ICT4D and the capability approach: understanding how freedom of expression on ICTs affect human development at the country-level

ABSTRACT

The purpose of this paper is to examine freedom of expression on ICTs as a conversion factor that impacts the success of information and communication technology for development (ICT4D). Prior country-level econometric research on ICT4Ds has measured development using resource- or utilitarian-based approaches. We utilize the capability approach, a people-centered approach which presents the opportunity to look at conversion factors in a country. Using the capability approach framework, four country-level conversion factors of ICTs are identified as enablers/restrictors of either opportunity or process freedoms, and then are hypothesized with relation to human development. Using archival data and a 2SLS model, we test ICT cost, ICT infrastructure, and the interaction effect between e-participation and freedom of expression on ICTs to predict a country’s human development. Results suggest that both ICT cost and infrastructure significantly affect human development. Furthermore, freedom of expression only impacts human development with high levels of e-participation.

KEYWORDS:

• ‌Freedom of expression
• capability approach
• human development
• two-stage least squares

Introduction

Information and communication technology for development (ICT4D) research has done little to understand how freedoms of expression on ICTs impacts development. The United Nations Development Programme (UNDP), who often reports on disparities in and between countries, last brought up the issue of ICT disparities, and more closely the freedom of expression on ICTs as a concern in 2010 along with issues over e-participation, ICT cost, and ICT infrastructure (Hamel, 2010). No country-level research, to our knowledge, has since reported on Hamel’s freedom of expression policy recommendation to:

… ensure that participation in the networked society is safe by defending freedom of expression; empowering users to make use of ICTs in such a way that does not result in backlash and censorship. (p. 58)

Since this report, global Internet freedom has declined every year (Shahbaz et al., 2022). Governments across the world are deconstructing the global Internet to control online spaces within their countries. This government takeover of the Internet can sway an individual’s choice to e-participate in government websites; the same government websites whose primary purpose is to reduce disparity and promote development in communities and national governments (Gorokhovskaia et al., 2023).

What is not fully understood is how freedom of expression on ICTs impacts development. The mere presence of ICTs within a country does not necessarily lead to quality-of-life improvements. Chacko (2005) highlights the debate on the role of ICTs in global, regional, and national human development. Critics believe that ICTs, as a resource, are not sufficient for development on their own, and at times can hamper development if not implemented with caution (Radovan, 2013; Rothe, 2020; Unwin & Unwin, 2017). Proponents believe that ICTs provide the ability to ‘leapfrog’ stages of development (Alhassan & Adam, 2021; Bankole et al., 2013; Lee et al., 2017). ICTs also can breakdown information barriers and bring forth economic opportunities (Chacko, 2005; Sein & Harindranath, 2004), as long as no unacceptable trade-offs occur in lesser developed countries (Bankole et al., 2013; Lee et al., 2017; Samoilenko & Ngwenyama, 2011). Both sides can agree that caution and context are needed when implementing ICT for development. The degree of freedom of expression for citizens is a contextual example that factors into the conversion of online platforms to development: Citizens will not participate in government services when governments have the power to censor their messages, ban their online accounts, restrict their access to services, or even throw them in jail. Thus, we propose the following research questions:

RQ1: How do the combined levels of freedom of expression and e-participation impact human development (i.e. interaction effect)?

The Capability Approach (CA) creates an appropriate framework for this study because it analyzes conversion factors through a human-centric lens (people are the agents of change) as opposed to a resource-centric lens (ICTs are the agents of change). The CA is often paired with human development for its measure of development because of this human-centric lens and understanding of human-based conversion factors, and in following with previous country-level research utilizing the CA-based model as an explanatory tool for empirical research (Zelenkov & Lashkevich, 2022). This paper makes a novel contribution by investigating the impacts freedom of expression on ICTs interacting with e-participation as well as ICT cost and ICT infrastructure as conversion factors in a CA-based model. Our dependent variable, in accordance with the CA, will use human development as measured through the human development index. Using archival data, the current study seeks to understand how each of these four conversion factors plays a role in a country’s human development. The results of this study can help support stakeholders’ claims towards ICT4D initiatives.

RQ2: How will Hamel’s four conversion factors (ICT cost, ICT infrastructure, e-participation, and freedom of expression on ICTs) impact a country’s human development?

The rest of this paper is organized as follows. The paper's second section discusses the literature review, which includes the capability approach as a basis for our theoretical model and hypotheses. The third section introduces data collection. The fourth section presents the analysis and results. Finally, the last section summarizes with discussion, limitations, and concluding remarks.

Literature review

Capability approach

Amartya Sen, an economist working for the World Bank in 1979, developed the capability approach (CA) to allow a multidisciplinary method of defining development (1979). He argues that development should not only include income but other important aspects of human development and agency that a person has reason to value. The CA was designed to measure development at the individual’s capability level, placing human freedoms in the front and center of analysis instead of ICTs. ICTs are then seen as a means to an end, rather than an end in itself (Zheng & Walsham, 2008).

As argued previously, the CA is a human-centric view of development. People are agents who actively use their capabilities to achieve goals. The Oxford Poverty & Human Development Initiative (OPHI) expanded upon Sen’s definition of development in a background paper for the 2010 Human Development Report, saying:

[The purpose of human development is to] … expand people’s freedoms – the worthwhile capabilities people value – and to empower people to engage actively in development processes … People are both the beneficiaries and the agents of long term, equitable human development, both as individuals and as groups. (Alkire, 2010)

This core definition is what drives the CA, a normative framework for evaluation and assessment of human development and social arrangements, as well as policy design in societies (Robeyns, 2005). It also incorporates the concept of freedom as the mechanism for development. The capability approach is made up of five core elements: (1) resources, (2) opportunity freedom, (3) process freedom, (4) functionings, and (5) conversion factors as summarized from the large volume of Sen’s work (Nyemba-Mudenda & Chigona, 2018; Robeyns, 2005; Zheng, 2009).

Resources, also sometimes called commodities, refer to the goods and services that individuals can turn into real opportunities for desired human development. These resources are the means to achieve for an individual. An example of this is a car as a means of mobility. In ICT4D research, the resources are ICTs that possess the characteristic of information. Information derived through ICTs can affect the decision-making process for all dimensions of human development, making ICTs a strong influencer of human-centric development.

Just the mere presence of a resource does not bring about freedoms or human development. The CA points to what individuals potentially can achieve. In Sen’s definition of development, ‘the worthwhile capabilities people value,’ he refers to opportunity freedom. He believed that development took place by expanding people’s choices (i.e. capability sets) to obtain their ideal lives. Opportunities are added when resources, such as goods and services, become present. Resources then can expand one’s capability set. A capability set refers to all available choices an individual may choose from to increase their human development. Importantly, this represents the opportunity freedom that gives rise to the potential of human development, not the actual choice (Nyemba-Mudenda & Chigona, 2018).

The third core element is process freedom, which ‘empower[s] people to engage actively in development processes’ (Sen, 2001). Process freedoms empower people to act as agents on behalf of what matters to them. This means, for example, that people can partake in political processes, whether that be through democracy in government, protesting on the streets for something they believe in, or voicing their opinion in a room with alternative viewpoints. An active agent believes their choice matters for their human development, as opposed to a passive agent who may be viewed as a beneficiary of a larger organization (e.g. markets or governments). Both active and passive agents derive human development from their capability sets within opportunity freedoms (Nyemba-Mudenda & Chigona, 2018). For process freedoms, human development is derived from the active selection of capabilities aligned with individual values (Nyemba-Mudenda & Chigona, 2018). Thus, while passive and active agents derive human development from opportunity freedoms, active agents may derive human development from process freedoms.

