• 1 One can distinguish between economic and social infrastructure. Recent evidence suggests heterogene (...)

1Since infrastructure investment is widely recognised as a crucial driver of economic development, while the quality, quantity and accessibility of economic infrastructure in developing countries lag considerably behind those in advanced economies, scaling up infrastructure investment is widely seen as a key pillar in national development strategies in low-income developing countries (LIDCs).1 In fact, in recent years, many developing countries have been scaling up infrastructure investment, mostly through public spending, but also with a growing participation of the private sector. The growth dividend and the distributional effect of this investment push cannot be taken for granted, as past experiences suggest (see Section II), and many challenges lie ahead: infrastructure gaps are still large and bridging those gaps will require tackling several problems, in terms of additional financing and project selection and implementation.

• 2 This analysis complements and extends previous regional studies, such as IMF (2014b) and IMF (2016) (...)

2This paper reviews infrastructure investment in LIDCs, focusing on the last 15 years, which have been characterised by a rising importance of non-official financing flows. Our main objective is to provide a multi-faceted picture of infrastructure development in LIDCs, covering the evolution of several physical indicators of infrastructure, the role of public and private sectors in delivering infrastructure, and its financing, including traditional and new sources.2

• 3 Enhancing infrastructure is a key component of the 2030 Development Agenda, mentioned explicitly in (...)

5Infrastructure investment is a key component of the 2030 Development Agenda.3 However, the interest around infrastructure is not new (Rosenstein-Rodan, 1943). Since the 1990s there has been a wide body of literature looking at the possible development gains from investing in infrastructure (World Bank, 1994). The economic importance of infrastructure investment has been analysed both at project and macro levels. At a project level, the focus is on the social cost-benefit of infrastructure projects and their implied internal rate of return (see Marcelo et al., 2016, for example). The social cost-benefit analysis often tries to account for negative externalities. At a macro level, the impact of infrastructure investment is analysed using aggregate production function with the assumption that infrastructure is complementary to other inputs in the production function (see Aschauer, 1989; Gramlich 1994). The macro literature shows that improvements in infrastructure could raise productivity, stimulate private investment (Cavallo and Daude, 2011), and facilitate domestic and international trade (Bougheas et al., 1999; Vijil and Wagner, 2012), thereby promoting sustainable growth (Esfahani and Ramı́rez, 2003; Agénor, 2010; Calderón and Servén, 2010).

• 4 In addition, agricultural productivity could decline in the upstream because of flood, increased sa (...)

8Increased access to essential infrastructure services could reduce inequality, foster inclusion and support poverty reduction efforts (Calderón and Chong, 2004; Calderón and Servén, 2010). Micro-level evidence shows that the distributional effect of infrastructure investment could vary. For instance, Khandker et al. (2009) look at road improvement projects in Bangladesh and find overall positive effects on output and poverty reduction. They also show that the poorest households are those benefiting the most. Similarly, Jedwab and Storeygard (2016) point to the importance of taking the local context into consideration, given the evidence of heterogeneous effects of transportation investments in Africa—which seem to favour small and remote cities. The evaluation of programs of infrastructure rehabilitation in Georgia and Vietnam also shows positive average effects, with some evidence of a stronger effect on the poor (Lokshin and Yemtsov, 2005; Mu and van de Walle, 2011). Duflo and Pande (2007) look at large public infrastructure investments—specifically, dams in India—and find a bleaker picture as poverty, in the aggregate, rises. Moreover, they point out significant distributional implications, as agricultural productivity increases in downstream districts, which benefit from irrigation, but not in those where dams are built, where construction activities causes loss of agricultural land and expose the population to diseases, resulting in higher poverty rates4. Similarly, the extensive highway network built in China since the 1990s has complicated spatial effect on economic activity, with winners and losers. Large cities in the centre of a dense regional highway network grow faster and specialise in business services and manufacturing, while the hinterlands grow more slowly, and become relatively more specialised in agriculture (Baum-Snow et al., 2017). This points to the importance of anticipating distributional effects of infrastructure projects and planning offsetting measures if such effects are expected to be negative.

• 5 Consistent with the aggregate evidence, firm level data show that the share of firms that indicate (...)

