asia-pacific s d journal · a sia-p acific s ustainable d evelopment j ournal vol. 26, no. 1, june...

ASIA-PACIFIC

SUSTAINABLE DEVELOPMENT

JOURNAL

Vol. 26, No. 1, June 2019

IN THIS ISSUE:

Valuing the digital economyof New ZealandJonathan Millar and Hamish Grant

Petroleum consumption and economicgrowth relationship: evidence fromthe Indian statesSeema Narayan, Thai-Ha Le,Badri Narayan Rath andNadia Doytch

Current trends in private financingof water and sanitationin Asia and the PacificHongjoo Hahm

Impact of food inflation on headline inflationin IndiaAnuradha Patnaik

Tapping capital markets and institutionalinvestors for infrastructure developmentMathieu Verougstraete and Alper Aras

The Economic and Social Commission for Asia and the Pacific (ESCAP) serves as theUnited Nations’ regional hub, promoting cooperation among countries to achieve inclusiveand sustainable development. As the largest regional intergovernmental platform with53 member States and 9 associate members, ESCAP has emerged as a strong regionalthink-tank, offering countries sound analytical products that shed light on the evolvingeconomic, social and environmental dynamics of the region. The Commission’s strategicfocus is to deliver on the 2030 Agenda for Sustainable Development, which it does byreinforcing and deepening regional cooperation and integration in order to advanceconnectivity, financial cooperation and market integration. The research and analysisundertaken by ESCAP coupled with its policy advisory services, capacity building andtechnical assistance to governments aims to support countries’ sustainable and inclusivedevelopment ambitions.

*The designations employed and the presentation of material on this map do not imply the expressionof any opinion whatsoever on the part of the Secretariat of the United Nations concerning the legalstatus of any country, territory, city or area or of its authorities, or concerning the delimitation ofits frontiers or boundaries.

The shaded areas of the map indicate ESCAP members and associate members.*

United NationsNew York, 2019

ASIA-PACIFICSUSTAINABLE DEVELOPMENT

JOURNAL

ii

ASIA-PACIFICSUSTAINABLE DEVELOPMENT JOURNAL

Vol. 26, No. 1, June 2019

United Nations publication

Sales No. E.20.II.F.98

Copyright © United Nations 2019

All rights reserved

Printed in Thailand

ISBN: 978-92-1-120797-2

e-ISBN: 978-92-1-004570-4

ISSN (print): 2617-8400

ISSN (online): 2617-8419

ST/ESCAP/2871

Cover design: Nina Loncar

This publication may be reproduced in whole or in part for educational or non-profit purposes withoutspecial permission from the copyright holder, provided that the source is acknowledged. The ESCAPPublications Office would appreciate receiving a copy of any publication that uses this publication asa source.

No use may be made of this publication for resale or any other commercial purpose whatsoeverwithout prior permission. Applications for such permission, with a statement of the purpose and extentof reproduction, should be addressed to the Secretary of the Publications Board, United Nations,New York.

iii

Editorial Advisory Board

Kaushik BasuProfessor of Economics and C. Marks Professor of International StudiesDepartment of Economics, Cornell University

Martin RavallionEdmond D. Villani Professor of EconomicsGeorgetown University

Sabina AlkireDirector of the Oxford Poverty and Human Development Initiative (OPHI)Oxford Department of International Development, University of Oxford

Naila KabeerProfessor of Gender and Development, Department of Gender StudiesLondon School of Economics and Political Science

Li XiaoyunChief Senior Advisor at the International Poverty Reduction Centre in China, andDirector of OECD/China-DAC Study Group, Chair of the Network of SouthernThink Tanks (NeST) and Chair of China International Development Research Network

Shigeo KatsuPresident, Nazarbayev University

Ehtisham AhmadVisiting Senior FellowAsia Research Centre, London School of Economics and Political Science

Myrna S. AustriaSchool of Economics, De La Salle University

Chief Editors

Srinivas TataDirector, Social Development Division (SDD) of ESCAP

Hamza Ali MalikDirector, Macroeconomic Policy and Financing for Development Division (MPFD)of ESCAP

Editors

Patrik AnderssonChief, Sustainable Socioeconomic Transformation Section, SDD

Cai CaiChief, Gender Equality and Social Inclusion Section, SDD

Sabine HenningChief, Sustainable Demographic Transition Section, SDD

Oliver PaddisonChief, Countries with Special Needs Section, MPFD

Sweta SaxenaChief, Macroeconomic Policy and Analysis Section, MPFD

Tientip SubhanijChief, Financing for Development Section, MPFD

Editorial Assistants

Gabriela Spaizmann

Pannipa Jangvithaya

iv

EDITORIAL STATEMENT

The Asia-Pacific Sustainable Development Journal (APSDJ) is published twice

a year by the Economic and Social Commission for Asia and the Pacific. It aims to

stimulate and enrich research in the formulation of policy in the Asia-Pacific region

towards the fulfillment of the 2030 Agenda for Sustainable Development.

APSDJ welcomes the submission of original contributions on themes and issues

related to sustainable development that are policy-oriented and relevant to Asia and the

Pacific. Articles should be centred on discussing challenges pertinent to one or more

dimensions of sustainable development, policy options and implications and/or policy

experiences that may be of benefit to the region.

Manuscripts should be sent to:

Chief Editors

Asia-Pacific Sustainable Development Journal

Social Development Division and

Macroeconomic Policy and Financing for Development Division

United Nations Economic and Social Commission for Asia and the Pacific

United Nations Building, Rajadamnern Nok Avenue

Bangkok 10200, Thailand

Email: [email protected]

For more details, please visit www.unescap.org/publication-series/APSDJ.

v

ASIA-PACIFIC SUSTAINABLE DEVELOPMENT JOURNALVol. 26, No. 1, June 2019

CONTENTS

Page

Jonathan Millar and Hamish Valuing the digital economy of New Zealand 1

Grant

Seema Narayan, Thai-Ha Le, Petroleum consumption and economic 21

Badri Narayan Rath and growth relationship: evidence from

Nadia Doytch the Indian states

Hongjoo Hahm Current trends in private financing of water 67

and sanitation in Asia and the Pacific

Anuradha Patnaik Impact of food inflation on headline inflation 85

in India

Mathieu Verougstraete and Tapping capital markets and institutional 113

Alper Aras investors for infrastructure development

vi

Explanatory notes

References to dollars ($) are to United States dollars, unless otherwise stated.

References to “tons” are to metric tons, unless otherwise specified.

A solidus (/) between dates (e.g. 1980/81) indicates a financial year, a crop year or an

academic year.

Use of a hyphen between dates (e.g. 1980-1985) indicates the full period involved,

including the beginning and end years.

The following symbols have been used in the tables throughout the journal:

Two dots (..) indicate that data are not available or are not separately reported.

An em-dash (—) indicates that the amount is nil or negligible.

A hyphen (-) indicates that the item is not applicable.

A point (.) is used to indicate decimals.

A space is used to distinguish thousands and millions.

Totals may not add precisely because of rounding.

The designations employed and the presentation of the material in this publication do

not imply the expression of any opinion whatsoever on the part of the Secretariat of the

United Nations concerning the legal status of any country, territory, city or area or of its

authorities, or concerning the delimitation of its frontiers or boundaries.

Where the designation “country or area” appears, it covers countries, territories, cities or

areas.

Bibliographical and other references have, wherever possible, been verified. The United

Nations bears no responsibility for the availability or functioning of URLs belonging to

outside entities.

The opinions, figures and estimates set forth in this publication are the responsibility of

the authors and should not necessarily be considered as reflecting the views or carrying

the endorsement of the United Nations. Mention of firm names and commercial products

does not imply the endorsement of the United Nations.

VALUING THE DIGITAL ECONOMY OF NEW ZEALAND

Jonathan Millar and Hamish Grant*

The present paper provides estimates of the value of the digital economy ofNew Zealand through the use of the supply-use tables. By design, nochanges are made to the production boundary as the products beingassessed are already included within the production boundary and grossdomestic product (GDP). The approach is a practical attempt at using theframework first presented in the paper entitled “Measuring digital trade:towards a conceptual framework”, and in particular, the “nature”component of the framework. This is extended to the whole economy toidentify “digital” transactions in the country’s National AccountsCommodity Classification. The main finding from this paper is that the“digitally ordered” and “digitally delivered” aspects of the framework wereable to be broadly applied. However, the significant material assumptionsand the broad nature of the product classification at the aggregate levelmeant that our estimates were not of high quality. For the year endingMarch 2015, the estimate of the value of gross output of New Zealand thatcan be delivered digitally was 27.9 billion New Zealand dollars (NZ$)(US$18.8 billion), while for digitally ordered gross output, it was NZ$109.2billion

JEL classification: E01

Keywords: digital economy, supply-use tables, digitally ordered, digitally delivered,

platform enabled, national accounts, gross domestic product, Statistics New Zealand

1

* Jonathan Millar and Hamish Grant (email: [email protected]), National Accounts, StatisticsNew Zealand, Wellington, New Zealand. The opinions, findings, recommendations, and conclusionsexpressed in the present paper are those of the authors. They do not represent those of Statistics NewZealand, which takes no responsibility for any omissions or errors.

Asia-Pacific Sustainable Development Journal Vol. 26, No. 1

2

I. BACKGROUND

National statistical offices play a key role in providing information that supports and

informs efforts aimed at making progress towards achieving the Sustainable

Development Goals of the 2030 Agenda for Sustainable Development. A critical part of

this role is to ensure that the relevance of the information is maintained. Otherwise,

there is a risk that the information would be misleading, and any decisions based on this

may not lead to the desired outcomes. The digital economy is an area that has

developed quickly. The international statistical community, including national statistical

offices, are exploring ways to show this development and understand the impacts on

economic statistics as part of maintaining relevance and supporting the implementation

of the 2030 Agenda.

The present paper entails a discussion on an attempted application of parts of the

framework first proposed in the paper entitled “Measuring digital trade: towards

a conceptual framework” (OECD, 2017b). It is assumed that readers are familiar with the

framework and the work previously done by the OECD Informal Advisory Group on

Measuring of GDP in a Digital Economy. The expanded framework is shown in figure 1.

Source: OECD (2017c).

Figure 1. Digital economy conceptual framework

Producers(‘who’)

Product(‘what’)

Nature(‘how’)

Users(‘who’) Enablers

Digital

products

Producers of

information

and/or

Digitallyordered

Platformenabled

and/or

Digitaldelivered

and/or

Goods

Services

Information Government

Consumer

Business

Definitions of concepts, digitally ordered, platform enabled and digitally delivered,

are taken from the aforementioned paper. These definitions along with the definitions

others concepts mentioned in this paper are given in appendix 3.

The digital economy is a growing area of interest for Statistics New Zealand (Stats

NZ) customers, such as the Organization for Economic Cooperation and Development

(OECD), government departments, and the private sector. The appetite for the

measurement of the digital economy of New Zealand is driven by the desire to improve

the understanding of its role within the country’s economic and social context.

Valuing the digital economy of New Zealand

3

Stats NZ has been working with the Ministry of Innovation and Employment in

developing a “digital domain plan”. This project is focused on how New Zealanders,

businesses and the public sector use digital technologies and will formalize questions of

common interest across government to support better management of the digital

economy. This initial scoping will be used to guide the country’s approach for measuring

the digital economy.

Stats NZ continues to conduct research on the digital economy. Two particular

areas mentioned here are on the consumer price index (CPI) and national accounts.

Investigations within national accounts have been ad hoc and mainly in response to

OECD requests. This paper is our first attempt at measuring the digital economy from

a high-level macroeconomic perspective. Research on CPI is often focused on an

individual enterprise or transaction basis as transactions representative of a larger group

through weighting.

II. APPLYING THE FRAMEWORK

Methodology

The framework first proposed in the paper entitled “Measuring digital trade:

towards a conceptual framework” and since adapted by the OECD Informal Advisory

Group on Measuring GDP in a Digital Economy, has been used to compile initial

estimates of gross output from the digital economy in New Zealand. The approach has

been to classify products as digitally ordered, platform enabled and digitally delivered

from the New Zealand supply-use tables.

The analytical interest in these estimates may be in understanding the extent to

which a digital element is present in economic production. For the most part, the value

for each product included represents the full value of production in the economy for that

product, instead of only the part that has been digitally ordered, platform enabled or

digitally delivered. As such, it could be interpreted that the product values presented in

this paper are the maximum potential values, if those who do not already use digital

ordering or digital delivery in production of the product, move to digital ordering or digital

delivery.

In our calculation of gross output, we have excluded changes to work in progress

and finished goods stock change, as stock changes are not split by product level, but we

have included own account capital formation (OAKF) for software and information

technology design and development-related services. Most of the products of interest

are services and do not have a stock change element, but we assume that most

industries have some level of in-house, capitalized information technology systems.


4

Appendix 1, table A.1 shows how we have applied the nature section of the

framework in our estimates. This involved identifying products within the National

Accounts 2006 Commodity Classification (NA06CC) that were digitally ordered, platform

enabled and digitally delivered.

