this gap: (1) access to the Internet, (2) frequency of use of
254
Social Science Quarterly
the Internet, and (3) scope of use of the Internet. Access refers to the opportunity for individuals to use the web because they can utilize a computer
in a public or private setting and have connections to the Internet. The
frequency of use refers to the amount of time that an individual devotes to
the use of the Internet. The scope of use refers to the variety of websites (e.g.,
financial, health, hobby, commercial) used by an individual. With regard to
scope of use, it is necessary to determine whether some sets of sites are more
likely to be utilized by certain categories of Internet users.
With regard to access, current studies (Bimber, 2000; Ono and Zavodney,
2003) have found little significant variation in access by gender. In the early
stages of home computer use, the new technology was popularly portrayed
as a male domain; at that time, woman were more likely to be technophobic
(Levine and Donista-Schmidt, 1998). However, recently, women have begun to use computers in home and office settings to make consumer purchases and to exchange email messages with friends and colleagues. Given
their emotive role in family matters, women are more likely than men to use
email messages to maintain long-distance social network ties with friends
and relatives; by contrast, men and women differ little in their use of email
communication at the local level (Boneva, Krout, and Frohlich, 2001). As
they become more experienced with the Internet, women become increasingly competent in its use (Schumacher and Morahan-Martin, 2001). All
these findings would lead one to believe that there is little gender variation in
access to the web.
Although one can access the Internet through free sources at universities
and public libraries, for most individuals access usually requires the ownership of a home computer and the payment of a monthly service fee to an
Internet provider. Given this fact, one would expect socioeconomic and
cultural factors to influence Internet access. Thus, one would expect those
with lower incomes (Martin, 2003), and those individuals with less formal
education to have less access to the web. One might also expect African
Americans (Hoffman, Novak, and Schlosser, 2001) and Hispanics to have
less access, given their socioeconomic disadvantages, which limit their web
connectivity at home, work, and at public facilities not located in their
neighborhood. In the case of Hispanics, language might serve as a barrier to
their use of the web (Wilhelm, 2000).
Geographic location may also serve as a barrier to Internet use. One
would expect economic barriers to limit rural access to the web. Strover
(1999) and Bell, Reddy, and Rainie (2004) found that rural residents are
limited with regard to the quality of Internet providers. Regional geography
may also influence Internet use. Spooner (2003) found that web use was
highest in the New England states,1 with 66 percent of the residents having
1
These states are Connecticut, Maine, Massachusetts, New Hampshire, Vermont, and
Rhode Island.
Gender and the Internet
255
Internet access, and the Pacific Northwest,2 with 68 percent of these residents having web access. By contrast, Internet access was lowest in the
south,3 with only 48 percent of the residents having web access. It is likely
that regional variations in education and income may partially explain the
regional variation in web access.
The frequency of use of the Internet involves the amount of time an
individual uses the web for social and/or professional activity. Many individuals use the Internet for social entertainment, to play games, and for
hobby interests. By contrast, other individuals use this new technology for
business and commercial activities, such as banking and stock transactions.
One would expect individuals of higher socioeconomic status, and those
involved in the economic system (often men, for financial reasons) to make
greater use of the Internet.
The scope of use of the Internet involves the employment of millions of
websites currently available, with the number of these websites expanding at
an astronomical rate. In principle, it is possible to categorize all these different websites into various broad groupings (Web Bound, 2002). Government websites would include the sites created by the federal, state, and local
governments. Websites related to science, including biology, chemistry, geology, physics, and weather forecasting, have also expanded on the web.
Some of these sites (e.g., sports, sexually explicit materials) (Mehta, 2001)
are more likely to be male oriented; others (e.g., cooking, religious) can be
classified as female oriented, while a vast majority of them (e.g., health and
fitness, games) might be classified as androgynous. One would expect the use
of these various types of sites to vary among men and women. Individuals
are unlikely to utilize all of these sites, but are more likely to employ sets of
sites for their activities.
The variation in the scope and frequency of use of these various sites by
gender may be caused by (1) socioeconomic differences between men and
women and (2) gender-specific differences related to the Internet (Bimber,
2000:870–71). Socioeconomic differences between men and women are
related to the fact that men have higher income levels and, at present,
slightly higher educational levels. These differences may also be related to
work and home activity by men and women that influence the availability of
the web. Being at work full time, and especially in professional and administrative work, gives men greater access to the web, as well as to technical
experts who can effectively advise them on its use—a resource that women
who are at home lack.
A second reason for any variation in frequency and scope of use may be
gender-specific differences caused by lifetime experience with technology.
Men have been more familiar with computers and the Internet than women,
2
These states are Oregon and Washington.
These states are Alabama, Arkansas, Kentucky, Louisiana, Mississippi, Tennessee, and
West Virginia.
3
256
Social Science Quarterly
and therefore they possess more information and skills about how to use this
new technology (Goulding, 2003). At present, women are less experienced
with regard to this new technology and this may influence their level of use
of it. Under situations (e.g., classrooms, work activity) where men and
women both use the web, studies (e.g., Martin, 1998; Wei, 1998) have
found that women are just as proficient as men in its use. Thus, genderspecific differences appear to be related to historical differences between men
and women with regard to the use of technology and female preferences for
face-to-face social interaction. One would predict that controlling for socioeconomic differences between men and women, as well as gender-specific
differences in experience, should reduce the influence of gender on the
frequency and scope of Internet use.
