Internet Use and Economic Development: Evidence and Policy Implications Assignment Sample
In this dissertation, four different economic development outcomes from 202 nations are examined by our Economics assignment help during an eleven-year period, from 1996 to 2007, to determine the effects of Internet use on economic development. Panel data and panel data estimate methods are used. The main theory being examined is that rising Internet usage has a beneficial impact on the following development outcomes: GDP, exports, the growth of local equity markets, and overall welfare as assessed by the UN Human Development Index (HDI). However, depending on the country's income level, Internet use has different effects on economies. Compared to low- or high-income countries, the effect is probably more pronounced in middle-income countries. This may be partially attributed to the low adoption of the Internet in low-income nations and the declining marginal product of Internet use in developed nations.
Therefore, in order to construct the four samples of interest—the whole sample of all nations and samples for each income category—countries are grouped according to the World Bank's definition of income class. To investigate the effects of extra Internet use at various developmental stages, the effectiveness of Internet use is assessed on each sample, and the findings are compared. A production function framework with Internet usage as an additional input will be used for the econometric analysis.
As important a development in human history as the invention of the printing press may have been the rise and expansion of the Internet, a global network of computing resources. The way we communicate, conduct business, learn, and govern ourselves has been permanently altered by widespread and consistent access to the world's collective computer resources and Internet-enabled gadgets. The effects of the Internet on the economy are astounding. By lowering transaction and search costs and improving access to information about goods and services offered in domestic and global marketplaces, internet use improves market efficiency. Innovation is sparked by information access, which also makes it simpler to embrace new technology and boosts output.
There are numerous stories in the news about how technological advancements are altering how business is done in the developing countries. The evidence of the effects of Information and Communication Technologies (ICT) on economic growth abounds, and the possibilities seem limitless. From farmers in Kenya using cell phones for Internet access connecting with insurance agents to protect their crops (DAWN Media Group 2009), to Chinese Internet centers for remote villages (Fong 2009), to Indian fishermen using cell phones to locate markets with the greatest demand for their products (Jensen 2007), the examples of ICT's effects on economic growth are numerous.
As people and organizations use the Internet more frequently in developing economies' private and public sectors, this opens up potential for the development of new goods and services. In the end, more information availability leads to market transformation, where new markets may appear and current markets may change, enlarging their size, sphere of influence, and interacting with markets in other nations or regions. Internet access has the immense potential to revolutionize governance through information use, a prospect that plainly disturbs autocratic leaders because they promptly halt Internet access at times of popular disturbance, as was the case during the unrest in the Arab world in early 2011.
Figure 1 shows how quickly the Internet has expanded and become available to users all around the world in the past 15 years. Computers with network capabilities are now commonplace in offices, homes, and educational institutions all throughout the developed world. Almost every element of industrialised economies has been impacted by computers and the Internet. Similar to electrical electricity, municipal water, and public roads, Internet connectivity is increasingly seen by the populace in industrialised nations as a basic service. Internet access is widely available, dependable, and, for the most part, affordable. Both locals and visitors may anticipate being able to access the Internet on a regular basis, just like they do with power outlets, contemporary transportation options, and sanitary services.
If it is available at all, Internet connection in a substantial portion of the developing world is only found in densely populated areas. Only as improvements are made to the necessary electrical power and communication infrastructures is it becoming more generally available to connect to the global communication networks that permit Internet use in less developed countries (LDCs). It is difficult to connect to the Internet without restrictions without a wide range of electrical and communication resources. However, without access to this technology, the potential of the Internet to boost productivity and human welfare cannot be realised. The population's access to and affordability of computers and other devices with Internet capabilities is necessary for the Internet to have a materially positive impact on the economy.
In LDCs, political and social structures are frequently unstable, and incomes are insufficient to support the investment required to establish even the most fundamental reliable public utilities. Poor yet stable nations cannot maintain the infrastructure necessary to provide broad Internet access. In a developing nation, solid political and social structures, fundamental public utilities like the distribution of electricity, and educational resources are more basic requirements, while Internet use is a higher order need, according to psychologist Abraham Maslow (1943). Internet connectivity will not be a top priority, if it is even possible, in nations where the populace must prioritise basic survival owing to extreme poverty, a lack of or failure of public health institutions, an ongoing violent conflict, or generally unstable political systems.
The "One Laptop Per Child" initiative's underwhelming results in LDCs offer some anecdotal evidence of the inadequacy of technical solutions to produce replicable and significant benefits in nations lacking the requisite infrastructure and fundamental political, social, and economic institutions (Shaikh 2009). Of course, those who lack access to the Internet or who are illiterate will find little use in the information supplied there. This dissertation offers empirical proof that expanding access to and utilisation of the information and knowledge made available by Internet connectivity has a significant positive impact on development outcomes, and it shows how the size of this impact varies for economies at various stages of development.
In the literature on economic development, there is little empirical research on how Internet use affects development outcomes. The few studies that have been conducted on the topic have narrowly concentrated on either income growth or the productivity of a specific industry. However, economic development encompasses much more than just income growth rate. The goal of this dissertation is to empirically examine how using the Internet affects a variety of developmental outcomes. The literature is summarised in Section 2 along with a description of the research hole that this dissertation fills. There has not yet been a thorough empirical investigation of the potential productivity-enhancing effects of the global improvements in information dissemination that the Internet offers, particularly in LDCs.
In this research, I offer a straightforward theoretical framework for examining how Internet use affects economic growth. I then apply the model to a panel of data on numerous nations at various stages of development. ICT studies in the past have focused solely on growth effects, but development encompasses more than just GDP growth. By extending the scope of development outcomes evaluated, this study offers a more thorough understanding of how Internet use affects development.
Additionally, earlier research frequently makes use of dynamics-impervious cross-country regression approaches. I employ panel data and econometric estimators that can take into account dynamics and unexplained heterogeneity in this analysis. As a robustness test, I additionally estimate the models using mixture modelling methods. Studies already conducted on the effects of ICT on all countries or just one particular country presume linear responses. This is unlikely to be the case because, as was mentioned before, the ability to use and benefit from Internet use may vary according to the level of development. As a result, I divide the world into different income classes and take into account the effects on each class.
202 countries' worth of data from the World Bank, the International Telecommunication Union, and the United Nations are combined to create a rich data set for the econometric analysis using dynamic panel estimation techniques. These data cover the years 1996 to 2007 and come from a variety of sources. The analysis shows that in nations with sufficiently high income levels, Internet use has a considerable positive and possibly incidental impact on a number of development indicators. GDP, export revenues, and equity market capitalization per capita are the development outcomes taken into account in this research. A full examination of economic development must take this measure of economic production into account. Per capita GDP is maybe the most popular development outcome metric in the development literature.
Some significant studies on ICT and exports may be found in the literature on development and growth. To other development outcomes, there isn't much extension, though. By examining the connection between the Internet and exports using a model that accounts for dynamics, this dissertation expands on existing research. The comparison and assessment of the findings are presented in Chapter 5. A growing body of research on domestic financial markets and economic development has raised development financing as its main focus. This dissertation adds to that body of work by examining how Internet use affects domestic financial markets. It specifically looks into the connection between Internet usage and capitalization rates of domestic financial markets, which is undoubtedly a sign of how far along these markets are.
The United Nations Human Development Index (HDI) is used as a supplementary indicator of economic development in order to examine how Internet use affects a broader measure of societal well-being. The HDI is used in this study, as well as many others, as a stand-in for the population's overall wellbeing within a nation. Front-line healthcare professionals will have increased access to information, including innovative techniques and best practises, for handling urgent health crises as Internet connectivity becomes more widely available in developing nations. Internet usage can also improve academic results as indicated by literacy rates. The HDI index incorporates these elements, making it a unique proxy for overall wellbeing.
The empirical study tests the hypothesis by contrasting the outcomes of fitting various econometric models to each sample. In order to account for or adjust for endogeneity resulting from the potential simultaneity issue of jointly determined dependent and explanatory variables, dynamic panel data and finite mixture model estimate techniques are used. A case for the causal effects of Internet use on economic outcomes can be made by accounting for the potential endogeneity of the measure of Internet use.
The use of the Internet may have a favourable impact on economic outcomes via a variety of processes. By reducing the distances between people, businesses, and nations and increasing information flow, access to the Internet lowers the cost of transactions and transportation. The sharing of knowledge and ideas is essential for the advancement of technology. According to Granovetter (2005), social networks have a significant impact on economic outcomes, particularly when it comes to supporting innovation. Websites that offer social networking features, like Facebook, are one of the fastest-growing types of Internet communication. Access to the Internet has expanded information availability, which could improve institutions' efficiency and transparency, resulting in greater governance. As Acemoglu et al. (2001) have demonstrated, better institutions can have favourable economic results. Kalathil (2003) makes a compelling case for how Internet use might promote effective governance.
By lowering search and transaction costs, information exchange increases market efficiency, which in turn can enhance demand, domestic production, and trade opportunities since it connects buyers and sellers and therefore expands marketplaces. New markets and industries can be created using the internet and other information technology. Access to educational resources via the internet can result in the production of additional human capital and an increase in labour productivity. Similar to this, access to health care is being made possible for people who might not have had it otherwise thanks to online health information. Economic development outcomes can be stimulated and amplified by these kinds of effects in economies at all stages of development.
Studies emphasising the ability of growing Internet use to modify society are starting to appear more frequently in the literature. In order to access new markets, learn new skills through shared experiences, and create more robust supply chains, small rural farmers in Central America and India can use ICT, according to a detailed analysis provided by Parikh et al. (2007). Advanced ICT networks encourage Foreign Direct Investment (FDI), which enhances economic opportunities in nations of all development levels, as Reynolds et al. (2004) show.
Although the effects will probably be favourable at all income levels, they will probably vary based on the countries' economic classes. Before many of the positive effects of Internet accessibility can have a large impact on the economy, developing countries must have basic public infrastructure, such as energy distribution, sanitation, primary care, and education systems. It's possible that low-income nations lack the political and social institutions needed for their populace to truly benefit from Internet access. High- and middle-income nations have attained levels of economic stability (they are higher on the Maslow's hierarchy of needs applied to economies pyramid) and are able to draw investment in the communications infrastructure, which can help educational systems operate more effectively and result in a population with a higher level of literacy.
In the context of this investigation, investing in Internet use can be seen as a complement to both investments in physical capital and human capital, as it makes workers more efficient by giving them access to knowledge that improves their skills and more rapid ways to create, disseminate, and assimilate information. The evidence in this dissertation shows that Internet use has a favourable impact on economic growth and other development outcomes, even if it does not give a growth story per se.
This dissertation's main goal is to present empirical proof of the causal link between rising Internet usage and rising economic activity across a spectrum of development outcomes. I use the per capita GDP, exports, the size of domestic equity markets, and overall wellbeing as assessed by the HDI as development indicators in this dissertation. Domestic developing economies will change when more people and enterprises in those areas start accessing the Internet. The way people learn about goods and services through websites and email will change how things are produced and consumed. A greater flow of information about available jobs, production methods, and new goods will make resource allocation and human capital deployment more effective. Although there are many development outcomes, I concentrate on these four in part because of their close connection to human welfare and in part because data are readily available.
Domestic markets that are expanding and mature draw consumers from other nations, enhancing trade prospects. Domestic equities markets become increasingly accessible to international capital markets, expanding the breadth and depth of the market's offerings as more economic agents from the public and private sectors establish online presences. The efficient operation of financial markets depends on information flow, which is crucial for developing equity markets. The usage of the internet has perhaps improved information production and distribution more than the use of the phone and fax machine combined. The effects of financial market structure and maturity on international Internet diffusion are examined by Yartey (2008), who discovers a strong correlation. This line of inquiry will be expanded upon in this dissertation in order to examine the effects of rising Internet usage on the size of the domestic equities market.
This dissertation shows that not all countries experience the same marginal advantages from Internet use. It depends on the specific outcome being researched whether increasing Internet access and Internet users in the least developed nations has a substantial impact on all aspects of economic development. All of the study's examined metrics of economic activity show a significant increase in middle-income nations as the number of Internet users increases.
The findings indicate that although middle-income nations have strong positive and significant effects, low-income countries exhibit lower, less significant effects of increased Internet use on exports and market capitalization. Middle-income nations see a rise in per capita export earnings of 2.3% as a result of a 10% increase in Internet users, but low-income countries see no change (fail to reject the null hypothesis of no change). The results are significant in terms of per capita GDP. Grows in per capita GDP of 3.2% in middle-income nations and even higher 3.5% in low-income countries are statistically significant when Internet usage increases by 10%.
