Does the Association of Social Media Use with Problematic Internet Behaviours Predict Undergraduate Students’ Academic Procrastination? L'association de l'utilisation des médias sociaux avec les comportements problématiques sur Internet permet-elle de prédire la procrastination académique des étudiants de premier cycle?

.

Recently, some researchers have extended their models to the use of new technologies and students' academic procrastination (Yang, Asbury, & Griffiths, 2018). Yang, Asbury, and Griffiths (2018) demonstrated the relationships existing among problematic smartphone use, academic anxiety, academic procrastination, self-regulation, and subjective wellbeing using a hypothetical model. Others have provided only individual variable correlations between social media use and academic procrastination (Muslikah, Mulawarman, & Andriyani, 2018) notwithstanding the continuous under-exploration, and the poor understanding, of the mechanism behind procrastination (Procee, Kamphorst, vanWissen, & Meyer, 2014). Delaying the execution of tasks could be better understood when a range of factors are considered that are capable of redirecting students' attention to pleasurable activities with short-term utilities (Siaputra, 2010). The allure that the Internet and social media present has the tendency to result in academic procrastination. Given that adolescents are heavy users of the social media and the fact that it could lead to task delay, it becomes imperative to understand the associations among social media use, problematic Internet behaviours, and academic procrastination.
Does the Association of Social Media Use with Problematic Internet Behaviours Predict Undergraduate Students' Academic Procrastination? 3

Literature Review Theoretical Framework
Theories stemming from psychodynamic, behavioural, cognitive, and the latest temporal motivation theory (TMT) have been advanced to explain procrastination (Siaputra, 2010). Siaputra explained that the psychodynamic perspective situates procrastination as a defence mechanism resulting from failure to complete a task which in turn threatens the ego of the individual. The behavioural theorists postulate that procrastination is reinforced when an individual repeatedly gets away with dilatory behaviour, while the cognitivists present procrastination as resulting from faulty thought processes and belief systems that lead to delay in task execution (Siaputra, 2010). However, researchers are of the view that TMT is more comprehensive to address the mechanisms of procrastination (Siaputra, 2010;Steel & Konig, 2006;Steel, Svartdal, Thundiyil, & Brothen, 2018). From the TMT perspective, procrastination is linked with the gap existing between an individual's intention and his/her actions indicating that procrastinators are "susceptible to proximity of temptation and to the temporal separation between their intention and the planned act; the more distal, the greater the gap" (Steel et al., 2018, p. 1). TMT is an eclectic motivational theory that lays emphasis on time as a fundamental component of motivation (Steel & Konig, 2006). Steel and Konig (2006) situated TMT in the explanation of situations in which expectancy, value, and time, which are the core of the theory, simultaneously impact decision making and are in turn impacted by individual differences. For Steel and Konig (2006), one of the suitable examples of a prototypical motivational problems that can be explained using TMT is procrastination which is seen as a 'motivational failure' (Steel, et al, 2018:1). Steel and Konig (2006) acknowledged how complex the mechanism of procrastination is and stated that TMT has the capacity to explain procrastination given that procrastinators are entangled in preference reversal. However, Siaputra (2010) is of the view that theories that have been advanced in the explanation of procrastination, including TMT, have demonstrated their weaknesses. In a bid to make up for these weakness, Siaputra (2010) theorized that procrastination could be comprehensively understood from the angle of delaying or avoiding doing activities with higher long-term utility value to engage in more pleasurable activities with ultimately short-term utility. Siaputra's (2010) opinion can be merged with the propositions in TMT given that these more pleasurable activities with ultimately short-term utility value may stand as ingredients for preference reversal in TMT. This theoretical concern is not fully understood. Modelling social media use and procrastination could give an insight into this concern.

