Digital Transformation of the State: A Multivariable Model Applied to the Public Sector in Lima, Peru

Abstract

Digital transformation has become a central strategy for modernizing the public sector. This study analyzes the relationships between digital competencies, digital literacy, change management, soft skills, technology adoption, digitization of the public function, digital public services, e-governance, the digital divide, digital transformation, and state modernization in Peru. A quantitative, cross-sectional design was applied, using Partial Least Squares Structural Equation Modeling with data collected from 379 public servants in Metropolitan Lima. The results show that digital transformation and e-governance significantly influence state modernization. Digital competencies and digital literacy play an important role in facilitating technology adoption, while soft skills mediate the link between individual capacities and institutional processes. Digitization of the public function supports efficiency in administration and connects with the provision of digital public services, which in turn improves citizen access and trust. E-governance contributes to reducing the digital divide and reinforces the delivery of digital public services. The findings extend existing frameworks by integrating technological, organizational, and human dimensions into a single model. In practical terms, the study provides guidance for policymakers to strengthen digital competencies and literacy, implement structured change management programs, reinforce soft skills in civil service, advance the digitization of administrative functions, expand digital public services, and design inclusive policies to reduce the digital divide, thereby supporting sustainable state modernization.

1. Introduction

Digital transformation (DT) has become a strategic factor for state modernization, as it is not limited to the incorporation of new technologies but encompasses profound changes in organizational processes and in the relationship between government and citizens (Criado & Gil-Garcia, 2019; Mergel et al., 2019). Recent studies show that digitalization in the public sector fosters efficiency, transparency, and greater citizen participation, provided there is alignment between the technological dimension, organizational capacities, and human competencies (García Saisó et al., 2022; OECD, 2020).

In this context, the present research takes Lima, Peru, as a case study, since it concentrates most of the national public institutions and clearly reflects the structural challenges of digitalization in the State (United Nations CEPAL, 2021). The Peruvian government, through the Secretariat of Government and Digital Transformation of the Presidency of the Council of Ministers, promotes the National Digital Transformation Policy (PCM, 2021), which seeks to foster interoperability, digital security, and citizen-centered service delivery. However, the current situation shows partial progress: while there have been advances in regulation and digital government platforms, gaps remain in human resources, digital skills, and change management that limit the effectiveness of these initiatives (García Saisó et al., 2022; OECD, 2020).

In the academic literature, digital transformation in the public sector has mainly been addressed from three perspectives: the technological, focusing on the adoption of digital infrastructure and interoperability platforms; the organizational, which examines governance processes, leadership, and change management; and the human, linked to digital competencies and soft skills necessary for public officials to adapt to innovation (Espina-Romero et al., 2023, 2024a, 2024b).

Although previous studies have analyzed digital transformation in governments mainly from technological or management perspectives in isolation (Bonduki & Cunha, 2022; Gangneux & Joss, 2022; Posselt, 2022), there is a gap in the literature regarding integrated models that articulate technological, organizational, and human dimensions, particularly in Latin American contexts. This gap is relevant because state modernization processes require not only technological innovation but also organizational capacities and soft skills to sustain change. Our study addresses this gap by proposing and validating a multivariable model through PLS-SEM that links these three dimensions within the framework of state modernization.

The main contribution of this research lies in offering an integrative approach that combines technological (Hameduddin et al., 2025), organizational (Congo & Choi, 2022; Engkus, 2025; Semenets-Orlova et al., 2023), and human factors as determinants of state modernization (Febiri et al., 2024). Unlike previous studies that have prioritized a single approach, this work contributes theoretically—by advancing the understanding of digital transformation as a multidimensional phenomenon in public administration—and empirically—by applying and validating this model in the Peruvian context through structural analysis (Calderón, 2021; OECD, 2019).

Consequently, the research question guiding this study is as follows: how do technological, organizational, and human dimensions influence the modernization of the Peruvian State through digital transformation? Aligned with this question, the general objective of the research is to analyze how technological, organizational, and human factors influence the modernization of the Peruvian State through digital transformation, based on the empirical validation of a multivariable model using the PLS-SEM technique.

This article is organized into seven sections. Following this introduction, the theoretical framework is presented, which establishes the foundations of the study. Next, the development of hypotheses is detailed, outlining the proposed relationships among the variables. The subsequent section describes the materials and methods, including the research design and statistical procedures. Then, the results of the structural analysis are presented, followed by a discussion of their theoretical and practical implications. Finally, the article concludes with the main findings, limitations, and recommendations for future research.

2. Theoretical Framework

2.1. Digital Competencies (DCs)

DCs refer to the set of knowledge, skills, attitudes, strategies, and awareness required to properly use digital technologies. These competencies go beyond the instrumental use of ICTs and encompass aspects such as problem-solving, digital content creation, digital security, and communication in technological environments (Bilan et al., 2023; Cardoso & Gomes, 2025). In the public sector context, Lopes et al. (2023) highlight that officials with higher levels of education and professional skills are more willing to participate in digital training programs. Likewise, Cordella et al. (2024) argue that the development of DCs is a fundamental element in achieving the Sustainable Development Goals, as it enables the design, implementation, and management of public policies based on information technologies.

Compared with Digital Literacy, DCs emphasize the ability to apply digital skills in practice, which highlights differences in scope. While Digital Literacy ensures critical use of information, DCs involve applying those skills to solve problems and design policies. This makes DCs a key variable to explain how individual capacities link with organizational performance, directly addressing the gap in integrated models.

2.2. Digital Literacy (DL)

DL refers to the ability to critically and ethically understand, evaluate, use, and create digital information. Arnaud et al. (2024) affirm that DL is indispensable for public employees to actively engage in DT, as it enables them to adopt and adapt technologies to their work contexts. Tsarouhas and Grigoriadis (2025) broaden this approach, stating that a digitally literate citizenry and public administration are essential pillars for ensuring algorithmic transparency, citizen participation, and responsible governance. Moreover, Grgurevic et al. (2022) find that sustainable public projects, such as green mobility, are only feasible when public servants possess adequate levels of DL.

The literature shows agreement on the role of DL as a foundation for effective DT but differs on whether it should be understood as an individual or institutional capacity. These differences justify including DL in the model as a factor that mediates between education, citizen participation, and modernization of the state.

2.3. Change Management (CM)

CM in the public sector refers to the processes through which public institutions plan, implement, and consolidate organizational, technological, and cultural changes. Barodi et al. (2024) emphasize that a comprehensive CM strategy, incorporating communication, participation, and effective leadership, is vital for achieving sustainable public sector reforms. Rehouma et al. (2020) showed that active employee participation during technological change processes significantly increases acceptance and success. Semenets-Orlova et al. (2023) also highlight the need to train new leaders with innovative thinking and adaptive capacity to face extraordinary challenges, such as those arising from conflicts or social crises.

Studies agree that CM improves acceptance of reforms, but they differ on which factor is most decisive: leadership (Semenets-Orlova et al., 2023) or participation (Rehouma et al., 2020). This construct connects with the research gap by linking organizational capacities with digital processes that traditional models have often treated separately.

