Abstract
This study evaluates the efficiency of educational infrastructure investment in China and Colombia between 2020 and 2025. The research aims to measure technical efficiency, identify institutional and contextual determinants, and assess whether infrastructure investment causally improves educational outcomes. Using a dual-method approach—Data Envelopment Analysis and Stochastic Frontier Analysis—combined with panel Granger causality tests, the study analyzes provincial and departmental data from multilateral sources. Results show substantial efficiency variation within and between countries, with Chinese provinces operating closer to the efficiency frontier than Colombian departments. Diminishing returns to infrastructure investment are evident, while governance quality, urbanization, and budget execution rates significantly shape efficiency patterns. Causality tests confirm that infrastructure investment precedes improvements in educational outcomes.
Keywords
Introduction
Educational infrastructure investment constitutes one of the largest components of public expenditure in developing and middle-income countries, yet its effectiveness in transforming financial inputs into measurable educational outcomes remains insufficiently clarified across contrasting institutional settings. Governments continue to face rising fiscal pressures, intensifying the need for efficiency-oriented policy decisions. Studies have addressed efficiency patterns within national contexts such as China (Gu & Ayala Garcia, 2021) and Colombia (Faguet & Sánchez, 2008), but cross-national comparisons with shared analytical structures remain scarce. This limitation is particularly relevant because educational systems often operate under different technological conditions, degrees of decentralization, and administrative capacities that affect the use of infrastructure investments (Chan & Karim, 2013; Hu et al., 2009; Li & Wye, 2023). The lack of harmonized approaches weakens efforts to understand why similar spending levels generate dissimilar performance in countries with comparable development challenges.
A second challenge concerns the limited incorporation of institutional and territorial heterogeneity in efficiency assessments. Governance quality, administrative capacity, and urban–rural disparities shape how public resources translate into learning outcomes. Although studies in Latin America and Asia document heterogeneous spending patterns and local capacity constraints, consistent methodological strategies for comparing institutional environments remain uncommon. Educational systems also differ in technological structures and investment cycles, indicating that efficiency should not be assumed to follow uniform pathways across countries.
This study employs an integrated methodological strategy inspired by advanced efficiency assessments of higher education systems in China and European countries, where dual-method approaches have helped clarify technological differences, environmental drivers, and productivity patterns. The analysis merges Data Envelopment Analysis and Stochastic Frontier Analysis to evaluate technical efficiency and technological gaps, combined with panel Granger causality models to identify directional relationships between investment and outcomes. This strategy addresses calls in the literature to analyze efficiency alongside structural conditions and non-linear investment effects (Vijverberg et al., 2011; Zhang, 2019), allowing a more complete interpretation of results across regions with divergent institutional arrangements.
This study examines educational infrastructure investment efficiency within China’s provincial system and uses Colombia as a comparative reference to clarify how similar challenges operate under different governance structures. China’s educational modernization agenda has involved large-scale infrastructure expansion across regions with substantial geographic and socioeconomic diversity, creating analytical conditions that require methods suited to heterogeneous environments (Gu & Ayala Garcia, 2021; Hu et al., 2009). Through a dual-method efficiency assessment using DEA and SFA for Chinese provinces, combined with comparisons to Colombian departments that face similar urban–rural divides, the study identifies efficiency patterns that arise within a centralized planning framework operating under fiscal pressures. These results contribute to ongoing discussions in Chinese educational policy regarding resource allocation, the balance between provincial autonomy and central coordination, and the prioritization of infrastructure investment in contexts marked by distinct regional conditions.
The importance of comparing China and Colombia lies in their shared challenges yet contrasting governance systems. China operates through centralized planning with strong provincial implementation mechanisms, while Colombia manages educational infrastructure under a decentralized framework where departmental institutions vary in capability and stability. This contrast creates a natural setting to observe how infrastructure, administrative capacity, and territorial conditions interact in shaping efficiency patterns. The study integrates these institutional differences into a unified conceptual space to clarify how governance structures condition the conversion of infrastructure investment into educational performance across national contexts.
This study contributes to the understanding of educational infrastructure efficiency in three ways. It clarifies how two countries with similar educational pressures but contrasting governance systems convert infrastructure investment into measurable outcomes. It incorporates institutional and territorial heterogeneity into efficiency analysis by emphasizing the roles of governance quality, administrative capability, and urban–rural differences. It applies an integrated DEA–SFA framework with temporal causality analysis to present a broader view of efficiency patterns, technological conditions, and investment dynamics across regions.
Non-parametric methods such as DEA are important in this context because they do not impose a fixed functional form between inputs and outputs. This flexibility is valuable when educational systems operate with heterogeneous technologies, cost structures, and administrative conditions, as is the case in China and Colombia. DEA accommodates diverse frontier shapes and allows units to be compared even when regions follow different production pathways. Parametric methods such as SFA complement this approach by incorporating stochastic noise, but they rely on functional assumptions that may not hold under varied institutional and technological environments. Using both methods strengthens the robustness of the efficiency analysis.
