Stages and Determinants of E-Government Development: A Twelve-Year Longitudinal Study of Global Cities

ABSTRACT:

Global e-government innovations are at the forefront of municipal efforts to be better organized and more efficient in delivering services and improving outcomes for the public. Scholars have argued that such innovations are embedded in institutional and environmental factors, and municipal e-government growth evolves through stages as a result of the effects of these factors. However, existing studies rarely model the distinct success factors of the different stages. This article addresses that shortcoming with data from the largest cities in the world’s top 100 “most wired” countries from 2003 to 2016. Cluster analysis addresses whether there are any consistent growth trends, and finds that there are four clusters of e-government development. Regression analysis tests whether stages may be driven by specific factors, and findings reveal that e-government stages mostly have uniform drivers. Population size, GDP, and regional competition have a positive association across all stages. However, democracy level appears to have a more ambiguous status, as it influences some higher stages in large countries but has a negative association in small countries.

INTRODUCTION

E-government is the electronic delivery of information and services to citizens, business, and public administration (Lee et al. 2008). Scholars of e-government have learned how modern information and communication technologies (ICTs) spread around the world through public and private sector innovation, and have become relied upon by governments as tools of service provision and organization (Dunleavy 1994; Pratchett 1999). The stages model theories of e-government have emerged to address e-government diffusion and evolution (e.g., Andersen and Henriksen 2006; Lee, Chang, and Berry 2011; Layne and Lee 2001; Reddick 2004). In this theoretical vein, e-government research can sometimes seem to imply a kind of technological determinism, as if the maturity of e-governments through stages of development is inevitable. In fact, the picture is more mixed. According to a large number of empirical studies, determining factors do appear to be strongly influential in e-government development. Chief among these determining factors are financial resources and gross domestic product (GDP); socio-economic factors of the local citizen body, such as education, wealth, and technology access; and political factors, such as the role of elected officials (Gallego-Álvarez, Rodríguez-Domínguez, and García-Sánchez 2010; Manoharan 2013b; Rodríguez, Sánchez, and Álvarez 2011). Additionally, determinants of e-government maturity vary considerably within or between countries, levels of government, or types of e-government (Bonsón et al. 2012; Gallego-Álvarez et al. 2010; Holden, Norris, and Fletcher 2003; Moon 2002; Norris and Moon 2005), which means that e-government development is certainly neither inevitable nor unilinear.

Despite evidence of variation, research has not yet tested analytical models that explain antecedents of the e-government stages over time. E-government might mature on multiple, interrelated fronts with different variables influencing each stage, and not necessarily in a linear order. While a minimum set of economic, institutional, political, and social conditions must be present for the early stages, it may be that more advanced levels of e-government success later rely on different sets of factors rather than maturing uniformly (Ahn 2011). However, understanding of such variations in determinants at different e-government stages has been limited by a lack of multi-country longitudinal studies. Such studies can account for the factors of time and global variation to make stronger arguments about causality and to explain what stages exist and how they come about. The authors of this article suggest that there are two main reasons why such longitudinal and multi-country analyses are vital for advancing the field. The first is that knowledge of human phenomena relies upon theories that stand up to scrutiny in a range of country and cultural contexts, with the possibility of accounting for trends over time (Cetina 2009; Merton 1968). The second is that e-government is a phenomenon that lends itself to global forms of policy innovation. E-government is based on the global network of the Internet, which encourages global patterns of change. In this article, we explore these global patterns in e-government development and test findings from earlier studies on the determinants of e-government stages, with greater attention to specific dimensions and country contexts. Our research question is therefore two-fold: (1) What evidence exists for global trends in e-government stages? (2) What factors explain the development of the different e-government stages?

In order to address these questions, the article draws on cluster analysis and panel analysis of the Global E-Governance Survey, which is a biennial evaluation of official websites of the largest cities of the 100 “most wired” countries in the world. The survey started in 2003 and there have been seven subsequent surveys, the last covering 2016. The range of countries and the date ranges therefore make the survey an ideal data source for investigating global patterns of e-government development using a longitudinal perspective. The theoretical section of the article will present stage model theories of how e-government evolves and how a government’s social, political, and economic context shapes success in different stages of e-government. A systematic literature review that brings together research in public administration, public management, and information science presents theoretical arguments and hypotheses for determining factors in e-government development worldwide. The article then outlines the research methodology and describes the data settings and the measures we are using for analyses. After the presentation of the results, the final section covers discussion of the findings, presents best practice recommendations, refers to the contributions and implications of our study, and points to directions for future research.

THEORETICAL FOUNDATIONS

E-Government Stages

Many e-government authors have investigated the development of e-government and proposed different types of stage models. Layne and Lee (2001) were among the first to describe e-government as “an evolutionary phenomenon,” and proposed four stages of e-government. In the first stage, cataloguing, governments provide citizens with access to online information. The second stage, transaction, involves adoption of interactive processes between government and citizens, which results in higher efficiencies in service delivery. The third and fourth stages, vertical integration and horizontal integration, cover the transformation and re-conceptualization of government services; for example, such changes may occur through a portal where citizens can access services from different governments, and thus require the collaboration of governments from different levels.

Since Layne and Lee’s work, a civic dimension has been added to the online information and service integration stages of e-government so as to create three major classifications for e-government dimensions: e-information, e-services, and e-participation (Lee, Chang, and Berry 2014; Manoharan 2013a; Nam 2014). For example, Moon (2002) incorporated a civic stage that includes political participation through e-government, and thus marks an important step in recognizing political maturity alongside administrative maturity of e-government. In subsequent literature, the political and social elements of e-government maturity gained more prominence. Andersen and Henriksen (2006) proposed an extension of the model by Layne and Lee, involving a first phase of “cultivation” encompassing vertical and horizontal integration within government. A second phase focuses on the importance of a personalized web interface for customer processes. A third phase moves towards an increased level of transparency and problem-solving mechanisms in citizen exchange, and a fourth phase involves greater citizen-orientation through systematic data mobility across organizations and transfer of data ownership.

Some authors have also analyzed the extent to which governments have progressed through the stages thus far. Moon (2002) assessed the adoption of e-government in 1,471 municipal governments with populations over 10,000. He concluded that only a few governments have moved into the participative stage, and fewer integrate electronic service delivery across levels of government. Similarly, Reddick’s (2004) study of American municipalities showed that the majority of cities have already implemented services of Stage 1, such as online provision of news and information. However, interactive services of Stage 2, such as the provision of online procurement tools, have been marginally adopted by municipalities. Coursey and Norris (2008) also found that, while the percentage of cities offering financial transactions (e.g., tax payments) and nonfinancial transactions (e.g., service requests) increased between 2000 and 2004, advanced stages of e-government adoption remained low. Based on these results from previous empirical studies, we expect that distinct stages of e-government around the world evidence incrementally limited development of later stages of e-government adoption. That is, the curve of e-government stages is positively skewed, with the majority of world cities in lower stages of e-information with limited integration of service levels or citizen participation.

H1: The e-government level of municipalities around the world shows distinct stages of development, with the majority of municipalities concentrated in information provision, with lesser levels of progression to service and then participation stages.

Factors Influencing E-Government Development

Beyond assessing stages of digital maturity, we examine which factors influence the three different dimensions of the e-government stages. Prior literature on e-government offers a large range of factors that influence its adoption and development, which are often referred to as “success factors” (Gil-Garcia and Pardo 2005; Jacobson and Ok Choi 2008). These are shown in Table 1. A systematic review of articles, book chapters, and conference papers from public administration, public management, and information science was carried out using the Web of Science electronic library, and the conference proceedings of specialist e-governance conferences, the conference of the Digital Government Society, Electronic Government conference (EGOV-IFIP-CeDEM-ePart), and the United Nations E-Gov conference (full bibliography in Appendix A). In our theoretical section, we have included as many of the variables as possible that have been found in prior works. In some cases, we have had to choose between two very similar variables from the literature that are either worded almost identically or operationalized in similar ways. In other cases where there are disagreements in the literature or where evidence is sparse, we lean on other logical and empirical arguments to make decisions or to theorize more explicitly about causal mechanisms in e-governance stages.

