Whose Space? Differences Among Users and Non-Users of Social Network Sites

Introduction

Social network sites (SNSs) have become some of the most popular online destinations in recent years (comScore, 2007a, 2007b). Not surprisingly, this level of user attraction has been accompanied by much coverage in the popular press, including speculations about the potential gains and harms stemming from the use of SNS services (Hempel, 2005; Magid, 2006; Stafford, 2006). Academic researchers have started studying the use of SNSs, with questions ranging from their role in identity construction and expression (boyd & Heer, 2006) to the building and maintenance of social capital (e.g., Ellison, Steinfeld, & Lampe, 2007) and concerns about privacy (e.g., Gross & Acquisti, 2005; Hodge, 2006). While these areas of inquiry are all important and worthy of exploration, a significant antecedent question has been largely ignored: Are there systematic differences between who is and who is not a SNS user, and are people equally likely to join the various types of services that exist? This article sets out to address this question.

A significant challenge for studies trying to answer questions about who is and is not using SNSs is that the samples on which they are based (e.g., Ellison et al., 2007) typically include such a small number of non-users that there is little variance present to explain differentiated basic adoption of the services. On the rare occasions when data have been available on non-users in addition to users, the focus of the studies has been elsewhere. For example, Pasek, More, and Romer (2007) have disaggregated data by site and variance on the usage of SNSs, but they look at the predictive power of SNS usage on civic engagement, employing SNSs as an independent variable, rather than exploring what explains their use in the first place. This article fills a gap in the literature by: (1) explaining differences in SNS adoption and (2) disaggregating SNS usage by specific service to see whether it is possible to predict use of one service over another based on the background characteristics of the user, information about the social context of use, and experiences with the medium.

Disaggregating usage by site also makes an important methodological contribution to the study of SNSs. As the results show, disaggregating which specific site one is researching is important, because people do not randomly select into their uses, and aggregate analyses of SNS use may make it difficult to identify important trends. This suggests that researchers should tread lightly when generalizing from studies about the use of one SNS to the use of another such service. While these sites do share commonalities, they also have distinct features—whether at the level of site design or the particular communities who comprise their user base—that may attract different populations and may encourage different types of activities. Thus, an examination of SNSs both in the aggregate and with respect to specific sites is important in order to gain a better understanding of how use of such sites is spreading across various population segments and the social implications of their usage.

Differentiating Types of Internet Uses

The New Yorker's now-classic cartoon proclaimed in 1993 that "[o]n the Internet, nobody knows you're a dog" (Steiner, 1993), suggesting that identity was so hidden online that opportunities would be widely open to all, regardless of background characteristics that may have traditionally disadvantaged some people compared to others. The idea that people would be on an equal footing online assumes that offline characteristics are not mirrored in people's online pursuits. However, subsequent research has found this not to be the case, for example, with respect to gender identity (Herring, 1993). Researchers have observed that despite initial impressions and arguments about how users shed their offline identities in online interactions (Turkle, 1995), offline identities very much carry over to online behavior (boyd, 2001; Smith & Kollock, 1999). This suggests that the Internet is not necessarily leveling the playing field in the way that the above-mentioned cartoon would have us believe, given that people bring constraints and opportunities from their offline lives with them to their online interactions and activities.

Indeed, studies looking at how different people use the Internet in their everyday lives have found systematic differences across types of users. For example, even after women caught up with men (in the United States) concerning basic connectivity statistics, their uses continued to differ. Men have been shown to spend more time online and claim higher-level skills than women (Bimber, 2000; Hargittai & Shafer, 2006; Jackson, Ervin, Gardner, & Schmitt, 2001; Ono & Zavodny, 2003), consistent with earlier literature on women and technology use more generally (Frissen, 1995; Hall & Cooper, 1991; Herring, 1994; Livingstone, 1992). Factors such as socioeconomic status have also been shown to predict types of Internet uses (Howard, Rainie, & Jones, 2001; Livingstone & Helsper, 2007; Madden & Rainie, 2003). For example, so-called "capital-enhancing" activities (DiMaggio & Hargittai, 2002), such as looking for financial, political, or government information online, are associated with socioeconomic status (Howard et al., 2001). Moreover, the circumstances under which people use the medium—such as their autonomy (Hassani, 2006) and experience of use (Howard et al., 2001)—are also related to the purposes to which they put the medium. Research has shown that more locations where one has Internet access and more time spent online are associated with more diverse types of uses (Hargittai & Hinnant, 2005).

