Internet use that reproduces educational inequalities: Evidence from big data

Computers & Education 86 (2015) 212e223

Contents lists available at ScienceDirect

Computers & Education journal homepage: www.elsevier.com/locate/compedu

Internet use that reproduces educational inequalities: Evidence from big data Meilan Zhang* College of Education, University of Texas at El Paso, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 1 January 2015 Received in revised form 11 August 2015 Accepted 13 August 2015 Available online 17 August 2015

Although the Internet has become ubiquitous in students' lives in school and at home, little is known about whether the Internet is used to close or reproduce educational inequalities. Drawing upon Bourdieu's notion of capital, there are two kinds of Internet use: capitalenhancing versus entertainment. This study used two big data analytic tools to examine interest in and usage of two highly popular websites that primarily target children and adolescents: KhanAcademy.org and CartoonNetwork.com. The former represents a capitalenhancing use of the Internet, while the latter represents an Internet use for entertainment. Data analysis revealed that high sociodemographic status was positively correlated with interest in Khan Academy, while low sociodemographic status was positively correlated with interest in Cartoon Network. This study provided some evidence that existing educational inequalities may be reproduced through unequal Internet use. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Internet use Big data Digital divide Educational inequalities

The Internet has become ubiquitous in students' lives in school and at home (Census Bureau, 2011; National Center for Education Statistics, 2010). As of 2008, nearly all (98%) public schools in the United States had Internet access, as opposed to only 8% in 1995 (National Center for Education Statistics, 2010). Remarkable progress has been achieved in closing the gap in Internet access among different socioeconomic groups. As of 2008, there was virtually no difference in Internet access rates (97%e99%) across elementary and secondary, large and small, rural, urban, and suburban schools (National Center for Education Statistics, 2010). A recent study by the Pew Research Center showed that 95% of teenagers aged 12 to 17 had Internet access at home, including 89% of those whose household income was less than $30,000 (Madden, 2013). However, equal access does not ensure equal use. Not all types of Internet use are equally beneficial (van Deursen & van Dijk, 2014). Differences in Internet use among various socioeconomic groups may lead to a digital divide beyond Internet access. 1. Internet use and the digital divide Drawing upon Bourdieu's capital theory (Bourdieu, 1990), some types of Internet use may offer more opportunities than others to help individuals gain economic (financial resources), social (relationships and networks), and cultural (knowledge and dispositions) capital and move forward in their education, career, and social status. In line with this view, there are two kinds of online activities based on their potential for enhancing one's economic, social, and cultural capital (van Deursen & van Dijk, 2014). The first kind is capital-enhancing use, which involves using resources on the Internet to improve education, find jobs, advance career, and enhance physical and mental health. Another kind of Internet use mainly involves using the

* EDUC 801D, 500 West University Ave., El Paso, TX, 79968, USA. E-mail address: [email protected]. http://dx.doi.org/10.1016/j.compedu.2015.08.007 0360-1315/© 2015 Elsevier Ltd. All rights reserved.

M. Zhang / Computers & Education 86 (2015) 212e223

213

Internet for entertainment, such as playing games, surfing for fun, and gambling online. Such use is generally believed to have little potential for increasing economic, social, and cultural capital (van Deursen & van Dijk, 2014). Bourdieu's research on social inequalities suggests that individuals tend to develop practices and dispositions that accommodate their social positions and thereby reproduce existing advantages and disadvantages (Bourdieu, 1990). If Bourdieu's theory applies in Internet use, disadvantaged youth may be less likely to use the Internet to enhance their education, but more likely to use the Internet for entertainment. Prior research suggested three possible reasons that disadvantaged youth may be more prone to Internet use for entertainment. First, children in disadvantaged families may lack supervision and guidance in enriching use of the Internet because their parents have limited technological skills themselves (Hollingworth, Mansaray, Allen, & Rose, 2011). Second, children in disadvantaged families may lack opportunities to attend more intellectually stimulating extracurricular activities because of financial restrictions or time constraints due to parents' work schedules (Gershenson, 2013). Third, disadvantaged children may play games to escape from the drudgery of schoolwork, poor peer interaction, and other life stresses (Han et al., 2009; Jackson et al., 2005). Prior research provided some evidence for the tendency to use the Internet for entertainment by disadvantaged youth. Rideout, Foeh, and Roberts (2010) found that Black and Hispanic youth aged 8 to 18 spent three more hours per day on media use, including playing computer games, than their White peers. Gershenson (2013) found that children in poor families whose parents had low educational levels spent more time watching television in summer and less time reading than peers in affluent families whose parents were highly educated. In a similar vein, disadvantaged children may be more likely to use the Internet for entertainment. Jackson et al. (2005) found that African American children in poor families used computers primarily to play games. Hollingworth et al. (2011) found that middle-class parents were better able to guide their children's Internet use for educational purposes, while some working-class parents showed discomfort with their own lack of technological skills. Chang and Kim (2009) discovered a positive association between home computer access and science performance for English-speaking students, but a negative association for English language learners. Han et al. (2009) found that about half (52%) of children in their study with attention-deficit hyperactivity disorder were also diagnosed with Internet gaming addiction. In addition, disadvantaged students are also likely to attend high-need schools, as a substantial portion of public school funding in the United States is from property taxes. Such schools are more likely to be staffed with teachers who lack teaching experience and technical skills needed for effective technology use in classrooms (Chapman, Masters, & Pedulla, 2010). Prior research showed that technology use in schools serving disadvantaged students tends to be less sophisticated than in schools serving affluent students (Reinhart, Thomas, & Toriskie, 2011). Wood and Howley (2012) reported that teachers in affluent suburban schools had more technological training opportunities than did their peers in rural and urban schools. Similarly, Reinhart et al. (2011) found that teachers in schools with many students from poor families lacked instructional support of technological specialists. As a result, students with disadvantaged backgrounds may be less likely to use quality Internet resources for learning due to their teachers' lack of technological training and support. In summary, prior research indicated a possibility that the Internet may be used to reproduce, rather than close, educational inequalities. However, few studies have examined to what extent this hypothesis may hold true. In particular, little research has examined this issue by analyzing the actual usage of youth-oriented educational and entertainment websites. The paucity of research in this area is largely due to the lack of an effective and efficient method for tracking Internet use by millions of users. Two big data analytic tools that have rarely been used by educational researchers offer great potential for understanding current Internet use by youth. 2. Tracking internet use with two big data analytic tools In recent years, big data (data at a very large scale) has received significant attention in various fields, including education. Nevertheless, emerging discussions on big data in education are mainly focused on educational data mining and learning analytics (Bienkowski, Feng, & Means, 2012), which typically produces data controlled by local institutions. In the present study, two publicly available big data analytic tools are introduced: Google Trends and web analytics. Search engine use is one of the most common online activities, thanks to its ease of use and convenience in access. Internet users rely on search engines to meet their information needs. The majority (67%) of Internet searches are conducted on Google, which receives over 12 billion search queries each month (comScore, 2014a). Students are heavy Google users to the extent that they are called “the Google Generation” (Rowlands et al., 2008). According to Purcell et al. (2012), 94% of teachers stated that their students were most likely to use Google or other search engines to look for information for school projects, while only 18% reported their students used books for that purpose. Accordingly, analyzing search volumes and trends for search terms related to youth-oriented websites may provide important insight into Internet use by students. Google provides public access to its search data via a free tool called Google Trends (www.google.com/trends). 2.1. Google Trends Google Trends provides normalized search volumes for queries that Internet users have conducted on Google. The highest search volume is set at 100. Other volumes are divided by the highest volume and multiplied by 100. Zero represents either a lack of searches or insufficient search volumes. Google Trends provides monthly search volume data with the earliest date being January 2004. Google Trends also provides search volume data by country and region. To generate a