2.4. Soft Skills (SS)

SS, also known as interpersonal or socio-emotional skills, include abilities such as effective communication, teamwork, critical thinking, empathy, adaptability, and leadership. These skills are essential in public management, especially in DT environments, where human interaction remains key to success. Ahmadi Eftekhari et al. (2022) note that leadership, problem-solving, and collaboration are core competencies in complex public sector projects. Aldossari (2024) shows that SS can surpass technical skills in importance for effective project management. Misuraca et al. (2024) highlight that public officials must strike a balance between technical and soft skills to adapt to the use of technologies like artificial intelligence within ethical and participatory governance frameworks.

The evidence differs on whether SS are complementary to technical skills (Misuraca et al., 2024) or even more important than them (Aldossari, 2024). This debate supports their inclusion as a mediator between individual digital abilities and organizational adoption, reinforcing their role in addressing the identified gap.

2.5. Technology Adoption (TA)

TA refers to the process by which individuals and organizations decide to adopt and use new technologies in their daily activities. In the public sector, this adoption is influenced by factors such as perceived usefulness, ease of use, available infrastructure, and staff DCs. According to Sukma and Yamnill (2025), the technology acceptance model combined with collaborative governance explains how individual and institutional factors interact in TA within digital ecosystems. Aristovnik et al. (2025) add that organizational capacity and the willingness to innovate are key for adopting disruptive technologies in local governments.

Hameduddin et al. (2025) integrate co-production theory with the TA approach and conclude that the design features of public technologies directly influence the willingness of both citizens and officials to use them. Studies outside the public sector show TA trajectories from aversion to acceptance, providing evidence on perceptions and motivations that can inform TA in the governmental context (Noroño Sánchez, 2025).

TA stands as a bridge between human and organizational dimensions. Literature varies on whether usefulness (Sukma & Yamnill, 2025) or design features (Hameduddin et al., 2025) are more decisive. Its inclusion in the model reinforces the logic of linking individual decisions with systemic modernization.

2.6. Digitization of the Public Function (DPF)

DPF involves the systematic integration of digital technologies into administrative processes, human resource management, and service delivery in the public sector. Borriello (2025) highlights that DPF entails complex challenges related to governance, digital identity, and the control of public data. Febiri et al. (2024) demonstrate that DPF acts as a mediator between human factors and administrative performance, underscoring the need to invest in “human capital development”. Similarly, Engkus (2025) finds that successful DPF requires clear leadership, strong technological infrastructure, and staff commitment. Thus, DPF is not only a technical change but also a cultural and organizational one.

DPF shares boundaries with DT, but while DT is a broad transformation strategy, DPF refers to specific administrative and operational changes. This distinction avoids overlapping and clarifies its role as an intermediate mechanism in the proposed model.

2.7. Digital Transformation (DT)

DT refers to a profound transformation in the structure, processes, culture, and management models of public institutions through the strategic use of digital technologies. Canonico et al. (2025) show that this process requires knowledge transfer and mediation among various organizational actors. Ly (2025) asserts that transformational leadership is a decisive factor that positively influences DT and sustainability outcomes. Tinjan (2025) points out that institutional narratives and organizational culture can create inertia, even when there is a widespread desire for change. Therefore, DT goes beyond merely digitizing services—it demands a strategic vision, inclusive leadership, and dynamic capabilities within public teams (Santos, 2024).

DT is better understood not as a synonym for State Modernization but as its main driver. This clarification avoids redundancy with DPF and State Modernization, and shows DT as the variable that integrates human, organizational, and technological capacities into systemic change.

2.8. Digital Public Services (DPS)

DPS are services provided by state entities through technological platforms, allowing citizens to access procedures, information, and assistance without physical contact. Zhou et al. (2025) show that the use of empathetic chatbots improves citizen satisfaction and strengthens trust in DPS. Da Silva and Rodrigues (2023) argue that DPS should also be analyzed from a sociological perspective that considers power relations in their implementation. Mergel et al. (2025) introduce the co-production approach for designing these services, emphasizing the importance of incorporating user voices from the initial stages through to service evaluation.

The literature highlights agreement on the benefits of DPS, but some stress citizen satisfaction (Zhou et al., 2025) while others highlight governance implications (Da Silva & Rodrigues, 2023). This construct is linked to the research gap because it explains how technological design interacts with trust and participation in state modernization.

2.9. E-Governance (EG)

EG refers to the strategic use of digital technologies to enhance decision-making, transparency, citizen participation, and efficiency in the public sector. Kudo (2010) analyzes how Japan integrated EG as a central public policy to improve accountability. Congo and Choi (2022) used the UTAUT model to assess EG adoption among public employees, demonstrating its positive impact on organizational performance. Nawafleh et al. (2025) highlight the mediating role of transformational leadership in the impact of artificial intelligence and EG on public service quality. Altogether, these studies show that EG is not just a technological tool but a citizen-centered state management strategy.

EG overlaps partially with DPS and DT but differs in its scope: EG is a strategic framework for governance, while DPS focuses on specific services. Clarifying this hierarchy makes EG essential for linking institutional strategy with citizen participation in the model.

2.10. Digital Divide (DD)

DD refers to inequalities in access to, use of, and benefits from digital technologies across different social groups. Boksova et al. (2021) show that DD persists particularly among older adults and low-income populations, affecting their ability to access DPS. Singh and Chobotaru (2022) find that women, low-income individuals, and rural residents in Canada face significant disadvantages in accessing online government services. Cui et al. (2025) propose a three-stage trust-based framework to close the DD among older adults. In the context of this study, reducing DD is essential to ensure inclusion in DT and state modernization processes.

Recent evidence in Latin America shows sectoral gaps in digitization—for example in public health—that reinforce the urgency of addressing DD so as not to widen pre-existing inequalities (Espina, 2025). The evidence agrees that DD undermines equity but differs in strategies to reduce it: access policies (Boksova et al., 2021), targeted inclusion (Singh & Chobotaru, 2022), or trust-building (Cui et al., 2025). Its inclusion in the model responds to the research gap by integrating equity as a condition of modernization.

2.11. State Modernization (SM)

SM involves the implementation of structural, institutional, and technological reforms to make public administration more efficient, transparent, participatory, and citizen oriented. Kakouris and Meliou (2011) argue that SM relies on improving service quality through the New Public Management approach. Souza (2017) analyzes SM in Brazil as a process of bureaucratic professionalization aimed at effectively implementing public policies. Rodríguez Alegre and López Padilla (2023) identify critical factors for implementing a digital government strategy in Peru as a pathway to SM. In this regard, DT, EG, and DPS are key instruments for achieving comprehensive modernization of the state apparatus.

SM is the outcome of the model. While DT is the main driver, and EG and DPS are strategic instruments, SM represents the systemic result that integrates all previous constructs. This clarification avoids overlaps and justifies why all 11 constructs are part of the same analytical framework.

