The World Bank Economic Review, 38(1), 2024, 1–23 https://doi.org10.1093/wber/lhad021 Article Infrastructure Quality and FDI Inflows: Evidence from Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 the Arrival of High-Speed Internet in Africa Justice Tei Mensah and Nouhoum Traore Abstract Does ambient infrastructural quality affect foreign direct investment (FDI) in developing countries? This pa- per investigates how the arrival of high-speed internet in Africa triggered FDI into the region. It also explores the role of complementary infrastructure, such as access to electricity and road connectivity, in amplifying the impact of internet connectivity on investment. To causally estimate impacts, the paper exploits plausibly ex- ogenous variations in access to high-speed internet induced by the staggered arrival of submarine fiber-optic internet cables and spatial variations in terrestrial fiber cable networks across locations on the continent. Find- ings from the paper indicate that access to high-speed internet induces FDI, particularly in the service sector, with the finance, technology, retail, and health services subsectors as the main beneficiaries. Access to (hard) infrastructure, such as electricity and roads, amplifies the impact of internet connectivity on FDI, thus highlight- ing the role of complementarities in the impact of infrastructure. Further, the results suggest that improvement in quality of governance and increased performance of incumbent firms are plausible mechanisms. JEL classification: O12, O16, F21, F23, G21, N77 Keywords: internet, FDI, services, banking, technology 1. Introduction Foreign direct investment (FDI) is an important driver of economic growth as it facilitates technology transfer and job creation in host economies (Smarzynska Javorcik 2004; Toews and Vézina 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 Justice Tei Mensah (corresponding author) is an economist in the Office of the Chief Economist for Africa at the World Bank. His email address is jmensah2@worldbank.org. Nouhoum Traore is an economist at the International Finance Corporation of the World Bank Group. His email address is ntraore3@ifc.org. Funding for this research was provided by the World Bank’s Africa Chief Economist Office, and the Development Impact Department at IFC. We wish to express our profound gratitude to the editor, Nina Pavcnik, and two anonymous reviewers for their constructive comments on the paper. In addition, comments and support from Moussa Blimpo, Mark Dutz, Albert Zeufack, Valentina Saltane, Camilo Mondragon-Velez, Issa Faye, Andrew Dabalen, and seminar/conference participants at IFC, World Bank, PSDRN (2021), and CSAE Conference on African Development (2022) are deeply appreciated. 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 or the World Bank. A supplementary online appendix is available with this article at The World Bank Economic Review website. 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, McMillan, and Serafinelli (2018), and Crescenzi and Limodio (2021). C 2023 International Bank for Reconstruction and Development / The World Bank. Published by Oxford University Press. 2 Mensah and Traore Figure 1. Trends in Submarine Internet Cable Connectivity in Africa. Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 Note: 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. 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 investment locations (Reinikka and Svensson 1999; Harding and Javorcik 2011; Blonigen, Oldenski, and Sly 2014) 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 investment flows to developing countries by leveraging the arrival of high-speed internet to Africa as a unique natural exper- iment. Between 2000 and 2012, Africa experienced a significant improvement in its internet landscape with the arrival of several submarine fiber-optic internet cables linking the continent to the global inter- net network via Europe (fig. 1). Prior to this, internet transmissions to the continent were mainly through satellite, which was characterized by low bandwidth, low speed, and high cost of service. The submarine cables brought faster internet speeds at a relatively low cost, thereby increasing uptake of internet for commercial and household uses. The arrival of fast internet via submarine cables was particularly benefi- cial to the banking and finance sector, as it enabled the adoption of financial technologies (Fintech) such as internet banking and the real-time gross settlement system (RTGS), leading to an increase in economic activities in the sector (D’Andrea and Limodio 2023). The technology sector also benefited, as the avail- ability of high-speed internet created supply and demand for technology services.2 Similarly, as shown by Hjort and Poulsen (2019), manufacturing and service sector firms experienced increased performance following the arrival of the submarine internet cables and subsequent connection to inland towns and cities via the terrestrial fiber backbone networks. We interpret high-speed internet connection as a posi- tive technological shock that led to an expansion in the productive sectors, with possible implications on demand for investments in these sectors. Therefore, our goal is to show how the arrival of high-speed internet triggered FDI in the various sectors in Africa. In addition, we explore the potential role of complementary infrastructure, such as access to electricity and roads, in amplifying the impact of internet connectivity. To this end, we assemble unique project-level data on sectoral FDI, together with data on internet infrastructure, to estimate the spillover effects of fast internet connectivity on FDI flows. The spatial and time granularities of our data sets allow 2 See for example https://www.bbc.com/news/technology-46055595. The World Bank Economic Review 3 us to undertake the analysis at the subnational level. Our identification strategy is a difference-in-difference (DiD) design that relies on the plausibly exogenous variation in access to high-speed internet induced by the staggered arrival of submarine internet cables across countries and the spatial variation in access to the terrestrial fiber backbone network within countries. Essentially, we compare districts connected to the terrestrial fiber network to those unconnected, and explore the changes in the FDI flows to these districts before and after the arrival of the submarine cables in the respective countries. Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 Three main findings emerge from the analysis. First, the arrival of high-speed internet played a key role in stimulating FDI to Africa. However, the effects are largely concentrated in the service sector, with finance, technology, health, and retail subsectors as the main beneficiaries. The effect on FDI in services is statistically and economically significant in both the intensive and extensive margins. On the extensive margin, the probability of a subnational district receiving FDI in services increases by 5.7 percentage points (pp) following the arrival of high-speed internet connectivity. On the intensive margin, access to high-speed internet connectivity is associated with a 9.6 percent and 23.6 percent increase in the number and value of FDI in services. Within the service sector, we find positive and statistically significant effects of internet connectivity on FDI in the various subsectors, that is, finance, technology, health, and retail. The effects of internet connectivity on FDI in other sectors such as manufacturing, energy, and construction are not statistically significant. Second, we provide suggestive evidence that the effects of high-speed internet connectivity on FDI per- tain largely to subnational districts with high access to complementary infrastructure such as electricity and roads. For instance, districts with high electricity grid density are shown to attract greater FDI fol- lowing internet connectivity compared to districts with relatively low grid infrastructure density. This is possibly due to the fact that access to electricity is essential in powering devices such as computers and mobile phones that use the internet, hence greater availability of the electricity grid infrastructure signals low connection cost and greater reliability of electricity services, thereby attracting investments. These findings speak to the potential role of complementary infrastructure, such as access to electricity and roads, in amplifying the economic impact of internet services in developing countries. Finally, we provide suggestive evidence on two potential mechanisms: (a) Access to high-speed inter- net connectivity is associated with improvement in the quality of governance. Access to internet reduces information friction and enhances reporting and scrutiny, thereby enabling citizens to demand greater accountability from their governments. The improvement in the quality of governance makes countries attractive for investment, as it reduces uncertainties and the cost of doing business. (b) Access to the inter- net is associated with an increase in the performance of incumbent firms due to the expansion in market opportunities for both firms and consumers enabled by internet connectivity. This increase in firm per- formance (proxied by sales) is a potential signal of the perceived high returns on investment to investors, thereby resulting in high FDI inflows. 