Investigating the barriers that intensify undergraduates’ unwillingness to online learning during COVID-19: A study on public universities in a developing country

Abstract Online learning has been extensively conducted to continue the academic activities in the universities transversely the realm during the pandemic instigated by COVID-19. Like other countries’ universities across the world, universities in a developing country such as Bangladesh are going through the online learning phenomenon. However, the current virtual platform of learning implies enormous challenges for undergraduates to participate in the online learning process due to numerous barriers. Thus, the current paper intends to examine barriers that intensify unwillingness to online education at the university level in the context of Bangladesh. Based on the underpinning theories of TAM3 and UTAUT, this study identified four major barriers such as financial, insufficient institutional support, technological, and individual. Furthermore, gender is considered as moderating variable in the model. To inspect such barriers along with the moderating variable, this study employs the Partial Least Squares (PLS) technique to estimate the measurement and structural model parameters and to generate the coexisting bootstrap assessments. The bootstrapping procedure has been initiated to test the statistical significance of the model parameter estimations. The findings confirm that technological, and insufficient institutional support barriers are statistically significant whereas the other two barriers have been revealed as non-significant to intensify learners’ reluctance to the online study. Furthermore, current findings also ratify no significant effect of gender as a moderating variable. Finally, this study augments specific policy implications for diverse stakeholders in current settings.


PUBLIC INTEREST STATEMENT
The education system across the world has been severely disrupted by the COVID-19 pandemic and has almost shifted from traditional face-toface learning methods to online learning platforms. This study has examined the significant obstacles that intensify learners' unwillingness to participate in online classes during the pandemic. The research has also inspected whether gender impacts the relationship between learners' reluctance to online classes and the different barriers faced by them. The research model was developed using the established theories, namely, TAM3 and UTAUT model. Several hypotheses were advanced to attain the research objectives. Major barriers faced by learners while participating in online classes were categorized into financial, insufficient institutional support, technological, and individual. The findings of the current study reveal that technological and insufficient institutional support barriers were the most significant. Moreover, gender had no significant effect on the association between learners' unwillingness and different obstacles they faced during online classes. Hence, this study yields augmented policy implications for diverse stakeholders in current settings.
technological, and individual. Furthermore, gender is considered as moderating variable in the model. To inspect such barriers along with the moderating variable, this study employs the Partial Least Squares (PLS) technique to estimate the measurement and structural model parameters and to generate the coexisting bootstrap assessments. The bootstrapping procedure has been initiated to test the statistical significance of the model parameter estimations. The findings confirm that technological, and insufficient institutional support barriers are statistically significant whereas the other two barriers have been revealed as non-significant to intensify learners' reluctance to the online study. Furthermore, current findings also ratify no significant effect of gender as a moderating variable. Finally, this study augments specific policy implications for diverse stakeholders in current settings.