The level of detail presented in appendix 1 is the level that is available in the

supply-use system, although we were able to further split the retail and wholesale trade

classifications. Some products are sufficiently detailed to identify them as being mostly

digital. Others are far broader and include significant non-digital output. The products

captured are kept constant over the time series.

We also use the industrial sector when assessing the products identified in

appendix 1. We remove industries that recorded some output of a product, which are not

likely to have the digital aspect that is of interest to us.

Products sold from retail trade and wholesale trade industries are recorded only as

margin and not as the gross product value. This is a divergence from the methodology

of all other products where values are the gross total amount. This methodology has

been selected because the range of goods able to be sold through retail and wholesale

trade is vast, in particular for department stores and supermarkets. To include the value

of all underlying products, we need to work with a much larger number of products,

which, in turn, would resulted in a less useful final figure for digitally ordered.

Digitally ordered

Our estimate of digitally ordered goods and services produced in the New Zealand

economy were 109.2 billion New Zealand dollars (NZ$) (US$73.2 billion) in 2015. This is

up from NZ$81.6 billion in 2007, and has increased at an annual pace of 3.8 per cent

over the period, largely mirroring growth in total gross output of New Zealand, at 4 per

cent from 2007 to 2015.

The value of digitally ordered gross output makes up approximately 20 per cent of

total gross output of New Zealand. This proportion remained consistent over the

observed period. Future estimates must account for the introduction of products over the

time series as they became digitally ordered. This is likely to be difficult to estimate with

any certainty, which is why we have kept our products constant.

In estimating the value of digitally ordered, we assume a product is digitally

ordered if it is likely that online orders make up a non-insignificant portion of the

industries’ output. This is a subjective estimate without the use of a “percentage sales

made digitally” rule given that these data are not available.

Most products could feasibly be ordered digitally, but only some products are likely

to have been commonly purchased digitally. From a New Zealand perspective, this


5

equates to 49 products within our classification, 44 of which are services and the

remaining five are goods (appendix 1).

Of the 49 digitally ordered products included here, there are 112 subindustries

contributing to the NZ$109 billion in total. Most of these values are small and relate to

only one product. Appendix 2, table A.2, shows the industries that contribute more than

NZ$1 billion in gross output and that retail trade and financial services are the largest

contributors.

These values of digitally delivered gross output illustrate that digitization is

prevalent in a large part of New Zealand production. It shows that digitization in this form

is not necessarily tied to innovation, but that the kind of digital ordering is the new

normal for many industries. Most of the output included in the value of digitally ordered

would still occur without the presence of digitization.

In terms of understanding the overall extent to which digital ordering is common

among New Zealand industries, this measure is useful, especially for answering

questions about the value of sales digitally versus brick and mortar store sales. How we

have applied it in this paper has not resulted in accurate figures to answer this question;

further splits are needed to be applied to product data.

Digitally delivered

Gross output of digitally delivered products rose 39 per cent over the period

2007-2015 from NZ$20 billion to 27.9 billion. On an annual basis, digitally delivered

products increased more than the total economy gross output, with an average increase

of 4.3 per cent over the period, as compared to 4 per cent for gross output. Digitally

delivered products contributed between 5.7 and 6.1 percent of total gross output over

the observed period.

This increase in digitally delivered products was driven in part by a 10 per cent

annual average change in the value of mobile and Internet telecommunications services

and online content, information technology design and development-related services,

and licensing services for the right to use computer software and databases.

The digitally delivered dimension narrowed the focus of the digital economy and

resulted in the gross output associated with digital delivered being slightly more than

25 per cent of the value of digitally ordered transactions.

The major contributing industries were the telecommunications services industry,

financial intermediation services directly measured from the banking and financing

industry, and the computer system design and related services industry. The industry

dimension is shown in figure 2. Digitally delivered services are dominated by a few large

industries, with all other industries contributing a negligible amount.


6

The scope of digitally delivered products is an area that is challenging to interpret

in some cases.

Figure 2. Digitally delivered industry composition

The note to OECD Informal Advisory Group members accompanying the second

digital economy questionnaire indicated that digitally delivered would include examples,

such as downloadable products and database services. Beyond these examples,

however, there are activities that could be included as digitally delivered.

An example of a product that was reviewed was fixed telecommunications

services. In practice, this classification includes telecommunications package deals,

which are effectively delivered over the Internet via Voice over Internet Protocol (VoIP),

but are classified as fixed telecommunications services. It was decided that while this

service (telecommunications) is not downloaded, its delivery through the Internet

qualifies it as a digitally delivered product.

Another example is financial Internet services indirectly measured (FISIM). The

degree to which the financial sector is delivered digitally is likely to differ by country. In

New Zealand, the banking sector is fairly digitalized; consumer research agency Canstar

Blue reported in 2015 that between 49 and 57 per cent of New Zealand consumers used

online banking, depending on generational position (Davies, 2015). It is our assumption

that this proportion of users has grown and that FISIM is a service that has

a significantly digitally delivered element.

2007 2008 2009 2010 2011 2012 2013 2014 2015

NZ

$ m

illio

ns

30 000

25 000

20 000

15 000

10 000

5 000

0

JJ Telecommunications services industry KK Financial and insurance services

MN Professional, scientific and technical services

RS Arts and recreation services All other industries

OO Public administration and safety


7

The scope of digitally delivered production in this paper is more about including

production which likely would not take place, or would be significantly different, without

digitization.

This dimension of the framework could also be useful in identifying industries that

may experience changes in the short term with evolving technologies and potential

disruptions to the way production is delivered and consumed.

Platform enabled

Through our classifications, we are not able to provide the level of detail required

to adequately identify output from platform enabled means. Platform enabled activities

can be and are present in many industries still predominantly non-platform enabled.

Using the methodology as presented above, we have not included any products

because the platform enabled aspect of these activities is still likely to represent

a relatively small proportion of total output for these sectors.

In this sense, it is not effective to include these transactions, as they would not

provide any useful narrative. As was expected, to get reliable data on platform enabled

production require additional data sources or breakdowns that are currently not

available within the existing national accounts in New Zealand.

We still, however, consider this to be a useful part of the framework. Stats NZ is

engaging and developing relationships with many digital intermediaries to obtain usable

admin data. While these relationships are still very new, this presents a promising

development for going forward.

Other potential methods for gathering data on platform enabled production are

web-scraping and the use of application programme interfaces (APIs). Digitization

makes it possible to use new methods of gathering information on this kind of activity,

which we hope can improve deflators and estimates of household consumption

expenditure within national accounts.

III. DIRECT VERSUS INDIRECT

We found digitally delivered to be the easiest to identify when attempting to

estimate the digital economy from existing macroeconomic data in the national

accounts. At an aggregate level, there are few indicators of digital businesses and

transactions.

Comparing the gross output of digitally delivered products with those that are

digitally ordered leads us closer to a representation of the gross output directly

attributable to the digital economy and the value indirectly attributable to the digital

economy.


8

Direct contribution of the digital economy would likely include activities only made

possible through digital means. For our purposes, digital means includes production in

and over computer networks which cannot be produced through a non-computerized

mechanism. The types of products included within digitally delivered and platform

enabled dimensions of the framework would contribute most to a direct measure of the

digital economy. Digitally ordered direct contribution is likely to be much smaller

compared to its indirect contribution to the digital economy.

Products that may be included in a direct contribution estimate from our list of

digitally delivered products are, for example, packaged software; mobile

telecommunications services; telecommunications services and online content; and

licensing services for the right to use computer software and databases.

The framework set out above is a useful concept in practically determining different

types of digital economy production; however, the level of detail required to accurately

identify and classify these activities is not available within current national accounts

data.

Any potential digital economy satellite would benefit usability if it were to follow

other satellites that estimate direct and indirect contributions to value added or gross

output. The delineation of digitally ordered, digitally delivered and platform enabled is

effective within a satellite account as additional estimates for aiding in analysis of the

digital economy.

IV. ENABLERS

With the recent redevelopment of the ICT Supply (ICTS) survey at Stats NZ, we

decided to also look at the enablers dimension of the framework, which was added to

the framework by the OECD Advisory Group.

The ICTS survey run by Stats NZ is naturally suited to support this dimension. Until

2014, the biennial survey had been a census targeting all economically significant

resident New Zealand businesses involved in producing and supplying information and

communications technology (ICT) goods and services. In 2017, a redesigned ICT survey

was put into the field, and renamed ICT Software and Services because of its stripped

back nature. This survey is now a sample of businesses instead of a census and only

focuses on the sales of software and services.

We find that in 2010 and 2012, the rolling mean employment (RME) group of 500+

employees contributed the most to total sales, but in 2014 the RME group of 50 to 249

contributed the most. This is reflective of the RME grouping of 50-249 being the second

fastest growing segment behind businesses with an RME of 1-9 (figure 3).


9

Figure 3. Total sales by rolling mean employment count

The forthcoming data from the redesigned ICT Software and Services survey, due

early 2018, will provide new insights on the changing nature of the companies acting as

enablers of the digital economy in New Zealand. We will also be interested in the new

figures relating to information telecommunication hosting and cloud computing services,

which are expected to have increased since the survey was last conducted in 2014.

V. OTHER DEVELOPMENTS AT STATS NZ

Within Stats NZ, work is also centred on the digital economy in particular, from the

prices unit, and the International Business Statistics team from their contribution to the

Digital Nation Domain Plan.

The prices development team at Stats NZ are looking at several additions from the

digital economy to the consumer price index (CPI). The team is considering to add

“private accommodation rented from others” and ride-sharing to the CPI basket of

goods. The prices for this will most likely be collected through APIs and web-scraping

with weights determined from a mixture of household expenditure surveys and market

research.

The prices development team are also interested in digital downloads of films,

music, and video games, which are currently included in the CPI basket, but are likely to

be under-reported.

35

30

25

20

15

10

5

0Sole proprietor 1 to 9 10 to 49 50 to 249 250 to 499 500+

2010 20142012

Per cent


10

Over the past two years, the prices development team at Stats NZ increased their

focus on new tools, such as web-scraping and APIs, to increase their coverage of not

only new data sources, such as APIs, but also on traditional activities. It is thought that

their work in this area and on the digital economy will be able to be incorporated into the

national accounts in the future.

The Digital Nation Domain Plan is another significant activity that Stats NZ is

involved in along with the Ministry of Business Innovation and Employment. To date, this

work has involved a stock take of the enduring questions facing government. These

questions have focused on what New Zealanders are doing with digital technologies;

what New Zealanders want to do with digital technologies; and what policymakers would

like New Zealanders to do with digital technologies. The Digital Nation programme has

been developed across the government, with the support of country’s digital community

and is aligned closely with the OECD “Going Digital” projects Pillar 1, Horizontal

activities.1

The OECD Going Digital project and the Digital Nation programme are aimed at

increasing the accessibility and effective use of digital technologies to drive innovation,

improve productivity and enhance quality of life. The cross government and stakeholder

approach, suggested by the Going Digital project and already set in motion by Stats NZ

and MBIE, positions New Zealand well in the digital policy environment internationally.

While the Digital Nation programme may not result in statistics that will be

immediately implementable in the national accounts, it will help to further understand the

digital economy within Stats NZ and across the government.

VI. CONCLUSION

The central theme of this paper is to simply apply the framework we received with

the second OECD questionnaire in May 2017. The digital economy is not only an area of

interest to Stats NZ, but it is also of interest to many areas, which could benefit from

additional focus and research related to it. The work presented in this paper is an

interesting exercise at applying a very useful framework for understanding the digital

economy in New Zealand using existing national accounts data.

Our work highlights the need for further discussion to improve the understanding

around the scope of digitally ordered and digitally delivered. It also highlights one way of

how this framework may be implemented on existing national accounts data, and the

data gaps that hamstring these estimates to a low level of quality.

1 For more details, see www.oecd.org/going-digital/project/.


11

Areas for future research that would be beneficial are exploring direct data

collection from platforms to cover the facilitation part and the actual provider of the

service. This would enable better estimates of the value of this activity and be more

efficient than surveying a large number of households or businesses. Another approach

to investigate is web-scraping for specific products, which may help to elaborate which

products to include.

The continued work on the digital economy across Stats NZ and the developing

data sources for key activities taking place in this area are exciting potential

developments, which is intended to improve our coverage and ability to measure the

digital economy in New Zealand.


12

APPENDICES

Appendix 1

National Accounts 2006 Commodity Classification (NA06CC) –

selection of products

The table shows the products selected as having a significant digital aspect. That

is, the nature of the transactions is digitally ordered, platform enabled or digitally

delivered.

The approach taken for each product is to investigate the firms that supply this

product and any existing New Zealand or international studies along with news articles

to understand the nature of how a particular product is ordered, delivered or facilitated

through a platform. As an example, NA06CC 322.00 includes “books” and “music”.

These products can be purchased as a physical CD or book though retail firm’s websites

or as a digital download or for music, as part of a streaming service from international

companies. Hence, the product can be digitally ordered and digitally delivered. Note,

platform enabled has not been considered as part of this work because of the underlying

information not having sufficient detail to separate this out.