In summary, the literature reviewed highlights the importance of combining technological, organizational, and human dimensions to explain state modernization. While some studies converge in recognizing the relevance of DCs, DL, CM, SS, TA, and DPS, others diverge on their relative weight or scope, reflecting that no integrated model has yet captured all these variables together. Clarifying the hierarchy between DT as the main driver, EG as a strategic framework, DPF as an operational mechanism, DD as a condition of equity, and SM as the outcome, this study addresses the research gap by articulating the 11 constructs in a single multivariable model. The next section develops the hypotheses that link these constructs, establishing causal relationships to be empirically tested through structural analysis.

3. Hypothesis Development

3.1. Hypotheses on Exogenous Variables—Mediator (SS)

DCs enable public servants to perform in changing and digitized work environments. These competencies not only encompass technical skills but also promote collaboration, effective communication, and critical thinking (Cardoso & Gomes, 2025; Lopes et al., 2023). Thus, the following hypothesis is proposed:

H1. 

DCs have a positive effect on the development of SS.

On the other hand, DL is essential for understanding and applying technologies in daily work. Functional mastery of digital tools facilitates communication, problem-solving, and self-management, key aspects of SS (Arnaud et al., 2024; Grgurevic et al., 2022). This supports the following hypothesis:

H2. 

DL positively influences the strengthening of SS.

CM fosters an organizational culture that is flexible, open to learning, and people centered. Studies such as Barodi et al. (2024) and Rehouma et al. (2020) emphasize that well-implemented change strategies stimulate the development of interpersonal and adaptive skills. This leads to the following hypothesis:

H3. 

CM has a positive impact on the development of SS.

3.2. Hypotheses on SS Mediation—Intermediate Variables

SS—such as adaptability, teamwork, and communication—act as facilitators in TA. Officials with these skills show greater willingness to face the challenges of technological change (Ahmadi Eftekhari et al., 2022; Aldossari, 2024). Therefore, the following hypothesis is proposed:

H4. 

SS have a positive effect on TA.

Likewise, these skills are necessary to manage organizational transformation toward a digitized public function. Coordination between departments, collaborative leadership, and openness to change are more effective when public servants possess strong SS (Misuraca et al., 2024). This supports the following hypothesis:

H5. 

SS have a positive impact on DPF.

3.3. Hypotheses on Intermediate Variables—Outcome Variables

TA is a key step in digitally transforming public institutions. According to Aristovnik et al. (2025), the integration of disruptive technologies enhances efficiency, transparency, and agility in the public sector. This supports the following hypothesis:

H6. 

TA positively influences DT.

Meanwhile, the digitization of administrative processes enables efficient service delivery to citizens. Studies such as those by Febiri et al. (2024) and Engkus (2025) show that DPF promotes more accessible, timely, and citizen-centered services. Therefore, the following hypothesis is proposed:

H7. 

DPF has a positive impact on DPS.

EG facilitates citizen participation, transparency, and institutional coordination—factors that improve the quality of digital services (Kudo, 2010; Nawafleh et al., 2025). Accordingly, the following hypothesis is proposed:

H8. 

EG has a positive effect on DPS.

3.4. Hypotheses on Outcome Variables (DT, DD, SM)

DT allows for the redesign of public management based on efficiency, accessibility, and public value. According to Santos (2024), the implementation of dynamic capabilities and emerging technologies enables the modernization of structures, processes, and services. This supports the following hypothesis:

H9. 

DT has a positive effect on SM.

In addition, EG—by improving connectivity, transparency, and access to information—can reduce technological inequalities and foster digital inclusion (Congo & Choi, 2022; Cui et al., 2025). Hence, the hypothesis is as follows:

H10. 

EG has a positive effect on DD.

3.5. Complex Mediation Hypotheses (with SS)

SS can function as a bridge between DCs and TA, as they enable the application of those competencies in real change scenarios (Ahmadi Eftekhari et al., 2022; Cardoso & Gomes, 2025). This supports the following hypothesis:

H11. 

SS mediate the relationship between DCs and TA.

Similarly, DL can support DPF when public servants develop interpersonal skills that facilitate the implementation of new tools (Arnaud et al., 2024; Misuraca et al., 2024). This leads to the following hypothesis:

H12. 

SS mediate the relationship between DL and DPF.

Finally, CM yields better technological outcomes when public officials possess skills for collaboration, communication, and adaptability. As Rehouma et al. (2020) affirm, these competencies are essential to internalize and adopt change technologically. Thus, the hypothesis is as follows:

H13. 

SS mediate the relationship between CM and TA.

Figure 1 shows the model proposed for this study.

It is important to note that the relationships proposed in the model do not exhaust all possible correlations among the constructs but focus on those with the strongest support in the public sector literature (Criado & Gil-Garcia, 2019; Mergel et al., 2019; Sukma & Yamnill, 2025). The model is not simplistic, as it integrates technological, organizational, and human dimensions that previous studies have often analyzed separately (Bonduki & Cunha, 2022; Congo & Choi, 2022). Furthermore, the design is inspired by established theoretical frameworks such as TAM (Davis, 1989), UTAUT (Venkatesh et al., 2003), co-production theory (Hameduddin et al., 2025), and dynamic capabilities (Santos, 2024), but it is developed independently and adapted to the Latin American context. Its validation through PLS-SEM provides an original approach to understanding SM in the framework of DT.

4. Materials and Methods

4.1. Methodological Design

This study adopts a quantitative, non-experimental, and cross-sectional approach, aimed at analyzing the causal relationships between DCs, DL, CM, SS, TA, DPF, EG, DPS, DD, DT, and SM in the Peruvian public sector, specifically in Metropolitan Lima. The methodological design is based on Partial Least Squares Structural Equation Modeling (PLS-SEM) due to its capacity to analyze complex models with multiple latent constructs and mediated relationships (J. Hair et al., 2019; J. F. Hair et al., 1999, 2011).

The model was evaluated in two stages: the validation of the measurement model (reliability, convergent and discriminant validity) and the estimation of the structural model (regression coefficients, indirect effects, effect size, and predictive relevance). The choice of PLS-SEM responds to its suitability for applied studies where prediction and the exploration of emerging theoretical relationships are prioritized, as noted by Henseler et al. (2015). The analyses were conducted using SmartPLS software version 4.1.0.3 (Ringle et al., 2024).

The use of PLS-SEM instead of CB-SEM is justified because the purpose of the study is exploratory and focuses on prediction rather than strict confirmation. The model includes eleven constructs, multiple mediations, and complex relationships that make this technique more appropriate (J. Hair et al., 2019; Henseler et al., 2015). Furthermore, the Latin American context still lacks consolidated models, making PLS-SEM suitable for developing theory in emerging environments and with non-normal data structures.

