- Email: [email protected]
Ex-post evaluation of illegalizing juvenile online game after midnight: A case of shutdown policy in South Korea
Telematics and Informatics 34 (2017) 1597–1606
Contents lists available at ScienceDirect
Telematics and Informatics journal homepage: www.elsevier.com/locate/tele
Ex-post evaluation of illegalizing juvenile online game after midnight: A case of shutdown policy in South Korea Changjun Lee a, Hongbum Kim b, Ahreum Hong c,⇑ a
Lee Kuan Yew School of Public Policy in National University of Singapore, Singapore Korea National Industrial Convergence Center, Korea Institute of Industrial Technology, South Korea c Graduate School of Technology Management, Kyung Hee University, 1732 Deokyoung-ro Giheung-gu, Youngin-Si Gyunggi-Do, South Korea b
a r t i c l e
i n f o
Article history: Received 15 June 2017 Received in revised form 10 July 2017 Accepted 11 July 2017 Available online 14 July 2017 Keywords: Media policy Internet hours Sleep duration Policy evaluation Youth policy
a b s t r a c t In November 2011, the Korean government legalized blocking access to online games for youths younger than age 16 late at night; this is called the shutdown policy. Using multiple regressions we examined how the compulsory block affected youths’ Internet hours and sleep duration. Data were drawn from the 2011, 2012 Korea Youth Behavior Risk Factor Survey, a cross-sectional online survey of middle and high school students aged 13–18 years. Legalizing a ban of online gaming late at night for youths caused an increase in the predicted probability of being in a high-ranked Internet user group by 1.6 percent points, a decrease in the predicted probability of Internet addiction by 0.7 percent points, and an increase in sleep duration of 1.5 min. All results showed a gender difference in the effect of the policy. Although the net effect of the shutdown policy was statistically significant, the small effect size, the partial effect on female youths, and the side effects related to human basic rights and inappropriate regulation of the game industry made the effectiveness of the policy arguable. Ó 2017 Elsevier Ltd. All rights reserved.
1. Introduction The Internet is now a part of most people’s daily lives. In 2016, the rate of Internet accessibility in households reached a high proportion in East Asian countries (96.5% in Japan, 89.5% in Singapore, and 98.8% in South Korea; hereafter, Korea) as well as in Western countries (91.3% in England, 82.2% in the United States, and 90% in Germany) (International Telecommunication Union, 2016; Ministry of Science, ICT and Future Planning, 2017). While many people can access the Internet easily by many means, the rate of overdependence on the Internet, especially using smartphones, has been increasing. According to a national survey, the rate of people in the danger zone of overdependence on the Internet has been increased by 9.4 percent points, from 8.4% in 2011 to 17.8% in 2016 (Ministry of Science, ICT and Future Planning, 2017). Among the people in the danger zone, teenagers were at 30.6%, ranking as the highest proportion. Usage of computers and the Internet has many benefits. It helps workers build required skill sets for jobs (Turow and Lilach, 2000), and students get higher scores in math and reading given appropriate time use (Battle, 1999). However, excessive use is negatively associated with physical and psychological well-being, especially for youths (Wight et al., 2009). The excessive online time also correlates to poor lifestyle habits in youths (Wang et al., 2012).
⇑ Corresponding author. E-mail addresses: [email protected] (C. Lee), [email protected] (H. Kim), [email protected] (A. Hong). http://dx.doi.org/10.1016/j.tele.2017.07.006 0736-5853/Ó 2017 Elsevier Ltd. All rights reserved.
1598
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
Among the side effects of excessive Internet usage, the lack of sleep duration raises a serious concern for youths, and this is mostly caused by online gaming at night because late-night playing of online games inevitably crowds out youths’ time for adequate sleep. Sleep duration of youths in Korea amounts to 31% of the day, which is less than the 33–37% per day in Japan and the United States (National Youth Policy Institute, 2014). Youths’ lack of sleep duration causes behavioral, cognitive, and mood impairments (O’Brien, 2009) as well as depression and anxiety (Reigstad et al., 2010). To protect youths from playing online games excessively and to increase their sleep duration, in November 2011 the Korean government legalized the blocking of access to online games for users below 16 years of age late at night, the so-called shutdown policy. Instead of the autonomous regulation applied in most countries such as the United States, European Union, and Japan, only a few countries like Korea, China, Thailand, and Vietnam chose compulsory ways such as the shutdown system (Kim et al., 2015a; Park and Ahn, 2010). Even if the shutdown policy can encourage youth to sleep more by prohibiting online gaming at night, scholars have argued that forbidding Internet games at night is not the ultimate solution to alleviate Internet or online game addiction or lack of sleep duration (e.g., Jeong et al., 2016; Jin, 2016; Lee and Kim, 2017). The compulsory regulations like the shutdown policy have been shown not to be a solution for relieving game addiction (Jeong et al., 2016). According to Lee and Kim (2017), the policy should not be legitimized as youths also have their right to do gaming at night even if online game activity late at night is correlated with game addiction. Moreover, there are many ways for youths to avoid the regulation. For example, they can route through an overseas server or use their parents’ account to game late at night (Jin, 2016). In other aspects, some scholars have argued that the policy could hinder the growth of game and software industries due to unnecessary intervention on the demand side (Kim et al., 2015b; Sang et al., 2017). Because of such negative aspects, a shutdown policy is not widely implemented worldwide, so related studies examining the policy’s effect itself are not prevalent. The Korea Institute for Industrial Economics and Trade (2013) is the only literature addressing the effect of the shutdown policy empirically, to the best of the authors’ knowledge. However, it has some limitations such as small observations and unbalanced samples, less various of estimation, and limited focus on the gender effect, which is one of the primary factors affecting online gaming behavior in youth groups (Lucas and Sherry, 2004; Ogletree and Drake, 2007). To fill this research gap, this research aims to examine whether or not the shutdown policy actually affected a reduction in juvenile Internet hours and an increase in juvenile sleep duration, and, if it did, to estimate the size of the effect. Evaluating policy is always important even if the policy is abolished because policy makers may refer to the evaluation in many aspects when they need to build a policy in similar situations (Vedung, 1997). Therefore, this study will provide empirical evidence for political debates for future practical actions. In the rest of the paper, we firstly introduce the shutdown policy in details and review literature on the policy. Next, we define the key independent and the outcome variables, then using the difference in difference approach, examine how the shutdown policy effect on juvenile sleep duration and Internet hours. In result and analysis section, we show the significance and the size of the effect, and lastly, we discuss the results with regard to the previous discourse and gender difference to conclude a policy evaluation of legalizing a ban of online gaming late at night for youths. 2. Background 2.1. Shutdown policy The shutdown policy aimed to protect Korean youth from playing online games excessively, which can lead to addiction to the games, violence and a lack of social activities. For the implementation of shutdown policy, Juvenile Protection Act was amended in May 2011, initiated by Ministry of Gender Equality and Family. Newly added Article – Restriction on Hours Provided for Internet Games in Late Night Time – became the legal basis of the regulation. The policy protects by blocking access to online games for those under 16 from midnight to 6 a.m. However, due to the lack of ex-ante evaluation of the policy, it has caused many of controversies (Kim et al., 2015a; Sang et al., 2017). First, the shutdown policy is targeting only games made in Korea. Second, the system violates the constitutional principle of family autonomy and infringes upon parents’ right to educate and foster their children as they see fit. Finally, government authorities do not distinguish between online games and other games with the online game shutdown policy. In 2016, another government department, Ministry of Culture, Sports and Tourism announced a plan to abolish the shutdown policy because the firms in the gaming industry insisted that the policy was a kind of the double regulation added on to the existing policy, the Protection of Adolescents Act (Kim, 2016). In addition, the validity of the shutdown policy has yet to become clear. Youths could borrow identities of adults and play games if they were eager to do so. Also, there are other ways to reduce the time of playing online games even if there is a way to stop playing online games from midnight to 6 a.m. 2.2. Previous studies of the shutdown policy Because online game addiction and its negative effects are frequently discussed (Allison et al., 2006; Liu and Chang, 2016; Wenzel et al., 2009), many countries are implementing appropriate regulation of online game usage. In the United States, European Union, and Japan, the game industry is restricting their services by autonomous regulation, thus promoting the
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
1599
online gaming industry as well as securing legitimacy. In contrast, Korea, China, Thailand, and Vietnam have implemented compulsory shutdown systems (Kim et al., 2015a; Park and Ahn, 2010). Because such shutdown systems’ law and policy are not prevalent worldwide, existing studies examining the effect of the shutdown law and policy are not easily found. Even many domestic studies regarding the shutdown law in Korea are dealing with the legality of the law. Some previous studies addressed questions regarding the effect of a shutdown policy. Jeong et al. (2016) showed that compulsory regulation is not a solution for relieving game addiction. Lee and Kim (2017) emphasized that although online game activity late at night is correlated with game addiction, this cannot legitimize the existence of the shutdown system. By using an analytical framework regarding modalities of regulation, Sang et al. (2017) examined the regulatory framework of the shutdown law. After a case analysis, they noted problems regarding not only basic human rights, but also the game industry context. Specifically, the failure of the age verification system was not acceptable to the regulatory framework, meaning that the policy should be fully reconsidered. Kim et al. (2015b) also stated that two different online game policies, promotion policy and regulation policy, conflicted with each other, causing the game industry to be confused. Although those studies dealt with the shutdown policy and its contextual implications, empirical evaluations on the effect of the policy are missing in the current literature. To the authors’ best knowledge, the Korea Institute for Industrial Economics and Trade (2013) alone evaluated the effect of the shutdown law empirically. The study showed that although the policy was effective in terms of securing juveniles’ right to sleep, the playing time of online games increased. This implies the essential objective of the shutdown policy for alleviating gaming addiction was not achieved. As an alternative, Park and Ahn (2010) proposed two different regulation approaches to solve the online game addiction problem and evaluate them with a system dynamics approach: a self-regulation policy and a tax and rebate policy. Their results showed that a tax and rebate policy is more effective in terms of sales growth, improvement of social image, and especially decrease of game user addiction. Scholars have also focused on gender differences in the time consumption of youths. In terms of time consumption for gaming, they found that male youths spent much more time online gaming than female youths. Ogletree and Drake (2007) found that male youths played video games two or more hours a week than female youths, so male students were more likely to experience sleeping and class preparation interference due to playing games. Lucas and Sherry (2004) also found a gender difference in the time spent video gaming. They showed that female respondents reported less frequent play and less motivation to play in social situations in a survey of youths. Those studies hint that the effects of policy might be different by gender. While some discussions on the shutdown policy have been carried out, the effect of the shutdown policy on juvenile Internet usage and sleep time has not been critically or empirically assessed. In light of this research gap, this study evaluated the effect of the shutdown policy on Internet usage time and sleep duration in Korean youths.