Methodology
The data for this study are drawn from the General Social Survey (GSS)
for the year 2000 (Davis and Smith, 2000). The GSS has been conducted
for a number of years between 1972 and 2000, with the survey results being
accumulated over the years. The survey was administered to a representative
national sample between 1972 and 2000. For the year 2000, the GSS
included, for the first time, a series of questions concerned with computer
and Internet use; these computer and Internet questions were administered
to 2,362 randomly selected individuals. From this total sample for the year
2000 it was possible to determine whether an individual had access to the
World Wide Web (WWW). For those individuals who answered the questions related to web use, it was found that 992 individuals had access,
whereas 1,318 individuals did not. Independent of this web use, the 2000
GSS questioned 1,098 respondents regarding their level of use of email and
the locations (i.e., home, work, public facilities) where they utilized email.
The extent of email use was determined from the hours per week that the
respondents used this technology. Since this variable was skewed, it was
transformed (Neter et al., 1996:129–34) to:
emailtr ¼ lne ð1 þ emailÞ:
ð1Þ
Among web users, the 2000 GSS determined the extent of their use of
this new technology. The number of web users who were questioned was
reduced by one-third by only surveying those individuals who used the web
at some time during the previous week. Frequency of web use was determined by ascertaining how many hours per week the respondent used the
Internet. Since this variable was skewed like email use, it was transformed in
a manner similar to Equation (1). To ascertain the scope of use of the
Internet, the GSS considered a series of 21 generic websites (e.g., health,
sports, sexually explicit materials), which will be examined in the analysis
section of this article, and determined whether a respondent used the various
Gender and the Internet
257
types of websites within the past 30 days and their degree of use of that type
of site (National Opinion Research Center, 2001:794–97). For each type of
website, the respondent was asked whether in the past 30 days they had not
used the site, used it one to two times, three to five times, or more than five
times.
The 21 generic categories do not cover all websites, but they do include a
large proportion of them and it is unlikely that excluded sites would be used
more frequently by women than men, thereby biasing our findings. For
example, one might think that excluding commercial sites would bias the
findings in favor of men, but other studies ( Jorgensen, 2001; Kennedy,
Wellman, and Klement, 2003:85) have found that men shop online more
than women.
To determine the scope of use of the Internet, a factor analysis was
performed on the 21 websites. Factor analysis (Harman, 1976), also known
as latent variable analysis, examines a set of variables, and considers the
underlying and unobserved factors that may explain and summarize complex
phenomenon. For example, one might expect a social conservative to
strongly oppose abortion, gay marriage, and affirmative action. Exploratory
factor analysis (Nunnally, 1978:327–404), the type utilized in this study,
considers the 21 variables, which have a score of 0 if the respondent did not
utilize them in the past 30 days and 1 if they did use them in that time, and
determines which factors belong to which groups. Employing a varimax
rotation technique, the most widely used orthogonal rotation method (Nie
et al., 1975:485), a set of factor loadings was computed using the SPSS
Factor program. For each factor loading, the study selected only variables
with factor scores greater than 0.50 (the usual cut-off point), and then
created a factor score using the following formula (Nie et al., 1975:488):
Fi ¼ f 1 Z1 þ f 2 Z2 þ . . . ¼ f i Zi ;
ð2Þ
where fi 5 factor loading greater than 0.50, Zi 5 standard score for Variable
I 5 (Variable I mean Variable I)/(SD Variable I).
In the analysis section of this article, four factor scores will be created and
related to the other independent variables.
Past studies (Bimber, 2000; Ono and Zavodny, 2003) have employed
education, income, housewife status, and working full time as measures of
SES. However, most studies (Nie and Erbring, 2000; Wilhelm, 2000:25;
Mossberger, Tolbert, and Stansbury, 2001; National Telecommunications
and Information Administration, 2001; Lenhart, 2003) use education and
income as measures of socioeconomic status. A problem with using income
as a measure of SES for this study is that the relation between income and
web use is not linear but curvilinear, with low-income individuals having
little use of the web, and higher-income individuals having significant use.
To account for this curvilinear relation, it is necessary to relate income and
income2 to web use, creating a second-order nonlinear model. A difficulty
with doing this is that there is a high correlation between income and
258
Social Science Quarterly
income2, introducing multicollinearity into the model. To reduce this multicollinearity in the model, the income variable is transformed to:
Incometr ¼ Income Mean Income:
ð3Þ
This centering technique substantially reduces multicollinearity in the model, and allows these two measures of income to be utilized in the study. For
this article, Education (years) and Incometr and Income2tr will be utilized as a
measure of SES.
In relation to geography, it was first ascertained whether an individual
lived in a rural area or not, being given a score of 1 if he or she resided there,
and 0 otherwise. Since Internet use was highest in the New England and
Northwest Pacific states,4 and lowest in the south,5 three regional geographic variables will be created, including New England states, Pacific
states, and southern states. For each of these three categories of states, an
individual will be given a score of 1 if he or she resides in the respective
regions, and a score of 0 otherwise.
Other studies have found that knowledge of the computer and the Internet influences its use. Mossberger, Tolbert, and Stansbury (2001:15–37)
found that there was a computer skill divide in relation to Internet use, while
Hargittai (2002) found that there were differences in people’s online skills
that influenced their proficient use of the Internet. Given these findings, an
Internet knowledge score, titled Knowledge, was created for each respondent
from a set of questions related to the utilization of the web. Four items
related to Internet use were considered, and they concerned whether the
respondent (1) knew how to download information, (2) knew how to upload information, (3) knew how to use a hyperlink, and (4) knew the name
of five search engines (e.g., Google, Yahoo, Copernicus, AltaVista). For each
of the four items, the respondent was given a score of 0 of he or she did not
know how to accomplish each of the tasks, and 1 if he or she did know how.