Research on refined understandings of the digital divide has found that even once people go online, differences exist among their online pursuits (DiMaggio, Hargittai, Celeste, & Shafer, 2004; Hargittai, 2002, 2007; Livingstone & Helsper, 2007; Mossberger, Tolbert, & Stansbury, 2003; van Dijk, 2005). Given that various background characteristics of people, the context of their Internet uses, and their level of experience have all been shown to influence types of Web uses in general, it is worth considering whether they may also relate to social network site usage in particular. That is, given earlier work on differentiated Internet use among people from different backgrounds, there is no reason to assume equal adoption of SNSs across population segments. Work that focuses solely on users of social network sites excludes, by definition, people who are not SNS users. Insofar as these people are systematically different from those who embrace these services, it is problematic not to know anything about them, since researchers thereby risk unintentionally excluding entire groups of people from discussion about SNSs.

The Challenges of Studying SNS Adoption

An important reason for the scarcity of work that predicts SNS usage is the lack of appropriate data necessary to address such questions. Despite Internet user studies starting to focus on particular online behaviors, rather than considering all online actions to be uniform (Howard & Jones, 2004; Wellman & Haythornthwaite, 2002), categorizations of online activities have remained relatively broad, making it difficult to understand who does what online, why, and how this influences the rest of people's lives. Additionally, because the popularity of SNSs is relatively recent, initial data collection efforts about Web uses did not focus on them. It is more customary to ask about the topics people encounter on websites (e.g., Internet use for the purposes of gathering information about news or health matters) than to inquire in detail about the particular sites and communities in which people may be participating. Moreover, because individuals' goals and activities on SNSs are extremely varied, investigating their uses through traditional survey instruments poses several new and distinct challenges. Perhaps due to such methodological challenges, most related work has focused on more exploratory questions regarding SNS usage, typically relying on qualitative methods (e.g., boyd, 2008; Dwyer, 2007).

Another challenge in studying social network site usage stems from the fact that large-scale questionnaires (e.g., the Current Population Survey and the General Social Survey) have mainly focused on adult populations, with relatively few young people represented in their samples. Yet, young people are known to be some of the most likely to participate on some SNSs (e.g., Facebook's initial focus on college students and then high school students left out older people by design), suggesting that concentrating on adolescents and young adults is especially important if researchers are to gain a better understanding of how such sites are being incorporated into people's lives. Moreover, because young adults are much more wired than their older counterparts (Fox, 2004; Madden, 2006), it can be beneficial to focus studies on this population, especially if the goal is to understand refined measures of use once basic access and connectivity are controlled for.

One study has addressed questions similar to those raised here, although it focused on a somewhat different age group (12-17 year olds) and different aspects of SNS use. The Pew Internet and American Life Project administered a survey on the social network site usage of teens in late 2006 (Lenhart & Madden, 2007). Although the survey did not ask about social network site usage by service (except to inquire on which service users updated their profiles the most often), the study offers helpful insight into differences in various young people's adoption of such sites. Namely, the data suggest different uptake by age and gender within the group of 12-17 year olds in the sample and also some differences by race and ethnicity. However, the study does not present more detailed analyses and also lacks the data that would allow comparison of SNS adoption by service.

College students in the U.S. constitute an ideal population in which to study differences in particular types of digital media uses, given their high connectivity levels. Often, the lack of data on young people's experiences with information and communication technologies makes it difficult to know whether assumptions about their active online participation are warranted. It would be incorrect to assume that simply using the medium can be equated with equal use of all sites in similar ways. A systematic study of everyday digital media practices is essential to understanding how communication and information technologies are affecting the lives of different types of young adults. The next section introduces the unique data set used in this study to address these questions, followed by findings from bivariate and logistic regression analyses explaining differential social network site adoption.

