Acceptance of Smart-City Technologies: Some Evidence on the Role of Perceptions and Demographics from a Municipality of Athens, Greece

Highlights

What are the main findings?

• The SSA model explains citizens’ intention to use smart-city technologies in a municipality of the greater Athens Metropolitan Area, Greece.
• Self-efficacy, price value, and trust in technology strongly predict intention; effort expectancy does not.

What is the implication of the main findings?

• Tailor smart-city policies and services to demographic groups, especially by age and education.
• Use targeted design interventions and communication initiatives to raise adoption and ensure inclusiveness.

Abstract

The rise of the smart city reflects a transformative shift in urban development, defined by the integration of advanced technologies and data-driven solutions seeking to address rapid urbanization, environmental externalities, and the ever-increasing pressing need for optimal resource use. Nevertheless, a better understanding of the factors that shape citizens’ behavioral intentions towards smart-city living is becoming a sheer necessity. This study is among the first to empirically examine determinants describing the propensity to use smart-city services in an urban setting of south-eastern Europe. In this regard, we employ the smart-city stakeholders’ adoption (SSA) model in order to shed light on smart-city technology acceptance, further focusing on the underlying impact of demographics in shaping citizen attitudes and perceptions. Findings suggest that key predictors of acceptance (latent variables describing self-efficacy, price value, and trust in technology), all positively affect behavioral intention while the non-significance of effort expectancy contradicts the relevant results of previous studies and warrants further investigation. Furthermore, the analysis supports the theorized indirect effects of the model, whereas perceived privacy and perceived security both influence behavioral intention via trust in technology, while price value mediates the effect of citizen’s trust in government. The role of demographics was examined for potential moderating effects and was found to be significant, particularly in the case of age and education. Even though the demographic moderators we opted for do not substantially affect the explanatory power of the model, they seem to improve its specificity, particularly regarding perceptions on effort expectancy across the different demographic groups. Such results offer actionable insights on the relevance of smart-city acceptance models to the different demographic groups and in tailoring policies according to demographic segmentation groups with common characteristics.

1. Introduction

The rise of the smart city reflects a transformative shift in urban development, defined by the integration of advanced technologies and data-driven solutions seeking to address rapid urbanization, environmental externalities, and the ever-increasing pressing need for optimal resource use. Smart cities incorporate digital solutions such as the Internet of Things (IoT), artificial intelligence (AI), and big data analytics with the overarching aim of upgrading city operations and having a positive impact on the urban quality of life. Such innovative interventions provide enabling conditions for a more efficient and inclusive urban environment by redefining—among others—transportation, energy consumption, public services, and waste management [1].

The International Telecommunication Union (ITU) pinpoints the strong links between smart-city initiatives, quality of life, and sustainability transitions by defining these cities as innovative settlements “that use Information and Communication Technologies (ICTs) and other means to improve quality of life, efficiency of urban operation and services, and competitiveness, while ensuring that they meet the needs of present and future generations with respect to economic, social, environmental as well as cultural aspects” [2]. Moreover, international organizations such as the UN and the OECD point out critical issues in the deployment of smart technologies in cities by stressing the essential role of engaging citizens in decision-making processes and ensuring that these innovations are people-centered and respond to local needs. In the UN’s ‘Global Assessment of Responsible AI in Cities’ report, the importance of inclusive governance frameworks that reflect both the needs and concerns of local communities is underscored [3]. Likewise, according to the [4], cities must strike a balance between leveraging AI’s potential and ensuring the citizens’ fundamental rights and data security. In a similar vein, the European Union (EU) emphasizes inclusive citizen engagement as a cornerstone of its smart-city initiatives, promoting collaboration between governments, businesses, and communities to ensure technology serves people effectively. Through initiatives like the European Innovation Partnership on Smart Cities and Communities (EIP-SCC), the EU promotes the co-development of urban solutions with the citizens being at the center. This partnership encourages participatory approaches, in terms of co-design and co-production of solutions, to ensure transparency and social inclusion. The EU integrates privacy-by-design principles to maintain public trust and seeks to promote inclusivity by engaging marginalized communities and supporting gender-balanced participation so that the benefits of smart-city solutions are equitably distributed and the public’s buy-in is accelerated [5,6].

Likewise, numerous scholars argue that smart-city technologies are necessary but insufficient solutions to address the multifaceted challenges facing urban areas, and that citizen involvement is crucial for identifying and addressing such challenges effectively (e.g., [7,8]). Citizens’ attitudes towards these pace-setting solutions significantly influence their effectiveness, particularly regarding the potential users’ behavioral intentions in a smart-city living environment. The success of these initiatives largely depends on the acceptance, engagement, and attitudes of citizens (users), who are both beneficiaries and participants in these digital transformations [9]. It is parameters such as the citizens’ perceived usefulness, (dis)trust of data-driven technologies, ease of use, and privacy concerns that can have catalyzing effects on public receptiveness and the relative willingness to actively support and participate in a smart-oriented urban environment [10,11].

Research suggests that without broad public acceptance, even the most promising technological innovations may face resistance and opposition, ultimately undermining their penetration and/or leading to redundant efforts as well as failure costs (e.g., [12,13,14]). Supporting arguments for this claim can also be found in the work of Lytras and Visvizi [15] who report that even highly educated citizens may have reservations and express concerns with regard to the utility, accessibility, safety, and/or efficiency of smart-city services. Likewise, less-densely populated urban settlements, characterized by a strong digital divide, digital illiteracy, and/or a population dominated by certain demographic groups may face significant challenges towards the ‘smartification’ of the city [1,16,17].

