Scholarly article on topic 'Does infrastructure really explain economic growth in Sub-Saharan Africa?'

Does infrastructure really explain economic growth in Sub-Saharan Africa? Academic research paper on "Economics and business"

Share paper
Academic journal
Review of Development Finance
OECD Field of science
{"Infrastructure access" / "Infrastructure stock" / "Infrastructure quality" / "Economic growth" / "Sub-Saharan Africa"}

Abstract of research paper on Economics and business, author of scientific article — Odongo Kodongo, Kalu Ojah

Abstract In light of Africa’s palpable deficit in public infrastructure, we use System GMM to estimate a model of economic growth augmented by an infrastructure variable, for a panel of 45 Sub-Saharan African countries, over the period 2000–2011. We find that it is the spending on infrastructure and increments in the access to infrastructure that influence economic growth and development in Sub-Saharan Africa. Interestingly, these significant associations, especially those of infrastructure spending, are more important for lesser developed economies of the region than for the relatively more developed economies, which uncommonly have better than near-zero access to infrastructure. In addition to these robust direct links between the target variables, we find importantly that infrastructure access, and quality, also relate to economic growth indirectly via export diversification (trade competitiveness), and cross-border capital flows and trade competitiveness, respectively. Among other important policy derivatives of our findings, we emphasize that efforts aimed at reversing Africa’s pervasive infrastructure deficit, in ways that enable economic growth and development, must be carefully nuanced.

Academic research paper on topic "Does infrastructure really explain economic growth in Sub-Saharan Africa?"



Available online at


Review of Development Finance xxx (2016) xxx-xxx

Review of



Does infrastructure really explain economic growth in Sub-Saharan Africa?

Odongo Kodongo *, Kalu Ojah

Wits Business School, 2 St. David's Place, Parktown, Johannesburg 2193, South Africa


In light of Africa's palpable deficit in public infrastructure, we use System GMM to estimate a model of economic growth augmented by an infrastructure variable, for a panel of 45 Sub-Saharan African countries, over the period 2000-2011. We find that it is the spending on infrastructure and increments in the access to infrastructure that influence economic growth and development in Sub-Saharan Africa. Interestingly, these significant associations, especially those of infrastructure spending, are more important for lesser developed economies of the region than for the relatively more developed economies, which uncommonly have better than near-zero access to infrastructure. In addition to these robust direct links between the target variables, we find importantly that infrastructure access, and quality, also relate to economic growth indirectly via export diversification (trade competitiveness), and cross-border capital flows and trade competitiveness, respectively. Among other important policy derivatives of our findings, we emphasize that efforts aimed at reversing Africa's pervasive infrastructure deficit, in ways that enable economic growth and development, must be carefully nuanced.

©2016 Africagrowth Institute. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (

JEL classification: H54; O11; 040; O55

Keywords: Infrastructure access; Infrastructure stock; Infrastructure quality; Economic growth; Sub-Saharan Africa

1. Introduction1

An AfDB/UNDP/OECD (2014) report called for a comprehensive strategy to promote sustainable inclusive growth in Africa. Specifically, the report called on countries to, among others, ensure political and macroeconomic stability, create favorable conditions for doing business, encourage the proper functioning of financial markets, and provide adequate human capital and infrastructure investment. Calls of this nature, coming from the African Development Bank (AfDB), currently an important infrastructure financier in Africa, might be construed as self-serving and intended to safeguard the bank's key business segment. This paper is therefore an attempt to understand whether investment in public infrastructure does indeed play an

* Corresponding author.

E-mail addresses: (O. Kodongo), (K. Ojah).

1 Parts of the background information draw from a commissioned work

done by Odongo Kodongo for the African Capacity Building Foundation: Occasional%20Paper%2025%20English.pdf.

important role in fostering economic growth in the Sub-Saharan Africa (SSA) region.

Public infrastructure (defined in this paper as physical installations such as highways and roads, airports, telecommunication facilities, water supply systems, electricity, waste treatment facilities and the like) is believed to provide services that form a part of residents' consumption bundles and augments capital and labor as an input in the production process (Ayogu, 2007). Access to infrastructure provision promotes human development, and betters quality of life through improved productivity and sustainable economic growth (Sanchez-Robles, 1998; Egert et al., 2009; Ajakaiye and Ncube, 2010). Specifically, public infrastructure provisioning may enhance trade and commerce (Mbaku, 2013) and play an important role in alleviating poverty and inequality (Ndulu, 2006; World Bank, 2006). We view infrastructure deficit as an important binding constraint on some traditional antecedents of economic growth, in addition to the fact that it may also directly enable growth.

Despite its apparent importance, public infrastructure stocks of most African countries are believed to be way below the capacity required to support their required levels of production. At the end of the 2000s decade, annual infrastructure invest-

1879-9337/© 2016 Africagrowth Institute. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (

Fixed line telephone (per 1000 people)

O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx-xxx

Internet density (per 1000 people)

2007 2008 2009 2010 2011 2012 2013 EAP ECA LAC MNA SSA

15 10 5 0

Mobile telephone (per 1000 people)

2007 2008 2009 2010 2011 2012 2013 EAP ECA LAC MNA SSA

2007 2008 2009 2010 2011 2012 2013 EAP ECA LAC MNA SSA

Improved sanitation (% of popn. with access)

Clean water (% of population with access)


2007 2008 2009 2010 2011 2012


2007 2008 2009 2010 2011 2012


Electricity production (MWh per million people)

2007 2008 2009 2010 2011


Fig. 1. Trends in SSA infrastructure deficit.

The figure covers the following developing regions: SSA - Sub-Saharan Africa; MNA - Middle East & North Africa; LAC - Latin America & Caribbean; ECA -Developing Europe & Central Asia; EAP - East Asia & Pacific.

Source: Authors' construction using World Bank's World Development Indicators data.

ment needs for Africa was estimated at USD 93 billion (15% of GDP), a third of which would be for operations and maintenance in various sectors (Foster and Briceno-Garmendia, 2010; AfDB, 2011). It was also estimated that annual expenditure on infrastructure going forward would range from 9% to 13% of GDP (Sachs et al., 2004; ECA, 2005).

To put the region's infrastructure deficit to perspective, we present in Fig. 1, a comparative evolution of infrastructure endowments for the world's developing regions over 2007-2013. The figure shows that the SSA region trails its peers on every key metric. For instance, fixed line telephone coverage has stagnated over time at below 0.5 lines per 1000 people compared to other developing which have at least 1.5 lines per 1000 people. One may argue that the apparent low fixed line telephone reach has been partially mitigated by the expanding mobile telephone coverage. However, even on the mobile telephony score, as Fig. 1 shows, SSA fares poorly relative to other regions. In fact, African Union (2014) estimates that internet penetration rate for SSA was only 6% in 2012, substantially lower than developing regions' average penetration rate of 40%. If the region were to achieve universal rural access, for both voice service and limited broadband service based on WiMAX technology,2 it would need an investment of USD 1.7 billion a year. Overall, the estimated ICT sector annual investment need is USD 9 billion, including USD 2 billion for maintenance (Foster and Briceno-Garmendia, 2010).

Power appears to be the worst affected infrastructure sector in SSA. In addition to the evidence in Fig. 1, various accounts have documented the region's critical power shortage situation: over a five-year cycle (2008-2012), SSA's total primary energy consumption increased only by 0.9% from 10.23 QBtu to 10.32 QBtu and its share of the world energy consumption was only 1.9%.3 In most countries, a good proportion of power generation is from hydro-power, coal and gas plants, with the more expensive thermal generation (diesel turbines) often used to boost supply during low base load generation. To meet suppressed demand, provide additional capacity and support projected economic growth, installed electricity production capacity would have to grow more than 7 GW per year, requiring annual funding of USD 41 billion over 5 years, with USD 14 billion for operations and maintenance (Eberhard et al., 2011).

Per available estimates, SSA has ample water resources that are, however, underdeveloped, unsustainably managed and underutilized, with only 5% of agriculture under irrigation (African Union, 2014). UNICEF data4 show that, as of 2012, SSA, with 64% access to improved drinking water, lags behind all regions and also falls below the least developed countries average of 66% (Fig. 2). To deal with this dismal performance, Africa Water Vision 2025 called for increased development of

2 WiMax is a standardized wireless version of Ethernet intended primarily as an alternative to wire technologies (such as Cable Modems, DSL and T1/E1 links) to provide broadband access.

3 "QBtu" refers to quadrillions of British Thermal Units. All World Bank data used in this paper were accessed from World Development Indicators on various dates between March and May 2015 on the website: =world-development-indicators.

4 Accessed from on April 12, 2015.

O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx—xxx

Developed areas Latin America and Caribbean South East Asia East Asia ex China

Sub-Saharan Africa , —

-20 0 20 40 60 80 100 120

Perecentage Change, 1990-2012 Total improved drinking water (%)

Fig. 2. Access to improved water, by region. Source: Authors' construction using WHO/UNICEF data.

water resources by 10% in 2015 and 25% by 2025 to meet increased demand from agriculture, power, tourism and transport. Annual funding gap in the sector was estimated at USD 11 billion (AfDB, 2011).

In the transport sector (not covered in Fig. 1), SSA's road network comprises strategic trading corridors of not more than 10,000 km that carry about USD 200 billion of trade annually; road access rate is only 34% compared to 50% in other developing regions and transport costs are 100% higher than in other regions (African Union, 2014). The region has ambitions for a Trans-African Highway, which remains a pipedream due to poor maintenance on key segments. Such a network would require construction of 60,000-100,000 km of paved road (Foster and Briceno-Garmendia, 2010). The region's road density (204 km of road per 1000 sq. km of land area) is substantially lower than the world average of 944 km and is less than 30% of the next-lowest region, South Asia. As of 2011, only about 15% of SSA's roads were paved compared to 26% in Latin America and the Caribbean, 65% in East Asia and the Pacific, 76% in Middle East and North Africa, and 86% in developing Europe and Central Asia (computed using World Bank data).

Railway infrastructure is not that much better: the spatial density, which compares track mileage with the size of a country, ranges from 1 to 6 for most countries (16 for South Africa); 13 countries have no operating railway at all (Bullock, 2009). A transport network with sufficient regional, national, urban, and rural road connectivity, accompanied by adequate rail, port, and airport infrastructure, requires spending of approximately USD 18 billion a year, half of which would be for maintenance only (Foster and Briceno-Garmendia, 2010).

SSA governments and their development partners are cognizant of the region's colossal infrastructure development need and huge funding gap. It is understood, however, that mobilization of investment funds for various public infrastructure sectors is made difficult by the region's financially weak public institutions that, collectively, have an aggregate annual revenue shortfall of approximately USD 8 billion (AfDB, 2011). Growth economists and policy-makers are concerned that continued underinvestment in infrastructure might derail development efforts in various economic sectors and adversely impact regional economic growth. Although agreement exists on the need to improve SSA's infrastructure endowment if longer term impact on economic growth is to be realized, there appears not to be a clear understanding of the nature of the relationship between infrastructure and economic growth. Our postulation, however,

is that infrastructure is to a large extent both an enabler of growth antecedents as well as, if not more so, a direct antecedent of growth in the SSA kind of environment.

We explore an initial mapping of the possible nature of the relationship between several infrastructure indicators and aggregate income, using energy consumption and number of telephone lines as two key infrastructure metrics. Fig. 3 graphs a 12-year (2000-2011) relationship between energy consumption per capita and GDP per capita for several SSA countries. Energy consumption data is obtained from the USA Energy Information Administration (EIA) database; GDP per capita is from World Bank's WDI. In general, it is reasonably evident in the figure that countries, such as Seychelles, with higher levels of infrastructure also have higher income levels; similarly, countries with low infrastructure levels also have low income levels. Subsequently, we stack the countries into a panel and run a pooled correlation - the coefficient of 0.66 indicates a potential positive relationship between energy consumption and income; however, the correlation coefficient between telephone lines and income is relatively low at 0.37. These two observations, combined, point to the possible impact infrastructure provision could have on economic growth in SSA.

