ADB Economics Working Paper SeriesStéphane Straub and Akiko Terada-Hagiwara November 2010 Stéphane Straub is Professor at the Toulouse School of Economics, Arqade; Akiko Terada-Hagiwara

ADB Economics Working Paper Series

Infrastructure and Growth in Developing Asia

Stéphane Straub and Akiko Terada-HagiwaraNo. 231 | November 2010

ADB Economics Working Paper Series No. 231

Infrastructure and Growth in Developing Asia

Stéphane Straub and Akiko Terada-Hagiwara November 2010

Stéphane Straub is Professor at the Toulouse School of Economics, Arqade; Akiko Terada-Hagiwara is Economist in the Macroeconomics and Finance Research Division, Economics and Research Department, Asian Development Bank. This paper was prepared as a background material for the theme chapter, “The Future of Growth in Asia” in the Asian Development Outlook 2010 Update, available at www.adb.org/economics. The authors thank Douglas Brooks, Biswanath Bhattacharyay, and other participants in the workshops in Manila for their helpful comments; and Shiela Camingue for her superb assistance.

http://www.adb.org/economics

http://www.adb.org/economics

Asian Development Bank6 ADB Avenue, Mandaluyong City1550 Metro Manila, Philippineswww.adb.org/economics

©2010 by Asian Development BankNovember 2010ISSN 1655-5252Publication Stock No. WPS102881

The views expressed in this paperare those of the author(s) and do notnecessarily reflect the views or policiesof the Asian Development Bank.

The ADB Economics Working Paper Series is a forum for stimulating discussion and eliciting feedback on ongoing and recently completed research and policy studies undertaken by the Asian Development Bank (ADB) staff, consultants, or resource persons. The series deals with key economic and development problems, particularly those facing the Asia and Pacific region; as well as conceptual, analytical, or methodological issues relating to project/program economic analysis, and statistical data and measurement. The series aims to enhance the knowledge on Asia’s development and policy challenges; strengthen analytical rigor and quality of ADB’s country partnership strategies, and its subregional and country operations; and improve the quality and availability of statistical data and development indicators for monitoring development effectiveness.

The ADB Economics Working Paper Series is a quick-disseminating, informal publication whose titles could subsequently be revised for publication as articles in professional journals or chapters in books. The series is maintained by the Economics and Research Department.

Contents

Abstract v

I. Infrastructure: A Review of Issues 1

II. Where Does Developing Asia Stand? 3

A. Stylized Facts 3 B. Demand for and Gaps in Infrastructure Services 8

III. Theory and Existing Evidence 10

A. Theory 10 B. Empirics 12

IV. Empirical Analysis 18

A. Cross-country Regressions 18 B. Growth Accounting 32

V. Conclusion 39

Appendix 41

References 43

Abstract

This paper presents the state of infrastructure in developing Asian countries. It applies two distinct approaches (growth regressions and growth accounting) to analyze the link between infrastructure, growth, and productivity. The main conclusion is that a number of countries in developing Asia have significantly improved their basic infrastructure endowments in the recent past, and this appears to correlate significantly with good growth performances. However, the evidence seems to indicate that this is mostly the result of factor accumulation (a direct effect), while the impact on productivity is inconclusive.

I. Infrastructure: A Review of Issues

The relevance of infrastructure for development outcomes comes from the fact that it provides both final consumption services to households and key intermediate consumption items for production. Crude estimates from the literature indicate that between one third and one half of infrastructure services are used as final consumption by households (Prud’homme 2005, Fay and Morrison 2007), while the rest corresponds to firms.

Infrastructure should thus play an important role to support growth and poverty reduction. On the supply side, there is both a direct channel (infrastructure capital stock serves as a production factor), and an indirect one (improved infrastructure affects technological progress). An increase in the stock of infrastructure capital is argued to have a direct, increasing effect on the productivity of the other factors. It is also believed to generate important externalities across a range of economic activities, which could possibly have a larger net effect than what is expected from a simple factor accumulation effect. These indirect effects could operate through various channels; among others, through labor productivity gains resulting from improved information and communication technologies; reductions in time wasted commuting to work and stress; improvements in health and education; and through improvements in economies of scale and scope throughout the economy.

From a demand side point of view, infrastructure provides people with services they need and want—water and sanitation; power for heat, cooking, and light; telephone and computer access; and transport. The absence of some of the most basic infrastructure services is an important dimension of what we often mean when we talk about poverty. Increasing level of infrastructure stock, therefore, has direct implication for poverty reduction.

Infrastructure appears significantly related to per capita gross domestic product (GDP) growth in the past decades mainly through accumulating infrastructure capital stock as a production factor. Yet, the assumed link between infrastructure and growth relies mostly on the assumption that variations of infrastructure services availability and quality are one of the main drivers of differences in firms’ productivity across regions and countries. The link between the two, however, is not particularly clear from the data (Figure 1). Linking the availability and the quality of infrastructure with economic growth or productivity is, therefore, still subject to considerable debate and uncertainty.

Figure 1: Average Productivity Growth, 2005–2009 versus Infrastructure Quality, 2005

ARM

BAN

PRC

HKG

INOIND KOR

SRIMON

PAK

SINTHA

VIE TAP

MALPHI

CAM

KGZ

-10

-5

0

5

10

Aver

age

Ann

ual C

hang

e in

Fac

tor P

rodu

ctiv

ity, 2

005–

2009

2 3 4 5 6 7

Quality of Overall Infrastructure, 2005

Developing Asia Rest of the World

ARM = Armenia; BAN = Bangladesh; CAM = Cambodia; HKG = Hong Kong, China; IND = India; INO = Indonesia; KGZ = Kyrgyz Rep.; KOR = Rep. of Korea; MAL = Malaysia; MON = Mongolia; PAK = Pakistan; PHI = Philippines; PRC = People’s Rep. of China; SIN = Singapore; SRI = Sri Lanka; TAP = Taipei,China; THA = Thailand; VIE = Viet Nam.

Sources: Authors’ calculations using data from World Economic Forum (2005), and Barro and Lee (2010).

There are many ways infrastructure shortcomings translate into productivity efficiency losses. For example, access to markets and interactions with potential clients rely on the existence and reliability of transport and telecommunication networks, and when these fail, firms may suffer from lack of access to market opportunities, higher logistic costs and inventory levels, or information losses.1 Similarly, investment and technological choices may be affected by the efficiency of electricity networks, in the sense that frequent power outages and unstable voltage induce high costs and greater risk of machinery breakdown. Growing evidence shows that firms respond by making suboptimal technological choices, by investing in remedial equipment such as power generators, or by deferring some types of investments (Alby, Dethier, and Straub 2009).

1 See for example Guasch and Kogan (2001), Li and Li (2009), and Jensen (2007) .

2 | ADB Economics Working Paper Series No. 231

The objective of the present paper is twofold. First, it presents a detailed update of the state of infrastructure in developing Asian countries. Then it applies two distinct approaches (growth regressions and growth accounting) to analyze the link between infrastructure, growth, and productivity. The main conclusion is that a number of countries in developing Asia have significantly improved their basic infrastructure endowments in the recent past, and this appears to correlate significantly with good growth performances. However, the evidence seems to indicate that this is mostly the result of factor accumulation (a direct effect), while the impact on productivity is rather inconclusive.

The rest of the paper is organized as follows. The next section provides an overview of past developments in infrastructure capital accumulation in developing Asia. Section III reviews the existing empirical literature.2 Section IV presents empirical results investigating the theoretical effects discussed in Section III. Section V concludes.

II. Where Does Developing Asia Stand?

A. Stylized Facts

While some developing Asian countries have far better infrastructure than others, overall, the region remains below the world average in terms of both its quantity and quality. Quality of infrastructure as well as access to infrastructure services is also uneven across region and country. Except for the five relatively advanced economies—Hong Kong, China; the Republic of Korea; Malaysia; Singapore; and Taipei,China (or Asia-5)—the quality of infrastructure in developing Asia is largely lagging behind that of industrialized economies (Figure 2).

2 This is by no mean an exhaustive review of empirical contributions. For surveys of the empirical literature, see Gramlich (1994), Sturm et al. (1998), Romp and de Haan (2005), and Straub (2008 and 2010).

Infrastructure and Growth in Developing Asia | 3

Figure 2: Quality of Overall Infrastructure and GDP per Capita, Selected Economies, 2008

BRU

PRC

HKG

SRI

MON

MAL

SIN

THA

TIM

KOR

CAM

VIE INO

PHI

BANNEP

IND

TAP

GEO

0

2

4

6

8

Qua

lity

of O

vera

ll In

fras

truc

ture

, 200

8

6 8 10 12

Log per Capita GDP, 2008

Developing Asia Rest of the World OECD

BAN = Bangladesh; BRU = Brunei Darussalam; CAM = Cambodia; GEO = Georgia; HKG = Hong Kong, China; IND = India; INO = Indonesia; KOR = Rep. of Korea; MAL = Malaysia; MON = Mongolia; NEP = Nepal; PHI = Philippines; PRC = People’s Rep. of China; SIN = Singapore; SRI = Sri Lanka; TAP = Taipei,China; THA = Thailand; TIM = Timor-Leste; VIE = Viet Nam.

Note: Per capita GDP is based on purchasing power parity (PPP) in current international dollars. Quality of overall infrastructure refers to assessments of the quality of general infrastructure (e.g., transport, telephony, and energy) in an economy. 1 indicates extremely underdeveloped while 7 indicates extensive and efficient by international standards. Data cover 132 economies.

Sources: Author's calculation using data from World Economic Forum (2009) and International Monetary Fund (2010).

While aggregate measures of infrastructure stock are not systematically available, the quality measure based on a perception survey (qualitative data) from World Economic Forum (WEF 2009) suggests a significant improvement over the past 4 years. In particular, significant upgrading in the quality of transport, telephony, and energy infrastructure was reported for Cambodia, Georgia, and Sri Lanka. But there are countries such as Bangladesh and Mongolia that registered worsening of the overall infrastructure quality over the same period despite the fact that these countries should only improve given the very low level of existing infrastructure stock (Figure 3). Further, the same survey indicates a worrisome situation in terms of the quality of electricity supply in developing Asia. Close to half of the 25 surveyed Asian countries reported that the quality of electricity supply worsened since 2005. The worsening of the situation appears particularly severe in some South Asian economies including Bangladesh, Nepal, and Pakistan.


Figure 3: Change in Quality of Overall Infrastructure, 2009 versus 2005

-1 0 1 2

USA JPN

World

MAL INO

BAN MON

SIN THA TAJ PAK VIE

TIM KAZ IND NEPTAP

HKG PHI

KGZKOR ARMPRC AZE

CAM SRI

GEO

Percentage Points ARM = Armenia; AZE = Azerbaijan; BAN = Bangladesh; CAM = Cambodia; GEO = Georgia; HKG = Hong Kong, China; IND = India; INO = Indonesia; KAZ = Kazakhstan; KGZ = Kyrgyz Rep.; KOR = Rep. of Korea; JPN = Japan; MAL = Malaysia; MON = Mongolia; NEP = Nepal; PAK = Pakistan; PHI = Philippines; PRC = People’s Rep. of China; SIN = Singapore; SRI = Sri Lanka; TAJ = Tajikistan; TAP = Taipei,China; THA = Thailand; TIM = Timor-Leste; USA = United States; VIE = Viet Nam. Note: Quality of overall infrastructure refers to assessments of the quality of general infrastructure (e.g., transport, telephony, and energy) in an economy. 1 indicates extremely underdeveloped while 7 indicates extensive and efficient by international standards. 2009 data cover 133 economies, while that of 2005 covers 117 countries.Source: World Economic Forum (2005 and 2009).

A closer look at different types of infrastructure capital over a longer period reveals a similar trend—robust growth but uneven across countries and regions. Electricity generation capacity in developing Asia grew by 4.3% annually (Figure 4) and more than doubled during the period between 1990 and 2007. But there is a divergence within developing Asia. Central Asia and Pacific economies were an exception to the rapid growth trend and achieved only marginal growth, if not contraction, during this period.


Among the good performers, particularly the Asia-5, capacity growth slowed down by 2000, as the economies matured. Meanwhile some lower-income countries such as Cambodia (since 1995), the People's Republic of China (PRC), and Viet Nam continued with a robust expansion in generation capacity in the 2000s (Figure 5).

Figure 4: Annual Growth in Electricity Generation, 1990–2007

-5 0 5 10

World

Advanced Economies

Rest of the World

Developing Asia

Central Asia

East Asia

South Asia

Southeast Asia

Asia-5

Average Annual Change (percent)

Note: Starting year for Cambodia is 1995. Asia-5 refers to Hong Kong, China; the Republic of Korea; Malaysia; Singapore; and Taipei,China.

Source: Authors’ calculations using data from World Development Indicators online database, accessed 19 August 2010 (World Bank 2010).

Increase in other infrastructure stocks such as in telecommunication and internet connections has also been significant. Internet users, for example, more than tripled since 2000, or from five users per 100 persons in 2000, to 16 users by 2008 (Figure 6). This growth, however, has again been largely driven by the Asia-5. Nevertheless, the current level lags far behind that of Latin American economies, where for example, 28 people out of 100 have access to the internet in 2008, a level comparable to the world average.

Share of paved road (as a percentage of the total) increased from less than half in 1990 to reach almost 60% of the total roads by mid-2000 (Figure 7). But again the dispersion of this infrastructure stock index is huge across economies. Many of developing Asia’s roads such as in Bangladesh, Cambodia, Mongolia, and Papua New Guinea are hardly paved. Meanwhile almost all roads are paved in Armenia; Hong Kong, China; Kazakhstan; and Singapore.