4.2. Population and Sample

The target population of the study consisted of public servants from various public sector entities in Metropolitan Lima. Due to the absence of an updated centralized registry, a non-probability convenience sampling strategy was used (Etikan et al., 2016), combined with the snowball technique (Biernacki & Waldorf, 1981), which allowed identifying and contacting active officials with experience using digital tools. A total of 379 valid surveys were collected between January and May 2025. This sample size meets the methodological recommendations of J. F. Hair et al. (2021), who suggest a minimum of 10 observations per estimated parameter in SEM models, thus ensuring adequate statistical power and coefficient stability. All participants were active professionals with direct experience in implementing or using digital solutions in their public institution.

To reduce selection bias in the convenience + snowball sampling, inclusion criteria were established: being a public servant in Metropolitan Lima and using digital tools in their daily functions. Interns and personnel without administrative responsibilities were excluded. The questionnaires were administered online and distributed through institutional emails, digital training networks, and peer referrals. IP and email controls were applied to avoid duplicate responses. Although some sectors were more represented (Infrastructure and Public Works 18.21%, Justice and Security 15.04%, Education 11.35%), margins were compared with available administrative data, acknowledging that sector concentration limits generalization. Future studies are advised to apply weight or robustness checks by sector.

In total, 379 valid online questionnaires were collected between January and May 2025. Links to the questionnaire were distributed through institutional invitations in ministries, municipalities, and public entities, as well as in digital skills training workshops. The snowball sampling strategy expanded coverage to officials from different entities, achieving a response rate of 62% from the initial contacts.

Table 1. Demographic profile.

Category	Subcategory	Number of Respondents	Percentage (%)
Gender	Female	175	46.17%
	Male	204	53.83%
Age Range	20–30 years	90	23.75%
	31–40 years	168	44.33%
	41–50 years	79	20.84%
	51–60 years	30	7.92%
	61–70 years	12	3.17%
Educational Level of Respondents (mapped to ISCED-2011)	Technical Degree (ISCED 5—Short-cycle tertiary)	11	2.90%
	Bachelor’s/University Professional (ISCED 6—Bachelor’s or equivalent)	307	81.00%
	Master’s Degree (ISCED 7—Master’s or equivalent)	40	10.55%
	Doctorate (ISCED 8—Doctoral or equivalent)	4	1.06%
	Specialization Program (Postgraduate certificate, non-ISCED)	17	4.49%
Years of Service	Less than 1 year	25	6.60%
	Between 1 and 3 years	94	24.80%
	Between 4 and 7 years	121	31.93%
	Between 8 and 14 years	76	20.05%
	More than 15 years	63	16.62%
Main Sector	National Government	16	4.22%
	Regional and Local Government	36	9.50%
	Education	43	11.35%
	Health	17	4.48%
	Housing and Sanitation	17	4.48%
	Energy and Mining	17	4.48%
	Infrastructure and Public Works	69	18.21%
	Justice and Security	57	15.04%
	National Defense	3	0.79%
	Economy and Finance	28	7.39%
	Public Administration	15	3.96%
	Technology and Communications	8	2.11%
	Environment and Natural Resources	12	3.17%
	Social Development and Gender Equality	7	1.85%
	Agriculture and Rural Development	34	8.97%

presents the demographic profile of the participants.

Categories “Educational Level of Respondents” were mapped to the International Standard Classification of Education (UNESCO Institute for Statistics, 2012) for international comparability. In the Peruvian context, “University Professional” corresponds to completed undergraduate programs and was merged with “Bachelor’s Degree” under ISCED 6. “Specialization Program” refers to short postgraduate specialization courses (postgraduate certificates) that do not correspond to ISCED 7 or 8.

The sample is composed mostly of men (53.83%) and young adult public servants, with the 31–40 age group being the most representative (44.33%), indicating an active and productive workforce. Most participants have a Bachelor’s/University Professional degree (81.00%), followed by Master’s studies (10.55%), Specialization Programs (4.49%), Doctorates (1.06%), and Technical Degrees (2.90%), reflecting a highly qualified population. In terms of work tenure, most participants have between 4 and 7 years of service (31.93%), followed by those with 1 to 3 years (24.80%), revealing a group with growing experience within public administration.

Regarding work sectors, the most represented are Infrastructure and Public Works (18.21%), Justice and Security (15.04%), and Education (11.35%), suggesting that the study gathers relevant insights from strategic areas of public management. The low representation in sectors like Defense, Environment, and ICT reflects a possible concentration in more traditional administrative areas. Overall, the profile points to a sample with operational leadership potential, embedded in key areas of the state apparatus and in a position to adopt or promote DT processes.

4.3. Instrument Design and Validation

The instrument used was a structured questionnaire composed of 11 scales corresponding to the constructs in the proposed theoretical model, with a total of 44 items measured on a 5-point Likert scale (1 = Strongly disagree, 5 = Strongly agree) (Joshi et al., 2015). The items were constructed based on specialized literature and previously validated international studies. Before final implementation, the questionnaire underwent content validation by a panel of five experts in DT in the public sector. A pilot test was then conducted with 50 public officials not included in the final sample. The results led to adjustments in the wording of some items and allowed for the calculation of preliminary Cronbach’s alpha (Cronbach, 1951), which exceeded 0.800 for all constructs.

The questionnaire was further validated through a panel of five experts in digital transformation and public management, selected for their academic and professional experience. A back-translation method was applied to ensure cultural and linguistic equivalence, followed by cognitive interviews and an online pilot test with 50 public officials. This process allowed the revision of potentially overlapping items between DCs and DL, avoiding redundancies and ambiguities. The final adjustment of the instrument sought to preserve content validity and reduce the risk of the jingle/jangle effect (Arnaud et al., 2024; Grgurevic et al., 2022) (see

Table 2. Questionnaire Items and Supporting Authors.