Methods

The analyses presented here are based on data representing a diverse group of mainly 18- and 19-year-old college students. The study was conducted in February and March of 2007 at the University of Illinois, Chicago, which is a U.S. urban public research university.¹ U.S. News and World Report (2006) ranked this campus among the top 10 national universities as regards campus ethnic diversity, suggesting that this school offers an ideal location for studies of how different kinds of people use online sites and services.

The project had the support of the First-Year Writing Program at the university, ensuring that a representative sample of the school's undergraduate student body would participate. The writing course offered through this program is the only course on campus that is required of all students; thus, enrollment in it does not pose any selection bias. Out of the 87 sections offered as part of this course, 85 took part in the study, constituting a 98% participation rate on the part of course sections. Overall, there was a final response rate of 82% based on all of the students enrolled in the course. In order to control for time in the program, this article focuses on students in the first-year class.

The survey was administered on paper instead of online. Relying on an online questionnaire when studying Internet uses could create a bias toward people who spend more time online, given that they may be more inclined to fill out the questionnaire and also, perhaps, more inclined toward higher rates of participation on the sites of research interest. The average survey completion time was approximately 30 minutes. The survey included detailed questions about respondents' Internet uses (e.g., experience, types of sites visited, and online activities) and their demographic background.

Basic demographic information was measured using standard modes of operationalization. Students were asked their year of birth, and this information was used to calculate their age, which is included in the models as a continuous variable. Male is the base gender category (male=0, female=1). Information about race and ethnicity was collected using the U.S. Census Bureau (2000) questionnaire format, and dummy variables are used in the statistical model, with White as the omitted category. Consistent with work by others, parental education was used as a measure of socioeconomic status (e.g., Carlson, Uppal, & Prosser, 2000; Lamborn, Mounts, Steinberg, & Dornbusch, 1991; Stice, Cameron, Hayward, Taylor, & Killen, 1999). Since asking about household income has limited utility with such an age group (both because students do not know their parents' income and because those who live in dorms may not know how to interpret "household"), and since educational level is constant in this group (every respondent is in the first year of college), parental schooling is a helpful measure. This information is included in the model as dummy variables, with some college education (but no college degree) as the base.

Both the question about living at home with parents and the question about having access to the Internet at a friend's or family member's house is included as a dummy variable, where 1 signals yes to that question, and 0 stands for no. Finally, figures for both hours spent online per week and number of years a respondent has been an Internet user are logged in the analyses, given that an additional hour or year, respectively, likely has diminishing returns as the values increase. The analyses first consider only the core background characteristics of the user (age, gender, race and ethnicity, parental education). Then, a second model includes information about context and experience, with use supplementing the core demographic variables.

The 1,060 first-year students included in these analyses represent a diverse group of people.² Fifty-six percent of the respondents are female, 44% are male. Almost all are 18 or 19 years old, with a mean age of 18.4 and a median of 18. Fewer than half are White and non-Hispanic. Slightly less than 8% claim African or African-American descent, almost 30% are of Asian or Asian American ancestry, and just under one-fifth are of Hispanic origin. These students come from varied family backgrounds. Over a quarter of respondents have parents whose highest level of education is high school, with an additional 20% whose parents do not have a college degree. While it may seem that sampling from a college population assumes a highly educated group, 25% of first-years at this university drop out of college by their second year (Ardinger et al., 2004) and fewer than half (43.6%) will graduate within six years of enrollment (University of Illinois-Chicago, 2004). Unlike many U.S. colleges, over half of the students at this university commute from home and live with their parents (53.1%).

Baseline access and use statistics (Table 1) for the sample suggest that the Internet is not a novel concept in most of these students' lives. On average, participants have access to the Internet at over six locations and have been users for over six years. When asked how often they go online, the vast majority report doing so several times a day. They estimate spending 15.5 hours visiting Web sites weekly (excluding email, chat, and VoIP). While there is certainly some amount of variation in access and use, there are no basic barriers standing in the way of these young adults accessing the Internet. Limits may be put on their uses due to other factors (e.g., the need to share resources at home, limited hours of access due to employment), but they all have basic access. This suggests that traditional concerns about the so-called digital divide do not apply to these students as regards basic availability of the Internet. Thus looking at such a wired group of users allows us to hold basic access to digital media constant and focus on differences in details of use instead.