The vital importance of citizen-centric smart-city development is further underscored by evidence highlighting unintended social, economic, and environmental impacts of technological solutions implemented without sufficient public input [7,18,19] and delineated by Ianculescu et al. [20] who coined the term ‘smart citizens’ to describe individuals who not only benefitted from using digital technologies in their daily lives but were also empowered to actively participate in the design and refinement of smart public services. Stressing the pivotal role of governance in the acceptance of smart-city applications, Manogaran and Teoh [21] maintain that the success of such technological advancements is contingent upon the citizens’ trust in governance structures, while Neupane et al. [22] propose a trust-based framework for adopting smart-city technologies, emphasizing the importance of securing citizens’ active support to facilitate widespread adoption. Such views are also echoed in Cardullo and Kitchin [23] who argue that smart-city initiatives must shift from rigid ‘top-down’ approaches, primarily driven by tech corporations and governmental bodies, to ‘bottom-up’ models that prioritize and actively engage citizens as primary stakeholders. Thus, as user engagement within the smart-city ecosystem is deemed to be pivotal for the success and sustainability of these initiatives, the importance of better understanding the factors that influence citizens’ behavioral intentions towards smart-city living becomes a sheer necessity [9]. By incorporating citizen input and feedback into the planning and management of smart-city services, these initiatives will be better tailored to the specific needs and priorities of the local population [24]. This is also suggested by Lim et al. [25] who explored the understanding and receptiveness of smart-city policies from the practitioners’ perspective, denoting that future research should place more emphasis into employing both qualitative and quantitative methods for disambiguating citizens’ attitudes toward smart-city development. In a similar vein, Dirsehan and Zoonen [26] stress the (so far) limited understanding of the underlying dynamics describing citizens’ acceptance of smart-city development, urging for a more comprehensive exploration of technology acceptance models.

Against this background, this study is among the first to empirically examine determinants describing the propensity to use smart-city services in an urban setting of the south-eastern Europe. Using an adapted specification of well-established models assessing the acceptance and use of technological interventions we determine the impact of behavioral factors and demographic variables on the intention to embrace smart-city technologies.

The rest of the paper is structured as follows. Section 2 reviews current knowledge on assessing citizens’ acceptance of smart-city technologies and sets forth the conceptual framework of this study. Section 3 and Section 4 outline the material and methods and present the findings, respectively. The paper concludes with a discussion of the findings and final remarks in terms of research and policy implications.

2. Literature Review

The theoretical framework we adopted for studying citizens’ acceptance of smart-city technologies is the smart-city stakeholders’ adoption (SSA) model [27]. The SSA model is based on the unified theory of acceptance and use of technology (UTAUT2), developed by Venkatesh et al. [28]. Albeit introduced a little over 10 years ago, the UTAUT2 has become one of the most frequently cited and employed theoretical approach to better understand and explain technology-adoption-related issues by consumers [29]. Our decision to opt for the SSA model for this study is due to the fact that it has been developed explicitly for studying smart-city technology acceptance by its residents. This model builds directly on UTAUT2, which provides a broad consumer technology-adoption framework through constructs such as performance expectancy, effort expectancy, social influence, and price value. However, UTAUT2 does not sufficiently account for governance-related and rights-based aspects, which are particularly salient in smart-city contexts. The SSA adapts UTAUT2 by integrating constructs specifically relevant to citizen–government–technology interrelations in urban settings, i.e., namely trust in government, trust in technology, perceived privacy, and perceived security. These additional constructs to a certain degree reflect the unique interplay of institutional trust and digital rights in determining citizens’ acceptance of smart-city technologies. In this regard, the European Union’s governance frameworks strongly shape these construct boundaries. For instance, EU policy emphasizes inclusive citizen participation, privacy-by-design, and data protection through the General Data Protection Regulation (GDPR). The European Data Protection Supervisor [5] stresses the safeguarding of personal rights and security as prerequisites for citizen trust. Similarly, the European Climate, Infrastructure, and Environment Executive Agency [6] and the UN-HABITAT [3] highlight that citizen trust in institutions is critical to accelerate the adoption of smart-city technologies. These principles reinforce the SSA’s focus on trust in government and technology as mediating factors that shape how individuals assess value and security in smart applications. Thus, the SSA model positions itself as both an extension and contextual refinement of the UTAUT2: it retains the predictive power of UTAUT2 core constructs (e.g., price value, effort expectancy, self-efficacy), while embedding them within a governance- and rights-sensitive framework tailored for examining smart-city interventions. Such a dual orientation enhances the explanatory power in settings where citizens’ trust in institutional and technological domains is deemed to be crucial for adoption rates.

The SSA model consists of seven constructs: Self-Efficacy, Effort Expectancy, Perceived Privacy, Perceived Security, Trust in Technology, Trust in Government, and Price Value—all of which are expected to positively affect a citizen’s Behavioral Intention (BI) to use smart-city technologies. Four of these seven constructs are hypothesized to have a direct effect on BI: Self-Efficacy (SE, the user believing themselves to be capable of using the smart-city technologies） [27]; Effort Expectancy (EE, ‘the degree of ease associated with consumers’ use of technology’); Price Value (PV, ‘the consumers’ cognitive tradeoff between the perceived benefits of the applications and the monetary cost for using them’) [28]; and, Trust in Technology (TT, the extent that the user trusts that smart-city technologies are adequate and sufficient in protecting his personal data/information) [27]. The remaining three constructs are supposed to indirectly influence one’s BI. Trust in Government (TG, ‘the public’s assessment of government based on their perceptions of political authorities’, agencies’, and institutions’ integrity and ability to provide services according to the expectations of citizens’ ([30], p.17) is expected to affect one’s PV. If residents are skeptical about their local government’s prioritization of their interests and rights， in general, as well as of its motivation and capabilities when it comes to smart technologies, in particular, it is more likely that the perceived ‘costs’ will outweigh the potential ‘benefits’ of these technologies, thus reducing one’s PV. Similarly, Perceived Privacy (PP, ‘the degree to which users believe a given technology is sage and will protect their personal information’ [31], and Perceived Security (PS, ‘the degree to which users believe that smart-city services are secure platforms for storing and sharing sensitive data’ [27]) are expected to influence one’s TT. If the city dwellers believe that the proposed smart technologies will not adequately/securely protect their personal information, it is logical to assume that their Trust in that particular technology will diminish. Figure 1 depicts the theorized relations.