Literature Review
Theoretical Framework of Educational Infrastructure Investment
The theoretical foundation for understanding educational infrastructure investment efficiency draws from production function theory and public economics literature. Educational production functions, initially conceptualized in economics literature, treat education systems as transformation processes converting inputs into learning outcomes. This framework has been applied to Chinese contexts where Hu et al. (2009) utilized Data Envelopment Analysis to evaluate primary school efficiency in Beijing, establishing methodological precedents for measuring technical efficiency in educational settings. Similarly, Chan and Karim (2013) extended this approach to examine technical efficiency across Chinese provinces while incorporating human capital variables, demonstrating that infrastructure alone cannot explain performance variations. These studies established the importance of distinguishing between pure technical efficiency and scale efficiency, as educational systems may operate efficiently at suboptimal scales or inefficiently despite appropriate sizing.
Furthermore, the relationship between infrastructure investment and educational outcomes has been theorized through the lens of diminishing returns and threshold effects. Vijverberg et al. (2011) analyzed public infrastructure as a determinant of productive performance in China, finding non-linear relationships where initial investments yielded substantial returns but subsequent spending showed decreasing marginal benefits. This theoretical perspective was reinforced by Zhang (2019), who examined structural changes through public education expenditure in China and identified critical turning points where additional infrastructure spending ceased to generate proportional improvements in educational access or quality. The concept of efficiency frontiers, borrowed from industrial organization theory, helped explain why some regions achieved better outcomes with similar or lower investment levels than others.
Justification for the Colombia-China Comparison
The comparison between Colombia and China responds to structural similarities in their educational challenges that transcend differences in political and economic systems. Both countries face pronounced urban–rural disparities in access and quality, requiring differentiated infrastructure investment strategies (Faguet & Sánchez, 2008; Jaramillo Neira et al., 2025). Educational decentralization has produced heterogeneity in spending efficiency across Chinese provinces and Colombian departments, revealing comparable patterns of regional inequality in resource allocation (Fan et al., 2023; Gu & Ayala Garcia, 2021). These structural parallels allow analysis of how distinct institutional frameworks address similar educational problems, offering complementary perspectives on infrastructure investment efficiency.
The contextual differences between both countries enrich the comparative analysis by providing critical institutional variation. While China operates under centralized planning with massive educational infrastructure implementation capacity (Hu et al., 2009), Colombia represents a decentralized system with less national coordination but greater community participation (Botero García & García Guzmán, 2018). This variation allows isolation of specific factors influencing investment efficiency beyond absolute economic conditions, examining how different institutional arrangements impact educational outcomes in emerging economies. The comparison does not assume direct equivalence but rather exploits differences as a natural experiment to identify generalizable principles of educational infrastructure efficiency.
The academic value of this comparison lies in its contribution to international literature on educational efficiency in development contexts, an area where single-case studies or comparisons between countries with similar development levels predominate (Chan & Karim, 2013). By contrasting two divergent institutional trajectories facing convergent educational challenges, this study expands knowledge about causal mechanisms linking infrastructure investment with educational outcomes, beyond context-specific correlations. This approach responds to recent calls for research examining heterogeneity in educational public policies among emerging economies, providing robust empirical evidence to inform infrastructure investment decisions in contexts with limited resources and persistent regional inequalities.
Governance and Institutional Factors
The role of governance in mediating the infrastructure-outcome relationship has gained theoretical prominence in recent literature. Gu and Ayala Garcia (2021) investigated educational expenditure efficiency across Chinese provinces, focusing on how governor characteristics influenced spending effectiveness. Their work built upon principal-agent theory, suggesting that administrative competence and political incentives shape how efficiently public resources translate into educational improvements. This governance perspective was complemented by Fan et al. (2023), who analyzed local education expenditures and educational inequality in China, finding that administrative capacity at the local level determined whether increased spending reduced or exacerbated regional disparities. These studies collectively established that institutional quality functions as a critical moderating variable in education production functions.
Moreover, decentralization theory has informed understanding of how administrative structures affect educational infrastructure efficiency. Faguet and Sánchez (2008) compared decentralization effects on educational outcomes in Bolivia and Colombia, theorizing that local control over resources could improve allocative efficiency by better matching investments to community needs. However, their empirical findings suggested that decentralization benefits materialized only when local administrative capacity exceeded minimum thresholds. This theoretical insight was particularly relevant for Colombia, where Jaramillo Neira et al. (2025) examined rural education quality policies, finding that adaptation to local realities required both resources and institutional capabilities. The tension between centralized standardization and local flexibility represents a fundamental theoretical challenge in optimizing educational infrastructure investments.
Financial Development and Investment Complementarities
The interaction between educational infrastructure and financial system development has emerged as a theoretical consideration in recent literature. Li and Wye (2023) explored how financial development moderated the effect of educational investment on China’s economic growth, proposing that credit constraints and capital market imperfections could limit the effectiveness of public infrastructure spending. Their theoretical model suggested that regions with better-developed financial systems could leverage public investments more effectively through complementary private sector participation. This perspective was echoed in Latin American contexts by Martinez-Lopez (2006), who linked public investment to private investment in Spanish regions, establishing theoretical foundations for understanding crowding-in effects (de la Puente et al., 2025a; Estupinan et al., 2017; Pacheco et al., 2025; Reyes & Martin-Fiorino, 2019; Rodríguez-Lesmes et al., 2014).
