The Infrastructure Paradox in Ecuador: Public Investment and the Persistence of Territorial Disparities in Cantons with Low Initial Development

Abstract

This paper examines the effects of public infrastructure spending across Ecuadorian cantons on adequate employment and the multidimensional poverty rate over the period 2008–2022, assessing whether such spending operated as a mechanism of convergence. The analysis is grounded in the hypothesis that infrastructure investment yields stronger effects in cantons characterized by lower initial levels of development. The coefficient of primary interest, associated with the interaction between low initial development and infrastructure expenditure, takes a value of −0.9542, which theoretically indicates a substantially larger impact on outcome variables in lagging cantons. This pattern is consistent with convergence theory and with the notion of investment spillovers, often described as a trickle-down process of development. Nevertheless, the estimated effect is marginally significant at the 10% level. In light of these results, the discussion revisits the convergence hypothesis by emphasizing the role of endogenous and institutional factors in shaping inclusive development across Ecuadorian cantons. The discussion underscores the importance of public expenditure quality in Latin America (LATAM) as a critical factor impacting the region’s economic growth, social equity, and overall development.

1. Introduction

1.1. The Dual Paradox of Public Infrastructure Investment

This research examines the impact of public infrastructure spending across Ecuadorian cantons on adequate employment and multidimensional poverty from 2008 to 2022. It investigates whether such investment functioned as a mechanism of convergence, disproportionately benefiting the country’s most socioeconomically disadvantaged territories in areas including employment, poverty reduction, access to education, and the satisfaction of basic needs. Regional economic convergence remains a challenge for developing countries, particularly those with significant territorial heterogeneity. Strategic investment in public infrastructure is a primary macroeconomic policy instrument for driving this process, aiming to improve productivity, connectivity, and access to markets and services to mitigate territorial disparities (Solow, 1956; Swan, 1956). This need is particularly relevant in Latin America, where spatial disparities persist not only between regions but also within subnational units such as provinces and cantons (Galperin et al., 2022). Economic convergence theory posits that less developed regions will tend to catch up with their more advanced counterparts. Strategic public infrastructure investment is a central mechanism hypothesized to drive this process. Indeed, empirical evidence from 17 Latin American countries indicates that, at the regional level, a 10% increase in infrastructure investment is associated with a 1.42% reduction in poverty and a 1.66% to 2.02% decline in extreme poverty (Durán & Saavedra, 2014). Consequently, these results problematize the simplistic trickle-down narrative of development (Roberts, 2023). They indicate that infrastructure investment is a necessary but insufficient condition, constituting just one component within a broader set of factors required for development. Historically, decision-making analyses have relied on the imprecise metric of Gross Domestic Product (GDP) as a proxy for development, thereby overlooking critical dimensions that more accurately reflect socioeconomic realities and the true territorial impact of infrastructure.

Despite the pressing need for infrastructure investment to foster development, the quality of public expenditure in LATAM suffers from chronic inefficiencies, misallocation, and opacity (S. Almeida et al., 2025), which waste up to 4% of regional GDP (World Bank Group [WBG], 2025). Addressing Latin America’s infrastructure deficit is hindered by economic volatility, political instability, and a complex regulatory framework that stifles financing (World Bank Group [WBG], 2020; Allianz Global Investors, 2025). The historical deficit in infrastructure investment across the region has served to deepen, rather than mitigate, existing economic inequalities, including interpersonal poverty gaps and spatial/regional disparities. Studies indicate that insufficient investment in infrastructure can lead to a loss of up to 15 percentage points in GDP growth over a decade (Banco Interamericano de Desarrollo [BID], 2019), particularly affecting the poorest households, which may see an 11-percentage-point decrease in real income over the same period (Serebrisky & Suárez-Alemán, 2021). This substantial drain has made structural reforms to improve investment efficacy an urgent priority for the region in order to achieve inclusive growth. This potential exists in tension with a chronic reality of inefficient public expenditure, which can waste substantial resources and, perversely, deepen the very inequalities it seeks to remedy.

This scenario reveals a double paradox at the core of contemporary development policies, particularly evident in Latin American economies such as Ecuador. First, the GDP paradox: despite decades of theoretical and empirical critiques, GDP remains the predominant criterion for public investment decisions. This over-reliance on aggregate indicators perpetuates an ‘information gap,’ often overlooking multidimensional and territorially specific socioeconomic outcomes—such as decent employment and multidimensional poverty—which better reflect development and the local distributive effects of infrastructure (Roberts, 2023). Second, and central to this study, is the territorial paradox: the risk that investments aimed at spatial integration may, under certain political and institutional conditions, trigger opposing forces that reinforce political fragmentation or fail to reach disadvantaged areas. How is this paradox reflected in public infrastructure? The central empirical question addressed in this article is whether large-scale spending fosters genuine territorial convergence or, conversely, risks entrenching existing privileged geographies.

Ecuador serves as a critical case study for examining the paradoxes inherent in public investment and territorial development. Despite substantial investment in physical capital, such as road networks, hydroelectric plants, schools, and hospitals intended to unify the national territory, significant and persistent cantonal disparities remain. This study argues that a comprehensive understanding of infrastructure requires moving beyond a basic input–output framework to consider infrastructure as a quasi-public good with contested territorial outcomes. Although infrastructure theoretically exhibits public-good characteristics, including non-rivalry and non-excludability, its planning, allocation, and localized impacts are shaped by political and spatial dynamics. The potential for infrastructure to promote territorial convergence or divergence depends on institutional quality, pre-existing local conditions, and the alignment of investment with local productive capacities (Rodríguez-Pose, 1999). Then, the theoretical framework juxtaposes convergence theory (Abramovitz, 1990), which emphasizes capital accumulation to close development gaps, with endogenous growth models (Romer, 1990; Labarca et al., 2021), which stress innovation and local capabilities.

