Digital Infrastructure and the Limits of Smart Urbanism: Evidence from a Panel Analysis and the Case of Wang Chan Valley

Highlights

What are the main findings?

• Digital infrastructure investment is strongly and positively associated with smart city performance in emerging economies.
• The impact of digital infrastructure is piecewise-linear, showing high return at a threshold, and is most effective when embedded in a collaborative Quadruple Helix governance model.

What is the implication of the main finding?

• Smart city policy should prioritize creating participatory innovation ecosystems over focusing on technology investment alone.
• Sustainable smart urbanism requires an integrated approach that combines digital infrastructure with adaptive governance, institutional learning, and inclusive stakeholder engagement.

Abstract

This study investigates how digital infrastructure contributes to smart city performance in emerging economic contexts and whether its impact is shaped by governance models. We estimate the effect of a Digital Technology Index on a composite Smart City Index, employing a generalized least squares (GLS) random-effects model to address heteroskedasticity and serial correlation. The analysis reveals a robust and statistically significant relationship: a one-standard-deviation increase in digital infrastructure corresponds to a 0.7-standard-deviation rise in smart city performance. The relationship is piecewise-linear, stagnating in the early stage before rising sharply after a threshold. To interpret these results, we draw on a qualitative case study of Wang Chan Valley (WCV), a science and innovation hub in Thailand’s Eastern Economic Corridor. WCV exemplifies how early-stage digital investment can amplify smart development outcomes and generate spillover effects across the broader urban region. The case reinforces the hypothesis that digital infrastructure embedded within participatory innovation ecosystems yields greater and more sustainable smart-city gains than technology investment alone. Taken together, the findings contribute to the understanding of how governance mediates the effectiveness of digital infrastructure in driving smart urban transformation within emerging economies.

1. Introduction

Over the past decade, smart city strategies have become central to urban development agendas, particularly in regions seeking to balance sustainability, livability, and innovation [1,2]. As cities embrace digital transformation, investments in digital infrastructure−such as broadband, sensor networks, and data platforms−are often assumed to directly enhance urban performance. However, growing evidence shows that technology alone is not sufficient; its effectiveness depends on the institutional, economic, and societal systems within which it is embedded [3,4]. The debate has therefore shifted from how much digital infrastructure is required to under what governance conditions digitalization generates lasting and inclusive benefits.

This study contributes to that debate by examining the relationship between digital infrastructure and smart city performance in the context of emerging economies. Using a balanced panel dataset of 12 cities observed over 11 years, we estimate the average effect of digital infrastructure on smart city outcomes and explore whether marginal returns to digitalization follow a piecewise-linear pattern. Our results suggest that early-stage investments yield limited gains, while sustained digitalization leads to substantial improvements. This pattern reflects broader evidence of threshold effects and non-linear dynamics in innovation and urban transformation processes [5,6,7].

Beyond estimating average effects, the paper introduces a governance-sensitive perspective rooted in the Quadruple Helix (QH) model, which emphasizes co-evolutionary innovation through collaboration among government, industry, academia, and civil society [8,9]. When applied to urban transformation, this perspective suggests that digital infrastructure delivers the greatest returns when embedded within participatory, cross-sectoral governance ecosystems rather than through purely top-down or market-driven approaches.

The Wang Chan Valley (WCV) case exemplifies how early-stage digital investment can amplify smart development outcomes and generate spillover effects across the broader urban region, reflecting similar dynamics found in prior studies of smart city diffusion and innovation ecosystems [1,3]. Located within the Eastern Economic Corridor (EEC), WCV is a national innovation hub integrating advanced ICT infrastructure with educational institutions, research facilities, and community co-creation platforms. Although WCV is not part of the econometric sample due to its classification as an innovation district rather than a municipality, it provides a compelling example of how digital investment, when coupled with QH-oriented governance, can catalyze transformation beyond project boundaries. As such, WCV serves as an interpretive case that helps contextualize the quantitative findings and links them to broader debates on innovation ecosystems and place-based policy [10,11].

By integrating econometric analysis with a governance-aware interpretation, this paper moves beyond linear and technology-determinist accounts of smart city development. It argues that understanding when and how digital infrastructure translates into urban transformation requires attention to institutional design, governance capacity, and civic engagement. The paper proceeds as follows: Section 2 details the methodology and data sources; Section 3 presents the quantitative results; Section 4 discusses the WCV case as an illustrative complement; and Section 5 concludes with theoretical and policy implications for smart city governance in the Global South.

2. Literature Review

2.1. Digital Infrastructure and Smart City Performance

Smart city development is widely conceptualized as the integration of physical and digital infrastructure to enhance urban services, governance, and quality of life [1,2]. A significant strand of the literature has examined how investments in information and communication technology (ICT), such as broadband networks, digital platforms, and sensor grids, support key domains including energy, mobility, environment, and governance [3,9]. Digital systems are argued to improve responsiveness, efficiency, and inclusiveness in urban management.

Empirical research in advanced economies provides strong evidence of these linkages. For example, ref. [6] finds that broadband networks and digital public services are positively associated with competitiveness and citizen satisfaction in European cities. Similarly, ref. [4], employing a composite Smart City Impact Index, shows that higher ICT deployment levels correlate with stronger innovation ecosystems and improved governance outcomes.