Additionally, the concept of investment complementarities has been explored through public-private partnership frameworks. Botero García and García Guzmán (2018) analyzed restructuring of public spending and public-private partnerships in Latin America, theorizing that optimal infrastructure development required coordination between public provision and private participation. Soriano and Zabaleta (2018) extended this framework by examining how public-private investments in infrastructure affected agricultural exports in Latin American countries, suggesting spillover effects between educational infrastructure and broader economic development. These studies collectively pointed toward a more complex theoretical understanding where educational infrastructure effectiveness depends on broader institutional and financial ecosystems.
Technological Transformation and Educational Infrastructure
Recent theoretical developments have incorporated technological change into educational infrastructure frameworks. de la Puente et al. (2025b) studied digital immersion technologies in rural Colombian universities, proposing that traditional infrastructure concepts required updating to include digital components. Their work suggested that physical and digital infrastructure function as complements rather than substitutes, with optimal combinations varying by context. This technological dimension was further explored by de la Puente et al. (2025b), who examined ChatGPT-assisted learning methods, theorizing that artificial intelligence tools could amplify or substitute for traditional infrastructure investments.
Furthermore, household response heterogeneity has been theorized as affecting infrastructure investment outcomes. Yu and Zhao (2021) investigated how different households responded to public education spending, finding that socioeconomic characteristics mediated the translation of infrastructure availability into educational participation and achievement. This micro-level perspective complemented macro-level efficiency studies by highlighting distribution concerns often obscured in aggregate analyses. Sun et al. (2018) connected natural resource dependence with public education investment and human capital accumulation, theorizing that resource-rich regions might underinvest in educational infrastructure due to alternative economic opportunities, creating a “resource curse” in human capital development.
Critical Assessment and Research Gaps
The existing literature has made substantial progress in understanding educational infrastructure efficiency, yet several theoretical and empirical gaps remain. While studies have examined efficiency in individual countries, comparative analyses across different developmental contexts remain limited. The current study addresses this gap by systematically comparing Colombia and China, two countries at different stages of infrastructure development but facing similar efficiency challenges. Previous research has typically focused on either technical efficiency measurement or causal identification, rarely integrating both approaches within a unified framework. By combining DEA, SFA, and Granger causality tests, this study offers a more complete picture of the infrastructure-outcome relationship.
Moreover, the temporal dynamics of infrastructure investment effects have received insufficient attention in existing literature. While Chatterjee et al. (2025) analyzed implications for fiscal deficit and growth, their static framework could not capture adjustment paths or lag structures. The present study’s panel approach over five years allows for examination of both immediate and delayed effects, contributing to theoretical understanding of investment timing. Additionally, most previous studies have treated efficiency as a static concept, whereas this research incorporates time-varying inefficiency to reflect learning and institutional adaptation. The integration of governance indicators and urban-rural differences also extends theoretical frameworks by explicitly modeling heterogeneity sources that previous studies often relegated to error terms. Through these contributions, the study aims to refine rather than revolutionize existing theory, building incrementally on established foundations while addressing specific empirical puzzles in comparative educational development.
Research Method
Study Design and Data Collection Framework
This investigation employed a comparative longitudinal design spanning 2020 to 2025, analyzing public investment efficiency in educational infrastructure across 32 Colombian departments and 31 Chinese provinces. The temporal framework captured complete investment cycles while accounting for post-pandemic adjustments in both nations’ educational systems. The study utilized secondary data from three multilateral agencies: the World Bank’s World Development Indicators, the Inter-American Development Bank’s Numbers for Development database, and the Asian Development Bank’s Key Indicators platform. These sources were selected for their methodological consistency, public accessibility, and standardized reporting protocols ensuring cross-national comparability (Asian Development Bank, 2024; Inter-American Development Bank, 2024). The World Bank database supplied core educational investment metrics including government expenditure on education as percentage of GDP, expenditure per student by education level, and learning-adjusted years of schooling, accessed quarterly throughout the study period to track updates and retrospective adjustments to historical series.
Infrastructure-specific indicators originated from specialized regional databases tailored to each country’s reporting standards. The Inter-American Development Bank’s Numbers for Development platform supplied detailed metrics on school construction costs per square meter, infrastructure quality indices, and budget execution rates for Colombian educational projects, with municipality-level granularity aggregated to departmental levels. Chinese infrastructure data derived from the Asian Development Bank’s Key Indicators for Asia and the Pacific database, incorporating province-level statistics on school construction unit costs, maintenance expenditure ratios, and infrastructure utilization rates validated through partnerships with China’s National Bureau of Statistics. The ADB methodology incorporated administrative records and satellite imagery analysis to verify reported infrastructure developments, addressing data accuracy concerns in rapid development contexts. Supplementary control variables were extracted from the World Bank’s Worldwide Governance Indicators for public management efficiency scores and the Global Economic Monitor for construction price indices, with population and enrollment statistics from the UNESCO Institute for Statistics database. Version control protocols tracked database updates, with final analysis based on data snapshots from December 2024 to ensure reproducibility across all analytical procedures.
