Challenges of online education /distance learning for university students during the COVID-19 pandemic: An observation from Bangladesh

Following the spread of COVID-19, online education/distance learning (OE/DL) was opted by the governments across the world. However, many challenges remained unexplored as OE/DL was quite an unfamiliar teaching approach, particularly in Bangladesh. This study was aimed at identifying the major challenges of OE/DL in Bangladesh from the perspectives of students. Data for this web-based cross-sectional study were collected using an e-questionnaire based on Google Form. Out of 451 initial responses, 419 were retained based on some specications. The univariate analyses identied the nancial, technical, and pedagogical issues as major problems. Exploratory factor analysis and conrmatory factor analysis suggest a two-factor solution – ‘nancial-technical obstacles’ and ‘pedagogical obstacles’ – with overall good reliability (Cronbach’s α = 0.881; McDonald’s ω = 0.880). The two-factors, with strong intercorrelation (r = 0.552), meet the criteria of discriminant validity. To overcome the shortfalls of OE/DL in Bangladesh, inclusiveness with sustainable educational and infrastructural approaches was recommended.


Introduction
Following the con rmed coronavirus cases in the second week of March, the government of Bangladesh declared a nationwide 'general holidays' on March 26, 2020 to enforce the non-therapeutic measures, advocated by the World Health Organization, to minimize the spread of COVID-19 (Jahid, 2020;Opu, 2020;Shammi et al., 2020;World Health Organization, 2020a). However, the educational institutions, including the primary, secondary, and tertiary of both public and private, were closed on March 18, 2020 initially for two weeks (Shawon, 2020). As the number of both infections and deaths increased in the following months (World Health Organization, 2020b, 2020c, 2020d, the closure of educational institutions and 'general holidays' extended on a regular interval (Dhaka Tribune, 2020aTribune, , 2020b. Although the government lifted the 'general holidays' in early September (The Daily Star, 2020b), the schools and colleges will remain closed till the 29 March 2021 (The Daily Star, 2021a), while universities will re-open on 24 May 2021 (The Daily Star, 2021b).
The prolonged closure of educational institutions has adversely affected not only the academic activities of students, but also their mental well-being during the ongoing COVID-19 pandemic. Studies suggest that restricted social interaction, insu cient mental refreshment, uncertainties over academic and professional career together with prolonged home con nement have ampli ed the mental problems Islam et al., 2020;Lu et al., 2020;Shovo et al., 2021), intensi ed the screen time (Gao et al., 2020;Hossain et al., 2020) and increased the substance use among students (Gritsenko et al., 2020). Moreover, the exposure to 'infodemic' through social and mass media (Gao et al., 2020;Hossain et al., 2020;Xie et al., 2020) along with the concern over the social and physical well-being of relatives Odriozola-González et al., 2020) led to an overall degradation of mental health. The world has witnessed heightened mental health problems, including anxiety, depression, stress, fear, poor sleep, and suicidal behavior, particularly among students Cao et al., 2020;Islam et al., 2020;Kaparounaki et al., 2020; Odriozola-González et al., Shovo et al., 2021;Zou et al., 2020).
to plan and implement policies and strategies to ensure an effective and quality OE/DL and to allow the students to complete their higher education successfully. The complementary sub-objectives are: 1. Specifying the challenges faced by university students in Bangladesh to participate in OE/DL.
2. Constructing and con rming a classi cation of the challenges and validate their replicability. 3. Making some suggestions or possible way out to overcome those challenges to ensure all-inclusive and quality OE/DL for all in Bangladesh.

