January 2022

ASIAN INFRASTRUCTURE FINANCE 2021 ANNEX

MEASURING TRANSPORT
CONNECTIVITY FOR
TRADE IN ASIA

WRITTEN BY
© Economist Intelligence Unit 2021

Asian Infrastructure Investment Bank

AIIB Headquarters, Tower A, Asia Financial Center
No. 1 Tianchen East Road, Chaoyang District, Beijing 100101 China
Tel: +86-10-8358-0000
[email protected]

CC BY-NC-ND 3.0 IGO. Some rights reserved.

Asian Infrastructure Investment Bank (AIIB) uses a Creative Commons

license for all intellectual property contained in this work except where
specified. Copyright designations may or may not apply to third-party
content found herein. AIIB does not necessarily own each component
of the content contained in or linked from this work and will not be held
accountable for any such third-party content. AIIB does not guarantee
the accuracy of the data contained in or linked from this work and
accepts no responsibility for any consequences of their use. The mention
of companies or any trademarked entity or object in this workdoes
not imply that they are being endorsed or recommended by AIIB in
preference to others that are not mentioned. The contents of this work
do not necessarily represent the views or policies of AIIB, its Board of
Directors or its members. Any designation of or reference to a specific
territory or geographic area, or the use of the term “country” in this work
does not constitute and shall not be construed as constituting an express
or implied position, endorsement, acceptance or expression of opinion by
AIIB as to the legal or other status of any territory or area.
ASIAN INFRASTRUCTURE FINANCE 2021 ANNEX

MEASURING TRANSPORT
CONNECTIVITY FOR
TRADE IN ASIA
i || CONTENTS

CONTENTS
ABOUT THE REPORT 1

EXECUTIVE SUMMARY 2

INTRODUCTION 3

METHODOLOGY 4

Computing fastest paths 4

From fastest paths to network efficiency 5

Isochrones 6

Using isochrones to measure connectivity 7

ROAD CONNECTIVITY IN ASIA 8

Domestic road connectivity analysis 8

Case study – Bangladesh’s bridge problem 12

Cross-border road connectivity analysis 13

Border road connectivity and trade 14

PORT CONNECTIVITY IN ASIA 16

Industry access to ports 17

Case Study - Why port connectivity matters for India’s exports 19

SPECIAL ECONOMIC ZONES IN ASIA 23

Case Study – China’s SEZ export performance and

transport infrastructure 28
LIST OF TABLES,
FIGURES AND BOXES
TABLES

Table 1: Speed assumptions by road type 5

Table 2: Data sources for port connectivity analysis 7

Table 3: Network efficiency ratios for AIIB regional member countries 8

Table 4: Shapley value regression decomposing NER determinants 10

Table 5: Ranking of Asian countries by network efficiency and associated factors 10

Table 6: Cross-border Network Efficiency Ratios in Asia 14

Table 7: Regression output for drive time specification 29

FIGURES

Figure 1: Illustration of fastest and ideal paths 5

Figure 2: 4-hour isochrone from Shanghai port 6

Figure 3: Road Network Efficiency Ratios and GDP per capita in Asia 9

Figure 4: Bangladesh’s lack of downstream bridges 12

Figure 5: Simulating new connections in Bangladesh 12

Figure 6: Bilateral trade and network efficiency in Asia 15

Figure 7: Isochrones of liner ports in Asia 16

Figure 8: 4-hour and ideal isochrones for ports in Southeast Asia 17

Figure 9: Current and Potential Port-Factory Connectivity Scores by Country 18

Figure 10: Isochrones of India’s top 20 ports by export value in 2019 19

Figure 11: Correlation of port exports in 2019 and factory count by isochrone 20

Figure 12: 4-hour Port-Factory Connectivity by major port 20

Figure 13: 4-hour isochrones for select ports 21

Figure 14: Industrial clusters and port connectivity in Mumbai and Mangalore 21

Figure 15: Manufacturing exports and port connectivity ratios 22

Figure 16: SEZs by year of establishment 23

iii || LIST OF TABLES, FIGURES AND BOXES

Figure 17: Share of new SEZs within 100km of coastline 24

Figure 18: SEZs locations in Asia 24

Figure 19: Share of new SEZs by sector across Asia 25

Figure 20: Share of new services SEZs by country 25

Figure 21: Sector contributing to the service sector boom 26

Figure 22: Country share of new SEZs in ICT industries 26

Figure 23: Country share of textiles, clothing and leather SEZs 27

Figure 24: Distribution of SEZs in China 28

 COUNTRY ABBREVIATIONS || iv

COUNTRY
ABBREVIATIONS
ISO 3 country code Country AIIB Member
AFG Afghanistan Afghanistan
ARE United Arab Emirates United Arab Emirates
ARM Armenia
AZE Azerbaijan Azerbaijan
BGD Bangladesh Bangladesh
BGR Bulgaria
BLR Belarus Belarus
BRN Brunei Brunei Darussalam
BTN Bhutan
CHN China China
DEU Germany Germany
EGY Egypt Egypt
EST Estonia
GEO Georgia Georgia
GRC Greece Greece
IDN Indonesia Indonesia
IND India India
IRN Iran Iran
IRQ Iraq
ISR Israel Israel
JOR Jordan Jordan
KAZ Kazakhstan Kazakhstan
KGZ Kyrgyz Republic Kyrgyz Republic
KHM Cambodia Cambodia
KOR South Korea Korea
KWT Kuwait
LAO Laos Lao PDR
LBN Lebanon
LTU Lithuania
LVA Latvia
MMR Myanmar Myanmar
MNG Mongolia Mongolia
MYS Malaysia Malaysia
NLD Netherlands Netherlands
NPL Nepal Nepal
OMN Oman Oman
PAK Pakistan Pakistan
POL Poland Poland
QAT Qatar Qatar
RUS Russia Russia
SAU Saudi Arabia Saudi Arabia
SGP Singapore Singapore
SYR Syria
THA Thailand Thailand
TJK Tajikistan Tajikistan
TKM Turkmenistan
TUR Turkey Turkey
UZB Uzbekistan Uzbekistan
VNM Vietnam Viet Nam
YEM Yemen
 About the report

1 ||

ABOUT THE REPORT

Measuring Transport Connectivity for Trade in Asia was commissioned by the Asian Infrastructure
Investment Bank (AIIB) as an annex to Asia Infrastructure Finance 2021: Sustaining Global Value
Chains, the AIIB’s annual flagship publication. It employs geospatial techniques to understand transport
infrastructure development in Asia and the link between connectivity and global value chains. The report
was written by The Economist Intelligence Unit (EIU), which is solely responsible for the content of this
report. The EIU research team includes Liu Weisi, Tang Jie, Minakshi Barman and Alexander van Kemenade.
 || 2

EXECUTIVE SUMMARY
Asia has made significant strides in building its physical transport network in recent decades. At the
same time, businesses in the region have accelerated their integration into global value chains. However,
significant infrastructure imbalances remain across the region and present obstacles for international
trade to realise its full potential in boosting economic growth.

This report employs novel geospatial techniques to measure transport connectivity in AIIB member
countries, focusing on road and seaport connections. These new measures give rise to the following findings:

• There is substantial variation in domestic road connectivity across countries, with the highest-scoring
(China and the Gulf states) performing on par with or even exceeding the best-connected countries in
Europe, such as Germany and the Netherlands.
• 
Countries experiencing lagging connectivity, such as Kyrgyzstan and Bangladesh, tend to be
challenged by either rugged terrain or the presence of large rivers. Yet, well-targeted infrastructure
projects, such as Bangladesh’s Padma Bridge, carry the potential to significantly boost these countries’
connectivity scores.
• Highway networks are the main missing ingredient in Asia’s road connectivity, with highways forming
the most important determinant of a country’s connectivity score. National highway networks range
from relatively comprehensive to non-existent.
• Cross-border road connectivity between Asian countries is significantly behind that found in more
developed regions such as Europe. Only two cross-border highway connections are currently in place
(Malaysia-Singapore and China-Vietnam).
• Improving road connections between ports and industrial clusters has the potential to boost participation
in global value chains. Landlocked countries, or those with significant industrial activity in non-coastal
areas, such as Laos, Cambodia, Georgia and Bangladesh, stand to benefit significantly from reduced
travel time to ports.

The report explores the links between trade, economic development and connectivity. Income is positively
correlated with road connectivity. Seaports with surrounding road connections to industry play a more
prominent role in trade. Special Economic Zones with shorter travel times to airports and seaports also
export more. Countries with better cross-border road connections tend to trade more with each other.

