Introduction

In an era characterized by rapid technological advancements and global connectivity, the role of research and innovation in shaping the future cannot be overstated (Sarpong et al., 2023; Perez-Alaniz et al., 2023; Gkoumas et al., 2021). Among the diverse landscapes of academia worldwide, South Asian universities have emerged as critical hubs for generating knowledge and fostering innovation. The research and innovation productivity of these institutions have gained increasing attention due to their potential to drive economic growth, social progress, and technological breakthroughs in the region and beyond (Cho, 2014; Krishna, 2019; Reichert, 2019; Meek et al., 2009), these institutions encapsulate an essence of transformative potential.

While the prominence of research and innovation productivity in these universities garners focus, an intrinsic paradox arises. A report by Nilofer (2020) shows that South Asian countries have a pattern of allocating a small proportion of their investment funds to higher education. If this trend persists, inadequate human capital development will further widen the economic imbalance between developed and South Asian countries. This will result in wealthier nations maintaining their affluence while poorer nations continue to struggle, leaving the gap between them unaddressed. In recent years, it has been observed that the level of accessibility to higher education in South Asian countries is comparatively lower than in other Asian nations, specifically East Asian countries like China, the Republic of Korea, Singapore, and Malaysia (Zamir et al., 2023). On the other hand, the research and innovation outputs of universities from this region have increased in recent years. Figure 1 shows a visual representation of research and innovation metrics of South Asian universities, respectively.

Fig. 1
Fig. 1
Full size image

Research and innovation metrics of South Asian universities.

Given this backdrop of challenge and opportunity, the pressing need to strengthen the research and innovation framework in South Asia becomes irrefutable. Thus, the following questions arise with urgency: What potential factors could be geared up to improve higher education in South Asian countries, focusing on university research and innovation productivity? Secondly, to what extent do universities’ research and innovation productivities promote the economic growth of South Asian countries? Consequently, there is a critical need to explore the factors that boost universities’ productivity in research and innovation. With a cross-country perspective, this exploration addresses the persistent and expanding disparity in higher education across South Asian countries. The insights derived from this investigation are poised to guide policymakers, authorities, stakeholders, and other vested parties in making strategic investments and interventions to enhance higher education within the region. This collective endeavor, in turn, promises to drive rapid economic growth across these nations (Eze et al., 2020; Cao et al., 2023; Abbasi et al., 2023). For instance, according to Drèze and Sen (2013), in most East Asian nations, achieving rapid economic growth was contingent upon a rise in human capability, which was accomplished through higher education. The nation’s ability to move to more complex manufacturing methods and techniques, leading to increasing productivity, was facilitated by the enhanced quantity and quality of higher education, which in turn resulted in a superior workforce. As per the 2017 Global Competitiveness report by Schwab and Sala-i-Martín (2017), the interconnected factor plays a pivotal role in boosting productivity. Singapore ranked third in global economic competitiveness. Hong Kong secured the sixth position among 144 economies. Meanwhile, South Asian countries had comparatively lower rankings. India stood at 68th, Sri Lanka at 84th, Bangladesh at 105th, Nepal at 108th, and Pakistan at 110th (Nilofer, 2020). Therefore, South Asian universities have exhibited poor performance in global rankings, necessitating efforts to mitigate political influences within the sector and enhance the standards of educational offerings and quality assurance methods. Furthermore, governments and industry must devote more resources to education and research if South Asia is to achieve sustainable economic growth. While South Asia’s research capacity has been growing, the proportion of its research output relative to the rest of the globe remains extremely low. According to scholarly sources (Badat, 2016; Ra et al., 2021), the nations belonging to the South Asian Association for Regional Cooperation (SAARC) have contributed only around 2.86 percent of the global research output during the span of the last five decades.

However, while numerous studies have explored determinant factors for boosting research and innovation in higher education, few studies centered on the university level, which is the peak institution for educational learning. Instead, the majority of existing research has either addressed higher education in general or investigated the performance of university students. Furthermore, the factors or variables investigated by existing studies have neglected other important variables such as patent rights, tertiary school enrollment, government funding for tertiary education, government funding per tertiary students, information technology, published articles (journal publications, trade journal publications, conference publications, book series publication), and number of citations. Moreover, only a few studies, if any, gave such attention to South Asian countries. In addition, although there have been limited cross-country studies conducted using panel model techniques such as mean group, pooled mean group, and augmented mean group, these studies have not accounted for the presence of cross-sectional dependence among the units. Failing to address this issue could lead to potentially misleading results, indicating a potential methodological limitation in these studies. Therefore, this study aims to contribute to the current body of literature in the following manners. First, as a strategy for unleashing the potential, this study has launched a quest to understand and examine the factors enriching research and innovation productivity of South Asian universities. This includes incorporating some neglected factors that could have the potential to promote research and innovation productivity of universities in South Asian countries. The study employs a cross-country analysis using updated data, which is very important considering the nature of the region’s higher education issue. This issue requires a general overhaul and reconsideration for improvement to attain a high standard of university productivity. This, in turn, will lead to rapid economic growth in the countries. Second, this study introduces a fresh perspective by exploring the potential role of wider contextual factors in the research and innovation productivity of South Asian universities. Unlike previous research that mostly looked at direct influences, this study examines how broader aspects like policies, collaborations, and regional conditions might affect the connection between determinants and research outcomes. Third, this study explores the potential for misleading results when assessing the relationship with models such as mean group, pooled mean group, and enhanced mean group in the presence of cross-sectional dependence among the units. Fourth, this study will employ a rigorous and cutting-edge model to execute its estimation, i.e., the newly developed technique known as dynamic common correlated effects (DCCE), established by Chudik and Pesaran (2015a). The study chooses to employ this model because of its novelty which could provide more insight from such investigation, and its ability to provide significant robust outcomes in the presence of cross-sectional dependence among the cross-sectional units, while in most of the early panel research studies, researchers ignored the issues of cross-sectional dependence among the cross-sectional units and assumed homogenous slopes by just proceeding to use models that are sensitive to cross-sectional dependence (Turkay, 2017; Arain et al., 2019).