Since Internet use has different effects on economic development outcomes depending on the country's income level, the policy recommendations must likewise differ. In low-income nations, increased Internet use has a strong positive impact on both the HDI average and per capita GDP. In order to provide Internet access at a reasonable cost, policymakers should concentrate on cellular phone networks. Additionally, foreign direct investment can be used to build the infrastructure required for increased Internet deployment while promoting foreign aid for health and education initiatives.
The effects of increased Internet usage are especially noticeable in nations with middle-income status. All four indicators of economic development in these nations are positively and significantly impacted by rising Internet usage. This argues that middle-income country policymakers should concentrate on giving the service industry institutional and legal support so that mobile banking, insurance, and other Internet-enabled technical solutions can be given to the populace via the mobile Internet. Widespread Internet connectivity is necessary for developing exports and domestic financial markets in nations with economies that are only starting to function and expand.
The rest of this dissertation is divided into the following sections: In order to comprehend the significance of this dissertation, Chapter 2 surveys the literature of pertinent earlier works on economic development, ICT, and economic growth. While Chapter 4 analyses the data sources and details the variables used for the empirical analysis, Chapter 3 presents the broad theoretical model and empirical technique for this study. The econometric findings of each investigation are presented and discussed in Chapter 5, and a summary, policy repercussions, and prospective future research possibilities are explored in Chapter 6.
1. Literature Review
Although there is a wealth of literature on economic growth, academic academics are just now starting to pay more attention to the connection between economic development and ICT. Existing research only offers a narrow and fragmentary perspective on the contribution that Internet use makes to economic growth. The literature has largely ignored other facets of development in favour of concentrating on the connection between information use and income increase. By examining many processes via which Internet use can affect the well-being of nations at various stages of economic growth, this dissertation offers a far broader perspective on the relationship between Internet use and development. The current research that serves as the backdrop for this dissertation is discussed in this section of the dissertation.
A foundation for understanding the contributions made by this study can be found in the numerous significant publications on the relationship between telecommunications generally and economic growth. While there are few published comprehensive empirical analyses of the role that Internet use plays on economic development, this study adds to a body of knowledge on this topic. I quickly highlight a few research that look at the connections between various facets of development. The review is organised into sections that examine how ICT impacts growth, factor productivity, welfare, trade, and equity markets, among other economic consequences.
1.1 ICT, Development, and Economic Growth
The methodology used in this study is an extension of the ground breaking research of Papakek (1973), which used cross-country panel methodologies to isolate the effects of FDI, aid, and exports on economic growth. It is also comparable to Barro's landmark study from 1991, which found a positive relationship between human capital and political stability, which in turn promotes economic growth. Before the advent of the Internet, there are outstanding evaluations of the state of empirical cross-country growth research by Levine and Zervos (1993) and later Sachs and Warner (1997). ICT, the Internet, or information mechanisms are not examined in these research' analysis of growth determinants. According to the enlarged Solow model-based neo-classical growth theory, long-term growth is reliant on technical advancement and labour force growth (Grossman and Helpman 1994). By examining how Internet use affects various outcomes, including economic growth, as a gauge of technical advancement, this dissertation adds to the body of scholarship.
Studies have just started to look into how important institutions are to economic growth. Although Rodrik et al. (2004) and Acemoglu et al. (2001) concur that institutional quality is a crucial component in economic development; neither study takes technological aspects into account that can improve productivity, trade, and the development of human capital. It is crucial to take institutional quality into account since it explains why technical advancement varies between nations. This dissertation isolates the effects of Internet use on development outcomes while controlling for institutional quality and other variables. It's interesting to note that Audretsch and Keilbach (2007) introduce the concept of entrepreneurship capital, or "the capacity for economic agents to generate new firms," as a crucial component of economic growth in Germany and discover that ICT infrastructure is a sizable component of entrepreneurship capital.
While this dissertation examines the effects of new technology (particularly, Internet use) on development outcomes by using institutions as a control variable, the research discussed above all employ institutions to explain differences in technical advancement.
The challenges faced by the world's poor people are depicted in great depth by Banerjee and Duflo (2007). Technology-based solutions are frequently advocated as a tool to help low-income people find ways to raise their income and raise their standard of living. In low-income nations, Gulati (2008) argues that investing in ICT infrastructure for educational purposes may primarily help the already wealthy, and that there may be a greater immediate need for fundamental educational services. By concentrating on potential distributional issues, Gulati's opinion ignores the total welfare benefit that ICT access and Internet use bring to the population of developing countries. It can be challenging to envision how relief programmes that give people access to ICT and the Internet will be of immediate assistance to a population that is already battling to survive. Although the poorest people of a population might not immediately gain from ICT use, overall wellbeing is increased. Redistributive measures can be put in place once welfare is increased to aid the poor.
Rajan and Subramanian (2008) use dynamic panel data approaches to further explore the issue of how foreign assistance transfers affect economic growth and come to the conclusion that aid transfers have little impact on growth across a range of nations. They employ a range of measurement techniques to control for endogeneity. To counteract this, Mishra and Newhouse (2009) found that, as measured by a decline in newborn death rates, direct economic aid does have a significant impact on health outcomes. In nations with underdeveloped health care systems, expanding access to information through ICT and the Internet will probably have favourable effects on health outcomes that may directly and indirectly affect economic development. In neither of these cases, the use of technology is the type of Internet infrastructure in particular and ICT infrastructure in general. Although economic aid is not directly examined in this dissertation, it serves as a crucial control in the empirical analysis and is viewed as endogenous.
In the mid to late 1990s, the increased accessibility to computers and Information Technologies (IT) in general started to have a revolutionary impact on industrialised economies. In their non-empirical study, DePrince and Ford (1999) concentrate on the economic growth that has resulted from the rapidly developing Internet economy in the United States and draw the following conclusion: "The emergence of the Internet economy is a Schumpeterian event that may rival the introduction of printing, steam power, the telephone, and the assembly line as a growth enhancing innovation." Madon (2000) acknowledges that the Internet will have significant societal effects in underdeveloped nations and offers a conceptual framework for comprehending how the Internet and economic growth interact. In poor countries, he argues that there are "six important application areas of the Internet, namely economic productivity, health, education, poverty alleviation and empowerment, democracy, and sustainable development." I am not aware of any published research that empirically investigate how using the Internet affects any of these factors. I experimentally research a number of Madon's hypothesised Internet-related effects on economic development with an emphasis on low- and middle-income nations as separate classes. Therefore, this dissertation offers the most thorough empirical analysis of the connection between Internet use and development that has ever been conducted in a single study.
The impact of telecommunications infrastructure on economic growth in Organization for Economic Co-operation and Development (OECD) nations over a 20-year period is examined by Röller and Waverman (2001). They discover a significant causal effect of communications infrastructure on aggregate output after controlling for country fixed-effects and simultaneity. Despite the fact that this study concentrated on OECD nations, it's plausible that these effects exist in nations with various levels of development. In this study, I use government effectiveness as a proxy for institutional quality, which can affect the deployment of infrastructure.
Three distinct methods that ICT can increase economic output in developed economies are outlined by Jalava and Pohjola (2002). First, ICT products immediately boost production. Second, ICT capital is employed in the manufacture of different products. Third, the industries that produce ICTs themselves contribute goods and services. They show large productivity-enhancing effects of ICT in the United States in the 1990s using a macroeconomic growth accounting model, but weaker evidence in the other G7 nations. The effects are attributable to the US service sector's intense international competition. They don't try to look at the relationship in less developed nations.
Attempts to comprehend the slow economic growth rates in Africa and other low growth regions have been the focus of a lot of economic development study in the literature. The development literature hasn't typically focused on how ICT and the Internet might aid Africa's economic development. Even though the research of the effects of Internet use on developing economies is not primarily focused on Africa in this dissertation, the findings offer new information that can help in understanding the developing economies in Africa.
Ernst and Lundvall (1997) study whether new institutions may be required for developing countries to utilise cutting-edge IT solutions that may improve learning. Their stylized models are based on data from the United States and Japan. "For the majority of developing countries, the fundamental priority is to construct the essential institutions that provide incentives and externalities necessary for domestic learning," they propose understanding the impact of functioning institutions on economic growth in LDCs is made possible by significant studies like Acemoglu et al. (2001), Rodrik et al. (2004), and Banerjee and Duflo (2004). Using dynamic panel estimating approaches, Gyimah- Brempong (2000) discovers negative effects of corruption on growth in African nations during the 1990s. According to Kalathil (2003), having access to the Internet can aid in the formalisation of informal institutions and foster freedom in totalitarian regimes, which are frequently present in LDCs. It will be interesting to watch events in the Middle East and North Africa in 2011 to determine if the upheaval, sparked by communication over the Internet, results in more transparently operating institutions or if autocratic regimes just block Internet access to maintain power.
According to Decker and Lim (2008), political institutions play a key role in determining growth. These studies don't look into how communication affects institutions or how ICT influences income growth. The development and upkeep of efficient institutions depend heavily on the open flow of information. For instance, effective and responsible governance depends on both the open flow of information and transparency. Utilizing ICT and the Internet allows access to information that makes communication easier, and this information flow can only help institutions.
Czernich et al. (2009) examine the impact of broadband Internet infrastructure on economic growth in OECD nations in a recent study. They show the beneficial causal effects of more broadband Internet infrastructure in industrialised countries using a technology diffusion model and instrumental variable methodologies. Although there is no justification for limiting the application of this paradigm to just industrialised countries, the analysis is not extended to developing nations. The correlation between increased Internet bandwidth and economic growth in wealthy nations is highlighted by this result. The number of Internet users in poor nations might be viewed of as the sole indicator of ICT capacity. The direct impacts of increased Internet use on outcomes for development in underdeveloped nations are examined in this dissertation.
1.2 ICT, Internet, and Productivity
Numerous studies look into how ICT, IT investments, and information networks might increase productivity. Dedrick et al. (2003) give an extensive assessment of the literature encompassing fifty past firm and country level research on IT investment and draw the conclusion that IT investment significantly boosts economic growth and labour productivity. Other research have shown inconsistent results, notwithstanding Engel Brecht and Xayavong's (2006) conclusion that ICT does not definitely increase worker productivity in New Zealand. Using a stochastic-frontier production function estimation technique, Thompson and Garbacz (2007) discover that information networks, such as mobile and fixed-line telephones and the Internet, benefit both the "whole world" and some of the "poorest nations" by enhancing institutional functionality and business efficiency. In order to close the productivity and production gaps that exist between industrialised and developing countries, Stein Mueller (2001) draws the conclusion that investments in ICT may enable "leapfrogging" or "bypassing some of the processes of accumulation of human talents and fixed investment." The quick uptake and spread of cell phone networks to areas in developing nations with no history of fixed-line telephone infrastructure is one illustration of this leapfrogging. Ngwainmbi (2000) draws attention to the adjustments Africa must do in order to keep up with the technological advancements happening at the time of his study. Africa has to have access to energy resources and telecommunications infrastructure in order to compete in the global information industry. These studies, as well as numerous others, present evidence for the economic impacts of the Internet through casual observation, logic, and descriptive statistics. To better understand how the Internet affects economic development, this dissertation offers an empirical investigation utilising econometric techniques.
Goel and Hsieh (2002) argue that the Internet has the ability to increase competition and make markets more competitive by helping to eliminate information asymmetries, but they do not offer any empirical analysis to support their argument. Parikh et al. (2007) discuss how small-scale farmers might use a range of technical solutions to better integrate themselves into "global value chains," but they do so without offering any concrete evidence. Goyal (2010) demonstrates that when farmers are given access to information about market prices via Internet kiosks, rural soybean markets in India become more effective.
By enabling producers and consumers to interact and transact in novel ways, the availability of the internet can aid in the creation and expansion of markets. These studies offer convincing justifications for how and why access to ICT and the Internet might improve economic outcomes. The lack of micro level data in these investigations may be the cause of the absence of empirical proof. In order to remedy this vacuum in the literature, this dissertation uses aggregate country-level data to present empirical evidence of the effects of Internet use on economic development.