Figure 5: Electricity Growth by Selected Asian Developing Economies, 1990–2007

0 5 10 15 20

Hong Kong, China

Philippines

Singapore

Myanmar

India

Brunei Darussalam

Thailand

Korea, Rep. of

Indonesia

Malaysia

China, People's Rep. of

Viet Nam

Cambodia*

Average Annual Change (percent)

*Change over 1995.Source: Authors’ calculations using data from World Development Indicators online database, accessed 19 August 2010 (World Bank

2010).

Figure 6: Internet Users by Region, 2008

0 20 40 60 80

World

Advanced Economies

Rest of the World

Developing Asia

Central Asia

East Asia

South Asia

Southeast Asia

Paci�c

Asia-5

Number per 100 Inhabitants

Note: Asia-5 refers to Hong Kong, China; the Republic of Korea; Malaysia; Singapore; and Taipei,China.Source: Authors’ calculation using data from World Development Indicators online database, accessed 19 August 2010 (World Bank

2010).


Figure 7: Paved Roads by Region, 2006

0 20 40 60 80 100

World

Advanced Economies

Rest of the World

Developing Asia

Central Asia

East Asia

South Asia

Southeast Asia

Paci�c

Asia-5

Percent of Total Roads

Note: Asia-5 refers to Hong Kong, China; the Republic of Korea; Malaysia; Singapore; and Taipei,China. Data for the Pacific pertain to share in 2000. Paved roads are those surfaced with crushed stone and hydrocarbon binder or bituminized agents, with concrete, or with cobblestones, measured in length and as a percentage of all the country’s roads.

Source: Authors’ calculation using data from World Development Indicators online database, accessed 19 August 2010 (World Bank 2010).

Are these developments sufficient to improve productivity? WEF (2009) finds that the inadequate supply of infrastructure is a problem for doing business in countries such as Bangladesh, Nepal, and Viet Nam. This closer look at various types of infrastructure indexes in developing Asia reveals that overall the improvement has not been enough to catch up with developed economies, and is not sufficient to translated into an important productivity-enhancing factor in many developing Asian countries.

B. Demand for and Gaps in Infrastructure Services

1. Urbanization

The demand for infrastructure is increasing and expected to soar further, in particular because of increasing urbanization. Except for Hong Kong, China; the Republic of Korea; and Singapore where more than 80% of the total population already live in urban areas, a majority of people in developing Asia live in rural areas. As economies grow rapidly and urbanization advances, however, the demand for more and highquality infrastructure is expected to further accelerate.


The proportion of a country’s population living in urban areas is highly correlated with its level of income (Bloom et al. 2008). As economies grow, the urban share of developing Asia’s population is rising exponentially (Figure 8). From 28.9% in 1960 to 43.0% in 2000, it is expected to rise to 56.9% by 2030 (data from United Nations population projections). Urban areas offer richer market structures, and there is strong evidence that workers in urban areas are individually more productive and earn more than rural workers. Densely populated urban areas provide markets for output, inputs, labor, and other services and allow firms to profit from economies of scale and scope, specialization, and rapid diffusion of knowledge innovation.

Figure 8: Urbanization and Income per Capita, Selected Economies, 2005

KOR

SRI

SIN

THA

VIE

HKG

MAL

PRCIND

PAK PHIINO

BRU

KAZ

0

20,000

40,000

60,000

80,000

Per C

apita

GD

P, PP

P (c

onst

ant 2

005

inte

rnat

iona

l $)

0 20 40 60 80 100

Urban Population Shares (percent)

Developing Asia Rest of the World

BRU = Brunei Darussalam; HKG = Hong Kong, China; IND = India; INO = Indonesia; KAZ = Kazakhstan; KOR = Rep. of Korea; MAL = Malaysia; PAK = Pakistan; PHI = Philippines; PRC = People’s Rep. of China; SIN = Singapore; SRI = Sri Lanka; THA = Thailand; VIE = Viet Nam.

Note: Data cover 181 economies.Source: World Development Indicators online database (accessed 19 August 2010).

Challenges are mounting, however. Rapid urbanization is associated with crowding, environmental degradation, and other impediments to productivity unless accompanied by appropriate infrastructure services. Poorly conceived investments today will have costly environmental, economic, and social impacts later. If appropriately designed and implemented, infrastructure should play an enormous role in maintaining the competitiveness of cities.


In particular, developing Asia needs to connect cities to international markets through enhanced transport infrastructure, telecommunications, and logistic services. Exploiting the region’s comparative advantage in high-value services and high-tech industry requires providing advanced communications and just-in-time delivery for integrated production chain to spread across a number of countries in developing Asia (Brooks and Hummels 2009). The urban investment climate, and hence employment prospects, can depend critically on the quality of urban infrastructure.

2. Urban–Rural Divide

While developing Asia is rapidly urbanizing, about 62.6% (or 2.2 billion people) still lived in rural areas as of 2008. Moreover, developing Asia’s poverty is overwhelmingly rural, and rural–urban disparities—across income as well as access to services—provoke political concerns and demands for inclusion in economic development.

In rural areas, poor road conditions for example restrict the all-weather access of local communities within and between towns/townships, while the lack of clean and reliable water supply inhibits economic growth and poverty alleviation. The proportion of population using an improved drinking water source in 2006, for example, is 90% for those living in urban areas as opposed to 68% for those in rural areas. The gap is particularly significant in economies such as the Lao People's Democratic Republic, Mongolia, Papua New Guinea, and Vanuatu. Urban–rural integrated infrastructure development is crucial to achieve inclusive growth. Failing to provide basic needs in infrastructure means disconnect from markets, basic education, medical services, and jobs, which often leads to a poverty trap.

III. Theory and Existing Evidence

This section reviews theory underlying the link between the infrastructure capital stock and growth and productivity, and also discusses the existing empirical literature.

A. Theory3

In a standard production function where factors are gross complements, an increase in the stock of infrastructure capital would have a direct, increasing effect on the productivity of the other factors. This is particularly clear if one thinks of cases of strong complementarities,4 for example, if roads or bridges investment provide access to previously inaccessible areas, enabling productive investment there, or if improvements of the electricity or telecommunications networks make the use of certain types of 3 See Straub (2010) for a detailed review.4 The theoretical formulation is in Kremer (1995) O-ring production function.


machineries possible. But because infrastructure capital is also believed to generate important externalities across a range of economic activities, it is possible that its net effect is larger than expected from a simple factor accumulation effect. The theoretical literature has discussed a number of channels for these indirect effects.

The first one is maintenance, private capital durability, and adjustment costs. There is growing evidence that infrastructure policy is biased toward the realization of new investments at the detriment of the maintenance of the existing stock. The main reasons appear to be political economy ones.5 As a consequence, the life span of the stock of both the infrastructure itself and of private capital that makes use of it such as trucks operating on low-quality roads or machines connected to unstable voltage lines is reduced, and operating costs increase.6 The case of palliative private investments in devices such as electricity generators is an extreme example of this.

Second, infrastructure appears to have a microeconomic impact through a number of channels, including labor productivity gains resulting from improved information and communication technologies, reductions in time wasted commuting to work and stress, and improvements in health and education among others. Moreover, such improvements are likely to induce additional investment in human capital in the medium and long term.

Finally, infrastructure may be the source of economies of scale and scope throughout the economy. For example, as roads and railroads improve, lowering transport costs, private firms benefit from economies of scale and more efficient inventory management.7 Similarly, enhanced access to communication devices, as was the case across the developing world in the last 2 decades with the growth of mobile telephony, is likely to result in efficient market clearing and enhanced competition as a result of improved information flows.8 Economic geography tells us that, as a result, we should expect different patterns of agglomeration, as well as changes in the pattern of specialization of agents and in their incentives to innovate (see for example Baldwin et al. 2003). In other words, changes in the nature and availability of key types of infrastructure are likely to have profound structural effects on the whole economy through agents’ and firms’ decisions.

Infrastructure investments, however, do not occur in isolation from other economic constraints. Unlocking some of the direct and indirect effects mentioned above may be conditioned by a number of practical issues, which are at present less well understood. First of all, infrastructure investments are often at least partly publicly financed, either through taxation or borrowing on financial markets, with the consequent risk of crowding out private investments. Another key question refers to sequencing. Which type of

5 Either on the financing side or linked to pork-barrel arguments. See Rioja (2003), Maskin and Tirole (2006), and Dewatripont and Seabright (2006), among others.

6 See Engel, Fischer, and Galetovic (2009) for a detailed analysis of the case of roads.7 See Li and Li (2009) for evidence in the case of the PRC.8 See Jensen (2007) for a striking example in the case of Indian fishermen.


infrastructure is more effective in supporting growth and should be prioritized? And which type of reforms supporting these investments, such as privatization, restructuring measures, regulation and introduction of competition, should be pursued and how? Because of their obvious context-dependence, the answers to these questions have to be empirical.

The next section reviews the existing empirical literature, with a special emphasis on contributions focusing on Asian countries.9 It discusses the theoretical effects identified above, including practical issues of financing and investment sequencing, showing that on all these aspects empirical evidence is somewhat lagging.

B. Empirics

1. Macroeconomic Level

The first generation of studies applying a production function approach augmented with some measure of infrastructure capital to state-level data for the United States (US) provided estimates of the output elasticity of infrastructure capital, varying between 0.3 and 0.5.10 Because these numbers imply a marginal product of around 100%, they have often been dismissed as unrealistic.

The subsequent literature has responded to these critics in several ways. On one hand, focus shifted to industry-level studies, such as Fernald (1999), who analyzed the impact of the large US interstate highway network built in the 1950s and 1960s. Although Fernald (1999) finds an output elasticity of road investment of around 0.35 for US industries that use roads more intensively, he also argues that this was a one-off effect, corresponding to a one-time boost in productivity and explaining the post-1973 slowdown in productivity.11

On the other hand, there were a number of econometric weaknesses in the early contributions that were pointed out by later studies.12 These weaknesses include the existence of state- or region-level unobserved effects, and potential reverse causality between output and infrastructure investment, which may generate an upward bias in the estimated coefficients. Unsurprisingly, a second generation of studies taking these concerns into account came up with significantly smaller estimates. For example, the survey by Romp and de Haan (2005) reports elasticities between 0.1 and 0.2; and Bom and Ligthart (2008) in their meta-analysis of 67 studies using public capital measures report an unconditional output elasticity of public capital of around 0.15. A more recent

9 This is by no mean an exhaustive review of empirical contributions. For surveys of the empirical literature, see Gramlich (1994), Sturm et al. (1998), Romp and de Haan (2005), and Straub (2008 and 2010).

10 The most widely quoted paper is Aschauer (1989). Gramlich (1994) provides a review of this early empirical literature.

11 Fernald estimates a 1% boost in productivity, while the post-1973 productivity slowdown was 1.3%. 12 See Straub (2010) for a detailed analysis of these aspects.


paper by Calderón et al. (2009) estimates the output elasticity of a synthetic infrastructure index to be between 0.07 and 0.10.13

This evolution in the magnitude of estimates is also partly explained by the fact that physical indicators of infrastructure have progressively replaced monetary measures of public infrastructure capital investment as the main proxies used in the empirical literature. The growing share of private investment in infrastructure and the large measurement and reliability problems of public capital indicators14 were probably instrumental in favoring the emergence of physical indicators databases, such as the one initially developed by Canning (1998), which tracks country-level kilometers of road and railroads, number of phone lines, and electricity generating capacity.

In this strand of literature, specific evidence on East Asia can be found in both Seethepalli et al. (2008) and Straub et al. (2008). Seethepalli et al. (2008) do find a positive effect of all dimensions of infrastructure stocks on growth, using standard growth regressions in a panel of 16 East Asian countries at 5-year intervals. They also conclude that these significant effects vary with a number of country-level characteristics. For example telecom and sanitation are found to have a greater effect in countries with better governance, higher income level, and low inequality in the access to infrastructure. In contrast, Straub et al. (2008) find much weaker results in their cross-country growth regressions, when using a sample of 93 developing or emerging countries, including 16 East Asian countries. The number of phone lines has a positive effect on growth, and some evidence emerges of an above-average effect for East Asia and high income countries. However, most results are not robust to using panel techniques or to controlling for an endogenous response of infrastructure to growth. As a matter of fact, while Seethepalli et al. (2008) argue that the use of infrastructure stocks rather than flows alleviates the problem of reverse causation, they fail to control for the potential endogeneity of infrastructure stocks due to countries’ unobserved characteristics, leading them both to have higher infrastructure stocks and higher growth; and do not include country fixed effects (see Holtz-Eakin 1994).

The existence and magnitude of indirect effects is a more complex issue at the empirical level, which as a result has been rarely addressed. Most of the existing contributions have used a growth accounting framework, for example Hulten et al. (2000 and 2005) and La Ferrara and Marcelino (2000). Assuming that the share of output of intermediate inputs is constant over time, Hulten and Schwab (2000) conclude to the absence of infrastructure externalities on growth in the US case, while Hulten et al. (2005) find highways and electricity to account for about half of total factor productivity (TFP) growth across Indian states in the period 1972–1992. Straub et al. (2008) growth accounting exercise on five East Asian countries yields few significant results: telecommunications investment has contributed to TFP growth more than other types of capital in Indonesia

13 Note that such estimates still imply high rates of return, between 25% and 50%.14 See for example Pritchett (1996).


and the Philippines, while roads have had a positive influence only in Thailand. No significant effect is found in the Republic of Korea and Singapore.

