Territorial patterns of open e-government: evidence from Chilean municipalities

ABSTRACT

This article analyses the development of open e-government between 2019 and 2021 in Chile’s 345 municipalities. We aggregated an e-government index (EGi) to measure the provision of local digital services for citizens. We then combined this with indicators of transparency and access to public information to create an open e-government index (OEGi). Our empirical strategy is based on geospatial econometric analysis in two stages: first, we describe and georeference our index, estimating the level of spatial autocorrelation and then fit different econometric models to measure the impact of the degree of Internet use, socioeconomic dynamism and management capacity on the municipalities’ development of open e-government. Our main findings indicate that monetary poverty has a negative effect on the index, while the municipal government’s budget has a positive effect.

KEYWORDS:

• Local governments
• e-government
• open government
• transparency
• management capacity

Introduction

On 4 September 2013, the Municipal Council of the town of Algarrobo (Valparaíso Region, Chile) agreed, with the unanimous approval of the councillors present,¹ that, between 8 and 22 October 2013, five councillors and an official from the municipal government’s Directorate of Infrastructure would attend the International Seminar on Municipal Management in Spain, Italy and France. The cost to the municipal government in fares and other travel expenses was over 40 million pesos (some US$40,000). Three years later, following a television report entitled ‘Councillors on Tour’, the case became a public scandal. The report found that, between 2014 and 2015, municipal governments had spent over 8,000 million pesos (around US$8 million) on training, in Chile and abroad, for councillors and mayors, despite the limits placed on activities of this type by Chilean administrative jurisprudence.

This case serves as an excellent example of the situation of open government at the municipal level. In it, journalists from a television channel used Chilean Law 20 285 on Transparency and Access to Public Information, enacted in 2008, to investigate the conduct of the country’s 345 municipal governments as regards their spending on training for mayors and councillors. However, it could have been a resident of any municipal district who requested the information, and the outcome would still have permitted the identification of the five municipal governments that spent the most on this type of training between 2014 and 2015: Lampa (200 million pesos), Ñuñoa (181 million), Alto Hospicio (165 million), Chimbarongo (161 million) and Calbuco (153 million). The list does not discriminate by political party, geographical location or size of the municipal budget.

It is precisely some of these characteristics that the literature often identifies as determinants of the success of open and electronic government at the local level. As well as the size of the municipal budget, the income level of the district’s inhabitants tends to be relevant (Andrews and Entwistle 2015; Andrews et al. 2013; Boyne 1995; Moon 2002; Pina, Torres, and Royo 2010). Higher indices of economic development are, in general, positively correlated with household Internet access and demand for access to public information, which may, in turn, foster the development of open e-government (Alcaide-Muñoz, López Hernández, and Caba-Pérez 2014; Lowatcharin and Menifield 2015).

In this context, our main question is: What factors determine the development of open e-government in Chilean municipalities? To answer this question, we constructed an index that combines elements of e-government with transparency indicators for all Chilean municipalities. This represents both an empirical and theoretical contribution in that we approach electronic and open government, concepts that, albeit interconnected, are usually studied separately. We adapted the e-services model of Esteves (2005), which comprises a number of items grouped into five dimensions to assess government websites (Fath-Allah et al. 2017) and the dimensions of Dias’ (2020) local e-government empirical model. We carried out measurements between 2019 and 2021 in the 345 Chilean municipalities to create an e-government index (EGi), into which we incorporated transparency indicators to create an open e-government index (OEGi). This aggregation reflects how e-government and open government have gone through a complementary process during the last decade. With our index, we evaluated digital development at the local level and then analysed its determinants using geospatial econometric models, which allows us to account for territorial differences and diffusion of innovation at the local level.

Answering our question and conducting this research using the Chilean case is relevant due to its leadership in e-government, open government and transparency in Latin America. An example of this is implementing a number of digital platforms to increase citizen participation and transparency as of the abovementioned Law 20 285, one of the first regulations in Latin America to strengthen a legal framework to access public information 15 years ago. These characteristics make Chile a relevant case for scholars and practitioners interested in understanding how diverse initiatives can effectively be implemented in a developing country.

This article is divided into four sections. The first offers a theoretical reflection on open e-government before, in the second, addressing the determinants of digital development and presenting our empirical expectations. The third section, of a methodological nature, provides information about the construction of the indices, the measurement and operationalisation of the variables and the econometric strategy. The fourth section presents the main results of the georeferencing process and the econometric estimates and, finally, is followed by a brief discussion that returns to the initial question and our empirical expectations in light of the evidence presented.

Unpacking conceptualisations on electronic and open government

The term electronic government began to be used in the late 1990s, related mainly to transparency and the efficient use of information and communications technologies (ICT) in the public sector (Barría, González-Bustamante, and Araya 2017; González-Bustamante, Carvajal, and González 2020; Concha and Naser 2012). Indeed, according to Dias (2019b), the first indexed article on e-government was published by the end of the 1990s, however, the term is older and could be tracked to U.S. government documents at the beginning of that decade. Although significant research has been published since then, the concept is dynamic and could be associated with various meanings (Dias 2019b; Relyea 2002).