This research makes its contribution at this intersection. We conduct a canton-level longitudinal analysis of Ecuador (2008–2022) to empirically test whether public infrastructure spending has functioned as a convergent mechanism for disadvantaged areas. By doing so, it challenges trickle-down arguments (Komlos, 2019) that large-scale investment primarily benefits already-developed areas. Such potential asymmetric effects necessitate a critical reassessment of infrastructure as a panacea for regional development and underscore the need for robust empirical evidence based on official public data and spatial econometric methods. From a policy perspective, the findings advocate for public interventions adapted to specific territorial conditions, thereby contributing to the achievement of Sustainable Development Goal (SDG) 10 and the principle of leaving no one behind (United Nations Development Programme [UNDP], 2015).

1.2. Literature Review

1.2.1. From Public Goods to Spatial Outcomes

The theoretical rationale for public infrastructure investment originates in public goods theory, defined by non-rivalry and non-excludability, which justifies state provision due to market failure (Samuelson, 1954; Musgrave, 1959). Endogenous growth theory expands this, positing that such investment directly drives long-term growth and innovation, supporting the neoclassical convergence hypothesis that lagging regions should catch up due to higher returns on capital (Solow, 1956; Romer, 1990).

However, evidence from applied territorial studies suggests that this mechanism is far from automatic. In practice, infrastructure often functions as a quasi-public good with spatially bounded benefits. Its allocation is a political process, and access is often uneven, leading to geographic disparities (Tiebout, 1956). This dynamic engages a “territorial paradox”, whereby integrative investments may paradoxically entrench fragmentation rather than reduce it (Zhang, 2025). Furthermore, class-based spatial segregation shapes political coalitions for public goods, affecting their equitable provision (Xu, 2024).

Empirical evidence from Ecuador confirms these complexities. Public investment has supported slight convergence but coexists with persistent segregation, limiting productivity gains in smaller provinces (Correa-Quezada et al., 2019). The impact is also multidimensional; for example, sanitation infrastructure has proven critical for building health resilience to climate shocks (Lipscomb et al., 2025). Taken together, this literature indicates that infrastructure’s territorial effects depend critically on political, spatial, and institutional conditions, challenging its simplistic treatment as an automatic tool for equitable development.

1.2.2. Empirical Literature on Allocation and Impact

Empirical evidence from Latin America demonstrates that the governance and targeting of public investment are critical determinants of its effectiveness. Persistent inefficiency, misallocation, and corruption substantially reduce potential returns, as wasteful expenditures divert resources from projects with proven impact (R. Almeida et al., 2020; Cavallo & Daude, 2011; Izquierdo et al., 2021).

Research conducted in Ecuador presents a complex landscape, indicating that outcomes are highly contingent on project design and resource allocation. Targeted investments in digital and social infrastructure have demonstrated effectiveness by promoting local economic diversification and enhancing human capital (Galperin et al., 2022; Navarro Silva & Flores Tabara, 2017). Nevertheless, these positive results are frequently eclipsed by a prevailing preference for large-scale, politically driven “flagship” projects. This tendency perpetuates a territorial paradox, as investment is disproportionately directed toward already-developed areas, thereby intensifying spatial inequalities instead of promoting regional convergence (Carrión & Herrera, 2020; Gwilliam, 2011).

Furthermore, the type of infrastructure determines its developmental impact. While connective infrastructure like broadband can be transformative, investments primarily serving extractive industries often generate localized social and environmental costs without fostering broad-based local development, underscoring infrastructure’s role as a contested political good (Bebbington et al., 2018).

1.2.3. Between Potential and Paradox

The synthesis of the literature reveals a fundamental tension that this study is uniquely positioned to address. While theoretical frameworks and specific micro-studies highlight the strong potential of infrastructure as a driver of equitable growth (Galperin et al., 2022), broader empirical evidence from Ecuador and Latin America consistently reveals paradoxical outcomes. Between 2008 and 2022, Ecuador underwent substantial changes in its development model. The initial years were marked by elevated public investment, primarily funded by commodity revenues, especially from oil exports. This increase in infrastructure spending took place within a state-led development framework that prioritized redistribution and territorial integration. After the decline in commodity prices and the emergence of fiscal constraints, public investment became more volatile and subject to frequent budgetary adjustments. These macroeconomic dynamics influenced both the scale and territorial distribution of infrastructure spending, which may have contributed to the varied development outcomes observed across cantons.

Research in Ecuador confirms this disconnect, showing that public investment has fostered only a “slight” process of convergence, failing to significantly boost productivity in smaller provinces and occurring alongside persistent territorial “segregation” (Correa-Quezada et al., 2019). This paradox is further evidenced at the cantonal level, where critical inputs like human capital can actually increase inequality due to geographic concentration, while micro-enterprises help mitigate it (Ponce et al., 2024). This pattern is reinforced by findings of persistent “poverty traps” in Ecuadorian municipalities (Sánchez et al., 2025) and highly uneven social vulnerability (Sánchez et al., 2018), demonstrating how spatial disadvantage can resist aggregate investment.

This dynamic reflects a broader regional pattern. Latin America exhibits a pronounced “innovation paradox,” in which high potential returns on technology and capability-building are constrained by systemic underinvestment, weak business environments, and a lack of frontier firms (Maloney, 2025). This phenomenon is evident even in digital development, where increased access is accompanied by persistent inequalities in quality and affordability. As a result, a “half-finished” technological leap may reinforce existing disparities within new systems (J.P. Morgan Private Bank, 2025). Despite these challenges, few studies systematically examine whether public infrastructure investment at the national sub-unit level over time leads to convergence or divergence. The application of robust spatial econometric methods to canton-level data in Ecuador addresses this research gap. This approach advances beyond isolated case studies or macro-level critiques by providing a systematic assessment of whether investment fulfills its theoretical promise as a public good or, consistent with the territorial paradox (Zhang, 2025), perpetuates existing spatial inequalities.