By contrast, rigorous evidence from emerging economies remains limited, where digital infrastructure is often less mature and institutional capacity more uneven. In China, for instance, ref. [10] demonstrates that the returns to digitalization vary sharply depending on local governance readiness. This gap motivates the present study’s focus on developing contexts, where investments in digital infrastructure may have heterogeneous outcomes.

2.2. Non-Linearity and Diminishing Returns in Smart City Development

Recent scholarship challenges the assumption of a linear, ever-increasing benefit from digital investment. Instead, many studies highlight non-linear or threshold dynamics, where early ICT deployment delivers substantial gains, but marginal benefits diminish once foundational infrastructure is in place [8,11]. This pattern echoes broader economic theories of diminishing returns and inverted-U relationships in innovation studies [12,13]. Econometric approaches such as threshold regression and panel smooth transition regression (PSTR) have been developed to capture these dynamics [14,15].

In urban terms, after cities establish core ICT systems (fiber networks, open-data portals, smart grids), additional digitalization yields benefits only if accompanied by strong governance and service co-production capacities. Without such institutional support, high-investment innovation districts risk technological redundancy or saturation effects. The literature therefore suggests that robust methods beyond simple quadratic specifications are necessary to test non-linearity in digital-performance relationships, particularly in emerging contexts with limited high-maturity observations.

2.3. The Quadruple Helix Model and Innovation Ecosystems

To explain why digital infrastructure translates into performance in some places but not others, scholars increasingly draw on innovation systems theory, particularly the Quadruple Helix (QH) model. Expanding the Triple Helix of government–industry–academia, the QH includes civil society as a co-creator of knowledge and innovation [6,7].

In the smart city literature, the QH framework is used to highlight how participatory innovation ecosystems enable digital investments to produce sustainable and inclusive outcomes [3]. Cities with stronger institutional coordination and civic engagement consistently outperform those that adopt technology without supportive governance. Recent critiques of “smartness without democracy” reinforce this point: smart city projects often underperform where community inclusion, regulatory readiness, or human capital alignment are lacking [16].

This governance-oriented lens is especially relevant for emerging economies, where fiscal, institutional, and human capacity constraints often limit digital infrastructure effectiveness. By embedding technology deployment within cross-sectoral governance structures, the QH framework provides both an analytical model and a practical roadmap for sustainable transformation.

2.4. Positioning This Study

While prior research has examined the link between digitalization and smart city performance, non-linear effects, and the importance of governance, few studies integrate these strands empirically in emerging economic contexts. This study contributes in three ways:

1.

It estimates the marginal effects of digital infrastructure on smart city performance using an 11-year, 12-city panel dataset. Although modest in scale, this dataset provides rare longitudinal evidence from secondary cities.

2.

It explores potential non-linearity in the relationship, acknowledging the limitations of quadratic models and calling for future robustness checks with threshold or non-parametric approaches.

3.

It interprets findings through the case of Wang Chan Valley (WCV), a flagship innovation hub in Thailand, to illustrate how Quadruple Helix governance can amplify and sustain digital investments.

By integrating econometric evidence with governance-sensitive analysis, this paper responds to recent calls to link technology metrics with institutional dynamics in assessing smart city outcomes [4,9]. It advances the hypothesis that digital infrastructure, when guided by QH principles, generates positive spillovers that extend beyond individual projects, shaping broader urban ecosystems in emerging contexts.

3. Method

3.1. Data Collection

This study utilizes a provincial-level panel dataset for Thailand, shown in

Table 1. Shows sources and descriptions of data.

Variable/Index	Description	Source	Temporal Coverage
Index Score	A composite index measuring the level of digital technology adoption and development. It is composed of several sub-indices that capture different dimensions of the digital economy.	Digital Economy Promotion Agency (DEPA), Thailand	2014–2024
Gross Provincial Product (GPP) per capita	The total value of goods and services produced within a province, divided by its population. This is a measure of the average economic output per person.	Office of the National Economic and Social Development Council (NESDC), Thailand	2014–2024
Gini Coefficient	A measure of income inequality within each province, ranging from 0 (perfect equality) to 1 (perfect inequality).	Office of the National Economic and Social Development Council (NESDC), Thailand	1988–2024
Government R&D Expenditure	The total amount of money spent by the provincial government on research and development activities.	Office of the National Economic and Social Development Council (NESDC)	2014–2024

, covering the period from 2018 to 2025. The dataset was constructed by compiling information from various official sources to analyze the relationship between digitalization and smart city performance. The dataset for this study was constructed based on a census approach, encompassing all provinces in Thailand for which consistent and reliable data were available for the period of 2014 to 2024.

The primary dependent variable, the Smart City Index, is sourced from the Digital Economy Promotion Agency (DEPA) [17]. This composite index measures provincial performance across several key dimensions of urban development, including economy, governance, environment, energy, living, mobility, and people [17]. The main independent variables include Gross Provincial Product (GPP) per capita and the Gini Coefficient, which were obtained from the Office of the National Economic and Social Development Council (NESDC) to control economic development and income inequality, respectively [18]. Government R&D expenditure data were sourced from internal budget proposals submitted by state-owned enterprises to the NESDC [19].