Introduction
Higher education that leads learners at the tertiary level to be awarded an academic degree has been drastically affected due to the COVID-19 pandemic. This ongoing pandemic has disrupted the whole education system (Macken et al., 2021;Sahlberg, 2021) and affected approximately 220 million tertiary-level students around the globe (World Bank, 2020). However, the effects are even more severe for the developing, emerging, and underdeveloped economies (Aristovnik et al., 2020;Khan, 2021;Olaniran & Uleanya, 2021;Qazi et al., 2020;Zarei & Mohammadi, 2021). For combating the adverse effects, like other institutions across the world, universities of those economies have initiated online teaching and learning facilities (Baticulon et al., 2021;Dhawan, 2020;Zalat et al., 2021). Online learning has numerous advantages such as convenience, ease of participation, less expensive, availability of blended learning prospects, and technology-driven benefit (Alam et al., 2021;Dumford & Miller, 2018;Fedynich, 2013;Gherheș et al., 2021;Kumar, 2010;Palvia et al., 2018). However, online learning has several shortcomings and challenges like lack of internet access, the high price of ICT devices, lack of social interaction, lack of IT skills, and ineffective assessments, etc. (Adedoyin & Soykan, 2020;Barrot et al., 2021;Fedynich, 2013;Jordan et al., 2021;Kumar, 2010).
A study conducted by Adarkwah (2021) found that traditional teaching is considered more suitable than online learning by more than half of the study responders. Only a few participants believe that outcomes of online learning would be better than the traditional teaching approach. As students face challenges, they prefer the traditional learning approach rather than online learning (Adarkwah, 2021). These findings are also supported by another research where more than half of the study responders pronounced that they do not want to attend online learning in the future if they have options (Chung et al., 2020). Most participants believe online learning has failed to benefit intermediate and secondary schools (Rouadi & Anouti, 2020). However, Khalil et al. (2020) conducted a thematic content analysis, and their research results reveal that most preclinical students prefer the online learning method for the subsequent educational sessions. In contrast, Nambiar (2020) exhibits that most respondents (59%) are unwilling to attend online classes.
Developed countries can cope up with these challenges because of government support (United Nations Educational, Scientific and Cultural Organization-UNESCO, 2021a). Besides, due to governmental support, innovation, and internal cooperation, some emerging and developing economies have transformed the educational challenges into opportunities. In addition, several universities and research centers have offered and conducted different training sessions on online learning. Adequate infrastructure, more accessible access to the internet and ICT devices, sufficient IT skills among learners and instructors, and a blended teaching method are emphasized during this pandemic (United Nations Educational, Scientific and Cultural Organization-UNESCO, 2021b).
However, like other developing and lower-middle-income countries worldwide, Bangladesh is facing challenges to run and execute adequate online teaching-learning facilities at the tertiary level. Among the students at the tertiary level in Bangladesh, undergraduates who are studying at different public universities have been badly affected due to the epidemic (Ela et al., 2021;Sarkar, 2012). Moreover, the prolonged closure of the public universities since 18 March 2020 has also momentously affected the students' willingness to the online studies and learning. Thus, unwillingness to online learning has been a major challenge in the context of Bangladesh (Khan & Abdou, 2021;Jeenia et al., 2021;Rahman, Uddin, & Dey, 2021;Begum et al., 2020;Dutta & Smita, 2020) and hence, Tang et al. (2021), Kundu and Bej (2021), and Thapa et al. (2021) have called for further research to identify the possible barriers that intensify undergraduates' unwillingness to online learning during COVID-19.
Accordingly, the research aims to identify the barriers that intensify undergraduates' unwillingness to online learning. In addition, this paper will also investigate the most significant barrier(s) that intensify undergraduates' unwillingness to online learning. However, respondents' gender is considered as a moderating variable between the relationship of different barriers and reluctance to online education.