3. Method 3.1. Data and analysis sample We used data drawn from the 2011–2012 Korea Youth Risk Behavior Web-based Survey (KYRBS), an annual crosssectional online survey of middle and high school students aged 13–18 years. KYRBS collects data on health behavior in the form of an anonymous and self-reported survey. To minimize sampling error, KYRBS divides the population into 129 strata by using 43 regional and school classes as strata variables with regard to geographical accessibility, the number of schools and population, the life environment, and the rate of smoking and drinking (Ministry of Health and Welfare, 2013). The data set enables researchers to determine whether a respondent was subject to the shutdown policy in a given year because it contains detailed information on the exact age and date of an interview each year. This study constructed the analytic sample in the following ways. It restricted the sample to respondents who were 13– 18 years old in each wave. For the purpose of this study to examine the marginal impact of the policy on sleep and Internet hours for juveniles, the sample consisted of students who participated in 2011 and 2012 to exclude respondents who were already adjusted to the policy after 2012.
3.2. Key independent variables The study included three key independent binary variables. It determined the value of each variable in each year considering the exact age and date of the interview in the KYRBS. The first one was a treatment variable, which indicated whether a respondent was eligible for the shutdown policy or not. Specifically, the variable was equal to 1 if the age of a respondent was below 16 and 0 if he or she were not eligible for the policy each year. The second was a post policy variable, which indicated whether the time a respondent surveyed had passed during the time the shutdown policy applied or not. The variable was equal to 1 if the exact surveyed date was after November 2011, and 0 otherwise. The third was the interaction term multiplying the first and second variables, which indicated the genuine difference due to the policy considering the existing difference between the treatment and the control groups.
1600
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
3.3. Outcome variables This study examined two outcome variables for juveniles’ health behavior. First, we analyzed how long respondents slept on average in a year. We only used the data from 2011 to 2012 to avoid measurement error because the KRYBS has asked about sleep hours separately on weekdays and weekends since 2013. Fig. 1 shows the distribution of juvenile sleep duration by gender. The distribution is normal for being a dependent variable in an ordinary least squares (OLS) regression. The second outcome variable was Internet hours, how long respondents used the Internet on average by purpose and days of the week. Purposes consisted of two categories: for study and for other activities. We used only Internet hours not for the purpose of study because Internet hours for studying were not counted in the shutdown policy. To test the impact of the shutdown policy, we created an extreme variable, Internet addiction, which indicated whether a respondent belonged to the Internet addicted or alarming group or not. The variable was equal to 1 if a respondent’s Internet hours were over 3.15 h except hours used for study purposes and 0 if below 3.15 h according to the result of an investigation of Internet addiction in Korea showing that the average Internet hours of the addicted alarming users and the addicted users were 3.15 and 3.75 h a day, respectively (Ministry of Science, ICT and Future Planning, 2014). Fig. 2 shows the distribution of juvenile Internet hours only including hours not for study purposes by gender. Because the distribution of Internet hours did not follow a normal distribution, to conduct ordered logit and logit regression we categorized Internet hours into five groups: (1) light user, someone who used the Internet for 0–1 h, (2) normal user, someone who used the Internet for 1–2 h, (3) heavy user, someone who used the Internet for 2–3.15 h, (4) addicted alarming user, someone who used the Internet for 3.15– 3.75 h, and (5) addicted user, someone who used the Internet for 3.75 or more hours. Female youths comprised a larger proportion of the light user group than males, whereas males comprised a larger proportion of the heavy user group and groups four and five than females. 3.4. Research design To estimate the effect of the shutdown policy on respondents’ sleep and Internet hours, we conducted multivariate regression analysis in a difference research design. This design used OLS regression for the continuous dependent variable of sleep duration. It used ordered logit regression for the ordered categorical variable, Internet hours, and logit regression for the binary variable, Internet addicted. The equation below shows the estimation model. In the equation, yi was sleep duration, Internet hours, and Internet addicted; i indexed each respondent, and t indexed each year.
yi ¼ a þ b1Treati þ b2Posti þ b3ðTreati Posti Þ þ b4X i þ b5City:FEsi þ ei
ð1Þ
0
.1
Density
.2
.3
The key independent variable was the interaction term, which was obtained by multiplying the treatment group and post policy. Then the coefficient, b3, captured the effect of the shutdown policy per se on the outcomes. We also included various covariates, denoted by the vector X i . We controlled for additional exogenous outcome-related characteristics, such as grades, the class of schools, parents’ educational attainment, and living arrangements (categorized into no parent alive, living with
.5
2.8
5 7.3 Sleep duration Female youths
9.5 Male youths
Fig. 1. Distribution of juvenile sleep duration by gender.
11.75
1601
0
.2
Density
.4
.6
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
0
2
4 6 Internet hours Female youths
8
10
Male youths
Fig. 2. Distribution of juvenile Internet hours excluding hours for a purpose on study by gender.
parents, only father, and only mother) at each year. In order to capture city- or region-specific trends in the dependent variables, in all models we included city fixed effects, denoted by City.FEs, a vector consisting of dummy variables for each year (12 cities and regions are controlled). We also conducted subgroup analysis to see whether effects varied by gender. 4. Results 4.1. Descriptive results Table 1 shows descriptive statistics for variables in this study. The mean value of Internet hours excluding hours for the purpose of study was 1.53 h with a standard deviation (SD) of 1.52. The mean value of sleep duration was 6.39 h with an SD of 1.39. When comparing male and female samples, it was clear males had longer Internet hours (1.68 h with an SD of 1.61 h in the male sample and 1.38 h with an SD of 1.40 h in the female sample), as well as sleep duration (6.60 h with an SD of 1.39 h in the male sample and 6.18 h with an SD of 1.36 h in the female sample). The ratio of the treatment group whose age was below 16 at each wave of the survey was 0.58 and the post-shutdown policy, which indicated whether or not the survey dates were later than the date the shutdown policy was applied, was 0.50. The ratio of the interaction term between the treatment group and the post-shutdown policy was 0.30. 4.2. Results of regression analysis Table 2 reports the result of the multiple regression analyses of juvenile Internet hours and sleep duration. For all outcome variables, we report regression coefficients of the main variable, which indicates the difference in the difference effect of the shutdown policy (that is, estimated effects of the policy per se), although the aforementioned other variables are also controlled in the regressions (full regression results are shown in the Appendix A and B). For each outcome variable, the table shows results from models with total, male and female samples. We present OLS regression coefficients for the continuous dependent variables and marginal effects (with all other variables are assumed to be at their mean values) for ordered categorical and binary dependent variables. The first and second rows of Table 2 show the effect of the shutdown policy on the juvenile Internet hours and Internet addiction sleep duration, respectively (for details, see Appendix A and B). The shutdown policy increased the predicted probability of being a high-ranked Internet user group by 1.6 percent points (b = 0. 016, p-value < 0.01). This had a significant impact on female (b = 0. 030, p-value < 0.01), however, with no significant impact on male. When it came to the effect of the shutdown policy on Internet addiction, it decreased the predicted probability of addiction by 0.7 percent points (b = -0.007, p-value < 0.1). It also had no significant impact on male but decreased the probability of being in the Internet-addicted group by 1 percent point at 5% significance level in the female’s sample. The third row of Table 2 shows the effect of the shutdown policy on juvenile sleep duration, which increased 0.025 h, which equals 1.5 min (b = 0.025,
1602
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
Table 1 Descriptive statistics of key variables. Variables
Mean (SD)
Internet hours (except a purpose of study) Internet hours (in five categories) (1) 0–1 h (Light user) (2) 1–2 h (Normal user) (3) 2–3.15 h (Heavy user) (4) 3.15–3.75 h (Addicted alarming user) (5) 3.75 or above hrs (Addicted user) Sleep duration Treatment group (age < 16) Post shutdown policy (year > 2011) Treatment group Post policy Grade 1 2 3 4 5 6 Special high school Father’s educational level Below high school High school College or above Mother’s educational level Below high school High school College or above Coresidence with Both Only father Only mother No one
Total (N = 116,908)
Male youths (N = 58,446)
Female youths (N = 58,462)
1.53 (1.52)
1.68 (1.61)
1.38 (1.40)
0.41 0.27 0.20 0.04 0.07 6.39 (1.39) 0.58 0.50 0.30
0.37 0.28 0.21 0.05 0.09 6.60 (1.39) 0.58 0.50 0.30
0.45 0.27 0.18 0.04 0.06 6.18 (1.36) 0.57 0.49 0.29
0.16 0.16 0.17 0.17 0.17 0.17 0.11
0.16 0.17 0.17 0.17 0.16 0.16 0.11
0.16 0.16 0.17 0.17 0.17 0.17 0.12
0.17 0.36 0.47
0.18 0.35 0.47
0.16 0.38 0.46
0.17 0.47 0.36
0.19 0.45 0.36
0.16 0.49 0.35
0.88 0.03 0.07 0.02
0.88 0.03 0.07 0.02
0.88 0.03 0.07 0.02
Notes: The analytic sample is restricted to respondents who were 13–18 years old in each wave. Grade 1–3 represent middle school students and 4–6 represent high school students in Korea. Special high school includes special-purpose high school such as science- or language-specialized schools.