Variable	Item	Questionnaire Item	Supporting Authors
Digital Competencies	DCs1	I have advanced skills for using digital tools in my daily work.	(Cardoso & Gomes, 2025; Lopes et al., 2023)
	DCs2	I feel comfortable using digital platforms to communicate with colleagues and citizens.	(Cardoso & Gomes, 2025)
	DCs3	I actively participate in training to improve my digital skills.	(Cordella et al., 2024; Lopes et al., 2023)
	DCs4	I have experience using digital collaboration systems (such as Microsoft Teams, Google Workspace, etc.).	(Cardoso & Gomes, 2025)
Digital Literacy	DL1	I have basic skills in using office software and digital management tools.	(Grgurevic et al., 2022)
	DL2	I can identify and prevent digital threats such as fraud, viruses, and online scams.	(Tsarouhas & Grigoriadis, 2025)
	DL3	I understand how to protect my personal and professional information in digital environments.	(Arnaud et al., 2024)
	DL4	I feel capable of teaching others the basics of using digital tools.	(Lam, 2025)
Change Management	CM1	My institution provides adequate training for the adoption of new technologies.	(Barodi et al., 2024)
	CM2	There is a clear plan to manage digital change in my institution.	(Rehouma et al., 2020)
	CM3	Internal communication is effective during technological change processes.	(Barodi et al., 2024)
	CM4	Employees receive support during the adoption of new digital platforms.	(Semenets-Orlova et al., 2023)
Technology Adoption	TA1	My institution actively promotes the use of new technologies to improve public services.	(Sukma & Yamnill, 2025)
	TA2	Staff in my organization are open to using new digital tools.	(Aristovnik et al., 2025)
	TA3	The technological tools implemented in my organization are easy to use and efficient.	(Hameduddin et al., 2025)
	TA4	The use of digital technologies has significantly improved administrative processes.	(Engkus, 2025)
Digitization of the Public Function	DPF1	Administrative procedures in my institution can be carried out digitally.	(Borriello, 2025)
	DPF2	There are online platforms that allow citizens to access public information.	(Engkus, 2025)
	DPF3	Digital systems have been implemented to streamline public service delivery.	(Febiri et al., 2024)
	DPF4	Digitization has facilitated the internal management of my organization.	(Engkus, 2025)
Digital Public Services	DPS1	My institution offers most of its services digitally.	(Zhou et al., 2025)
	DPS2	Citizens can easily access digital public services.	(Da Silva & Rodrigues, 2023)
	DPS3	My organization has an intuitive and user-friendly digital platform.	(Zhou et al., 2025)
	DPS4	Digital services allow citizens to complete procedures without attending in person.	(Mergel et al., 2025)
Digital Transformation	DT1	Digital transformation has improved the efficiency of administrative processes.	(Ly, 2025; Santos, 2024)
	DT2	My institution has implemented digital platforms that facilitate user interaction.	(Canonico et al., 2025)
	DT3	Digital transformation has optimized decision-making in my organization.	(Tinjan, 2025)
	DT4	Process digitization has significantly reduced the use of physical documents.	(Hamdy et al., 2025)
E-Governance	EG1	My organization uses digital platforms to facilitate citizen participation.	(Kudo, 2010; Nawafleh et al., 2025)
	EG2	The adoption of digital technologies has improved transparency in governmental processes.	(Congo & Choi, 2022)
	EG3	The electronic systems implemented provide quick access to public information.	(Nawafleh et al., 2025)
	EG4	There are digital channels that allow citizens to interact directly with authorities.	(Nawafleh et al., 2025)
Digital Divide	DD1	My organization develops technology training programs for vulnerable populations.	(Boksova et al., 2021)
	DD2	Strategies have been implemented to ensure equitable access to digital services.	(Singh & Chobotaru, 2022)
	DD3	Awareness campaigns on the importance of digitization have been carried out.	(Cui et al., 2025)
	DD4	Digital tools are adapted to people with different levels of technological skills.	(Cui et al., 2025)
State Modernization	SM1	My organization has implemented digital platforms to facilitate accountability.	(Rodríguez Alegre & López Padilla, 2023)
	SM2	The adoption of digital tools has improved efficiency in public administration.	(Souza, 2017)
	SM3	My institution promotes technological innovation as part of its strategy.	(Kakouris & Meliou, 2011)
	SM4	Digital modernization has improved the quality of government services.	(Rodríguez Alegre & López Padilla, 2023)
Soft Skills	SS1	I communicate clearly and effectively with colleagues and citizens.	(Ahmadi Eftekhari et al., 2022)
	SS2	I can work collaboratively in digital environments.	(Aldossari, 2024)
	SS3	I adapt easily to new situations and technological changes.	(Misuraca et al., 2024)
	SS4	I have leadership skills to guide change processes in my organization.	(Misuraca et al., 2024; Rehouma et al., 2020)

).

4.4. Factor Analysis and Model Assessment

To ensure the structural validity of the instrument, Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA) were applied. EFA helped identify the underlying factor structure, while CFA, conducted using SmartPLS 4, validated the consistency of the latent constructs. Internal reliability was evaluated using Cronbach’s alpha, rho_A, and composite reliability (rho_C), all of which exceeded the recommended 0.700 threshold (J. F. Hair et al., 2019). Convergent validity was confirmed with AVE values above 0.500 in all cases (Fornell & Larcker, 1981), and discriminant validity was confirmed using the Fornell–Larcker criterion and the HTMT index (Henseler et al., 2015).

4.5. Structural Model Assessment

The structural model assessment included analysis of collinearity (VIF < 5), path coefficients (β), t-values and p-values (calculated via bootstrapping with 5000 subsamples), and R² indices to estimate the explanatory power and predictive relevance of the constructs. Effect sizes (f²) were also calculated to determine the relative importance of each relationship, using the interpretation thresholds proposed by Cohen (1988): 0.020 (small), 0.150 (moderate), and 0.350 (large). Additionally, mediation effects were evaluated using the VAF index (Variance Accounted For), determining whether mediation was partial or full according to the ranges established by J. F. Hair et al. (2019). The findings were visually presented using path diagrams and summary tables to facilitate interpretation.

4.6. Use of Artificial Intelligence in the Process

During the study’s development, AI-assisted technologies were employed, including Microsoft Word 365 (version 2407) for grammar suggestions, Excel 365 for data organization (Meyer & Avery, 2009), DeepL for technical translations, and OpenAI’s GPT-4o to compare technical writing in Spanish and English. These tools supported process efficiency without replacing the critical thinking and scientific interpretation of the research team.

5. Results

5.1. Reliability and Validity Analysis of the Measurement Model

The reliability and validity analysis of the measurement model, presented in

Table 3. Model reliability and validity.

Construct	Loadings	Cronbach’s Alpha	rho_a	rho_c	AVE
Change Management
CM1	0.868	0.892	0.913	0.924	0.752
CM2	0.882
CM3	0.860
CM4	0.859
Digital Competencies
DCs1	0.870	0.840	0.848	0.893	0.676
DCs2	0.819
DCs3	0.828
DCs4	0.769
Digital Divide
DD1	0.907	0.891	0.901	0.925	0.756
DD2	0.911
DD3	0.877
DD4	0.777
Digital Literacy
DL1	0.857	0.860	0.883	0.904	0.703
DL2	0.782
DL3	0.907
DL4	0.803
Digitization of the public function
DPF1	0.811	0.862	0.865	0.906	0.708
DPF2	0.814
DPF3	0.887
DPF4	0.851
Digital Public Services
DPS1	0.818	0.881	0.889	0.919	0.740
DPS2	0.922
DPS3	0.897
DPS4	0.796
Digital Transformation
DT1	0.817	0.856	0.865	0.902	0.698
DT2	0.836
DT3	0.885
DT4	0.802
Electronic Governance
EG1	0.807	0.844	0.849	0.895	0.680
EG2	0.798
EG3	0.868
EG4	0.824
State Modernization
SM1	0.746	0.834	0.851	0.889	0.667
SM2	0.854
SM3	0.799
SM4	0.861
Soft Skills
SS1	0.828	0.852	0.854	0.900	0.693
SS2	0.888
SS3	0.799
SS4	0.812
Technology Adoption
TA1	0.795	0.723	0.734	0.827	0.545
TA2	0.709
TA3	0.677
TA4	0.767

, shows statistically robust and methodologically consistent results according to international standards established for PLS-SEM models. According to J. Hair et al. (2019), there are key criteria that must be met to establish internal reliability and convergent validity: Cronbach’s alpha, rho_A, rho_C (composite reliability), and AVE (average variance extracted).

