Learning Continuity during COVID-19: An Analysis of the Higher Education Sector of Bangladesh

Aim. This study aims to understand the factors determining university students’ behavioural intentions toward online learning in Bangladesh. Specifically, this study investigates the relationship between performance expectancy (PE), effort expectancy (EE), social influence (SI), facilitating conditions (FC), and behavioural intention (BI). Moreover, this study explores the influence of pandemic fear (PF) as a moderator in the relationship between exogenous and endogenous factors. Methods. The study is cross-sectional and followed a quantitative research approach with purposive sampling. Data were collected at a single point using a sample size of 578 respondents who studied online during the various phases of lockdown at five public and five private universities in Bangladesh. Regarding multivariate analysis, the Partial Least Squares - Structural Equation Modeling (PLS-SEM) is applied in this study to test the causal relationships in the structural model, as it is considered a second-generation technique. Results. Statistically, a positive significance was found between PE, EE, SI, and BI in online learning participation. Whereas the FC and the BI exhibited a negative relation - ship, a positive relationship was found between PE, EE, and the SI on BI. In addition, a moderating role for PF was investigated, and EE and FC were found to influence BI significantly. Conclusion. This study presents an extended UTAUT model by integrating pandemic fear as the moderator to study students’ behavioural intention to adopt


Introduction
T he COVID-19 outbreak forced nearly 1.6 billion learners across 200 countries to discontinue physical classes and move to online courses (Shahzad et al., 2020). Online education incorporates peoples' homes, turning them into classrooms and offices (Hasan et al., 2021;Mukherjee & Hasan, 2020). Now teachers and students have various options like Google Meet, Microsoft Teams, Zoom, Skype, WhatsApp, and others for online education (Sangeeta & Tandon, 2020). To stop the spread of the Coronavirus, the Government of the People's Republic of Bangladesh initially declared all educational institutions to remain closed until March 31, 2020, intermittently extended for another two years (Islam et al., 2020). The temporary closure affected almost a million teachers and 36 million students in Bangladesh (Uddin, 2020).
Technology intervention in the teaching-learning process took a phenomenal surge where the usage of learning technologies was augmented to fulfill the need for emergency remote teaching (Rapanta et al., 2020 ;Wang et al., 2020). The inclusion of media into instructional activities has become an essential element. All instructional activities are coordinated through a centralised communication platform. Students must make considerable progress in learning activities using online knowledge (Ali, 2021). Low cost and support infrastructure to promote learning benefits are two essential elements to justify the development of online learning (Altameemi & Al-Slehat, 2021).

Background of the Study
In Bangladesh, 90% of students participated physical classes before the COVID-19 era (Rahman et al., 2021). As online teaching platforms are growing in Bangladesh at the institutional level, it is paramount to assess the primary stakeholders' interest level and engagement level, i.e., students enrolled for formal education. Mukherjee and Hasan (2020) identified that students select online courses for convenience, flexibility, and easy access to online classes. Researchers found that 55% of students could not attend classes due to poor internet connections, and 44.7% lacked the necessary equipment to participate in online classes among 2038 students drawn from Bangladesh's private and public universities (Islam et al., 2020).
Interestingly students found the practice of evaluation in the online medium is less practical than the physical ones offering lesser impact (Rouf et al., 2022). In the realm of Education 4.0, where the novel andragogical approaches ensure uninterrupted learning even in the formal setting despite hindrances such as natural calamities perpetuated for a long haul, the students attain higher learning gain, which is the prime objective of the learning continuity approach. Switching to different learning modes as the requirement of the recent moment only augments the impact of learning and ensures the students learn well even under stressful situations like the COVID-19 pandemic (Guppy et al., 2022).

Problem statement
The Government of Bangladesh suspended all academic activities issuing a notice followed by the COVID-19 outbreak. Initially, learning continuity suffered due to inadequate technological infrastructure, but a few universities in Bangladesh gradually launched online learning platforms, like IBA, the University of Dhaka (Mukherjee & Hasan, 2020). However, many institutions in Bangladesh still need to evolve to online learning due to the need for teaching-learning technologies and a general willingness of students arising out of various issues, including psychological and financial constraints (Hasan et al., 2021).