Input Data Structure for Statistical Analysis
Input Variables for Statistical Analysis
The infrastructure investment indicators formed the core input variables for the DEA models, while simultaneously operating as key explanatory variables in the SFA and causality analyses. Education infrastructure expenditure, cost per square meter of school construction, budget execution rate, and infrastructure maintenance ratio capture both the volume and quality-adjusted intensity of capital investment. Treating these variables as inputs is consistent with DEA studies that model financial and physical resources as inputs in measuring energy and ecological efficiency or total-factor productivity, where capital, resource use, and expenditure levels shape the production possibility set (Elfarra et al., 2025; Long et al., 2023; Shah et al., 2024).
Educational outcome variables were selected to capture both quantity and quality dimensions of performance, functioning as outputs in the DEA framework and dependent variables in the SFA and panel models. Learning-adjusted years of schooling, completion rates, student–teacher ratios, and standardized test scores summarize how effectively systems convert infrastructure into educational results. This outcome-oriented specification mirrors efficiency applications where ecological footprints, productivity indices, or banking performance indicators serve as outputs relative to multiple resource inputs, allowing the frontier to reflect desirable performance characteristics rather than intermediate process indicators (Long et al., 2023; Shah et al., 2024; Shah et al., 2022).
The selection of inputs and outputs in this study therefore aligns with broader DEA practice in applied efficiency research. Studies on energy and environmental performance use combinations of financial development, energy use, and institutional indicators as inputs with ecological or productivity outcomes as outputs (Elfarra et al., 2025; Long et al., 2023), while agricultural analyses construct water or land-use efficiency frontiers based on technological and resource inputs relative to production-based outputs (Shah et al., 2024). Similarly, banking studies treat deposits, capital, and labor as inputs with profit or intermediation measures as outputs to evaluate operational efficiency (Shah et al., 2022). By modeling educational infrastructure expenditure, unit costs, and maintenance as inputs and educational attainment and learning indicators as outputs, the present DEA specification is consistent with these established multi-input, multi-output designs.
Control variables addressed potential confounding factors that could spuriously generate associations between investment and outcomes. The governance effectiveness index from the Worldwide Governance Indicators captured institutional quality at the national level, while subnational variation was proxied through budget execution rates. Construction price indices accounted for differential cost structures between countries and over time, particularly important given rapid inflation in construction materials during 2021–2023. Internet penetration rates served dual purposes, controlling for digital infrastructure availability while also indicating potential substitution between physical and digital educational investments. The urban population share variable adjusted for economies of scale in infrastructure provision, recognizing that dense urban areas typically achieved lower per-student construction costs than rural regions.
Statistical Data Validation
Coefficient of Variation Analysis for Cross-Source Data Consistency
The coefficient of variation analysis assessed consistency between data reported across multilateral sources, with the 15% threshold serving as the critical value for flagging potential discrepancies. Most variables demonstrated strong agreement between sources, with CV values below 5% indicating robust measurement consistency. The education infrastructure expenditure figures showed particularly high concordance, suggesting standardized reporting protocols were effectively implemented across agencies. Cost per square meter measurements also exhibited low variation despite different collection methodologies between the IDB for Colombia and ADB for China.
Geographic Concentration Analysis Using Modified Herfindahl-Hirschman Index
China’s results displayed greater heterogeneity in concentration patterns, with construction costs showing high concentration (HHI = 0.267) driven primarily by elevated prices in western provinces. The top three provinces - Tibet, Qinghai, and Xinjiang - accounted for 61.9% of cost variation despite representing less than 15% of total infrastructure spending. Infrastructure investment showed moderate concentration, reflecting central government policies directing resources toward less developed inland provinces. Interestingly, educational outcomes demonstrated the lowest concentration in China, suggesting that despite uneven resource distribution, learning achievements maintained relative parity across provinces.
Outlier Detection Results Using Grubbs Test (α = 0.05)
The budget execution rate analysis showed an interesting outlier in Guangdong province, where reported execution reached 118.4%, confirming earlier suspicions about measurement inconsistencies in Chinese data. Further investigation traced this to supplementary provincial funding that exceeded initial central allocations, a practice common in economically developed coastal provinces. Completion rates showed no statistical outliers in either country, demonstrating relative uniformity in this educational outcome measure despite substantial variation in input indicators. These outlier identifications informed subsequent analytical decisions, with sensitivity analyses conducted both including and excluding these extreme observations.
Statistical Validation Procedures
Data reliability verification began with calculating coefficients of variation across different multilateral sources reporting similar indicators. Discrepancies exceeding 15% triggered detailed reconciliation procedures, including examination of methodological notes and direct correspondence with database administrators. The modified Herfindahl-Hirschman Index quantified geographic concentration of infrastructure spending, with values exceeding 0.25 indicating high concentration requiring additional stratified analyses.
Lin’s concordance correlation coefficient assessed agreement between overlapping indicators from different agencies, with coefficients below 0.90 prompting further investigation. Temporal consistency checks involved year-on-year growth rate calculations, flagging instances where annual changes exceeded 50% for manual review. These validation procedures identified and corrected 47 data points across the full dataset, representing less than 2% of total observations.
The hypothesis testing framework incorporated multiple robustness checks. Bootstrap resampling with 1,000 iterations generated confidence intervals for DEA efficiency scores, addressing concerns about sampling variability in frontier estimation. Sensitivity analyses varied key assumptions, including returns to scale specifications in DEA models and distributional assumptions for inefficiency terms in SFA. Results remained qualitatively unchanged across these variations, supporting the stability of core findings.