Social Media Use, Problematic Internet Behaviours, and Procrastination
The active use of digital technologies at an early age helps children and adolescents in their cognitive development, as they learn to think creatively (Nikolopoulou, 2018;Yushau, Mji, & Wessels, 2005). Innovations in digital technologies include social media. Social media are online interaction sites that foster interactive participation among people with the intention to build, share, and change their idea and opinions on a wide range of issues (Manning, 2014;Zamri, Zaihan, & Samat, 2018). Social media use is currently the most predominant activity among adolescents (BaÂnyai, et al, 2017) and refers to the access of different social media platforms for diverse reasons and activities (BaÂnyai et al., 2017). Researchers see Internet use by students as comprising the average time spent on the internet, number of visited web pages, and the most regularly visited websites (Oskouei & Sharma, 2013). Advantages of using social media include social connections and skills, improvement in learning opportunities, and exchange of ideas (Zamri et al., 2018). In view of the above, Fernandes, Maia, and Pontes (2019) have noted that the use of the Internet itself is not inherently evil given its undeniable benefits.
However, behavioural tendencies in online activities that could interfere with adolescents' mental health and academic successes have been documented (Deschamps & McNutt, 2016;Nixon, 2014). Field (2018) is of the view that the widespread use of electronic devices and social networks has resulted in Internet addiction and cyberbullying. Negative side effects result from problematic Internet use (PIU). Recent concern in literature shows that there is not yet consensus on what PIU is (Anderson, Steen, & Stavropoulos, 2017;Fernandes et al., 2019). This has led to the use of different terms to refer to PIU (Anderson et al., 2017;Fernandes, et al, 2019). PIU has remained an "ill-defined and heterogeneous construct" (Spada, 2014, p. 5) with two fundamental features: the aspects of excessive, compulsive Internet use, and the "various adverse consequences of spending too much time on the Internet, such as neglecting social activities, relationships, health and work or school duties, and altering sleep and eating habits in a detrimental way" (Spada, 2014, p. 5). Problematic Internet behaviours in this study go beyond the compulsive use that is detrimental but include unhealthy online behaviours or antisocial Internet use (Ma, 2011) in the sense that they are capable of affecting the well-being of the individual and other online users connected to him/her. These problematic Internet behaviours include cyber bullying and victimization and Internet addiction.
Cyber bullying and victimization have been studied as online antisocial Internet use (Lapierre & Dane 2020;Ma, 2011) connoting intentional and repeated harm inflicted through electronic means (Brown, Demaray, & Secord, 2014). A study in a Nigerian university has documented that a range of 48% -57% undergraduates have been bullied through various cyber media and 28.6% -40.0% have bullied others through these devices (Nwosu, Ementa & Ejikeme, 2018). Research has shown that 'between 20% -50% of adolescents have been victims of peer aggression by electronic means at some time, and between 2% -7% have suffered severe victimization' (Garaigordobil, 2011as cited in Álvarez-García, Perez, Gonzale, & Perez, 2015. Recent studies have documented the negative impacts cyberbullying and aggression could have on adolescents including suicidal ideation, social isolation, mental illness, low selfesteem, poor self-concept, loneliness, and school phobia (Deschamps & McNutt, 2016;Nixon, 2014;Sourander et al., 2010). Apart from cyber victimization and aggression, another online problem is Internet addiction which results in neuro-psychological disturbances (Cash, Rae, Steel, & Winkler, 2012). Shaw and Black (2008), along with Sutton and Subbarao (2011), are of the opinion that Internet addiction is manifested in excessive and/or regulatory failure in the use of the Internet which leads to a negative effect in the well-being of the individual. This could result in interference with other activities in the life of adolescents due to excessive online engagement. Studies have shown that between 8% -34% of undergraduate students are Internet addicted (Krishnamurthy & Chetlapalli, 2015;Shao et al., 2018;Xin et al., 2018). This takes up most of their time that could otherwise be used in a meaningful academic venture. Some Internet users tend to develop an emotional attachment to their online friends and activities. This allows them to use functions of the Internet to meet, socialize, and exchange ideas in different chat rooms, social networking websites, or virtual communities (Alassiri, Muda, Ghazali, & Ahamefula, 2014;Mustafa & Hamzah, 2011;Ridings & Gefen, 2004). Reinecke et al. (2018) Does the Association of Social Media Use with Problematic Internet Behaviours Predict Undergraduate Students' Academic Procrastination? 5 noted that literature is replete with studies that demonstrate that "Internet users use online media as tools for procrastination" (p. 862). Similarly, Lian et al. (2018) acknowledged that the popularity of social media has led to increasing problematic social media behaviours and use, both of which have been associated with irrational procrastination. Using qualitative research design, Powell, Gray, and Reese (2013) found that students use social media as tool for procrastination.
Empirical studies have shown that Internet addiction has been found to be negatively related to the academic achievement of undergraduate students (Akhter, 2013) and positively related to academic procrastination of undergraduate student teachers (Kandemir, 2014;Saleem, Owaisi, & Tufail, 2015). Other factors that could be associated with problematic digital technology use include excess time spent online (Savoldi & Ferraz de Abreu, 2016) and number of social media sites engaged in, since absorption by the social websites could lead to addiction (La Barbera, La Paglia, & Valsavoia, 2009). However, the relationships existing among these problematic Internet-related behaviours and students' academic procrastination seem to be underexplored. More so, how the association between social media use and problematic Internetrelated behaviours could predict students' academic procrastination among Nigerian undergraduates is yet to be understood.
The purpose of this correlational study is to build a model that will examine the effect of students' social media use as mediated by problematic Internet behaviours on their academic procrastination using a path analysis. We specifically set out to: 1. Develop a causal model to explain how the association between social media use and problematic Internet-related behaviours predicts students' academic procrastination.
2. Estimate the standardized regression weights of the exogenous variables on the endogenous variable.
3. Estimate the indirect and total effects of the exogenous variables on the endogenous variable.

