Unleashing the potential: a quest to understand and examine the factors enriching research and innovation productivities of South Asian universities

The study addresses the growing importance of research and innovation for economic and social progress in South Asia, where universities are pivotal yet constrained by low higher-education investment, limited accessibility, and persistent quality gaps relative to East Asia. Against this backdrop, the paper asks: (1) Which factors can be leveraged to improve university-level research and innovation productivity in South Asian countries? (2) To what extent do university research and innovation productivities promote economic growth in the region? Motivated by cross-country disparities and policy relevance, the study focuses on Bangladesh, India, Nepal, Pakistan, and Sri Lanka, aiming to identify determinants of university productivity (including often-neglected variables like patent rights, tertiary funding levels, IT infrastructure, publications, and citations) and to quantify their links to GDP per capita. The work also interrogates methodological issues, noting that ignoring cross-sectional dependence can bias conventional panel estimators, and thus adopts advanced methods (DCCE) to obtain robust inferences.

The review covers theoretical and empirical perspectives. Research productivity is defined as the quantity and quality of scholarly outputs (publications, citations, patents, funds), crucial for institutional assessment and improvement. Innovation in higher education spans technological and non-technological change and is driven by ICTs, policy, funding, and industry collaboration. Mechanisms linking research productivity and university rankings emphasize funding, collaboration, ICT, staff qualifications, and satisfaction with KPIs; international comparisons show regional specializations that reflect policy and economic contexts. Innovation productivity’s link to rankings is supported by evidence that knowledge sourcing (R&D, IP acquisition), appropriability regimes (IP protection), and ICT investments enhance productivity and competitiveness. On education and economic growth, the review invokes augmented Solow and endogenous growth frameworks where human capital—proxied here by university research and innovation productivities—supports output growth. Empirically, prior studies identify funding mechanisms, institutional legitimacy, collaboration, infrastructure, human capital, and IP as salient drivers of research and innovation; however, gaps persist: few studies focus specifically at the university level in South Asia, many omit variables like patent rights and tertiary funding intensity, and most cross-country panel studies neglect cross-sectional dependence, risking misleading results. This study addresses these gaps by incorporating broader determinants and employing DCCE.

Design and data: A panel of five South Asian countries (Bangladesh, India, Nepal, Pakistan, Sri Lanka) from 2009–2021 is assembled due to comparable higher-education structures and data availability. Bhutan, Maldives, and Afghanistan are excluded due to data limitations/classification issues. Dependent variables: (a) University-level research productivity ranking positions; (b) University-level innovation productivity ranking positions, both from SCImago Institutions Rankings (derived from Scopus-based indicators; research includes normalized impact, leadership output, publications, etc.; innovation includes publication output cited in patents, technological impact, and patent applications). For macro impact, GDP per capita (constant LCU) is used. Independent variables: patent_rights; tertiary_enrollment; tertiary_funding (government funding for tertiary education); funding_per_tertiary_student; information technology (internet_users, bandwidth in Mbit/s); published_articles (journals, trade journals, conferences, book series); citations. Data sources: World Bank WDI, SCImago statistical bulletins, WIPO patent statistics. Model specifications: (1) Research = f(patent_rights, tertiary_enrollment, tertiary_funding, funding_per_tertiary_student, internet_users, bandwidth, published_articles, citations). (2) Innovation = same determinant set. (3) GDP_per_capita = f(research, innovation). Econometric strategy: The study emphasizes cross-sectional dependence and heterogeneous dynamics. Preliminary steps include graphical exploration and descriptive statistics; Pesaran CD test for cross-sectional dependence; panel unit-root tests—LLC and IPS where applicable, and Pesaran CIPS (CADF-based) robust to cross-sectional dependence; multicollinearity diagnostics via VIF (leading to dropping citations where VIF>10 with published_articles); cointegration tests (Westerlund and Pedroni). Main estimator: Dynamic Common Correlated Effects (DCCE) of Chudik and Pesaran (2015a), which augments PMG/MG ideas with cross-sectional means and lags to handle unobserved common factors, dynamics, heterogeneous slopes, and small samples (with jackknife correction). Robustness/benchmarking: Mean Group (MG), Pooled Mean Group (PMG), and Augmented Mean Group (AMG) are estimated for comparison despite their sensitivity to cross-sectional dependence. Causality: Dumitrescu-Hurlin panel Granger causality tests are applied to explore directional relationships among variables.