Table 2 Regression estimates on the effect of the shut-down policy on Juveniles’ Internet hours and sleep duration.
Internet hours (in five ordered categories) Ordered logit coefficients (S.E.) Internet addicted Logit marginal coefficients (S.E.) Sleep duration (in hours) OLS regression coefficients (S.E.) Note: Robust standard errors in parentheses
***
Total (N = 116,908)
Male youths(N = 58,446)
Female youths (N = 58,462)
0.016*** (0.005)
0.002 (0.007)
0.030*** (0.008)
0.007* (0.004)
0.001 (0.006)
0.010** (0.005)
0.025* (0.013)
0.001 (0.019)
0.045** (0.019)
p < 0.01, **p < 0.05, *p < 0.1.
p-value < 0.1). The total impact was shown to increase the sleep duration owing to the effect of the female’s sample. The policy increased the female’s sleep duration by 0.045 h, which equals 2.7 min with a more statistically significant p-value, <0.05. 5. Concluding remarks In this paper, we examined the effect of the shutdown policy on juvenile Internet use hours and sleep duration. Firstly, the results showed that the shutdown policy increased juvenile Internet hours, instead of reducing them, which was supposed to be one of the goals of the policy. However, the policy did reduce the possibility of juvenile Internet addiction. It decreased the
1603
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
predicted probability of Internet addiction by a 0.7 percentage points but was only significant in the female’s sample. As far as the sleep duration is concerned, the policy increased sleep duration 1.5 min in the total sample, and 2.7 min in the female youths’ sample. There was no statistically significant result when conducted only with the male’s sample. The policy was shown to reach the original goals when focusing on only the directions of the coefficients. It decreased the rate of Internet addiction and increased sleep duration. However, even if the coefficients were significant, they were too small to regard them as an effect of the policy (0.7 percent points decrease of the probability of Internet addiction and 1.5 min increase of sleep duration). Those are not large enough to offset the other issues related to the side effects of the policy related to basic human rights and inappropriate regulation for the game industry. The partial effect of the shutdown policy was also considered as a limitation of the policy. It only affected female youths but had no effect on male youths at all. This confirmed the existence of a gender difference in gaming behavior due to different life behaviors with the same results as in Ogletree and Drake (2007) and Lucas and Sherry (2004). As the descriptive statistics show that they are different in daily life including sleep duration and Internet hours (see Table 1), male youths have longer Internet hours and longer sleep duration than female youths. Hence, we assume that male youths are more likely to become addicted to online gaming. Moreover, it might be possible that male youths adapted to the policy faster than female youths, thereby having a higher chance to find a way to avoid the policy. In short, the results of this study provide evidence regarding the effect of the shutdown policy itself. According to the existing empirical literature on the shutdown policy conducted by the Korea Institute for Industrial Economics and Trade (2013), while the shutdown policy was partly effective on juveniles’ game-playing time, the effect of the policy per se was not statistically significant. The analysis of the current study, however, found the significant net effect of the shutdown policy. Yet, with considering the small and partial effects of the policy, the efficacy of the policy remains questionable. Such results can support the arguments of previous literature addressing the potential problems of the shutdown policy. This paper has several limitations. First, this study used Internet hours as a proxy variable for online gaming hours, which could cause a measurement error in estimation. Second, there is limited information on the exact time of using Internet. This hindered us from examining the effect of the shutdown policy on Internet hours late at night. Despite these limitations, this study highlights valuable policy implications for policy makers who are considering a way of regulating juvenile online gaming late at night. The current study based on the exogenous policy change in Korea suggests that the compulsory way of regulating might be an effective policy tool to encourage youths to have enough sleep duration, but it only effects female youths. This is not enough to justify the infringement of fundamental human rights and the criticism that the regulation might violate fair competition in the gaming industry. Appendix A. Difference in difference analysis of juvenile Internet hours.
Variables
Treatment group Post policy Treatment group (age < 16) Post shutdown policy (year > 2011) Male Grade (ref: 1) 2 3 4 5 6 Special high school
Internet hours(in five ranked orders)
Internet addicted
Total
Males
Females
Total
Males
Females
0.065⁄⁄⁄ (0.022) 0.009 (0.028) 0.412⁄⁄⁄ (0.017) 0.065⁄⁄⁄ (0.011)
0.007 (0.031) 0.010 (0.039) 0.292⁄⁄⁄ (0.024) 0.007
0.121⁄⁄⁄ (0.031) 0.018 (0.040) 0.545⁄⁄⁄ (0.024) 0.121⁄⁄⁄
0.073⁄ (0.039) 0.033 (0.049) 0.305⁄⁄⁄ (0.030) 0.073⁄ (0.019)
0.007 (0.050) 0.077 (0.064) 0.269⁄⁄⁄ (0.039) 0.007
0.125⁄⁄ (0.061) 0.002 (0.075) 0.383⁄⁄⁄ (0.049) 0.125⁄⁄
0.347⁄⁄⁄ (0.019) 0.531⁄⁄⁄ (0.019) 0.003 (0.024) 0.116⁄⁄⁄ (0.033) 0.020 (0.033) 0.531⁄⁄⁄ (0.019)
0.332⁄⁄⁄ (0.026) 0.562⁄⁄⁄ (0.026) 0.022 (0.033) 0.153⁄⁄⁄ (0.045) 0.022 (0.045) 0.663⁄⁄⁄ (0.027)
0.358⁄⁄⁄ (0.027) 0.498⁄⁄⁄ (0.027) 0.022 (0.034) 0.064 (0.047) 0.075 (0.047) 0.405⁄⁄⁄ (0.026)
0.308⁄⁄⁄ (0.033) 0.457⁄⁄⁄ (0.033) 0.133⁄⁄⁄ (0.044) 0.111⁄ (0.058) 0.217⁄⁄⁄ (0.059) 0.789⁄⁄⁄ (0.029)
0.289⁄⁄⁄ (0.045) 0.516⁄⁄⁄ (0.043) 0.089 (0.057) 0.006 (0.076) 0.072 (0.077) 0.852⁄⁄⁄ (0.038)
0.321⁄⁄⁄ (0.050) 0.376⁄⁄⁄ (0.049) 0.202⁄⁄⁄ (0.067) 0.278⁄⁄⁄ (0.091) 0.436⁄⁄⁄ (0.092) 0.714⁄⁄⁄ (0.045)
(continued on next page)
1604
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
Difference in difference analysis of juvenile Internet hours. (continued)
Variables
Father’s educational attainment (ref: Below high school) High school College or above Mother’s educational attainment (ref: Below high school) High school College or above Coresidence with (ref: no parents) Only father Only mother Both parents Constant cut 1 Constant cut 2 Constant cut 3 Constant cut 4
Internet hours(in five ranked orders)
Internet addicted
Total
Males
Females
Total
Males
Females
0.028 (0.019) 0.211⁄⁄⁄ (0.020)
0.005 (0.026) 0.262⁄⁄⁄ (0.027)
0.060⁄⁄ (0.027) 0.157⁄⁄⁄ (0.028)
0.085⁄⁄⁄ (0.030) 0.371⁄⁄⁄ (0.032)
0.080⁄⁄ (0.039) 0.424⁄⁄⁄ (0.042)
0.087⁄ (0.047) 0.293⁄⁄⁄ (0.050)
0.080⁄⁄⁄ (0.019) 0.233⁄⁄⁄ (0.020)
0.078⁄⁄⁄ (0.025) 0.264⁄⁄⁄ (0.028)
0.078⁄⁄⁄ (0.027) 0.198⁄⁄⁄ (0.030)
0.088⁄⁄⁄ (0.030) 0.259⁄⁄⁄ (0.034)
0.072⁄ (0.039) 0.229⁄⁄⁄ (0.044)
0.122⁄⁄⁄ (0.046) 0.308⁄⁄⁄ (0.054)
0.320⁄⁄⁄ (0.053) 0.184⁄⁄⁄ (0.048) 0.153⁄⁄⁄ (0.044) 0.554⁄⁄⁄ (0.058) 0.638⁄⁄⁄ (0.058) 1.939⁄⁄⁄ (0.059) 2.467⁄⁄⁄ (0.059)
0.380⁄⁄⁄ (0.074) 0.207⁄⁄⁄ (0.068) 0.199⁄⁄⁄ (0.062) 0.861⁄⁄⁄ (0.080) 0.320⁄⁄⁄ (0.080) 1.599⁄⁄⁄ (0.081) 2.115⁄⁄⁄ (0.081)
0.246⁄⁄⁄ (0.076) 0.172⁄⁄ (0.068) 0.109⁄ (0.063) 0.644⁄⁄⁄ (0.084) 0.570⁄⁄⁄ (0.084) 1.908⁄⁄⁄ (0.085) 2.457⁄⁄⁄ (0.085)
0.197⁄⁄⁄ (0.075) 0.077 (0.070) 0.435⁄⁄⁄ (0.064)
0.242⁄⁄ (0.099) 0.099 (0.094) 0.413⁄⁄⁄ (0.086)
0.144 (0.114) 0.052 (0.104) 0.462⁄⁄⁄ (0.097)
1.700⁄⁄⁄ (0.093) Yes 116,908
1.340⁄⁄⁄ (0.121) Yes 58,446
1.532⁄⁄⁄ (0.143) Yes 58,462
Constant City controls Observation Note: Robust standard errors in parentheses
Yes 116,908 ***
Yes 58,446
Yes 58,462
p < 0.01, **p < 0.05, *p < 0.1.