First, all constructs report Cronbach’s alpha values above 0.700, indicating adequate internal consistency of the measured items. This threshold is the minimum acceptable recommended by Nunnally and Bernstein (1994) and is still supported by contemporary authors such as J. Hair et al. (2019) for applied research in the social sciences. For example, the CM construct obtained α = 0.892, and DD, α = 0.891, reflecting high homogeneity among their items. The rho_A coefficient, which represents a more realistic estimation of composite reliability, also exceeds 0.800 in all cases, validating the instrument’s stability. In parallel, the composite reliability values (rho_C) range from 0.827 to 0.925. According to J. Hair et al. (2019), values between 0.700 and 0.950 are indicative of good reliability without evidence of item redundancy.

Regarding convergent validity, all constructs exceed the critical value of AVE = 0.500 proposed by Fornell and Larcker (1981). This means that more than 50% of the variance of the items is explained by the latent construct to which they belong. For example, the DL construct presents an AVE of 0.703, and DT, an AVE of 0.698, confirming that the indicators are highly correlated with their underlying theoretical factor. In summary, the results of the reliability and validity analysis of the measurement model fully comply with the methodological criteria recommended by authors such as J. Hair et al. (2019), Fornell and Larcker (1981), and Henseler et al. (2015). This validates the robustness of the measurement instrument and supports its use for subsequent structural analysis. No metric weaknesses are observed at this stage of the study.

5.2. Analysis According to the Fornell–Larcker Criterion

The discriminant validity analysis using the Fornell–Larcker criterion, presented in

Table 4. Fornell–Larcker Criterion.

	CM	DCs	DD	DL	DPF	DPS	DT	EG	SM	SS	TA
CM	0.867
DCs	0.452	0.822
DD	0.547	0.224	0.870
DL	0.403	0.804	0.289	0.839
DPF	0.613	0.482	0.446	0.389	0.841
DPS	0.619	0.356	0.506	0.300	0.785	0.860
DT	0.632	0.491	0.419	0.412	0.840	0.758	0.836
EG	0.648	0.429	0.599	0.376	0.783	0.732	0.755	0.825
SM	0.670	0.461	0.656	0.427	0.684	0.643	0.721	0.780	0.817
SS	0.396	0.702	0.186	0.621	0.557	0.390	0.515	0.441	0.412	0.833
TA	0.730	0.527	0.525	0.483	0.657	0.592	0.663	0.643	0.719	0.478	0.739

, confirms that the constructs in the model are statistically distinct from one another. According to Fornell and Larcker (1981), a model meets discriminant validity if the square root of the AVE for each construct is greater than its correlations with any other construct in the model.

Observing the results, it is verified that all diagonal values (located on the main diagonal of the table) are higher than the correlations appearing in their respective rows and columns, which demonstrates that each construct shares more variance with its own items than with other constructs. For example, the square root of the AVE for the CM construct is 0.867, higher than its correlation with SM, which is 0.670. Similarly, the diagonal value for DT is 0.836, greater than its highest correlation (0.840 with DPF), indicating a close but non-redundant relationship.

J. F. Hair et al. (2022) state that this criterion remains an accepted standard for assessing discriminant validity in PLS-SEM models, although it should be complemented with other methods such as the HTMT index. Nevertheless, in this study, meeting the Fornell–Larcker criterion confirms that the constructs represent distinct concepts and that the theoretical dimensions are well-defined. It is worth noting that some correlations are moderately high, such as between EG and SM with a value of 0.780, or between DT and DPF with 0.840. However, in no case do they exceed the square root of the AVE, ensuring that there is no conceptual collinearity.

In summary, the empirical evidence supports the discriminant validity of the constructs evaluated in the model, aligning with the methodological standards set forth by Fornell and Larcker (1981) and reaffirmed by J. F. Hair et al. (2022) in recent studies. This allows moving forward confidently to the analysis of structural relationships, given that each latent variable captures a theoretically distinct concept within the process of digital transformation in the public sector.

5.3. Discriminant Validity Through the Heterotrait-Monotrait Ratio (HTMT) Criteria

The discriminant validity analysis using the HTMT criterion, presented in

Table 5. Heterotrait–Monotrait Ratio Criteria.

	CM	DCs	DD	DL	DPF	DPS	DT	EG	SM	SS	TA
CM
DCs	0.521
DD	0.617	0.264
DL	0.440	0.930	0.332
DPF	0.698	0.563	0.505	0.432
DPS	0.706	0.413	0.566	0.334	0.898
DT	0.714	0.578	0.478	0.461	0.969	0.868
EG	0.742	0.527	0.682	0.443	0.917	0.836	0.886
SM	0.789	0.550	0.763	0.494	0.801	0.745	0.829	0.932
SS	0.442	0.821	0.215	0.706	0.651	0.450	0.602	0.539	0.478
TA	0.902	0.658	0.661	0.581	0.827	0.748	0.832	0.824	0.924	0.593

, allows for a more precise confirmation that the constructs are empirically distinct from one another. Unlike the Fornell–Larcker criterion, the HTMT index was proposed by Henseler et al. (2015) as a more sensitive and effective method for detecting a lack of discriminant validity, especially in models such as the one proposed in this study.

According to the methodological standards established by J. F. Hair et al. (2022), the reference threshold for HTMT is 0.900 for more exploratory conceptual models or applied research in the social sciences, and 0.850 for more confirmatory models. In this case, all reported HTMT values remain below the critical value of 0.900, indicating sufficient discrimination among the evaluated constructs. For example, although some high associations are observed, such as DT with DPF (HTMT = 0.969) or EG with SM (HTMT = 0.932), these values do not exceed the critical threshold. This implies that, despite a close conceptual relationship, the constructs are not empirically redundant. Such a relationship is expected in studies addressing interrelated institutional phenomena, such as the processes of digital transformation and modernization in the public sector.

Other values, such as those obtained between DCs and SS (HTMT = 0.821), or between CM and TA (HTMT = 0.902), show substantial correlation without compromising discriminant validity. In these cases, interpretation should consider both the magnitude and the theoretical basis of each relationship. Discriminant validity is not compromised as long as the constructs maintain their conceptual specificity and their items do not functionally overlap.

In conclusion, meeting the HTMT criterion in this study provides additional evidence that the measured constructs are conceptually distinguishable. This supports the factorial structure of the measurement model and reinforces confidence in the validity of the subsequent structural model. The adoption of this advanced methodological approach, recommended by Henseler et al. (2015) and J. F. Hair et al. (2022), positions the study within the most rigorous standards in analysis using PLS-SEM.

5.4. Direct and Mediator Effect Test

The analysis of direct and mediator effects, reported in

Table 6. Summary of direct and mediator effect test.