TABLE 1 Summary of Findings on Antecedents of E-Government Success across Three Stages

	E-government (general)	E-information	E-services	E-participation
Environmental factors
Population size (large)	Brudney and Selden, 1995; Das, Singh, and Joseph, 2017; Gulati, Williams, and Yates et al., 2014; Ho and Ni, 2004; Holden, Norris, and Fletcher, 2003; Homburg and Dijkshoorn, 2012; Manoharan 2013b; Moon, 2002; Norris and Moon, 2005; Nelson and Svara, 2012; Stier, 2015	Ahn, 2011; Feeney and Brown, 2017; Gallego-Álvarez et al., 2010; Reddick and Norris, 2013	Feeney and Brown, 2017; Gallego-Álvarez et al., 2010; Reddick, 2009; Reddick and Norris, 2013	Feeney and Brown, 2017; Gallego-Álvarez et al., 2010; García-Sánchez et al., 2011; Guillamón et al., 2016; Reddick and Norris, 2013
Internet and ICT infrastructure	Azad et al., 2010; Gulati et al., 2014; Holzer and Kim, 2008; Holzer, You, and Manoharan 2010; Holzer and Manoharan, 2012; Ifinedo, 2012; Singh, Das, and Joseph, 2007; Stier, 2015; Titah and Barki, 2006	Gallego-Álvarez et al., 2010	Bhatti, Olsen, and Holm Pedersen, 2011; Gallego-Álvarez et al., 2010	Åström et al., 2012; Gallego-Álvarez et al., 2010; Gallego-Sánchez et al., 2011; Krishnan, Teo, and Lymm, 2017
Socio-economic characteristics (Education)	Gulati et al. 2014; Holzer, Zheng, and Manoharan, 2014; Jun and Weare, 2010; Manoharan, 2013b; Nelson and Svara, 2012; Pick and Azari, 2008; Stier, 2015	Reddick and Norris, 2013	Reddick and Norris, 2013.	Krishnan, Teo, and Lymm, 2017; Reddick and Norris, 2013
Citizen demand	Faniran and Olaniyan, 2009; Ho and Ni, 2004; Lee, Chang, and Berry, 2011; Solvodelli et al., 2012; Zhao et al., 2014	Ahn, 2011	Bhatti et al., 2011; Wang and Feeney, 2016	Guillamón et al., 2016
Institutional factors
Political competition	Lee et al. (2011)	Ahn, 2011
Political support	Ho and Ni, 2004; Puron-Cid, 2012; Schwester, 2009; Yun and Opheim, 2010			Gallego-Álvarez et al., 2011
Democratic institutions and norms	Aladwani, 2016; Azad et al., 2010; Basu, 2004; Berntzen and Karamagioli, 2010; Eom, 2013; Gulati et al., 2014; Ho and Ni, 2004; Ifinedo, 2012; Khalil, 2011; Lee et al. 2011; Puron-Cid, 2012; Singh, Das, and Joseph, 2007; Stier, 2015; Zhao, Shen, and Collier, 2014
Level of development	Holzer and Kim, 2004; Holzer and Kim, 2006; Holzer and Kim, 2008; Holzer, You, and Manoharan, 2010; Holzer and Manoharan, 2012; Holzer, Zheng, and Manoharan, 2014; Holzer and Manoharan, 2016
Elected municipal manager		Feeney and Brown, 2017	Feeney and Brown, 2017	Feeney and Brown, 2017; Zheng, Schachter, and Holzer, 2014
Appointed municipal manager	Brudney and Selden, 1995; Moon, 2002; Nelson and Svara, 2012	Reddick and Norris, 2013	Reddick, 2009
Internal capacity factors
Technical capacity	Ifenido, 2012; Jun and Weare, 2010; Norris and Moon, 2005; Puron-Cid, 2012; Schwester, 2009; Tolbert, Mossberger, and McNeal, 2008	Ahn, 2011	Reddick, 2004	Ma, 2014
Administrative professionalism	Brudney and Selden, 1995; Manoharan, 2013b; Norris and Moon, 2005; Titah and Barki, 2006; Van Veenstra et al., 2010; Yun and Opheim, 2010		Bhatti 2011; Wang and Feeney, 2016	Lau et al., 2008; Ma, 2014
Wealth of public organizations	Azad et al., 2010; Das, Singh, and Joseph, 2017; Gulati et al., 2014; Ifenido, 2012; Pick and Azari, 2008; Puron-Cid, 2012; Schwester, 2009; Singh, Das, and Joseph, 2007; Tolbert, Mossberger, and McNeal, 2008	Ahn, 2011	Bhatti, et al. 2011; Gallego-Álvarez et al., 2010	Gallego-Álvarez et al., 2010
Years of experience	Jun and Weare, 2010; Manoharan, 2013b; Yun and Opheim, 2010	Reddick and Norris, 2013	Reddick and Norris, 2013	Reddick and Norris, 2013

We have chosen to present our model using a three-part institutionalist framework that sees government innovation as the result of internal and external characteristics of public organizations (Frumkin and Galaskiewicz 2004; Manoharan and Ingrams, 2018; Moon 2002). According to Eom (2013:880), this “institutional approach helps e-government researchers not only to locate constraints on actors in e-government policy processes and structural factors that influence the interactions among them, but also to focus on mediators of the impact of information technologies and various environmental factors.” Following the literature review, we found a difference between internal technical factors, on the one hand, and legal and political factors on the other. We therefore discuss internal and institutional factors separately, in addition to the environmental factors. In our discussion of the determinants, we specifically mention the variables that have an effect on all e-government dimensions and those that are expected to have an effect on just one or two specific dimensions.

Environmental factors

Population size is a major environmental factor that helps to encourage municipalities to develop their e-government provisions (e.g., Ahn 2011; Brudney and Selden 1995; Lee et al. 2011; Rogers 1995). This appears to hold across e-information (e.g., Ahn 2011; Gallego-Álvarez et al. 2010; Reddick and Norris 2013), e-services (e.g., Gallego-Álvarez et al. 2010; Reddick 2009; Reddick and Norris 2013), and e-participation (e.g., Gallego-Álvarez et al. 2010; García-Sánchez et al. 2011; Guillamón et al. 2016). Government plays the role of a supplier and must meet the expectations of a densely populated area with a large number of people by being more innovative (Rogers 1995; Weare, Musso, and Hale 1999). Even across major global cities, population size varies widely.

Another function of the responsiveness of government policymakers to public characteristics concerns the characteristics that are often grouped under socio-economic variables or “human capital,” such as education level. Education has been shown to correlate with higher demand for public information, government services, and a willingness to be more participative in the governance of public affairs (Putnam 1995). According to Tolbert, Mossberger, and McNeal (2008), e-government innovation is determined by innovation capacity, reflected largely in local citizen demographics such as level of urbanization, wealth, and education. Similar to wealth and population size, education is expected to influence all three e-government dimensions, which has been confirmed in a prior study by Reddick and Norris (2013) of local governments in the United States.

H2: Population size is positively associated with higher e-information, e-services, and e-participation development.

H3: Education level is positively associated with higher e-information, e-services, and e-participation development.

Internal capacity factors

Perhaps one of the most intuitive and widely confirmed findings about the antecedents of e-government development is the relationship between e-government development and the wealth and financial resources of public organizations, as measured by GDP or taxable wealth (e.g., Bhatti et al. 2011; Gallego-Álvarez et al. 2010). Regardless of environmental or leadership pressure to innovate, all e-government dimensions require significant investment in technology. Therefore, wealthy governments with a healthy financial situation are more likely to adopt innovative e-government tools (Edmiston 2003; Ma 2014; Tolbert et al. 2008).

In addition to the financial resources, e-government planning and implementation involve knowledge gained through learning and leadership (Lee et al. 2011). In trying to explain why governments may evolve slowly through the e-government stages, Ho (2002) suggests that there is a learning curve that is driven by the gradual accumulation and application of new knowledge. Several studies have found that governments with more years of experience working on e-government platforms such as websites develop a better capacity and skill (e.g., Jun and Weare 2010; Manoharan 2013b; Yun and Opheim 2010). Layne and Lee’s (2001) original model suggested that the stages are sequential because e-government capacities build on prior stages; transactions depend on an organizational foundation of electronic information sharing and integration.

