Underlying Structure of Online Risks and Harm among Bangladeshi Teenagers S

7 http://journal.uinjkt.ac.id/index.php/aism Abstract—This paper gives an account of the consequences of the procedure of EFA (exploratory factor analysis) invoking online risks as well as harm information acquired by an overview of 443 adolescents in Bangladesh. The information was gathered utilizing a 42-item, adjusted Likert survey investigating the critical variables. PCA (Principal Component Analysis) with Varimax rotation was chosen by the researcher to accomplish it. Based on data, the procedure took out the crucial factors social, religiosity, psychological, online risks as well as harm. 65.594% of the variance was clarified by these nine dimensions together. The procedure of reliability analysis generated internal consistency estimates which may be considered acceptable. The range was found from 0.625 which belongs to Emotional Problems to 0.930 which belongs to Online Risks. These discoveries give comprehensive justification to build legitimacy for the items. The presence of the components influencing altogether the young people online in Bangladesh has been identified as well.


I. INTRODUCTION
ince 1980s Internet has been influencing the lifestyle of society. From that period, it has been dominating the whole world. Through its dominance not only a lot of positive but also negative impacts have been observed in the society as well. One of the mentionable blessings of the internet is acting as a gateway to knowledge. No one can deny its role as a facilitator of the process of learning and education. But, the negative impacts of the internet are also noteworthy. Among all age cohort groups of the people, teenagers are the most primitive as well as passionate users of the internet. As a consequence, both negative and positive impacts affect them most. They were found using the internet for learning purposes, entertainment, meeting with friends, online games which can be taken as productive uses of it. But, on the other hand, they were found abusing and harassing others, watching pornography, sexting, etc. These can be taken as serious bad habits practiced by them. At present, child abuse is becoming serious in developing worlds, especially in Bangladesh. It is raising with time at a steady rate [1]. Before going to an alarming stage, precautions must be taken to sustain the society in this region. Keeping it in mind, this research was designed to investigate the factors which are making Bangladeshi teenagers exposed to online risks and eventually affecting harm. However, it is mentionable here that the study solely focused on the viewpoints of 13-18 years of old teenagers.

II. LITERATURE REVIEW
Several factors have already been investigated by many researchers regarding this aspect. Psychological factors are one of them [2][3][4]. According to Livingstone and Görzig (2014), children with psychological difficulties reported being more vulnerable to online risks comparing with the normal ones [5]. d 'Haenens, Vandoninck, and Donoso (2013) found them more vulnerable and affected by internet use as well. Teenagers with lower self-efficacy reported being more exposed to online risks in a series of studies. They got more affected by the harm of internet use as well [6]. Livingstone and Görzig (2014) also identified risky behaviour and practices conducted by teenagers as one of the predictors of online risks [5]. This predictor was also found playing a vital role in having bitter experiences among teenagers from Facebook in another study [7]. So, from the previous studies, it is expected that the psychological factors may cause a significant impact on online risks and harm among Bangladeshi teenagers.
The role of social factors specifically parents, teachers, and peers were thoroughly scrutinized among most European nations [8][9][10][11]. Majority of the cases, the respondents reported that whenever they were guided, monitored, supported and advised by their parents, teachers, as well as peers, they could avoid those type of risks mostly, minimize the effects of the risks and overcome them (i.e. not being affected by the harm) successfully. So, from the past studies conducted by many  Stack, Wasserman, and Kern (2004) found the teenagers who attended church regularly were less addicted to internet pornography [12]. Abell, Steenbergh, and Boivin (2006) also got similar results. They found teenagers possessing a higher level of religiosity less prone to abusive activities [13]. According to Hoffman (2009), college students practicing Christian spiritual disciplines were engaged less in watching internet pornography as well [14]. Lastly, Murray (2014) found Catholic high school students more aware of online risks and harm and they took proactive and coping strategies more to deal with them [15]. So, it is expected from previous studies that religiosity level might play a significant role in online risks and harm among Bangladeshi teenagers.
According to several studies, the relationship between online risks and harm fluctuates by region in a very complicated way [9,[16][17][18]. Lobe, Livingstone, Ólafsson, and Vodeb (2011) found North-East European countries specifically Lithuania and Estonia most vulnerable in terms of online risks experienced by teenagers compared to other European countries. They were immediately followed by Norway and Sweden. South, as well as West European countries such as Italy, Portugal, Greece, and Turkey, were found as the safest zones [9]. According to Livingstone and Das (2010), new risks, as well as new harm, were observed among teenagers specifically from Southern and Eastern Europe. The authors held the arrival of new technologies as well as their availability to the society responsible for this [19].
A large number of people in Bangladesh have started using the internet recently regularly. Usually, upper-middle-class families are blessed with uninterrupted high-speed internet facilities. To maintain their lifestyle and status in society as well as meet up the expenses, both parents in these families are forced to work day and night. As a result, their children are forced to stay in their homes alone with sophisticated modern technologies and devices. In addition to that, due to the lack of places for playing as well as other outdoor activities, these children normally pass their leisure time by engaging themselves in two major indoor activities. The first is watching television; the second one is browsing the internet. Both tasks, normally prefer to do alone [1].
With the increase in their internet use, many are suffering from psychological problems. They are becoming more and more reluctant about their responsibilities especially in case of their homework, studies and other household activates. They become so hooked to it that they now are unwilling and feel uncomfortable in passing time with their friends and families. Even many have become so addicted to it that they often forget to take enough food as well as sleep properly from time to time [1].
To avoid the interruption from family members, 12% of the respondents used to remain online from 12 to 2 at night. Because most of the members in their usually go to sleep at that time. Consequently, they are not having enough sleep. Recently, psychologists warned that sleep deprivation can cause serious psychological effects among teenagers. When teenagers become so much addicted to the internet and thus spending time on it in an unjustified manner, the first thing that happens is they harm themselves. Afterward, they harm their families and lastly, the whole society [1].
Growing addiction to Facebook has become obvious among teenagers in Dhaka. More and more time passing with it has become a norm among them. This is not the main problem. The main problem is, they are engaging themselves more and more in abusive and risky activities using Facebook. Seeking relationships, amusement, higher personal status among peers may be mentionable vulnerable practices conducted by them. Going to meet the strangers after the development of online relationships, bothering others, having false identities (male presenting himself as female and vice versa), spying on friends, hacking a private account, cyberbullying, stalking others online are the abusive activities practiced by teenagers found in the literature [1].
Teenagers risky online behavior and practices have become so important issues now in these days that it has gained a lot of media coverage in Bangladesh. The two prominent daily newspapers -The Daily Observer and The Daily Star are mentionable among them. According to their report, approximately half of the teenagers in the country are encountering attacks from internet predators. And they felt completely hopeless since they were unable to do anything against them [20,21].