Insights from recent contributions to the mobility-sector digital transformation literature further contextualize smart-city adoption dynamics. Particularly, Saeedikiya et al. [32] apply a dynamic-capabilities lens (in terms of sensing, seizing, and reconfiguring) to examine how digital innovations (re-)configure service ecosystems, with mobility identified as a highly visible and citizen-facing layer of transformation. Such perspectives can be usefully aligned with SSA constructs; for instance, self-efficacy maps onto digital literacy in using journey planning and e-ticketing applications, while effort expectancy reflects frictions of onboarding, authentication, or payment during service use. Trust in technology closely relates to perceptions of reliability, safety, and fairness in routing algorithms and dynamic pricing, whereas privacy and/or security concerns become particularly salient around location data traces generated by mobility apps. Moreover, price value is interpreted by citizens through the travel cost and time saved when adopting these services. Integrating such lens helps to illustrate how the SSA framework operates within a concrete service domain and provides additional insight into why demographic ‘sensitivities’ (such as age and education’s moderating effort expectancy) emerge, since ease of onboarding and digital fluency are disproportionately challenging for certain groups in mobility contexts.

In their study of citizens’ intention to uptake smart-city technologies in Denton, USA, Habib et al. [27] found that all of the SSA model’s assumptions were largely confirmed. The relations between the various constructs and behavioral intention were all found to be statistically significant and having the anticipated signs. Over 55% of the variance in the TT construct (and about 20% of the variance in the PV) was accounted for by Perceived Security and Perceived Trust (and Trust in Government, respectively), while the full model accounted for almost 41% of the dependent construct’s (i.e., Behavioral Intention) variance. Largely, the same results were reported by Hamamurad et al. [33] who employed a highly similar model to Habib et al.’s [27] for studying smart technologies’ adaptation in Kuala Lumpur, Malaysia. (While Hamamurad et al. [33] use the exact same predictor names, relations and hypotheses as Habib et al. [27], the fact that they do not provide the questionnaire they used, does not allow us to determine whether the former is a replication of the latter).

Accordingly, the first aim of our research is to replicate the SSA model in a different context, the city of Athens, Greece, an issue of great importance especially for a newly proposed theory. In the words of the NASEM (National Academies of Sciences, Engineering, and Medicine) [34], “replication is one of the key ways scientists build confidence in the scientific merit of results” (p. 71), while a recent review article appearing in the Annual Review of Psychology approvingly commented on the fact that “Replication—an important, uncommon, and misunderstood practice—is gaining appreciation in psychology” [35]. If we obtain similar results to Habib et al. [27] and Hamamurad et al. [33], then the applicability of the SSA model will be further validated.

While based on the UTAUT2, the SSA model does not take into account the possible effects of demographic variables (age, gender, and education). Given the fact that many models of technology adoption (including the UTAUT2) have suggested that demographics should influence (most likely indirectly) one’s intention to adopt new technologies, the second aim of our research is to examine to which extent the relations proposed in the SSA model are moderated by the respondent’s characteristics. In particular, we will examine the moderating effects of age, gender, and educational attainment on the SSA model’s four main predictors of behavioral intention: Effort Expectancy, Price Value, Self-Efficacy, and Trust in Technology.

3. Hypothesis Development

3.1. The Moderating Effect of Gender

While it has long been considered an important determinant of the digital divide, recent research suggests that gender differences have diminished in importance. Hence Qazi et al.’s [36] meta-analysis revealed a small and positive, yet not significant, effect in favor of boys regarding ICT use and skills, while Shin et al. [37] found no gender differences when it came to smart-city innovation engagement (see also Hou et al. [38] for similar results). That said, a gender divide may still be present regarding the factors affecting smart-city engagement.

Thus, in the original formulation of the UTUA, Venkatesh et al. [39] suggested that Effort Expectancy would be more salient for women than for men, “driven by cognitions related to gender roles” (p. 450). Thus, they argued that “The influence of effort expectancy on behavioral intention …will be stronger for women”. Nevertheless, both Ventakesh et al. [28] and Teng et al. [40] found no gender differences regarding Effort Expectancy’s effect on an individual’s willingness to engage in IT/smart-city technologies. On the contrary, Nusir et al. [41] found that gender significantly moderated the relationship between ‘Perceived Ease of Use’ (‘defined as less effort required by people to use a new system or technology’) and the behavioral intention to adopt smart-city technologies in Jordan.

It was similarly argued by Venkatesh et al. [28] that gender would moderate the relationship between Price Value and Behavioral Intention. According to the authors, this may be the case because women are supposed to “be more cost conscious than men” and “given the penchant of men to play with technologies, the price value assigned by men to technologies will likely be higher than the value assigned by women” (p. 163). Yet, Venkatesh et al. [28] failed to find any moderation effects.