Theoretical background and inferential objectives
The research model of this study has been adapted from the Technology Acceptance Model (TAM3; Venkatesh & Bala, 2008) and prior literature studies. As a theoretical model, TAM3 explains that users' intention to use IT depends on the perceived usefulness and ease of use. In the TAM3 model, several external factors such as individual differences, system characteristics, social interaction, and facilitating conditions impact perceived usefulness and usability.
For theoretical underpinning, as an assimilated model of technology acceptance (Al-Emran et al., 2018;Tan, 2019;Yu, 2020), TAM3 offers an inclusive and related network of the factors of individuals' IT adoption and use (Al-Gahtani, 2016;Chang et al., 2017;Hamutoglu, 2020;Venkatesh & Bala, 2008). Therefore, TAM3 yields coherent clarifications into how and why individuals make decisions on the adoption and use of IT, specifically on the dynamics of perceived usefulness and perceived ease of use.
The Unified Theory of Acceptance and Use of Technology (UTAUT) model developed by Venkatesh et al. (2003) aims to clarify individual intentions to use information systems and succeeding usage behavior. In the UTAUT model, predictors such as performance expectancy (PE), effort expectancy (EE), social influence (SI) determine users' behavioral intention. In articulating the UTAUT model, Venkatesh et al. (2003) argued that individuals would anticipate facilitating conditions (FC) to foresee behavioral intention (BI) if effort expectancy (EE) was not incorporated in the model Baabdullah et al., 2019;Dwivedi et al., 2019).
From TAM3, we adapted user intention (learners' unwillingness to online learning) and external variables (different types of barriers) as these are relevant to our study topic. From the UTAUT model, we adapted users' behavioral intension (learners' unwillingness to online learning) and predictors (technological barriers, financial barriers, insufficient institutional barriers, and individual barriers) that determine users' behavioral intention. After studying prevalent research works (Gaur et al., 2020;Ivala, 2013;Muilenburg & Berge, 2005;Muthuprasad et al., 2021;Sarkar et al., 2021;Xu & Jaggars, 2013), gender is taken as moderating variable that is incorporated in the research model. Figure 1 shows the adapted research model of the study. Nambiar (2020) showed that satisfaction with online classes depends on several factors such as quality and timely interaction between learners and educators, technological aid, focused online class modules, and modification of practical classes for conducting online. Among these factors, technical support is broadly impacting the satisfaction of learners with online classes. Though convenience and flexibility are positive outcomes of online learning, lack of inadequate network access (Aboagye et al., 2020;Adnan & Anwar, 2020;Alchamdani et al., 2020;Farooq et al., 2020;Hussein et al., 2020), lack of technological knowledge and experience (Bean et al., 2019;Nambiar, 2020;Rajab et al., 2020;Sinha & Bagarukayo, 2019;Srichanyachon, 2014), and electricity (Simamora, 2020) hinder the benefits of online learning (Adarkwah, 2021). Lack of previous experience in attending online classes as well as expertise to operate and use online tools create barriers (Muflih et al., 2020). However, Adarkwah (2021) demonstrated mixed findings regarding the effect of electricity availability on online learning. Thus, from the existing literature, we can hypothesize that H1: The higher the technological barriers, the higher the learners' unwillingness to online learning. Simamora (2020) demonstrated that economic condition is also a barrier for students participating in online classes. Adnan and Anwar (2020) concluded that online learning could not generate expected outcomes in underdeveloped countries. Their study identified that financial issues are a major factor that hinders the benefit of online learning (Adnan & Anwar, 2020;Sarkar et al., 2021). For successfully implementing e-learning, financial issue is one of the major critical factors (Sarkar, 2012). The high price of ICT equipment is another hindrance reported by learners (Bean et al., 2019;Sinha & Bagarukayo, 2019). In addition, the respondents (73%) opined that the outbreak had decreased their income level. Therefore, students were facing a financial crisis as they lost their part-time earning sources. Moreover, around 47% of respondents stated that they could not purchase the internet bundle (Begum et al., 2020). Hence, the following hypothesis can be derived from previous studies: Simamora (2020) demonstrated that economic condition is also a barrier for students participating in online classes. Adnan and Anwar (2020) concluded that online learning could not generate expected outcomes in underdeveloped countries. Their study identified that financial issues are a significant factor that hinders the benefit of online learning (Adnan & Anwar, 2020;Sarkar et al., 2021). For successfully implementing e-learning, financial issue is one of the major critical factors (Sarkar, 2012). The high price of ICT equipment is another hindrance reported by learners (Bean et al., 2019;Sinha & Bagarukayo, 2019). In addition, the respondents (73%) opined that the outbreak had decreased their income level. Therefore, students were facing a financial crisis as they lost their part-time earning sources. Moreover, around 47% of respondents stated that they could not purchase the internet bundle (Begum et al., 2020). Hence, the following hypothesis can be derived from previous studies:

Financial barriers (FB) and unwillingness to online learning (UOL)
H2: The higher the financial barriers, the higher the learners' unwillingness to online learning.

Insufficient institutional support barriers (IISB) and unwillingness to online learning (UOL)
An online medical education study found that inadequate training and institutional support are significant challenges of online learning (Farooq et al., 2020). Education institutes can support learners gain skills during orientation activities on ICT (Sharpe & Benfield, 2012). To get benefits from online learning, workshops, technical support (Nambiar, 2020), and ICT training are important for digital users (Forsyth et al., 2010;Muhammad, 2017). To ensure effective involvement with e-learning tools, education organizations can establish e-learning centers (Piña et al., 2018) and provide professional training to instructors (Adarkwah, 2021) to improve their digital literacy. E-learning centers provide support and training for e-learning with the help of technology experts and instructor support experts (Piña et al., 2018). Effective training programs influence online learning usage (Alhabeeb & Rowley, 2018;Solangi et al., 2018). Therefore, it can be hypothesized: H3: The higher the insufficient institutional support barriers, the higher the learners' unwillingness to online learning. Baticulon et al. (2021) demonstrate that learners face problems coping with new learning modes and heavy workloads (Hussein et al., 2020). Besides these, simultaneously performing classwork and household responsibilities are another cause of unwillingness to online study. Anxiety is created among students due to online lectures (Rajab et al., 2020). However, Nambiar (2020) finds that approximately 46% of respondents feel less anxious during online classes. Stowell and Bennett (2010) argue that learners feel less anxious as the examination method is open book and is conducted in unrestricted mode. Nevertheless, they also face physical problems such as eye strain because of participating in online classes (Octaberlina & Muslimin, 2020).