Indirect effects matter because they are instrumental in determining whether the effects of infrastructure investments are permanent or transitory. The belief that indirect effects may be sufficient to generate constant returns on aggregate and lead to endogenous growth underlies for example the claim in Calderón and Servén (2004) that raising Latin American quantities and qualities of infrastructure stocks to East Asian tigers’ level would have generated long-term per capita growth gains of around 3%. Straub (2010) claims that an alternative take on these numbers is that the different overall incentive structures prevailing in Latin America and East Asian countries imply that their economies display, in equilibrium, important gaps both in their infrastructure stocks and in output growth. However, it is not clear which part of these incentive differences has to do with infrastructure, and the lack of an empirical method to disentangle externalities from different sources imply that Calderón and Servén (2004) estimates might be biased upward. Indeed, several contributions have shown that infrastructure investments returns were highly sensitive to the quality of the regulatory or contractual framework, to the political economy environment, including aspects such as the efficiency of the local bureaucracy and potential corruption. Guasch, Laffont, and Straub (2007 and 2008) show that concession renegotiations in Latin America in the 1990s were to a large extent due to regulatory and institutional failures, and there are reasons to believe that this contributed to the subsequent backlash against private participation and discouraged private investment in infrastructure there in more recent years. The fact that East Asia focused mostly on build–operate–transfer (BOT) schemes for wholesale facilities, instead of concessioning of retail and distribution facilities as in Latin America, and could mobilize its higher savings, may have created fewer incentive and information problems and reduced to some extent direct political concerns.

2. Microeconomic Level

Microeconomic data first provide a wealth of descriptive evidence. For example, firm-level surveys on the investment climate provide a window on the extent to which infrastructure deficiencies constitute barriers to entrepreneurial development. For example, the World Bank Enterprise Surveys (World Bank 2009) indicate that a large proportion of respondents (between 20% in East Asia and the Pacific, and 55% in the Middle East and North Africa, as well as Latin America) view any of electricity, telecommunications, or transport as a major or severe obstacle to doing business. In Asia, electricity is considered as a major constraint by 78% of firms in Bangladesh, 76% in Nepal, 75% in Pakistan, 68% in Afghanistan, 61% in Timor-Leste, 44% in Samoa, 32% in India, and 30% in the PRC, among others. For transport, high proportions of firms viewing it as a major problem are found in Nepal (33%), Afghanistan (30%), Samoa (29%), and Thailand (21%), among others.


Moreover, a growing amount of microeconometric evidence is providing insights into a number of specific channels linking infrastructure investment and development outcomes, such as household income and poverty, welfare, health issues such as child mortality, and gender empowerment, for example, in the labor market.

Transportation infrastructure development has been shown to favor poverty reduction. Gibson and Rozelle (2003), using the timing of progressive coast to inland highway penetration in Papua New Guinea to instrument for road density, show that reducing access time to less than 3 hours where it was above this threshold leads to a decrease of 5.3% in the headcount index. Fan, Nyange, and Rao (2005) similarly find very positive effects of public investment and roads on household level income and poverty using Tanzanian household survey data. In an historical context, Donaldson (2009) shows that railroad development in colonial India significantly reduced trade costs, increased commercial exchanges between regions, with large overall welfare effects.

Thomas and Strauss (1992) analyze the determinants of child height in Brazil, a standard development outcome, and include electricity, water and sewerage connections as explanatory variables. They find the number of electricity connections per capita in the community to be positively correlated with height of babies, with a complementary effect of mother education. Positive results are also found for water and sanitation, with variations by children’s age and level of mother’s education. Dinkelman (2009) uncovers large effect on women employment of rural electrification in South Africa, which she relates to an increase of their labor supply as a result of more efficient home production technologies, and of a job expansion in new, small-scale market-based services.

In the case of water, in a widely quoted paper, Galiani et al. (2005) show that in Argentina, increased household access to the water network following privatization significantly reduced the incidence of water-borne diseases related child mortality. Duflo and Pande (2007) analyze the link between irrigation dams and agricultural production and poverty in India using geographic information system data on land inclination as instruments for dams placement choices, and claim that dams have low rates of return and negative distributional effects. Ilahi and Grimard (2000) find a significant impact of water infrastructure development on women’s time allocation in Pakistan, while Koolwal and van de Walle (2010), looking at micro data across nine countries, including three in South Asia (India, Nepal, and Pakistan), find that better access to water improves both boys’ and girls’ enrollments in countries with substantial gender gaps in schooling (among which Nepal and Pakistan), but no significant effect on women’s off-farm work.

Finally, a few papers, such as the study by Alby, Dethier, and Straub (2009) on the impact of electricity deficiencies have used enterprise survey data to assess the impact of infrastructure constraints on firms’ choices and performance.15 Additionally, microeconomic firm-level data are interesting because they can be used to analyze the

15 Dethier, Hirn, and Straub (2008) review this literature.


sequencing of reforms supporting infrastructure investment, in particular when the private sector is involved, and these reforms may include restructuring measures, regulation and the introduction of competition (see for example Andres et al. 2007). Of course, these studies generally deal with outcomes such as operating performance, quality, employment, prices, output and coverage, while policy makers care about final outcomes, such as output growth, welfare, and poverty reduction. This signals the need for more large-scale welfare assessments based on micro-data, as performed for example by McKenzie and Mookherjee (2003) in the case of four Latin American countries.

3. Geographic Evidence

It is difficult to underestimate the importance of the spatial nature of infrastructure investment. First, investment decisions imply rival choices on the geographical areas to be served. Second, spatial variations in the availability and quality of infrastructure are likely to have an impact on individuals’ and firms’ decisions, such as migration, location of new firms, etc.

Some contributions indirectly account for the spatial implications of public capital investment. Jacoby (2000) analyzes the distributional consequences of rural roads in Nepal, a country in which a large proportion of farmers live far from the main road giving access to market centers, using household survey and plot value data. It concludes that a 10% increase in travel time to market reduces the value of land by 2.2%, but that these benefits are “more like a tide that lifts all boats rather than a highly effective means of reducing income inequality”.

Recently, empirical contributions directly inspired by the new economic geography literature have explicitly included spatial variables. They are based on the use of an accessibility indicator, which measures, for households and firms in a given geographic area, the opportunities available at other locations in terms of employment opportunities or market potential by inversely weighting the sum of some destinations’ indicators (GDP or employment for example) by a proxy of the costs involved in reaching them.

Examples for developing countries include Deichman, Fay, Koo and Lall (2002) for Southern Mexico; Lall, Funderburg, and Yepes (2004) for Brazil; Lall, Shalizi, and Deichmann (2004) for India; Deichmann, Kaiser, Lall, and Shalizi (2005) for Indonesia; and Lall, Sandefur, and Wang (2009) for Ghana. All these studies find that accessibility is a major determinant of firm productivity.

Looking at Asian countries in more detail, Lall, Shalizi, and Deichmann (2004) examine the effects of agglomeration economies on plant level productivity, distinguishing between economies of scale at the plant level (related to increases in market access), at the industry level (from localization economies), and at the regional level (from urbanization economies across industries). They find that market access and proximity to transport hubs provide net benefits in four out of 11 industrial sectors. As for localization


economies, net benefits are found in only two industrial sectors. Finally, firms’ benefits of locating in dense urban areas appear be smaller than associated costs, a finding that the authors attribute to the unequal spatial distribution of transport infrastructure, which makes it difficult for firms to locate in cheaper secondary urban centers while maintaining linkages with final buyers and suppliers.

Deichmann, Kaiser, Lall, and Shalizi (2005) estimate the impact of factors influencing location choice at the firm level to explain the aggregate and sectoral geographic concentration of manufacturing industries in Indonesia. They distinguish between natural advantage, such as infrastructure endowments, wage rates, and natural resource endowments, and production externalities linked to the co-location of firms in the same or complementary industries. They conclude that because of the strong relevance of agglomeration economies in explaining location choices, investment in infrastructure in lagging regions would have failed to attract firms away from established “leading” regions, especially in mainstream sectors characterized by a high concentration in these leading regions. One of the policy lessons is that the best strategy may not be to divert industries already concentrated in large agglomerations away to lagging regions, but rather to exploit existing location endowments to stimulate the development of alternative industries.

4. Practical Issues

Public financing often raises the concern of the potential crowding out of private investment. Of course, the extent to which this happens is an empirical issue, and should in particular be sensitive to the business cycle. One potential channel for this is the limited capacity of domestic financial markets, which might be overburdened by public borrowing, especially in times of crisis when spending on infrastructure is likely to build up, and the consequent deterioration of banks’ portfolios. The optimistic view is that in an economic downturn, public financing for large infrastructure projects of the type sustained by most large stimulus plans would in fact sustain the credibility of the projects and may crowd-in private investors (Lardy 2010). Unfortunately, systematic empirical evidence has yet to be produced.

As for sequencing of investments and reforms, little guidance is available for policy makers. While virtually no evidence exists to assess the comparative growth dividends of investments in different infrastructure sectors, insights about the sequencing of reforms supporting infrastructure investment are found in relation with discussions of private sector involvement. In this case, the accumulated wisdom points to the need of restructuring and the implementation of a regulatory framework prior to the transfer to the private partners. Moreover, institutional arrangements aiming at ensuring transparency and the independence of the agencies in charge seem warranted. Finally, competition should be introduced whenever the characteristics of the activity make this possible (World Bank 2004).16

16 See also Straub (2008) for more discussion.


IV. Empirical Analysis

In this section, we extend the analysis in Straub et al. (2008) along several lines. The first part performs a growth regression analysis to examine specifically interactions with several subgroups of Asian countries. The second part performs a growth accounting analysis, including longer TFP time series and more countries (14). The description of the theoretical framework used in both sections is in the Appendix.

A. Cross-country Regressions

1. Data

We use physical infrastructure indicators used extensively in recent literature. This allows direct comparisons with the results from the growth accounting exercise below.

Physical indicators for four different sectors (telecommunications, energy, transport, and water) are taken from the World Development Indicators (WDI) database (unless specified otherwise) covering the 1971–2006 period. Specifically, we use the following series:

(i) Telecommunications

– main telephone lines – number of mobile phones – internet coverage: users per 100 inhabitants (from the International Telecommunication Union website, www.itu.int/ITU-D/icteye/ Indicators/Indicators.aspx#, downloaded 3 August 2010).

(ii) Energy

– a “quality proxy” is computed using electricity generating capacity (in million kilowatts) and electric power transmission and distribution losses (as a percentage of output).

(iii) Transport

– road total network (in kilometers) – paved roads (as a percentage of total network), as a quality proxy – rail route length (in kilometers)

(iv) Water

– improved water source (percent of population with access)


Additional variables used include GDP per capita, gross fixed capital formation (investment/GDP), and primary and secondary school enrolment (all from WDI).

2. Sample

We rely on a sample of 102 developing or emerging countries, of which 17 (the PRC; the Fiji Islands; Hong Kong, China; Indonesia;the Republic of Korea; Democratic People's Republic of Korea; the Lao PDR; Malaysia; Mongolia; Myanmar; Papua New Guinea; the Philippines; Singapore; Thailand; Tonga; Vanuatu; Viet Nam) belong to the East Asia and Pacific region. Five (Bhutan, India, Nepal, Pakistan, Sri Lanka) belong to the South Asia area.

3. Results

In what follows we present the results from cross country estimations based on the collapsed dataset17 for 1975–2006 for fixed plus mobile telephony (1980–2006 for mobile telephony alone and 1999–2006 for internet usage); 1980–2006 for rail; 1990–2006 for roads; and 1990–2006 for water, using the rate of growth of GDP per capita as dependent variable and standard controls (initial level of GDP, investment, proxies for human capital). Note that we introduce infrastructure proxies in two ways: first we use the average levels over the relevant period; second, we use the average growth rates of these variables over the same period. In each case, after testing simple OLS specifications, we instrument potentially endogenous infrastructure indicators and perform related tests. In all cases, the instruments are beginning of the period indicators for the relevant infrastructure variable, the share of agriculture over GDP, population density, and total population.

17 Using averages over the period.


We also test specifications with interactions between both an East Asia and Pacific (EAP) and a South Asia (SA) dummy and the alternative infrastructure indicators mentioned above. Results for each infrastructure dimension are included in a separate table: Table 1 for electricity, Table 2 for telecommunications (both fixed + mobile phone lines and mobile phones alone), Table 3 for internet, Table 5 for railroads, Table 6 for roads, and Table 7 for water.18

Note first that the control variables yield results consistent with the empirical growth literature present: initial per capita GDP is negative and significant, which denotes convergence conditional on the other variables, while education and investment variables are positive and generally significant throughout Tables 1 to 7. Table 1 presents the specific results for electricity.

Table 1: Cross-section, Electricity

(1) (2) (3) (4) (5) (6)

pcgdpgrowth pcgdpgrowth pcgdpgrowth pcgdpgrowth pcgdpgrowth pcgdpgrowthpcgdp71 -1.57e-06*** -1.24e-06*** -1.16e-06*** -1.07e-06*** 2.48e-06 -9.96e-07*

(4.91e-07) (1.85e-07) (4.21e-07) (1.90e-07) (4.87e-06) (5.59e-07)School enrollment secondary

0.000577*** 0.000485*** 0.000482*** 0.000420*** 0.00119 0.000215

(0.000149) (0.000138) (0.000165) (0.000121) (0.000799) (0.000296)School enrollment primary

-0.000293 -0.000154 -0.000168 -0.000115 -0.000293 -7.43e-05

(0.000186) (0.000188) (0.000200) (0.000151) (0.000580) (0.000371)Inv/gdp 0.149*** 0.110** 0.120** 0.0687* 0.149 0.00608

(0.0523) (0.0482) (0.0520) (0.0390) (0.122) (0.107)pcegc 0.00121 -5.49e-05 -0.0170

(0.00187) (0.00164) (0.0209)pcegc_gr 0.222*** 0.181*** 0.245

(0.0646) (0.0444) (0.203)pcegc_eap 0.00765*** -0.0220

(0.00258) (0.0361)pcegc_sa 0.0476*** 0.281

(0.0132) (0.251)pcegc_greap 0.0198*** 0.0475

(0.00664) (0.0563)pcegc_grsa 0.0155*** -0.00262

(0.00403) (0.0562)Constant -0.0149 -0.248*** -0.0177 -0.201*** -0.0353 -0.252

(0.0162) (0.0713) (0.0170) (0.0528) (0.0430) (0.234)Observations 48 46 48 46 38 37R-squared 0.509 0.635 0.582 0.763Wu-Hausman F test, p-value

0.001 0.377

*** p<0.01, ** p<0.05, * p<0.1.pcegc = per capita electricity generation capacity, net of transmission and distribution losses; pcegc_gr = rate of growth; pcegc_eap = interaction with East Asian and Pacific dummy; pcegc_sa = interaction with South Asia dummy.Note: Robust standard errors in parentheses.