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 have focused on the role of deliberate government policies, such as investment promotion initiatives (Harding and Javorcik 2011, 2012) and business regulations (Arita et al. 2013), in attracting 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 resource booms trigger FDI in non-extractive sectors. Preferential tariffs and trade agreements have also been cited as potential incentives to attract FDI (Helpman 2006; Feinberg and Keane 2006, 2009; McCaig, Pavcnik, and Wong 2022). McCaig, Pavcnik, and Wong (2022) for instance, show that the United States–Vietnam bilateral trade agreement and the consequent reduction in tariffs on Vietnamese exports to the United States led to an increase in FDI in the Vietnamese manufacturing 3 That may be induced by language and cultural distance. 4 Mensah and Traore sector. Aside from these, evidence from other studies also underscores the influence of non-economic factors, such as cultural ties between countries, on FDI flows across countries. For instance, 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 a high concentration of refugees. In spite of these compelling pieces of evidence, little is known about the premium that foreign investors place on the quality of infrastructural services in destination markets when Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 they are assessing investment opportunities. 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 impacts of mod- ern infrastructure, such as internet and electricity, in developing and emerging economies (Hjort and Tian 2021). The arrival of high-speed internet in Africa has been a topic of growing interest to many researchers (Hjort and Poulsen 2019; D’Andrea and Limodio 2023; Ouedraogo et al. 2020; Goldbeck and Lindlacher 2021). 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 (2023) show that access to high-speed internet led to a boom in the financial sector, as it enabled technology adoption among African banks. These positive impacts of internet connectivity on job creation (Hjort and Poulsen 2019) and technology adoption (D’Andrea and Limodio 2023) translated into high economic growth in Africa (Goldbeck and Lindlacher 2021). These induced impacts on job creation and technology adoption also translated into economic growth, as shown by Goldbeck and Lindlacher (2021). Thus, consistent with D’Andrea and Limodio (2023), 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 inter- net stimulates financial development. Findings on the complementary role of access to electricity 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 framework linking internet to investment in service (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. We discuss potential mechanisms and robustness checks in Sections 6 and 7 respectively. Section 8 concludes the paper with a summary of the main findings. 2. Conceptual Framework In this paper we hypothesize that access to high-speed internet is likely to trigger investments in low- income countries, with the service sector as the most likely beneficiary, as internet connectivity will enable faster adoption of internet in the sector. Subsectors such as retail, finance, and technology services are also potential beneficiaries. In this section we describe how internet connectivity increases technology adoption in the financial services and technology sectors, and consequently its effects on the development of these sectors, to illustrate the spillover effects of access to high-speed internet on investments. 2.1. High-Speed Internet and FDI in Financial Services Financial technologies, popularly referred to as “Fintech,” have changed the banking and finance land- scape around the world by reducing costs and improving efficiencies of financial intermediation. The banking sector has historically been organized around manual processing, with customers going to bank- ing halls to make deposits and withdrawals. Interbank payments 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 The World Bank Economic Review 5 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). The RTGS enables a one-to-one exchange of money and securities between banks (financial institu- tions) in real time without needing to bundle or net them with other trades (D’Andrea and Limodio 2023). 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 associated with a significant reduction in Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 the cost of interbank settlements and improves efficient integration of the financial sector (Townsend 1978; Guerrieri and Lorenzoni 2009; D’Andrea and Limodio 2023). 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 2023). Thus, the RTGS plays an important role in the development of the financial sector. Adoption of the RTGS is relatively 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 the RTGS. What influenced adoption of the RTGS across countries? Aside from regulatory and policy reforms, ac- cess to high-speed internet is a key determinant of RTGS uptake across the world. As a real-time exchange, the RTGS requires fast and reliable interconnection among parties. D’Andrea and Limodio (2023), for instance, show that the arrival of high-speed internet in Africa led to an increase in adoption of the RTGS. As a result, banks significantly reduced their liquidity hoarding, while interbank transactions and lending to customers increased. Essentially, the findings from D’Andrea and Limodio (2023) 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 access 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 “bricks-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 the RTGS and internet bank- ing 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 adoption of the 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 entertainment, social network- ing, online banking, e-payments, telemedicine, and e-commerce. Demand for these services is particularly high when they are made in Africa or have content tailored for the African market, creating opportunities for digital firms that provide these services. 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 Nigeria with the aim of tapping into Africa’s movie industry. Social media companies like Twit- ter 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 continent’s share of population with an online presence, there are huge economic prospects for the development of a digital economy on the 6 Mensah and Traore continent. These trends are expected to catalyze investments into the technology services sector to meet the growing demand for digital services. Second, from the supply side, access to fast internet has created a “launch pad” for digital entrepreneur- ship in Africa (Manyika et al. 2013; Houngbonon, Mensah, and Traore 2022). With the rising demand for digital services and the low concentration of tech companies on the continent, a new generation of tech entrepreneurs is rising in Africa. In many countries across the continent, young and vibrant entrepreneurs Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 are establishing start-ups in such areas as mobile payments, online retail, e-transport and logistics, and e-learning. For instance, in Nigeria alone, Konga and Jumia have become major online retailers, while Kobo360—an e-logistics company—is expanding and providing 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 vendors, 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 responding 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 investments 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 infrastructure is also a catalyst for investment in the technology services sector. 3. Data Our main data are project-level data on FDI from fDiMarkets,8 a subsidiary of the Financial Times. fDi- Markets monitors cross-border investments around the globe and is a major source of data used by the World Bank and UNCTAD9 to measure cross-border financial flows. It provides detailed information re- lating to date of announcement, sector, subsector, whether the investment is greenfield (new) or brownfield (expansion),10 names of investors, source country and city (state), destination country and city, capital in- vestment, and anticipated direct jobs created, among other things. Data on FDI from 2003 to 2018 are used in this study. 4 Jobberman is a leading digital marketplace for employers and prospective employees, 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 experiments. 6 Morocco and South Africa now export significant services in the BPO area. The sector generates more than US$1.5 billion of revenues and accounts for 54,000 jobs in South Africa (Manyika et al. 2013). 7 The company counts Google, Microsoft, Salesforce, and Yahoo among its clients. See https://www.bbc.com/news/techn ology-46055595. 8 https://www.fdimarkets.com/. 9 The United Nations Conference on Trade and Development. 