The groups of products selected tend to be predominantly service-related products

rather than raw goods or intermediate goods. Goods are generally those with an

NA06CC code of less than 500, while 500-999 codes are mostly service related. The

three main reasons for services rather than goods being selected are the following:

• Selecting margin products only for products sold through retail and wholesale

firms

• Digitally delivered or platform enabled tend to relate to services

• Many goods are used by firms as part of intermediate inputs to produce other

goods

One of the main reasons for services being predominantly selected is because

only the retail and wholesale margin value is included when a product is purchased

through a retailer or wholesaler. For example, many food and beverage products are in

the NA06CC 200 codes; however, none of these products are included as digital even

though many of them can be digitally ordered though retailers, such as supermarkets

that offer online shopping. Instead, the purchase of the good through a retailer or

wholesaler is split into two products: the underlying good and the retail or wholesale

margin product. Only the additional mark up or margin the supermarket adds on is

recorded as being digitally ordered; this is shown as a retail or wholesale margin product


13

in the national accounts supply-use tables. Including the full value and the underlying

products would significantly increase the value and number of products for digitally

ordered, which limits the usefulness of this information.

Digitally delivered requires that the product being delivered can be received

digitally only. This is not the case for goods. The only area where there has been some

discussion internationally around digitally delivered goods is related to 3D printing. The

argument for excluding this is that the plans are delivered digitally, but the actual printing

part only would be a good. A platform enabled activity also tends to be service related,

as it is easier to facilitate than the facilitation of actual goods.

Another reason that there does not tend to be many goods selected is that often

these goods are purchased by firms as intermediate inputs into subsequent processes

to produce other goods. These purchases are much less likely to be done digitally than

directly to consumer activity, as firms already have direct relationships with suppliers

and there is less agglomeration benefit from setting up digital ordering. An additional

factor here is that it is more difficult to identify these digital transactions for firms based

on the current information available.


14

Table A.1. National Accounts 2006 Commodity Classification (NA06CC) –

selection of products

NA06CC DescriptionDigitally Platform Digitally

ordered enabled delivered

322.00 Books, maps, music, cards, pictures and plans; Y N Y

excluding advertising material

323.00 Newspapers and periodicals, in print Y N N

493.00 Games and toys; roundabouts, swings and other Y N N

fairground amusements

710.00 Wholesale trade services Y N N

720.00 Retail trade services Y N N

730.00 Accommodation services Y N N

741.00 Meal serving services Y N N

751.10 Road transport services of freight; transport Y N N

services via pipeline

751.20 Road passenger transport Y N N

752.10 Railway transport services of freight Y N N

752.20 Railway passenger transport Y N N

753.20 Water passenger transport Y N N

754.10 Air transport services of freight Y N N

754.20 Air passenger transport Y N N

755.00 Scenic and sightseeing transportation services Y N N

756.00 Postal and courier services Y N N

768.00 Freight transport agencies and other supporting Y N N

transport services

781.00 Publishing, printing and reproduction services Y N N

782.00 Packaged software Y N Y

783.00 Audio, video and other disks, tapes and other Y N N

physical media, recorded

784.00 Audio-visual and related services Y N Y

785.00 Broadcasting, programming and programme Y N Y

distribution services

786.10 Fixed telecommunications services Y N Y

786.20 Mobile telecommunications services Y N Y

789.00 Internet telecommunications services and Y N Y

online content


15

791.20 Library and archive services Y N Y

811.10 Financial intermediation services directly measured Y N Y

811.11 Financial intermediation services, insurance Y N N

services and pension services

812.10 Life insurance Y N N

812.20 Accident and health insurance services Y N N

812.30 Other insurance services Y N N

813.00 Services auxiliary to financial services other than Y N N

to insurance and pensions

814.00 Services auxiliary to insurance and pensions Y N N

821.10 Leasing or rental services concerning transport Y N N

equipment without operator

822.10 Licensing services for the right to use computer Y N Y

software and databases

831.10 Real estate services involving own or leased Y N N

residential property

915.00 Accounting, auditing, bookkeeping, insolvency, Y N Y

receivership and taxation services

916.00 Advertising services and provision of advertising Y N Y

space or time

917.00 Market research and public opinion polling services Y N Y

923.10 Information technology design and development Y N Y

related services

924.00 Travel arrangement, tour operator and Y N N

related services

925.00 Employment services Y N N

931.10 Local government administration services Y N N

932.10 Central government administrative services Y N N

961.00 Live entertainment event presentation and Y N N

promotion services; services of performing and

other arts; museum and preservation services

962.10 Sports and recreational sports facility operation Y N N

services

Table A.1. (continued)




16

963.20 Lottery services Y N N

963.30 Racing and sports betting services Y N Y

963.40 Online gambling services; gaming machines Y N N

outside of casinos; other gambling services

Notes: Y, yes; N, no. The full classification can be found in the tab labelled “NA06CC to CPC”. Available at http://

archive.stats.govt.nz/~/media/Statistics/browse-categories/economic-indicators/national-accounts/supply-use-

tables/na-input-output-tables-ye-mar13.xlsx.





17

Appendix 2

Table A.2. Digitally ordered industries Industries >NZ$1 billion

Digitally ordered industries2015 2014 2013 2008 2007

(NZ$ millions)

GH Retail trade 20 643 19 217 18 449 16 117 15 279

KK Financial and insurance services 19 899 18 314 16 822 16 403 15 442

II Transport, postal and warehousing 17 487 16 511 15 691 14 946 13 866

MN Professional, scientific and 12 558 11 950 11 199 8 436 7 823

technical services

LL Rental, hiring and real estate 11 586 10 799 10 001 7 540 7 235

services

JJ Information media and 11 812 11 926 11 929 10 377 9 993

telecommunications

FF Wholesale trade 7 601 7 324 7 274 6 662 6 039

RS Arts and recreation services 3 411 3 258 3 201 3 098 3 013

CC Manufacturing 1 746 1 656 1 763 1 522 1 498


18

Appendix 3

Definitions

Digitally ordered

“An e-commerce transaction is the sale or purchase of a good or service,

conducted over computer networks by methods specifically designed for the

purpose of receiving or placing orders. The goods or services are ordered by those

methods, but the payment and ultimate delivery of the goods or services do not

have to be conducted online. An e-commerce transaction can be between

enterprises, households, individuals, governments, and other public or private

organizations. To be included are orders made over the web, extranet or electronic

data interchange. To be excluded are orders made by phone, fax or manually

typed email.” (OECD, 2011).

Platform enabled

An important characteristic of digitalization is peer-to-peer services intermediated

by digital intermediary platforms (“sharing economy”, “gig economy”, “collaborative

economy”), such as Airbnb, Uber and eBay, that facilitate transactions in goods and

services (OECD, 2017a, p. 5).

Digitally delivered

The third dimension is referred to as digitally delivered; in other words, it captures

those services and data flows that are delivered digitally as downloadable products.

Examples include software, e-books, data and database services. Goods, as physical

items, are not very likely to be digitally delivered en masse. However, 3D printing may

possibly result in a (future) category of transactions that could be classified under

digitally delivered goods, if these transactions are deemed to be fundamentally different

from trade in services (of 3D blueprints) transactions.

Direct versus indirect contribution

Direct contribution is where the use of digital mediums is the reason for the activity

and accounts for all or most of the value of the activity. Indirect contribution is simply

activity facilitated by digital mediums where the product or service is carried out

physically (non-digitally).

Calculation used for gross output

Gross output (GO) = Sales + Margin on goods purchases for resale + Own account

capital formation (OAKF) + Service for own use (SFOU) + Fringe benefit value excluding

GST (FBVEXGST) + Work in progress and finished good stock change


19

REFERENCES

Davies, J. (2015). How much banking do we do online? Canstar Blue, 18 May. Available atwww.canstarblue.co.nz/banking-insurance/banking/how-much-online-banking/.

Organization for Economic Cooperation and Development (OECD) (2011). Guide to Measuring the

Information Society. Paris: OECD Publishing.

(2017a). Issue paper on a proposed framework for a satellite account for measuring the digitaleconomy. Paper prepared for the Meeting of Advisory Group on Measuring GDP ina Digitalised Economy. OECD Conference Centre, Paris, 10 November. Available atwww.oecd.org/off ic ia ldocuments/publ icdisplaydocumentpdf /?cote=STD/CSSP/WPNA(2017)10&docLanguage=En.

(2017b). Measuring digital trade: towards a conceptual framework. Working Paper onInternational Trade in Goods and Trade in Servicer. Available at www.oecd.org/off ic ia ldocuments/publ icd isplaydocumentpdf /?cote=STD/CSSP/WPTGS(2017)3&docLanguage=En.

(2017c). Summary of responses of the Advisory Group: survey on digital economy typology.Paper prepared for the Committee on Statistics and Statistical Policy. Paris, 9-10 November.

PETROLEUM CONSUMPTION AND ECONOMIC GROWTHRELATIONSHIP: EVIDENCE FROM THE INDIAN STATES

Seema Narayan, Thai-Ha Le, Badri Narayan Rath and Nadia Doytch*

This paper reveals that over the period 1985-2013, the wealthier states ofIndia experienced a prevalence of the feedback hypothesis between realgross domestic product growth and petroleum consumption in the shortrun and the long run. Over the short term, the whole (major) 23 Indian statepanels show support for the conservative hypothesis. Regarding the panelscomprising low- and middle-income Indian states, although there appearedto be significant bidirectional effects in the long run, none of the resultssuggest that energy consumption increases economic growth. This impliesthat growth in energy demand can be controlled without harming economicgrowth. The results, however, indicate that for the low- and middle-incomestates, increases in petroleum consumption could adversely affecteconomic activity in the short and long run. These findings relate to theaggregate data on petroleum. Examining the short-run and long-runenergy-growth linkages using disaggregated data on petroleumconsumption reveals that only a few types of petroleum products havestable long-run relationships with economic growth. In fact, withdisaggregated petroleum data, the vector error correction model (VECM)and cointegration results support the neutral hypothesis for high-incomesstates. For the low- and middle-income groups, while the conservationeffect is found to prevail in the short run and the long run, higher economicgrowth appears to reduce consumption of selected types of petroleumproducts.

JEL classification: O13, Q43, C33

Keywords: petroleum consumption, economic growth, feasible generalized least squares

(FGLS), cross-sectional dependence, Indian states

21

* Seema Narayan, RMIT University, Australia. Thai-Ha Le, corresponding author, RMIT University,Viet Nam (email: [email protected]). Badri Narayan Rath, Indian Institute of Technology Hyderabad,India. Nadia Doytch, City University of New York, Brooklyn College and Graduate Center.


22

I. INTRODUCTION

Energy is an inseparable component of economic development. Among the

different energy sources, such as coal, oil, natural gas, electricity, solar, wind and

nuclear energy, oil continues to play a vital role in a country’s economy, supporting, for

example, transportation, industries, and households. In this regard, India is no

exception, however, oil is the largest energy source of the country, accounting for 31 per

cent of primary energy consumption. In 2018, oil consumption in India was 239.1 million

tons oil equivalent, an increase of 5.3 per cent compared to the previous year, and

represented a 5.1 per cent share of total world oil consumption (British Petroleum, 2019,

p. 21). In terms of barrels per day, the country consumed 5,156,000 barrels per day

(bpd), increased by 5.9 per cent compared to the previous year, and accounted for

5.2 per cent of world oil consumption in 2018, according to British Petroleum (2019).

India was the third largest consumer of crude oil in the world during the year, only

behind the United States of America (20,456,000 million bpd) and China (13,525,000

million bpd) in terms of consumption (British Petroleum, 2019).

According to Reuters, in 2017, India became the third largest net oil importer in the

world, with imports averaging 4.37 million barrels per day (Verma, 2018). Because of its

fast growing economy, energy demand in India rose rapidly over the years, in terms of

per capita energy consumption and oil consumption. This is attributable to the increased

affordability of oil (on the back of the drop in the price of oil) for a large section of its

population who previously could not afford it, as is evident in the motorization of the

Indian economy (Sen and Sen, 2016).

In per capita terms, however, oil consumption in India remains relatively low in

comparison to the world’s largest consuming economies and to other non-Organization

for Economic Cooperation and Development (OECD) countries (Sen and Sen, 2016).

Interestingly, even though the population of India is 1.3 billion, the country still lags other

emerging market powerhouses in oil consumption per capita, giving it room for rapid

growth. In September 2014, a policy initiative, the “Make in India” programme, was

launched by Prime Minister Narendra Modi.1 The objective of the programme is to put

manufacturing at the heart of the country’s growth model. A government target of

increasing the manufacturing sector’s share of gross domestic product (GDP) from

approximately 15 per cent to 25 per cent by the beginning of the next decade can be

expected to equate to a significant increase in demand for energy, and higher oil

consumption in manufacturing (Sen and Sen, 2016). Also of note, a programme

involving infrastructure construction (roads and national highways), which is being partly

funded through revenue from higher taxation of oil and oil products, is likely to support

oil demand growth in the country.

