Influence of Institutional Characteristics on Academic Staff’s Research Productivity: The Case of a Vietnamese Research-oriented Higher Education Institution

As research productivity (RP) can be an indicator of a university’s reputation, it is necessary to examine factors within a higher education institution that contribute to promoting RP. This article aims to examine the relationship between a university’s institutional characteristics and its academics’ RP within a research-oriented university, as regards publications in international and Vietnamese journals. The study was conducted with a survey questionnaire with 96 observations (96 participating academics among approximately 220 academics). The data was analysed through descriptive and inferential statistics including exploratory factor analysis and Poisson regression. The results showed that institutional climate was a significant predictor of both international and domestic journals; for every one unit increase on the predictor institutional climate, the count increased on the number of both international and domestic journals. In addition, the regression coefficients for coordinating goals and research culture were negative, suggesting that academics scoring higher on coordinating goals and research culture were more likely to exhibit a lower count for the number of international and domestic journals respectively than academics scoring lower on the measure. The findings suggested that greater attention should be paid to establishing appropriate research goals, communication, and improving aspects of a research culture within the university so as to enhance academics’ research productivity.


INTRODUCTION
The function of research in education is to offer information crucial for making relevant judgments on educational issues such as teaching and learning and national policies (Morris, 1967) with a general aim to improve the quality of education (Nguyen et al., 2021). Knowledge created through research can be disseminated through teaching, and research productivity can be an indicator of a university's prestige. Education, research, and service are the three major academic roles of higher education institutions (Edgar & Geare, 2013). When research is integrated into teaching, it can help enhance teaching quality (Brew, 2003). For example, it can help enhance academics' disciplinary knowledge and research competence which in turns improve academics' teaching and supervising students' research projects (Lindsay et al., 2002). Doing research has been considered one of academics' responsibilities in research-oriented universities (Cummings, 2014). Research productivity (especially publications) are considered one of the key criteria for recruitment and promotion to meet the research orientation of the universities (Cummings & Shin, 2014).
Research productivity (RP) is a concept that has many varied meanings in different contexts. In general, it is the overall sum of faculty publications over a given period of time (Print & Hattie, 1997), which can include conference Hypothesis 2: Coordinating goals (CG), research culture (RC), and institutional climate (IC) correlate positively to academics' publications in domestic journals.

MATERIALS AND METHODS
The study primarily employed quantitative methods to examine the relationship between factors concerning institutional characteristics and academics' RP regarding the number of publications in international and domestic journals. The participants were purposefully selected from University A. The survey sample included 96 observations: 96 academics, including lecturers (85) across faculties and departments, and research staff at several centres in the university. Purposive sampling allowed selecting people who provided the information the researcher needed to gain insight into the phenomenon under study (Marshall, 1996). Data was collected through a survey questionnaire composed of 15 items from nine institutional characteristics selected from Bland et al.'s (2005) model for components of productive organization. These 15 items were grouped into three components called 'coordinating goals; research culture; and institutional climate', which were to be examined in relation to academics' RP.
Quantitative analysis helps to determine the relationship between two or more quantitative variables, measuring the extent of the relation between the independent and dependent variables of the study and interpreting the relation by a number of statistical tests (Creswell, 2009). Several hypothesis tests are performed. Data from the questionnaire was analysed through descriptive statistics and inferential statistics based on SPSS 22.0 software including exploratory factor analysis (EFA), correlation analysis, and Poisson regression analysis.
EFA was used to identify the factors concerning institutional characteristics, including coordinating goals (CG); research culture (RC); and institutional climate (IC). EFA was performed with the absolute value for factor analysis below 0.55, with a total variance explained (greater than 60%) for a number of observations close to 100 (Hair et al., 2014). Regression analyses were performed to examine the predictive role of these three factors (independent variables) on the number of articles published in international and domestic journals (dependent variables). Poisson regression was used to model events where the outcome was observed with count data according to the Poisson distribution. This method allowed analysis of count data by explaining which independent variable has an effect on the dependent variable whose values are non-negative integers and small numbers (Koletsi & Pandis, 2017;Trivedi, 2014). Since each dependent variable (the number of publications) is a non-negative integer (count data), Poisson regression analysis would be appropriate (Coxe et al., 2009).