The literature has proposed several plausible theoretical explanations of infrastructure's impacts on economic growth. One school of thought regards infrastructure as part of a country's physical stock of capital and therefore a factor of production (e.g., Aschauer, 1993; Gramlich, 1994). Proponents of this school argue that infrastructure stock changes impact national output and directly induce economic growth. The second viewpoint is that infrastructure complements other factors of production: the argument here is that infrastructure may improve total factor productivity by lowering input costs or by expanding the production frontier or the set of profitable investment opportunities (e.g., Barro, 1990).5

In the third school of thought, infrastructure is believed to stimulate the accumulation of factors of production. Advocates of this view argue that accumulation and productivity of a factor (e.g., labor) is incentivized by infrastructure (e.g., educational and health facilities and roads to access those facilities) (Fedderke and Garlick, 2008). In this sense, infrastructure indirectly affects economic growth. Furthermore, it has also been argued that infrastructure investment can affect economic growth by stimulating aggregate demand or by directing industrial policy toward a desired path.6

Our postulation in this paper aligns with these latter views on the nature of relations between infrastructure and economic growth. This view is particularly more germane for less developed regions, such as SSA, where traditional antecedents of economic growth are either significantly inadequate or lacking.

5 For instance, availability of safe Internet may increase speed of communication and timeously open up profitable trade opportunities just like reliable electricity may reduce manufacturers' energy (electricity) costs.

6 Opponents of these arguments have question its efficacy as a policy tool pointing out that expanding aggregate demand without an accompanying change in productive capacity may cause long-term inflation (e.g., Canning and Pedroni, 2004).

O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx-xxx



' ? "I

jBurkina FJso ^^ur



Energy consumption per capita (million BTU)



FSaosuoth Africa


• Cami^on



Telephone lines per 1GG people

Fig. 3. Energy consumption, telephone lines and nominal GDP per capita in SSA. Source: Authors' construction. GDP and telephone lines data are from WDI; energy data are from the USA Energy Information Administration (EIA).

Therefore, an erroneous consideration of infrastructure as only directly related to growth may yield insignificant estimate or a weak reflection of infrastructure's effects on growth. As a result of this tendency, care must be taken in specifying appropriately the test of the relation of infrastructure to economic growth (i.e., considering both direct and, the more plausible, indirect effects).

Empirical tests of the effects of infrastructure on economic growth use various econometric specifications that depend on the underlying theoretical argument(s). In the Sub-Saharan Africa region, the preponderance of evidence from various studies of individual countries (e.g., Reinikka and Svensson, 1999; Fedderke et al., 2006; Estache and Vagliasindi, 2007) and panels (Estache et al., 2006; Calderón and Servén, 2010) point to the SSA's lost growth opportunities which they variously attribute to low infrastructure investment that has jeopardized the region's international competitiveness, increased the cost of doing business, inhibited foreign direct investment and derailed trade (emphasis ours). These explanations imply an indirect link between infrastructure development/access and economic growth through the highlighted variables. We test the indirect relationship empirically. Boopen (2006) likewise presents panel estimates of the output contribution of transport infrastructure using a similar dataset as Estache et al. (2006) and Calderón and Servén (2010). Further, geographical impediments, such as landlocked nature of many SSA countries, which might potentially be redressed through adequate provision of transport and telecommunications infrastructure, puts the region at a disadvantage in attracting trade and investment (Acemoglu et al., 2002; Redding and Venables, 2004; Cinyabuguma and Putterman, 2011).

South Africa has attracted more attention in this literature, partly reflecting the significantly better quality of its data relative to that of other countries in the region. Fedderke et al. (2006) use a detailed database on infrastructure investment and capital stocks, spanning as long as a 100 years, to test for the existence

of a long-run relation between different infrastructure measures and GDP. Their results suggest a bidirectional relation in most cases. Kularatne (2006) explores the effects of economic infrastructure investment (as well as social spending on health and education) on GDP. He also finds bidirectional effects, although the impact of infrastructure investment appears to occur indirectly through private investment. Dinkelman (2011) finds a significant impact of household electrification on employment in South Africa's rural labor markets.

Most of the Africa-focused studies have sought to understand the effects of infrastructure investment on economic development on a country-by-country basis. However, a majority of them focus on one element of infrastructure (e.g., telephone, roads) in disregard of the multidimensional nature of public infrastructure and commonly find that infrastructure stocks are positively related to economic development (measured as levels of income). In this study, we take a different approach: first, we examine both the direct and indirect relationships between economic infrastructure and economic growth for 45 countries in SSA as a group using alternative measures of infrastructure that combine several of its dimensions. In contrast to the only known study in the literature that has examined the relationship between infrastructure stocks/quality and economic growth (Calderón and Servén, 2010), our study explores the more plausible indirect effect of infrastructure on economic growth, in addition to the direct effect, which has been almost the exclusive focus of past related studies.

Further, Calderón and Servén (2010) deal with the issue of cyclicality by using 5-year rolling averages of infrastructure variables. Our view is that this approach causes smoothing in the infrastructure data and, in cases of missing data (an unavoidable problem for African countries for which data are always scarce), distorts observations by repeating one year's values for several years. To deal with this important concern, we use non-averaged annual data to address this apparent data distortion.

O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx—xxx 5

Furthermore, their study does not examine the possible linkage between infrastructure spending and economic growth; neither does it examine the relationship between incremental infrastructure stocks/quality and economic growth. Our study attempts to fill these gaps. Importantly, in an unprecedented analysis, we explore potential channels (highlighted in our foregoing literature review) through which economic infrastructure might impact economic growth in the SSA region; we believe that such nuanced relationship is more plausible in the African context.

To mention briefly, we document, contrary to the extant literature, that rather than the sheer stock and/or quality of infrastructure being relevant for economic growth in the SSA region, which is so legendarily bereft of basic infrastructure, it is the spending on infrastructure and increments in access to infrastructure that influence growth. Interestingly, these significant associations, especially those of infrastructure spending, are more important for lesser developed economies in the region than for the relatively advanced economies, which typically have better than near-zero infrastructure endowments.

In addition to these robust direct links between the target variables, we find importantly that infrastructure access, and quality, also relate to economic growth indirectly via trade competitiveness (export diversification), and cross-border capital flows and export diversification, respectively. This finding supports our postulation that analysis of the infrastructure—growth nexus ought to be a bit more nuanced than simply examining only their direct association. Among factors that have been documented in the literature as relevant growth factors, we found human capital and institutions to feature consistently significantly in most of our model estimations.

2. Empirical design

2.1. Determinants of economic growth

Neo-classical models of economic growth predict that countries with lower initial income levels relative to their long-run steady-state position should grow faster - i.e., the convergence hypothesis. This prediction derives from diminishing returns to capital: economies with lower capital per capita relative to their steady state tend to have higher rates of return on capital (Barro, 1996). The theoretical argument is that poorer countries lack in some critical exogenous factors, which, once supplied, cause exceptionally high returns on capital during the catch-up period (Roberts and Fagernas, 2004); and they can draw from international capital markets to supplement domestic savings (Rodrik, 2014). These predictions have been confirmed by some studies (e.g., Ghura and Hadjimichael, 1996) which also found that male secondary and higher education, life expectancy, fertility rates, the terms of trade and external competitiveness, and democracy and the rule of law, all have significantly positive impacts on per capita GDP growth and investment expenditure; and that inflation and government consumption (excluding expenditure on education and health) have significantly negative effects.

In another strand of the literature, economists have tried to explain the role of accumulation of factors of production (especially capital) in economic growth. Easterly and Levine (2001)

have observed that the accumulation of factors of production, including human capital, tends to be persistent and progressive through time, while economic growth seems irregular over time and similarities in the accumulation of factors among countries have not been matched by convergence in the countries' growth rates. In contrast to this observation, some researchers argue that growth can be explained by changes in total factor productivity; in fact, several studies find that a host of policy variables (e.g., openness, government expenditure, macroeco-nomic stability, control of corruption, bureaucratic effectiveness, and mitigation of expropriation risk), which might improve the efficiency of factor-utilization, significantly explain economic growth (Ghura and Hadjimichael, 1996).

Other variables that have been found to have a significant effect on growth include natural resource endowment, distance from external markets (Redding and Venables, 2004), transport costs, climate, disease incidence (Sachs and Malaney, 2002) and institutional quality, proxied by a composite index of the rule of law and property rights (Rodrik et al., 2002). Rodrik (2003) argues that initiating growth does not necessitate a set of growth-inducing policies and institutions; rather, a government's favorable attitude to the private sector and the elimination of impediments to enterprise, the creation of domestic policies and institutionalization of respect for property rights and contracts, sound monetary and fiscal management, and market-oriented incentives are important for determining sustainable growth. However, Rodrik (2014), analyzing the recent high growth in African economies, argues that although institutional quality can generate greater economic stability and prevent deep crises from arising out of past mismanagement, institutional quality may not, on its own, be deemed a driver of economic growth. We revisit this point briefly in our discussion of findings.

Financial development has also received a lot of attention in economic growth literature. For instance, Calderón and Liu (2003), using Geweke decomposition test on pooled data of several developing and industrial countries, find that financial development generally leads economic growth, with stronger effect observed for developing countries than industrialized ones. Importantly, their findings suggest that financial deepening works through rapid capital accumulation and productivity growth to influence economic growth; the influence through productivity growth being the stronger of the two. Similarly, Hassan et al. (2011) find a positive relationship between financial development and economic growth in developing countries; they conclude that a well-functioning financial system is a necessary but not sufficient condition to reach steady economic growth path in developing countries. In Africa, Nyamongo et al. (2012) find that Diaspora remittances are not only an important source of economic growth but also appear to complement financial development in enabling it.

In an important recent critique of the economic growth literature, Mthanti and Ojah (2016) argues that institutional variables are partially shaped by local context and culture which makes it difficult to predict their temporal evolution; they propose aggregate entrepreneurship orientation, a mélange of aggregate risk taking, innovation and proactive-ness, as an important economic growth determinant. Additional critique is that variables such

O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx—xxx

as institutional quality, infrastructure investment and financial development might as well be explained by economic development dynamics, making them endogenous to the growth process (Glaeser et al., 2004; Lee and Kim, 2009; Rodrik, 2014).

2.2. Econometric specification

Generally, theoretical economic growth thinking begins with the Solow model, which explains aggregate income by aggregate capital and labor. And because capital exhibits diminishing marginal returns, long-run growth is explained by population growth and technological progress, both of which are exogenous. This general model, which has been revised to include several variables, notably government spending (infrastructure), human capital, protection of property rights and market distortions (see Barro, 1996), has been criticized on several grounds including its failure to explain technological progress and cross-country income differences. These deficiencies have motivated the development, and burgeoning empirical applications, of endogenous growth models. Barro (1990), one of the earliest contributors to theoretic endogenous growth modeling, argues that the government's contribution to current production is driven by its flow of productive (infrastructure) expenditure, which can prevent diminishing private-sector capital returns, raise the marginal product of private-sector capital, and these in turn raise the rate of output growth. This motivates our study's focus on incremental infrastructure investment/development.

Thus, to appreciate the effect of infrastructure development/investment on economic growth, we formulate, in the spirit of Barro (1990), an endogenous economic growth model of the form:

yit = ayt,t -1 +P'Xtt + £it, (1)

where the disturbance term eit = qi + eit, such that n are individual fixed effects and eit are the idiosyncratic shocks. y is the natural log of real GDP per capita; X is a vector of growth determinants as discussed in Section 2.1. Subtracting yi,t-1 fromEq. (2) yields the following growth equation:

Ayit = 8yu-1 + P'Xit + Sit (2)

where A is the difference operator and 8 = a — 1. We augment Eq. (2) with an infrastructure variable, z, as follows:

Ayit = 8yi,t—1 + P'Xit + YZit + Sit (3)

The variable, z, includes infrastructure development indices AIDI and AIQI (defined in Section 3.1 as African Infrastructure Development Index and African Infrastructure Quality Index) and infrastructure investment variable, namely, public sector gross fixed capital formation, separately entering the Eq. (3). Because the lagged dependent variable is endogenous to the stochastic individual fixed effects (n), the disturbances (sit) are serially correlated (see e.g., Bond, 2002). Therefore, estimating Eq. (3) in its current form would bias the estimated coefficients. A common way of dealing with individual fixed effects (n) is to transform Eq. (3) by differencing it once (Holtz-Eakin et al., 1988). However, this procedure is inappropriate for unbal-

anced panels in which one missing variable may generate several missing differenced variables in the transformed data.