Further analysis of the causal relationship between infrastructure and economic growth would provide
greater confidence and accuracy in the economic implications of improving road connectivity. This report
presents the possibilities for generating insights using new data and tools to assess connectivity. Such
insights may provide a useful future basis for policymakers in Asia to engage in infrastructure planning.
3 ||

INTRODUCTION
Transport connectivity is essential for trade in 2. Port access and connectivity: As the main
physical goods. As increasing specialisation of conduits of international trade, access to
tasks across countries creates ever-more-complex seaports is a critical component for a country’s
international supply chains, the need for seamless participation in global value chains. New
connectivity has never been greater. At the same country-level measures of port connectivity are
time, the emergence of better digital datasets constructed with a view to understanding where
on global transport infrastructure, coupled with improved road connections to ports can benefit
faster software and computing capabilities, is firms, with a special focus on India’s ports.
enabling the study of transport connectivity to an
unprecedented degree. 3. 
Special economic zones (SEZs): For many
countries, SEZs are an important policy tool
This section of the Asia Infrastructure Finance (AIF) used to accelerate industrial and infrastructure
report focuses on using geospatial datasets to build development in a targeted region. A new
a deeper understanding of transport connectivity spatial dataset on SEZs is presented and links
across Asia, particularly as it pertains to global between various forms of transport
trade. The analysis primarily aims to compile new infrastructure and SEZ exports are explored in
data and calculate new measures of connectivity, a case study on China.
focusing on the Asian Infrastructure Investment
Bank’s (AIIB) regional member countries. The analysis presented in the following sections is
A secondary objective is to explore linkages largely descriptive, seeking to paint an Asia-wide
between connectivity and trade. The emphasis is portrait of transport connectivity. The authors
on assessing: hope it provides a starting point for policymakers
to assess transport infrastructure needs, and how
1. Road connectivity: Roads remain the arteries such infrastructure may facilitate participation
of global transport, yet little is understood about in global value chains, rather than presenting
road connectivity across and within country strong recommendations for new infrastructure
borders in Asia. New measures of cross-border investment. Every country and region will have
and domestic road connectivity in Asia are economic, social, and environmental nuances that
constructed. The potential impact of new road should be considered in much greater depth before
improvements is illustrated with a case study on undertaking any individual project.
Bangladesh’s bridges.
 || 4

METHODOLOGY
Traditional measures of connectivity tend to rely to plot the most expeditious route to the desired
on density-based concepts, such as the length of destination. In this study, extensive use is made
highways per thousand square kilometres. These of fastest path algorithms1 to construct measures
measures tend to be problematic because they of connectivity.
do not measure how easy it is to get from one
place to another, which is the ultimate goal of To plot a fastest path, four elements are required:
connectivity. For example, imagine a situation an origin, a destination, a set of possible paths,
where road networks are dense on both sides of a and a set of travel speeds associated with each
river, but the absence of bridges makes traversing path (some paths may allow faster driving than
the river challenging. others). In this study, urban settlements are used
as the basic unit of origins and destinations. We
Another problem associated with density-based are thus measuring, for all countries and border
measures is that they do not take into account areas, the degree of connectivity between cities, as
demand for transport. The Gobi Desert will opposed to, inter alia, how easy it is to get around
exhibit significantly lower road density than within cities. All cities2 with a population larger than
Greater Tokyo, but of course, it is preposterous to 50,000 are covered, resulting in a total of 7,891
claim on this basis that more roads need to be built cities across Asia. The city centroids are used as
in the desert! the point of origin/destination.

Thankfully, the availability of a rich library of For the road network, OpenStreetMaps (OSM), the
geospatial data and algorithms allow for the most comprehensive open-source data currently
computation of more intelligent measures of available, is used. As open-source data, OSM
connectivity today. This section first introduces may not be as comprehensive in its coverage as
the basic geospatial tools and concepts used to other commercially available road network layers.
compile the connectivity measures presented, then However, the gaps in OSM’s coverage are largely
explains how these concepts are incorporated into confined to lower-level streets, which may be
a composite measure of connectivity. important for navigating to one’s favourite local
restaurant but are less so for driving to another
city. For the latter, OSM has more than sufficient
Computing fastest paths coverage of major roads, such as motorways,
trunks, primary and secondary roads, which
Fastest path algorithms have become a part are the road type layers used in this study.
of daily life in the 2020s. Every time someone Driving speed assumptions (see Table 1)
searches for directions on their smartphone, the are assigned to each road layer to allow for
map application will use a fastest-path algorithm differentiation between different types of road—a

1 The OD Matrix algorithm of the QNEAT3 Plugin in QGIS is used for fastest path computation in this study.
2 Pesaresi, Martino; Florczyk, Aneta; Schiavina, Marcello; Melchiorri, Michele; Maffenini, Luca (2019): GHS settlement grid, updated and refined
REGIO model 2014 in application to GHS-BUILT R2018A and GHS-POP R2019A, multitemporal (1975-1990-2000-2015), R2019A. European
Commission, Joint Research Centre
5 || METHODOLOGY

country with more highways will tend to have Figure 1: Illustration of fastest and ideal paths
better connectivity. This presents a methodological
challenge as there can be a lot of variation in
speed limits, congestion and road quality between
countries. However, as the purpose of this study is
to assess physical transport networks, the authors
chose to adopt uniform speed assumptions across
all countries to emphasise variation in the quality
and geometry of the road network itself.

Table 1: Speed assumptions by road type

Fastest travel times alone are insufficient to assess
OSM road class Speed connectivity, as this would be unfair to large
countries. In Russia, for example, it takes 29 hours
Motorway 100km/h
to drive between a random pair of cities, compared
Trunk 75km/h with 17 minutes for Kuwait. Using “ideal travel time”
as a lower bound to normalize the data makes for
Primary/secondary 65km/h
fairer comparison across countries.
Data source: Approximations based on averages taken from OSM Wiki.
Ideal travel time is constructed simply by calculating
It is true that safe maximum driving speeds the time it would take to drive at maximum speed
(and thus connectivity) depend, to some (100km/h) in a straight line between two cities.
extent, on the quality of the road surface. The The ratio of the fastest path travel time to its ideal
authors acknowledge that such nuances are not counterpart forms the basis of the connectivity
reflected in the connectivity measures presented. measure. If ideal time requires a 30-minute drive,
Accounting for such differences would make compared with one hour in actuality, then the ratio
for a vastly more complex exercise. For the fastest is 0.5. This ratio is computed for all city pairs within a
path analyses, all highways are assumed to offer country or a border area.
the same driving speeds.
Before summarising the ratios into a national
From fastest paths to network indicator, a final step is necessary. Not all city pairs
efficiency are equal in terms of travel demand. Travel demand
tends to be highest between large cities, and so
The first step in compiling the country- and border- the ratios should be weighted according to the
level measure of road connectivity is to compute population of the city pairs. This ensures that the
fastest path travel times for all possible pairs of final indicator, called the Network Efficiency Ratio
cities. This is done at two levels: national level (all (NER)3, is appropriately weighted to emphasise the
city pairs within a country) and border level, (for all more important connections. In formal terms, the NER
cross-border city pairs for cities within 200km of for country z is expressed as:
a border).