An example of this would be two people with nearly identical capability sets. One may choose to live in government-provided housing as a means of safety, while the other chooses to be homeless for fear of constantly being monitored by their government. This second individual became an active agent for themselves and chose to forgo the safety benefits of government housing in return for what they believe is privacy. Process freedom is then a decision for individuals to either choose opportunities that would maximize their human development or act as agents for what they believe matters.

A functioning is the intended, and sometimes unintended, human development outcome as a result of a chosen capability option (Sen, 1999). Examples of functioning include having a good job, being nourished, being safe, expressing oneself, being literate, and being part of a community. For each listed functioning, there is an actively chosen opportunity derived from a resource to go along with it, and as such, increasing opportunity and process freedoms become the key to development instead of the functionings themselves.

Whether a country can achieve a certain level of development depends on the context within the country. Because no country is the same, resources that are implemented in one country will be affected differently through its opportunity and process freedoms than in another country. Resources, such as ICTs, are studied in different contexts to see relations between new capability options introduced and the peoples’ preferred functionings. The CA conceptualizes context as conversion factors. Three kinds of conversion factors can enable or restrict a country’s opportunity and process freedoms: personal characteristics which refer to physical condition, biological health, and psychological well-being (e.g. age, gender, literacy, mental health, metabolism, mobility, religion); social arrangements which refer to social norms, social hierarchies, cultural traditions, and institutional policies (e.g. rules & regulations, gender inequality, power relations); and environmental dynamics which refer to physical and organizational environments (e.g. climate, geography, infrastructure) of a country (Robeyns, 2005).

Developing capabilities from resources such as goods and services requires conversion factors to be met, otherwise the resource is restricted by the conversion factor. For example, physical disabilities (personal) may prevent some citizens from utilizing a traditional car that could have expanded their opportunities within their capability set. If a country has laws that restrict certain genders (social) from driving, then a car may not expand that gender’s opportunities. If a country has no road infrastructure (environmental), then a car may be restricted in its mobility. Likewise, conversion factors also enable or restrict a country’s choices in process freedoms. For example, some resources, such as the government housing example used earlier in this paper, may increase one dimension of human development (security) but diminish another dimension of human development (privacy) due to personal, social, or environmental conversion factors. This may lead to a decision to reject the resource based on human development preferences. Thus, conversion factors enable or restrict people’s opportunity and process freedoms by imposing boundary conditions on both resources and the process freedom’s choice enactment (see Figure 1).

Figure 1. Sen’s Capability Approach Framework, adapted from Nyemba-Mudenda and Chigona (2018).

Conversion factors on human development

There is a belief among scholars that if development is to take place, factors beyond economic growth and subjective measures of development should be considered, such as factors that promote people’s choices (Peet & Hartwick, 1999) and account for various dimensions of human development. Moving away from resource- or utility-centric views to a human-centric one provides researchers with a means to empirically study the effects of a country’s personal, social, and environmental conversion factors on their human development (Adaba et al., 2019; Dasuki & Effah, 2021; Faith, 2018; Hatakka et al., 2020; Hatakka & Lagsten, 2012; Nambiar, 2013; Nyemba-Mudenda & Chigona, 2018). As a human-centric theory, the CA posits that conversion factors enable or restrict people’s ability to convert resources/commodities into an achieved state of human development (Sen, 1999).

Previous ICT4D research has studied various conversion factors within the CA at the individual-level. Adaba et al. (2019) studied mobile money users within northern Ghana and found that ICT skills (personal) and ICT Infrastructure (environmental) were factors for users wanting to use mobile phones in remote regions. Dasuki and Effah (2021) studied mobile phone use of internationally displaced people in Nigeria and found many inhibitors of capabilities such as lack of identification/bank account, personal income, and literacy (personal); social inclusion, criminal activity, and gender bias (social); and ICT Infrastructure (environmental). Faith (2018) focused on mobile phone use by low-income women in the UK, finding ICT skills (personal) and gender bias (social) to be the main inhibitors for this group. Hatakka and Lagsten (2012) analyzed students using Internet resources in developing countries, discovering many restrictors: PC cost, literacy, motivation, interest, trust, personal use, and security (personal); copyright infringement, censorship, and educational support (social); and of course, connectivity cost and ICT infrastructure (environmental). Also, within education, Hatakka et al. (2020) observed inhibitors of ICT within Kenyan study circles to promote economic opportunities, finding limited income and ICT skills (personal); gender bias, available services, and educational certification (social); and ICT infrastructure (environmental) as limitations. Nambiar (2013) studied how non-governmental organizations within Malaysia can improve members capabilities, finding that institutional changes required adjustments with social justice and gender bias. Nyemba-Mudenda and Chigona (2018) observed pregnant mothers from Malawi using mobile technology for healthcare services. Major inhibitors for use of the service included ICT skills (personal), gender bias (social), and ICT infrastructure (environmental).

More recently, ICT4D research has studied conversion factors at the country-level. Zelenkov and Lashkevich (2022) analyzed three pillars of the Legatum Prosperity Index as conversion factors which measures empowered people (personal), inclusive societies (social), and open economies (environmental). ICTs were significant on all 3 conversion factors in a 115-country sample except for social factors within developing countries. Looking across the ICT4D research on conversion factors, most of the focus has been at the individual-level with an emphasis on personal and social conversion factors. We believe that the current study will help the conversion factor research by including freedom of expression on ICTs as a new conversion factor and by bolstering the understudied country-level research. You may review Table 1 for the simplified list of previously used conversion factors.

Table 1. Studied conversion factors.

Authors	Type	Concept
Adaba et al. (2019); Faith (2018); Hatakka et al. (2020); Nyemba-Mudenda and Chigona (2018)	Personal	ICT skills
Dasuki and Effah (2021); Hatakka and Lagsten (2012); Hatakka et al. (2020)	Personal	Personal income
Dasuki and Effah (2021); Hatakka and Lagsten (2012)	Personal	Literacy
Hatakka and Lagsten (2012)	Personal	Motivation, interest, trust, personal use, security
Dasuki and Effah (2021)	Personal	Identification/bank account
(Zelenkov & Lashkevich, 2022)	Personal	Empowered people
Dasuki and Effah (2021); Faith (2018); Hatakka et al. (2020); Nambiar (2013); Nyemba-Mudenda and Chigona (2018)	Social	Socio-cultural norm (gender bias)
Nambiar (2013)	Social	Social justice
Hatakka and Lagsten (2012); Hatakka et al. (2020)	Social	Educational certificates/support
Dasuki and Effah (2021); (Zelenkov & Lashkevich, 2022)	Social	Social inclusion, criminal activity
Hatakka and Lagsten (2012)	Social	Legal (copyright), Openness (censorship)
Adaba et al. (2019); Dasuki and Effah (2021); Hatakka and Lagsten (2012); Hatakka et al. (2020); Nyemba-Mudenda and Chigona (2018)	Environmental	ICT infrastructure
Hatakka and Lagsten (2012)	Environmental	Connectivity cost
(Zelenkov & Lashkevich, 2022)	Environmental	Open economies