The forward orthogonal deviations transformation procedure (Arellano and Bover, 1995) performs better in such situations because it subtracts the average of all future available observations of a variable from levels of that variable.7 Because lagged values of untransformed endogenous variables do not enter the transformation mechanism, they can be used as valid instruments, to control for endogeneity. This is the basis of the system generalized method of moments (GMM) for dynamic panel data models (Blundell and Bond, 1998), which we use to estimate Eq. (3). In our estimation, we include time dummies to control for the potential effect of2007/8 financial crisis on economic growth in Africa. We test Eq. (3) on a panel of 45 SSA countries, using annual data for the period 2000 through 2011.8 Our data come from World Bank's African Development Indicators.

3. Data and measurement

3.1. Measuring infrastructure

Although many studies have found a positive long-run relationship between infrastructure and economic growth, interpretation of these results is always complicated by infrastructure measurement issues. According to Calderón and Servén (2010), measuring infrastructure as a single variable, either in physical or monetary unit fails to capture the multidimensional nature and heterogeneity of infrastructure across time periods and countries, and does not properly distinguish between quality/productivity and bulk of infrastructure. Additionally, simultaneity can be a serious econometric problem in infrastructure—growth studies because countries with faster growing output may spend more on infrastructure while infrastructure provision may also positively mediate the relationship between aggregate input and output, and hence foster output growth. Indeed, Kumo (2012), using South African data, confirms strong bidirectional causality between infrastructure investment and economic growth.

These flagged issues inform our variable measurement and choice of econometric procedures. That is, we try to respond to the criticism about the use of single variable measures by applying an index of various infrastructure measures. The African Infrastructure Development Index (AIDI), developed by AfDB (2013b), is a weighted average of nine indicators of infrastructure covering four key components: electricity, transport, information and communications technology (ICT), and water and sanitation. Although the index emphasizes measures of infrastructure "bulk", it also captures some aspects of infras-

7 An important advantage of this transformation is that, because the rows of the transformation matrix are orthogonal to each other, the transformed data maintain the properties of the original data. For instance, if the original data were identically and independently distributed (iid), the transformed data will be iid as well (see Roodman, 2006).

8 Serious attempts at involving the private sector in infrastructure financing started with market reforms in the mid-1990s. This investigation seeks to capture the relationships of interest post reforms.

O. Kodongo, K. Ojah/Review of Development Finance xxx (20l6) xxx—xxx

tructure "quality". For instance, bulk of transport infrastructure is captured through total road network in km (per square km of exploitable land area) while transport infrastructure quality is addressed through total paved roads (km per 10,000 inhabitants).

Water and sanitation is covered only through quality measures: improved water source and improved sanitation facilities (both as percentage of population with access) while electricity and ICT are only represented through bulk measures. The methodology used to develop AIDI is discussed in detail in AfDB (2013b). We interpret the AIDI index as infrastructure access and use it as our core measure of infrastructure development in the SSA region.9 To our knowledge, this would be the first time the AIDI index is used in a study relating infrastructure development to economic growth.

Table 1 shows the basic relationship between AIDI and real GDP in local currency units for SSA countries over the 2000-2011 decade. In the cross-sectional units, the two variables exhibit very strong correlation (negative in some cases) in levels (1 & 3). With such high levels of association, it is no surprise that studies examining the role of infrastructure in explaining economic development in Africa have recorded high levels of statistical significance of the infrastructure coefficient. We believe that such results can be partly ascribed to coincidence of statistical association. Because of negative correlations in a few countries, the pooled (panel) correlation is only 0.015 (see bottom right of the table). The table shows that infrastructure access (levels) seem to have a poor relationship with economic growth (1 & 4), with a substantial number of countries reporting low and/or negative correlations.

Finally, we see that growth in infrastructure access seems to have some reasonable relationship with growth in real GDP in several countries, albeit with a negative sign in some cases (2 & 4). Because of its fairly close relationship to economic growth at this basic testing level, we conjecture that incremental infrastructural investment/development is more relevant to economic growth in SSA than current levels of available infrastructure. Accordingly, we use the change in infrastructure access [ln (AlDlt) - ln (AIDIt-i)] as our dominant infrastructure proxy. However, for comparability with existing studies (e.g., Calderón and Servén, 2010 and Fedderke et al., 2006), we estimate some equations with levels of infrastructure as an alternative measure.

Third, because our infrastructure access measure, the AIDI index described above, puts more emphasis on infrastructure stock than it puts on infrastructure quality, we develop a synthetic index of infrastructure quality as an alternative measure of infrastructure development. Our African Infrastructure Quality Index (AIQI) is constructed from six variables: telecommunications (secure internet per million people and telephone faults

9 AfDB's (2013b) measure contains index values for the period between 2000 and 2010 only. To estimate the 2011 index values, we work with the values for 2009 and 2010 and the 2009-2011 average obtained from the AfDB website: http://www.

Table 1

Correlations between infrastructure indexes (AIDI) and real GDP in Sub-Saharan Africa.

Country Correlation between

1 &3 1 &4 2 & 4

Angola G.94 -G.11 G.44

Benin G.84 -G.33 G.G3

Botswana G.92 -G.19 G.31

Burkina Faso G.81 -G.G4 G.2G

Burundi G.47 G.G9 G.G3

Cameroon G.92 -G.54 -G.44

Cape Verde 0.96 -G.G2 G.38

Cen. Afr. Rep. -0.26 G.57 G.G7

Chad G.73 -G.32 -G.47

Comoros -G.9G -G.22 G.23

Congo, DR. G.97 0.56 G.48

Congo, Rep. G.95 G.18 G.54

Cote d'Ivoire -0.61 G.1G G.2G

Djibouti G.95 0.64 G.74

Eq. Guinea 0.64 -G.43 -G.35

Eritrea -G.85 G.G7 -G.17

Ethiopia G.97 G.34 -0.56

Gabon 0.63 G.58 G.28

Gambia G.51 -G.G4 G.21

Ghana G.98 G.74 G.38

Guinea G.83 -G.39 -G.15

Guinea-Bissau G.37 0.16 G.17

Kenya 0.86 G.24 G.G5

Lesotho G.99 G.39 G.28

Liberia G.59 -G.G4 -G.G5

Madagascar G.14 -G.13 0.26

Malawi G.91 G.55 G.33

Mali G.91 -G.34 -G.84

Mauritania G.89 0.06 0.16

Mauritius G.99 G.G3 G.2G

Mozambique 0.96 0.06 G.5G

Namibia G.92 G.14 G.54

Niger G.79 0.16 G.2G

Nigeria G.93 G.22 -G.G1

Rwanda G.94 G.G7 G.G8

Senegal G.88 -G.25 -G.G5

Seychelles G.82 G.44 G.23

Sierra Leone G.84 G.G1 G.1G

South Africa G.77 -G.14 G.13

Swaziland G.94 -G.37 -G.5G

Tanzania G.94 -G.19 -G.1G

Togo G.5G 0.66 G.G8

Uganda G.92 G.1G -G.G2

Zambia G.99 0.56 G.66

Zimbabwe -G.25 G.8G G.52

Panel G.G15 G.G18 G.G78

Data sources: World Development Indicators (real GDP) and African Development Bank (infrastructure index).

1 is African Infrastructure Development Index (AIDI) - proxy for infrastructure development.

2 is growth in infrastructure (annual percentage change in AIDI).

3 is real GDP per capita (logs).

4 is growth in real GDP per capita (annual percentage).

per 100 mainlines), power (percent of transmission and distribution losses in electricity production), transport (share of paved roads in total roads) water (percent of population with access to improved water source) and sanitation (percent of population with access to improved sanitation). We normalize each vari-

O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx-xxx

Table 2

Correlations between infrastructure quality indices and infrastructure variables.


with AIQI

Secure internet per million people 0.2353

Telephone faults per 100 mainlines -0.0985

Percentage of transmission and distribution losses in -0.0738

electricity production

Share of paved roads in total roads 0.7034

Percent of population with access to improved water 0.5383


Percentage of population with access to improved 0.4540


Africa Infrastructure Development Index - AIDI 0.4724

(AfDB, 2013a,b)

able using the formula di = (Ai — mi) / (Mi — mi) where Ai is the actual value of variable i, mi is the lower limit (empirical minimum) for variable i, and Mi is the empirical maximum for variable i. This procedure ensures that observed values retain their rank in the normalized series. We restate two variables -electric power transmission and distribution losses (kWh) and telephone faults (per 100 mainlines) - by subtracting the normalized values from unity to ensure that higher values for each normalized variable represents higher quality infrastructure.

To ensure consistency between our infrastructure quality variable and the infrastructure access variable (AIDI), we calculate the weight of variable i using the method used in AfDB (2013b)10 as: wi = aTot/ai, such that 1/aTot = 2i 1/°i- We start by applying this procedure to compute the dimension index for telecommunications, which is represented by two variables. The resulting telecommunications dimension is then treated as a distinct variable for purposes of constructing our synthetic quality index. With the resulting weights, we use the following linear transformation to construct the synthetic African Infrastructure Quality Index (AIQI):

AIQI = ^ WiX (4)

The resulting index compares favorably with AfDB's (2013b) Africa Infrastructure Development Index (AIDI) as well as with its component variables as shown in Table 2. A word of caution in the interpretation of the negative correlation coefficients is necessary: the two variables in their original forms imply poor quality infrastructure; in our index construction (as already explained), they are transformed so that large values imply good quality infrastructure. The transformed series correctly correlate negatively with the original unstandardized series (as reported in Table 2).

Due to the multiplicity of infrastructure services (and the accompanying complexity in measurement), many studies (e.g., Barro, 1990; Barro, 1996) use monetary values of investment in public assets to proxy infrastructure. Again, for ease of

10 The reasoning behind using standard deviations as the weighting basis is

to adjust for volatility in the original series and thereby reduce volatility in the index ranking of countries.

comparison with such studies, we also use the (log of) World Bank's public sector gross fixed capital formation of individual countries, as an additional proxy more specifically representing infrastructure investment. This variable is defined as outlays of expenditures to increase the stock of fixed assets and is, therefore, a flow, akin to the percentage change in the infrastructure access defined above.

3.2. Description of variables

The control variables included in our estimations are as follows. First is human capital, measured as the United Nations Development Program (UNDP) human development index reported in World Bank's World Development Indicators. Human capital is important because it enables a country's pool of labor resources to acquire hard skills (e.g., ability to operate machines) and soft skills (e.g., for teamwork and effective communication) which can potentially improve the productivity of capital. UNDP's human development index is composed of life expectancy, national income, and average and expected years of schooling.

Second, from Section 2.2, it is clear that financial development is considered important for economic growth especially in low-income countries (e.g., Hassan et al., 2011; Menyah et al.,

2014). Commonly used proxies of financial development include the ratio of broad money supply (M2) to GDP which indicates the size of financial intermediation (Calderón and Liu, 2003; Hassan et al., 2011) and ratio of credit to the private sector to GDP which represents the actual resources that are channeled to the private sector (Hassan et al., 2011; Menyah et al., 2014). We use both measures interchangeably. Third, we control for terms of trade shocks, defined, following Calderón and Servén (2010), as log differences in terms of trade indices; terms of trade shocks can promote economic growth in the long-run by impacting on savings/investment rate and capital accumulation (Chowdhury,


Other control variables that have been used by previous studies (e.g., Barro, 1990; Nyamongo et al., 2012) and which we included here are governance (proxied, in turns, by control of corruption and rule of law), government consumption as a percentage of GDP, inflation (GDP deflator) and exchange rates (local currency units per unit of the USA dollar). As a measure of price stability, inflation is expected to adversely affect consumer demand and adversely affect economic growth.