2. Materials and Methods

The analysis is grounded in the hypothesis that the impact of infrastructure investment has been significantly stronger in cantons with lower initial levels of development, a pattern that would be consistent with convergence theory rather than with endogenous growth explanations. To empirically assess this premise, the study relies on a panel data framework at the cantonal level in Ecuador, examining the evolution of adequate employment and multidimensional poverty in relation to infrastructure investment across cantons over the period 2008–2022. The results are subsequently discussed in light of contrasting development theories, drawing on national and regional evidence to reflect on the determinants of territorial development that are not explicitly captured by the model. This broader discussion seeks to frame inclusive territorial development as a pathway toward comprehensive economic growth, benefiting both poorer populations and structurally lagging cantons in Ecuador. Ultimately, the adopted methodology addresses a critical need in LATAM to systematically evaluate the role of public expenditure quality and policy effectiveness as drivers of development, with specific attention to their impact on poverty alleviation and adequate employment generation. By providing a replicable framework for spatially disaggregated analysis, this approach offers a model with significant relevance for other developing regions facing similar challenges of territorial inequality and inefficient public investment.

2.1. Data Collection

For this research, a longitudinal panel dataset was constructed covering all 221 cantons of Ecuador between 2008 and 2022. Information on public infrastructure spending was obtained from official records of the Ministry of Economy and Finance, while socioeconomic indicators—adequate employment, multidimensional poverty, unmet basic needs, and educational attendance—were sourced from the Instituto Nacional de Estadísticas y Censos (INEC) based on the Encuesta Nacional de Empleo Desempleo y Sub-empleo (Enemdu) translated National Survey of Employment, Unemployment, and Underemployment. A substantial data cleaning and harmonization process was required to reconcile information from these two institutions. In particular, territorial codes were standardized to account for administrative changes during the study period, such as the creation of the province of Santa Elena with its respective cantons and the reassignment of the canton of La Concordia. This process ensured temporal consistency across panel series and preserved data completeness when merging information from different public sources using cantonal identifiers.

The dataset captures executed annual infrastructure spending disaggregated by canton, providing a more accurate measure of actual fiscal effort than planned allocations. However, it does not directly measure the quality of public expenditure or distinguish between well-executed and inefficient spending, as it lacks project-level performance indicators or direct metrics of institutional efficiency. This limitation is relevant from an institutional perspective, as variations in spending quality are closely linked to local governance capacity and policy design. Consequently, part of the estimated effects may reflect unobserved institutional heterogeneity rather than the impact of expenditure levels per se. Future research should address this gap by incorporating indirect proxies for spending quality, such as indices of government effectiveness, control of corruption, or project execution rates, which would allow for a more explicit assessment of institutional efficiency—a necessary refinement for models applied to Ecuador and the broader Latin American context. Furthermore, the analysis does not account for other unobserved local factors, such as social capital or the degree of productive diversification, nor does it include cantonal GDP data. Including these variables in future model extensions could help explain residual heterogeneity in development outcomes.

2.2. Research Design

Low Initial Development and Infrastructure Effects

The study examines cantonal data over time to address two central questions: first, whether infrastructure investment effectively contributed to the generation of quality employment, taking into account the level of multidimensional poverty in each canton; and second, whether this effect was homogeneous across territories or whether less developed cantons experienced a significantly different impact on multidimensional poverty reduction. To address these questions, two theoretically grounded models were specified in the results to capture the transmission channels linking infrastructure investment to socioeconomic outcomes, and a description of the difference between planned and executed budgets was provided to introduce efficiency as a proxy for quality expenditure. The discussion presents the counter position in theory, and the conclusion underscores the importance of public expenditure quality in LATAM as a critical factor impacting the region’s economic growth, social equity, and overall development.

In order to build a model from the perspective of public capital theory (Aschauer, 1989), infrastructure investment is treated as a productive asset that enhances private marginal productivity, reduces transport and energy costs, and facilitates both direct employment creation—such as in construction—and indirect employment through sectors like trade and logistics. Complementarily, the multidimensional poverty framework emphasizes that deprivation extends beyond income to include access to basic services such as education, drinking water, and public health (Alkire & Foster). In addition, inequality theory suggests that infrastructure can help narrow urban–rural gaps, particularly in cantons with limited initial access to infrastructure and a high density of working-age population (Glaeser, 2014).

Based on these conceptual foundations, two equations are specified:

$${Ea}_{it}={{\beta }_0+\beta }_1{GPInf}_{it}+{\beta }_2{TPM}_{it}+{\propto }_i+y_t+e_{it}$$

(1)

where

• ${Ea}_{it}=$ Adequate employment indicator for canton i in year t.
• ${GPInf}_{it}=$ Public infrastructure expenditure (executed).
• ${TPM}_{it}$ = Control variable vector capturing the multidimensional poverty rate.
• ${\propto }_i$ = Canton fixed effects (controlling for time-invariant characteristics).
• $y_t=$ Year fixed effects (controlling for common shocks, such as crisis).
• $e_{it}=$ Error term.

A description of the variables used in this study, along with their expected signs, is provided in

Table 1. Variables for Model 1.