12Progress has not been uniform across sectors, partly because differences in the regulatory frameworks and institutional quality across sectors and countries could impede private investment (Banerjee et al., 2006; Cubbin and Stern, 2006). Information and communication technology (ICT) has expanded dramatically, with the number of Internet servers growing from near zero in 2005 to the average of 6 servers per million people in 2015. Electricity generation per capita has increased by 57 per cent on average, with over 300 per cent increase in a few countries, such as Bhutan and Vietnam.5 Access to improved water and sanitation facilities rose on average by around 20 per cent from 2000 to 2014. On the other hand, improvements in transport infrastructure have been relatively minor, even though transportation is typically the largest item in LIDC capital budgets. Firm-level data from the World Bank Enterprise Survey confirm these trends, as the share of firms identifying electricity and water insufficiencies as major constraints to their business activity sharply decreased over the last decade, while almost no progress is observable on transportation infrastructure (Table 1, bottom panel).

• 6 The analysis in this section is based on 47 LIDCs for which the IMF’s World Economic Outlook databa (...)
• 7 The current median level of public investment in LIDCs is similar to that observed in the present-d (...)

16Public investment in LIDCs is higher as a percentage of GDP than in emerging and advanced economies and has followed a general upward trend since 2000, first surging before the Global Financial Crisis (GFC) and then picking up again until 2015.⁶ By contrast, trends in emerging markets and advanced economies had been downward sloping in the 2010s. Median public investment in LIDCs rose significantly from 5.5 per cent of GDP in 2000 to a peak of 7.1 per cent of GDP in 2010. Following a temporary slowdown in 2011, public investment picked up again and stood at 6.7 per cent of GDP in 2015, before declining to 6.4 per cent in 2016 (Figure 3).7 As documented in the previous section, this scaling-up has resulted in a broad enhancement of economic infrastructure in LIDCs, although this relationship is far from tight and exhibits significant variation across countries and sectors. Moreover, a large gap still remains compared to emerging and advanced economies.

• 8 Five-year averages are used to smooth over yearly fluctuations.

18Diversity is notable not only between but also within groups. In every category, one can find examples of countries that achieved or maintained high public investment levels and examples of those that failed to do so. A large majority increased the public investment-to-GDP ratio in the 2011-2015 period compared to 2001-05 (Figure 5).⁸

• 9 High public investment in Ethiopia reflects the government’s national development agenda with a foc (...)

19Some countries stand out with substantial scaling-up, with highest levels reached in Djibouti, Congo (Alter et al., 2017, discuss Congo’s experience) and Ethiopia9, which have been pursuing national development agendas centred on improving infrastructure. Public investment rose steadily in several commodity exporters, including Bolivia, Mongolia, Mozambique, Niger, and Tajikistan, until a drop in 2014-15 following a negative commodity price shock. However, in some other countries, the ratio of public investment to GDP has declined significantly over time, reflecting, for example, intensified fragility in Eritrea and Yemen, and fiscal pressures in Nigeria and Uzbekistan. A few countries have not experienced a pronounced scaling-up, but have maintained fairly high levels of public investment throughout the past 15 years. For example, Bhutan and Vietnam averaged 13 and 9 per cent of GDP, respectively, since 2000. On the other hand, in several countries, public investment has been quite low over the whole period (e.g., never exceeding 5 per cent of GDP in Nepal, primarily because of implementation capacity constraints and frequent government turnover).

• 10 Among the 33 countries where the public investment/GDP ratio increased between 2001-05 and 2011-15, (...)
• 11 The correlation coefficient between median public investment and saving is 0.74 with p-value of 0.0 (...)

20Public saving has generally not been scaled up commensurately with the increase in public investment. Over the last decade and a half, there has been a clear correlation (with correlation coefficient of 0.68) between changes in public investment and in public saving (Figure 6). However, the former was greater than the latter in most countries, especially in most recent years.¹⁰ As a result, the gap between public investment and saving—which narrowed just before the GFC—started widening in subsequent years, indicating increasing recourse to debt financing (Figure 7).11 Median public saving as a share of GDP rose 2.9 percentage points between 2000 and 2007—twice as much as public investment. Median public saving declined sharply during the GFC, and, after a brief rebound, started slipping again, with the latest slide reflecting lower commodity prices. As a result, median public saving has dropped 2.4 percentage points of GDP since its 2007 peak, returning to levels from the early 2000s, even as median public investment eked out a small increase. In 2015, public investment exceeded public saving in 42 out of 46 LIDCs and the gap between median public investment and median public saving reached 4.8 per cent—the widest it has been since 2000.