Notably, an additional set of variables - i.e., trade competitiveness, cost of doing business and foreign direct investment - are presented separately because they speak to a key innovation of our study. We postulate that infrastructure, importantly, also impacts economic growth via indirect channels, especially given the very low levels of infrastructure endowment in SSA countries. So the little infrastructure endowment available likely strengthens and/or enables growth antecedents. The economics of how these growth antecedents works with infrastructure to impact growth, are fully explored in Section 4.4 where the analysis of the indirect relations is developed further and tested.

Descriptive statistics of the variables are presented in Table 3. The maximum values of the AIDI infrastructure index is fairly

O. Kodongo, K. Ojah/Review of Development Finance xxx (20l6) xxx—xxx

Table 3

Summary statistics.

Mean Median SD Min. Max. #Obs.

GDP per capita growth 2.G7 2.G3 5.41 -41.17 45.91 538

Africa infrastructure index (AIDI) 16.2G 12.85 13.72 G.37 87.GG 54G

Africa infrastructure quality index (AIQI) 5.93 5.82 3.15 G.15 16.61 54G

GFCF (public sector) (% of GDP) 17.9G 9.55 27.22 G.25 227.47 481

Human development index G.44 G.43 G.12 G.23 G.8G 343

Credit to private sector (% of GDP) 599 18.33 3932 -34.19 46,875 423

Money supply (M2) (% of GDP) 3G.GG 22.44 24.42 2.83 148.71 521

Terms of trade shocks G.G2 G.G1 G.12 -G.98 G.52 537

Inflation (GDP deflator) 15.34 6.69 114.53 -29.55 263G 54G

Exchange rate (LCU units per USD) 5.G6 6.1G 2.28 -3.11 22.63 536

Government consumption (% of GDP) 4G8 15.28 2474 G.GG 23,461 449

Control of corruption -G.58 -G.68 G.58 -1.71 1.25 495

Rule of law -G.69 -G.7G G.63 -2.12 1.06 495

SD is standard deviation; Min and Max denote minimum and maximum; #Obs. is number of observations.

high - an examination shows that these values relate to the higher income economies of Seychelles (range: 47.43-87.00) and South Africa (46.07-77.91), whose infrastructure are relatively more advanced than that of the rest of the region. The general observation is that richer countries in the region seem to have better infrastructure endowments than poorer countries. We interpret this observation to mean that infrastructure development is endogenous to economic growth and accordingly treat the variables as such in our estimations.

Further, we run our empirical tests initially with all forty five countries in the panel and then run additional tests excluding all economies defined by the World Bank11 as upper middle income and high income: Botswana, Equatorial Guinea, Gabon, Mauritius, Namibia and South Africa. However, the median infrastructure development index of 12.85 out of a possible 100, confirms Africa's much-discussed poor infrastructure endowment. As expected, the infrastructure quality index (AIQI) is even poorer, with a maximum of 16.61 and a median of only 5.82.

On the average, terms of trade in Africa have improved over the 2000-2011 period as shown by the positive mean and median of the shock variable, which suggests a general weak but ascending favorable balance of trade position. The medium credit to the private sector and broad money supply stand at 18.33% and 22.44%, respectively, relative to GDP, both of which speak to the low financial depth in the Sub-Saharan Africa region. The distribution of the credit to the private sector is very broad (standard deviation of 3932%) with the minimum and a maximum values being, respectively, -34.19% and 46,875%. A closer look at the data shows that the high values belong invariably to Mozambique, a fast growing economy in the SSA region whose growth seems to be largely anchored on an expansionary

monetary policy, public spending, and foreign direct investment (FDI) flows.12

Government spending to GDP also varies widely, a sign that economies in the region operate different macroeconomic management policies: this suggests the need to control for cross-sectional heterogeneity in our estimation and justifies the fixed effects procedure applied through the system GMM, discussed in Section 2.2. Institutional quality, represented by control of corruption and rule of law, appears weak, with both the means and medians recording negative results. A final observation is in respect of missing variables. Data is scarce on several of our variables for the SSA region. Although this is largely taken care of by the forward orthogonal deviations transformation in the system GMM, we nonetheless repeat the estimations excluding all countries with more than four missing observations for any of the variables. That gives us a smaller sample of 31 countries.

4. Empirical results

4.1. Causality between infrastructure development and economic growth and development

We begin our empirical section with Geweke-Granger causality tests as a prelude to understanding the nature of the relationship between various infrastructure development metrics and economic growth and development. The test begins from the standard Granger-causality test, which hypothesizes that variable x causes/informs another variable, y, if lagged values of x can explain contemporaneous values of y. The standard Granger-causality procedure has one major drawback: if the data are not observed with high frequency (e.g., if data are observed annually as is the case in this paper), the test may leave much of the contemporaneous correlation between variables unexploited (Dicle and Levendis, 2013). In particular, if after controlling for their lags, there is still correlation between contemporaneous

11 Upper middle income countries were defined, as of July 2011, as countries with GNI per capita of USD 3976 or higher. See World Bank's website: (accessed April 7, 2016).

12 This is according to a recent AfDB/OECD/UNDP report, available at: _Long_EN/Mozambique_GB^2015.pdf (accessed April 7, 2016).

O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx-xxx

Table 4

Causality test results.

Panel A: infrastructure levels and economic growth

Fx^y 4.157 (0.042)** 2.223 (0.136)

Fy^x 0.872 (0.350) 0.067 (0.797)

Fx^y 11.35 (0.000)*** 1.287 (0.257)

Cxy 16.38 (0.001)*** 3.577 (0.311)

Panel B: infrastructure changes and economic growth

Fx^y 0.199(0.656) 0.002(0.962)

Fy^x 0.719 (0.396) 0.001 (0.973)

Fx^y 16.13 (0.000)*** 2.178(0.140)

Cxy y 17.05 (0.001)*** 2.182(0.536)

Panel C: infrastructure levels and economic development

Fx^Y 10.02(0.002)*** 1.1491 (0.284)

FY^x 0.217 (0.641) 0.3269 (0.568)

Fx^Y 12.33 (0.000)*** 1.0711 (0.301)

Cx Y 22.57 (0.000)*** 2.5471 (0.467)

3.308 (0.069) 8.560 (0.003)*: 0.935 (0.334) 12.80 (0.005)*

3.623 (0.057)* 13.60 (0.000)* 5.332 (0.021)* 22.56 (0.000)*

2.391 (0.122) 0.098 (0.754) 4.415 (0.036)*: 6.904 (0.075)*

0.116(0.734) 0.487 (0.485) 4.902 (0.027)* 5.504 (0.138)

0.404 (0.525) 1.447 (0.229) 6.368 (0.012)* 8.219 (0.042)*

The table presents chi-square statistics (p-values in parentheses) for test of the hypothesis that variable x [variable y] does not Granger-cause variable y [variable x] (in first two rows in each panel), that there is no linear feedback between variables x and y (row 3) and that there is no contemporaneous correlation between variables x and y (row 4). We use one lag for each test, the lag structure chosen by the SIBC criterion. The test statistic follows a chi-square distribution with the degrees of freedom (DF) shown. x is the infrastructure variable in level or in first difference; y is change in economic growth (log difference in GDP per capita); Y is economic development (level of GDP per capita). We do not run causality tests with gross fixed capital formation (GFCF) in first difference because they are flows by definition. AIDI is Africa infrastructure development/access index; AIQI is Africa infrastructure quality index; AISI is Africa infrastructure stock index.

values of x and y, then the instantaneous correlation between the two variables will not have been fully accounted for.

Geweke (1982) proposed that the variance-covariance matrix of residuals from the vector autoregression estimation (that yields the Granger-causality test output) can be used to estimate the linear feedback between y to x and x to y, and the instantaneous linear feedback between x and y. We use the Geweke decomposition methodology to test for causality between economic growth and various infrastructure indicators. Results of the test are presented in Table 4.

The results show that the pertinent variables do not have a systematic tendency to Granger-cause each other but where causality exists, it tends to be unidirectional and runs frominfras-tructure development to economic growth [development]. The few cases of causality seem to be dominated by infrastructure access (AIDI); there is no evidence of causality between infrastructure quality (AIQI) and economic growth [development]. Further, there are instantaneous (contemporaneous) feedback effects between levels of and changes in infrastructure access and gross fixed capital formation (GFCF) on the one hand and economic growth [development] on the other. Our pure infrastructure stock (AISI) variable shows bidirectional causality with economic growth and instantaneous feedback with economic development.

The mixed causality results, which seem to suggest a lack of clarity in the nature of the relationship between infrastructure development and economic growth [development], is the first of the many reasons for which we raise the issue as to whether we should be circumspect about past studies of the relationship between these two variables, which primarily find that infrastructure development explains economic growth [development] in Sub-Saharan Africa. For this reason, we not only examine the direct link between infrastructure and economic growth but

also the potential role of indirect channels, hypothesized in the literature, through which infrastructure may explain economic growth.

4.2. Infrastructure access, infrastructure quality and growth in public fixed capital formation

We estimate the model in Eq. (3) using a panel of all 45 countries in our sample. We use the Arellano and Bond (1991) instrument set in the first test (Table 5) and then, given the instrument proliferation observation of Roodman (2009), we restrict the instrument set to between 2 and 5 lags of the dependent variable (Table 6). In all estimations, we instrument infrastructure development and investment by population growth. The population variable is transformed in the same way as the explanatory variables. The appropriateness of population growth as an instrument is rooted in theory and supported by several empirical analyses that have reported a strong relationship between population growth and density and both national income and infrastructure. In their theoretical model, Becker et al. (1999) argue that the net relation between increasing population and per capita incomes depends on whether inducements to human capital and expansion of knowledge enabled by larger populations are stronger than the traditionally acclaimed diminishing returns to land and other natural resources.

In an empirical analysis with data of developing economies, Simon and Gobin (1980) find a positive relationship between population density and economic growth and conclude that over the long term, population growth has a positive effect on per capita income. Similarly, Weinhold and Reis (2001) show a strong positive contemporaneous relationship between infrastructure and urban population as well as provide evidence supporting the hypothesis that growing urban populations lead

^ o re

Table 5

Estimation results with full instrument set.

Eq. (l)a Eq.(2)a Eq. (3)a Eq. (4)a Eq.(1)q Eq. (2)q Eq. (3)q Eq. (4)q Eq. (1)i Eq. (2)i Eq. (3)i Eq. (4)i

Lagged GDP per capita -3.818"' (0.790) -4.190''' (0.376) -3.860''' (0.809) -4.290''' (3.782) -3.601''' (1.202) -4.936''' (0.724) -3.957''' (0.976) -5.086''' (0.824) -4.760''' (0.756) -3.959''' (0.672) -4.546''' (0.840) -4.000''' (0.786)

Africa infrastructure index -3.332(5.785) -5.377''' (2.434) -4.079(5.312) -5.117'' (2.385)

Infrastructure quality index -0.164(2.302) -3.108''' (0.850) -0.559(2.183) -1.483(1.431)

GFCF (public sector) 0.036' (0.021) 0.060''' (0.017) 0.033 (0.022) 0.060''' (0.020)

Human development index 7.261'*' (2.703) 5.050'' (1.065) 7.760''' (2.237) 5.286''' (9.536) 5.751 (4.143) 3.596'' (1.656) 7.423''' (2.321) 5.059''' (1.366) 7.137''' (1.470) 2.288(2.187) 7.095''' (1.502) 2.789(1.889)

Credit to private sector -0.434 (2.508) -0.601 (2.323) -2.420 (2.617) -2.716(2.230) 1.809(2.359) 1.962 (2.610)

Money supply (M2) -5.497''' (1.434) -5.438''' (1.310) -6.847''' (2.226) -8.035''' (2.408) -3.840'' (1.578) -4.697'' (2.067)

Terms of trade shocks 1.011 (2.438) -0.721''' (2.724) 1.125 (2.424) -1.116(2.851) 0.091 (3.796) - 1.761 (4.506) 1.216(2.782) -1.607(5.186) 5.390''' (2.007) 6.228' (3.562) 5.533'' (2.381) 7.323 (4.909)

Inflation 0.627 (0.500) 0.439 (0.293) 0.652 (0.477) 0.426 (0.427) 0.956' (0.543) 0.781' (0.402) 1.027' (0.521) 0.556 (0.525) 0.218 (0.360) -0.335 (0.601) 0.375 (0.372) -0.458 (0.647)

Exchange rate -7.032''' (2.422) -3.616(3.970) -6.781'' (2.740) -3.036 (4.049) -7.671 (2.758) 1.451 (7.320) -6.355'' (2.802) 5.995 (10.72) -9.578''' (2.223) -0.903 (7.298) -8.520''' (2.613) 0.827 (8.070)

Government consumption 1.276(1.951) 1.509(1.522) 1.401 (2.059) 1.596(1.800) 1.491 (2.194) 0.862 (2.083) 1.899 (2.325) 1.122 (2.356) 0.872(1.858) 0.750 (2.220) 1.470(1.676) 0.342(2.159)

Control of corruption 1.325(1.508) 2.078' (1.253) 2.721 (3.412) 7.987''' (2.960) 2.274'' (1.124) 3.661' (2.084)

Rule of law 0.635 (1.987) 1.677(1.885) 2.049 (3.806) 5.275' (2.974) 2.093 (1.698) 3.230 (2.418)

p-value of Sargan test 0.487 0.414 0.472 0.400 0.429 0.635 0.392 0.538 0.495 0.633 0.501 0.615

Robust standard errors are in parentheses. "a" stands for infrastructure "access"; "q" represents "quality" of infrastructure and "i" represents infrastructure "spending". ***,**, and * represent statistical significance at 1%, 5% and 10% respectively.