3.2. GLS Methodology

This study employs a balanced panel dataset of 12 cities observed over 11 years provided by Digital Economy Promotion Agency [17] to examine the relationship between digital infrastructure and smart city performance in emerging economic contexts. The analysis focuses on two standardized composite indicators: the Smart City Index (ranging from 0 to 5) and the Digital Technology Index (ranging from 0 to 1). The Smart City Index captures multidimensional urban performance across key domains, including economy, environment, governance, mobility, living, and people [1,2]. The Digital Technology Index reflects the degree of digital infrastructure adoption within each city, including broadband coverage, ICT service availability, and digital platform implementation [4]. For clarity, the underlying concept measured by this index will be referred to as digital infrastructure throughout the paper. The baseline model specification is a linear panel data model:

$$SC{I}_{it}={\beta }_0+{{\mathbf{x}}^{\prime }}_{it}\beta +v_{it}$$

(1)

where $SC{I}_{it}$ is the Smart City Index for city i at time t, ${{\mathbf{x}}^{\prime }}_{it}$ is a vector of explanatory variables, $\beta$ is the vector of coefficients to be estimated, and $v_{it}$ is the composite error term.

A key advantage of panel data is the ability to model unobserved city-specific heterogeneity. We decompose the error term into two components:

$$v_{it}=u_i+{ϵ}_{it}$$

(2)

Here, $u_i$ represents the unobserved, time-invariant specific effect for each city (e.g., stable governance quality, geographical location), and ${ϵ}_{it}$ is the idiosyncratic error term that varies across both city and time.

All variables were normalized to a 0–1 scale to ensure comparability across cities and over time. In addition to the main explanatory variable, the model incorporates several controls to address potential confounding factors: (1) economic prosperity, proxied by gross product per capita (GPP per capita), (2) income inequality, measured by the Gini coefficient, and (3) public investment in research and development (R&D). Due to strong multicollinearity between GPP per capita and the Digital Technology Index (correlation coefficient ≈ 0.97), GPP was excluded from models 2 and 3 to preserve estimator stability and interpretability [14,15].

To estimate the effect of digital infrastructure on smart city performance, we apply a generalized least squares (GLS) regression model suitable for panel data. The GLS estimator is given by:

$${\widehat{\beta }}_{GLS}={\left({\mathbf{X}}^{\prime }{\mathsf{\Omega }}^{-1}\mathbf{X}\right)}^{-1}{\mathbf{X}}^{\prime }{\mathsf{\Omega }}^{-1}\mathbf{y}$$

(3)

where $\mathsf{\Omega }$ is the variance-covariance matrix of the error term. The GLS procedure effectively transforms the data to eliminate the serial correlation in the errors, thereby producing estimates that are both consistent and efficient [20,21]. The GLS estimator is selected to correct for heteroskedasticity across cities and serial correlation within them—issues that may compromise the efficiency of ordinary least squares (OLS) estimates [22,23]. In practice, this corresponds to a random-effects model, under the assumption that city-specific unobserved effects are uncorrelated with the regressors.

The primary specification regresses the Smart City Index on the Digital Technology Index, controlling for inequality and government R&D, with robust standard errors to account for model uncertainty. In supplementary models, we explore piecewise-linear dynamics by estimating a segmented panel regression. By including the square term of Digital Index, diminishing return is explored. All estimations were conducted in a statistical computing environment, with standard diagnostic checks (residual analysis, goodness-of-fit tests) employed to validate model assumptions.

3.3. Random Effect Model

To select the most appropriate panel data specification, a robust, regression-based Hausman test, based on the formulation by Mundlak [24], was conducted. This approach was chosen over the standard Hausman test [25] due to its robustness to heteroskedasticity and its ability to handle the panel structure of our data correctly. The test involves augmenting a random-effects model with the time-means of all time-varying regressors and using a Wald test to assess their joint significance.

The results of this test, presented in

Table 2. Robust Hausman Test Results for Model Specification.

Dependent Variable	Chi-Squared	p-Value	Decision
Smart Cities	7.2833	0.2004	Random-Effects
Supporting Smart Cities	4.9206	0.4257	Random-Effects
Environment	8.8242	0.1163	Random-Effects
Energy	6.8403	0.2328	Random-Effects
Economy	12.5236	0.0283 *	Fixed-Effects Preferred
Governance	9.1074	0.1049	Random-Effects
Living	15.9456	0.0070 *	Fixed-Effects Preferred
Mobility	8.6905	0.1221	Random-Effects
People	11.4927	0.0424 *	Fixed-Effects Preferred

* Note: Significant at $p<0.05$.

, showed that the time-mean variables were not jointly significant across most model specifications ($p>0.05$). This outcome indicates that the unobserved, time-invariant provincial characteristics are not systematically correlated with the explanatory variables in the model. The exceptions were the Economy, Living, and People dimensions, which showed significant p-values, suggesting a Fixed-Effects model might be more suitable for them. However, for model consistency and to retain the explanatory power of time-invariant variables across all dimensions, the Random-Effects (RE) model was selected as the primary specification for this study. A key advantage of the RE model in this context is its ability to estimate the coefficients of important time-invariant variables, which would be omitted in a Fixed-Effects specification, thereby allowing for a more comprehensive analysis of the factors influencing smart city performance.