The main objective of this study, as a strategy for unleashing the potential, is to launch a quest for understanding and examining the factors enriching research and innovation productivities in universities of South Asian countries. However, the specific objectives are as follows:

  1. i.

    To understand and analyze the factors enhancing research and innovation productivities of universities in South Asian countries;

  2. ii.

    To explore the performance of research and innovation productivities of universities in promoting the economic growth of South Asian countries; and

  3. iii.

    To investigate whether examining the relations with models such as MG, PMG, and AMG yields misleading results if cross-sectional dependence exists among the units.

The findings of this research will aid policymakers, educators, student researchers, practitioners, and financers. Their insights will help bridge the higher education gap between South Asian countries and other more advanced Asian or high-income countries.

The remainder of this study is organized into four sections: a review of the relevant literature, a description of the methodology employed to achieve the study’s aims, an analysis and discussion of the results, and finally, a conclusion and policy suggestions based on the study’s findings.

Literature review

Theoretical review

Research productivity

Research productivity is a fundamental aspect of academic endeavors, serving as a measure of the quantity and quality of scholarly output within a given academic field or institution (Dundar, and Lewis, 1998). In any production system, whether within institutions or organizations, productivity serves as a key indicator of efficiency. According to Abramo and D’Angelo (2014) and Simisaye (2019), “research productivity” can be defined as “the totality of research performed by academics in universities.” The amount of productive research that universities conduct is an essential component. A study published on JSTOR found that the research production of institutions in the United States for eight different scientific categories demonstrates that research output follows a process consistent with a return to scale. According to Adams and Griliches (2007), a university’s research productivity can be understood as a measurement of the amount of research output generated by universities. Because it enables universities to evaluate their research performance and locate areas where they may improve, it is an essential indicator for educational institutions. A university’s research production can be evaluated using a variety of metrics, including the number of publications, citations, patents, and research funds it receives (Demeter et al., 2022).

Research innovation

Innovation is commonly defined as the introduction of new products or services, new processes, the opening of new markets, and the utilization of new resources to create value in the market, as stated by Obunike and Udu (2019) and Wang and Ahmed (2017). Both technological and non-technological forms of inventiveness can be found in the world of academics (Tseng and Lee, 2014; Pisano, 2015; Rahman et al., 2016). These two types of inventiveness are often contrasted with one another. Innovation in higher education encompasses various approaches, including competency-based learning, adaptive learning, and gamification. These innovations are spurred by a multitude of factors, ranging from technological advancements to shifting student demographics and the imperative to improve student outcomes. Moreover, external influences such as government policies and funding significantly shape the direction of innovation in higher education (Innovation in Higher Education—TeachThought, 2020). Recognizing its critical importance, innovation in higher education is essential for meeting the evolving needs of students and society at large. It plays a pivotal role in enhancing access to education, elevating the quality of learning experiences, and fostering greater student engagement and success (Exploring higher education innovation and trends | Deloitte US, 2023). Innovation is an essential engine for expanding productivity and economic development, where a more progressive economy is created when technology and non-technological innovation are combined (Zawawi et al., 2016). The production of new ideas and the acceleration of economic growth are both significantly influenced by universities. As we move into a more tech-driven economy, institutions will need to adapt at an unprecedented rate in the history of higher education, according to the World Economic Forum (2018). Universities are tasked with the dual role of education and innovation, acting as key drivers of economic development while prioritizing knowledge cultivation. The “University Knowledge Production and Innovation” report outlines four strategies for enhancing their impact: fostering entrepreneurship, enabling research innovation, developing talent for the innovation economy, and partnering with industry. In the face of digital era challenges, it’s crucial for these institutions to actively engage in innovation, encompassing both the creation of new products or services and the enhancement of existing ones. By pioneering technologies that improve productivity, universities not only contribute to economic growth but also fulfill their educational mission more effectively. (How could universities more effectively commercialize their work, 2021).

Mechanism between research productivity and university ranking

Research productivity often quantified through publications, citations, and the quality of research outputs, plays a critical role in determining university rankings. Factors contributing to research productivity in higher education include age cohort, highest qualification, satisfaction with key performance indicators (KPIs), and policy satisfaction, which influence the research productivity of academic staff. Personal, environmental, and behavioral factors were found to significantly affect research productivity, suggesting the need for strategic management and proper monitoring systems to enhance it (Henry et al., 2020). Another study emphasizes the importance of funding, collaboration, information and communication technology (ICT), and job satisfaction in boosting research productivity. Specifically, funding was highlighted as having the highest impact, indicating that university management should focus on enhancing research funding opportunities, fostering collaboration among researchers, enabling ICT, and improving job satisfaction to promote research productivity (Jameel and Ahmad, 2020). Research productivity is also linked to teaching effectiveness, suggesting a reciprocal relationship where improvements in one area could enhance performance in the other. This relationship emphasizes the importance of a balanced approach in valuing both teaching and research within higher education institutions (Centra, 1981). Furthermore, international comparisons of research productivity reveal differences in priorities and outcomes between regions. For instance, East Asian countries demonstrate higher productivity in natural sciences and technology, while North European countries excel in clinical medicine. These disparities highlight the influence of regional educational policies, economic indicators, and cultural factors on research productivity and, by extension, university rankings (Kivinen et al., 2013).