Objectives of the Study
A limited amount of research has been conducted on online learning deployment in developing countries and specialised fields. The present study aims to explain the BI of university students in Bangladesh toward the OL. The objectives of the study are as follows: • To identify the present learning continuity situation in Bangladesh during the pandemic. • To examine the factors determining university students' Behavioural Intentions (BI) by investigating the relationship between Performance Expectancy (PE), Effort Expectancy (EE), Social Influence (SI), and Facilitating Conditions (FC) to adopt online learning during COVID-19. • To explore the influence of Pandemic Fear (PF) as a moderator in the relationship between exogenous and endogenous factors.

Identification of Research Variables
The variables used to develop the research model are summarised in Table 1.

Gap Analysis
A review of the existing literature was conducted to identify the gaps in the research domain. Table 2 below represents a focal analysis of the research gaps drawn from eight key research papers.

Unified Theory of Acceptance and Use of Technology (UTAUT) The UTAUT model was developed by Viswanath Venkatesh et al. in 2003.
The following theories collectively comprise the UTAUT: The theory of Reasoned Action (Fishbein & Ajzen, 1977), the Theory of Planned Behaviour (Ajzen, 1985), the Social Cognitive Theory (Bandura, 1986), the Technology Acceptance Model (Venkatesh & Davis, 1996), Diffusion of Innovation Theory (Rogers, 1983)

Relationship between PE and BI
Due to online learning-based platforms, PE is the degree to which students and teachers anticipate their performance in teaching, learning, and assessing will improve. PE is an influential factor when evaluating the effectiveness of online learning systems (Abbad, 2021). Therefore, PE is one of the significant factors influencing students' intention to engage in online learning, according to Balakrishnan et al. (2022). Further, Teo et al. (2019) found that perceived usefulness is a significant predictor of online learning intentions among university students. Based on prior research, it is possible to hypothesise that PE relates to BI regarding online learning acceptance. It is therefore posited that, H1: PE significantly and positively impacts the students' BI to adopt OL. found that EE is consistently and convincingly associated with BI. It is therefore posited that, H2: EE significantly and positively impacts the students' BI to adopt OL.

Relationship between SI and BI
SI is defined as a person's belief that the people close to them significantly impact whether or not they can effectively utilise new technologies (Venkatesh et al., 2003). In previous studies, essential persons such as friends, family, colleagues, teachers, and teaching assistants have recommended using online learning platforms (Lolic et al., 2020). This study will explore how much respondents are influenced to use online learning by their family, friends, and teachers. Dhawan (2020) has found that SI is associated with a high intention to use the Internet for health promotion purposes. The perceptions of others regarding online learning during the pandemic are influenced by socially significant individuals (Mahmood et al., 2021).
It is therefore posited that, H3: SI significantly and positively impacts the students' BI to adopt OL.

Relationship between FC and BI
Facilitating conditions can be considered into four categories: procedural support, human support, technical support, and organisational support. An online learning system must have a technical support system to be accepted and used (Lai,

PF Moderates the Relationship between EE and BI
Another essential factor influencing technology usage and acceptance is the fear associated with its use, which is influenced by anxiety and illiteracy. This physiological aspect must therefore be paid particular attention by teachers and students so that students can freely adopt the technology by developing the necessary skills and information.

Study of Research Model
UTAUT is a research framework that integrates determinants from various technology acceptance theories. Figure 1.1 shows four antecedents of BI, and the PF moderates the relationships between all antecedents and outcomes (intention).

Figure 1
The proposed research model Source. Own research.

Study Area
A sampling frame of 10 Bangladeshi (five public and five private) universities approved by the UGC of Bangladesh was considered for the target population. This research involves the collection of 578 samples, which is more than the recommended number per numerous criteria in the literature, out of which 78 responses of spurious nature were discarded. For the analysis, 500 responses were considered.