Method
The research design is a correlational research design with two independent (exogenous variables) and four dependent (endogenous) variables with three serving as mediator variables. The study sought to establish a causal model for students' academic procrastination. Hence, path analysis was adopted to develop the causal model. Given the available literature, the researchers theorized that students' unhealthy engagement in online activities could impact on their academic procrastination. The independent variables were the number of social media sites (SMS) used and time spent (TIME) on social media, while the dependent variables are cyber victimization (CV), cyber aggression (CA), and Internet addiction (mediators), and academic procrastination. It was conceptualized that exogenous variables could have direct impacts on students' academic procrastination as well as indirect impacts through Internet addiction, cyber victimization, and cyber aggression (see Figure 1). The population of the study consisted of the 1,462 first year undergraduate students in the Faculty of Education at Nnamdi Azikiwe University, Awka, who registered for the first-year introductory course in Educational Psychology in the 2018/2019 academic year. The sample size comprised of 500 students who attended a second semester course titled Introduction to Educational Psychology. The sampling technique adopted was the convenience sampling in which those who volunteered to participate in the study were given a questionnaire to fill in. First, the researchers explained to the students the essence of the study and assured them that information collected from them was for research purposes and would be kept confidential. They filled in the questionnaire after their classes and it was retrieved from them on the spot.
Four instruments were used in the collection of the data. They are: cyber victimization questionnaire (15 items scaled never = 0, sometimes = 1, several times = 2, always = 3), cyber aggression questionnaire (15 items scaled as: never = 0, sometimes = 1, several times = 2, always = 3). These two instruments were adopted from Garaigordobil's (2015) cyberbullying tests. Garaigordobil (2015) reported good indices of the psychometric properties of the test. However, these instruments were revalidated in Nigeria and the reliability coefficients for the two instruments using Cronbach's alpha test were 0.82 and 0.88 respectively. Other instruments are Internet addiction scale (AIS) developed by Young (1998;20 items scaled as: does not apply = 0, rarely = 1, occasionally = 2, frequently = 3, often = 4, always = 5) which was also revalidated in Nigeria and its reliability coefficient using Cronbach's alpha test was 0.92. The academic procrastination scale was 21 items scaled as: strongly disagree = 1, disagree= 2, agree = 3, strongly agree = 4 with Cronbach's alpha coefficient of 0.75. This instrument is an adaptation of the scale developed by Abu Ghazal (2012). This instrument has been used by several authors in measuring students' academic procrastination and good psychometric indices have been reported (Mahasneh, Bataineh, & Al-Zoubi, 2016). We relied on the English version used by Mahasneh, Bataineh, and Al-Zoubi (2016) in our adaptation of Abu Ghazal's (2012) procrastination scale.
Data was analysed for the demographic variables using percentages while Pearson and path analysis were used to establish the relationships and the casual model. Path analysis was conducted using SPSS AMOS version 24, and the fit indices and the casual relationships were determined. Because there were missing values, data was imputed, and the Bootstrap bias confidence level (Bollen-Stine bootstrap) was used to establish the significance of the direct and indirect effects of the predictor variables on the independent variable.

