Research and innovation are crucial for shaping the future, particularly in South Asia where universities are key drivers of economic growth and social progress. However, South Asian universities face challenges, including underinvestment in higher education (Nilofer, 2020) and lower accessibility to higher education compared to East Asian countries (Zamir et al., 2023). Despite these challenges, research and innovation outputs from South Asian universities have increased in recent years (Figure 1). This study aims to identify factors that enhance research and innovation productivity in South Asian universities. Specifically, it addresses the following research questions: What factors can improve higher education, focusing on university research and innovation productivity in South Asian countries? To what extent do universities' research and innovation productivities promote economic growth in these countries? The study seeks to bridge the existing higher education gap between South Asian countries and other more developed Asian nations. The findings will be valuable for policymakers, stakeholders, and those invested in improving higher education in the region to drive rapid economic growth. Existing research has limitations: a focus on higher education generally or on student performance, neglect of crucial variables (patent rights, tertiary school enrollment, funding, information technology, publications, citations), and lack of cross-country analysis accounting for cross-sectional dependence. This study addresses these limitations through a cross-country analysis using updated data, incorporating neglected factors, exploring contextual factors, and using the robust DCCE model to account for cross-sectional dependence. The main objective is to understand and examine factors enriching research and innovation productivities in South Asian universities, with specific objectives to analyze the factors enhancing research and innovation productivities, explore their contribution to economic growth, and investigate the potential for misleading results from MG, PMG, and AMG models in the presence of cross-sectional dependence.

The literature review examines research productivity, defined as the quantity and quality of scholarly output (Dundar and Lewis, 1998; Abramo and D'Angelo, 2014; Simisaye, 2019). Research productivity is measured using various metrics including publications, citations, patents, and research funds (Demeter et al., 2022). Innovation, defined as the introduction of new products, services, or processes (Obunike and Udu, 2019; Wang and Ahmed, 2017), includes technological and non-technological forms relevant to higher education (Tseng and Lee, 2014; Pisano, 2015; Rahman et al., 2016). Innovation in higher education involves various methods, including competency-based learning, adaptive learning, and gamification, influenced by technology, demographics, and policies (Innovation in Higher Education -TeachThought, 2020; Exploring higher education innovation and trends | Deloitte US, 2023). The literature reviews mechanisms between research productivity and university rankings (Henry et al., 2020; Jameel and Ahmad, 2020; Centra, 1981; Kivinen et al., 2013), innovation productivity and university rankings (Cainelli et al., 2006; Musolesi and Huiban, 2010; Yeo, 2018; Hall and Sena, 2014; Liu et al., 2021), and education and economic growth (Mankiw et al., 1992; Tallman and Wang, 1992). Empirical studies examine factors influencing research, innovation, and academic performance (Martín et al., 2015; Cattaneo et al., 2016; Barrichello et al., 2020; Henry et al., 2020; Ryazanova and Jaskiene, 2022; Bate et al., 2023; Liu et al., 2023; Odei and Novak, 2023). However, existing research lacks focus on the university level, neglects key variables, and often ignores cross-sectional dependence in panel data analyses. This study aims to fill these gaps.

This study uses panel data from 2009 to 2021 on research and innovation productivity of universities in Bangladesh, India, Nepal, Pakistan, and Sri Lanka. The dependent variables are university rankings for research and innovation productivity (from SCImago) and GDP per capita. Independent variables include patent rights, tertiary school enrollment, government funding for tertiary education (overall and per student), internet usage, international bandwidth, publications, and citations. Bhutan, Maldives, and Afghanistan were excluded due to data limitations. Three models were specified: Research productivity as a function of independent variables (Eq. 4), Innovation productivity as a function of independent variables (Eq. 5), and GDP per capita as a function of research and innovation productivity (Eq. 6). The study employed the DCCE method (Chudik and Pesaran, 2015a) for estimation due to its ability to handle cross-sectional dependence and heterogeneity. Before applying DCCE, several preliminary tests were performed: graphical representation of the series, descriptive statistics, Pesaran test for cross-sectional dependence, Levin, Lin, and Chu/Im, Pesaran, and Shin panel unit root tests (for no cross-sectional dependence), Pesaran CIPS panel unit root test (for cross-sectional dependence), multicollinearity test, Pedroni and Westerlund panel co-integration tests. Following DCCE estimation, MG, PMG, and AMG models were estimated for robustness checks. The Dumitrescu-Hurlin causality test was used to examine causal relationships between variables. The DCCE methodology was chosen for its robustness to cross-sectional dependence and its suitability for both balanced and unbalanced panel data, even with small sample sizes, overcoming limitations of previous approaches such as PMG, which ignores cross-sectional dependence.