Organizational leadership also shapes the planning and capacity that leads to adoption of innovations in e-government (Bhatti et al. 2011; Wang and Feeney 2016). Sophisticated innovations, such as micro-blogging and publishing financial reports online, depend on a government leader who is internally motivated to develop such programs to improve efficiency or transparency (Ho and Ni 2004; Ma 2014). Eom (2013) found that e-government development resulted from having officials in favorable positions of legal authority and technological expertise, and Norris and Moon (2005) found that organizations with shortages of such expertise by way of inadequate technology and web skills experience barriers to adopting better e-government applications. These qualities of leadership are normally subsumed under the responsibility of a chief information officer (CIO).

H4: Financial capacity (GDP per capita) is positively associated with higher e-information, e-services, and e-participation development.

H5: Years of experience are positively associated with e-information, e-services, and e-participation development.

H6: Technological leadership is positively associated with higher e-information, e-services, and e-participation development.

Institutional factors

According to Gallego-Álvarez et al. (2010), different types of government organizations and politics play a role in shaping the e-government innovations. Prior work suggests that competition among governments and pursuit of peer-to-peer legitimacy (Jun and Weare 2010) affect leadership responsiveness and the ability to adopt new policies and programs (Zhang and Feiock 2009). Governments compete with other neighboring governments to gain the respect of citizens and their satisfaction with services (Ho and Ni 2004; Lee et al. 2011). Lee, Chang, and Berry’s study considers the competition between neighboring country governments as a key driver of e-government diffusion, but it makes sense that major cities of neighboring countries would also compete in this way. Neighboring cities compete by developing e-government infrastructure to become more economically attractive to investors (Jun and Weare 2010). Governments are thus expected to be responsive to other governments with which they share geographic proximity and to innovate more in e-government if those neighbors are strong performers.

A substantial amount of theoretical and empirical research has also been undertaken linking characteristics of democratic countries, such as “effective governance,” and e-government (Azad et al. 2010; Eom 2013; Gulati et al., 2014; Stier 2015). Findings suggest that institutional arrangements can make the difference between a highly successful e-services initiative and a failing initiative (Eom 2013; Gulati et al. 2014). On the other hand, Gulati et al. (2014) argue that evidence of the effect of democratic political structures is mixed, but further attention is needed on the relationships between institutional arrangements and specific types of e-government. Despite extensive research on e-democracy and e-participation, there is very little work on the connection between democratic characteristics and transactional and interactional dimensions of e-government.

According to Stier (2015), certain democratic and legal prerequisites need to be in place before e-government initiatives can be successful, but autocratic governments are also making progress. Some cities that score quite low on global measures of democracy, such as Singapore, attain high e-government levels (Calista and Melitski 2007; Lee et al. 2011). Calista and Melitski (2007) suggest that autocratic countries may eagerly embark on e-government initiatives, but have little interest in the interactional dimensions. However, in empirical findings, the relationship between democracy and higher stages of development remains unclear. Historically, democratic countries have led innovation in e-government. So, while evidence is mixed, it is likely, in the long term, that giving citizens more of a say and facilitating open communication in government policies and programs drive higher success in e-services and e-participation adoption.

H7: Regional competition is positively associated with e-information, e-services, and e-participation development.

H8: Democracy is positively associated with e-services and e-participation development.

The hypothesized determinants of the three stages of e-government development are shown in Figure 1. Environmental and internal capacity factors are influential at all stages of development. Institutional factors are shown at the bottom of the diagram in two tracks. One consists of mimetic processes of competition and learning from other governments in the region. Like environmental and internal factors, these are continuous throughout. The second institutional factor is the governance factors, including democratic level and transparency norms. It is these institutional factors which make a difference at the higher levels of e-government development, e-services, and e-participation.

Figure 1. Conceptual model of e-government determinants.

DATA AND METHODS

Setting and Data Collection

The survey instrument used in the research is based on the Rutgers E-Governance Performance Index (Holzer and Manoharan 2016), one of the most comprehensive instruments for e-government research, with 104 measures and five distinct categorical areas. The survey assessed the official websites of the largest cities in the world in five categories: two of the categories relate to website e-information (information content and usability); two relate to website e-services (privacy and e-services); and one category relates to website e-participation (citizen engagement). The five categories aim to capture the e-government levels of maturity in the design of municipal websites as an interface for government-citizen information and interaction (Layne and Lee 2001; Moon 2002; United Nations 2010).

Each category included 17–23 measures, which were coded into ordinal scales where the high number shows full success in a website’s adoption of a feature such as a privacy statement (in the privacy category) or a clear menu (in the usability category). The municipalities evaluated were selected based on population size and the percentage of individuals using the Internet. Using data from the International Telecommunication Union (ITU), the top 100 wired nations were identified, and the largest cities by population in each of those nations were selected for the study as surrogates for their countries (shown in Appendix B). To ensure inter-rater reliability, each municipal website was evaluated in the native language by two individuals, and websites with significant variation (more than 10%) were analyzed by a third evaluator. Evaluators were also provided with sufficient examples to guide how the variables needed to be measured and were given comprehensive written instructions for assessing the websites.

Analytical Methods

Two methods are employed in the analysis: (1) cluster analysis and (2) time series regression. To derive patterns of cities in e-government stages, the method first involved conducting a hierarchical cluster analysis using five e-government scores. Similar to other studies on e-government (e.g., Holzer, Manoharan, and Van Ryzin 2010), cluster analysis is applied to group cities to construct types which should reflect the level of e-government sophistication. The variables were standardized by z-score transformation and used Euclidean distances with Ward’s method of clustering (Aldenderfer and Blashfield 1984). To investigate the groups of e-government across time, a cluster analysis was run using data from 2003, 2009, and 2016.

Second, we conducted analyses on unbalanced panel data to identify the factors influencing e-government stages across municipalities. The time series analysis uses random effects estimates, taking cities as units of analysis across biannual survey years (2003–2016). Random effects estimates are ideal for models that may include unobserved heterogeneity in the units of analysis and when predictors remain constant over time. Further, a Hausman test supported the null hypothesis that a random effects model is a better estimator than fixed effects. Newey West robust standard errors are clustered at the city level to control for unit fixed effects. Robust standard errors also provide wider confidence intervals, which is a precaution here, given slightly high levels of multicollinearity in one model with a 9.88 variance inflation factor (VIF). The other 17 models were in a normal VIF zone of 0.85–4.80. A further reason for robust estimates is that Breusch-Pagan tests in five of the models suggested that variance of the error terms is not constant.

Measures

Dependent variables

Content. Content is the first dependent variable that measures the level of e-information. It uses 25 items to assess five e-information aspects: contact information, public documents, disability access, multimedia materials, and time-sensitive information. The content for contact information included the hours of operation for government offices, access to city codes and city charters, agency mission statements, minutes of public meetings, and budget information. The website was evaluated for the availability of multilingual content, and how it addresses the requirements of disabled users. The website was also evaluated for the presence of audio or video files of public meetings or speeches, calendar of events, and to what extent it integrated the emergency alert systems.

Usability. The category of usability is the second dependent variable for e-information level. It determined the “user-friendliness” of websites’ information based on 19 indicators. These include specific features such as the use of hypertext markup language (HTML), branding and structure (e.g., color consistency, font, graphics, and page length), and if the website indicated the system hardware and software requirements for the users. Related features, such as targeted audience links, use of navigation bars, hyperlinks, and links to the homepage on every web page, were also assessed during the evaluation. Finally, the website search tool was checked in terms of ease of use and the capacity to sort the results as required by the users.

Privacy and security. This category is the first e-services dependent variable. It included 19 measures that focused on two particular aspects of user transactions: the presence of privacy policies and the possibility of user authentication. The evaluation assessed whether users who make an online transaction have access to personal information to report inaccurate information, as well as administrative steps taken by the government to address these privacy concerns, such as encryption, secure servers, and digital signatures. Finally, the survey assessed the intention of websites to monitor citizen activities through cookies and web beacons.