To address the high correlation between the digital index and GPP per capita ($r=0.968$) and the potential for endogeneity, we re-estimated the model using an instrumental variable (IV) approach [26]. Specifically, a Two-Stage Least Squares (2SLS) model was employed, using the one-year lag of the digital index as an instrument for its current value. This technique helps to isolate the causal impact of digital infrastructure on smart city performance by purging the estimate of endogeneity bias [21].

The results of the 2SLS estimation are presented in

Table 3. Two-Stage Least Squares (2SLS) estimation results.

Smart Cities
Variable	Coefficient	Std. Err.
Digital Index	3.2348 ***	0.9284
GPP per capita	0.0286	0.5867
GINI country	−0.1026	0.1673
Government R&D total	−0.4301	0.3929
const	0.1273	0.3702
Observations	120
R-squared	0.5089
First-Stage F-statistic	958
Clusters	12

Note: *** $p<0.001$. Clustered standard errors are in parentheses.

. The findings reveal a strong and highly significant positive effect of the digital index on the smart cities index ($\beta =3.23,p<0.001$). A crucial finding is that after instrumenting, the coefficient for GPP per capita becomes statistically insignificant. This provides strong evidence that it is the investment in digital infrastructure itself, rather than just the general economic wealth of a province, that is the key driver of smart city outcomes.

The validity of this approach is strongly supported by first-stage diagnostics. The partial F-statistic of 958.00 is exceptionally high and far exceeds the conventional threshold of 10, confirming that the lagged instrument is both relevant and strong [27].

3.4. Case Study Protocol

Complementing the panel analysis, this study incorporates a qualitative case study of Wang Chan Valley. The case was selected instrumentally to provide an in-depth examination of the Quadruple Helix model in practice. Data for the case study was drawn from semi-structured interviews with key stakeholders, as detailed in Section 4.2, and were analyzed thematically to identify the governance mechanisms linking digital investment to performance outcomes. This qualitative component is not intended for statistical generalization but to provide explanatory depth to the quantitative findings, a common and powerful strategy in mixed-methods research [28]. The integration of these two data types allows for a more complete picture than either method could provide alone, a key strength of mixed methods designs.

4. Results

4.1. Results from Models

Table 4. Descriptive statistics of the data.

Variables	Obs.	Mean	Standard Deviation	Min	Max
Smart city index	132	1.409	0.873	0.0	5.0
Supporting smart city index	132	4.0	3.057	1.0	15.0
Environment index	132	2.197	1.244	1.0	7.0
Energy index	132	1.0	1.091	0.0	5.0
Economy index	132	1.735	1.341	0.0	7.0
Governance index	132	1.78	1.333	0.0	7.0
Living index	132	2.485	1.037	1.0	7.0
Mobility index	132	0.818	1.247	0.0	5.0
People index	132	1.833	2.45	0.0	11.0
Digital technology index	132	0.449	0.207	0.052	0.937
GPP per capita	132	0.223	0.256	0.0	1.0
Annual GINI value	132	0.49	0.346	0.0	1.0
Government R&D	132	0.266	0.269	0.0	1.0

displays the descriptive statistics for 132 samples collected from 12 different cities over 11 years.

Table 5. Smart City Index and supporting Smart City Index correlations.

	1	2	3	4	5	6
1. Smart city index	1.0
2. Supporting smart city index	0.878	1.0
3. Digital technology index	0.702	0.744	1.0
4. GPP per capita	0.775	0.809	0.968	1.0
5. Annual GINI value	0.092	0.113	0.131	0.146	1.0
6. Government R&D	−0.105	−0.135	−0.185	−0.204	−0.07	1.0

shows the pairwise correlations among the Smart City Index, its Supporting Index, the Digital Technology Index, GPP per capita, Annual GINI, and Government R&D. The Smart City Index strongly correlated with the Supporting Index ($r=0.878$), confirming their conceptual alignment. Its correlation with the Digital Technology Index is also substantial ($r=0.702$), indicating that smarter cities tend to have more advanced digital infrastructures. The Digital Technology Index and GPP per capita exhibit an exceptionally high correlation ($r=0.968$), suggesting that economic growth and digital capacity advance in correlation. In contrast, Annual GINI and Government R&D show only weak associations with the Smart City Index ($r=0.092$ and $r=-0.105$, respectively), implying that neither income inequality nor public research spending bears a strong direct relationship to overall smart city performance.

A more detailed sub-index correlation matrix (

Table 6. Sub-index correlation matrix.

