Researchers’ self-archiving behavior towards open-access institutional repositories

The increasing interest of authors in self-archiving their scientific works is essential in efforts to maximize the use of open-access repositories (Nazim & Ashar, 2023; Stieglitz et al., 2020). This is evidenced by previous studies that have identified the lack of self-archiving as one of the factors that can influence the failure to utilize open-access institutional repositories (Hadad & Aharony, 2024; Ntim & Fombad, 2021). Efforts to overcome this issue are by creating policies that can encourage authors, researchers, or related parties to actively carry out self-archiving of their scientific works (Hadad & Aharony, 2024; Nazim & Ashar, 2023). As a government research institution established through Presidential Regulation Number 74 of 2019, the National Research and Innovation Agency is responsible for organizing the national research and development agenda. One of the main tasks of this institution is to manage the national science and technology information system, with the aim of facilitating the protection of intellectual property and managing the storage of all primary data and research outputs conducted within the institution (Presiden Republik Indonesia, 2019). Primary data and research outputs are important assets that must be available in the long term (Oberhiri-Orumah & Baro, 2023). To ensure the availability of primary data and research outputs, the institution issued regulation number 18 of 2022 concerning the mechanism of data submission and storage obligations so that data can be accessed and used by the public (Badan Riset dan Inovasi Nasional, 2022). In fulfilling its responsibilities and recognizing the need for open-access scientific repositories, particularly for students and researchers, the National Research and Innovation Agency has established the National Scientific Repository as a platform for sharing research data and promoting the accessibility of scientific resources. This National Scientific Repository enhances data availability for individuals, supports the replication of others' work, and protects intellectual property rights for researchers. To maximize the utilization of the platform and self-archiving intention, the National Research and Innovation Agency issued the latest regulation, number 12 of 2023, concerning the mandatory submission and storage of primary data and research outputs (Badan Riset dan Inovasi Nasional, 2023). It revises the previous regulation concerning the mandatory submission and storage of primary research data and outputs. This new regulation specifically highlights the provision of RIN-Dataverse, the National Scientific Repository. The presence of open-access platforms in scientific publications cannot be separated from economic benefits (Ejikeme & Ezema, 2019; Stieglitz et al., 2020). Open-access journals have gained global popularity as they are considered a solution to the limitations inherent in fee-based journals, which charge authors a publication fee (Utulu & Ngwenyama, 2021). Although there are avenues for researchers to make their research results freely accessible, several factors have influenced the decision-making process regarding self-archiving, including personal choice, institutional mandates, awareness of repository systems, skepticism toward the repository concept, and technical skills required for self-archiving (Lee et al., 2019). Another interesting finding is that many researchers demonstrate a lack of concern for institutional mandates related to self-archiving (Posigha & Eseivo, 2024). Previous studies have also highlighted challenges such as limited access to information about self-archiving and the burden of complicated and time-consuming administrative procedures (Ten Holter, 2020). In addition, the adoption of technology is also closely related to demographic factors. Research conducted by Makinde et al. (2022) evaluated the influence of demographic factors, such as gender, age, education level, and field of study, on the use of electronic information sources. Furthermore, Onyebinama et al. (2022) conducted a study examining the influence of demographic factors such as type of institution, discipline of study, level of education, job position, and teaching experience on research output submission to institutional repositories. The original Unified Theory of Acceptance and Use of Technology (UTAUT) model is a widely adopted framework in research on the factors influencing technology adoption. It comprises four exogenous and two endogenous constructs and also incorporates four moderating factors (Békés et al., 2022). Additionally, studies by Mbughuni (2023); Mutsvunguma (2019); Shivdas et al. (2020); Zia and Nazim (2023) have successfully adapted the UTAUT model to meet the specific needs of their respective case studies on institutional repository adoption factors. However, these studies generally focused on open-access repositories rather than on self-archiving specifically. Mbughuni et al. (2024) and Wang (2022) have conducted studies evaluating self-archiving behavior in open-access