Dias’ (2019b) bibliometric analysis of e-government research describes the term as a fuzzy construct that could be conceptualised as the study of governments’ ICT usage in different dimensions. This work is key to understanding how the topics usually studied are interconnected since it identified the main lines of research in the Ibero-American agenda and compared them with the most relevant worldwide research. The topics of e-participation and transparency are more common in Ibero-America. In contrast, the rest of the world tends to focus on issues such as quality of service, security, privacy and policy adoption. In particular, e-participation, transparency and local governments are transversal topics for the first three of the four clusters of countries identified by Dias (2019b).²

Indeed, the study of e-government has grown considerably in the past decade, with annual conferences and specialised journals on the subject, as indicated by Gil-García and Catarrivas (2017). However, most lines of research have focused on the provision of digital services and participation or electronic democracy (i.e. e-participation or e-democracy; see Barría, González-Bustamante, and Araya 2017; Criado 2004; Prieto-Martín 2012). The provision of e-services is related to the information supplied and consumed through electronic channels, particularly the Internet (González-Bustamante, Carvajal, and González 2020). In this framework, a co-creation process is generated between the public administration and citizens to create public value (Gatautis et al. 2015). In turn, e-democracy is understood as the incorporation of technologies in decision-making (Barría, González-Bustamante, and Araya 2017; Vicente and Novo 2014).

The sustained increase in Internet penetration has fostered the consolidation of e-services, especially government portals that offer users personalised attention for official procedures, queries or complaints. There are different models for evaluating government portals, for example, Fath-Allah et al. (2017; see also González-Bustamante, Carvajal, and González 2020) evaluate 25 recent models found in the academic literature. On this basis, it is possible to identify five phases of digital maturity: (i) presence, which is limited to the provision of information; (ii) interaction, where citizens have channels to interact with government bodies; (iii) transaction, where exchanges can take place; (iv) integration; and (v) participation (Luna 2017; González-Bustamante, Carvajal, and González 2020; Sandoval-Almazán and Gil-García 2008).

In line with the stages identified by Fath-Allah et al. (2017) and the components of the evolutionary model, Esteves (2005) developed a model of e-services grouped into five dimensions. It has been adapted to study municipal governments in Argentina, Chile and Venezuela (Bayona and Morales 2017; Gómez 2007; González-Bustamante, Carvajal, and González 2020). These dimensions represent phases of digital maturity at the local level: (i) presence, in which services are based on municipal information that is available online using basic search systems; (ii) urban information, which provides territorial information about the municipal district and means of transport; (iii) interaction, which permits simple communication between the municipal government and citizens; (iv) transaction, which includes electronic interactions such as paying for services and applying for permits; and (v) e-democracy, a phase that includes instances of digital participation and electronic voting on issues of local interest. However, most interactions between citizens and governments have not progressed beyond the informative early stage (Sandoval-Almazán and Gil-García 2012).

Consequently, we focus on a classic operational definition of local e-government, in line with Dias (2019a, 2020), which relates to the study of how local governments use ICT to deliver public services and engage citizens. In this way, we are focusing on an important line of research in the Ibero-American agenda related to the study of the subnational levels. We connect this with other relevant topics in the study of e-government, such as e-participation, since our empirical strategy adapts Esteves’ (2005) e-services model, which incorporates e-democracy as its highest service component. However, at this point, we nuanced our argument by incorporating transparency and open government elements into our empirical analysis. In this way, we connect the three cross-cutting research topics in e-government publications in Ibero-America between 2003 and 2017 (Dias 2019b).

In particular, open government is associated with the release of public information on the Internet to strengthen transparency and is, therefore, linked to concepts such as transparency, participation and e-government (Barría, González-Bustamante, and Cisternas 2019; Dawes 2010; González-Bustamante, Carvajal, and González 2020; Ramírez-Alujas 2012; Rodríguez-Bolívar, Alcaide-Muñoz, and López 2012; Valenzuela 2014). In this sense, the Memorandum on Transparency and Open Government, released by the government of President Barack Obama in January 2009 as a way to strengthen democracy and promote his administration’s efficiency, summarises in three principles what it means to adhere to this culture of governance: transparency (to know), participation (to take part) and collaboration (to contribute) (Barría, González-Bustamante, and Cisternas 2019; Cruz Meléndez and Zamudio Vázquez 2017).

Since then, although they are different concepts, e-government and open government have undergone a process of complementarity in which technology facilitates a new form of linkage between government and society. Examples include the tools and platforms that tend to be used most frequently: websites, transparency and open data portals, public procurement sites, mobile applications and digital social networks such as Facebook and Twitter (Cruz Meléndez and Zamudio Vázquez 2017).

Determinants of open e-government at the local level

The literature identifies a series of factors as determinants of the success of local e-government. They include elements related to the availability of technological infrastructure (Alcaide-Muñoz, López Hernández, and Caba-Pérez 2014; Lowatcharin and Menifield 2015); the characteristics or size of the population (Dias and Costa 2013; Maciel, Gomes, and Dias 2016; Tejedo-Romero et al. 2022); organisational factors or local management capacity (Andrews and Entwistle 2015; Andrews et al. 2013; Manoharan 2013b; Norris and Moon 2005; Puron-Cid and Rodríguez-Bolívar 2018); and political factors (Sicáková-Beblavá, Kollárik, and Sloboda 2016; Tejedo-Romero and Ferraz Esteves de Araujo 2018).