A combined Knowledge score was then created for each respondent by
adding together the scores for the four items. The Knowledge variable had a
range of values from 0 to 4, and an alpha value of 0.6250, where the alpha
value is a measure of the consistency of answers to the four items. An alpha
value above 0.60 indicates that the items form a consistent scale.
A series of other control variables were also created from the 2000 GSS
data set. The respondent’s age in years was determined, and six qualitative
variables were created. If a respondent was African American, Asian or other
related group, Latino, female, married, or divorced, he or she was given a
score of 1, and 0 otherwise.
4
The 2000 GSS includes California in the Pacific states, not separating the northwest states
from California. Our measure will include all three Pacific states.
5
The GSS does not include all the southern states in the Spooner (2003) study, but has
most of his states in the south central area of the nation, including the states of Alabama,
Kentucky, Mississippi, Tennessee, and West Virginia. This study will equate the south
central states with the southern states in the Spooner (2003) study.
Gender and the Internet
259
Analysis of Results
To determine access to the web, a qualitative variable was created from the
2000 GSS data, where a respondent was given a score of 1 if he or she used
the web in any setting, including at home, work, school, and/or public
location, and 0 otherwise. Since this measure concerns web and nonweb use,
it was not possible to relate the measure to Internet knowledge. Web access
was statistically related to a set of variables that involve SES, social and
geographical variables, and racial and ethnic variables, using logistic regression to estimate the coefficients in the model. The coefficients provide an
indication as to whether individuals in all the previous social categories are
more or less likely to use the web (Hosmer and Lemeshow, 1989). Table 1
indicates the estimates for the various control variables. The computed Exp
(B) is an odds ratio, where a value of 1 specifies that the variable has no
impact on web use, a value less than 1 shows that individuals with those
social characteristics are less likely to use the web, and a value greater than 1
demonstrates that individuals with those social characteristics are more likely
to use the web.
The findings in Table 1 are consistent with previous findings related to
gender and access, since they point out that there is no gender gap with
regard to web access. With regard to SES, Nie and Erbring (2000) found
TABLE 1
Logistic Regression Analysis Relating a Set of Independent Variables to Web
Access for 2000 General Social Survey
Variables
Socioeconomic Variables
Education (years)
R’s Income 1998tr
R’s Income 19982tr
Social and Geographic Variables
Female
Age
Married
Divorced
Rural
New England
South central
Pacific
Racial and Ethnic Variables
African American
Asian and other
Latino
Constant
Slope B
Exp(B)
0.352 n n
0.037 n n
0.002
1.423
1.037
1.002
0.029
0.040 n
0.434 n n
0.231
0.357 n
0.052
0.349
0.120
0.972
0.961
1.539
1.260
0.700
1.054
0.706
1.128
0.670 n n
0.110
0.395 n
3.243 n n
0.512
0.896
0.675
0.039
N 5 2,310; model chi-square 5 383.140; po0.01;
po0.01; n0.01opo0.05.
nn
260
Social Science Quarterly
education and age to be the most important determinants of access, while
Mossberger, Tolbert, and Stansbury (2001:35), examining three other recent studies, found a significant educational and income gap. The findings
in Table 1 are consistent with these previous findings. Age and marital status
are statistically related to access, with younger and married individuals having more access. As predicted earlier in the article, geographical rural location significantly influenced access. Regional geographical location,
although it is statistically related to access by itself, does not influence access when controls are introduced into the model. Consistent with the
findings of Mossberger, Tolbert, and Stansbury (2001:35), African Americans and Latinos have less access to the Internet than whites, even with
controls for education and income. One would expect that cultural lags in
web use by racial and ethnic minorities would explain this discrepancy and
should diminish over time. However, findings by Lenhart (2003:8) between
2000 and 2002 suggest that this gap may persist, especially for African
Americans.
Before examining web use in relation to the content of the web, it will be
useful to study communication on the web in relation to email and chatroom use. A chat-room variable that determines the amount of hours per
week that respondents spend in chat-room activity is utilized, and is transformed like the email variable (Equation (1)) to take account of skewness in
this measure. These two communication measures on the Internet are related to the independent variables in the study by using ordinary least
squares (OLS) to determine their influence on the two dependent variables
(Table 2).
With regard to email communication, previous studies (Boneva, Krout,
and Frohlich, 2001; Nie and Erbring, 2000) have shown that women are
more likely to use email for long-distance personal communication, but not
for other forms of communication. The findings in Table 2 are consistent
with these results, since women use email more than men, but the differences in use are not significant. The socioeconomic variables are the most
significant for explaining email use, with those individuals with higher levels
of education and income using email more frequently. Email is used for
business and professional activities to main social contacts. Internet users
and email users are wealthier and more highly educated (Hoffman, Novak,
and Schlosser, 2001), and the findings in Table 2 reflect this fact.