1.3 ICT, Internet, and Welfare
The body of knowledge about how ICT and the Internet affect societal well-being is growing. Both Crandall and Jackson (2001) and Prahalad and Hammond (2002) describe the large possibilities for commercial enterprises to profit from offering goods and services to rising markets in poor countries around the world, particularly by utilising ICT infrastructure. While it's possible that the poor areas of Chicago are not entirely comparable to developing nations, Masiet al(2003) findings that access to health information made the poor citizens of Chicago more powerful is an intriguing conclusion that might be equally applicable to LDCs. This dissertation makes an effort to investigate this link in the context of the developing world by looking at the effects of Internet use on several economic measurements and the UN HDI.
According to Jensen's fascinating and significant study from 2007, when Indian fishermen are given cell phones, price-dispersion in neighbourhood fish marketplaces is drastically decreased, improving both producer and consumer welfare. As a result, markets grow more effective as knowledge becomes more widely available. Using the internet is a useful way to spread information more widely. This is one of the ways that Internet use affects the state of the economy in developing nations.
ICT has a beneficial effect on advancing democracy and freedom of expression, according to Shirazi (2008), who examined eleven Middle Eastern nations. The widespread upheaval that occurred in North Africa and the Middle East in the early part of 2011 may be explained by the strong transformative power of populations connected through the Internet. Future research would be interesting in delving deeper into how institutions and corruption are affected by access to ICT.
Some researchers contend that ICT will have minimal effect on reducing poverty and promoting economic growth in LDCs using only reason and descriptive statistics. According to Kenny (2002), programmers shouldn't be used to create ubiquitous access to the Internet until it is made easier, less expensive, and literacy rates rise. His assertion that "LDCs appear ill-prepared to benefit from the potential that the Internet does present—they lack the physical and human resources, as well as the institutions required to leverage the e-economy" is reiterated (Kenny 2003). Kenny makes the claim that there is data showing that Internet use has little to no impact on the rate of income development in the lowest-income countries, but he does not conduct any actual research to support this claim. He contends that assisting with health, telephony, and literacy may benefit LDCs more than providing Internet access. In order to more fully explore these arguments, this dissertation extends them by examining the effects of the Internet using a thorough empirical study that isolates countries by income level.
Like Kenny indicates, Thompson and Garbacz (2007) show that expanding telecommunications networks improves organisational efficiency in nations of all economic development levels. More crucially, they clarify why the combination of institutional reform and growing information networks appears to help the poorest nations the most. This is probably a result of poor penetration, which raises the use of communications technology's marginal product. These findings support the study's central thesis, which holds that before the benefits of Internet accessibility can be realised, fundamental institutions and a strong social infrastructure must exist.
Chinn and Fairlie (2007) find that per capita income has a positive relationship with Internet use and raise the possibility of a simultaneity or reverse-causality issue with regard to income and Internet use by using "a technique of decomposing inter-group differences in a dependent variable into those due to different observable characteristics across groups." Neglecting this potential endogeneity will lead to skewed and contradictory empirical findings. This dissertation use dynamic panel data econometric methodologies and controls for the endogeneity concerns inherent in such cross-country empirical development research, whereas Chinn and Fairlie do not address the potential endogeneity problem. Chapter 5 provides a comprehensive description of the empirical findings.
Aker and Mbiti (2010) investigate the effects of increased mobile phone availability on the standard of living in Africa's low-income nations. They draw the conclusion that, despite not being an empirical study, the availability of mobile technology has the potential to advance economic development in sub-Saharan Africa. The potential advantages of expanding ICT accessibility to developing communities are explained by studies like this one, however they do not offer strong empirical support. This dissertation produces a thorough empirical examination where none now exists, so filling this gap in the literature.
1.4 Internet, Trade, and Investment
International trade, and more specifically exports, are one of the more well-developed fields of academic research on the economic effects of the Internet. Exports have a significant role in helping developing nations achieve rapid economic growth.
An analytical framework is provided by Feder (1982) for investigating the growth effects of exports on a cross-country panel of LDCs between 1964 and 1973. His findings imply that export-oriented policies help governments allocate resources more effectively and raise marginal factor productivity, which boosts economic growth in developing nations. Edwards (1998) examines the effects of economic openness (as determined by trade and policy indicators) on total factor productivity using a panel of 93 countries between 1960 and 1990. He discovers that economies with greater openness improve their production more quickly.
In a study concentrating on Canada and the United States, Zestos and Tao (2002) found causal correlations between the growth rates of exports and imports and the GDP of these two nations. This finding suggests that exports may be a substantial determinant of economic growth. Increased trade has been shown to significantly improve social welfare as evaluated by the UN HDI in a cross-country panel by Davies and Quinlivan (2006). It makes sense that Internet use would boost income growth through commerce if Internet use boosts trade.
Freund and Weinhold (2002) found that as Internet penetration rises in a nation, both import and export growth increase in the study of US trade in services. In particular, a 10% increase in Internet usage abroad is linked to a 1.7% rise in service exports to the US. In a later study on global trade (2004), they discover that having connection to the Internet—using Internet hosts8 as a proxy—helps to account for the expansion of trade. A 10 percentage point increase in the number of web hosts in a country causes around a 0.2 percentage point rise in export growth, according to this 2004 study's calculation of trade elasticity. They claim that using the Internet effectively eliminates transportation costs for services and reduces fixed costs. Both of these studies do not address potential endogeneity issues or focus on underdeveloped nations, thus they cannot draw any conclusions about causality. In this dissertation's empirical analysis of the effects of Internet use on exports, endogeneity is attempted to be controlled for utilising instrumental variables and dynamic panel estimating methods.
The study by Clarke and Wallsten (2006) aims to comprehend the extent to which the Internet encourages trade between industrialised and poor nations. Although they acknowledge that the direction of causality is ambiguous, their study reveals that having access to the Internet encourages exports from developing to developed nations. They provide a tool (a regulation dummy) for Internet users to account for this potential endogeneity, and they discover that their findings are unaffected by endogen zing Internet penetration. Using firm-level data from low- and middle-income nations in Europe and Central Asia, Clark's most recent work (2008) supports a causal connection between Internet use and exports and offers additional evidence of the high positive link.
To examine how FDI affects growth, Nair-Reichert and Weinhold (2001) used a panel of 24 developing nations using a mixed fixed and random effects estimate approach. They find evidence that growing FDI promotes growth in emerging nations, but the effects vary depending on the nation. This implies that research that presumes homogeneous effects could produce biassed results. This research investigates countries in sub-samples divided by socioeconomic level in order to examine the possibility of diverse effects on Internet use.
According to Choi (2003), there is substantial evidence that rising Internet usage encourages FDI inward. He argues that increased productivity as a result of Internet use makes the nation more alluring to foreign companies eager to invest. Choi makes no attempt to account for the possibility that FDI and Internet use could be determined concurrently or for the possibility that growing FDI could result in rising Internet use.
In their preliminary analysis of all nations from 1975 to 1998, Reynolds et al. (2004) found that information infrastructure is a significant driver of foreign direct investment. They employ the quantity of telephone lines as their infrastructure measure and apply a two-step residual estimator to try and account for endogeneity. Ko (2007) expands on this by utilising dynamic panel estimators and comes to the conclusion that rising Internet usage draws FDI when there are favourable network externalities in industrialised nations, such as reduced connectivity costs and new electronic markets. Negative network externalities like network congestion and rising Internet usage do not significantly boost FDI in developing nations. Ko's strategy of classifying nations into developed and developing samples is comparable to the strategy applied in this research. FDI is a key control variable in this dissertation's empirical investigation of how Internet use affects economic outcomes, even though the majority of research in the literature concentrate on the factors that influence FDI.
1.5 Internet and Equity Markets
The relationship between domestic capital markets and economic growth has been the focus of recent empirical research on equity markets in emerging nations. Levine and Zervos (1996) used cross-country panel methodologies to examine the impact of a market strength index on GDP growth and found indications of a positive association, but they were unable to establish a definitive causal link. In contrast, Arestis et al. (2001) used a vector auto-regression (VAR) framework to analyse five developed economies in order to investigate the connection between stock market development and economic growth. They discover that while stock markets may support global economic expansion, the banking sector has a higher overall impact. For the advanced economies under study, their findings imply that "bank-based financial systems may be more able to foster long-term growth than ones based on capital markets."
Bekaert et al. (2001) look into how the liberalisation of the financial sector affects the chances of economic growth in a sample of thirty emerging nations. According to their research, financial system liberalisation is linked to actual economic growth. These findings imply that there might be a connection between financial market size and economic outcomes, even if that was not the study's main aim. This dissertation investigates how Internet usage affects the size of the financial markets and makes the argument that the size of the capital markets is a sign of strengthening economic circumstances.
By defining financial globalisation as "the integration of a country's local financial system with worldwide financial markets and institutions," Schmukler (2004) expands on the concept. He argues that although financial globalisation might assist emerging nations, successful integration into the international financial system depends on robust institutions. The overall integration of developing nations into the global economy depends on their capital markets. This dissertation demonstrates how rising Internet usage can support the development of domestic equities markets, which in turn can support the acceleration of economic integration into the larger global economy.
Shirai (2004) examines the function of the domestic equities market and economic development in a thorough case study of China. He comes to the conclusion that China's market does not fulfil the requirements for supporting development because it falls short in three key areas: money raised from market issues is not used productively; state ownership is excessive overall; and questionable accounting practises render firm reports unreliable. This outcome might be a result of restrictions put in place by the Chinese government on the material that can be found online. Access to information is made possible via the Internet, and it is this open exchange of information that enables more effective resource and financial allocation.
Yartey (2008) finds that cross-country ICT diffusion is highly correlated with stock market development (measured by market capitalization to GDP) when investigating the factors influencing technology diffusion using a panel analytical technique. According to the report, ICT development funding is attracted to countries with substantial domestic market capitalizations from neighbouring ICT-enabled nations.
By demonstrating how rising Internet usage raises market capitalisation in developing nations, this dissertation offers fresh analyses to the literature. This could be attributed to the growing accessibility and transmission of information on domestic enterprises through Internet use, which piques the interest of foreign investors. Another mechanism might be the development of an appealing environment for new, tech-savvy enterprises that use domestic equity markets as a result of rising Internet usage.
1.6 Basis for this Dissertation
A foundation for understanding the contribution of this dissertation is provided by the research discussed in this succinct survey of the literature. These studies draw attention to the crucial connections between ICT, economic development, and economic growth that are made in this dissertation's investigation of Internet use and its effects on economic growth. This paper is the first comprehensive empirical investigation that offers a rich understanding of the causal relationship between Internet use and a diverse array of economic and welfare measures in developing countries, despite the fact that there are many academic studies exploring aspects of economic development and ICT.
2 Theoretical Framework and Estimation Methods
The two primary empirical techniques used to estimate the equations are presented in this chapter together with the theoretical framework that takes into account Internet use in a production functional form that can be used to investigate the relevant metrics of economic development. The theoretical framework is presented in the first section, and the estimating techniques are presented in the second.
2.1 Model Specification
Due to the ease of access, communication, and use of information offered by the internet, it has an impact on economic development. The efficient creation, improvement, and dissemination of information on the Internet has a variety of direct and indirect effects on economic development. Access to information can be used as a direct input to improve production decisions and labour and capital allocation. By lowering search and transaction costs and increasing export options, information can give businesses the chance to take advantage of economies of scale. The Internet offers a new mechanism for gathering and exchanging information that enables businesses to learn about new potential markets for finished goods and services, discover new inputs and production methods, and locate competitive prices for inputs that lead to more effective input ratios.
According to studies that have already been done, using the Internet to obtain more information increases both labour and capital productivity. Access to the Internet offers cutting-edge communication technologies like email and instant messaging that make the sharing of ideas easier. Knowledge is produced as a result of this information exchange, which is essential for technological advancement. Information therefore advances knowledge, which advances human capital.
I start with a country's production function, F, for an outcome Y, throughout period t, in a manner similar to the presentation in Barro and Salai-Martin (2004) that comes after Solow-Swan and Ramesy:
Yt = At · F (Kt, Lt, INETt, Xt) (1)
Yt is an economic outcome, At is total factor productivity (TFP), Kt is the capital stock, Lt is labour, INETt is the number of Internet users, and Xt is a vector of controls that might include a variety of elements, such as institutional policy initiatives. In this dissertation, I argue that Internet use, or INETt, has a favourable impact on the results of development.
∂Yt/∂INETt > 0. (2)
I model at as having both a deterministic and a stochastic component in accordance with what can be seen in the data. The stochastic component, ezt, produces random fluctuations around the trend growth path that are expected to follow an MA(1) process, whereas the deterministic component, egt, corresponds to an underlying trend with a constant rate of growth.