214

M. Zhang / Computers & Education 86 (2015) 212e223

search volume for a keyword in a specific region, the total search volume for that keyword is divided by the total Google search volume from the same region during a given period of time. This method takes into account the differences in population and Internet adoption rates in different geographic regions. Therefore, search volumes for keywords in larger and smaller regions are relative and comparable. Within the United States, search volume data is available for the 50 states. Google Trends also eliminates repetitive searches within a short period from the same user, so that this type of search does not impact the search trend. Google Trends has been increasingly used by researchers to study public interest and opinions in different topics, such as political election, automobile sales, consumer confidence, and the stock market (Choi & Varian, 2012). In addition, researchers in different fields have compared Google Trends data and data gathered from traditional methods, such as reports by the Centers for Disease Control and Prevention (McCarthy, 2010) and transaction volumes in the stock market (Preis, Reith, & Stanley, 2010). These studies have generally yielded positive evidence for the validity of Google Trends data. However, Google Trends has rarely been used in educational research for studying Internet use by students. Moreover, the recent controversy concerning the validity of Google Flu Trends (Lazer, Kennedy, King, & Vespignani, 2014), a different tool from Google Trends, suggests that more evidence is needed to support the use of Google Trends as a research tool. 2.2. Web analytics tools Web analytics tools report traffic data to a website. Such data typically includes the number of visors; the number of visits; time spent on a site per visit; number of pages viewed per visit; traffic sources; and search keywords that refer traffic to a site (Beasley, 2013). Several independent web analytics services, such as Compete (www.compete.com), Alexa (www.alexa.com), and SimilarWeb (www.similarweb.com), provide traffic data for millions of websites on the Internet. This study used Compete for the following reasons: 1) It is the only tool that reports the volumes of U.S. visitors and visits in actual numbers, rather than in estimated percentages relative to total Internet traffic, for the two websites examined in this study; 2) It focuses on Internet users in the United States; 3) It provides traffic data over a relatively long period of time (25 months from the retrieval date); and 4) It reportedly has the largest and most diverse sample size (2 million Internet users in the United States, accounting for approximately 1% of the total U.S. Internet population).1 Compete gathers clickstream traffic data, such as the order and timing of websites that users visit and searches they conduct, through Internet service providers. In addition, Compete gathers income data from participants in user registration and subsequent surveys. Based on data from its sample of two million U.S. Internet users, Compete provides estimated traffic data for the total U.S. Internet user population. To date, web analytics has mainly been used in Internet advertising and marketing research, but rarely in educational studies. While Google Trends reveals search interest, web analytics reports site usage. Together these tools can provide valuable information for educational research on Internet use. These tools were used in the present study to examine whether Internet use may close or enlarge educational inequalities. As discussed earlier, prior research has led to the hypothesis that high socioeconomic status is associated with capital-enhancing use of the Internet, while low socioeconomic and minority status are associated with entertainment use of the Internet. Therefore, unequal use of the Internet may reinforce existing educational advantage or disadvantage. To test this hypothesis, case studies were conducted on two representative websites, KhanAcademy.org and CartoonNetwork.com. The former represents educational use of the Internet that enhances cultural capital, while the latter represents Internet use for entertainment that is believed to have little potential for enhancing capital (van Deursen & van Dijk, 2014). These two websites were chosen for the following reasons: 1) Both websites primarily target children and adolescents; 2) Both are enormously popular, which reflects online interest and behavior of millions of Internet users, most of whom are presumably youth; 3) Both websites clearly lie at opposite ends of the Internet use continuum (educational versus entertainment), whereas other websites may fall somewhere in between (e.g., educational gaming sites); 4) Google search is one of the major traffic sources for both websites, enabling the use of Google Trends; and 5) The majority of search traffic for each website is generated by specific search terms that are unique to each site (khan academy and cartoon network, respectively), which further enables the use of Google Trends. 3. One-man university: Khan Academy Aiming to provide “a free, world-class education for anyone, anywhere” (Khan, 2012, p. 1), Khan Academy is a not-forprofit educational organization founded by Salman Khan in 2008.2 Khan Academy originated from YouTube videos that Salman Khan made to tutor his cousins remotely in 2006. Khan Academy currently offers more than 5000 educational videos, most of which were created by Khan himself with a tablet, a natural drawing application, and a screen recording program. The majority of these videos focus on pre-college mathematics, while others teach biology, chemistry, physics, finance, economics, and history. A typical Khan Academy video shows drawings and diagrams on an electronic blackboard along with Khan's explanation, akin to a one-on-one tutoring experience.

1 2

Information about Compete.com is retrieved from https://www.compete.com/about-compete/our-data/. Information about Khan Academy was retrieved from https://www.khanacademy.org/about and http://khanacademy.desk.com/.

M. Zhang / Computers & Education 86 (2015) 212e223

215

Khan Academy offers more than 100,000 exercise problems, and presents problems to students that are adaptive to their current levels. If students need help, each problem can be broken down into smaller steps or they can view a related video. Khan Academy keeps track of student progress and reports the statistics in a personalized learning dashboard. Detailed records of student performance are available by topic. Teachers, parents, and coaches can access the learning dashboard to track what students are doing and learning on Khan Academy, what concepts they have mastered, and where they struggle. Khan Academy offers badges and points as rewards to motivate students to solve exercise problems, a technique often used in games. Khan Academy has been featured as an exemplar of online education, personalized education, and flipped classroom models in high-impact media, such as the New York Times (Sengupta, 2011). Few educational resources have received as much media attention as Khan Academy. Khan Academy and Salman Khan had received multiple prestigious awards and recognitions, including the Tech Award in 2009,3 the Time 100 Most Influential People in the World in 2012,4 and the Heinz Award in 2014.5 However, despite the enormous fame and popularity, little empirical research exists regarding who is using Khan Academy and whether this “free, world-class” educational tool is equally used by individuals from different racial and socioeconomic backgrounds to improve their education.

