Facebook Usage and its Impact on Academic Performance of University Students in Bangladesh

Abstract

The prompt rise of digital technologies and social networking podiums has changed the way students interact, communicate, and access information. Facebook, amongst those podiums, has become exclusively prevalent in university students' day-to-day practices in Bangladesh. Facebook deals with several educational advantages, like group discussions, knowledge sharing, and staying informed etc. So, it is a growing concern over its impact on students’ academic attention, time management, and largely academic performance. This study carried out with the intent to identify the Facebook usage and its impact on academic performance of university students in Bangladesh. The primary objectives of this study are threefold: (1) To identify the usage patterns of Facebook among university students in Bangladesh, (2) To examine how Facebook usage influences academic performance, and (3) To analyze the specific effects of such usage on university students’ academic achievement. The research also marks three research questions affiliated with the objectives and tests three null hypotheses relating collaboration, communication, and resource sharing over Facebook to academic performance. This cross-sectional study involved a mixed-methods approach, incorporating both qualitative and quantitative. Total of 462 students from three public universities (University of Dhaka, Jahangirnagar University, and Comilla University) and three private universities (North South University, Daffodil International University, and Bangladesh University of Business and Technology) were surveyed using a structured questionnaire. Convenience sampling was applied. The data were analyzed using SPSS 26.0, SMART PLS 4.0, and VOSViewer, incorporating Descriptive statistics, Confirmatory Factor Analysis (CFA), and Structural Equation Modeling (PLS-SEM). The major findings of the study are summarizing as follows based on the analyzed data: Based on survey 57 percent respondents were male and 43 percent respondents were female. Eighty six percent respondents belong to age group 22 to 25 years, and 14 percent respondents belong to age group 18 to 21 years. Forty three percent respondents studying fourth year, 29 percent respondents studying second year, 14 percent respondents studying first year and 14 percent respondents studying fifth year. Eighty six percent respondents studying undergraduate level and 14 percent respondents studying master’s level. Forty three percent respondents studying public universities, 57 percent respondents studying private universities. Fifty seven percent respondents visited Facebook, 29 percent respondents vii visited Whatsapp, 14 percent respondents visited you tube. Forty three percent respondents use mobile for browsing internet, 43 percent respondents use laptop for browsing internet and 14 percent respondents use desktop for browsing internet. Seventy one percent respondents use internet at home and 29 percent respondents use internet at university premises. All respondents use internet daily basis. Seventy one percent respondents spent 0-5 hrs for academic purpose and 29 percent respondents spent 5-10hrs for academic purpose. Forty three percent respondents use internet to visit social sites, 43 percent respondents use internet for group study and 14 percent respondents use internet for sharing files. Fifty seven percent respondents usage internet sharing and downloading lecture notes, 29 percent respondents usage internet sharing and downloading journal articles and 14 percent respondents usage internet sharing and downloading e-books. Fifty seven percent respondents spent 1-3 hrs daily basis to visit social media sites, 29 percent respondents spent 3-6 hrs daily basis to visit social media sites and 14 percent respondents spent more than 6 hrs daily basis to visit social media sites. Eighty six percent respondents spent less than 5 hrs on Facebook for academic related activities weekly basis and 14 percent respondents spent 5-10 hrs on Facebook for academic related activities weekly basis. Average Facebook friends of the respondents around 953 and they spent time on Facebook average 3.43 hrs daily basis. The mean of the individual items' scores was used to calculate the four constructs or components. The means of all four constructs are near the central scale point (four on a five-point Likert scale) and have an SD of less than two. All four constructs have skewness and kurtosis coefficients that are less than two. Every factor loading is higher than 0.5. Academic performance's R2 score is 0.229.Cohen (1988) states that an R2 value of 2 percent indicates a modest influence, an R2 value of 13 percent indicates a medium effect, and an R2 value of 26 percent indicates a strong effect. Accordingly, the model had a medium impact on the endogenous characteristics and explained 22.9 percent of academic performance. Factor loading above 0.5, which is adequate for the construct and convergent validities, is demonstrated by the CFA approach. All independent variables have factor loadings greater than 0.5. Cronbach's alpha and composite reliability (CR) values for each construct were evaluated, and they were all above Cohen's recommended critical level of 0.7 (1988). All constructs had average variance extracted (AVE) values that were higher than the threshold value of 0.50 proposed by Hair et al. (2017). Heterait-monotrait, or simply HTMT, is a novel approach to obtaining the discriminant validity that was offered by Heseler et al. (2015) as an additional method for proving discriminant validity. According to Heseler et al. (2015), the strict HTMT threshold value of 0.85 must be less than 0.9, which is a loosen criterion. All of the values fall below 0.85, meeting the necessary threshold value and hence viii establishing discriminant validity. According to the Fornell and Larcker technique, a variable must differ more from its own items or self than from other variables. According to Hair et al. (2011), the diagonal values in the table must be bigger than the column and row they represent because they are the square root of AVE. All of the diagonal values satisfy the required standards, guaranteeing discriminant validity. When two variables have a strong relationship or when one predictor variable in multiple regression can be linearly predicted from another, this is known as collinearity (Hair et al, 2014). Collinearity can be checked using the inner VIF. Since the VIF values for every component were less than five (Hair et al., 2017), the results show that collinearity was not an issue. The cross-loading method is another way to measure discriminant validity; although though it is not employed nowadays, it is still used to assess validity. Each item's loading in its own variable must be bigger than the item's value in cross loadings in other variables, and the cross-loading method ensures that every item is loaded (Hair et al 2011; Hair Jr et al 2014). According to Gefen and Straub (2005), there must be a difference of more than 0.1 between the cross loadings of the item on its own variable and other constructs. The researchers used PLS-SEM to examine the structural model after evaluating the measurement model (Hair et al. 2019). The PLS approach (Partial Least Square) was used to test the hypothesis, and boot strapping was used to run the results (Haenlein and Kaplan 2004). From the original data set, this method produced 5000 subsamples in total (Hair Jr et al 2014). The model predictive quality (R2) was examined in order to assess the structural model. R2 has three possible values: high, moderate, and low. A number greater than 0.6 is regarded as high, a value between 0.3 and 0.6 as moderate, and a value less than 0.3 as low (Sanchez 2014).It displays a low R2 value of 0.229. The findings indicate that there is a positive significant relationship between collaboration and academic performance (β=0.286, t=9.332, p<0.000), communication and academic performance (β=0.364, t=9.562, p<0.000), and resource sharing and academic performance (β=-0.319, t=9.572, p<0.000).