3 Gutiérrez, J., Monzón, A., & Piñero, J. M., 1998. “Accessibility, Network Efficiency, and Transport Infrastructure Planning.” Environment and Planning
A, 30(8), 1337–1350. doi:10.1068/a301337, cited by Christodoulou et al. (2019), “Road Accessibility In Border Regions.” European Commission,
Regional and Urban Policy, https://ec.europa.eu/regional_policy/sources/docgener/work/2019_01_road_access.pdf.
 METHODOLOGY || 6

where, for city pairs I and j, ATT is the actual travel Isochrones
time for the fastest path, IDT represents ideal travel
time, and POP stands for population. While a fastest path represents the quickest route
between two points, it does not help in answering
The interpretation of the NER is straightforward. questions concerning the area of coverage. The
If a country were to attain the maximum value of owner of a shopping mall, for example, may want
1, its road network would have to consist of direct to know how many people live within a 20-minute
highway connections between all cities. Naturally, drive of the mall. Likewise, an infrastructure planner
such a road network is unattainable and even may ask how many factories are located within a
undesirable—it would be absurd for every village, 1-hour drive of a new port.
town and city to have a straight highway to every
other one as the entire country would likely be Answering these questions requires computing
covered by roads. An NER of 0.5 would mean that an isochrone (see Figure 2). Isochrones can be
it would take twice as long for a randomly selected based on any time limit. A two-day isochrones, for
person in the country to travel to another randomly instance, would be extremely large and cover most
selected city compared with the ideal case. countries. A one-minute isochrone may cover a few

Figure 2: 4-hour isochrone from Shanghai port

Data source: Isochrone: HERE Routing API, Road and water: OSM, Land polygon: Natural Earth.
7 || METHODOLOGY

blocks. The road network is also an important factor an “ideal”. The ideal isochrone is essentially the
in determining the extent of an isochrone. An area two-dimensional counterpart of the ideal path—a
with extensive highways will likely have much larger circle whose radius is determined by a combination
isochrones than one with few roads. of speed and travel time. The ideal counterpart of
a 1-hour isochrone where the maximum speed is
For this study, isochrones are computed using HERE 100km/h is thus a circle with a 100km radius, often
Technologies’ Routing API. As with the fastest referred to as an “as-the-crow-flies” circle.
path analysis, a “no traffic” assumption is used to
create as controlled an environment as possible The ratio of the area of the isochrone to the
and isolate variability to the characteristics of the circle then becomes a natural measure of the
physical road network. connectivity associated with a given point, such
as a port. The measure can be given even more
meaning by including other data of interest.
Using isochrones to measure One study assesses transport performance in
connectivity Europe by measuring the ratio of the population
within the isochrone to that within the circle.
Following the spirit of using actual versus ideal travel Since this study focuses on connectivity in the
times in constructing the Network Efficiency Ratio, context of trade, factories and ports are our main
it is also possible to benchmark isochrones against variables of interest.

Table 2: Data sources for port connectivity analysis

Ports Population4 Factory

UNCTAD regularly publishes a The Global Human Settlement For India, the HERE Map India dataset
Port liner shipping connectivity Layer population grid data (2019) provides information on industrial
index (PLSCI)5 to evaluate the is an open-source population raster zones and industrial complexes,
connectivity of worldwide container layer that provides population defined as non-residential areas
ports in the global liner shipping distribution estimates in 2015 dedicated to industrial/storage
network. Geolocation information on at high geographical resolution. activities. Centroids of 19,452
495 ports in Asia collated in This layer is the basis for population industrial complex polygons are used
the PLSCI from a range of coverage. for factory counts in isochrones.
open-source port location
databases6,7,8,9 form the scope Factory data for the rest of Asia is
of the analysis. assembled from the OSM database,
consisting of 338,987 centroid
points from polygons under the
“landuse=industrial” tag.

4 Schiavina, M, Freire, S, & MacManus, K (2019) “GHS population grid multitemporal (1975-1990-2000-2015), R2019A”. European Commission,
Joint Research Centre (JRC) [Dataset] doi:10.2905/0C6B9751-A71F-4062-830B-43C9F432370F PID: http://data.europa.eu/89h/0c6b9751-
a71f-4062-830b-43c9f432370f;
5 Port liner shipping connectivity index (PLSCI), UNCTAD, https://unctadstat.unctad.org/wds/TableViewer/summary.aspx
6 International Ports, World Bank, https://datacatalog.worldbank.org/dataset/global-international-ports
7 World Port Index, National Geospatial-Intelligence Agency, https://data.humdata.org/dataset/world-port-index
8 Global Ports, World Food Programme, https://geonode.wfp.org/layers/esri_gn:geonode:wld_trs_ports_wfp
9 World Port Source, www.worldportsource.com
10 Dijkstra, L, Poelman, H & Ackermans, L (2019), “Road Transport Performance In Europe, Introducing a new accessibility framework”, European
Commission, Regional and Urban Policy, https://ec.europa.eu/regional_policy/sources/docgener/work/2019_02_road_transport.pdf
 || 8

ROAD CONNECTIVITY
IN ASIA
Roads are arguably the most fundamental excluded from the analysis, and analysed road
component of transport networks. Without an networks exclude tertiary roads and streets. Intra-
extensive supporting road network, air, sea and rail city connectivity is also excluded from the analysis.
ports would all be rendered inaccessible. Following The measures compiled thus place emphasis
the methodology outlined in the previous section, squarely on inter-city road networks.
this section presents the findings of the road
connectivity analysis across Asia, first at a national
level and then for cross-border connections. Domestic road connectivity analysis

It should be noted that the basic unit of analysis for The analysis shows a high degree of variation
connectivity in this section is an urban settlement, in road connectivity, measured according
defined as a contiguously-built urban area with to the Network Efficiency Ratio (NER), across
a minimum population of 50,000. Villages are Asia (see Table 3). The highest-scoring

Table 3: Network efficiency ratios for AIIB regional member countries

(Germany and the Netherlands added for comparison)

Expected Ideal travel Network Expected Ideal travel Network

travel time time Efficiency travel time time Efficiency
(hours) (hours) Ratio (hours) (hours) Ratio

China 14.32 11.54 0.80 Iran 10.26 6.22 0.61

Germany 3.99 3.16 0.78 Azerbaijan 2.79 1.72 0.60
South Korea 2.69 2.14 0.78 Russia 29.04 16.93 0.60
Netherlands 1.11 0.87 0.76 Sri Lanka 2.54 1.51 0.60
Saudi Arabia 9.10 7.03 0.76 Jordan 1.27 0.74 0.56
UAE 1.61 1.22 0.76 Uzbekistan 6.78 3.83 0.56
Qatar 0.56 0.42 0.75 Cambodia 4.17 2.36 0.55
Malaysia 3.17 2.35 0.74 Vietnam 12.59 6.67 0.55
Israel 0.98 0.72 0.73 Myanmar 6.94 4.00 0.53
Indonesia 4.98 3.59 0.69 Tajikistan 3.06 1.47 0.50
Egypt 3.41 2.36 0.67 Kazakhstan 22.93 10.61 0.50
Pakistan 8.24 5.58 0.66 Afghanistan 7.68 3.82 0.49
Philippines 2.24 1.42 0.65 Nepal 4.77 2.37 0.48
Georgia 3.13 2.00 0.64 Laos 8.06 3.76 0.48
Thailand 6.60 4.03 0.63 Mongolia 4.98 2.37 0.48
Turkey 8.29 5.33 0.63 Bangladesh 4.07 1.88 0.44
Oman 5.08 3.14 0.61 Kyrgyz Republic 9.87 2.32 0.25
India 16.34 10.00 0.61

Data source: Road network: OSM, Population: Global Human Settlement Layer, Calculations: EIU.
9 || ROAD CONNECTIVITY IN ASIA

Figure 3: Road Network Efficiency Ratios and GDP per capita in Asia

100
QAT
KWT
GDP per capita (thousand PPP US$), log scale

ARE
SAU
EST
GRC POL KOR
ISR
OMN LTU
KAZ TUR
RUS MYS
LVA BGR
THA BLR
ARM TKM AZE GEO EGY CHN
MNG LBN
BTN VNM LKA IRN IDN
IRQ
10 JOR
LAO UZC PHL
IND
MMR
KGZ
BGD PAK
KHM SYR
TJK
NPL

YEM AFG

1
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Network efficiency

Data source: EIU, OSM.

countries in the region are world-leading, less wastage, lower inventories and so forth, which
with a number of countries outscoring the in turn boosts productivity and GDP.
most connected economies in Europe. Using
the same methodology, the road networks of Yet, GDP per capita explains only 37% ofthe variation
Germany and the Netherlands produce scores in NERs, suggesting a diverse range of factors
of 0.78 and 0.76, respectively. that determine road connectivity. A Shapley
value regression analysis helps to shed some light
This may come as a surprise, especially to Germans on factors that influence the NER. Variables
accustomed to driving on the autobahn at speeds explored include:
up to 200km/h. It is worth remembering that a
uniform driving speed of 100km/h is imposed for • Length and density of the road network
highways in the analysis to enable international • Share of highways in road network length
comparison of physical road networks, rather than
• Terrain ruggedness, measured as variation in
speed limits.
elevation across one sq km cells

In general, Asia’s more developed economies tend • Water bodies, measured as the share of water
to exhibit a higher NER—the correlation coefficient bodies of the country’s surface area
between GDP per capita (PPP$) and national NERs • Boundary constraints, measured as the share of
is high at 0.57. The relationship between economic straight paths between cities that intersect with
development and road connectivity is complex, a land boundary (thus preventing construction
with causality running in both directions. More of straight roads)
advanced economies have more resources to invest
in building high quality infrastructure. At the same Results of the regression are shown in Table 4.
time, better road networks mean firms have better The Shapley values decompose the relative
access to internal markets, lower transport costs, importance of each variable in the overall fit of the
 ROAD CONNECTIVITY IN ASIA || 10

Table 4: Shapley value regression decomposing NER determinants

Variables coef se T pval r2 Relative Relative
importance importance
(%)

Intercept 0.460 0.045 10.250 0.000 0.777

Share of highways 0.006 0.001 4.392 0.000 0.777 0.314 40.4%
Road density 0.044 0.010 4.304 0.000 0.777 0.215 27.6%
Ruggedness -0.0003 0.000 -4.938 0.000 0.777 0.170 21.9%
Water bodies -0.006 0.002 -2.455 0.018 0.777 0.060 7.8%
Boundary constraints -0.001 0.000 -1.819 0.076 0.777 0.018 2.3%

Data source: Road network: OSM, Ruggedness: Nunn, Nathan, and Diego Puga, Water bodies: OECD, Boundary constraints: Natural Earth, GHS urban
centre layer, Calculations: EIU.