4. Cartoon Network Launched in July 1998, CartoonNetwork.com is the official website of Cartoon Network, a popular television channel that offers entertaining cartoon animations primarily aimed at children and adolescents. As of August 2013, the Cartoon Network channel had reached 86% of households with television in the United States, making it one of the most popular children's television programs (Seidman, 2013). Accordingly, CartoonNetwork.com is a highly popular website in terms of visitors and time spent on the site. In July 2007, Nielsen reported that visitors spent an average of 77 min per person on CartoonNetwork. com, ranking it 26th among all U.S. websites in terms of the total time spent on the site (Ball, 2007). On CartoonNetwork.com, visitors can view full episodes and video clips of Cartoon Network shows, such as Adventure Time, Ben 10, Ninjago, and Pokemon. In addition, visitors can play “all types of games online including sports, action, arcade, and adventure games for kids.”6 The website provides over 350 free online games featuring popular characters on the Cartoon Network shows, such as Tom and Jerry, and Scooby-Doo. Despite the fact that children and adolescents are the primary audience of CartoonNetwork.com and that a tremendous amount of time is spent on this website, few educational studies have examined how this website is used, who is more likely to visit, and how its usage relates to academic performance. A literature search on major databases including the Education Resources Information Center, Web of Science, Academic Search Complete, PsycINFO, and Google Scholar found no empirical studies about children's use of this website. The purpose of this study was as follows: (a) to understand the extent to which Internet users are interested in Khan Academy, a representative case of educational use of the Internet, and Cartoon Network, a representative case of entertainment use of the Internet; (b) to verify whether there is a good match between Internet search and site traffic data from two independent sources, Google Trends and Compete; (c) to understand the relationship between search interest in the two websites and academic performance; and (d) to understand whether socioeconomic and minority status can predict interest in the two websites. Accordingly, this study examined the following research questions: 1) How many Internet users are interested in a popular educational website and a popular entertainment website, and to what extent? 2) To what extent do the search trends for the two websites correspond to the site visit trends? 3) To what extent is interest in the two websites associated with academic performance at the national level? And 4) To what extent can sociodemographic characteristics predict interest in the two websites?

5. Method 5.1. Data sources 5.1.1. Google Trends Based on Compete data showing that khan academy and cartoon network were the most popular search keywords that generated traffic to KhanAcademy.org and CartoonNetwork.com, respectively, a search for each of these terms was performed using Google. The search results confirmed that KhanAcademy.org and CartoonNetwork.com were listed in first place out of millions of search results for each of their respective search terms. These findings demonstrate that khan academy and cartoon network are the most representative terms reflecting Internet users' interest in KhanAcademy.org and CartoonNetwork.com, respectively.

3 4 5 6

Information Information Information Information

was was was was

retrieved retrieved retrieved retrieved

from from from from

http://thetechawards.thetech.org/laureate/archive. http://content.time.com/time/specials/packages/article/0,28804,2111975_2111976_2111942,00.html. http://www.heinzawards.net/recipients. http://www.cartoonnetwork.com/games/index.html.

216

M. Zhang / Computers & Education 86 (2015) 212e223

Google Trends data showed that search interest in khan academy emerged around January 2010, while search interest in cartoon network emerged prior to January 2004, the earliest date for which Google Trends data is available. Accordingly, search volume data for the 50 U.S. states was obtained for khan academy for the period of January 2010 to May 2014 (53 months). Also, search volume data for the 50 states for cartoon network was obtained from Google Trends for the period of January 2004 to May 2014 (125 months). 5.1.2. Web analytics The following monthly data for KhanAcademy.org and CartoonNetwork.com were obtained from Compete for the period of May 2012 to May 2014 (25 months): number of visitors and visits per month to each website, time on site and number of page views per visit, traffic sources, and keywords that Internet users searched for finding the website. Finally, this study used Compete data to determine the annual income of users for each website, based on four categories: $0e30 K, $30e60 K, $60e100 K, and more than $100 K. 5.1.3. Academic performance This study used state-level academic performance data from the National Assessment of Educational Progress (NAEP), which is the only available educational assessment in the United States for state comparison. NAEP evaluates students in mathematics and reading at Grades 4 and 8 every two years and in science in some years. NAEP reports four measures of academic performance for each subject by state: the average score and the percentage of students attaining basic, proficient, and advanced levels. The present study used the four measures for mathematics and reading in 2009, 2011, and 2013 for both grades. For science, this study used the four measures for fourth grade students in 2009 and eighth grade students in 2009 and 2011.7 5.1.4. Sociodemographic characteristics Prior research suggests that sociodemographic characteristics, including minority status (Jackson et al., 2006), income (Wainer et al., 2008), language status (Kim & Chang, 2010), disability (Han et al., 2009), and parental education (Hollingworth et al., 2011), play an important role in mediating how children use the Internet. Therefore, the following data for the 2010e2011 school year were retrieved from the U.S. Department of Education, National Center for Education Statistics for the 50 states: the percentages of White, Black, and Hispanic students in public elementary and secondary schools; the percentage of students in public schools eligible for lunch subsidies; the percentage of children aged 3 to 21 served under the Individuals with Disabilities Education Act (IDEA); and the percentage of students participating in programs for English Language Learners (ELL).8 In addition, the 3-year-average percentage of adults 25 years or older with a bachelor's degree or higher was retrieved by state from the U.S. Department of Commerce, Census Bureau for the years 2008 to 2010.9 The Census Bureau used a 3-year average to increase the sample size and reduce sampling errors. These sociodemographic characteristics were used as independent variables in regression analysis to predict interest in Khan Academy and Cartoon Network at the state level. 5.2. Data analysis First, descriptive data from Compete was used to illustrate how many and to what extent Internet users were using the two websites. In addition, traffic sources and search terms for the two websites were described using the web analytics data from Compete. Second, to determine whether Google Trends data matched Compete data, search volumes for Khan Academy and Cartoon Network from Google Trends for the period of May 2012 to May 2014 were correlated with site visit volumes from Compete over the same time period for the two websites. To illustrate the search trend and site visit trend data in a graph, site visit volumes from Compete were normalized to a scale of 0e100, similar to search volumes on Google Trends. The highest visit volume was normalized to100, and other monthly visit volumes were divided by the highest volume and multiplied by 100. Third, Pearson correlations were conducted between the search volumes for khan academy and cartoon network by state and NAEP academic performance data in 2009, 2011, and 2013 for Grades 4 and 8. To verify the linear relationships between these variables, scatter plots were produced for each correlation. The scatter plots showed that all relationships were approximately linear. Thus it is appropriate to use the Pearson correlational analysis for these variables. In addition, linear regression was conducted to determine whether sociodemographic characteristics can predict search volumes for khan academy and cartoon network at the state level. In the regression model, the dependent variable was search volumes for khan academy or cartoon network. The independent variables were the percentages of White, Black, and Hispanic students; students eligible for free or reduced-price lunch; students served under IDEA; ELL students, and adults with at least a bachelor's degree. For the regression models, the DurbineWatson test was used to verify independence of observations. The DurbineWatson statistic for the Khan Academy analysis was 1.834 and for the Cartoon Network analysis was 2.226. The DurbineWatson

7 8 9

Science was not assessed at grade 4 in 2011 and 2013 and at grade 8 in 2013. Data was retrieved from the Digest of Education Statistics in 2011 and 2012 from http://nces.ed.gov/programs/digest. Data was retrieved from http://nces.ed.gov/programs/digest/d12/tables/dt12_013.asp.

