Policy Research Working Paper 9946 Infrastructure Quality and FDI Inflows Evidence from the Arrival of High-Speed Internet in Africa Justice Tei Mensah Nouhoum Traore International Finance Corporation February 2022 Policy Research Working Paper 9946 Abstract Does ambient infrastructural quality affect foreign direct staggered arrival of submarine fiber-optic internet cables investment (FDI) in developing countries? This paper and the subsequent rollout of terrestrial fiber cable networks investigates how the arrival of high-speed internet in Africa across locations on the continent. Findings from the paper triggered FDI into the banking and technology services show that access to high-speed internet induces FDI into sectors. It also explores the role of complementary infra- the banking and technology sectors. However, the impact structure, such as access to reliable electricity, in amplifying pertains mainly to countries with access to reliable electric- the impact of internet connectivity on investment. The ity, thus highlighting the role of complementarities in the identification strategy exploits plausibly exogenous vari- impact of infrastructure. ations in access to high-speed internet induced by the This paper is a product of the Development Impact Department, International Finance Corporation. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at jmensah2@ifc.org and ntraore3@ifc.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Infrastructure Quality and FDI Inflows: Evidence from the Arrival of High-Speed Internet in Africa* Justice Tei Mensah† Nouhoum Traore‡ JEL: O12, O16, F21, F23, G21, N77 Key words: High-speed Internet, FDI, Banking, Technology, Africa * We thank seminar participants at the IFC and World Bank for their helpful comments. Comments and support from Moussa Blimpo, Mark Dutz, Albert Zeufack, Valentina Saltane, Camilo Mondragon-Velez, and Issa Faye are deeply appreciated. This paper was supported by the research program at IFC’s Sector Economics and Develop- ment Impact Department. The paper is also background paper for the World Bank Flagship Report on "Digital Africa for Inclusive Growth". All errors and views expressed in this paper are those of the authors. They do not necessarily represent the views of the International Finance Corporation/World Bank Group. † International Finance Corporation, The World Bank Group. Email:jmensah2@ifc.org ‡ International Finance Corporation, The World Bank Group. Email:ntraore@ifc.org 1 Introduction Foreign direct investment (FDI) is an important driver of economic growth as it facilitates tech- nology transfer and job creation in host economies (Smarzynska Javorcik, 2004; Toews and Véz- ina, 2021).1 As a result, countries embark on massive investment promotion drives aimed at reducing the transaction costs of investors by providing them with easy access to information on investment opportunities in the country, including understanding and overcoming the legal and regulatory hurdles involved in setting up businesses in the country (Harding and Javorcik, 2011). In addition, some countries undertake structural reforms to liberalize their economies, offering, among other things, tax holidays for investors to make their economies attractive in- vestment locations (Reinikka and Svensson, 1999; Harding and Javorcik, 2011) for investors who are attracted to countries with low costs of doing business (Amiti and Javorcik, 2008). In spite of these incentives, some countries still struggle to attract FDI due to low-quality infrastructure and its implications for the cost of doing business (Reinikka and Svensson, 1999). In this paper, we show causal evidence of the impact of infrastructure quality on invest- ment flows to developing countries by leveraging the arrival of high-speed Internet to Africa as a unique natural experiment. Between 2000 and 2012, Africa experienced a significant im- provement in its Internet landscape with the arrival of several submarine fiber-optic Internet cables linking the continent to the global Internet network via Europe. Prior to this, Internet transmissions to the continent were mainly through satellite, which was characterized by low bandwidth speed and high cost of service. The submarine cables brought faster Internet speeds at relatively low cost, thereby increasing uptake of the Internet for commercial and household uses. The arrival of fast Internet via submarine cables was particularly beneficial to the banking and finance sector, as it enabled the adoption of financial technologies (Fintech) such as Inter- net banking and real-time gross settlement system (RTGS), leading to an increase in economic activities in the sector (D’Andrea and Limodio, 2019). The technology sector also benefited, as the availability of high-speed Internet created supply and demand for technology services.2 We interpret high-speed Internet connection as a positive technological shock that led to an expansion in the finance and technology sectors with possible implications on demand for in- vestments in these sectors. Therefore, our goal is to show how the arrival of high-speed Internet triggered FDI in the technology and finance sectors in Africa. In addition, we explore the role of complementary infrastructure, such as access to reliable electricity, in amplifying the impact of Internet con- 1 There is also a large body of literature showing mixed evidence on the impact of FDI on destination markets. See for instance, Edwards and Jenkins (2015), Abebe et al. (2018) and Crescenzi and Limodio (2021) 2 See for example: https://www.bbc.com/news/technology-46055595 2 nectivity. To this end, we assemble unique project-level data on FDI in the finance and tech- nology sectors together with data on Internet infrastructure to estimate the spillover effects of fast Internet connectivity on FDI flows. The spatial and time granularity of our datasets allows us to undertake the analysis at both national and subnational levels. Our identification strategy is a difference-in-difference (DiD) design that relies on the plausibly exogenous variation in ac- cess to high-speed Internet induced by the: (i) staggered arrival of submarine Internet cables across countries for the cross-country analysis and (ii) staggered arrival of submarine Internet cables and the subsequent rollout of terrestrial fiber backbone network within countries for the subnational analysis. Two main findings emerge from the analysis. First, the arrival of high-speed Internet played a key role in stimulating FDI in the banking and technology sectors in Africa. The effect is statis- tically and economically significant in both the intensive and extensive margins. On the exten- sive margins, high-speed Internet connectivity via the submarine cables is associated with an 18 and 12 percentage point (pp) increase in the probability of FDI in financial and technology services sectors, respectively. On the intensive margin, the number of FDI projects in financial services increased by 28.6% upon the arrival of the submarine Internet cables. The value (size) of FDI in technology and financial services increased as well. To isolate the impact of Internet connectivity from other underlying drivers of FDI, we show that our estimated impacts only hold for the technology and finance sectors. FDI in other sectors was not affected by connectiv- ity to the submarine Internet cables. Further, we show that our results hold at the subnational level, with implementation of FDI projects in these sectors within