On the contrary, gender is likely to moderate the effect of Self-Efficacy. A recent meta-analysis of existing research [42] found that, albeit less prominent than in the past, males have a higher self-efficacy compared with females regarding technology, technology use, computers, etc.

Regarding Trust in Technology, Beştepe and Yildirim’s [43] research for Turkey, and Teng’s et al. [40] for Malaysia, found no moderating effects for Gender.

Accordingly, we posit the following hypotheses:

Hypothesis 1a: 

Gender will not moderate the relation between Effort Expectancy and Behavioral Intention to adopt smart-city technologies.

Hypothesis 1b: 

Gender will not moderate the relation between Price Value and Behavioral Intention to adopt smart-city technologies.

Hypothesis 1c: 

The effect of Self-Efficacy on Behavioral Intention to adopt smart-city technologies will be stronger for females.

Hypothesis 1d: 

Gender will not moderate the relation between Trust in Technology and Behavioral Intention to adopt smart-city technologies.

3.2. The Moderating Effect of Age

Available research has established that younger individuals are more likely to use the internet [44] as well as smart-city applications [37,38]. Venkatesh et al. [39] suggest that Effort Expectancy would be more salient for older individuals, since “increased age has been shown to be associated with difficulty in processing complex stimuli and allocating attention to information on the job … both of which may be necessary when using software systems” (ibid.). Yet, available research has not supported this claim [28,40], contrary to Nusir et al. [41]. Regarding Price Value, while it was hypothesized that older individuals would be more cost conscious—since they were more likely to be burdened with the task of providing for their families [39]—no such effect was found by Venkatesh et al. [28] (see also [41] for similar results). Regarding the moderating effect of Age on Trust in Technology, available research has returned mixed results. While a survey of UK residents found that younger individuals are more trusting regarding the ‘benevolence’, ‘integrity’, ‘privacy’, and ‘security’ of smart home technologies [45], Beştepe & Yildirim [43] for Turkey, and Teng et al. [40], for Malaysia, found no moderation effects of Age on trust. Finally, concerning Self-Efficacy, we are unaware of any studies that examined the possible moderating effect of Age. Arguably, younger individuals would find it easier to engage in smart-city technologies, due to their greater use of IT in general.

Accordingly, we suggest the following hypotheses:

Hypothesis 2a: 

Age will not moderate the relation between Effort Expectancy and Behavioral Intention to adopt smart-city technologies.

Hypothesis 2b: 

Age will not moderate the relation between Price Value and Behavioral Intention to adopt smart-city technologies.

Hypothesis 2c: 

The effect of Self-Efficacy on Behavioral Intention to adopt smart-city technologies will be stronger for younger individuals.

Hypothesis 2d: 

Age will not moderate the relation between Trust in Technology and Behavioral Intention to adopt smart-city technologies.

3.3. The Moderating Effect of Education

Available research suggests that higher educational attainment has a positive effect on smart-city application usage [37,38], arguably because higher education is related to higher internet usage [44].

Arguably, individuals with a higher educational attainment will find it easier to adopt smart-city technologies because they possess both the knowledge and the experience (through their longer formal training) in dealing with ICTs. While it was suggested that Effort Expectancy would be more salient for less experienced users [39], available research has not supported this claim [28]. We also anticipate that Self-Efficacy will be more important for better-educated individuals since they have (acquired) the skills and knowledge necessary for engaging with smart-city technologies. On the contrary, we anticipate a negative moderating effect of Educational Attainment on the Trust in Technology, since available research has shown that lower educated individuals are overall more trusting regarding the ‘integrity’, ‘security’, and ‘privacy’ of smart home technologies [45]. Finally, following Venkatesh et al. [28], who anticipated no moderating effect of ‘experience’ on Price Value, we similarly suggest that Education will not have a moderating effect on Price Value.

Accordingly, we will test the following hypotheses:

Hypothesis 3a: 

The effect of Effort Expectancy on Behavioral Intention to adopt smart-city technologies will be stronger for more educated individuals.

Hypothesis 3b: 

Education will not moderate the relation between Price Value and Behavioral Intention to adopt smart-city technologies.

Hypothesis 3c: 

The effect of Self-Efficacy on Behavioral Intention to adopt smart-city technologies will be stronger for more educated individuals.

Hypothesis 3d: 

The effect of Trust in Technology on Behavioral Intention to adopt smart-city technologies will be stronger for more educated individuals.

4. Material and Methods

4.1. Context of the Study and Research Instrument

Our study focuses on a municipality of the Attica regional unit, Greece. It is located in the southern suburbs of the greater Athens metropolitan area, with a population of approximately 50,000 inhabitants, and is an area where, in the last few years, a growing number of smart-city initiatives have been deployed towards the integration of smart, citizen-centered, and sustainability-oriented services, in line with international initiatives and standards. Such applications include, among others, (a) a sensor-based IOT network system to refine the daily operations of municipal vehicles and buildings, school units, public parks, and open areas, (b) smart bins for improved waste management and advanced telematic techniques for route optimization in municipal waste collection, (c) smart irrigation management systems of green spaces, (d) provisions for the utilization of smart parking sensors for improving parking efficiency and smartphone GPS data to map and mitigate traffic congestion, (e) the development and deployment of a digital platform for analysis and prediction of tourists’/visitors’ flow in the area that will propose automated measures and interventions across critical areas of the municipality’s public services (i.e., waste management, security, and transportation), (f) the development of smart, AI-based, governance tools to endorse the citizens’ participation in decision-making processes and the formulation of the municipality’s policies, and (g) a smart detection and monitoring system equipped with drones with thermal imaging cameras to report early warnings of temperature change, allowing immediate responses to fire threats.