Variable	Description and Measurement	Theoretical Justification and Literature	Expected Empirical Relationship with ${\mathit{E}\mathit{a}}_{\mathit{i}\mathit{t}}$ (Adequate Employment)
${Ea}_{it}$ (Dependent Variable)	Rate of adequate employment in canton i, year t. Measures employment meeting criteria for sufficiency, security, and rights (International Labour Organization [ILO], 2008).	Central to endogenous growth and human development frameworks. Quality employment signals productive economic absorption and is a key development outcome (Anand & Sen, 2000; Labarca et al., 2021; Romer, 1990).	Not applicable—This variable is specified as the dependent variable in the model.
${GPInf}_{it}=$ (Key Independent Variable)	Executed public infrastructure expenditure (log-transformed) in canton i, year t. Captures actual investment in transport, utilities, etc.	Rooted in public goods theory and endogenous growth. Infrastructure boosts productivity, reduces costs, and can stimulate job creation in formal sectors (Aschauer, 1989; Romer, 1990).	Positive (β1 > 0). Higher investment should correlate with higher adequate employment via productivity and market access channels.
${TPM}_{it}$ (Control Variable)	Multidimensional Poverty Index rate for canton i, year t. Captures simultaneous deprivations in health, education, and living standards (Alkire & Foster, 2011).	Poverty constrains human capital and labor participation (structuralist/dependency views). A critical control to isolate infrastructure’s effect from a core socioeconomic correlate (Ravallion & Datt, 2002).	Negative. Higher poverty is expected to correlate with lower adequate employment due to constrained capabilities and labor informality.
${\propto }_i$ (Canton Fixed Effects)	Dummy variable for each canton i. Controls for all unobserved, time-invariant canton characteristics.	Standard panel data method to address omitted variable bias. Controls for geographic, cultural, or historical factors constant over time (Wooldridge, 2010).	Not applicable (Control). Absorbs level differences, ensuring β1 estimates within-canton changes over time.
$y_t$ (Year Fixed Effects)	Dummy variable for each year t. Controls for common shocks affecting all cantons in a given year.	Controls for macroeconomic cycles, national policy shifts, or common external shocks (e.g., oil price crashes, the COVID-19 pandemic).	N/A (Control). Isolates the effect from nationwide temporal trends.
$e_{it}$ (Error Term)	Idiosyncratic error term capturing all other unobserved factors influencing ${Ea}_{it}$ in canton i, year t. Assumed to be independently and identically distributed for consistent estimation. Spatial correlation may be a concern (Anselin, 1988). Captures random variation and model misspecification.

.

Based on the foregoing, the second model is specified as follows:

$${TPM}_{it}={\beta }_0++{\beta }_1{(\mathrm{l}\mathrm{n}\_ID}_{it})+{\beta }_2\left({\mathrm{l}\mathrm{n}\_DI}_{it}\times {Bajodesarrollo}_i\right)+{\alpha }_i+y_t+e_{it}$$

(2)

where

• ${TPM}_{it}=$ Multidimensional poverty indicator for canton i in year t.
• ${\mathrm{l}\mathrm{n}\_ID}_{it}=$ Executed infrastructure investment in logarithmic form.
• ${Bajodesarrollo}_i=$ Dummy variable (1 = lower percentile of the initial development index).
• ${\mathrm{l}\mathrm{n}\_DI}_{it}\times {Bajodesarrollo}_i=$ Differential effect in lagging cantons.
• ${\propto }_i$ = Canton fixed effects (controlling for time-invariant characteristics).
• $y_t=$ Year fixed effects (controlling for common shocks).
• $e_{it}=$ Error term.

A description of the variables used in the model, along with their expected signs, is provided in

Table 2. Variables for Model 2.

Variable	Description and Measurement	Theoretical Justification and Literature	Expected Empirical Relationship with ${\mathit{T}\mathit{P}\mathit{M}}_{\mathit{i}\mathit{t}}$ (Multidimensional Poverty)
${TPM}_{it}$ (Dependent Variable)	Multidimensional Poverty Index rate for canton i, year t.	Central to the human development approach and SDG 1, capturing simultaneous deprivations in health, education, and living standards (Alkire & Foster, 2011). It is the key territorial equity outcome.	Not applicable (Dependent variable).
${ID}_{it}$ (Key Independent Variable)	Natural logarithm of executed infrastructure investment in canton i, year t.	Rooted in public goods theory. Log transformation is standard for investment variables to interpret elasticity and address skewed distributions.	Negative (β1 < 0). The baseline hypothesis is that higher investment reduces poverty.
$BajoDesarrolloi$ (Moderator Variable)	Dummy variable (1 = canton in the lower percentile of the initial development index; 0 otherwise).	Captures initial conditions critical to convergence theory and the concept of spatial poverty traps. It operationalizes the distinction between “lagging” and other territories.	Not applicable (Moderator). Its coefficient in isolation is subsumed by canton fixed effects ${\propto }_i$.
${\mathrm{l}\mathrm{n}\_DI}_{it}\times {Bajodesarrollo}_i$ (Key Interaction Term)	Interaction between logged investment and the low-development dummy.	Tests for conditional convergence (Barro & Sala-i-Martin, 1992) and the territorial paradox. A negative β2 would signal stronger poverty reduction in lagging cantons (convergence). A positive β2 would signal weaker or adverse effects (divergence/paradox).	β2 ≠ 0. The sign and significance of β2 directly test your core research question.
${\propto }_i$ (Canton Fixed Effects)	Dummy for each canton i.	Controls for all time-invariant canton characteristics (e.g., geography, persistent cultural factors). Critically, it absorbs the standalone effect of BajoDesarrollo_i.	N/A (Control). Ensures the interaction term captures only the differential effect of investment over time.
$y_t$ (Year Fixed Effects)	Dummy for each year t.	Controls for common temporal shocks (e.g., national economic cycles, oil price fluctuations, the COVID-19 pandemic) affecting all cantons equally.	Not applicable (Control).
$e_{it}$ (Error Term)	Idiosyncratic error term.	Assumed to be i.i.d. Spatial autocorrelation in poverty outcomes is a potential concern that may require spatial econometric techniques for robustness.	Not applicable (Statistics).

.

In both specifications, the econometric strategy allows the isolation of the specific effect of infrastructure investment from other territorial characteristics. By using a panel structure with fixed effects, the models control for unobserved heterogeneity—namely, endogenous and place-specific factors that remain constant over time.

2.3. Index Construction

To classify cantons according to their initial level of socioeconomic development in 2008, a composite multidimensional index was constructed with a theoretical range from 0 to 400 points. The index combines four components: (1) the percentage of adequate employment, (2) the multidimensional poverty rate (inverted to ensure directional consistency, such that lower poverty corresponds to higher initial development), (3) the schooling rate, and (4) the proportion of the population with unmet basic needs. After normalizing each component using the min–max method, the four variables were summed to obtain an overall score for each canton. The distribution of these scores was then used to divide the sample into two groups: 111 cantons with high initial development and 110 with low initial development, using the median value (177.92 points) as the cutoff. This index-based classification was adopted because a simple rural–urban distinction is not feasible at the cantonal level, given that each of the 221 cantons comprises both urban areas, organized into neighborhoods and parishes, and surrounding rural territories.