countries responding to the availability of fast Internet connectivity across their respective locations. Second, our findings show that the effects of high-speed Internet connectivity on FDI per- tain largely to countries with reliable electricity supply. In other words, while high-speed con- nectivity is associated with increased investment in the finance and technology sectors, the effect is largely concentrated in countries with reliable supply of electricity. This is largely due to the fact that access to electricity is essential in powering devices such as computers and mo- bile phones that use the Internet. Although reliance on generators by firms during periods of outages can provide the needed electricity to power these appliances, the associated high cost constrains technology adoption and utilization of the Internet for productive uses by commer- cial entities. Thus, the findings from this paper underscore the important role of complemen- tary infrastructure, such as access to reliable electricity, in amplifying the economic impact of Internet services in developing countries. This paper contributes to two strands of the literature. First, it offers a novel contribution to the literature on the drivers of FDI in developing countries. A growing number of studies 3 have focused on the role of deliberate government policies, such as investment promotion ini- tiatives (Harding and Javorcik, 2011, 2012), and business regulations (Arita et al., 2013), in at- tracting FDI. Harding and Javorcik (2011), for instance, show that investment promotions are effective in attracting FDI, particularly in countries with cumbersome business regulations and information asymmetries.3 . Toews and Vézina (2021) also present evidence that natural re- source booms trigger FDI in non-extractive sectors. Evidence from other studies also shows that non-economic factors, such as cultural ties between countries, influence FDI flows across countries. Using a quasi-random allocation of refugees in the United States induced by the 1980 Refugee Act, Mayda et al. (2019) show evidence of greater outward FDI flow to the origin countries of refugees from US states with higher concentration of refugees. In spite of this com- pelling evidence, little is known about the premium that foreign investors place on the quality of infrastructural services in destination markets when they are assessing investment opportu- nities. Findings from this study therefore provide an important contribution to the literature by demonstrating the infrastructure premium in attracting FDI. Second, this paper contributes to the emerging strand of literature on the economic im- pacts of modern infrastructure, such as Internet and electricity, in developing and emerging economies. The arrival of high-speed Internet in Africa has been a topic of growing inter- est to many researchers (Hjort and Poulsen, 2019; D’Andrea and Limodio, 2019; Ouedraogo et al., 2020). By leveraging the gradual arrival of submarine Internet cables in Africa, Hjort and Poulsen (2019) show that access to high-speed Internet increases employment for skilled and unskilled labor. Firms also benefited through productivity improvements. D’Andrea and Limodio (2019) show that access to high-speed Internet led to a boom in the financial sector as it enabled technology adoption among African banks. Consistent with D’Andrea and Limodio (2019) , our findings of a positive impact of high-speed Internet on FDI in the banking sector provide suggestive evidence on the channels through which access to high-speed Internet stim- ulates financial development. Findings on the complementary role of access to reliable electric- ity are also consistent with Andersen and Dalgaard (2013) and Mensah (2018), who show that unreliable electricity provision is a challenge to economic development in Africa. The rest of the paper proceeds as follows. In Section 2, we present the conceptual frame- work linking Internet to investment in finance and technology sectors. Section 3 describes the data used in the analysis while the identification strategy is outlined in Section 4. We present and discuss the findings in Section 5. Section 6 concludes the paper with a summary of main findings. 3 That may be induced by language and cultural distance 4 2 Conceptual Framework To understand the spillover effects of access to high-speed Internet on investments, in this sec- tion we describe how Internet connectivity increases technology adoption in the banking and technology sectors, and consequently its effects on the development of these sectors. 2.1 High-Speed Internet and Banking Financial technologies, popularly referred to as "Fintech" have changed the banking and fi- nance landscape around the world by reducing costs and improving efficiencies of financial intermediation. The banking sector has historically been organized around manual process- ing, with customers going to banking halls to make deposits and withdrawals. Interbank pay- ments were settled via a netting system: a system of deferred payments, usually until the end of a business day, before all payments were tallied and funds exchanged among participating institutions (Bech et al., 2007). This affected the pace and volume of interbank transactions. The advent of modern technologies such as the Internet enabled the adoption of Fintech, such as Internet banking and the real-time gross settlement system (RTGS). RTGS enables a one-to-one exchange of money and securities between banks (financial in- stitutions) in real time without needing to bundle or net them with other trades (D’Andrea and Limodio, 2019). The real time nature of interbank settlement under the RTGS means that huge volumes of transactions can be made in an efficient manner. Thus, RTGS adoption is associ- ated with a significant reduction in the cost of interbank settlements and improves efficient integration of the financial sector (Townsend, 1978; Guerrieri and Lorenzoni, 2009; D’Andrea and Limodio, 2019). This has positive implications for the performance of the banking sectors, as a reduction in the cost of interbank lending leads to an increase in interbank borrowing as well as an expansion in credit to the private sector (D’Andrea and Limodio, 2019). Thus, RTGS plays an important role in the development of the financial sector. The adoption of RTGS is rel- atively recent. As of 1986, only four countries (Denmark, Netherlands, Sweden, and the United States) had adopted the RTGS, increasing to 93 (including 11 in Africa) by 2006 (Bech et al., 2007). Today, almost all countries have adopted some form of RTGS. What influenced the adoption of RTGS across countries? Aside from regulatory and policy reforms, access to high-speed Internet is a key determinant of RTGS uptake across the world. As a real-time exchange, RTGS requires fast and reliable interconnection among parties. D’Andrea and Limodio (2019), for instance, show that the arrival of high-speed Internet in Africa led to an increase in the adoption of RTGS. As a result, banks significantly reduced their liquidity hoard- ing, while interbank transactions and lending to customers increased. Essentially, the findings 5 from D’Andrea and Limodio (2019) suggest that access to high-speed Internet in Africa led to a boom in the financial sector. Automation of banking sector operations is another channel through which Internet ac- cess influences the banking sector. Internet banking and other seamless banking operations (e.g., ATMs, debit/credit cards) have enabled banks to overcome the constraints of traditional "brick-and-mortar" banking to provide fast and cost-effective services to clients. This reduces overhead costs, allowing banks to stay competitive. In this paper, we hypothesize that the positive impacts of Fintech, such as RTGS and Internet banking uptake by banks, are channels through which access to high-speed Internet influences investment in the banking (financial) sector. For instance, a booming financial sector via the adoption of RTGS increases competition and creates new opportunities for investment in the sector. 