Considering the abundant and somewhat fuzzy empirical evidence and literature on the topic over the last two decades, we have decided to structure our empirical expectations based on some of the dimensions and categories of determinants of Dias’ (2020) empirical model on local e-government elaborated from a systematic revision of studies published between 2002 and 2018. This model identified three e-government dimensions: (i) e-participation associated with citizens’ engagement; (ii) e-transparency relates to e-disclosure and the provision of public information; and (iii) e-services that refers to the provision of digital services at the local level. As we will detail in our methodology, we elaborated our index on Esteves’ (2005) e-services model, which incorporates e-participation at its highest level, and then we extend it by including the e-transparency dimension. In addition, Dias’ (2020) model identifies four categories of determinants of local e-government: (i) local socioeconomic determinants; (ii) internal determinants; (iii) local political determinants; and (iv) other environmental determinants. Our empirical expectations focus on the categories related to socioeconomic and internal determinants since both are considered transversal e-government determinants.

The local socioeconomic determinants focus on demography, Internet use indicators and socioeconomic dynamism (Dias 2020). Indeed, several studies suggest that the number of inhabitants may influence local e-government by creating scale economies that enable municipal governments to distribute the costs of technological projects more efficiently (Andrews and Entwistle 2015; Andrews et al. 2013; Boyne 1995; Moon 2002; Pina, Torres, and Royo 2010). Population density, in turn, impacts the amount of information and forms of participation used by municipal governments to create trust in their local management (Maciel, Gomes, and Dias 2016). As explained in the next section, these variables are used as controls in our models.³

Surprisingly, there is no consensus about the magnitude of the influence of Internet infrastructure on local e-government and transparency. Sicáková-Beblavá, Kollárik, and Sloboda (2016) found that, in the case of Slovak municipalities, it is not possible to confirm that greater Internet access leads to a higher level of government transparency. Lowatcharin and Menifield (2015) also reached a similar conclusion for counties in the Midwest of the United States. More complex analyses, incorporating other factors such as the population’s educational level and density, suggest that Internet access is not statistically significant (Lowatcharin and Menifield 2015). Even so, our empirical expectation is to find a positive relationship between infrastructure and the development of open e-government in Chilean municipalities (González-Bustamante, Carvajal, and González 2020).⁴ Therefore, our first hypothesis is:

• Internet Hypothesis. The number of fixed Internet connections in a municipal district increases the municipality’s development of open e-government.

On the socioeconomic dynamism construct, a study of Portuguese municipalities by Dias and Costa (2013) concluded that the population’s income level explains the demand for access to information and that the size of the population and the resources at the disposal of the municipal government were the factors that most boosted the development of local e-government. Indeed, various socioeconomic characteristics such as the percentage of young and educated people, human development, income and employment indicators, among others, can be used to measure local socioeconomic dynamism (Dias 2020). In light of the above, we focus on the aggregate socioeconomic level measured inversely using monetary poverty indicators. Consequently, our second hypothesis is:

• Socioeconomic Dynamism Hypothesis. A municipal district’s level of monetary poverty reduces the municipality’s development of open e-government.

Focusing on the second category of determinants elaborated by Dias (2020), we can find a number of relevant indicators such as the size of local government, financial capacity, management capacity, technical capacity, leadership, organisational culture or experience.⁵ We are particularly interested in some of the most common, such as local government size and management capacity measured using municipal budget and human resources indicators, respectively.

Just as the population’s income or different socioeconomic dynamism indicators are important for evaluating the development of open e-government, the municipal government’s budget should also be important (Dias and Costa 2013; González-Bustamante, Carvajal, and González 2020). Indeed, the budget reflects the size of the local government (Dias 2020). In this context, we assert that the greater a municipal government’s autonomous financing, the higher the level of open e-government will be. There will, therefore, be a difference with municipalities that do not have sufficient sources of revenues and, probably, have weaker organisational capacities. Consequently, our empirical expectation is as follows:

• Budget Hypothesis. A municipality’s budget increases its development of open e-government.

Finally, we also consider the management capabilities of municipal personnel as a determinant of the development of open e-government (Dias 2020). Chapman (2017), for example, identifies a municipal government’s professionalisation as explaining its level of adoption of e-government and innovation. Similarly, Manoharan (2013a) studied the institutional variables of US counties, such as the size of the organisation, its technical capacity and the presence of specialised ICT personnel, to determine their influence, concluding that they are an excellent predictor of the counties’ adoption of e-government. Therefore, our fourth hypothesis is:

• Management Capacity Hypothesis. The level of professionalisation of municipal personnel increases the municipality’s development of open e-government.

As indicated above, Dias’ (2020) empirical model also identified additional categories, such as local political and environmental determinants, which do not impact e-government transversally. For example, part of the literature focuses on political factors such as electoral competitiveness. This factor may favour transparency and the implementation of technological and innovation projects since greater competitiveness implies more propitious environments for innovation (Alcaide-Muñoz, López Hernández, and Caba-Pérez 2014; González-Bustamante, Carvajal, and González 2020; Grimmelikhuijsen and Welch 2012). The capacity for local collective action and the complexity of the population have also been examined as variables affecting the performance of local governments and the diversity of services provided, conditioning innovation projects (Andrews 2007; Andrews et al. 2013; Walker, Berry, and Avellaneda 2015).

Empirical strategy

Index of open e-government

We started by using the e-services model of Esteves (2005), which contains different items grouped into five dimensions, to evaluate government websites (Fath-Allah et al. 2017) and some dimensions of Dias’ (2020) local empirical e-government model. Salazar, Ubeda-Medina, and Fernández (2010) used Esteves’ (2005) model in the Los Ríos Region of Chile, while González-Bustamante, Carvajal, and González (2020) carried out measurements in five of the country’s most populous regions in 2016.