M. Zhang / Computers & Education 86 (2015) 212e223

217

statistic can range from 0 to 4. A value of approximately 2 indicates that there is no correlation between residuals. Both values were very close to 2, so it can be accepted that there was independence of errors (residuals) in both regression models. In addition, the multicollinearity test was conducted to examine whether there were two or more independent variables that were highly correlated with each other in the regression model. If the Tolerance value is less than 0.1, or a VIF is greater than 10, there may be a collinearity problem. The multicollinearity test found that all the Tolerance values were greater than 0.1 and VIFs were less than 10, so it is safe to conclude that there was no problem with collinearity. Finally, t-tests were conducted to examine the differences between the percentages of KhanAcademy.org and CartoonNetwork.com users per month at each of the four income levels ($0e30 K, $30e60 K, $60e100 K, and more than $100 K). The assumption for t-test is that data should be normally distributed. The ShapiroeWilk test was conducted to examine the normality of the data. None of the results in the ShapiroeWilk test were significant, suggesting that the data was normally distributed. 6. Results 6.1. Website usage 6.1.1. Khan Academy Table 1 presents an overview of U.S. traffic to the Khan Academy and Cartoon Network websites during a one-year period from January to December 2013 using data from Compete. Approximately 1.4 million Internet users in the United States made 2.8 million visits to the Khan Academy website per month. An average user spent 12 min on the website and viewed 6.7 pages per visit. Collectively, U.S. Internet users spent an average of 569,001 h on this website per month. Data from Compete showed that direct traffic to the Khan Academy website constituted about 22% of the total traffic in April 2014. About 41% of traffic came from search engines including Google.com (33.7%), Yahoo.com (4.3%), and Bing.com (3.7%). Between March 2, 2014 and May 31, 2014, 3146 search terms sent traffic to this website. The term khan academy accounted for 39.9% of traffic from searches. Other five most popular search terms were khan (2.0%), kahn academy (1.7%), khanacademy.org (1.3%), khanacademy (1.1%) and www.khanacademy.org (1.1%). Clearly, khan academy was the most representative term for search interest in this website. 6.1.2. Cartoon Network Approximately 5.9 million Internet users in the United States made 17 million visits to the Cartoon Network website per month. An average user spent 18 min on the website and viewed 8.4 pages per visit. Collectively, U.S. Internet users spent an average of 5,257,363 h on this website per month. Direct traffic to the Cartoon Network website constituted about 27% of the total traffic in April 2014. About 42% of traffic came from search engines including Google.com (18.6%), Bing.com (8.8%), Yahoo. com (7.4%), Ask.com (1.8%), and MSN.com (0.7%). Between March 2, 2014 and May 31, 2014, 1933 search terms sent traffic to this website. The term cartoon network accounted for 41.2% of traffic from searches. Other five most popular search terms were cartoon network games (15.5%), cartoonnetwork.com (2.8%), cn (2.4%), cartoonetwork (1.7%), and ben 10 games (1.7%). Clearly, cartoon network was the most representative term for search interest in this website. 6.2. Correlation between search trends and site visit trends Overall, the search trends for khan academy and cartoon network from Google Trends closely matched visit trends from Compete for each website, respectively. The Pearson correlational coefficient for the monthly search volumes for khan

Table 1 Monthly US traffic statistics for KhanAcademy.org and CartoonNetwork.com in 2013. KhanAcademy.org

Jan 13 Feb 13 Mar 13 Apr 13 May 13 Jun 13 Jul 13 Aug 13 Sep 13 Oct 13 Nov 13 Dec 13 Average

CartoonNetwork.com

Unique visitors

Visits

Average stay (mm:ss)

Page view

Hours on site per month

Unique visitors

Visits

Average stay (mm:ss)

Page view

Hours on site per month

1,318,493 1,260,540 1,071,354 978,039 820,445 710,804 640,508 926,699 2,140,307 2,341,200 2,104,136 1,961,293 1,356,152

2,582,916 2,626,941 2,356,352 1,814,939 1,499,094 1,370,324 1,527,327 2,028,947 4,329,755 4,776,072 4,278,565 3,996,970 2,765,684

12:14 11:38 12:01 11:40 11:51 11:03 11:13 10:07 11:23 12:47 14:01 14:18 12:01

6.8 6.2 6.0 6.2 6.0 6.1 5.9 5.9 6.0 7.3 8.8 9.1 6.7

526,628 509,335 471,925 352,905 296,071 252,368 285,525 342,103 821,451 1,017,569 999,520 952,611 569,001

6,862,269 6,804,500 6,867,385 6,702,450 6,491,698 6,246,488 6,548,799 5,854,228 4,792,069 3,863,369 5,271,778 4,480,251 5,898,774

20,701,976 19,463,194 20,604,186 20,384,243 19,795,686 19,417,674 20,153,904 17,285,169 12,247,465 9,856,221 13,785,641 11,170,102 17,072,122

19:39 18:15 18:19 19:13 19:11 19:45 19:04 18:55 17:41 16:45 15:31 16:29 18:14

9.8 9.2 8.8 8.6 8.4 8.7 8.2 8.4 7.7 7.7 7.3 7.9 8.4

6,779,897 5,920,055 6,290,000 6,528,620 6,329,121 6,391,651 6,404,463 5,449,630 3,609,600 2,751,528 3,565,120 3,068,675 5,257,363

218

M. Zhang / Computers & Education 86 (2015) 212e223

100 80 60 40 20

Search volume

Apr 14

May 14

Mar 14

Jan 14

Feb 14

Dec 13

Oct 13

Nov 13

Sep 13

Jul 13

Aug 13

Jun 13

May 13

Apr 13

Mar 13

Jan 13

Feb 13

Dec 12

Oct 12

Nov 12

Sep 12

Jul 12

Aug 12

May 12

Jun 12

0

Visit volume

Fig. 1. Normalized monthly search volumes for Khan Academy and site visit volumes for KhanAcademy.org.