2.2 High-speed Internet and FDI in Technology Services Further, we argue that the demand and supply of digital services associated with Internet use are the key channels through which high-speed Internet connectivity influences investment in the technology services sector. From a demand perspective, increasing uptake of broadband Internet, particularly among Africa’s growing middle class, has fueled demand for digital services, such as online entertain- ment, social networking, online banking, e-payments, telemedicine, and e-commerce. De- mand for these services is particularly high when they are made in Africa or have content tai- lored for the African market, creating opportunities for digital firms that provide these ser- vices. For instance, content aggregation companies like iROKOtv in Nigeria have been providing streaming services to viewers in Africa and the diaspora over the past decade focused solely on African movies. Netflix has recently entered the space by establishing its Africa office in Nige- ria with the aim of tapping into Africa’s movie industry. Social media companies like Twitter and Facebook have also set up offices in Africa, while ride-sharing companies like Uber and Bolt are present in many African countries. Thus, as Africa’s young and booming population, rapid urbanization, and expansion in Internet connectivity contribute to increasing the conti- nent’s share of population with an online presence, there are huge economic prospects for the development of a digital economy on the continent. These trends are expected to catalyze in- vestments into the technology services sector to meet the growing demand for digital services. Secondly, from the supply-side, access to fast Internet has created a "launch pad" for digital entrepreneurship in Africa (Manyika et al., 2013). With the rising demand for digital services and the low concentration of tech companies on the continent, a new generation of tech en- 6 trepreneurs is rising in Africa. In many countries across the continent, young and vibrant en- trepreneurs are establishing start-ups in such areas as mobile payments, online retail, e- trans- port and logistics, and e-learning. For instance, in Nigeria alone, Konga and Jumia have be- come major online retailers, while Kobo360—an e-logistics company—is expanding and pro- viding freight transport services to clients in the country.4 In Kenya, Twiga Foods operates a digital platform that connects manufacturing firms and consumers with informal market ven- dors, thereby reducing the inefficiencies in the agricultural value chain. Further, access to fast Internet creates opportunities for business process outsourcing (BPO) firms. BPOs cover a range of activities, from data collection, entry, and processing to respond- ing to calls from customers. With relatively low labor costs and a rising share of the educated labor force, access to quality infrastructure, such as high-speed Internet, provides a competitive advantage for African firms to attract outsourcing contracts from multinationals in developed economies (Mann and Friederici, 2015). Notable examples of successful BPOs include the Sama (formerly Samasource) and Busara Lab,5 both based in Nairobi, Kenya.6 Sama, a San Francisco- based company, has established its presence in Kenya, where it employs people to generate and process data for cutting-edge AI projects for Silicon Valley tech firms.7 As these tech start-ups (including incumbent firms) grow, they require significant invest- ments in order to survive and scale up. However, given the low levels of capital in many African countries, the importance of FDI in providing the needed capital to African technology firms cannot be overemphasized. In addition, big Silicon Valley firms like Google and Microsoft are investing directly in incubator hubs in Africa to train future tech entrepreneurs. Thus, aside from the direct impact of access to high-speed Internet on socioeconomic outcomes, the in- frastructure is also a catalyst for investment in the technology services sector. 3 Data Our main data is project-level data on FDI from fDiMarkets,8 a subsidiary of the Financial Times.fDiMarkets monitors cross-border investments around the globe and is a major source 4 Jobberman is a leading digital marketplace for employers and prospective employees by providing employment-oriented services for job applicants in Nigeria. 5 The Busara Lab is used by researchers around the world to conduct surveys, interactive games, and experi- ments. 6 Morocco and South Africa now export significant services in the BPO area. The sector generates more than US$1.5 billion of revenues and accounts for 54,000 jobs in South Africa (Manyika et al., 2013). 7 The company counts Google, Microsoft, Salesforce, and Yahoo among its clients. See: https://www.bbc.com/ news/technology-46055595 8 https://www.fdimarkets.com/ 7 of data used by the World Bank and UNCTAD9 to measure cross-border financial flows. It pro- vides detailed information relating to date of announcement, sector, subsector, whether the investment is greenfield (new) or brownfield (expansion),10 name of investors, source country and city (state), destination country and city, capital investment, and anticipated direct jobs created, among other things. Data on FDI from 2003 to 2018 are used in this study. For each country and year, we compute the total number of FDI projects and value of in vestment for each sector. We focus primarily on investment in financial services and tech- nology services. For a robustness test, we also complement our analysis with data on FDI in sectors excluding technology and financial services. This allows us to test our hypothesis that the spillover effects of access to fast Internet are primarily driven by investment in the finan- cial and technology sectors, which directly depend on the Internet for service delivery. Aside from investment at the country level, we also compile the FDI data at the subnational level by leveraging detailed data on the destination of the projects. Our final datasets are therefore a country-year panel between 2003 and 2018 and a subnational-year panel over the same period. In addition, we compile data on the staggered arrival of submarine Internet cables to the coasts of various African countries, as well as the connections of landlocked countries, us- ing data from www.africabandwidthmaps.com,11 www.submarinecablemap.com, and other country-specific sources. While connection to the submarine cable network signals the avail- ability of high- speed Internet in a country, the gradual deployment of terrestrial fiber backbone network within countries also created a second layer of access. Therefore, we exploit the grad- ual rollout of the terrestrial fiber backbone network, together with the staggered variation in connectivity to the submarine Internet cables, in our subnational analysis to identify access to high-speed Internet at the local level. Data on terrestrial fiber backbone network comes from www.africabandwidthmaps.com. Figure 1 shows trends in the number of countries connected to at least one submarine cable between 2003 and 2018. Finally, we rely on macroeconomic indicators such as real GDP growth and GDP per capita, from the World Development Indicators (WDI) database, and data on quality of electricity pro- vision proxied by the System Average Interruption Frequency Index (SAIFI) from the World Bank Doing Business database. Table 1 presents the summary statistics of the indicators used in the analysis. 