10 About 95 percent of projects in the database are greenfield. The World Bank Economic Review 7 Table 1. Summary Statistics Variable Mean Std. dev. Min. Max. FDI in all sectors (0/1) 0.192 0.394 0 1 # of FDI in all sectors 0.628 3.165 0 73 FDI in all sectors ($ million) 39.9 241.5 0 7,457.1 FDI in manufacturing (0/1) 0.076 0.265 0 1 Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 # of FDI in manufacturing 0.104 0.427 0 6 FDI in manufacturing ($ million) 14.296 142.608 0 7,058.8 FDI in services (0/1) 0.14 0.347 0 1 # of FDI in services 0.492 2.871 0 69 FDI in services ($ million) 10.705 84.169 0 4,237.326 FDI in other sectors (energy & real estate) (0/1) 0.027 0.161 0 1 # of FDI in other sectors (energy & real estate) 0.033 0.217 0 4 FDI in other sectors (energy & real estate) ($ million) 12.881 155.925 0 6,811.5 FDI in financial services (0/1) 0.072 0.259 0 1 # of FDI in financial services 0.117 0.548 0 11 FDI in financial services ($ million) 1.518 9.106 0 319.7 FDI in tech services (0/1) 0.06 0.238 0 1 # of FDI in tech services 0.203 1.492 0 38 FDI in tech services ($ million) 3.942 63.584 0 4,127.7 FDI in other services (health & retail) (0/1) 0.07 0.255 0 1 # of FDI in other services (health & retail) 0.173 1.15 0 33 FDI in other services (health & retail) 5.244 40.826 0 1,595.7 Submarine 0.544 0.498 0 1 Connected 0.741 0.438 0 1 Submarine × connected 0.409 0.492 0 1 Electricity grid network (km) 443.4 774.7 1.117 5697.4 Distance to the coast (km) 427.17 378.76 0 1690.34 Road length (km) 1,149.439 2,374.847 0.001 22,051.2 We conduct our analysis at the subnational level, specifically at the second administrative level, here- after referred to as the district. For each district and year, we compute the total number of FDI projects and the value of investment for each sector. In this paper, we focus on three main sectors: manufacturing,11 services, energy and real estate (construction).12 We exclude FDI in telecom and internet construction, and extractives (mining and petroleum) given that (a) investments in telecom and internet construction could be associated with the internet rollout, thereby resulting in reverse causality, and (b) investments in the extractives are largely influenced by discoveries of deposits and less of ambient infrastructure quality. Within the service sector, we leverage the detailed nature of the data to classify the data into three cate- gories (subsectors): FDI in financial services, technology services, and other services.13 Our final data set is a district-year panel between 2003 and 2018. 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 to landlocked countries, using data from 11 Garments and apparel, machinery, agro-processing, etc. 12 Electricity generation, transmission & generation, real estate construction. 13 Financial services include activities such as retail banking, investment banking, and insurance. The technology services classification includes FDI in business services (computer systems design, software publishers, motion picture & sound recording industries, custom computer programming services, travel and ticketing, rental and leasing services), software and IT services, biotechnology, paper printing & packaging, communications, consumer electronics, semiconductors, aerospace. Other services include retail, health care, hospitality, and real estate services. 8 Mensah and Traore www.africabandwidthmaps.com,14 www.submarinecablemap.com, and other country-specific sources. We complement this with data on the terrestrial fiber backbone network from www.africabandwidthm aps.com. Figure 1 shows trends in the number of countries connected to at least one submarine cable between 2003 and 2018. We complement the internet data with additional data on infrastructure such as the electricity grid network from https://gridfinder.org/ and road density from the Global Roads Inventory Project.15 Also, Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 we use data on the quality of governance from the World Bank’s World Governance Indicator database16 and, finally, firm-level data in nine countries17 from the World Bank Enterprise Survey (WBES). Our final data are a balanced panel of 401 subnational districts in 32 countries over a 16-year period.18 Summary statistics of the full data are presented in table 1. 4. Empirical Strategy In the 1990s, satellite transmission was the main source of Africa’s internet connection. It was charac- terized by high connection costs and slow speeds, which affected the utilization of internet services for productive uses. To address these issues, several countries partnered with telecom companies to undertake massive infrastructural investments by laying fiber-optic submarine cables (SMC) 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. We refer to these early cables as the “first-generation” SMCs to Africa. Also, the band- width capacity of these early submarine cables was still relatively low, limiting high-speed internet access to a few locations. The construction of the undersea cables slowed between 2002 and 2008, partly due to the global financial crisis and dot-com bubble (D’Andrea and Limodio 2023). However, between 2009 and 2012, Africa witnessed a boom in the arrival of (“second-generation”) 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 fig. 1).19 Figure 2 shows the gradual arrival of the submarine cables between 2000 and 2018. Further, the landing points of the submarine cables were mostly cities along the coast. However, to get this internet available to end users, it has to travel through the “middle-mile” and “last-mile” networks.20 The middle mile consists of the national (fiber) “backbone” infrastructure, which transmits high-capacity internet from the landing points of the submarine cables throughout the do- mestic network. Along these fiber backbone networks, there are several connection points, referred to as “internet nodes,” that connect the backbone to the last-mile infrastructure–which is essentially the form of infrastructure that households and firms connect directly to for internet. In Africa, the last mile network usually consists of fiber broadband cables, and mobile internet (via mobile phone towers) 14 A digital consulting firm focused on publishing maps on internet infrastructure in Africa. 15 https://www.globio.info/download-grip-dataset. 16 https://info.worldbank.org/governance/wgi/. 17 These include Ethiopia, Ghana, Kenya, Mauritania, Nigeria, Senegal, Tanzania, Uganda, and Zambia. 18 These include Angola, Benin, Botswana, Burundi, Cameroon, Central African Republic, Democratic Republic of Congo, Eritrea, Eswatini, Ethiopia, Gabon, Ghana, Kenya, Lesotho, Madagascar, Mali, Mauritania, Mozambique, Namibia, Niger, Nigeria, Republic of the Congo, Rwanda, Senegal, Somalia, South Africa, Sudan, Tanzania, Togo, Uganda, Zam- bia, and Zimbabwe. 19 As of the end of 2018, only three African countries (Central African Republic, Eritrea, and South Sudan) were uncon- nected to at least one submarine cable. Also, every seaboard country in Africa, except Guinea-Bissau and Eritrea, has at least one submarine cable landing. See https://www.broadbandcommission.org/Documents/working-groups/DigitalMo onshotforAfrica_Report.pdf. 20 The “middle-mile” can be likened to the network of electricity transmission lines that carries high-voltage energy from generation plants to substations. The “last-mile” on the other hand can be likened to the network of distribution lines that carry low-voltage electricity from substations to customers. The World Bank Economic Review 9 Figure 2. Arrival of Submarine Internet Cables in Africa. Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 Source: Authors’ constrcted using spatial data from https://cablemap.info/_default.aspx. Note: The figure shows the gradual arrival of submarine internet cables in Africa. The black lines show additional cables that are “in-service” at the respective years. Pink lines are existing cables. 10 Mensah and Traore (Hjort and Poulsen 2019). The national backbone infrastructure also played an important role in con- necting (neighboring) landlocked countries to the submarine cables. As highlighted by Hjort and Poulsen (2019), in almost all African countries, the backbone networks were built by national telecom companies, albeit, in some instances, private telecomunication companies (telcos) extended the network. Also, many of these backbone networks were built many years before the arrival of the submarine cables, as internet was hitherto transmitted via telephone cables. Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 We treat the arrival of the SMCs as a positive shock in internet connectivity. This is because internet via these cables is (a) faster and (b) 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.21 This obviously implies a high user cost for customers who rely 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 commercial and household uses. In this paper, we follow the approach of Hjort and Poulsen (2019) and exploit the plausibly exogenous variations in the arrival of the SMCs in Africa in a difference-in-difference design. The timing of the arrival of the SMCs is arguably exogenous given the fact that (a) the construction of the cables was undertaken through a consortium of countries and hence delays in coordination among the project partners could induce significant delays; (b) even in the absence of delays, a country’s relative location along the coastline influenced the timing of SMC connection; and (c) access to the SMCs by landlocked countries depend to a large extent on the speed of construction (expansion) of national backbone networks of neighboring countries (Goldbeck and Lindlacher 2021). Essentially, we estimate the effect of internet connectivity on FDI using a double difference design that leverages two main sources of variations: first, the staggered arrival of SMCs along the African coastlines, and second comparing subnational districts connected to the terrestrial backbone network with those unconnected to the network, focusing exclusively on backbone networks constructed prior to the arrival of the SMCs. Consequently, we specify our baseline equation as follows: Yict = φ · 1(Submarinect × Connectedi ) + β · Xi × T + i + ηct + μict , (1) where Yct is a placeholder for the sector-specific FDI receipts in subnational district i, country c, and year t. Three main measures of FDI are used: (a) a dummy variable equal to 1 if there was any FDI project in the district in year t, and 0 otherwise; (b) the number of FDI projects in the relevant year, and (c) the total size (monetary value) of FDI projects in the relevant year. It is important to emphasize that FDIs in internet infrastructure are excluded due to the potential of reverse causality. Likewise, FDIs in extractives are also excluded. The term 1(Submarinect ) is an indicator variable equal to 1 if a country is connected to at least one submarine internet cable at time t, and 0 otherwise. The term 1(Connectedi ) is an indicator variable equal to 1 if a district is connected to the fiber backbone network in the country, and 0 otherwise. Specifically, we consider a district to be connected if the terrestrial backbone line is within the boundaries of the district, and 0 otherwise.22 We also include district-level controls (distance to the coast, road density, and intensity of electricity grid network), interacted with time trends. The terms i and ηct represent fixed effects for district and country × year, respectively. The inclusion of district fixed effects absorbs time-invariant district effects that may influence FDI flows. Country-year fixed effects, on the other hand, enable us to absorb country-specific temporal shocks that may influence the outcome variable. The main parameter of interest is φ , which represents the causal impact of access to high-speed 21 See the World Bank’s World Development Report 2021; https://openknowledge.worldbank.org/bitstream/handle/109 86/35218/9781464816000.pdf. 22 That is, either at least one fiber line passes through the district boundary or the boundary contains at least one endpoint of the network. The World Bank Economic Review 11 internet connectivity on FDI flows. Standard errors are clustered at the district level to allow within-district correlation among the residuals (Abadie et al. 2022). The inverse hyperbolic sine (IHS) transformation was applied to continuous variables, such as investment size and the 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 2023; Hjort and Poulsen 2019). In other words, absent the arrival of submarine cables, trends in FDI receipts between connected Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 and unconnected districts would have evolved along a similar pattern (i.e., 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 connected and unconnected districts. Meanwhile, recent advances in the literature suggest that the two- way fixed effects (TWFE) estimator is likely to be biased due to the possibility of heterogeneity in the treatment effects across groups (De Chaisemartin and D’Haultfoeuille 2018, 2020a,b; Goodman-Bacon 2021; Callaway and Sant’Anna 2021; Borusyak, Jaravel, and Spiess 2021). This implies that the regular event studies estimates based on the TWFE estimator are plausibly biased. Therefore, we present event study estimates using the “imputation” estimator by Borusyak, Jaravel, and Spiess (2021), which ensures consistency in the presence of the TWFE and also deals with issues of spurious identification and negative weighting (Acemoglu, He, and le Maire 2022; De Chaisemartin and D’Haultfoeuille 2020b). Finally, in the main analysis, we focus exclusively on the second-generation SMCs to Africa. This is motivated by the following three facts: (a) The first-generation SMCs had relatively low bandwidth capacity and hence low internet speeds. Bandwidth cost also remained relatively expensive. On the other hand, the second-generation SMCs brought faster internet speeds at low cost to the continent and marked the beginning of rapid expansion in internet connectivity in Africa (Hjort and Poulsen 2019). (b) The arrival of the first-generation SMCs (1997–2002) predates the start of our FDI data (2004), hence limiting the application of our before-and-after design. (c) The staggered arrival of the second-generation SMCs also enables causal identification of impact as used in Hjort and Poulsen (2019). Therefore, in our quest to examine the effects of high-speed internet connectivity on FDI, focusing on second-generation SMCs offers the best option. However, including countries that had already received the first-generation SMCs in the analysis, while assigning treatment status to the arrival of the second-generation SMCs, may pose challenges to causal identification.23 Therefore, our baseline analysis focuses solely on countries without prior connection to the first-generation SMCs.24 Nonetheless, we show in Section 7.2 that our results are robust to the inclusion of these countries. 5. Results We begin by presenting descriptive evidence on the trends in FDI receipts between connected (treated) and unconnected (control) districts before and after the arrival of the first SMC cable in their respective countries. This allows us to check the gap in FDI receipts between treated and control districts before and after the arrival of high-speed internet connection, thus providing a validity check of our identifying assumption that requires that FDI receipts in the two groups would have evolved along a similar pattern in the absence of treatment. As shown in fig. 3, we observe a similar upward trend in the number of FDI projects between treated and control districts within five years prior to the SMC arrival. However, the gap in FDI receipts between treated and control districts begins to increase after the arrival of the SMC, suggesting a potentially positive impact of high-speed internet connectivity on FDI receipts. 23 For example, countries already connected to the first-generation SMCs may have a high (low) probability of receiving the second-generation SMCs. 24 Thus the sample for the main (baseline) analysis consists of data from 237 districts in 21 countries. 12 Mensah and Traore Figure 3. Trends in FDI Receipts in Connected and Unconnected Districts before and after the Arrival of High-Speed Internet. Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 5.1. Main Results We now turn our attention to the main results. We present our DiD estimates of the effects of high-speed internet connectivity on sectoral FDI using the baseline specification in equation (1). Table 2 presents the results showing how high-speed internet connection affects FDI receipts in all sectors, services, man- ufacturing, and other sectors (energy and real estate). Across these sectors, we evaluate the effects of internet connectivity on three outcomes: “FDI (0/1),” a dummy variable set equal to 1 if a district in year t received FDI in the respective sector, and 0 otherwise; “# FDIs (IHS),” the IHS transformation of the number of FDI projects received in year t; and “FDI size ($, IHS),” the IHS transformation of the size (monetary value) of FDIs received. Also, for each of these outcomes we estimate two variant speci- fications: with and without baseline district controls interacted with (linear) time trends. Our preferred specifications are those with the baseline district controls interacted with time trends (even numbered columns). Starting with the aggregate FDI receipts across all sectors,25 our results suggest that high-speed internet connectivity increases the probability of receiving FDI by about 6.3 pp (column 2). Also, the arrival of high-speed internet is associated with an increase in the number and size of FDI receipts by 10.5 percent (column 4) and 33.4 percent (column 6) respectively.26 In which sectors are the impacts of internet connec- tivity on FDI concentrated? The results in columns 7–12 show a positive impact of internet connectivity on FDI into services. The probability of receiving FDI in the services sector increases by 5.7 pp, while the number and size of FDI receipts in services increase by 9.6 percent (column 10) and 23.6 percent (column 12) respectively following the arrival of high-speed internet. In terms of FDI in manufacturing, we do not find any statistically significant effect of high-speed internet connectivity, albeit the estimates are positive and relatively sizeable. Similarly, we do not find any consistent evidence of an impact of internet con- nectivity on FDI in the energy and real estate (sub)sectors. Overall, the results herein suggest that while access to fast internet is associated with an increase in FDI receipts, the effect is mainly concentrated in FDI in service-related activities. This falls in line with our a priori expectations that internet-dependent sectors, like services, are plausibly the most likely direct beneficiaries of investments induced by high-speed internet connectivity. 25 Excluding investments in extractives, telecom, and internet construction. 