Freedom of expression on ICTs has not been studied in ICT4D research as a conversion factor under the CA lens. Hamel’s (2010) UNDP report on the potentials of ICTs for development points toward freedom of expression on ICTs as a country-level conversion factor that can enable or restrict human development along with e-participation, ICT cost, and ICT infrastructure. Even though Hamel’s report is more than 10 years prior, research has only focused on e-participation (Chohan & Hu, 2022; Karkin & Janssen, 2020; Mahmood et al., 2019; Pirannejad, 2017), ICT usage cost (Hasan et al., 2022; Jayaprakash & Pillai, 2022; Qureshi & Najjar, 2017), and ICT infrastructure (Alderete, 2017; Chatterjee, 2020; Pradhan et al., 2022) from his report. The authors here suggest empirically studying freedom of expression on the net, along with e-participation, ICT cost, and ICT infrastructure in a country-level model. The latter 3 factors are already established conversion factors from research, and will be kept within the model as a basis to extend the nomological network. We intend to fill the gap in the research by analyzing freedom of expression on ICTs in our country-level model. This current study empirically tests these four conversion factors to understand their relationship on human development. Figure 2 displays the research model and hypotheses.

Figure 2. Research model and hypotheses.

In the next few sections, we will map out each policy recommendation to one of the three categories of conversion factors (personal, social, or environmental) along with its association to either enabling or restricting opportunity and/or process freedoms and hypothesize their relationship with human development.

ICT cost and human development

ICTs are often priced above the means of marginalized groups within developing countries, thereby negating any benefits that could be brought. Some of the poorest countries only have a single provider (Guida & Crow, 2009) and they provide access only to the wealthiest who can afford it. High telecommunications costs that inhibit Internet use are most apparent in these poorest countries (Mann, 2003). In a meta-analysis of 253 ICT4D studies, less than 20% reported the costs associated with the interventions, begging the question of ‘how much good evidence exists about the impacts, or effectiveness, of ICTD interventions?’ (Brown & Skelly, 2019). Survey results from 460 respondents identified ICT cost as the second-largest issue to users, just behind electricity, especially in developing nations (Hosman & Armey, 2017). Their open-ended responses included the need for finding a ‘sweet spot’ that balances quality through durability and functionality with the correct expense. Otherwise, marginalized groups get caught in higher long-run expenses associated with charging, connectivity, and repairs of lower-quality items. Simple ICTs, such as mobile phones and broadband Internet, are a means of providing knowledge and information necessary to reduce poverty, reallocate resources, and empower citizens. Mobile phones, for example, have become the most significant ICT to bridge the digital divide in developing countries (Rashid & Diga, 2008). ICT costs then become a problem when ICTs are implemented in these often low populated and impoverished communities that may need ICTs the most. If citizens are unable to access ICTs because ICT costs are too high, then nothing is added to their capability set. The citizens are marginalized even further no human development is achieved.

We have categorized ICT cost as an environmental conversion factor that may restrict a resource from being added to a capability set (i.e. opportunity freedom). Here, we predict that higher levels of ICT cost can restrict disadvantaged groups from even accessing or acquiring ICTs. Therefore, we hypothesize:

H1: Higher levels of ICT costs will negatively affect human development in a country.

ICT infrastructure and human development

An ICT infrastructure is necessary for citizens to access and use ICTs. ICT infrastructure includes the accessibility and usage of computers/phones and broadband Internet (Chatterjee, 2020) and the skills necessary to use them (Dedrick et al., 2013; Dewan et al., 2010). Part of building a good infrastructure for ICT means also investing in education projects and literacy campaigns that will help foster the skills needed for ICT use. The combination of ICT accessibility, use, and e-skills is the core of ICT infrastructure.

Several studies have investigated the impacts of ICT infrastructure within initiatives of e-banking in Ethiopia (Borena & Negash, 2016), immunization programs in Kenya (Holeman & Barrett, 2017), and country-level development (Andoh-Baidoo et al., 2014; Ganju et al., 2016; Meso et al., 2009). All the reported studies above have suggested that lifestyles and living standards are greatly impacted through ICT infrastructure as an enabler of change. Past studies have shown a positive statistical correlation between telephone and Internet access and the success of entrepreneurship, business development, and the incomes of the poor (Forestier et al., 2002). Even more so in developing countries, mobile phones have become the most important ICT in bridging the digital divide (Rashid & Diga, 2008). More recently, a study measuring the impact of high-speed broadband on the innovation capabilities of rural firms, showing increases to firm performance (Rampersad & Troshani, 2020).

We have categorized ICT infrastructure as an environmental conversion factor that can enable or restrict a resource from being added to the citizens opportunity freedom. Without access to core ICT infrastructure, citizens are unable to convert ICT resources into their capability set and no human development is achieved. Conversely, a nation that has high levels of ICT infrastructure has little worries when implementing an ICT for its community to access. As such, we hypothesize:

H2: Higher levels of ICT infrastructure positively affect human development in a country.

E-Participation, freedom of expression, and human development

Citizen participation is critical in establishing and maintaining effective democratic governance (Teorell, 2006; Verba et al., 1995). When a citizen participates in governmental decision-making processes, they enhance their capacity to make effective choices and translate the choices to desired actions and outcomes (Alsop & Heinsohn, 2005). Tai et al. (2020) refer to e-participation as the ‘citizens’ use of ICTs to engage with public affairs and democratic processes. E-participation is a way to empower citizens by allowing them to be involved in decision-making processes that they would otherwise not be involved in. Higher levels of e-participation give access to higher transparency and public information within governments, provides ways of giving feedback, and allows the opportunity to cooperate with government officials in making policies and services through ICT platforms. E-participation enables all stakeholders, policymakers and citizens alike, to interact and better understand the obstacles that exist in people’s lives (Day & Greenwood, 2009). Policymakers need feedback from their communities if they are truly going to understand the needs of their citizens, and the citizens are empowered by partaking with public affairs and democratic processes through e-participation.

However, e-participation alone is not sufficient to generate human development. Freedom of expression also empowers citizens by allowing the means to express alternative views and information. Freedom of expression, defined by article 19 in the United Nations ‘Universal Declaration of Human Rights’ (1948), is the right ‘to hold opinions without interference and to seek, receive, and impart information and ideas through any media and regardless of frontiers.’ The true purpose of freedom of expression is to call attention to the desired change for what one believes in. Freedom of expression within ICTs is, then, a virtual method of expressing one’s opinion that transcends the confines of physical distances.

We have categorized e-participation and freedom of expression within ICTs as social conversion factors that will enable or restrict the population’s freedom to choose (i.e. process freedoms) to be involved in decision-making processes dependent on the level of freedom of expression on ICTs in the country. This is because e-participation and freedom of expression on ICTs are closely tied to each other. Both e-participation and freedom of expression on ICTs are tools to help empower people by allowing them to become agents for themselves and to fight for what they believe matters. If the citizens in a country fear persecution from freely saying what they believe, then their process freedoms are restricted, and they are much less likely to participate at all in government programs that promote human development. Thus, we hypothesize:

H3: The level of e-participation will impact human development in a country depending on the level of freedom of expression on ICTs in a country.