Services. The services category is the second e-services measure. It is composed of 20 items that gauge the extent to which official websites facilitated the provision of public services to citizens, which include those that allow the users to communicate and interact with the city and those that allow users to register online for public events. The initial types of questions focused on the interactivity of websites in terms of reporting crimes or violations, accessing private information (such as medical records or education background), and web portal customization. The questions pertaining to online registration included permits, licenses, online procurement, and the ability to access and bid requests for proposals online. Finally, the availability of transactional services, such as online payments of bills, parking tickets, fees and fines, and online tax payment systems, was determined.

Citizen and social engagement. This category is the single measure used in the research for municipal e-participation level. It employs 17 items to evaluate users’ ability to provide feedback to public administrators and elected officials, along with the presence of discussion boards, blogs, and policy forums. The category also included questions that pertain to the provision of current and accurate information by governments through online newsletters, listservs, and social media channels. Surveys are another important avenue for gathering citizen feedback on government policy and implementation process, and these were also determined on the official websites.

Overall score. Finally, one dependent variable is an overall e-government score that is calculated by adding the five categories together.

Independent variables

Each of the seven independent variables is derived from a different secondary data source. Measurement items are shown in Table 2, and descriptive statistics are shown in Table 3. Measurements were taken every other year in seven surveys between 2003 and 2016. Due to slight changes in the 100 cities evaluated in each survey year, there are 118 cities that have been evaluated, but only 82 of these have been consistently part of at least five of the seven surveys. Therefore, in order to minimize the estimation problems of missing data, we chose to focus only on these 82 cities in the panel regression.

TABLE 2 Descriptive of Data Operationalization and Sources

	Measure	Source
Dependent variables
Total score	Combined score of other scores (privacy, usability, content, services, and engagement)	Global E-governance survey
Privacy score	Score on website privacy features	Global E-governance survey
Usability score	Score on website usability features	Global E-governance survey
Content score	Score on website content features	Global E-governance survey
Services score	Score on website services features	Global E-governance survey
Engagement score	Score on website participation features	Global E-governance survey
Independent variables
Population size (natural log)	Number of people living in the municipality	City population size data from the World Bank
Education level	Continuous variable measuring the number and quality of higher education institutions in the city	Own scale using the QS Global University Rankings
Years of experience	Continuous variable measuring age of website	The Wayback Machine
GDP per capita (natural log)	Extrapolated from year 2014 using average annual growth rate for 2000–2014	Brookings Global Metro Monitor
Chief information officer (CIO)	Binary variable of whether the city government has a chief information officer	City website organizational chart
TABLE 2 Continued
	Measure	Source
Regional competition	Average e-government score for the geographic region of the city	United Nations E-government Index
Democratic level	Categorical variable measuring country democracy with three levels (Free, Partially free, and Not free)	Freedom House measure of “Freedom in the World”
Control variables
Development status	Dichotomous variable measuring country stage of development (Developed or Developing)	United Nations classification of developed and developing countries
Region	Seven world regions: Africa; Central Asia; Western Europe; Eastern Europe; Middle East; Oceania; North America; South and Central America; South and East Asia (North America is the reference category)	United Nations classification of major world regions

TABLE 3 Summary Statistics

Variable	N	Mean	Std. Dev.	Min	Max
Dependent variables
Total score	534	33.68	15.70	3.73	87.74
Privacy score	534	4.66	4.78	0	18.80
Usability score	534	12.06	3.13	2.82	19.38
Content score	534	7.61	3.84	0.16	18.80
Services score	534	5.84	3.88	0	19.83
Engagement score	534	3.53	3.10	0	18.75
Independent variables
Population size (natural log)	511	0.65	0.48	0	1
Education level	457	3.40	3.15	0	15
GDP (natural log)	356	10.23	1.49	8.57	11.15
Experience	498	15.2	5.09	2	21
CIO	461	0.09	0.29	0	1
Regional competition	534	0.55	0.14	0.17	0.89
Democratic level	534	2.46	0.13	1	3
Control variables
Development level
Developed	309
Developing	225
Total	534
Region
Africa	29
Latin America	93
North America	16
Central Asia	16
East Asia	27
South East Asia	49
Eastern Europe	110
Western Europe	115
Middle East	65
Oceania	14
Total	534

Environmental variables. Among the environmental variables, education level was operationalized using data from the QS Global University Rankings to establish the number of universities from the Global E-Governance Survey cities that feature in the top 500 universities in the world. In order to weight the data for better-quality universities as opposed to the quantity, two points were given for a university in the top 100 of the ranking and one point was given for a ranking below 100. Each city’s score was therefore an additive score of the number of quality universities in the city. Finally, population size was measured using World Bank statistics on city population size.

Internal capacity variables. GDP per capita is used as a proxy for the financial capacity of the local government and comes from the Brookings Global Metro Monitor. The Brookings data are incomplete for years prior to 2013, and so figures from earlier years are extrapolated using the average annual rate of GDP growth between the years 2000 and 2014 according to the Global Metro Monitor. The years of experience variable was operationalized using the age of the city’s website. Older websites indicate longer experience of creating e-government programs through websites. The online tool “Wayback Machine” was used to identify the year when the website was launched. The data for whether the city has a CIO were obtained from organizational charts publicly available on the city website. Occasionally, organizational charts could not be found, and thus 73 observations are not included.

Institutional variables. Among the institutional variables, regional competition is operationalized by the mean e-governance index score of the geographic regions where the cities are located. This measure relied on the regional average scores from the UN E-government index for 10 regions according to the UN classification. The higher the average score of the region, the higher the level of regional competition. Democratic level was operationalized using the Freedom House index of democracy (“Free,” “Partly free,” or “Not free”) that assesses seven items on political and civil liberties, such as electoral process, the rule of law, and political participation.

Control variables. Two controls were used: first, the geographic region of the country where the city is located, and second, the development status of the city’s country. The regional control aimed to account for variation in e-government development effects at the regional level. Development status controlled the role of unobserved factors resulting from economic and political status that influence e-government level. Another important area of sample variance comes from the range of sizes in the countries. To distinguish country size effects, all of the variables are estimated in three separate models: (1) cities in small countries; (2) cities in large countries; and (3) all cities together. In the absence of official definitions of “large” and “small” countries, we used modal figures in our data and geographic knowledge to adopt our own classification where a large country has a population of at least 50 million and a land area of at least 200,000 square kilometers according to the CIA World Factbook. This resulted in sample sizes of 39 observations for cities in small countries, and 43 cities in large countries.

Table 3 shows that the number of observations varies between 534 and 356 for the predictor variables due to missing data at the city level. To address this problem, the study included a statistical method of imputing missing data using the data matrix’s observed values. The statistical package missForest with R tests imputed values by averaging random forests (multiple “trees” of variables in a data matrix) with non-parametric assumptions, and has been shown to outperform other methods such as maximum likelihood and multiple imputation (Overton 2016; Stekhoven and Bühlmann 2012). MissForest then computes an out-of-bag (OOB) estimate for the true imputation error. Error estimates approaching 0 show a good level of imputation, while estimates approaching 1 are considered poor. OOB estimates for models estimated in this research ranged from 0.001044 to 0.001236, indicating an excellent OOB score.

RESULTS

Cluster Analysis

The cluster analysis on the five e-government categories estimates the clustering of each city in relation to one of four cluster groups. Based on inspection of the dendrograms, a four-cluster solution was selected as most meaningful. Table 4 illustrates the distribution of cities across the clusters using 2003, 2009, and 2016 data. In order to identify the characteristics of all cluster groups, we estimate the mean score for each cluster in relation to each of the five e-government categories. For each e-government category, we use one-way analysis of variance (ANOVA) to test for significant differences in means across the cluster groups. Table 4 further reports the statistically significant differences between the cluster groups as determined by one-way ANOVA.

