Analysis of Gender Differences in Internet Use of Children : Evidence from Turkey

In this study, factors in Internet use of female and male children in Turkey were determined with probit regression model by using micro data set in Household Information Technologies Usage Research of 2013 carried out by Turkish Statistical Institute. Dependent variable of the study is two category variable, namely Internet use and non-use of female and male children. Independent variables are socio-economic and demographic variables. According to chi-square analysis, there is a relation between Internet use of female and male children and socio-economic and demographic characteristics. According to probit regression analysis results, for female children, region, educational status, having computer or mobile phone on their own, frequency of watching TV, watching movie, series; floor show, music, game show; watching educational programs such as documentaries, culture, art, reading newspaper and journal in printed media, using mobile phone and frequency of using computer are variables effective in Internet use. Region, rural-urban difference, age, being literate, educational status, having mobile phone or game console on their own, watching entertainment, music, competition programs, reading newspaper and journal in printed media, using mobile phone and, frequency of using computer are variables effective in Internet use among male children. Frequency of using computer is the most effective variable in Internet use and it is more effective among female children compared to male children.


Introduction
Today, digital media and Internet became a part of communication and information tools of many children.Internet access and use of children has been increasing throughout the world (McQuillan & O'Neill, 2009).Internet which was founded in 1960's in America has spread out all around the world in the last 50 years and according to data of June 2012, it has reached 2,405,518,376 people around the world (Mayda & Yılmaz, 2015).The section who is especially born after 1979 and named as "Generation Y" is on the frontline both due to their population and the new technology they use.This section and children and young people of today are primary actors of technology in general sense and in parts which require interaction and communication of computer and Internet.Considering in this sense, children are the factors who use technology well and progress to future (Canbek & Sağıroğlu, 2007).87% of 5-7 year olds are known to use the Internet in UK, 21% of the 6-7 years old and 48% of the 8-9 year old use the Internet "at least rarely" in Germany, 64% of 7 year olds use the Internet in Finland, 70% of Flemish pre-schoolers are online, usually from the age of 3 to 4 onwards, and mostly on a regular basis of at least several times a month in Belgium, 70% of 3 to 4 year olds go online at least sometimes in Sweden, 78% of Dutch toddlers and pre-schoolers are already online and 5% of babies under 1 are going online in Netherlands, almost half of 3-6 year olds use the Internet on a regular basis in Austria, 58% of 0-6 year olds go online in Norway, 93% of 3-9 year olds go online for an average of 8-9 hours a week in South Korea, 25% of 3 year olds go online daily, rising to about 50% by age 5 and nearly 70% by age 8 in the US and 79% of children aged between 5-8 years go online at home in Australia (Holloway, Green, & Livingstone, 2013).
Research across Europe provides evidence that gender gaps in access to the Internet are closing in nearly all countries, as home and school access becomes more commonplace.Although in most countries the percentage of boys using the Internet is slightly higher than girls, in only three countries are the differences statistically significant: Austria (boys: 62 per cent, girls: 45 per cent), Italy (boys: 41 per cent, girls: 30 per cent), and Germany (boys: 52 per cent, girls: 43 per cent) (McQuillan & O'Neill, 2009).
Many factors such as demographic, geographic and economic factors are effective in Internet use of children (Cleary, Pierce, & Trauth, 2006).Internet use and Internet access are important factors in the sense of equality.Internet can be used for entertainment and educational purposes and may also provide social benefits via communication mediums such as e-mail and social media (Madell & Muncer, 2004).Young adults are avid Internet users.Online social media, such as social networking sites (e.g., Facebook, MySpace), blogs, status updating sites (e.g., Twitter) and chat rooms, have become integral parts of childrens' and young adults' lives (Pujazon-Zazik & Park, 2010).Internet use influence sleeping behavior ( Van den Bulck, 2004), health and social life (Koyuncu, Unsal, & Arslantas, 2014), familial communication and time spent with family (Lee & Chae, 2007), academic performance (Jackson et al., 2008;Kim & So, 2012) of children.Numerous surveys have attempted to measure how frequently children use the Internet at home.Estimates vary from as high as several hours a day to as low as 3 hours a week, depending on how Internet use is measured (e.g., self-report, automatically recorded), age of children sampled, and the year data were collected.Despite high variability in empirical estimates, public perception is that children spend a great deal of time online (Jackson, von Eye, & Biocca 2003).In a research carried out on 9-19 years old age group children in UK in 2004, it was determined that 84% of the children used Internet at least one time in a week (Livingstone, 2011).
In Turkey, internet use has been increasingly become wide and physical and mental health problems as relevant to internet addiction has been increasing at youths (Kayıran, Comert, Kose, & Gurakan, 2012).Misuse of internet may affect especially social life of the children in various ways.Important problems occur at social development of the children who use internet very frequently and play computer games.Having low self-esteem, increasing their social concerns and making peevish behaviors are one of these problems seen on children.While it is provided that children and youths are getting benefit from computer and internet opportunities in the right, effective and efficient way, their safeties are given particular importance (Harman, Hansen, Cochran, & Lindsey, 2005;Holloway et al., 2013;Livingstone, Görzig, & Ólafsson, 2011).The basic aim of this study is to determine effect size of socio-economic and demographic factors which are effective in Internet use of children considering gender difference of them.Therefore, by using cross-section data in Household Information Technologies Usage Research carried out by Turkish Statistical Institute in 2013, factors which are effective in Internet use of female and male children in 6-15 age group will be analyzed with chi-square and probit regression analysis.The rest of this study was designed as such.In the second section, literature review was carried out about Internet use of children.In the third section, the method to be used in the study was explained and data, variables to be used were expressed.In the fourth section, analysis results were given in details.In the fifth section, discussion of findings was given.