At = egtezt ; At > 0 (3)
zt = ρzt−1 + t; t ∼ iid. (4)
As a result, egt+zt explain how technical advancements outside of the Internet behave.
This concept includes the consequences of both exogenous technological advancement and Internet use. Learning about new technology is made possible by having access to the internet. Additionally, having access to the Internet makes it easier for new technologies to be adopted and disseminated because it offers a variety of channels for discussing their training and uses, such as email, websites, online forums, and shared academic courseware.
ICT in general, and Internet use in particular, have been found to have a positive impact on economic production and factor productivity in nations at all levels of development to varied degrees, as can be observed from the earlier studies in the literature review. I build on this justification to investigate how Internet use affects a variety of outcomes related to economic growth.
Given that (1) lacks a clear functional form in economic theory, I take into account the widely-used Cobb-Douglas intensive (or per capita) type production function that takes into account the effects of Internet use, other technical advancements, and the vector of controls:
You'll see that this equation keeps a stochastic component and a trend component, gt component, zt. I model the trend and cyclical behavior as a time-invariant constant, α, plus the log of lagged realizations of the dependent variable, lnyt−1, and an iid stochastic shock, t:
gt + zt = gt + ρzt−1 + t = α + βlnyt−1 + t. (7)
Now substituting (7) into (6) and rearranging, I arrive at a log-linear model specifi- cation for an outcome yj, in country i, during time period t:
The major explanatory variable of interest in this generalised log-linear model specification is the number of Internet users in country I at time t, with coefficient. yjit is a specific per capita development outcome that is indexed by j, and inetit is now per capita Internet use. The lagged outcome variable lnyji, coefficient t1's is. Kit is the per-capita capital stock with coefficient, Xnit is one of the n row control vectors with coefficient, is the intercept, and it is a stochastic error term.
The variables in the vector X, which are defined for each equation, can change based on the specific development outcome being studied. The control variables are the same for three of the four models examined in this dissertation, though. This model employs log-linear models because the log transformation minimises the sensitivity of the resulting estimates to outliers and reduces the range of the data. Importantly, the log-log equation coefficients may be simply translated into elasticities, enabling direct comparison of the effects of Internet use on various outcomes of economic development.
In all of the estimation equations for each development outcome, the coefficient, on the measure of Internet users is of particular interest in this research because of its sign, magnitude, and statistical significance. In the presence of dynamics, the coefficient on the lag-dependent variable is anticipated to be positive, less than 1, and statistically significant. The elasticity of Internet use, will be non-negative with varying magnitudes and significance depending on the development outcome being researched and the specific sub-sample used, according to the study's premise.
he models used in the literature on economic development frequently use GDP as a function of the outcome under study. In this dissertation, I offer a general model specification that can be applied to empirical research on a variety of GDP-related economic development outcomes. As I view GDP as an economic development outcome measure that can be explained by the same controls as other measures, it is not included in this method as a control (for other outcomes). In fact, I think that this model's design might be helpful for investigating other development outcomes that I shall take into account in Chapter 6.
For the three per capita outcomes being examined here, the general log-linear estimating equation is:
lnyjit = α+βlnyji,t−1+γlinetit+δlcapit+ζ0laidit+ζ1secschit+ζ2lifexpit+ζ3instit+it. (9)
In this definition, lny is a logged per capita economic outcome that is indexed by j, linet is a measure of institutional quality, lcap is a natural log of per capita fixed capital formation, laid is a natural log of per capita net foreign aid, secsch is the length of secondary school in years, lifexp is a measure of life expectancy at birth, and laid is a natural log of per capita fixed capital formation. The variables chosen as controls are those that are frequently employed as proxies for the essential elements of economic development in growth and development empirics: fixed capital formation, foreign aid, education, health, and institutions. Since these models are estimating per capita results, there is no labour control.
Numerous empirical studies in the literature support the selection of control variables. At least since Pa- panek's ground breaking paper on the causes of economic development, controls for capital formation and aid have been in place (1973). Studies that include controls for human capital using educational attainment and health indices as proxies are common in the literature on economic growth. For instance, secondary education was taken into account in two important growth papers: Barro (1991) and Mankiw et al (1992). According to Sachs and Warner, life expectancy is a frequent proxy for health status (1997). Several significant articles have underlined the role of institutions in economic development, as was previously discussed in the literature review above, including Acemoglu et al. (2001), Rodrik et al. (2004), and Banerjee and Duflo (2004). Chapter 4 contains a detailed overview of the data sources and the particular control variables used.
The general model in (9) is followed by the log-linear model for logged per capita GDP, lgdp:
lgdpit = α+βlgdpi,t−1+γlinetit+δlcapit+ζ0laidit+ζ1secschit+ζ2lifexpit+ζ3instit+it
This approach differs somewhat from those frequently used in the literature for estimating the impact of ICT on exports. Typically, some measure of GDP is used as a control in empirical estimations of export growth. In this dissertation, I examine a variety of economic outcomes that are conditioned by the same variables as GDP. As a result, I apply the same model to estimate how using the Internet affects each of these metrics. Thus, the generic estimate equation (9) and the per capita GDP equation (10) above are both used in the log-linear model for per capita exports, lexp:
lexpit = α+βlexpi,t−1+γlinetit+δlcapit+ζ0laidit+ζ1secschit+ζ2lifexpit+ζ3instit+it
This expands on well-known models to investigate the impact of Internet use on exports.
The factors that influence capital markets in emerging nations have received scant empirical study. The majority of researches, as shown in the literature review, concentrate on financial markets as a predictor of economic growth rather than looking into the variables that influence financial market expansion. This dissertation introduces a novel idea: the size of domestic equities markets as an outcome of economic development. I contend that the same methodology that was used to examine other outcomes, like GDP and exports, may also be used to examine how the Internet affects market capitalization. The Capital Asset Pricing Model (CAPM) was expanded by the theory of arbitrage pricing in order to investigate the numerous factors that affect the pricing of individual stocks. By extending this notion and incorporating it into existing models, such as those used by Holzmann (1997), Perotti and van Oijen (2001), and Bekaert et al. (2001), I aim to investigate the effects of Internet use as a determinant of market capitalization, conditional on a number of economic development factors. As a result, the estimation equation for market capitalization per person, or lmcp, is represented similarly to the equations above:
lmcpit = α+βlmcpi,t−1+γlinetit+δlcapit+ζ0laidit+ζ1secschit+ζ2lifexpit+ζ3instit+it
I anticipate that the coefficients for fixed capital formation, education, health, and institutions will all be positive for all three of the aforementioned estimate models. Although there is still discussion over the impact of help on economic growth, I predict a negative correlation because underperforming nations receive more aid. One of the endogeneity issues covered in the section below on estimation methodology is this one. With the exception of low-income nations, I anticipate that the Internet use coefficient in all equations will be positive and statistically significant across all economic strata.
The UN Human Development Index (HDI) is a composite index that takes into account a number of welfare indicators, including GDP per capita, adult literacy rate, and life expectancy at birth (United Nations 2008). From 1980 to 2000, it was calculated every ten years, then starting in 2005, the UN started calculating the metric annually. In order to fill in the data for the sample's missing years, this dissertation added interpolated HDI values. The HDI cannot be represented in the same way as the other equations since it is a mix of several different development indices.
Due to the structure of the index and the timing of the measurement, the model for estimating the effects of Internet use on HDI takes a different form. Although the HID's underlying components clearly exhibit dynamism, the index has historically not been monitored annually or at set intervals; instead, the measurement period has changed during the course of the index's development. As a result, the lagged dependent variable is not used in this model. As per capita control measures may result in skewed estimates because we lack the necessary data to reliably assess variables in terms of per capita, the HDI equation is log-linear and the controls are not. Here, the terms "linetto" (total Internet users), "lcaptot" (total fixed capital formation), and "laidtot" (total overseas assistance) are all expressed in current US dollars. A control for the size of the labor force, labor, is introduced in place of the proxies for health and educational attainment since those are factors in the index.
hdiit = α + γlinettotit + δlcaptotit + ζ0laidtotit + ζ1llaborit + ζ2instit + it (13)
I anticipate that the coefficient on in the middle-income sample will be positive and significant, much like in the estimate equations for GDP, exports, and market size. While those for aid and labour are predicted to be negative, the elasticity for capital formation and institutional strength are predicted to be positive.
The model specifications employed in this dissertation are aesthetically pleasing on first glance since they incorporate key components from significant growth and development empirical investigations. The purpose of these specifications is to examine the causal relationships between Internet use and various metrics of economic development activities. They adopt, adapt, and extend the common standards. I offer a consistent empirical framework for assessing the variety of economic development outcomes reported in this dissertation and others that will be investigated in the future by merging the key components of the model specifications from across the growth and development literatures.
2.2 Estimation Methods
In these models, there are two possible sources of endogeneity. They are the endogeneity of economic assistance and the direction of causality of Internet use. As people with more disposable income become more sophisticated in their demands for goods and services and businesses modernise using increased profits, rising domestic production and productivity, as measured by the per capita GDP, can undoubtedly lead to an increase in Internet availability and use. Similar to this, low productivity (low GDP per capita) will lead to increased help. In order to identify a causal relationship between Internet use and economic development, estimate approaches that take endogeneity into account are required by these models' sources of endogeneity. By addressing these endogeneity issues as well as the presence of dynamics, which complicates the application of the standard panel estimators, this research adds to the earlier development and ICT literature.
For empirical economic analysis, estimation using Ordinary Least Squares (OLS) is the standard starting point. It can offer accurate initial estimates of the marginal effects of the variables being studied. The equations can be stacked, and pooled ordinary least-squares can be used to estimate this model specification with ease. If the conditional mean is accurately specified, the errors are independent and identically distributed, and there is no multicollinearity in the regresses, OLS estimation can identify the parameters of interest. However, OLS estimation is biased and inconsistent when endogeneity and dynamics are present—when there is a correlation between the error term and any of the regresses—which is likely to happen in cross-country panels when examining aggregated macroeconomic measures because some of these measures are probably determined simultaneously. The aforementioned equations are dynamic because they contain lag dependent variables and potentially endogenous control variables.
Because the lagged realisation of the dependent variable is linked with the nation fixed-effects, using OLS to estimate (9) will result in issues. Both the lagged realisation and the country effect will be impacted by a shock to a country in the prior period. This goes against an OLS presumption. In the social science literature, temporally demeaning the data and then estimating the model on the time-demeaned data using OLS is a typical method for solving this problem. This method is known as a fixed-effects (FE) estimator. Since yit is correlated with it, the error term in the FE regression model, (it I will also be correlated with the response variable, (yit-yi), which is the same issue as with the OLS estimation. Fixed-effects estimates for each of the models are presented for comparison even though they do not solve the endogeneity issue.
More sophisticated econometric methods are required to account for the endogeneity contained in the model. Two more estimating techniques are applied in the empirical study in this dissertation using equations derived from (8). The first addresses Internet use and helps endogeneity and dynamism of the response variable by using Dynamic Panel Data estimation, the most common cross-country panel data estimate technique utilised in the literature. The second method examines the response variables as draws from a distribution made up of groups of unique continuous distributions of subpopulations using Finite Mixture Model estimation.
2.2.1 Dynamic Panel Data (DPD) Estimation
When dynamics and endogeneity are present, the Dynamic Panel Data estimators—more specifically, the Dynamic Panel System and Difference General Method of Moments (GMM) estimators—are frequently employed in the literature to estimate models on cross-country data. According to Bandyopadhyay et al., there are three reasons why these estimators are favoured (2011). The first step is to incorporate the long-lasting effects of Internet use into a dynamic framework. Second, there are significant endogeneity issues with regard to Internet use, financial help, and the metrics of development outcomes being taken into account. Two-stage least squares methods cannot be used without clearly defined and understood instrumental variables. Third, it's critical to account for unobserved national heterogeneity that can be linked to the outcomes of the study.
If we refer back to the general autoregressive model mentioned in equation (8) from Cameron and Trivedi (2005):
yit = α + βyi,t−1 + γlinetit + δkit + ζnXnit + it (14)
where the error term it is composed of ηi, representing time-invariant country specific e?ects, and uit, an idiosyncratic error that varies across countries and time periods:
it = ηi + uit (15)
E[ηi] = E[it] = E[ηiuit] = 0 (16)
then the autoregressive model can be specified:
yit = α + βyi,t−1 + γlinetit + δkit + ζnXnit + ηi + uit. (17) n
The fixed effects (FE) and random effects (RE) models are the two frameworks that are utilized to account for i. While random effects, or between, estimation implies that I is a country-specific disturbance in each time period, fixed effects, or within, estimation assumes that I is a country-specific constant in the regression model. While the random effects approach implies that I is uncorrelated with X at all times, the fixed effects approach presupposes correlation between the unobserved heterogeneity and the regressors in X. When all explanatory variables are strictly exogenous, which is not the case in this situation; these panel estimating procedures produce estimates that are consistent.