9 The United Nations Conference on Trade and Development 10 About 95% of projects in the database are greenfield. 11 A digital consulting firm focused on publishing maps on Internet infrastructure in Africa 8 4 Empirical Strategy In the 1990s, satellite transmission was the main source of Africa’s Internet connection. It was characterized by high connection costs and slow speeds, which affected utilization of Internet services for productive uses. To address these issues, several countries partnered with telecom companies to undertake massive infrastructural investments by laying submarine fiber-optic cables from Europe to Africa. The cables began arriving on the African coastline between 1997 and 2002, but only a few countries had access to them. Also, the bandwidth capacity of these early submarine cables was still relatively low, limiting high-speed Internet access to a few loca- tions. The construction of the undersea cables slowed between 2002 and 2008, partly due to the global financial crisis and dotcom bubble (D’Andrea and Limodio, 2019). However, between 2009 and 2012, Africa witnessed a boom in the arrival of the fiber-optic submarine cables, with the number of African countries with at least one submarine cable connection increasing from about 20 in 2008 to 52 in 2012 (see Figure 1).12 Figure 2 shows the gradual arrival of the subma- rine cables between 2000 and 2018. Further, the landing points of the submarine cables were mostly cities along the coast. Hauling Internet to inland towns and cities required the construc- tion of terrestrial fiber backbone networks. The rollout of these networks gradually induced within-country variations in the timing of high-speed Internet connections. Landlocked coun- tries were connected to the submarine cables via the backbone networks of neighboring coastal countries. The way the submarine cables arrived and the subsequent rollout of the terrestrial fiber net- work induced two sources of variation: (i) Cross-country variations in access to high-speed In- ternet induced by the staggered arrival of the submarine cables and (ii) within-country variation resulting from the staggered rollout of the terrestrial fiber network. We explore these sources of variation in the analysis, using a difference-in-difference design to estimate the impact of ac- cess to fast Internet. Aside the spatial and time variations in connectivity, we treat the arrival of submarine ca- bles as a positive shock in Internet connectivity, as Internet via the submarine cables is much cheaper than satellite-based Internet. In Togo and Mauritania, for instance, connections to the submarine cables reduced the cost of international bandwidth for 1 Mb/s from $1,174 to $73 and $3,500 to $29 respectively.13 This obviously implies a high user cost for customers who rely 12 As of the end of 2018, only three African countries (Central African Republic, Eritrea and South Sudan) were unconnected to at least one submarine cable. Also every seaboard country in Africa, except Guinea-Bissau and Eritrea, have at least one submarine cable landing. See https://www.broadbandcommission.org/Documents/ working-groups/DigitalMoonshotforAfrica_Report.pdf 13 See the World Bank’s World Development Report 2021. https://openknowledge.worldbank.org/bitstream/ handle/10986/35218/9781464816000.pdf 9 on satellite transmission for Internet services. Hence, coupled with the relatively slow speed of satellite connections, submarine connectivity brought faster and cheaper Internet to many African countries, thereby enabling greater uptake and utilization of the Internet for commer- cial and household uses. To this end, our empirical analysis is structured in two stages. First, in our cross-country analysis, we exploit the plausibly exogenous variations in access to high-speed Internet across countries and time, induced by the staggered arrival of submarine Internet cables to the con- tinent. Our second strategy involves analysis at the subnational level that exploits plausibly exogenous variations in access to high-speed Internet induced by the staggered rollout of the fiber Internet backbone network across locations in the country as well as the gradual arrival of submarine Internet cables that connect domestic fiber Internet networks to the rest of the world. These approaches essentially rely on a difference-in-difference design to estimate the impact of access to fast Internet. Starting with the cross-country analysis, our baseline equation is expressed as: Yct = φ × 1(Submar i ne ct ) + Xct β + Γc + η t + µct (1) where Yc t is a placeholder for the sector-specific FDI receipts in country c in year t . Three main measures of FDI are used: (i) a dummy variable equal to 1 if there was any FDI project registered in the country in year t and 0 if otherwise; (ii) the number of FDI projects in the relevant year; and (iii) the total size (monetary value) of FDI projects in the relevant year. It is important to emphasize that FDI in Internet infrastructure is excluded due to the potential of reverse causal- ity. 1(Submar i ne ct ) is an indicator variable equal to 1 if a country is connected to at least one submarine Internet cable at time t and 0 if otherwise. X is a vector of macroeconomic controls such as real GDP growth and real GDP per capita. We estimate variant specifications while ex- cluding these macro-control variables. Meanwhile Γc and η t represent country and year fixed effects respectively. The inclusion of country fixed effects ensures that we compare changes in FDI receipts before and after the arrival of submarine Internet cable in the same country. Year fixed effects, on the other hand, enable us to absorb contemporaneous shocks that may have been common across countries. Standard errors are clustered at the country level to allow for within-country correlation among the residuals. The inverse hyperbolic sine (IHS) transforma- tion was applied to continuous variables, such as investment size, and number of FDI projects. In this setup, causal identification of impact relies on the assumption that the timing of the arrival of (connection to) submarine cables is plausibly exogenous (D’Andrea and Limodio, 2019; Hjort and Poulsen, 2019). In other words, absent the arrival of submarine cables, trends in FDI receipts between early- and late-connected countries would evolve along a similar pat- 10 tern (i.e., the so-called the so-called "parallel trends" assumption). To assess the validity of the identifying assumption, we estimate an event-study to trace the trends in FDI receipts before and after connection to the submarine fiber-optic network between early- and late-connected countries: 10 Yct = φτ × 1(Submar i ne ct τ ) + Γc + η t + µct (2) τ=−5;τ=−1 where τ is the year relative to when a country got connected to the submarine Internet cable network. We set the year before the arrival of the submarine cable , τ = −1, as the reference year. Also, we