Mechanism between innovation productivity and university ranking

In today’s knowledge-driven economy, the role of universities as hubs of innovation and drivers of societal progress cannot be overstated. The mechanism between innovation productivity and university rankings is complex. One strand of research highlights the positive impact of innovation activities, particularly investments in Information and Communication Technologies (ICTs), on growth and productivity. Cainelli et al. (2006) elucidated that innovation acts as a self-reinforcing mechanism, enhancing economic performance and indicating the endogenous nature of innovation within the services sector. This perspective underscores the importance of innovation as a competitive and selection mechanism, suggesting that universities engaged in high levels of innovation activities may enjoy improved economic outcomes and, potentially, higher rankings. Furthermore, the study by Musolesi and Huiban (2010) on Knowledge Intensive Business Services (KIBS) emphasizes the critical link between sources of knowledge, innovation, and productivity. The research demonstrates that formal knowledge acquisition, such as R&D or purchases of equipment, patents, or licenses, has a significant and positive effect on productivity. This relationship highlights the value of innovation outputs and suggests that universities that facilitate such knowledge exchanges are likely to see a direct impact on their productivity measures, subsequently affecting their rankings. The research conducted by Yeo (2018) investigated the societal impact of university innovation, presenting evidence that while university innovation plays a role in driving economic performance, it constitutes only a fraction of the economic performance drivers. This finding points to the need for a broader set of factors and policies to fully leverage the impact of university innovation on economic productivity and, by extension, on university rankings. Hall and Sena (2014) explored the relationship between appropriability mechanisms, such as intellectual property protection, innovation, and productivity at the firm level. Their findings indicate that universities prioritizing innovation and intellectual property protection may see greater productivity gains. This suggests that such practices can influence university rankings by enhancing their reputation and demonstrating their commitment to safeguarding and commercializing innovative outputs. Moreover, the work of Liu et al. (2021) introduced a novel approach to university ranking through the application of a complex system model that examines the “winning and losing” relationships between universities. This method provides a new lens to assess the relative competitiveness of universities, emphasizing that innovation capacity is a key factor in determining a university’s standing.

Mechanism between education and economic growth

The relationship between education and economic growth is a pivotal area of study within the field of economics, revealing intricate mechanisms through which education contributes to the economic development of a society. The impact of higher education productivities (research and innovation productivities of the universities) on economic growth, the mechanism follows the augmented Solow endogenous growth model (Mankiw et al., 1992) as:

$$(t)=(t){\rm{H}}{(t)}^{1-\alpha }$$
(1)

where in the given context, the variable y represents the output, K(t) represents the stock of physical capital, and H(t) represents the stock of human capital. According to the findings of Tallman and Wang (1992), human capital can be conceptualized as being dependent on education:

$$(t)={(t)}^{\varphi }$$
(2)

where it is assumed that the variable φ has a value of unity. Additionally, the variable (t) is assumed to grow at a constant and exogenous rate of n, which may be mathematically represented as (t) = (0)ent. This function can be expressed in its simplest form as:

$$(t)=(t)$$
(3)

This shows that the source of human capital development (research and innovation productivity of universities) is a function of output.

Empirical review

In the field of higher education, a plethora of studies have empirically examined the complex factors influencing research, innovation, and academic performance. However, a number of studies have been empirically conducted to examine the factors that influence research, innovation, and academic performance in higher education. One such study by Martín et al. (2015) investigated the primary determinants of innovation behaviors among university students. The study adopts a longitudinal perspective and utilizes a sample of 78 students from various disciplines, including psychology, management, fine arts, and education. The study’s findings indicate a favorable correlation between innovation behaviors exhibited by university students during their freshman year, their present degrees of autonomy, cognitive demands, and individual innovation. Cattaneo et al. (2016) conducted a study that revealed that implementing competitive financing mechanisms within the realm of higher education tends to impact research output positively. Nevertheless, the presence of variety within higher education institutions might result in universities exhibiting varying behaviors when faced with the implementation of competitive financing requirements. The reaction of universities is significantly influenced by the legitimacy they possess, which is determined by the extent to which they conform to socially recognized norms and expectations. Barrichello et al. (2020) conducted a study on 137 nations using multivariate data analysis techniques to examine the determinants of innovation growth. Their findings highlighted a marked correlation between the quality of scientific research institutions and the number of patent applications made under the Patent Cooperation Treaty. This suggests that the robustness of scientific institutions and active patenting are significant contributors to a nation’s level of innovation. Henry et al. (2020) conducted a study to ascertain the determinants influencing research productivity in higher education. Their empirical findings underscored the significance of various factors, such as age cohort, highest academic qualification, disciplinary cluster, and track emphasis, in shaping the research outcomes. In a rigorous exploration of the dynamics affecting research productivity, Ryazanova and Jaskiene (2022) elucidated seven pivotal insights intended to inform institutional strategies and managerial decision-making of paramount significance among their findings is the observation that a pronounced augmentation in external financial allocations does not necessarily engender a commensurate increase in research publications or a proliferation in citation frequency, particularly when such capital is allocated in a non-cohesive manner. Bate et al. (2023) analyzed the drivers of innovation performance across 63 nations using multiple linear regressions, hierarchical regression, and analysis of variance (ANOVA). They explored the relationship between national income and innovation outcomes. The study pinpointed human capital, research capability, infrastructure, and business sophistication as core innovation determinants. Key predictors included innovation linkages, knowledge absorption, research and development efforts, and both physical and digital infrastructure components. Using a cross-sectional questionnaire survey, Liu et al. (2023) explored scientific creativity and innovation capacity among 1,241 medical postgraduate students in Fujian, China. Employing descriptive statistics, multiple regression, and logistic regression analyses, the study highlighted medical specializations, types of master’s degrees related to cognition and abilities, academic achievement, and creativity as key factors influencing innovation. Odei and Novak (2023) examined the determinants of knowledge transfer activities within higher educational institutions. Relying on data from the higher education and business survey (HESA-BCI) from the 2017/18 academic period in the United Kingdom, they adopted the partial least squares structural equation model for analytical purposes. The study unequivocally indicated that financial provisions, the safeguarding of intellectual property through patenting, and the acquisition of institutional awards play instrumental roles in facilitating the inception of university spin-off enterprises.