• 12 According to the latest debt sustainability analyses conducted by IMF and World Bank country teams, (...)

21In the most recent years, the negative public saving-investment balances have contributed to higher government debt-to-GDP ratios, following a notable drop in debt ratios in most LIDCs over the course of the 2000s, mostly driven by multilateral and bilateral debt relief initiatives (Figure 9). Fiscal vulnerabilities have increased recently, particularly among commodity exporters. Budget deficits have gone up, interest rates have risen, and local currency depreciation has increased the burden of external debt. As a result, the median general government debt ratio went up from 34 per cent in 2013 to 43 per cent in 2016.12 In some frontier markets, rising debts are also the result of access to international capital markets: since 2010 LIDCs issued more than USD 22 billion in sovereign bonds, in many cases with the aim of using part of the proceeds to finance new infrastructures (Presbitero et al., 2016). For instance, in 2014 Ethiopia issued a USD one billion Eurobond to finance imports related to export-oriented projects such as investment in the power transmission infrastructure, sugar factories, and the development of industrial parks (IMF, 2015). More recently, in May 2017 Senegal issued its third Eurobond (USD 1.1 billion) with the intent to finance a series of infrastructure and power production projects.

• 13 See Annex Table A1 for details. For some countries, coverage included only budgetary central govern (...)

22As noted above, internationally comparable data on infrastructure investment for a broad set of LIDCs is lacking. To fill that gap, we have conducted a survey of the IMF’s LIDC country desks. Thirty-two country teams were able to provide information on public investment in economic infrastructure over the last five years, typically in consultation with the authorities.13 Twenty-three of them had data by sector. This information offers valuable insights, even though the results should be taken with a grain of salt as quality and comparability of data cannot be assured.

• 14 The correlation between public infrastructure investment and total public investment is 0.8 in this (...)

23For a median LIDC in the sample, investment in economic infrastructure accounted for about half of total public investment.14 The median investment level stood around 3 per cent of GDP in 2011–14, but dropped below 2½ per cent in 2015 as commodity exporters were hit by falling export prices (Figure 8). Looking across country groupings, frontier market economies had somewhat higher levels of investment, facilitated by easier access to financing and stronger economic prospects. Investment levels in fragile states were typically lower than average, likely reflecting limited fiscal space and weak institutional capacity (Collier and Cust, 2015).

• 15 The World Bank’s PPI database (Ruiz-Nunez and Harris, 2016) records total investment in infrastruct (...)
• 16 Lao PDR stands out as the LIDC with the largest volume of infrastructure PPPs. This can be largely (...)

25Private participation in infrastructure investment is quite limited in LIDCs. Since 2000, LIDCs accounted for 6.5 per cent of the value and 10.5 per cent of the number of Public-Private Partnership (PPP) projects in all emerging market and developing economies (Figure 9).15 In the last five years PPP volume amounted on average to about 0.4 per cent of LIDC GDP—a ratio similar to EMs. After a sharp acceleration in the early 2010s, PPP flows have declined in the most recent years. Of the USD 43 billion in LIDC PPP projects since 2010, more than half has been invested in Asia and one third in Sub-Saharan Africa—see also the recent report by the Infrastructure Consortium for Africa (ICA, 2017) for recent data and examples of large PPP projects in Africa. Vietnam and Bangladesh have the largest number of projects (Table 2), while Lao PDR is an undisputed leader in terms of volume.16 Public-private partnerships have also been used to finance regional projects. Across Africa there are several examples of regional infrastructure projects, especially in the energy and transport sectors (UNCTAD, 2016). For instance, the Central Corridor is an integrated transport program across five countries (Burundi, DR Congo, Rwanda, Tanzania, and Uganda) with an investment of about USD 18 billion involving local and international actors from the public and private sectors (WEF, 2015).

• 17 Changes in methodology amplify the rise and fall in telecom PPPs in Figure 10. Coverage of the tele (...)
• 18 Micro-grids are an example of private production and distribution of electricity. Micro-grids deliv (...)
• 19 At the regional level, in Asia the vast majority of PPI are in the energy sector, while in Sub-Saha (...)

