Browsing the Web for School: Social Inequality in Adolescents’ School-Related Use of the Internet

Inequality in Adolescents’ ICT Usage

The study of different usage patterns by sociodemographic characteristics has increased as opportunities for Internet access have risen, thus, research in this field started around the millennium (Attewell, 2001; DiMaggio & Hargittai, 2001; DiMaggio, Hargittai, Celeste, & Shafer, 2004). Most studies in this field have differentiated between informational versus entertaining Internet usage and examined the social stratification of usage patterns. These studies found that adolescents from families with higher social backgrounds tend to use the Internet for informational purposes more often than adolescents from lower socioeconomic backgrounds (Micheli, 2015; Notten et al., 2009; Peter & Valkenburg, 2006), while those with fewer socioeconomic resources use ICT more often for entertainment (Peter & Valkenburg, 2006) and playing games (Koivusilta, Lintonen, & Rimpelä, 2007; Notten et al., 2009). These findings are consistent with research conducted on adults’ online activities (Bonfadelli, 2002; Van Deursen, van Dijk, & ten Klooster, 2015; Zillien & Hargittai, 2009). Bonfadelli (2002) notes, “More educated people use the Internet more actively and their use is more information oriented, whereas the less educated seem to be interested particularly in the entertainment functions of the Internet” (p. 65).

Other studies which use clustering methods have identified different usage types among children and young people (Eynon & Malmberg, 2012; Livingstone & Helsper, 2007; Van den Beemt, Akkerman, & Simons, 2010). They were able to distinguish specific online user types, as exemplified by van den Beemt et al.: “traditionalists” who mainly engage with the Internet through browsing, “gamers,” “networkers,” and “producers” who use the web for a wide range of activities and upload content. It has been demonstrated that middle class children participated in more online activities than their peers from working class backgrounds (Livingstone & Helsper, 2008).

Research focusing on content production and sharing activities revealed that the link between parental resources and these activities is not that clear. College students with higher educated parents were described as more likely to be producers of content (Hargittai & Walejko, 2008). In contrast, Correa (2010) found no relation between parental educational levels and content creating. Other studies showed that the type of content produced should be considered: There is no association between individuals’ educational level and the creation of entertainment content or social network site usage (Blank, 2013; Micheli, 2015). In contrast, political content is more often created by highly educated people (Blank, 2013) and students with better educated parents are more likely to produce skilled content (their own website, blog, post created content, online forum, tweets on twitter) than their peers with less educated parents (Micheli, 2015).

To sum up, previous research has clearly demonstrated socially stratified ICT usage patterns among adolescents and young people. Adolescents from higher social backgrounds use the Internet more often for informational purposes and are more likely to produce “skilled content.” However, ICT usage for “informational purposes” is a very broad category, which encompasses very heterogeneous activities (e.g., reading news, obtaining practical information). Therefore, we shall now look more specifically at adolescents’ educational ICT use.

Inequality in Adolescents’ Educational ICT Usage

A growing body of research examines ICT usage at school and digital learning opportunities (Bulman & Fairlie, 2016; Falck, Mang, & Woessmann, 2018; Fraillon et al., 2014; Patterson & Patterson, 2017). In school, students use ICT most frequently in “information technology or computer studies,” followed by the natural sciences and human sciences (Fraillon et al., 2014, p. 151). In addition, students’ ICT educational usage outside school is receiving increasing attention (Ben-David & Kolikant, 2012; Furlong & Davies, 2012). Research on the social selectivity of using ICT for educational purposes outside school shows inconsistent results. Some studies found differences by social background regarding computer use for educational computer activities (Steffens, 2014; Vekiri, 2010), whereas others did not (Cranmer, 2006; Eamon, 2004; Gümüş, 2013). Gümüş (2013) analyzed the Turkish sample of the PISA 2009 study and found no significant link between family wealth, parental education, or cultural possessions with students’ ICT use for school-related activities at home. Eamon (2004) reported that children from poor and non-poor families did not differ in their home computer use for academic purposes in a U.S. sample of children aged 10 to 14 years. A qualitative study by Cranmer (2006) on use of the Internet for homework found no differences between family types within a sample of 11- to 15-year-old children from 17 families in the United Kingdom.

On the contrary, Vekiri (2010) showed that educational computer activities are carried out more often by students from better-off families in Greece. Higher percentages of children from better-off households are reported for the activities drill and practice, writing, and Internet search. Steffens (2014) showed that adolescents’ (from Germany, Finland, and the Netherlands) Internet use at home for school functions increases as socioeconomic status rises, but does not control for other variables. In addition, better-off parents are more likely to regard ICT as an educational tool (Linebarger & Chernin, 2003), which could result in different ICT mediation practices (Clark, 2011).