code τ = −5 and τ = 10 to include all periods at the extremes of the window (i.e., τ = −5 ∀ τ ≤ −5 and τ = 10 ∀ τ ≥ 10 ). Thus, conditional on plausibly exogenous rollout ˆ in equation 1 recovers the causal impact of high-speed Internet of submarine Internet cables, φ access on FDI. Our second strategy moves beyond the cross-country DiD regressions to a subnational level analysis. As highlighted earlier, once the submarine cables reached landing points on the coast, they were connected to the fiber backbone infrastructure that links several cities and towns to provide connectivity to the last-mile networks of terrestrial fiber broadband and aerial net- works. The gradual construction (rollout) of the inland fiber networks, together with the stag- gered variations in submarine fiber-optic connectivity across countries, provides a unique quasi- natural experiment to estimate the effects of access to high-speed Internet on the direction of FDI. Consequently, we harness the spatial granularity of the FDI and Internet data to conduct the sub-national analysis using the following DiD design: Yi c t = φ × 1(Submar i ne ct × C onnec t ed i ct ) + β × 1(C onnec t ed i ct ) + Γi + η c ×t + µi ct (3) where i indexes a subnational unit, hereafter referred to as a district. 1(C onnec t edi ct ) is an indicator variable that is equal to 1 if a district is connected to the fiber backbone network in the country and 0 if otherwise. Specifically, we consider a district to be connected if it contains at least one active fiber node connected to the backbone network and 0 if otherwise. Γi and η c ×t represent fixed effects for district and country×year, respectively. Thus in this DiD setup, we exploit plausible spatial and temporal variation in the roll out of the fiber network across locations to identify the effect of high-speed Internet access on FDI receipts. Standard errors are clustered at the district level to allow within-district correlation among the residuals. 11 5 Results 5.1 Country-Level Analysis As indicated earlier, our theory of change suggests that access to high-speed Internet would likely induce FDI in the financial services and technology services sectors, as the Internet plays a key role in technology adoption in these sectors relative to other sectors. In Table 2, we present our baseline analysis, showing results from estimating equation 1 using FDI receipts in technology services, financial services, other sectors excluding technology and financial services, and all FDIs as outcomes. For each sector, we present three sets of results with varying outcome variables: (i) a dummy variable equal to 1 if there was any FDI project registered in the country-sector and 0 if (ii) the number of FDI projects, and (iii) the total size (monetary value) of FDI projects. Starting with FDI in the financial services sector, the estimates in column 1 indicate that connection to the submarine fiber-optic cable is associated with an 18.4 pp increase in the probability of attracting FDI in the financial services sector. Relative to the sample mean, this corresponds to an increase of about 44%. In column 2, the connection to the submarine fiber- optic cable is associated with a 28.6% increase in the number of FDI projects in financial ser- vices.14 Similar to an increase in the number of FDI receipts, there was an overall increase in the size (monetary value) of FDI into the sector. Based on the results in column 3, connection to the submarine cable is associated with a significant increase (approximately 110%) in the value of FDI receipts to the financial sector.15 In columns 4-6, we explore the effects of high-speed Internet availability on FDI flows to the technology services sector. The results (column 4) suggest that connectivity to the submarine cable is associated with an 11.6 pp increase in the probability of receiving FDI in technology services. While the effect on the number of FDI projects in the sector is statistically insignificant (column 5), we find that the connectivity is associated with an increase in the value of FDI in the sector of approximately 66%. Interestingly, we do not find any statistically significant effect on FDI in other sectors (e.g., manufacturing, energy and extractives), as shown in columns 7-9. Similarly, we do not find any significant effect on total FDI receipts.16 This falls in line with our a priori expectations that only Internet-dependent sectors such as the financial and technology sectors are plausibly the most likely direct beneficiaries of investments induced by high-speed Internet connectivity. ˆ 14 In a log-linear model, this effect is given as e β − 1. Hence e 0.2514 − 1. 15 0.7443 e − 1 = 1.10 16 Excluding investment in Internet construction 12 Taken together, our results provide suggestive evidence that the provision of high-speed Inter- net through connection to the submarine Internet cables have significant impact on FDI flows in the financial and technology services sectors, but not other sectors. Further, in Table A1 in the appendix, we present results from estimating the baseline model without macroeconomic controls. The results are qualitatively and quantitatively similar to those in Table 2. Next, we use an event-study analysis to show the dynamics in FDI receipts before and after the arrival of the submarine cables. Given the findings from the baseline estimates in Table 2, we focus exclusively on FDI receipts in financial and technology services as shown in Figure 3. The top-left and top-right panels show, respectively, the trends in the number and size (mone- tary value) of FDI projects in financial services. In each case, we see an absence of pre-trends, as the estimates before the arrival of high-speed connectivity are statistically and economically no different from zero, suggesting that FDI in treated and control units prior to the arrival of the submarine cables trended in similar patterns. The absence of pre-trends gives confidence to the DiD (two-way fixed effects) estimates. However, as shown in the top-left panel, the coefficients become (increasingly) positive and sta tistically significant after the arrival of the submarine ca- bles. In particular, we see an uptick in the number of FDIs in financial services after the second year. A similar pattern is observed in the top-right panel, though the statistical significance of the estimates is mixed. In the bottom-left and bottom-right panels, we see similar patterns for FDI in technology services. Once again, the plots show the absence of pre-trends. In the bottom-left panel, for instance, while the estimates prior to the arrival of the submarine cables are close to 0 and sta- tistically insignificant, they trend upwards after the arrival and become statistically significant, suggesting a significant increase in the number of FDI projects in technology services induced by the availability of high-speed Internet via the submarine cables. A common observation among the four plots is the delayed response of FDI to the arrival of high-speed Internet. This can be plausibly explained by the fact that it took some countries sometime (1-2 years) to de- ploy the fiber broadband Internet from the landing points to other parts of cities and towns.15 Hence, in some cases there may have been a time lag between the arrival of the submarine cable connection and high-speed Internet availability in parts of the country. 