1 For more information on “Make in India” scheme, see www.makeinindia.com/about.

Petroleum consumption and economic growth relationship: evidence from the Indian states

23

Against this background, for this paper, we use state-wise petroleum consumption

and economic growth data for 23 Indian states. Our study relates to the voluminous

literature that examines the role of the energy consumption (E) and economic growth (Y)

nexus in the cases of a single country and multiple countries (Akarca and Long, 1980;

Asafu-Adjaye, 2000; Fang and Le, forthcoming; Kraft and Kraft, 1978; Le, 2016; Le and

Nguyen, 2019; Le and Quah, 2018; Lee and Chang, 2005; Apergis and Payne, 2009a;

2019b; Narayan, Narayan and Popp, 2010a; 2010b; Narayan, 2016; Oh and Lee, 2004;

Proops, 1984; Rafiq and Salim, 2009; Stern, 1993; and Yang, 2000). The E-Y nexus is

governed by four hypotheses: the growth hypothesis; the conservation hypothesis; the

feedback hypothesis; and the neutrality hypothesis.2

A number of recent studies have analysed the relationship of oil consumption and

economic growth in India. The E-Y literature on India has been based on gas (Akhmat

and Zaman, 2013); oil (Akhmat and Zaman, 2013); nuclear energy (Akhmat and Zaman,

2013; Wolde-Rufael, 2010); coal (Govindaraju and Tang, 2013); electricity (Abbas and

Choudhury, 2013; Akhmat and Zaman, 2013; Cowan and others, 2014; Ghosh, 2002;

Nain, Ahmad and Kamaiah, 2015) and aggregate energy consumption (Pao and Tsai,

2010; Vidyarthi, 2013; Yang and Zhao, 2014) (table 1).

As indicated earlier, we examine the state data for 23 states as a panel and also

divide the states by income in order to account for some heterogeneity that arises as

a result of income (see section II). As explained by the International Energy Agency

(IEA) (2015, p. 21), “(t)he widespread differences between regions and states within

India necessitate looking beyond national figures because of the country’s size and

heterogeneity, in terms of demographics, income levels and resource endowments, and

also because of a federal structure that leaves many important responsibilities for

energy with individual states.” While our study is predominantly based on aggregate

data, we also check the robustness of our findings using disaggregated petroleum data3

and have found the disaggregated data to be informative and useful because of the

importance of each petroleum product tends to vary across states.

Foreshadowing our key results, in the long run, we find evidence in favour of the

feedback effect for the all states panel in addition to all the subpanels of states at

different income levels. In the short run, we find that while the all states panel shows

support for the conservative hypothesis, all income panels seem to show the presence

of the feedback effect. Regarding the signs of the effects, however, we find that while

petroleum consumption and economic growth are positively related for the high-income

2 The growth hypothesis indicates that E causes Y; the conservation hypothesis indicates that Y causes E;the feedback hypothesis treats both E and Y as leading each other; and the neutrality hypothesis relatesno linkage between E and Y.

3 We are thankful to an anonymous reviewer for the suggestion of introducing disaggregated data in thestudy.


24

states in the short run and the long run, they can be negatively linked for the middle-

and low-income states. The use of disaggregated petroleum products data in the

analysis reveals that cointegration between petroleum products and income is missing

for the high-income states and only present for selected petroleum products in the case

of low- and middle-income states.

The remainder of the study is organized as follows. Section II includes a review of

the related literature with a focus on India. Section III contains an explanation of the

aggregate petroleum consumption and economic growth patterns for 23 Indian states. In

section IV, the econometric methods and models used to examine the four hypotheses

associated with the petroleum consumption-economic growth nexus are presented.

Section V includes a discussion of the key findings relating to the aggregate data on

petroleum consumption, while section VI presents the results derived using the

disaggregated data on petroleum consumption. Section VII provides a discussion on the

key findings and their implications relating to aggregate and disaggregated data on

petroleum consumption. Section VIII concludes the study with policy implications.

II. LITERATURE REVIEW

A handful of studies have investigated the link between energy consumption and

economic growth in India (Paul and Bhattacharya, 2004; Vidyarthi, 2013; Tiwari,

Shahbaz and Hye, 2013; Shahbaz and others, 2016; Nain, Bharatam and Kamaiah,

2017). Paul and Bhattacharya (2004) find the prevalence of the feedback hypothesis for

the Indian economy over the period 1950-1996, when energy consumption leads to

economic growth in the short run and economic growth leads to higher energy

consumption in the long run. Vidyarthi (2013) shows evidence of the feedback effect for

electricity consumption, although the casual effects in the short run and the long run

were different from Paul and Bhattacharya (2004) (see table 1). Nasreen and Anwar

(2014) find that the feedback effect is prevalent in the short run and long run over the

period 1983-2011. Tiwari, Shahbaz and Hye (2013) examine the Environmental Kuznets

Curve (EKC) hypothesis of India using aggregate coal consumption and economic

growth data along with carbon dioxide (CO2) emissions. They find feedback hypothesis

between economic growth and CO2 emissions. The same interpretation is drawn

between coal consumption and CO2 emissions.

Abbas and Choudhury (2013) concur when looking at electricity consumption in

India and agricultural GDP over the period 1972-2008. Some authors find evidence of

a unidirectional relationship relating to the growth hypothesis, which suggests that

energy consumption drives economic growth in the long run (Pao and Tsai, 2010) and in

the short run (Yang and Zhao, 2014; Nain, Ahmad and Kamaiah, 2015). Akhmat and

Zaman (2013) suggest a unilateral link for electricity and gas consumption in India in

the long run. Wolde-Rufael (2010) shows the same linkage for nuclear energy in the


25

long run. Other studies on India show evidence of the conservative hypothesis, or

a unidirectional link flowing from economic growth to energy consumption, for different

sources of energy: electricity consumption (Ghosh, 2002 (in the short run); Abbas and

Choudhury, 2013 (in the short run and the long run)); nuclear energy in the long run

(Akhmat and Zaman, 2013); and coal consumption in India in the short run (Govindaraju

and Tang, 2013). Similarly, Shahbaz and others (2016) examine the relationship

between globalization and energy consumption in India and have found acceleration of

globalization results in a decline in energy consumption, but economic growth increases

energy demand in the long run.

In the literature, we find that there is also evidence in favour of the neutrality

hypothesis for India. Akhmat and Zaman (2013), for instance, find a relationship

between fuel and oil consumption and economic growth over the period 1975-2009.

Similarly, Govindaraju and Tang (2013) find evidence supporting the neutrality

hypothesis in the case of coal in the long run for the period 1965-2009; and Cowan and

others (2014) find this for electricity consumption over the period 1990-2010.

Almost all these studies come up with short-term and long-term inferences from

Granger causality tests drawing on the vector autoregressive (VAR) model or the vector

error correction model (VECM), depending on whether a cointegration relationship

between non-stationary variables, E and Y, is established. The key variations are in the

datasets in terms of panel or time series (aggregate or disaggregated), and sample

periods; and the techniques (cointegration and causality tests) (see table 1). Naser

(2015) finds that a long-run impact of oil is associated with nuclear energy consumption

on economic growth in India, along with China, the Republic of Korea and the Russian

Federation. Bildirici and Bakirtas (2014) argue that for China and India, this relationship

is bidirectional.

Regarding the cointegration tests, several studies have used the time series

Engle-Granger univariate cointegration approach (see, for instance, Paul and

Bhattacharya, 2004); others have used the time series Johansen multivariate

cointegration method (Paul and Bhattacharya, 2004). Furthermore, to address the issue

of a small sample, some authors use the autoregressive distributed lag (ARDL) bounds

test (such as Nain, Bharatam and Kamaiah, 2017); others have tackled the small

sample problem by including more countries in the study. This gives them the benefit of

taking advantage of a larger dataset and using panel-based cointegration methods, such

as the Pedroni (1999; 2004) cointegration test, the Kao (1999) test, or the Johansen/

Fisher test, to derive results from a larger dataset (Nasreen and Anwar, 2014; Pao and

Tsai, 2010). Instead of applying the standard Granger causality test, Kónya (2006)

employs the bootstrap panel causality approach to allow for cross-section dependence

and heterogeneity within the panel. Yang and Zhao (2014), in place of the usual

in-sample Granger causality tests, apply an out-of-sample Granger causality test to

better gauge the out-of-sample forecasting performance of models. Wolde-Rufael (2010)


26

Tab

le 1

. A

su

mm

ary

of

recen

t lite

ratu

re o

n I

nd

ian

en

erg

y c

on

su

mp

tio

n a

nd

eco

no

mic

gro

wth

Stu

dy

Sam

ple

Data

Tech

niq

ue

Va

riab

les

Re

su

lt

Paul and

1950-1

996

Tim

e s

eries

Engle

-Gra

nger

Energ

y c

onsum

ption;

LR

: Y

-> E

; S

R: E

-> Y

Bh

att

ach

ary

aco

inte

gra

tio

n a

nd

GD

P;

gro

ss c

ap

ita

l

(20

04

)G

ran

ge

r ca

usa

lity;

form

ation; popula

tion

Johansen m

ultiv

ariate

coin

tegra

tion

Nasre

en a

nd

1980-2

011

Panel data

:P

edro

ni

Energ

y c

onsum

ption,

LR

and S

R: E

<->

Y

Anw

ar

(2014)

15 A

sia

n c

ountr

ies

conin

tegra

tion

PG

DP

; tr

ade o

penness;

energ

y p

rices

Tiw

ari (

20

11)

19

70

-20

07

Tim

e s

erie

sG

ran

ge

r ca

usa

lity

LR

: Y

->E

(VA

R);

Dola

do a

nd

Lü

tke

po

hl a

pp

roa

ch

En

erg

y c

on

su

mp

tio

n w

ith

carb

on

em

issio

ns a

nd

oth

er

vari

ab

les

Pao a

nd T

sai

1971-2

005

Panel in

clu

din

gK

ao, Johansen/F

isher;

Energ

y c

onsum

ption;

LR

: E

->Y

(20

10

)B

RIC

na

tio

ns

Pe

dro

ni co

inte

gra

tio

n;

real G

DP

; carb

on

(Bra

zil,

Russia

nG

ranger

causalit

yem

issio

ns

Fe

de

ratio

n,

Ind

ia

an

d C

hin

a)

Yang a

nd Z

hao

1970-2

008

Tim

eO

ut-

of-

sam

ple

Gra

nger

Energ

y c

onsum

ption;

SR

: E

-> Y

and C

O2;

(20

14

)se

rie

s/a

gg

reg

ate

ca

usa

lity t

ests

an

dre

al G

DP

; carb

on

trade o

penness->

E

directe

d a

cyclic

gra

phs

em

issio

ns; tr

ade

(DA

G)

openness

Vid

ya

rth

i (2

01

3)

19

71

-20

09

Tim

eJo

ha

nse

n a

pp

roa

ch

;E

nerg

y c

onsum

ption;

LR

:E->

Y; S

R: Y

->E

se

rie

s/a

gg

reg

ate

Gra

ng

er

ca

usa

lity

real G

DP

; carb

on

em

issio

ns


27

Tab

le 1

. (

continued)

Stu

dy

Sam

ple

Data

Tech

niq

ue

Va

riab

les

Re

su

lt

Ahm

ad a

nd

1971-2

014

Tim

eA

RD

LTo

tal energ

y, g

as, oil,

E->

CO

2; Y

<->

CO

2

oth

ers

(2

01

6)

se

rie

s/a

gg

reg

ate

(au

tore

gre

ssiv

ee

lectr

icity a

nd

co

al

dis

trib

ute

d la

g b

ou

nd

s)

co

nsu

mp

tio

n;

RG

DP

;

carb

on e

mis

sio

ns

Ele

ctr

icit

y

Ab

ba

s a

nd

19

72

-20

08

Tim

e s

erie

sJo

ha

nse

n a

pp

roa

ch

Ele

ctr

icity c

onsum

ption

Aggre

gate

: G

DP

- L

R:

Ch

ou

dh

ury

– a

gg

reg

ate

- G

DP

an

d G

DP

; P

GD

P; A

GD

PY

-> E

; S

R: Y

-> E

;

(2013)

and p

er

capita G

DP

PG

DP

- L

R:

E ≠

Y;

(PG

DP

); a

nd

SR

: Y

->E

.

dis

aggre

gate

-D

isaggre

gate

:

agriculture

GD

PA

GD

P -

LR

: Y

<->

E;

(AG

DP

)S

R: Y

<->

E

Akhm

at and

1975-2

010

Tim

e s

eries –

Gra

nger

causalit

yE

lectr

icity,

PG

DP

gro

wth

LR

: E

->Y

Zam

an (

2013)

aggre

gate

- G

DP

(VA

R)

and p

er

capita G

DP

(PG

DP

); a

nd

dis

ag

gre

ga

te -

agriculture

GD

P

(AG

DP

)

Ghosh (

2002)

1950-1

997

Tim

eE

ngle

and G

ranger

Ele

ctr

icity c

onsum

ption

LR

: Y

->E

se

rie

s/a

gg

reg

ate

(1987);

Gra

nger

an

d e

co

no

mic

gro

wth

ca

usa

lity

(pe

r ca

pita

)