Appendix B. Difference in difference analysis of juvenile sleep duration.
Variables
Sleep duration Total
Treatment group Post policy Treatment group (age < 16) Post shutdown policy (year > 2011) Male Grade (ref: 1) 2 3 4
Males
Females
0.025 (0.013) 0.001 (0.017) 0.002 (0.010) 0.403⁄⁄⁄ (0.007)
0.001 (0.019) 0.028 (0.024) 0.032⁄⁄ (0.014)
0.045⁄⁄ (0.019) 0.025 (0.025) 0.023 (0.015)
0.294⁄⁄⁄ (0.012) 0.642⁄⁄⁄ (0.012) 1.637⁄⁄⁄
0.271⁄⁄⁄ (0.016) 0.635⁄⁄⁄ (0.016) 1.677⁄⁄⁄
0.317⁄⁄⁄ (0.017) 0.651⁄⁄⁄ (0.017) 1.594⁄⁄⁄
⁄
1605
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606 Difference in difference analysis of juvenile sleep duration. (continued)
Variables
Sleep duration
5 6 Special high school Father’s educational attainment (ref: Below high school) High school College or above Mother’s educational attainment (ref: Below high school) High school College or above Coresidence with (ref: no parents) Only father Only mother Both parents Constant City controls Observation Adjusted R-squared Note: Robust standard errors in parentheses
***
Total
Males
Females
(0.014) 1.790⁄⁄⁄ (0.020) 2.105⁄⁄⁄ (0.020) 0.585⁄⁄⁄ (0.011)
(0.020) 1.892⁄⁄⁄ (0.028) 2.233⁄⁄⁄ (0.028) 0.612⁄⁄⁄ (0.016)
(0.021) 1.692⁄⁄⁄ (0.028) 1.976⁄⁄⁄ (0.029) 0.555⁄⁄⁄ (0.016)
0.066⁄⁄⁄ (0.012) 0.181⁄⁄⁄ (0.012)
0.044⁄⁄⁄ (0.016) 0.163⁄⁄⁄ (0.017)
0.089⁄⁄⁄ (0.017) 0.199⁄⁄⁄ (0.017)
0.069⁄⁄⁄ (0.011) 0.142⁄⁄⁄ (0.013)
0.069⁄⁄⁄ (0.016) 0.148⁄⁄⁄ (0.017)
0.067⁄⁄⁄ (0.017) 0.138⁄⁄⁄ (0.018)
0.049 (0.032) 0.010 (0.029) 0.025 (0.026) 7.445⁄⁄⁄ (0.035) Yes 116,908 0.340
0.021 (0.044) 0.040 (0.041) 0.027 (0.037) 7.828⁄⁄⁄ (0.048) Yes 58,446 0.360
0.074 (0.046) 0.014 (0.041) 0.022 (0.038) 7.468⁄⁄⁄ (0.051) Yes 58,462 0.294
p < 0.01, **p < 0.05, *p < 0.1.
References Allison, S.E., von Wahlde, L., Shockley, T., Gabbard, G.O., 2006. The development of the self in the era of the internet and role-playing fantasy games. Am. J. Psychiatry 163 (3), 381–385. Battle, P.A.J., 1999. Home computers and school performance. Inform. Soc. 15 (1), 1–10. International Telecommunication Union (2016). Measuring the Information Society Report 2016. Jeong, E.J., Kim, D.J., Lee, D.M., Lee, H.R., 2016. A Study of Digital Game Addiction from Aggression, Loneliness and Depression Perspectives. Paper presented at the System Sciences (HICSS), 2016 49th Hawaii International Conference on. Jin, D.Y., 2016. Mobile gaming in Asia: politics, culture and emerging technologies. Mobile communication in Asia: local insights, global implications. Simon Fraser University: Springer. Kim, S.H., 2016. A study on Shutdown system legal and political of game industry development. Soong-Sil Law Rev. 35, 93–130. Kim, I.H., Kim, J., Kim, M.S., Hong, S.H., 2015a. Status and problems of online game regulations for juvenile protection-centered on the online game shutdown system of korea. KSII Trans. Internet Info. Syst. 9 (4), 1548–1568. Kim, H., Shin, D.-H., Lee, D., 2015b. A socio-technical analysis of software policy in Korea: towards a central role for building ICT ecosystems. Telecomm. Policy 39 (11), 944–956. Korea Institute for Industrial Economics and Trade, 2013. The impact of regulation on the development of industrial ecosystem in gaming market and its implications. KIET Issue Paper, 2013–325. Lee, C., Kim, O., 2017. Predictors of online game addiction among Korean adolescents. Addict. Res. Theory 25 (1), 58–66. Liu, C.-C., Chang, I.-C., 2016. Model of online game addiction: the role of computer-mediated communication motives. Telematics Inform. 33 (4), 904–915. Lucas, K., Sherry, J.L., 2004. Sex differences in video game play: a communication-based explanation. Commun. Res. 31 (5), 499–523. Ministry of Health and Welfare, 2013. User guide for Korea Youth Risk Behavior web-based Survey. Ministry of Science, ICT and Future Planning, 2014. 2014 An investigation of Internet addiction in Korea. Ministry of Science, ICT and Future Planning, 2017. 2016 Survey on the Internet Usage. National Youth Policy Institute, 2014. Korea Children and Youth Panel Survey V: An Analysis on Daily Activities of the Children and the Youth. O’Brien, L.M., 2009. The neurocognitive effects of sleep disruption in children and adolescents. Child Adolesc. Psychiatr. Clin. N. Am. 18 (4), 813–823. Ogletree, S.M., Drake, R., 2007. College students’ video game participation and perceptions: gender differences and implications. Sex Roles 56 (7–8), 537– 542. Park, B.W., Ahn, J.H., 2010. Policy analysis for online game addiction problems. Syst. Dynam. Rev. 26 (2), 117–138. Reigstad, B., Jørgensen, K., Sund, A.M., Wichstrøm, L., 2010. Prevalences and correlates of sleep problems among adolescents in specialty mental health services and in the community: what differs? Nord. J. Psychiatr. 64 (3), 172–180. Sang, Y., Park, S., Seo, H., 2017. Mobile Game Regulation in South Korea: A Case Study of the Shutdown Law Mobile Gaming in Asia. Springer, pp. 55-72.
1606
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
Turow, J., Lilach, N., 2000. The Internet and the Family 2000: The View From Parents, the View From Kids. From Report Series. No. 13, The Annenberg Public Policy Center of the University of Pennsylvania. Vedung, E., 1997. Public policy and program evaluation: Transaction Publishers. Wang, L., Luo, J., Gao, W., Kong, J., 2012. The effect of Internet use on adolescents’ lifestyles: a national survey. Comput. Hum. Behav. 28 (6), 2007–2013. Wenzel, H., Bakken, I., Johansson, A., Götestam, K., Øren, A., 2009. Excessive computer game playing among Norwegian adults: self-reported consequences of playing and association with mental health problems. Psychol. Rep. 105, 1237–1247. 3_suppl. Wight, V.R., Price, J., Bianchi, S.M., Hunt, B.R., 2009. The time use of teenagers. Soc. Sci. Res. 38 (4), 792–809.