100 80 60 40 20

Search volume

May 14

Apr 14

Feb 14

Mar 14

Jan 14

Dec 13

Nov 13

Oct 13

Sep 13

Aug 13

Jul 13

Jun 13

May 13

Apr 13

Mar 13

Feb 13

Jan 13

Dec 12

Nov 12

Oct 12

Sep 12

Aug 12

Jul 12

Jun 12

May 12

0

Visit volume

Fig. 2. Normalized monthly search volumes for Cartoon Network and site visit volumes for CartoonNetwork.com.

academy from May 2012 to May 2014 and the visit volumes for the KhanAcademy.org website for the same period was 0.934 (n ¼ 25, p < .001). The search trend and site visit trend well matched one another, as shown in Fig. 1. The Pearson correlational coefficient for the monthly search volumes for cartoon network from May 2012 to May 2014 and the visit volumes for the CartoonNetwork.com website for the same period was 0.918 (n ¼ 25, p < .001), respectively. Fig. 2 shows the two trends.

6.3. Correlation with academic performance 6.3.1. Khan Academy As shown in Table 2, positive correlations were found between the search volumes for khan academy and all NAEP performance measures in all years (2009, 2011, and 2013) for Grades 4 and 8, although not all correlations were statistically Table 2 Correlation of the state-level search index for Khan Academy and Grades 4 and 8 mathematics, reading, and science performance in NAEP 2009, 2011, and 2013. Grade 4

2013

2011

2009

Average score % of students at % of students at % of students at Average score % of students at % of students at % of students at Average score % of students at % of students at % of students at

or above basic or above proficient advanced or above basic or above proficient advanced or above basic or above proficient advanced

Grade 8

Math

Reading

0.195 0.120 0.237 0.245 0.216 0.139 0.267 0.265 0.251 0.182 0.306* 0.305*

0.080 0.112 0.151 0.173 0.047 0.081 0.111 0.104 0.071 0.086 0.158 0.160

Science

Math

Reading

Science

0.181 0.216 0.180 0.083

0.303* 0.292* 0.316* 0.292* 0.295* 0.265 0.329* 0.371** 0.272 0.252 0.299* 0.314*

0.321* 0.290* 0.348* 0.283* 0.293* 0.258 0.327* 0.349* 0.204 0.172 0.245 0.241

0.314* 0.331* 0.287* 0.133 0.292* 0.296* 0.284* 0.169

*p < .05, **p < .01. The bold signifies that the probability of rejecting the null hypothesis when it is true is 5% or smaller.

M. Zhang / Computers & Education 86 (2015) 212e223

219

Table 3 Correlation of the state-level search index for Cartoon Network and Grades 4 and 8 mathematics, reading, and science performance in NAEP 2009, 2011, and 2013. Grade 4

2013

2011

2009

Average score % of students at % of students at % of students at Average score % of students at % of students at % of students at Average score % of students at % of students at % of students at

or above basic or above proficient advanced or above basic or above proficient advanced or above basic or above proficient advanced

Grade 8

Math

Reading

¡0.412** ¡0.392** ¡0.440** ¡0.340* ¡0.375** ¡0.372** ¡0.409** 0.233 ¡0.432** ¡0.434** ¡0.450** ¡0.288*

0.207 ¡0.283* 0.248 0.126 0.180 0.256 0.236 0.106 ¡0.280* ¡0.347* ¡0.327* 0.200

Science

Math

Reading

Science

¡0.503*** ¡0.550*** ¡0.478*** 0.050

¡0.450** ¡0.502*** ¡0.444** ¡0.304* ¡0.473*** ¡0.516*** ¡0.459*** ¡0.339* ¡0.496*** ¡0.532*** ¡0.485*** ¡0.329*

¡0.417** ¡0.466*** ¡0.425** 0.211 ¡0.471*** ¡0.516*** ¡0.480*** 0.235 ¡0.478*** ¡0.522*** ¡0.449** 0.155

¡0.604*** ¡0.633*** ¡0.562*** 0.138 ¡0.613*** ¡0.629*** ¡0.583*** ¡0.286*

*p < .05, **p < .01, ***p < .001. The bold signifies that the probability of rejecting the null hypothesis when it is true is 5% or smaller.

significant. For fourth graders, only correlations between search volumes and the percentage of students attaining proficient or advanced levels in mathematics in 2009 were statistically significant (p < .05). For eighth graders, correlations between search volumes and reading or mathematics performance were all statistically significant (p < .05) for 2013. For 2011, correlations with reading and mathematics performance of eighth grade students were all significant, except for the percentage of students attaining the basic level. For 2009, correlations between search volumes and the percentage of students at proficient or advanced levels in mathematics were significant. For Grade 8 science, all correlations with search volumes were significant except for the percentage of students at the advanced level. Overall, Internet users in states with more highperforming eighth grade students were more likely to search for khan academy. 6.3.2. Cartoon Network As shown in Table 3, negative correlations were found between the search volumes for cartoon network and all NAEP performance measures in all years for Grades 4 and 8, most of which were statistically significant. For fourth graders, correlations between search volumes and mathematics performance measures were all significant for all three years except for the percentage of students at the advanced level in 2011. Correlations with reading performance measures in 2009 were all significant except for the percentage of students at the advanced level. For Grade 4 reading in 2013, correlation with the percentage of students at or above the basic level was significant. For eighth graders, correlations with mathematics performance measures were significant for all three years. Correlations with reading performance measures were all significant except for the percentage of students at the advanced level for each year. Correlations with science performance measures were all significant, except for the percentage of students at the advanced level in 2011. Overall, Internet users in states with more high-performing students at Grades 4 and 8 were less likely to search for cartoon network. 6.4. Predicting search interest by sociodemographic characteristics 6.4.1. Khan Academy The regression model for search volumes and sociodemographic variables was significant (R2 ¼ .394, p < .01), suggesting that the sociodemographic variables included in the model can explain 39% of the variance in the search volumes for khan academy from January 2010 to May 2014 among the 50 states (see Table 4). Specifically, the percentage of adults with at least a bachelor's degree was a positive predictor of search interest in khan academy (B ¼ 1.400, p < .05). On the other hand, the Table 4 Predicting search interest in Khan Academy by sociodemographic variables.

(Constant) % of adults with a bachelor's degree or higher % of White students % of Black students % of Hispanic students % of ELL students % of IDEA students % of students eligible for lunch subsidies

B

SE

3.752 1.400* .256 ¡.608* .157 .405 ¡2.151* .525

39.347 .646 .199 .242 .242 .745 1.033 .386

*p < .05. The bold signifies that the probability of rejecting the null hypothesis when it is true is 5% or smaller.

b

t

Sig.

.450 .314 .508 .145 .137 .296 .340

.095 2.167 1.286 2.513 .649 .543 2.082 1.362

.924 .036 .206 .016 .520 .590 .044 .181

220

M. Zhang / Computers & Education 86 (2015) 212e223

Table 5 Predicting search interest in Cartoon Network by sociodemographic variables.