Cow

an a

nd

1990-2

010

Panel – B

RIC

S/

Bo

ots

tra

p p

an

el

Ele

ctr

icity,

GD

P g

row

th,

LR

: E

≠Y

oth

ers

(2

01

4)

ag

gre

ga

teca

usa

lity a

pp

roa

ch

;C

O2

Kó

nya

(2

00

6)


28

Tab

le 1

. (

continued)

Stu

dy

Sam

ple

Data

Tech

niq

ue

Va

riab

les

Re

su

lt

Nain

, Ahm

ad

1971-2

011

Tim

eA

RD

L b

ounds test;

Secto

ral and a

ggre

gate

Ag

gre

ga

te -

LR

:

an

d K

am

aia

hse

rie

s/a

gg

reg

ate

Tod

a a

nd

Yam

am

oto

ele

ctr

icity c

on

su

mp

tio

n;

E ≠

Y; S

R: E

->Y

;

(20

15

)a

nd

dis

ag

gre

ga

te:

(19

95

)R

GD

Pd

isa

gg

reg

ate

:

secto

ral

agriculture

- E

≠ Y

;

industr

ial -

LR

: E

≠ Y

;

SR

: E

->Y

; dom

estic a

nd

com

merc

ial -

LR

and

SR

: Y

->E

Co

al

Govin

dara

ju a

nd

1965-2

009

Tim

eB

ayer

and H

anck

Coal consum

ption;

LR

: E

≠ Y

; S

R: Y

->E

Tan

g (

20

13

)se

rie

s/a

gg

reg

ate

(2009)

coin

tegra

tion

real G

DP

per

capita

test;

Gra

ng

er

ca

usa

lity

Nu

cle

ar

en

erg

y

Akh

ma

t a

nd

19

75

-20

10

Tim

eG

ran

ge

r ca

usa

lity

Coal consum

ption;

LR

: Y

->E

Zam

an (

2013)

series/a

ggre

gate

real G

DP

per

capita

- P

RG

DP

; a

nd

dis

ag

gre

ga

te -

agriculture

GD

P

(AG

DP

)

Wold

e-R

ufa

el

1969-2

006

Tim

eA

RD

L b

ounds tests

;N

ucle

ar

energ

y; R

GD

PL

R:

E->

Y

(20

10

)se

rie

s/a

gg

reg

ate

Toda a

nd Y

am

am

oto

per

capita; re

al gro

ss

(19

95

)fixe

d c

ap

ita

l fo

rma

tio

n


29

Tab

le 1

. (

continued)

Stu

dy

Sam

ple

Data

Tech

niq

ue

Va

riab

les

Re

su

lt

Oil

Akh

ma

t a

nd

19

75

-20

10

Tim

eG

ran

ge

r ca

usa

lity

Oil

consum

ption

LR

: Y

≠ E

Za

ma

n (

20

13

)se

rie

s/a

gg

reg

ate

-

per

capita G

DP

(PG

DP

); a

nd

dis

ag

gre

ga

te -

agriculture

GD

P

(AG

DP

)

Gas

Akh

ma

t a

nd

19

75

-20

10

Tim

eG

ran

ge

r ca

usa

lity

Gas c

onsum

ption

LR

: E

->Y

Za

ma

n (

20

13

)se

rie

s/a

gg

reg

ate

-

per

capita G

DP

(PG

DP

); a

nd

dis

ag

gre

ga

te -

agriculture

GD

P

(AG

DP

)

Co

mb

inati

on

of

dif

fere

nt

en

erg

y s

ou

rces

Bild

iric

i and

1980-2

011

Tim

eA

RD

L (

auto

regre

ssiv

eC

oa

l, n

atu

ral g

as a

nd

oil

LR

: E

<->

Y (

for

coal and

Ba

kirta

s (

20

14

)se

rie

s/a

gg

reg

ate

dis

trib

ute

d la

g b

ou

nd

s)

consum

ption; R

GD

Poil)

Na

se

r (2

01

5)

19

65

-20

10

Tim

eJo

ha

nse

n c

oin

teg

ratio

nO

il co

nsu

mp

tio

n,

nu

cle

ar

LR

: E

->Y

se

rie

s/a

gg

reg

ate

techniq

ue

consum

ption; R

GD

P

No

tes:

E,

energ

y c

onsum

ption; Y,

econom

ic g

row

th;

GD

P,

gro

ss d

om

estic p

roduct;

PG

DP,

per

capita g

ross d

om

estic p

roduct;

RG

DP,

rea

l gro

ss d

om

estic p

rod

uct;

LR

, lo

ng r

un; S

R, short

run.


30

apply the multivariate Toda and Yamamoto (1995) approach, which is often employed in

the case of a small sample.

A sectoral perspective on the manufacturing sector of India suggests that the three

dominant and highly energy-intensive manufacturing industries are steel, aluminium and

cement. Dutta and Mukherjee (2010) suggest that unless these sectors innovate in the

way they are using energy, India will lose global competitiveness in related industries.

Innovation in the energy sector of India is also necessary because of the impact of

oil and gas energy consumption on CO2 emissions. Ahmad and others (2016) find that

energy consumption from oil and gas, electricity and coal consumption contributes to

carbon emissions in India. The question of energy consumption in the country as

a determinant of growth is inevitably intertwined with the issue of raising CO2 emissions.

A series of papers that examine various scenarios for future energy consumption

indicate that none of the traditional sources of energy, oil, gas, coal, hydrocarbon,

nuclear, hydrogen, hydro and renewables, will be sufficient to meet the future energy

demands and that India would have to rely on imports for a significant portion of its

energy supply (Parikh and others, 2009; Parikh and Parikh, 2011). At the same time, the

most feasible scenario for CO2 emissions reduction is to cut energy demand and boost

energy efficiency in production and consumption. That would make it possible to meet

environmental conservation goals without compromising on economic development and

future growth (Parikh and Parikh, 2011).

While the overall energy consumption of the country is estimated to rise sharply in

the next decade, energy inequalities in the country are rampant. Saxena and

Bhattacharya (2018) examine the role of caste, tribe, and religion as determinants of

energy inequality in India. Using data at the household level for 2011-2012, the authors

estimate the energy inequalities stemming from differential access to liquid petroleum

gas and electricity, focusing on disadvantaged groups, such as castes, tribes, and

religious denominations, and find that these factors are relevant to energy access. Even

though the above-mentioned social inequalities in energy access exist, residential

energy consumption in India is expected to quadruple in the next decade because of

lifestyle changes related to the county’s recent economic growth (Bhattacharyya, 2015).

Urbanization, a fast-growing middle class and western-style consumerism are factors

behind the expected overbearing residential energy consumption expansion in the near

future. A large part of the energy supply burden on liquefied petroleum gas is expected

to fall (Bhattacharyya, 2015). This makes the unveiling of the link between petroleum

consumption and economic growth in the context of India even more pressing.

The expected rapid growth in energy consumption, in conjunction with the above

described energy inequalities and contribution to carbon emissions, make India a prime

candidate for the development of renewable energy technologies (Singh, 2018). In

addition to coping with the energy deficits, transitioning to renewables would reduce the

exposure of India to variations in the price of crude oil. A recent study by Mallick,


31

Mahalik and Sahoo (2018) finds that crude oil price reduces significantly private

investment, whereas economic growth and globalization tend to boost it. Economic

growth and urbanization are the key factors pushing energy demand higher in the long

run, Shahbaz and others (2016) argue that transitioning to renewables would allow for

supporting raising energy demand without the negative side effects on pollution and of

energy access inequality in India.

III. DATA

Our study covers 23 Indian states,4 which in total encompasses approximately

95 per cent of the national area. We collected the petroleum consumption and its

by-products consumption data for the states from the States of India database,

a comprehensive compilation of state-level statistics published by the Centre for

Monitoring Indian Economy. The only problem with this is related to the state-wise

population data for each year spanning from 1985/86 to 2013/14. The petroleum

product-wise data referred to in each state over the sample period are available in the

absolute value (in thousand tonnes). Therefore, in order to convert the data to per capita

term, we have collected state-wise population data from the Economic and Political

Weekly Research Foundation database for the same period and then divided the

aggregate petroleum consumption and the various by-products by the population for

each state. Furthermore, we note that this is an unbalanced panel data, as there are

missing observations for a number of states. All of the per capita variables (petroleum

products and the by-products consumption) that we converted are in kilograms. For the

by-products of the petroleum data not available for some states for different years, the

per capita term becomes zero for those observations.

State-wise income per capita is defined as real per capita net state domestic

product at factor cost data, with a base year of 2004/05 and is sourced from the Reserve

Bank of India.5 We divided these 23 states into three panels based on their level of

income. For this classification, we calculated the average per capita income of each

state over the study period 1985-2013 and categorized the states by high, middle, and

low income, presented in table 2.6

4 Andhra Pradesh, Arunachal Pradesh, Assam, Bihar, Delhi, Gujarat, Haryana, Himachal Pradesh, Jammuand Kashmir, Karnataka, Kerala, Madhya Pradesh, Maharashtra, Manipur, Meghalaya, Nagaland,Odisha, Punjab, Rajasthan, Tamil Nadu, Tripura, Uttar Pradesh, and West Bengal.

5 Real gross domestic product (RGDP) data are extracted from Indiastat. Available at Indiastat.com.6 Our classification of the Indian states by income closely follows Narayan, Rath and Narayan (2012) for at

least 15 states.


32

Table 2. Panels by income

High-income states Middle-income states Low-income states

States Delhi, Gujarat, Haryana, Andhra Pradesh, Assam, Bihar,

Maharashtra, Punjab, Arunachal Pradesh, Madhya Pradesh,

Tamil Nadu Himachal Pradesh, Manipur, Meghalaya,

Jammu and Kashmir, Odisha, Rajasthan,

Karnataka, Kerala, Uttar Pradesh,

Nagaland, Tripura,

West Bengal

Table 3. Descriptive statistics

All states High-income states Middle-income states Low-income states

PEC PRGDP PEC PRGDP PEC PRGDP PEC PRGDP

Mean 92.4 24 534.1 173.1 36 996.8 71.2 24 450.8 55.7 15 280.8

Median 70.7 20 711.0 159.6 30 808.2 62.3 22 376.9 48.8 14 333.0

Maximum 399.3 118 411.0 399.3 118 411.0 189.5 58 961.0 159.3 37 154.0

Minimum 18.8 2 728.0 72.8 12 736.7 18.8 8 275.4 24.4 2 728.0

Std. dev. 63.6 15 109.5 63.0 19 815.8 33.1 10 456.4 26.1 6 170.1

Skewness 1.5 2.1 0.9 1.6 0.8 0.9 1.9 0.5

Kurtosis 5.5 10.3 3.8 6.2 3.3 3.3 7.0 3.9

Jarque-Bera 437.5* 1 984.9* 29.2* 150.0* 30.9* 35.4* 295.8* 17.8*

Observations 667 667 174 174 261 261 232 232

Notes: *Normality is rejected at the 1 per cent level. The mean values of the per capita real GDP (PRGDP) are in

Indian rupees while petroleum is measured in terms kg per capita; PEC, per capita energy consumption.

The preliminary observations indicate a strong positive correlation between income

and energy consumption, at least in the average data in per capita terms. In table 3, we

display the average per capita income and per capita energy consumption. Note that for

the high-income states, which are also the most industrially developed ones (Delhi,

Gujarat, Haryana, Maharashtra, Punjab, and Tamil Nadu) the average per capita income

is 36,997 Indian rupee (Rs) (US$537) and their average petroleum consumption stands

at 173 kg of oil equivalent per capita, which is also the highest. The middle-income

states (Andhra Pradesh, Arunachal Pradesh, Himachal Pradesh, Jammu and Kashmir,

Karnataka, Kerala, Nagaland, Tripura, and West Bengal) have an average per capita

income of Rs24,451 and petroleum consumption is the second largest on average

at 71.2 kg of oil equivalent per capita. The low-income states (Assam, Bihar, Madhya

Pradesh, Manipur, Meghalaya, Odisha, Rajasthan, and Uttar Pradesh) on average

show a per capita income of Rs15,281 and consume the least amount of petroleum


33

(56 kg of oil equivalent per capita) in comparison to the other two income groups (see

figure 1).

In the figure, we display energy consumption and real gross domestic product

(RGDP) in per capita terms. For the high-income states (with the exception of Delhi),

per capita RGDP is closely tracked by petroleum consumption per capita and thus this

relationship seems to be positive. We find a similar pattern for middle- and low-income

panels, with the exception of a few states. For instance, for the middle-income states,

including Arunachal Pradesh, Nagaland, Kerala, and West Bengal, and more recently

Jammu and Kashmir, the plots show a decline in petroleum consumption amid steady

growth in income per capita. Of the low-income states, for Bihar, an agriculture-based

state and the third largest in terms of population, a significant decline in petroleum

consumption per capita in the 2000s is shown even though per capita income has been

increasing steadily. For other low-income states, including Assam, Madhya Pradesh,

Manipur, and Uttar Pradesh, similar relationships are shown on a year-to-year basis,

although the long-term trend is upward.