Study sites and samples
This web-based cross-sectional study was conducted a straight three weeks, starting from the third week of November to that the rst of December 2020. This study was carried out to identify the predicaments of OE/DL in Bangladesh from university students' perspective. The inclusion criteria for university students were (i) any on-seat university students -both private and public, (ii) participating in the online classes during the COVID-19 pandemic in Bangladesh. A self-reported e-questionnaire, using the online Google Form, was forwarded by the researchers to the university students through professional networks as well as different social media platforms, such as Facebook, Messenger, and WhatsApp. Thus, the participants were selected conveniently through snowballing from the primary contacts Four hundred and nineteen responses (419), out of the initial 451 responses, were deemed suitable to retain in the study after careful and rigorous scrutiny. It is important to note that using a population proportion equation with a margin of error of 5% (Yamane, 1967), the recommended sample size was 384 for a population of 5,94,493 (University Grants Commission of Bangladesh, 2020).

Ethical issues
This study was ethically approved by the institutional ethical clearance committee. The participants responded voluntarily by lling up a written informed consent letter in the rst section of the equestionnaire, con rming their anonymity and right to withdraw. Thus, the participants were free to decline from the web-based survey at any moment without prior justi cation.

Measures
Following the consent of the participants, the survey commenced. The e-questionnaire was divided into three separate but interrelated sub-sections, each focused on a different issue. For example, the section one extracted relevant demographic information, including age, sex, location, level of education, discipline/department, and type of university; the section two highlighted questions regarding access to and use of internet connection and digital devices to participate in OE/DL; and the last section contained 18 closed-ended Likert-scale questions divided into ve sub-categories (see Table 1 for details), to identify the experience and perception of the participants regarding the OE/DL.

Socio-demographics
The socio-demographic information of the participants includes age, sex, levels of education, residence, stream of education and type of university. The participants, with an average age around 22 years (SD = 2.00), were almost equally distributed between male (56.6%) and female (43.4%), and about 61% of participants were students of First Year (26.3%) and Second Year (33.9%). Around 45% of the participants resided in urban areas, and 46.3% were studying courses of humanities and social sciences. Around three out of ve participants were from general universities, followed by 28.9% from engineering universities, while 7.6% of participants enrolled in other forms of university in Bangladesh, including private universities.

Analysis
Data were analyzed in three consecutive stages: rstly, the descriptive analysis, including the percentage, mean (M) and standard deviation (SD), was estimated; secondly, the exploratory factor analysis (EFA) was executed; and nally, con rmatory factor analysis (CFA) was performed. The analyses were executed by SPSS 25.0 and AMOS 23.0 for a windows software package, respectively.
The goodness-of-t of the CFA models were assessed by the comparative t index (CFI), the Tucker-Lewis index (TLI), the root-mean-square error of approximation (RMSEA) and the standardized-root-mean-square residual (SRMR) (DiLalla, 2000;Hu & Bentler, 1999). These indices suggest an excellent t by CFI = ≥ 0.95, TLI = ≥ 0.90, RMSEA = ≤ 0.06 and SRMR ≤ 0.08 (Hu & Bentler, 1999;Tabachnik & Fidell, 2013), minimize the errors of deciding the good t of a model. The reliability was assessed by the Cronbach's alpha (α) and McDonald's omega (ω) to measure the internal consistency of the constructs; however, the latter was used for a construct with error covariance (Hayes & Coutts, 2020). Finally, the model was validated by the average variance extracted (AVE) and construct validity (CR), with an excellent indicator of AVE = 0.50 and CR = 0.70 (Hair Jr. et al., 2014).