5.2 Subnational Analysis We next turn our attention to the subnational level and analyze the effects of high- speed In- ternet availability on FDI receipts at the district level. Specifically, we ask to what extent does high-speed Internet availability at the local level explain within-country variations in the lo- cation of FDI projects? This is particularly important for our understanding of the localized 13 impact of infrastructure services. Also to the extent that FDI contributes to the local economy, in particular jobs (Toews and Vézina, 2021), knowledge of the drivers of FDIs is of particular importance to policy makers. Table 3 presents our DiD estimates of the effects of high-speed Internet on FDI at the sub- national level. Similar to the cross-country regressions, we split our analysis into four groups: technology services, financial services, other sectors excluding technology and financial ser- vices, and FDI in all sectors. Our parameter(s) of interest is the coefficient of Submar i ne XC onnec t ed which shows the pre-post differences in the outcome between treated and control (subnational) units. In column 1, the results suggest that access to high-speed Internet is associated with a 3.7 pp increase in the likelihood of a district receiving an FDI project related to financial ser- vices. Relative to the mean, this corresponds to about 54%. In columns 2 and 3, the results suggest that high-speed Internet access is associated with an increase in the number and value of financial services related FDI by 5% and 14%, respectively. The effects on investments in technology services (columns 4-5) are similar to those in fi- nancial services (columns 1-3), albeit with slightly higher magnitudes. The probability of tech- nology services-related FDI receipts increases by 5 pp following the connection to high-speed Internet at the district level. Meanwhile, the number and value (monetary) of technology- related FDIs received at the district level increases by 7.8% and 19%, respectively.17 Once again, we do not find any impact on FDI in non-technology and non-financial services sectors (columns 7-9). Interestingly, we find statistically significant effects on FDI in all sectors, but this is most likely driven by investments in financial and technology services. Overall, the subnational anal- yses are in line with the cross-country analysis. Access to high-speed Internet is associated with an increase in FDI in financial and technology services, as the Internet provides an enabling technology to increase productivity in these sectors. 5.3 Role of Complementary Enablers This section explores the role of complementary infrastructure, such as access to reliable elec- tricity, in amplifying or reducing the effects of access to high-speed Internet in attracting FDI in the finance and technology sectors. In other words, is access to high-speed Internet enough to drive FDI or is the effect conditional on having complementary infrastructure, such as reliable electricity? To address this question, we revisit the cross-country analysis and leverage data from the World Bank Doing Business database to classify countries into two groups: countries with re- 17 0.174 e − 1 = 0.19 14 liable versus unreliable electricity provision. Administrative data on electricity outages are scant in many developing countries, particularly in Africa. Where available, they are often not reported on a consistent basis, and varying methodologies are used across countries. The World Bank’s Doing Business database has attempted to measure the quality of electricity sup- ply across countries since 2005 by surveying managers of the major utilities on the SAIDI and SAIFI measures.18 These indicators essentially give an indication of the frequency and duration of outages in electricity systems. While these measures are not available over the entire study period, and also are not exhaustive of the countries in this study, they nonetheless provide a good representation. To this end, for each country with available data, we compute the aver- age SAIFI index between 2005 and 2018, classifying a country as having an unreliable supply of electricity if its average SAIFI index is above the average for all African countries and vice versa. We match this data with our cross-country dataset19 and estimate the following specification: Yc t = φ × 1(Submar i ne ct ) + θ × 1(Submar i ne ct × Unr el i abl eEl ec ) + Xct β + Γc + η t + µc t (4) where Unr el i abl eEl ec is a dummy variable coded 1 if the country’s average SAIFI index is above the African average and 0 if otherwise. All else remains as previously defined. The co- efficient of interest, θ , measures the effect of having a high-speed Internet connection via the submarine cables in countries with unreliable electricity supply relative to countries with reli- able electricity supply. Results are presented in Table 4. For each outcome, we estimate variant specifications with and without controls for macroeconomic variables (GDP growth and GDP per capita). Columns 1-6 show the results of the interactions between submarine cables and reliability of electricity in relation to FDI in financial services, while columns 7-12 show the corresponding estimates for investments in technology services. Starting with FDI in financial services, the results on the coefficient(s) of 1(Submar i ne ct ×Unr el i abl eEl ec ) paint an interesting picture.In column 2, for instance, the effect of high-speed connectivity on the probability of receiving an FDI in financial services is about 30 pp lower in countries with unreliable electricity supply relative to countries with reliable electricity supply. To put this in perspective, high-speed Internet con- nectivity increases the probability of FDI in financial services by 29 pp in countries with reliable electricity supplies, while FDI in financial services in countries with unreliable electricity sup- 18 System Average Interruption Duration Index (SAIDI) and System Average Interruption Frequency Index (SAIFI) 19 In all we have data on electricity reliability for 28 out of 54 countries in our dataset 15 ply reduces by 1 pp,20 although the effect is statistically insignificant. Similarly, the effects on the number and value (size) of FDI in financial services are 41%21 (column 4) and 150% (col- umn 6) higher in countries with reliable electricity access relative to countries with unreliable supply of electricity. The results on FDI in technology services are largely consistent with the financial services findings, indicating that high-speed Internet is associated with higher invest- ments in technology services in countries with reliable electricity supply relative to countries with unreliable access to electricity. Interestingly, as shown by the F-statistics of the joint test,22 the net effect of high-speed Internet connectivity in countries with unreliable electricity provi- sion is not statistically different from zero, implying that Internet connectivity does not induce FDI in countries with unreliable supply of electricity. Overall, the main takeaway from the analysis in this section is that while access to high- speed Internet is associated with increased investment in the finance and technology sectors, the effect is largely concentrated in countries with reliable access to electricity. This is plausibly due to the key role electricity plays in powering digital equipment and the general effect of electricity on technology adoption. Therefore, the findings herein provide suggestive evidence that reliable electricity services amply the role of high-speed Internet in attracting investment in the finance and technology sectors. 