Drawing from the aforementioned empirical reviews, it is evident that while there is research on determinant factors for enhancing research and innovation in higher education, limited attention has been given specifically to the university level. However, the majority of current research predominantly focuses on general higher education factors or specifically on university students’ academic outcomes. Furthermore, the factors or variables investigated by existing studies have neglected other important variables such as patent rights, tertiary school enrollment, government funding for tertiary education, government funding per tertiary students, information technology, published articles (journal publications, trade journal publications, conference publications, book series publication), and number of citations. Moreover, none of the studies is on South Asian countries, whether on the same variables or related to ours. In addition, though there are few cross-country studies executed using panel model techniques, such studies have not checked whether there is cross-sectional dependence among the units.

Methodology

Data and materials

The study will utilize panel data concerning research productivity and innovation productivity of universities in emerging South Asian countries. These countries include Bangladesh, India, Nepal, Pakistan, and Sri Lanka. The reason for choosing South Asian universities is that the countries share a higher education structure that is identical in terms of entry standards, age, course duration, and instructional management system. Furthermore, their educational system is of poor quality, and education spending ranges from 1–4% of GNP, which is less than the UNESCO target of 6 percent for developing countries. The study’s selected determinant factors align with theoretical frameworks and practical considerations. The dependent variables for achieving the study’s objectives include the research and innovation productivities of the universities (universities ranking on research and innovation productivities) sourced from the SCimago ranking. This ranking is used because it covers outputs recorded/published in journals and country scientific indicators developed from the information contained in the Scopus database (Elsevier) which is the largest curated abstract and citation database of research literature in the world today, and often selected by customers for the breadth and depth of its content (Schotten et al., 2023). Moreover, Scopus is considered as the most authentic database of scholarly publications because it only indexes high-quality curated content selected by an external content selection and advisory board (CSAB) of subject matter experts (Scispace, 2023). However, the ranking is based on the annual global ranking position of the universities in terms of their research and innovation productivity. See Table 1 for the indicators included in the SCimago ranking. The other dependent variable for achieving the study’s objectives is GDP per capita (constant value in local currency units). Factors such as patent rights, funding for tertiary education (government funding for tertiary education and government funding per tertiary students), information technology (percentage of individuals using the internet and international bandwidth in Mbit/s), tertiary school enrollment, publications (number of articles published including journals, trade journals, conferences, and book series publications), and number of citations were used as independent variables. However, these factors and their range—2009 to 2021—are done based on the availability of data. Following the approach of Arain et al. (2019) and other research on the region, Bhutan and Maldives were excluded due to data limitations. Furthermore, Afghanistan was left out, not only for the lack of data but also for its occasional exclusion from South Asian classifications. The data were sourced from the World Bank Development Indicators, statistical bulletin, SCIMAGO statistical bulletin, and World Intellectual Property Organization (WIPO), WIPO patent report: statistics on worldwide patent activity.

Table 1 Indicators included in the SCimago ranking.
Full size table

Functional forms of the model

Based on the nature of the study’s objectives, three different specifications will be stated. These are where factors including the patent right (patent_rights), tertiary school enrollment (tertiary_enrollment), government funding for tertiary education (tertiary_funding), government funding per tertiary student (funding_per_tertiary_student), information technology (percentage of individuals using internet “internet_users” and international bandwidth “bandwidth” in Mbit/s), published articles (published_articles), and number of citations (citations) as a function of research productivity of the universities (research) and innovation productivity of the universities (innovation), respectively, i.e., Eq. 4 and Eq. 5, as well as the research and innovation productivities of the universities as a function of economic growth (GDP_per_capita), i.e., Eq. 6, as follows:

$$\begin{array}{c}Researc{h}_{it}={\beta }_{0}+{\beta }_{1it}patent\_rights+{\beta }_{2it}tertiary\_enrollment+{\beta }_{3it}tertiary\_funding\\ +\,{\beta }_{4it}funding\_per\_tertiary\_student+{\beta }_{5it}internet\_users+{\beta }_{6it}bandwidth\\ +\,{\beta }_{7it}published\_articles,+{\beta }_{8it}citations+{e}_{it}\end{array}$$
(4)
$$\begin{array}{c}Innovatio{n}_{it}={\beta }_{0}+{\beta }_{1it}patent\_rights+{\beta }_{2it}tertiary\_enrollment+{\beta }_{3it}tertiary\_funding\\ +\,{\beta }_{4it}funding\_per\_tertiary\_student+{\beta }_{5it}internet\_users+{\beta }_{6it}bandwidth\\ +\,{\beta }_{7it}published\_articles,+{\beta }_{8it}citations+{e}_{it}\end{array}$$
(5)
$$GDP\_per\_capit{a}_{it}={\beta }_{0}+{\beta }_{1it}research+{\beta }_{2it}innovation+{e}_{it}$$
(6)

where β0 represents constant, β1, β2, β3, β4, β5, β6, β7, and β8 are the respective coefficients, i represents the panel of the individual countries at time t, and e denotes the error term of the respective models.