With regard to chat-room use, previous studies (e.g., Nie and Erbring,
2000) have found users to be young and anonymous. The results in Table 2
support this conclusion, although it shows that these users are not significantly younger. The findings in the table suggest that they are more likely
to be males with lower income levels, and with a significantly higher probability of being divorced. Chat rooms are used for a variety of purposes,
including social and sexual contacts, the swapping of news and political
information, and the interaction of professionals (e.g., librarians). When
social groups of men and women interact, men tend to dominate the
Gender and the Internet
261
TABLE 2
Ordinary Least Square Estimates of Independent Variables Related to Email Use
and Chat-Room Use on the Internet
Email Use
Variable
Slope B
Socioeconomic Variables
Education
0.081 n n
R’s Income 1998tr
0.020 n n
R’s Income 19982tr
0.002 n n
Social and Geographic Variables
Female
0.069
Age
0.001
Married
0.080
Divorced
0.034
Rural
0.100
New England
0.037
South central
0.149
Pacific
0.063
Racial and Ethnic Variables
African American
0.156
Asian and other
0.170
Latino
0.222
Constant
0.252
N
1,098
R2
0.083
nn
Chat-Room Use
(S.E.)
Slope B
(S.E.)
(0.012)
(0.006)
(0.001)
0.015
0.016 n n
0.001
(0.013)
(0.006)
(0.001)
(0.060)
(0.003)
(0.071)
(0.088)
(0.111)
(0.124)
(0.128)
(0.082)
0.174 n n
0.004
0.109
0.219 n
0.022
0.071
0.199
0.017
(0.062)
(0.003)
(0.074)
(0.093)
(0.117)
(0.134)
(0.134)
(0.082)
(0.089)
(0.159)
(0.115)
(0.204)
0.126
0.160
0.016
0.882
555
0.080
(0.106)
(0.160)
(0.115)
(0.212)
po0.01; n0.01opo0.05.
conversation because they are likely to use their social power to control the
social interaction (Carroll, 2002). This social domination is also present in
many chat rooms, which discourages women from participating in them.
The findings in Table 2 support the previous findings that chat rooms are
male dominated. Women have attempted to increase their chat-room use by
forming feminine online chat groups.
Next, we examine the frequency of web use by considering the number of
hours per week that respondents are online, transforming the measure like
the previous two measures to take account of skewness. Since the knowledge
variable was also skewed, it was transformed in a similar manner. Table 3
indicates the OLS estimates for the independent variables in the study.
Consistent with previous studies (Nie and Erbring, 2000; Ono and
Zavodny, 2003), it is found that women use the Internet less frequently than
men. Socioeconomic status does not significantly influence frequency of use,
nor does geography. The lack of any causal relationship between socioeconomic status and frequency of use is surprising, but understandable when
we realize that there are so many types of web users. Asians and others tend
to use the web more frequently than whites and other racial and ethnic
262
Social Science Quarterly
TABLE 3
Ordinary Least Square Estimates of Independent Variables for Level of Web Use
on the Internet for the 2000 GSS
Variable
Socioeconomic Variables
Education
R’s Income 1998tr
R’s Income 19982tr
Social and Geographic Variables
Female
Age
Married
Divorced
Rural
New England
South central
Pacific
Racial and Ethnic Variables
African American
Asian and other
Latino
Internet Skills
Knowledgetr
Constant
N
R2
Slope B
(S.E)
0.020
0.001
0.000
(0.015)
(0.007)
(0.001)
0.163 n
0.002
0.128
0.009
0.193
0.187
0.126
0.154
(0.074)
(0.003)
(0.087)
(0.109)
(0.137)
(0.164)
(0.165)
(0.094)
0.102
0.399 n n
0.076
(0.123)
(0.178)
(0.131)
0.661 n
0.816 n
553
0.107
(0.123)
(0.326)
po0.01; n0.01opo0.05.
nn
groups. Internet knowledge does significantly influence Internet use, with
those who are more knowledgeable using it more frequently. The causal
direction is difficult to ascertain with this cross-sectional data, since it is hard
to know whether Internet knowledge increases level of use, or whether use of
the web increases expertise in this new technology. The issue of causal
direction will be explored further when the article considers the scope of web
use.
The influence of lifetime experience on the frequency of Internet use by
gender was further explored by defining three age groups (i.e., 18–35,
36–60, and 61–89), and determining the slope and standard error estimates
for these three age groups for the variables in Table 3. In calculations not
shown here, it was found that only for the elderly age group were females
significantly less likely to use the Internet less frequently. For the other two
age groups, women did use the web less frequently than men, but the results
were not significant. The findings suggest that it is greater lifetime experience with technology that explains the greater use of the web by men, since
there was no significant variation in frequency of use by gender for the
young and middle-age sample.
Gender and the Internet
263
Previous studies of the Internet have shown that websites tend to be male
oriented, female oriented, or androgynous. For example, an Australian study
in 2000 (Australian Broadcast Authority, 2001) found that men were more
likely to use the web for financial trading (23 percent vs. 14 percent),
accessing the news (58 percent vs. 38 percent), and looking at sexually
explicit materials (25 percent vs. 6 percent). Using this a priori information,
the 21 websites were classified into these three categories, and their use by
gender was examined for the 2000 GSS data (Table 4). The findings indicate the level of use of the Internet for the various specific websites by
males and females, and compute a chi-square value for each of the sites to
determine variation in use by gender, with higher chi-square values indicating greater gender variation in use. Male, female, and androgynous sites
are classified in terms of whether there is a significant gender difference in
use, with androgynous sites being classified as those where there is no significant difference in use by gender.
The male- and female-oriented sites are those that one would expect from
prior information. Men were more likely to use websites that provided
financial information, government information, news and current events,
and sexually explicit information. By contrast, women were significantly
more likely to use religious and church sites, as well as cooking and recipe
sites. In general, men were more likely than women to use most of the 21
categories of websites, a finding that one would expect, given the findings in
Table 3.