TABLE 4 Comparison of Clusters and Analysis of Variance

	N	Privacy	Usability	Content	Services	Engagement
2003	80	2.53 (3.52)	11.45 (3.59)	6.43 (3.60)	4.82 (3.48)	3.26 (3.04)
Cluster 1	7	10.51 (2.81)	16.48 (1.82)	12.52 (1.89)	13.01 (1.58)	9.16 (3.42)
Cluster 2	27	3.12 (3.21)	13.66 (2.16)	8.91 (2.33)	6.24 (1.60)	5.16 (1.93)
Cluster 3	16	2.19 (1.67)	12.01 (1.66)	5.98 (1.69)	4.13 (1.74)	1.38 (.75)
Cluster 4	30	.31 (.70)	7.98 (2.36)	3.01 (1.16)	2.01 (1.44)	1.19 (.98)
F (3,76)		41.17***	49.82***	80.69***	104.42***	65.67***
2009	87	5.57 (5.09)	11.96 (3.09)	8.21 (3.78)	6.68 (3.94)	3.50 (3.11)
Cluster 1	21	9.68 (4.75)	14.23 (2.34)	12.20 (2.46)	11.09 (3.55)	7.46 (3.30)
Cluster 2	13	12.46 (1.22)	12.99 (2.50)	8.71 (2.33)	7.21 (2.29)	2.84 (1.73)
Cluster 3	16	2.35 (2.08)	13.60 (2.43)	10.10 (2.68)	6.99 (2.47)	3.49 (1.67)
Cluster 4	37	2.22 (1.94)	9.61 (2.25)	4.96 (2.06)	3.87 (2.48)	1.49 (1.21)
F (3,83)		64.13***	22.39***	48.80***	31.09***	38.38***
2016	97	5.55 (4.65)	12.38 (2.84)	8.22 (3.85)	6.82 (4.00)	3.87 (3.24)
Cluster 1	12	10.74 (4.05)	16.88 (.93)	13.88 (1.84)	12.46 (2.46)	9.71 (3.14)
Cluster 2	34	8.87 (2.81)	13.05 (1.90)	10.34 (2.77)	9.33 (2.31)	5.01 (1.72)
Cluster 3	20	2.41 (1.88)	12.50 (1.95)	7.20 (1.86)	5.61 (2.10)	2.82 (1.82)
Cluster 4	31	1.92 (2.94)	9.84 (1.97)	4.36 (1.27)	2.67 (1.28)	1.04 (.71)
F (3,93)		52.65***	44.86***	78.01***	95.61***	78.77***

Notes: Please note that the sample size varies among 2003, 2009, and 2016 data. Significance levels are indicated as follows: ***p <.001.

The results of all cluster analysis suggest four levels of world cities. The profiles are shown in Figure 2. Similar to Holzer et al. (2010), the cluster profiles indicate that groups differ regarding scores but are similar in terms of shapes. All cluster groups show high means for usability and indicate low degrees for service and citizen engagement and privacy. The first group of “digitally mature cities” shows relatively high mean scores for all five categories. The second group of “digitally moderate cities” suggests decreased mean scores for the e-government categories, especially for service and citizen engagement. In terms of 2009 data, the second cluster group indicates higher scores for privacy than the first one. However, it performs poorly with regard to service and citizen participation. The third group of “digitally minimal cities” shows a similar mean value for usability as the second group. The 2009 data further point to higher means for usability and content. However, there is a large gap between the mean values of privacy in contrast to digitally moderate cities. Finally, the last group of “digitally marginal cities” suggests low values for all categories, with the exception of usability.

Figure 2. Cluster profiles for 2003, 2009, and 2016.

Further results on the performance of countries according to the e-government dimensions are shown in the appendix. Appendix C sheds light on the top performing cities in e-government based on the 2016 overall e-government score. It further outlines the values of each of the five categories of municipal e-governance for these 20 cities.

Multivariate Results

The results of the time series regression are shown in Tables 5 and 6. Table 5 shows the overall score across e-government categories, while Table 6 shows estimates for each of the five categories. There is some notable variation in antecedents across the e-government categories, as well as between cities in small and large countries. In Table 5, for example, GDP is a consistent predictor across cities in small and large countries. In contrast, regional competition and population size only make a difference in the large countries. But these results are much more nuanced when we look into the different models for each of the e-government categories. In Table 6, interpretation of the models for small countries should be taken cautiously, as three of the models have insignificant F-statistics and are therefore not significant overall. However, the Wald tests suggest that the models are still better than a constant-only model, except for the “participation” model. GDP has a significant effect on e-information and e-services stages in small countries, as does regional competition in e-services stages. Small countries also evidence a surprising finding regarding democratic level, which is that the latter is negatively associated with e-services components of privacy and services.

TABLE 5 Regression Results for Overall E-government Level (Robust Standard Errors in Parentheses)

	Small countries	Large countries	All countries
Population size	–01.39 (6.23)	5.75 (3.05)	5.26** (1.93)
Education	–0.63 (2.81)	–0.33 (0.51)	–0.41 (0.49)
GDP	24.66*** (6.79)	9.80* (4.52)	17.16*** (3.84)
Experience	–0.40 (0.39)	–0.49 (0.40)	0.17 (0.30)
CIO	–3.99 (6.43)	0.96 (7.73)	4.01 (4.41)
Regional competition	22.23 (14.24)	52.34*** (11.80)	27.51** (9.63)
Democracy	–13.86 (7.53)	1.39 (1.92)	1.49 (2.38)
Development level:
Developed	4.87 (4.83)	0.96 (4.22)	–0.48 (3.96)
Region:
Western Europe	1.96 (9.07)	2.13 (8.38)	–2.89 (5.41)
Eastern Europe	4.30 (8.50)	–5.79 (8.61)	–4.56 (5.98)
Middle East	5.08 (10.52)	8.15 (12.23)	–6.44 (6.69)
East Asia	2.60 (9.74)	–4.73 (7.95)	–5.38 (7.21)
Oceania	–5.26 (9.98)	5.41 (4.90)	2.20 (4.12)
South America	–3.54 (8.05)	2.40 (7.99)	1.75 (6.93)
Central America	6.22 (3.85)	–20.03* (7.99)	–3.60 (6.95)
Southeast Asia	1.56 (5.12)	–2.56 (8.15)	–0.26 (6.69)
Africa	–1.32 (3.11)	–7.02 (9.01)	–4.82 (8.31)
Central Asia		–7.34 (8.77)	10.12 (9.74)
Constant	–171.39 (133.86)	–180.41* (83.23)	–240.00*** (59.91)
R-square	0.296	0.452	0.237
F-score	3.294***	6.966***	4.749***
Wald Test	45.05***	154.87***	88.177***
Breusch-Pagan Test	17.37	43.71***	19.01
Panels	17	24	41
N	116	161	277

Notes: Significance levels are indicated as follows: *p <.05; **p <.01; ***p <.001.

Referent region is North America. No Central Asian countries included in the small country sample.

TABLE 6 Regression Results for Five E-government Components (Robust Standard Errors in Parentheses)