6 Conclusion This paper provides causal evidence of the role of infrastructure quality in driving FDI in de- veloping economies. To this end, it explores how the arrival of high-speed Internet in Africa via submarine Internet cables stimulated FDI in the technology and financial services sectors, which rely on Internet services for operations. It also examines the influence of complemen- tary infrastructure, such as access to reliable electricity, in amplifying the impact of Internet connectivity. We use granular data on FDI projects in several African countries matched with spatial data on Internet infrastructure roll out to estimate the effects of Internet connectivity on FDI in the banking and technology services sectors in Africa. The spatial and time granularity of our datasets allows us to undertake the analysis at both national and subnational levels. First, we show that the arrival of high-speed Internet played a crucial role in stimulating FDI in the banking and technology sectors in Africa. The probability of receiving FDI, as well as the number and value of FDI in the financial and technology services sectors, increased with access 20 0.29 − 0.30 = −0.1 21 0.3428 e − 1 = 0.409 22 Submarine Internet + SubmarineXUnreliable Electricity=0 16 to fast Internet. Second, we show that the effects of high-speed Internet connectivity on FDI pertain largely to countries with reliable electricity supply. In other words, while high- speed connectivity is associated with increased investment in the finance and technology sectors, the effect is largely dependent on access to reliable electricity. Overall, the findings of the paper underscore the importance of quality infrastructure provi- sion in attracting investments to developing and emerging economies. In addition, the results highlight the complementarities in the economic impact of infrastructural services. Thus, fu- ture research on the impact of infrastructural services should pay particular attention to these complementarities. 17 References Abebe, G., McMillan, M. S., and Serafinelli, M. (2018). Foreign direct investment and knowledge diffusion in poor locations: Evidence from Ethiopia. Technical report, National Bureau of Economic Research. Amiti, M. and Javorcik, B. S. (2008). Trade costs and location of foreign firms in China. Journal of Development Economics, 85(1-2):129–149. Andersen, T. B. and Dalgaard, C.-J. (2013). Power outages and economic growth in Africa. Energy Economics, 38:19–23. Arita, S., Tanaka, K., et al. (2013). FDI and investment barriers in developing economies. Tech- nical report, Institute of Developing Economies, Japan External Trade Organization (JETRO). Bech, M. L., Hobijn, B., et al. (2007). Technology diffusion within central banking: The case of real-time gross settlement. International Journal of Central Banking, 3(3):147–181. Crescenzi, R. and Limodio, N. (2021). The impact of Chinese FDI in Africa: Evidence from Ethiopia. Technical report. D’Andrea, A. and Limodio, N. (2019). High-speed Internet, financial technology and banking in Africa. Technical report, Centre for Applied Research on International Markets Banking. Edwards, L. and Jenkins, R. (2015). The impact of Chinese import penetration on the South African manufacturing sector. The Journal of Development Studies, 51(4):447–463. Guerrieri, V. and Lorenzoni, G. (2009). Liquidity and trading dynamics. Econometrica, 77(6):1751–1790. Harding, T. and Javorcik, B. S. (2011). Roll out the red carpet and they will come: Investment promotion and FDI inflows. The Economic Journal, 121(557):1445–1476. Harding, T. and Javorcik, B. S. (2012). Investment Promotion and FDI Inflows: Quality Matters. CESifo Economic Studies, 59(2):337–359. Hjort, J. and Poulsen, J. (2019). The arrival of fast Internet and employment in Africa. American Economic Review, 109(3):1032–79. Mann, L., G. M. and Friederici, N. (2015). The Internet and business process outsourcing in East Africa. 18 Manyika, J., Armando, C., Moodley, L., Moraje, S., Yeboah-Amankwah, S., Chui, M., and Antho- nyrajah, J. (2013). Lions go digital: The Internet’s transformative potential in Africa. Mayda, A. M., Parsons, C. R., Pham, H., and Vézina, P.-L. (2019). Refugees and foreign direct investment: Quasi-experimental evidence from US resettlements. Technical report. Mensah, J. T. (2018). Jobs! Electricity shortages and unemployment in Africa. Technical Report 8415, World Bank Policy Research Working Paper. Ouedraogo, R., Sy, M. A. N., et al. (2020). Can digitalization help deter corruption in Africa? Technical report, International Monetary Fund. Reinikka, R. and Svensson, J. (1999). How inadequate provision of public infrastructure and services affects private investment. The World Bank. Smarzynska Javorcik, B. (2004). Does foreign direct investment increase the productivity of do- mestic firms? In search of spillovers through backward linkages. American Economic Review, 94(3):605–627. Toews, G. and Vézina, P.-L. (2021). Resource discoveries, FDI bonanzas, and local multipliers: Evidence from Mozambique. Review of Economics and Statistics, pages 1–36. Townsend, R. M. (1978). Intermediation with costly bilateral exchange. The Review of Economic Studies, 45(3):417–425. 19 Figures Figure 1: Trends in Submarine Internet Cable Connectivity in Africa The figure shows the cumulative number of African countries with a connection to at least one submarine fiber- optic Internet cable between 2000 and 2018. 20 Figure 2: Arrival of Submarine Internet Cables in Africa 2000 2002 2008 2009 2010 2011 2012 2014 2018 The figure shows the gradual arrival of submarine Internet 21 cables in Africa. Blue lines show additional cables that are in-service at the respective years. Pink lines are existing cables. Maps are constructed using spatial data from https://cablemap.info/_default.aspx Figure 3: Event Study Analysis: High-speed Internet and FDI in Finance and Tech Services The figures show event-study plots (and 90% confidence intervals) of the effects of high-speed Internet access on FDI in financial and technology services sectors. These are estimated using the specification in equation 2. The top-left panel shows the trends in number of FDI in the financial services before and after the arrival high- speed Internet through a connection to a submarine fiber-optic Internet cable. The top-right panel however shows the trends in amount (USD) of FDI in financial services before and after the arrival high-speed Internet through a connection to a submarine fiber-optic Internet cable. The bottom-left and bottom-right panels replicates the same approach for FDI in the technology services sector. In each case the dependent variable is transformed by the inverse-sine hyperbolic transformation (IHS). 