Literatur Review
The Internet may be a global technology but children work in local/national contexts, and have differences in other aspects of their identities; one of the most important of which is gender identity (Li & Kirkup, 2007).Gender differences in the use of children's Internet have been well documented with studies in recent years (McQuillan & O'Neill, 2009;Madell & Muncer, 2004;Miliany, 2014;Talves & Kalmus, 2015).Although the Internet frequently has been characterized as male-dominated, recent evidence indicates that the gender differences in Internet use is rapidly diminishing (Weiser, 2000).Studies in recent years pay attention to gender differences.In these studies, gender differences are not very pronounced for interactive media use (McQuillan & O'Neill, 2009).Traditionally, technology is a male sphere, and research has previously shown that boys have a greater interest in technology itself than girls.Girls want to use the technology (Enochsson, 2005).
There are many academic studies which analyze gender differences in Internet use.In the study carried out for female and male children in Korea, Kim and So (2012) stated that school performance and success of female and male students who spend 3 or more hours on Internet on daily basis decrease compared to students who do not use Internet.Miliany (2014) found that Saudi females are less likely to engage in certain online activities such as social networking.Li and Kirkup (2007) indicated that gender difference still existed in college students or adult users in terms of access to and use of the Internet, attitudes towards the Internet, frequency of Internet use and self-assessment of Internet competency.Enochsson (2005) indicated that boys talk about their knowledge to a greater extent, and this interplays with their reflections about the Internet's reliability.Joiner et al. (2005) found that males were proportionally more likely to have their own web page than were females.They used the Internet more than females, in particular, they were more likely to use game websites, to use other specialist websites, and to download material from the Internet.However, females did not use the Internet for communication more than males.Jackson, Ervin, Gardner and Schmitt (2001) predicted and found that females used e-mail more than did males, males used the Web more than did females, and females reported more computer anxiety, less computer self-efficacy, and less favorable and less stereotypic computer attitudes.Jackson et al. (2008) indicated race and gender differences in the intensity of IT use; African American males were the least intense users of computers and the Internet, and African American females were the most intense users of the Internet.Males, regardless of race, were the most intense videogame players, and females, regardless of race, were the most intense cell phone users.Schumacher and Morahan-Martin (2001) found that males were more experienced with computers, more likely to have taken high school courses requiring computer use, and reported higher skill levels in applications such as programming, games and graphics than females.Martin (1998) indicated that girls were at least as able as boys to complete the Internet activity.Furthermore, girls displayed greater enthusiasm for the task than did their male classmates.Durndell and Haag (2002) indicated that gender effects were found throughout, with males tending to report greater computer self-efficacy, lower computer anxiety, more positive attitudes towards the Internet and longer use of the Internet than females.Jackson et al. (2010) found that boys played videogames far more than did girls.Videogame playing was associated with a lower behavioral self-concept and lower self-esteem.