18 Note that results from panel regressions on 5-year subperiod averages, using alternatively fixed and random effects, as well as instrumental variable estimations, yield not significant results overall. These results are omitted for the sake of space.


Source: Authors' calculations.

The per capita electricity generating capacity appears to have no significant effect on growth in column 1, while in column 2 its growth rate has a positive and significant effect, such that an additional point in the average growth rate of electricity generating capacity net of losses results in 0.22 additional average per capita growth over the period in our sample of developing countries. The introduction of the interactions with the EAP and the SA dummies in columns 3 and 4 show that positive and significant effects arise for both groups of countries. As for electricity generating capacity, the effects are now 0.008 for EAP and 0.05 for SA. The effects when variables are expressed in growth rates are 0.2 for EAP and 0.197 for SA.19 These numbers mean that an additional point in the average growth rate of electricity generating capacity net of losses results in 0.18 additional average per capita growth over the period in our sample of developing countries, but that this effect is about 11% stronger among Asian countries, indicating that investments in electricity across the region have been effective in relieving infrastructure bottlenecks and complementing productive investment. In columns 5–6, a Wu-Hausman endogeneity test rejects exogeneity for the electricity variable in level, which probably corresponds to the fact that countries have characteristics unobserved to the econometrician such that they have both faster growth and higher electricity generation. When instrumented, electricity in level loses significance. On the other hand, exogeneity is accepted for the growth rate, suggesting that these results are more robust than those in levels.

Table 2 presents the results for the telecommunications variables, which are respectively the sum of fixed and mobile phone lines in columns 1–6, and the number of mobile phones lines alone in columns 7 to 9.20

19 Marginal effects are computed as follows: for example the effect in growth rates for EAP = 0.181 + 0.0198 ≈ 0.2.20 The series of mobiles phone growth rates are characterized by very high rates in the first years as they start from

very low levels, so we do not use it. The growth effect resulting from the introduction of mobile phones in the 1980s is however reflected in the growth of the total number of fixed and mobile phones.


Tabl

e 2:

Cro

ss-s

ecti

on, T

elec

omm

unic

atio

ns

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

p75

-1.5

6e-0

6***

-1.1

1e-0

6***

-1.2

7e-0

6***

-9.3

7e-0

7**

-2.2

7e-0

6-3

.99e

-07

(2.7

5e-0

7)(3

.81e

-07)

(2.1

9e-0

7)(3

.90e

-07)

(8.0

1e-0

6)(5

.28e

-06)

Scho

ol e

nrol

lmen

t s

econ

dary

0.00

0304

**0.

0006

05**

*0.

0003

13**

0.00

0533

***

0.00

135

-3.6

6e-0

50.

0002

05*

0.00

0200

**-0

.002

40

(0.0

0013

3)(0

.000

110)

(0.0

0012

2)(9

.87e

-05)

(0.0

0287

)(0

.002

04)

(0.0

0010

8)(9

.93e

-05)

(0.0

0641

)Sc

hool

enr

ollm

ent

prim

ary

-0.0

0018

6-0

.000

243*

-0.0

0015

8-0

.000

210*

-0.0

0088

0-0

.000

264

-0.0

0015

9-0

.000

128

0.00

0403

(0.0

0013

4)(0

.000

136)

(0.0

0013

0)(0

.000

111)

(0.0

0220

)(0

.001

27)

(0.0

0012

9)(0

.000

126)

(0.0

0152

)In

v/gd

p0.

110*

**0.

109*

**0.

0986

***

0.09

07**

*0.

285

-0.1

010.

135*

**0.

120*

**0.

119

(0.0

293)

(0.0

310)

(0.0

285)

(0.0

297)

(0.5

39)

(0.5

83)

(0.0

231)

(0.0

228)

(0.2

09)

pcte

l0.

0367

**0.

0173

0.06

94(0

.017

2)(0

.012

7)(0

.702

)pc

tel_

eap

0.04

18**

*-0

.869

(0.0

145)

(2.4

11)

pcte

l_sa

0.53

3***

8.92

8(0

.197

)(1

9.95

)pc

tel_

gr0.

231*

**0.

197*

**0.

0004

25(0

.077

5)(0

.052

0)(1

.172

)pc

tel_

grea

p0.

0157

***

0.19

7(0

.005

68)

(0.5

46)

pcte

l_gr

sa0.

0120

**-0

.157

(0.0

0500

)(0

.513

)pc

gdp8

0-1

.46e

-06*

**-1

.25e

-06*

**-1

.47e

-05

(3.3

3e-0

7)(2

.98e

-07)

(3.6

0e-0

5)pc

mob

0.08

24**

0.05

50**

2.46

7(0

.031

1)(0

.023

7)(6

.058

)pc

mob

_eap

0.06

97**

-1.4

75(0

.027

5)(3

.805

)pc

mob

_sa

0.90

6**

23.2

5(0

.387

)(4

5.92

)Co

nsta

nt-0

.009

26-0

.276

***

-0.0

0994

-0.2

36**

*-0

.033

10.

0423

-0.0

136

-0.0

133

-0.0

503

(0.0

114)

(0.0

953)

(0.0

113)

(0.0

630)

(0.1

02)

(1.4

59)

(0.0

121)

(0.0

119)

(0.1

13)

Obs

erva

tions

6657

6657

4846

7272

60R-

squa

red

0.46

90.

545

0.53

10.

640

0.50

00.

552

Wu-

Hau

sman

F

tes

t, p-

valu

e0.

001

0.08

0.00

0

***

p<0.

01, *

* p<

0.05

, * p

<0.1

. pc

tel =

per

cap

ita n

umbe

r of fi

xed

plus

mob

ile p

hone

s lin

es; p

ctel

_gr =

rate

of g

row

th; p

ctel

_eap

= in

tera

ctio

n w

ith E

ast A

sian

and

Pa

cific

dum

my;

pct

el_s

a =

inte

ract

ion

with

Sou

th A

sia

dum

my.

pcm

ob =

per

cap

ita n

umbe

r of m

obile

pho

nes

lines

; pcm

ob_e

ap =

in

tera

ctio

n w

ith E

ast A

sian

and

Pac

ific

dum

my;

pcm

ob_s

a =

inte

ract

ion

with

Sou

th A

sia

dum

my.

Not

e:

Robu

st s

tand

ard

erro

rs in

par

enth

eses

.So

urce

: A

utho

rs' c

alcu

latio

ns.


This total telecom variable is positive and significant both when introduced in level in column 1 and in growth rate in column 2 (with a marginal effect indicating that 1 additional point in the average growth rate of the number of per capita phone lines generates 0.23 additional average per capita growth over the period). The interactions with the regional dummies show much stronger and significant effects for Asian countries, both in levels and in growth rates. Focusing on these last ones, the overall effect is 0.213 for EAP countries and 0.209 for SA countries, compared with only 0.197 for the overall sample. These results are supported by the outcome of the estimations focusing on mobile phones in columns 7 and 8. The positive and significant effect on growth remains, and again is much stronger in the EAP and SA subsamples (0.125 and 0.961 respectively, versus 0.055 in the overall sample). Together, these results support the idea that telecom development was instrumental in boosting growth in the Asian region, and especially the rapid spread of mobile telephony, probably facilitated by well-designed regulatory frameworks.21

The introduction of internet services in the 1990s is an additional aspect that is often thought to have been instrumental for private sector development. To test this hypothesis, we perform a similar analysis using as independent variable the average growth rate of the number of internet users, as provided by the International Telecommunication Union (ITU).22 Table 3 shows that while the effect across the whole sample is negative, the net effect for the two groups of Asian countries under study is positive and significant. Specifically, an additional point in the average growth rate of internet coverage implied an additional 0.01 average per capita growth over the period for EAP countries, and 0.003 for SA countries, respectively. Given that the average annual growth of the number of internet users over this period was 18% worldwide, and close to 35% for the subsample of available Asian and Pacific countries (see Table 4), these elasticities are far from negligible. Exogeneity of the internet usage growth rate is not rejected.

In Table 5, we introduce the railroads variable. While no results are significant in the overall sample, the interactions with the regional dummies show again much stronger and significant effects for Asian countries, both in levels and in growth rates. For example in column 4, an additional point in the average growth rate of railroads line results in 0.25 additional average per capita growth over the period in our sample of developing countries, but this effect is about 10%–12% stronger among Asian countries (0.283 and 0.276 for EAP and SA respectively, versus 0.248 in the overall sample). Note that exogeneity is not rejected by the Wu-Hausman test instrumented for the railroads variables in levels and in growth rates.

21 Exogeneity is rejected by the Wu-Hausman test for both the variables in level and in growth rate, although only marginally in this last case. However, when instrumented, they are no longer significant, which may indicate that our instruments are weak.

22 Other indicators, such as the number of broadband subscriptions, were not available for a long enough period to perform this analysis.


Table 3: Cross-section, Internet

(1) (2) (3)

pcgdpgrowth pcgdpgrowth Pcgdpgrowthpcgdp99 -5.56e-08 -9.63e-08 -1.03e-07

(5.17e-07) (4.04e-07) (1.12e-06)School enrollment secondary 9.34e-05 8.44e-05 -1.07e-05

(0.000116) (0.000112) (0.000193)School enrollment primary -0.000187 -0.000218 -0.000143

(0.000150) (0.000151) (0.000277)Inv/gdp 0.150*** 0.117*** 0.0820

(0.0368) (0.0334) (0.0689)pcinternet_gr -0.00271 -0.00420 -0.0458

(0.0117) (0.0116) (0.0568)pcinternet_greap 0.0154*** 0.0119

(0.00317) (0.0204)pcinternet_grsa 0.00706* 0.0625

(0.00421) (0.0488)Constant 0.00570 0.0161 0.0784

(0.0202) (0.0199) (0.0882)Observations 77 77 75R-squared 0.272 0.372Wu-Hausman F test, p-value 0.12

*** p<0.01, ** p<0.05, * p<0.1.pcinternet_gr = rate of growth in percent of number of users per 100 inhabitants between 1999 and 2006; pcinternet_greap =

interaction with East Asian and Pacific dummy; pcinternet_grsa = interaction with South Asia dummy.Note: Robust standard errors in parentheses.Source: Authors' calculations.

Table 4: Average Growth of Internet Usage, 1999–2006

Economy Growth of Internet UsersChina, People’s Rep. of 40.1Fiji Islands 33.7Hong Kong, China 15.0India 34.0Indonesia 34.7Iran 58.8Korea, Rep. of 15.4Lao PDR 52.5Malaysia 19.6Nepal 28.9Pakistan 81.5Papua New Guinea 12.8Philippines 19.0Singapore 12.0Thailand 27.7Tonga 24.4Vanuatu 34.7Viet Nam 84.2World 18.0

Source: International Telecommunication Union (2010).


Table 5: Cross-section, Railroads

(1) (2) (3) (4) (5) (6)

pcgdpgrowth pcgdpgrowth pcgdpgrowth pcgdpgrowth pcgdpgrowth pcgdpgrowth

pcgdp80 -1.65e-06** -2.54e-06*** -1.08e-06 -1.91e-06** 1.53e-06 -1.03e-06

(6.22e-07) (7.07e-07) (6.67e-07) (8.43e-07) (2.06e-05) (1.72e-06)

School enrollment secondary

0.000393* 0.000246 0.000282 0.000332 0.00126 0.000337

(0.000230) (0.000279) (0.000179) (0.000222) (0.00610) (0.000357)

School enrollment primary

-0.000128 -0.000260 -2.32e-05 -0.000113 -0.000194 0.000204

(0.000239) (0.000318) (0.000203) (0.000327) (0.00263) (0.000574)

Inv/gdp 0.155** 0.161** 0.0954* 0.0694 0.778 0.0312

(0.0613) (0.0738) (0.0509) (0.0614) (4.105) (0.100)

pcrail 3.968 9.105 -36.46

(7.129) (6.600) (320.1)

pcrail_eap 443.8*** -5,175

(146.4) (33,066)

pcrail_sa 306.8*** 3,457

(87.71) (17,518)

pcrail_gr 0.397 0.248 -0.199

(0.272) (0.259) (0.674)

pcrail_greap 0.0346*** 0.0489

(0.0114) (0.0379)

pcrail_grsa 0.0281*** 0.0390

(0.00763) (0.0435)

Constant -0.0257 -0.391 -0.0243 -0.253 -0.183 0.159

(0.0217) (0.258) (0.0204) (0.237) (0.933) (0.623)

Observations 39 26 39 26 32 24

R-squared 0.456 0.508 0.650 0.726

Wu-Hausman F test, p-value

0.13 0.58

*** p<0.01, ** p<0.05, * p<0.1.pcrail = per capita kilometer of railroad lines; pcrail_gr = rate of growth; pcrail_eap = interaction with East Asian and Pacific dummy; pcrail_sa = interaction with South Asia dummy.Note: Robust standard errors in parentheses.Source: Authors' calculations.