(Constant) % of adults with a bachelor's degree or higher % of White students % of Black students % of Hispanic students % of ELL students % of IDEA students % of students eligible for lunch subsidies

B

SE

96.129 .943 .174 .924*** .418* .282 .203 .449

28.878 .474 .146 .177 .178 .547 .758 .283

b

t

Sig.

.298 .210 .760 .381 .094 .027 .286

3.329 1.988 1.191 5.206 2.351 .516 .267 1.587

.002 .053 .240 .000 .024 .608 .791 .120

*p < .05, ***p < .001. The bold signifies that the probability of rejecting the null hypothesis when it is true is 5% or smaller.

percentage of Black students in public schools was a negative predictor of search interest in khan academy (B ¼ .608, p < .05), as was the percentage of students served under IDEA (B ¼ 2.151, p < .05). In other words, Internet users in states with fewer college graduates, more Black students, and more students served under IDEA were less likely to search for the Khan Academy site. The other variables, which are the percentages of White and Hispanic students, ELL students, and students eligible for free or discounted lunch, were not significant predictors of search interest in khan academy. 6.4.2. Cartoon Network The regression model for search volumes and sociodemographic variables was significant (R2 ¼ .684, p < .001), suggesting that the sociodemographic variables included in the model can explain 68% of the variance in the search volumes for cartoon network from January 2004 to May 2014 among the 50 states (see Table 5). Specifically, the percentage of Black students in public schools was a positive predictor of search interest in cartoon network (B ¼ .924, p < .001), as was the percentage of Hispanic students (B ¼ .418, p < .05). In addition, the percentage of adults with at least a bachelor's degree was a marginally significant negative predictor of search interest in cartoon network (B ¼ .943, p < .06). In other words, Internet users in states with more Black students, more Hispanic students, and fewer college graduates were more likely to search for the Cartoon Network site. The other variables included in the model were not significant predictors of search interest in cartoon network. Fig. 3 presents the differences between the percentages of KhanAcademy.org and CartoonNetwork.com users by income using data from Compete. The most striking differences were from the highest income group (Khan Academy: M ¼ 19.97, SD ¼ 2.63; Cartoon Network: M ¼ 13.85, SD ¼ 1.18) and lowest income group (Khan Academy: M ¼ 29.08, SD ¼ 2.67; Cartoon Network: M ¼ 37.00, SD ¼ 1.85). That is, 20% of Khan Academy users were from the highest income group ($100 Kþ), which is 6% more than the percentage of Cartoon Network users at that income level (t(24) ¼ 14.09, p < .001). In contrast, 29% of Khan Academy users were from the lowest income group ($0e30 K), which is 8% less than the percentage of Cartoon Network users at that income level (t(24) ¼ 23.81, p < .001). In addition, there were significantly more Khan Academy users than Cartoon Network users in the second highest income group ($60e100 K) (Khan Academy: M ¼ 25.28, SD ¼ 1.45; Cartoon Network: M ¼ 22.91, SD ¼ 1.23; t(24) ¼ 7.21, p < .001). The difference in the second lowest income group ($30e60 K) was not statistically significant (Khan Academy: M ¼ 25.67, SD ¼ 2.00; Cartoon Network: M ¼ 26.25, SD ¼ 1.66; t(24) ¼ 1.61, p ¼ .121). 7. Discussion This study provided some evidence that the two focused websites may be used to reproduce existing educational inequalities. Overall, high academic performance and high sociodemographic status were positively correlated with interest in 40%

30%

37% 29% 26% 26%

25%

23% 20%

20% 14% 10%

0% 0-30k

30-60k

KhanAcademy.org

60-100k

100k+

CartoonNetwork.com

Fig. 3. Average percentages of Khan Academy and Cartoon Network users by income per month (May 2012eMay 2014).

M. Zhang / Computers & Education 86 (2015) 212e223

221

Khan Academy, an educational use of the Internet. Also, the correlations were stronger for eighth grade students than fourth grade students, perhaps due to the fact that Khan Academy was more often used by middle school students than elementary students, particularly in its early years (Khan, 2012). On the other hand, low academic performance and low sociodemographic status were positively correlated with interest in Cartoon Network, an entertainment use of the Internet. The scale of the unequal Internet use identified in this study is enormous when considering the number of visitors to the two websites in relation to the total student population. There are approximately 35 million K-8 students in public schools in the United States (Keaton, 2012), among which about 32 million are Internet users (Census Bureau, 2011). If not considering the overlap between users of the two websites, Khan Academy and Cartoon Network reach 7.3 million users in the United States per month, equivalent to 23% of the total K-8 Internet population. This study is among the first to report the unequal Internet use on a scale that involves millions of people, most of whom are presumably youth. The percentage of highly educated adults and the percentage of Black students were important indicators of interest in Khan Academy and Cartoon Network, but in opposite directions. Internet users in states with more highly educated adults and fewer Black students were more likely to search for Khan Academy and less likely to search for Cartoon Network. In addition, the percentage of Hispanic students was a positive indicator of search interest in Cartoon Network, and Internet users in states with more students served under IDEA were less interested in Khan Academy. Compete data also showed that high-income Internet users were more likely to use Khan Academy, while low-income Internet users were more likely to use Cartoon Network. Prior research suggested that excessive online game play is associated with attention deficit and learning difficulty, aggressive behaviors, and sleep deprivation (Kuss & Griffiths, 2012). Moreover, fast-paced cartoons present events rapidly, which captures the sensory attention of viewers and may lead to rapid depletion of cognitive resources. Lillard and Peterson (2011) found that watching 9 min of a fast-paced television cartoon had an immediate detrimental impact on children's executive function, such as self-regulation and working memory. The authors warned that watching a full episode of fastpaced cartoons could be even more detrimental. It is important to note that the findings on the use of Khan Academy in this study are consistent with previous findings on the use of PhET science simulations using a similar method (Zhang, 2014). PhET science simulations were developed based on sound design principles by researchers with high scientific credentials (Wieman, Adams, & Perkins, 2008). However, similar to Khan Academy, this valuable educational resource is not equally used by individuals from different racial and socioeconomic backgrounds. Internet users in states with more high-performing students, more White students, fewer students qualified for lunch subsidies, and fewer Black students were more likely to search for PhET science simulations. The results from the present study and the previous study on PhET simulations suggest that providing Internet access to racially and socioeconomically unequal groups may not level the playing field (Goolsbee & Guryan, 2006; National Center for Education Statistics, 2010). Instead, given equal Internet access, different ways of using the Internet may widen, rather than narrow, the achievement gaps between White and Black students, White and Hispanic students, and students with high and low socioeconomic status (Sirin, 2005). Thus, more attention should be paid to closing the usage gap, which is less obvious than the access gap. It is interesting to note that the two opposite associations, advantaged status and educational use of the Internet, and disadvantaged status and entertainment use, were not equal in magnitude and impact. First, far more users made more visits to and spent much longer time on Cartoon Network than Khan Academy. In 2013, the number of visitors, the number of visits, and the amount of time on the Cartoon Network site was 4.3, 6.2, and 9.2 times greater than that for Khan Academy, respectively. Second, as shown in Tables 2 and 3, correlations with NAEP performance indicators were stronger for Cartoon Network than for Khan Academy. In addition, the sociodemographic factors explained only 39% of the variance in search volumes for Khan Academy across states, compared to 68% of the variance in search volumes for Cartoon Network. These results suggest that the unequal Internet use may have a stronger negative impact on disadvantaged youth. That is, an initial disadvantage may beget bigger disadvantages derived from the Internet use than the advantage accrued by an initial advantage. It should be noted that the associations between interest in Khan Academy and Cartoon Network and academic performance may be mediated by socioeconomic status, which itself is an important predictor of academic achievement. Sirin (2005) reviewed 207 correlations between socioeconomic status and academic achievement in 58 journal articles published from 1990 to 2000 and found the average correlation was .299. Therefore, it may be possible that although the two websites were used by youth with different socioeconomic backgrounds, the usage did not have impact on academic performance. In addition, it is possible that affluent Internet users may be more likely to use paid online services such as Netflix for entertainment, which low-income families cannot afford, who thus turn to free entertainment resources such as Cartoon Network. Netflix has a large collection of children-oriented video content, and a substantial portion of their subscribers are families with children (comScore, 2014b). Although the monthly subscription fee for Netflix is relatively cheap, financially challenged families may still prefer free entertainment services. Several limitations of this study should be noted. First, the findings of this study are limited to the two websites examined. It is unclear to what extent these findings can be generalized to other educational and entertainment websites. Second, it was assumed that the audiences of Khan Academy and Cartoon Network are primarily children and adolescents based on the content of these two websites and media reports (Seidman, 2013; Sengupta, 2011). However, the aggregated data from Google