34

Figure 1. Per capita energy consumption

and real gross domestic product by state

High-income states

Gujarat

320

280

240

200

160

1201985 1990 1995 2000 2005 2010

80 000

60 000

40 000

20 000

0

160

140

120

1001985 1990 1995 2000 2005 2010

Maharashtra80 000

60 000

40 000

20 000

0

1985 1990 1995 2000 2005 2010

Tamil Nadu

200

120

80

40

160

80 000

60 000

40 000

20 000

0

Per capita energy consumption (left-hand side) Per capita gross domestic product (right-hand side)

1985 1990 1995 2000 2005 2010

1985 1990 1995 2000 2005 2010

1985 1990 1995 2000 2005 2010

300

280

260

240

220

200

240

200

160

120

80

500

400

300

200

100

0

80 000

60 000

40 000

20 000

0

50 000

40 000

30 000

20 000

120 000

100 000

80 000

60 000

40 000

20 000

Delhi

Punjab

Haryana


35

Middle-income states

Andhra Pradesh

Jammu and Kashmir

Nagaland Tripura

Karnataka

Arunachal Pradesh

West Bengal

Kerala

Himachal Pradesh



36

Assam

Meghalaya

Uttar Pradesh Bihar

Madhya Pradesh

Manipur

RajasthanOdisha


Low-income states


37

IV. EMPIRICAL METHODS

Our models for long-run inferences are as follows:

LPECi,t = α1i + δ1it + β1LPGDPi,t + ε1,it (1)

LPGDPi,t = α2 + β2LPECi,t + ε2,it (2)

where i = 1,...,N for each country in the panel and t = 1,...,T refers to the time period.

The parameters αi and δi allow for country-specific fixed effects and deterministic

trends, respectively. Deviations from the long-run equilibrium relationship are

represented by the estimated residuals, εit, LPEC and LPGDP are petroleum

consumption per capita and economic growth per capita, respectively, expressed in log

form.

Our estimation of short-run models consists of two steps. The first step relates to

the estimation of the residual from the long-run relationship as in equations (1) and (2).

Incorporating the residual as a right-hand side variable, the short-run error correction

model is estimated at the second step. We then get the dynamic error correction model

of our interest for estimation. Specifically, causality (short-run) inferences are made by

estimating the parameters of the following VECM equations.

DLPEC = α3 + ΣK=1β31kDLPECt–k + ΣK=1 β32kDLPGDPt–k + β33Z3,t–1 + ε3,it (3)

DLPGDP = α4 + ΣK=1β41kDLPECt–k + ΣK=1 β42kDLPGDPt–k + β43Z4,t–1 + ε4,it (4)

where DLPEC and DLPGDP denote petroleum consumption per capita and economic

growth per capita, expressed in log-first-difference form and Z3,t–1 and Z4,t–1 are the

error correction terms which are the lagged residual series of the cointegrating

vector (1) and (2), respectively.

From equation (4), the null hypothesis that LPEC does not Granger-cause LPGDP

is rejected, therefore supporting the growth hypothesis, if the set of estimated

coefficients on the lagged values of LPEC is jointly significant. Furthermore, in instances

where LPEC appears in the cointegrating relationship, the growth hypothesis is also

supported if the coefficient of the lagged error correction term is significant. Changes in

an independent variable may be interpreted as representing the short-run causal impact,

while the error correction term provides the adjustment of LPEC and LPGDP towards

their respective long-run equilibrium. The vector error correction model (VECM)

representation, therefore, allows us to differentiate between the short- and long-run

dynamic relationships.

Models (1), (2), (3) and (4) are estimated using the feasible generalized least

squares (FGLS). In cross-sectional analysis, the error variance is likely to vary across

the groups affecting the consistency of the estimators. Using the generalized least

m m

m m


38

squares (GLS) method in the estimation could solve this issue. The proposed analysis

nested within the GLS model can be stated as the following:

Yit = α + X'itβ + δi + γi + εit (5)

where i = 1,N, t = 1,T,Y is a dependent variable (LPEC or LRGDP), α is a constant, X is

a vector of explanatory variables, β represents a vector of coefficients to be estimated,

εit represents the residual terms, δi and γi are the cross-section and, respectively period

fixed or random effects, the GLS estimator is based on the following moments:

g(β) = Σi=1gi (β) = Σi=1Z'i Ω εi (β) (6)

where Z'i is the instrument matrix for the i-th cross-section, εi (β) = (Yit – α – X'

itβ) and Ωis a consistent estimation of the variance-covariance matrix Ω. In cross-sectional

analysis, the error variance may vary across the groups, affecting the consistency of the

estimators. GLS in the estimation can solve this issue, although other sources of

variance variability may still exist.

To explore the FGLS model with the best fitted error process for the data, we test

for heteroskedasticity using the modified Wald test proposed by Greene (2008). This has

a null hypothesis in that there is homoskedasticity in the error term. The results reported

in table 4 confirm the rejection of this null hypothesis at a 1 per cent significance level

M M

Table 4. Evidence of heteroskedasticity

DPEC = f(DPGDP)

Test name Error process Test (1) (2) (3) (4)

statistic All states High-income Middle-income Low-income

states states states

Modified Heteroskedasticity Chi(2) 720.92*** 194.01*** 115.84*** 378.32***

DPGDP = f(DPEC)

Test name Error process Test (1) (2) (3) (4)

statistic All states High-income Middle-income Low-income


Modified Heteroskedasticity Chi(2) 5 942.32*** 531.06*** 506.85*** 1 680.17***

Notes: The modified Wald statistic for group-wise heteroskedasticity in the residuals of a fixed effect model is

calculated following Greene (2008, p. 598). The most likely deviation from homoskedastic errors in the context

of pooled cross-section time-series data (or panel data) is likely to be error variances specific to the cross-

sectional unit. xttest3 tests the hypothesis that H0: sigma(i)^2 = sigma^2 for all i, N_g, where N_g is the

number of cross-sectional units. The resulting test statistic is distributed Chi-squared(N_g) under the null

hypothesis of homoskedasticity. ***, ** and * indicate rejection of the null hypothesis at 1 per cent, 5 per cent

and 10 per cent significance levels.

^

^ –1


39

for all the panels, including those with the dependent variables as petroleum

consumption per capita (PEC) and as economic growth per capita (PGDP).

Next, we apply the Pesaran (2004) test that examines the null hypothesis of cross-

sectional independence for the PEC and PGDP models (Pesaran, Ullah and Yamagata,

2008). We present the cross-sectional dependence statistics for the PEC and PGDP

models, respectively, in panels 1 and 2 in table 5. The hypothesis that the innovations

relating to energy consumption or economic growth equations are cross-sectionally

independent is rejected for all panels. Not surprisingly, the all states panel shows the

greatest cross-sectional dependence. This is followed by the middle-income states in

panel 1 and high-income states in panel 2. On the basis of this result, we proceed to use

the FGLS model with an error process that assumes heteroskedasticity and panels that

are cross-sectionally dependent. The econometric models were estimated using Stata.

Table 5. Evidence of cross-sectional dependence

Pesaran (2004) Statistic p-value

Panel 1: DPEC = f(DPGDP)

All states 80.02*** 0.0007

High-income states 18.81*** 0.0004

Middle-income states 31.61** 0.0253

Low-income states 29.4*** 0.0000

Panel 2: DPGDP = f(DPEC)

All states 27.59*** 0.0007

High-income states 15.26*** 0.0004

Middle-income states 3.496** 0.0253

Low-income states 2.468*** 0.0000

Notes: The Pesaran (2004) test was applied for the cross-sectional dependence (also see

Pesaran, Ullah and Yamagata, 2008). H0: cross-sectional independence. ***, ** and *

indicate rejection of the null hypothesis at 1 per cent, 5 per cent and 10 per cent

significance levels.

V. EMPIRICAL RESULTS

Panel unit root and cointegration tests and the vector error correction model

The panel unit root tests, namely, Im, Pesaran and Shin (2003); Levin, Lin and Chu

(2002); and panel augmented Dickey-Fuller (ADF) (Maddala and Wu, 1999) are

performed. These tests have the common null hypothesis of unit root. The test results


40

are presented in table 6. Petroleum consumption per capita (PEC) and economic growth

per capita (PGDP), expressed in log form, are integrated of order 1. This applies to all

the panels.

Table 6. Unit root test results

All states

High-income Middle-income Low-income


PEC I(0) I(1) I(0) I(1) I(0) I(1) I(0) I(1)

LLC - t* -1.671 -14.369* -0.845 -6.594* -1.185 -9.315* -0.839 -8.764*

0.047 0.000 0.199 0.000 0.118 0.000 0.201 0.000

IPS - W-stat. 1.490 -13.390* -0.248 -6.325* 1.592 -8.557* 1.052 -8.151*

0.932 0.000 0.402 0.000 0.944 0.000 0.854 0.000

ADF - Fisher Chi-square 32.208 254.551* 15.555 60.761* 7.449 101.843* 9.204 91.948*

0.939 0.000 0.213 0.000 0.986 0.000 0.905 0.000

PGDP I(0) I(1) I(0) I(1) I(0) I(1) I(0) I(1)

LLC - t* 5.750 -10.385* 2.856 -4.699* 3.749 -9.610* 4.351 -2.741*

1.000 0.000 0.998 0.000 1.000 0.000 1.000 0.003

IPS - W-stat. 11.652 -12.868* 6.024 -6.121* 7.359 -9.071* 6.736 -6.896*

1.000 0.000 1.000 0.000 1.000 0.000 1.000 0.000

ADF - Fisher Chi-square 1.550 245.771* 0.183 59.088* 0.847 108.921* 0.520 77.762*

1.000 0.000 1.000 0.000 1.000 0.000 1.000 0.000

Notes: The table covers the Im-Pesaran-Shin (IPS) (Im, Pesaran and Shin, 2003); Levin-Lin-Chu (LLC) (Levin, Lin

and Chu, 2002); and augmented Dickey-Fuller (ADF) (Maddala and Wu, 1999) test results. * suggests

statistical significance at 1 per cent level. PEC is the petroleum consumption in kilogram of oil equivalent per

capita; PGDP is the real per capita net state domestic product at factor cost data with a base year of 2004/05.

As the panels comprise I(1) variables, they all are fit for three panel cointegration

tests: Kao (1999), Pedroni (1999; 2004), and the Fisher type-test from Maddala and Wu

(1999). The test of Pedroni (1999; 2004) is a panel cointegration test that extends the

Engle and Granger method to a system of multivariate independent variables for

homogeneous and heterogeneous properties across individuals for the panel data. The

Kao (1999) test is a residual-based panel test that applies the Dickey-Fuller and

augmented Dickey-Fuller type tests and considers homogeneous properties across

individuals. The Kao (1999) test focuses on both strict endogenous regressors and strict

exogenous regressors.

The Pedroni tests, unlike those of Kao, allow for heterogeneity among individual

units of the panel and no exogeneity requirements are imposed on the regressors in the

cointegrating regressions. The Maddala and Wu (1999) test is a different method that

applies the combination test from Fisher (1932) to derive the test statistics for panel

estimation. The combination statistic is constructed from various individual statistics, this


41

combination statistic follows the Chi-square distribution rule, in which individual test

statistic is computed by Johansen (1988).

Of these tests, the Pedroni (1999; 2004) test allows for cross-sectional

dependence. Such test uses the fully modified ordinary least squares (FMOLS)

estimator that deals with possible autocorrelation and heteroskedasticity of the

residuals, taking into account the presence of nuisance parameters, which is

asymptotically unbiased and deals with potential endogeneity of regressors. As our

panel is burdened by all these three problems, we take this as the superior test of

cointegration.

The results from the three cointegration tests are captured in table 7, panels 1-3.

Pedroni test results (panel 1) suggest at least one cointegrating relationship for all

panels. When compared against the Kao and Fisher test results, we find that the results

for all Indian states and the middle- and low-income states are the same.7

The relationship between petroleum and economic growth within the long-run

models and vector error correction models (VECMs)

Next, we estimate the long-run models and VECMs for the all states and income-

based panels. This approach differs from the literature on the long run and VECM in that

we estimate the long run and VECM nested within the FGLS model relating to petroleum

consumption and economic growth. The long-run results are presented in table 8. The

influence of income on petroleum consumption on per capita is examined in panel 1 and

the impact of petroleum consumption on per capita real income is examined in panel 2.

In the long run, we see signs of a feedback effect for the Indian states at the higher end

of the income spectrum. In this regard, our findings are consistent with only two out of

16 studies on energy-economic growth that support the feedback hypothesis.

Per capita real income is found to have a positive and significant influence on

petroleum consumption for all the states in the long run (table 8, panel 1). Petroleum

consumption positively affects per capita income of the high-income states (table 8,

panel 2). However, for the all states panel, and the middle- and low-income Indian

states, we find that petroleum consumption reduces per capita real income in the long

run. Hence, while the bilateral link exists between the two variables, it is clear that we

fail to find evidence on the feedback hypothesis in its true form.