Individual barriers (IB) and unwillingness to online learning (UOL)
H4: The higher the individual barriers, the higher the learners' unwillingness to online learning.

Gender
From previous studies, it is observed that gender has not been examined as a moderator regarding the relationship of technological barriers and learners' reluctance to online learning as well as the association between financial barriers and students' unwillingness to participate in the online class. Hence, we propose the following hypothesis: H5: Gender significantly moderates the relationship between technological barriers and learners' unwillingness to online learning H6: Gender significantly moderates the relationship between financial barriers and learners' unwillingness to online learning Xu and Jaggars (2013) find that learners' demographic characteristics such as age, gender, location, and past academic performance affected the adaptability to online learning. Another study (Muilenburg & Berge, 2005) was conducted to determine fundamental constructs that constitute barriers to online learning from learners' perspectives. The result showed that gender significantly impacted when learners rated different barrier factors (Muilenburg & Berge, 2005). Moreover, Baticulon et al. (2021) add that female respondents have more mental health concerns than males. Gaur et al. (2020) also stated that a significant difference is found in the respondent's gender when learners address online learning barriers. Therefore, the following hypotheses can be derived from prior studies.

H7: Gender significantly moderates the relationship between insufficient institutional support barriers and learners' unwillingness to online learning H8: Gender significantly moderates the relationship between individual barriers and learners' unwillingness to online learning
Thus, the conceptual framework demonstrates that financial issues, technological barriers, individual obstacles, and lack of institutional support increase learners' unwillingness to online learning. In addition, demographic characteristic such as gender played a moderating role as shown in Figure 1.

Study area
The target population of the study is undergraduate learners who participate in online classes during the pandemic. Due to resource constraints, we focus only on current undergraduates at selected public universities, namely the University of Dhaka, Jahangirnagar University, and Jagannath University.

Data collection and sample size
For conducting the study, data were collected from currently enrolled students at the undergraduate programs. G*power 3.1 was used to determine the sample size of the research model (Faul et al., 2009). The software recommended 129 sample to conduct this study with configuration (effect size f2 = 0.15, power = 0.95, α err prob = 0.05, number of predictors = 4). The actual sample size was 150 which was higher than the recommended sample size. Although the survey questionnaire was sent to 210 respondents through e-mail, we got completed questionnaires from 150; therefore, the response rate was about 71.4%. Here, a non-probability sampling approach was used to select elements from the sample where researchers' judgment was crucial. In addition, the participants with particular characteristics or qualities were chosen from the sample to attain the research objectives; hence, this study applied the non-probability purposive sampling technique. Responses were collected from June 15 to 31 July 2021.

Measurement of the research constructs
The questionnaire was adapted from earlier studies (Baticulon et al., 2021;Aboagye et al., 2020;Begum et al., 2020;Dost et al., 2020;Rafi et al., 2020;Gaur et al., 2020;Muflih et al., 2020;Adarkwah, 2021) and a pre-testing was conducted to ensure clarity, accuracy, and readability before distributed questionnaire to the target respondents. The questionnaire was divided into three sections: demographics section, statements associated with barriers to online learning, and learners' willingness to online learning. A 5-point Likert scale was used to assess the obstacles and learners' unwillingness towards online learning, ranging from "1 = Strongly Disagree" to "5 = Strongly Agree".
The measurements of the major constructs are described in the following sections:

Financial barriers (FB)
For assessing financial barriers, three indicators were adapted from literature reviews, and these are (1) Unable to bear the internet expense; (2) Unable to buy ICT devices due to high price; and (3) Lack of basic needs due to the pandemic (Aboagye et al., 2020;Baticulon et al., 2021;Begum et al., 2020).

Insufficient institutional support barriers (IISB)
Three components were used to measure insufficient institutional support barriers: (1) Insufficient support from institutions for online classes; (2) Lack of training about how to operate online platforms; and (3) Instructors have inadequate skills (Aboagye et al., 2020;Baticulon et al., 2021).