Table 5: Ranking of Asian countries by network efficiency and associated factors

Road network Terrain and borders

Country Network Share of Road Rugged-ness Water Boundary

efficiency ratio highways density bodies constraints

Relative indicator importance* 40% 28% 22% 8% 2%

China 1 2 7 20 19 13
South Korea 2 4 1 22 20 16
Saudi Arabia 3 6 21 9 1 6
UAE 4 3 5 8 6 28
Qatar 5 1 3 1 9 33
Malaysia 6 5 13 12 12 14
Israel 7 7 2 19 21 20
Indonesia 8 14 20 11 16 15
Egypt 9 8 23 7 10 4
Pakistan 10 10 17 21 27 10
Philippines 11 22 9 24 18 26
Georgia 12 11 15 30 4 5
Thailand 13 23 11 14 28 29
Turkey 14 12 10 29 17 18
Oman 15 9 22 17 2 7
India 16 25 8 13 26 11
Iran 17 19 16 26 13 8
Azerbaijan 18 13 12 18 15 9
Russia 19 24 31 10 24 25
Sri Lanka 20 15 4 6 22 22
Jordan 21 20 18 16 3 1
Uzbekistan 22 27 27 4 30 30
Cambodia 23 27 26 5 32 12
Vietnam 24 17 6 25 31 32
Myanmar 25 16 25 23 23 19
Tajikistan 26 27 24 33 14 27
Kazakhstan 27 18 32 3 29 17
Afghanistan 28 27 30 27 7 3
Nepal 29 27 19 32 5 21
Laos 30 21 28 28 11 31
Mongolia 31 26 33 15 8 1
Bangladesh 32 27 14 2 33 23
Kyrgyzstan 33 27 29 31 25 24

Data source: Road network: OSM, Ruggedness: Nunn, Nathan, and Diego Puga, Water bodies: OECD, Boundary constraints: Natural Earth, GHS urban
centre layer, Calculations: EIU.
11 || ROAD CONNECTIVITY IN ASIA

model (R squared). This is done by taking cross Laos. A south-bound road from Bangkok
the average R-squared contribution of the variable would need to curve to fit the narrow contours of
in all possible variable combination specifications peninsular southern Thailand.
(shown in the final column). The coefficient estimates
and associated diagnostic statistics shown are Finally, it is noteworthy that there are no apparent
for the all-variable regression specification. The (dis)economies of scale in the NER. Larger or
model’s high R-squared indicates that a substantial smaller road networks, measured either by total
amount of the variation in the NER is captured by road network length or number of cities, play an
the five variables. unimportant role in driving variation. It can thus be
said that the NER is a “size agnostic” measure.
Highways stand out as the dominating
factor driving variation in the NER, given the
speed improvements enabled by traffic-light-
free, multi-lane carriageways. It is notable that
increasing the share of highways in the road network
has more impact on the NER than increasing the
length of the network. Generally, countries gain
more connectivity from upgrading roads than
building new ones, though there are exceptions,
particularly when important links such as bridges
are missing.

Terrain clearly also matters. In particular, ruggedness

explains 22% of the (explained) variation in the
NER. Rugged terrain substantially increases the
costs of road development and maintenance and
severely limits routing options, leading to inefficient
networks. Nepal, Kyrgyzstan and Tajikistan are
among the world’s most mountainous countries and
hence show low connectivity scores.

Table 5 shows the ranking of countries for the

variables included in the regression, which helps to
understand the relative performance of countries.
China performs well largely due to its extensive
highway network, second only to Qatar (which also
happens to be the flattest country studied).

Interestingly, boundary constraints are a

statistically significant variable in the regression but
do not contribute much to NER variation. Countries
where the spatial pattern of urban settlements
does not permit for direct, straight roads suffer a
mild penalty in the NER. For instance, a straight line
between Hanoi and Ho Chi Minh City, Vietnam’s
two largest metropolitan areas, would have to
 ROAD CONNECTIVITY IN ASIA || 12

Case study – Bangladesh’s bridge problem

Bangladesh stands out as a country with Figure 4: Bangladesh’s lack of
particularly challenging natural conditions downstream bridges
for road connectivity. The densely-populated
country’s land area (see Figure 4) is split into
three sections by the mighty Padma (Ganges)
and Jamuna (Brahmaputra), the world’s 3rd
and 9th largest rivers in terms of discharge. At
the time of writing, only two bridges spanning
these rivers were in operation: Lalon Shah and
Jamuna. No bridges are currently in operation
on the lower stretch of the two rivers after
they join. However, the Padma Bridge, under
construction, aims to address this issue.
Naturally, the shortage of bridges creates a
large drag on Bangladesh’s NER. Data source: Country polygon: Natural Earth, Rivers: OSM.

Where would a new bridge create the largest boost to Bangladesh’s NER? To answer this question,
263 hypothetical bridges are generated across pairs of all 47 riverside major urban centres. The
impact of each of these hypothetical bridges on Bangladesh’s NER is computed to identify the top
efficiency-improving new links.

Unsurprisingly, new connections on the lower stretch of the Padma yield the most significant
improvement to the country’s NER. While it is currently possible to cross the lower Padma by ferry, the
boat trip adds up to 3 hours to the journey. By road, the river crossing would be achievable in under 15
minutes, making it clear why the upcoming Padma Bridge is sorely needed. When in operation, the new
bridge will shorten the travel time of 8,500 out of 37,000 inter-city paths in Bangladesh and increase
the country’s NER from 0.44 to 0.50, boosting Bangladesh by five places in the NER rankings.

This case study illustrates how singular improvements in a country’s road network can bring
large improvements in connectivity on a nationally significant scale, with evident implications for
economic efficiency.

Figure 5: Simulating new connections in Bangladesh

Connectivity
Bad Padma Bridge

Good

Data source: Country polygon: Natural Earth, Rivers: OSM, Cities: Global Human Settlements Layer, Calculations: EIU.
13 || ROAD CONNECTIVITY IN ASIA

Cross-border road connectivity analysis The overall correlation coefficient appears relatively
weak at 0.19 due to a number of outliers in the
The analysis points to significant room for Middle East and Central Asia. Countries whose
improvement in cross-border connectivity in the exports are predominantly energy products, such as
region. Cross-border NERs are notably lower than Saudi Arabia and Qatar, may not trade much with
the domestic NERs presented in the previous each other even if cross-border road connections
section. The average cross-border NER is 0.51, are good. But there is a much tighter relationship
compared with 0.61 for the domestic measure. between a subset of countries towards the top of
Moreover, unlike domestic connectivity where the the chart. Interestingly, the two country pairs with
best-connected countries in Asia outperformed the most trade are also the countries with cross-
their advanced economy peers, the highest-scoring border highway links in place.
borders in Asia fall short of those in Europe. The
NER for the Netherlands-Germany border is 0.79, The relationship between cross-border road
compared with 0.75 for the Jordan-Syria border connectivity and trade may be direct (via overland
(see Table 6). transport) or indirect, exerting itself through other
means not immediately observed. Internationally
Why does Asia underperform on cross-border comparable statistics on global trade by mode of
connectivity? A quick examination of the features transport remain hard to find, though it is clear that
of the continent’s physical road network is revealing. shipping via sea remains the preferred mode of
While most countries in the region are in the early- transport for international trade. In the European
to-mid stages of building national highway networks Union (EU), seaborne transport accounted for 47%
(some, such as South Korea, China and Japan are and 70% of extra-EU trade, in value and volume
already quite advanced), complete cross-border terms respectively, in 2018, according to the
highways links are non-existent except for in the European Commission. The same figures for road
cases of Malaysia-Singapore and China-Vietnam transport are 20% and 3%.
(near complete). There are also countries that lack
cross-border road connections altogether. There are The economics of transport change significantly,
no mapped roads between Myanmar and Bangladesh. however, as distances get shorter and border
procedures are simplified. Overland transport
Similar to the domestic connectivity analysis, offers more flexibility and shorter delivery times
developed economies and those in the Middle East than seaborne shipping, which may outweigh the
(thanks partly to more accommodating terrain) higher costs. For intra-EU freight that requires no
score relatively well in cross-border network customs clearance, road transport accounted for
efficiency. South and Southeast Asian countries 51% of freight transport in kilometre tonnes in
tend to score lower. 2018. Better road connectivity can thus reduce
transport time and costs and be a powerful catalyst
for trade, especially when supported by streamlined
Border road connectivity and trade border processes.