Results
Description of the challenges of online education/distance learning Table 2  Exploratory factor analysis Maximum likelihood extraction with varimax rotation was executed on 18 items from a sample of 419 university students. Initially principal components extraction was performed before maximum likelihood extraction, to estimate the number of factors, presence of multicollinearity and factorability of the correlation matrices (Field, 2013;Tabachnik & Fidell, 2013), to determine the number of factors with three principles: the Kaiser's criterion (Kaiser, 1960), the Cattell's scree plot test (Cattell, 1966) and the Horn's parallel analysis (Horn, 1965). The Kaiser's criterion based on eigenvalues was inconclusive as it suggests 5 factors with an eigenvalue of 1 and above. The scree plot test endorses a three-factor solution by accepting the 'higher scree' and ignoring the 'lower scree' (Horn, 1965). The suitability of a three-factor solution was further assessed by comparing the eigenvalues from the extracted eigenvalues generated from the same size of random data set, and it also suggested a three-factor solution as the rst three factors with the eigenvalues exceed the values from randomized data (Horn, 1965;Watkins, 2000).
However, the pattern coe cient of ≥ 0.40, and an internal consistency of ≥ 0.70 (DeVellis, 2003) were considered for a meaningful and consistent factor structure. The items were sorted and grouped by size of loading, and three items were deleted as their loadings were under 0.40, while two others were deleted for high cross-loading (Hair Jr. et al., 2014). Later, ve more items were deleted from the extraction as three of them had low commonalities (≤ 0.50), while the other two were eliminated because of cross-loading and loading under 0.40, respectively (Hair Jr. et al., 2014).
Based on the criteria, a two-factor solution was retained from the exploratory factor analysis by using maximum likelihood extraction with varimax rotation -the rst factor contained six items, while the second factor consisted of two items (see Table 3). The two-item factor was retained because the items were highly correlated (r = 0.70) (Bollen, 1989;Worthington & Whittaker, 2006). The Kaiser-Meyer-Olkin (KMO) for the two-factor solution was 0.86 ('meritorious' according to Kaiser and Rice (1974)) with Bartlett's test of sphericity of χ 2 [36] = 1962.225, p < 0.001 signifying the sampling adequacy. Factor 1labeled as 'Financial-Technical obstacles' -explaining 50% of the total variance entailed the items associated predominantly with nancial and technical aspects. Factor 2 -labeled as 'Pedagogical obstacles' -explaining 10% variance -referred to pedagogical issues associated with OE/DL.
Cumulatively the two-factor measurement of challenges of OE/DL, explaining 60% of the total variance (Hair Jr. et al., 2014) indicates that the measurement could be applicable for similar studies.
Based on the results of the EFA, the scores of the two-factors were measured by unit-weighted items together. The Cronbach's α coe cients of the ' nancial-technical obstacles' and 'pedagogical obstacles' were 0.882 and 0.819, respectively, and the overall coe cient was 0.881 ('good' according to DeVellis (2003)) (see Table 3). Table 4 presents the correlation among the factors, and the results suggest that the factors are intercorrelated (r = 0.522). These ndings indicate that the challenges assessed in the twofactor solutions form a coherent structure.

Con rmatory factor analysis
The Model 1 of CFA, presented in  (Hu & Bentler, 1999). The SRMR suggest an excellent t, however, the CFI, TLI and RMSEA all were greater than their respective cut-off points. The covariance in the modi cation indices suggests a covariance between FINO 1 and FINO 2, between TECHO 1 and TECHO4, between FINO 1 and TECHO4, and between PO2 and PEDO1. A study, during the swine u pandemic, suggests a strong relationship between nancial resources with access to and use of technology (Van et al., 2010), and studies during the COVID-9 pandemic are not the exceptions. A study in Nigeria identi ed electricity and access to digital devices and internet connectivity as important determinants for OE/DL (Azubuike et al., 2020). Likewise, the high costs of digital devices, lack of accessibility to the internet and electricity connection are also being identi ed in a study in Ghana as well as in Bangladesh (Adarkwah, 2020;Al-Amin et al., 2021). A qualitative investigation on both teachers and students in Bangladesh revealed that the nancial constraints limited the access to and use of both digital devices and internet connectivity for university students, particularly in remote rural areas. The study further added that university students often struggle during OE/DL, especially in the practical classes where a face-to-face interaction with both teachers and classmates would be a tting solution (Ela et al., 2021). Based on these empirical underpinnings, the error covariances were correlated. After correlating the error variances, the Model 2 of CFA, presented in Table 5 The two-factor solution with eight items was subsequently validated by calculating the AVE and CR (  (Hu & Bentler, 1999).