Finally, for this index, data quality constraints must be acknowledged, given by the initial absence of national household survey (Enemdu) sampling in 36 cantons for the 2008 baseline. Although this was resolved by imputing data from the nearest available year—enabling full territorial coverage for all 221 cantons—the procedure may introduce measurement noise in the baseline year, a factor to consider when interpreting disaggregated results.

2.4. Econometric Analysis

The empirical analysis relies on the estimation of two fixed-effects models designed to isolate the causal impact of infrastructure spending. Such causal interpretations are often subject to debate, as econometric models, by their inherently “flat” structure, cannot fully capture the complex and multidimensional nature of development processes. For this reason, theoretical contrast plays a central role in interpreting the estimated causal relationships. The discussion focuses on the estimated β coefficients and their associated p-values. The coefficients indicate the magnitude and direction of the relationship between independent and dependent variables, holding other factors constant, while p-values assess statistical significance. Interpretation of the interaction coefficient follows directly from the model specification and the nature of the variables involved.

In addition, clustered robust standard errors are employed in both fixed-effects models, as this approach is methodologically preferable when working with panel data characterized by territorial heterogeneity and intra-group temporal correlation. Conventional standard errors would underestimate true variance by incorrectly assuming homoscedasticity and independence across observations. The use of clustered robust errors captures this structural complexity more adequately and yields more conservative and reliable inferences for public policy design. This is illustrated by cases in which public infrastructure spending shifts from being statistically significant at the 5% level to marginally significant at the 10% level, ensuring that identified relationships reflect robust patterns rather than statistical artifacts driven by restrictive parametric assumptions.

2.5. Theoretical and Comparative Assessment of Territorial Disparities

The results are interpreted against the limited empirical support for economic convergence theory, which would predict stronger effects of public spending or infrastructure investment in initially less developed cantons. The findings are contrasted with evidence from comparable contexts in Latin America, such as Mexico, where studies testing the neoclassical convergence hypothesis have reported similar outcomes (Carrillo, 2001). In a comparable vein, the simple injection of physical capital into infrastructure has proven insufficient to generate convergence in neighboring countries. The absence of a statistically significant, albeit differentiated, effect of infrastructure investment on recent poverty outcomes suggests the presence of regional poverty traps, as documented in Colombia (Galvis & Meisel, 2012).

These inconsistencies between investment inputs and development outcomes align more closely with the predictions of endogenous growth theory, which conceptualizes development as a systemic process shaped by institutional contexts and emphasizes the importance of local dimensions and innovative relational dynamics (Brunet Icart & Baltar, 2010). Empirical studies further indicate that the limited impact of labor productivity and human capital on poverty reduction in Ecuador may stem from the absence of critical complements, including competitive human capital, strong institutional capacity, and diversified productive linkages in lagging or low-development cantons (Jiménez & Alvarado, 2018). Building on these determinants, the study advances policy-oriented recommendations aimed at fostering inclusive territorial development.

2.6. Cantonal Scale

While much of the development literature emphasizes regional (provincial) scales, this study adopts the cantonal level for several empirically grounded reasons.

Ecuador is administratively organized into three main territorial levels: provinces (first level), cantons (second level), and parishes (third level). Cantons function as the primary intermediate administrative unit and represent the most detailed territorial scale for which consistent fiscal and socioeconomic data are systematically available nationwide. Additionally, due to their political, administrative, and financial autonomy, cantons are legally mandated to design and implement independent territorial planning strategies.

It should be noted that the historical formation of these units has not followed uniform demographic or territorial criteria. Consequently, cantons exhibit significant disparities in population size, geographic extent, and economic structure. Each canton further comprises a mix of urban and rural parishes, increasing internal heterogeneity. This diversity renders the canton an analytically robust unit for examining territorial inequalities and the dynamics of conditional convergence (Código Orgánico de Organización Territorial, Autonomía y Descentralización [COOTAD], 2010).

Recent empirical work underscores the analytical power of the cantonal scale, demonstrating its ability to reveal the spatial determinants of income inequality and to map the highly uneven distribution of social vulnerability across Ecuador. The use of a can-tonal panel for this analysis expands the macro-regional narratives to a systematic, micro-diagnosis of how public investment is connected to local conditions and, therefore, generates more specific evidence for place-based policymaking.

3. Results

Table 3. Descriptive statistics.

Variable		Mean	Std. Dev.	Min	Max	Observations
${GPInf}_{it}$	overall	8.276036	9.106056	0	2544.559	N = 2896
	between		57.3555	0.0093747	840.1961	n = 221
	within		67.50636	−826.2486	1712.639	Tbar = 13.1041
${TPM}_{it}$	overall	54.70442	25.40285	0	100	N = 2671
	between		19.40979	5.175444	98.44248	n = 221
	within		16.64797	−17.2291	117.41	Tbar = 12.086
${Ea}_{it}$	overall	30.02815	14.3788	0	100	N = 2722
	between		11.51512	5.877547	67.57996	n = 221
	within		8.832751	−12.15427	77.51021	Tbar = 12.3167
$BajoDesarrolloi$	overall	0.5973757	0.490511	0	1	N = 2896
	between		0.4886621	0	1	n = 221
	within		0	0.5973757	0.5973757	Tbar = 13.1041

presents descriptive statistics for the primary variables. The panel includes 221 cantons observed over an average period of approximately 13 years, resulting in between 2671 and 2896 observations per variable. Infrastructure expenditure demonstrates substantial dispersion, with significant variation both within and between cantons, indicating considerable territorial and temporal heterogeneity. Multidimensional poverty has an average value of 54.7 and ranges from 0 to 100, highlighting pronounced disparities among cantons. Adequate employment has a mean of 30.0 and also displays notable cross-sectional and temporal variation. The indicator BajoDesarrollo remains constant over time, with approximately 60% of cantons classified as initially low development. The observed variability across and within cantons justifies the application of a fixed-effects panel specification.