Graphical representations (Figure 2) and descriptive statistics (Table 2) showed trends in the variables. The Pesaran test (Table 3) indicated cross-sectional dependence in most variables. Panel unit root tests (Tables 4 and 5) determined the order of integration for each variable. Multicollinearity tests (Table 6) led to the removal of the 'citations' variable from the research and innovation productivity models. Co-integration tests (Tables 7 and 8) confirmed long-run relationships among the variables in all three models. The DCCE estimation results (Tables 9, 10, and 11) are the primary focus, as they address cross-sectional dependence. For research productivity, patent rights, tertiary funding, internet users, bandwidth, and publications showed positive but insignificant relationships. Tertiary enrollment and funding per tertiary student showed negative but insignificant relationships. For innovation productivity, tertiary enrollment and publications had significant but negative impacts, while patent rights, tertiary funding, and internet users showed positive but insignificant relationships. Funding per tertiary student and bandwidth showed negative but insignificant relationships. Regarding GDP per capita, research productivity showed a positive and innovation productivity a negative relationship with GDP per capita, although neither was significant. Causality tests (Tables 12, 13, and 14) revealed unidirectional or bidirectional causalities between various factors and the dependent variables.

The findings indicate that while several factors are positively related to research and innovation productivity, their impact is negligible at the South Asian university level. The lack of significant impact of tertiary enrollment on research productivity suggests potential inefficiencies in resource allocation, while the negative impact of enrollment on innovation productivity might indicate a need for more targeted interventions. The insignificant impact of innovation on GDP per capita suggests that other factors beyond university innovation may be crucial for stimulating economic growth. The better performance of the PMG model compared to AMG and MG emphasizes the importance of considering cross-sectional dependence in panel data analysis for this specific context. The causality test results highlight the importance of considering the direction of influence between variables in policymaking. The findings regarding the impact of individual factors on research and innovation productivity are largely in line with existing literature (Azoulay et al., 2006; Patrick, 2017; Heidi, 2017; Jacob and Lefgren, 2012; Abbadia, 2022; Fu, 2022; Sattari et al., 2022; Heyard and Hottenrott, 2021; Yigitcanlar et al., 2018; Trinugroho et al., 2021; Chen et al., 2022; Gökalp, 2010; Hawajreh and Sharabati, 2012; Berchane, 2018; Verma, 2019; Kejawa, 2020; Alam et al., 2019; Chu et al., 2019; Romer, 1987, 1990; Aghion and Peter, 1992; Grossman and Helpman, 1991; Barro and Martin, 2004; Zaman et al., 2018; Sweet and Eterovic, 2019).

This study highlights the need for improvements in patent rights, tertiary education funding, information technology, and publications to enhance research and innovation productivity in South Asian universities. However, the insignificant impact on GDP per capita underscores the complexity of economic growth and the need for broader policy interventions. Future research could explore the impact of other variables (year of establishment, university type, etc.) and investigate the interaction effects between the variables studied here. Further research is needed to explore deeper the complex relationships between university outputs and economic indicators.

The study's findings may not be generalizable to non-South Asian countries. The focus on specific variables might have neglected other important factors influencing research and innovation productivity. The use of SCImago ranking, while comprehensive, may be subject to biases inherent in citation-based metrics. The data availability also limited the scope of the study.