We applied the model to Chile’s 345 municipalities. To this end, we carried out a binary measurement for each item from the first to the fifth phase in Table 1 and an aggregation, derived from the original formula of Esteves (2005), which takes into account the weights theoretically associated with the level of sophistication or digital maturity of each phase.⁶ This measurement of municipal e-government was carried out in February 2019 and November 2021.

Table 1. Local open e-government index aggregation.

	Phase	Item	Weight
1st	Presence	Forms (downloadable documents)	0.25
		Municipal newsletter	0.25
		Online map	0.25
2nd	Urban information	Urban map	0.50
		Transport information	0.50
3rd	Interaction	Municipal social networks	0.75
		Municipal telephone	0.75
		Mobile	0.75
4th	Transaction	Online official procedures	1.00
		Follow-up	1.00
		Digital certificates	1.00
		Register of residents	1.00
		Online payment	1.00
5th	E-democracy	Citizen participation	1.25
6th	E-transparency	Transparency request response within 30 working days	1.00
		Transparency request response within 20 working days	1.25

Note: Adapted from Esteves (2005), Bayona and Morales (2017) and González-Bustamante, Carvajal, and González (2020). Some items in the original model, such as the presence of a search engine or the option of personalising the interface with the website, have been omitted so as to give priority to the incorporation of more important elements that are now of daily use, such as digital social networks. E-transparency items were incorporated as an additional phase.

We then expanded the EGi by incorporating transparency indicators to arrive at the OEGi. For this, we used the requests for access to public information filed to the Council for Transparency (CPLT 2022).⁷ With this information, we constructed an indicator of the average number of working days that each municipal government took to respond to the transparency requests. This average is relevant to measure response since the Chilean legal framework is stringent concerning deadlines and imposes fines on public servants in case of non-compliance. In this way, we are underlyingly considering environmental factors, such as normative regulations and directives, which may affect e-transparency (Dias 2020). Therefore, we elaborated two binary variables per year reflecting whether a municipal government’s average response time was within the legally stipulated 20 working days or within the legally permitted extension of ten additional working days. Both variables were incorporated into the aggregate index with a weighting equivalent to the fourth phase of the EGi for responses within the extended legal period and the fifth phase for responses within the strict legal period of 20 days (Table 1).⁸

We calculated Cronbach’s α with a bootstrapping of 1,000 samples and a 95% confidence interval. Our aggregate OEGi has a Cronbach’s α of 0.681 for all Chile’s municipalities (95% CI [0.635, 0.719]). In the municipal districts of the most populous regions, with over one million inhabitants, the indicator increases to 0.713 (95% CI [0.658, 0.758]).

Independent variables

We integrated the OEGi with geospatial information and local government indicators compiled using the following open data and public information: (i) geospatial shapefiles and data from the Infrastructure of Geospatial Data of Chile portal and the Undersecretariat for Regional Development (IDE-Chile and SUBDERE 2018, 2020); (ii) data for the number of fixed Internet connections at the municipal district level from the Undersecretariat for Telecommunications (SUBTEL 2022); (iii) the rate of monetary poverty⁹ from the Social Observatory Division of the Ministry of Social Development and the Family and the UN Economic Commission for Latin America and the Caribbean (DOS-MDSF and ECLAC 2021); (iv) indicators of municipal budget with monetary correction and the professionalisation of municipal personnel based on the proportion holding a professional qualification, in both cases using data from SUBDERE’s National System of Municipal Information (SINIM 2022); and (v) estimated population of the municipal district by the National Institute of Statistics (INE 2022).

Spatial autocorrelation and econometric strategy

Using a geospatial econometric strategy is particularly useful considering that our local socioeconomic and internal determinants, particularly those related to demographic indicators we use as controls and characteristics of municipalities, can also be determinants of innovation (Dias 2020). In this context, relying on this empirical strategy that accounts for territorial dynamics allows us to assess spatial autocorrelation as an indicator of diffusion of adoption at the local level since neighbouring municipalities could influence the development of open e-government.

We began by using Moran’s Index (Moran’s I), which is analogous to the Pearson coefficient for spatial units (Goodchild 2008). We used one variant under randomisation and another with Markov chain Monte Carlo (MCMC) simulations. The first variant involves a random permutation of the spatial units to obtain the spatial autocorrelation values ⁣⁣and their levels of statistical significance (Celemín 2009). Values close to zero indicate randomness in the spatial pattern, while values close to 1 and close to −1 indicate perfect correlation and perfect dispersion, respectively. In turn, MCMC simulations permit the identification of random variables in a sample based on their probability distribution (Jackman 2004). We ran 10,000 Bayesian simulations in each analysis.

To calculate both indices, we worked with vector data from IDE-Chile and SUBDERE (2018, 2020) and spatial contiguity matrices based on the sphere of influence (SOI) model, which is derived from the Delaunay triangulation of neighbours in which the relationship between points is based on the Euclidean distance. In SOI matrices, triangulation is reduced by eliminating the longest links (Bivand, Pebesma, and Gómez-Rubio 2013) (Figure 1).

Figure 1. a. Queen-style contiguity matrix. b. Matrix with Delaunay triangulation. c. Matrix with SOI model, highlighted in colour. d. Matrix based on neighbours by distance with k = 4. e. Matrix based on neighbours by distance with k = 2.

Note: Santiago Metropolitan Region has been used as an example. Based on data from IDE-Chile and SUBDERE (2018, 2020).