It is necessary to use an estimator that generates reliable results even when dynamics and endogeneity are present. Dynamic GMM panel estimation methods that make use of the linear moment limitations implied by the aforementioned dynamic panel estimation equation have been introduced and extended by Arellano and Bond (1991), Arellano and Bover (1995), and Blundell and Bond (1998). Lagged differences of the dependent variables, endogenous regressors, and current values of strictly exogenous regressors are used as instruments in the DPD, an instrumental variable GMM estimator.
Since it can amplify gaps in unbalanced panels, the system DPD GMM estimator is preferable in this application over the difference estimator. This serves as the inspiration for the forward orthogonal deviations transformation (used in the estimation of the models in this dissertation), which "subtracts the average of all future accessible observations of a variable instead of the prior observation from the contemporaneous one. It is computable for all observations except the last for each individual, regardless of the number of gaps, minimizing data loss (Roodman 2006). In the presence of non-spherical errors, the two-step estimator is recommended over the less effective one-step estimator, and I choose the two-step form because I believe models with endogenous regresses and dynamics will have non-spherical errors.
The robustness of the obtained estimates is assessed using the common DPD statistical tests of over identification limits and serial correlation of the errors terms. Following the regression findings for each model in the appendices are tables with the test statistics.
2.2.2 Finite Mixture Model (FMM) Estimation
According to the World Bank's classification of countries by income, there are three distinct categories of development that can be identified: low, middle, and high income. In this dissertation, I make the assumption that Internet use affects countries differently depending on their level of economic development. This suggests that there may be some uniformity within each of the three distinct income classes that would enable a separate analysis of each income class. This estimate problem is ideally suited to use Finite Mixture Model (FMM) techniques since it is assumed that the distribution of outcomes under consideration is distributed into a finite number of reasonably homogeneous classes.
The FMM estimation technique assumes that the variable of interest is derived from a distribution that is made up of an additive mixture of distributions from several sub-populations or classes in order to model unobserved heterogeneity. FMM estimate is employed in various economic applications even if it is not yet widely used in development literature. In that it offers an alternative method of modeling heterogeneity, mixture modeling is appealing.
FMM estimation is used by Owen et al. (2009) to investigate the issue of country growth rates. They examine a number of latent class predictors, including models with two to five different classes and measures of latitude, settler mortality, and country landlocked status. They discover that, among the variables considered, institutional quality is the best predictor, and that a two-class model best fits the data. They come to the conclusion that country growth rate heterogeneity is ignored by single class pooled analysis. I propose that, contrary to their strategy, the latent class membership is determined by one's degree of income.
The identification of the precise, perhaps latent classes can be challenging without some natural interpretation, Cameron and Trivedi (2005) indicate, despite the fact that I have assumed a finite number of classes that account for country het- erogeneity. The latent classes have a theoretical foundation because they directly match to the World Bank country income classifications. This is because I am assuming that the effects of Internet use on economic development outcomes differ depending on the income level of the country.
The FMM estimate approach may model the various elasticity between income class even though it does not directly address endogeneity and dynamics issues. It models a different distribution for each class. Low, medium, and high income component densities are each represented by proportions c, where:
A probabilistic mixture of the densities from the three income classifications can be used to generalize the densities of the economic metrics in this study (or components.) This is represented by the equation below:
g (y | Θ) = π1g1 (y | Θ1)+ π2g2 (y | Θ2)+ π3g3 (y | Θ3) (20)
where gc (y | c) is the individual class density of the variable y given the parameter vector c and c is the percentage of the mixture density as defined previously. The following equation, where the mean of the distribution for the income class is discovered using equation (8), can be used to explain each distinct class density since each one is normal by construction:
This equation is not explicitly estimated; rather, the parameters for the means of the various class distributions used for mixing are provided by the linear estimation equations deriving from (8). The standard deviations, c, the regression coefficients, and the mixing probabilities can all change for each class when using this estimate method. Based on the income classification of each country, the model is estimated using maximal maximum likelihood estimation with set mixing probabilities.
The elasticity (or marginal effects), which are calculated at the means of the covariates indicated in the models, are then determined after the models have been estimated. Similar to how the coefficient estimates from the OLS and DPD estimators are interpreted, these values. As I anticipate finding a comparable sign, magnitude, and significance for the elasticity on Internet use regardless of the estimation method, using FMM to estimate each of the models provides a robustness check to the model parameters.
3 Data Sources and Panel Construction
A variety of socioeconomic variables are needed in order to conduct an empirical examination of the effects of Internet use on global economic development. To conduct this inquiry, data on Internet usage, macroeconomic indicators, institutional effectiveness of the government, and population health status are all required. The panel utilized in the empirical analysis had to be constructed from a variety of data sources.
This study focuses on the period from 1996 to 2007, which saw a dramatic increase in global Internet usage. There are two major reasons why this time frame was chosen. First, due to the low penetration rate of the Internet outside of a few wealthy nations, data on Internet use in general, and on low and medium income countries particularly, prior to 1996 are not generally available. Second, as of the date of writing, the data sources that are currently available only provide information on Internet usage for a wide number of nations as of 2007. Although there are some observations for 2008 and 2009 in the data used, most countries' data coverage is incomplete. Therefore, observations made after 2007 are not included in the economic projections.
The number of Internet users in a nation is the primary explanatory variable of interest for this study. The metrics for GDP, exports, assistance, and fixed capital creation are all expressed in current US dollars. All four of these variables are normalized by the population to obtain the per capita measurements. The life expectancy at birth in years serves as a proxy for health, while the number of years spent in secondary school serves as a proxy for educational achievement. Both the Institutional Quality Index and the HDI are indices that are used to assess the general wellbeing and quality of institutions, respectively. In the sections that follow, each variable and the specific data sets it was obtained from will be examined in detail.
The quality of the data available determines the robustness of any empirical research, and this study is no exception. Aggregated country level statistics are present in all the data sources used in this inquiry. Data that has been combined for each nation and year is relevant because this study examines phenomena at the aggregate country level. Undoubtedly, there is measurement error in these data, but since they are the main sources for cross-country panel studies in the economic literature, we rely on the supposition that the errors are random and not systemic and do not, thus, add bias into the samples.
In order to offer the macroeconomic measures required for this study, four primary data sources were chosen. The panel that results from fusing these data sources includes details on 202 nations and observations on economic, sociological, governance, communications, and health characteristics for the years 1996 to 2007. The identification, description, and examples of use in the literature for the variables used in this investigation are provided in Table 1. Appendix B contains descriptive data for both the complete sample and the low- and middle-income sub-samples. In Appendix C, you'll find histograms of the major metrics applied in this investigation. The four datasets used in this dissertation's analysis are briefly described in the sections that follow, along with the variables that were taken from each.
3.1 International Telecommunications Union: ICT Indicators
One of the most complete sources of statistics on telecommunications is the 2008 International Telecommunications Union World Telecommunication ICT Indicator (ITU/ICT). This database offers comprehensive global information and communication technology indicators. The International Telephony Union (2008) states that "the data are gathered by an annual questionnaire submitted to official country contacts, typically the regulatory authority or the ministry in charge of telecommunication and ICT."
The number of Internet users in a nation, the primary explanatory variable of interest for this study, is taken from this data collection. The number of Internet users per capita (or the percentage of the population having Internet access) gives the level of detail required for this inquiry, even if other measures are also employed in the literature to comprehend the implications of the Internet's rapid growth.
3.2 World Bank: World Development Indicators
The World Bank World Development Indicators (WDI) 2009 (World Bank 2009) was chosen to provide the comprehensive measures of economic activity covering hundreds of development indicators on 208 countries covering the years of interest: 1996-2007. It is the leading data source for empirical development economic investigation. The macroeconomic variables and income groupings for the 202 nations utilized in this analysis are provided by this data set. Due to insufficient information or uncertainty surrounding their classification as independent nations, the following six nations—Hong Kong, Macao, Puerto Rico, the US Virgin Islands, American Samoa, and West Bank and Gaza—were removed from the WDI data collection.
This data set is where the majority of the economic metrics used in the study were found. A thorough investigation into the effects of Internet use can be done using a variety of measures of economic activity, such as GDP, export value, and domestic equity market capitalization. As controls, a number of the data set's measurements are used. Net foreign aid offers a way to counteract outside economic impacts. Fixed capital formation is used to account for domestic investment. To reduce the potential skewing effect of very big countries, all of these indicators are normalized by population size to provide per capita measures.
Every financial indicator is expressed in dollars of the current year. The local currency is converted into US dollars using current year exchange rates. The World Bank's national accounts data are the source of the GDP, exports, and fixed capital creation metrics. Information from the Organization for Economic Cooperation and Development (OECD) is used to calculate net official aid amounts. Statistics on market capitalization are sourced from Standard & Poor's.
I utilize the life expectancy at birth expressed in years to account for health status in the investigated nations. This metric is based on population data from the United Nations. The length of secondary schooling in a nation, which is also measured in years, serves as a proxy for educational achievement (or grade levels.) Data from the Institute for Statistics of the United Nations Educational, Scientific, and Cultural Organization (UNESCO) are used by the World Bank to measure educational outcomes. Two often used indicators of health and education in the development literature are life expectancy and the number of years spent in secondary school.
In order to isolate the effects of the Internet on nations at various stages of development, the sample was stratified using the World Bank's classification of income for each nation. Three income levels are used to categories nations, with upper and lower income levels being subdivided into high and medium.
The World Bank Atlas method21 was used to determine the 2009 Gross National Income per capita for each economy. Low income is defined as $995 or less; lower middle income is defined as $996 to $3,945; upper middle income is defined as $3,946 to $12195; and high income is defined as $12196 or more (2009).
These income categories were used to construct three sub-samples for this study: low, middle (consisting of lower and upper middle income), and high income.
3.3 World Bank: Worldwide Governance Indicators
The World Bank Worldwide Governance Indicators (WGI) compile aggregated data on 212 countries between 1996 and 2007 on six aspects of governance: voice and accountability, political stability and lack of violence, government effectiveness, regulatory quality, rule of law, and corruption control. These aggregate indicators, according to Kaufmann et al. (2009), are weighted averages of the underlying data, with the weights representing the accuracy of the various data sources.
In line with Decker and Lim (2008), this study substitutes the maturity and stability of governmental institutions with an equally weighted average of five WGI components. The Voice and Accountability metric, which may be a better proxy for democratic representation in government than the strength of national institutions, is the component left out of the weighted average.
The body of empirical evidence in the literature makes it abundantly evident that effective governmental institutions are essential for economic development in emerging nations. A government needs functioning institutions in order to offer ICT infrastructure and make Internet access available to its citizens. The created index serves as a gauge of institutional efficiency and a crucial check for the empirical studies.
3.4 United Nations: Human Development Index
The Human Development Index (HDI), produced by the United Nations Development Programme in 1995, 2000, and 2003–2007 and released in 2008, is a composite indicator of income, education, and literacy for 176 countries.
The HDI, or human development index, analyses a nation's average performance in three fundamental areas of human development: health, knowledge, and a respectable quality of life. A person's level of knowledge is determined by their life expectancy at birth, adult literacy rate, and combined primary, secondary, and tertiary gross enrollment ratio, while their standard of living is determined by their GDP per capita (PPP USD) (United Nations 2008).
In order to assess the broadest measure of economic progress and population well-being, the UN HDI is employed.
The sample size is greatly decreased when utilizing the UN HDI as a response variable because it is not calculated annually. Cubic spline interpolation techniques are utilized to fill in the missing observations for years where the UN did not compute HDI values in order to increase the number of observations available for the empirical study. Both the original indices and the extended measures following interpolation and extrapolation are covered in the descriptive statics supplied in Appendix B.
3.5 Panel Construction
Variables from all four of the aforementioned data sets are present in the panel that was developed for the empirical inquiry in this dissertation. In order to enable merging, each country observation in each data set is given a consistent number value based on the UN country name and the three-character World Bank country code identification. Prior to panel building, each individual panel is first processed to remove unnecessary variables, fix formatting issues, and set the shared numerical country identifiers.