Results using the kilometers of roads per capita are in Table 6. Most of the results fail to be significant, with two exceptions. In column 4, the interaction of the quality proxy (percentage of the road network that is paved) with the EAP and SA dummy indicates that the quality of the road network has a positive effect in the Asia region: the implied marginal effects are that an increase of 10% in the proportion of paved roads


corresponds to an additional average per capita growth over the period of 0.3% in EAP and 0.4% in SA. In column 5, an additional point in the average growth rate of roads results in significantly higher average per capita growth over the period for EAP and SA than in the overall sample of developing countries (0.071, versus 0.056 in the overall sample).23

Finally, Table 7 reports the results using percentage of the population with access to an improved water source. While overall results are not significant, interactions with the EAP and the SA dummies are again positive and significant, supporting the idea of a positive link between the number of connections and growth in the Asian region. In levels, the results indicate that an additional 10% in the rate of population coverage translates into 0.06% additional per capita growth in the EAP region and 0.04% in SA. In growth rates, an additional point in the average growth rate of water coverage results in significantly higher average per capita growth over the period for SA than in the overall sample of developing countries (0.172, versus 0.158 in the overall sample). This is in line with some previous studies that found that countries with a well-functioning water sector also experimented stronger growth. Possible channels include an indirect impact through external effects such as better health and better productivity of workers.

23 Note that exogeneity is rejected by the Wu-Hausman test for roads, both when considering level and growth rate, but instrumental estimations fail to be significant for growth rates.


Tabl

e 6:

Cro

ss-s

ecti

on, R

oads

(1)

(2)

(3)

(4)

(5)

(6)

(7)

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

p90

-6.4

4e-0

7*-7

.03e

-07*

-5.5

4e-0

7-7

.61e

-07*

*-5

.82e

-07

9.91

e-07

-1.6

4e-0

6(3

.66e

-07)

(3.9

8e-0

7)(6

.86e

-07)

(3.7

8e-0

7)(5

.83e

-07)

(3.9

7e-0

6)(2

.08e

-06)

Scho

ol e

nrol

lmen

t s

econ

dary

0.00

0213

*0.

0001

98*

0.00

0188

0.00

0233

**0.

0001

75-4

.55e

-05

0.00

0312

(0.0

0010

8)(0

.000

113)

(0.0

0012

2)(0

.000

114)

(0.0

0012

6)(0

.000

897)

(0.0

0023

1)Sc

hool

enr

ollm

ent

prim

ary

2.69

e-05

3.24

e-05

-5.9

9e-0

51.

41e-

05-3

.71e

-05

0.00

0690

-8.7

1e-0

6

(0.0

0013

0)(0

.000

132)

(0.0

0014

3)(0

.000

135)

(0.0

0014

5)(0

.001

59)

(0.0

0018

8)In

v/gd

p0.

144*

**0.

140*

**0.

154*

**0.

116*

**0.

113*

**-0

.090

40.

0806

(0.0

258)

(0.0

272)

(0.0

368)

(0.0

306)

(0.0

301)

(0.3

69)

(0.0

934)

pcro

ad-0

.570

-0.5

18-0

.357

-4.5

87(0

.568

)(0

.561

)(0

.552

)(7

.252

)pc

road

_eap

-0.8

47-3

4.80

(1.7

42)

(110

.3)

pcro

ad_s

a-1

.797

61.0

8(2

.598

)(1

07.4

)ro

adqu

al3.

29e-

05-3

.98e

-05

3.93

e-05

(6.6

6e-0

5)(6

.60e

-05)

(7.0

9e-0

5)ro

adqu

al_e

ap0.

0002

41**

(0.0

0010

6)ro

adqu

al_s

a0.

0003

45*

(0.0

0018

4)pc

road

_gr

0.09

560.

0562

-0.2

17(0

.077

5)(0

.063

1)(0

.421

)pc

road

_gre

ap0.

0149

**0.

0296

(0.0

0649

)(0

.037

7)pc

road

_grs

a0.

0145

**0.

0645

(0.0

0605

)(0

.040

7)Co

nsta

nt-0

.023

3**

-0.0

236*

*-0

.113

-0.0

181

-0.0

708

-0.0

0981

0.19

7(0

.011

6)(0

.011

6)(0

.077

4)(0

.012

8)(0

.063

1)(0

.059

6)(0

.412

)O

bser

vatio

ns79

7959

7959

6756

R-sq

uare

d0.

468

0.46

90.

448

0.52

30.

532

Wu-

Hau

sman

F

tes

t, p-

valu

e0.

001

0.03

3

***

p<0.

01, *

* p<

0.05

, * p

<0.1

.pc

road

= p

er c

apita

kilo

met

er o

f roa

d lin

es; p

croa

d_gr

= ra

te o

f gro

wth

; pcr

oad_

eap

= in

tera

ctio

n w

ith E

ast A

sian

and

Pac

ific

dum

my;

pcr

oad_

sa =

inte

ract

ion

with

Sou

th A

sia

dum

my;

road

qual

= p

aved

road

s in

per

cent

of t

otal

net

wor

k; ro

adqu

al_e

ap =

in

tera

ctio

n w

ith E

ast A

sian

and

Pac

ific

dum

my;

road

qual

_sa

= in

tera

ctio

n w

ith S

outh

Asi

a du

mm

y.N

ote:

Ro

bust

sta

ndar

d er

rors

in p

aren

thes

es.

Sour

ce:

Aut

hors

' cal

cula

tions

.


Table 7: Cross-section, Water

(1) (2) (3) (4) (5) (6)

pcgdpgrowth pcgdpgrowth pcgdpgrowth pcgdpgrowth pcgdpgrowth pcgdpgrowthpcgdp90 -7.57e-07* -5.99e-07* -5.30e-07 -3.79e-07 3.75e-07 4.85e-07

(4.12e-07) (3.54e-07) (3.82e-07) (4.05e-07) (1.21e-06) (2.12e-06)School enrollment secondary

0.000213* 0.000217* 0.000233** 0.000211 0.000406 0.000271

(0.000117) (0.000122) (0.000111) (0.000127) (0.000338) (0.000337)School enrollment primary

3.10e-05 -2.32e-05 2.95e-05 -4.23e-06 -8.94e-05 -1.47e-05

(0.000132) (0.000159) (0.000128) (0.000159) (0.000263) (0.000327)Inv/gdp 0.149*** 0.213*** 0.120*** 0.183*** 0.0663 -0.0595

(0.0276) (0.0363) (0.0283) (0.0416) (0.224) (0.518)pcwater -5.74e-05 -0.000122 -0.000390

(0.000126) (0.000132) (0.000420)pcwater_eap 0.000179** 0.000716

(7.07e-05) (0.00117)pcwater_sa 0.000164*** 0.000366

(5.92e-05) (0.000675)pcwater_gr 0.111 0.158 0.319

(0.141) (0.143) (0.531)pcwater_greap 0.00897 0.0975

(0.00674) (0.193)pcwater_grsa 0.0142*** 0.00458

(0.00363) (0.105)Constant -0.0226* -0.146 -0.0150 -0.192 0.0120 -0.319

(0.0129) (0.149) (0.0140) (0.150) (0.0662) (0.497)Observations 77 60 77 60 61 57R-squared 0.461 0.550 0.518 0.580Wu-Hausman F test, p-value

0.094 0.202

*** p<0.01, ** p<0.05, * p<0.1.pcwater = percentage of population with access to an improved water source; pcwater_gr = rate of growth; pcwater_eap = interaction with East Asian and Pacific dummy; pcwater_sa = interaction with South Asia dummy.Note: Robust standard errors in parentheses.Source: Authors' calculations.

One possibility that is often mentioned is that infrastructure would require a suitable institutional environment to generate sizable growth dividends (e.g., see De 2010). To test this hypothesis in the context of Asian countries, we augment our standard specifications by introducing several indices capturing institutional quality. The first one is an index of regulatory quality taken from Kaufmann, Kraay, and Mastruzzi (2009). We focus on this index because the overall quality of the regulatory environment should matter for infrastructure development. Alternatively, we use, from the same source, an index computed as the average of government effectiveness, rule of law, and control of corruption. This is meant to capture the overall institutional quality prevailing in a given country. These indicators are measured in units ranging from about –2.5 to 2.5, with higher values corresponding to better governance outcomes. We use 2006 values.24

24 Unfortunately, these indices are only available from 1996, preventing us from using similar period-average values as for other variables.


We also experiment with an index of infrastructure performance taken from the World Bank. The specific infrastructure subcomponent of this index measures the “quality of trade and transport related infrastructure (e.g., ports, railroads, roads, information technology)”. Consistently with its definition, this index is used only for estimation including transportation indicators (railroads and roads). The scores are from 1 to 5, with 1 being the worst performance. Table 8 displays Asian countries scores, as well as regional and income groups averages.

Table 8: Infrastructure Performance Index

Economy Infrastructure Performance Index

Singapore 4.22Hong Kong, China 4.00Korea, Rep. of 3.62China, People’s Rep. of 3.54Malaysia 3.50Thailand 3.16India 2.91Philippines 2.57Viet Nam 2.56Indonesia 2.54Pakistan 2.08Fiji Islands 1.98Lao PDR 1.95Mongolia 1.94Myanmar 1.92Papua New Guinea 1.91Sri Lanka 1.88Bhutan 1.83Nepal 1.80East Asia and Pacific average 2.46South Asia average 2.13High income: all 3.56Upper middle income 2.54Lower middle income 2.27Low income 2.06

Source: World Bank (2010), availble: info.worldbank.org/etools/tradesurvey/mode1b.asp, downloaded 3 August 2010.

Table 9 presents results where the standard specifications have been extended to include our institutional indices of interest, as well as additional interactions with our Asian country groups, i.e., we introduce, on top of the standard interactions of previous tables, triple interactions between infrastructure indices, regional dummies, and institutional indices. A significant coefficient would therefore indicate that the additional effect of infrastructure in the Asian subgroups, as implied by the significant interactions found in most previous regressions, can be partly attributed to institutional quality in the sense that within EAP or SA country groups, countries with better environment would experiment stronger links between infrastructure investment and growth.


Tabl

e 9:

Inst

itut

iona

l Ind

icat

ors (1

)(2

)(3

)(4

)(5

)(6

)(7

)

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

pcgd

pgro

wth

Infr

astr

uctu

re in

dica

tor

Elec

tric

ityPh

ones

Railr

oads

Railr

oads

Road

sRo

ads

Wat

erIn

stitu

tiona

l ind

exRe

gula

tory

qu

ality

Regu

lato

ry

qual

ityRe

gula

tory

qu

ality

Infr

astr

uctu

re

perf

orm

ance

Regu

lato

ry

qual

ityIn

fras

truc

ture

pe

rfor

man

ceRe

gula

tory

qu

ality

pcgd

p71

-1.0

6e-0

6***

(2.1

7e-0

7)pc

gdp7

5-1

.10e

-06*

**(3

.37e

-07)

pcgd

p80

-2.3

9e-0

6**

-1.8

8e-0

6(8

.05e

-07)

(1.2

0e-0

6)pc

gdp9

0-1

.41e

-06*

*-1

.07e

-06

-7.9

5e-0

7**

(6.0

7e-0

7)(7

.80e

-07)

(3.2

8e-0

7)Sc

hool

enr

ollm

ent s

econ

dary

0.00

0230

0.00

0372

***

0.00

0242

0.00

0114

0.00

0122

6.96

e-05

8.49

e-05

(0.0

0015

6)(0

.000

112)

(0.0

0022

6)(0

.000

326)

(0.0

0014

4)(0

.000

163)

(0.0

0014

8)Sc

hool

enr

ollm

ent p

rimar

y-5

.01e

-05

-0.0

0014

9-0

.000

348

5.88

e-05

-0.0

0013

1-5

.23e

-05

2.12

e-05

(0.0

0020

4)(0

.000

134)

(0.0

0029

2)(0

.000

345)

(0.0

0016

4)(0

.000

189)

(0.0

0016

3)In

v/gd

p0.

0830

**0.

0924

***

0.03

750.

0470

0.11

6***

0.14

4***

0.15

9***

(0.0

374)

(0.0

239)

(0.0

481)

(0.0

706)

(0.0

306)

(0.0

388)

(0.0

401)

Regu

lato

ry Q

ualit

y0.

0379

-0.0

902

0.55

10.

259*

*-0

.166

(0.0

600)

(0.0

770)

(0.3

24)

(0.1

22)

(0.3

29)

Infr

astr

uctu

re p

erfo

rman

ce0.

831

0.11

3(0

.706

)(0

.101

)pc

infr

a_gr

0.16

3***

0.19

5***

0.28

72.

696

0.11

20.

407

0.25

3(0

.045

8)(0

.048

6)(0

.227

)(2

.156

)(0

.071

9)(0

.331

)(0

.298

)pc

infr

a_gr

*eap

0.01

71**

0.01

20*

0.04

56**

*-0

.117

0.01

32**

-0.0

231

0.00

916

(0.0

0809

)(0

.006

25)

(0.0

0898

)(0

.145

)(0

.005

80)

(0.0

197)

(0.0

0651

)pc

infr

a_gr

*sa

0.01

04**

0.00

518

0.03

32**

*0.

0392

0.02

50**

*-0

.011

80.

0150

**(0

.004

46)

(0.0

0713

)(0

.006

15)

(0.0

382)

(0.0

0544

)(0

.022

5)(0

.006

48)

pcin

fra_

gr*i

nst_

inde

x-0

.028

80.

0890

-0.5

51-0

.846

-0.2

51**

-0.1

060.

176

(0.0

597)

(0.0

707)

(0.3

31)

(0.7

18)

(0.1

24)

(0.1

000)

(0.3

27)

pcin

fra_

gr*e

ap*i

nst_

inde

x-0

.003

500.