26 ˆ − 0.5Var(β In an IHS-linear model, this effect is given as ≈ exp(β ˆ ) is the estimated variance of β ˆ )) − 1, where Var(β ˆ (Bellemare and Wichman 2020). Hence, in column 4, the estimate corresponds to exp (0.1005 − 0.5(0.0334)2 ) − 1 = 0.105. Table 2. High-Speed Internet and FDI (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) FDI (0/1) # FDIs (IHS) FDI size ($, IHS) FDI (0/1) # FDIs (IHS) FDI size ($, IHS) All sectors Services The World Bank Economic Review Submarine × connected 0.0791∗∗∗ 0.0632∗∗ 0.1394∗∗∗ 0.1005∗∗∗ 0.4226∗∗∗ 0.2957∗∗ 0.0703∗∗∗ 0.0574∗∗ 0.1229∗∗∗ 0.0918∗∗∗ 0.3099∗∗∗ 0.2164∗∗ (0.0264) (0.0283) (0.0333) (0.0334) (0.1090) (0.1244) (0.0244) (0.0255) (0.0315) (0.0305) (0.0890) (0.0920) Mean dep. var 0.1645 0.1697 0.2104 0.2187 0.7148 0.7396 0.1202 0.1239 0.1534 0.1595 0.4171 0.4299 R-squared 0.4173 0.4306 0.6149 0.6320 0.4377 0.4515 0.4564 0.4751 0.6449 0.6638 0.5272 0.5517 (13) (14) (15) (16) (17) (18) (19) (20) (21) (22) (23) (24) Manufacturing Other sectors (energy & real estate) ∗ Submarine × connected 0.0111 0.0009 0.0194 0.0116 0.0752 0.0306 0.0224 0.0164 0.0199∗ 0.0129 0.1162 0.0886 (0.0170) (0.0193) (0.0163) (0.0183) (0.0873) (0.0976) (0.0122) (0.0143) (0.0118) (0.0139) (0.0723) (0.0862) Mean dep. var 0.0588 0.0608 0.0615 0.0638 0.2861 0.2962 0.0212 0.0226 0.0208 0.0222 0.1242 0.1326 R-squared 0.2914 0.2959 0.3384 0.3453 0.2819 0.2866 0.2441 0.2491 0.2490 0.2538 0.2311 0.2348 Baseline ctrls × trend No Yes No Yes No Yes No Yes No Yes No Yes District FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Country × year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 3,776 3,536 3,776 3,536 3,776 3,536 3,776 3,536 3,776 3,536 3,776 3,536 Source: own estimations Note: Submarine × connected is a dummy variable equal to 1 if a subnational district is connected to a fiber-optic backbone network that is connected to the submarine cable in year t, and 0 otherwise. Baseline ctrls × trend represents the interaction between time trend and distance from the district centroid to the nearest coastline, total length of roads, and electricity grid network. FDI (0/1) is a dummy variable equal to 1 if there was any FDI (foreign direct investment) project in the district in year t, and 0 otherwise; # FDIs (IHS) is the inverse hyperbolic sine (IHS) transformation of the number of FDI projects in the district in year t; and FDI size ($, IHS) is the IHS transformation of the amount (constant USD) of FDI in the district. Robust standard errors are clustered at the district level in parentheses. ∗ Significant at 10 percent level. ∗∗ Significant at 5 percent level. ∗∗∗ Significant at 1 percent level. 13 Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 14 Mensah and Traore To further understand the internet–FDI nexus, we drill deeper into the service(s) sector to explore the effects of internet connectivity on FDI in the various subsectors, namely, financial services, technology services, and others (health and retail). In table 3 (columns 7–12), we find that access to high-speed internet leads to an increase in FDI in financial services: the probability of receiving FDI into the subsector increases by 5.7 pp, while the number and size of the investments into the subsector increase by 6.2 percent (column 10) and 20.4 percent (column 12) respectively. FDI into technology services also improves with Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 connectivity to high-speed internet. The probability of receiving FDI in technology services increases by 2.7 pp (column 14), while the number of FDI projects in technology services increases by 3.4 percent (column 16). The effect on the size of FDI into technology services is also positive and significant (column 17), albeit it loses significance once we control for district-level characteristics interacted with time trends (column 18). We also find positive effects on FDI in other service subsectors (including health and retail). These effects are statistically and economically meaningful. The probability of receiving FDI increases by 2 pp (column 20), while the number and size of FDI in the subsector increase by 4.1 percent (column 22) and 11.5 percent (column 24) respectively. Taken together, the main finding from the results is that high-speed internet connectivity increases FDI, although the effect is mainly concentrated in the services sector (e.g., finance, technology, retail, and health). 5.1.1. Event Study Aside from the overall positive average effect of internet connectivity on FDI, how does the effect vary over time? This is important for at least two reasons: (a) dynamic estimates from an event-study analysis allow us to provide an assessment of the parallel trends assumption—a fundamental assumption behind the DiD design, and (b) to assess whether the impacts of internet connectivity on FDI are instantaneous or accumulate over time. Also, motivated by the recent advances in the literature in relation to the limitations of the regular TWFE estimator in producing consistent estimates in the presence of differential weight- ing and heterogeneous treatment effects, we use the Borusyak, Jaravel, and Spiess (2021) “imputation” estimator to conduct event study analysis. This Borusyak, Jaravel, and Spiess (2021) imputation estimator follows a three-step procedure: (a) The coefficients of control variables, and fixed effects (time and unit) are estimated from a regression based solely on untreated observations. (b) Using the regression results from step one, the treatment effect of each observation is computed as the difference between the observed (actual) outcome and the potential outcome of the untreated. (c) Using the individual treatment effect from step two, the average treatment effect on the treated is computed.27 Figure 4 presents event study estimates from the Borusyak, Jaravel, and Spiess (2021) estimator, show- ing trends in the number of FDI projects (all sectors) before and after the arrival of high-speed internet at the district level. Before the arrival of fast internet enabled by SMC connectivity, point estimates are close to zero and statistically insignificant, suggesting that connected and unconnected districts were not on differential trends in FDI receipts prior to the arrival of SMCs at the country level, conditional on our fixed effects. This provides support for our parallel trends assumption, despite the seeming presence of an anticipation effect at time t − 1. However, once connected to the backbone network, and the subsequent availability of high-speed internet to treated districts, FDI receipts into connected districts began to rise at a faster rate relative to unconnected districts. The effect persists throughout the five years after the SMC’s arrival. We also conduct similar event study analyses focusing on the service sector (and subsectors) as shown in fig. 5. In fig. 5(a) for instance, we find a sustained increase in the number of FDI projects in (all) services in 27 It is important to note that, unlike the conventional event study estimators, the reference group for pretrends test in the Borusyak, Jaravel, and Spiess (2021) imputation estimator is all k periods prior to the event date and all never-treated observations. Table 3. High-Speed Internet and FDI in the Services Sector (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) FDI (0/1) # FDIs (IHS) FDI size ($, IHS) FDI (0/1) # FDIs (IHS) FDI size ($, IHS) Services (all) Financial services The World Bank Economic Review Submarine × connected 0.0703∗∗∗ 0.0574∗∗ 0.1229∗∗∗ 0.0918∗∗∗ 0.3099∗∗∗ 0.2164∗∗ 0.0512∗∗ 0.0566∗∗ 0.0582∗∗ 0.0609∗∗ 0.1692∗∗ 0.1887∗∗ (0.0244) (0.0255) (0.0315) (0.0305) (0.0890) (0.0920) (0.0216) (0.0233) (0.0230) (0.0249) (0.0704) (0.0770) Mean dep. var 0.1202 0.1239 0.1534 0.1595 0.4171 0.4299 0.0659 0.0676 0.0706 0.0729 0.2156 0.2223 R-squared 0.4564 0.4751 0.6449 0.6638 0.5272 0.5517 0.3774 0.3951 0.4372 0.4519 0.4096 0.4268 (13) (14) (15) (16) (17) (18) (19) (20) (21) (22) (23) (24) Technology services Other services (health & retail) Submarine × connected 0.0340∗∗ 0.0273∗∗ 0.0464∗∗∗ 0.0335∗∗ 0.0754∗ 0.0542 0.0480∗∗∗ 0.0201∗ 0.0712∗∗∗ 0.0408∗∗ 0.2466∗∗∗ 0.1103∗∗ (0.0148) (0.0131) (0.0179) (0.0161) (0.0384) (0.0390) (0.0158) (0.0120) (0.0225) (0.0203) (0.0838) (0.0519) Mean dep. var 0.0400 0.0421 0.0512 0.0541 0.1095 0.1155 0.0567 0.0591 0.0670 0.0702 0.2098 0.2171 R-squared 0.4312 0.4443 0.6096 0.6200 0.4844 0.4966 0.4194 0.4368 0.5321 0.5462 0.4494 0.4733 Baseline ctrls × trend No Yes No Yes No Yes No Yes No Yes No Yes District FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Country × year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 3,776 3,536 3,776 3,536 3,776 3,536 3,776 3,536 3,776 3,536 3,776 3,536 Source: own estimations Note: Submarine × connected is a dummy variable equal to 1 if a subnational district is connected to a fiber-optic backbone network that is connected to the submarine cable in year t, and 0 otherwise. Baseline ctrls × trend represents the interaction between time trend and distance from the district centroid to the nearest coastline, total length of roads, and electricity grid network. FDI (0/1) is a dummy variable equal to 1 if there was any FDI (foreign direct investment) project in the district in year t, and 0 otherwise; # FDIs (IHS) is the inverse hyperbolic sine (IHS) transformation of the number of FDI projects in the district in year t; and FDI size ($, IHS) is the IHS transformation of the amount (constant USD) of FDI in the district. Robust standard errors are clustered at the district level in parentheses. ∗ Significant at 10 percent level. ∗∗ Significant at 5 percent level. ∗∗∗ Significant at 1 percent level. 