Estimation methods

We will employ the recently developed approach DCCE, as proposed by Chudik and Pesaran (2015a). The selection of this model is motivated by its novelty, which offers the potential for greater insights into the investigation. Furthermore, this model can generate robust outcomes even in the presence of cross-sectional dependence among the studied units. It is worth noting that previous panel research studies often overlooked the issue of cross-sectional dependence and made the assumption of homogenous slopes. Consequently, they employed models sensitive to cross-sectional dependence without addressing it explicitly. However, prior to the estimation of the DCCE model, the techniques that this study will employ include the graphical representation of the series; the descriptive statistics of the series; the Pesaran test, which examines cross-sectional dependence among cross-sectional units; the Levin, Lin, and Chu panel unit root test and the Im, Pesaran, and Shin panel unit root test on the panels where there is no cross-section dependence; the Pesaran cross-sectional augmented Im, Pesaran, and Shin (CIPS) panel unit root test, which is derived from the cross-sectionally augmented Dickey-Fuller (CADF) test, and this test is specifically designed to be insensitive to cross-section dependence; multicollinearity test; the Pedroni panel co-integration test and the Westerlund panel co-integration test which is not sensitive to cross-section dependence. However, after the estimation of the DCCE, the study will proceed to estimate the MG model, PMG model, and AMG model for robustness check and to verify if examining the relationships with these models yields misleading results in the presence of cross-sectional dependence among the units of the countries. In addition, the study will conduct the Dumitrescu–Hurlin causality test to understand the direction of the causality between each pair of variables.

The DCCE technique is derived from the PMG estimate, which was originally proposed by Pesaran et al. (1999), also incorporates elements from MG estimation, as introduced by Pesaran and Smith (1995a, b), common correlated effects (CCE) estimation, as developed by Pesaran (2006), and the estimation method proposed by Chudik and Pesaran (2015a). Blackburne and Frank (2016) employed a novel xtpmg command designed for non-stationary and heterogeneous large panel datasets to estimate the PMG estimator. However, it is crucial to acknowledge that the PMG estimate fails to consider the presence of cross-sectional dependence among the individual units in the dataset. In the study conducted by Eberhardt (2012), an estimation was made about commonly correlated effects. On the other hand, the analysis did not consider pooled coefficients or dynamic common linked effects. According to Chudik and Pesaran (2015a), the prior estimation of the CCE model did not include the lagged value of the endogenous variable as an independent variable. This is something that was overlooked throughout the estimation process. The DCCE technique, on the other hand, considers both homogeneous and heterogeneous coefficients, as well as dynamic common correlated effects, and it also considers cross-sectional dependency. By adding cross-sectional means and lags, this technique accommodates the presence of varied slopes and cross-sectional dependency. Additionally, the importance of this methodology lies in its equal applicability to situations involving a limited sample size through the utilization of the jackknife correction approach (Chudik and Pesaran, 2015b). This assertion is supported by the findings presented in the research conducted by Ditzen (2016). The DCCE technique has been found to possess a notable advantage in terms of its robustness as an estimator, particularly when confronted with structural fractures (Kapetanios et al., 2011). In addition, the DCCE model has demonstrated strong performance in the context of unbalanced panel data, as noted by Ditzen (2016). The subsequent equation represents the dynamic equation of the DCCE model proposed by Chudik and Pesaran (2015a, b):

$${Z}_{it}={\alpha }_{i}{Z}_{it-1}+{\delta }_{i}{X}_{it}+\mathop{\sum }\limits_{p=0}^{{P}_{T}}{\gamma }_{xip}{\bar{X}}_{t-p}+\mathop{\sum }\limits_{p=0}^{{P}_{T}}{\gamma }_{yip}{\bar{y}}_{t-p}+{\mu }_{t}$$
(7)

In Eq. (7), the symbol Z represents the dependent variables in models 1, 2, and 3. The term αiZit-1 denotes the lag of the dependent variables in each model, treated as independent variables. Additionally, δiXit refers to the set of independent variables. However, the variable PT represents the number of lags used to calculate cross-sectional averages.

Results and discussions

Representation of the variables involved in the analysis was graphically made, as shown in Fig. 2, from 2009 to 2021. From the figure, the research productivity of the universities (research), tertiary school enrollment (tertiary_enrollment), and percentage of individuals using the internet (internet_users) showed an upward trend throughout the period. Innovation productivity of the universities (innovation), patent rights (patent_rights), and international bandwidth in Mbit/s (bandwidth) displayed partially constant trends for about half of the period, and then the trend continued upward up to the end. Government funding for tertiary education (tertiary_funding), government funding per tertiary student (funding_per_tertiary_student), published articles (published_articles), and the number of citations (citations) exhibited fluctuating trends throughout the period. Economic growth (GDP_per_capita) showed a rising trend except in the year 2020 but then picked up. However, in recent years all the variables have increased except funding_per_tertiary_student, published_articles, and citations, which are at a decrease.

Fig. 2
Fig. 2
Full size image

Graphical representations.