Smart Index Correlation Matrix	1	2	3	4	5	6	7	8	9	10	11
1. Environment index	1.0
2. Energy index	0.866	1.0
3. Economy index	0.855	0.876	1.0
4. Governance index	0.859	0.861	0.881	1.0
5. Living index	0.842	0.809	0.883	0.867	1.0
6. Mobility index	0.825	0.903	0.916	0.867	0.895	1.0
7. People index	0.885	0.899	0.892	0.933	0.873	0.887	1.0
8. Digital technology index	0.706	0.751	0.662	0.635	0.61	0.745	0.671	1.0
9. GPP per capita	0.745	0.778	0.715	0.672	0.682	0.802	0.705	0.968	1.0
10. Annual GINI value	0.091	0.133	0.029	0.095	0.04	0.138	0.067	0.131	0.146	1.0
11. Government R&D	−0.111	−0.141	−0.105	−0.121	−0.09	−0.1	−0.126	−0.185	−0.204	−0.07	1.0

) reveals that the seven domain sub-indices are themselves highly intercorrelated (pairwise $r\approx 0.809-0.933$). The Digital Technology Index maintains moderate correlations with each domain, ranging from 0.610 (Living) to 0.751 (Energy), showing that while digital capacity is integral to every smart city dimension, it does not fully subsume them. GPP per capita again correlates strongly with both digital capacity and each sub-index. Annual GINI and Government R&D remain only marginally related to the sub-indices ($\left|r\right|<0.14$), reinforcing the view that these factors play a limited cross-sectional role in shaping individual smart city domains.

The GLS estimation in

Table 7. GLS result on Smart City Index and supporting Smart City Index using different models.

	Smart City						Supporting Smart City
Variables	Model 1		Model 2		Model 3		Model 1		Model 2		Model 3
	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.
Digital technology index	2.948	0.504	2.648	0.475	3.606	0.48.	12.226	1.586	11.496	1.488	13.717	1.478
Digital technology index squared	4.681	1.388	4.315	1.399			10.322	4.368	9.276	4.382
GPP per capita	0.052	0.516					0.647	1.624
Annual GINI value	−0.104	0.166			−0.091	0.172	−0.355	0.523			−0.34	0.528
Government R&D	−0.496	0.213			−0.427	0.219	−1.436	0.669			−1.27	0.674
Model Metrics
R-Squared	0.381		0.349		0.325		0.454		0.431		0.429
Chi-Squared	77.58		69.181		61.69		104.573		97.582		96.08

reveals a positive and statistically significant relationship between the Digital Technology Index and the Smart City Index, supporting prior research that highlights digital infrastructure as crucial to urban innovation and effectiveness [8,13]. In practical terms, cities with higher levels of digital infrastructure implementation tend to score better on the smart city scale. The coefficient on the Digital Technology Index is positive and significant ($p<0.01$), confirming that digital infrastructure is a key driver of smart city performance. Substantively, the magnitude of this effect is considerable: a one-standard-deviation increase in the Digital Technology Index is associated with roughly a 0.7 standard deviation increase in the Smart City Index. This finding aligns with prior research suggesting that digital advancement propels smart urban development. It also corresponds to the strong bivariate correlation ($r\approx 0.70$) observed between the two indices in Table 5, indicating that nearly half of the variance in smart city outcomes can be explained by variation in digital infrastructure alone. Results of sub-index can be found in

Table A1. GLS result on environment and energy indices.

	Environment						Energy
Variables	Model 1		Model 2		Model 3		Model 1		Model 2		Model 3
	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.
Digital technology index	4.354	0.716	4.156	0.669	4.404	0.654	3.755	0.635	3.726	0.595	3.744	0.58
Digital technology index squared	0.13	1.973	−0.264	1.969			−0.358	1.749	−0.702	1.753
GPP per capita	0.415	0.734					0.492	0.651
Annual GINI value	−0.192	0.236			−0.203	0.234	−0.31	0.209			−0.325	0.207
Government R&D	−0.568	0.302			−0.555	0.298	−0.568	0.268			−0.56	0.265
R-Squared	0.289		0.267		0.287		0.291		0.263		0.288
Chi-Squared	51.226		47.057		51.58		51.757		46.074		51.695

,

Table A2. GLS result on economy and governance indices.

	Economy						Governance
Variables	Model 1		Model 2		Model 3		Model 1		Model 2		Model 3
	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.
Digital technology index	4.759	0.875	4.422	0.82	4.833	0.798	4.661	0.932	4.557	0.866	4.368	0.853
Digital technology index squared	0.54	2.411	0.005	2.415			−2.299	2.567	−2.686	2.548
GPP per capita	−0.012	0.897					0.358	0.955
Annual GINI value	−0.291	0.289			−0.289	0.285	−0.304	0.307			−0.322	0.305
Government R&D	−0.799	0.37			−0.791	0.364	−0.62	0.393			−0.643	0.389
R-Squared	0.248		0.219		0.248		0.203		0.187		0.197
Chi-Squared	41.51		36.092		42.101		32.172		29.716		31.473

,

Table A3. GLS result on living and mobility indices.

	Living						Mobility
Variables	Model 1		Model 2		Model 3		Model 1		Model 2		Model 3
	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.
Digital technology index	3.189	0.699	2.87	0.655	3.721	0.647	4.896	0.718	4.498	0.673	5.343	0.661
Digital technology index squared	3.728	1.926	3.288	1.93			2.987	1.977	2.538	1.981
GPP per capita	0.139	0.716					0.393	0.735
Annual GINI value	−0.159	0.231			−0.151	0.231	−0.028	0.237			−0.03	0.236
Government R&D	−0.612	0.295			−0.554	0.295	−0.548	0.303			−0.494	0.302
R-Squared	0.245		0.216		0.223		0.366		0.342		0.353
Chi-Squared	40.94		35.593		36.663		72.789		66.924		69.962

and

Table A4. GLS result on people index.