To conclude, the findings on the social selectivity of educational ICT are not consistent. As demonstrated, some studies have found a relationship between social origin and educational ICT activities, whereas other researchers reported that adolescents do not differ in their educational ICT use by social background. These studies are carried out on relatively small samples and focus on single regions or countries only. In addition, previous studies that examined online content production and sharing of young people focused mainly on college students (Correa, 2010; Hargittai & Walejko, 2008). Thus far, educational sharing practices have not been analyzed from a social stratification perspective.

Measurements

The dependent variables are measured by the self-reported time spent on browsing and sharing/uploading Internet activities with school-related and entertainment purposes (Figure 1). The students were asked how often they used a computer outside school for different activities. They could choose between doing the activity never or hardly ever, once or twice a month, once or twice a week, almost every day, or every day. We selected the following items to represent the different categories from Figure 1: “browsing the Internet for fun (such as watching videos, e.g., <YouTube™>),” “uploading your own created content for sharing (e.g., music, poetry, videos, computer programs),” “browsing the Internet for schoolwork (e.g., for preparing an essay or presentation)” and “sharing school related materials with other students” (OECD, 2011).

To measure parental cultural capital, the number of books at the student’s home and the highest educational qualification of the parents are used. The number of books in the parental home is measured using three categories: (0) 0 to 25 books, (1) 26 to 200 books, and (2) more than 200 books. Parental education is measured according to the International Standard Classification of Education (ISCED). In our analysis, we distinguish three categories: (0) up to lower secondary education (ISCED ≤ 2), (1) upper secondary education and postsecondary nontertiary education (ISCED 3 or 4), and (2) tertiary education (ISCED 5 or 6).

Previous research has shown that Internet activity can be influenced by diverse factors. We, therefore, include various control variables: migration background (1: the student or at least one parent was born outside the test country, 0: the student and both parents are born in the test country), gender (1: female, 0: male), ICT possession (number of ICT devices available for students’ use at home, e.g., a desktop computer, laptop or tablet), Internet experience (1: student was 9 years old or younger at first access to the Internet, 0: student was 10 years old or older), ICT usage in school (index of nine specific activities at school, see OECD, 2014), reading literacy (results of students’ reading test score). In all models, we also control for country dummy variables.

We removed from the data set all students who indicated that they do not have an Internet connection at home or never used the Internet (3.10%) because we are interested in students who are able to use the Internet autonomously. Moreover, we removed respondents with a missing value for one of the model variables (8.05%). However, we tested the robustness of our results using multiple imputations with chained equations (with 10 imputed data sets) and exclusion of observations with initial missing values on the dependent variables from the imputed data sets (Von Hippel, 2007)—our main results remained unchanged. We shall, therefore, only report the results from non imputed data. This analysis sample includes 154,539 respondents in 7,064 schools in 25 countries. Table 1 shows the distribution of the dependent variables. Further descriptive statistics concerning the sample can be found in the Appendix (Tables A1-A3).

OPEN IN VIEWER

Table 1.

Distribution of the Dependent Variables.

	Browse for fun	Upload content	Browse for school	Share for school
1. Never or hardly ever	3.18	48.60	11.86	40.47
2. Once or twice a month	4.65	17.57	32.33	23.65
3. Once or twice a week	14.87	13.61	37.63	20.46
4. Almost every day	29.11	10.15	12.94	10.54
5. Every day	48.18	10.07	5.23	4.88
Total	100.00	100.00	100.00	100.00

Source. Programme for International Student Assessment 2012.

Note. Figures in percent, n = 154,539.

Browsing activities are carried out more frequently than activities, which involve uploading or sharing materials. Nearly half of the respondents (over 48%) use the Internet every day to browse for fun, whereas less than 6% use the Internet daily to browse for school purposes.

Analytical Plan

The respondents are nested in countries and schools, and we take this structure into account using country fixed effects and clustering of the school variable. Standard errors are, therefore, adjusted for 7,064 school clusters. Due to the ordinal nature of our dependent variables, we estimate partial proportional odds models with the Stata add-on gologit2 (Williams, 2006). ¹ Models, which are fitted with this method, do not violate the parallel line assumption of the ordinal logistic regression model. The coefficients can vary across the different categories of the dependent variable, but do not necessarily do so. We use the autofit option to test for coefficients for which the parallel line assumption cannot be held. A partial proportional odds model with more than two categories “becomes equivalent to a series of binary logistic regressions where categories of the dependent variables are combined” (Williams, 2006, p. 59). In our case, we first contrast never/hardly ever with monthly or more often. In the second step, the categories never and once or twice a month are contrasted with the remaining categories weekly or more often. Then, never, monthly, once or twice a week are compared with almost every day and daily. Finally, we contrast the categories never, monthly, weekly, and almost every day with the category every day. We use such an elaborated method so that we do not need to shed information and take the structure of the data into account. An alternative method would have been to reduce the models to a single contrast, for example, “at least weekly” versus “monthly, hardly ever or never.” We have refrained from this because this contrast is included in our models, and additionally all the other possible contrasts can also be considered. For binary independent variables, a positive coefficient states that the probability of a respondent being in a higher category for the outcome variable than for the current one rises, a negative coefficient reveals that a respondent is more likely to be “in the current or a lower category” (Williams, 2006, p. 63). To report the coefficients of interest, we use the Stata add-on coefplot (Jann, 2014). Some “typical cases” are shown by calculating predicted probabilities. To handle the plausible values for reading literacy, the models are fitted with the Stata add-on PV (Macdonald, 2014).