0020

6-0

.043

0**

0.04

500.

0019

50.

0114

-0.0

0559

(0.0

0706

)(0

.005

42)

(0.0

179)

(0.0

441)

(0.0

0738

)(0

.008

05)

(0.0

0783

)pc

infr

a_gr

*sa*

inst

_ind

ex-0

.015

0-0

.026

80.

0346

-0.0

0481

0.04

90**

*0.

0085

10.

0058

9(0

.018

1)(0

.018

8)(0

.023

0)(0

.014

9)(0

.014

6)(0

.009

34)

(0.0

203)

Cons

tant

-0.1

79**

*-0

.229

***

-0.2

51-2

.656

-0.1

09-0

.434

-0.2

74(0

.055

1)(0

.062

2)(0

.209

)(2

.104

)(0

.069

2)(0

.325

)(0

.304

)O

bser

vatio

ns46

5726

2559

5260

R-sq

uare

d0.

813

0.71

60.

862

0.76

10.

604

0.62

20.

637

***

p<0.

01, *

* p<

0.05

, * p

<0.1

.pc

infr

a _g

r = ra

te o

f gro

wth

of p

er c

apita

infr

astr

uctu

re in

dica

tor;

pcin

fra_

gr*e

ap =

inte

ract

ion

with

Eas

t Asi

an a

nd P

acifi

c du

mm

y;

pcin

fra_

gr*s

a =

inte

ract

ion

with

Sou

th A

sia

dum

my;

pci

nfra

_gr*

inst

_ind

ex =

inte

ract

ion

of p

cinf

ra _

gr w

ith in

stitu

tiona

l ind

icat

or;

pcin

fra_

gr*e

ap*i

nst_

inde

x an

d pc

infr

a_gr

*sa*

inst

_ind

ex =

inte

ract

ions

of p

cinf

ra_g

r*in

st_i

ndex

with

Eas

t Asi

an a

nd P

acifi

c du

mm

y,

and

Sout

h A

sia

dum

my.

N

ote:

Fo

r eac

h co

lum

n, th

e in

fras

truc

ture

and

the

inst

itutio

nal i

ndic

ator

s us

ed a

re in

dica

ted

abov

e. R

obus

t sta

ndar

d er

rors

in

pare

nthe

ses.

Sour

ce:

Aut

hors

' cal

cula

tions

.


Table 9: continued.

(8) (9) (10) (11) (12)

pcgdpgrowth pcgdpgrowth pcgdpgrowth pcgdpgrowth pcgdpgrowthInfrastructure indicator Electricity Phones Railroads Roads WaterInstitutional index Institutional

qualityInstitutional




qualitypcgdp71 -1.15e-06***

(1.97e-07)pcgdp75 -1.15e-06***

(3.26e-07)pcgdp80 -2.55e-06***

(7.52e-07)pcgdp90 -1.34e-06** -7.45e-07**

(5.97e-07) (3.44e-07)School enrollment secondary

0.000116 0.000239* 0.000128 0.000135 4.65e-05

(0.000201) (0.000131) (0.000262) (0.000154) (0.000160)School enrollment primary 2.75e-05 -5.88e-05 -0.000284 -8.41e-05 7.04e-05

(0.000247) (0.000149) (0.000274) (0.000178) (0.000182)Inv/gdp 0.0638* 0.0668*** 0.0247 0.104*** 0.147***

(0.0338) (0.0222) (0.0493) (0.0324) (0.0413)Institutional quality 0.0527 -0.0784 0.428 0.231** -0.523

(0.0644) (0.0631) (0.314) (0.108) (0.568)pcinfra_gr 0.159*** 0.213*** 0.195 0.0953 0.491

(0.0420) (0.0501) (0.266) (0.0698) (0.369)pcinfra_gr*eap 0.0170** 0.0132** 0.0432*** 0.0153** 0.0114*

(0.00678) (0.00509) (0.0129) (0.00575) (0.00673)pcinfra_gr*sa 0.0112*** 0.00762 0.0219*** 0.0170*** 0.0110**

(0.00251) (0.00488) (0.00482) (0.00392) (0.00452)pcinfra_gr*inst_index -0.0405 0.0813 -0.423 -0.227** 0.532

(0.0642) (0.0580) (0.322) (0.109) (0.564)pcinfra_gr*eap*inst_index -0.00262 0.00154 -0.0360* 0.00605 0.000994

(0.00599) (0.00450) (0.0201) (0.00638) (0.00774)pcinfra_gr*sa*inst_index -0.00805 -0.0117 0.00512 0.0181** -0.00281

(0.00976) (0.00849) (0.00886) (0.00708) (0.00841)Constant -0.171*** -0.243*** -0.155 -0.0955 -0.512

(0.0541) (0.0629) (0.253) (0.0673) (0.373)Observations 46 57 26 59 60R-squared 0.819 0.741 0.854 0.588 0.638

*** p<0.01, ** p<0.05, * p<0.1.pcinfra _gr = rate of growth of per capita infrastructure indicator; pcinfra_gr*eap = interaction with East Asian and Pacific dummy; pcinfra_gr*sa = interaction with South Asia dummy; pcinfra_gr*inst_index = interaction of pcinfra _gr with institutional indicator; pcinfra_gr*eap*inst_index and pcinfra_gr*sa*inst_index = interactions of pcinfra_gr*inst_index with East Asian and Pacific dummy, and South Asia dummy. Note: For each column, the infrastructure and the institutional indicators used are indicated above. Robust standard errors in

parentheses.Source: Authors' calculations.


Note that the Kaufmann et al. (2009) indices are commonly used in cross-country applications and are built by aggregating most of the existing country-level institutional indicators along each dimension (corruption, regulatory quality, government effectiveness, etc.). As such, the results obtained here can be considered pretty reliable, in the sense that using alternative indicators would most probably produce very consistent outcomes.

We focus on interactions with infrastructure indicators rates of growth, as these have both shown consistently significant results and are also much less subject to endogeneity problems than indicators in levels.

The results are disappointing, as almost no significant effects are found. The interpretation is that the comparatively higher impact of infrastructure on per capita growth in the EAP and SA groups is not particularly driven by the subset of countries in these regions that have better institutional environment. The only exception is the case of roads in the South Asia region, which therefore indicates that the fact that the growth of the roads network has a stronger effect on per capita growth in this group of countries than in the whole sample is driven by the countries in the group with a better institutional environment (therefore Bhutan, India, and Sri Lanka). The interpretation cannot be pushed too far, however, as only five South Asian countries are included in the analysis.

Next, we turn to the growth accounting analysis to delve further into the interpretation of our results.

B. Growth Accounting

1. Data and Estimation

There are two main options for estimating the impact of infrastructure on TFP growth as summarized in equations (7) or (8) in the Appendix. One is based on regional panel data, while the other one is a country-per-country approach based on time series data.

We first perform individual country estimations, which more realistically do not assume that there is a common underlying technology for all countries. This has been the approach used by most noninfrastructure growth accounting studies (see for example Barro and Sala-i-Martin 2005). The panel estimation technique, on the other hand, rests on the assumption that a common production function exists for the countries under analysis, with individual country effects to be controlled for. While this approach has been extensively used with state/provincial panel data for India (Hulten et al. 2005), Italy (La Ferrara and Marcellino 2000), and the US (Holtz-Eakin 1994), it remains to be seen whether it can work when applied to a set of countries, albeit in the same region. We report below panel estimations suggesting that indeed this modeling of growth accounting runs into the problem of country-level heterogeneity and adds little to individual country estimations.


Concerning any possible simultaneity in the estimation of (8) we cannot rule out a

priori an influence of TFP growth on investment in infrastructure, XX . Possible causes

of simultaneity include endogenous responses of infrastructure policies to TFP growth, making it necessary to test the presence of reverse causation in the data.

Country-specific estimations, as opposed to panel estimations, call for longer time series in order to produce efficient estimators. We concentrate again on physical indicators of infrastructure that allow for longer time coverage.

With respect to explanatory variables, we use data from WDI, covering a time period up to 2006:

(i) Number of mobile and fixed-lines telephone subscribers.

(ii) Electricity: Computed from electricity generation in kilowatt-hours, net of electricity power transmission and distribution losses; this provides a quality-adjusted time series, which is used in all the estimations below.

(iii) Railway lines, total route in kilometers.

(iv) Roads data, total route in kilometers. Note that this variable could only be used for two countries.

(v) Water data could not be used as the corresponding time series are too short for all Asian countries.

The TFP growth rates, to be used as the dependent variable as in equation (8), were collected from two sources:

(i) Asian Productivity Organization (APO 2010) provides TFP growth rates for six Asian and Pacific countries (the PRC, Indonesia, the Fiji Islands, the Republic of Korea, the Philippines, Thailand) over the period 1970–2007. These are calculated (not estimated) TFP growth rates following the standard methodology in equation (3) in the Appendix and, in addition, taking into account changes in labor quality.

(ii) UNIDO’s World Productivity Database (WPD) (UNIDO 2010) provides information on levels and growth of aggregate TFP for eight other Asian and Pacific countries between 1960 and 2000 (see Isaksson 2008). We use TFP growth estimated with a nonparametric technique using Data Envelopment Analysis with Long Memory (3).25

25 See Isaksson (2008) and Noumba et al. (2009) for details. Essentially the LM DEA method relies on linear programming estimators.


Crossing available infrastructure data series, which in general are good for telephones (fixed and mobiles) and electricity (production); mediocre to bad for railroads and roads; and inexistent for water, with TFP data, we get the following coverage:

(i) For the period 1970–2007 (APO data):

(a) the PRC (telecom, electricity, railroads) (b) Indonesia (telecom, electricity) (c) Philippines (telecom, electricity) (d) Thailand (telecom, electricity, railroads) (e) Fiji Islands (telecom) (f) Republic of Korea (telecom, electricity, railroads, roads)

(ii) For the period 1961–2000 (UNIDO data):

(a) Hong Kong, China (telecom, electricity) (b) India (telecom, electricity, railroads) (c) Nepal (telecom, electricity) (d) Pakistan (telecom, electricity, railroads, roads) (e) Sri Lanka (telecom, electricity) (f) Malaysia (telecom, electricity, railroads) (g) Singapore (telecom, electricity) (h) Papua New Guinea (telecom)

2. Results

Tables 10a, 10b, and 11 report the results from individual growth accounting regressions. The results of regressions including only the electricity and telecom variables for the 14 countries included in our sample are in Tables 10a and 10b, while Table 11 adds railroads and roads when available.

First, note that in 11 out of 14 countries, we cannot reject the hypothesis that the coefficients of both electricity and telecom are zero. Again, recall that the interpretation for this result is not that infrastructure is not productive but rather that there is no evidence that it is more productive than other types of capital. This conclusion is not modified for the two countries of this group also included in Table 3, where neither railroads nor roads are significant.


Table 10a: Individual Country Regressions (dependent variable: TFP growth rate)

(1) (2) (3) (4) (5) (6) (7)

China, People’s Rep. of

Fijiislands

Hong Kong,China

India Indonesia Korea,Rep. of

Malaysia

Growth rate of mobile and fixed-line telephony per capita

0.0497* 0.0913 -0.280 0.0230 0.0750 -0.0757 -0.00822

(0.0272) (0.107) (0.234) (0.0818) (0.0537) (0.0820) (0.101)

Growth rate of per capita electricity production net of losses

0.346** 0.113 0.288 0.0491 0.279** 0.0311

(0.152) (0.0827) (0.367) (0.111) (0.126) (0.123)

Constant -0.00374 -0.00860 0.0344* -0.00144 -0.0181 -0.00622 0.0131

(0.0182) (0.0138) (0.0197) (0.0314) (0.0265) (0.0153) (0.0176)

Observations 31 31 25 25 31 28 23

R-squared 0.234 0.030 0.175 0.077 0.053 0.228 0.003

*** p<0.01, ** p<0.05, * p<0.1.TFP = total factor productivity.Note: Robust standard errors in parentheses.Source: Authors' calculations.

Table 10b: Individual Country Regressions

(8) (9) (10) (11) (12) (13) (14)

Nepal Pakistan Papua New

Guinea

Philippines Singapore Sri Lanka Thailand


-0.00635 -0.0515 -0.0395 0.0675 -0.0666 0.000378 0.0238

(0.0347) (0.0536) (0.212) (0.0485) (0.104) (0.0405) (0.0435)


0.00182 0.0719 0.107 0.314 -0.151 0.498**

(0.0277) (0.106) (0.192) (0.243) (0.121) (0.197)

Constant 0.00543 0.0160 -0.00398 -0.0193 0.0104 0.00894 -0.0285

(0.0109) (0.00954) (0.0126) (0.0125) (0.0158) (0.00880) (0.0253)

Observations 25 25 25 31 20 25 31

R-squared 0.001 0.043 0.001 0.118 0.074 0.051 0.253

*** p<0.01, ** p<0.05, * p<0.1.Note: Robust standard errors in parentheses.Source: Authors' calculations.


As for the significant results, they concern the PRC, the Republic of Korea, and Thailand. For the PRC, the total number of telephones (fixed plus mobiles) has a positive coefficient of 0.02, significant at the 1% level, while electricity generating capacity adjusted for losses also appears to be more productive than standard capital, with a 5% level of significance. This can be interpreted as an externality effect expressed as an output elasticity of 0.12 for telecom and 0.35 for electricity.

In the Republic of Korea and Thailand, the electricity variable also has a positive coefficient of 0.28 and 0.50 respectively, significant at the 5% level, again supporting externalities from this variable. Finally, note that in these three countries, the growth of infrastructure indicators explains close to one fourth of the TFP growth (R2 of 0.23 for the PRC, 0.23 for the Republic of Korea, and 0.25 for Thailand).