We then used ordinary least squares (OLS) models. Our baseline model considers i-th municipalities (n = 345) and regresses our index of open e-government Y_i = OEGi_[i] on the number of fixed Internet connections¹⁰ per municipal district X_1[i]. Since the number of Internet connections is very high, we used a logarithmic transformation. As fixed effects, we then incorporated the municipal district’s estimated population and population density, also as logarithmic transformations, before also incorporating the municipal district’s rate of monetary poverty X_2[i] and, finally, each municipal government’s budget X_3[i], with logarithmic transformation, and the rate of professionalisation of municipal personnel X_4[i]. As result, our final model is expressed as follows: $Y_i=\alpha +{\text{β}}_1\log \left(X_{1\left[i\right]}\right)+{\text{β}}_2{X}_{2\left[i\right]}+{\text{β}}_3\log \left(X_{3\left[i\right]}\right)+{\text{β}}_4{X}_{4\left[i\right]}+{\gamma }_1\log \left(\mathrm{po}{p}_i\right)+{\gamma }_2\log \left(\mathrm{densit}{y}_i\right)+{ϵ}_i$

With the model’s residuals, spatial autocorrelation can be evaluated statistically with Moran’s I. If the test is statistically significant at 95% confidence, the residuals are spatially grouped, and it is appropriate to fit spatial econometric models. The options include spatial autoregressive (SAR) models in which the ρW parameter measures the spatial autocorrelation of the dependent variable. Considering our vector of the j-th main independent variables (J = 4), the equation is as follows, applying the transformation of the vector’s variables when necessary: $Y_i=\mathrm{\rho W}{Y}_i+\alpha +{\sum }_{j=1}^J{\text{β}}_j{X}_{j\left[i\right]}+{\gamma }_1\log \left(\mathrm{po}{p}_i\right)+{\gamma }_2\log \left(\mathrm{densit}{y}_i\right)+{ϵ}_i$

Another option corresponds to spatial error models (SEM), where the parameter λWu measures the spatial dependence of the errors for a latent continuity variable u. $Y_i=\alpha +{\sum }_{j=1}^J{\text{β}}_j{X}_{j\left[i\right]}+{\gamma }_1\log \left(\mathrm{po}{p}_i\right)+{\gamma }_2\log \left(\mathrm{densit}{y}_i\right)+\lambda W{u}_i+{ϵ}_i$

Reliability and robustness checks

For the MCMC analyses, we performed a convergence diagnostic of the Markov chains, specifically the trace and distribution (see Supplementary Material file). In the econometric models, we ran multicollinearity tests and, as indicated above, calculated Moran’s I with the residuals of each regression to evaluate the conformation of spatial clusters.

In addition, we applied a series of robustness checks, alternating the variable of the municipal budget according to tax collection rates and financial dependency using data from SINIM (2022). The former was calculated based on revenues from local tax collection over total annual revenues, and the latter used the revenues the municipal government received from the Common Municipal Fund (FCM).¹¹ There is some low level of collinearity between revenues from the FCM and a municipal district’s poverty rate because the criteria for the FCM’s distribution give a weight of 10% to the poverty rate (DOS-MDSF and ECLAC 2021). We also incorporated a binary variable that reflects the existence of a protocol of citizen participation at the local level¹² in order to carry out additional tests with a variable from the category of local political determinants addressed in the empirical model of Dias (2020).

Results

Georeferencing and spatial autocorrelation

We detected patterns of weak spatial dependence at the national level, with a 95% statistical significance and a Moran’s I of 0.183 under randomisation (p ≤ 0.001) and for the MCMC simulations (p ≤ 0.001). Table 2 shows the average values of the OEGi by region and level of autocorrelation, while choropleth maps for the regions where significant patterns were detected are provided in the Supplementary Material file.

Table 2. Open e-government at the local level and spatial autocorrelation, Chile, 2019–2021.

Region	OEGi	Moran’s I	MCMC
Arica y Parinacota	4.229	−0.378
Tarapacá	5.226	0.311**
Antofagasta	5.778	−0.176	−0.176
Atacama	5.245	−0.320	−0.320
Coquimbo	5.686	−0.064	−0.064
Valparaíso	6.432	0.121	0.121
Santiago Metropolitan	6.722	0.092	0.092
O’Higgins	5.953	−0.223	−0.223
Maule	5.965	0.082	0.082
Ñuble	5.199	0.087	0.087
Biobío	5.554	0.058	0.058
Araucanía	5.319	0.150*	0.150*
Los Ríos	5.717	−0.198	−0.198
Los Lagos	5.389	0.203**	0.203*
Aysén	5.298	0.103	0.103
Magallanes	5.692	0.046	0.046

Note: Compiled with data from IDE-Chile and SUBDERE (2018, 2020) and CPLT (2022). The Arica y Parinacota and Tarapacá Regions have very few municipal districts to perform MCMC simulations. MCMC convergence plots can be found in the Supplementary Material file. Significance: * p ≤ 0.100; ** p ≤ 0.050; *** p ≤ 0.010.

As seen in Table 2, the Santiago Metropolitan Region, the most populous of all Chile’s regions, has the country’s highest average index out of a maximum possible of 12.5 points. However, the OEGi’s spatial distribution does not show significant patterns at the regional level (Moran’s I p = 0.105, MCMC p = 0.110), as seen in Figure 2. Another region that stands out for its high average is the Valparaíso Region, the third most populous, while the areas in the extreme north and south of the country have comparatively lower indices.

Figure 2. Distribution of the open e-government index in the Santiago Metropolitan Region.