7 Before the estimation, we conduct the Di Iorio and Fachin (2007) test for breaks in cointegrated panelsto examine the stability of the relationship between our variables of interest. The results support theacceptance of the null hypothesis of no break. That is, the relationship among the investigated variablesis stable and not subject to structural breaks during the investigation period. The results are notpresented here to conserve space, but they are available upon request.


42

Table 7. Cointegration results

All statesHigh-income Middle-income Low-income


Panel 1: Pedroni residual

cointegration test Stat. Prob. Stat. Prob. Stat. Prob. Stat. Prob.

Panel v 4.242* 0.000 2.665* 0.004 2.347* 0.010 2.431* 0.008

Panel rho -4.905* 0.000 -2.099* 0.018 -1.723* 0.043 -4.251* 0.000

Panel PP -5.616* 0.000 -2.274* 0.012 -2.091* 0.018 -4.734* 0.000

Panel ADF -3.393* 0.000 -1.898* 0.029 -1.987* 0.023 -2.070* 0.019

W. Stat. Prob. W. Stat. Prob. W. Stat. Prob. W. Stat. Prob.

Panel v 3.686* 0.000 2.080* 0.019 2.596* 0.005 1.754* 0.040

Panel rho -4.173* 0.000 -1.841* 0.033 -2.077* 0.019 -3.178* 0.001

Panel PP -5.389* 0.000 -2.341* 0.010 -2.648* 0.004 -4.147* 0.000

Panel ADF -3.476* 0.000 -1.590* 0.056 -2.531* 0.006 -1.866* 0.031

Stat. Prob. Stat. Prob. Stat. Prob. Stat. Prob.

Group rho -2.056* 0.020 -0.621* 0.267 -0.577 0.282 -2.336* 0.010

Group PP -4.940* 0.000 -1.942* 0.026 -2.215* 0.013 -4.344* 0.000

Group ADF -3.070* 0.001 -1.087 0.138 -2.083* 0.019 -2.054* 0.020

Panel 2: Kao residual

cointegration test t-Stat. Prob. t-Stat. Prob. t-Stat. Prob. t-Stat. Prob.

ADF -1.643* 0.050 -0.327 0.372 -2.579* 0.005 -0.262 0.397

Panel 3: Fisher statistics Trace Prob. Trace Prob. Trace Prob. Trace Prob.

test test test test

None 87.130* 0.000 15.740 0.204 34.450* 0.011 36.950* 0.002

At most 1 53.890 0.198 12.990 0.370 21.160 0.271 19.740 0.232

Max-eigen Prob. Max-eigen Prob. Max-eigen Prob. Max-eigen Prob.

test test test test

None 81.740* 0.001 15.320 0.225 32.050* 0.022 34.360* 0.005

At most 1 53.890 0.198 12.990 0.370 21.160 0.271 19.740 0.232

Notes: The table presents the results from three cointegration tests: Pedroni, Kao, and Fisher. For the Pedroni test,

the first eight statistics refer to homogenous test – the alternative hypothesis: common AR coefficients (within-

dimension) while the last three statistics refer to heterogeneous test with the alternative hypothesis: individual

AR coefficients (between-dimension). * suggests statistical significance at the 1 per cent level.


43

Next, we report the results on VECMs selected using the usual selection criteria

between models with one to six lags. The VECM results relating to per capita petroleum

consumption and economic growth models are presented, respectively, in tables 8

and 9.

The key findings are as follows. First, the error correction model (ECM) has the

expected negative sign and is significant for all the models with petroleum consumption

(or economic growth) as the dependent variable. The implications are twofold. First,

there is a two-way long-run relationship, or a feedback effect, between economic growth

and petroleum consumption, as suggested by the preliminary observations. Second,

after a shock related to economic growth (petroleum consumption), petroleum

consumption (economic growth) bounces back towards equilibrium.

Furthermore, the VECM results point towards a bidirectional association between

economic growth and petroleum consumption in the short run for all the panels, except

the all states panel. For the high-income Indian states, the feedback hypothesis in its

true form is found for the short run as well. This implies that higher petroleum

consumption predicts higher economic growth, and in return past economic growth

encourages petroleum consumption in the following year. However, for the middle-

income states, while a previous year’s economic growth is a precursor for a positive

change in petroleum consumption in the following year, a previous year’s increase in

petroleum consumption does not mean higher economic growth in the following year.

Table 8. Long-run models

(1)(2) (4) (5)

All statesHigh-income Middle-income Low-income


Panel 1:LPEC = f(LPGDP)

LPGDP 0.812*** 0.556*** 0.650*** 0.550***

(0.028) (0.036) (0.057) (0.038)

Observations 667 174 261 232

Number of crossid 23 6 9 8

Panel 2: LPGDP = f(LPEC)

LPEC -0.682*** 1.030*** -0.511*** -0.867***

(0.024) (0.067) (0.045) (0.06)

Observations 667 174 261 232

Number of crossid 23 6 9 8

Notes: Using the feasible generalized least squares (FGLS) methodology, we estimate the long-run relationship

between petroleum consumption and economic growth. Standard errors are reported in the parentheses. ***,

** and * indicate rejection of the null hypothesis at 1 per cent, 5 per cent and 10 per cent significance levels.


44

Table 9. State-wise economic growth and petroleum consumption:

feasible generalized least squares (FGLS) results

Dependent variable: Dependent variable:

(1) (2) (3) (4) (1) (2) (3) (4)

Variables All States High- Middle- Low- All States High- Middle- Low-

income income income income income income

States States States States States States

DLPGDPt–1 -0.0471 0.0348*** 0.366*** -0.140 -0.187*** -0.244** -0.00776 -0.295***

(0.0665) (0.0114) (0.121) (0.0893) (0.0440) (0.0774) (0.0634) (0.0707)

DLPGDPt–2 -0.0217 -0.234** 0.121*** 0.0828

(0.0683) (0.0953) (0.0453) (0.0760)

DLPGDPt–3 0.245*** 0.120 0.0457 0.0567

(0.0650) (0.0939) (0.0430) (0.0749)

DLPGDPt–4 0.112* 0.141***

(0.0650) (0.0430)

DLPGDPt–5 0.0930 -0.0373

(0.0641) (0.0423)

DLPECt–1 -0.0972** 0.0760 0.0685 -0.234*** -0.00431 0.0299*** -0.0182*** -0.0288***

(0.0425) (0.0780) (0.0622) (0.0719) (0.0277) (0.0052) (0.00322) (0.00548)

DLPECt–1 -0.0657 -0.130* 0.0206 0.0162

(0.0423) (0.0716) (0.0276) (0.0553)

DLPECt–1 0.00278 -0.0315 -0.0393 -0.0367

(0.0412) (0.0690) (0.0270) (0.0540)

DLPECt–1 -0.0288 -0.0347

(0.0417) (0.0273)

DLPECt–1 0.0561 -0.0358

(0.0420) (0.0275)

ECMt–1 -0.0213** -0.0624** -0.0256** -0.0189** -0.0950*** -0.0258** -0.0333*** -0.0205***

(0.00930) (0.0286) (0.0129) (0.0053) (0.00500) (0.01433) (0.0076) (0.0015)

Observations 529 162 243 200 529 162 243 200

No. of crossid 23 6 9 8 23 6 9 8

Notes: Using the feasible generalized least squares (FGLS) methodology, we estimate the short-run relationship

between petroleum consumption and economic growth. Lag length selection for each panel is based on

Akaike information criterion (AIC) and Bayesian information criterion (BIC). ***, ** and * indicate rejection of

the null hypothesis at 1 per cent, 5 per cent and 10 per cent significance levels. Standard errors are reported

in the parentheses.

In addition, for the low-income states, higher growth in previous years predicts reduced

demand for petroleum consumption. What is puzzling is that higher petroleum

consumption predicts a fall in the short-term real income growth. Unsurprisingly, for the

all states panel, we find an unidirectional link in the short run, with the effect running

from economic growth to petroleum consumption. This supports the prevalence of the

conservative hypothesis for the short run. The finding suggests that a reduction in the


45

use of petroleum and a switch to cleaner and cheaper alternatives will not harm

economic growth.

VI. THE ENERGY CONSUMPTION AND ECONOMIC GROWTH (E-Y)

CONNECTIONS WITH DISAGGREGATED PETROLEUM

We examine the relationship between state-wise data on petroleum consumption

and income using the disaggregated data on petroleum consumption by state. We

classified the different types of petroleum consumption into six energy sources:

(a) liquefied petroleum gas (LPG); (b) petrol (PET); (c) superior kerosene oil (SKO);

(d) diesel/high speed diesel (HSD); (e) furnace oil (FO); and (f) naptha; aviation turbine

fuel; light diesel oil; low sulphur heavy stock/hot heavy stock; lubes and greases;

itumen; others (OTHERS). The disaggregated petroleum consumption data are sourced

from the States of India database. The disaggregated petroleum consumption data are

converted into per capita terms using population data on the Indian states attained from

the Economic and Political Weekly Research Foundation database. We conducted the

same tests for the aggregate data and the disaggregated data. The results for the

disaggregated data are presented in the appendix.

We begin with the descriptive statistics in appendix table A.1. Notice that, with the

exception of HSD, the petroleum disaggregates vary in terms of importance for each

state. Out of all petroleum products, the average consumption of HSD is consistently the

strongest type of consumption in all states. In the high-income states, the consumption

of HSD is followed by PET, SKO, LPG, and FO. In the middle-income states, HSD

consumption is trailed by SKO, PET, LPG, and FO. In the low-income states,

consumption of SKO, PET, LPG, and FO are, on average, lower than that of HSD.

The unit root tests of the disaggregated petroleum data are presented in appendix

table A.2. As the disaggregated petroleum types are found to be stationary at I(1), we

proceed with the cointegration tests. The cointegration test results indicate rather limited

cases of cointegration between the disaggregated petroleum types and economic

growth. The full sample, comprising of all the Indian states, indicates that petroleum

disaggegates SKO and OTHERS, possibly having a stable long-run association with

income (appendix table A.3). For the high-income states panel, none of the petroleum

types are cointegrated with the state income (appendix table A.4). For the middle-

income Indian states panel, PET, LPG, and OTHERS may have stable long-run relations

with income (appendix table A.5). For the low-income states panel, only LPG has

a possible cointegration link with income (appendix table A.6).

The causal relationships and the direction of the causation between these

cointegrated relationships are examined using VECMs (appendix table A.7). Estimation

methods were similar to those discussed in the previous sections. For VECM, when the


46

state-wise economic growth is the dependent variable, we find VECM to be valid in two

instances — the link between LPG and economic growth of the middle-income and

low-income states (appendix table A.7, panel 1). The long-run linkage between these

variables are positive and significant (appendix table A.8, panel 2). This means that LPG

has a positive effect on income of the middle-income states and low-income states.

Returning to VECMs, when different petroleum types are alternated as dependent

variables, all cointegrated relations produce valid VECMs (appendix table A.7, panel 2).

These findings imply that LPG and economic growth of the low-income states have

a bidirectional or a feedback relationship. However, the rest of the valid relationships

discussed here satisfy the conservative hypothesis. In the conservative hypothesis,

economic growth is a good predictor of use of petroleum disaggegates, namely, SKO,

and OTHERS (for the full sample); PET (for the middle-income sample); and LPG (for

the low-income sample).

While in the long run economic growth is predicted to have a positive effect on the

disaggregated energy consumption, in the short run economic growth is found to reduce

consumption of SKO (for the all states panel) and LPG (for the low-income states

panel).

VII. FURTHER DISCUSSIONS

This study shows different results regarding the nexus between energy

consumption and economic growth across the 23 selected Indian states grouped in

different panels based on their income level. This suggests that an appropriate approach

for India should be to adopt state-specific policies in lieu of an integrated policy for all

states.

For the high-income (and most industrialized) states of India, we find a prevalence

of the feedback effect in the long and short run using aggregate petroleum data. This

finding implies that energy supply shock may have a significant impact on economic

growth (and vice versa). As such, adopting a general energy conservation policy may

have a detrimental impact on the economic growth process in high-income states in

India. Energy policy targeted towards higher petroleum usage is critical for the economic

growth of these states. In this regard, it is suggested that the Government of India

encourages the use and development of more advance and eco-friendly technologies by

providing an array of energy tax credits as incentives for use of alternative energy

resources. By so doing, it can minimize the energy supply shock effect on the output

and reduce the unfavourable effects on the environment.


47

The Government of India has achieved significant milestones in building nuclear

power plants. For instance, the Russian Federation-backed 2,000 megawatt

Kudankulam Nuclear Power Plant in Tamil Nadu was completed in 2013; it has become

the single largest nuclear power station in India. In addition, India also signed the Civil

Nuclear Cooperation Agreement with the United States in 2008. This initiative is

expected to foster the growth of the country’s civil nuclear sector and consequently

enhance its energy security. India would greatly benefit from a stable clean energy

source for its large and rapidly growing economy, which also would have favourable

environmental effects. Our use of disaggregated data indicates insignificant effects of

short-term and long-term linkages between petroleum and economic growth. This

suggests that the use of aggregate data is more appropriate for modelling the linkages

between petroleum consumption and income in high-income states.