E-participation levels are dependent on the overall levels of freedom of expression on ICTs. Freedom of expression can be an enabler of dissent towards authorities when this form of expression is interpreted as a threat toward current administrative governments. This is known as ‘the dictator’s dilemma,’ where adoption of ICTs can promote both wealth and influence while also increasing the likelihood of dissent and civil mobilization (Kedzie & Aragon, 2002). This situation can sometimes lead to censorship and monitoring, which are fundamental issues related to the use of ICTs (Jordan, 1999). Countries may even choose to control the use of ICTs either through physical control or by blocking online networks (Castells, 2001). Some public debate has come up over legitimate control on child pornography and terrorism by taking away the anonymity of online users, but advocates for online freedom suggest that these issues are scapegoats utilized for further political control over the people (Slevin, 2000). When governments or organizations are suspected of using personal information to harm (physical or non-physical) users of an ICT, those users are less likely to use said ICT, thereby restricting their agency. Given this negative outcome for those in power, a government may choose to limit freedom of expression. This action has a suppression effect whereby the lowering of freedom of expression on ICTs suppresses the impact of e-participation on human development because citizens will be dissuaded from using government ICTs when they are likely to be censored or blocked. We look to our data to provide two example states that encourage e-participation, yet where the freedom of expression is low. We identify China and Russia as exhibiting this condition, two countries known for censorship of negative government-related rhetoric (Troianovski, 2021). The e-participation scores for China and Russia are 0.8 and 0.7, respectively (dataset mean = 0.46, SD = 0.27), which are greater than the average. The freedom of expression scores for China and Russia are 0.07 and 0.27, respectively (dataset mean = 0.53, SD = 0.211), both much lower than the average. According to a report, Russia’s recent shutdown of Western social media and Russian independent news outlets has cost the country an estimated $861 million in 2022 as compared to the prior year in which it was negligible (Daniel, 2022). This example shows that a lack of freedom of expression can lead to a decrease in the impact e-participation has on human development. Therefore, we hypothesize the following moderating effect:

H3a: The lower the freedom of expression on ICTs, the less the impact of e-participation on human development in a country.

While lowering the freedom of expression can dampen the effect of e-participation on human development, increasing freedom of expression can have the opposite effect. Freedom of expression is also enhanced through ICTs by giving the citizen’s voice a larger microphone to speak through. During the e-participation phase, when a lack of consensus occurs between the current administration and citizens or among citizens themselves, freedom of expression on ICTs can then enhance the citizens’ process freedoms by vocalizing their positions and gathering agreement among stakeholders. Freedom of expression on ICTs is then seen to amplify the effects of e-participation’s feedback and decision-making processes. Given this, we hypothesize:

H3b: The greater the freedom of expression on ICTs, the greater e-participation will impact human development in a country.

Methodology

Econometric methods were used to test the relationship between the independent and dependent variables in the proposed research model. We utilized country-specific fixed effects and controlled for potential endogeneity using instrumental variables. This method has been utilized in previous research involving ICT and development (Ganju et al., 2016).

Data collection

Our data was collected from several databases and merged based on country ID. Following previous research on ICT4D, we selected country-level data as our unit of analysis (Andoh-Baidoo et al., 2014; Baliamoune-Lutz, 2003; Chatterjee, 2020; Dewan & Kraemer, 2000; Ganju et al., 2016; Kiiski & Pohjola, 2002; Lee et al., 2017; Leoz et al., 2015; Mayer et al., 2020; Meso et al., 2009; Ngwenyama et al., 2006; Sağlam, 2018; Shirazi et al., 2009, 2010). This selection allows generalizability to previous research.

Measuring e-participation required us to find data on e-government, and more specifically, governments participating with the community through electronic platforms. We used data from the United Nations’ department of economic and social affairs (2020). They measure citizen engagement with governments through ICTs. There we pulled data on 191 countries for the 2016 E-Participation Index. The index is comprised of E-information: enabling participation by providing citizens with public information and access to information without or upon demand; E-consultation: engaging citizens in contributions to and deliberation on public policies and services; and E-decision-making: empowering citizens through co-design of policy options and co-production of service components and delivery modalities. This index adequately measures our intended participation construct as it measures the level of information shared between governments and citizens, along with the agency component that empowers citizens to be involved in policy discussions and decisions. Values were standardized between zero and one for analysis.

Freedom of expression is pulled from Freedom House’s annual Freedom on the Net report (Kelly et al., 2016). They give a score for Internet Freedom to 63 countries around the world based on obstacles to access, limits on content, and violations of user rights. An aggregated score of the three mentioned measures is used for their freedom on the net score. To obtain this data, a 21-question survey is administered to a trained researcher (or organization) within each country. Following the survey is a two-day rating review meeting, held with the rater to critique and adjust the draft scores following the project’s comprehensive research methodology. Careful attention is paid to current events, laws, and practices. A final score is then submitted after extensive review and fact-checking from Freedom House staff. Obstacles to access are measured by infrastructural and economic barriers, legal and ownership control over ISPs, and independence of regulatory bodies. Limits on content are measured by legal regulations on content, technical filtering and blocking of websites, self-censorship, vibrancy, and diversity of online news media, and the use of digital tools for civic mobilization. Violations of User Rights are measured by laws that give governments power to survey, infringe on privacy, and penalize users for online speech and activities through imprisonment, extra-legal harassment, or cyberattacks. We decided to drop the obstacles to the access category as the questions directly related to our infrastructure measures. Limits on content and violations of user rights were then aggregated for a max score of 75, instead of the previous 100 score range which included obstacles to access questions. Scores are then inverted, as following with the 2019 Freedom on the Net report, to align with other development indexes, normalized, and standardized between zero and one for further analysis. This means that higher scores now represent more freedom of expression.

ICT cost is collected from the International Telecommunication Union’s ICT Prices 2017 report (2017) which gives consumer costs for four ICT categories in the year 2016. The categories include mobile-cellular, mobile-broadband prepaid, mobile-broadband post-paid, and fixed-broadband. Each of the services’ prices corresponds to the least expensive plan offered by the dominant operator who fulfills the usage requirement. The mobile-cellular category includes only voice and SMS, comprising of about 30 calls (approximately 50 min each) and 100 SMS messages per month, with the inclusion of both on-net/off-net pricing and peak/off-peak weekend pricing variations. Mobile-broadband pre-paid and post-paid plans are separated with the former having a data allowance of 500 megabytes per month and the latter having one gigabyte per month. For our data selection, we chose to go with the cheaper of the two mobile-broadband options for each country. This was done to exemplify the more practical decision an individual would make within their country. Fixed-broadband values are based on the monthly price for an entry-level fixed-broadband plan with a monthly allowance of one gigabyte and a download speed of 256 kilobits per second. To compare prices between nations with varying income levels, we collected and used the percentage of gross national income per capita (% of GNI p.c.). This measure is the annual utility price in US dollars divided by GNI p.c. in US dollars for each country.