Model 1: Effect of executed infrastructure spending on adequate employment in Ecuadorian cantons, controlling for multidimensional poverty (2008–2022).

As shown in

Table 4. β coefficients and p-values for Model 1.

Variable	Coefficient	Standard Error	t-Value	p-Value	[95% Interval]
GPInf_millons	0.0039	0.0020	1.95	0.051	[−0.0000, 0.0077]
TPM	−0.2781	0.0091	−30.53	0.000 ***	[−0.2959, −0.2602]
Year (Reference: 2008)
2009	1.2339	0.7400	1.67	0.096	[−0.2172, 2.6849]
2010	0.4493	0.7519	0.60	0.550	[−1.0252, 1.9237]
2011	1.2176	0.8791	1.38	0.166	[−0.5063, 2.9415]
2012	1.1471	0.7866	1.46	0.145	[−0.3954, 2.6895]
2013	0.2606	0.7474	0.35	0.727	[−1.2051, 1.7262]
2014	−0.2996	0.7304	−0.41	0.682	[−1.7319, 1.1326]
2015	−2.3346	0.7282	−3.21	0.001 **	[−3.7626, −0.9066]
2016	−6.7342	0.7515	−8.96	0.000 ***	[−8.2079, −5.2605]
2017	−5.9581	0.7498	−7.95	0.000 ***	[−7.4285, −4.4877]
2018	−5.2366	0.7418	−7.06	0.000 ***	[−6.6911, −3.7821]
2019	−8.2546	0.7598	−10.86	0.000 ***	[−9.7446, −6.7646]
2020	−17.1701	0.7905	−21.72	0.000 ***	[−18.7202, −15.6200]
2021	−11.5787	0.7981	−14.51	0.000 ***	[−13.1437, −10.0137]
2022	−10.6627	0.7812	−13.65	0.000 ***	[−12.1946, −9.1309]
Constant	49.6025	0.7820	63.43	0.000 ***	[48.0691, 51.1360]

Note. Asterisks indicate statistical significance levels: *** p < 0.01, ** p < 0.05.

, the longitudinal panel model yields statistically meaningful relationships between the explanatory variables and adequate employment. Executed public infrastructure spending (GPInf_millons) is associated with a positive, borderline significant effect (β = 0.0039; p = 0.051). Substantively, this estimate implies that an additional USD 1 million in cantonal infrastructure spending is linked to an increase of 0.0039 percentage points in adequate employment over the period studied—an economically modest effect that sits close to conventional significance thresholds, yet still consistent with a small positive association between infrastructure investment and job quality. By contrast, the multidimensional poverty rate (TPM) displays a large and highly robust negative relationship (β = −0.2781; p < 0.001), indicating that a one-percentage-point increase in multidimensional poverty is associated with an approximately 0.28 percentage-point decrease in adequate employment, making it the strongest and most precisely estimated predictor in the model.

Regarding temporal dynamics, Table 4 suggests an overall deterioration in adequate employment over time. While the 2008–2014 period shows small and statistically insignificant year effects—generally on the positive side—coefficients become negative and statistically significant from 2015 onward, with a sustained decline that reaches its sharpest drop in 2020, coinciding with the COVID-19 pandemic. In that year, adequate employment is estimated to be 17.17 percentage points lower than in the base year (2008). Although a partial recovery appears in 2021–2022, the estimates do not return to pre-2015 levels.

As reported in

Table 5. Model 1 statistics: coefficients of determination.

Parameter	Value	Description
Within R-squared	0.4024	40.24% of within-canton variation explained
Between R-squared	0.5712	57.12% of between-canton variation explained
Overall R-squared	0.5037	50.37% of total variation explained
F-statistic	F(16.2434) = 102.43

, model fit is supported by a within R² of 0.4024, indicating that the specification explains 40.24% of the temporal variation within cantons, alongside an F-test that rejects the null hypothesis of jointly zero coefficients. The between-canton component also accounts for a substantial share of variance (57.12%), reinforcing the overall strength of the first model’s estimates.

Model 2: Effect of infrastructure spending on multidimensional poverty in low-initial-development cantons in Ecuador (2008–2022).

Table 6. β coefficients, p-values, and key statistics for Model 2.

Parameter	Value	Relevance to Hypothesis
${\beta }_2$: Interaction BajoDesarrollo X ln_Gasto	−0.9542	Direct parameter of interest
Interaction p-value	0.065	Marginally significant at the 10% level
95% confidence interval	[−1.9686, 0.0601]	Includes zero
Spending effect in developed cantons	0.6466	Reference baseline effect
Spending effect in underdeveloped cantons	−0.3076	Computed net effect
R2 within	0.1830	Low explanatory power

reports the interaction between cantons’ initial development level and public infrastructure spending over the study period, which is central for assessing whether investment effects differ systematically between initially developed and lagging territories. The interaction term—the parameter of primary interest—takes a value of −0.9542, which, from a theoretical standpoint, would suggest a substantially larger effect of infrastructure investment on the outcome in lagging cantons. Such a pattern would be broadly consistent with convergence arguments and with the notion that investment benefits diffuse over time across territories. However, the interaction term is marginally significant at the 10% level, a threshold commonly accepted in applied regional and policy-oriented research using administrative panel data.