Very minimal processing is necessary to prepare the two World Bank data sets for incorporation into the panel. Prior to the integration, region identifiers that are missing for some nations in the WDI database must be added.
Both the UN WDI and ITU ICT databases need to be redesigned before merging. Reshaping is the process of rearranging panel data so that all variables for a given subject are contained in a single observation rather than in a lengthy format, where variable entries are distinct observations. Wide panels are typically used for panel econometric analysis, as they are in this study. A custom Python script was used to reshape the data by reading comma-delimited files containing information from the relevant sources, compiling and rearranging the data, and producing new comma-delimited files suitable for econometric analysis.
The indicator variables for area and country income levels were produced after the four main data sets were combined. The logged per capita values of the relevant variables were then generated.
3.6 Variables and Sample Stratification
With the countries stratified by income level, all estimations are performed on both the whole sample and reduced samples. The countries are divided into high, medium, and low income groups based on the 2007 World Bank income level. This coding is reflected by the introduction of a variable income code. I employ the three main categories, High, Medium, and Low Income, to identify the sub-samples for this study, despite the fact that the World Bank also defines five additional income sub-categories in addition to the three basic categories.
For the entire sample as well as each of the three sub-samples, each model is calculated. The coefficients from the regressions on each sample are then compared to the variable of interest, linet, which is the natural log of the number of Internet users per capita in a particular nation and year. In this study, log transformations are employed for a number of significant reasons. By using the observation logs, one can lower the sensitivity of the models to outliers and the data range. Additionally, elasticity can be inferred from the coefficients of the log converted approximated equations.
The availability of data poses a significant challenge, particularly for empirical assessments of developing nations. In the early years of the sample used for this study, this issue exists for several of the nations. Additionally, data on a few economic variables from various years within the study period is unavailable.
3.7 Summary Statistics
The whole data collection spans the years 1996 to 2007 and includes indicators for 202 nations. If all of the indicators were accessible for all nations and all years, there might be a maximum of 2,424 observations. With the exception of the market capitalization measure and the UN HDI, the majority of variables had decent coverage across the study period.
The summary statistics indicate that there are considerable differences in Internet use across the panel and sub-samples because the standard deviation is much higher than the mean. This is probably because Internet usage has changed quickly over time.
A list of the nations covered in this study is provided in Appendix A, arranged according to World Bank income categories. The three samples under investigation—the overall data set, the sub-sample of middle-income countries, and the sub-sample of low-income countries—are all summarised statistically in Appendix B. Histograms of the data distributions for the main variables of interest may be found in Appendix C.
4 Empirical Results
This dissertation's major goal is to examine how Internet use affects four different development outcomes: per capita GDP, exports, equity market size, and a composite index of human welfare, or UN HDI, as a stand-in. The data from the complete multi-country panel data set and three sub-samples based on income level (high, middle, and low) from the data set mentioned in Chapter 4 are presented in this chapter together with the empirical findings of estimating the four models indicated in Chapter 3 on them.
I covered the theoretical underpinnings of the empirical inquiry in Chapter 3 and demonstrated the presence of dynamics and endogeneity in the estimated equations that were provided. As a result, the DPD estimator is the most suitable because it can take into consideration these circumstances. The effects of Internet use on these outcomes may have error disturbances that vary depending on income class. In these circumstances, the FMM estimator offers an alternative method for assessing the effects of Internet usage by modeling the distribution of the outcomes as a combination of the component distributions and using the income classes as distributions. So as a robustness check, I estimate the equations using the FMM estimator. I've also included the outcomes of all equations' OLS and FE estimations for comparison's sake.
The presence of first order autocorrelation and the absence of second order autocorrelation in the error terms are critical conditions for the consistency of the DPD GMM estimator. For each estimated equation, the outcomes of the Arellano-Bond tests for first and second order autocorrelation in the errors are presented.
Lagged values of the endogenous variables as well as current and lagged values of the exogenous regressors are used as instruments in the differenced equation by the Arellano and Bond system GMM DPD estimator. The use of the Internet, assistance, and lifespan are all considered to be endogenous in all of the models that follow. All of the exogenous variables, as well as the lags of the endogenous and predetermined variables, are used as instruments in the DPD estimator.
The Sargan, Hansen, and Difference-in-Hansen tests are three examples of statistical tests that can be used to assess the validity of the instruments. The over identification limits are tested using both the Sargan and Hansen statistics, which serve as joint tests of model specification and instrument validity. The erogeneity of the instruments is examined using the Difference-in-Hansen statistics. The outcomes of these tests are presented for each DPD estimation in the table that follows the regression results in order to assess the model specifications in the estimated equations.
The estimated elasticity on Internet use are introduced for each of the four models shown below in a table with a column for each estimation technique: The OLS estimate is shown in column (1), followed by the FMM marginal effects estimate in column (2), the panel fixed effects (FE) estimate in column (3), and the DPD estimate in column (4). Under each coefficient, heteroskedasticity-robust standard errors are presented in parenthesis. Appendices D–G contains comprehensive tables of regression results and test data for each estimated equation for each model.
There are five sections in this chapter. The GDP equation is discussed in Section 5.1, the exports equation is presented and discussed in Section 5.2, the market size equations are presented and discussed in Section 5.3, and the HDI equation is discussed in Section 5.4. The statistical findings are compiled in the final subsection, 5.5.
4.1 GDP Per Capita (lgdp)
Estimates of the impact of Internet use on per capita GDP are provided in this subsection. A scatterplot of the correlation between log per capita GDP and log Internet use is shown in Figure 2. Table 2 shows, for the entire sample and for each of the three subsamples of low, middle, and high income countries using all four estimators, the elasticities of log per capita Internet use on log per capita GDP. Appendix D contains all of the regression test statistics as well as the complete regression results.
The model fits the data reasonably well, according to the DPD test statistics for the entire sample (Table D2). According to the Arellano-Bond tests, there is first-order autocorrelation present in the mistakes but not the necessary second-order autocorrelation. Valid instrument over identification limits are not ruled out by the Sargan or Hanson tests, and the Difference-in-Hansen measures do not rule out instrument erogeneity. These tests show that the model is recognised and the instruments selected are reliable.
Starting with the estimation on the entire sample, the coefficient on Internet use is positive and statistically significant for each estimation technique; as predicted, the only difference between the estimates is their absolute magnitude. The results consistently demonstrate that growing Internet use considerably has favourable effects on per capita GDP, regardless of the model estimation technique. The latter techniques, which account for country-specific effects, dynamics, and endogeneity, result in a larger rise in the elasticity on Internet use between the OLS and FMM estimators and the FE and DPD estimators. For instance, the absolute magnitude difference between the DPD estimate and the OLS estimate is 27%. Due to endogeneity, it is plausible that OLS is underestimating the effects of Internet use. According to estimates made using the DPD approach on the entire sample, a 10% increase in Internet users results in an average 3.2% rise in per capita GPD. This supports the fundamental thesis of this dissertation, which holds that rising Internet usage improves both overall development outcomes and GDP per capita in particular.
The coefficient on the lagged realisation of per capita GDP is positive and significant, demonstrating the presence of dynamics, as expected, as seen in the full regression findings of the DPD estimator shown in Table D1. Even though the elasticity of help is negligible and modest, it has a negative sign. The idea that aid does not contribute to economic progress as measured by per capita GDP may be poorly supported by this. Again as predicted, the coefficient for per capita fixed capital formation is positive and substantial. Secondary education has favourable effects, but it doesn't appear to have any impact on life expectancy (statistically indistinguish- able from zero.)
Surprisingly, the estimate of the effects of improving institutional quality, The elasticity is considerable and negative.
This may imply that offering high-quality schools has a direct cost that may be calculated as a drop in per capita GDP. It's also conceivable that more advanced institutions could result in lower revenue at less developed stages of development. Before increases in institutional quality result in revenue increases, nations might need to reach a certain institutional threshold. Present research is being done in this area.
I estimate the model with three extra control variables: time, income level, and an interaction term for time and income level taken together in order to separate the effects of Internet use from any temporal or income class effects. These findings are displayed in Table D1's column (5). We can observe that the model is tolerant of the inclusion of these controls because the coefficient on Internet usage is still appreciably positive. This lends more credence to my claim that there is a connection between Internet use and per capita GDP.
All of the coefficients on the model's variables have consistent signs and magnitudes, as shown by comparing the findings of the OLS, FMM, and FE estimates to the DPD estimates. Obviously, depending on the estimator, a particular coefficient's absolute value and importance will vary. This indicates that the model is resistant to the different estimating techniques.
There are quantitative differences even if all of the coefficient estimates for Internet use are qualitatively identical. The FMM estimation method appears to have captured the average marginal effect across all income levels because the results of the FMM closely resemble the OLS results across samples and in the middle-income estimates of the other estimators as well. Although the DPD estimator's estimates of elasticity tend to be bigger in size, they share the same sign, accuracy, and significance as those from the other estimators. The FE estimates' coefficients' signs likewise line up with those of the other estimations.
Depending on the level of development, Internet use may have different effects on income. So, using three subsamples of rich, middle, and low income nations, I estimate the equations. The output of these estimations is displayed in Table 2's panels (b) through (d). High-income country estimates are shown in panel (b), middle-income country estimates are shown in panel (c), and low-income country estimates are shown in panel (d).
I generate indicator variables for the high and middle income classes and interact these with the measure of Internet use to test the hypothesis that the effects of Internet use vary by income class. The high-income interaction term is found to be significant after I estimate the entire sample GDP equation with these interaction terms. This demonstrates that Internet effects differ depending on income level.
We observe varying degrees of the effect for each sub-sample when we examine the estimates of Internet use on the three sub-samples in Table 2 (the complete results are in Tables D3-D8). Contrary to the OLS and FE results, the magnitude of the Internet use coefficient in high-income nations is bigger than that in low- and middle-income countries, as would be expected, but it is not statistically significant. This may be partially attributable to the DPD estimator's shortcomings due to the DPD estimator's tiny sample size.
Surprisingly, increasing Internet use by 10% raises per capita GDP in low-income nations by 4.2% on average, a higher effect than the 3.2% rise in middle-income nations. This might be because Internet use is more prevalent in low-income nations, where there are labour and resource surpluses, but the information flow required to make efficient allocations is constrained. Using the Internet more frequently gives you access to information that can enable more effective allocations. The DPD estimates of Internet use on per capita GDP are compared in Figure 3 for each income class, with the statistically negligible high-income estimate shown as 0.
The results reported here are consistent with the expanding body of research on the advantages to economic growth from ICT and Internet deployment, even if there are no other empirical studies looking into the direct causal effects of Internet use on GDP at the time of writing. According to Röller and Waverman's (2001) research of telecommunication infrastructure in OECD nations, there is a causal relationship between the expansion of ICT infrastructure and the increase of total production here sented.
More intriguingly, according to Czernich et al. (2009)'s analysis of broadband Internet penetration in OECD nations, a 10% increase in broadband Internet access improves per capita GDP growth by 0.9% to 1.5% on average. The magnitude and importance of the effects of Internet use on per capita GDP show a positive effect, even though it is challenging to directly compare this estimate to those provided in this dissertation.
Overall, these findings provide a consistent picture of the benefits of increased Internet usage on per capita GDP, independent of the nation's income level. According to expectations, the amount of the effect varies by the country's income class. This firmly backs up the thesis statement for this dissertation. Chapter 6 presents assessments of the results' policy ramifications.
4.2 Per Capita Exports (lexportspc)
The estimated impacts of Internet use on real export income in US dollars are covered in this subsection. The scatter plot of the correlation between Internet usage and export revenue for the entire sample in 2007 is shown in Figure 4. The estimated elasticities of the log of per capita Internet use on the log of per capita export revenue, calculated using all four estima- tors, are shown in Table 3 for both the entire sample and each subsample. In Appendix E, the complete set of regression findings and regression test statistics are shown.
The Arellano-Bond tests for autocorrelation in the errors discover the sufficient first order autocorrelation and the necessary lack of second order autocorrelation in the error terms, according to a review of the DPD test statistics for the entire sample estimation shown in Table E2. The exogeniety and identification measures all show correct specification, according to the results of the Sargan, Hansen, and Difference-in-Hansen tests for instrument validity. These statistical tests indicate that the instruments and model are both reliable.