To measure ICT infrastructure, we used data from the International Telecommunication Union’s World Telecommunications/ICT database (2019). There we pulled data on 176 countries for the 2016 ICT Development Index, which gives data on ICT access, use, and skills of a population for each country. The specific measures used for ICT access were fixed-telephone subscriptions (per 100 inhabitants), mobile-cellular subscriptions (per 100 inhabitants), Internet bandwidth (bits/s) per Internet user, percentage of households with a computer, and percentage of households with Internet access. The measures used for ICT use include the percentage of individuals using the Internet, fixed-broadband subscriptions (per 100 inhabitants), and active mobile-broadband subscriptions (per 100 inhabitants). Finally, ICT skills were measured from mean years of schooling, secondary gross enrollment ratio, and tertiary gross enrollment ratio. The three sub-indexes are combined using a weighted scale of 40/40/20, with ICT skills as the latter. This lower weight on ICT skills is due to the use of proxy indicators. All scales were normalized before combining them. The final score for the ICT Development Index is a value between zero and ten that we standardized between zero and one with the latter having the highest level of development. Hamel (2010) notes that electrification is another important infrastructure measure, but we decided not to include access to electricity due to the other infrastructure measures acting as a proxy for electricity. Scores are then normalized for analysis testing.

Matching all datasets and excluding countries with missing data, we total 63 countries for use in the analysis. Of the 63 countries, 22 were classified as developed and 41 were classified as developing countries by the United Nations Development Programme (2018). A post-hoc analysis was run on the final model by including an independent variable indicating whether the country was developed or developing. Results show that this independent variable is not a significant predictor of HDI above and beyond the variables in the theoretical model, providing greater credence to the generalizability of the results. Furthermore, the correlation between this variable measuring development and inequality-adjusted human development index (IHDI) is around 80 percent, providing some credence that the results would be similar if IHDI had been used. It is important to note that if we were to only use the first three independent variables (i.e. exclude freedom of expression), the sample size becomes 186 countries. Our focus on understanding freedom of expression on ICTs and its interaction with e-participation warranted the smaller dataset. This is in accordance with other country-level research given the limited sample size. Given this, the ratio of a sample size to the number of items within our study is 13:1, which still exceeds recommendations from previous research of 5:1 (Gorsuch, 1983), 6:1 (Cattell, 1978), and 10:1 (Everitt, 1975). More recent research uses ratios similar to those previously recommended:

• Chatterjee (2020):

  • o Model 1: n = 41, 9 variables (5:1)

  • o Model 2: n = 41, 8 variables (5:1)

  • o Model 3: n = 41, 10 variables (4:1)

  • o Model 4: n = 41, 9 variables (5:1)

  • o Model 5: n = 41, 7 variables (6:1)

• Dewan et al. (2010): n = 26, 3 variables (9:1)

• Ganju et al. (2016): n = 94, 9 variables (10:1)

• Lee et al. (2017): n = 102, 3 variables (34:1)

For our dependent variable, we have selected the human development index (Bordé et al., 2009) as it complements the CA. This measure uses the geometric mean of three indices: Life Expectancy Index, Education Index, and Income Index. We collected data on HDI from the United Nations Development Program’s annual World Development Reports (Lopez-Calva et al., 2017). All 63 countries from our independent variables were accounted for.¹

Instrument variables

To address possible endogeneity, several country-level characteristics were identified as instruments. Instrumental variables are used in regression when you have endogenous variables (unobserved correlations) that influence other variables in the model. Instrumental variables are correlated with an explanatory variable but not correlated with the error term. The instrumental variables include ICT competition level, terrain ruggedness, e-skills, and the freedom of expression and information index. Further details on each instrumental variable are written below.

Prior research has implemented an instrumental variable approach to ICT costs with ICT competition intensity as the instrumental variable (Iacovone et al., 2016). The International Telecommunications Union has advocated for increased levels of competition within ICT markets as the best means to lower costs (2008). Monopolies on ICT infrastructure and services keep prices in several developing countries above the viable range of many users, thereby keeping demand low and limiting the impacts derived from such markets. Therefore, we argue that the level of competition in ICT markets will impact the infrastructure and service costs of ICT, but not correlated with the error term in our model. We utilize competition data from ITU’s World Telecommunications/ICT Database (2019). The data given is categorical (Monopoly, Duopoly, Partial competition, and Full competition). The variables chosen were mobile cellular, mobile broadband, and fixed wireless broadband. Inputs were converted to an ordinal scale (i.e. 1, 2, 3, 4) and standardized. By including three different competition levels and then standardizing to a zero to one scale, this helps to remove issues relating to the conversion of each of the three individual competition levels from categorical to ordinal.

Prior studies have argued that the extent of the sloping terrain in a country will impact the difficulty of providing ICT services (Ganju et al., 2016; Kolko, 2012). Also, hilly terrains are argued to reduce the broadcasting range in mobile devices (Arokiamary, 2009). We argue that slope ruggedness will impact a country’s ability to provide ICT infrastructure, but the ruggedness is not correlated with the error term in our model. Slope ruggedness data is obtained from (Nunn & Puga, 2012). We specifically used the population-weighted ruggedness variable to capture the importance of areas that are more densely populated. The variable is then standardized.

In studying e-participation, researchers have utilized e-skills as an instrumental variable (Tai et al., 2020). The difference between participation and e-participation being the use of ICTs for online participation versus offline participation. Citizens need knowledge of electronic skills to participate in society online. We argue that e-skills will impact e-participation, but e-skills alone are not correlated with the error term in our model. The level of e-skills is collected from the ITU’s World Telecommunication/ICT Database (2019). E-skills are made up of mean years of schooling, secondary gross enrollment ratio, and tertiary gross enrollment ratio. Values range from zero through ten. The variable is then standardized.

Freedom of Expression and Information is an index reported by the Fraser Institute’s Human Freedom Index report (Vásquez et al., 2018). The Freedom of Expression and Information index specifically targets press freedoms and access to their reported information. The press are considered citizens of a nation and are, therefore, given similar rights as other citizens within a nation (e.g. freedom of expression). This index correlates with a citizen’s freedom of expression in the sense that the press is considered to be arbiters of information and pushback on government mismanagements. We argue that press freedoms influence human development only through the freedom of expression given to all citizens of a nation, and therefore press freedoms are not directly correlated with human development. The index is then standardized.

Analysis

We utilize two-stage least squares (2SLS) to analyze our model using SAS. This decision is based on previous work in the ICT literature using simple least squares and regression at the country level (Andoh-Baidoo et al., 2014; Baliamoune-Lutz, 2003; Chatterjee, 2020; Dewan & Kraemer, 2000; Ganju et al., 2016; Kiiski & Pohjola, 2002; Lee et al., 2017; Leoz et al., 2015; Loh & Chib, 2019; Mayer et al., 2020; Ngwenyama et al., 2006; Sağlam, 2018; Shirazi et al., 2009, 2010). One alternative in the literature is PLS (Meso et al., 2009; Venkatesh & Sykes, 2013), though it is rarely used. We note that while Meso et al. (2009) analyzed data at the country-level, Venkatesh and Sykes (2013) analyzed the family-unit-level; thus, PLS is given little attention to analyzing country-level data.