	People
Variables	Model 1		Model 2		Model 3
	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.
Digital technology index	9.602	1.602	9.24	1.493	9.093	1.465
Digital technology index squared	−3.779	4.414	−4.565	4.395
GPP per capita	0.24	1.642
Annual GINI value	−0.534	0.529			−0.553	0.524
Government R&D	−1.225	0.677			−1.272	0.669
R-Squared	0.264		0.245		0.26
Chi-Squared	45.248		41.861		44.924

, and the interpretation can also be found in Appendix A.

The GLS model’s constant term is positive, implying that even at very low levels of digitalization there is a baseline smart-city capacity, likely reflecting foundational urban factors or other dimensions of smartness not entirely captured by digital infrastructure. Crucially, even after controlling other factors, the Digital Technology Index remains a robust predictor. Neither the inequality measure (Gini) nor government R&D spending shows a significant effect on the Smart City Index in this specification. Their coefficients are small and not statistically distinguishable from zero. This suggests that, within our sample, socio-economic inequality and public R&D investment do not directly translate into measurable differences in overall smart city performance (at least not in the short-run or in the presence of high multicollinearity among development indicators). The dominance of digitalization variables in explaining smart city rankings underscores the central role of technological implementation in urban innovation outcomes, consistent with the notion that smarter cities are fundamentally enabled by ICT infrastructure and digital services [11].

Figure 1 shows box plot of the Smart City Index across binned levels of the Digital Index. The Digital Index has been segmented into ten deciles (0–9) to show the distribution of smart city performance at each level of digital maturity. The y-axis represents the Smart City Index, a composite score ranging from 0 to 5. The plot illustrates a clear positive trend, where provinces in higher deciles of digitalization also exhibit higher median smart city scores and greater overall performance.

Figure 2 shows heatmap of the conditional probability of Smart City Index levels. Each column represents a decile of the Digital Index, and each row represents a discrete level of the Smart City Index (0–5). The color and number in each cell indicate the probability of a province falling into a specific Smart City Index level, given its Digital Index bin. The concentration of higher probabilities in the top-right portion of the map reinforces the strong positive relationship between the two indices.

Figure 3 shows the relationship between the Digital Index and Smart City Index with a segmented regression line. The red points represent the actual observations for each province-year. The solid blue line represents the fitted relationship from the segmented regression model, with the “kink” occurring at the median value of the Digital Technology Index (0.42). The flat initial slope followed by a steep increase provides strong visual evidence of a critical mass or tipping point effect, where the benefits of digitalization only accelerate after a foundational threshold has been crossed.

While the linear GLS model confirms a positive linkage, the overall goodness-of-fit is only modest, and diagnostic plots shown in Figure 1 and Figure 2 suggest that the relationship may not be strictly linear. The model’s R-squared is moderate, indicating that a substantial portion of variability in the Smart City Index remains unexplained by our linear specification. More revealingly, a visual examination of the data points in Figure 3 show a piecewise pattern. The slope of the relationship appears steeper at lower to mid-range values of the Digital Technology Index and then rises at higher values.

The preceding analysis establishes a strong, piecewise-linear relationship between digital infrastructure and smart city performance, with evidence of accelerating returns after a critical mass threshold is reached. However, these models do not explain the underlying governance and institutional factors that enable cities to cross this threshold. To explore these mechanisms, we now turn to our explanatory case study of Wang Chan Valley, a site where these dynamics can be observed in detail.

4.2. Non-Linearity Analysis: Evidence of Accelerating Returns

To formally test the hypothesis of a piecewise-linear relationship between digitalization and smart city performance, a segmented panel regression was estimated. This approach addresses the potential for a non-linear impact more robustly than a simple quadratic model [29]. The sample was split into low and high digitalization groups based on the median value of the digital index (0.4259). An interaction term was then introduced to test if the effect of the digital index was statistically different for the high-digitalization group.

The results, presented in

Table 8. Segmented panel regression results for non-linearity.

Smart Cities
Variables	Coefficient	(Std. Err.)
Digital Index	1.4320	0.8176
High Digitalization (Dummy > Median)	−1.7815 ***	0.4369
Interaction Term	3.7103 ***	0.9513
GPP per capita	0.6196	1.4164
GINI country	−0.0926	0.1912
Government R&D total	−0.5020	0.4130
const	0.5606	0.3397
Observations	132
R-squared (Within)	0.3465
F-statistic (robust)	31.136
Included Effects	Entity

Note: *** $p<0.001$. Clustered standard errors are in parentheses.

, reveal a significant non-linear relationship consistent with an acceleration effect. The coefficient on the interaction term is positive and highly significant ($\beta =3.71$, $p<0.001$), indicating that the benefits of digital infrastructure are substantially greater in provinces that have already achieved a critical mass of digitalization.