The data for this study was collected during the first quarter of 2024 from citizens of the aforementioned municipality, using snowball-convenience sampling and through an online questionnaire that was developed in Greek. This resulted in 572 usable responses after incomplete records were discarded. Adopting the measurement items of Habib et al. [27] for latent variables explaining smart-city technology acceptance, the data collection instrument consists of 30 items (rated on a five-point scale where ‘1’ stands for ‘Strongly disagree’ and ‘5’ for ‘Strongly Agree’) that evaluate an individual’s propensity to adopt and promote smart-city applications (see Appendix A) along with three questions to describe the sample in terms of age, gender, and educational attainment.

Table 1. Demographic information (and coding category).

	Frequency	Percentage
Gender
Male (1)	288	50.3
Female (2)	284	49.7
Age
<24 years (1)	100	17.5
25–44 years (2)	176	30.8
45–64 years (3)	212	37.0
>65 years (4)	84	14.7
Education
Secondary (1)	288	50.3
Vocational (2)	22	3.9
Undergraduate (3)	220	38.5
Postgraduate (4)	42	7.3

summarizes these basic demographic characteristics of the respondents.

4.2. Data Analysis and Structural Equation Modeling Approach

To validate the theoretical framework of this study, we employed structural equation modeling (SEM), a statistical technique that enables the simultaneous examination of multiple relationships among latent variables [46]. SEM is widely used in empirical research to assess complex causal relationships, ensuring a systematic representation of constructs within a theoretical model. Our analysis involved SEM, aligning with contemporary approaches to modeling nonlinear systems and intricate interactions among factors.

Moreover, to examine the moderating effects of demographic variables on key relationships within the model, we employed a moderation analysis approach within the structural equation modeling (SEM) framework, by extending the standard SEM model utilizing moderating terms of main demographic variables. Specifically, we tested whether age, gender, and education moderated the influence of self-efficacy (SE), effort expectancy (EE), trust in technology (TT), and price value (PV) on the behavioral intention (BI) to adopt smart-city applications. Interaction terms between the moderator variables and the respective predictors were incorporated into the model, and their statistical significance was assessed.

Before proceeding with SEM, we ensured the model met the necessary reliability and validity criteria. Construct validity, convergent validity, and discriminant validity were examined to confirm that measurement items effectively captured the intended latent constructs. Cronbach’s alpha coefficient [47] was employed to assess the internal consistency of constructs, ensuring values met the threshold for reliability. Convergent validity was assessed through the average variance extracted (AVE) and composite reliability (CR) measures. Additionally, the explanatory power of each construct was evaluated by the proportion of variance explained, ensuring a minimum of 50% variance capture for validity. Finally, discriminant validity is assessed by the utilization of the Fornell–Larcker criterion [48] and the heterotrait–monotrait ratio (HTMT) [49]. Fornell–Larcker criterion compares the square root of AVE for each construct with its correlations and with any other construct. A construct should share more variance with its own indicators than with other constructs. HTMT measures the ratio of between-construct correlations to within-construct correlations. A HTMT value should be <0.90.

Prior to SEM, the data were also checked for the potential problem of multicollinearity (i.e., via inter-construct correlations, HTMT, and variance inflation factors).

To assess the overall adequacy of the model, several goodness-of-fit (GoF) indices were utilized [50]. These included the χ²/df (chi-square divided by degrees of freedom) ratio, the root mean square error of approximation (RMSEA), the goodness-of-fit index (GFI), the adjusted goodness-of-fit index (AGFI), the parsimonious goodness-of-fit index (PGFI) and the comparative fit index (CFI). Standard benchmarks suggest that GFI, AGFI, CFI, and PGFI should ideally exceed 0.8–0.9 for an acceptable model fit, whereas RMSEA values should be below 0.08 for a well-fitting model [51]. A χ²/df value of ≤3 suggests an acceptable model fit.

The statistical analysis, including reliability/validity checks, was conducted using SPSS 21.0 [52]. For SEM analysis, model estimation, and hypothesis testing, we employed the AMOS 7.0 software [53], which facilitated model-fitting, path analysis, and inferential testing.

5. Results

5.1. Reliability and Validity Analysis

Prior to SEM analysis, reliability and validity tests were conducted for the latent variables incorporated in the SEM models (

Table 2. Reliability and convergent validity measures for latent constructs used in SEM analysis (Cronbach’s α, % of explained variance, average variance extracted, composite reliability, variance inflation factor).

Latent Structure	Cronbach’s Alpha	% of Variance	AverageVariance Extracted (AVE)	Composite Reliability (CR)	Variance Inflation Factor (VIF)
Perceived Security	0.927	76.83	0.733	0.933	3.184
Perceived Privacy	0.768	63.11	0.594	0.856	3.676
Trust in Government	0.800	56.41	0.671	0.859	2.336
Self-Efficacy	0.701	53.94	0.460	0.783	2.159
Effort Expectancy	0.755	55.87	0.534	0.684	2.320
Price Value	0.832	64.09	0.657	0.801	1.466
Trust in Technology	0.734	51.67	0.617	0.689	1.858
Behavior Intention	0.876	74.79	0.761	0.848	2.293

).

Cronbach’s alpha values were used to evaluate internal consistency, while the percentage of explained variance from factor analysis provided insights into construct validity. The results indicate that all constructs exhibit acceptable to excellent reliability, with Cronbach’s alpha values exceeding the commonly accepted threshold of 0.70. Regarding construct validity, the percentage of explained variance exceeded the recommended 50% threshold for all factors, confirming their ability to capture a substantial portion of variance in the observed indicators.