15 Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 16 Mensah and Traore Figure 4. Event Study: High-Speed Internet and FDI. Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 Note: The figure shows event study estimates from the Borusyak, Jaravel, and Spiess (2021) estimator. The outcome variable is the number of FDI projects in the district at year t. Standard errors are clustered at subnational district level. treated districts relative to control districts after the arrival of fast internet. Meanwhile, before the arrival of fast internet, we do not find any statistical difference between the inflow of FDI to control and treated districts, albeit there seems to be an anticipation effect at time t − 1. In relation to financial services, the results in fig. 5(b) show an increase in the number of FDI receipts to treated districts in the first three years after connection to a high-speed internet network. Beyond the third year, the effect is still positive, albeit statistically insignificant at the 95 percent confidence interval. Figure 5(c) presents the event study results for FDI into technology services. The results show a clear increase in FDI into technology services after the arrival of fast internet connection, albeit the coefficients are marginally statistically insignificant at the 5 percent error level. Finally, as regards FDI into other services (retail & health), we once again see a sharp increase in FDI following high-speed internet connectivity (see fig. 5(d)). Overall, the results from these event study analyses provide further support for the robustness of our results in tables 1 and 3. 5.2. Role of Complementary Enablers In this section we explore the role of complementary infrastructure, such as access to roads and electricity, in amplifying or reducing the effects of access to high-speed internet in attracting FDI. In other words, is access to high-speed internet enough to drive FDI, or is the effect conditional on having complementary (hard) infrastructure (road and electricity)? To address this question, we revisit the baseline analysis and leverage spatial data on access to road and electricity grid networks: two critical infrastructures that support the productive sectors of the economy by aiding transport and providing needed energy for the production of goods and services. Specifically, we compute the density of electricity grid and road networks at the district level and classify districts into high and low (grid/road) density based on the sample median. Thus “high electrification (high road density)” is a dummy equal to 1 if the electricity grid (road) density is above the sample median, and 0 otherwise. The density of the electricity grid is used as a measure (proxy) of the availability of electricity and hence the cost of electricity connection (supply). Places with high grid intensity plausibly have low connection costs, all else equal. Similarly, high road density is used as a proxy for transport cost, that is, high road density implies low transport cost and vice versa. The World Bank Economic Review 17 Figure 5. Event Study Analysis: High-Speed Internet and FDI at the Subnational Level. Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 Note: The figure shows event study estimates from the Borusyak, Jaravel, and Spiess (2021) estimator. The outcome variable is the number of FDI projects in the respective sectors in the district at year t. Standard errors are clustered at subnational district level. We combine these infrastructure data with our baseline data and estimate the following specification: Yict = φ · 1(Submarinect × Connectedi ) +γ · 1(Submarinect × Connectedi × Infra) +β · Zi × T + i + ηct + μict , (2) where Infra is a placeholder for the (high) electrification (road) dummy. The term Zi is a placeholder for baseline measures such as distance to the coast interacted with time trends. All else remains as previously defined. The coefficient of interest, γ , measures the effect of having a high-speed internet connection via submarine cables in districts with high electrification (road) density. Based on the results in tables 1 and 3, we do this analysis for total FDI and FDI in services. Results are presented in tables S1.1 and S1.2 in the supplementary online appendix. Starting with table S1.1, columns 1–9 show the interaction effects on FDI in all sectors. Across the various specifications, we find positive and statistically significant effects on the interaction between high-speed internet connectivity 18 Mensah and Traore and the high electrification dummy.28 Interestingly, the effect of internet connectivity in places with low electrification is not statistically significant. This provides suggestive evidence of the complementary role that access to electricity plays in amplifying the effects of internet connectivity. The interaction between road density and internet connectivity is largely (statistically) insignificant, albeit positive. Next we focus on FDI in the services sector. Once again, the interaction with electrification is highly significant across all specifications. In column 10 for instance, the results show that high-speed internet Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 connectivity in districts with high electrification footprints increases the probability of attracting FDI into services by 8.1 pp relative to internet connectivity in districts with low electrification. The interaction with road density is also positive and statistically significant in all cases except columns 12, 15, and 18, where we estimate the joint interaction effects together in the same specification. Nonetheless, the results here suggest that access to road networks also amplifies the effects of internet connectivity in attracting FDI into the services sector. Further, in table S1.2 in the supplementary online appendix, we examine the role of access to electricity and roads in amplifying the effects of internet connectivity on FDI into finance and technology services, as well as the other (retail and health) service categories. Again, high access to electricity plays a crucial role in enhancing the effects of internet connectivity on FDI in these service subsectors. For instance, internet connectivity increases the probability of attracting FDI to finance, technology, and health and retail services by 5.3 pp (column 1), 6.9 pp (column 10), and 4.3 pp (column 19) respectively in districts with high electrification rates relative to districts with low electrification rates. Interestingly, the interactive effect of road networks on the internet–FDI nexus is largely evident with respect to FDI in technology services. 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 services sector, the effect is largely concentrated in districts with better access to infrastructure, particularly electricity. This is plausibly due to the key role electricity plays in powering digital equipment and the general effect of electricity on technology adoption. There- fore, the findings herein provide suggestive evidence that electricity services amplify the role of high-speed internet in attracting investment. 6. Potential Mechanisms The preceding analysis shows a positive effect of high-speed internet connectivity on FDI, particularly in the service sector. In this section, we provide evidence on (some) potential pathways via which high- speed internet connectivity influences FDI arrivals. In this paper, we hypothesize at least three channels: (a) a reduction in the cost of doing business, including, but not limited to, transport costs and other trade barriers; (b) improvement in the quality of governance; and (c) market expansion (Hjort and Tian 2021). Empirical evidence on these channels, particularly changes in the cost of doing business following the arrival of fast internet, is constrained by the lack of data. However, we show some (suggestive) evidence of the governance and market expansion channels. 