Note: Based on data from IDE-Chile and SUBDERE (2018, 2020) and CPLT (2022).

Econometric models

Table 3 shows our econometric models. Model I corresponds to a linear regression estimated with OLS to predict the OEGi with the number of fixed Internet connections. The model is significant and explains a moderate proportion of the variance of the index (R² = 0.245, F(1, 342) = 111.01, p ≤ 0.001, adj. R² = 0.242). In this baseline model, the effect of Internet connections is statistically significant (β = 0.285, 95% CI [0.232, 0.338], p ≤ 0.001). This implies that the more fixed Internet connections a municipal district has, the higher its level of open e-government.

Table 3. Determinants of open e-government at the local level, Chile, 2019–2021.

	Model I	Model II	Model III	Model IV	Model V
Internet connections (log)	0.285*** (0.027)	0.193*** (0.073)	0.077 (0.075)	0.048 (0.074)	0.049 (0.074)
Poverty rate			−0.083*** (0.017)	−0.049*** (0.019)	−0.049** (0.019)
Budget (log)				0.382*** (0.098)	0.382*** (0.098)
Personnel professionalisation rate					0.002 (0.006)
Constant	3.773*** (0.206)	2.827*** (0.956)	3.511*** (0.937)	0.528 (1.193)	0.425 (1.241)
Population (log)	No	Yes	Yes	Yes	Yes
Population density (log)	No	Yes	Yes	Yes	Yes
Estimation method	OLS	OLS	OLS	OLS	OLS
VIF	1.324	1.331	1.421	1.486	1.486
B-P/Cook-Weisberg	0.031	0.001	1.364	1.443	1.437
Shapiro-Wilk	0.991**	0.991**	0.995	0.994	0.994
Moran’s I	0.081**	0.082**	0.033	0.016	0.018
N	345	345	345	345	345
R^2	0.245	0.249	0.296	0.327	0.327
Adj. R^2	0.242	0.242	0.288	0.317	0.315

Note: Beta coefficients, standard errors in parentheses. Compiled with data from IDE-Chile and SUBDERE (2018, 2020), CPLT (2022), SUBTEL (2022), DOS-MDSF and ECLAC (2021), SINIM (2022) and INE (2022). Significance: * p ≤ 0.100; ** p ≤ 0.050; *** p ≤ 0.010.

Subsequently, in model II, we incorporated logarithmic transformations of the municipal district’s estimated population and population density as fixed effects. The result is quite similar to the baseline model (R² = 0.249, F(3, 341) = 37.81, p ≤ 0.001, adj. R² = 0.242) and the number of Internet connections remains statistically significant at 95% with a positive effect (β = 0.193, 95% CI [0.050, 0.337], p = 0.009).

In model III, we incorporated the municipal district’s monetary poverty rate. This regression is also significant (R² = 0.296, F(4, 340) = 35.82, p ≤ 0.001, adj. R² = 0.288). However, the number of Internet connections ceases to be statistically significant (p = 0.307). The poverty rate has a significant negative coefficient (β = −0.083, 95% CI [−0.117, −0.049], p ≤ 0.001), implying that the level of open e-government decreases the higher the poverty rate.

Finally, in models IV and V, we sequentially incorporated the municipal government’s budget and a rate that measures the professionalisation of its personnel. Model IV is significant and explains a substantial part of the variance of the OEGi (R² = 0.327, F(6, 338) = 27.39, p ≤ 0.001, adj. R² = 0.315). Again, the effect of the number of Internet connections is not significant (p = 0.517), while the coefficient of the poverty rate varies slightly but remains negative (β = −0.049, 95% CI [−0.087, −0.012], p = 0.010). The effect of a municipal government’s budget is statistically significant and positive (β = 0.382, 95% CI [0.190, 0.575], p ≤ 0.001). In model V, the results are very similar (R² = 0.327, F(5, 339) = 32.93, p ≤ 0.001, adj. R² = 0.317), the coefficients of the monetary poverty rate and a municipal government’s revenues do not change, and the rate of professionalisation of its personnel is not statistically significant (p = 0.760).

Based on the results of the first two models, we could evaluate the Internet Hypothesis. However, problems in normality and a significant level of spatial autocorrelation in the residuals are apparent. In response, we fitted SAR and SEM models to adequately measure the connections’ effect on the OEGi (Table 4). The SAR model confirms the results of models I and II, the spatial autocorrelation in the residuals is controlled, and the significant positive coefficient of rho indicates spatial autocorrelation in the dependent variable. In turn, the SEM model shows a very similar coefficient, and the value and significance of lambda show spatial dependence in the errors. With this evidence, we could accept the Internet Hypothesis. However, models III, IV and V suggest that the number of Internet connections in a municipal district is statistically significant only in the absence of variables related to the socioeconomic dynamism construct and the municipal government’s management capacity. This is in line with the findings of Lowatcharin and Menifield (2015) and Sicáková-Beblavá, Kollárik, and Sloboda (2016). In any case, it is interesting that, in the absence of other dimensions, the relationship between Internet connections and digital development is subject to spatial autocorrelation.

Table 4. Spatial models of Internet connections and open e-government at the local level, Chile, 2019–2021.