The statistical characteristics of the series under consideration are presented in Table 2 where the variables include research, innovation, patent_rights, tertiary_enrollment, tertiary_funding, funding_per_tertiary_student, internet_users, bandwidth, published_articles, citations, and gdp_per_capita from 2009 to 2021. From the table, the highest and lowest variability variables are citations and tertiary_enrollment, respectively. All the variables’ distributions are positively skewed. The p-value of the Jarque-Bera statistic is insignificant for research, innovation, tertiary_enrollment, tertiary_funding, and funding_per_tertiary_student which means the distribution of those variables is normal while that for patent_rights, internet_users, bandwidth, published_articles, citations, and gdp_per_capita is significant and thus, the distribution of those variables is explosive.

Table 2 Descriptive statistics of key variables.
Full size table

The subsequent stage of the study will involve conducting the Pesaran test to estimate cross-sectional dependence. This test determines whether cross-sectional dependence exists among the units by evaluating the null hypothesis of no cross-section dependence against the alternative hypothesis of its presence.

The results of the Pesaran tests for cross-sectional dependence with regard to the variables are presented in Table 3. Based on the data presented in the table, it can be concluded that the null hypothesis, which posits the absence of cross-section dependency among the units within a variable, is rejected at a significance level of 1% for all variables, with the exception of published_articles. Hence, it can be inferred that cross-sectional dependence exists among the units in all variables, with the exception of the variable published_articles.

Table 3 The Pesaran tests for cross-sectional dependence (CD) results.
Full size table

Table 4 shows the LLC and IPS panel unit root test results at the level and at first difference. From the table, it shows that the variable published_articles is stationary at the first difference, i.e., I(1).

Table 4 LLC and IPS panel unit root tests results.
Full size table

Table 5 presents the results of the Pesaran CIPS panel unit root test. The test was executed both at the level and first difference as indicated by the decision based on the null hypothesis of homogeneous non-stationary. From the table, the null hypothesis was rejected for research, innovation, tertiary_enrollment, and bandwidth all at 1% level of significance; thus, they are I(0), while for patent_rights, tertiary_funding, funding_per_tertiary_student, internet_users, citations, and gdp_per_capita rejected it at first difference and at 5%, 5%, 5%, 1%, 1%, and 10% levels of significance, respectively; thus, they are I(1).

Table 5 Pesaran CIPS panel unit root test results.
Full size table

Table 6 displays the results of the multicollinearity test according to the explanatory variables of each of the three models, i.e., where research, innovation, and gdp_per_capita are the dependent variables. According to Table 6a, for independent variables of research and innovation productivities models, there is multicollinearity between citations and number of published articles as their respective variance inflation factors (VIFs) exceeds 10; hence, there is a need to drop either of the two variables; henceforth, the study dropped the citations variable. Table 6b shows the multicollinearity test after dropping the citations variable, which shows no more multicollinearity problems among the explanatory variables. Furthermore, Table 6c reports the multicollinearity test for independent variables of the gdp_per_capita model, showing no evidence of multicollinearity among the explanatory variables within the model.

Table 6 Multicollinearity test results (a): For independent variables of research and innovation productivities models, (b): For independent variables of research and innovation productivities models when citations were dropped, (c): For independent variables of gdp_per_capita model.
Full size table

Table 7 displays the results of the Westerlund panel co-integration test for each of the three models, i.e., where research, innovation, and gdp_per_capita are the dependent variables. From Table 7a–c, the null hypothesis of no co-integration was rejected at a 5% significance level; therefore, evidence of co-integration among the variables in the three models.

Table 7 Westerlund panel co-integration test results (a): For research productivity model, (b): For innovation productivity model, (c): For gdp_per_capita model.
Full size table

Table 8 displays the results of the Pedroni panel co-integration test for models 1 through 3, where research, innovation, and gdp_per_capita are the dependent variables. When observing from Table 8a, b, it can be seen that the null hypothesis of no co-integration among the variables was rejected at a 1% level of significance. Therefore, the three models have a long-run relationship among the variables.

Table 8 Pedroni panel co-integration test results (a): For research productivity model, (b): For research productivity model, (c): For research productivity model.
Full size table

Table 9 presents the results obtained from MG, PMG, AMG, and DCCE estimates where the research served as the dependent variables. From the table, according to the MG model’s result, patent_rights is positively and significantly related to research productivity at a 10% significance level, while the remaining variables are insignificant. Based on the PMG model’s result, tertiary_enrollment, tertiary_funding, tertiary_funding, and funding_per_tertiary_student are positively and significantly related to the research productivity, all at a 1% significance level while the remaining variables are insignificant. From the AMG result, only the variable internet_users is significant, and it is found to be positively related to the research productivity at a 5% significance level, while the other variables are insignificant. However, the prime objective of this study is to make use of the DCCE model to overcome some flaws as discussed (under the estimation techniques of this study) of earlier approaches, including MG, PMG, and AMG models, now the focus and decision of the impact will be on the DCCE result where it shows that none of the determinant factors is significantly promoting the research productivity, however, patent_rights, tertiary_funding, internet_users, bandwidth, and published_articles are positively related with it while tertiary_enrollment and funding_per_tertiary_student are negatively related with it.