repositories. Both studies have unique limitations related to the role of demographic data in the models used. Mbughuni et al. (2024) used demographic data such as gender, age, work experience, and education level as moderating variables for self-archiving scientific publications. Meanwhile, Wang (2022) used demographic data such as gender, age, work experience, and voluntary use as moderating factors of the relationship between independent variables such as performance expectancy, effort expectancy, social influence and facilitating conditions with behavioral intention and usage behavior. In Indonesia, Nurdin and Muchlis conducted research on open-access institutional repositories, focusing on the implementation of institutional repositories in universities. This study constructs a conceptual model that emphasizes institutional repositories as an infrastructure for scholarly communication (Nurdin & Mukhlis, 2019). This model includes several key elements that interact with each other, including scientific paper development, scientific document processing and recruitment, institutional repository promotion, and scientific paper distribution and dissemination. In this case, the self-archiving activity is only associated with the role of the library in providing training or assistance without discussing the factors that can encourage people's desire to archive their academic works (Nurdin & Mukhlis, 2019). Based on the existing literature, there is still potential for empirical research on the use of open-access institutional repositories that specifically focus on self-archiving, particularly in relation to the role of demographic data. In addition, no study has examined the self-archiving behavior of researchers in the context of government efforts to increase the use of national research repositories. This gap indicates the need for further research to understand the factors that influence researchers’ participation in self-archiving in national repositories and to evaluate the impact of demographic factors and the effectiveness of government policies in encouraging the use of such repositories. Therefore, this study aims to comprehensively examine the complex landscape of researchers' motivations to engage with open-access institutional repositories. The objective is to reveal the diverse factors that influence the decision-making process. RQ: “What factors influence researchers' decisions to engage in self-archiving of their research in open-access institutional repositories?”

This study used a research model that adapted the UTAUT concept with a quantitative assessment method, as the purpose of this study was to identify factors that influenced researchers’ behavior towards self-archiving in open-access institutional repositories. The UTAUT model's comprehensive nature and its recognition in many studies in the context of technology adoption have caused the researcher to use this model as the primary analytical framework. Furthermore, this study integrated the UTAUT model with additional variables, particularly those related to demographic factors and Government regulations. This study followed the flow depicted in Figure 1, starting by identifying gaps in current understanding by reviewing existing literature. This understanding served as the basis for constructing a conceptual framework with several hypotheses. Figure 1. Research Methodology Source: Original by author, 2023 The authors identified Behavioral Intention and Use Behavior as endogenous constructs and highlighted Behavioral Intention as a mediating factor. The exogenous construct variables included four constructs of the original model, which incorporated 30 manifest variables and served as measurement indicators. This study hypothesizes several significant correlations within the adapted UTAUT model regarding researchers' propensity to self-archive data in open-access institutional repositories. First, H1 hypothesizes a positive correlation between Performance Expectancy and Behavioral Intention. H2 hypothesizes a similar positive correlation between Effort Expectancy and Behavioral Intention. H3 hypothesizes a positive correlation between Social Influence and Behavioral Intention. H4 hypothesizes a positive correlation between Facilitating Conditions and Usage Behavior. Finally, H5 hypothesizes a positive correlation between Behavioral Intention and Usage Behavior. These hypotheses collectively form the basis for understanding the factors influencing researchers’ self-archiving behavior in open-access institutional repositories. Additionally, this study includes demographic factors such as Gender, Age, and Experience as moderating variables, in line with the original model proposed by Venkatesh (Mutsvunguma, 2019). Furthermore, the authors include Regulatory Awareness and Research Discipline as control variables. A previous study by Mbughuni et al. (2024) supports the inclusion of these control variables, which highlights the use of regulatory awareness. In addition, Zia and Nazim (2023) employ research discipline. The researcher also validated the model used with experts, especially regarding the