Probit Regression
Probit regression is one of the qualitative methods which has been widely used in health, science and social sciences fields from 1933 up to now (Cramer, 2003).In probit regression model, dependent variable is two-category variable which has the values of 0 and 1.While 1 shows that mentioned case has occurred, 0 shows that it has not.In other words, in probit regression two category dependent variable is explained such as yes-no, successful-unsuccessful, wrong-correct.In probit regression, probability is used while dependent variable values are estimated.Probit model uses cumulative normal distribution function.According to this, probit model is written as; When the reciprocal of cumulative normal distribution function; equation is obtained (Aldrich & Nelson, 1984).When parameters of probit regression model are estimated, maximum likelihood method is used.In Probit regression, each observation value complies with Bernoulli distribution.Then probability distribution for each observation is, Since observations are independent, likelihood function is, Following necessary mathematical processes, log-likehood function through which model parameters were estimated as such (Montgomery, Peck, & Vining, 2012), By using this function, model parameters which would maximize lnL value are estimated.
Since marginal effects are dependent on independent variables in regression model, marginal effects vary according to different observations.In a probit regression model where there is n observation and k independent variables, the effect of j.Independent variable can be calculated through average marginal effects.In other words, marginal effect for each independent variable, equation is obtained (Heij, De Boer, Franses, Kloek & Van Dijk, 2004).

Data
Data used in this study was obtained from the survey of Household Information Technologies Usage Research which was carried out by TUIK in 2013.Household Information Technologies Usage Research is carried out annually on regular basis since 2004 (except for 2006) by TUIK compatible to EU regulations, with the help of model question form developed in close association of European Union Statistical Office with EU member states offices and OECD.Every settlement place was included in the scope for sample choice throughout Turkey.Sampling method of the research is two stage stratified cluster sampling.In the first stage, clusters (blocks) composed of 100 houses on average were chosen into (PPS) sample in proportion to the size, in the second stage by using systematic choice method from clusters chosen for sample, sample addresses were determined.In this study, cross-section data of 6-15 age group children were used (TUIK).

Measures and Variables
Dependent variable of the study is the status of Internet use which is measured with the question "Do you use Internet?"(Yes/No).Dependent variable is two category variable in this study.In probit regression model, dependent variable categories are 1 if the child is using Internet and 0 is the child is not using Internet.
Under the name of Statistical Territorial Units Classification (STUC), Turkey was divided into 12 regions at Section 1.These regions and cities in these regions are given in details in Table 1.In the basis of forming STUC regions in Turkey, there is the compulsion of founding Development Agencies.Since accession partnership contract signed with EU and national program prepared later on regard STUC regions as a precondition to found Development Agencies, it was a compulsion to found STUC regions.Available geographical regions were not considered in forming STUC regions in Turkey, region boundaries were determined depending on different criteria.One of the most important factor is population.Apart from population, cultural structure and development level of cities were also considered (Taş, 2006).

Estimated Model
Probit regression model was used in order to determine factors effective in Internet use of children according to gender.In order to analyze effects of categories belonging to all variables to be included in probit regression model, ordinal and nominal variables were defined as dummy variable.It was tested whether there is multiple linear relation between independent variables to be included in probit regression model.It is thought that if variance inflation factor (VIF) values are 5 and above, it results in medium level, if it is 10 and above it results in high level of multiple linear relation (Bagheri, Habshah, & Imon, 2012).As it is seen in Results and marginal effects of estimated probit regression models are given in Table 3. Models formed were found statistically significant (P<0,000).Pseudo R 2 values of models were calculated as 0.5728, 0.5853 and 0.5679 respectively.