Statistical Analysis
SmartPLS (version 3.0) statistical approaches were adopted for data analysis and hypothesis testing via PLS-SEM (Partial Least Squares -Structural

Common Method Bias/ Variance
The "Harman single-factor test" was employed to address this potential concern. Five factors with eigenvalues greater than one were identified through a factor analysis of all measuring items. The following table shows that the first component explains only 34% variance, which is below the recommended threshold of 50%. Therefore, Common Method Variance (CMV) was shown to be of minor importance in this investigation since no one component appeared and did not account for a significant portion of the variation. A summary of the study variables and the mean and standard deviation can be found in Table 4. All items were evaluated using a Likert scale of seven points. All variables had an average greater than 4.0. SI had the highest mean value of 4.809 and the lowest standard deviation score of 1.625 of the seven scale values. BI has a mean of 3.886 and a standard deviation of 1.945.

Goodness of Measurement Model
The measurement model was tested for individual loading, consistency, composite reliability, and discriminant validity (Hair et al., 2021).

Convergent Validity
Analysing item loadings and calculating average variance extracted (AVE) was used to determine convergent validity. The AVE value should be greater than 0.5. All variables have an AVE of at least 0.5. It is shown in Table  4 that the AVE for each construct was achieved in this investigation.
The researchers recommended composite reliability (CR) cutoff value of 0.70 (Hair et al., 2017). CR values range from 0.945 to 0.815, as shown in   Table 6, the HTMT must be less than or equal to 0.90 (Hair et al., 2019).

Direct Effect
Structural model coefficients were estimated using regression equations. Variance inflation factor (VIF) is typically used to determine whether an error was caused by collinearity when examining structural relationships . Each construct's VIF value was less than or equal to 3, suggesting no collinearity problems existed. Path coefficients were tested for statistical significance using the bootstrapping approach (minimum resampling = 5,000) (Hair et al., 2017). Correlations were calculated between endogenous and exogenous components based on a 0.05 (p<0.05) level of statistical significance.

Testing the Moderating Effect
An investigation was conducted to determine whether PF moderated the relationships between PE, EE, SI, FC, and BI to adopt online learning.  The graphical representation of the moderation interaction plot is more important than the calculation, according to Ahmed et al. (2022). The results indicate that students with low PF perceived higher intention than students with high PF when their level of EE increased, and students with high PF perceived higher BI than those with low PF when the level of FC increased.  When the Q 2 value exceeds zero, exogenous constructs are found to be predictive of endogenous constructs (Hair et al., 2013). As the Q2 value is 0.439, the model is statistically highly predictive. The findings of Table 8 indicate that the model is predictively relevant.