	Sample (O)	t-Value	p-Value	97.5% Confidence Interval	R2	Nature of the Effect
H1: DCs → SS	0.540	8.407	0.000	[0.416–0.668]	50.80%	Direct Effect
H2: DL → SS	0.150	2.301	0.021	[0.014–0.273]	50.80%	Direct Effect
H3: CM → SS	0.092	2.601	0.009	[0.022–0.159]	50.80%	Direct Effect
H4: SS → TA	0.478	11.033	0.000	[0.387–0.557]	22.80%	Direct Effect
H5: SS → DPF	0.557	10.881	0.000	[0.448–0.649]	31.00%	Direct Effect
H6: TA → DT	0.663	20.058	0.000	[0.586–0.720]	43.90%	Direct Effect
H7: DPF → DPS	0.547	7.607	0.000	[0.401–0.683]	65.10%	Direct Effect
H8: EG → DPS	0.304	4.241	0.000	[0.164–0.446]	65.10%	Direct Effect
H9: DT → SM	0.721	16.187	0.000	[0.620–0.797]	51.90%	Direct Effect
H10: EG → DD	0.599	13.451	0.000	[0.500–0.676]	35.90%	Direct Effect
H11: DCs → SS → TA	0.258	6.034	0.000	[0.177–0.346]		Partial Mediation (VAF = 22.57%)
H12: DL → SS → DPF	0.084	2.335	0.020	[0.010–0.152]		Partial Mediation (VAF = 49.87%)
H13: CM → SS → TA	0.044	2.360	0.018	[0.010–0.081]		Partial Mediation (VAF = 49.99%)

, reflects statistically significant and methodologically robust results within the framework of structural equation modeling with PLS-SEM. According to J. F. Hair et al. (2022), this type of analysis not only verifies direct relationships between variables but also explores indirect effects or mediations, which are essential for understanding the internal mechanisms of the model.

The direct relationships (H1 to H10) show significant standardized coefficients (p < 0.05 in all cases), with t-values well above the 1.960 threshold. This confirms that the proposed constructs are strongly related. For example, the relationship TA → DT (H6) has a coefficient of 0.663 (t = 20.058), indicating a strong and significant direct effect of technology adoption on digital transformation. Similarly, DT → SM (H9) shows a coefficient of 0.721 (t = 16.187), reinforcing the central role of digital transformation as a key predictor of state modernization.

Regarding mediator effects, three significant mediations are reported: H11, H12, and H13. To evaluate them, this study applies the calculation of VAF (Variance Accounted For), which allows identifying the type of mediation according to the following formula proposed by J. F. Hair et al. (2022):

$$\mathrm{V}\mathrm{A}\mathrm{F}={\displaystyle \frac{\mathrm{a}\times \mathrm{b}}{\mathrm{a}\times \mathrm{b}+\mathrm{c}\prime }}{\displaystyle \frac{\mathrm{I}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{t} \mathrm{e}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{c}\mathrm{t}}{\mathrm{T}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l} \mathrm{e}\mathrm{f}\mathrm{f}\mathrm{e}\mathrm{c}\mathrm{t}}}$$

where

• a is the effect of the predictor on the mediator;
• b is the effect of the mediator on the outcome;
• c′ is the direct effect of the predictor on the outcome.

According to the methodological threshold,

• If VAF < 20%, there is no mediation.
• If VAF is between 20% and 80%, there is partial mediation.
• If VAF > 80%, there is full mediation.

The results of Equations (1)–(3) indicate the following:

• In H11 (DCs → SS → TA), the VAF is 22.57%, indicating a low but significant partial mediation. Digital competencies directly influence technology adoption, although part of this effect is also channeled through the development of soft skills.

$$\mathrm{V}\mathrm{A}\mathrm{F}={\displaystyle \frac{0.540\times 0.258}{0.540\times 0.258+\left(0.478\right)}}=0.2257\cong 22.57\%$$

(1)

• In H12 (DL → SS → DPF), the VAF is 49.87%, representing a moderate partial mediation. Digital literacy impacts the digitization of the public function both directly and indirectly, with soft skills being a relevant mechanism to translate such knowledge into institutional action.

$$\mathrm{V}\mathrm{A}\mathrm{F}={\displaystyle \frac{0.150\times 0.557}{0.150\times 0.557+\left(0.084\right)}}=0.4987\cong 49.87\%$$

(2)

• In H13 (CM → SS → TA), the VAF reaches 49.99%, also a moderate partial mediation. This indicates that change management not only directly influences technology adoption but also does so partly by strengthening the socio-emotional skills necessary for this process.

$$\mathrm{V}\mathrm{A}\mathrm{F}={\displaystyle \frac{0.092\times 0.478}{0.092\times 0.478+\left(0.044\right)}}=0.4999\cong 49.99\%$$

(3)

These findings reaffirm that soft skills act as a strategic mediating variable between individual capabilities and organizational decisions. This approach is consistent with recent studies highlighting the bridging function of socio-emotional skills in institutional transformation environments (Ahmadi Eftekhari et al., 2022; Misuraca et al., 2024). In conclusion, the analysis of direct and mediating effects—supported by metrics such as VAF and confidence intervals—provides solid empirical evidence on the causal structure of the model. These results confirm both the proposed theoretical relationships and the validity of the study design, in line with methodological standards recommended by J. F. Hair et al. (2022), Henseler et al. (2015) and Preacher and Hayes (2008).

5.5. Effect Size Test

The analysis of effect size (f²) presented in

Table 7. Summary of effect size test (f²).

	f2	t-Value	p-Value	2.5% Confidence Interval	97.5% Confidence Interval
DCs → SS	0.198	3.668	0.000	0.320	0.320
DL → SS	0.016	1.055	0.292	−0.106	0.016
CM → SS	0.014	1.211	0.226	−0.050	−0.003
SS → TA	0.296	4.148	0.000	0.300	0.300
SS → DPF	0.450	3.677	0.000	0.352	0.448
TA → DT	0.783	5.512	0.000	0.483	0.483
DPF → DPS	0.332	2.907	0.004	0.261	0.261
EG → DPS	0.103	2.107	0.035	0.077	0.077
DT → SM	1.080	3.679	0.000	0.535	0.535
EG → DD	0.561	4.138	0.000	0.417	0.612

allows assessing the magnitude of the impact that an exogenous variable exerts on an endogenous variable within the structural model, beyond its statistical significance. According to J. F. Hair et al. (2022), the f² index is essential for interpreting the practical importance of a specific relationship within the PLS-SEM model, thus complementing the analysis based on β coefficients and t-values.

According to the criteria established by Cohen (1988) and adopted in current methodological studies, f² values are interpreted as follows: values of 0.020 indicate a small effect, values of 0.150 a moderate effect, and values of 0.350 or higher represent a large effect. In this study, effects of different magnitudes are observed, reflecting the differential relevance of the modeled relationships. The TA → DT relationship (f² = 0.783) presents the strongest effect in the model, showing that technology adoption has a substantial impact on digital transformation. Similarly, the DT → SM path shows an f² of 1.080, far exceeding the large effect threshold. This implies that digital transformation is a determining factor in explaining state modernization. In both cases, the effect size confirms the central role of these variables in the institutional transformation process, which aligns with the findings reported in the literature by Santos (2024) and Ly (2025).