Two-stage least squares (2SLS) modeling was utilized to test for endogeneity. Regression analysis has been utilized previously in related research and also offers generalizability with this research (Andoh-Baidoo et al., 2014; Chatterjee, 2020; Dewan & Kraemer, 2000; Ganju et al., 2016; Lee et al., 2017; Ngwenyama et al., 2006; Shirazi et al., 2010). Preliminary analysis was performed to first identify if endogeneity was truly present or if traditional ordinary least squares (Hölscher & Tomann, 2016) modeling would be sufficient. The Hausman test was used with a significant result indicating endogeneity is present (1978). Results in Table 2 shows that while Cost and Infrastructure displayed no significant endogeneity, e-participation showed substantial endogeneity and ICT freedom of expression showed marginally significant endogeneity. Given this, 2SLS modeling was utilized to control for this endogeneity.

Table 2. Endogeniety analysis.

Independent Variable	Instrumental Variable	Hausman Test
ICT Cost	Competition Level	0.21 (p = 0.90)
ICT Infrastructure	Terrain Ruggedness	0.16 (p = 0.92)
E-participation	E-skills	51.94 (p < 0.001)
ICT Freedom of Expression	Freedom of Expression and Information	5.02 (p = 0.08)

The 2SLS model was evaluated using the equations below. (1) $\mathrm{HDI}={\beta }_0-{\beta }_1\mathrm{cost}+{\beta }_2\mathrm{infr}+{\beta }_3\mathrm{epart}+{\beta }_4\mathrm{ICTfoe}+{\beta }_5\mathrm{epart}\ast \mathrm{ICTfoe}+{ϵ}_1$(1) (2) $\mathrm{epart}={\beta }_6+{\beta }_7\mathrm{eskills}+{ϵ}_2$(2) (3) $\mathrm{ICTfoe}={\beta }_8+{\beta }_9\mathrm{foei}+{ϵ}_3$(3)

Equation (1) specifies the primary model of country-level human development (Bordé et al., 2009) regressed on ICT Cost, ICT Infrastructure (infr), e-participation (epart), ICT Freedom of Expression (ICTfoe), and the interaction of e-participation and ICT Freedom of Expression (epart*ICTfoe). Equations (2)–(4) specify each of the endogenous variables regressed on their respective instrumental variable. Ruggedness, while found not significant for endogeneity using the Hausman test above, was still conservatively included as an instrumental variable of ICT Infrastructure given the extensive analysis in previous research purporting such (Ganju et al., 2016).

Results

Overall, the model is highly significant (F_(5,57) = 138.30, p < 0.001), explaining over 92 percent of the variance in HDI (R² = 0.924). The 2SLS results are shown in Table 3. One-tailed t-tests show that ICT Cost negatively impacts HDI, and ICT Infrastructure positively impacts HDI, supporting H1 and H2 respectively. While e-participation and ICT Freedom of Expression do not significantly impact HDI, the interaction of the two does positively impact HDI, supporting H3.

Table 3. 2SLS regression results.

Hypos	Independent Variable	Estimate	t-value	p-value	Supported
H1	ICT Cost	−0.09	−2.16	0.02	YES
H2	ICT Infrastructure	0.57	1.73	0.04	YES
	E-participation	−0.01	−0.02	0.49
	ICT FoE	−0.04	−0.30	0.38
H3	E-part * ICT FoE	0.10	2.33	0.01	YES

Notes: For the analysis, an alpha level threshold of 0.05 was used for the hypothesized single-sided tests. The estimates for E-participation and ICT FoE are not hypothesized given the interaction effect that supersedes these main effects in H3.

Looking more closely, at the interaction of e-participation and ICT FoE, a simple slope analysis was conducted to analyze simple main effects within the interaction. Figure 3 graphs this interaction. Simple main effects show that, when ICT FoE is low, the level of e-participation has no significant impact (β = 0.56, p = 0.21), supporting H3a. Conversely, when ICT FoE is high, the level of e-participation is magnified such that higher levels of e-participation have a more significant impact as compared to lower levels (β = 0.22, p = 0.04), supporting H3b.

Figure 3. Impact of interaction of E-participation and ICT FoE on HDI.

Discussion

This paper seeks to understand the impact of freedom of expression with e-participation on human development using the CA as a theoretical lens. The model includes the four conversion factors (i.e. ICT cost, ICT infrastructure, e-participation, and freedom of expression on ICTs) as recommended by Hamel’s (2010) report on the potentials of ICTs on human development. Our empirical research of 63 countries finds that freedom of expression on ICTs is significantly impacts human development only with high levels of e-participation in government by citizens. According to the CA, freedom of expression on ICTs and e-participation are social conversion factors that enable or restrict a country’s process freedoms. That is, as a new ICT resource is added to a population’s capability set, the people’s decision to utilize the resource will be impacted by both the levels of e-participation and freedom of expression on ICTs within the country. Citizens are not likely to use new ICTs if they believe the government is restricting their use and freedoms on said ICTs (Kedzie & Aragon, 2002). However, when governments and citizens cooperate through e-participation, there are reinforced reasons to believe freedoms of expression exist, such that government ICTs (e.g. e-government programs) and the-like are more likely to succeed.

As with previous research, ICT cost (Brown & Skelly, 2019; Hosman & Armey, 2017; Kiiski & Pohjola, 2002) and infrastructure (Borena & Negash, 2016; Forestier et al., 2002; Holeman & Barrett, 2017; Mayer et al., 2020; Rampersad & Troshani, 2020; Rashid & Diga, 2008) were included in our model. Utilizing the CA, ICT cost and infrastructure were labeled environmental conversion factors that enable or restrict a country’s opportunity freedoms. As a new resource is introduced to a country, the cost and infrastructure required to access that resource can act as a barrier to its citizens. Citizens will be unable to access ICT resources if their costs and infrastructure are too high or inadequate respectfully. Policymakers could take note of these four country-level conversion factors and refocus policies on expanding both opportunity and process freedoms as a means to development.

A final finding was the ability of the overall model to explain a high amount of variance in the dependent variable, human development. A similar study had 80-90.7% explanatory power in their model for human development (Zelenkov & Lashkevich, 2022), where our model showed a 1% improvement in variance explained. This provides a stable foundation for future research and provides authors with a solid starting point from which to undertake their research.

An interesting surprise from our results is no significant difference was found between developing and developed countries. Extant research has found differences, especially when technology plays a central role. For example, in developing countries, national information infrastructure influences the quality of governance as well as the level of socio-economic development (Meso et al., 2009). Regarding mobile devices, less developed countries derive increases in human development primarily from mobile devices (Ganju et al., 2016) while the diffusion of mobile devices enhances the economic freedom of citizens for both developing and developed countries (Lee et al., 2017) with this lack of diffusion of mobile technologies in developing countries being linked to a lack of banking education in the populace (Chatterjee, 2020). Additionally, for developing countries, the complementarity of the diffusion of PCs and the Internet is greater (Dewan et al., 2010). Income and government trade policies in developing countries have also been shown to influence ICT diffusion fostering economic development and enhancing political rights and civil liberties (Baliamoune-Lutz, 2003). Furthermore, the lack of institutional maturity in nations may see a negative impact from ICTs in upper middle and lower middle-income groups (Zelenkov & Lashkevich, 2020). Overall, while we find no significant difference between developing and developed countries, we acknowledge differences have been found in other research.