Table 9 Results of MG, PMG, AMG, and DCCE estimates for the research model.
Full size table

Results of MG, PMG, AMG, and DCCE models where innovation served as the dependent variable are presented in Table 10. From the table, the MG result shows that the variable patent_rights is positively and significantly related to innovation productivity at a 1% level, while other variables are insignificant. From the table, the result from the PMG model revealed that patent_rights, tertiary_enrollment, tertiary_funding, and internet_users are positively and significantly related to innovation productivity at 1%, 1%, and 10% levels, respectively, while funding_per_tertiary_student is negatively related though significant at 1% level, while the other variables are not significant. According to the AMG model, the result shows that patent_rights, tertiary_enrollment, and tertiary_funding are positively and significantly related to innovation productivity at 5%, 1%, and 5%, respectively, whereas the other variables are insignificant. However, the motive of this study is to make use of the DCCE model to overcome some defects of models, including MG, PMG, and AMG models, now the focus and decision of the impact will be on the DCCE result show that tertiary_enrollment and Published_Articles have a significant effect at 1 and 5%, respectively, but negative and thus, not influencing the innovation productivity, while patent_rights, tertiary_funding, and internet_users are positively related with it whereas funding_per_tertiary_student and bandwidth are negatively related with it though none is significant.

Table 10 Results of MG, PMG, AMG, and DCCE estimates for innovation model.
Full size table

A presentation of the results of MG, PMG, AMG, and DCCE models where gpd_per_capita is the dependent variable is made in Table 11. According to the table, the MG estimate found research and innovation productivities are positively and negatively related to the gdp_per_capita, respectively, but neither is significant. Furthermore, following the PMG result, research and innovation productivities are positively and negatively related to the gdp_per_capita, respectively, but only the former is significant and at a 1% level. Moreover, the AMG estimate shows that both the variables are positively related to the gdp_per_capita, but none is significant. However, the estimate of the decision model, i.e., the DCCE result, shows the same result pattern as that of the MG model but with extremely different coefficients.

Table 11 Results of MG, PMG, AMG, and DCCE estimates for GDP per capita model.
Full size table

Table 12 is the representation of the causality test between each pair of variables concerning research productivity. Following the table, there is unidirectional causality running from patent_rights, tertiary_funding, funding_per_tertiary_student, and bandwidth to research productivity, while there is bidirectional causality between internet_users and research productivity, but there is no causality between tertiary_enrollment and published_articles and research productivity.

Table 12 Pairwise Dumitrescu–Hurlin panel causality tests for research model.
Full size table

Table 13 is the representation of the causality test between each pair of variables concerning innovation productivity. Following the table, there is unidirectional causality running from patent_rights, tertiary_enrollment, funding_per_tertiary_student, internet_users, bandwidth to research productivity, and from research productivity to published_articles while there is bidirectional causality between tertiary_funding and research productivity.

Table 13 Pairwise Dumitrescu–Hurlin panel causality tests for innovation model.
Full size table

Table 14 represents the causality test between each pair of the variables with respect to GDP_per_capita. Following the table, there is unidirectional causality running from research productivity to GDP_per_capita but bidirectional causality between innovation productivity and GDP_per_capita.

Table 14 Pairwise Dumitrescu–Hurlin panel causality tests for GDP per capita model.
Full size table

Based on the results of this study, certain factors have been identified as contributing to research productivity, although their impact is minimal. These factors include patent rights, funding for tertiary education, information technology, and publications. However, tertiary school enrollment does not significantly affect research productivity. On the other hand, when it comes to innovation productivity, patent rights, funding for tertiary education, and information technology have a negligible impact, whereas tertiary school enrollment and publications do not. Azoulay et al. (2006) aim to resolve a longstanding dispute on the impact of patenting among academic scientists, specifically focusing on university professors, on the quantity and caliber of academic productivity. Several prior research studies have indicated that there was either a lack of impact or a potentially favorable correlation between patenting and publication. The allocation of funding is anticipated to significantly influence the selection of research topics, as well as the breadth, substance, trajectory, outcomes, and potential ramifications of scholarly investigations. According to Aagaard et al. (2021), the provision of funding is crucial for the sustenance of researchers and serves as an essential requirement for nearly all research endeavors. The advent of information technology has resulted in notable advancements in the field of research. The potential for researchers to engage in broader and more streamlined collaborations is evident. A significantly larger amount of data is accessible for the purpose of analysis. The advancement of analytic capabilities has experienced notable enhancements, accompanied by the increased capacity to represent findings (Verma, 2019) visually. The impact of school enrollment on university productivity is likely to be negative. This is due to the increasing global enrollment in schools, leading to a strain on resources such as lecturers, furniture, and facilities like white/black boards, floor tiles, and air conditioners. The efficiency of a school is contingent upon the nature, sources, availability, and utilization of both people and physical resources (Akinyemi and Lawal, 2021). However, when effectively administered, it can positively impact research productivity in terms of both research output and innovation. This is due to the huge pool of individuals who can contribute productively to such outcomes.

Furthermore, the research productivity of South Asian universities has a negligible impact on the countries’ GDP per capita, while innovation productivity does not. According to economic theory, education serves as the primary institutional mechanism for human capital accumulation, production, and dispersion. It is also considered an externality in relation to the dissemination of both market and non-market interests. This perspective is supported by scholars such as Schultz (1961), Becker (1964), Romer (1986, 1990), and Lucas (1988). The significance of education or human capital in the growth process is emphasized in the macro- and microeconomic literature by Acemoglu (2012), Arjun et al. (2020), Campbell and Üngör (2020), Castilla-Polo and Sánchez-Hernández (2020), Fatima et al. (2020), Rico and Cabrer-Borrás (2020), Oyinlola and Adedeji (2021), and Braunerhjelm (2022).

Moreover, examining these relations using models such as MG, PMG, and AMG could yield misleading results if there is cross-sectional dependence among the units; nonetheless, PMG performed better than AMG, and then MG.