Using factor analysis, it is possible to construct four factors related to these
21 types of website.6 The first factor involves entertainment and personal
interaction, with the variables that constitute the factor being humor, sexually explicit materials, and personal home page sites. The second factor
involves government and politics, with the variables that constitute the factor
being government information, news, and political information. The third
factor involves art and education, with the variables that constitute this factor
being art, music, school, and other educational sites. The fourth factor
involves hobbies and practical information, with the variables that constitute
this factor being cooking sites, hobby sites, and health sites. The first factor
explains 20.72 percent of the variance in total scores, the second factor 8.11
6
The following formulas were used to compute the four factors:
1. Factor 1 5 0.629 n((Humor Site 0.4027)/0.4905)10.640 n((Sexually Explicit Materials 0.1353)/0.3423)10.573 n((Personal Home Page 0.3176)/0.4659) 5 Entertainment
and Personal Interaction Sites.
2. Factor 2 5 0.727 n((Government Information 0.4605)/0.4988)10.551 n((News
0.7705)/0.4208)10.683 n((Political Information 0.2948)/0.4583) 5 Government and
Political Information Sites.
3. Factor 3 5 0.588 n((Art 0.2776)/0.4477)10.518 n((Music 0.4529)/0.4982)10.595 n
((School Information 0.2371)/0.4526)10.623 n((Other Educational Sites 0.5653)/0.4961)
5 Educational and Art Sites.
4. Factor 4 5 0.722 n((Cooking 0.3328)/0.4716)10.516 n((Hobby 0.5002)/0.5004)1
0.556 n((Health 0.5046)/0.5004) 5 Hobby and Practical Activity Sites.
264
Social Science Quarterly
TABLE 4
Variation in Type of Website Used by Gender in 2000 GSS
Website
Male-Oriented Websites
Financial
News and current events
Government
Sports
Sexually explicit information
Science
Humor
Personal home page
Hobbies, crafts
Sites related to work
Female-Oriented Websites
Religious and church related
Cooking, recipes
Androgynous Websites
School you or children attend
Other educational sites
Travel
Music, concerts
Visual art/art museums
Television/movie
Health and fitness
Games on computer
Political information
% Male Use (N1) % Female Use (N2) Chi-Square
62.5
84.3
53.5
58.7
23.9
49.8
45.3
36.6
54.7
66.1
(331)
(331)
(331)
(329)
(331)
(331)
(331)
(331)
(331)
(330)
47.0
70.0
38.9
25.2
4.2
37.5
35.2
26.4
46.3
55.5
(338)
(337)
(337)
(337)
(337)
(337)
(335)
(331)
(331)
(331)
16.207 n n
19.231 n n
14.239 n n
76.553 n n
54.144 n n
10.339 n
7.054 n
7.076 n
4.704 n
7.817 n
6.3
23.9
(331)
(331)
23.7
42.4
(337) 5.704 n
(337) 25.941 n n
23.0
55.2
67.7
48.3
26.3
29.3
48.6
42.7
33.0
(330)
(330)
(331)
(331)
(331)
(331)
(331)
(337)
(330)
24.3
57.9
63.6
42.4
29.1
33.2
51.9
43.6
27.0
(337)
(337)
(338)
(337)
(337)
(337)
(337)
(337)
(337)
0.156
0.499
1.221
2.349
0.657
2.637
0.722
0.054
2.885
po0.01; n0.01opo0.05.
nn
percent of the variance, the third factor 6.37 percent, and the fourth factor
5.59 percent of the variance in total scores. OLS is used to relate these four
factors individually to our independent variables.
Table 5 shows the estimates for the four factors for the various independent variables. Gender is significant only for the first and fourth factor,
with women being less likely to use the web for entertainment and personal
interaction, and more likely to use it for hobbies and practical matters. With
regard to Factor 1, entertainment and personal interaction, it is the less
educated, males, African Americans, and those with more Internet knowledge who tend to use these type of sites. For Factor 2, related to government
and news information, only Internet knowledge was significant. For Factor
3, related to art and education, it was the more educated, younger, and those
with more Internet knowledge who were more likely to use these sites. For
Factor 4, related to hobbies and practical activities, gender, being married,
and Internet knowledge were positively related to this factor. For all these
constructed types of sites, Internet knowledge significantly explains variation
in the constructed scores.
Gender and the Internet
265
TABLE 5
Ordinary Least Square Estimates of Independent Variables Related to Four
Factors Linked with the Scope of Web Use
Entertainment and Government and
Personal Activity
Political
Art and Education
Slope B
(S.E.)
Socioeconomic Variables
Education
0.068 n n (0.023)
R’s Income 1998tr 0.014 (0.011)
R’s Income 19982tr 0.002 (0.001)
Social and Geographic Variables
Female
0.582 n n (0.112)
Age
0.009 (0.005)
Married
0.369 (0.131)
Divorced
0.057 (0.166)
Rural
0.253 (0.208)
New England
0.016 (0.248)
South central
0.173 0(.250)
Pacific
0.200 (0.142)
Racial and Ethnic Variables
African American 0.372 n (0.186)
Asian and other
0.279 (0.269)
Latino
0.020 (0.198)
Internet Skills
0.871 n n (0.186)
Knowledgetr
Constant
0.221 (0.494)
N
(552)
0.167
R2
nn
Slope B
(S.E.)
Slope B
(S.E.)
Hobbies and
Practical
Slope B
(S.E.)