Contents lists available at ScienceDirect
Telematics and Informatics journal homepage: www.elsevier.com/locate/tele
Ex-post evaluation of illegalizing juvenile online game after midnight: A case of shutdown policy in South Korea Changjun Lee a, Hongbum Kim b, Ahreum Hong c,⇑ a
Lee Kuan Yew School of Public Policy in National University of Singapore, Singapore Korea National Industrial Convergence Center, Korea Institute of Industrial Technology, South Korea c Graduate School of Technology Management, Kyung Hee University, 1732 Deokyoung-ro Giheung-gu, Youngin-Si Gyunggi-Do, South Korea b
a r t i c l e
i n f o
Article history: Received 15 June 2017 Received in revised form 10 July 2017 Accepted 11 July 2017 Available online 14 July 2017 Keywords: Media policy Internet hours Sleep duration Policy evaluation Youth policy
a b s t r a c t In November 2011, the Korean government legalized blocking access to online games for youths younger than age 16 late at night; this is called the shutdown policy. Using multiple regressions we examined how the compulsory block affected youths’ Internet hours and sleep duration. Data were drawn from the 2011, 2012 Korea Youth Behavior Risk Factor Survey, a cross-sectional online survey of middle and high school students aged 13–18 years. Legalizing a ban of online gaming late at night for youths caused an increase in the predicted probability of being in a high-ranked Internet user group by 1.6 percent points, a decrease in the predicted probability of Internet addiction by 0.7 percent points, and an increase in sleep duration of 1.5 min. All results showed a gender difference in the effect of the policy. Although the net effect of the shutdown policy was statistically significant, the small effect size, the partial effect on female youths, and the side effects related to human basic rights and inappropriate regulation of the game industry made the effectiveness of the policy arguable. Ó 2017 Elsevier Ltd. All rights reserved.
1. Introduction The Internet is now a part of most people’s daily lives. In 2016, the rate of Internet accessibility in households reached a high proportion in East Asian countries (96.5% in Japan, 89.5% in Singapore, and 98.8% in South Korea; hereafter, Korea) as well as in Western countries (91.3% in England, 82.2% in the United States, and 90% in Germany) (International Telecommunication Union, 2016; Ministry of Science, ICT and Future Planning, 2017). While many people can access the Internet easily by many means, the rate of overdependence on the Internet, especially using smartphones, has been increasing. According to a national survey, the rate of people in the danger zone of overdependence on the Internet has been increased by 9.4 percent points, from 8.4% in 2011 to 17.8% in 2016 (Ministry of Science, ICT and Future Planning, 2017). Among the people in the danger zone, teenagers were at 30.6%, ranking as the highest proportion. Usage of computers and the Internet has many benefits. It helps workers build required skill sets for jobs (Turow and Lilach, 2000), and students get higher scores in math and reading given appropriate time use (Battle, 1999). However, excessive use is negatively associated with physical and psychological well-being, especially for youths (Wight et al., 2009). The excessive online time also correlates to poor lifestyle habits in youths (Wang et al., 2012).
⇑ Corresponding author. E-mail addresses: [email protected] (C. Lee), [email protected] (H. Kim), [email protected] (A. Hong). http://dx.doi.org/10.1016/j.tele.2017.07.006 0736-5853/Ó 2017 Elsevier Ltd. All rights reserved.
1598
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
Among the side effects of excessive Internet usage, the lack of sleep duration raises a serious concern for youths, and this is mostly caused by online gaming at night because late-night playing of online games inevitably crowds out youths’ time for adequate sleep. Sleep duration of youths in Korea amounts to 31% of the day, which is less than the 33–37% per day in Japan and the United States (National Youth Policy Institute, 2014). Youths’ lack of sleep duration causes behavioral, cognitive, and mood impairments (O’Brien, 2009) as well as depression and anxiety (Reigstad et al., 2010). To protect youths from playing online games excessively and to increase their sleep duration, in November 2011 the Korean government legalized the blocking of access to online games for users below 16 years of age late at night, the so-called shutdown policy. Instead of the autonomous regulation applied in most countries such as the United States, European Union, and Japan, only a few countries like Korea, China, Thailand, and Vietnam chose compulsory ways such as the shutdown system (Kim et al., 2015a; Park and Ahn, 2010). Even if the shutdown policy can encourage youth to sleep more by prohibiting online gaming at night, scholars have argued that forbidding Internet games at night is not the ultimate solution to alleviate Internet or online game addiction or lack of sleep duration (e.g., Jeong et al., 2016; Jin, 2016; Lee and Kim, 2017). The compulsory regulations like the shutdown policy have been shown not to be a solution for relieving game addiction (Jeong et al., 2016). According to Lee and Kim (2017), the policy should not be legitimized as youths also have their right to do gaming at night even if online game activity late at night is correlated with game addiction. Moreover, there are many ways for youths to avoid the regulation. For example, they can route through an overseas server or use their parents’ account to game late at night (Jin, 2016). In other aspects, some scholars have argued that the policy could hinder the growth of game and software industries due to unnecessary intervention on the demand side (Kim et al., 2015b; Sang et al., 2017). Because of such negative aspects, a shutdown policy is not widely implemented worldwide, so related studies examining the policy’s effect itself are not prevalent. The Korea Institute for Industrial Economics and Trade (2013) is the only literature addressing the effect of the shutdown policy empirically, to the best of the authors’ knowledge. However, it has some limitations such as small observations and unbalanced samples, less various of estimation, and limited focus on the gender effect, which is one of the primary factors affecting online gaming behavior in youth groups (Lucas and Sherry, 2004; Ogletree and Drake, 2007). To fill this research gap, this research aims to examine whether or not the shutdown policy actually affected a reduction in juvenile Internet hours and an increase in juvenile sleep duration, and, if it did, to estimate the size of the effect. Evaluating policy is always important even if the policy is abolished because policy makers may refer to the evaluation in many aspects when they need to build a policy in similar situations (Vedung, 1997). Therefore, this study will provide empirical evidence for political debates for future practical actions. In the rest of the paper, we firstly introduce the shutdown policy in details and review literature on the policy. Next, we define the key independent and the outcome variables, then using the difference in difference approach, examine how the shutdown policy effect on juvenile sleep duration and Internet hours. In result and analysis section, we show the significance and the size of the effect, and lastly, we discuss the results with regard to the previous discourse and gender difference to conclude a policy evaluation of legalizing a ban of online gaming late at night for youths. 2. Background 2.1. Shutdown policy The shutdown policy aimed to protect Korean youth from playing online games excessively, which can lead to addiction to the games, violence and a lack of social activities. For the implementation of shutdown policy, Juvenile Protection Act was amended in May 2011, initiated by Ministry of Gender Equality and Family. Newly added Article – Restriction on Hours Provided for Internet Games in Late Night Time – became the legal basis of the regulation. The policy protects by blocking access to online games for those under 16 from midnight to 6 a.m. However, due to the lack of ex-ante evaluation of the policy, it has caused many of controversies (Kim et al., 2015a; Sang et al., 2017). First, the shutdown policy is targeting only games made in Korea. Second, the system violates the constitutional principle of family autonomy and infringes upon parents’ right to educate and foster their children as they see fit. Finally, government authorities do not distinguish between online games and other games with the online game shutdown policy. In 2016, another government department, Ministry of Culture, Sports and Tourism announced a plan to abolish the shutdown policy because the firms in the gaming industry insisted that the policy was a kind of the double regulation added on to the existing policy, the Protection of Adolescents Act (Kim, 2016). In addition, the validity of the shutdown policy has yet to become clear. Youths could borrow identities of adults and play games if they were eager to do so. Also, there are other ways to reduce the time of playing online games even if there is a way to stop playing online games from midnight to 6 a.m. 2.2. Previous studies of the shutdown policy Because online game addiction and its negative effects are frequently discussed (Allison et al., 2006; Liu and Chang, 2016; Wenzel et al., 2009), many countries are implementing appropriate regulation of online game usage. In the United States, European Union, and Japan, the game industry is restricting their services by autonomous regulation, thus promoting the
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
1599
online gaming industry as well as securing legitimacy. In contrast, Korea, China, Thailand, and Vietnam have implemented compulsory shutdown systems (Kim et al., 2015a; Park and Ahn, 2010). Because such shutdown systems’ law and policy are not prevalent worldwide, existing studies examining the effect of the shutdown law and policy are not easily found. Even many domestic studies regarding the shutdown law in Korea are dealing with the legality of the law. Some previous studies addressed questions regarding the effect of a shutdown policy. Jeong et al. (2016) showed that compulsory regulation is not a solution for relieving game addiction. Lee and Kim (2017) emphasized that although online game activity late at night is correlated with game addiction, this cannot legitimize the existence of the shutdown system. By using an analytical framework regarding modalities of regulation, Sang et al. (2017) examined the regulatory framework of the shutdown law. After a case analysis, they noted problems regarding not only basic human rights, but also the game industry context. Specifically, the failure of the age verification system was not acceptable to the regulatory framework, meaning that the policy should be fully reconsidered. Kim et al. (2015b) also stated that two different online game policies, promotion policy and regulation policy, conflicted with each other, causing the game industry to be confused. Although those studies dealt with the shutdown policy and its contextual implications, empirical evaluations on the effect of the policy are missing in the current literature. To the authors’ best knowledge, the Korea Institute for Industrial Economics and Trade (2013) alone evaluated the effect of the shutdown law empirically. The study showed that although the policy was effective in terms of securing juveniles’ right to sleep, the playing time of online games increased. This implies the essential objective of the shutdown policy for alleviating gaming addiction was not achieved. As an alternative, Park and Ahn (2010) proposed two different regulation approaches to solve the online game addiction problem and evaluate them with a system dynamics approach: a self-regulation policy and a tax and rebate policy. Their results showed that a tax and rebate policy is more effective in terms of sales growth, improvement of social image, and especially decrease of game user addiction. Scholars have also focused on gender differences in the time consumption of youths. In terms of time consumption for gaming, they found that male youths spent much more time online gaming than female youths. Ogletree and Drake (2007) found that male youths played video games two or more hours a week than female youths, so male students were more likely to experience sleeping and class preparation interference due to playing games. Lucas and Sherry (2004) also found a gender difference in the time spent video gaming. They showed that female respondents reported less frequent play and less motivation to play in social situations in a survey of youths. Those studies hint that the effects of policy might be different by gender. While some discussions on the shutdown policy have been carried out, the effect of the shutdown policy on juvenile Internet usage and sleep time has not been critically or empirically assessed. In light of this research gap, this study evaluated the effect of the shutdown policy on Internet usage time and sleep duration in Korean youths.