The finding of this study with respect to the impact of patent rights on research productivity is in line with the findings of Patrick (2017) in the U.S. and EU countries, Heidi (2017); with respect to the impact of patent rights on innovation productivity, it is in line with the finding of Jacob and Lefgren (2012), Abbadia (2022). On the side of funding on research productivity, it is in line with the finding of Fu (2022) on Chinese universities, Sattari et al. (2022), and Heyard and Hottenrott (2021); while with respect to the impact of funding on innovation productivity, it is in line with the finding of Yigitcanlar et al. (2018), Trinugroho et al. (2021), and Chen et al. (2022) in universities of the United States of America and China. The finding with regard to the impact of information technology on research productivity is in line with the finding of Gökalp (2010), Hawajreh and Sharabati (2012), Berchane (2018) in the University of Verma (2019), Kejawa (2020); while with respect to the impact of information technology on innovation productivity, it is in line with the finding of Alam et al. (2019) in South Sulawesi, and Chu et al. (2019). Furthermore, the finding of this study with respect to the impact of research productivity on economic growth was confirmed in the models like Romer (1987), Romer (1990), Aghion and Peter (1992), Grossman and Helpman (1991) and Barro and Martin (2004), Zaman et al. (2018) in United States, Italy, Spain, Australia, India, Netherlands, Brazil, Switzerland, Taiwan, and Poland; however, with respect to the impact of innovation productivity on economic growth is in line with the absorptive capacity theory, indicating that the stimulation of economic growth is not solely dependent on the discovery and ownership of new products and processes at the forefront of innovation. Instead, the capacity to adapt, replicate, and disseminate innovations throughout the global production chain drives productive growth (Sweet and Eterovic, 2019).

Conclusion and policy recommendations

The current study addresses the imperative to enhance research and innovation productivity within South Asian universities. Given the pivotal role of higher education institutions in fostering development, understanding the determinants that drive research and innovation outcomes becomes crucial. Therefore, this study is an empirical investigation to understand and examine the factors enriching research and innovation productivity of South Asian universities. The results revealed that the factors found to be enriching research productivity, though the impact is negligible, are patent rights, funding for tertiary education, information technology, and publications, but tertiary school enrollment is not. However, for innovation productivity, though the impact is negligible are patent rights, funding on tertiary education, and information technology while tertiary school enrollment and publications are not. Furthermore, the research productivity of the universities in South Asian countries is negligibly stimulating the countries’ GDP per capita, while the innovation productivity of the universities is not. Moreover, examining these relations using models such as MG, PMG, and AMG could yield misleading results if there is cross-sectional dependence among the units; nonetheless, PMG performed better than AMG, and then MG. Additionally, based on the results of the causality test, it is evident that patent rights, tertiary education funding, and information technology play significant roles in predicting research productivity. On the other hand, when it comes to innovation productivity, all the variables exhibit predictive capabilities. Notably, both research productivity and innovation productivity demonstrate potential for predicting GDP per capita.

Based on these findings, to promote the research and innovation productivity of universities in South Asian countries, the examined determinant variables need to be thoroughly improved, re-strategized, and restructured where possible. For instance, implementing well-defined intellectual property (IP) and technology transfer policies can enhance patent rights within universities. Furthermore, universities can monitor the dissemination of knowledge from their intellectual contributions and subsequent integration into the patent system. The utilization of patent citation analysis has the potential to serve as a means of quantifying the influence and caliber of patents. In addition, the identification of emerging research and technological trends can prove beneficial. Second, with regard to improving funding for tertiary education, it can be done in various ways. One way is to increase government funding for tertiary institutions. This can be done by increasing the budget allocation for education in the national budget. Another way is to encourage private sector investment in tertiary education. Tax incentives can do this for companies investing in tertiary institutions. Additionally, tertiary institutions can also increase their revenue by offering more courses and programs that are in demand by students and the industry. This can help attract more students and increase revenue for the institution. Moreover, on funding per tertiary student can be improved by providing financial assistance to students pursuing tertiary education. One potential approach to achieve this objective is implementing public subsidies for tertiary education studies conducted in public and private institutions. Third, on information technology, improving internet use in universities can be done by optimizing Wi-Fi networks. This can be done using Wi-Fi automation platforms that monitor and analyze the entire network every second of every day. Fourth, tertiary school enrollment can be improved by providing financial aid and scholarships to students who cannot afford tuition fees. This can be achieved by partnering with organizations that offer scholarships and grants to students. In addition, tertiary schools can provide career counseling services to students to help them make informed decisions about their future careers. This can help students understand the value of education and how it can help them achieve their goals. Fifth, university publications can be improved by incentivizing faculty members to publish their research. This can be achieved by offering financial rewards or commendation letters to faculty members who publish their research in reputable journals. In addition, universities can support faculty members by offering free or low-rate fee writing workshops and editing services. This can help faculty members improve the quality of their research and increase their chances of getting published. Sixth, with respect to how to improve research and innovation productivity for the development of GDP per capita or rather income per capita, it is important to provide adequate funding and resources to support research activities. This can include funding for equipment, facilities, and personnel. In addition, universities can encourage collaboration between researchers and industry partners to help identify new research opportunities and facilitate the transfer of knowledge and technology. So, universities can provide training and support for researchers to help them develop the skills they need to conduct high-quality research and innovation.

The limitation of this study is that its findings may not be applicable to other non-South Asian countries due to their level of development. Furthermore, incoming studies might consider the impact of other variables that have been neglected/ignored or not thoroughly investigated by the existing studies in evaluating research and innovation productivities of universities, such as year of establishment, whether a university is private or public, whether a university is conventional or specialized. Moreover, like any impact metric, the SCimago ranking has a limitation that it may be skewed by citation outliers (e.g., a single article may receive the vast majority of citations).