Results
• Relation between institutional characteristics and academics' research productivity Before performing quantitative analyses, the data was tested and filtered for outliers in the sample using the Little's MCAR test (Table 1). Hypothesis H0 for Little's MCAR test is completely random missing data (MCAR). The data is MCAR when the sample is missing a value independent of the data value (Little, 1988). Because the Sig-value was greater than 0.05, there was no statistical significance, so the hypothesis H0 could not be rejected. Thus, it could be concluded that the missing data was completely random and suitable for analysis. • Exploratory factor analysis Exploratory Factor Analysis (EFA) was performed to determine the factors of the institutional characteristics that may impact academics' research productivity, providing evidence for the validity of the questionnaire structure. The assumptions of data suitability for factor analysis were met as the Kaiser Meyer Olkin (KMO) measures for sampling adequacy was 0.880, the Bartlett's test was significant (p < 0.000) (Field, 2009), which met the condition for EFA when 0.5 < KMO < 1 and Sig. < 0.05, as shown in Table 2. Sig. 0.000

VIETNAM JOURNAL OF EDUCATION
Principal components factoring was used for factor extraction, and varimax was used for simplifying the column of the factor matrix so that the factor extracts were clearly associated. The number of factors to be extracted was determined by considering the eigenvalues together (greater than 1), with the absolute value for factor analysis less than 0.55 for an observed sample close to 100, with explaining cumulative variance (71.47% greater than 60%) (Hair et al., 2014). When performing EFA, the rotated component matrix showed that the variables were clearly distributed in three components, and there was no loading of variables on two factors. Performing EFA produced three factors with a total variance explained of 71.47%, that is, 71.47% of extracted factors were associated with the institutional characteristics that affected academics' research productivity, and 28, 53% was due to other factors. The results of EFA are shown in Table 3. 0.610 Table 3 shows there were three factors affecting academics' research productivity. The descriptive statistics of these factors are presented in Tables 4-6.
-Factor 1: coordinating goals  The reliability of the internal construct of items was expressed by the Cronbach's Alpha value. For the reliability test, it was required to remove the variable when Cronbach's Alpha < 0.6. The results of Cronbach's Alpha of these three groups of items were all > 0.8 (Table 7). Therefore, the observed variables in these three factors were accepted and have high reliability. The descriptive statistics of CG, RC and IC are presented in Table 8. The multicollinearity test was performed in two steps: examining the correlation between the independent variables and testing the tolerance and Variance Inflation Factor (VIF) values. Table 9 shows that the correlation coefficient between these two independent variables was low (< 0.5), which was the first sign that there was no multicollinearity.   Based on EFA and the conditions applied to the dependent variable, Poisson regression model was applied to find the relationship between the three factors (CG, RC, IC) of the institutional characteristics and two dependent variables of RP, respectively: articles in international journals (IAR), articles in domestic journals (DAR). Table 11 describes parameter estimates in the relationship between institutional characteristics and international publications.  Table 11 shows that the variable IC (institutional climate) had an influence on the international publication performance at 0.05, Sig values. = 0.035. Omnibus Test gave p = 0.024 < 0.05, which means the model was predictive at 0.05. The positive coefficient (the beta value B) indicates that as scores increased on the predictor IC, the count increased on the number of IAR. IC was a good predictor of IAR (B = 0.276; S.E = 0.1308; p = 0.035 < 0.05): for each unit increase in the variable IC the predicted number of IAR increased by 31.7%. It is necessary to mention that the incidence rate ratio (IRR) is found in the Exp(B) column. The Exp(B) value for this variable is greater than one, which means that with an increasing score on the predictor IC, the count for the number of IAR changed by a factor of 1.317 (IRR).
The negative coefficients for CG and RC indicate that as scores increased on the predictors CG and RC, the count decreased on the number of IAR. In the model, CG was a negative and significant predictor of the count for the number of IAR (B = -0.255; S.E = 0.1227; p = 0.038). The IRR suggests that for every one unit increase on the predictor CG, the count for the number of IAR changed by a factor of 0.774 (meaning that the count was decreasing).
The regression coefficient for RC was negative, suggesting that academics scoring higher on RC were more likely to exhibit a lower count for the number of IAR than academics scoring lower on the measure. Nevertheless, RC was not a significant predictor in the model (B = -0.053; S.E = 0.1371; p = 0.701); the IRR indicates that for every one unit increase in RC, the count for the number if IAR changed by a factor of 0.949 (meaning that the count was decreasing). Table 12 describes Poisson regression of three predictors, parameter estimates in the relationship between institutional characteristics and DAR.  Table 12 shows that the variable IC had an influence on DAR at 0.05, Sig values. = 0.000. Omnibus Test gave p = 0.000 < 0.01, which means the model was predictive at 0.01.
The positive coefficient for IC indicates that as scores increased on the predictor IC, the count increased on the number of DAR. IC was a good predictor of DAR (B = 0.440; S.E = 0.0871; p = 0.000 < 0.01): for each unit increase in the variable IC, the predicted number of DAR increased by 55.3%. The Exp(B) value for this variable was greater than one, which means that with an increasing score on the predictor IC, the count for the number of DAR changed by a factor of 1.553.
The negative coefficients for CG and RC indicate that as scores increased on the predictors CG and RC, the count decreased on the number of DAR. In the model, RC was a negative and significant predictor of the count for the number of DAR (B = -0.317; S.E = 0.0855; p = 0.000 < 0.01). The IRR suggests that for every one unit increase on the predictor RC, the count for the number of DAR changed by a factor of 0.728 (meaning that the count was decreasing).
The regression coefficient for CG was negative, suggesting that academics scoring higher on CG were more likely to exhibit a lower count for the number of DAR than academics scoring lower on the measure. Nevertheless, CG was not a significant predictor in the model (B = -0.054; S.E = 0.0776; p = 0.487); the IRR indicates that for every one unit increase in CG, the count for the number if DAR changed by a factor of 0.948 (meaning that the count was decreasing).