3. Method 3.1. Data and analysis sample We used data drawn from the 2011–2012 Korea Youth Risk Behavior Web-based Survey (KYRBS), an annual crosssectional online survey of middle and high school students aged 13–18 years. KYRBS collects data on health behavior in the form of an anonymous and self-reported survey. To minimize sampling error, KYRBS divides the population into 129 strata by using 43 regional and school classes as strata variables with regard to geographical accessibility, the number of schools and population, the life environment, and the rate of smoking and drinking (Ministry of Health and Welfare, 2013). The data set enables researchers to determine whether a respondent was subject to the shutdown policy in a given year because it contains detailed information on the exact age and date of an interview each year. This study constructed the analytic sample in the following ways. It restricted the sample to respondents who were 13– 18 years old in each wave. For the purpose of this study to examine the marginal impact of the policy on sleep and Internet hours for juveniles, the sample consisted of students who participated in 2011 and 2012 to exclude respondents who were already adjusted to the policy after 2012.
3.2. Key independent variables The study included three key independent binary variables. It determined the value of each variable in each year considering the exact age and date of the interview in the KYRBS. The first one was a treatment variable, which indicated whether a respondent was eligible for the shutdown policy or not. Specifically, the variable was equal to 1 if the age of a respondent was below 16 and 0 if he or she were not eligible for the policy each year. The second was a post policy variable, which indicated whether the time a respondent surveyed had passed during the time the shutdown policy applied or not. The variable was equal to 1 if the exact surveyed date was after November 2011, and 0 otherwise. The third was the interaction term multiplying the first and second variables, which indicated the genuine difference due to the policy considering the existing difference between the treatment and the control groups.
1600
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
3.3. Outcome variables This study examined two outcome variables for juveniles’ health behavior. First, we analyzed how long respondents slept on average in a year. We only used the data from 2011 to 2012 to avoid measurement error because the KRYBS has asked about sleep hours separately on weekdays and weekends since 2013. Fig. 1 shows the distribution of juvenile sleep duration by gender. The distribution is normal for being a dependent variable in an ordinary least squares (OLS) regression. The second outcome variable was Internet hours, how long respondents used the Internet on average by purpose and days of the week. Purposes consisted of two categories: for study and for other activities. We used only Internet hours not for the purpose of study because Internet hours for studying were not counted in the shutdown policy. To test the impact of the shutdown policy, we created an extreme variable, Internet addiction, which indicated whether a respondent belonged to the Internet addicted or alarming group or not. The variable was equal to 1 if a respondent’s Internet hours were over 3.15 h except hours used for study purposes and 0 if below 3.15 h according to the result of an investigation of Internet addiction in Korea showing that the average Internet hours of the addicted alarming users and the addicted users were 3.15 and 3.75 h a day, respectively (Ministry of Science, ICT and Future Planning, 2014). Fig. 2 shows the distribution of juvenile Internet hours only including hours not for study purposes by gender. Because the distribution of Internet hours did not follow a normal distribution, to conduct ordered logit and logit regression we categorized Internet hours into five groups: (1) light user, someone who used the Internet for 0–1 h, (2) normal user, someone who used the Internet for 1–2 h, (3) heavy user, someone who used the Internet for 2–3.15 h, (4) addicted alarming user, someone who used the Internet for 3.15– 3.75 h, and (5) addicted user, someone who used the Internet for 3.75 or more hours. Female youths comprised a larger proportion of the light user group than males, whereas males comprised a larger proportion of the heavy user group and groups four and five than females. 3.4. Research design To estimate the effect of the shutdown policy on respondents’ sleep and Internet hours, we conducted multivariate regression analysis in a difference research design. This design used OLS regression for the continuous dependent variable of sleep duration. It used ordered logit regression for the ordered categorical variable, Internet hours, and logit regression for the binary variable, Internet addicted. The equation below shows the estimation model. In the equation, yi was sleep duration, Internet hours, and Internet addicted; i indexed each respondent, and t indexed each year.
yi ¼ a þ b1Treati þ b2Posti þ b3ðTreati Posti Þ þ b4X i þ b5City:FEsi þ ei
ð1Þ
0
.1
Density
.2
.3
The key independent variable was the interaction term, which was obtained by multiplying the treatment group and post policy. Then the coefficient, b3, captured the effect of the shutdown policy per se on the outcomes. We also included various covariates, denoted by the vector X i . We controlled for additional exogenous outcome-related characteristics, such as grades, the class of schools, parents’ educational attainment, and living arrangements (categorized into no parent alive, living with
.5
2.8
5 7.3 Sleep duration Female youths
9.5 Male youths
Fig. 1. Distribution of juvenile sleep duration by gender.
11.75
1601
0
.2
Density
.4
.6
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
0
2
4 6 Internet hours Female youths
8
10
Male youths
Fig. 2. Distribution of juvenile Internet hours excluding hours for a purpose on study by gender.
parents, only father, and only mother) at each year. In order to capture city- or region-specific trends in the dependent variables, in all models we included city fixed effects, denoted by City.FEs, a vector consisting of dummy variables for each year (12 cities and regions are controlled). We also conducted subgroup analysis to see whether effects varied by gender. 4. Results 4.1. Descriptive results Table 1 shows descriptive statistics for variables in this study. The mean value of Internet hours excluding hours for the purpose of study was 1.53 h with a standard deviation (SD) of 1.52. The mean value of sleep duration was 6.39 h with an SD of 1.39. When comparing male and female samples, it was clear males had longer Internet hours (1.68 h with an SD of 1.61 h in the male sample and 1.38 h with an SD of 1.40 h in the female sample), as well as sleep duration (6.60 h with an SD of 1.39 h in the male sample and 6.18 h with an SD of 1.36 h in the female sample). The ratio of the treatment group whose age was below 16 at each wave of the survey was 0.58 and the post-shutdown policy, which indicated whether or not the survey dates were later than the date the shutdown policy was applied, was 0.50. The ratio of the interaction term between the treatment group and the post-shutdown policy was 0.30. 4.2. Results of regression analysis Table 2 reports the result of the multiple regression analyses of juvenile Internet hours and sleep duration. For all outcome variables, we report regression coefficients of the main variable, which indicates the difference in the difference effect of the shutdown policy (that is, estimated effects of the policy per se), although the aforementioned other variables are also controlled in the regressions (full regression results are shown in the Appendix A and B). For each outcome variable, the table shows results from models with total, male and female samples. We present OLS regression coefficients for the continuous dependent variables and marginal effects (with all other variables are assumed to be at their mean values) for ordered categorical and binary dependent variables. The first and second rows of Table 2 show the effect of the shutdown policy on the juvenile Internet hours and Internet addiction sleep duration, respectively (for details, see Appendix A and B). The shutdown policy increased the predicted probability of being a high-ranked Internet user group by 1.6 percent points (b = 0. 016, p-value < 0.01). This had a significant impact on female (b = 0. 030, p-value < 0.01), however, with no significant impact on male. When it came to the effect of the shutdown policy on Internet addiction, it decreased the predicted probability of addiction by 0.7 percent points (b = -0.007, p-value < 0.1). It also had no significant impact on male but decreased the probability of being in the Internet-addicted group by 1 percent point at 5% significance level in the female’s sample. The third row of Table 2 shows the effect of the shutdown policy on juvenile sleep duration, which increased 0.025 h, which equals 1.5 min (b = 0.025,
1602
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
Table 1 Descriptive statistics of key variables. Variables
Mean (SD)
Internet hours (except a purpose of study) Internet hours (in five categories) (1) 0–1 h (Light user) (2) 1–2 h (Normal user) (3) 2–3.15 h (Heavy user) (4) 3.15–3.75 h (Addicted alarming user) (5) 3.75 or above hrs (Addicted user) Sleep duration Treatment group (age < 16) Post shutdown policy (year > 2011) Treatment group Post policy Grade 1 2 3 4 5 6 Special high school Father’s educational level Below high school High school College or above Mother’s educational level Below high school High school College or above Coresidence with Both Only father Only mother No one
Total (N = 116,908)
Male youths (N = 58,446)
Female youths (N = 58,462)
1.53 (1.52)
1.68 (1.61)
1.38 (1.40)
0.41 0.27 0.20 0.04 0.07 6.39 (1.39) 0.58 0.50 0.30
0.37 0.28 0.21 0.05 0.09 6.60 (1.39) 0.58 0.50 0.30
0.45 0.27 0.18 0.04 0.06 6.18 (1.36) 0.57 0.49 0.29
0.16 0.16 0.17 0.17 0.17 0.17 0.11
0.16 0.17 0.17 0.17 0.16 0.16 0.11
0.16 0.16 0.17 0.17 0.17 0.17 0.12
0.17 0.36 0.47
0.18 0.35 0.47
0.16 0.38 0.46
0.17 0.47 0.36
0.19 0.45 0.36
0.16 0.49 0.35
0.88 0.03 0.07 0.02
0.88 0.03 0.07 0.02
0.88 0.03 0.07 0.02
Notes: The analytic sample is restricted to respondents who were 13–18 years old in each wave. Grade 1–3 represent middle school students and 4–6 represent high school students in Korea. Special high school includes special-purpose high school such as science- or language-specialized schools.