26Private participation varies greatly across sectors (Figure 10). The telecom sector attracted considerable private participation in the 1990s, following liberalisation and technological advances, and has ultimately moved toward mostly purely private provision (particularly for mobile services), with the government’s role limited to regulation and licensing.17 There is very little purely private provision of infrastructure services outside the telecom sector, although some small-scale successful models can be found, and private sector involvement is channelled predominantly via PPPs.18 Currently, the energy sector attracts the bulk of PPPs, with transportation a distant second, and a small share allocated to water and sanitation.19 This likely reflects the fact that it might be easier—both technically and politically—to charge end users for electricity than for roads or water (Swaroop, 1994). The vast majority of projects are greenfield projects (87 per cent since 2000) and brownfields (8 per cent) and almost all the contracts (97 per cent) have been with the central government. There is considerable variation in the size of PPP projects, and some of them are very large, such as a coal plant in Laos with an investment of USD 3.7 billion, the expansion of the Onne port complex in Nigeria (USD 2.9 billion), and a thermal power generation project in Vietnam (USD 2 billion). Nine projects started since 2010 are valued over USD 1 billion.

• 20 On average, about 5 per cent of the projects recorded in the PPI database are cancelled or under di (...)

27Multilateral development banks (MDBs) are involved in a significant share of PPPs to provide operational assistance, financial support and risk mitigation (Ruiz-Nunez and Harris, 2016). More than a quarter of the projects in LIDCs involve MDB support in the form of direct loans, syndication, equity investment, partial credit guarantees, and political risk coverage. The presence of MDBs is associated with a lower probability that a project comes under distress or is cancelled, even after controlling for a set of project-specific variables and for year and country fixed effects.20 This likely reflects a combination of careful project selection by MDBs and the impact that MDB involvement—through a thorough preparation and a strengthened oversight—has on the quality of the project (Jandhyala, 2016).

• 21 Multilateral support accounted for 57 per cent of ODF, bilateral for 43 per cent. The World Bank is (...)
• 22 According to Miyamoto and Chiofalo (2016), ODF covers 6-7 per cent of infrastructure investment in (...)

28Official development finance (ODF) is a major source of infrastructure financing in LIDCs. Detailed data obtained from OECD show that LIDCs received nearly USD 17 billion in project finance from MDBs and OECD members in 2014.21 While the total value of infrastructure investment in LIDCs is not known, ODF certainly covers a much larger share of investment in LIDCs than in other developing countries.22 In this respect, the Aid for Trade Initiative disbursed over 300 billion USD since 2007—27 per cent of which was directed to least developed countries—to improve trade connectivity through ICT, physical and network infrastructures (OECD/WTO, 2017). Moreover, 87 per cent of ODF for LIDCs consisted of grants and concessional loans, in contrast to only 56 per cent for all developing countries. The bulk of the money went to public projects, with direct support to the private sector amounting to USD 0.9 billion. The share of projects in water and transportation sectors in total infrastructure ODF declined steadily since 2006, while the share of energy increased to about 30 per cent in 2014 (Figure 11).

• 23 The figures of the box are based on the AidData’s Global Chinese Official Finance Dataset, Version (...)
• 24 Figures based on realised foreign direct investment of China show that China only accounted for aro (...)

29Some emerging donors, notably China and India, have also become important providers of infrastructure financing to LIDCs. These countries direct a considerable share of their development financing to infrastructure. China, in particular, committed billions of dollars of infrastructure investment under the “Belt and Road” initiative, an ambitious plan to boost trade and global development, strengthening the links between Asia, Europe, and Africa. According to the data on Chinese official development assistance published by AidData, between 2000 and 2014 infrastructure projects accounted for about 70 per cent of total Chinese financing.23 Gutman et al. (2015) calculate that China contributes about 20 per cent of external finance for infrastructure projects in Sub-Saharan Africa, with most of that financing provided by China's EXIM Bank.24 India’s development financing for infrastructure is estimated at USD 1.3 billion in 2014, with most of it going to neighbouring countries, primarily for energy and transportation. There is considerable dispersion across countries in the amounts of ODF received. In 2014, for all LIDCs, the median ratio of ODF to GDP equalled 1.3 per cent, the simple average 2.0 per cent, and the GDP-weighted average 0.9 per cent. As expected, grants accounted for the bulk of financing in fragile states, while frontier markets and commodity exporters received less ODF (relative to their GDP) than other country groups as they have a higher domestic revenue base and greater access to commercial borrowing.