0.058
0.013
0.002
(0.026) 0.068 n n (0.027) 0.036 (0.023)
(0.013) 0.028 (0.013) 0.015 (0.011)
(0.002) 0.001 (0.002) 0.003 n n (0.001)
0.198
0.011
0.030
0.066
0.241
0.275
0.313
0.177
(0.127)
(0.006)
(0.149)
(0.188)
(0.236)
(0.283)
(0.285)
(0.161)
0.344
0.185
0.072
0.236 (0.133) 0.261 n (0.114)
0.023 n n (0.006) 0.007 (0.005)
0.106 (0.156) 0.331 n n (0.134)
0.357 (0.196) 0.082 (0.169)
0.125 (0.246) 0.098 (0.211)
0.172 (0.294) 0.058 (0.253)
0.386 (0.297) 0.242 (0.255)
0.022 (0.169) 0.046 (0.144)
(0.212) 0.381
(0.307) 0.349
(0.225) 0.200
(0.223) 0.041
(0.319) 0.517
(0.235) 0.051
(0.190)
(0.275)
(0.202)
1.326 n n (0.212) 1.287 n n (0.220) 0.733 n n (0.189)
3.778 n n (0.561) 2.914 n n (0.585) 1.391 n n (0.502)
(552)
(551)
(553)
0.153
0.126
0.066
po0.01; n0.01opo0.05.
To determine the direction of the causal relation between Internet
knowledge and frequency and scope of web use, it will be useful to use OLS
estimates to relate Internet knowledge to these two types of web use, as well
as the other independent variables in the study (Table 6). With regard to
gender and race, it is clear from Table 6 that women possess less Internet
knowledge than do men, as do African Americans in relation to whites and
other racial groups. For many of the dependent variables, race and gender
have an indirect effect on them through web knowledge. For the factor
scores, rural location has an inverse impact on Internet knowledge.
Frequency of web use has a significant impact on web knowledge, suggesting that the frequent use of the web increases Internet knowledge—
certainly not an unexpected finding. Only Factors 2 and 3 significantly
influence web knowledge. Thus, people who access these categories of sites
acquire increased web knowledge, which is not the case for Factors 1 and 4.
These type of sites may require greater technical sophistication to sift
through the multiple layers of information. Using sites related to Factors 1
and 4, which concern entertainment and personal interaction, as well as
266
Social Science Quarterly
TABLE 6
Ordinary Least Squares Estimates Relating Internet Knowledge to Other
Independent Variables and Level of Web Use and Factor Scores for
2000 GSS
Frequency of Web Use
Variable
Slope B
Socioeconomic Variables
Education
0.027 n n
R’s Income 1998tr
0.005 n
2
R’s Income 1998tr
0.000
Social and Geographic Variables
Female
0.106 n n
Age
0.001
Married
0.021
Divorced
0.021
Rural
0.091
New England
0.028
South central
0.100
Pacific
0.055
Racial and Ethnic Variables
African American
0.090 n
Asian and other
0.040
Latino
0.027
Level of Internet Use
Web use (hrs/week)
0.077 n n
Entertainment and personal
—
Government and political
—
Art and educational
—
Hobbies and practical
—
Constant
1.564 n n
N
(553)
R2
0.230
(S.E.)
Factor Scores
Slope B
(S.E.)
(0.005)
(0.003)
(0.000)
0.022 n n
0.006 n
0.000
(0.005)
(0.002)
(0.000)
(0.025)
(0.001)
(0.030)
(0.037)
(0.047)
(0.056)
(0.056)
(0.032)
0.102 n n
0.001
0.026
0.009
0.110 n
0.026
0.066
0.040
(0.026)
(0.001)
(0.030)
(0.037)
(0.046)
(0.055)
(0.055)
(0.032)
(0.042)
(0.061)
(0.005)
0.112 n n
0.013
0.026
(0.042)
(0.060)
(0.044)
(0.014)
—
—
—
—
(0.089)
—
0.019
0.033 n n
0.028 n n
0.008
1.750 n n
(549)
0.269
—
(0.010)
(0.009)
(0.009)
(0.010)
(0.087)
po0.01; n0.01opo0.05.
nn
hobbies and practical information, has no significant impact on web
knowledge, perhaps because these sites are more user friendly. It is probably
the case that the use of these latter categories of sites does not require that
the individual user master this new technology to the same degree as the
second and third categories of sites.
Discussion
This article has been concerned with the causes for gender variation in
Internet use. Most prior studies of this question (e.g., Bimber, 2000; Ono
and Zavodny, 2003) have focused on issues related to web access and frequency of web use. This study, employing nationwide GSS data for the year
Gender and the Internet
267
2000, extended the analysis to consider email and chat-room use, as well as
the scope of web use by employing four constructed factors from 21 categories of websites. Unlike previous studies, this one considered the importance of web knowledge for explaining variation in the frequency and
scope of Internet use.
With regard to web access, the findings in the article are consistent with
other studies that found no gender variation in access. The web has become
an integral part of the home and work environment, which has acted to give
women access to this new technology. As men purchase home computers
and obtain web access related to their work activity, it gives married persons
in the household equal access to the web. Race and rural location did
influence access, with African Americans and rural residents having lower
levels of access. The racial disparity in access is likely caused by the work,
home, and educational disadvantages suffered by African Americans in
American society. The rural difference is due to the higher costs of access in
rural areas, as well as the level of choice of providers.
With regard to types of use, email and chat-room use involves personal
communication between social actors, social professionals (e.g., social scientists), and social organizations (e.g., corporation actors), with email communication involving all three types of personal communication, and chatroom interaction involving mainly personal, anonymous communication.