222

M. Zhang / Computers & Education 86 (2015) 212e223

Trends and web analytics did not provide evidence for the identity of the users; possibly, many adults use the two websites. Third, there is a general lack of transparency regarding how the companies obtain and process data, which makes it difficult for researchers to verify the data. Despite these limitations, Google Trends and web analytics remain valuable tools that complement existing methods for studying Internet use (Beasley, 2013; Choi & Varian, 2012). These third-party tools provide free or low-cost access to realtime, large-scale data for understanding online interest and behaviors, which would otherwise be difficult to obtain using traditional survey methods. In addition, this study found a close match between Google Trends and web analytics data, lending additional support for using these tools to investigate the complex and dynamic Internet use in educational research. 8. Conclusion The findings of this study have important implications for policy makers and educational researchers. The priority in the national agenda for the Internet and education in the United States still focuses on Internet access. The Obama's administration has recently initiated a new program called ConnectED to provide Internet access through high-speed broadband and wireless to 99% of students in the United States within five years (The White House, 2013). However, without understanding and addressing the hidden digital divide in Internet use, the multibillion-dollar initiative may not achieve its goal to close the achievement gaps and educational inequalities, if not enlarging. Therefore, policy makers should pay close attention to the large-scale unequal Internet use found in this study and invest resources to address the usage gap. In addition, this study offers an innovative approach for educational researchers to studying the role of the Internet in academic performance. Prior research has reported conflicting results regarding the relationship between Internet use and students' academic performance, including positive relationship (e.g., Bebell & Kay, 2010), no relationship (e.g., Goolsbee & Guryan, 2006), and negative relationship (e.g., Wainer, et al., 2008). One problem is that the definition of Internet use in these studies tended to be vague and general, without much detail on how students used the Internet. These mixed results may be explained by the different types of Internet use. More use of high-quality educational websites may have a positive impact on academic performance, while the opposite may be true for entertainment websites. Thus, educational researchers can better understand the relationship between Internet use and academic performance by examining the nature of the websites that are widely used. Future research should examine whether the findings of this study apply to other youth-oriented websites that focus on education or entertainment. In addition, empirical studies should be conducted to examine whether and why youth with different socioeconomic backgrounds tend to use the Internet in a way that reproduces existing inequalities. What are the underlying mechanisms behind the phenomenon revealed in this study? More importantly, what intervention programs can be developed to encourage disadvantaged youth to use the Internet more productively? It is possible that educational websites are more likely to be used in schools and entertainment websites are more likely to be used at home. If so, interventions should be created for both schools and families to promote productive Internet use by disadvantaged youth. This study also has implications for research on Internet use in other countries. According to Alexa, in April 2015, 41% of Khan Academy users and 50% of Cartoon Network users came from countries other than the United States (Alexa Internet, 2015a; 2015b). Therefore, it is important to examine whether findings in this study can be generalized to Internet users in other countries. Google Trends provides data for searches in languages other than English (Google, 2015). Educational researchers in non-English speaking countries can choose their native language to examine search trends for youth-related websites in their country. In summary, the Internet has become an integral part of youth's personal and academic lives. Thus it is important to understand whether Internet use may alleviate or contribute to existing educational inequalities. This study introduced two innovative tools, Google Trends and Web analytics, for researchers to monitor large-scale Internet use for education.

References Alexa Internet. (2015a). Cartoonnetwork.com site overview. Retrieved May 18, 2015 from http://www.alexa.com/siteinfo/cartoonnetwork.com. Alexa Internet. (2015b). Khanacademy.org site overivew. Retrieved May 18, 2015 from http://www.alexa.com/siteinfo/khanacademy.org. Ball, R. (2007, August 23). CartoonNetwork.com hits visitor high. Retrieved from http://www.animationmagazine.net/internet/cartoonnetwork-com-hitsvisitor-high/. Beasley, M. (2013). Practical web analytics for user experience: How analytics can help you understand your users. Waltham, MA: Elsevier. Bebell, D., & Kay, R. (2010). One to one computing: a summary of the quantitative results from the Berkshire wireless learning initiative. The Journal of Technology, Learning and Assessment, 9(2). Bienkowski, M., Feng, M., & Means, B. (2012). Enhancing teaching and learning through educational data mining and learning analytics. Washington, D.C.: U.S. Department of Education, Office of Educational Technology. Bourdieu, P. (1990). The logic of practice. Stanford, CA: Stanford University Press. Census Bureau. (2011). Current population survey (CPS), table 16. Number and percentage of persons 3 years old and over using the internet and percentage distribution by means of internet access from home and main reason for not having high-speed access, by selected characteristics of students and other users: 2010. U.S. Department of Commerce. Chang, M., & Kim, S. (2009). Computer access and computer use for science performance of racial and linguistic minority students. Journal of Educational Computing Research, 40(4), 469e501. Chapman, L., Masters, J., & Pedulla, J. (2010). Do digital divisions still persist in schools? Access to technology and technical skills of teachers in high needs schools in the United States of America. Journal of Education for Teaching, 36(2), 239e249. Choi, H., & Varian, H. (2012). Predicting the present with Google Trends. Economic Record, 88(s1), 2e9.