Table 6

Estimation results with restricted instrument set.

Eq. (1)a Eq. (2)a Eq. (3)a Eq. (4)a Eq. (1)q Eq.(2)q Eq. (3)q Eq. (4)q Eq. (1)i Eq. (2)i Eq. (3)i Eq. (4)i

Lagged GDP per capita -4.825''' (0.832) -4.662''' (0.905) -4.887 (5.980) -4.861''' (0.617) -4.921''' (8.522) -5.266''' (1.094) -4.904''' (0.776) -5.332''' (7.393) -5.728'' (1.040) -3.518''' (0.562) -6.088''' (1.198) -3.811''' (0.692)

Africa infrastructure index -3.514 (3.868) -4.333 (3.910) 7.204 (25.89) -4.323 (2.914)

Infrastructure quality index -0.600(1.227) -0.739(1.379) -0.363 (1.091) -0.355(1.215)

GFCF (public sector) 0.061 (0.040) 0.045 (0.027) 0.059 (0.045) 0.031 (0.030)

Human development index 9.560''' (2.553) 2.519 (3.045) 17.86(12.63) 5.597'' (2.171) 8.804''' (2.301) 2.778 (3.530) 9.066''' (1.967) 5.835'' (2.745) 6.374''' (2.283) -2.085 (2.203) 7.245''' (2.557) -1.258(2.035)

Credit to private sector -0.805 (2.390) -6.728 (9.529) -0.723 (2.341) -0.866 (2.284) 0.748 (2.991) 0.205 (3.400)

Money supply (M2) -3.848''' (1.234) -3.395''' (1.030) -2.952'' (1.313) -2.995'' (1.246) -2.556''' (0.787) -2.749''' (1.000)

Terms of trade shocks 3.765 (2.441) 0.407 (2.546) -3.599 (6.285) 0.828 (2.294) 4.402 (2.838) 0.363 (3.543) 4.450 (2.737) 0.363 (2.763) 3.734 (2.354) 1.571 (1.677) 3.873 (2.963) 1.143 (1.931)

Inflation -0.205 (0.612) 0.591 (0.538) 1.032(1.320) -0.059 (0.440) -0.223 (0.664) 0.618 (0.599) -0.228 (0.601) 0.128 (0.561) -1.203 (0.801) 1.052' (0.567) -1.193(0.952) 0.860 (0.527)

Exchange rate -7.996''' (2.850) 3.657 (8.827) -1.428 (2.080) -1.115(6.588) -8.780''' (2.878) 3.876 (10.858) -8.756''' (2.671) -0.863 (9.227) -9.577'' (4.560) 0.739 (1.075) -9.957'' (4.940) 5.486(11.67)

Government consumption 0.330 (1.795) -0.797 (2.797) 6.900(1.032) -0.858(1.793) 0.233 (1.740) -0.399(3.075) -0.088(1.441) -1.339(2.184) 0.058 (3.250) -3.218 (2.790) -0.188(3.195) -4.489 (2.799)

Control of corruption -0.569 (2.041) 5.248' (2.700) 0.741 (1.707) 7.395' (3.830) 3.725' (2.244) 4.516'' (1.930)

Rule of law -7.265 (8.060) 1.379(2.089) 1.783 (2.315) 3.989 (2.760) 3.998 (3.049) 2.108(1.918)

J-statistic p-value 0.280 0.472 0.240 0.356 0.301 0.358 0.325 0.723 0.550 0.681 0.503

Robust standard errors are in parentheses. "a" stands for infrastructure "access"; "q" represents "quality" of infrastructure and "i" represents infrastructure "spending". ***,**, and * represent statistical significance at 1%, 5% and 10% respectively.

O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx—xxx

to more infrastructure development (in the Granger-causality sense). Further, Esfahani and Ramírez (2003) find a strong positive association between returns to net infrastructure investment and population density. In our analysis, we also use population density as an instrument and find qualitatively similar results.13 We include time dummies in all equations.

Results, reported in Table 5, show a weak relationship, which is sometimes negative, between the target variable -infrastructure indexes - and economic growth. Alongside this weak relationship, human (capital) development index, financial development, terms of trade, exchange rate, and control of corruption, which had been documented as important in previous economic growth studies, also register statistically significant relations to economic growth. However, as we flagged in the foregoing discussion, these relationships may be spurious; and indeed when we remove potential instrument proliferation (Table 6), none of the infrastructure variables remains significantly related to economic growth. Thus, we base our discussion on the empirically better set of results presented in Table 6.

Apart from showing a strengthened convergence, with higher coefficients of lagged GDP per capita than those reported in Table 5, all three measures of infrastructure (Africa infrastructure index, infrastructure quality index and public sector GFCF) are statistically insignificant in relation to economic growth. This outcome suggests that the level of access to infrastructure or its quality, as were then available in African countries, did not affect economic growth. Recall, going by our preliminary analysis in Section 3.1 (and Table 1), and our postulation about SSA's level of infrastructure endowments, infrastructure increments and consideration of channels that transmit the effects of infrastructure to economic activity are our important areas of focus. We will turn to these areas shortly. First, we explore briefly how the control variables impact economic growth in Africa, especially for potential useful policy guides, and highlight the possible channels, among antecedents of growth, through which infrastructure impacts economic growth (see Table A1 in the Appendix A).

In support of findings by Ghura and Hadjimichael (1996), Easterly and Levine (2001), Pritchett (2004) and others, human capital is highly statistically relevant and economically important to economic growth in Africa (with an average coefficient of about 7.000 in economic growth per unit change in human capital). Similarly, the coefficient of the control of corruption variable supports Rodrik et al.'s (2002), Rodrik's (2003) and, Lee and Kim's (2009) finding of how institutions matter for both developing and developed economies in driving economic growth. Not only do these two help in eliciting incremental returns from Africa's limited capital stock, they also, especially institutions, minimize the distortionary effects their absence would have on markets.

Though a preponderance of studies about finance—growth nexus finds a positive relationship (Calderón and Liu, 2003; Hassan et al., 2011; Ojah and Kodongo, 2015), on occasions

13 These are not presented in the paper but are available with the authors upon request.

the positive relation has not been straightforward, suggesting the likelihood of a nonlinear relationship. A recent study by Aizenman et al. (2015) has evidently confirmed this likelihood with data from developing Asia and Latin America. They documented the kind of negative relations between financial deepening and economic growth, which we report in Table 6. Therefore, it appears, using the ratio of M2 to GDP as a proxy of financial depth, that African countries have not attained a threshold level of financial development that would support incremental economic activity. Alternatively, it could be that the implied increased level of money supply subsumes the effect of inflation on economic activity instead of representing financial intermediation.

In fact, the more reflective proxy, credit supply to the private sector, has no statistically significant effect on economic growth (perhaps because a good proportion of credit supplied, especially to individual borrowers, which is largely in the form of unsecured personal loans, could be financing consumption, rather than investment activities). Unlike in developing East Asia, which has a significant endowment of process technology and thus export a great deal, as an economic growth strategy, relative currency value depreciation in Africa yields a significant negative relation to economic growth. Though African countries export a significant amount of its natural resources (which yields favorable terms of trade during commodities' price spikes), the burden of importation with highly weaker currencies more than negatively overwhelm the terms of trade effect (Muhanji and Ojah, 2011,2016).

Some studies present institutions as the ultimate economic growth determinant (e.g., Acemoglu et al., 2005; Kimenyi, 2007). According to this view, our finding of a largely insignificant effect of infrastructure access and quality on economic growth might potentially be attributed to Africa's relatively low institutional quality which might have made infrastructure less effective as a growth catalyst. However, many studies appear not to concur with the institutions-drive-economic-growth view, often pointing to the difficulty of measuring institutional quality and historical evidence showing that many of today's advanced economies acquired their stellar institutions (e.g., good governance, property rights, securities regulation, etc.) after they developed rather than before (e.g., Chang, 2011).

Notwithstanding these views, we have controlled for institutional quality through two commonly used proxies - rule of law and control of corruption in our estimation. We, therefore, proceed to explore the suspected areas that sensibly reflect the link between our variables of interest. These areas have also been theoretically hypothesized as relevant (see Section 1). Before that, however, we attempt, in Sections 4.2 and 4.3, to establish the robustness of the foregoing findings.

4.3. Infrastructure access/quality and economic growth in low-income SSA countries

Let us first consider the likely effects of the heterogeneity of African economies. It is possible that mixing richer countries with a relatively superior infrastructure endowment with poorer countries with near-zero endowments may distort the relation-

O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx—xxx 13

Table 7

Estimation results for low-income Sub-Saharan African countries.

Panel A (full instruments set) Panel B (restricted instruments set)

Eq. (1) Eq. (2) Eq. (3) Eq. (1) Eq. (2) Eq. (3)

Lagged GDP per capita -4.385*** (0.672) -4.166 *** (0.960) -3.820*** (0.943) -5.840*** (0.869) -5.781 *** (1.000) -4.751*** (0.569)

Africa infrastructure index -8.415 (5.591) -3.419 (2.835)

Infrastructure quality index -3.706 (2.307) -0.999 (1.356)

GFCF (public sector) 0.049** (0.024) 0.074** (0.035)

Human development index 8.310*** (1.436) 6.721** (2.641) 3.547* (1.934) 8.194*** (2.459) 7.110** * (2.156) -1.312(1.495)

Money supply (M2) -0.515 (4.978) -4.784 (3.877) -3.251 (4.753) -1.224 (1.284) -0.979 (1.926) -1.128 (1.778)

Terms of trade shocks 2.033 (5.014) -2.585 (5.465) 4.955 (4.527) 3.407 (2.452) 3.520 (3.436) 2.904 (2.331)

Inflation 0.284 (0.558) 1.187** (0.585) 0.310(0.750) 0.750 (0.476) 0.329 (0.445) 0.461 (0.349)

Exchange rate -6.359 (6.430) -1.996 (1.246) -3.345 (7.174) -7.405 (8.810) -5.928 (9.631) 1.113 (6.514)

Government consumption 1.750(1.890) 3.432(2.116) 2.634 (2.114) 1.322(2.482) 1.681 (2.355) 0.987 (2.285)

Control of corruption -2.014 (2.933) -1.938 (2.151) 1.714 (2.087) 4.911 (3.268) 5.911 (4.316) 8.204*** (2.567)

J-statistic p-value 0.252 0.479 0.348 0.611 0.466 0.756

Robust standard errors are in parentheses. ***, **, and * represent statistical significance at 1%, 5% and 10% respectively.

ship between infrastructure and economic growth in the panel. Consequently, in this section, we remove seven countries classified by the World Bank as upper middle income or high income countries14 and then rerun our tests on 38 poorer countries in the SSA region. Because of space constraints, we report only money supply as the financial depth variable and control of corruption as the proxy for institutional quality.