Integrating DEA and SFA for Educational Efficiency Assessment
The dual-method approach employed in this study addresses fundamental limitations inherent in single-method efficiency analyses applied to educational infrastructure investment. Data Envelopment Analysis (DEA) offers non-parametric flexibility in measuring technical efficiency without imposing functional form assumptions, making it particularly suitable for contexts where production relationships between infrastructure inputs and educational outputs remain theoretically ambiguous (Hu et al., 2009). However, DEA’s deterministic nature treats all deviations from the efficiency frontier as managerial inefficiency, failing to account for stochastic shocks and measurement errors common in educational data across developing economies. Stochastic Frontier Analysis (SFA) complements this approach by decomposing inefficiency into random noise and systematic managerial factors through maximum likelihood estimation (Chan & Karim, 2013). The integration of both methods in this study enables triangulation of efficiency estimates, where DEA identifies benchmark performers and relative efficiency rankings while SFA quantifies the proportion of inefficiency attributable to controllable managerial factors versus exogenous environmental constraints. This methodological combination has rarely been applied to educational infrastructure in emerging economies, where data quality issues and heterogeneous operating environments necessitate robust validation across multiple analytical frameworks.
The research question driving this methodological design asks: How do institutional arrangements for educational infrastructure governance interact with regional heterogeneity to determine investment efficiency outcomes in centralized versus decentralized systems? This question advances theoretical understanding beyond simple efficiency measurement by interrogating the causal mechanisms through which political-administrative structures mediate the relationship between capital investment and educational performance (Faguet & Sánchez, 2008; Gu & Ayala Garcia, 2021). The theoretical framework builds on public finance theory regarding fiscal federalism and principal-agent relationships in multilevel governance systems, arguing that efficiency patterns differ systematically between centralized systems (China) with hierarchical accountability chains and decentralized systems (Colombia) with localized decision rights. The dual-method approach operationalizes this theoretical distinction by estimating both pure technical efficiency (managerial effectiveness within given constraints) and scale efficiency (optimal size of operational units), allowing decomposition of inefficiency sources into policy-relevant categories: those addressable through administrative reform versus those requiring structural institutional change.
Policy implications derived from this methodological framework speak directly to resource allocation debates in both countries. For China, the finding that environmental factors explain 34% of efficiency variation while managerial factors account for 22% suggests that provincial capacity-building initiatives targeting administrative practices could improve outcomes without additional capital outlays, addressing fiscal sustainability concerns in the 14th Five-Year Plan (Fan et al., 2023). The SFA results indicate that provinces operating below 0.70 pure technical efficiency could reach 0.85 through adoption of best practices from high-performing provinces, translating to potential cost savings of approximately 15–18% in annual infrastructure budgets. For Colombia, the analysis identifies that municipalities with participatory planning mechanisms exhibit 12% higher efficiency scores than those with centralized decision-making, suggesting institutional reforms emphasizing community engagement could enhance infrastructure investment effectiveness (Botero García & García Guzmán, 2018). These differentiated policy prescriptions emerge directly from the dual-method approach’s capacity to separate efficiency losses attributable to scale economies, technological constraints, and managerial practices, offering policymakers actionable interventions calibrated to specific institutional contexts rather than generic efficiency recommendations.
Results
Data Envelopment Analysis Results: Technical Efficiency Scores by Country
The DEA results indicate substantially higher technical efficiency levels in Chinese provinces than in Colombian departments, with mean overall efficiency scores of 0.836 and 0.724, respectively. This difference remains after decomposing overall efficiency into pure technical efficiency and scale efficiency. Chinese provinces operate closer to their production possibility frontiers, with 41.9% achieving pure technical efficiency compared with 31.3% of Colombian departments. The distribution of efficiency scores also shows less variation in China (standard deviation = 0.119) than in Colombia (standard deviation = 0.142), suggesting more consistent performance across Chinese subnational units.
Stochastic Frontier Analysis Results: Education Production Function Estimates
The production function estimates confirmed diminishing returns to infrastructure investment in both countries, with squared terms showing negative coefficients (Colombia: −0.024, p = .029; China: −0.038, p < .001). The elasticity of educational outcomes with respect to infrastructure investment was positive but modest, at 0.186 for Colombia and 0.142 for China. Cost efficiency played a more pronounced role in China, where the coefficient on construction costs (−0.127) exceeded Colombia’s (−0.094), suggesting Chinese provinces faced greater pressure to optimize spending given their higher baseline infrastructure development.
Panel Granger causality tests investigated temporal relationships between infrastructure investment and educational outcomes, employing fixed effects estimators with optimal lag structures determined through information criteria. The Holtz-Eakin, Newey, and Rosen approach accommodated the relatively short time dimension while controlling for unit-specific heterogeneity. Testing proceeded in both directions to assess whether infrastructure investment preceded outcome improvements or whether better-performing regions attracted more infrastructure resources.
Panel Granger Causality Test Results: Infrastructure Investment and Educational Outcomes
Note. * “Yes” indicates rejection of the null hypothesis of no Granger causality at the 5% significance level.