Marginal Effects
Marginal effects show change in the probability of estimated Internet use against one unit of change in independent variable.According to probit regression models, estimated possibility of Internet use of female and male children in TR1 (İstanbul) region is 8.4% and 4.9% more respectively compared to female and male children in TRC (Southeast Anatolia) region.Estimated possibility of Internet use of female and male children in TR2 (Western Marmara), TR3 (Aegean), TR4 (Southern Marmara), TR5 (Western Anatolia), TR6 (Mediterranean), TR7 (Central Anatolia) and TR9 (Eastern Black Sea) is (11.4%, 2.8%), (7%, 9.4%), (9.1%, 12.8%), (5.4%, 6%), (4.9%, 5.7%), (5.6%, 4.6%) and (10.2%, 6.9%) more compared to female and male children in TRC (Southeastern Anatolia) region.It was determined that place of residence, age and being literate are not effective in Internet use of female children.Estimated possibility of Internet use of male children in urban area is 4.6% more than those in rural area.Estimated possibility of Internet use of male children in 11-15 age group is 3.6% more than male children in 6-10 age group.Estimated possibility of Internet use of male children who are literate is 6.9% more than those who are illiterate.Estimated possibility of Internet use of male children who attend general secondary and general high-school is respectively 3.9% and 11.7% more compared to male children who attend elementary school.For female children, estimated possibility of Internet use of those who attend general secondary school is 4.2% more compared to those who attend elementary school.Estimated possibility of Internet use of female children who has computer and mobile phone on their own is 2.8% and 3.9% more compared to those who do not have.Estimated possibility of Internet use of male children who have mobile phone and game console is respectively 4.1% and 6.8% more compared to those who do not have.Estimated possibility of Internet use of female children who watch television nearly every day is 4.2% more compared to those who watch at least once in a week.Estimated possibility of Internet use of female children who watch movies, series, entertainment, music, competition; educational programs such as documentaries, culture, art is respectively 2.9%, 2% and 2.7% more compared to those who do not watch.Estimated possibility of Internet use of male children who watch entertainment, music, competition programs is 2.3% more compared to those who do not watch.Estimated possibility of Internet use of female and male children who read newspaper and journal in printed media is respectively 4.4% and 4.8% more compared those who do not read.Estimated possibility of Internet use of female and male children who use mobile phone is respectively 4.9% and 3.1% more compared to those who do not use.Frequency of using computer is the most effective variable in Internet use of female and male children.Estimated possibility of Internet use of female and male children who use computer nearly every day is respectively 44.7% and 41.6% more compared to those who do never use.Estimated possibility of Internet use of female and male children who use computer at least once in a week is respectively 37.3% and 38.4% more compared to those who do not use.Estimated possibility of Internet use of female and male children who use computer at least once in a month is respectively 36% and 33.1% more compared to those who do not use.Estimated possibility of Internet use of female and male children who use computer less than at least once in a month is respectively 31.7% and 26.4% more compared to those who do not use.

Discussion
In this study, socio-economic and demographic factors which effect Internet use of female and male children in Turkey were determined by using probit regression model.Moreover it was determined how much these factors are effective in Internet use of female and male children.
It was determined that possibilities of Internet use of female and male children in economically more developed regions was higher.High level of wealth of regions has positive effect on Internet use possibility of female and male children.Regional geography may also influence Internet use.Spooner (2003) found that web use was highest in the New England states, 1 with 66 percent of the residents having Internet access, and the Pacific Northwest, 2 with 68 percent of these residents having web access.By contrast, Internet access was lowest in the south, 3 with only 48 percent of the residents having web access.It is likely that regional variations in education and income may partially explain the regional variation in web Access (Wasserman & Richmond-Abbott, 2005).Urban-rural region discrimination is only effective in Internet use of male children.Age and status of being literate also increase only possibility of Internet use of male children.As the age increase among male children, so does the possibility of Internet use.Similarly, as the educational status increase among male children, so does the possibility of Internet use.Having computer and mobile phone increases possibility of using Internet among female children, and having mobile phone and game console increases possibility of using Internet among male children.In a study carried out on Korean children, it was found that possibility of playing online game among male children was higher than female children (Lee & Chae, 2007).Similar results were found in another study (Jackson, 2008).Frequency of watching TV is not effective in Internet use of male children.However, it was determined that possibility of Internet use of female children who watch television nearly every day is higher.
While it was determined that possibility of Internet use of male children who watch entertainment, music, competition programs is higher, watching movies, series; entertainment, music, competition; educational programs such as documentaries, culture and are increase possibility of Internet use among female children.
Reading newspaper and journal in printed media has positive effect on Internet use of female and male children.Status of using mobile phone is more effective in Internet use of female children compared to male children.Frequency of using computer is the most effective variables in Internet use of female and male children.As the frequency of computer using increases for both female and male children, so does possibility of estimated Internet use.It was determined that frequency of computer using is more effective in Internet use of female children compared to male children.
This study has specific limitations.Since cross-section data are used, inference cannot be done for the future.
Estimations only for the current status were obtained.Since secondary data are used; variables which would be important such as monthly income, occupations of parents, educational status of the household were not included in the study.
In this study, factors which only effect Internet use were analyzed.In studies which would be carried out in the future, factors which effect Internet addiction of children in Turkey can be analyzed in details.Similarly, frequency of Internet use of children can be analyzed in more than two categories by using ordered probit regression model.