Results of the rerun test are reported in Table 7. Panel A uses the full set of instruments; panel B uses the restricted set. Once more, we discuss the estimates of the better technically developed 'restricted instrument set' (Panel B of Table 7). These results largely resemble those in Table 6 (and our interpretation there suffices here), except for an important difference regarding GFCF (public sector): expectedly, convergence is robustly strong for the set of lesser developed countries; secondly, variables we would deem basic drivers of economic growth (i.e., human capital and institutions) remain significant and positively related to economic growth.

Interestingly, however, none of the other variables is significantly important, in contrast to Table 6 which documented negative effects of money supply and currency depreciation. Most interestingly and in confirmation of our suspicion that sample heterogeneity may have affected the results in Table 6, infrastructure, in the form of gross fixed capital formation (GFCF), is found to enter significantly positively the model of the lesser developed countries sample. That is, increases in the capital base of lesser developed economies are relevant for growth than increases in capital base of the relatively more advanced economies, which perhaps have already hit the threshold of requisite infrastructure stock (Reinikka and Svensson, 1999; Estache et al., 2006; Kumo, 2012).

4.4. Improved infrastructure access and infrastructure quality

Because current flow of money into infrastructure development may indirectly contribute to economic growth by increasing consumption through higher labor employment and attendant high income, we test whether the observed effect of public spending on fixed capital assets on growth can also be observed in the change in access and quality of infrastructure that the spending creates. We use log differences in infrastructure access and quality indices for this purpose.

Results are in reported in Table 8. Focusing on the restricted instruments set, the results show that improved access to infrastructure is important for economic growth, and thus buttress the finding in Table 7. Infrastructure quality improvement, though not statistically significant, are positive and larger when the relatively more advanced African economies are included in the sample (coefficient of 0.678 of Eq. (3) in Panel A) than when they are excluded (coefficient of -0.070 of Eq. (3) in Panel B). This overall non-effect of infrastructure quality is not entirely surprising considering the poor quality and low infrastructure endowment in the continent. On the other hand, marginal improvements in infrastructure access go a long way in alleviating their 'legendary shortage' and thus reflect in economic growth. Finally, note that both convergence and human capital remain robust and are stronger statistically and economically for lesser developed economies than for the sample that includes more advanced African economies, an outcome that aligns, broadly, to findings of Lee and Kim (2009).

4.5. Mediating effect(s) of pertinent growth antecedents

14 We use the World Bank's classification as of January 2011, obtained from their website: Jan%202011%20World%20bank%20list%20of%20Economies.PDF.

As we argued in Section 1, it is very likely in the SSA context that infrastructure might indirectly impact economic growth by improving total factor productivity. According to Barro (1990) and others, this is possible if infrastructure provisioning acts by lowering input costs or expanding the production frontier or the set of profitable investment opportunities. For instance,

O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx-xxx

w N ^ C „ ^ ^ ^ ^ ^

O » N ®

I ^o I

m c\ o 00

00<N 00

, in Q\ on <N

m m cK I

t in N I ^ I

^ ^ ^ O ' 00 00 (N on p.

t^ \n ' ; ^ >r> . (N 00 ^ O

o I o o iri I

o in 00 o o

o o o o

o o\ o inï ^t oo o

c^ on ;

^ N № tn

(D -S !i 1

Q • O

.2 S j3

availability of safe internet may increase the speed of communication and timeously open up profitable trade opportunities. To illustrate this point in a more practical sense, consider the Sub-Saharan African region, where the electric power grid is often unstable due to insufficient generation, and transmission and distribution inefficiencies, forcing manufacturers to operate expensive standby thermal power generators. In such situations, the provision of, and access to, reliable electricity would obviate the need for emergency thermal generation thereby reducing manufacturers' energy costs and overall cost of doing business. The resultant reduced business cost should improve countries' trade competitiveness, improve their attractiveness as investment destinations, and increase economic output.

The literature is rich with suggestions that public infrastructure significantly affects countries' trade competitiveness and attractiveness as an investment destination. For example, Kumar (2001), using a synthetic infrastructure index capturing transport, telecommunications, information and energy infrastructure availability for 66 countries over the 1982-94 period, finds that infrastructure availability contributes to a country's attractiveness for foreign direct investment and promotes export-orientation of production of multinational enterprises. Similarly, Obwona (2001) argues that poor infrastructure facilities, especially in transport, communications and information technologies, are major impediments to investment in lower income countries. Many studies have also found a role for trade competitiveness, foreign direct investment flows, and cost of doing business in explaining economic growth (see e.g., Li and Liu, 2005). Given these findings in the literature, it is not implausible to argue that infrastructure can indirectly influence economic growth by acting via specific growth antecedents.

In this section, therefore, we attempt to understand whether infrastructure availability and quality may act through the mediating effects of some of these variables to impact economic growth. We examine the cost of doing business, trade competitiveness, and foreign direct investment (cross-border capital flows), all of which have featured prominently in the literature. We proxy cost of business by the cost of starting a business, reported by World Bank15; trade competitiveness is proxied by export diversification indices reported by UNCTAD; and foreign direct investment is the aggregate inward and outward FDI flows as a proportion of GDP, also reported by UNCTAD.

As a prelude to testing one of our key postulations, we conduct a preliminary analysis of how growth antecedents - of trade competitiveness, cost of doing business and cross-border capital flow - relate to economic growth/development, and how infrastructure relate to these growth antecedents, respectively. The results, reported in Table A1 in the Appendix A, provide additional justification, to theoretical and anecdotal motivations in

15 The statistics are reported in the ease of doing business indicators. The "ease of doing business" is captured by an index called distance to frontier (DTF) that covers several aspects of "doing business" including starting a business, getting electricity, dealing with construction permits, paying taxes etc. The comprehensive index is available for many African countries only from 2010. Thus, we use the "cost of starting a business" data which are available for a longer period (from 2004).

O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx—xxx

the preceding paragraphs of this section, of why these variables are likely channels by which infrastructure indirectly impacts growth.

Tables 9 (for infrastructure access) and 10 (for infrastructure quality) report our regression results, with Panel A reporting results of models using level of infrastructure access/quality and Panel B reporting results of models using increments in infrastructure access/quality. As in previous estimations of our baseline model, convergence remain relevant, human capital and institutions remain largely positive though not all statistically significant, financial development remains negatively related to growth; and most importantly, improvements in infrastructure access remains robustly related to economic growth (Table 9). Moreover, the view from extant literature that availability of infrastructure could affect economic growth via other drivers of economic growth is given credence in our documented result (Panel A). Specifically, a unit of the combination of infrastructure access and export diversification (i.e., trade competitiveness) yields an average of 1.614 units increase in economic growth in SSA.

This additional finding (i.e., indirect effects of infrastructure access), plausibly suggests that even for economies of SSA which rely largely on harvesting their natural resources for export earnings (without value addition by way of processing/manufacturing) and, thus economic growth enhancement, they need and accordingly value basic public infrastructure availability such as electricity, roads and ports. Further, note that interaction terms of the other two intermediating variables considered - i.e., cost of doing business and cross-border capital flows (net foreign direct investments (FDI)) - have no statistically significant effect in the models in Table 9.

The results in Table 10, though essentially similar to those of Table 9, reveal a few interesting and noteworthy findings. These results are based on analysis of intermediating variables through which infrastructure quality could work to affect economic growth (results in Panel A of Table 10 uses quality of available public infrastructure while results in Panel B uses increments in the quality of available public infrastructure). Focusing on the interaction term variables, a unit of combination of infrastructure quality and export diversification on one hand, and a unit of combination of infrastructure quality and cross-border capital flows on the other hand, yield average units increases of 4.594 and 0.452 of economic growth, respectively (Panel A). The combined effects of the same intermediating variables and increments in quality of available public infrastructure yield the corresponding average units increases of economic growth of 5.985 and 0.406, respectively (Panel B).

Judging by the magnitude of the coefficients reported above, it is clear that these indirect effects are both statistically and economically important. Further, the intermediating effects through trade competitiveness are more economically important than those through cross-border capital (coefficients of 4.594 and 5.985 versus 0.452 and 0.406). That is, the quality of available infrastructure appears more impactful on economic growth by leveraging the effects of trade competitiveness, which has been documented as a significant economic growth driver across many countries (e.g., Muhanji and Ojah, 2011), than the likely

^ N I H

^ Ö TT ^

m c^ I I

m m o m

O 2 o «a

U^Ttr^ _ _ _

OOOOO r-.^ ^



Ó 00 ^ Ö <N

m c^ o m <N c^

^ (N vq TT in o

<N 00 Ö <N en -H


<N m I <N

t-^ooeo • -H^OOO MD<Neo


en m Ó Ó I I


on CD hOO

^ CDS CD ^Doomm

l^i—: t'Jl'J

-rf- O

O ^o ^ o ^ 1 O » C In M ® B ^ N IO Ü

1 "I^IS

ï ö. s

3 * t^ ^


№ o o

l> " VO

1 № O -h VO

c ° O,

O * T3 <D

'S i-i C

.tí ^ <3> _

x o o ~ ~

su—article in presS

O. Kodongo, K. Ojah /Review ofDevelopment Finance xxx (2016) xxx-xxx

Ooo^Ococo VT VT r^.

m c^ 01

oo^oh^r-c^ ooa .r^^t moo-HMDMD

^laco^Ooo I -^000*0

6 0 .2 6 6 0 0 .5 .0 2 .9 .4 .0 .4

1. 2. 0. 1. 1. 1.

.4 0 8 0 0 1. 2 0 0 1


.1 2 .5 .0 .4 .1 .5 .0 .1 .4

I doddr^-^ood

m <N 4.

(Nmoom 'hoooon ooinom


o\\o oooo h -Hooc^r-^m 0<NC^CO<N


xxx xe e e edd

ty t ) k it n ) c

ye ld pa pr pt

of n g

or g r ono

c in p f oi xp

o n rn ol

cS^-H ^¿H S ^ * ^ qhhhh

impact of leveraging cross-border capital flows. Yet again, the intermediating effects of "cost of doing business" are statistically unimportant in both models (Tables 9 and 10).

Consistent with the effects of the control variables in Table 9, convergence remains relevant in Table 10, human capital and institutions remain largely positive, financial development remains negatively related to growth; and interestingly, quality of institutions are more statistically relevant when quality of available infrastructure are considered compared to when access to infrastructure are considered (Table 10). Again, as is consistent in all our previous results, both the level and increment of quality of available infrastructure themselves remained unimportant in these latest set of model estimates.

5. Additional tests

5.1. Do missing observations matter?

Critics may argue that missing data (although well-handled econometrically in our estimations) may have an effect on our findings. We rerun our tests, excluding countries for which at least four time periods of data for any variable are missing. The following countries are removed by this filter: Benin, Burundi, Cape Verde, Djibouti, Eritrea, Ghana, Niger, Nigeria, Rwanda, Seychelles, Tanzania, and Zimbabwe. We also exclude Equatorial Guinea (high income) and South Africa (relatively more advanced infrastructure) from the sample. This leaves a sub-sample of 31 countries. All remaining variables have two or less missing observations, except human development index which has 4 missing observations for all countries. Besides the expected weakening of some variable coefficients, due largely to smaller sample size, the result in Tables 6 and 8 are essentially upheld in the rerun tests results reported in Table 11.

5.2. Pure infrastructure stock index

Our analyses, thus far, have used the infrastructure access index (AIDI) developed by the African Development Bank and our own infrastructure quality index (AIQI). Although the AIDI index includes aspects of quality and aspects of stock as pointed out before, we find it necessary to include, as further robustness check, a separate analysis of the stock of infrastructure. In this section, we document that the relationship between stocks of infrastructure and economic growth is similar to that between quality of infrastructure and economic growth. For consistency, our 'pure' stock of infrastructure index, AISI, is developed using the same methodology as the other two indexes that we have used. The pure infrastructure stock index, is constructed using electricity production (kWh); road density (km of road per 100 square km of land area); telephone lines (per 100 people); and international internet bandwidth (bits per second per person) all obtained from World Development Indicators for 2000-2011.