Although the overall results do not provide a significant differential impact of total public spending on multidimensional poverty by initial development level, the decomposition of effects is informative. For developed cantons, the baseline coefficient is positive and sizeable (0.6466), whereas the computed net effect for underdeveloped cantons is negative (−0.3076). While this configuration could be read as suggestive heterogeneity, the lack of statistical precision precludes strong causal claims. In sum, the data do not offer firm evidence that the effect of infrastructure spending differs significantly by cantons’ initial development status, even if point estimates hint at potentially meaningful heterogeneity. This pattern warrants further examination using stronger identification strategies, additional controls—such as economic variables like GDP—and alternative specifications. To finalize, the model’s low within explanatory power (R² = 0.1830) underscores that most of the temporal variation in multidimensional poverty is likely driven by factors not captured in the specification, rather than by infrastructure spending alone.

This section also analyzes the efficiency of budget execution during the study period, as shown in

Table 7. Key data description: Planned vs. executed infrastructure spending for Ecuador’s cantons (2008–2022).

Year	Planed	Executed	Difference	Percentage of Non-Execution
2008	240,527,599.18	158,234,451.03	82,293,148.15	34.21%
2009	352,522,376.04	237,406,154.58	115,116,221.46	32.66%
2010	497,605,275.78	295,828,531.81	201,776,743.97	40.55%
2011	297,883,721.34	250,409,743.84	47,473,977.50	15.94%
2012	719,451,042.27	444,168,140.75	275,282,901.52	38.26%
2013	912,456,886.73	650,915,690.98	261,541,195.75	28.66%
2014	5,454,634,749.24	4,713,516,829.95	741,117,919.29	13.59%
2015	5,424,144,145.56	4,165,088,823.62	1,259,055,321.94	23.21%
2016	4,778,328,253.76	3,845,297,770.17	933,030,483.59	19.53%
2017	3,783,820,110.42	2,944,482,406.33	839,337,704.09	22.18%
2018	1,795,089,507.34	1,332,895,234.30	462,194,273.04	25.75%
2019	1,741,693,398.91	1,037,361,299.53	704,332,099.38	40.44%
2020	1,799,866,763.07	1,043,994,811.30	755,871,951.77	42.00%
2021	2,212,467,464.20	1,978,378,078.85	234,089,385.35	10.58%
2022	1,200,402,626.93	869,422,544.95	330,980,081.98	27.57%
Total	31,210,893,920.77	23,967,400,511.99	7,243,493,408.78	23.21%

. The data tell a clear story of ambition versus reality in public infrastructure investment. The cycle began in 2008 with a significant execution gap of 34.2 percent. Spending then surged dramatically, peaking between 2014 and 2016. The absolute highest budgets, both planned and executed, occurred in 2014 at USD 5.45 billion and USD 4.71 billion, respectively. However, the largest financial shortfall was recorded the following year. In 2015, a USD 1.26 billion difference opened between planned and executed budgets. In terms of relative efficiency, the most problematic year was 2020, which saw the widest percentage gap at 42.0 percent. This indicates that less than 60 percent of the planned infrastructure budget was actually spent, largely due to the pandemic. Conversely, 2021 was the most efficient year in the series, with the smallest percentage gap of just 10.6 percent. By 2022, the budget volume had fallen considerably from its mid-decade peak to a planned USD 1.20 billion. The execution gap widened again to 27.6 percent, representing a USD 331 million shortfall for that year.

These results provide an indication of spending inefficiency, but they do not serve as a strong indicator of heterogeneity in the preceding econometric model. The difference between planned and executed infrastructure budgets in low-development cantons versus high-development cantons is 22.95% and 24.99%, respectively. This narrow gap verifies that there is no significant difference in execution efficiency between the two groups, leaving other unobserved variables in the model as pending factors for explanation, but it supports the conclusion that average efficiency levels in Ecuador and in LATAM are relatively low, pointing to structural governance and institutional constraints rather than budgetary limitations.

4. Discussion

A territorial diagnostic of public infrastructure spending in Ecuador between 2008 and 2022 reveals a nuanced and limited relationship between investment and key socioeconomic outcomes. Although infrastructure spending demonstrates a small positive association with adequate employment, its differentiated impact on multidimensional poverty across cantons remains statistically weak. These results challenge the assumption that public investment necessarily leads to territorial convergence.

Adequate employment serves as a key indicator of territorial development. Within the framework of endogenous growth theory, infrastructure is expected to enhance productivity and support job creation. Nevertheless, the relatively modest estimated effect indicates that infrastructure investment alone does not lead to sustained improvements in labor-market quality. Ongoing structural disparities, especially between rural and urban regions, demonstrate that capital accumulation without corresponding institutional and productive capacity produces limited benefits. Recent data illustrate a stark divide, with national informal employment at 54% but rising to 77.9% in rural zones, while rural poverty reached 46.4% compared to 18% in urban areas (Cobena & Palacios, 2024). These patterns are difficult to reconcile with territorial equity and resonate with classic center-periphery concerns (Prebich, 1950).

The study examines multidimensional poverty—a key human-development indicator capturing deprivations that restrict life opportunities. Reducing such poverty is a primary objective of social infrastructure spending through schools, hospitals, and universities, which constitutes an investment in human capital and social cohesion, not mere consumption. Mapping these deprivations at the cantonal level helps identify exclusion hotspots, enabling more efficient resource allocation to close development gaps. The existing literature emphasizes that unemployment is both a cause of poverty and a symptom of structural labor-market weakness (Cobena & Palacios, 2024). While public investment plausibly channels into poverty reduction, the present evidence does not permit a definitive claim that infrastructure spending, on its own, reduces inequality.

The limited evidence of convergence among cantons with low initial development levels highlights the conditional nature of infrastructure’s territorial impact. While the interaction effect does not meet the conventional 5% threshold, its marginal significance at the 10% level and theoretically consistent sign suggest the presence of differentiated territorial dynamics that warrant policy attention. This pattern aligns with endogenous growth and new economic geography perspectives, which emphasize path dependence and initial advantages. Territories with greater institutional capacity and diversified economic structures appear better equipped to translate infrastructure into sustained development gains.