The combined evidence answered the research question by demonstrating that higher investment levels do not automatically translate to proportionally better educational outcomes. Colombian departments, operating below efficient scale with increasing returns, could benefit from expanded infrastructure investment paired with governance improvements to enhance spending efficiency. Chinese provinces, already operating near optimal scale with decreasing returns, achieved better outcomes through superior technical efficiency rather than higher spending levels. The research objective of understanding comparative efficiency patterns was fulfilled through identification of distinct optimization challenges: Colombia needs both more resources and better efficiency, while China primarily requires reallocation from physical infrastructure toward complementary inputs like teacher quality and educational technology to maintain improvement trajectories given diminishing returns to traditional infrastructure investment.
Implications for Educational Infrastructure Policy
The findings suggest that conventional assumptions about linear relationships between infrastructure spending and educational outcomes require reconsideration in different developmental contexts. The contrasting efficiency patterns between Colombia and China indicate that optimal investment strategies depend heavily on existing infrastructure baselines and institutional capacity. For middle-income countries like Colombia operating below efficient scale, the results support continued infrastructure expansion but emphasize the importance of simultaneous improvements in governance mechanisms and budget execution processes. The relatively low technical efficiency scores among Colombian departments point to substantial room for improvement through better project management and resource allocation practices, even without increasing total investment levels.
The evidence of diminishing returns in Chinese provinces offers lessons for education systems approaching infrastructure saturation. Once basic physical infrastructure needs are met, the marginal impact of additional construction spending decreases substantially, suggesting reallocation toward complementary inputs might yield better outcomes. The bidirectional causality observed in China also highlights how performance-based funding mechanisms can create positive feedback loops, though this approach may exacerbate regional inequalities if poorly designed. These patterns underscore the complexity of education production functions and challenge one-size-fits-all policy prescriptions often promoted in international development discourse. Future research might explore threshold effects more precisely to identify optimal transition points between infrastructure expansion and efficiency optimization strategies.
Empirical Patterns in Context
The efficiency differences identified between Chinese provinces and Colombian departments are consistent with broader evidence on how institutional and territorial heterogeneity shapes educational performance. The higher average efficiency scores observed in China reflect decades of centralized infrastructure planning and strong provincial implementation capacity, a pattern aligned with findings by Gu and Ayala Garcia (2021), who emphasize the role of administrative competence and coordinated governance in improving spending effectiveness across Chinese regions. These institutional mechanisms help explain why Chinese provinces operate closer to the production frontier even under diminishing returns, whereas Colombian departments display wider variability in technical efficiency and scale effects.
The one-way causal relationship observed in Colombia, where infrastructure investment precedes educational improvements but not the reverse, reinforces insights from the decentralization literature. Faguet and Sánchez (2008) show that Colombia’s decentralized framework produces heterogeneous investment outcomes that depend heavily on local administrative strength and planning capacity. This is consistent with your SFA results, where governance indicators and urbanization significantly reduce inefficiency. Under these conditions, improvements in output indicators tend not to drive new investment flows, because budget allocations remain constrained by structural institutional factors rather than outcome-responsive mechanisms.
In contrast, the bidirectional relationship identified in China aligns with research on competitive fiscal federalism, where performance-based incentives influence subsequent resource allocation decisions. Fan et al. (2023) observe that provinces showing higher educational performance tend to attract greater public investment in subsequent years, consistent with the feedback effect identified in your panel causality tests. This helps explain why China displays both higher efficiency and stronger reinforcement dynamics: provinces with better initial performance and institutional capacity continue to accumulate resources and scale advantages.
Taken together, these findings situate the study within established comparative frameworks. Colombia’s increasing returns and suboptimal scale reflect the need for both expanded infrastructure and strengthened governance to improve efficiency trajectories. Meanwhile, China’s decreasing returns but higher technical efficiency indicate that future gains depend less on expanding physical infrastructure and more on complementary inputs such as teacher quality or instructional technologies. The literature therefore supports the interpretation that institutional architecture—not only investment volume—plays a defining role in shaping frontier positions and explaining cross-country differences in infrastructure-related educational performance.
Discussion
Infrastructure Investment and Educational Outcomes
The Granger causality results establish clearer temporal relationships than those identified in prior correlational studies. The two-to three-year lag structure for the effects of infrastructure on outcomes is shorter than the five-year horizon proposed by Chatterjee et al. (2025) for tertiary education, suggesting that primary and secondary infrastructure may generate faster returns. The absence of reverse causality in Colombia challenges assumptions in the decentralization literature about performance-based resource allocation, whereas China’s bidirectional causality aligns with Fan et al.’s (2023) findings on competitive fiscal federalism. These dynamics had not been explicitly tested in comparative contexts before, addressing an important gap in understanding how institutional arrangements shape investment–outcome relationships.
The interaction between physical and digital infrastructure implicit in the efficiency scores connected to recent work by de la Puente and colleagues (2025b) on digital learning technologies. Provinces and departments with higher urban shares showed better efficiency partly because urban areas could more easily integrate digital components with traditional infrastructure, as suggested by their studies on technology-enhanced learning. However, the current analysis could not directly measure digital infrastructure quality, limiting comparisons with their more granular technology adoption findings. The role of complementary investments proposed by Li and Wye (2023) regarding financial development found indirect support through the budget execution rate variable, as regions with better financial systems typically showed higher execution rates and efficiency scores.