III. OBJECTIVES OF THE STUDY
The aims of this investigation were twofold. The first was to recognize the fundamental factor that are influencing young people online in Bangladesh. Afterward, validating the scale of online risks and harm based on the information researcher collected from the youngsters was the second target.

A. Respondents
A nation-wide survey was conducted by the researcher invoking 443 teenagers. The participants were chosen from different educational institutions such as universities, colleges as well as schools. The institutions were chosen from both rural and urban regions of the country as well. The sample comprised of 45.8% by the boys (n=203) and 54.2% by the girls (n=240). The mean age was found 15.44 years.

B. Instrument
The researcher employed an adapted survey questionnaire to get information regarding online risks as well as harm. The questionnaire, in turn, was divided into two parts. The first part was about the demographic information of the participants. The second one was designed to measure the level of agreement of the participants regarding 42 items. Those 42 items collectively covered five features: social (12 items), psychological (12 items), religiosity level (4 items), harm (6 items) as well as online risks (8 items). The researcher adopted these items from the work of the previous researchers regarding this topic. A 5-Applied Information Systems and Management (AISM) Volume 3, (1) 2020, hal 7-14 P-ISSN: 2621-2536; E-ISSN: 2621-2544 9 http://journal.uinjkt.ac.id/index.php/aism point Likert scale which ranges between Highly Untrue (1) and Highly True (5) was used to measure the constructs.

C. Data Collection and Analysis
The researcher distributed the questionnaire copies to 700 teenagers from rural and urban regions in Bangladesh. In every sector, the institutional authorities facilitated the process. Among 700 participants, 555 returned the questionnaire to the researcher. Among them, 112 copies of the questionnaire were rejected due to missing values and/or incomplete responses. Hence, 443 copies were taken eventually as usable responses. As a result, the response rate was found at 63.3%. An exploratory factor analysis procedure specifically, Principal Component Analysis (PCA) was applied to the data to extract the Factor structure of online risks and harm. Pituch and Stevens (2015) reported that exploratory factor analysis is mainly used to recognize the model or the underlying structure from a set of variables [22]. Since this study was designed to investigate the previously unsuspected relationships, PCA was found most suitable to the researcher with the target of accomplishing the research objectives. Stevens (2012) as cited by Pallant (2013) suggested PCA considering numerous motives in their studies, too. Two of them may be mentioned here because of their great significance. The first one is, it is mathematically simple; and the other one is, it is psychometrically sound. According to the authors, some complications always occur due to 'factor indeterminacy' that is related to factor analysis. Which can be overcome by PCA as well [23,24]. Among several rotational techniques, Varimax was selected by the researcher. The choice was made because of its utmost popularity as an orthogonal approach as well as its supreme capacity of minimizing the variables with high loadings on each factor [24]. This is the main objective for which the Varimax method was developed.

A. Test of Normality of Constructs
According to Kline (2015), the value of skewness should be less than 3 as well as that of kurtosis should be less than 10 for the normal distribution of data. Table 4.2 shows the mean, standard deviation, skewness, and kurtosis statistics of the components.
Since all the items fell in these ranges of permissible values, it is assumed that normality is achieved. The result encouraged that the data is suitable for further analysis as answers to the components are, accordingly, normally distributed. A positive skewness indicated a greater number of larger values and showed the distribution is to the right. Whereas, a negative kurtosis gives an indication of flatter than normal shape for the distribution. However, Pallant (2013) said that it is difficult to achieve normality for large samples. So, the rest of the analysis proceeds.