Women were more likely to use email communication than men, but the
difference in use was not statistically significant. It is likely that men are more
likely to use email communication for professional and commercial communication, while women use it more for personal, long-distance communication. This issue can be explored in future research by examining
variations in types of email use. With regard to chat-room use, it tends to be
male dominated, probably due to the fact that in society men use their higher
social and economic power to dominate conversations, and chat-room use on
the Internet reflects these conditions in the larger society (Carroll, 2002). The
greater use of chat rooms by divorced individuals may reflect their need for
increased social interaction after the rupturing of their family ties.
Consistent with previous studies, the article found that men use the web
more frequently than women, and this level of use was related to web
knowledge. It was also found that socioeconomic status had no significant
impact on the frequency of web use, probably because of the multiple uses of
the Internet for personal and professional activities. The findings raise the
question as to whether women are historically disadvantaged with their
knowledge of this new technology and if their disadvantage explains their
lower level of use. The article further explored the issue by considering the
scope of Internet use. Descriptive data, using chi-square statistics, showed
that men dominated certain types of websites (e.g., government, financial,
sexually explicit materials), and in general used more types of sites than
women. The article then utilized factor analysis to differentiate four types of
websites from among the 21 types of sites examined.
268
Social Science Quarterly
The findings in Table 5 show that Internet knowledge increases use of all
four types of sites. However, Table 6, examining the relation in the other
direction, showed that only two types of sites (i.e., governmental and educational) increased Internet knowledge, probably because the users of these
types of sites needed to know how to upload and download relevant information. Differences in Internet knowledge persist by gender, and may
continue to do so because of the different types of websites accessed by each
sex. The findings suggest that the higher level of use of a variety of websites
by men increases their web knowledge, which in turn causes them to utilize
this new technology more frequently than women. As women expand their
use of different types of websites, their web knowledge and their use of the
web should expand.
It is likely that prior to the introduction of web technology in the 1990s
women engaged less frequently in social activities (e.g., financial, scientific,
governmental) that influenced relative use of the various websites. The
growth of the web allowed women to more efficiently engage in these social
activities (e.g., trading stocks), which should expand their use of the web and
decrease the gender gap with regard to Internet knowledge. Future research
is needed to specify the activities involved in the use of the various categories
of websites and to identify the settings (e.g., work, home, school, public
facilities) where these different types of social activities occur.
REFERENCES
Australian Broadcast Authorities. 2001. Australian Families and Internet Use. Available at
hhttp://www.aba.gov.au/internet/research/families/index.htmi.
Bell, Peter, Pavani Reddy, and Lee Rainie. 2004. Rural Areas and the Internet. Washington,
DC: Pew Internet & American Life Project.
Bimber, Bruce. 2000. ‘‘Measuring the Gender Gap on the Internet.’’ Social Science Quarterly
81:868–76.
Boneva, Bonka S., Robert Krout, and David Frohlich. 2001. ‘‘Using E-Mail for Personal
Relationships: The Difference Gender Makes.’’ American Behavioral Scientist 45:530–49.
Carroll, Marnie Enos. 2002. Internet Chat Rooms: A Comparison of Conversations Among
Women’s, Men’s and Mixed Groups. Unpublished Ph.D. dissertation. University of Colorado.
Chisholm, Patricia. 1996. ‘‘Cyber-Sorority. Women Using the Internet.’’ Maclean’s 109:53–54.
Clemente, Peter. 1998. The State of the Net. The New Frontier. New York: McGraw-Hill.
Cummings, Jonathan N., and Robert Krout. 2002. ‘‘Domesticating Computers and the
Internet.’’ Information Society 18:221–31.
Davis, James Allan, and Tom W. Smith. 2000. General Social Surveys, 1972–2000 [machinereadable data file]. Principal Investigator James A. Davis and Co-Principal Investigator Tom
W. Smith; Co-Principal Investigator Peter V. Marsden, NORC ed. Chicago. National
Opinion Research Center Producer, 2000: Storrs, CT.
Gender and the Internet
269
Dhalokia, Ruby Ray, Nikhilesh Dhalokia, and Briget Pedersen. 1994. ‘‘Putting a Byte in the
Gender Gap.’’ American Demographics 16:20–21.
Durndell, Alan, and Zsolt Haag. 2002. ‘‘Computer Self-Efficacy, Computer Anxiety, Attitudes Toward the Internet and Reported Experience with the Internet, by Gender, in an
East European Sample.’’ Computers in Human Behavior 18:521–35.
Flagg, Jennifer L. 1999. ‘‘Women Joining Men in Droves on the Web.’’ Editor & Publisher,
the Fourth Estate 132:27–28.
Goulding, Anne. 2003. ‘‘Women and the Information Society: Barriers and Participation.’’
IFLA Journal 29:33–40.
Hargittai, Eszter. 2002. ‘‘Second-Level Digital Divide.’ First Monday (Web Journal) 7 (April).
Harman, Harry H. 1976. Modern Factor Analysis, 3rd ed. Chicago, IL: University of Chicago
Press.
Hilts, Elizabeth. 1997. ‘‘How Women are Changing the Web. Study by Net SmartResearch.’’ Editor & Publisher, the Fourth Estate 130:40.
Hoffman, Donna L., Thomas P. Novak, and Ann E. Schlosser. 2001. ‘‘The Evolution of the
Digital Divide: Examining the Relationship of Race to Internet Access and Usage Over
Time.’’ Pp. 47–97 in Benjamin M. Compaine, ed., The Digital Divide. Facing a Crisis or
Creating a Myth. Cambridge, MA: MIT Press.