Given our findings of no difference between developed and developing countries, we performed multiple post-hoc tests to verify that the inclusion of the developing versus developed variable did not significantly impact HDI above and beyond the primary model. First, we included just the developing dummy variable as a direct effect on HDI in addition to the other variables in the model and found no significant direct effect of developing on HDI (β = 0.08, p = 0.11). Additionally, we also included an interaction of the developing dummy variables with each of the primary relationships in the model and found no significant interaction with cost (β = −2.66, p = 0.63), infrastructure (β = −0.14, p = 0.85), e-participation (β = 0.80, p = 0.51), freedom of expression (β = 0.78, p = 0.56), or the three-way interaction of e-participation, freedom of expression, and developed (β = −1.08, p = 0.54). Even so, future research should be performed to provide a more in-depth assessment of the developed nature of a country and the impact not only on HDI but more nuanced impacts concerning the individual theoretical relationships within the model.

Although in our research we suggest ICTs benefit development, there have also been several ICT initiatives over the past couple of decades with varying levels of success (Ahmed, 2007; Heeks, 2002; Keniston, 2002; Keniston & Kumar, 2004; Lin et al., 2015). These findings have raised questions about how ICTs may negatively lead to changes in ethical behavior (Radovan, 2013); social, economic, and environmental implications (Rothe, 2020); and sustainable development (Unwin & Unwin, 2017). ICTs are potentially good and bad for development, and future research would benefit with refocusing ICT4D towards a wholistic human development (Chacko, 2005). Above and beyond infrastructure and ICT cost, our study found value by incorporating the conversion factors freedom of expression and e-participation. Our study indicates their interaction leads to increases in human development. While previous research has studied conversion factors, none has studied the involvement of freedom of expression along with e-participation.

Limitations

It must be noted that there are limitations to our study. Notably, the number of countries analyzed. To have the most robust results, we took out countries with missing data. This limited our sample to 63 countries. However, our study has a good sample of countries from all levels of development. The 2016 HDI values are categorized into low, medium, high, and very high development categories. The value thresholds for each are below 0.55, 0.55-0.699, 0.7-0.799, and 0.8 and above. In our dataset, we count nine low-developed, 13 medium-developed, 19 high-developed, and 22 very-high-developed countries. This limits bias that can be taken from a skewed dataset.

Another limitation of our study is the use of external data sources to run our analysis. Using indexes that were collected and stored by third parties can have implications with not knowing exactly how the data was collected. We performed our due diligence and researched the methodologies used to collect all external data sources used for this paper. Appropriate indexes were selected based on the criteria used to collect each variable. Indexes were also standardized across all variables in the model to ensure no bias in one variable over another. Furthermore, an added PLS analysis was run to add credence to our use of these indexes in the primary analysis (see Appendix B). Endogeneity is also an issue with using secondary data sources. We used several instrumental variables to check for and correct if any endogeneity was observed. We included instrumental variables that have been used in previous research. Additional econometric specifications could be examined in future research.

A third limitation is the selected ICTs used for this study do not capture all possible ICTs with regard to both ICT cost and ICT infrastructure. We measured the cost of mobile-cellular, mobile-broadband prepaid, mobile-broadband post-paid, and fixed-broadband, as well as infrastructure related to mobile cellular, mobile broadband, and fixed broadband as our collective ICT measures. This does not account for all possible ICTs, such as radio, television, and more, though current research in the area has also utilized these same measures providing some justification for their use.

A fourth limitation is that despite our high variance explained, our model does not include personal conversion factors. Previous research investigating personal conversion factors suggest a strong relationship with Human Development. For example, in a study including the relationship between personal conversion factors and human development, Zelenkov and Lashkevich’s (2022) model results in an R-Squared of 90.7%. Yet, despite such a high R-Squared, their model does not account for the conversion factors we studied. An important consideration for R-Squared is its definition. R-Squared is a measure of the variance explained in the dependent variable by the included independent variables and the data used to model those variables. R-Squared is not an indication of the lack of variance explained for variables not included in a model. Thus, a high R-Squared does not preclude other variables in a model. For example, this could be due to our selected sample of countries and year. Zelenkov and Lashkevich relied on a data set resulting in115 observations. Our sample size of 63 countries is robust enough for our study, yet the variance could be quite different in our model with other countries and a different year. A final limitation can be attributed to the lack of management between the data-gathering entities. Due to the non-overlapping nature of the collected datasets, our study utilizes a cross-sectional analysis of country-level data. Going forward, it may be beneficial for the various groups and entities that collect this data to better coordinate their years of collection to enable longitudinal analyzes to uncover potential trends in the data.

Implications

Country policy makers shouldn’t focus on implementing ICTs as a means of development, but instead focus on how they will expand the opportunity and process freedoms of their citizens within. We demonstrate how opportunity and process freedoms can be mapped to conversion factors within the CA. We also provide empirical evidence for four conversion factors (i.e. e-participation, freedom of expression on ICTs, ICT cost, and ICT infrastructure) that have a significant impact on human development. Countries should examine their strategies concerning these conversion factors to more positively serve their constituencies.

As a practical implication, we introduce freedom of expression on ICTs as a new conversion factor that works in combination with e-participation to improve a country’s human development. Both must exist in tandem to receive improvements in human development. More specifically, both high freedom of expression and high e-participation are needed to jointly impact human development, where a high degree on only one will not impact human development.

Methodologically, while our use of 2SLS increases comparability by providing a mechanism for comparing this research with previous research, our post-hoc analysis using PLS provides implications for future research (see Appendix B). Specifically, we introduce a recently established method for assessing endogeneity within PLS models utilizing a Gaussian Copula approach (Hult et al., 2018). This novel method has yet to be utilized within IS research and provides a much-needed mechanism for addressing the potential for endogeneity within these more complex modeling techniques. Future research within our field should utilize and further examine this method, especially the yet-to-be-solved problem of interactions within this context. Given this, we still encourage future reviewers to exercise caution regarding the use of PLS. Specifically, the Gaussian Copula approach for assessing endogeneity is quite novel and has not been extensively tested. This is especially true when assessing endogeneity in a PLS model that includes interaction terms where researchers still recommend the use of 2SLS (Hult et al., 2018; Sande & Ghosh, 2018). More generally, PLS has been shown to have drawbacks in other areas. First, given the biased estimators in PLS (Savalei & Bentler, 2007) due to model errors not being taken into account (Marcoulides et al., 2009), the models are maximized for prediction of endogenous variables as opposed to explaining what is happening within a theory, as is needed in this research (Fornell & Bookstein, 1982; Mathes, 1993; McDonald, 1996; Noonan & Wold, 1982). Second, the PLS bias (Hair et al., 2014) which yields biased estimates due to the lack of explicit modeling of measurement error (Goodhue et al., 2012, 2013; Marcoulides et al., 2009), the results produce different outcomes for each research context thereby limiting the accuracy of the theoretical model and also the ability to compare results to previous research (Dijkstra, 1983; McDonald, 1996), which we do in this study.

Conclusion

In summary, our findings provide a strong variance-explained model for ICT conversion factors that affect human development for both opportunity and process freedoms. We mapped out the whole capability approach theory and explained how conversion factors are associated with different parts of the model. We have also highlighted an interesting interaction effect between e-participation and freedom of expression on ICTs in the ICT4D literature. Our model shows how policymakers can further increase the success rate of ICT initiatives within their countries by increasing opportunity and process freedoms. Moreover, future country-level research can focus on additional conversion factors as ways to improve human development.