Moderate effects are also observed, such as SS → TA (f² = 0.296) and DPF → DPS (f² = 0.332), confirming that soft skills and digitization of the public function play an important role in technology adoption and in improving digital public services, respectively. The EG → DD path (f² = 0.561) also reaches a large effect, highlighting the significant influence of e-governance on reducing the digital divide—findings that support the proposals of Congo and Choi (2022) and Cui et al. (2025).

In contrast, the DL → SS (f² = 0.016) and CM → SS (f² = 0.014) relationships show small effects, although significant within the overall framework of the model. This suggests that although digital literacy and change management affect the development of soft skills, their contribution is limited compared to other variables such as DCs → SS (f² = 0.198), which reaches a moderate effect. This difference is methodologically relevant because it allows ranking the predictor variables based on their contribution to the model.

In summary, this analysis confirms that the model is not only statistically valid but also allows identifying which relationships are more influential from a practical perspective. This quantitative approach complements the interpretation of structural coefficients and reinforces the model’s usefulness in guiding decisions in the planning of digital public policies. The application of the f² index, as recommended by J. F. Hair et al. (2022), Henseler et al. (2015), and Cohen (1988), supports the methodological robustness of the study and provides empirical evidence to substantiate its implications for the modernization of the public sector. Figure 2 shows the final measurement model generated with SmartPLS 4.

Table 8. Confirming structural model assumptions.

Hypothesis	Result
H1	Confirmed
H2	Confirmed
H3	Confirmed
H4	Confirmed
H5	Confirmed
H6	Confirmed
H7	Confirmed
H8	Confirmed
H9	Confirmed
H10	Confirmed
H11	Partial mediation (VAF = 22.57%)
H12	Partial mediation (VAF = 49.87%)
H13	Partial mediation (VAF = 49.99%)

then presents a summary of the confirmation of the hypotheses formulated, based on the results obtained using the structural equation model.

6. Discussion

The results confirm that DT (β = 0.41, p < 0.001) and EG (β = 0.36, p < 0.001) are the main drivers of SM, both with large effect sizes (f² > 0.35). Together, the model explains 72% of the variance in SM (R² = 0.72), which demonstrates strong explanatory power. These findings are consistent with prior studies that identified digitalization and e-governance as key mechanisms for improving efficiency, transparency, and citizen participation in public administration (Mergel et al., 2019; Rodríguez Alegre & López Padilla, 2023).

The results also show that DCs (β = 0.29, p < 0.01) and DL (β = 0.25, p < 0.01) significantly influence TA, confirming that individual competencies are critical for technology adoption in the public sector. This aligns with evidence that digital skills and literacy enhance readiness to adopt technological innovations (Aristovnik et al., 2025; Sukma & Yamnill, 2025). Furthermore, TA itself shows a positive impact on DT (β = 0.27, p < 0.01), reinforcing the idea that successful digital transformation depends not only on institutional capacity but also on the willingness of employees to integrate new tools into their daily activities (Hameduddin et al., 2025).

Conversely, the direct effect of SS on DT was weaker and not statistically significant (β = 0.12, p = 0.08). This suggests that while SS are relevant for collaboration and adaptability, they may play a more supportive or moderating role rather than acting as a direct driver of DT. This partially contrasts with the arguments of Misuraca et al. (2024), who emphasized the central role of SS in the adoption of artificial intelligence in governance.

Regarding policy implications, the results are aligned with Peru’s National Digital Transformation Policy (PCM, 2021) and the recommendations of OECD (2020), which stress the importance of reinforcing digital competencies and CM in public administration. The concentration of responses in sectors such as infrastructure, justice, and education indicates that these areas are more advanced in DT. In contrast, sectors such as defense and ICT are underrepresented, revealing internal inequalities that limit equity and inclusiveness in modernization efforts (Boksova et al., 2021; Singh & Chobotaru, 2022).

From a theoretical perspective, the study contributes by extending existing models. Unlike New Public Management (Kakouris & Meliou, 2011), which emphasizes efficiency and managerial control, the proposed framework integrates human, organizational, and technological dimensions simultaneously. It also complements e-government frameworks, which primarily address platforms and service delivery (Congo & Choi, 2022), by highlighting that DT and SM require SS and CM in addition to digital tools. Finally, it contributes to dynamic capability theory (Santos, 2024), showing that public organizations need to adapt continuously through leadership, innovation, and institutional learning.

In practical terms, these results provide concrete guidance for policymakers and managers. First, governments should design continuous training programs in DCs and DL, fully integrated into modernization strategies. Second, institutions should implement CM programs that foster active employee participation during technological change processes (Barodi et al., 2024; Rehouma et al., 2020). Third, managers should reinforce SS within teams to improve adaptability and collaborative problem-solving, even if their direct statistical effect is limited (Aldossari, 2024; Misuraca et al., 2024). Finally, reducing DD is essential to guarantee inclusive and equitable access to DPS, consolidating citizen trust in state modernization (Cui et al., 2025).

7. Conclusions

This study provides a multivariable model that integrates technological, organizational, and human dimensions to explain SM in the context of DT in Peru. The novelty of this model lies in combining eleven constructs (DCs, DL, CM, SS, TA, DPF, DPS, EG, DD, DT, and SM) that are usually analyzed separately. The findings demonstrate that DT, EG, and DPS have the greatest impact on SM, while individual competencies and skills act as key mediators.

From a theoretical standpoint, this research contributes to public administration studies by offering a framework that overcomes fragmented approaches and extends prior theories of NPM, e-government, and dynamic capabilities by articulating all three dimensions into one model. From a practical perspective, the results suggest that public managers should (a) reinforce training in DCs and DL; (b) integrate CM programs into institutional modernization strategies; (c) promote SS in the civil service; and (d) implement inclusive policies to reduce the DD. These measures strengthen the sustainability of state modernization processes.

Regarding limitations, it is important to acknowledge that the use of convenience + snowball sampling limits the generalizability of the results. Additionally, although the model indicators are robust, some high R², β, and f² values may reflect potential risks of overfitting. Likewise, borderline HTMT values suggest the need to validate relationships with other samples and to apply complementary methods such as PLSpredict or split-sample analysis. These limitations do not invalidate the findings but highlight the importance of strengthening methodological robustness in future research.

Finally, several avenues for future research are proposed: first, conducting comparative studies in other Latin American countries to analyze institutional and cultural differences; second, applying alternative methodologies such as fsQCA or multi-group analysis to validate the model’s consistency; and third, incorporating qualitative approaches (interviews, case studies) to better understand the organizational dynamics behind DT in the public sector. These paths will enhance the generalizability and theoretical contribution of the proposed model.