For the low-digitalization group, the effect of the digital index is modest and only marginally significant ($\beta =1.43$, $p=0.08$). However, for the high-digitalization group, the effect is nearly four times larger and highly significant ($\beta =1.43+3.71=5.14$). This finding provides strong evidence that smart city policies may have the greatest impact when they can push developing regions across a key digital maturity threshold, unlocking accelerating returns.

4.3. Stakeholder Interview Insights

To complement the quantitative analysis, we conducted semi-structured interviews with key stakeholders involved in Thailand’s smart city initiatives between October and December 2024. Two interviews were particularly informative: one with a senior digital innovation official from the Eastern Economic Corridor Office (Respondent A), and another with a municipal smart city coordinator from the broader EEC region (Respondent B).

The interview data provides a compelling narrative for the critical mass effect observed in the quantitative results. Respondent A confirmed our quantitative findings regarding non-linear returns to digital investment, noting that while initial digital deployments produced rapid improvements in management and energy efficiency within 18 months, subsequent technology additions yielded progressively smaller gains. This aligns with the initial, flatter part of the curve in Figure 3. Critically, both respondents emphasized that the transition to the steeper, accelerating part of the curve required fundamental shifts toward collaborative governance. They described moving from “top-down technology deployment” to “collaborative innovation” involving universities, businesses, and residents. This institutional evolution supports our theoretical framework linking digital infrastructure effectiveness to Quadruple Helix governance arrangements and provides a tangible explanation for the acceleration of returns: the threshold is not merely technological, but institutional.

Respondent B provided evidence of positive externalities from Wang Chan Valley’s Quadruple Helix approach. They described how neighboring municipalities adapted WCV’s digital systems and, more importantly, its stakeholder engagement methods. Spillover mechanisms included knowledge transfer workshops, regional supply chain development, and workforce training programs confirming our hypothesis about innovation hubs generating broader urban benefits. This demonstrates that the QH model does not just optimize performance within a single entity but can create a learning ecosystem that diffuses innovation across a region.

Both respondents highlighted challenges in replicating WCV’s success across smaller municipalities, particularly regarding institutional capacity for multi-stakeholder collaboration. Respondent B noted that while technology costs are declining, the human capital and coordination requirements for effective implementation may increase as cities move beyond basic digitalization. These interviews validate key aspects of our theoretical framework while revealing important contextual factors specific to emerging economic contexts. The stakeholder perspectives confirm that digital infrastructure effectiveness is mediated by governance quality and that innovation hubs can generate positive spillovers through deliberate knowledge transfer mechanisms. However, they also suggest that the S-curve dynamic we observed quantitatively may be more pronounced in developing urban systems due to uneven institutional capacity, potentially creating new forms of digital divide based on governance capability rather than technological access alone.

5. Discussion

5.1. Analysis of Results

Our findings, drawn from both panel data analysis and an in-depth case study, converge on a central argument: while digital infrastructure is a powerful driver of smart city performance, its full potential is only unlocked when embedded within a mature, collaborative governance ecosystem. The piecewise-linear critical mass effect identified in our quantitative results is not merely a technological threshold but an institutional one, as illustrated by the experience of Wang Chan Valley.

Our panel analysis demonstrates a strong average positive effect of digitalization. The WCV case illuminates the ideal type of pathway for this effect, where a Quadruple Helix framework actively fosters collaboration between government, industry, academia, and civil society [30]. This contrasts with the likely realities of the 12 cities in our panel, where, as stakeholder interviews suggest, institutional constraints, departmental silos, and limited human capital often hinder the absorption of digital innovations. The varied performance in the panel data can thus be interpreted as reflecting different proximities to the collaborative WCV model. The WCV case provides the explanatory for the correlation found in the larger dataset, a key benefit of the mixed-methods approach [31].

Our quantitative models showed no significant effect for the Gini coefficient or government R&D spending. While seemingly counterintuitive, the WCV case offers a plausible explanation. In WCV, R&D investment is effective because it is channeled through a coordinated QH ecosystem that connects research with industry and community needs. This suggests that the impact of R&D is not automatic but is conditional on the existence of a governance framework capable of translating research into innovation. The null result in the panel data may reflect situations where such frameworks are absent, leading to inefficient or untargeted spending. Similarly, while broad income inequality (Gini) may not have a direct short-term effect, the WCV model’s emphasis on ’resident awareness campaigns’ and ’digital training centers’ suggests that proactive efforts to address digital inequality are a key component of its success [32]. This implies that governance is a crucial precondition; without a coherent ecosystem, R&D spending may be inefficient and the negative impacts of inequality on smart city adoption may be more pronounced.

The case of Wang Chan Valley (WCV) illustrates these findings in practice. Early investments in 5G networks, data platforms, and sensor grids generated rapid gains—mirroring the strong impacts observed in mobility and governance sub-indices. However, sustaining progress required complementary measures: digital training centers, resident awareness campaigns, and cross-institutional platforms. These initiatives illustrate why institutional capacity and human capital are essential for moving from early technological gains to longer-term performance.