Table 2 Regression estimates on the effect of the shut-down policy on Juveniles’ Internet hours and sleep duration.
Internet hours (in five ordered categories) Ordered logit coefficients (S.E.) Internet addicted Logit marginal coefficients (S.E.) Sleep duration (in hours) OLS regression coefficients (S.E.) Note: Robust standard errors in parentheses
***
Total (N = 116,908)
Male youths(N = 58,446)
Female youths (N = 58,462)
0.016*** (0.005)
0.002 (0.007)
0.030*** (0.008)
0.007* (0.004)
0.001 (0.006)
0.010** (0.005)
0.025* (0.013)
0.001 (0.019)
0.045** (0.019)
p < 0.01, **p < 0.05, *p < 0.1.
p-value < 0.1). The total impact was shown to increase the sleep duration owing to the effect of the female’s sample. The policy increased the female’s sleep duration by 0.045 h, which equals 2.7 min with a more statistically significant p-value, <0.05. 5. Concluding remarks In this paper, we examined the effect of the shutdown policy on juvenile Internet use hours and sleep duration. Firstly, the results showed that the shutdown policy increased juvenile Internet hours, instead of reducing them, which was supposed to be one of the goals of the policy. However, the policy did reduce the possibility of juvenile Internet addiction. It decreased the
1603
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
predicted probability of Internet addiction by a 0.7 percentage points but was only significant in the female’s sample. As far as the sleep duration is concerned, the policy increased sleep duration 1.5 min in the total sample, and 2.7 min in the female youths’ sample. There was no statistically significant result when conducted only with the male’s sample. The policy was shown to reach the original goals when focusing on only the directions of the coefficients. It decreased the rate of Internet addiction and increased sleep duration. However, even if the coefficients were significant, they were too small to regard them as an effect of the policy (0.7 percent points decrease of the probability of Internet addiction and 1.5 min increase of sleep duration). Those are not large enough to offset the other issues related to the side effects of the policy related to basic human rights and inappropriate regulation for the game industry. The partial effect of the shutdown policy was also considered as a limitation of the policy. It only affected female youths but had no effect on male youths at all. This confirmed the existence of a gender difference in gaming behavior due to different life behaviors with the same results as in Ogletree and Drake (2007) and Lucas and Sherry (2004). As the descriptive statistics show that they are different in daily life including sleep duration and Internet hours (see Table 1), male youths have longer Internet hours and longer sleep duration than female youths. Hence, we assume that male youths are more likely to become addicted to online gaming. Moreover, it might be possible that male youths adapted to the policy faster than female youths, thereby having a higher chance to find a way to avoid the policy. In short, the results of this study provide evidence regarding the effect of the shutdown policy itself. According to the existing empirical literature on the shutdown policy conducted by the Korea Institute for Industrial Economics and Trade (2013), while the shutdown policy was partly effective on juveniles’ game-playing time, the effect of the policy per se was not statistically significant. The analysis of the current study, however, found the significant net effect of the shutdown policy. Yet, with considering the small and partial effects of the policy, the efficacy of the policy remains questionable. Such results can support the arguments of previous literature addressing the potential problems of the shutdown policy. This paper has several limitations. First, this study used Internet hours as a proxy variable for online gaming hours, which could cause a measurement error in estimation. Second, there is limited information on the exact time of using Internet. This hindered us from examining the effect of the shutdown policy on Internet hours late at night. Despite these limitations, this study highlights valuable policy implications for policy makers who are considering a way of regulating juvenile online gaming late at night. The current study based on the exogenous policy change in Korea suggests that the compulsory way of regulating might be an effective policy tool to encourage youths to have enough sleep duration, but it only effects female youths. This is not enough to justify the infringement of fundamental human rights and the criticism that the regulation might violate fair competition in the gaming industry. Appendix A. Difference in difference analysis of juvenile Internet hours.
Variables
Treatment group Post policy Treatment group (age < 16) Post shutdown policy (year > 2011) Male Grade (ref: 1) 2 3 4 5 6 Special high school
Internet hours(in five ranked orders)
Internet addicted
Total
Males
Females
Total
Males
Females
0.065⁄⁄⁄ (0.022) 0.009 (0.028) 0.412⁄⁄⁄ (0.017) 0.065⁄⁄⁄ (0.011)
0.007 (0.031) 0.010 (0.039) 0.292⁄⁄⁄ (0.024) 0.007
0.121⁄⁄⁄ (0.031) 0.018 (0.040) 0.545⁄⁄⁄ (0.024) 0.121⁄⁄⁄
0.073⁄ (0.039) 0.033 (0.049) 0.305⁄⁄⁄ (0.030) 0.073⁄ (0.019)
0.007 (0.050) 0.077 (0.064) 0.269⁄⁄⁄ (0.039) 0.007
0.125⁄⁄ (0.061) 0.002 (0.075) 0.383⁄⁄⁄ (0.049) 0.125⁄⁄
0.347⁄⁄⁄ (0.019) 0.531⁄⁄⁄ (0.019) 0.003 (0.024) 0.116⁄⁄⁄ (0.033) 0.020 (0.033) 0.531⁄⁄⁄ (0.019)
0.332⁄⁄⁄ (0.026) 0.562⁄⁄⁄ (0.026) 0.022 (0.033) 0.153⁄⁄⁄ (0.045) 0.022 (0.045) 0.663⁄⁄⁄ (0.027)
0.358⁄⁄⁄ (0.027) 0.498⁄⁄⁄ (0.027) 0.022 (0.034) 0.064 (0.047) 0.075 (0.047) 0.405⁄⁄⁄ (0.026)
0.308⁄⁄⁄ (0.033) 0.457⁄⁄⁄ (0.033) 0.133⁄⁄⁄ (0.044) 0.111⁄ (0.058) 0.217⁄⁄⁄ (0.059) 0.789⁄⁄⁄ (0.029)
0.289⁄⁄⁄ (0.045) 0.516⁄⁄⁄ (0.043) 0.089 (0.057) 0.006 (0.076) 0.072 (0.077) 0.852⁄⁄⁄ (0.038)
0.321⁄⁄⁄ (0.050) 0.376⁄⁄⁄ (0.049) 0.202⁄⁄⁄ (0.067) 0.278⁄⁄⁄ (0.091) 0.436⁄⁄⁄ (0.092) 0.714⁄⁄⁄ (0.045)
(continued on next page)
1604
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
Difference in difference analysis of juvenile Internet hours. (continued)
Variables
Father’s educational attainment (ref: Below high school) High school College or above Mother’s educational attainment (ref: Below high school) High school College or above Coresidence with (ref: no parents) Only father Only mother Both parents Constant cut 1 Constant cut 2 Constant cut 3 Constant cut 4
Internet hours(in five ranked orders)
Internet addicted
Total
Males
Females
Total
Males
Females
0.028 (0.019) 0.211⁄⁄⁄ (0.020)
0.005 (0.026) 0.262⁄⁄⁄ (0.027)
0.060⁄⁄ (0.027) 0.157⁄⁄⁄ (0.028)
0.085⁄⁄⁄ (0.030) 0.371⁄⁄⁄ (0.032)
0.080⁄⁄ (0.039) 0.424⁄⁄⁄ (0.042)
0.087⁄ (0.047) 0.293⁄⁄⁄ (0.050)
0.080⁄⁄⁄ (0.019) 0.233⁄⁄⁄ (0.020)
0.078⁄⁄⁄ (0.025) 0.264⁄⁄⁄ (0.028)
0.078⁄⁄⁄ (0.027) 0.198⁄⁄⁄ (0.030)
0.088⁄⁄⁄ (0.030) 0.259⁄⁄⁄ (0.034)
0.072⁄ (0.039) 0.229⁄⁄⁄ (0.044)
0.122⁄⁄⁄ (0.046) 0.308⁄⁄⁄ (0.054)
0.320⁄⁄⁄ (0.053) 0.184⁄⁄⁄ (0.048) 0.153⁄⁄⁄ (0.044) 0.554⁄⁄⁄ (0.058) 0.638⁄⁄⁄ (0.058) 1.939⁄⁄⁄ (0.059) 2.467⁄⁄⁄ (0.059)
0.380⁄⁄⁄ (0.074) 0.207⁄⁄⁄ (0.068) 0.199⁄⁄⁄ (0.062) 0.861⁄⁄⁄ (0.080) 0.320⁄⁄⁄ (0.080) 1.599⁄⁄⁄ (0.081) 2.115⁄⁄⁄ (0.081)
0.246⁄⁄⁄ (0.076) 0.172⁄⁄ (0.068) 0.109⁄ (0.063) 0.644⁄⁄⁄ (0.084) 0.570⁄⁄⁄ (0.084) 1.908⁄⁄⁄ (0.085) 2.457⁄⁄⁄ (0.085)
0.197⁄⁄⁄ (0.075) 0.077 (0.070) 0.435⁄⁄⁄ (0.064)
0.242⁄⁄ (0.099) 0.099 (0.094) 0.413⁄⁄⁄ (0.086)
0.144 (0.114) 0.052 (0.104) 0.462⁄⁄⁄ (0.097)
1.700⁄⁄⁄ (0.093) Yes 116,908
1.340⁄⁄⁄ (0.121) Yes 58,446
1.532⁄⁄⁄ (0.143) Yes 58,462
Constant City controls Observation Note: Robust standard errors in parentheses
Yes 116,908 ***
Yes 58,446
Yes 58,462
p < 0.01, **p < 0.05, *p < 0.1.