In addition, we have checked for convergent validity through the average variance extracted (AVE) and composite reliability (CR) measures (see Table 2). The AVE and CR values for the latent constructs of our analysis exceeded the limits for a construct to considered reliable, with almost all AVEs being higher than 0.5, whereas majority of CI values were >0.7. Discriminant validity tests through the Fornell–Larcker criterion and HTMT are shown in

Table A1. Fornell–Larcker criterion values matrix.

Construct	SE	EE	PS	PP	TT	PV	TG	BI
SE	0.678	0.670	0.487	0.431	0.450	0.379	0.335	0.397
EE		0.731	0.655	0.450	0.300	0.558	0.484	0.300
PS			0.856	0.388	0.331	0.378	0.411	0.562
PP				0.771	0.395	0.410	0.409	0.445
TT					0.785	0.530	0.384	0.410
PV						0.810	0.383	0.340
TG							0.819	0.683
BI								0.872

Diagonal values are √AVE, off-diagonal values are correlations.

and

Table A2. Heterotrait–Monotrait ratio value (HTMT) matrix.

Construct	SE	EE	PS	PP	TT	PV	TG	BI
SE	0.847	0.726	0.662	0.624	0.495	0.495	0.467	0.701
EE		0.876	0.655	0.456	0.731	0.691	0.512	0.503
PS			0.503	0.538	0.602	0.575	0.776	0.627
PP				0.490	0.535	0.556	0.582	0.821
TT					0.768	0.599	0.622	0.463
PV						0.626	0.499	0.469
TG							0.915	0.730
BI								0.684

in Appendix B, respectively. According to the Fornell–Larcker criterion (Table A1), most diagonal values (√AVE) are higher than off-diagonal correlations, indicating adequate discriminant validity for most construct pairs. Similarly, according to the HTMT criterion, discriminant validity holds adequately, since almost all of the HTMT values are <0.9 (Table A2).

With regard to the problem of multicollinearity between the latent constructs, the AVE values indicate the absence of multicollinearity in the data, since all variance inflation factor values (VIF) are well below the threshold of 10 (see Table 2).

Overall, these results suggest that the latent factors demonstrate satisfactory reliability and validity, justifying their inclusion in the SEM model.

5.2. Original Model

As a first step, we re-ran Habib et al.’s [27] original SSA model. The results, presented in Figure 2 and

Table 3. Standard SSA (smart-city stakeholders’ acceptance) model results (standardized loadings, significance, 95% confidence intervals).

Association			Estimate	p-Value	95% Lower Bound	95% Upper Bound
Perceived Security	→	Trust in Technology	0.719	<0.001	0.699	0.737
Perceived Privacy	→	Trust in Technology	0.695	<0.001	0.676	0.715
Trust in Government	→	Price Value	0.994	<0.001	0.967	0.999
Self-Efficacy	→	Behavior Intention	0.440	0.004	0.374	0.513
Effort Expectancy	→	Behavior Intention	0.033	0.364	0.022	0.202
Price Value	→	Behavior Intention	0.526	<0.001	0.477	0.563
Trust in Technology	→	Behavior Intention	0.727	<0.001	0.697	0.743
R2 (explained variance) of dependent variable, Behavior Intention: 0.646

, are all largely as expected. Perceived Security and Privacy are important predictors of Trust in Technology (explaining 50.67% of the variance). Similarly, Trust in Government strongly impacts Price Value (44.09% of variance explained). Regarding the Behavioral Intention to engage in smart-city technologies, all predictors turned out to be statistically significant, except for Effort Expectancy (i.e., the perceived ease in engaging with technology), which turned out to be statistically non-significant. The model explains 64.6% of the variance for Behavioral Intention.

5.3. Moderated Models

5.3.1. The Demographics’ Moderating Effect on Self-Efficacy (SE)

In

Table 4. Results of the SSA model incl. the moderating effects of demographics on Self-Efficacy (standardized loadings, significance, 95% confidence intervals).

Interactions	Estimate	p-Value	95% Lower Bound	95% Upper Bound
PS → TT	0.937	<0.001	0.875	0.998
PP → TT	0.944	<0.001	0.880	0.999
TG → PV	0.945	<0.001	0.932	0.995
TT → BI	0.765	<0.001	0.662	0.843
PV → BI	0.669	<0.001	0.520	0.796
EE → BI	0.397	0.002	0.249	0.599
SE → BI	0.739	<0.001	0.636	0.838
Gender*SE → BI	−0.022	0.402	−0.178	0.021
Age*SE → BI	0.620	<0.001	0.567	0.668
Education*SE → BI	0.213	0.009	0.125	0.285
R2 of BI: 0.635
Model goodness-of-fit statistics: χ2/df: 1.99; RMSEA: 0.073; GFI: 0.941; CFI: 0.920; AGFI: 0.935; PGFI: 0.843

, we present the predictors’ effects on BI once the SE→BI relation is moderated by Age, Gender, and Education.

We find that the Gender*SE interaction is non-significant; thus, males and females are equally self-confident that they can use the smart-city technologies (Hypothesis 1c is rejected). On the contrary, Hypothesis 2c is supported: self-efficacy is a stronger determinant of behavioral intention for elder individuals. Similarly, self-efficacy is a stronger determinant of behavioral intention for the better educated (Hypothesis 3c confirmed).

5.3.2. The Demographics’ Moderating Effect on Effort Expectancy (EE)

In

Table 5. Results of the SSA model incl. the moderating effects of demographics on Effort Expectancy (standardized loadings, significance, 95% confidence intervals).