Individual barriers (IB)
Individual barriers were measured by three components adapted from (Aboagye et al., 2020;Baticulon et al., 2021). The components are: (1) Online classes usually lead to more physical problems like tiredness, eye pain, and headache; (2) Online classes trigger anxiety or stress in me; and (3) Doing simultaneously household responsibilities and classwork create obstacles in online learning.

Unwillingness to online learning (UOL)
From (Aboagye et al., 2020;Begum et al., 2020;Muflih et al., 2020) prior studies, three components were adapted to measure online learning unwillingness: (1) I prefer online learning should be stopped, (2) Online learning cannot attain learner goals, and (3) Lack of self-inspiration for online learning.

Results
The research model was examined by adopting the Structural Equation Modeling (SEM) technique using Partial Least Squares (PLS). SmartPLS 3 , a nonparametric second generation statistical software that is capable of examining the complex models containing latent variables, has been used. In the PLS-SEM model, the independent variables are technological barriers, financial barriers, insufficient institutional support barriers, and individual barriers, whereas the dependent variable is learners' unwillingness to online learning. There is one moderating variable, i.e., gender. After testing the measurement model, we also investigated the structural model as Anderson and Gerbing (1988) suggested two-stage analytical procedures (Hair et al., 2017;Ramayah et al., 2011). Podsakoff et al. (2003) recommended Harman's single factor test using SPSS to detect CMV. In Harman's single factor, a fixed component is extracted from all the main variables. This factor is expected to forecast less than 50% of the variance of the constructs. SPSS results showed that one factor could describe 23.06% of the variance, which is lower than the threshold value. Thus, this study has confirmed no common method bias (Bagozzi et al., 1991).

Demographic profile of the study participants
Of the 150 respondents, the male and female respondents were 76 and 74, respectively. The respondents' age was categorized, and 38 percent of them fell between the 17-21 years range. The majority (42%) of respondents were living in semi-urban areas. Approximately 45 percent of study participants had one ICT device, and 32.7 percent of respondents used mobile internet data.
Then, 40 percent of them reported that they are "studying" in terms of employment status. Further information about study participants is given in Table 1. Hair et al. (2019) suggested that examining the constructs' loadings is the first step to assess the measurement model. In addition, convergent validity and discriminant validity are used to The values of the loadings should be between 0.70 and 0.90, and most of the loadings of our study are greater than 0.70. When the value of loading is below 0.40, it should be considered removed from the model, and indicators with loadings between 0.40 and 0.70 should be considered for removal if this deletion increases the value of CR and AVE. According to Hair et al. (2011), weaker indicators could be retained if the deletion of the item affects content validity (Chin, 2010;Rasoolimanesh et al., 2017).

Measurement model assessment
As some outer loading values of the study range from 0.40 to 0.70, their deletion decision depends on the values of CR and AVE. All the CR values of constructs exceed 0.70, which conforms reliability of the measurement model. The majority of the AVE values are higher than 0.50, which is recommended threshold requirement for AVE. Therefore, the items with loading lower than 0.70 in the study can be retained. Table 2 shows the construct reliability of the study, including outer loadings, AVE, and CR. By following the three criteria, the discriminate validity test results are demonstrated in Tables 3, 4, and 5. and HTMT .85 Criterion are used to establish discriminant validity of the measurement model. In Fornell-Larcker Criterion, the diagonal value should be greater than its off-diagonal values, and the outcomes reveal that the study constructs attain sufficient discriminant validity (Fornell & Larcker, 1981). In Fornell-Larcker Criterion, the diagonal value should be greater than its off-diagonal values. The outcomes reveal that the study constructs attain sufficient discriminant validity (Fornell & Larcker, 1981), shown in Table 3.
This study follows HTMT. 85 criterion, which is the more conservative method to examine discriminant validity. As presented in Table 4, all the constructs have met the HTMT .85 criterion (Henseler et al., 2015;Kline, 2015), and thus, the study confirms discriminant validity from all three criteria. In the case of cross-loadings, Hair et al. (2011) suggested that an item's loadings should exceed all of its cross-loadings, and the study meets this requirement (see , Table 5).