M. Zhang / Computers & Education 86 (2015) 212e223

223

comScore. (2014a, September 17). comScore releases August 2014 U.S. search engine rankings. Retrieved from http://www.comscore.com/Insights/MarketRankings/comScore-Releases-August-2014-US-Search-Engine-Rankings. comScore. (2014b, October 14). The U.S. total video report. Retrieved from http://www.comscore.com/Insights/Presentations-and-Whitepapers/2014/TheUS-Total-Video-Report. Gershenson, S. (2013). Do summer time-use gaps vary by socioeconomic status? American Educational Research Journal, 50(6), 1219e1248. Google. (2015). Trends help. Retrieved August 8, 2015 from https://support.google.com/trends. Goolsbee, A., & Guryan, J. (2006). The impact of internet subsidies in public schools. The Review of Economics and Statistics, 88(2), 336e347. Han, D. H., Lee, Y. S., Na, C., Ahn, J. Y., Chung, U. S., Daniels, M. A., et al. (2009). The effect of methylphenidate on internet video game play in children with attention-deficit/hyperactivity disorder. Comprehensive Psychiatry, 50(3), 251e256. Hollingworth, S., Mansaray, A., Allen, K., & Rose, A. (2011). Parents' perspectives on technology and children's learning in the home: social class and the role of the habitus. Journal of Computer Assisted Learning, 27(4), 347e360. Jackson, L. A., Samona, R., Moomaw, J., Ramsay, L., Murray, C., Smith, A., et al. (2006). What children do on the internet: domains visited and their relationship to socio-demographic characteristics and academic performance. CyberPsychology & Behavior, 10(2), 182e190. Jackson, L., von Eye, A., Biocca, F., Barbatsis, G., Zhao, Y., & Fitzgerald, H. (2005). How low-income children use the internet at home. Journal of Interactive Learning Research, 16(3), 259e271. Keaton, P. (2012). Public elementary and secondary school student enrollment and staff counts from the common core of Data: School year 2010-11. Washington, D.C.: National Center for Education Statistics, U.S. Department of Education. Khan, S. (2012). The one world schoolhouse: Education reimagined. New York, NY: Twelve. Kim, S., & Chang, M. (2010). Computer games for the math achievement of diverse students. Educational Technology & Society, 13(3), 224e232. Kuss, D. J., & Griffiths, M. D. (2012). Online gaming addiction in children and adolescents: a review of empirical research. Journal of Behavioral Addictions, 1(1), 3e22. Lazer, D., Kennedy, R., King, G., & Vespignani, A. (2014). The parable of google Flu: traps in big data analysis. Science, 343(6176), 1203e1205. Lillard, A. S., & Peterson, J. (2011). The immediate impact of different types of television on young children's executive function. Pediatrics, 128(4), 644e649. http://dx.doi.org/10.1542/peds.2010-1919. Madden, M. (2013, May 29). Technology use by different income groups. Paper presented at the 16th Annual Welfare Research and Evaluation Conference, Washington, D.C. McCarthy, M. J. (2010). Internet monitoring of suicide risk in the population. Journal of Affective Disorders, 122(3), 277e279. National Center for Education Statistics. (2010). Internet access in U.S. public schools and classrooms: 1994-2005 and educational technology in U.S. public schools: Fall 2008. Washington, D.C.: U.S. Department of Education. Preis, T., Reith, D., & Stanley, H. E. (2010). Complex dynamics of our economic life on different scales: insights from search engine query data. Philosophical Transactions of the Royal Society A, 368, 5707e5719. Purcell, K., Rainie, L., Heaps, A., Buchanan, J., Friedrich, L., Jacklin, A., et al. (2012). How teens do research in the digital world. Washington, D.C.: Pew Research Center. Reinhart, J. M., Thomas, E., & Toriskie, J. M. (2011). K-12 teachers: technology use and the second level digital divide. Journal of Instructional Psychology, 38(3), 181e193. Rideout, V. J., Foeh, U. G., & Roberts, D. F. (2010). Generation M2: Media in the lives of 8- to 18-year-olds. Menlo Park, CA: Kaiser Family Foundation. Rowlands, I., Nicholas, D., Williams, P., Huntington, P., Fieldhouse, M., Gunter, B., et al. (2008). The google generation: the information behaviour of the researcher of the future. Aslib Proceedings, 60(4), 290e310. Seidman, R. (2013, August 23). List of how many homes each cable networks is in - Cable network coverage estimates as of August 2013. Retrieved from http:// tvbythenumbers.zap2it.com/2013/08/23/list-of-how-many-homes-each-cable-networks-is-in-cable-network-coverage-estimates-as-of-august-2013/ 199072/. Sengupta, S. (2011, December 4). Online learning, personalized. The New York Times. Retrieved from http://www.nytimes.com/2011/12/05/technology/khanacademy-blends-its-youtube-approach-with-classrooms.html. Sirin, S. R. (2005). Socioeconomic status and academic achievement: a meta-analytic review of research. Review of Educational Research, 75(3), 417e453. The White House. (2013, June 6). President Obama unveils ConnectED initiative to bring America's students into digital age. Retrieved July 20, 2013 from http:// www.whitehouse.gov/the-press-office/2013/06/06/president-obama-unveils-connected-initiative-bring-america-s-students-di. van Deursen, A. J., & van Dijk, J. A. (2014). The digital divide shifts to differences in usage. New Media & Society, 16(3), 507e526. http://dx.doi.org/10.1177/ 1461444813487959. Wainer, J., Dwyer, T., Dutra, R. S., Covic, A., Magalhaes, V. B., Ferreira, L. R. R., et al. (2008). Too much computer and internet use is bad for your grades, especially if you are young and poor: results from the 2001 Brazilian SAEB. Computers & Education, 51(4), 1417e1429. Wieman, C. E., Adams, W. K., & Perkins, K. K. (2008). PhET: simulations that enhance learning. Science, 322(5902), 682e683. Wood, L., & Howley, A. (2012). Dividing at an early age: the hidden digital divide in Ohio elementary schools. Learning, Media and Technology, 37(1), 20e39. Zhang, M. (2014). Who are interested in online science simulations? Tracking a trend of digital divide in internet use. Computers & Education, 76, 205e214. http://dx.doi.org/10.1016/j.compedu.2014.04.001. Meilan Zhang is an Assistant Professor of Educational Technology in the Department of Teacher Education at the University of Texas at El Paso. Her research interests focus on improving student learning using mobile technology and understanding Internet use and digital divide using big data from Internet search trends and Web analytics.