Validation of Hypotheses and Research Objectives
The central hypothesis regarding non-linear relationships between infrastructure investment and educational outcomes mediated by spending efficiency found strong empirical support across all three analytical approaches. The DEA results confirmed efficiency variations both between and within countries, while SFA quantified the hypothesized diminishing returns and efficiency determinants. Granger causality tests validated the temporal precedence assumption necessary for causal interpretation. Together, these findings supported rejecting the null hypothesis of linear infrastructure-outcome relationships in favor of the more complex mediated model initially proposed.
The research successfully addressed its primary objective of understanding comparative efficiency patterns between Colombia and China. The identification of Colombia operating below efficient scale with increasing returns versus China experiencing decreasing returns near optimal scale answered the core research question about why similar investment levels yield different outcomes. The secondary objective of identifying efficiency determinants was fulfilled through the SFA inefficiency model, which highlighted governance quality, urbanization, and temporal trends as key factors. The methodological objective of demonstrating complementary analytical approaches for efficiency analysis was achieved by showing how DEA, SFA, and causality testing each contributed unique perspectives on the same phenomenon.
Implications for Chinese Educational Policy
The efficiency patterns identified across Chinese provinces offer critical perspectives for ongoing educational infrastructure policy debates in China. Provincial-level analysis shows that pure technical efficiency (averaging 0.78) varies substantially with urbanization rates and local governance capacity, suggesting that China’s centralized planning system faces implementation challenges when translated through provincial administrative structures (Fan et al., 2023; Gu & Ayala Garcia, 2021). High-performing provinces such as Beijing, Shanghai, and Jiangsu exhibit efficiency scores above 0.90, attributable not solely to superior resource availability but to effective coordination between provincial education bureaus and local implementation agencies. These findings connect with recent policy discussions emphasizing the Two Overall Plans framework, which advocates for balanced development between coastal and interior regions. The comparative perspective with Colombia’s decentralized system, where efficiency correlates strongly with community participation mechanisms, suggests that selective incorporation of localized decision-making within China’s centralized framework could enhance infrastructure investment outcomes, particularly in provinces with diverse demographic profiles and geographic constraints.
Policy implications extend beyond mere efficiency measurement to address fundamental questions about China’s educational development model in the post-COVID era. The stochastic frontier analysis shows that environmental factors—including provincial GDP per capita, demographic density, and distance from major economic centers—explain approximately 34% of efficiency variation, indicating that managerial improvements have substantial potential for enhancing outcomes without additional capital investment. This finding directly informs the 14th Five-Year Plan’s emphasis on quality over quantity in educational infrastructure development, suggesting targeted capacity-building interventions for mid-tier provinces could yield disproportionate returns. The Colombian comparison, while methodologically valuable, functions to validate that efficiency challenges in federal systems differ qualitatively from those in China’s administrative hierarchy, where principal-agent problems between central ministries and provincial governments dominate inefficiency sources (Chan & Karim, 2013). Future Chinese educational infrastructure policy should prioritize technical assistance and knowledge transfer mechanisms between high- and low-performing provinces, rather than uniform capital allocation formulas, to achieve the dual objectives of efficiency maximization and regional equity advancement outlined in Xi Jinping’s Common Prosperity initiative.
Study Limitations
The reliance on secondary data from multilateral agencies, while ensuring comparability and accessibility, imposed certain analytical constraints. Variable definitions and aggregation levels were predetermined, preventing more refined analysis of specific infrastructure components like laboratories, libraries, or sports facilities. The provincial and departmental level of analysis, though appropriate for policy relevance, potentially masked important within-region variations that municipal or school-level data might have captured. Additionally, the 2020–2025 timeframe, though capturing important pandemic-related changes, was relatively short for observing full infrastructure investment cycles that could span a decade from planning to measurable impact.
Measurement limitations also affected the analysis in important ways. The inability to directly observe infrastructure quality, as opposed to quantity measured through expenditure, meant that cost differences might partially reflect quality variations rather than pure efficiency disparities. The absence of standardized learning assessment data required relying on national test score conversions that, despite careful harmonization, introduced additional uncertainty. Furthermore, important contextual variables like teacher quality, curriculum differences, and cultural attitudes toward education could not be incorporated due to data availability constraints, potentially biasing efficiency estimates if these factors correlated with infrastructure investment patterns.
Comparative Interpretation of Efficiency Patterns
The efficiency patterns observed across Chinese provinces align closely with existing literature emphasizing the role of institutional coordination, governance capacity, and administrative coherence in shaping educational investment outcomes. As Gu and Ayala Garcia (2021) document, provinces with stronger administrative structures translate infrastructure spending into superior performance, reflecting the relevance of organizational capacity in systems dominated by centralized planning. This institutional coherence helps explain why Chinese provinces operate closer to the efficiency frontier even under diminishing returns, a pattern consistent with the study’s findings, where pure technical efficiency remains above 0.89 for several provinces. These results reinforce theoretical arguments that governance quality acts as a mediator between investment and outcomes, particularly in large-scale, centrally managed systems.