Appendix B. Difference in difference analysis of juvenile sleep duration.
Variables
Sleep duration Total
Treatment group Post policy Treatment group (age < 16) Post shutdown policy (year > 2011) Male Grade (ref: 1) 2 3 4
Males
Females
0.025 (0.013) 0.001 (0.017) 0.002 (0.010) 0.403⁄⁄⁄ (0.007)
0.001 (0.019) 0.028 (0.024) 0.032⁄⁄ (0.014)
0.045⁄⁄ (0.019) 0.025 (0.025) 0.023 (0.015)
0.294⁄⁄⁄ (0.012) 0.642⁄⁄⁄ (0.012) 1.637⁄⁄⁄
0.271⁄⁄⁄ (0.016) 0.635⁄⁄⁄ (0.016) 1.677⁄⁄⁄
0.317⁄⁄⁄ (0.017) 0.651⁄⁄⁄ (0.017) 1.594⁄⁄⁄
⁄
1605
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606 Difference in difference analysis of juvenile sleep duration. (continued)
Variables
Sleep duration
5 6 Special high school Father’s educational attainment (ref: Below high school) High school College or above Mother’s educational attainment (ref: Below high school) High school College or above Coresidence with (ref: no parents) Only father Only mother Both parents Constant City controls Observation Adjusted R-squared Note: Robust standard errors in parentheses
***
Total
Males
Females
(0.014) 1.790⁄⁄⁄ (0.020) 2.105⁄⁄⁄ (0.020) 0.585⁄⁄⁄ (0.011)
(0.020) 1.892⁄⁄⁄ (0.028) 2.233⁄⁄⁄ (0.028) 0.612⁄⁄⁄ (0.016)
(0.021) 1.692⁄⁄⁄ (0.028) 1.976⁄⁄⁄ (0.029) 0.555⁄⁄⁄ (0.016)
0.066⁄⁄⁄ (0.012) 0.181⁄⁄⁄ (0.012)
0.044⁄⁄⁄ (0.016) 0.163⁄⁄⁄ (0.017)
0.089⁄⁄⁄ (0.017) 0.199⁄⁄⁄ (0.017)
0.069⁄⁄⁄ (0.011) 0.142⁄⁄⁄ (0.013)
0.069⁄⁄⁄ (0.016) 0.148⁄⁄⁄ (0.017)
0.067⁄⁄⁄ (0.017) 0.138⁄⁄⁄ (0.018)
0.049 (0.032) 0.010 (0.029) 0.025 (0.026) 7.445⁄⁄⁄ (0.035) Yes 116,908 0.340
0.021 (0.044) 0.040 (0.041) 0.027 (0.037) 7.828⁄⁄⁄ (0.048) Yes 58,446 0.360
0.074 (0.046) 0.014 (0.041) 0.022 (0.038) 7.468⁄⁄⁄ (0.051) Yes 58,462 0.294
p < 0.01, **p < 0.05, *p < 0.1.
References Allison, S.E., von Wahlde, L., Shockley, T., Gabbard, G.O., 2006. The development of the self in the era of the internet and role-playing fantasy games. Am. J. Psychiatry 163 (3), 381–385. Battle, P.A.J., 1999. Home computers and school performance. Inform. Soc. 15 (1), 1–10. International Telecommunication Union (2016). Measuring the Information Society Report 2016. Jeong, E.J., Kim, D.J., Lee, D.M., Lee, H.R., 2016. A Study of Digital Game Addiction from Aggression, Loneliness and Depression Perspectives. Paper presented at the System Sciences (HICSS), 2016 49th Hawaii International Conference on. Jin, D.Y., 2016. Mobile gaming in Asia: politics, culture and emerging technologies. Mobile communication in Asia: local insights, global implications. Simon Fraser University: Springer. Kim, S.H., 2016. A study on Shutdown system legal and political of game industry development. Soong-Sil Law Rev. 35, 93–130. Kim, I.H., Kim, J., Kim, M.S., Hong, S.H., 2015a. Status and problems of online game regulations for juvenile protection-centered on the online game shutdown system of korea. KSII Trans. Internet Info. Syst. 9 (4), 1548–1568. Kim, H., Shin, D.-H., Lee, D., 2015b. A socio-technical analysis of software policy in Korea: towards a central role for building ICT ecosystems. Telecomm. Policy 39 (11), 944–956. Korea Institute for Industrial Economics and Trade, 2013. The impact of regulation on the development of industrial ecosystem in gaming market and its implications. KIET Issue Paper, 2013–325. Lee, C., Kim, O., 2017. Predictors of online game addiction among Korean adolescents. Addict. Res. Theory 25 (1), 58–66. Liu, C.-C., Chang, I.-C., 2016. Model of online game addiction: the role of computer-mediated communication motives. Telematics Inform. 33 (4), 904–915. Lucas, K., Sherry, J.L., 2004. Sex differences in video game play: a communication-based explanation. Commun. Res. 31 (5), 499–523. Ministry of Health and Welfare, 2013. User guide for Korea Youth Risk Behavior web-based Survey. Ministry of Science, ICT and Future Planning, 2014. 2014 An investigation of Internet addiction in Korea. Ministry of Science, ICT and Future Planning, 2017. 2016 Survey on the Internet Usage. National Youth Policy Institute, 2014. Korea Children and Youth Panel Survey V: An Analysis on Daily Activities of the Children and the Youth. O’Brien, L.M., 2009. The neurocognitive effects of sleep disruption in children and adolescents. Child Adolesc. Psychiatr. Clin. N. Am. 18 (4), 813–823. Ogletree, S.M., Drake, R., 2007. College students’ video game participation and perceptions: gender differences and implications. Sex Roles 56 (7–8), 537– 542. Park, B.W., Ahn, J.H., 2010. Policy analysis for online game addiction problems. Syst. Dynam. Rev. 26 (2), 117–138. Reigstad, B., Jørgensen, K., Sund, A.M., Wichstrøm, L., 2010. Prevalences and correlates of sleep problems among adolescents in specialty mental health services and in the community: what differs? Nord. J. Psychiatr. 64 (3), 172–180. Sang, Y., Park, S., Seo, H., 2017. Mobile Game Regulation in South Korea: A Case Study of the Shutdown Law Mobile Gaming in Asia. Springer, pp. 55-72.
1606
C. Lee et al. / Telematics and Informatics 34 (2017) 1597–1606
Turow, J., Lilach, N., 2000. The Internet and the Family 2000: The View From Parents, the View From Kids. From Report Series. No. 13, The Annenberg Public Policy Center of the University of Pennsylvania. Vedung, E., 1997. Public policy and program evaluation: Transaction Publishers. Wang, L., Luo, J., Gao, W., Kong, J., 2012. The effect of Internet use on adolescents’ lifestyles: a national survey. Comput. Hum. Behav. 28 (6), 2007–2013. Wenzel, H., Bakken, I., Johansson, A., Götestam, K., Øren, A., 2009. Excessive computer game playing among Norwegian adults: self-reported consequences of playing and association with mental health problems. Psychol. Rep. 105, 1237–1247. 3_suppl. Wight, V.R., Price, J., Bianchi, S.M., Hunt, B.R., 2009. The time use of teenagers. Soc. Sci. Res. 38 (4), 792–809.