Structural model assessment
To assess the structural model of the study, the absence of collinearity issues is essential. Hair et al. (2019) recommended that Variance Inflation Factor (VIF) values be less than 3 to ensure non-collinearity in the study. The inner VIFs values of independent variables are 1.084, 1.461, 1.075, and 1.436 for technological barriers, financial barriers, insufficient institutional support barriers, individual barriers, respectively. Hence, the result indicates that the study has no major collinearity issues as all the values meet the requirement (Hair et al., 2019).
To test the hypotheses, we apply bootstrapping procedure with 5000 resamples to estimate co-efficient, t-values, p-values, and R 2 for the inner model (Hair et al., 2017). The results of the  PLS bootstrapping method support H1, H3. However, the results obtained from the survey data cannot support H2 and H4. The result indicates that technological barriers (β = 0.374, t = 5.474, p < 0.01, f 2 = 0.192) positively and significantly impact the learners' reluctance to online learning. On the contrary, as the p-value is greater than 0.05, financial barriers (β = 0.080, t = 1.033, p > 0.05, f 2 = 0.008) have no significant effect on the learners' unwillingness to online learning. Learners' reluctance to attend online classes is significantly impacted by insufficient institutional support barriers (β = 0.346, t = 5.255, p < 0.01, f 2 = 0.169). Nonetheless, the insignificant impact is found concerning the individual barriers effect on the students' negative perception of participating in online classes (β = 0.084, t = 1.034, p > 0.05, f 2 = 0.005). Following the suggestion of Cohen (1988), the effect size of technological barriers and insufficient institutional support barriers are 0.192 and 0.169, respectively, which are greater than 0.15 and indicate a medium effect on the learners' reluctance to online learning. Hair et al. (2014) recommended that to measure predictive accuracy, researchers should use R 2 , and for the model's predictive relevance, they should use Q 2 (Geisser, 1974;Stone, 1974). The R 2 value was 0.354, which represented that variation in the unwillingness to online learning was explained by technological barriers, financial barriers, insufficient institutional support barriers, and individual barriers. The Q 2 value was 0.197, which was greater than zero and thus assured that the model also had predictive relevance. Table 6 summarizes the results of path  coefficients and hypotheses testing results. In the following Figure 2, the structural model is shown with path coefficients and p-values.

The moderating role of gender
Hypotheses five, six, seven, and eight were developed to test the moderating effect of gender. For examining moderating effect, the interaction was created between (1) gender and technological barriers; (2) gender and financial barriers; (3) gender and insufficient institutional support barriers; and (4) gender and individual barriers. The R 2 value was increased to 0.374, with a change of 2%, when the model incorporated the interaction effect.
Gender as a moderator had no significant role to strengthen or weaker the relationship between technological barriers on the learners' unwillingness towards online classes (β = 0.062, t = 0.866, p > 0.05). Similarly, the link between financial barriers and reluctance to online class was not significantly moderated by gender (β = −0.087, t = 0.943, p > 0.05). The interaction between gender and insufficient institutional support barriers (β = −0.053, t = 0.791, p > 0.05) was non-significant as the p-value was greater than 0.1, which indicated a trivial effect. Besides, gender had an insignificant moderating role (β = 0.149, t = 1.587, p > 0.05) between the association of individual barriers and students' reluctance to attend online classes. Therefore, these findings show that H5 to H8 developed by the study are not supported by the data obtained from the survey. Table 7 illustrates the results of moderating role of gender.