The results of the four estimation methods are all consistently positive for the effects of Internet use on per capita export revenues on the entire sample, as shown in Table 3. However, the majority of sub-sample estimates are typically non-significant whereas the whole sample estimates are typically significant. While the OLS estimates are comparatively smaller and unimportant, the FMM, FE, and DPD estimates have equal absolute magnitudes and importance. According to the elasticity derived using the DPD estimates for the entire sample, a 10% increase in Internet use is expected to result in an average rise in per capita export income of 2.2%. These findings support the dissertation's claim that growing Internet usage increases export earnings, however the magnitude of the effects varies depending on the economic development of a nation.
The complete regression findings for the entire sample are shown in Table E1 for all four estimators. As we can see, the coefficient on the lagged realisation of per capita exports is considerable and positive, supporting the notion that dynamics are at play. Aid has a negligible, tiny, and minimal elasticity. The life expectancy coefficient has an extremely small, negligible, and negative sign. Theoretically, there is no justification for anticipating a substantial elasticity on this control. As might be predicted, both the fixed capital formation and secondary education coefficients are positive and substantial. The coefficient on institutional quality is interestingly negative and substantial once more, adding support to the claim made in the preceding section about the expense of building institutions.
To separate the effects of Internet use from any specific time period, as in the GDP model
I estimate the model with three extra control variables: time, income level, and a term that acts as an interaction between time and income level. These findings are displayed in Table E1's column (5). Given that the Internet use coefficient remains substantial and positive, it is obvious that the model is resistant to the inclusion of these controls. The claim that there is a causal link between Internet use and per capita Exports is strengthened by this estimate.
In this dissertation, I propose that, depending on the level of development, the impact of Internet use on per capita export revenues varies. As a result, I estimate the equation on three subsamples of high, middle, and low income nations using all four estimators. In Table 3, the results are displayed in panels (b) through (d). High-income country estimates are shown in panel (b), middle-income country estimates are shown in panel (c), and low-income country estimates are shown in panel (d).
Returning to Table 3, we can see that the DPD estimates for how Internet use affects exports varies significantly depending on the income class. While the estimate in low-income nations is negative and virtually equal to zero, it is marginally greater in high-income countries than in middle-income countries. The test statistics do not offer the same level of confidence with the complete and middle-income samples, therefore the DPD estimates for both the high and low income sub-samples are rather shaky. To highlight the findings on the full and middle income samples, which are of key importance, the results from the OLS, FMM, and FE estimators provide sufficient context.
The most remarkable finding is that a 10% increase in Internet usage is linked to a (statistically significant) 2.5% rise in per capita export earnings in middle-income nations. The premise of this dissertation, that Internet use will have different effects on export performance of countries at different levels of development, is supported by these estimates. Figure 5 illustrates how increased Internet usage has a considerable impact on export performance in middle-income nations, but has little to no effect on exports from low- or high-income countries.
These dissertation's findings are in good agreement with earlier studies in the literature. Freund and Weinhold (2004) discovered a correlation between an increase in Internet hosts of 10% and an increase in total export revenue growth of 0.2%. Although direct comparisons between the estimations are challenging, both point to a direct positive relationship between rising Internet usage and exports. In a previous investigation, Freund and Weinhold (2002) discovered that a 10% increase in Internet usage abroad is connected to a 1.7% rise in service exports to the US. Once more, despite these estimates.
Cannot be directly compared, the results discovered in this dissertation complement and extend Freund and Weinhold’s earlier studies.
The findings in this dissertation build on earlier investigations of exports and Internet usage. For instance, a 2006 study by Clarke and Wallsten found that developing nations with higher Internet access penetration export more to developed nations, and a 2008 study by Clarke discovered a strong correlation between Internet access and exports at the firm level in Eastern Europe and Central Asia. While earlier works gave examples of specific instances when using the Internet had a favourable impact on exports, this dissertation analyses the topic more generally and draws similar results.
In conclusion, this dissertation shows evidence that, under equal circumstances, Internet use favourably increases export profits, although the effects vary depending on the economic development of the countries under consideration. According to income class, countries that have advanced from the low level base to the middle level of a Maslow-like developmental hierarchy exhibit convincing evidence of the beneficial effects of Internet use on export performance.
4.3 Market Capitalization Per Capita (lmktcappc)
The estimated effects of Internet usage on the size of domestic equities markets (market capitalization) are shown in this subsection in US dollars. The scatter plot showing the correlation between market capitalisation and Internet use for the entire sample in 2007 is shown in Figure 6. For the entire sample and each subsample using all four estimators, Table 4 shows the elasticities of log per capita export revenue in US dollars on log per capita GDP. Appendix F contains the complete regression results as well as the regression test statistics.
We find consistent positive and substantial elasticities for per capita Internet use on per capita market capitalization starting with the full sample estimates using all four estimation methodologies presented in Table 4. According to the DPD estimate based on the entire sample, a 10% increase in Internet usage is correlated with an average increase in per capita market value of 24%. The statistically insignificant estimate for low-income nations is shown as zero in Figure 7, which compares the DPD estimates for the three subsamples. This finding is consistent with the dissertation's premise that, for nations with middle- and high-income levels, Internet use has a significant impact on domestic equities markets' size as shown by market capitalization per capita.
The full regression results from the model estimation on the entire sample using all four estimators are shown in Table F1. No matter the estimation technique, the coefficient on the lagged realisation of per capita market capitalization is large and positive, indicating that dynamics are present as anticipated. Additionally, just like in the previous two models, the elasticity of Internet use is constantly positive and significant. The only variation is in magnitude. The positive (albeit insignificant) coefficient on aid may indicate that increased aid may contribute to the growth and expansion of local capital markets.
Life expectancy's elasticity as a health proxy is negligible and inconsequential. While secondary education is insignificant and typically positive, but weakly negative (and insignificant) in the DPD estimate, the estimates of the effects of capital production are positive but insignificant, presumably indicating reverse causality. The reduced sample size used in this study may have contributed to this conclusion. In estimators that do not account for endogeneity and dynamics, institutional quality effects are positive but not statistically significant. The DPD assessment reveals a negligibly small, unimportant effect.
By evaluating the model with three extra control variables—time, income level, and an interaction term for time and income level combined—I isolate the effects of Internet use from any temporal or income class effects, just like in the GDP and export models. These findings are displayed in Table F1's column (5). We can observe that the Internet use coefficient is still substantial and positive, indicating that the model is resilient to the inclusion of these extra covariates. This lends more evidence to my claim that there is a link between per-person market capitalization and Internet use.
I analyse the idea that the impact of Internet use on income varies depending on the level of development, much like in each of the models examined in this dissertation. I re-estimate the equations on three subsamples as a result. The output of these estimations is displayed in Table 4's panels (b) through (d). High-income country estimates are shown in panel (b), middle-income country estimates are shown in panel (c), and low-income country estimates are shown in panel (d).
Tables F3–F8 show the results for the three subsamples. All income classes and estimators show positive significant effects of increased Internet use on market capitalization, with the exception of the low-income country sample where the elasticity from the DPD estimator is statistically insignificant. This is compelling evidence of the correlation between domestic market size and Internet use, which varies across income strata.
It is evident that, for high and middle income nations, Internet use consistently has a considerable positive impact on stock market size, whereas these effects are negligible across estimators for low-income countries. Due to the very small sample sizes, the test statistics on the DPD estimates for the high and middle income samples do not consistently show that the model specification is correct. However, these estimates serve as comparison points together with the OLS, FMM, and FE estimations.
The literature on the factors that influence the growth of equities markets is not very well-researched. The majority of currently conducted research focuses on either the connection between capital markets and economic growth or the role that capital markets play in the spread of technology. According to the reported findings, financial markets and ICT in general (and the Internet in particular) are interconnected and both contribute to economic growth. However, this dissertation is the only one that has looked at the connection between rising Internet usage and the size of domestic equities markets as assessed by per capita market capitalization as of the time of publication.
The findings in this subsection support the claim that Internet use has a favourable impact on domestic market capitalisation in nations of all income levels, but that these effects are only truly noticeable in middle- and high-income countries. Chapter 6 contains the outcomes' political ramifications.
4.4 UN HDI (hdi)
The Human Development Index (HDI), a combined measure of education, literacy, and income published by the United Nations Development Programme, is used in this subsection to quantify the effects of Internet use on overall welfare. The scatterplot association between Internet usage and the HDI is represented graphically in Figure 8. The elasticities of Internet use on the HDI for the entire sample and for each subsample using all four estimators are shown in Table 5. Appendix G contains all of the regression test statistics as well as the complete regression results.
The fit is not excellent, according to the test statistics for the DPD estimations on the entire sample. Although the Sargan and Hansen tests both reject the legitimate overidentification limits, the Arellano-Bond test statistics find evidence of first order autocorrelation and the absence of send order autocorrelation, as required. The exogeniety of the instruments divides the Difference-in-Hansen tests. Nevertheless, all four estimators have marginally significant and positive coefficients on Internet use, which lends some credence to the thesis of this dissertation—that rising Internet use will lead to higher HDI measures of welfare.
The sample sizes are significantly reduced when the UN HDI is used as a response variable because it is not calculated annually. Cubic spline interpolation and linear extrapolation were used to produce estimates for the missing observations in order to increase the number of observations available for the empirical study.
Since there is insufficient data to adequately measure these covariates in terms of per capita, per capita control measures are not employed for this model, as explained in Chapter 3. Furthermore, because the index includes measures for these determinants, the proxies for health and educational attainment (life expectancy and time spent in secondary school) are not included as controls. Since the level measures utilised are not per capita measures, an additional control for labour force size is included.
Starting with the estimates of the four estimation methods for the impacts of Internet use on the HDI for the entire sample, Table 5, we observe consistent positive significant effects. According to the elasticity estimates made by the DPD estimator for the entire sample, a 10% increase in Internet use should, on average, result in a modest but considerable increase in the HDI. A visual comparison of the DPD estimations for each subsample is shown in Figure 9. These findings offer some support for the dissertation's central claim that rising Internet use improves overall welfare as indicated by the HDI.
The complete regression findings for the entire sample are shown in Table G1. Aid has a negative and substantial co-efficient. This offers more evidence in favour of the claim that help does not increase general welfare and presents a promising subject for future research. Both the coefficient on fixed capital formation and the estimate of labour force size are positive and substantial, as would be expected. We can observe from this model that the impacts of institutional quality on the HDI are often estimated to be positive and significant. This is an intriguing counterweight to the findings presented above about the variable influences of institutional quality on the results of economic progress.
By estimating the model with three extra control variables—time, income level, and an interaction term for time and income level combined—I once more isolate the effects of Internet use from any temporal or income class effects. These findings are displayed in Table G1's column (5). We can observe that the Internet use coefficient is still substantial and positive, indicating that the model is resilient to the inclusion of these extra covariates. This lends more evidence to my claim that there is a causal connection between Internet use and overall welfare as determined by the HDI.
The GDP model has shown that the income level coefficient is significant and negative (increasing income levels corresponds to decreasing income class.) This suggests that the HDI is affected by income class.
By estimating the equations on the three subsamples of high, middle, and low-income nations using all four estimate methods, I continue to investigate the possibility that the impact of Internet use on income varies depending on the level of development. In Table 2, panels (b) through (d) show the estimates for high-income, middle-income, and low-income nations, respectively. Panel (b) in Table 2 depicts the estimates for high-income, middle-income, and low-income countries.
Tables G3–G8 illustrate the findings of the estimates of Internet use on the HDI for each of the subsamples, and they consistently demonstrate minor, favourable, and significant effects. The estimated effects are almost the same for low- and middle-income countries, but they are greater for high-income countries. These findings provide more proof for the dissertation's conclusion—that using the Internet significantly boosts economic growth.
Since there are no published studies on the welfare effects of ICT and the Internet in poor nations, there are no empirical studies to compare the findings of this dissertation with. Nevertheless, other researches contend that an increase in Internet users has a positive impact on welfare.
Prahalad and Hammond (2002) and Crandall and Jackson (2001) both discuss how businesses may gain from the development of ICT infrastructure in developing nations. Technology can boost welfare even in underdeveloped locations, according to a Jensen study published in 2007 that examined the influence of cell phones on rural Indian fisherman. According to Thompson and Garbacz (2007), developing information networks are beneficial for underdeveloped nations. Although Aker and Mbiti's (2010) argument lacks empirical backing, they contend that expanding mobile phone connectivity in Africa's low-income nations has a significant potential to improve welfare.
My initial attempt to research the impact of Internet use on economic welfare is presented in this subsection. Although it is challenging to estimate models using aggregate measures on an index that is made up of aggregates, this investigation does succeed in supplying some additional evidence in support of this dissertation's hypothesis that increased Internet use has a positive impact on economic development.