To understand the link between road connectivity and The extent to which cross-border trade in Asia
cross-border trade, NERs are plotted against bilateral takes place via overland transport is not entirely
total trade data from UN Comtrade. The correlation clear. Given the presence of border customs
shown is partial as it is necessary first to control for processes and lower levels of cross-border road
the economic size of countries (larger economies tend connectivity, trucking should be a less popular
to trade more with each other). The Y axis thus shows option in the region compared with the EU. The
the residual from the regression of total bilateral trade correlation shown in Figure 6 is striking.
on the product of country pair GDP.
 ROAD CONNECTIVITY IN ASIA || 14

Table 6: Cross-border Network Efficiency Ratios in Asia

Expected Ideal travel Network Expected Ideal travel Network

travel time time Efficiency travel time time Efficiency
(hours) (hours) Ratio (hours) (hours) Ratio

DEU-NLD 2.86 2.28 0.79 IRN-TUR 5.71 2.85 0.50

JOR-SYR 2.12 1.60 0.75 AZE-RUS 6.96 3.60 0.49
ISR-LBN 2.77 2.03 0.73 ARM-AZE 4.90 2.40 0.49
ARE-OMN 2.74 1.96 0.71 CHN-RUS 10.24 5.20 0.48
KWT-SAU 3.18 2.19 0.69 AFG-UZB 5.76 2.82 0.48
MYS-SGP 1.08 0.78 0.68 MMR-THA 11.13 5.73 0.48
LVA-RUS 3.74 2.46 0.66 KHM-THA 5.05 2.52 0.48
EST-RUS 4.72 3.07 0.65 AFG-TKM 5.77 2.51 0.47
CHN-VNM 5.65 3.66 0.64 IRN-TKM 6.73 3.31 0.47
MYS-THA 3.88 2.47 0.63 GEO-RUS 7.10 3.46 0.47
BLR-RUS 4.84 3.04 0.62 ARM-GEO 4.06 1.91 0.47
LTU-RUS 3.81 2.45 0.62 GEO-TUR 6.50 3.19 0.47
JOR-SAU 2.62 1.60 0.61 CHN-MMR 7.61 3.54 0.47
KAZ-RUS 19.93 12.00 0.60 BGD-IND 6.90 3.21 0.46
KHM-VNM 4.30 2.48 0.60 BTN-IND 4.41 2.00 0.45
AZE-GEO 3.75 2.21 0.59 AFG-TJK 5.52 2.54 0.45
BGR-TUR 4.93 3.05 0.59 CHN-KGZ 9.19 4.36 0.45
TKM-UZB 7.17 4.42 0.58 LAO-THA 7.85 3.37 0.44
IRQ-TUR 2.91 1.72 0.57 CHN-NPL 9.47 4.19 0.43
QAT-SAU 3.32 1.87 0.56 CHN-IND 14.81 6.49 0.42
AFG-IRN 6.72 3.71 0.56 EGY-ISR 3.52 1.52 0.42
SYR-TUR 4.96 2.85 0.54 IND-PAK 13.00 6.19 0.42
KGZ-UZB 3.90 2.15 0.54 AZE-TUR 4.50 1.90 0.42
ARM-TUR 3.24 1.72 0.53 CHN-KAZ 10.03 4.22 0.42
IND-NPL 6.49 3.56 0.53 IDN-MYS 7.80 3.01 0.41
KAZ-KGZ 5.56 2.86 0.52 CHN-LAO 9.46 3.88 0.41
KAZ-UZB 6.46 3.04 0.52 IRN-PAK 9.47 3.89 0.41
AZE-IRN 6.97 3.68 0.52 LAO-VNM 12.17 5.19 0.41
GRC-TUR 4.65 2.62 0.52 ISR-JOR 2.75 1.14 0.41
POL-RUS 3.35 1.77 0.51 KGZ-TJK 10.17 3.78 0.34
IRN-IRQ 8.25 4.38 0.51 MNG-RUS 12.03 4.09 0.34
TJK-UZB 5.32 2.71 0.51 IND-MMR 14.33 4.40 0.30
AFG-PAK 6.83 3.53 0.51 KHM-LAO 7.44 1.85 0.25
SAU-YEM 5.76 3.01 0.51 IRQ-SAU 14.64 4.34 0.23
BRN-MYS 3.51 1.77 0.50 BGD-MMR - - -

Data source: Road network: OSM, Population: Global Human Settlement Layer, Calculations: EIU.
15 || ROAD CONNECTIVITY IN ASIA

Figure 6: Bilateral trade and network efficiency in Asia

4.5 CHN-VNM MYS-SGP

4.0 BLR-RUS
3.5
Log of total bilateral trade, controlling for trade gravity

CHN-KAZ CHN-MMR POL-RUS KAZ-RUS MYS-THA

3.0 IDN-MYS IRQ-TUR
2.5 CHN-KGZ KHM-THA MMR-THA LTU-RUS ARE-OMN
2.0 LAO-THA IND-NPL BGR-TUR LVA-RUS
AZE-TUR BGD-IND CHN-RUS
1.5 CHN-LAO AZE-RUS
GRC-TUR
KAZ-UZB KHM-VNM EST-RUS
1.0 MNG-RUS
GEO-TUR IRN-TUR SUR-TUR JOR-SAU KWT-SAU
LAO-VNM
0.5 CHN-NPL AFG-PAK
AZE-GEO
IND-MNR GEO-RUS AFG-PAK
0 IND-PAK
BRN-MYS
IRQ-SAU BTN-IND
-0.5 ARM-GEO
KAZ-UZB
-1.0 ISR-JOR AFG-UZB SAU-YEM
-1.5 EGY-ISR TJK-UZB
-2.0 CHN-IND
-2.5 JOR-SYR
TKM-UZB
-3.0 KGZ-TJK AZE-IRN
-3.5 KHM-LAO
-4.0 AFG-IRN
-4.5
AFG-TJK
-5.0 IRN-PAK ARM-AZE
-5.5
-6.0 ARM-TUR
-6.5 AFG-TKM
QAT-SAU
-14.0
0.22 0.32 0.42 0.52 0.62 0.72 0.82
Network efficiency

Data source: Network efficiency and GDP: EIU, Trade: UN Comtrade (residual from regression of bilateral trade on the product of GDP
of country pairs, Calculations: EIU.

Overland transport may have a larger role to

play than previously expected. Another plausible
explanation is that countries with better economic
ties are more proactive in promoting bilateral trade,
including investing in cross-border infrastructure
links while simultaneously lowering other barriers
to trade. Better road connectivity can also lead
to greater mobility for people, which can also
enhance trade.

Clearly, the relationship between road connectivity

and trade is multi-faceted. Building more cross-
border highways alone may not guarantee
more trade and prosperity. But it is part of a
larger package that countries pursuing greater
participation in global value chains should consider.
 Road connectivity in Asia

|| 16

PORT CONNECTIVITY
IN ASIA
With maritime trade accounting for the bulk and time. But “close” need not refer to physical
of global merchandise trade, seaports are the distance. Firms can enjoy significantly lower travel
gateways of global value chain participation for times to a port that is well-connected by a highway
most firms. Export/import firms tend to be located network compared with one that lacks proper road
close to a seaport to minimise transport costs connectivity, even if they are further away.