Interactions	Estimate	p-Value	95% Lower Bound	95% Upper Bound
PS → TT	0.945	<0.001	0.892	0.999
PP → TT	0.938	<0.001	0.889	0.999
TG → PV	0.992	<0.001	0.944	0.999
TT → BI	0.719	<0.001	0.625	0.798
PV → BI	0.653	<0.001	0.538	0.776
EE → BI	0.365	0.014	0.227	0.492
SE → BI	0.679	<0.001	0.576	0.778
Gender*EE → BI	−0.242	0.020	−0.335	−0.161
Age*EE → BI	0.505	0.008	0.454	0.562
Education*EE → BI	0.129	0.046	0.049	0.217
R2 of BI: 0.655
Model goodness-of-fit statistics: χ2/df: 1.89; RMSEA: 0.069; GFI: 0.953; CFI: 0.961; AGFI: 0.948; PGFI: 0.854

, we present the predictors’ effects on BI once the EE→BI relation is moderated by Age, Gender, and Education.

All the interaction effects turned out statistically significant. Effort Expectancy exerts a stronger influence on adopting smart-city technologies for males (Hypothesis 1a is rejected), the elderly (Hypothesis 2a is rejected), and the better educated (Hypothesis 3a is confirmed).

5.3.3. The Demographics’ Moderating Effect on Trust in Technology (TT)

In

Table 6. Results of the SSA model incl. the moderating effects of demographics on Trust the Technology (standardized loadings, significance, 95% confidence intervals).

Interactions	Estimate	p-Value	95% Lower Bound	95% Upper Bound
PS → TT	0.938	<0.001	0.888	0.978
PP → TT	0.943	<0.001	0.904	0.999
TG → PV	0.944	<0.001	0.911	0.999
TT → BI	0.762	<0.001	0.645	0.840
PV → BI	0.670	<0.001	0.541	0.800
EE → BI	0.350	0.014	0.280	0.504
SE → BI	0.730	<0.001	0.609	0.816
Gender*TT → BI	0.006	0.457	−0.100	0.097
Age*TT → BI	0.582	<0.001	0.504	0.648
Education*TT → BI	0.338	0.031	0.257	0.400
R2 of BI: 0.637
Model GoF statistics: χ2/df: 2.08; RMSEA: 0.097; GFI: 0.886; CFI: 0.847; AGFI: 0.872; PGFI: 0.786

, we present the predictors’ effects on BI once the TT→BI relation is moderated by the three demographic variables.

The moderation effect of Gender is statistically non-significant (Hypothesis 1d is confirmed). On the contrary, Trust in Technology is more relevant for elderly individuals (Hypothesis 2d is rejected) and the more educated ones (Hypothesis 3d is confirmed).

5.3.4. The Demographics’ Moderating Effect on Price Value (PV)

In

Table 7. Results of the SSA model incl. the moderating effects of demographics on Price Value (standardized loadings, significance, 95% confidence intervals).

Interactions	Estimate	p-Value	95% Lower Bound	95% Upper Bound
PS → TT	0.952	<0.001	0.902	0.998
PP → TT	0.936	<0.001	0.895	0.989
TG → PV	0.973	<0.001	0.897	0.999
TT → BI	0.680	<0.001	0.507	0.708
PV → BI	0.634	<0.001	0.499	0.756
EE → BI	0.317	0.034	0.212	0.432
SE → BI	0.630	<0.001	0.502	0.748
Gender*PV → BI	−0.109	0.088	−0.210	−0.038
Age*PV → BI	0.410	0.009	0.315	0.501
Education*PV → BI	0.215	0.042	0.135	0.290
R2 of BI: 0.687
Model GoF statistics: χ2/df: 2.34; RMSEA: 0.091; GFI: 0.800; CFI: 0.822; AGFI: 0.775; PGFI: 0.710

, we present the predictors’ effects on BI once the PV→BI relation is moderated by the demographic variables.

The moderation effect of Gender is statistically non-significant (Hypothesis 1b is confirmed). On the contrary, Price Value is more relevant for elderly individuals (Hypothesis 2b is rejected) and the more educated ones (Hypothesis 3b is rejected).

6. Discussion and Concluding Remarks

In this paper we set out to examine the factors affecting citizens’ acceptance of smart-city technologies in a Greek town. Available research suggests that citizen involvement is crucial for delivering the potential benefits of smart-city technologies, yet our understanding of the underlying dynamics relating to citizens’ acceptance of smart-city development is still limited. Accordingly, in this study, we tested a model developed explicitly for studying smart-city technology acceptance by its residents, the smart-city stakeholders’ adoption (SSA) model [27]. To the best of our knowledge, this is the first time this model is tested for a south European country while available research on smart-city development in Greece has largely ignored the determining factors for citizens’ acceptance.

Our findings suggest that the SSA model adequately explains the citizens’ motivation to engage with smart-city technologies. For the Greek case study, the model explains over 64% of the variance in the respondent’s ‘Behavioral Intention’ which is similar (and actually higher) to what has been found in other studies employing the same model; thus, this research validates the applicability of the SSA model to a different cultural setting. Furthermore, our results also largely corroborate the theoretical relations suggested in the SSA model. Self-Efficacy (‘the user believing him/herself capable of using the smart-city technologies’); Price Value (‘the consumers’ cognitive tradeoff between the perceived benefits of the applications and the monetary cost for using them’); and Trust in Technology (‘the extent that the user trusts that the smart-city technologies are adequate and sufficient in protecting his/her personal data/information’), were all found to positively affect one’s behavioral intention. The only deviation from other studies’ results relates to Effort Expectancy (‘the degree of ease associated with consumers’ use of technology’), which turned out to be statistically non-significant. We will return to this unanticipated result further down in the text.