6.1. Quality of Governance Recent evidence from the literature shows that access to digital technologies, such as internet and mo- bile phones, affects political economy outcomes such as confidence in government (Guriev, Melnikov, and Zhuravskaya 2020), political mobilization (Manacorda and Tesei 2020), and corruption (Gonzalez 2021). For instance, Gonzalez (2021) shows that expansion in mobile phone connectivity reduced elec- toral fraud in Afghanistan, while Guriev, Melnikov, and Zhuravskaya (2020) also show that expansion in mobile internet connectivity is associated with a reduction in government approval by citizens. In all 28 The effects are insignificant in two out of six cases. The World Bank Economic Review 19 Table 4. Internet Connectivity and Governance Government Control of Regulatory quality effectiveness corruption Political stability Rule of law (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Submarine internet 0.0711∗ 0.0724∗ 0.1232∗∗ 0.1229∗∗ 0.1429 0.1433 0.0197 0.0318 0.0409 0.0449 Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 Connectivity (0.0381) (0.0393) (0.0577) (0.0566) (0.0967) (0.0969) (0.1115) (0.0979) (0.0728) (0.0750) Baseline ctrls × trend No Yes No Yes No Yes No Yes No Yes Country FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Mean dep. var −0.8157 −0.8157 −0.8931 −0.8931 −0.6927 −0.6927 −0.6136 −0.6136 −0.7835 −0.7835 R-squared 0.9472 0.9475 0.9465 0.9465 0.9365 0.9366 0.8457 0.8582 0.9462 0.9491 Observations 448 448 448 448 448 448 448 448 448 448 Source: own estimations Note: Dependent variables are indices on governance from the World Bank World Governance Indicators database. Submarine internet connectivity is a dummy variable equal to 1 if the country is connected to at least one submarine fiber internet cable, and 0 otherwise. Baseline ctrls × trend represents the interaction between a linear time trend and a dummy variable equal to 1 if the country is anglophone, and 0 otherwise. Robust standard errors are clustered at the country level in parentheses. ∗ Significant at 10 percent level. ∗∗ Significant at 5 percent level. ∗∗∗ Significant at 1 percent level. these, a principal mechanism is that access to these digital technologies reduces information friction, thus enabling citizens to demand greater accountability from governments. In other words, internet access en- hances reporting and scrutiny, for example via social media, and thus will induce governments to be more transparent and accountable to the populace, thereby improving the quality of governance (Khazaeli and Stockemer 2013). For investors, quality of governance is an important driver of investment decisions, as it largely influ- ences the cost of doing business and also reduces uncertainty in the business environment. To this end, we test the association between high-speed connectivity via the SMC arrivals and quality of governance using metrics such as regulatory quality, government effectiveness, control of corruption, political stability, and rule of law from the World Bank’s World Governance Indicator database.29 Specifically, we run the following specification at the country-level regression: Yct = φ · 1(Submarinect ) + β · Xc × T + ψc + ηt + ct , (3) where Yct is a placeholder for the measure of the quality of governance in country c at year t; 1(Submarinect ) an indicator variable set equal to 1 if the country is connected to at least one subma- rine internet cable, and 0 otherwise. We also control for baseline controls, such as whether the country is anglophone or otherwise, interacted with time trends represented by the term Xc × T.30 Standard errors are clustered at the country level. The results in table 4 show a positive relationship between the arrival of the SMCs and governance measures, albeit the relationship is statistically significant only with respect to regulatory quality, and government effectiveness. In column 2, for instance, high-speed internet connectivity is associated with a 0.072 standard deviation (sd) increase in the regulatory quality index. This result is particularly impor- tant to the findings of the paper since the regulatory quality index captures “perceptions of the ability of the government to formulate and implement sound policies and regulations that permit and promote private sector development.” Hence, access to the internet is associated with policies that support pri- vate sector development. Similarly, high-speed internet connectivity is associated with a 0.12 sd increase 29 https://info.worldbank.org/governance/wgi/. 30 The anglophone/francophone/Lusophone classification is used as a proxy for the legal origins of the country, that is, common law vs civil law, as the current legal framework in many African countries was inherited from their colonial rulers (Anderson 2018). 20 Mensah and Traore Table 5. High-Speed Internet Connectivity and Firm Performance Sales (log) Sales per worker (log) (1) (2) (3) (4) Submarine × connected 0.6709∗ 0.5796∗ 0.7357∗∗ 0.7437∗∗ (0.3520) (0.3253) (0.3457) (0.3379) Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 Firm ctrls No Yes No Yes City FE Yes Yes Yes Yes Industry × year FE Yes Yes Yes Yes Yr of est. FE Yes Yes Yes Yes Source: own estimations Note: Dependent variables are the natural logs of total firm sales (constant 2009) and sales per worker (constant 2009). Firm ctrls include a dummy equal to 1 if the firm is sole proprietorship, and 0 otherwise; firm size; and whether the firm has foreign ownership. Robust standard errors are clustered at the city level in parentheses. ∗ Significant at 10 percent level. ∗∗ Significant at 5 percent level. ∗∗∗ Significant at 1 percent level. in the government effectiveness index. Overall, the results herein provide suggestive evidence that inter- net connectivity is associated with an improvement in the quality of governance, which could plausibly incentivize foreign investments. 6.2. Market Expansion Another channel underlying the economic impact of internet connectivity is the expansion in market access for both firms and consumers (Hjort and Tian 2021). In addition to traditional “bricks-and-mortar” markets, internet access offers an additional market channel via “online markets,” thus allowing firms to sell products outside the boundaries of traditional markets, at relatively low cost. The expansion in market access via online platforms (in addition to offline/traditional platforms) will result in increased revenue for firms with positive implications on profitability. The potential for high revenue is particularly important for investors as it increases the expected return on investment. To test this channel, we use firm-level data from the WBES to explore the relationship between high- speed internet connectivity and firm sales by estimating the following equation:ok Ymict = φ · 1(Submarinect × Connectedi ) + β Xmt + i + ηct + μmict , (4) where Ymict is the outcome of firm m in city i, country c at time t; Xmt is a vector of firm controls (ownership type, firm size, and whether it is foreign or domestic), while i represent city fixed effects. All else remains as previously defined. Standard errors are clustered at the city level. In table 5, we find a positive effect of internet connectivity and the two measures of firm sales: log of total sales and sales per worker. These results provide evidence suggestive of the economic benefits of internet connectivity on revenue, potentially signaling a high return on investment. 7. Robustness Checks Before concluding the paper, we address two potential challenges to our baseline analysis, namely standard error clustering and extending the sample to include countries with prior connection to the first-generation SMCs. 7.1. Standard Errors We begin by showing the robustness of our analysis to alternative clustering. Given that one dimension of our treatment indicator—SMC arrival—varies at the country level, we conduct additional analysis by clustering our standard errors at the country level to account for serial correlation within countries. The The World Bank Economic Review 21 results are shown in tables S1.3 and S1.4 in the supplementary online appendix. The statistical significance remains robust as in the baseline results (tables 1 and 3), where we cluster the standard errors at the subnational level. We also show that our results are robust to spatial correlation in the residuals by replicating the baseline results (tables 1 and 3) using the Conley (1999) estimator. Results are shown in tables S1.5 and S1.6 in the supplementary online appendix. Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 7.2. Sample Extension As highlighted in Section 4, our baseline analysis focuses exclusively on the set of countries that got con- nected to the second-generation SMCs to Africa without prior connection to the first-generation SMCs. In this section we relax this criterion to include countries with a prior connection to an SMC before 2007, while still focusing on the effect of high-speed internet connectivity via the second-generation SMCs on FDI.31 Essentially, we classify these countries as though they were not “treated” prior to 2007. While, in principle, these countries were already connected to the early SMCs that arrived on the continent, internet speed in these countries was relatively low due to the low bandwidth capacity of these early cables. Hence, given our focus on “high-speed” internet connectivity, our focus on the second-generation SMCs should suffice. To this end, we replicate the baseline analysis in tables 1 and 3 using the extended sample and show the results in tables S1.7 S1.8 respectively in the supplementary online appendix. Compared to the baseline results (tables 1 and 3), our new results based on the extended sample are qualitatively and quantitatively similar (tables S1.7 and S1.8) albeit the latter are slightly smaller in magnitude. We also perform our event study analysis using the Borusyak, Jaravel, and Spiess (2021) estimator. Results in fig. S1.1 also confirm our earlier findings of an increase in FDI after connection to the SMCs. Thus, overall, we find consistent evidence that high-speed internet connectivity is associated with high FDI inflow, particularly, in the service sector. 