Figure 7: Isochrones of liner ports in Asia

Data source: Isochrones: HERE Routing API, Land polygons: Natural Earth.
Note: All isochrones shown in this section of the report have been generated from the same data source.
17 || PORT CONNECTIVITY IN ASIA

Industry access to ports poles will appear larger than those on the equator.
For instance, in reality, Russia, occupies roughly
To understand the access that firms across Asia half the surface area that Africa does, though it
have to seaports, the first step is to generate appears larger.
isochrones (see methodology section for detail) for
Asia’s ports. This is done at two-hour intervals up to Even despite the distortions, the benefit of an
8 hours, which represents a full day’s drive time.11 advanced highway network is visible from the map.
A total of 324 liner shipping ports in AIIB regional China’s port isochrones are visibly larger than in
member countries and other countries of interest other countries at similar latitudes. China also
are included, covering well over 90% of container benefits from freshwater shipping on the Yangtze,
traffic in the region. Figure 7 shows the isochrones the only river in Asia to have a liner shipping port,
for these ports across the region. located in the city of Wuhan.

The isochrones effectively capture the extent Having large isochrones, however, does not
of road connectivity in the vicinity of ports. automatically lead to high port accessibility for
In general, the larger the isochrone, the more firms. To measure the latter, it is first necessary
extensive the road network will be. However, two- to account for the location of factories and then
dimensional projections of the world map are to establish an upper bound for the maximum
often poor representations of reality. Pressing achievable degree of connectivity. The latter is
the globe onto a flat surface causes distortions— achieved by means of an “ideal isochrone” (see
the same surface areas and lengths closer to the methodology section). A 4-hour isochrone is used

Figure 8: 4-hour and ideal isochrones for ports in Southeast Asia

Data source: Isochrone: HERE Routing API, Land polygons: Natural Earth.

11 To ensure international comparability, the isochrones are generated under “no traffic” assumptions.
 PORT CONNECTIVITY IN ASIA || 18

Figure 9: Current and Potential Port-Factory Connectivity Scores by Country

Small/Island Landlocked Coastal

100%
Potential
90% Current

80%
70%
% of factories

60%
50%
40%
30%
20%
10%
0%
BHR
MYS
PHL
FJI
LKA
IDN
SGP
BRN
QAT
ISR
KAZ
KGZ
MNG
NPL
TJK
UZB
LAO
AFG
AZE
JOR
KOR
ARE
VNM
THA
TUR
BGR
EGY
CHN
MMR
OMN
GEO
KHM
SAU
IND
PAK
BGD
IRN
RUS
Data source: Isochrone: HERE Routing API, Factory count: OSM, Calculations: EIU.

as the main benchmark in this section (as shown densest highway network in continental Asia, also
in the next section, 4 hours correlates well with performs well. Of the coastal countries, Bangladesh
exports). The ideal isochrone is thus 400km in has the least-connected port areas, partly owing
radius, given a maximum driving speed of 100km/h. to challenging river delta terrain that complicates
road-building. Landlocked countries fall outside of
Next, location data covering 338,987 industrial any 4-hour port isochrone and hence score zero.
complexes across Asia is used to measure factory
locations. Comparing the number of factories Turning to potential improvements, the countries
located within the 4-hour isochrone to the number on the right of the chart have the greatest
of factories within the ideal isochrone by means opportunity to increase port access to industry,
of a simple ratio then provides a Port-Factory in particular, Bangladesh, Cambodia and Laos. For
Connectivity Score (PFCS). The intuition is the former two, over 50% of industrial facilities in
straightforward—a score of 1 means all factories the respective country can potentially be brought
within 400km can be reached within 4 hours, while within the 4-hour port isochrone. In Bangladesh, it
a score of 0.5 suggests only half can. currently takes nearly six hours to make the 250km
drive from Dhaka, the national capital around which
Figure 9 illustrates the PFCS for countries in Asia. much of the country’s manufacturing firms are
Small or island countries are grouped separately clustered, to Chattogram, which hosts the closest
given that their natural geography will tend to boost port. Likewise, Phnom Penh, the Cambodian capital,
the PFCS. Similarly among the coastal countries, is a four-and-half hour drive from Sihanoukville,
many perform well as a result of geographic the main coastal port city.
conditions—a relatively small land area combined
with extensive coastlines. This is the case for nearly While Laos is the country that stands to gain the
all countries with a PFCS higher than 0.85. most from road-to-port improvements, it faces a
more complicated situation as a landlocked country.
Of the larger coastal countries, Thailand scores While Vientiane, the capital, lies on the Mekong,
the highest, thanks largely to a combination of the Southeast Asian river does not accommodate
highway connectivity and the bulk of Thailand’s liner shipping. Currently, drive times to the closest
manufacturing industry being located in the coastal ports in Vietnam and Thailand are in excess
Bangkok-Laem Chabang area. China, with the of 10 hours. With better road links, this time can
19 || ROAD CONNECTIVITY IN ASIA

potentially be cut in half. Better roads to port can agreements with neighbouring countries are a
bring multiple benefits, including lower inventory common instrument for facilitating port access
costs, greater reliability in deliveries and lower for landlocked countries, the implementation of
likelihoods of theft, spoilage and road accidents. such agreements has seen mixed results. Carriers
in landlocked countries still face discrimination
It should be noted, however, that even with better and are subject to a variety of restrictions on the
cross-border hard infrastructure, multiple soft “number of trips, cargo volumes, carrying capacity
impediments remain. While cross-border transit or numbers of permits”.12

Case Study - Why port connectivity matters for India’s exports

Building on port-level export data in India, this section illustrates how improving road connectivity
to ports can potentially boost participation in global value chains. Figure 10 shows 2-hour to 8-hour
isochrones for India’s 20 largest ports in terms of export value in 2019. As introduced previously, an
isochrone13 is a geometric shape that maps temporal boundaries of possible travel paths.. The better
connected a port is in terms of the number and quality of road connections, the larger its associated
isochrones will be. Mumbai port, for instance, boasts relatively large isochrones compared with
Visakhapatnam’s, which is located in a more sparsely-populated region with fewer roads.

As Figure 11 shows, port-level exports are highly correlated with the number of factories14 located
within their 2- and 4- hour isochrones, which is intuitive—the more factories located near the port,
the more the facility exports. The correlation falls as travel time to the port increases beyond 4
hours since factories located further from the port are less likely to be engaging in export trade. 4
hours is thus used as the benchmark for the remaining analysis in this section.

Figure 10: Isochrones of India’s top 20 ports by export value in 2019

Data source: Isochrone: HERE Routing API, Factory count: OSM, Land polygon: Natural Earth, Calculations: EIU.

12 Chapter 6, Review of Maritime Trade 2013, UNCTAD

13 To ensure international comparability, the isochrones are generated under “no traffic” assumptions
14 Defined as the number of “industrial complexes”.
 ROAD CONNECTIVITY IN ASIA || 20

Figure 11: Correlation of port exports in 2019 and factory count by isochrone

0.6
0.53 0.52
0.5
Correlation of coefficient

0.4 0.36

0.29
0.3

0.2

0.1

0.0
2 hours 4 hours 6 hours 8 hours
Isochrone

Data source: Isochrone: HERE Routing API, Factories: HERE, Calculations: EIU.

Figure 12 shows the Port-Factory Connectivity Scores for each of India’s top 20 ports based on
the measure described above. India’s best-connected Port is Kolkata, with a ratio of 44% (in other
words 44% of factories within 400km can be reached in 4 hours), Mumbai’s two ports (Nhava
Sheva and Mumbai Port) follow. A relatively high-quality road network around Mumbai (the city is
linked by a highway to nearby Pune), coupled with clustering of industries near or in India’s economic
capital, account for this high performance.