	Information		Services		Participation	Information		Services		Participation	Information		Services		Participation
	Cont	Use	Priv	Serv	Part	Cont	Use	Priv	Serv	Part	Cont	Use	Priv	Serv	Part
Population size	−0.17 (1.58)	−2.83 (1.51)	2.57 (3.21)	−0.01 (1.54)	–0.14 (1.69)	1.12 (0.83)	0.16 (1.00)	–0.37 (1.43)	3.36** (1.09)	1.87** (0.59)	0.91 (0.49)	–0.32 (0.47)	1.73** (0.57)	1.72** (0.52)	0.87** (0.41)
Education	0.01 (0.72)	1.12 (0.67)	–2.31 (1.40)	0.05 (0.71)	0.98 (0.79)	–0.04 (0.14)	0.11 (0.17)	–0.03 (0.23)	–0.45* (0.18)	–0.12 (0.09)	–0.09 (0.11)	0.12 (0.11)	–0.14 (0.16)	–0.21 (0.12)	–0.11 (0.43)
GDP	3.98* (0.01)	2.68 (1.63)	12.02*** (3.16)	7.29*** (1.17)	3.31 (1.87)	1.08 (1.23)	0.38 (1.54)	1.49 (2.04)	6.31** (1.93)	2.85** (0.86)	2.93** (1.01)	1.97* (0.88)	5.02*** (1.44)	5.39*** (1.06)	3.02*** (0.89)
Experience	0.01 (0.09)	0.06 (0.09)	–0.45** (0.15)	–0.01 (0.11)	0.07 (0.11)	–0.12 (011)	–0.13 (0.14)	0.01 (0.18)	–0.03 (0.12)	–0.01 (0.06)	0.01 (0.07)	–0.03 (0.07)	–0.05 (0.10)	–0.09 (0.07)	0.00 (0.06)
CIO	–0.49 (1.63)	–2.26 (1.56)	1.72 (3.75)	–0.50 (1.73)	–3.35 (1.74)	0.55 (2.11)	1.83 (2.65)	1.81 (3.50)	–1.07 (3.35)	–2.35 (1.66)	1.77 (1.02)	1.28 (0.98)	1.32 (1.45)	1.80 (1.07)	–0.39 (0.89)
Regional competition	4.47 (3.73)	–7.58* (3.32)	13.63** (6.89)	9.16* (3.95)	0.21 (4.19)	13.73*** (3.16)	6.64 (3.43)	15.84** (4.83)	10.15** (3.48)	5.17* (2.40)	6.34** (2.42)	–0.41 (2.29)	12.27*** (3.58)	5.44* (2.55)	2.20 (2.28)
Democracy	–2.17 (0.26)	–0.81 (1.63)	–6.65* (4.12)	–2.67* (1.32)	0.18 (2.04)	0.46 (0.52)	–0.71 (0.65)	0.52 (0.87)	0.79 (0.57)	0.20 (0.38)	0.21 (0.53)	0.16 (0.50)	0.39 (0.76)	0.52 (0.55)	0.07 (0.47)
Development level:
Developed	0.38 (0.93)	0.57 (0.80)	5.48** (1.91)	–0.27 (0.95)	0.55 (1.06)	–0.51 (1.14)	–0.67 (1.15)	1.29 (1.81)	0.01 (1.24)	0.63 (0.83)	1.44 (1.73)	1.28 (1.10)	–0.30 (1.73)	–0.12 (1.23)	0.00 (1.12)
Region:
Western Europe	–1.06 (1.72)	–1.56 (1.45)	–3.21 (3.65)	–0.15 (1.73)	–1.41 (2.06)	0.61 (3.94)	1.11 (3.32)	4.64 (6.13)	–6.12 (4.67)	0.46 (2.68)	0.75 (2.68)	–0.91 (2.56)	1.45 (3.87)	1.39 (2.82)	–2.27 (2.43)
Eastern Europe	1.21 (1.12)	1.86 (0.95)	–0.57 (2.31)	1.60 (1.13)	2.22 (1.34)	–2.56 (2.90)	–2.10 (2.34)	–4.94 (4.70)	–4.17 (3.55)	–2.31 (1.86)	1.18 (2.53)	2.04 (2.39)	–5.79 (3.73)	–0.98 (2.82)	–1.88 (2.38)
Central Asia						–2.10 (3.78)	–2.44 (2.96)	–1.27 (6.21)	–1.64 (4.69)	0.34 (2.28)	–0.98 (2.10)	–2.24 (2.00)	–4.24 (3.02)	2.03 (2.21)	–2.94 (1.89)
Middle East	0.12 (2.08)	0.56 (1.75)	0.03 (4.54)	–1.46 (2.08)	1.38 (2.52)	–0.37 (3.47)	0.90 (2.88)	–0.12 (5.50)	–7.23 (4.17)	–0.03 (2.32)	0.43 (1.94)	1.83 (1.84)	–1.80 (2.85)	–0.59 (2.04)	–0.89 (1.59)
East Asia	2.04 (2.34)	–0.95 (2.00)	1.99 (4.54)	4.06 (2.37)	4.57 (2.72)	–0.54 (2.98)	1.29 (2.36)	0.41 (4.84)	–2.11 (3.66)	–1.41 (1.84)	–0.29 (2.01)	2.49 (1.92)	–4.74 (2.90)	0.37 (2.12)	–1.94 (1.82)
Oceania	0.05 (2.17)	0.28 (1.82)	–4.28 (4.74)	–0.95 (2.17)	–1.87 (2.62)	–0.62 (3.08)	0.23 (2.42)	–1.94 (5.06)	–7.36 (3.82)	–0.52 (1.87)	2.25 (2.70)	4.01 (2.57)	–2.33 (2.41)	–0.21 (2.84)	–3.22 (2.44)
Southeast Asia	1.49 (1.71)	–0.26 (1.44)	–2.02 (3.57)	3.47* (1.71)	3.76 (2.03)	–3.18 (3.30)	–2.65 (2.63)	–1.29 (5.36)	–6.02 (4.05)	–3.50 (2.06)	0.04 (1.66)	0.68 (1.54)	–2.55 (3.89)	0.47 (1.75)	–1.07 (1.51)
South America	2.48 (3.21)	2.69 (2.71)	4.65 (6.79)	1.41 (3.21)	0.33 (3.85)	–1.50 (3.00)	0.56 (2.42)	–2.36 (4.82)	–6.09 (3.65)	–0.69 (1.92)	1.99 (1.98)	2.50 (1.87)	–0.95 (2.90)	2.44 (2.08)	–1.37 (1.84)
Constant	–27.95 (37.31)	28.47 (31.74)	–62.17 (32.31)	–70.21** (21.54)	–32.83 (43.96)	–30.02* (22.23)	3.35 (27.67)	–14.35 (36.05)	–80.36** (24.68)	–54.97** (16.65)	–40.06*** (15.72)	4.27 (15.05)	–76.29*** (22.45)	–76.64*** (16.53)	–39.38** (13.95)
R–square	0.170	0.148	0.286	0.562	0.147	0.386	0.125	0.255	0.349	0.358	0.194	0.083	0.202	0.231	0.125
F-score	1.610	1.364	3.150***	10.107***	1.367	5.302***	1.21	2.891***	4.510***	4.785***	3.271***	1.227	3.427***	4.071***	1.930*
Wald Test	23.72*	20.96*	29.87*	18.109***	18.11	95.19***	22.95	51.237***	53.304***	86.147***	69.73***	29.10	67.28***	82.133***	38.54**
Breusch-Pagan Test	3.29	13.63	28.26**	8.14	27.19*	37.59**	16.299	13.47	31.72*	31.91*	29.04	24.95	40.76**	28.13	36.896**
Panels	17	17	17	17	17	24	24	24	24	24	41	41	41	41	41
N	116	116	116	116	116	161	161	161	161	161	277	277	277	277	277

Notes: Significance levels are indicated as follows: *p <.05; **p <.01; ***p <.001.

Referent region is North America. No Central Asian countries included in the small country sample.

Large countries are notably different from small countries. In large countries, population size is positively associated with the e-services and e-participation stages, while GDP is only influential at the higher stages. Regional competition is significant across all stages. Finally, democratic level is significant, but only for the privacy component in large countries. Considering the results for small and large countries, as well as the coefficients for all countries, both H2, for the effect of population size on e-information, e-services, and e-participation development, and H4, for the effect of GDP, are broadly supported, although there is substantial variation between small and large countries in terms of which stages are affected. H3, for education level, is not supported. In fact, in the case of e-services in large countries, the effect is significant and negative. The effects of other internal capacity factors, years of experience (H5) and technology leadership (H6), on e-information, e-services, and e-participation development are also not supported. H7, for the effect of regional competition on e-information, e-services, and e-participation development, is widely supported, and as expected, the effect is constant across all e-government stages. Also, like GDP and population, regional competition influences small and large countries in different ways. The positive association between regional competition and e-participation stages is only found in large countries. Finally, H8, for the variable of democracy level and its effect on the development of higher e-government levels of e-services, and e-participation, is not statistically supported.