Table 12 presents the relevant results. Like infrastructure quality, the stock of infrastructure appears not to inform eco-

O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx—xxx 17

Table 11

Robustness to missing observations.

Levels of infrastructure indices % change in infrastructure indices

Eq. (1) Eq. (2) Eq. (1) Eq. (2)

Lagged GDP per capita -7.279 (4.724) -5.903** (2.962) -3.436*** (0.833) -3.283*** (0.587)

Africa infrastructure index 1.813 (3.871) 2.б07*** (0.742)

Infrastructure quality index -2.504 (3.973) -0.109 (0.765)

Human development index 1.685 (1.163) 1.891* (0.991) -1.809 (2.183) -3.132(26.60)

Money supply (M2) -1.306* (0.748) -1.017 (0.867) -0.752(1.254) -2.573*** (0.884)

Terms of trade shocks -2.283* (1.266) -2.503 (1.625) 1.620 (2.368) 2.533 (1.777)

Inflation 0.303 (1.884) 0.481 (1.719) 0.285 (0.593) 0.485 (0.572)

Exchange rate -2.341 (107.2) -4.517 (15.04) 1.546(3.253) 8.491 (21.81)

Government consumption 3.029 (8.2б3) -2.331 (7.535) -2.549 (4.888) -3.848 (2.75б)

Control of corruption -0.012(15.12) -0.182(9.370) 4.825* (2.510) 2.620(1.883)

J-statistic p-value 0.747 0.б37 0.бб3 0.182

Robust standard errors are in parentheses. ***, and * represent statistical significance at 1%, and 10% respectively.

Table 12

Pure infrastructure stock and economic growth.

Level of infrastructure stock Growth in infrastructure stock

Full instrument set Restricted instrument set

Lagged GDP per capita -3.б98*** (0.558) -3.492*** (0.729) -4.38б*** (-0.б89)

Infrastructure stock index 0.188 (0.215) -0.166 (0.328) -0.089 (-0.430)

Human development index 4.270* (0.225) 2.760(4.109) 5.502* (0.165)

Money supply (M2) -4.002*** (1.272) -3.450*** (1.073) -3.287*** (-3.205)

Terms of trade shocks 2.74б (2.883) 2.бб1 (2.827) 1.258 (0.бб4)

Inflation Exchange rate -0.125 (0.513) -0.436 (0.394) -0.775* (-1.752)

Government consumption 0.631 (4.384) 4.165 (5.343) 8.571 (1.438)

0.685 (1.784) 0.706 (1.896) -1.054 (-0.837)

Control of corruption 6.132(4.391) 3.433 (3.485) 1.281 (0.354)

J-statistic p-value 0.б1 0.55 0.49

Robust standard errors are in parentheses. ***, and * represent statistical significance at 1%, and 10% respectively.

nomic growth in the Sub-Saharan African region, both in levels and in annual improvement. The implication of this result is that provision of infrastructure is not important if it is not significant as to reach most economic agents in order to foster widespread involvement in economic activity. This is critical in Sub-Saharan Africa where the provision of infrastructure is sometimes concentrated in urban areas where only a minority of the populace resides, leaving the bulk of the population with minimal access to infrastructure and hence excluded from the mainstream of their countries' economy. This result does not seem to speak directly to the role the extant literature has given physical infrastructure in influencing economic growth in the region (Calderón and Servén, 2010; Fedderke et al., 2006). This is an important aspect of our contribution to the literature: research effort needs to clearly delineate the aspects of infrastructure that are relevant to economic growth in order for it to inform policy in a productive manner.

5.3. Infrastructure and economic development

In Table 1, we saw that levels of infrastructure development are more highly correlated with levels of income than they are

correlated with economic growth. Therefore, in this section, we seek to establish whether a significant relationship exists between infrastructure variables (levels and increment/growth) and economic development (income level). In line with convention, we represent economic development by GDP per capita.

The pertinent results are reported in Table 13. The same baseline model specification deployed in the economic growth estimation is adopted in the determination of the extent, if any, to which public infrastructure relates to economic development. Unsurprisingly, the results of the economic development model estimation are largely similar to those of economic growth, particularly per the findings reported in Tables 7, 8 and 11. The results in Table 13 show that increments in infrastructure stock are important for economic development (coefficient of 0.127 at the 5% significance level). Similarly, increments in infrastructure quality, though not statistically significant, are positively related to economic development. Further, as in the case of the economic growth model estimation, human capital and institutions largely have significantly positive coefficients; financial development is statistically negatively related to economic development. However, at variance with the economic growth model result, inflation (which can be argued to reflect relative excess aggregate demand) statistically and positively relate to economic development.

18 O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx—xxx

Table 13

Infrastructure development and economic development.

Infrastructure level % change in infrastructure

Eq. (1) Eq. (2) Eq. (3) Eq. (1) Eq. (2) Eq. (3)

Lagged GDP per capita 0.446*** (0.091) 0.401*** (0.102) 0.560***(0.078) 0.460*** (0.118) 0.400*** (0.096) 0.467*** (0.072)

Africa infrastructure index -0.031 (0.050) 0.127* (0.069)

Infrastructure quality index 0.002 (0.013) 0.003 (0.011)

Infrastructure stock index 0.003 (0.003) 0.001 (0.003)

Human development index 0.491 (0.315) 0.578* (0.313) 0.357 (0.340) 0.558 (0.402) 0.532 (0.323) 0.691** (0.319)

Money supply (M2) -0.038*** (0.010) -0.024* (0.014) -0.029*** (0.009) -0.029** (0.015) -0.025* (0.014) -0.032*** (0.009)

Terms of trade shocks -0.006 (0.023) 0.003 (0.029) 0.021 (0.024) -0.017 (0.022) -0.001 (0.028) 0.001 (0.024)

Inflation 0.008** (0.004) 0.011** (0.005) -0.000 (0.005) 0.004 (0.005) 0.011* (0.006) -0.003 (0.005)

Exchange rate -0.068 (0.075) -0.022 (0.092) -0.045 (0.056) -0.072(0.123) -0.036 (0.098) 0.073 (0.058)

Government consumption 0.019 (0.020) 0.024 (0.021) 0.026 (0.027) 0.015 (0.025) 0.022 (0.022) 0.011 (0.016)

Control of corruption 0.054* (0.032) 0.060* (0.036) 0.067 (0.040) 0.037 (0.032) 0.060 (0.037) 0.035 (0.036)

J-statistic p-value 0.43 0.56 0.35 0.59 0.47 0.45

Robust standard errors are in parentheses. ***, **, and * represent statistical significance at 1%, 5% and 10% respectively.

6. Concluding remarks

Given the well-known public infrastructure deficit in Sub-Saharan African (SSA) countries, its speculated constraint on economic growth and development, and the many programs (especially at the regional level) that have been put forth in attempts to address these, we set out in this paper to explore the true nature of the relation between public infrastructure and economic growth in more comprehensive ways than have hitherto been done. We mapped, at both the aggregate as well as individual infrastructure scopes, the evolution of public infrastructure in SSA; and confirmed the extent to which the region substantially lags behind most other regions of the developing world in public infrastructure endowments. The preliminary outcome of this detailed mapping amply made a compelling case for further investigation of the relation between infrastructure and economic growth and development in SSA.

Taking a research methodology tack that would be most useful at informing and guiding cross-country regional effort at reversing public infrastructure deficit for the purpose of enabling economic growth, we constructed two multidimensional indexes reflective of the access to and quality of public infrastructure endowments in SSA countries. More specifically, we examined (1) the effects of infrastructure access and quality on economic growth and development, respectively; (2) the effects of increments in infrastructure access and quality on economic growth; and (3) the intermediating effects of these infrastructure measures on 'most infrastructure pertinent' drivers of economic growth; with these baseline model tests followed by a series of relevant robustness checks.

Our results show that rather than the sheer stock/access to or quality of infrastructure being relevant for economic growth in an environment characterized by low basic infrastructure endowments, such as Sub-Saharan Africa, it is the spending on infrastructure and increments in the access to infrastructure that unsurprisingly influence growth in the region. Interestingly, these significant associations, especially those of infrastructure spending, are more important for lesser developed economies of the region than for the relatively more developed economies,

which uncommonly have better than near-zero stock of infrastructure. In addition to these direct links between the target variables, we find that infrastructure access also relates significantly to economic growth indirectly via trade competitiveness (export diversification); and infrastructure quality also indirectly influences economic growth via cross-border capital flows and export diversification, both of which are fairly established drivers of economic growth.

Among factors that have been documented in the literature as relevant growth antecedents, which served as control variables in our model specifications, we found human capital and institutions to have featured consistently and mostly significantly positively in the majority of our models' estimations. Conversely and with fewer statistically significant coefficients, financial development featured negatively in many of the same model estimations.

The overarching policy upshot of our findings is that efforts aimed at reversing Africa's pervasive infrastructure deficit, especially in order to enable economic growth and development, must be carefully nuanced. Our results suggest that current endowments may be insufficient to meaningfully impact macroeconomic activity. Therefore, emphasis on quality of current infrastructure stock appears not to be that critical; unless, of course, it is in regards of a SSA country like South Africa which uncharacteristically possesses infrastructure endowment levels that have reached or surpassed an implied threshold level necessary for enabling incremental aggregate economic activity. Implicit in this latest observation is a call for extended research that would explicitly examine the importance of infrastructure endowment threshold in relation to economic growth.

This last observation flags an important area for a follow-up research on the topic of our study. It would be useful to ascertain whether or not a threshold level of infrastructure endowment is necessary before infrastructure could fulfill its touted huge promise of enabling economic growth and/or development. This quest is even much more relevant given the near-zero level of endowment of almost all forms of public infrastructure of SSA, relative to other regions, which we documented in the background to our analysis (Section 1). Further, the nature of results

O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx—xxx

from our infrastructure constructs, points to a need to explore further the issue of evolving a more robust proxy of infrastructure at the macro-level. Similarly, in the light of existing findings of country-specific studies which suggest that certain individual infrastructure, such as electricity and ICT, have significant effects on the economy, it would be a worthwhile exercise to ascertain whether some public infrastructure are more important than others, particularly in the SSA kind of environment. This quest will certainly be achievable as better quality individual infrastructure data become available.


We appreciate the useful comments of participants of the 2015 Forum for Economists Conference in Amsterdam, Netherlands. Funding support of Economic Research Southern Africa is gratefully acknowledged. All errors are ours.

Appendix A.

A.1 List of countries

The countries are: Angola, Benin, Botswana, Burkina Faso, Burundi, Cameroon, Cape Verde, Central African Republic, Chad, Comoros, Congo DR, Congo Republic, Cote d'Ivoire, Djibouti, Equatorial Guinea, Eritrea, Ethiopia, Gabon, Gambia, Ghana, Guinea, Guinea Bissau, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali, Mauritania, Mauritius, Mozambique, Namibia, Niger, Nigeria, Rwanda, Senegal, Seychelles, Sierra Leone, South Africa, Swaziland, Tanzania, Togo, Uganda, Zambia, and Zimbabwe.

A.2 Preliminary investigation of mediating variables Table Al.


Acemoglu, D., Johnson, S., Robinson, J., 2002. Reversal of fortune: geography and institutions in the making of the modern world income distribution. Q. J. Econ. 117, 1231-1294.

Acemoglu, D., Johnson, S., Robinson, J.A., 2005. Institutions as the fundamental cause of long-run growth. In: Aghion, P., Durlauf, S. (Eds.), Handbook of Economic Growth. North-Holland, Amsterdam, pp. 385-472.

AfDB, 2011. Africa's Infrastructure Outlook, 2040. Programme for Infrastructure Development in Africa (PIDA). African Development Bank, Tunis.

AfDB, 2013a. An Integrated Approach to Infrastructure Provision in Africa. African Development Bank.

AfDB, 2013b. The Africa Infrastructure Development Index (AIDI). African Development Bank, Tunis.