Assumption Testing
Sample size. Kline (2011) recommended that the ratio of the number of respondents to the number of parameters be estimated at 10:1, even though a more ideal sample size-toparameter estimate would be 20:1. Our hypothesized model had 20 distinct parameters with a sample size of 500 amounting to a ratio of 25:1 which falls within the acceptable range.

Missing values.
Missing values were handled using regression imputation in AMOS. Regression imputation takes into account "subject patterns of scores across all other variables" (Carter, 2006, p. 4) and predict "a score for each subject by using multiple regression based on their non-missing scores for other variables" (Carter, 2006, p. 4). The essence of this is to preserve the relationship between the variables since imputed values were based on regression models. Then a complete data set was generated for analysis. Multivariate normality test and outliers. Maximum likelihood estimation assumes variables exhibit multivariate normality. Violation of this could lead to a number of problems including the type 1 error and because chi square of fit index is sensitive to data deviation from the norm and significant outliers, multivariate normality test was conducted. Before this was ascertained, the univariate normality was inspected and based on the values of skewness and kurtosis, most of the individual distributions failed the normality test, resulting to the multivariate normality to deviate significantly. The recommendations of Kline (2011) were followed in which it was stated that no skewness values exceed the absolute value of 3 and that of kurtosis should not exceed the absolute value of 10. In the case of multivariate normality, the value of kurtosis is 64.534 showing substantial departure from normality since multivariate normality whose kurtosis value is greater than 5 deviated substantially from the norm. To handle the multivariate nonnormality, bootstrapping procedure available in AMOS was used and the p = 0.135 indicates that the null hypothesis that the model is correct be accepted since the value is greater than 0.05. This shows that our model is fit.

Linearity and multicollinearity.
To demonstrate that the variables are linearly dependent, deviation for linearity for all the predictor variables and the dependent variables were determined using SPSS. The values range between 0.075 -0.976, which are greater than 0.05. This shows that the variables are linearly related. Also, a multicollinearity test was conducted among the independent variables using SPSS to ascertain if they are highly correlated, the collinearity statistics showed that the VIF value ranged between 1.04 -1.43, which is less than 10. This assumption is not violated among the independent variables.   Table 1 shows the demographic data of the respondents analyzed with percentages. 27.8% are male students, 72.2% are female students, 99.4% are social media users, more than 56% of the respondents are users of four or more social media sites, more than 35% of the students spend more than four hours online daily. Table 2 Correlational Matrix of the Variables Note. ** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed). Table 2 shows the correlation matrix of both the dependent and independent variables. CVB correlated significantly with CAB, IA, AP, TIME, and SMS. CAB is significantly related to IA and AP; IA is significantly related to AP, TIME, and SMS; AP is significantly related to TIME, and TIME is related to SMS.  Figure 2. The standardized path model for students' academic procrastination.  Table 3 shows that the various tests are respectively consistent in suggesting that the hypothesized model represented an adequate fit to the data. Therefore, the hypothesized model proposed in the study fits the sample data satisfactorily.   Table 4 showed predictive powers of the independent variables on the dependent variables. SMS and TIME significantly predicted undergraduate students' IA (R 2 =.198, p<.05), and (R 2 =.295, p<.05) respectively. IA significantly predicted CVB (R 2 =.335, p<.05), and academic procrastination (R 2 =.339, p<.05) while CVB predicted significantly CAB (R 2 =.568, p<.05), and a negative non predictor of academic procrastination (R 2 =-.121, p>.05). CAB predicted academic procrastination (R 2 =.221, p<.05). Note. ** Significant at the 0.05 level (2-tailed). Table 5 shows the total, direct, and indirect effects of the variables on the endogenous (dependent) variables. TIME has a significant total and indirect effect (p<0.05) on their AP. It has no significant direct effect (p>0.05) on their AP. SMS has only a significant total indirect effect (p<0.05) on the AP without any significant direct and total effects on AP. CAB has significant total and direct effects (p<0.05) but no total significant indirect effect on their AP. CBV has significant total and direct effects (p<0.05) on AP. With regard to indirect effects, Table 4 represented the total indirect effects as computed by AMOS default and because AMOS by default does not have the capacity to estimate the specific indirect effects in multiple mediation (Perera, 2013), we estimated the specific indirect effects of the exogenous variables with the user defined estimand in AMOS as shown in Figure 3 and Table 6.  Note: CI = confidence intervals; LB = lower bound; UB = upper bound; SE = standard error. Table 6 showed that SMS and TIME had specific indirect effects on AP through IA while TIME had significant indirect effect on AP through CAB based on 1000 bootstrap sample at 95 bias-corrected confidence intervals. SMS and TIME had no significant indirect effects on AP through CBV and CAB.