Best Practices in Global Municipal E-Governance

The following presents some of the best practices from the 2016 Global E-Governance Survey—Seoul, Helsinki, Madrid, Hong Kong, and Prague. The city of Seoul ranked first in the survey, repeating its performance in previous surveys, along with high performance in content, service delivery, and citizen and social engagement. The website of Seoul is user-friendly, clearly formatted, and the homepage is within the appropriate length. These, along with relevant search tools, enable users to locate any required information and encourage them to post their questions and feedback. The website provides a simple “easy reads” section for citizens to quickly review important news and updates in the city. Seoul’s website is also a recognized model for enhancing privacy protection and Internet security, as well as offering a high level of usability. Helsinki ranked second and improved significantly from its sixteenth position in the previous 2014 survey, particularly due to its performance in privacy and citizen and social engagement. The website provides adequate content on its homepage in a clear and understandable manner, especially its privacy statement, which includes information on the use of cookies, and the collection and use of data. The website excels in providing opportunities for citizen participation online with interactive tools, discussion forums, and feedback forms, all of which helped elevate Helsinki to first place in the category of citizen and social engagement.

Madrid ranks third owing to high scores in the categories of content, services, and citizen and social engagement. The website enables citizens to utilize a wide range of services online, such as payment of taxes, license and permit applications, parking fees, fines, environmental and car control services, and other related social services. The license and permit application system is particularly well-integrated with citizens’ online municipal accounts, and citizens can both apply and track their applications online. The website of Madrid also offers a “Debates” section to provide opportunities for interactions between citizens and public administrators. Hong Kong ranks fourth overall and ranks highly in usability and service delivery, particularly due to its “one stop platform” that enables citizens to avail themselves of crucial services related to business, administration, and citizenship. The website provides well-organized content highlighted with important information related to budget and policy. The city takes strides to protect users’ privacy and meet security concerns, and also addresses the needs of non-citizens and expatriates, who form a sizable portion of Hong Kong’s residents. Prague took the fifth position, and also was highly ranked in privacy and content categories. The privacy statement addresses important questions related to users’ privacy and security, and the website enables users to locate relevant information in the content category, such as meeting minutes, mission statements, budget reports, and performance information.

DISCUSSION AND IMPLICATIONS

This article contributes to the stream of literature on e-government development through the following contributions, covered only to a marginal extent thus far. First, the article adds to the studies examining the development and current stages of e-government by extending the sample analyzed. Whereas Reddick (2004) and Coursey and Norris (2008) focused on American cities, this study analyzes e-government and e-governance from an international perspective, and provides evidence for the current stage of digital governance from 80 global cities. By applying cluster analysis, it provides an empirical classification of e-government types, reflecting previous studies on a staged adoption of e-government in terms of world cities. This typology provides a basis for comparing e-government stages and determinants across countries. Second, the study uses city-level data which have not been used before to analyze the stages of municipal digital governments. The municipal e-governance survey provides data for analysis of cities’ e-government development in three stages—e-information, e-services, and e-participation—and thus adds to the empirical literature that focuses on the earlier stages of e-government (e.g., Moon 2002; Reddick 2004; Coursey and Norris 2008).

The findings of the regression analysis in the article corroborate earlier findings about the different institutional antecedents associated with different dimensions or practice areas of e-government (e.g., Ahn 2011; Chen and Hsieh 2009; Lee et al. 2011; Wang and Feeney 2016), and generate stronger research insights by systematically drawing on information science literature, in addition to public administration and public management. In addressing these associations across three different stages of e-government across time, the results provide more granular evidence of the associations, and explain why some governments may move faster than others and in specific stages. Cross-dimensional approaches such as these are important for understanding broad e-government trends. Furthermore, as the cluster analysis has shown, the dimensions of e-government develop simultaneously; regardless of the level of digital maturity evidenced by a municipal government, the dimensions of e-government are derived in the same order, with usability in the lead and citizen and social engagement following behind. This means that digitally mature cities are ahead on all dimensions, rather than being ahead on some, but not others. The regression analysis supports this view by showing that many success factors, especially GDP, population size, and regional competition, are associated with all three stages. These support earlier findings on GDP per capita (e.g., Edmiston 2003; Ma 2014) and regional competition (e.g., Lee et al. 2011).

However, there are also some specific factors that can contribute uniquely to individual stages. The strongest effects of the antecedents on the higher levels of e-government development occur in larger countries, which echoes earlier views that complexity is a driving factor of higher e-government innovation (Gallego-Álvarez et al. 2010; Norris and Moon, 2005). We add to this theory of e-government innovation by testing the variable of regional competition, which points to a further driver of e-government development, especially in larger countries in the higher e-government stages. Cities in larger countries face more complex tasks demographically and geographically, and globalization forces of migration and international labor supplies lend themselves to increasing competition (Shipan and Volden 2012; Wang 2001; Weare et al. 1999). As we hypothesized, institutional factors such as democratic level also matter for the higher stages of e-government, but the relationships among these variables are more complex than we expected. Notably, democratic level, which does influence the e-services stage in one component (privacy and security), only seems to make a difference in large countries. In small countries, democratic level is inversely related to the e-services stage. This finding might suggest that small, democratic countries are better at providing traditional services than online services. While prior research does suggest that democratic countries tend to have better capacity and political support for e-government to be more fully developed, such research is mixed, or finds that differences depend on how democratic level is measured (e.g., Lee et al. 2011; West 2005). Small, non-democratic governments also have an advantage when it comes to building online services quickly across central agencies (Gulati et al. 2014), and may do so to enhance support and legitimacy among citizens and among international peers (Stier 2015). Large democratic countries, in contrast, may have a comparative disadvantage, as well as less incentive, in driving forward fast gains in e-government development. However, many of these interesting nuances of democratic variables are largely conjecture here, and further institutional analysis is needed to piece together the puzzle.

Other results from the research were surprising in that variables were not found to be supported, despite having support in earlier research. Notably among these are the internal variables relating to technical expertise and leadership, as well as education level. Each of these variables had evidence supporting them, but other studies also have failed to confirm them. Education, in particular, has been found to be interdependent on inclusion of wealth and GDP in models (Tolbert 2008; West, 2005), and the latter was a strongly significant variable in this study. Further, while internal variables are important, technology transformation and innovation theory tends to emphasize political and institutional factors over internal administrative or technical capacity (Jun and Weare 2010). Institutional and environmental variables tend to be more fixed compared to internal ones, and over the long term they may play the bigger role needed to embed technologies in processes of innovation (Fountain 2004; West 2005; Yang 2003).

Limitations and Directions for Future Research

This study is based on a purposive sample of cities that are in the most Internet-advanced countries in the world. However, the advantage of this is that, by selecting the largest cities in the most wired countries, we are comparing similar entities with a certain level of technological sophistication, and by selecting the largest city in each such country, we further underscore that attribute. Thus, by controlling for “most wired” and largest cities, we are making it possible to look for factors affecting e-government that are not simply a function of the available technology. The study has drawn conclusions from statistical analysis, but there is much more that can be learned from these cities and the determinants of e-government level through an explanatory case study approach. While the present study finds that statistical predictors are important, a case study approach can reveal what the specific mechanisms are, as well as detailing and identifying the common characteristics or distinguishing factors among groups of countries. Case study approaches may also address a limitation in this article that some of the variables identified in prior literature, such as citizen demand, political support, and political competition, could not be measured because of a lack of data for city-level characteristics. Cases focusing on small groups of cities could assess these factors in qualitative ways or through surveys to determine whether they also play important roles in e-government success.

As this study’s results suggest that the determinants of e-government level vary across the different dimensions of e-government, future research could also address the interdependency of these causal relationships. As the dimensions are not independent of one another, there must be interactions among e-governance predictors and outcomes. Case-based approaches are suitable for these types of questions, but so too are quantitative approaches that concentrate on the relationships among a limited number of moderators and outcome variables. There may be more specific associations or combinations of factors associated with specific dimensions. Citizen engagement is among the more critical areas to address in this respect, as the present work is not the first to suggest that government efforts in this area are frequently hampered.

This article offers a global perspective on e-governance development over time. The results of the cluster analysis revealed that there are four global clusters of e-government development relating to mature, moderate, minimal, and marginal stages. The results of the regression analysis show support for the findings of earlier empirical studies on the predictors of the level of e-government provision. However, the original contribution of this study is to show, at a global level, that these predictors are associated with the stages of e-government in different ways. Specific institutional and environmental factors may help to foster the growth of e-government, but this fostering does not occur equally across the spectrum of e-government. Rather, e-government performance responds to specific conditions.