role of demographic factors. The model used in this study is shown in Figure 2. Figure 2. UTAUT Model Adoption Source: Original by author, 2023 In addition to the hypotheses regarding the relationship between endogenous and exogenous factors, this study also tested hypotheses regarding significant differences caused by moderating and control variables. Hypotheses H6 to H9 examined the differential impact of Age on all construct factors, while hypotheses H10 to H12 investigated the differential impact of Gender on Performance Expectancy, Effort Expectancy, and Social Influence. Similarly, hypotheses H13 to H15 examined the differential impact of work experience on Effort Expectancy, Social Influence, and Facilitating Conditions. In addition, this study examined the influence of Regulatory Awareness on Behavioral Intentions and Usage Behavior (hypotheses H16 and H17) and explored differences in Usage Behavior based on Research Discipline (hypothesis H18). These hypotheses collectively aimed to reveal insights into how various demographic and experiential factors influenced perceptions and behaviors within the studied platforms. Data collection used an online questionnaire with closed-ended questions. The questionnaire was structured into two sections, and the data collection focused on respondents' details and their perspectives on the proposed model. In addition, the initial section aimed to collect basic information through five questions on the respondent’s demographics, including Gender, Age, Experience, Regulatory Awareness, and Research Discipline. The subsequent section encompassed the 30 indicators used to appraise and evaluate the proposed model using a Likert scale. The list of measurement items and their relationship to the model used in this study are presented in Appendices I and II. This study considered two minimum sample size rules for the Structural Equation Model (SEM) Partial Least Squares (PLS). The first rule is the 10-fold rule, where the minimum sample size is ten times the highest number of connections directed to a latent variable. The second is the Minimum R-squared method rule calculated using the least squares regression table (Hair et al., 2019). The respondents targeted by this study were researchers who had used the RIN-Dataverse platform, resulting in a relatively limited population. Therefore, the researcher used the PLS-SEM method as this method can be used to explore structural models with limited sample sizes (Hair et al., 2019). The stages in conducting PLS-SEM analysis begin with the Measurement Model Assessment to evaluate the reliability and validity of the model. This stage involved calculating Indicator Loadings (Outer loads), Cronbach's Alpha (CA), Composite Reliability (CR), and Average Variance Extracted (AVE). The next stage is the Structural Model Assessment, conducted to examine the relationship between latent variables, including the evaluation of Path Coefficients, R-squared (R²), and Significance Testing (Wibowo et al., 2023). Previous studies have complemented the structural model analysis with other statistical analysis methods, one of which is the correlation test between variables (Savić & Pešterac, 2019). The correlation test is a basic statistical procedure commonly used in exploratory data analysis (Makowski et al., 2020). Thus, this study conducted additional statistical analysis to investigate whether demographic variables have a moderating effect on the relationship between independent and dependent variables and whether regulatory awareness and research discipline have a controlling effect on the dependent variable. The t-test is performed to analyze the correlation significance of the moderating factor with two-level categorical variables, as it is commonly employed to compare the means of the two groups and ANOVA or F-test is used to analyze variables with more than two levels, such as age, work experience, and research discipline (Novak, 2022; Yu et al., 2022). This method is commonly used to compare means across multiple groups and detect significant differences among them (Yu et al., 2022).

Based on the determined research methodology, the next step was to collect data based on the research model and measurement indicators that had been prepared. Data collection was carried out on November 13 and 17, 2023, with a total of 75 responses. Detailed demographic data of respondents are presented in Table 1. Referring to the minimum sample size guidelines for PLS-SEM analysis by Hair et al. (2019), this study required at least 70 respondents. Table 1 Respondent Demographics Characteristic Total % Gender Male 44 59% Female 31 41% Age < 30 years 32 43% 31 - 40 years 33 44% 41 - 50 years 10 13% Experience < 3 years 16 21% 3 - 8 years 44 59% 9 - 13 years 13 17% 14 - 18 years 2 3% Research Discipline Engineering 24 32% Social Sciences 14 19% Business and Management 11 15% Computer & Information System 11 15% Agricultural and Environmental 8 11% Chemistry and Physics 7 9% Regulatory Awareness Aware 29 39% Unaware 46 61% Source: Data processing result, 2023 Wibowo et al. (2023) recommend a measurement model