The Quadruple Helix (QH) model provides a coherent lens for integrating these insights. By fostering collaboration among government, industry, academia, and civil society, QH arrangements embed digital investments within participatory ecosystems. Our interviews reinforce this governance pathway. A senior official from the Eastern Economic Corridor Office noted that early systems in WCV produced quick efficiency gains, but later additions yielded smaller returns, underscoring the need to shift from technology-focused to governance-focused strategies as digital maturity rises. A municipal coordinator further explained that neighboring municipalities replicated not only WCV’s digital tools but also its governance practices, such as stakeholder workshops and regional coordination meetings. These spillovers validate the quantitative finding that governance dimensions are particularly responsive to digitalization, but they also highlight persistent constraints: many smaller municipalities lack digital teams, adequate budgets, or inter-departmental coordination, limiting their ability to capture these gains.

Nonetheless, both respondents identified significant capacity gaps. Smaller municipalities often lacked dedicated digital teams, faced limited budgets, and struggled with departmental silos, which constrained their ability to absorb and adapt digital innovations. These findings suggest that without adequate governance capacity, digital investments risk entrenching inequalities or producing isolated enclaves of smartness rather than broad-based transformation.

5.2. Policy and Future Research Directions

These findings suggest that digital infrastructure should be treated as a strategic enabler within broader governance ecosystems, not an end. Policy responses should therefore emphasize operational strategies such as:

• Regional coordination hubs, measured by the number of joint projects, shared services, and inter-city workshops.
• Standardized stakeholder frameworks, tracking representation of government, industry, academia, and civil society in decision-making.
• Tiered implementation strategies, requiring digital readiness assessments before major investments.
• Financing mechanisms and dedicated digital units, enabling smaller municipalities to overcome fiscal and human resource constraints.

Future research should extend this agenda by employing expert surveys and resident questionnaires to capture governance and civic participation mechanisms beyond the reach of panel econometrics. In addition, spatial econometric approaches could quantify the regional spillover effects of innovation hubs like WCV, while comparative analyses across Global South contexts would clarify how institutional structures mediate the digitalization–performance nexus. Finally, disaggregated studies of emerging technologies such as Al, 5G, or digital twins could reveal domain-specific patterns of returns.

A key limitation of our quantitative model is its modest explanatory power, as indicated by the R-squared values. This suggests that while digital infrastructure is a significant predictor, a large portion of the variance in smart city performance is driven by other factors not included in our specification. Based on the broader literature, future models could be enhanced by incorporating variables from three key areas:

In terms of institutional variables, beyond the general concept of governance, quantitative measures of institutional capacity could provide deeper insights. These might include metrics on the effectiveness of local leadership, the degree of citizen participation in policymaking, or the presence of specific legal and regulatory frameworks that support innovation. Measuring a city’s capacity to collaborate, innovate, and orchestrate complex projects would likely improve the model’s fit.

Regarding demographic factors, characteristics of a city’s population can significantly influence the adoption and impact of smart city initiatives. Variables such as the average education level, age distribution, and labor conditions are known to affect citizens’ engagement with digital services. A higher level of human capital, for instance, may accelerate the benefits derived from digital infrastructure.

Pertaining to industrial structures of the city, its economic composition is another critical factor. A city with a strong concentration in IT manufacturing or knowledge-based industries may be better positioned to leverage digital infrastructure than one with a different economic base. State and national industrial policies can also play a crucial role in shaping the technological capabilities available for smart city development. Future research should aim to control these structural economic differences.

Addressing these factors would require the collection of more granular data, which was beyond the scope of this study but represents a crucial next step for research in this area.

6. Conclusions

This study investigated the relationship between digital infrastructure and smart city performance in an emerging economy context using a panel dataset of 12 cities over 11 years, complemented by qualitative evidence from the Wang Chan Valley innovation hub. Our results show that digital infrastructure has a strong and statistically significant association with smart city indicators across multiple domains. At the same time, the evidence points to a piecewise-linear relationship. The return can be hardly seen in the beginning before skyrocketing after constant investment.

The WCV case illustrates these patterns in practice. Initial infrastructure upgrades enabled rapid gains, but sustained progress required alignment with institutional capacity, human capital development, and participatory governance mechanisms. The experience demonstrates that smart urban transformation is most effective when guided by Quadruple Helix principles, which integrate government, industry, academia, and civil society into collaborative ecosystems. Importantly, WCV’s spillovers to neighboring municipalities show that governance practices such as stakeholder coordination and knowledge transfer can magnify the impact of local digital investments.

From these findings, several policy implications emerge. First, digital infrastructure should be seen not as an end but as a strategic enabler within broader innovation ecosystems. Second, governance capacity must be strengthened before or alongside technological investments, particularly in smaller municipalities where institutional constraints are acute. Third, regional frameworks should be established to capture and diffuse spillover benefits, preventing innovation hubs from becoming isolated enclaves.

Future research can extend these insights in multiple directions. Micro-level analyses, including surveys or process tracing, could identify specific mechanisms linking governance quality to digital outcomes. Spatial econometric approaches could help quantify spillover effects across jurisdictions. Disaggregated studies of emerging technologies such as Al, 5G, or digital twins may reveal domain-specific impacts. Finally, comparative research across Global South contexts would deepen understanding of how institutional structures mediate the trajectory of digital transformation.

In sum, the path to smart urbanism in emerging economies is neither linear nor purely technological. It demands an integrated approach that combines digital infrastructure with adaptive governance, inclusive participation, and long-term institutional resilience.