Figure 12: 4-hour Port-Factory Connectivity by major port

India Asia and Middle East
Kolkata 44 Singapore 99
Mumbai 43 Laem Chabang 94
Nhava Sheva 42 Tanjung Priok 89
Cochin 33 Port Klang 88
Sikka 32 Manila 84
Kakinada 31 Colombo 83
Vadinar 29 Saigon 82
Visakhapatnam 29 Tokyo 81
Kandla 26 Busan 74
Tuticorin 25 Port Said 65
Chennai 21 Shanghai 55
Kamarajar 21 Dubai 54
Kattupalli 21 Salalah 54
Paradip 20 Jeddah 48
Hazira 18 Nhava Sheva 44
Krishnapatnam 14
Pipavav 13
Mundra 11
Mormugao 10
New Mangalore 5
0 10 20 30 40 50 0 20 40 60 80 100

Connectivity ratio (%) Connectivity ratio (%)

Data source: Isochrones: HERE Routing API, Factory count in India chart: HERE, Factory data in Asia and Middle East chart:
OSM, Calculations: EIU.
Note: The minor difference in the ratio for Nhava Sheva Port between the India and Asia charts is due to the use of different datasets.
21 || ROAD CONNECTIVITY IN ASIA

Figure 13: 4-hour isochrones for select ports

30,859 sq km 37,923 sq km 80,521 sq km

Data source: HERE

However, there is significant room for improvement. The right side chart in Figure 12 shows the
connectivity ratios of Asia’s “prime” ports, defined as the busiest ports in each Asian economy,
selected from a list of the world’s top 50 ports ranked by container traffic. Nhava Sheva, India’s
busiest port, significantly lags behind its peers in terms of connectivity.

To be fair, not all port isochrones are comparable. In countries less expansive than India, many ports,
such as Singapore’s or Malaysia’s Port Klang, will have isochrones that are constrained by land area,
resulting in factories being located closer to the port and therefore a higher connectivity ratio. A
more suitable comparison would be the port of Shanghai or Thailand’s Laem Chabang, which do not
face such land constraints. For these ports, the isochrones are shown in Figure 13. The difference
in the extent of the isochrones areas is stark, with Shanghai’s over twice as large as the other
two, illustrating the impact an extensive highway network can have. There are 144km of highway
per thousand square kilometres in Shanghai’s isochrone, compared with 26km and 8km for Laem
Chabang and Nhava Sheva, respectively.

Figure 14: Industrial clusters and port connectivity in Mumbai and Mangalore

Data source: Isochrone: HERE Routing API, Factory count: OSM, Land polygon: Natural Earth, Calculations: EIU.
 ROAD CONNECTIVITY IN ASIA || 22

Figure 15: Manufacturing exports and port connectivity ratios

100,000
Manufacturing Exports 2019 (US$ m), log sclae

Nhava Sheva

Mundra Chennai
10,000 Sikka
Vadinar
Hazira Mumbai
New Mangalore Kattupalli Visakhapatnam
Tuticorin Kolkata
Pipavav
Kakinada Cochin
Krishnapatnam Kamarajar

1,000
Mormugao Paradip
Kandla

100
0 5 10 15 20 25 30 35 40 45
Connectivity ratio (4 hours, %)

Data source: Trade data: India Ministry of Commerce and Industry, Connectivity ratios: EIU (based on HERE isochrones and factory counts).

India’s least-connected of the top 20 ports, New Mangalore, only has a ratio of 5%. This is due to
the fact that the bulk of industry within 400km of Mangalore is clustered around Bengaluru, which
lies beyond the 4-hour isochrone (see Figure 14). Currently, it takes over 7 hours to reach Bengaluru
from Mangalore by car, covering a distance of approximately 350km. With a high-speed highway
link, it would be possible to nearly halve the drive time. Road trips from Paris to Brussels or from
Washington DC to New York, which span a similar distance, can both be done in under 4 hours.

There may also be deeper reasons behind why New Mangalore port lacks connectivity to
nearby industrial clusters. The port’s exports consist largely of commodities such as petroleum
products, iron ore, coffee and cashews. Bangalore’s high-tech orientation may be better
suited to air shipments. The connectivity measure may not capture all the historical and
economic complexities of the region which should be factored into any conclusions drawn.
It merely serves as a starting point for analysis. How does the road connectivity measure
fare in predicting port exports? The correlation between manufacturing exports and the connectivity
ratio for the top 20 ports is 0.39 (Figure 15). A notable outlier is Mundra, India’s second-largest
port by export value, which exports far more than its connectivity score would suggest. This is due
to the port’s status as the export hub for India’s industrial hinterland surrounding Delhi in the North.
Removing Mundra increases the correlation to 0.5.

It is important to note that correlation is not causation. Indeed, the direction of

causality can run both ways. Firms may be choosing to locate in better-connected areas, or the
Indian government may be prioritising road improvements in areas with more factories. Most likely,
both statements are true. What is clear is that there is ample room for improvement in India’s
road network. Given the benefits associated with tapping into global trade networks, as outlined in
Chapter 1, port connectivity may be a good place to start.
 Special Economic Zones in Asia

23 ||

SPECIAL ECONOMIC
ZONES IN ASIA
To understand the role of Special Economic Zones Currently, the countries with the largest number of
(SEZs) and infrastructure in advancing economic SEZs in Asia are the Philippines, India, and China,
growth in Asia, the EIU assembled a spatial dataset the latter being the lead adopter in terms of total
of 2,655 SEZs. The dataset included attributes number of SEZs to date. Established initially in
such as SEZ location (centroids), area, year of the late 1980s, China’s first SEZs functioned as
establishment, and sectoral classification. experimental zones for economic reform, notably
for attracting foreign direct investment (FDI). The
Asia is home to three-quarters of the SEZs success of the zones led to wider adoption across
in the world.15 The first SEZs in Asia were set the country, notably in 1992, the year of a major
up in the early 1960s in India, followed by the reform push to open a broad swath of coastal areas
Philippines and China in the 1970s and 1980s, to foreign trade and investment.
respectively. These were intended to serve as
focal points to spur economic growth, supported Early SEZs tended to be limited to coastal countries
by quality infrastructure and a variety of policy (see Figure 17). The emphasis on attracting FDI
incentives. The 1990s and 2000s, represented the and (export) manufacturing made port access an
boom years for SEZ establishment (see Figure 16), important factor in determining SEZ locations.
led by China and India. More recently, SEZs have increasingly been located
in inland regions and countries (see Figure 18).

Figure 16: SEZs by year of establishment

80 CHN
Number of new SEZs (5-year moving average)

IND
70 PHL
Rest of Asia
60

50

40

30

20

10

0
1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020

Data source: EIU

15 Chapter 4, World Investment Report 2019, UNCTAD.

 SPECIAL ECONOMIC ZONES IN ASIA || 24

Figure 17: Share of new SEZs within 100km of coastline

Share of coastal (5-year moving average) 80

70

60

50

40

30

20

10

0
1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020

Data source: EIU

Figure 18: SEZs locations in Asia

Data source: EIU, Land polygon: Natural Earth.

Two parallel trends have accompanied and such as software development and call centres
perhaps even supported this shift. The first is the focused on telemarketing and customer service,
increasing services as a sector (SaaS) focus of SEZ have proliferated. In this context, it is noteworthy
development. An emphasis on services allows for that many “micro SEZs”, often housed in a single
greater flexibility in location choices as proximity building, have emerged, primarily located in the
to ports becomes less critical. The second is the Philippines. These micro SEZs in the Philippines
Information and Communications Technology exported IT solutions and continued growing in
(ICT) manufacturing boom. ICT products have number from the late 90s up to the mid-2010s.
accounted for the bulk of growth in global trade in
the past decade. High-value-added components SEZ adoption in landlocked countries such as
of ICT products, such as microchips, are often Kazakhstan and Uzbekistan has also increased.
transported via air, which again decreases the The trend started in Kazakhstan in the early 2000s
relative importance of seaports. and since then SEZs in Kazakhstan have focused
on a variety of industries ranging from textiles to
The first trend is most prominent in countries such mining, construction and manufacturing, among
as India and the Philippines, where services SEZs, others. Uzbekistan’s experience with SEZs began in
25 || SPECIAL ECONOMIC ZONES IN ASIA

Figure 19: Share of new SEZs by sector across Asia

80 Manufacturing
Share of new SEZs (5-year moving average)
Services
70 Mixed
60

50

40

30

20

10

0
1990 1995 2000 2005 2010 2015 2020

Data source: EIU

the latter half of 2000s but the country has since These inland cities were able to attract sizable
built a significant number of SEZs (with 63 SEZs investments in ICT export manufacturing by
currently operational in the country). offering access to good infrastructure links and a
large, flexible workforce to accommodate volatility
In China, which hosts the lion’s share of Asia’s in labour demand that is typical in manufacturing of
SEZs, the shift inland came as part of a national electronics products. In particular, the high-value,
strategy emerging in the mid-2000s to spread low-bulk nature of ICT components enabled firms in
growth from wealthy coastal regions to less the sector to reduce reliance on maritime transport,
developed regions. Substantial investments in enabling inland cities to increase participation in
transport infrastructure, including highways, high- global value chains.
speed rail and airports were made to this end, which
dovetailed well with manufacturers in coastal areas Countries across Asia witnessed a surge in services
seeking lower-cost production sites. A number SEZs beginning in the late 1990s, with China,
of SEZs can be found in inland cities such as India and the Philippines leading the way (as seen
Chongqing, Chengdu and Zhengzhou. in Figure 20). In the beginning, China focused on