assessment as the initial stage of analysis to validate the model used. Hair et al. (2019) and Wibowo et al. (2023) recommend that the Indicator Loading (outer loads) of each indicator exceed 0.708. From the results of the initial analysis, it is evident that one latent variable recorded an outer loading of <0.708, UB2. In order to assess other indicators, the recommendation is to conduct a loading test by excluding UB2 and observing its impact on the path. Table 2 Measurement Model Assessment Item Loading Alpha CR AVE PE1 0,787 0,893 0,903 0,651 PE2 0,789 PE3 0,806 PE4 0,867 PE5 0,834 PE6 0,753 EE1 0,751 0,867 0,874 0,654 EE2 0,796 EE3 0,857 EE4 0,884 EE5 0,748 SI1 0,777 0,805 0,817 0,631 SI2 0,797 SI3 0,735 SI4 0,863 FC1 0,751 0,907 0,969 0,633 FC2 0,837 FC3 0,797 FC4 0,770 FC5 0,794 FC6 0,848 FC7 0,769 BI1 0,830 0,860 0,863 0,704 BI2 0,845 BI3 0,880 BI4 0,801 UB1 0,960 0,967 0,980 0,937 UB3 0,977 UB4 0,967 Source: Data processing result, 2023 Table 2 presents the results of the outer loading and internal integrity assessments after the author decided to remove UB2 from the proposed model. This action resulted in outer loading values exceeding 0.700 for each indicator. Furthermore, based on the values shown in Cronbach's Alpha and Composite Reliability columns, these values indicate the strong validity and reliability of the variables used, with acceptable values ranging from 0.70 to 0.90 (Hair et al., 2019). In addition, the Average Variance Extracted (AVE) metric shows that each indicator used can explain the variance of the related construct. AVE is determined by squaring the loading of each indicator on the construct and calculating its average value. An AVE of 0.50 or higher is considered to indicate that the construct explains at least 50 percent of the variance of its items (Hair et al., 2019). The structural model assessment is a method to explain the significant influence of variations from exogenous to endogenous variables. Hair et al. (2019) and Wibowo et al. (2023) recommend using path coefficient values to determine the direction of the hypothesis, with β > 0.1 or β < -0.1 indicating a significant influence and positive values reflecting a positive relationship. Next, this method employed the t-statistic test to determine significance. Since this study uses a directional test (one-tailed), the t-statistic value must exceed 1.64. Conversely, if the study is non-directional (two-tailed), the t-statistic value would need to exceed 1.96. Finally, the p-values are used as another criterion to assess the significance of the results, with a threshold of less than 0.05 (Hair et al., 2019; Wibowo et al., 2023). Table 3 shows the evaluation of the path coefficient test and reveals that three attributes show “rejected” results. This means that there is no significant relationship between the variables. These attributes are PE → BI (Performance Expectancy to Intention Self-Archiving), EE → BI (Effort Expectancy to Intention Self-Archiving), and BI → UB (Intention Self-Archiving to Self-Archiving Behavior). This attribute has a p-value greater than 0.05 and can also be assessed from a t-statistic value greater than 1.64. These indicators indicate that the path does not have a statistically significant relationship, and there are only two attributes that meet the requirements of all indicators. Table 3 Structural Model Assessment Attributes Path Coeff T-Stat value P-value Result PE → BI 0,172 1,165 0,244 Rejected EE → BI 0,194 1,642 0,101 Rejected SI → BI 0,388 3,315 0,001 Accepted FC → UB 0,218 2,156 0,031 Accepted BI → UB 0,182 1,508 0,132 Rejected Source: Data processing result, 2023 This study also applied the t-test method to the moderating factors and control variables with two levels and used ANOVA for moderating factors and control variables with more than two levels, with a significance criterion of p-value < 0.05. The results of the correlation tests are presented in Table 4. The t-test results show that gender has a moderating effect on social influence, and regulatory awareness has a controlling effect on self-archiving behavior. Meanwhile, the results of the ANOVA test found that age and experience only had a moderating effect on the social influence factor. Table 4 Demographic Factor Test Results Latent Var. P-value (T-Test) Result Gender Awareness PE 0.1895 Not Significant EE 0.0701 Not Significant SI 0.0470 Significant FC Not Checked BI 0.1842 Not Significant UB 0.0000 Significant Latent Var. P-value (ANOVA) Result Age Exp. Subject PE 0.0831 Not Significant EE 0.2680 0.1680 Not Significant SI 0.0018 0.0042 Significant FC 0.0598 0.3570 Not Significant BI Not Checked UB 0.6170 Not Significant Source: Data processing result, 2023 To answer the research question (RQ) regarding Factors Influencing Researchers’ Decision to Participate in Self-Archiving for Open-Access Institutional Repositories, the researcher employed five hypotheses (H1 – H5) to explore the relationships among the latent variables in the conceptual research model. First, the researcher focused on the link between performance expectancy and behavioral intention (H1). Performance expectancy