- Descriptives and dependence: Most variables trend upward over 2009–2021; strong cross-sectional dependence exists for nearly all variables (Pesaran CD significant at 1% for all except published_articles). VIFs indicated severe collinearity between citations and published_articles; citations was dropped in research and innovation models. - Stationarity and cointegration: CIPS tests indicate mixed integration orders (I(0) for research, innovation, tertiary_enrollment, bandwidth; I(1) for patent_rights, tertiary_funding, funding_per_tertiary_student, internet_users, citations, GDP per capita). Westerlund and Pedroni tests confirm cointegration across all three model specifications. - Determinants of research productivity (DCCE focus): None of the determinants achieve statistical significance. Signs suggest negligible positive associations for patent_rights, tertiary_funding, internet_users, bandwidth, and published_articles; tertiary_enrollment and funding_per_tertiary_student show negative associations. Benchmark models (MG/PMG/AMG) disagree and can be misleading under cross-sectional dependence (e.g., PMG shows significant positive effects of tertiary_enrollment, tertiary_funding, funding_per_tertiary_student at 1%; AMG shows internet_users positive at 5%). - Determinants of innovation productivity (DCCE focus): Tertiary_enrollment is significantly negative (coef ≈ -630.34, p≈0.004), and published_articles is marginally negative (p≈0.051). Patent_rights, tertiary_funding, and internet_users are positively signed but insignificant; funding_per_tertiary_student and bandwidth are negative but insignificant. MG/PMG/AMG frequently report positive significant effects for patent_rights, tertiary_enrollment, tertiary_funding, and internet_users, again underscoring potential bias when ignoring cross-sectional dependence. - Growth effects (GDP per capita model): Under DCCE, research productivity has a small, positive but insignificant association with GDP per capita (coef ≈ 2.06; p≈0.684); innovation productivity is negative and marginally insignificant (coef ≈ -11.24; p≈0.102). PMG indicates a significant positive effect of research (p<0.01) but innovation negative and insignificant. - Causality (Dumitrescu-Hurlin): For research productivity, unidirectional causality from patent_rights, tertiary_funding, funding_per_tertiary_student, and bandwidth to research; bidirectional causality between internet_users and research; no causality with tertiary_enrollment or published_articles. For innovation productivity, unidirectional causality from patent_rights, tertiary_enrollment, funding_per_tertiary_student, internet_users, and bandwidth to innovation, and from innovation to published_articles; bidirectional causality between tertiary_funding and innovation. For GDP per capita, unidirectional causality from research to GDP per capita; bidirectional causality between innovation and GDP per capita at the 10% level. - Overall: Patent rights, tertiary education funding, and IT show negligible positive associations with both research and (to a lesser extent) innovation productivity under DCCE; tertiary enrollment and (for innovation) publications can be counterproductive, potentially reflecting resource strain or quality issues. Research productivity modestly and weakly supports GDP per capita; innovation productivity does not robustly translate into growth within the study window.

The study directly addresses the research questions by identifying which macro-institutional and system-level factors correlate with university research and innovation productivity in South Asia once cross-sectional dependence is properly handled. The DCCE estimates suggest that traditional levers—patent rights protection, tertiary funding, and IT access—are directionally consistent with higher productivity but their effects are small and statistically fragile in this context. The significant negative effect of tertiary enrollment on innovation productivity and the lack of positive effect on research productivity likely reflect capacity constraints and quality dilution when enrollment rises faster than resources, faculty, and infrastructure. Similarly, the negative association of publications with innovation output may point to misalignment between publication-driven incentives and patentable or industry-relevant outputs. Regarding growth, the weak positive association from research productivity and lack of robust effect from innovation productivity suggest that universities’ outputs are not yet sufficiently translated into broader economic gains, possibly due to limited absorptive capacity, weak commercialization pathways, or misaligned incentives. Methodologically, the study shows that ignoring cross-sectional dependence (as in MG/PMG/AMG) can overstate effects; PMG appears to outperform AMG and MG among traditional estimators but still risks bias. These findings underscore the need for systemic reforms—IP and tech transfer frameworks, targeted funding, and digital infrastructure—coupled with mechanisms to connect university outputs to industry and policy demands.

The paper contributes by: (1) providing a South Asia-focused, university-level cross-country analysis of research and innovation productivity determinants using updated 2009–2021 data; (2) incorporating often-neglected variables (patent rights, tertiary funding intensity, IT, publication and citation metrics); (3) demonstrating that conventional panel estimators can mislead under cross-sectional dependence; and (4) applying DCCE for robust inference. Findings indicate that patent rights, tertiary funding, and IT are directionally supportive but exert negligible impacts; tertiary enrollment may depress innovation productivity; and research productivity only marginally supports GDP per capita, while innovation productivity does not. Policy recommendations include: strengthening university IP and technology transfer policies; tracking patent citations and tech impact; increasing and optimizing tertiary funding (including per-student support and private-sector participation); enhancing digital infrastructure (campus-wide high-quality internet/Wi‑Fi management); expanding financial aid and scholarships to improve access while safeguarding quality; incentivizing high-quality, impactful publications and offering writing/editing support; and fostering university–industry collaboration and researcher training to translate outputs into innovation and economic value. Future research should broaden determinants (e.g., university age, ownership, specialization) and refine productivity metrics to capture quality and commercialization outcomes.

Generalizability is limited to South Asian countries with similar development levels and higher-education structures. Data constraints led to the exclusion of Bhutan, Maldives, and Afghanistan, and to the 2009–2021 window. The SCImago-based productivity measures, while comprehensive, may be influenced by citation outliers and may not fully capture commercialization or societal impact. Some variables exhibited multicollinearity (necessitating the exclusion of citations). Unobserved institutional heterogeneity at the university level (e.g., age, public/private status, specialized vs. comprehensive) was not modeled explicitly and should be explored in future work.