Finally, by revealing the patterns of e-government performance across three stages of e-government, the article provides several specific points for intervention by e-government practitioners and policymakers in government. Such interventions may wish to focus on improving specific areas of e-government performance that are lacking. While governments can provide a conducive environment by focusing on a broad range of institutional and environmental factors, they may wish to improve in the important dimension-specific areas suggested in this article’s research. The knowledge of the factors and barriers of e-government success can help administrators to devise specific strategies to implement e-government projects that will meet their stakeholder demands, remain cost-effective, and reduce the possibility of IT project failure.

The phenomenon of e-government has multiple implications at all levels of government, particularly the local government level, where citizens tend to have greater interaction with public officials. This research examines the trends and best practices in municipal e-government and, more importantly, sets the tone for the transformation from e-governance to smart governance among global municipalities. The dimensions discussed in the article can contribute to crucial steps in the progress towards becoming a smart city. Future studies in municipal e-governance need to examine how governments are building smart cities, and to what extent they involve citizens and stakeholders in the process. What are the various stages of development of smart cities, and what are the critical success factors and barriers affecting these initiatives? More importantly, future research needs to focus on how to measure the performance of smart cities and analyze them from a global comparative perspective.

The best practices discussed in the article provide guidance and lessons for cities that have recently adopted e-government initiatives. Moreover, the comparison of the top-performing cities over the years shows that many cities are not able to repeat their success over consequent surveys. This trend suggests that administrators need to ensure that their e-government applications are not only effective and efficient, but also sustainable for long-term success by addressing specific environmental and institutional barriers to innovation. Perhaps the most important implication for cities adopting e-government is to begin to conceptualize e-government in terms of multiple dimensions, rather than an overall phenomenon. This will enable cities to implement e-government through incremental stages that will allow for citizens to provide their opinions, and for administrators to respond to them in a timely manner, as they implement the new technologies.

Additional information

Notes on contributors

Alex Ingrams

Alex Ingrams ([email protected]) is an assistant professor in the Institute of Governance at Tilburg University. His research interests are in digital government reform, and particularly impacts on transparency and accountability. He also researches other topics focused around the intersection of technology change and governance reform, such as open government, open data, and big data.

Aroon Manoharan

Aroon Manoharan ([email protected]) is an associate professor in the McCormack Graduate School of Policy and Global Studies at the University of Massachusetts, Boston. He received his Ph.D. from the School of Public Affairs and Administration, Rutgers University-Newark. His research interests include e-government, performance measurement, public reporting, citizen participation, strategic planning, and comparative public administration. His recent research focuses on the role of academic programs in strengthening the IT capacity of cities and local governments.

Lisa Schmidthuber

Lisa Schmidthuber ([email protected]) is a doctoral researcher in the Institute for Public and Nonprofit Management at Johannes Kepler University Linz, Austria. Her research interests include innovation management in the public sector in general, and focus on value creation by integrating citizens in innovation processes in particular.

Marc Holzer

Marc Holzer ([email protected]) is a distinguished professor of public administration at the Sawyer Business School, Suffolk University. He is a Past President of the American Society for Public Administration (ASPA) and the former founding dean of the School of Public Affairs and Administration at Rutgers University. He has published numerous articles, books, and monographs in public administration research and theory. He currently chairs the ASPA Endowment supporting education in strong, effective, and ethical public governance.

APPENDIX A

List of Works Cited in Table 1

APPENDIX B

100 Cities Selected by Continent (2016)

Table

Africa (7)
Accra (Ghana) Cairo (Egypt) Casablanca (Morocco) Johannesburg (South Africa)	Lagos (Nigeria) Nairobi (Kenya) Tunis (Tunisia)
Asia (30)
Almaty (Kazakhstan) Amman (Jordan) Baghdad (Iraq) Bangkok (Thailand) Colombo (Sri Lanka) Dhaka (Bangladesh) Dubai (United Arab Emirates) Ho Chi Minh City (Vietnam) Hong Kong (Hong Kong, China) Jakarta (Indonesia) Jerusalem (Israel) Karachi (Pakistan) Kathmandu (Nepal) Kuala Lumpur (Malaysia)	Kuwait City (Kuwait) Macao (China) Manama (Bahrain) Manila (Philippines) Mumbai (India) Muscat (Oman) Riyadh (Saudi Arabia) Seoul (Republic of Korea) Shanghai (China) Singapore (Singapore) Tashkent (Uzbekistan) Tbilisi (Georgia) Tehran (Iran) Tokyo (Japan) Yerevan (Armenia)
Europe (40)
Amsterdam (Netherlands) Athens (Greece) Belgrade (Serbia and Montenegro) Berlin (Germany) Bratislava (Slovak Republic) Brussels (Belgium) Bucharest (Romania) Budapest (Hungary) Chisinau (Moldova) Copenhagen (Denmark) Dublin (Ireland) Helsinki (Finland) Istanbul (Turkey) Kiev (Ukraine) Lisbon (Portugal) Ljubljana (Slovenia) London (United Kingdom) Luxembourg City (Luxembourg) Madrid (Spain) Minsk (Belarus)	Moscow (Russian) Nicosia (Cyprus) Oslo (Norway) Paris (France) Prague (Czech Republic) Riga (Latvia) Rome (Italy) San Marino (San Marino) Sarajevo (Bosnia and Herzegovina) Schaan (Liechtenstein) Sofia (Bulgaria) Stockholm (Sweden) Tallinn (Estonia) Tirane (Albania) Valletta (Malta) Vienna (Austria) Vilnius (Lithuania) Warsaw (Poland) Zagreb (Croatia) Zurich (Switzerland)
North America (11)
Castries (St. Lucia) Guatemala City (Guatemala) Hamilton (Bermuda) Mexico City (Mexico) New York (United States) Panama City (Panama)	Saint Joseph (Costa Rica) San Juan (Puerto Rico) San Salvador (El Salvador) Santo Domingo (Dominican Republic) Toronto (Canada)
South America (10)
Asuncion (Paraguay) Bogota (Colombia) Buenos Aires (Argentina) Caracas (Venezuela) Guayaquil (Ecuador)	Lima (Peru) Montevideo (Uruguay) Santa Cruz de la Sierra (Bolivia) Santiago (Chile) Sao Paulo (Brazil)
Oceania (2)
Auckland (New Zealand)	Sydney (Australia)

Source: Rutgers eGovernance survey, 2016; N = 97.

APPENDIX C

Top 20 Cities in Digital Governance (2016)

Table

Ranking	City	Total score	Five Dimensions of Municipal E-Governance
			Privacy score	Usability score	Content score	Services score	Engagement score
1	Seoul	79.92	13.33	15.94	17.30	16.89	16.46
2	Helsinki	69.84	14.44	17.50	13.17	11.80	12.92
3	Madrid	69.24	12.22	16.56	15.56	13.44	11.46
4	Hong Kong	67.56	12.59	17.81	13.65	14.75	8.75
5	Prague	66.48	14.44	15.31	15.08	11.64	10.00
6	Tallin	62.10	8.52	17.50	14.13	15.08	6.88
7	New York City	62.02	12.59	14.06	15.71	13.61	6.04
8	Bratislava	60.34	11.85	17.19	13.97	7.54	9.79
9	Yerevan	59.61	3.70	17.81	14.92	12.13	11.04
10	Vilnius	59.12	14.44	15.63	12.22	10.16	6.67
11	Buenos Aires	57.88	11.85	16.25	10.00	10.82	8.96
12	Tokyo	57.04	8.89	18.13	12.54	13.11	4.38
13	Singapore	56.03	9.63	14.38	10.16	13.11	8.75
14	Moscow	54.73	2.59	16.88	13.97	12.13	9.17
15	Oslo	54.37	14.07	10.94	14.44	10.33	4.58
16	Amsterdam	54.36	10.37	14.38	12.86	11.97	4.79
17	Auckland	54.27	8.89	14.06	12.22	11.80	7.29
18	London	52.54	12.22	15.00	10.00	11.15	4.17
19	Lisbon	51.68	9.26	12.50	11.90	8.85	9.17
20	Sydney	50.08	8.15	15.94	10.16	10.00	5.83