Page 8/23
The year 2020 is marked by the COVID-19 pandemic that ceased all forms of educational activities following the implementation of countrywide lockdown throughout the world. Governments of both developed and developing countries were compelled to suspend academic and administrative activities in schools, colleges, and universities out of a sheer fear after witnessing an unprecedented growth of infections as well as deaths from the COVID-19 Gritsenko et al., 2020;Odriozola-González et al., 2020). After breaking the initial spell of fear and panic, the OE/DL was promoted extensively by the government of Bangladesh using the online platforms as an alternative means to oncampus academic activities to avoid session jams, as well as to minimize uncertainties over academic and professional careers of young educands . The sudden move from traditional teaching and learning to OE/DL in Bangladesh has drawn mixed opinions from students, teachers as well as administrators, however, mostly negative (Jameel & Real, 2020). An online poll on both public and private universities in Bangladesh indicate a negative attitude of both educators and educands towards OE/DL (The Daily Star, 2020a), while in a developed country like Germany, the perception is another way around (Schlenz et al., 2020). The implication is that the students and teachers in developing countries could not adjust with the 'neo-normality,' i.e., the introduction of OE/DL, with ease compared to their equivalents from developed countries, and there are some valid reasons.
One of the major issues prompting the potential pessimist attitude towards OE/DL among students in Bangladesh is the nancial constraints (Ela et al., 2021;Mamun et al., 2020). The COVID-19 pandemic has been affecting not only the mental well-being of people, but also threatening the means of livelihood, especially in the developing and less developed countries, by reducing income, increasing unemployment and by bankrupting the small and medium-sized businesses (El Keshky et al., 2020;Pak et al., 2020). In Bangladesh, a lower-middle income country (Bhattacharya & Khan, 2019), the lower and lower-middle income families are also struggling nancially and could not afford the basic necessities (Ela et al., 2021).
The rst factor of the two-factor solution of our study, based on EFA and CFA, is comprised of six items, and it indicated that nancial and technical obstacles, interlinked with each other, were the major challenges for university students in Bangladesh to participate in the OE/DL.
Because nancial constraints are substantially in uencing the access to and use of technical or technological issues. Because the technologies and gadgets required for starting-up the OE/DL are not inexpensive (Ela et al., 2021;Singh & Thurman, 2019), thus, limited budgets may not allow people to get connected to the online platforms. In such case, frustration may lead to self-harm, like suicide. In India, for example, a female teenager committed suicide for being unable to participate in online classes as she did not have either a smartphone or television to connect with OE/DL (Lathabhavan & Gri ths, 2020). In Bangladesh, only 5.7% of households had a computer, while the internet was used by a mere 4.8% (Bangladesh Bureau of Statistics, 2015). Bangladesh stands 113 in the global networked readiness index (NRI) and 150 in the e-government development index (EGDI) (Access to Information Programme et al., 2013). Apart from the unavailability or inaccessibility to digital devices, the unstable and inconsistent internet connection and electricity could also be a major problem for students to participate in OE/DL. A survey on university students in Bangladesh suggests that the frequent load shedding and unstable internet connection were the two leading problems for OE/DL during the COVID-19 pandemic (Al-Amin et al., 2021). A study on university students in UAE, one of the richest countries with almost cent percent households having internet accessibility, also identi ed internet connectivity as a concerning issue to participate in OE/DL (Hussein et al., 2020). The simultaneous and overuse of internet connections by people during the lockdown (Alheneidi et al., 2021) may have led to the problems of internet connectivity.
In addition to nancial and technical issues, in the rst construct the di culties of students to understand the OE/DL, particularly the applied courses, in absence of classroom interaction, with both teachers and classmates, were also identi ed. Students across the globe reiterated the signi cance of interactions and communications with peers and teachers to improve their self-e cacy and learning motivation (Ela et al., 2021;Hussein et al., 2020). An experiment in Canada suggest that student performance decreased in online teaching environment compared to traditional face-to-face teaching environment (Lu & Lemonde, 2013).
The last factor of the two-factor solution -pedagogical obstacles -contained two items, and it was identi ed by the students as a major challenge for OE/DL in Bangladesh. Despite the initiatives of the government of Bangladesh to implement the OE/DL at the universities with the help of BdREN, the unfamiliarity with the online platforms and ineptitude to use it properly make it di cult for both students and teachers to interact and participate in the OE/DL (Ela et al., 2021). In Cameroon, the OE/DL was effectively used only by computer science teachers because of the familiarity with new technologies (Béché, 2020). In Estonia, only the privileged teachers with access to digital tools experienced 'a smoother transition' for OE/DL (Rannastu-Avalos & Siiman, 2020). Furthermore, the absence of effective pedagogical approaches and insu cient time to prepare the learning contents by teachers were several other challenges to materialize the bene ts of OE/DL for both students and teachers (Huang et al., 2020).
The ndings of our study complemented the previous studies.