African Union, 2014. Programme for Infrastructure Development in Africa (PIDA): addressing the infrastructure gap in Africa, to speed up regional integration. In: Seventh Conference of African Ministers in Charge of Integration, 14-18 July 2014. Swaziland.

Aizenman, J., Jinjarak, Y., Park, D., 2015. Financial development and output growth in developing Asia and Latin America: a comparative sectorial analysis. Natl. Bur. Econ. Res., NBER Working Paper w20917.

Ajakaiye, O., Ncube, M., 2010. Infrastructure and economic development in Africa. J. Afr. Econ. 19 (Suppl. 1), i3-i12, AERC.

Arellano, M., Bond, S., 1991. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Rev. Econ. Stud. 58, 277-297.

Arellano, M., Bover, O., 1995. Another look at the instrumental variable estimation of error-component models. J. Econom. 68, 29-51.

Aschauer, D.A., 1993. Genuine economic returns to infrastructure investment. Policy Stud. J. 21, 380-390.

Ayogu, M., 2007. Infrastructure and economic development in Africa: a review. J. Afr. Econ. 16 (Suppl. 1), 75-126, AERC.

Barro, R.J., 1990. Government spending in a simple model of endogenous growth. J. Political Econ. 98, S102-S125.

Barro, R.J., 1996. Determinants of economic growth: a cross-country empirical study. Natl. Bur. Econ. Res., NBER Working Paper 5698.

Becker, G.S., Glaeser, E.L., Murphy, K.M., 1999. Population and economic growth. Am. Econ. Rev. 89, 145-149.

Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel data models. J. Econom. 87, 11-143.

Boopen, S., 2006. Transport infrastructure and economic growth: evidence from Africa using dynamic panel estimates. Empir. Econ. Lett. 5, 37-52.

Bullock, R., 2009. Africa Infrastructure Country Diagnostic: Railways in Sub-Saharan Africa. World Bank, Washinton DC, Unpublished Manuscript.

Calderón, C., Servén, L., 2010. Infrastructure and economic development in Sub-Saharan Africa. J. Afr. Econ. 19 (Suppl. 1), i13-i87, AERC.

Table A1

Preliminary relations between infrastructure and potential intermediating variables.

Do these variables explain economic growth? Do these variables explain economic development?

Doing business cost Export diversification Net FDI inflows -0.150(0.184) 5.965* (3.338) 0.050** (0.025) Doing business cost Export diversification Net FDI inflows 0.190** (0.090) -0.577 (0.639) -0.021*** (0.008)

Does infrastructure explain these variables?

Infrastructure level Change in infrastructure

Access Quality Access Quality

Doing business cost Export diversification Net FDI inflows -1.130*** (0.077) -0.018*** (0.004) -1.413 (0.915) -2.149*** (0.326) 0.019 (0.016) -9.617*** (3.315) Doing business cost Export diversification Net FDI inflows 0.324 (0.801) 1.466** (0.568) 0.018 (0.053) 0.037 (0.029) 2.761 (11.017) 1.901 (4.689)

We run panel OLS regressions with one explanatory variable and a constant in each case. The table reports the coefficient estimates of the explanatory variable with their standard errors in parentheses. *"**. and * denote significance at 1%, 5% and 10% levels respectively.

O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx—xxx

Calderón, C., Liu, L., 2003. The direction of causality between financial development and economic growth. J. Dev. Econ. 72, 321-334.

Canning, D., Pedroni, P., 2004. The Effect of Infrastructure on Long-run Economic Growth. Mimeo. Harvard University.

Chang, Ha-Joon, 2011. Institutions and economic development: theory, policy and history. J. Inst. Econ. 7, 473-498.

Chowdhury, A., 2015. Terms of trade shocks and private savings in the developing countries. J. Comp. Econ. 43, 1122-1134.

Cinyabuguma, M., Putterman, L., 2011. Sub-Saharan growth surprises: being heterogeneous, inland and close to the equator does not slow growth within Africa. J. Afr. Econ. 20, 217-262.

Dicle, M.F., Levendis, J., 2013. Estimating Geweke's (1982) measure of instantaneous feedback. Stata J. 13 (1), 136-140.

Dinkelman, T., 2011. Effects of rural electrification on employment: new evidence from South Africa. Am. Econ. Rev. 101, 3078-3108.

Easterly, W., Levine, R., 2001. What have we learned from a decade of empirical research on growth? It's not factor accumulation: stylized facts and growth models. World Bank Econ. Rev. 15, 177-219.

Eberhard, A., Rosnes, O., Shkaratan, M., Vennemo, H., 2011. Africa's Power Infrastructure Investment, Integration, Efficiency. The World Bank, Washington DC.

Economic Commission for Africa (ECA), 2005. Assessing Regional Integration in Africa, ECA Policy Research Paper. ECA, Addis Ababa.

Egert, B., Kozluk, T., Sutherland, D., 2009. Infrastructure and Growth: Empirical Evidence. University of Michigan, William Davidson Institute Working Paper Number 957.

Esfahani, H.S., Ramírez, M.T., 2003. Institutions infrastructure, and economic growth. J. Dev. Econ. 70, 443-477.

Estache, A., Vagliasindi M., 2007. Infrastructure for accelerated growth for Ghana: needs and challenges, Unpublished Manuscript.

Estache, A., Speciale, B., Veredas, D., 2006. How Much Does Infrastructure Matter to Growth in Sub-Saharan Africa? The World Bank, Washington DC.

Fedderke, J., Perkins, P., Luiz, J., 2006. Infrastructural investment in long-run economic growth: South Africa, 1875-2001. World Dev. 34, 1037-1059.

Fedderke, J., Garlick, R., 2008. Infrastructure Development and Economic Growth in South Africa: A Review of the Accumulated Evidence. Policy Paper Number 12, School of Economics, University of Cape Town.

Foster, V., Briceño-Garmendia, C. (Eds.), 2010. Africa's Infrastructure: A Time for Transformation. The World Bank, Washington DC.

Geweke, J., 1982. Measures of conditional linear dependence and feedback between time series. J. Am. Stat. Assoc. 77, 304-313.

Ghura, D., Hadjimichael, M.T., 1996. Growth in sub-Saharan Africa. IMF Staff Pap. 43, 605-634.

Glaeser, E.L., La Porta, R., Lopez-de-Silanes, F., Sheifer, A., 2004. Do institutions cause growth? J. Econ. Growth 9, 271-303.

Gramlich, E., 1994. Infrastructure investment: a review essay. J. Econ. Lit. 32, 1176-1196.

Hassan, M.K., Sanchez, B., Yuc, J.-S., 2011. Financial development and economic growth: new evidence from panel data. Q. Rev. Econ. Finance. 51, 88-104.

Holtz-Eakin, D., Newey, W., Rosen, H.S., 1988. Estimating vector autoregressions with panel data. Econometrica 56, 1371-1395.

Kimenyi, M.S., 2007. Institutional Infrastructure to Support 'Super Growth' in Kenya: Governance Thresholds, Reversion Rates and Economic Development. Department of Economics, University of Connecticut, Economics Working Paper.

Kularatne, C., 2006. Social and economic infrastructure impacts on economic growth in South Africa. In: DPRU Conference, 18-20 October, Johannesburg.

Kumar, N., 2001. Infrastructure Availability, Foreign Direct Investment Inflows and Their Export-orientation: A Cross-Country Exploration. Research and Information System for Developing Countries, New Delhi.

Kumo, W.L., 2012. Infrastructure Investment and Economic Growth in South Africa: A Granger Causality Analysis. African Development Bank, Tunis, Working Paper No. 160.

Lee, K., Kim, B.Y., 2009. Both institution and policies matter but differently for different income groups of countries: determinants of long-run economic growth revisited. World Dev. 37, 533-549.

Li, X., Liu, X., 2005. Foreign direct investment and economic growth: an increasingly endogenous relationship. World Dev. 33, 393-407.

Mbaku, J.M., 2013. Building opportunities: addressing Africa's lack of infrastructure. The Brookings Inst., Afr. Growth Initiat.

Menyah, K., Nazlioglu, S., Wolde-Rufael, Y., 2014. Financial development, trade openness and economic growth in African countries: new insights from a panel causality approach. Econ. Model. 37, 386-394.

Mthanti, T., Ojah, K., 2016. Entrepreneurial orientation (EO): measurement and policy implications of entrepreneurship at the macroeconomic level. Res. Policy, Forthcoming.

Muhanji, S., Ojah, K., 2011. Management and sustainability of external debt: a focus on the emerging economies in Africa. Rev. Dev. Finance 1, 184-206.

Muhanji, S., Ojah, K., 2016. Governance infrastructure and indebtedness of African countries: do regional blocs matter? N. Am. J. Econ. Finance 36, 123-153.

Ndulu, B.J., 2006. Infrastructure, regional integration and growth in sub-Saharan Africa: dealing with the disadvantages of geography and sovereign fragmentation. J. Afr. Econ. 15 (Suppl. 2), 212-244, AERC.

Nyamongo, E.M., Misati, R.M., Kipyegon, L., Ndirangu, L., 2012. Remittances, financial development and economic growt in Africa. J. Econ. Bus. 64, 204-260.

Obwona, M., 2001. Determinants of FDI and their impact on economic growth in Uganda. Afr. Dev. Rev. 13, 46-81.

Ojah, K., Kodongo, O., 2015. Financial development in Africa: reflections and the way forward. In: Monga, C., Yifu-Lin, J. (Eds.), The Oxford Handbook of Africa and Economics: Policies and Practices II. Oxford University Press,

Pritchett, L., 2004. Towards a New Consensus for Addressing the Global Challenge of the Lack of Education. Center for Global Development, Working Paper No. 43.

Redding, S., Venables, A.J., 2004. Economic geography and international inequality. J. Int. Econ. 62, 53-82.

Reinikka, R., Svensson, J., 1999. How Inadequate Provision of Public Infrastructure and Services Affects Private Investment. The World Bank, Policy Research Working Paper Series 2262.

Roberts, J., Fagernas, S., 2004. Why is Bangladesh Outperforming Kenya? A Comparative Study of Growth and Its Causes Since the 1960. Overseas Development Institute, Economics and Statistics Analysis Unit, London, Working Paper No. 5.

Rodrik, D., 2003. Introduction. In: Rodrik, D. (Ed.), In Search of Prosperity: Analytic Narratives of Economic Growth. Princeton Universit Press, New Jersey, pp. 1-19.

Rodrik, D., 2014. An African growth miracle? Natl. Bur. Econ. Res., No. w20188.

Rodrik, D., Subramanian, A., Trebbi, F., 2002. Institutions rule: the primacy of institutions over geography and integration in economic development. Natl. Bur. Econ. Res., Working Paper No. 9305.

Roodman, D., 2006. How to do Xtabond2: An Introduction to Difference and System GMM in Stata. Center for Global Development, Working Paper Number 103.

Roodman, D., 2009. A note on the theme of too many instruments. Oxf. Bull. Econ. Stat. 71, 135-158.

Sachs, J., Malaney, P., 2002. The economic and social burden of malaria. Nature 415, 680-685.

Sachs, J., McArthur, J., Schmidt-Traub, G., Kruk, M., Bahadur, C., Faye, M., McCord, G., 2004. Ending Africa's poverty trap. Brook. Pap. Econ. Act., 117-240.

Sanchez-Robles, B., 1998. Infrastructure investment and growth: some empirical evidence. Contemp. Econ. Policy 16, 98-108.

Simon, J.L., Gobin, R., 1980. The relationship between population and economic growth in LDC's. In: Simon, J.L., DaVanzo, J. (Eds.), Research Population Economics: A Research Annual, vol. 2. JAI Press, Greenwich, Connecticut, pp. 215-234.

O. Kodongo, K. Ojah/Review of Development Finance xxx (2016) xxx—xxx 21

Weinhold, D., Reis, E.J., 2001. Model evaluation and causality testing in short World Bank, 2006. World Development Report: Equity and Development. World panels: the case of infrastructure provision and population growth in the Bank.

Brazilian Amazon. J. Reg. Sci. 41, 639-657.