had no significant effect on Behavioral Intention (β=0.172; t=1.165, p-value = 0.244). The researcher observed similar results in the relationship between Effort Expectancy and Behavioral Intention (H2). The impact of Effort Expectancy on Self-Archiving behavior did not significantly affect Behavioral Intention (β=0.194; t=1.642, p-value = 0.101). This finding aligns with a study by Nazim and Ashar (2023), which indicates that Performance Expectancy and Effort Expectancy do not have a significant relationship with the intention to self-archive. Factors such as low-quality repositories, lack of adequate skills to publish, and fear of openness contribute to this outcome. Diverse outcomes emerged in the analysis of Social Influence (H3) significantly affects self-archiving behavior, demonstrating a significant impact on Behavioral Intention (β=0.388; t=3.315; p-value = 0.001). However, this study is not in line with the findings of Nazim and Ashar (2023) & Shivdas et al. (2020), their findings showed no significant relationship between Social Influence and Behavioral Intention. This suggests that encouragement and support from colleagues and managers can help overcome issues related to incompetence and anxieties about self-archiving. Similarly, Facilitating Conditions (H4) significantly affect self-archiving behavior, showing a significant effect on Usage Behavior (β=0.218; t=2.156; p-value= 0.031). This is consistent with the study by Nazim and Ashar (2023), which also reported a significant positive relationship with publishing intention. However, it differs from the research results of Shivdas et al. (2020). It indicates that adequate infrastructure, as well as training support or assistance in performing self-archiving, will have a positive impact. On the other hand, Behavioral Intention (H5) showed no significant impact on Usage Behavior (β=0.182; t=1.508, p-value= 0.132). This divergence might be attributed to a lack of awareness among respondents regarding self-archiving features and regulations in open-access institutional repositories. From a demographic factors perspective, the results of hypothesis testing indicated that gender had no moderating effect on performance expectations (H6) and efficiency expectations (H7). However, gender was found to have a moderating effect on social influence (H8). Regarding regulatory awareness, no significant difference was found between individuals who were aware of the regulation and those who were unaware in terms of behavioral intention (H9). On the other hand, there was a large difference in usage behavior (H10) between these groups, which was likely due to the large number of people who were not yet aware of the regulation regarding self-archiving. This finding aligns with research conducted by Mbughuni (2023) & Zia and Nazim (2023), which also found that awareness of regulations requiring self-archiving has a significant impact on self-archiving practices. In other hypothesis tests encompassing multiple levels of variable categories, Age and Work Experience did not show a significant moderating impact concerning Performance Expectation (H11), Effort Expectation (H12) (H15), and Facilitating Conditions (H14) (H17). This is consistent with studies of Nazim and Ashar (2023) & Zia and Nazim (2023), which also did not find a significant positive relationship with usage intention. However, these two moderating factors showed a significant impact on Social Influence (H13) (H16).

This study successfully found the hypothesis that social influence significantly impacts self-archiving intention, and facilitating conditions affect self-archiving behavior. In addition, this study revealed moderating factors that influence the relationship between the social influence variables and self-archiving intention, including gender, age, experience with social influence factors, and the direct impact of awareness on regulations related to self-archiving behavior. The theoretical contribution of this study is the extension of the application of the UTAUT model, particularly in the context of open-access repositories. The findings enrich the existing literature by demonstrating the importance of social influence and regulatory awareness in shaping self-archiving behavior in open-access repositories. Furthermore, the practical implications of these findings are the need for policymakers and repository managers to create an environment that can encourage positive social influence on researchers' engagement in self-archiving activities, improve facilities in open-access repositories, and use specific policy initiatives related to self-archiving. Further research should explore the most effective social influence strategies to increase self-archiving intentions by analyzing social influence factors based on existing theories. It may involve exploring the impact of mentoring programs or institutional incentives on self-archiving intentions. In addition, researchers should examine the role of technological advances in facilitating self-archiving, such as the development of user-friendly repository interfaces or automated archiving tools.