Correlations
Does the Association of Social Media Use with Problematic Internet Behaviours Predict Undergraduate Students' Academic Procrastination? 12

Discussion
Our study proposed a model analysing the relationship patterns and directions between students' social media use (time spent on social media, the number of social media sites registered with), problematic Internet behaviours (cyber victimization, cyber aggression, Internet addiction), and their academic procrastination. The model fit indices showed that our hypothesized model fits the sample data satisfactorily. This shows that explaining academic procrastination in contemporary time could also be done when researchers look into what goes online where students spend most of their time either engaging in academics or social relationships. Engaging in social media activities, given its allure, could lead to procrastination when comprehensively examined from the angle of delaying or avoiding doing activities with higher long-term utility value to engage in more pleasurable activities with ultimately short-term utility (Siaputra, 2010). Our finding supports similar research which has shown that when social media use becomes problematic it positively predicts students' procrastination (Lian et al., 2018).
The correlation matrix in our study showed that there were significant positive relationships existing among time spent on social media, number of social media sites registered with, cyber victimization, cyber aggression, Internet addiction, and undergraduate students' academic procrastination. The finding indicated that time spent online is significantly positively related to Internet addiction. This shows that the more students spend their time in social media, the more they are likely to be addicted online. Similar findings, such as the findings in Al-Menayes' (2015) study, showed that time spent online is related to social media addiction. Simsek and Sali (2014) have also found that students who spend more time on social media have higher Internet addiction scores than those who do not. Although time students spent on social media predicted Internet addiction, it did not directly predict students' academic procrastination but had a significant positive indirect effect through Internet addiction on students' academic procrastination and a significant negative indirect effect through cyber-aggression. This has a lot to reveal concerning the relationship between time spent online and academic procrastination. It could mean that when time is spent appropriately online, for example, studying and doing assignments, it may not account for students' academic procrastination. This concurs with the fact that social media could enhance academic achievement when utilized as platforms for student's engagement (Mingle & Adams, 2015). However, when it becomes compulsive, it could derail students' attention to academic pursuits. This agrees with available literature that has shown that when time spent on social media becomes excessive or problematic, students procrastinate in their studies (Mingle & Adams, 2015). Positive relationship between Internet addiction and academic procrastination and the predictive power of Internet addiction on students' academic procrastination have been documented (Kandemir, 2014;Uzun, Unal, & Tokel, 2014). Furthermore, the significant negative indirect effect time has on students' AP through cyber-aggression indicate that proper use of time, such as productive interactions, could reduce students' aggression which in turn would reduce academic procrastination since TIME in our findings is a negative predictor of cyber-aggression.
We also found in our study that the number of social media sites used was significantly positively related to Internet addiction. It was also a significant predictor of Internet addiction. The significant positive relationship could be explained from the point that registering and using many social media sites could lead to prolonged use. This points to the pragmatic factor in social media sites addiction that "emphasizes access and habituation variables in the development of specific addictions" (Kuss & Griffiths, 2011, p. 3532). There is the tendency that an individual will satisfy his/her curiosity with what goes on in those social media sites. Furthermore, our findings showed that the number of social media used is not a significant predictor of students' academic procrastination. Researchers are of the opinion that it is when social media is poorly managed that it can lead to academic problems (Azizi, Soroush, & Khatony, 2019). This may imply that other beneficial use of social media in academics could be intervening variables. However, the number of social media sites had an indirect effect on students' academic procrastination through Internet addiction. What this means is that the number of social media sites used could predict students' academic procrastination through Internet addiction. This could be interpreted to mean that the more students are addicted as a result of the number of social media sites used, the more they are likely to procrastinate their academic works. In a bid to meet up with activities in these sites there is the likelihood that more time is expended resulting in the postponement of duties related to their academic works.