Figure 20: Share of new services SEZs by country

90 CHN
Country of new SEZs (5-year moving average)

IND
80
PHL
70 Rest of Asia
60
50
40
30
20
10
0
1990 1995 2000 2005 2010 2015 2020

Data source: EIU

 SPECIAL ECONOMIC ZONES IN ASIA || 26

Figure 21: Sector contributing to the service sector boom16

100 Business activities, services

90 Financial and insurance activities
Hotels and tourism accommodation
80 Information and communication
70 Real estate
Number of new SEZs

Software and IT services

60 Trade
50 Transportation, storage and communications
40
30
20
10
0
1980-1989 1990-1999 2000-2009 2010-2019

Data source: EIU

establishing a variety of economic and technical Similarly in the Philippines, the Special Economic
development zones along its coast and later shifted Zone Act of 1995 enabled private sector
towards technology-intensive industries, setting up investments in SEZ developments and the number
high-tech industrial zones in inland suburban cities of SEZs more than doubled between 1995 and
to utilise growing technical capacity and R&D. 1999. As the potential of the IT industry began to be
recognised elsewhere during this period, numerous
The liberalisation of the Indian economy in the IT parks and centres mushroomed in Manila.
1990s enabled private sector participation and
expansion of SEZs from agricultural sectors to Figure 21 illustrates the different sectors
services and manufacturing. Soon after, there was contributing to the post-1990s services SEZ
a surge in new SEZs, especially in services, as they boom in Asia. The ICT sector remained the highest
required less land compared with manufacturing contributor to this services boom until the late
SEZs and therefore had a lower chance of being 2000s. The countries where ICT industries were
disrupted by land acquisition issues. particularly flourishing in Asia are highlighted in

Figure 22: Country share of new SEZs in ICT industries

100% CHN 100% CHN
IND IND
90% PHL
90% MYS
80% TUR 80% PHL
RUS
70% 70% THA
60% 60%
50% 50%
40% 40%
30% 30%
20% 20%
10% 10%
0% 0%
1980-1989 1990-1999 2000-2009 2010-2019 1980-1989 1990-1999 2000-2009 2010-2019

ICT services industries ICT manufacturing industries

Data source: EIU

Note: An SEZ may be counted more than once if it is multisectoral.

16 An SEZ may be counted more than once if it is multisectoral.

27 || SPECIAL ECONOMIC ZONES IN ASIA

Figure 23: Country share of textiles, clothing and leather SEZs

100% BGD
CHN
90% IND
80% KHM
VNM
70% UZB
Country share

60%
50%
40%
30%
20%
10%
0%
1980-1989 1990-1999 2000-2009 2010-2019
Textiles, clothing and leather

Data source: EIU

Figure 22. While the contributing shares of China, case of Bangladesh and Vietnam. Bangladesh’s
India and the Philippines were high both in ICT burgeoning textile industry has been an important
services as well as ICT manufacturing industries, manufacturing growth driver in recent years.
countries like Thailand have significantly expanded Uzbekistan has also focused on its textile industry,
their ICT manufacturing offerings. which now contributes to a fifth of its GDP and
employs a third of all workers involved in industrial
As countries like China focused on climbing the jobs. Figure 23 illustrates the reduction in China’s
tech ladder, a number of countries in the region SEZ share of low-technology industries and the
stepped in hoping to capture opportunities created upswing in share of countries such as Bangladesh,
by lower-tech investment migrating out of China. Uzbekistan and Vietnam.
This transition is especially noticeable in the
 SPECIAL ECONOMIC ZONES IN ASIA || 28

Case Study – China’s SEZ export performance and transport infrastructure

With the highest SEZ count in Asia and a high degree of participation in global value
chains, China provides a rich data sample that enables deeper geospatial analysis of
the proximity effects of transport infrastructure on SEZ performance. The analysis draws on
2019 trade data in SEZs reported by the General Administration of Customs of China. 255
SEZs, accounting for approximately 30% of the country’s international trade in 2019, are geocoded
for analysis.

Figure 24: Distribution of SEZs in China

(size of bubble represents export value in 2019)

Data source: Export data: General Administration of Customs of China, Land polygon: Natural Earth.

Figure 24 shows the spatial distribution of SEZs in China, with circle sizes corresponding to 2019
export value. The largest concentrations of exports and SEZs are in the well-known manufacturing
clusters of the Pearl River and Yangtze River Deltas, centred in Guangdong province and around
Shanghai. A third, inland cluster can be found at the twin inland cities of Chongqing and Chengdu.

To perform a simple regression analysis, variables for the three main types of ports (air, sea, rail)
are constructed using two different spatial specifications. The first measures the drive time to the
nearest port, while the second counts the number of ports located within a one-hour isochrone. A
fourth infrastructure variable in the form of highway density within a 100km radius is computed.
Control variables include the age of the SEZ, the land area and provincial dummies.

Given the strong apparent effect of regional industrial clusters, shown in Figure 24, a
final variable is introduced to the model, namely, the number of other SEZs within a
100km radius. An SEZ could benefit from access to other SEZs in its vicinity, with synergies
29 || SPECIAL ECONOMIC ZONES IN ASIA

arising from a concentration of specialised labour supply, reduced logistics costs or well-
aligned business policies in a region, all of which can generate a positive agglomeration effect on
export performance.

Table 7 shows the regression outputs. Airports appear more robust than seaports to different
model specifications in the export model. This comes as a surprise, given that maritime transport
accounts for the bulk of international trade. The result can be interpreted in a number of ways.
The first is simply that airports matter more for China’s SEZ exports. This is not unlikely, given that
ICT products comprise the bulk of China’s exports and a considerable number of high-value, low-
bulk products are transported via air. The second interpretation is that proximity matters more for
airports than seaports. Indeed, air shipment is often preferred for just-in-time delivery in ultra-lean
supply chains. Hence, when hours matter, being close to an airport may be more important than
being close to a seaport.

It should be noted that when the dependent variable is switched to imports, seaports
start to become significant. This may be due to differences in the product composition
of imports and exports. China tends to import raw materials and intermediate goods
while exporting finished goods. Raw materials tend to be bulky and hence more economical to ship
via sea.

The second finding is the performance of the “number of other SEZs” variable, which, prima
facie, points to strong agglomeration effects. A popular saying in the Pearl River Delta,
arguably the world’s densest manufacturing cluster, is that you can source any electronic component
in the world within a one-hour drive. While that may be an exaggeration, the close proximity of
technology firms—both hardware and software—and a labour force the size of Germany’s in an area
the size of Luxembourg may be hard to beat.

Table 7 – Regression output for drive time specification

Data source: Calculations: EIU.

ASIAN INFRASTRUCTURE
ASIAN INFRASTRUCTURE FINANCE
FINANCE 2021
2021 ANNEX

SUSTAINING GLOBAL VALUE

MEASURING TRANSPORT CHAINS FOR TRADE IN ASIA
CONNECTIVITY

The Asian Infrastructure Finance 2021 report examines how Asian economies, to different
extents and in different ways, have integrated global value chains (GVCs) into their growth
models. It emphasizes how critical infrastructure quality and capacity are to the agility
and resilience of GVCs, as examined against the backdrop of the COVID-19 pandemic,
increased trade tensions, rapid technological development, environmental pressures
and other factors. Using case studies and research, the AIF 2021 report illustrates how
GVCs have provided opportunities for countries and companies to become internationally
competitive, in part through technological advancements and efficiency improvements.
It also examines how GVC engagement could reinforce existing inequalities and explores
possible paths for a just and inclusive transition, recognizing countries’ different starting
points and capacities. The report highlights how green infrastructure, consistent with net
zero transition, will become a source of competitive advantage and the key to sustaining
future GVCs.

Asian Infrastructure Investment Bank (AIIB)

AIIB Headquarters, Tower A, Asia Financial Center
No. 1 Tianchen East Road, Chaoyang District, Beijing 100101 China aiib.org