Discussion
Presently educational institutions in Bangladesh are improving the interface and ensuring that the new technology introduces learning continuity functionality to increase student success. When students' learning efficiency increases, they are more likely to succeed in their studies, especially during an outbreak such as PF. Students are improving the online learning system concerning the effort needed to ensure learning continuity. When students perceive technology as easy and beneficial, they are more likely to adopt it. As a result, students will remain flexible in the future because of learning continuity during pandemics when institutions are closed.
The first findings from the present study found that PE and BI showed a significant correlation with a solid positive path coefficient. As a result, PE is one of the strongest determinants of whether an individual will accept or reject OL. When perceived as beneficial, it will be more likely to be adopted. Based on such a finding, it is suggested that PE should be enhanced to increase students' engagement with OL and, thereby, learning continuity. The second important finding was that EE significantly influenced BI. Students found it easy to use to continue their OL as a result of this study. The result indicates that learning to be skillful in using OL is easy. To measure EE, it is essential to consider whether OL is perceived as easy or complicated. OL will be more likely to be used by students if they expect it to perform well during COVID-19. The third significant finding is that SI is significantly associated with the BI of students who adopt online learning. SI and BI were found to be significantly correlated, with a significant positive path coefficient. The fourth important finding is that FC does not significantly influence BI. Since OL students use their facilities outside of the campus, such as their homes or offices, campus facilities do not have an impact in this context, as they use their facilities. A previous study also found that FC does not influence learners' BI to adopt OL in Bangladesh (Amin & Zaman, 2021).
The moderating effect of PF does not support PE and BI in adopting OL. Using a specific technology is an individual's goal, and students believe it will allow them to perform better in class. If students believe that online learning enhances their performance in class, they are more likely to use it in the future. The PF determines a negative path coefficient that moderates the relationship between EE and BI to achieve OL adoption. In addition, the relationship between EE and BI is found to vary according to a student's PF level. Based on the highly collective culture of the population, it is likely that university students may use OL to the extent that their peers or instructors inform them that it is simple to use. The relationship between SI and BI is not affected by whether students are in the low-fear or highfear group. Instead, this relationship is the same for both groups of PF. This study found that students who experience less fear of COVID-19 are more likely to adopt OLS due to SI after listening to the perspectives of their peers, friends, instructors, and fellow students. Finally, the PF moderates the association between FC and the BI to adopt OL with a significant positive path coefficient. It reveals that the relationship is stronger for the high group of PF, suggesting that FC should be enhanced with more attention given to the high-fear group of students.

Implication
The study integrates pandemic fear with the UTAUT model while measuring the students' behaviour towards online learning. In this pioneering effort, the study also records the gradually receding fear of the pandemic among the students while resiliently returning to normalcy. There are several practical implications of the study for faculty members, decision-makers of tertiary education, and offices of teaching and learning. OL adoption in PE, EE, SI, FC, and BI is examined and PF acts as a moderator between PE, EE, SI, FC, and BI in this study. When implementing OL to promote learning continuity in Bangladesh's higher education sector, decision-makers must consider students' objectives and duties. By considering these responsibilities and priorities, decision-makers can make a more informed decisions regarding OL and enable students to utilise OL elements that are appropriate to their needs.

Conclusion
The study addresses the potential challenges of providing accountable and open information to the public (Schreiber et al., 2021). This pandemic has disrupted higher education faculty, academics, and university professionals. It is imperative that they work together to assess the disruption caused by it, document the best practices, note the increase in evidence--based practices, and simultaneously increase university students' learning experiences.
The present study examines how OL is accepted in a specific context of learning continuity, particularly during times of lockdown due to pandemics. This research indicates that students had a strong intention to adopt online learning, and their BI was driven by their opinions of online learning's efficacy in enhancing their class performance. In light of this, future research should focus on studying students' BI regarding the voluntary usage of OL regularly. A qualitative or mixed methodology may be used in future research to improve generalizability; this study used a quantitative methodology. The BI of students to accept OL can also be investigated using alternative frameworks and theories. In the future, OL acceptance may be examined using other constructs such as trust, culture, and experiences. Faculty perspectives may be explored in future research as the latest research focuses exclusively on learners.
Switching to emergency remote teaching (ERT) during the pandemic from physical classes was difficult for the students and teachers. However, the students exhibited the utmost grit and determination to overcome the teething challenges and resolutely subscribed that online learning was a valuable method for ensuring learning continuity, and the subsequent successful implementation is the way forward. This study offers a clear understanding how online learning is adopted in challenging times to improve learning. Students' perceptions of online learning may prove helpful to university administrators in improving existing installations and customising future deployments to meet students' needs and expectations. Additionally, universities must ensure that online learning platforms are practical and efficient and achieve students' best performance during this pandemic. Some limitations are associated with this study. The selected study samples were from a limited number of universities in Bangladesh, namely five public and five private. Future research should be conducted at universities and colleges throughout the country. Online learning will be intensified in tertiary education, and systematic coordination is needed across the universities to explore the most valuable aspects of professional development. Every organisation will benefit from the designed frameworks the ecosystem will place before the policymakers to continue its training and education missions in times of crisis.