• 25 A related issue is a frequent failure to allocate adequate funds to maintenance (Adam and Bevan, 20 (...)

33The path forward is not easy and the scope for increasing public investment in LIDCs is rather limited, even though the wide range of investment ratios shows that many countries may have some room for scaling up. Over recent years, public debt levels have risen, external financing conditions have tightened, and growth prospects have weakened for the LIDCs. These trends create a challenging environment for infrastructure investment. Countries with fiscal space should seek financing on the most concessional terms possible, with the support from the international community. Especially for countries where fiscal space is limited (but also for the others) there is a need to increase the efficiency of public investment—and considerable scope for it exists. The link between the amount of public investment (the input) and the quantity and quality of infrastructure in a country (the outputs and the outcomes) is not very tight and, although many factors may contribute to this variance, differences in investment efficiency are likely one of them. Several studies (Dabla-Norris et al., 2012) show that low-income countries have relatively weak public investment management institutions, and that improving those institutions could increase considerably the efficiency (i.e., the “value for money”) of public investment.25 In addition, mobilising domestic revenues and prioritising expenditures could provide more sustainable and reliable sources of development funding.

• 26 These include sound planning and project selection; strong fiscal institutions; strong legal framew (...)

39The balance between public and private financing depends to a large extent on the country context and, in particular, on the institutional weaknesses that are felt most acutely, as government could be affected by limited commitment, limited accountability, limited capacity, and limited fiscal efficiency (Estache et al., 2015). Macro-fiscal implications of PPP projects could be large and expose countries to fiscal risks. Thus, a strong regulatory environment and a robust institutional framework26 are essential to implement PPP infrastructure projects in a sustainable and efficient way, especially in developing countries, where public sector capacity constraints may be more severe (Romero, 2015). PPP use is correlated with domestic institutions, such as the rule of law and levels of corruption (Moszoro et al., 2014). In a broad sample of emerging market and developing economies, there is a positive association between PPP investment as a percentage of GDP and the Infrascope index developed by the Economist Intelligence Unit (EIU, 2010) to evaluate countries’ capacity to deliver efficient and sustainable infrastructure projects (Figure 14). The average Infrascope index for LIDCs is significantly lower than the one of EMs, and the gap is particularly strong for the legal regulatory framework and for the presence of financial facilities. A heavy reliance on external financing and lags in the implementation of the PPP legal framework have been identified as key constraints for the growth of PPPs in sub-Saharan Africa (EIU, 2015). In that perspective, there is scope to improve the collaboration between local governments and MDBs in the preparation, structuring and financing of infrastructure projects, to facilitate the participation of private long-term investors—the World Bank Global Infrastructure Facility and the EBRD Equity Participation Fund are infrastructure platforms that go in that direction (Arezki et al., 2017).

• 27 See https://goo.gl/zGh2Bk for list of available instruments.

40Beyond the PPPs, the large pool of resources held by institutional investors (such as insurers, pension funds, and sovereign wealth funds)— estimated at USD 88 trillion (McKinsey Global Institute, 2016)— could be an important source of private financing of infrastructure. Investment in infrastructure is attractive for institutional investors, as they look for stable, long-term, inflation adjusted return higher than government bonds to match their long-term liabilities. However, infrastructure investments are exposed to high market and non-market risks (Weber et al., 2016). The role of MDBs is thus crucial in de-risking infrastructure investment and crowding in private investment including from institutional investors through their various instruments27 such as partial risk guarantee and provision of subordinate loans. The African Development Bank (AfDB) partial risk guarantee for Kenya’s Lake Turkana Wind Farm Project in 2013 is an interesting example. The guarantee covered a potential delay in the construction of the transmission line by the government; off-take risks due to non-payment of monthly invoices; and termination of the power purchase agreement by the off-taker. The guarantee improved the risk-return mix and facilitated access to long-term debt and ‎financial close for the power plant (AfDB, 2014).