Summary

Our secondary analysis of the PISA 2012 data showed that differences by family background can be found for educational Internet usage; this is true for browsing and sharing activities. Previous research conducted on educational Internet usage was mostly unable to identify such a relationship (Eamon, 2004; Gümüş, 2013). However, the spread of ICT was smaller in the early 2000s and adolescents from less privileged families were prone to have a lower rate of Internet connection in their homes (Eamon, 2004). For Europe, high penetration of home Internet access can be observed in the PISA 2012 data, so our results indicated that cultural capital background matters for the educational use of the Internet outside school once access is provided.

Our results reveal that consuming activities are more common among adolescents than uploading and sharing activities. This is in line with previous research conducted on how young students use social media (Lu, Hao, & Jing, 2016). Using the Internet for educational sharing practices is an understudied area of research. The findings of this study revealed that differences exist by parental cultural capital level. In contrast, no crucial effect for parents’ cultural capital could be identified for uploading entertainment content. The variable used for measuring this activity was based on the question of how often respondents used the Internet for “uploading your own created content for sharing (e.g., music, poetry, videos, computer programs).” The findings, therefore, not only refer to sharing, but also include the production of entertainment content. In a former study on U.S. college students, differences could be found for creating content, but not for sharing practices (Hargittai & Walejko, 2008). In a more recent study, Lu and colleagues analyzed consumption and sharing practices in and outside school for a sample of Hong Kong students. They found no significant effect for parental education on content sharing and creating, but did not differentiate between educational and other sharing activities. The findings of our study are more in line with Blank’s outcome, which showed that political content creation is related to elite status for a sample of British Internet users, but that other uploading activities are not (Blank, 2013). Thus, not all kinds of content creation and sharing practices come along with higher social status or cultural capital background. In the PISA data, a sharing divide among European students could only be identified for school-related materials and not for entertainment content.

The presented results confirm a digital divide regarding the educational use of the Internet outside school. This is in line with the theoretical expectations derived from Bourdieu’s theory of cultural and social reproduction. However, the effect sizes seem rather moderate. In addition, we also want to stress that socially stratified educational Internet use is only one way (out of many other ways) in which parental cultural capital is transmitted to children and unequal educational opportunities are created.

Limitations

The most serious limitation of the data used in this study concerns the measurement of the dependent variables via students’ self-reporting. Such self-reporting may be biased by memory errors and may also be prone to social desirability effects. Also, the Internet activities considered in the analysis are rather broad. The PISA questionnaire is restricted regarding Internet use for school purposes. For this reason, we only analyzed two indicators for each concept of interest (school/entertainment). Better models could be fitted with tracked instead of self-reported data. However, such measurements are not available in the PISA data. As we have no data about the question of whether and how teachers include ICT in homework tasks, we could not control for those effects. Another limitation is that we were unable to control for school track effects, because the respective variable in the international PISA data set seems not valid for all countries (e.g., all respondents in Iceland were categorized as ISCED level 2). Thus, we control for reading literacy to control for students’ level of school achievement.

Implications and Future Research

Previous findings on Internet usage focused mainly on information-orientated online activities. Results indicate that there are different usage patterns depending on demographic and socioeconomic factors. If the assumption is applied that some activities confer greater advantages for users, those with higher socioeconomic status tend to benefit more from their media usage (Van Deursen et al., 2015; Zillien & Hargittai, 2009). This assumption can also be applied to school-related Internet activities out-of-school. If frequent use of the Internet for scholastic purposes is associated with educational success, the results of this study demonstrate that students from families with high cultural capital tend to benefit more from their Internet usage in terms of later life outcomes. School-related Internet usage might, therefore, contribute to the reproduction of social inequality. However, a longitudinal study in Switzerland by Camerini, Schulz, and Jeannet (2018) revealed that more frequent informational and educational use of the Internet is not always related to better school outcomes. Nevertheless, other studies were able to identify a positive relation between educational ICT usage and academic performance (Robinson et al., 2018). Thus, more long-term studies are necessary to investigate whether use of the Internet for school-related activities is connected with higher educational achievement and further success.

Different usage patterns according to users’ background characteristics indicate that there is a usage gap like that described by Van Deursen and van Dijk (2014). Not only is usage for informational purposes generally structured by users’ backgrounds, but we demonstrate that school-related Internet usage is also connected with cultural capital background. This is in contrast to previous research (Eamon, 2004; Gümüş, 2013) and, therefore, needs attention in further studies. The use of tracked data in such studies is desirable.