Furthermore, our results support the SSA model’s theorized indirect effects on behavioral intention. Perceived Privacy and Perceived Security affect behavior through Trust in Technology, while the effect of Trust in Government operates through the Price Value construct.

The second aim of our study was to examine the possible moderation effects of demographic variables (age, gender, and educational attainment) on the SSA model’s main predictors of behavioral intention (Effort Expectancy, Price Value, Self-Efficacy, and Trust in Technology), since many other technology acceptance models have anticipated that demographic characteristics would indirectly influence one’s intention to adopt new technologies. Our results suggest that this is a valid anticipation: 10 out of the 12 moderation effects turned out to be statistically significant, at least at the 0.05 level.

Including the demographic moderators does not alter the overall explanatory power of the SSA model. Thus, the ΔR²s between the standard SSA model-fitted and the -moderated ones is not statistically significant for all cases (based on the likelihood ratio test (LRT), p-value > 0.05). Nevertheless, taking into account the moderation effects improves the original model’s specification. This is especially evident for the case of the Effort Expectancy (EE) predictor. As shown above, in the standard SSA model, EE turned out to be statistically non-significant contrary to what was expected (see Table 3). Yet, in all the subsequent, moderated, models, EE is found to be statistically significant, suggesting that this particular predictor is significantly associated with all the moderator variables.

Our findings have practical implications for local government agencies, private sector firms, and any other stakeholder wishing to promote smart-city technologies (SCTs). We found that, as suggested in the SSA model’s original formulation, the more one believes that they have the necessary capacity for using the smart-city technologies (‘Self-Efficacy’), the easier they think using them is (‘Effort Expectancy’), the more they think that these technologies adequately protect their personal data (‘Trust in Technology’) and the more they are willing to accept a tradeoff in order to enjoy these technologies (‘Price value’), the more they are likely to engage in using SCTs. Nevertheless, our results suggest that these drivers’ effects are moderated by the citizens’ demographic characteristics. The ease of use plays a more important role in adopting SCTs for males compared with females. On the other hand, all four drivers exert a stronger influence in decision to engage with SCTs for older citizens compared with younger ones- and this is also the case for better-educated individuals compared with lesser ones. Thus, generic ‘one-size-fits-all’ information campaigns are unlikely to have an optimal effect. Therefore, local authorities should tailor their promotional and awareness-raising campaigns to the priorities of the different population segments, primarily focusing on the older and/or less digitally literate residents. Beyond such interventions for enhancing citizens’ digital confidence, local governments should place emphasis on transparency in data governance, the integration of privacy-by-design principles, and robust security standards to build trust. In addition, as price value emerged as a decisive parameter, they should explore subsidized access or tiered pricing to ensure equitable participation across the different population groups. Accordingly, developers of SCTs should ensure, and promoters should stress, that these technologies are indeed ‘a thing you (already) know, a ‘piece of cake’ to use, ‘waterproof’ against any breach, and available at a pinch of the user’s time or money. Companies developing or deploying smart-city solutions should tailor such services-products based on simplified, user-friendly interfaces, with perhaps more advanced features incorporated for younger, relatively more digitally skilled users. Likewise, clear(er) communication of tangible benefits (in terms of, e.g., time savings, efficiency, and/or energy reductions) could increase perceived value. In this regard, given trust’s centrality, compliance with EU-level data protection standards (i.e., GDPR) and adopting credibility-enhancing strategies (built around well-established third-party certifications and audits) will be crucial for business entities developing and promoting SCTs. Lastly, members of civil society (NGOs, citizen associations, academia) may act as intermediaries, advocating for inclusive SCT transitions and supporting those citizen groups in risk of being marginalized. Collaboration of such stakeholder groups with municipalities in co-designing services can potentially lead to better alignment of SCTs with local needs, securing citizen buy-in while offering additional pathways for increasing citizens’ digital literacy and long-term engagement with SCTs. Taken together, such implications indicate that successful smart-city development in Athens—and, by extension, in similar urban settings—depends not only on technological robustness but also on socially inclusive and targeted strategies that address demographic variation in order to achieve increased adoption rates and ensure equitable access.

As a final note, we would like to comment on the limitations of our research. To start, our sample comes from a particular Greek town in Greater Athens; thus, it is not representative of the entire country. Furthermore, this was a self-selected sample, accessible only to those who do have some competence in digital media. These limitations raise valid concerns about the existence of selection bias in our data and, ultimately, about the external validity of our findings. While we do acknowledge that our data limitations do not allow our findings to be generalized to the broader population, it is important to assess how these limitations affect our research’s prime goals. Thus, our first aim was to replicate the SSA model, employed by Habib et al. [27] and Hamamurad et al. [33], in order to assess its external validity. The fact that our results are highly consistent to the previous two studies (which have also employed non-random sampling) suggests that the SSA model’s theorization holds across different locations, towns and times. That said, further studies using more representative or randomized samples are necessary before deciding on the SSA model’s generalizability. Our study’s second aim was to examine whether the SSA model’s predictive power could be enhanced by taking into account the moderating effects of demographic variables—similar to what has been shown to be the case in other technology-adoption models. Due to the paucity of previous research on the matter, our primary goal was to provide ‘proof-of-concept’ evidence that the suggested moderations are indeed plausible, and our results lend credence to this claim. Obviously, further studies, using more representative designs, are needed to confirm these preliminary findings.