Conclusions
According to probit regression analysis results, for female children, region, educational status, having computer or mobile phone on their own, frequency of watching TV, watching movie, series; floor show, music, game show; watching educational programs such as documentaries, culture, art, reading newspaper and journal in printed media, using mobile phone and frequency of using computer are variables effective in Internet use.Region, rural-urban difference, age, being literate, educational status, having mobile phone or game console on their own, watching entertainment, music, competition programs, reading newspaper and journal in printed media, using mobile phone and , frequency of using computer are variables effective in Internet use among male children.Frequency of using computer is the most effective variable in Internet use and it is more effective among female children compared to male children.

Table 1 .
Statistical region units classification-Level 1Independent variables in the study, number of Internet use of children according to genders, percentages and chi-square test results related with status of Internet use are given in Table2.According to chi-square independent test results, it was determined that there was a significant relation between Internet use status of both females and males and socio-economic and demographic variables.The rate of female and male children who use Internet in more developed regions in economic sense is higher.While 16.41% of female children and 13.57% of male children who use Internet is in TR1 (İstanbul) region, 23.73% of female children and 22.69% of male children who do not use Internet is in TRC (Gaziantep, Adıyaman, Kilis, Şanlıurfa, Diyarbakır, Mardin, Batman, Şırnak, Siirt) region.80.9% and 81.01% of female and male children respectively who use Internet live in urban regions.Nearly all of the female and male children who use Internet (98%) are literate.It is seen in Table 2 that 66.95% of female children who use Internet are in 11-15 age group, 38.84% are at elementary school, 45.55% are at general secondary school, 66.23% of female children who use Internet are in 11-15 age group, 41.78% are at elementary school, 42.66% are at general secondary school.When Table 2 is analyzed, it is seen that 60.3% of female children who use Internet do not have computer on their own, 76.25% do not have mobile phone and 97.19% do not have a game console; 61.23% of male children who use Internet do not have computer on their own, 75.22% do not have mobile phone and 93.61% do not have game console.Respectively 89.67% and 92.26% of female and male children who use Internet watch TV nearly every day.It was determined that Internet use status of children varies according to status of watching television.It was determined that 12.97% of female children who use Internet watch news programs, 74.41% watch movies, series; 57.03% watch cartoons, 68.39% watch entertainment, music-competition programs, 10.27% watch sports programs and 25.76% watch educational programs such as documentaries, culture, art programs.It was determined that 15.98% of male children who use Internet watch news programs, 64.43% watch movies, series; 63.13% watch cartoons, 54.19% watch entertainment, music-competition programs, 51.69% watch sports programs and 28.58% watch educational programs such as documentaries, culture, art programs.When the rate of Internet use of children according to frequency of computer use is analyzed, 43.6% of female children who use Internet use computer nearly every day and 46.41% use computer at least once in a week.50.2% of female children who use Internet use computer nearly every day and 41.18% use computer at least once in a week.

Table 2 .
Internet use of children according to gender

Table 3 ,
none of the independent variables included in model have 5 or more variance inflation factor.According to this, there is no variable which cause multiple linear relation problem between variables in model.General probit regression model formed after independent variables of model are determined is written as such.

Table 3 .
Factors effective in Internet use of female and male children