	SAR	SEM
Internet connections (log)	0.182** (0.072)	0.180** (0.076)
ρ (rho)	0.157** (0.066)
λ (lambda)		0.149** (0.072)
Constant	1.842* (1.015)	2.762*** (0.976)
Population (log)	Yes	Yes
Population density (log)	Yes	Yes
Estimation method	ML	ML
VIF	1.324	1.324
Moran’s I	−0.005	−0.002
N	345	345
Pseudo-R^2	0.261	0.258
Log likelihood	−566.041	−566.663
AIC	1,144.083	1,145.325

Note: Beta coefficients, standard errors in parentheses. Compiled with data from IDE-Chile and SUBDERE (2018, 2020), CPLT (2022), SUBTEL (2022) and INE (2022). Significance: * p ≤ 0.100; ** p ≤ 0.050; *** p ≤ 0.010.

It is important to highlight that the different multicollinearity and homoscedasticity tests of the models were adequate. In addition, the calculation of Moran’s I with the residuals of each model to evaluate the conformation of spatial clusters and their normality analysis only evidenced a need to fit the additional models presented in Table 4.

The results of models III, IV and V consistently indicate the importance of the poverty level, with an inverse effect on the development of open e-government, and of the municipal government’s budget, in this case, with a positive coefficient. Based on model V, we estimated the marginal effects of the two determinants of the OEGi (Figure 3). In light of these results, we are able to accept the Socioeconomic Dynamism Hypothesis and the Budget Hypothesis, which is consistent with other studies such as those of Dias and Costa (2013) and González-Bustamante, Carvajal, and González (2020) and the relevance of local socioeconomic determinants and internal determinants in the empirical model of Dias (2020). On the other hand, we reject the Internet Hypothesis and the Management Capacity Hypothesis.

Figure 3. a. Effect of monetary poverty on open e-government at the local level in Chile. b. Effect of budget measured with own municipal revenues on open e-government at the local level in Chile.

Note: Based on data from IDE-Chile y SUBDERE (2018, 2020), CPLT (2022), SUBTEL (2022), DOS-MDSF and ECLAC (2021), SINIM (2022) and INE (2022).

In our robustness checks (see Supplementary Material file), we incorporated a binary variable that measures the existence of municipal protocols on citizen participation as a construct of the local political determinants category in Dias' (2020) empirical model. None of our tests showed it to be significant. We also alternated the municipal government’s revenues variable with measurements that reflect the level of local tax collection and the municipal government’s financial dependence. Only this last variable was significant, with a negative coefficient at 90% confidence (p = 0.055), which is consistent with financial dependence as an aspect of weakness in organisational capacity.

Discussion and conclusion

In this research, we present the OEGi, a novel open e-government index based on the e-services model of Esteves (2005) and transparency indicators. At the national level, a slight spatial autocorrelation is detected but, at the subnational level, only the Tarapacá, Araucanía and Los Lagos Regions show clustering of the indicator. This finding is interesting and requires further attention because if we consider that these regions have a particularly low index, below the national average, we could identify some clusters where backwardness occurs rather than a diffusion of innovation.

We also evaluated the determinants of open e-government in two categories proposed by Dias’ (2020) empirical model: local socioeconomic determinants and internal determinants. With these categories, we elaborated four hypotheses. The evidence of our econometric models allows us to accept the Socioeconomic Dynamism Hypothesis and the Budget Hypothesis since a municipal district’s monetary poverty rate causes decreases in our index while the level of the municipal government’s budget increases it. Our findings are consistent with the literature on electronic and open government and align with the transversal local e-government determinants categories of Dias’ (2020) empirical model drawn from a systematic review.

On the other hand, we reject the Internet Hypothesis and the Management Capacity Hypothesis. The number of fixed Internet connections in a municipal district is significant in the absence of other factors, and the econometric modelling reveals patterns of spatial autocorrelation. However, consistently with previous studies, the number of connections loses relevance when variables from other dimensions are incorporated.

Finally, the level of professionalisation of municipal personnel is seen not to be a significant variable. This suggests that management capacity at the aggregate level in municipalities may not influence open e-government, however, it would be interesting to evaluate microdata and the level of expertise of ICT teams in order to assess the potential effect of the technical capacity component.

It is relevant to consider some potential limitations of this study. First, the item e-democracy presented in Esteves’ (2005) classic model is considered part of e-services, however, much of the literature understands it as a separate dimension. Nevertheless, since we added transparency indicators, we could strictly indicate that our index covers three relevant dimensions of local e-government identified in the empirical model of Dias (2020) (i.e. e-services, e-participation and e-transparency). The potential problem here is that our index is inspired, like most quantitative studies assessing e-government implementation, by digital maturity models, thus mixing different dimensions that can be measured separately (Dias and Costa 2013) and presupposes a certain sequentially that does not necessarily occur (Dias 2020). On the other hand, our empirical strategy’s observational nature could be exposed to imbalances in determinants. In this context, we have some possibly relevant unobserved factors, such as local political variables or other environmental determinants, for example, stakeholder pressure.

Despite these limitations, this work constitutes a theoretical contribution that implies an interconnected approach to several dimensions of local e-government in Latin America. It is, moreover, a methodological and empirical contribution that can serve as a starting point for future research focusing on problems of non-random selection with observational data at subnational levels and measurement challenges on digital innovation implementation. It also presents an opportunity for in-depth evaluation of digital policies at the local level and guides decision-making in these spaces, in particular where innovation is lagging.