Disclosure statement

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

Additional information

Notes on contributors

Gabriel Bahr

Gabriel Bahr is a Visiting Professor at Oklahoma State University, Spears School of Business. He received a PhD in Business Administration from Oklahoma State University. He enjoys researching and writing about topics on information and communication technology for development, international technology diffusion, country-level analysis, and gamification. He has published in Transactions on Human–Computer Interaction and in the proceedings of the Hawaii International Conference on System Sciences and the Institute for Operations Research and Management Sciences.

Bryan Hammer

Bryan Iwata Hammer is an assistant professor at the University of Montana. He received a PhD in Business Administration from the University of Arkansas. His research interests include interpersonal trust and privacy in virtual work groups, heuristics and their psychophysiological basis in online information disclosure, technology’s impact on individuals with intellectual and developmental disabilities, and text-based deception detection. His research has appeared in outlets such as MIS Quarterly, Journal of the Association for Information Systems, Information Systems Journal, Journal of Strategic Information Systems, Communications of the Association for Information Systems, the International Conference on Information Systems, International Journal of People-Oriented Programming, and Proceedings of the NeuroIS Gmunden Retreat.

Andy Luse

Andy Luse is a William S. Spears Chair in Business and Associate Professor of Management Science and Information Systems in the Spears School of Business at Oklahoma State University. He received a B.A. degree in Computer Science from Simpson College, M.S. degrees in Information Assurance, Computer Engineering, Business Administration, and Psychology, and Ph.D. degrees in Human Computer Interaction, Computer Engineering, and Information Systems from Iowa State University. Andy has been published in the Journal of Management Information Systems, Journal of the Association for Information Systems, Decision Support Systems, Journal of Business Research, Communications of the Association for Information Systems, Computers in Human Behavior, and many other outlets.

Notes

1 Some previous research has also looked at the use of the inequality adjusted human development index (Andersson et al., 2012; Grimm et al., 2010). This measure utilizes the geometric mean of inequality-adjusted HDI by adjusting each dimension per the level of inequality (Atkinson, 1970). While useful, a google scholar search for HDI vs. IHDI found over 28,000 citations for the former and only 1200 for the latter since 2018. Furthermore, the number of citations for the articles utilizing the HDI are substantially higher than that of those articles using the IHDI, which only have citations in the single digits. For this reason, we have chosen to utilize the HDI in our research.

Appendices

Appendix A: Countries included in the analysis.

Table

Developing	Developed
Angola	Argentina
Armenia	Australia
Azerbaikijan	Bahrain
Bangladesh	Belarus
Brazil	Canada
Cambodia	Estonia
China	France
Colombia	Germany
Ecuador	Hungary
Egypt	Iceland
Ethiopia	Italy
Gambia	Japan
Georgia	Kazakhstan
India	Korea (Republic of)
Indonesia	Malaysia
Iran (Islamic Republic of)	Russian Federation
Jordan	Saudi Arabia
Kenya	Singapore
Kyrgyzstan	Turkey
Lebanon	United Arab Emirates
Libya	United Kingdom
Malawi	United States
Mexico
Morocco
Myanmar
Nigeria
Pakistan
Philippines
Rwanda
South Africa
Sri Lanka
Sudan
Syrian Arab Republic
Thailand
Tunisia
Uganda
Ukraine
Venezuela (Bolivarian Republic of)
Viet Nam
Zambia
Zimbabwe

Appendix B: PLS analysis

To provide greater support for our findings, the data were also analyzed using Partial Least Squares path modeling. The same model was used whereby ICT Cost, ICT Infrastructure (infra), and e-participation (epart) directly impact Human Development (Bordé et al., 2009), with Freedom of Expression on ICT’s (FoE) moderating the impact of epart.

Before analysis, endogeneity was assessed for the independent variables in the model. Endogeneity within PLS models has only recently been addressed in the literature (Hult et al., 2018). This novel Gaussian Copula approach involves a multi-part assessment of the independent variables in the model both for their ability to detect endogeneity if it exists, and subsequent testing if appropriate. First, the data was modeled according to the base model outlined above. Using SmartPLS (Hair Jr et al., 2016), the latent variable scores from the PLS algorithm are extracted from the output. Second, the R code provided for the method (see Hult et al., 2018) is used to run the Kolmogorov–Smirnov test with Lilliefors correction (Sarstedt & Mooi, 2014) to assess the normality of each independent variable where non-normality is required to further consider the variable for endogeneity in the model. Results found that cost (p = 0.002) and FoE (p = 0.037) were significantly different from a normal distribution and, therefore, these variables could be further assessed for endogeneity. Third, Copulas were created for each of these variables, and bootstrapped standard errors were used to calculate corrected p-values for the original model with each of the Copulas included for each of the variables individually. Results (see Table 4) did not find a significant corrected p-value for any of the tested variables. Given this, the results of this analysis would suggest using the original model as no endogeneity is present.

Table 4. M1 includes the Copula for cost (c_cost) and M2 the Copula for FoE (c_FoE).

	Original		M1		M2
	beta	p-value	beta	p-value	beta	p-value
Epart	−0.08	0.49	−0.06	0.75	−0.08	0.64
FoE	−0.18	0.18	−0.14	0.51	−0.19	0.51
Infra	0.78	< 0.001	0.73	< 0.001	0.78	< 0.001
Cost	−0.16	< 0.001	−0.10	0.30	−0.16	0.07
FoE*epart	0.30	0.06	0.27	0.38	0.30	0.28
c_cost			−0.09	0.34
c_FoE					0.01	0.93

Given the differing results regarding endogeneity between the traditional 2SLS instrumental variable approach and the Gaussian Copula approach, an important question is which is better suited for this research? First, the Gaussian Copula approach only enabled the assessment of endogeneity for two of the variables given its requirement that variables have a non-normal distribution to be tested. This left out both epart and FoE, both of which showed significant endogeneity using the Hausman test in the 2SLS model, which does not have these distributional restrictions. Another consideration is that, while the 2SLS instrumental variable approach is highly accepted and is an established assessment within econometrics, marketing, and other disciplines, the Gaussian Copula approach is relatively novel and somewhat rarely used, affecting the comparability of the study to previous studies using these methods. Finally, this study utilizes an interaction term in the base model. Currently, research does not yet have an accepted metric for assessing endogeneity in a PLS model that includes interaction terms (Hult et al., 2018), and researchers recommend instead using a 2SLS approach utilizing instrumental variables (Sande & Ghosh, 2018) until greater research into this area with PLS has been carried out. Overall, the 2SLS approach is preferable as a tried-and-true method for measuring endogeneity as well as the inability of PLS to work with endogeneity when interaction terms are involved. This 2SLS approach utilizing instrumental variables is what is utilized in this paper.

Even with the drawbacks of using PLS, when comparing the results of the two approaches (2SLS and PLS), the overall picture is highly similar. In both studies, there is a significant impact of cost, infrastructure, and the interaction of FoE and epart on HDI. Furthermore, the overall R² of both models is almost identical (0.92 in the 2SLS model and 0.91 in the PLS model). From this perspective, both methods point towards the same independent variables significantly affecting the dependent variable, with the explained variance around 91–92 percent.