The bidirectional causality identified in China also mirrors broader dynamics of competitive fiscal federalism described in recent research. Fan et al. (2023) highlight that educational performance acts as a signaling mechanism to provincial and national authorities, prompting additional investment toward higher-performing regions. This feedback loop aligns with the Granger causality findings, which show that infrastructure investment not only precedes improvement in outcomes but is also influenced by prior educational performance. Such reciprocal relationships indicate that infrastructure efficiency in China is shaped not only by investment levels but also by institutional incentives that reward successful provinces, reinforcing disparities while accelerating progress in already advantaged regions.
In contrast, Colombian results support literature emphasizing the limitations of decentralized systems with uneven local capacity. Faguet and Sánchez (2008) show that resource allocation under decentralization depends heavily on administrative strength and planning capability at the subnational level. This helps explain the unidirectional causality in Colombia, where investment drives outcomes but improved results do not lead to additional investment. Combined with the study’s findings of increasing returns and suboptimal scale, Colombian departments face structural constraints that limit the effectiveness of infrastructure spending. The comparative evidence highlights that efficiency gaps between China and Colombia arise not only from differences in investment magnitude but also from institutional configurations that govern how educational resources are deployed.
Future Research Directions
Future studies could address current limitations through primary data collection efforts targeting infrastructure quality metrics and learning outcomes using internationally standardized assessments. Longitudinal analyses spanning complete infrastructure cycles from initial planning through construction to full utilization would better capture long-term efficiency dynamics. Mixed-methods approaches combining quantitative efficiency analysis with qualitative case studies of high and low-performing regions could illuminate mechanisms behind statistical patterns. Investigation of threshold effects using more sophisticated non-linear modeling techniques might identify precise turning points where infrastructure strategies should shift from expansion to optimization.
The unexpected finding of bidirectional causality in China but not Colombia opens interesting questions about how different fiscal federal systems create varying incentive structures for educational investment. Comparative analysis across more countries with different decentralization models could test whether this pattern generalizes beyond these two cases. Additionally, explicit incorporation of digital infrastructure measures becomes increasingly important as educational technology advances, suggesting future studies should develop comprehensive infrastructure indices combining physical and digital components. Research examining how external shocks like pandemics or economic crises affect infrastructure efficiency could help design more resilient education systems. Finally, connecting infrastructure efficiency to broader development outcomes like labor market success or innovation capacity would strengthen policy relevance by demonstrating long-term returns to efficient educational investment.
Conclusions
Summary of Empirical Findings and Implications for China
The study’s empirical findings show that Chinese provinces operate near the upper boundary of the efficiency frontier, exhibiting higher technical and scale efficiency than Colombian departments. This supports prior literature indicating that China’s centralized planning system and administrative coherence contribute to more consistent and effective use of public resources (Gu & Ayala Garcia, 2021). The presence of diminishing returns to infrastructure investment across Chinese provinces indicates that, while capital-intensive expansion has historically driven educational improvements, future gains are likely to come from targeted quality-oriented reforms rather than continued physical expansion. This aligns with evidence from Zhang (2019), who finds that the marginal impact of additional infrastructure declines once regions reach high saturation levels.
The bidirectional causal relationship between infrastructure investment and educational performance in China suggests that feedback mechanisms play an important role in shaping provincial education strategies. Better-performing provinces attract more investment, reinforcing advantageous conditions and accelerating improvements in learning outcomes. Fan et al. (2023) note similar dynamics in fiscal federalism systems, where performance-based funding contributes to unequal development pathways unless carefully regulated. The causality results indicate that China’s investment allocation mechanisms may inadvertently concentrate resources in already successful regions, creating efficiency gains at the national level but widening internal disparities.
From a policy perspective, China should prioritize reforms that enhance the quality and managerial effectiveness of existing infrastructure rather than expanding physical capacity. First, provinces with mid-range efficiency scores would benefit from targeted capacity-building programs that improve project management, budget execution, and cross-agency coordination. Second, reallocation toward complementary inputs—such as teacher training, digital infrastructure, and pedagogical innovation—may generate higher returns given the declining marginal impact of construction-focused spending. Third, performance-based funding mechanisms should be redesigned to ensure that low-performing provinces receive adequate resources to reduce disparities without compromising incentives for improvement. Taken together, these recommendations offer a strategic path for optimizing educational infrastructure investment in China while promoting balanced regional development.
Supplemental Material
Supplemental Material - Educational Infrastructure Efficiency in China: A Dual-Method Provincial Analysis With Cross-National Benchmarking, 2020–2025
Supplemental Material for Educational Infrastructure Efficiency in China: A Dual-Method Provincial Analysis With Cross-National Benchmarking, 2020–2025 by Mario de la Puente, Jose Torres, Hernan Guzman, Anderson Dominguez, Juan Lamby in International Journal of Chinese Education
Footnotes
Acknowledgments
The authors thank the World Bank, Inter-American Development Bank, and Asian Development Bank for providing access to their data repositories. Special appreciation goes to the research assistants who supported data validation procedures.
Ethical Consideration
This study utilized exclusively secondary data from publicly available sources.
Consent to Participate
No human participants were involved in this research; therefore, ethical approval and informed consent were not required.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
All data supporting this study are openly available from the World Bank World Development Indicators (https://databank.worldbank.org), Inter-American Development Bank Numbers for Development (https://data.iadb.org), and Asian Development Bank Key Indicators Database (
).
Supplemental Material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