B. Reliability of the Data
The coefficients of Cronbach's alpha were considered to estimate the reliability of data intended for the extracted structure. Table II shows the results of the reliability analysis.
Fair internal consistency was shown by all the factors having coefficients larger than 0.62. In the measurement scale, an index of 0.6 or more than that designates a satisfactory internal consistency [24]. Since the results of the extracted factor structure generated estimates that range from 0.625 to 0.930, it can be inferred that the scale utilized as a part of this investigation exhibited high internal consistency.

C. Underlying Structure of Online Risks and Harm
Acceptable results were generated after applying the procedures of Principal Component Analysis on the data specified in the cases of sampling adequacy as well as interitem correlations. Sampling adequacy is usually measured by the Kaiser-Meyer-Olkin (KMO). The value of KMO lies between 0 and 1. Normally 0.7 is recommended as an acceptable cut-score. KMO was found in this case 0.871, which refers that the sampling frequency is adequate for the employment of PCA upon the data. An identity matrix namely the correlation matrix was scrutinized by Bartlett's test of sphericity. The value was found significant (χ2=8915.5, df=741, p=.000). This shows that overall correlations within the correlation matrix were adequate. In brief, the factorability of the data was proven by the results produced. Hence, the researcher's choice of selecting PCA in the data analysis phase was justified.
The screen plot was depicted firstly based on the data collected from the respondents (Figure 1). It can be suggested based on the screen plot that the corresponding items embodied the underlying factors. Subsequently, eigenvalues generated from the factors and the total variance explained both reinforced this underlying factor structure.
An assessment of the eigenvalue criteria having a value greater than 1 demonstrated the presence of factors. The factors that came out through this process aggregately clarified 65.594% of the variance by no cross-loading. Since the values of commonalities were found between 0.478 and 0.789, it can be mentioned that significant correlations existed among items. It is also noted that no item existed with a value of communality less than 0.4. Aggregately 42 items were allotted to corresponding factors depending on the peak level of their loading. The first factor extracted was teachers' help and guidance while something affected the youngsters online. Four items were included in this factor and all of them had loadings of more than 0.667. The second one was peers' help and support while they were bullied online in the past. This one also included four items; each of them having a loading value larger than 0.651. Parental monitoring and guidance were found as the third factor. Which contained four items, too. All items connected to this factor had loadings of more than 0.653 as well.
The fourth factor extracted was the religiosity level possessed by Bangladeshi teenagers. Four items were included in this factor and all of them had loadings of more than 0.623. The fifth one was the self-efficacy of those teenagers. This one also included four items; each of them having a loading value larger than 0.662. Emotional problems were found as the sixth factor. Which contained three items, too. All items connected to this factor had loadings of more than 0.641 as well. The seventh factor is about risky behaviour and practices conducted by Bangladeshi teenagers while they are online. Though this factor invoked only toe items, both possessed loading values larger than 0.67.
Finally, eight items were contained in online risks all of which had loadings larger than 0.742 as well as six items in harm all of which had loading values larger than 0.625.

D. Reliability of the Data
The coefficients of Cronbach's alpha were considered to estimate the reliability of data intended for the extracted structure. Table II shows the results of the reliability analysis.
Fair internal consistency was shown by all the factors having coefficients larger than 0.62. In the measurement scale, an index of 0.6 or more than that designates a satisfactory internal consistency [24]. Since the results of the extracted factor structure generated estimates that range from 0.625 to 0.930, it can be inferred that the scale utilized as a part of this investigation exhibited high internal consistency.

VI. CONCLUSION
Entirely a reliable underlying structure was generated by the outcomes regarding online risks harm the Bangladeshi teenagers are facing and additionally this upheld the factors to be specific, social, psychological as well as religiosity level. The PCA came out by demonstrating all the items having high factor loadings to the validity of the scale. In addition to that, they were classified successfully and significantly into some specific logical factors, which in turn collectively clarified 65.594% of the variance. The scale used in this study was proven successful by factor loadings and the total variance explained for apprehending the fundamental extents of online risks sand harm the Bangladeshi teenagers are facing. Along these lines, it can be sensibly contended that the scale offers a substantial measure regarding the construct.
What's more, the Cronbach's alpha esteems were found high (ranging between 0.625 and 0.930) for the generated factor structure. In this way, this exhibited a high internal consistency of the scale. Last, of all, the results of validity as well as reliability analysis also exhibited that the instrument designed and employed for the investigation is a reliable and legitimate estimation tool. Hence, that can be utilized to evaluate further the dimensions of online risks and harm in future investigations.
In brief, online risks and harm is a highly complex phenomenon involving a complex interplay of independent and dependent variables. To develop a better understanding of the problem, it is highly recommended to conduct a qualitative study afterward. Hopefully, rich data will be derived in this way to address the issue more deeply.