Table 11: Individual Country Regressions with Railroads and Roads (dependent variable: TFP growth rate)

(1) (2) (3) (4) (5) (6)

China, People’s Rep. of

India Korea,Rep. of

Malaysia Pakistan Thailand


0.0395 -0.0452 -0.211*** -0.123 0.00528 0.0228

(0.0282) (0.0545) (0.0525) (0.111) (0.0806) (0.0496)


0.274** 0.0356 0.216* 0.0956 0.0175 0.461**

(0.123) (0.249) (0.111) (0.113) (0.0710) (0.221)

Growth rate of per capita kilometer rail -0.906 0.196 0.0460 -0.614 -0.0809 -0.170

(0.539) (0.473) (0.0617) (0.939) (0.0947) (0.418)

Growth rate of per capita kilometer roads

0.00146 0.0346

(0.0293) (0.104)

Constant 0.00498 0.0290 0.0118 0.00477 -0.00242 -0.0263

(0.0170) (0.0202) (0.00882) (0.0335) (0.0169) (0.0291)


R-squared 0.191 0.044 0.758 0.160 0.058 0.233

*** p<0.01, ** p<0.05, * p<0.1.TFP = total factor productivity.Note: Robust standard errors in parentheses.Source: Authors' calculations.

In Table 3, the results on electricity are robust to including the railroads variable in the specifications, while the telecom variable becomes insignificant for the PRC and negative for the Republic of Korea. However, this could be due to the fact that including


railroads (and roads in the case of the Republic of Korea) reduces the number available observations, from 31 to 25 for the PRC, and from 28 to 12 for the Republic of Korea.

These individual country regressions could be missing cross-country variations explaining differences in TFP growth. Table 12 reports the results from panel regressions.

Table 12: Panel Regressions

(1) (2) (3) (4) (5) (6)

tfp_g tfp_g tfp_g tfp_g tfp_g tfp_gGrowth rate of per capita electricity production net of losses

0.0910** 0.0924** 0.228***

(0.0462) (0.0471) (0.0722)Growth rate of mobile and fixed-line telephony per capita

0.0283* 0.00663

(0.0160) (0.0186)Growth rate of per capita kilometer rail -0.0460

(0.0500)Growth rate of electricity production, net of losses

0.0653 0.0643 0.220***

(0.0458) (0.0522) (0.0556)Growth rate of mobile and fixed-line telephony

0.0293* 0.0102

(0.0161) (0.0189)Growth rate of kilometer rail -0.0281

(0.0590)Constant 0.00468* 0.00134 -0.00410 0.00514 0.00116 -0.00765

(0.00283) (0.00426) (0.00889) (0.00383) (0.00594) (0.00811)Observations 378 320 178 378 320 178Number of id_country 12 12 9 12 12 9R-squared (overall) 0.0266 0.0362 0.0730 0.0183 0.0259 0.0654Rho 0.0454 0.0734 0.152 0.0461 0.0727 0.128Hausman test (FE vs RE) p-value 0.85 0.86 0.52 0.78 0.84 0.00*

*** p<0.01, ** p<0.05, * p<0.1.Note: Full set of period dummies included. Robust standard errors in parentheses.Source: Authors' calculations.

The electricity variable comes out positive and significant in columns 1–3, i.e., when per capita growth rate of the infrastructure indicators is used. Additionally, in column 2 the telecom variable is also positive and significant. In columns 4–6, similar albeit slightly weaker results are obtained when using aggregate rather than per capita growth rates. Note that a Hausman specification test supports the random effect specification in all but the last column, which is not surprising considering that the countries included (between 9 and 12) are drawn from a larger population of countries. However, the fixed effect estimation of the specification in column 6 yields very similar results and is not reported. Also, the overall R² (measuring goodness of fit for both between and within variations) is rather small in all specifications.


Overall, the rho-statistic and the results from a between-effect version of the panel regressions in Table 12, not shown here to save space, indicate that most of the significant results are driven by the “within” (that is, individual country-level) variation. The panel analysis does not add anything to the results in Tables 10a, 10b, and 11, probably because the assumption needed to run such estimations, i.e., that countries in the sample share some common technology, is not warranted in our sample, which contains very heterogenous countries, from large emerging Asian countries such as the PRC, to small Pacific island economies.

Finally, because TFP series are by construction noisy, in Table 13 we turn to 5-year averages to smooth them out. The resulting panel size is reduced to between 38 and 77 observations. The results corresponding to the total number of phone lines are maintained, while electricity is no longer significant. However, the small panel size (N between 9 and 12, t between 3 and 7) makes these results obviously fragile.

Table 13: Panel Regressions, 5-year Averages

(1) (2) (3) (4) (5) (6)

tfp_g tfp_g tfp_g tfp_g tfp_g tfp_g


0.0734 0.0122 0.0279

(0.0655) (0.0772) (0.137)


0.0609* 0.0582

(0.0318) (0.0393)

Growth rate of per capita kilometer rail 0.167

(0.358)

Growth rate of electricity production, net of losses

0.0699 0.00902 0.0327

(0.0646) (0.0882) (0.134)

Growth rate of mobile and fixed-line telephony

0.0613** 0.0571

(0.0309) (0.0414)

Growth rate of kilometer rail 0.139

(0.462)

Constant 0.0182*** 0.00836 0.00894 0.0175*** 0.00738 0.00603

(0.00271) (0.00959) (0.00782) (0.00356) (0.0108) (0.00932)


Period dummies Yes Yes Yes Yes Yes Yes

Number of id_country 12 12 9 12 12 9

R-squared (overall) 0.180 0.222 0.329 0.181 0.221 0.310

Rho 0.0998 0.217 0.172 0.104 0.218 0.161

*** p<0.01, ** p<0.05, * p<0.1.Note: Robust standard errors in parentheses.Source: Authors' calculations.


V. Conclusion

Infrastructure stocks in developing Asia have been growing at a significant pace. We find, however, that their levels remain well below corresponding world averages in terms of both quantity and quality. As infrastructure-stock accumulation (in electricity, telecommunications, transport, and water supply) has a positive impact on economic growth, a massive build-up of these stocks is needed, but may well be beyond the financing capacities of many governments.

Moreover, demand for infrastructure services is expected to soar in cities due to rapid urbanization. In order to keep cities in developing Asia competitive, investments in infrastructure need to be designed to take account of congestion, environmental degradation, and other impediments to productivity that are associated with urban agglomeration. The urban–rural divide in access to infrastructure services detracts from the inclusiveness of growth in developing Asia. Improving access to basic infrastructure services, such as the provision of potable water and sanitation, in rural areas is crucial for poverty alleviation.

After reviewing the state of infrastructure development in developing Asia and the literature on infrastructure and development, we have performed a number of empirical exercises to assess the contribution of infrastructure to growth and productivity across a number of dimensions (electricity, telecommunications, railroads, roads and water).

Cross-country estimations show that for most infrastructure indicators, the growth rate of stocks has a positive and significant impact on per capita GDP average growth rate in the subgroups of EAP and SA countries. The growth accounting exercise, on the other hand, shows that positive and significant effects of infrastructure on TFP growth are only observed in a few countries (the PRC, the Republic of Korea, Thailand), for telecom and electricity indicators. At face value, a possible interpretation is that in most Asian countries, the observed effect on growth was simply the results of higher than average infrastructure capital accumulation (a “direct effect”), but that additional productivity enhancing (“indirect”) effects were rather rare.

East Asia’s economic history seems to give credit to that claim. As discussed in Straub, Vellutini, and Warlters (2008), between 1975 and 1995 East Asia accumulated infrastructure at a rate that outpaced other regions (see Table 14). The PRC and Viet Nam, the two fastest-growing economies in the region invest around 10% of GDP in infrastructure, and other countries, for example those in the Greater Mekong countries (the PRC, Cambodia, Indonesia, the Lao PDR, Myanmar, Thailand, and Viet Nam) are planning to reach investment levels above 5%–6%.


Table 14: 1995 infrastructure Stocks as Multiples of 1975 Levels

Electricity Roads TelecommunicationsEast Asia 5.9 2.9 15.5South Asia 4.4 2.5 8.2Middle East and North Africa 6.1 2.1 7.2Latin America and Caribbean 3.0 1.9 5.1OECD 1.6 1.4 2.2Pacific 2.0 4.3Sub-Saharan Africa 2.6 1.7 3.9Eastern Europe 1.6 1.2 6.9

OECD = Organisation for Economic Co-operation and Development.Note: Electricity = generating capacity in megawatts; roads = paved roads in kilometers; Telecommunications = number of main

lines. See Appendix for construction. Sources: Straub, Vellutini, and Warlters (2008); World Development Indicators (World Bank, various years); Canning (1998).

One must be cautious however, because due to the limited availability of data, there is no perfect match between the samples and the indicators used in both exercises. In particular, not all countries have long enough TFP data, and not all infrastructure series are long enough to be included in the growth accounting estimations.

One final comment can be made regarding convergence in the region. As shown by the cross-country regressions, initial values of per capita GDP are consistently negative and significant, indicating indeed the existence of convergence in our sample. Furthermore, specific results regarding the interaction between income levels and infrastructure endowments from Straub et al. (2008) are relevant. As shown there, infrastructure indicators appear to have a significantly lower impact among low- and middle-income countries, compared to high-income ones. That means that when we interact the specific infrastructure indicators with dummy variables for low-, middle-, and high-income countries, the results show that the net effect of infrastructure is lower in the low- and middle-income groups than in the high-income one (being actually negative in some cases for low-income countries).

The first possibility, consistent with some of the existing literature on telecommunication (for example, see Röller and Waverman 2001), is simply a network effect type of explanation. In the case of roads, a similar argument could maybe be made referring to the importance of regional integration to potentiate physical infrastructure investments among the fastest growing countries, for example in the Greater Mekong subregion (see Stone et al. 2009). Another line of explanation is that more developed countries also have a more favorable institutional environment, e.g., better property rights, which boosts the impact of infrastructure investments or facilitate their implementation. Although our data does not allow us to isolate such an effect, it may be the case that a more microeconomic approach to institution and infrastructure measurement is needed to illustrate such channels.


AppendixCross-country Regressions

This section is adapted from Straub et al. (2008). Cross-country regressions use as dependent variable real per capita GDP growth, and as explanatory variables the initial level of real per capita GDP and other additional factors such as physical investment, human capital, etc.26 Infrastructure capital can be included as an additional explanatory factor, yielding a specification of the form:

g yi i0= + + +α β γ νK ZiI

i i (1)

where gi is the growth rate of real per capita GDP for country i, yi0 is initial income (possibly in log form), KI

i is a measure of infrastructure capital, and Zi is a vector of controls.27

Growth Accounting Framework

The growth accounting framework is based on a standard production function of the type

Y A F K L= . ( , ) (2)

where Y is aggregate GDP, A is the time-varying total factor productivity (TFP) and K and L are (total) capital and labor, respectively. Taking logs and differentiating with respect to time yields

YY

AA

F LF

LL

F KF

KK

L K= + +. .

(3)

Assuming that marginal factor productivities equal factor prices, we get the standard formula for growth accounting, where the growth of TFP is computed as the residual between the growth of GDP and the growth of factors:

AA

YY

SLL

SKK

L K= − − (4)

In this equation SL and Sk are the respective observed shares of income. Importantly, equation (4) is typically implemented through direct calculation, as all the variables on the right-hand side are observed, rather than through econometric estimation, and is commonly used in country-specific studies to calculate TFP growth (see Barro and Sala-i-Martin 2005).

Assuming that infrastructure (denoted X) influences output through two channels, it can be added straightforward to this framework in the following way (Hulten et al. 2005). First, it impacts TFP through

A A X A X= =( ) . η

(5)

26 See for example Barro and Sala-i-Martin (2005).27 See Straub (2010) for a discussion of the limitations of such estimations in the context of infrastructure.


where A is the « true » TFP and η is the elasticity of A with respect to X. This channel captures the externality aspect of infrastructure, as the output-increasing effect occurs without any payments by firms for infrastructure services. Note that η is therefore not observable.

Second, infrastructure enters the production function as an additional factor of production:

Y A X F K L X= . . ( , , )η

(6)

where K is the stock of noninfrastructure capital. The presence of infrastructure as one more factor reflects its market-mediated (“direct”) impact, whereby firms pay for infrastructure services. This implies:

& %&

%

& & & %&

%%Y

YAA

XX

SXX

SLL

SKK

X L K= + + + +η (7)

where S X is the share of GDP accruing to market-mediated infrastructure and SK the share of revenue that accrues to noninfrastructure capital. If S X was observable we could disentangle the direct influence of infrastructure from its indirect effect. In practice data on infrastructure prices are not available in a consistent way and available data on capital do not distinguish between different types of capital, such as infrastructure. We can rewrite the model so as to fit the available data, as:

Y A X F K L= . . ( , )η

(6’)

which implies

& %&

%

& & &YY

AA

XX

SLL

SKK

L K= + + + +η ε

(7’)

Substituting equation (3) into (6’), and appending an error term, we get:

& %&

%

&AA

AA

XX

= + +η ε (8)

The left-hand side of equation (8) is TFP growth computed (not estimated) from the standard growth accounting approach. Rather than performing a full estimation of equation (7’), we can estimate the reduced form equation (8), using the (year by year) results of equation (3) for TFP

growth rates (AA ), which are usually available from standard growth accounting exercises for a

number of countries.

Either equation (8) or equation (7’) provide an estimation of η, the externality effect of infrastructure, as opposed to the full elasticity of output with respect to infrastructure. If an estimation of equation (8) produces a value for η not significantly different from zero, we conclude that infrastructure is not more productive than other types of capital, in the sense that it exerts no specific externality.