Acknowledgements

This work is dedicated to the memory of Eduardo Araya. We appreciate the feedback from participants at the ALACIP Spatial Symposium held virtually in July 2022, at the Smart Cities and Open Government Seminar in Madrid in September 2022, and at Digital Democracy Workshop 2022 in Zurich in October 2022. We also thank the support for revising municipalities’ websites to aggregate our index provided by Rodrigo Cuevas, Berenice Orvenes and Elinor Luco.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data can be obtained on Zenodo: González-Bustamante, B., and D. Aguilar. 2023. “Data Set on Local Government Indicators in Chile (Version 0.21.15 – Late Sky)” (dataset, pre-release version). University of Oxford, Universidad de Santiago de Chile (USACH) and Training Data Lab. DOI: 10.5281/zenodo.7568387.

Additional information

Funding

This work was supported by the Universidad de Santiago de Chile (USACH) under the Grant USA2156 POC2022_FAE2; the International Political Science Association (IPSA) under the IPSA RC05-RC10-RC22-GIGAPP-AECPA Travel Grant; and the Universität Zürich under the Digital Democracy Workshop 2022 Travel Grant.

Notes

1 Directly elected representatives who participate in the Municipal Council, a body of a normative, deliberative and supervisory nature, chaired by the municipal district’s mayor, that is responsible for materialising the effective participation of the local community. Each Municipal Council has between six and ten members, depending on the size of the municipal district. Algarrobo, for example, has a six-member Municipal Council.

2 The first cluster includes Spain, Brazil, Portugal and Mexico and has the most publications, citations, and H-indexes. Together, these countries make up 85.9% of the documents published, and 92.9% of the citations received. The second cluster comprises Argentina, Chile, Ecuador, Colombia and Uruguay and has fewer publications and citations than the first cluster. These countries constitute 14.5% of the documents and 7.4% of the citations. Then, the third cluster includes Costa Rica, Peru, Venezuela, Puerto Rico and the Dominican Republic and has even fewer publications and citations than the second cluster. The third cluster sums up 3.4% of the documents and 0.8% of the citations. Finally, the fourth cluster includes Bolivia, El Salvador, Honduras, Paraguay, Cuba, Guatemala, Nicaragua and Panama and has the fewest publications and citations of all the clusters.

3 In the case of the population’s characteristics and their effect on the development of e-government at the local level, the variables highlighted by the literature include age. For instance, van Dijk, Peters, and Ebbers (2008) found that, among the Dutch population, younger people are the most open to the use of technology. Budding, Faber, and Gradus (2018) reached a similar conclusion after analysing the maturity of the e-government sites of approximately 400 Dutch municipalities, finding that demographic, rather than socioeconomic, characteristics explain their level of e-government.

4 It is important to note that different studies have suggested that a population’s per capita income explains its use of the Internet (Lowatcharin and Menifield 2015) and demand for access to public information (Dias and Costa 2013), both at the local and national levels. Income is associated with the local socioeconomic dynamism construct in the framework of the local socioeconomic determinants category in Dias’ (2020) empirical model.

5 The internal determinants of local e-government, especially the size of municipalities, tend to be highly correlated with the determinants of innovation (Dias 2020). In this context, the diffusion of innovation theory from the 1960s explains adoption as a process in which innovativeness is passed or communicated through specific channels over time in a systemic context (Rogers 2003; see also Dias 2020).

6 Although the e-democracy phase is considered in Esteves’ (2005) model and updated versions, it is relevant to note this could be associated with potential conceptualisation and measurement limitations. This concept, on the one hand, could be related to the e-participation dimension in the empirical model of Dias (2020), which is not part of the e-services dimension. On the other hand, some concerns about measurements could arise. For example, the phenomenon’s complexity may not be reflected in the binary codification in this model, which has measured online forums and referendums, mainly e-voting processes on local issues.

7 We downloaded 1,406,466 observations on 9 July 2022 in comma-separated format. We then restricted the date of filing of the requests to the study period, obtaining 690,746 observations. Finally, we discarded all requests that were not addressed to a municipal government, reducing the set to 233,196 observations, equivalent to 33.8% of the requests filed during the period.

8 We carried out imputations for 12 missing cases in 2019 (3.5% of the municipalities) and eight in 2021 (2.3%), as well as 17 cases of municipal governments that did not receive transparency requests in 2019 (4.9% of the municipalities) and 15 in 2020 and 2021 (4.4%) (see Supplementary Material file for details of this procedure). We then assembled an average estimate of the EGi for 2020 and, finally, aggregated all these variables into OEGi.

9 The indicator used corresponds to monetary poverty in 2020 with a small area estimation (SAE) for Chile’s 345 municipal districts. In 256 districts, it corresponds to a Fay-Herriot estimator that combines direct estimates from the National Socioeconomic Characterisation Survey of Chile (CASEN) with a synthetic estimate calculated using econometric models with the administrative variables of wages, occupation, health and education. For the other municipal districts, only the synthetic estimate was used. For more details, see DOS-MDSF and ECLAC (2021).

10 For all the indicators, in general, we used the most recent information available except in the case of Internet connections by municipal district, where updated information was not available for all municipal districts and we used 2018 information for 22 districts (6.4% of the total number).

11 Through this fund, the country’s municipal governments redistribute their own revenues in a solidary manner. The FCM’s mission is to reduce inequalities between municipalities and provide additional resources for those with the lowest tax revenues. It consists primarily of contributions from those municipal governments that collect the most resources in the form of business licences, vehicle taxes, real estate taxes and fines.

12 Under the Constitutional Organic Law on Municipalities, each municipal government must establish a protocol on forms of local citizen participation, indicating the type of organisations that must be consulted and informed, as well as the dates or times at which these processes must take place.