Discussion
The first major finding is that IC was a significant predictor of both IAR and DAR with positive regression coefficients and Exp(B) values greater than one, meaning that for every one unit increase on the predictor IC, the count increased on the number of both IAR and DAR. The institutional climate where research collaboration was fostered among academics as experiences of research socialisation (Jalloun, 2010) facilitates sharing ideas and learning new knowledge. Within this community of practice (Wenger, 1998), positive group climate whereby academics were offered collegial support appeared to be conducive to academics' RP (Chen et al., 2006;Sulo et al., 2012). It should be noted that the university had a policy to offer financial incentives for publications in international journals and formal funding for strong research groups. Table 6 indicates that academics had strong attachment with colleagues within a well-developed research network and shared research interests. Such quality of researcher networks could make academics productive with research outcomes (Besancenot et al., 2017).
As the second major finding, the regression coefficients for CG and RC were negative, suggesting that academics scoring higher on CG and RC were more likely to exhibit a lower count for the number of IAR and DAR respectively than academics scoring lower on the measure. First of all, it should be noted that clear coordinating research goals and communication which were established at the university forced academics to do research within their discipline and published their works in prestigious international journals. This planning somehow hindered multidisciplinary research outputs. Then, data on Table 4 also showed that academics were not well-informed with the present communication system within the department. This study's results are consistent with Nguyen et al.'s (2021) study in that communication policies were negatively correlated with academics' research outcomes.
In addition, research culture was negatively correlated with academics' RP. The descriptive statistics of research culture in Table 5 also showed low ratings for items related to the availability of research resources, culture, sufficient work time for research, mentoring, and research emphasis (with means ranging from 2.6 to 3.2 on a 5-point Likert scale). Previous studies argue that research infrastructure considerably influences university academics' research productivity (Kang et al., 2017) because a lack of adequate research facilities would restrain their research activities (Alrahlah, 2016). It should be noted that internal funding was limited at the university, and data on Table 5 shows that academics had difficulty to receive external funding to do research, without which they could not enhance the amount of their research outputs (Hottenrott & Lawson, 2017;Sulo et al., 2012). In some cases, they had to spend more time to locate and secure research funding (Rawls, 2018). The academics in this study expressed low ratings for uninterrupted time for research because of their heavy teaching load. Consequently, teaching load significantly influenced their RP (Quimbo & Sulabo, 2014) because academics' RP depended on the time they spent on researchrelated activities with balancing the time they spent on their other responsibilities (Smeltzer et al., 2016). Mentoring was also an issue to academics in this study because their research outputs depended on experience and competence in research without which their RP could not be enhanced. In mentoring, for example, young lecturers received support and advice from lecturers with research experience (e.g., mentoring junior academics to do research) or between lecturers who collaborated in research (e.g., reviewing a manuscript for a colleague) (Nguyen, 2015). Another explanation for this finding could be issues with research interest and commitment. Only when academics have high commitment to undertake research because academics have research interest and are aware of their research responsibility, they are willing to be involved in research activities to achieve research goals (Nguyen, 2015).

CONCLUSION
The study results show that institutional climate was a significant predictor of both IAR and DAR; for every one unit increase on the predictor IC the count increased on the number of both IAR and DAR. In another finding, the regression coefficients for coordinating goals and research culture were negative, suggesting that academics scoring higher on CG and RC were more likely to exhibit a lower count for the number of IAR and DAR respectively than academics scoring lower on the measure. The findings suggest that greater attention should be paid to establishing appropriate research goals and communication. For example, research goals should focus on multidisciplinary research so that academics can collaborate with colleagues from different disciplines and they need to be wellinformed with research activities within the institution. There are also implications that the aspects of a research culture (such as the availability of research resources such as library resources and well-equipped research space, external research funding, sufficient work time for research, research training and mentoring, and research emphasis such as shared interest and commitment) within the university should be improved so as to enhance academics' RP.