Discussion and implications
The present study has revealed that technological barriers such as poor internet accessibility, electricity disruption, and lack of technological knowledge positively significantly affect the learners' reluctance to online learning. This finding is also supported by several prior studies (Aboagye et al., 2020;Baticulon et al., 2021;Dost et al., 2020;Gaur et al., 2020;Muflih et al., 2020;Rafi et al., 2020;Rajab et al., 2020;Sarkar et al., 2021). In consistence of this study results, Gaur et al. (2020) also found that around 40 percent of respondents claimed that technology accessibility and poor internet speed were major barriers to online learning. Likewise, Aboagye et al. (2020) demonstrated that most respondents stated that internet accessibility was not adequate and they found this a major challenge.
Although studies (Adnan & Anwar, 2020;Begum et al., 2020;Sarkar et al., 2021;Simamora, 2020) found financial barriers hindered online learning benefit, the result of the study revealed that financial barriers were non-significant. This output of the study is consistent with the result of Muilenburg and Berge (2005), where they found that cost was a less significant barrier to online learning. Contrary to this, a study conducted by Gaur et al. (2020) showed that study participants stated that online classes were costly due to internet purchasing costs. Moreover, Aboagye et al. (2020) also argued that the cost of the internet bundle was so high which hinder the readiness of students to attend online classes.
In addition to that, insufficient institutional support barriers also have a significant positive influence on the learners' unwillingness to participate in the online class, and this is consistent with the findings of other research work (Alhabeeb & Rowley, 2018;Farooq et al., 2020;Solangi et al., 2018). The possible grounds of this finding might be that in Bangladesh, most students are not familiar with online learning; therefore, they need training and workshops on how to participate and operate an online platform. The online class is a new phenomenon for which lack of facilities and supports from educational institutes leads to an unwillingness to online learning among the learners.
Several studies recognized individual barriers like physical, psychological, and family disturbance as challenges to online classes (Baticulon et al., 2021;Dost et al., 2020;Gaur et al., 2020;Rajab et al., 2020;Sarkar et al., 2021). In Japan, Arima et al. (2020) conducted a large-scale study on medical students and found that students experience some extent of psychological health problems during COVID-19. In addition, students reported that they were worried about their career, family wellbeing, and online class assessments. Arima et al. (2020) suggested providing mental health wellness training programs to improve medical students' self-confidence. Furthermore, Rafi et al. (2020) showed that students face some degree of eye problems and headaches as they use mobile phones to participate in online classes. Moreover, respondents asserted that they feel stressed due to the overload of online classes. Nonetheless, the result obtained from research data indicated that individual barriers were insignificant.
The result obtained from the study showed that gender had a non-significant moderating role in the relationship among different barriers and students' reluctance towards online classes. This finding contradicts the study of Muilenburg and Berge (2005), where they found differences in gender while addressing administrative barriers to online learning. Xu and Jaggars (2013); Gaur et al. (2020) found significant differences in the respondent's gender when learners address online learning barriers. Moreover, Baticulon et al. (2021) found different opinions in gender variables when participants reported individual barriers, which was inconsistent with the result of this study.
From the outcomes of the current study, policymakers for the development of higher education, university authorities, respective ministry of the government, and other related stakeholders should take appropriate initiatives; consequently, learners become willing to participate in online learning and continue their studies during the pandemic. Developing adequate IT infrastructure (Hoque et al., 2016;Rahman et al., 2017;Saif et al., 2021) and providing adequate training regarding technological issues and online platforms would inspire learners to participate in online classes. It will avert the breakdown of studies for the undergraduates at the university level in a developing country like Bangladesh. Besides, this study will assist the learners in getting acquainted with the remote working environment  that is going to gain much popularity in the coming days.

Conclusions
Students of public universities are currently continuing their academic studies on online platforms. Since online learning is a new phenomenon in Bangladesh, learners face several obstacles and are reluctant to continue online learning; this research paper addresses the significant barriers that intensify learners' reluctance toward online learning. The findings of the study suggest that technological barriers, insufficient institutional support barriers are significantly associated with learners' unwillingness to online learning. Internet accessibility, electricity interruption are still challenging to online learning. Some of the obstacles cannot be resolved quickly, but taking appropriate actions could minimize the problem in the long run. Stakeholders in education should take an integrated approach that ensures sufficient IT infrastructure, adequate training, and workshops. However, currently, Bangladesh is observing the demographic dividend (Jafrin et al., 2021). Hence, to utilize it, these identified barriers need to be addressed for the undergraduates. Otherwise, the undergraduates will not be that productive and effective to capitalize the demographic bonus in the near future.
This study extended the research of Baticulon et al. (2021), which was conducted among medical students in the Philippines. Baticulon et al. (2021) categorized online learning barriers during COVID-19. This study identifies the most significant effect of online learning barriers that lead to intensify the reluctance of students to attend online classes. Besides, the present research has augmented the outcome yielded by Balderas-Solís et al. (2021) in the context of Mexico.
However, this study assesses the online barriers from the perspective of learners and excludes instructors. Furthermore, it does not include secondary schools, colleges, and private universities. Only ongoing undergraduate learners are taken into consideration as study participants whereas students of post-graduation programs are excluded from the study. Thus, future studies can take these issues to examine the effects of different barriers on the online learning process to yield a more in-depth and comprehensive representation.