4.5 Concluding Observations
The premise of this dissertation that increased Internet use has favourable effects on economic development as assessed by GDP, exports, market size, and the UN HDI, is supported by evidence across all of the models and estimation findings. When estimates are run on stratified groups of countries—high, middle, and low-income countries as separate subsamples—it is evident that the effects of more Internet users vary in absolute magnitude and significance across income classes. Results across the full sample of countries frequently show significant positive effects.
My initial thought was that more Internet users wouldn't have much of an impact on developing nations. A surprise outcome was produced by the model's results when it was applied to per capita GDP. The elasticity on the low-income sample was positive and slightly bigger, however the elasticity on the complete sample estimate closely matched the estimate for the middle-income sample as was expected. This unexpected finding implies that even in the least developed nations, increased Internet use affects per capita income.
The model's projections of per-capita export earnings confirmed my suspicions that middle-income countries would benefit most from increased Internet usage. Once more, the elasticity calculated for the entire sample was fairly close to the calculation for the middle-income sample. The elasticity of Internet use in low- and high-income countries was negligible, but it was considerable and positive in middle-income nations.
My only prediction for the effects of internet use on market size was that lower income countries would likely show no effect because newly created markets might not be integrated into the global financial system. However, there has been little research on the factors that determine the size of markets in developing countries. This was supported by the results, which showed that the elasticity in middle- and high-income countries was significantly positive whereas it was negligible in low-income countries. The estimate for middle-income nations was higher than the elasticity of Internet use in high-income countries. This may imply that as economies develop and wealth levels rise, more people will have access to the Internet and domestic financial markets, opening up new investment options.
Unexpected model estimations resulted from the investigation of Internet usage on the UN HDI. Although I have less faith in the model fit in this instance, the data suggest that there might be intriguing effects that call for more investigation. Similar to the other models, both the full sample's and the middle-income sample's estimates of elasticity were positive and statistically significant. My broad expectations for this inquiry were met by this.
Surprisingly, the low-income sample's estimated elasticity was positive and substantial. It is reassuring in a way that this conclusion is supported by the findings in the model outlined above regarding the effects of Internet use on GDP since the HDI is an aggregate measure that also accounts for per capita GDP. Additionally, given that the UN also takes life expectancy and literacy rates into account when calculating the HDI, possibly increased Internet usage will help low-income nations have better access to information about health and education. Future research in this area should be exciting.
An extensive framework for comprehending the effects of Internet use on economic development outcomes is provided by the results of these four models of Internet use on four different economic development outcomes, applied to the entire sample of countries and to each of the three subsamples stratified by income class.
It is challenging to deny the evidence showing the significant beneficial effects of more Internet users on all four economic outcomes for middle-income nations, as well as the evidence of the clear differences in the degree and significance of the effects depending on the income class. These findings support the research hypothesis of this dissertation, which holds that Internet use is a significant positive predictor of economic development. However, the magnitude of these effects varies by country's position along a Maslow-like hierarchy of economic development stratified by income level.
5 Policy Implications and Future Research
5.1 Dissertation Summary
The impact of Internet use on economic growth is examined in this dissertation. According to the hypothesis under consideration, growing Internet usage has a beneficial impact on a number of development outcomes, including per capita GDP, export income for products and services, the size of domestic equity markets as evaluated by the UN Human Development Index, and overall welfare. However, depending on the country's income level, Internet use has different effects on economies.
I take into account an economic hierarchy of demands (based on Maslow's famous work on human psychology) where Internet use will have a measurably different impact at each degree of development. It is unlikely that Internet would be accessible at the lowest level of development where the population struggles to survive because of extreme poverty, absent or failing public health institutions, on-going violent conflict, or generally unstable political systems. If it were, however, people could use it to better their circumstances.
Countries that have made some progress toward economic growth (and into the middle-income bracket) are more likely to have the infrastructure and operating institutions required to sustain extensive Internet use. In this group of nations, I anticipate finding the strongest effects of Internet usage on outcomes of economic development. In the most developed nations, the marginal impact of increased Internet usage might not have any discernible effects.
The World Bank World Development Indicators and World Governance Indicators (WGI), the International Telecommunications Union World Telecommunication ICT Indicators, and the United Nations Human Development Index were used to create a panel data set that was used to conduct this empirical investigation. Over the course of an eleven-year period, from 1996 to 2007, these data were gathered and merged to generate a panel of economic statistics on 202 countries. Four econometric methodologies (Ordinary Least Squares, Finite Mixture Modeling, and Dynamic Panel Data) are used to analyse each sample's effectiveness in utilising the Internet, and the findings are compared to examine the effects of increased Internet use at various developmental stages.
For each of the four economic outcomes—per capita GDP, per capita export revenue, per capita equity market capitalization, and UN HDI—the econometric study employs a production function framework with Inter-net use as an additional input. The main controls for this framework are sourced from the academic literature and include life expectancy at birth, secondary school attendance duration in years, per capita fixed capital formation, per capita net foreign aid, and a measure of institutional quality for the WGI. The variables chosen as controls are those that are frequently employed as proxies for the essential elements of economic development in growth and development empirics: fixed capital formation, foreign aid, education, health, and institutions. Since these models are estimating per capita results, there is no labour control.
For the four outcomes, four different econometric estimate methods are applied to each of the four samples. The baseline estimating method is Ordinary Least Squares (OLS), but in the presence of endogeneity and dynamics, this estimator will yield skewed and contradictory findings. As a result, estimators for Finite Mixture Modelling (FMM) and Dynamic Panel Data (DPD) are also utilised.
The FMM estimate approach may model the various elasticities between income classes even though it does not directly address endogeneity and dynamics issues. It models a different distribution for each class. When dynamics and endogeneity are present, the DPD General Method of Moments (GMM) estimators are frequently employed in the literature to estimate models using cross-country data. There are three advantages to using these estimators. The first step is to incorporate the long-lasting effects of Internet use into a dynamic framework. Second, there are significant endogeneity issues with regard to Internet use, financial help, and the metrics of development outcomes being taken into account. Two-stage least squares methods cannot be used without clearly defined and understood instrumental variables. Third, it's critical to account for unobserved national heterogeneity that can be linked to the outcomes of the study.
The premise of this dissertation that increased Internet use has favourable effects on economic development as assessed by GDP, exports, market size, and the UN HDI welfare index, is supported by evidence across all of the models and estimation findings. When estimates are run on stratified groups of countries - high, middle, and low income countries as separate subsamples - results across the full sample of countries frequently show significant positive effects, but the effects of additional Internet users vary in absolute magnitude and significance across income classes.
A startling finding emerged from the calculation of the impact of Internet use on per capita GDP. The elasticity on the low-income sample was positive and slightly bigger, however the elasticity on the complete sample estimate closely matched the estimate for the middle-income sample as was expected. This implies that even in the least developed nations, increased Internet use has a beneficial effect on per capita income. Future studies should look into the process by which this occurs.
The model's projections of per-capita export earnings confirmed my suspicions that middle-income countries would benefit most from increased Internet usage. Once more, the elasticity calculated for the entire sample was fairly close to the calculation for the middle-income sample. The elasticities for low- and high-income countries were negligible, while for middle-income countries, the elasticity was both positive and considerable.
The effects of Internet use on market size indicate that there is no effect in lower-income countries. However, the findings show that Internet use is far more elastic in middle- and high-income countries. The estimate for middle-income nations was higher than the elasticity for high-income countries. This may imply that as economies develop and countries move toward higher income levels, increased Internet connectivity opens up more domestic capital markets and creates more chances for investment.
Unexpected model estimations resulted from the investigation of Internet usage on the UN HDI. Despite the fact that the model fit is not as good as it was for the previous three models, the findings suggest that there may be intriguing effects that call for more investigation. The elasticity estimate for the entire sample was similar to that of the middle-income sample, and both were positive and significant, as observed in the three other estimations. Surprisingly, the low-income sample's estimated elasticity was positive and substantial. The findings from the estimation of Internet use on GDP confirm this result because the HDI is an aggregate measure that also accounts for per capita GDP.
The results of this dissertation confirm findings from earlier studies that looked into the effect of Internet use on comparable economic outcomes. The positive and significant effects of Internet use on per capita GDP are particularly intriguing because they are consistent with findings from Czernich et al. (2009) that demonstrate a comparable impact of broadband Internet on GDP growth in OECD nations. Additionally, the findings in Freund and Winhold’s studies on the effects of broadband internet on exports are supported by the results of this dissertation, which demonstrate the effect of the Internet on export revenues.
It is significant to remember that Internet usage estimates are resilient to different specifications and estimating techniques. Regardless of the specific estimating method, the estimates of the effect of Internet use are constant for each sample and specification.
5.2 Policy Implications
The policy recommendations drawn from these results vary according to the country's income level since Internet use has different effects on economic development based on the country's income level and the particular outcome.
Increased Internet usage generally improves per capita GDP and overall welfare as assessed by the HDI in low-income nations. This argues that in order to improve job development, health care services, and programmes to raise literacy, policy initiatives should concentrate on providing more Internet connectivity, probably based on new or current cellular phone infrastructure. Importantly, given that I have demonstrated the benefits of ICT and Internet use for development, LDCs can benefit from the chance to use aid and FDI to quickly deploy new technological infrastructure, such as wireless telephone and Internet, in order to advance from outmoded technologies.
Institutions are required to provide and manage these services, but as we have seen, in certain nations, establishing effective institutions may come at a preliminary expense. Policymakers might encourage foreign aid to support services like health and education while directing local and international capital deployment toward building out the Internet infrastructure, for instance. Additional Internet usage had no appreciable effects on exports or market size in low-income nations. As a result, until the economy has reached higher income levels, policymakers should refrain from Internet-based development programmes aimed at developing exports or capital markets.
The most dramatic increases in economic and societal well-being outcomes are seen in middle-income nations. The findings of this dissertation imply that the four indicators of economic development in these nations are positively and significantly impacted by rising Internet usage. As a result, policymakers in nations that have started to leave the lowest level of economic development may want to promote the widespread installation of Internet connection. In order for the service sector to offer the populace mobile banking, insurance, and other Internet-enabled technical solutions, policymakers must provide the appropriate institutional and legal backing.
5.3 Future Research
The goal of this dissertation was to establish a comprehensive framework for future study on the impact of Internet use on economic growth. The findings provide convincing proof of the favourable effects, which vary depending on the amount of development attained. However, this dissertation has omitted a number of intriguing and crucial research issues and areas that are worthwhile exploring, as in any developing field of study.
Although I have uncovered evidence that internet use has an impact on a number of development outcomes, I do not look into the factors that influence internet use. What promotes or restricts Internet usage? What effects do the political, institutional, and economic circumstances in a given country have on not only the likelihood of access to the Internet but also its actual use? The process of ex- act transfer is another intriguing query. What are the actual mechanisms by which using the Internet affects these results?
This work has uncovered new fields of study that will expand on the findings. The additional information that becomes available on the years after 2007 will be useful for future analyses into the effects of Internet use in developing countries. Spatial econometric techniques, involving contiguity measures based on data and voice transmission networks, may offer another way to look into how Internet use affects economic development because the Internet is delivered to developing economies through fibre optic backbones that cross specific borders.
The findings of the effects of Internet use on per capita GDP reveal an intriguing growth story. This relationship might be examined in a later work using a more conventional macroeconomic growth modelling approach.
Remote areas are now receiving health care services because to technology. Telemedicine enables untrained doctors in rural Kenya to communicate with specialised experts who are hundreds of kilometres away (Independent News and Media 2010). These findings may imply that in developing nations, Internet access may have a more direct impact on particular health and educational outcomes. According to the UN HDI, there may be a direct relationship between Internet use and outcomes in terms of education and health.
Technology adoption in emerging nations will inevitably take place in phases rather than all at once across the entire nation. Randomized controlled trials, or random evaluations, may be an effective method for examining the effects of Internet use in field investigations. The same ethical issues that plague health or education treatments do not apply to providing access to the Internet as a treatment. In addition, it will be exciting to investigate these models using firm-level data as data from developing nations becomes accessible. Data at the country level may hold the key to unlocking some of the mysteries surrounding transmission techniques or the likelihood of Internet use.
The most fascinating arena to investigate how technology affects developing economies is quickly developing as the discipline of economic development study. My modest attempt to advance the discipline through this dissertation is to present thorough empirical proof of the obvious economic advantages of growing Internet usage in developing nations.