8. Conclusion This paper provides causal evidence of the role of infrastructure quality in driving FDI in developing economies. It explores how the arrival of high-speed internet in Africa via submarine internet cables stimulated FDI into the continent. It also examines the influence of complementary infrastructure, such as electricity and road connectivity, in amplifying the impact of internet access, and also explores potential mechanisms underlying the impact. We use granular data on FDI projects in several African countries, matched with spatial data on inter- net infrastructure rollout, to estimate the effects of internet connectivity on sectoral FDI in Africa. Our identification strategy leverages the plausibly exogenous variations in the staggered arrival of submarine fiber internet cables that brought high-speed internet to the continent and the spatial variations in access to the terrestrial cable network. First, we show that the arrival of high-speed internet played a crucial role in stimulating FDI to Africa. The effects are, however, largely concentrated in the service(s) sector, with the finance, technology, health and retail subsectors as the main beneficiaries. The probability of receiving FDI, as well as the number and value of FDI in these services (sub)sectors, increased with access to fast internet. Second, we find suggestive evidence that the effect of high-speed internet connectivity on FDI may pertain largely to subnational dis- tricts with better access to complementary infrastructure, such as roads and electricity. Third, we provide suggestive evidence that an increase in the quality of governance and market expansion—resulting in high 31 The additional countries included are Senegal, Ghana, Nigeria, Benin, South Africa, Angola, Gabon, Cameroon, and Sudan. 22 Mensah and Traore sales (return on investments)—are potential mechanisms through which high-speed internet connectivity induces FDI. Overall, the findings of the paper underscore the importance of quality infrastructure provision in attracting investments to developing and emerging economies. In addition, the results highlight the com- plementarities in the economic impact of infrastructural services. Thus, future research on the impact of infrastructural services should pay particular attention to these complementarities. Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 Data Availability Statement The FDI data used in this study is proprietary data from https://www.fdimarkets.com/ and hence cannot be shared without the express permission of the data provider. Interested users may purchase license to the data from the provider. References Abadie, A., S. Athey, G. W. Imbens, and J. M. Wooldridge. 2023. “When Should You Adjust Standard Errors for Clustering?” Quarterly Journal of Economics. 138(1): 1–35. Abebe, G., M. S. McMillan, and M. Serafinelli. 2022. Foreign Direct Investment and Knowledge Diffusion in Poor Locations: Evidence from Ethiopia Journal of Development Economics 158: 102926. Acemoglu, D., A. He, and D. le Maire. 2022. “Eclipse of Rent-Sharing: The Effects of Managers’ Business Education on Wages and the Labor Share in the US and Denmark.” Technical report, National Bureau of Economic Research. Amiti, M., and B. S. Javorcik. 2008. “Trade Costs and Location of Foreign Firms in China.” Journal of Development Economics 85(1–2): 129–49. Andersen, T. B., and C.-J. Dalgaard. 2013. “Power Outages and Economic Growth in Africa.” Energy Economics 38: 19–23. Anderson, S. 2018. “Legal Origins and Female HIV.” American Economic Review 108(6): 1407–39. Arita, S., K. Tanaka, et al. 2013. “FDI and Investment Barriers in Developing Economies.” Technical report, Institute of Developing Economies, Japan External Trade Organization (JETRO). Bech, M. L., B. Hobijn, et al.. 2007. “Technology Diffusion within Central Banking: The Case of Real-Time Gross Settlement.” International Journal of Central Banking 3(3): 147–81. Bellemare, M. F., and C. J. Wichman. 2020. “Elasticities and the Inverse Hyperbolic Sine Transformation.” Oxford Bulletin of Economics and Statistics 82(1): 50–61. Blonigen, B. A., L. Oldenski, and N. Sly. 2014. “The Differential Effects of Bilateral Tax Treaties.” American Economic Journal: Economic Policy 6(2): 1–18. Borusyak, K., X. Jaravel, and J. Spiess. 2021. “Revisiting Event Study Designs: Robust and Efficient Estimation.” arXiv:2108.12419. Callaway, B., and P. H. Sant’Anna. 2021. “Difference-in-Differences with Multiple Time Periods.” Journal of Econo- metrics 225(2): 200–30. Conley, T. G.. 1999. “GMM Estimation with Cross Sectional Dependence.” Journal of Econometrics 92(1): 1–45. Crescenzi, R., and N. Limodio. 2021. “The Impact of Chinese FDI in Africa: Evidence from Ethiopia.” Technical report. D’Andrea, A., and N. Limodio. 2023. “High-Speed Internet, Financial Technology and Banking in Africa.” Manage- ment Science De Chaisemartin, C., and X. D’Haultfoeuille. 2018. “Fuzzy Differences-in-Differences.” Review of Economic Studies 85(2): 999–1028. ———. 2020a. “Difference-in-Differences Estimators of Intertemporal Treatment Effects.” Technical report. ———. 2020b. “Two-Way Fixed Effects Estimators with Heterogeneous Treatment Effects.” American Economic Review 110(9): 2964–96. Edwards, L., and R. Jenkins. 2015. “The Impact of Chinese Import Penetration on the South African Manufacturing Sector.” Journal of Development Studies 51(4): 447–63. The World Bank Economic Review 23 Feinberg, S. E., and M. P. Keane. 2006. “Accounting for the Growth of MNC-Based Trade Using a Structural Model of US MNCs.” American Economic Review 96(5): 1515–58. ———. 2009. “Tariff Effects on MNC Decisions to Engage in Intra-Firm and Arm’s-Length Trade.” Canadian Journal of Economics/Revue canadienne d’économique 42(3): 900–29. Goldbeck, M., and V. Lindlacher. 2021. “Digital Infrastructure and Local Economic Growth: Early Internet in Sub- Saharan Africa.” Technical report, Working Paper. Downloaded from https://academic.oup.com/wber/article/38/1/1/7240369 by Joint Bank/Fund Library user on 02 February 2024 Gonzalez, R. M. 2021. “Cell Phone Access and Election Fraud: Evidence from a Spatial Regression Discontinuity Design in Afghanistan.” American Economic Journal: Applied Economics 13(2): 1–51. Goodman-Bacon, A. 2021. “Difference-in-Differences with Variation in Treatment Timing.” Journal of Economet- rics.255 2 254–277. Guerrieri, V., and G. Lorenzoni. 2009. “Liquidity and Trading Dynamics.” Econometrica 77(6): 1751–90. Guriev, S., N. Melnikov, and E. Zhuravskaya. 2021. “3G Internet and Confidence in Government.” Quarterly Journal of Economics 136 4 2533–2613. Harding, T., and B. S. Javorcik. 2011. “Roll Out the Red Carpet and They Will Come: Investment Promotion and FDI Inflows.” Economic Journal 121(557): 1445–76. ———. 2012. “Investment Promotion and FDI Inflows: Quality Matters.” CESifo Economic Studies 59(2): 337–59. Helpman, E. 2006. “Trade, FDI, and the Organization of Firms.” Journal of Economic Literature 44(3): 589–630. Hjort, J., and J. Poulsen. 2019. “The Arrival of Fast Internet and Employment in Africa.” American Economic Review 109(3): 1032–79. Hjort, J., and L. Tian. 2021. “The Economic Impact of Internet Connectivity in Developing Countries.” INSEAD Working Paper Houngbonon, G. V., J. T. Mensah, and N. Traore. 2022. “The Impact of Internet Access on Innovation and En- trepreneurship in Africa.” World Bank Policy Research Working Paper. 9945d Javorcik, B. S. 2004. “Does Foreign Direct Investment Increase the Productivity of Domestic Firms? In Search of Spillovers through Backward Linkages.” American Economic Review 94(3): 605–27. Khazaeli, S., and D. Stockemer. 2013. “The Internet: A New Route to Good Governance.” International Political Science Review 34 (5): 463–82. Manacorda, M., and A. Tesei. 2020. “Liberation Technology: Mobile Phones and Political Mobilization in Africa.” Econometrica 88(2): 533–67. Mann, L., G. M., and N. Friederici. 2015. “The Internet and Business Process Outsourcing in East Africa.” Oxford: Oxford Internet Institutehttps://opendocs.ids.ac.uk/opendocs/handle/20.500.12413/11949 Manyika, J., C. Armando, L. Moodley, S. Moraje, S. Yeboah-Amankwah, M. Chui, and J. Anthonyrajah. 2013. “Lions Go Digital: The Internet’s Transformative Potential in Africa.” Mckinsey Global Insti- tutehttps://www.mckinsey.com/∼/media/mckinsey/industries/technology%20media%20and%20telecommuni cations/high%20tech/our%20insights/lions%20go%20digital%20the%20internets%20transformative%20pot ential%20in%20africa/mgi_lions_go_digital_full_report_nov2013.pdf Mayda, A. M., C. R. Parsons, H. Pham, and P.-L. Vézina. 2022. “Refugees and Foreign Direct Investment: Quasi- Experimental Evidence from US Resettlements.” Journal of Development Economics 156: 102818. McCaig, B., N. Pavcnik, and W. F. Wong. 2022. “FDI Inflows and Domestic Firms: Adjustments to New Export Opportunities.” Technical report, National Bureau of Economic Research. 30729 Mensah, J. T. 2018. “Jobs! Electricity Shortages and Unemployment in Africa.” Technical Report 8415, World Bank Policy Research Working Paper. Ouedraogo, R., M. A. N. Sy, et al. 2020. “Can Digitalization Help Deter Corruption in Africa?” Technical report, International Monetary Fund. Reinikka, R., and J. Svensson. 1999. How Inadequate Provision of Public Infrastructure and Services Affects Private Investment. World Bank Policy Research Working Paper 2262 Toews, G., and P.-L. Vézina. 2022. “Resource Discoveries, FDI Bonanzas, and Local Multipliers: Evidence from Mozambique.” Review of Economics and Statistics: 104(5): 1046–1058. Townsend, R. M. 1978. “Intermediation with Costly Bilateral Exchange.” Review of Economic Studies 45(3): 417–25.