Additionally, execution inefficiencies in public spending, which are observed in both developed and lagging cantons, indicate systemic governance constraints rather than solely fiscal limitations. The ongoing disparity between planned and actual expenditures suggests that the quality of public investment, rather than its total volume, is a critical determinant of outcomes. This observation aligns with broader trends in Latin America, where institutional effectiveness mediates the conversion of capital investment into equitable growth. The quality of public spending is defined not only by the volume of resources but by their effectiveness, efficiency, and alignment with development objectives, all of which are conditioned by institutional capacity (Economic Commission for Latin America and the Caribbean [ECLAC], 2014). Significant inefficiencies persist, suggesting comparable outcomes could be achieved with lower expenditure under more efficient institutional arrangements (Herrera et al., 2025). Countries characterized by a robust rule of law and stringent anti-corruption measures exhibit higher spending quality in critical sectors, particularly infrastructure, where institutional efficiency is essential (Herrera et al., 2025). Therefore, the relationship between public expenditure and development outcomes cannot be fully understood without considering institutional quality. Further empirical research is required to examine how governance frameworks influence the effectiveness of public investment.

These findings are in line with structuralist and dependency arguments that emphasize the need for changes in productive structures, rather than the mere increase of physical capital. Infrastructure investment, if not integrated into comprehensive processes of productive diversification and institutional development, can reinforce existing territorial inequalities, especially when associated with extractive or centralized modes of development. In line with this, key recommendations include fostering innovative strategies that align public and private interests. Studies confirm that the lack of a cohesive approach to resource management and private participation has complicated infrastructure development (Pelaez, 2021; Castillo & Simón, 2021).

From a policy perspective, the evidence indicates that infrastructure spending needs to be part of integrated development strategies at the place-specific level. Without a transformative industrial and technological policy, large-scale infrastructure can merely reinforce extractivist models by improving the efficiency of commodity export corridors without fostering broad-based, integrated territorial development (Cypher, 2023; Svampa, 2015). In lagging cantons, it should be complemented by additional measures (local enterprise support, human capital development, governance strengthening) to boost absorptive capacity. In more developed regions, it should be used more as a lever for competitiveness and cluster formation. In the end, convergence will not only be on the scale of investment but also on its institutional quality, territorial targeting and relevance to local productive structures.

5. Conclusions

Therefore, interpreting the econometric evidence—particularly the weak or heterogeneous effects of investment on adequate employment and multidimensional poverty in less developed cantons—benefits from this theoretical lens. It suggests that overcoming territorial inequality requires more than capital expenditure; it demands a structural change in the productive matrix. The cantonal scale of this study is particularly apt for this inquiry, as it reveals whether infrastructure helps territories break from a dependent, primary-goods specialization or consolidates it, offering a sub-national test of these enduring theoretical propositions. In conclusion, the descriptive results show that execution efficiency does not vary by development level. The coefficient of primary interest, associated with the interaction between low initial development and infrastructure expenditure, takes a value of −0.9542, which theoretically indicates a substantially larger impact on outcome variables in lagging cantons. This pattern is consistent with convergence theory.

The analysis underscores that adequate employment is a central marker of labor-market health and a core development objective. Employment represents progress at both individual and territorial levels. From an endogenous growth perspective, it emerges as an economy expands through knowledge and productivity. Therefore, a key purpose of public spending is to create the conditions that generate and sustain quality jobs. Higher employment can strengthen local fiscal capacity through increased revenues, potentially creating a virtuous cycle of further investment. Conversely, territorial disparities in employment both reflect and deepen inequality. Cantons with higher employment rates tend to exhibit greater well-being and investment capacity, while those with lower rates face constrained development prospects.

Second, multidimensional poverty—closely linked to unmet basic needs—is a key human-development indicator. These poverty indicators provide a direct window into territorial inequality. Mapping them at the cantonal level helps identify exclusion hotspots, enabling more efficient resource allocation and targeted policies to close development gaps.

By combining econometric evidence on these dual outcomes with infrastructure data, this study interrogates the tension between convergence narratives and persistent regional disparities. The fixed-effects framework, which tests for different effects based on a canton’s initial development level, provides a robust territorial diagnostic for planning. It can inform differentiated policies: catch-up strategies for lagging cantons and competitiveness policies for more advanced ones. This approach, which can be refined with additional data like cantonal GDP, illustrates how integrating data and academic analysis can build a coherent narrative to guide public policy. Theoretically, the findings relate to path dependence and initial advantage. Territories with a stronger starting base typically have greater capacity to attract further investment and human capital, leading to cumulative growth. This process highlights a potential disconnect: although infrastructure investments are designed to narrow disparities, they may overlook distinct regional dynamics, leading to uneven outcomes and limited effectiveness in reducing entrenched inequality.

For policymakers in Ecuador and across Latin America, these findings underscore a central lesson: infrastructure spending alone is insufficient to guarantee territorial convergence. While public investment remains necessary, its developmental impact depends critically on institutional quality, effective targeting, and integration with broader productive strategies.

The evidence suggests that future infrastructure policy should move beyond aggregate expenditure metrics and incorporate systematic evaluation mechanisms capable of assessing multidimensional outcomes, including employment quality and poverty reduction. Strengthening data-driven decision-making processes and improving execution efficiency are essential to ensure that public resources translate into measurable territorial gains.

Moreover, policy design should adopt a differentiated, place-based approach. In lagging cantons, infrastructure investment should be embedded within broader catch-up strategies that enhance local productive capacity and absorptive potential. In more advanced territories, investment may focus on competitiveness and cluster development. Such differentiation acknowledges that convergence is conditional, not automatic.

Ultimately, the results highlight that the effectiveness of public investment depends not only on financial magnitude but on governance capacity, institutional coordination, and alignment with local economic structures. Without these complementary conditions, infrastructure risks reinforcing existing territorial disparities rather than reducing them.

Furthermore, strategic institutional reforms are essential. These should include establishing independent bodies to monitor spending efficiency and improving public procurement processes. Such measures are crucial for optimizing resource allocation, fostering better public service delivery, and ultimately promoting equitable growth through infrastructure investment.