Strengths And Limitations
There are several issues that are determining the strengths and limitations of this study. This study was conducted through web-based survey -a tool proved effective and widely used during health emergencies -to reach out and cover university students at regional and national levels by maintaining the 'social/physical distance' for rapid assessment of the emerging situation. The samples cover heterogenous population, ranging from rural, sub-urban and urban residents, from freshers to seniors, from different streams of education including science, humanities, and commerce, from public and private universities of different types, such as general as well as specialized universities. Moreover, the data were collected using globally standardized and validated research tools and methods to assess the opinions of the university students regarding OE/DL. Hence, the experiences and opinions of diversi ed samples could lead to the generability of the ndings. Yet, there are some limitations that should be kept in mind when generalizing the results. First, the data were collected from university students participating in OE/DL during the pandemic and could not reach to the students who did not participate in OE/DL due to poor or no internet connectivity. The study did not cover other educational institutions, including schools, colleges, Madrasahs in Bangladesh, and could lead to the limited representation of a speci c age group. Moreover, the cross-sectional nature of the samples together with a tendency to provide socially desirable response could also produce biasness, and such could limit the interpretation of the ndings.

Future Research
The future research, in the context of Bangladesh, is expected to address the limitations of the current study. The future studies should be able to collect the information from both the participants and nonparticipants of OE/DL, on a nationally representative samples and should not rely on online-platforms only for data collection. Moreover, the researchers are expected to reach out students at schools, colleges, and Madrasahs to get an all-inclusive view regarding the OE/DL to formulate the national education policy and strategy of Bangladesh to cope with the unprecedented condition such as the COVID-19 pandemic.

Conclusion
Despite the shortcomings, there is no denial that OE/DL is the only options for the government of Bangladesh to maintain the teaching and learning process without pushing the student and teacher to participate in the class physically during the COVID-19 pandemic. Thus, with its other development stakeholders, the government should plan and implement an appropriate guideline to adopt OE/DL for active teaching and learning and for better educational outcomes in future. However, the initiation of OE/DL under the emergent health crisis could not meet the requirements for an all-inclusive and quality education for all -a goal set forth by the sustainable development goals (SDGs). The university students in Bangladesh, as reported in this study, were struggling to participate in OE/DL largely due to nancial, technical as well as pedagogical limitations. The integration of OE/DL in the educational systems in Bangladesh will highly be appreciated and embraced only if it ensures the necessary provisions. For inclusive, secured, and improved educational outcomes, which the current government is thriving for, the policymakers must make some adjustment in the current educational systems, including (i) development and implementation of online platform-based curriculum; (ii) provide extensive training and other necessary facilities for both educators and educands to familiarize with OE/DL; (iii) enhance the capacity of universities and other educational institutions to provide OE/DL by increasing the educational budget; (iv) modernize the existing technological infrastructure to strengthen the internet connectivity and universalize the use of modern means of education at urban, sub-urban and rural educational institutions, (v) introduce online platform-based network among teachers and students at all levels to increase mutual trust and tp make learning enjoyable and e cient, (vi) provide educational loans for teachers and grants for students nancially struggling to afford the digital devices and internet connections, and nally (vii) reduce the existing income inequality between different social strata.     Model 1 is the baseline with no correlation among covariance