Hosmer, David W., and Stanley Lemeshow. 1989. Applied Logistic Regression. New York:
John Wiley & Sons.
Howard, Philip E. N., Lee Rainie, and Steve Jones. 2001. ‘‘Days and Nights on the Internet:
The Impact of Diffusing Technology.’’ American Behavioral Scientist 45:383–404.
Jackson, Linda A., Kelly S. Ervin, Philip S. Gardner, and Neal Schmitt. 2001. ‘‘Gender and
the Internet: Women Communicating and Men Searching.’’ Sex Roles 44:363–74.
Jorgensen, Barbara. 2001. ‘‘Gap? What Gap?’’ Electronic Business 27:42.
Kaplan, Rachel. 1994. ‘‘The Gender Gap at the PC Keyboard. Women Use Office Computers More than Men, But Computer Companies Sell to Men.’’ American Demographics
16:18.
Kennedy, Tracy, Barry Wellman, and Kristine Klement. 2003. ‘‘Gendering the Digital
Divide.’’ IT & Society 1:72–96.
Lenhart, Amanda. 2003. The Ever-Shifting Internet Population. A New Look at Internet Access
and the Digital Divide. Washington, DC: Pew Internet & American Life Project.
Levine, Tamor, and Smador Donita-Schmidt. 1998. ‘‘Computer Use, Confidence, Attitudes,
and Knowledge: A Causal Analysis.’’ Computers in Human Behavior 14:125–46.
Martin, Shelley. 1998. ‘‘Internet Use in the Classroom: The Impact of Gender.’’ Social
Science Computer Review 16:411–18.
Martin, Steven P. 2003. ‘‘Is the Digital Divide Really Closing? A Critique of Inequality
Measurement in a Nation Online.’’ IT & Society 1:1–13.
Mehta, Michael D. 2001. ‘‘‘Pornography in Usenet’: A Study of 9,800 Randomly Selected
Images.’’ Cyber Psychology & Behavior 4:695–703.
Morahan-Martin, J. 1998. ‘‘Males, Females and the Internet.’’ Pp. 169–97 in J. Gackenback,
ed., Psychology and the Internet: Intrapersonal, Interpersonal and Transpersonal Applications.
San Diego, CA: Academic Press.
270
Social Science Quarterly
Mossberger, Karen, Caroline J. Tolbert, and Mary Stansbury. 2001. Virtual Inequality.
Beyond the Digital Divide. Washington, DC: Georgetown University Press.
National Opinion Research Center. 2001. General Social Surveys, 1972–2000 Cumulative
Codebook. Storrs, CT: Roper Center for Public Opinion Research, University of Connecticut.
National Telecommunications and Information Administration. 2001. ‘‘Falling Through the
Net: Defining the Digital Divide.’’ Pp. 17–46 in Benjamin M. Compaine, ed., The Digital
Divide: Facing a Crisis or Creating a Myth. Cambridge, MA: MIT Press.
Neter, John, Michael H. Kutner, Christopher J. Nachtsheim, and William Wasserman.
1996. Applied Linear Statistical Models, 4th ed. Chicago, IL: McGraw-Hill.
Newton, Stephen. 2001. ‘‘Breaking the Code: Women Confront the Promises and the Perils
of High Technology.’’ Women’s Studies Quarterly 29:71–79.
Nie, Norman, and Lutz Erbring. 2000. Internet and Society: A Preliminary Report. Stanford,
CA: Stanford Institute for the Quantitative Study of Society.
Nie, Norman, Lutz Erbring, Hull C. Hadlai, Jean G. Jenkins, Karen Steinbrenner, and Dale
H. Bent. 1975. Statistical Package for the Social Sciences, 2nd ed. New York: McGraw-Hill.
Norris, Pippa. 2001. Digital Divide. Civic Engagement, Information Poverty and the Internet
Worldwide. New York: Cambridge University Press.
Nunnally, Jum C. 1978. Psychometric Theory, 2nd ed. New York: McGraw-Hill.
Ono, Hiroshi, and Madeline Zavodny. 2003. ‘‘Gender and the Internet.’’ Social Science
Quarterly 84:111–21.
Raymond, Joan. 2000. ‘‘For Richer & for Poorer.’’ American Demographics 22:58–64.
Schumacher, P., and J. Morahan-Martin. 2001. ‘‘Gender, Internet and Computer Attitudes
and Experiences.’’ Computers in Human Behavior 17:95–110.
Spooner, Tom. 2003. Internet Use by Region in the United States. Regional Variations in
Internet Use Mirror Differences in Educational and Income Levels. Washington, DC: Pew
Internet & American Life Project.
Strover, Sharon. 1999. Rural Internet Connectivity Report. P. 99-13. September, 1999.
Columbia, MO: Rural Policy Research Institute.
U.S. Department of Commerce, National Telecommunication and Information Administration. 2000. Falling Through the Net. Toward Digital Inclusion. Washington, DC: U.S.
Government Printing Office.
Web Bound. 2002. The Internet Atlas of Web Sites. 150,000 America’s Best Web Sites.
Knoxville, TE: Web Bound, Inc.
Wei, Liming. 1998. ‘‘Women and the Internet.’’ Beijing Review 41:37.
Wilhelm, Anthony G. 2000. Democracy in the Digital Age. Challenges to Political Life in
Cyberspace. New York: Routledge.
Wilson, Ernest. 2000. Closing the Digital Divide: An Initial Review. Briefing the President.
Washington, DC: Internet Policy Institute.