ReferencesAlby, P., J.-J. Dethier, and S. Straub 2009. “Firms Operating under Infrastructure and Credit

Constraints in Developing Countries. The Case of Power Generators.” World Bank and Toulouse School of Economics. Mimeo.

Andres L., J. L. Guasch, and S. Lopez Azumendi. 2007. “Regulatory Governance and Sector Performance: Methodology and Evaluation for Electricity Distribution in Latin America.” World Bank, Washington, DC. Mimeo.

APO. 2010. APO Productivity Databook 2010. Asian Productivity Organization, Tokyo.Aschauer, D. A. 1989. “Is Public Expenditure Productive?” Journal of Monetary Economics 23:177–

200.Baldwin, R., R. Forslid, P. Martin, G. Ottaviano, and F. Robert-Nicoud. 2003. Economic Geography

and Public Policy. New Jersey: Princeton University Press.Barro, R. J., and J.-W. Lee. 2010. A New Dataset of Educational Attainment in the World, 1950–

2010. NBER Working Paper No. 15902, National Bureau of Economic Research, Cambridge.Barro R., and X. Sala-i-Martin. 2005. Growth Theory, 2nd ed. Boston: MIT Press.Bloom, D. E., D. Canning, and G. Fink. 2008. “Urbanization and the Wealth of Nations.” Science

319:772–75.Bom, P., and J. Ligthart. 2008. How Productive is Public Capital? A Meta-Analysis. Cesifo Working

Paper No. 2206, Ifo Institute for Economic Research, Munich.Brooks, D. H., and D. Hummels, eds. 2009. Infrastructure’s Role in Lowering Asia’s Trade Costs:

Building for Trade. Cheltenham: Edward Elgar Publishers.Calderón, C., and L. Servén. 2004. The Effects of Infrastructure Development on Growth and

Income Distribution. Working Paper No. 270, Central Bank of Chile, Santiago.Calderón, C., E. Moral, and L. Servén. 2009. “Is Infrastructure Capital Productive? A Dynamic

Heterogeneous Approach.” World Bank, Washington, DC. Mimeo.Canning, D. 1998. A Database of World Infrastructure Stocks, 1950-95. Policy Research Working

Paper 1929, World Bank, Washington, DC.De, P. 2010. Governance, Institutions, and Regional Infrastructure in Asia. ADBI Working Paper

Series No. 183, Asian Development Bank Institute, Tokyo.Deichmann, U., M. Fay, J. Koo, and S. Lall. 2002. “Economic Structure, Productivity, and

Infrastructure Quality in Southern Mexico.” The World Bank, Washington, DC. Mimeo.Deichmann, U., K. Kaiser, S. V. Lall, and Z. Shalizi. 2005. Agglomeration, Transport and

Regional Development in Indonesia. World Bank Policy Research Working Paper No. 3477, Washington, DC.

Dethier, J-J., M. Hirn, and S. Straub. 2008. Explaining Enterprise Performance in Developing Countries with Business Climate Survey Data. World Bank Policy Research Working Paper No. 4792, World Bank, Washington, DC.

Dewatripont, M., and P. Seabright. 2006. “Wasteful Public Spending and Public Control.” Journal of the European Economic Association 4(2-3):513–22.

Dinkelmann, T. 2009. “The Effects of Rural Electrification on Employment: New Evidence from South Africa.” Quarterly Journal of Economics. Forthcoming.

Donaldson, D. 2009. “Railroads of the Raj: Estimating the Impact of Transportation Infrastructure.” London School of Economics. Mimeo.

Duflo, E., and R. Pande. 2007. “Dams.” Quarterly Journal of Economics 122(2):601–46.Engel, E., R. Fischer, and A. Galetovic. 2009. “On the Efficient Provision of Roads.” Paper

prepared for the Project Development Workshop on Varieties of Governance: Effective Public Service Delivery, 1–2 February 2009, Kuwait.

Fan, S., D. Nyange, and N. Rao. 2005. Public Investment and Poverty Reduction in Tanzania: Evidence from Household Survey Data. DSGD Discussion Paper 18, International Food Policy Research Institute, Washington, DC.


Fay, M., and M. Morrison. 2007. Infrastructure in Latin America and the Caribbean. Recent Development and Key Challenges. The World Bank, Washington, DC.

Fernald, J. G. 1999. “Roads to Prosperity? Assessing the Link between Public Capital and Productivity.” American Economic Review 89(3):619–38.

Galiani, S., P. Gertler, and E. Schargrodsky. 2005. “Water for Life: The Impact of Privatization of Water Services on Child Mortality.” Journal of Political Economy 113:83–120.

Gibson J., and S. Rozelle. 2003. “Poverty and Access to Roads in Papua New Guinea.” Economic Development and Cultural Change 52(1):159–85.

Gramlich, E. M. 1994. “Infrastructure Investment: A Review Essay.” Journal of Economic Literature 32:1176–96.

Guasch, J. L., and J. Kogan. 2001. Inventories in Developing Countries: Levels and Determinants—A Red Flag for Competitiveness and Growth. Policy Research Working Paper 2552, World Bank, Washington, DC.

Guasch, J. L, J. J. Laffont, and S. Straub. 2007. “Infrastructure Concessions in Latin America: Government-led Renegotiations.” Journal of Applied Econometrics 22(7):1267–94.

———. 2008. “Renegotiation of Concession Contracts in Latin America.” International Journal of Industrial Organization 26(2):421–42.

Holtz-Eakin, D. 1994. “Public-Sector Capital and the Productivity Puzzle.” Review of Economics and Statistics 76:12–21.

Hulten, C. R., E. Bennathan, and S. Srinivasan. 2005. “Infrastructure, Externalities, and Economic Development: A Study of Indian Manufacturing Industry.” World Bank, Washington, DC. Mimeo.

Hulten, C. R., and R. Schwab. 2000. “Does Infrastructure Investment Increase the Productivity of Manufacturing Industry in the US?” In L. J. Lau, ed., Econometrics and the Cost of Capital. Cambridge, MA: MIT Press.

Ilahi, N., and F. Grimard. 2000. “Public Infrastructure and Private Costs: Water Supply and Time Allocation of Women in Rural Pakistan.” Economic Development and Cultural Change 49(1):45–75.

International Monetary Fund. 2010. World Economic Outlook Database. Available: www.imf.org/external/pubs/ft/weo/2010/01/weodata/index.aspx.

International Telecommunication Union. 2010. ICT Statistics Database. Available: www.itu.int/ITU-D/icteye/Indicators/Indicators.aspx#, downloaded 3 August 2010.

Isaksson, A. 2008. “World Productivity Database: A Technical Description.” United Nations Industrial Development Organization, Vienna.

Jacoby, H. 2000. “Access to Markets and the Benefits of Rural Roads.” The Economic Journal 110(465):713–37.

Jensen, R. 2007. “The Digital Provide: Information (Technology) Market Performance, and Welfare in the South Indian Fisheries Sector.” Quarterly Journal of Economics 122(3):879–24.

Kaufmann, D., A. Kraay, and M. Mastruzzi. 2009. Governance Matters VIII: Aggregate and Individual Governance Indicators, 1996–2008. World Bank Policy Research Working Paper No. 4978, Washington, DC.

Koolwal, G., and D. Van de Walle. 2010. Access to Water, Women’s Work and Child Outcomes. World Bank Policy Research Working Paper No. 5302, Washington, DC.

Kremer, M. 1995. “The O-ring Theory of Economic Development.” Quarterly Journal of Economics 108(3):551–75.

La Ferrara, E., and M. Marcelino. 2000. “TFP, Costs and Public Infrastructure: An Equivocal Relationship.” Bocconi University and IGIER, Milan. Mimeo.

Lall, S., J. Sandefur, and H. G. Wang. 2009. “Does Improving Market Access Help De-industrialize Lagging Regions? Leaking by Linking in Ghana.” World Bank. Mimeo.

Lall, S. V., R. Funderburg, and T. Yepes. 2004. Location, Concentration, and Performance of Economic Activity in Brazil. World Bank Policy Research Working Paper WPS 3268, Washington, DC.


Lall, S. V., Z. Shalizi, and U. Deichmann. 2004. “Agglomeration Economies and Productivity in Indian Industry.” Journal of Development Economics 73:643–73.

Lardy, N. 2010. “Yes, China Does Need That Infrastructure.” Asian Wall Street Journal. 24 June.Li, Z., and H. Li. 2009. “Road Investment and Inventory Reduction: Evidence from a Large

Developing Country.” Paper presented at the International Conference on Infrastructure Economics and Development organized by the World Bank, Agence Française de Développement, the Public Private Infrastructure Advisory Facility and the Institut D'économie Industrielle, 14–15 January 2010, Toulouse.

Maskin, E., and J. Tirole. 2006. Public-Private Partnerships and Government Spending Limits. IDEI Working Paper No. 439, Institut d'Economie Industrielle, Toulouse. Available: idei.fr/doc/by/tirole/ppp.pdf.

McKenzie, D., and D. Mookherjee. 2003. “The Distributive Impact of Privatization in Latin America: Evidence from Four Countries.” Economía 3:161–233.

Noumba, U. P., S. Straub, and C. Vellutini. 2009. Infrastructure and Economic Growth in the Middle East and North Africa. Policy Research Working Paper Series 5105, World Bank, Washington, DC.

Pritchett, L. 1996. Mind your P’s and Q’s. The Cost of Public Investment is Not the Value of Public Capital. Policy Research Working Paper 1660, World Bank, Washington, DC.

Prud’homme, R. 2005. “Infrastructure and Development.” In F. Bourguignon and B. Pleskovic, eds., Lessons of Experience, Proceedings of the 2004 Annual Bank Conference on Development Economics. Washington, DC: The World Bank and Oxford University Press.

Rioja, F. 2003. “Filling Potholes: Macroeconomic Effects of Maintenance versus New Investments in Public Infrastructure.” Journal of Public Economics 87:2281–304.

Röller, L. H., and L. Waverman. 2001. “Telecommunications Infrastructure and Economic Development: A Simultaneous Approach.” American Economic Review 91:909–23.

Romp, W., and J. de Haan. 2005. “Public Capital and Economic Growth: A Critical Survey.” European Investment Bank Papers 2/2005:40-70.

Seethepalli, K., M. C. Bramati, and D. Veredas. 2008. How Relevant is Infrastructure to Growth in East Asia? Policy Research Working Paper No. 4597, World Bank, Washington, DC.

Stone, S., and A. Strutt. 2009. Transport Infrastructure and Trade Facilitation in the Greater Mekong Subregion. ADBI Working Paper No. 130, Asian Development Bank Institute, Tokyo.

Straub, S. 2008. Infrastructure and Growth in Developing Countries: Recent Advances and Research Challenges. Policy Research Working Paper 4460, World Bank, Washington, DC.

———. 2010. “Infrastructure and Development: A Critical Appraisal of the Macro-level Literature.” Journal of Development Studies. Forthcoming.

Straub, S., C. Vellutini, and M. Warlters. 2008. Infrastructure and Economic Growth in East Asia. Policy Research Working Paper 4589, World Bank, Washington, DC.

Sturm J.-E., G. H. Kuper, and J. de Haan. 1998. “Modelling Government Investment and Economic Growth on a Macro Level: A Review.” In S. Brakman, H. van Ees, and S. K. Kuipers, eds., Market Behaviour and Macroeconomic Modelling. London: MacMillan Press Ltd.

Thomas D., and J. Strauss. 1992. “Prices, Infrastructure, Household Characteristics and Child Height.” Journal of Development Economics 39:301–31.

UNIDO. 2010. World Productivity Database. Available: www.unido.org/data1/wpd. World Bank. 2004. Reforming Infrastructure. Privatization, Regulation and Competition.

Washington, DC.———. 2009. Enteprise Surveys. Available: www.enterprisesurveys.org/, downloaded 4 November

2009.———. 2010. Logistics Performance Index. Available: info.worldbank.org/etools/tradesurvey/

mode1b.asp.World Economic Forum. 2005. Global Competitiveness Report 2005–2006. Geneva.———. 2006. Global Competitiveness Report 2006–2007. Geneva.———. 2009. Global Competitiveness Report 2009–2010. Geneva.


http://www.enterprisesurveys.org/

About the PaperStéphane Straub and Akiko Terada-Hagiwara present the state of infrastructure in developing Asian countries, and analyze the link between infrastructure, growth, and productivity. The main conclusion is that a number of countries in developing Asia have significantly improved their basic infrastructure endowments in the recent past, and this appears to correlate significantly with good growth performances. However, the evidence seems to indicate that this is mostly the result of factor accumulation (a direct effect), while the impact on productivity is rather inconclusive.

About the Asian Development BankADB’s vision is an Asia and Pacific region free of poverty. Its mission is to help its developing member countries substantially reduce poverty and improve the quality of life of their people. Despite the region’s many successes, it remains home to two-thirds of the world’s poor: 1.8 billion people who live on less than $2 a day, with 903 million struggling on less than $1.25 a day. ADB is committed to reducing poverty through inclusive economic growth, environmentally sustainable growth, and regional integration. Based in Manila, ADB is owned by 67 members, including 48 from the region. Its main instruments for helping its developing member countries are policy dialogue, loans, equity investments, guarantees, grants, and technical assistance.

Asian Development Bank6 ADB Avenue, Mandaluyong City1550 Metro Manila, Philippineswww.adb.org/economicsISSN: 1655-5252Publication Stock No. WPS102881 Printed in the Philippines

< 0 1 0 2 8 8 1 2 >

ADB Economics Working Paper SeriesStéphane Straub and Akiko Terada-Hagiwara November 2010 Stéphane Straub is Professor at the Toulouse School of Economics, Arqade; Akiko Terada-Hagiwara

Documents