22 Tables Table 1: Summary Statistics Variable Mean Std. Dev. Min. Max. N Country Level FDI in Financial Services (0/1) 0.4 0.49 0 1 864 # of FDI in Financial Services 1.27 2.60 0 25 864 FDI in Financial Services ($ million) 17.80 43.88 0 571.44 864 FDI in Technology Services (0/1) 0.33 0.47 0 1 864 # of FDI in Technology Services 1.73 5.21 0 60 864 FDI in Technology Services ($ million) 62.69 369.76 0 6057.20 864 FDI in All Sectors (0/1) 0.68 0.46 0 1 864 # of FDI in All Sectors 8.14 17.53 0 155 864 FDI in All Sectors ($ million) 940.48 2714.30 0 35351.68 86 GDP per Capita 5576.55 6351.84 718.33 41249.43 806 GDP Growth (%) 4.43 7.10 -62.07 123.14 821 Submarine Internet (0/1) 0.596 0.491 0 1 864 Subnational Level FDI in Financial Services (0/1) 0.06 0.25 0 1 10240 # of FDI in Financial Services 0.10 0.51 0 11 10240 FDI in Financial Services ($ million) 1.50 10.72 0 518.67 10240 FDI in Technology Services (0/1) 0.04 0.21 0 1 10240 # of FDI in Technology Services 0.14 1.13 0 37 10240 FDI in Technology Services ($ million) 5.29 100.62 0 5940.93 10240 FDI in All Sectors (0/1) 0.22 0.41 0 1 10240 # of FDI in All Sectors 0.68 3.09 0 75 10240 FDI in All Sectors ($ million) 79.17 610.69 0 30612.24 10240 Submarine (0/1) 0.748 0.434 0 1 10240 Connected (0/1) 0.581 0.493 0 1 10240 Submarine × Connected (0/1) 0.467 0.499 0 1 10240 23 Table 2: High-speed Internet and FDI: Country Level Analysis Financial Services Technology Services (1) (2) (3) (4) (5) (6) FDI (0/1) # FDIs (IHS) FDI Size ($,IHS) FDI (0/1) # FDIs (IHS) FDI Size ($,IHS) Submarine Internet 0.1841*** 0.2514* 0.7443** 0.1157** 0.0873 0.4446* (0.0681) (0.1393) (0.2800) (0.0566) (0.1107) (0.2615) Mean dep. var 0.4186 0.6583 1.6666 0.3478 0.6310 1.4912 R-squared 0.4081 0.5655 0.5098 0.5511 0.7882 0.6532 Other Sectors All Sectors (7) (8) (9) (10) (11) (12) FDI (0/1) # FDIs (IHS) FDI Size ($,IHS) FDI (0/1) # FDIs (IHS) FDI Size ($,IHS) Submarine Internet 0.0158 0.4419 0.4419 0.0369 0.2098 0.4900 (0.0623) (0.4397) (0.4397) (0.0582) (0.1741) (0.3906) Mean dep. var 0.6087 3.9352 3.9352 0.7019 1.6561 4.3603 R-squared 0.4748 0.6242 0.6242 0.4396 0.8178 0.6469 Macro Ctrls Yes Yes Yes Yes Yes Yes Country FE Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Observations 805 805 805 805 805 805 Notes: Submarine Internet is a dummy variable equal to 1 if a country is connected to at least one submarine Internet cable at time t and 0 if otherwise. Continuous variables, such as the number and size of FDI projects, are transformed using the inverse hyperbolic sine (IHS) transformation. Macro controls include real GDP per capita and real GDP growth. Estimations are done using OLS. Robust standard errors clustered at the country level in parentheses. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 24 Table 3: High-speed Internet and FDI: A Subnational Analysis Financial Services Technology Services (1) (2) (3) (4) (5) (6) FDI (0/1) # FDIs (IHS) FDI Size ($,IHS) FDI (0/1) # FDIs (IHS) FDI Size ($,IHS) Submarine X Connected 0.0372* 0.0484** 0.1319** 0.0500*** 0.0751*** 0.1740*** (0.0196) (0.0229) (0.0641) (0.0162) (0.0249) (0.0590) Connected -0.0225 -0.0343 -0.0702 -0.0484** -0.0736*** -0.1649** (0.0212) (0.0237) (0.0686) (0.0194) (0.0264) (0.0705) Mean dep. var 0.0682 0.0767 0.2326 0.0477 0.0691 0.1815 R-squared 0.4140 0.5154 0.4653 0.5479 0.7495 0.6055 Other Sectors All Sectors (7) (8) (9) (10) (11) (12) FDI (0/1) # FDIs (IHS) FDI Size ($,IHS) FDI (0/1) # FDIs (IHS) FDI Size ($,IHS) Submarine X Connected 0.0389 0.2519 0.2519 0.0608** 0.1387*** 0.3416** (0.0279) (0.1542) (0.1542) (0.0306) (0.0452) (0.1521) Connected -0.0560* -0.3341* -0.3341* -0.0670* -0.1479*** -0.3621** (0.0309) (0.1726) (0.1726) (0.0347) (0.0506) (0.1766) Mean dep. var 0.1767 0.9306 0.9306 0.2215 0.3031 1.0945 R-squared 0.3958 0.4170 0.4170 0.4069 0.6828 0.4463 District FE Yes Yes Yes Yes Yes Yes Country X Year FE Yes Yes Yes Yes Yes Yes Observations 10240 10240 10240 10240 10240 10240 Notes: Submarine X Connected is a dummy variable equal to 1 if a district is connected to a node along the terrestrial fiber-optic backbone network that is connected to the submarine cable at time t and 0 if otherwise. Connected is a dummy equal to 1 if a district is connected at least one active node along the terrestrial fiber-optic backbone network. Continuous variables such as the number and size of FDI projects are transformed using the inverse-sine hyperbolic transformation (IHS). Estimations are done using OLS. Robust standard errors clustered at district level in parentheses. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 25 Table 4: High-speed Internet and FDI: The Role of Reliable Electricity Provision Financial Services (1) (2) (3) (4) (5) (6) FDI (0/1) # FDIs (IHS) FDI Size ($,IHS) Submarine Internet 0.3290*** 0.2916*** 0.5036** 0.4275* 1.3061*** 1.1492*** (0.0954) (0.0951) (0.2142) (0.2192) (0.3920) (0.3921) X Unreliable Electricity -0.3578*** -0.3003** -0.4450*** -0.3428** -1.1846*** -0.9197** (0.1280) (0.1259) (0.1598) (0.1542) (0.4122) (0.3966) Submarine Internet + SubmarineXUnreliable Electricity=0 F test 0.0522 0.0049 0.1188 0.2442 0.0763 0.2842 p-value 0.8210 0.9446 0.7330 0.6255 0.7845 0.5986 Mean dep. var 0.5357 0.5577 0.9408 0.9809 2.2583 2.3552 R-squared 0.4626 0.4638 0.5867 0.5948 0.5444 0.5523 Technology Services (7) (8) (9) (10) (11) (12) FDI (0/1) # FDIs (IHS) FDI Size ($,IHS) Submarine Internet 0.1507* 0.1433* 0.2097 0.1768 0.7304* 0.6844 (0.0817) (0.0835) (0.1774) (0.1882) (0.3878) (0.4099) X Unreliable Electricity -0.1672** -0.1436 -0.3599** -0.3248** -0.8842** -0.7649** (0.0787) (0.0875) (0.1442) (0.1542) (0.3211) (0.3539) Submarine Internet + SubmarineXUnreliable Electricity=0 F test 0.1209 0.0000 1.6737 1.3320 0.2915 0.0686 p-value 0.7308 0.9954 0.2067 0.2594 0.5937 0.7955 Mean dep. var 0.4888 0.5216 0.9768 1.0464 2.2254 2.3779 R-squared 0.5907 0.5795 0.8062 0.8034 0.6751 0.6729 Macro Ctrls No Yes No Yes No Yes Country FE Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Observations 448 416 448 416 448 416 Notes: Submarine X Connected is a dummy variable equal to 1 if a location is connected to a node along the terrestrial fiber-optic backbone network that is connected to the submarine cable at time t and 0 if otherwise. Unreliable Electricity is a dummy variable equal to 1 if the average System Average Interruption Frequency Index (SAIFI) in the country between 2004 and 2018 is above the sample (African) average and 0 if otherwise. Continuous variables such as the number and size of FDI projects are transformed using the inverse hyperbolic sine (IHS) transformation. Macro controls include real GDP per capita and real GDP growth. Estimations are done using OLS. Robust standard errors clustered at country level in parenthesis. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 26 A Appendix Table A1: High-speed Internet and FDI: Country Level Analysis without controls Financial Services Technology Services (1) (2) (3) (4) (5) (6) FDI (0/1) # FDIs (IHS) FDI Size ($,IHS) FDI (0/1) # FDIs (IHS) FDI Size ($,IHS) Submarine Internet 0.1757** 0.2471* 0.7071** 0.1192** 0.1007 0.4395* (0.0663) (0.1320) (0.2699) (0.0526) (0.1019) (0.2430) Mean dep. var 0.4005 0.6299 1.5935 0.3299 0.5936 1.4069 R-squared 0.4115 0.5607 0.5082 0.5464 0.7860 0.6496 Other Sectors All Sectors (7) (8) (9) (10) (11) (12) FDI (0/1) # FDIs (IHS) FDI Size ($,IHS) FDI (0/1) # FDIs (IHS) FDI Size ($,IHS) Submarine Internet 0.0301 0.5054 0.5054 0.0612 0.2462 0.5928 (0.0570) (0.4025) (0.4025) (0.0542) (0.1638) (0.3595) Mean dep. var 0.5914 3.8030 3.8030 0.6840 1.5863 4.2218 R-squared 0.4572 0.6051 0.6051 0.4302 0.8076 0.6280 Macro Ctrls No No No No No No Country FE Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Observations 864 864 864 864 864 864 Notes: Submarine Internet is a dummy variable equal to 1 if a country is connected to at least one submarine Internet cable at time t and 0 if otherwise. Continuous variables, such as the number and size of FDI projects, are transformed using the inverse hyperbolic sine (IHS) transformation. Estimations are done using OLS. Robust standard errors clustered at country level in parentheses. ∗ Significant at 10 percent level ∗∗ Significant at 5 percent level ∗∗∗ Significant at 1 percent level 27