An interesting finding was that the direct paths between students' social media use (TIME and SMS) and their academic procrastination were not significant but there existed a significant indirect effect among them through IA (positive effects) and CAB (negative effects for TIME only). This indicates that social media use is not inherently evil (Fernandes et al., 2019) and that a lot of advantages that can facilitate learning are attributed to proper use of social media (Mingles & Adams, 2015;Zamri et al., 2018).
Furthermore, our findings revealed that cyber victimization is significantly related to cyber aggression. Cyber victimization and cyber aggression appear to go hand-in-hand, indicating that a student who gets bullied might end up being so aggressive that he becomes a bully himself. This agrees with the conclusion of Ildirim, Calici, and Erdoğan (2017) when they found in their study that cyberbullying and cyber victimization are positive correlates of such psychological problems as hostility and impulsivity. There is the tendency that, without much thought, a student who is victimized could become enraged and revenge on the aggressor and may even become aggressive to others. Results showed that cyber victimization positively predicted cyber aggression among the students indicating that cyber victimization could account for students' aggressive behaviours online. Similarly, cyber aggression and cyber victimization had significant positive correlations with Internet addiction. Athanasiades, Kamariotis, Psalti, Baldry, and Sorrentin (2015) found a positive relationship between problematic Internet use and cyberbullying. Students who are cyber bullies may spend more time on the Internet as they are always on the lookout for who they will bully, resulting in this significant relationship.
More so, our analysis revealed that Internet addiction significantly positively predicted cyber victimization but not cyber aggression. This shows that the more one is addicted online, the more one is likely to be victimized online. Several studies have demonstrated how Internet addiction could negatively impact the psychological well-being of a child (Hou, Xiong, Jiang, Song, & Wang, 2019). There is the likelihood that the more time one spends online the more likely they are to meet bullies online (BioMed Central, 2018). Cyber aggression positively predicted academic procrastination. This could be explained by the fact that the time that should be devoted to studies may be directed to bullying online. Though researchers have concentrated on the impact of cyberbullying on victims (Myers & Cowie, 2019), given that findings showed that cyber aggression is linked to academic procrastination, the cyberbully may delay academic work to engage in cyber aggression.

Conclusions and Limitations
This study aimed at building a casual model and estimating the direct, indirect, and total effects of social media use, problematic Internet related behaviours, and students' academic procrastination. Findings showed that social media use did not directly account for students' academic procrastination but indirectly did so through Internet addiction. It is concluded that the use of social media may not be inimical to students' academic endeavour until it is related to unhealthy online behaviours. This is very useful for intervention and prevention of academic procrastination among undergraduate students. Stakeholders should endeavour to build students' self-regulatory skills, encourage them to visit more educational sites that will enhance their academic activities, and advocate the establishment of healthy child-parent relationship that will enhance students' productive use of social media sites.
Although our findings have made significant contributions, our study is limited by a number of factors. First, our sample is primarily made up of female students who accounted for 72.2% of the sample size. The generalization of our findings could be limited since there might be a substantial gender difference and this study was based on a sample from a single university. This limits its generalizability. Based on this, further research from other universities needs to be conducted to confirm these findings. Second, newer instruments have emerged in Internet addiction evaluations given the advancement in knowledge. IAS may appear dated given the emergence of newer instruments. Though the IAS is widely validated and used (Anderson et al., 2017;Fernandes et al., 2019) newer instruments could be adopted in further studies.