For the middle- and low-income states, we are unable to find a feedback effect

between petroleum consumption and economic growth in the aggregate data. For the

middle-income states, economic growth is able to predict higher petroleum consumption

but past increases in petroleum consumption does not predict future economic growth.

We find this to be the case in the short run and in the long run. However, when we

consider disaggregated petroleum consumption data, we find that LPG and economic

growth show the feedback effect.

For the low-income state panel, in the long run, economic growth increases

aggregate petroleum consumption, but increased aggregate petroleum consumption

reduces economic growth. In the short run, economic growth reduces petroleum

demand and lower petroleum consumption translates into higher economic growth. For

the all states panel, there is a prevalence of the unidirectional link, with the effect

running from economic growth to aggregate petroleum consumption. This supports the

conservative hypothesis for the short run. These findings suggest that a reduction in the

use of petroleum and switching to cleaner and cheaper alternatives (here, abundant and

cheap labour should not be ruled out) will not harm economic growth. In fact, in the case

of low- (and middle-) income states, economic growth is encouraged, with a reduction in

petroleum usage. Our study of the disaggregated petroleum consumption suggests that

petroleum products relating to superior kerosene oil and others are also influenced by

economic growth.

While our analysis gives strong support for the feedback hypothesis for the richer

states of India, our results also show two points of interest to policymakers: (i) petroleum

is affecting growth negatively in the middle- and low-income states in India; and

(ii) economic growth can be promoted even with lower petroleum consumption. These

results have not been observed in the Indian literature or any other study to date.


48

VIII. CONCLUDING REMARKS

We examined the energy consumption and economic growth (E-Y) nexus for a

panel of 23 Indian states and the subpanels of these Indian states classified by high,

middle, and low income on the basis of their average per capita real GDP over the

period 1985-2013. Upon finding the presence of cross-sectional dependence in the

panels and heteroskedasticity in the relationships, we use the FGLS methodology to

examine the long-run and short-run relationships.

Our key findings are as follows. For the country’s high-income (and most

industrialized) states, we find evidence of the feedback effect in the long run and the

short run. For the middle- and low-income states, however, we do not find this feedback

effect between petroleum consumption and economic growth in neither the short run nor

the long run. Similarly, for the low-income state panel, in the long run, economic growth

appears to increase petroleum consumption but higher petroleum usage seems to

reduce economic growth. In the short run, we find that economic growth reduces

petroleum demand while lower petroleum consumption leads to higher economic

growth. For the all states panel, there is evidence of the unidirectional effect running

from economic growth to petroleum consumption in the short run. This supports the

prevalence of the conservative hypothesis. These results are also confirmed by using

disaggregated data on petroleum consumption.

Some of the distortions we notice may be because the economies of the middle-

and low-income Indian states have been chiefly informal and therefore statistically

unaccounted for. A large part of agriculture, construction and manufacturing are

comprised of informal sectors that consume petroleum but are largely missing in the

GDP statistics.

At play here could be other features of the poorer states that do not show clear E-Y

linkages. For instance, the informal sectors rely heavily on unskilled labour. We suspect

that increased use of imported and expensive petroleum in place of abundant unskilled

workers is to some degree also leading to a misallocation of resources in these poorer

states. However, exploring this issue is not within the scope of the study. We leave this

as part of a future research agenda.


49

Ap

pe

nd

ix

Ta

ble

A.1

. D

es

cri

pti

ve

sta

tis

tic

s

Th

is t

ab

le p

rovid

es t

he

de

scri

ptive

sta

tistics f

or

the

pe

tro

leu

m t

yp

es (

in lo

g f

orm

): f

urn

ace

oil

(FO

); d

iese

l/h

igh

sp

ee

d

die

se

l (H

SD

); liq

ue

fie

d p

etr

ole

um

ga

s (

LP

G);

Pe

tro

l (P

ET

); s

up

eri

or

ke

rose

ne

oi l

(SK

O);

an

d n

ap

tha

; a

via

tio

n t

urb

ine

fu

el ;

l igh

t d

iese

l o

i l ; lo

w s

ulp

hu

r h

ea

vy s

tock/h

ot

he

avy s

tock;

lub

es a

nd

gre

ase

s;

bi tu

me

n;

oth

ers

(O

TH

ER

S).

Inco

me g

rou

ps

Hig

h in

co

me

Mid

dle

in

co

me

Lo

w in

co

me

Petr

ole

um

types

FO

HS

DLP

GP

ET

SK

OO

TH

ER

SF

OH

SD

LP

GP

ET

SK

OO

TH

ER

SF

OH

SD

LP

GP

ET

SK

OO

TH

ER

S

Mean

2.3

04.0

86

2.3

42.5

22.4

53.6

91.0

83.5

71.7

61.9

72.1

72.3

70.8

83.1

21.0

01.1

61.9

91.9

5

Sta

nd

ard

de

via

tio

n0.9

00.4

33

0.7

30.6

70.5

20.5

61.1

70.4

70.8

60.5

90.2

90.6

81.0

70.4

80.8

10.5

80.2

50.6

1

Co

effic

ien

t o

f va

ria

tio

n0.3

90.1

10.3

10.2

70.2

10.1

51.0

80.1

30.4

90.3

00.1

40.2

91.2

10.1

50.8

00.4

90.1

20.3

1


50

Ta

ble

A.2

. U

nit

ro

ot

tes

t: d

isa

gg

reg

ate

pe

tro

leu

m v

ari

ab

les

Th

is t

ab

le c

ove

rs t

he

Im

, P

esa

ran

an

d S

hin

(2

00

3);

Le

vin

, L

in a

nd

Ch

u (

20

02

); a

nd

AD

F (

Ma

dd

ala

an

d W

u,

19

99

)

test

resu

lts.

F

ull s

am

ple

Inco

me g

rou

p 1

Inco

me g

rou

p 2

Inco

me g

rou

p 3

Va

ria

ble

Me

tho

d I

(0)

I(1

)I(

0)

I(1

)I(

0)

I(1

)I(

0)

I(1

)

Sta

t.P

rob

.S

tat.

Pro

b.

Sta

t.P

rob

.S

tat.

Pro

b.

Sta

t.P

rob

.S

tat.

Pro

b.

Sta

t.P

rob

.S

tat.

Pro

b.

PE

TLevin

, Lin

and C

hu t*

2.8

95

0.9

98

-3.2

61

0.0

01

0.8

97

0.8

15

0.0

17

0.5

07

1.3

07

0.9

05

-3.3

52

0.0

00

2.0

76

0.9

81

-2.0

61

0.0

20

Im

, P

esara

n a

nd S

hin

W-s

tat.

6.2

65

1.0

00

-6.6

34

0.0

00

2.9

06

0.9

98

-3.4

08

0.0

00

3.1

42

0.9

99

-4.3

29

0.0

00

3.4

07

1.0

00

-3.6

99

0.0

00

A

DF

- F

isher

Chi-square

18.0

99

1.0

00

125.1

43

0.0

00

4.6

25

0.9

69

32.1

91

0.0

01

9.0

98

0.6

95

41.5

79

0.0

00

3.5

92

0.9

99

41.2

30

0.0

01

LP

GLevin

, Lin

and C

hu t*

-7.4

45

0.0

00

-6.4

12

0.0

00

-3.1

51

0.0

01

-4.5

63

0.0

00

-3.6

27

0.0

00

-5.4

81

0.0

00

-5.5

71

0.0

00

-0.4

96

0.3

10

Im

, P

esara

n a

nd S

hin

W-s

tat.

-1.8

36

0.0

33

-8.1

51

0.0

00

-0.7

67

0.2

22

-3.6

85

0.0

00

-0.7

03

0.2

41

-4.8

48

0.0

00

-1.6

19

0.0

53

-3.6

36

0.0

00

A

DF

- F

isher

Chi-square

72.2

78

0.0

08

153.1

16

0.0

00

17.9

57

0.1

17

34.9

52

0.0

01

21.6

99

0.0

41

45.3

63

0.0

00

27.1

99

0.0

39

40.6

26

0.0

01

HS

DLevin

, Lin

and C

hu t*

-0.0

60

0.4

76

-4.0

45

0.0

00

0.3

17

0.6

25

-0.7

82

0.2

17

-0.6

50

0.2

58

-2.1

55

0.0

16

0.2

39

0.5

95

-2.7

90

0.0

03

Im

, P

esara

n a

nd S

hin

W-s

tat.

2.7

67

0.9

97

-6.0

57

0.0

00

1.3

99

0.9

19

-2.3

52

0.0

09

1.6

85

0.9

54

-3.5

40

0.0

00

1.4

81

0.9

31

-4.0

32

0.0

00

A

DF

- F

isher

Chi-square

28.9

75

0.9

77

115.8

76

0.0

00

5.4

35

0.9

42

23.6

81

0.0

23

8.2

77

0.7

63

34.0

57

0.0

01

11.3

38

0.7

88

45.4

79

0.0

00

SK

OLevin

, Lin

and C

hu t*

3.6

57

1.0

00

-4.4

81

0.0

00

4.8

90

1.0

00

-1.5

01

0.0

67

1.0

98

0.8

64

-1.7

12

0.0

44

-0.7

70

0.2

21

-2.6

46

0.0

04

Im

, P

esara

n a

nd S

hin

W-s

tat.

3.0

75

0.9

99

-6.1

50

0.0

00

5.2

40

1.0

00

-1.8

25

0.0

34

0.7

05

0.7

60

-2.8

72

0.0

02

0.4

10

0.6

59

-4.2

12

0.0

00

A

DF

- F

isher

Chi-square

32.3

92

0.9

36

118.3

49

0.0

00

0.7

55

1.0

00

19.5

26

0.0

77

8.6

98

0.7

29

28.6

24

0.0

05

13.9

90

0.6

00

46.7

29

0.0

00

FO

Le

vin

, L

in a

nd

Ch

u t

*-0

.906

0.1

82

-6.6

45

0.0

00

-1.9

95

0.0

23

-8.3

18

0.0

00

1.0

98

0.8

64

-1.7

12

0.0

44

-0.7

70

0.2

21

-2.6

46

0.0

04

Im

, P

esara

n a

nd S

hin

W-s

tat.

-0.0

82

0.4

67

-9.3

80

0.0

00

-1.1

01

0.1

35

-8.0

36

0.0

00

0.7

05

0.7

60

-2.8

72

0.0

02

0.4

10

0.6

59

-4.2

12

0.0

00

A

DF

- F

isher

Chi-square

45.0

63

0.2

00

171.7

20

0.0

00

16.8

67

0.1

55

76.9

97

0.0

00

8.6

98

0.7

29

28.6

24

0.0

05

13.9

90

0.6

00

46.7

29

0.0

00

OT

HE

RS

Levin

, Lin

and C

hu t*

0.4

60

0.6

77

-9.8

12

0.0

00

2.0

96

0.9

82

-5.3

70

0.0

00

0.3

94

0.6

53

-5.0

50

0.0

00

-0.9

02

0.1

84

-5.3

70

0.0

00

Im

, P

esara

n a

nd S

hin

W-s

tat.

0.8

70

0.8

08

-12.2

93

0.0

00

0.6

92

0.7

56

-6.0

51

0.0

00

2.0

26

0.9

79

-6.2

91

0.0

00

-0.5

65

0.2

86

-7.1

45

0.0

00

A

DF

- F

isher

Chi-square

36.5

13

0.8

40

231.1

07

0.0

00

11.4

95

0.4

87

58.1

44

0.0

00

3.5

73

0.9

90

60.3

35

0.0

00

15.5

05

0.4

88

78.9

94

0.0

00

No

tes:

PE

T,

petr

ol; L

PG

, liq

uefied p

etr

ole

um

gas;

HS

D,

die

sel/hig

h s

peed d

iesel; S

KO

, superior

kero

sene o

il; F

O,

furn

ace o

il; O

TH

ER

S,

napth

a,

avia

tion t

urb

ine

fuel, lig

ht die

sel oil,

low

sulp

hur

heavy s

tock h

ot

heavy s

tock,

lubes a

nd g

reases,

itum

en,

and o

thers

.


51

Ta

ble

A.3

. C

oin

teg

rati

on

te

st

res

ult

s:

dis

ag

gre

ga

te e

ne

rgy

da

ta (

full

sa

mp

le)

Th

e t

ab

le p

rese

nts

re

su

lts f

rom

th

ree

co

inte

gra

tio

n t

ests

: P

ed

ron

i, K

ao

, a

nd

Fis

he

r . F

or

the

Pe

dro

ni

test,

th

e f

irst

eig

ht

sta

tistics r

efe

r to

ho

mo

ge

ne

ou

s t

est

–-

the

alte

rna

tive

hyp

oth

esis

: co

mm

on

au

tore

gre

ssiv

e (

AR

) co

ef f

icie

nts

(w

ith

in-

dim

en

sio

n)

wh

ile th

e la

st

thre

e sta

tistics re

fer

to h

ete

rog

en

eo

us te

st

with

th

e a

lte

rna

tive

h

yp

oth

esis

: in

div

idu

al

AR

co

effi c

ien

ts (

be

twe

en

-dim

en

sio

n).

Va

ria

ble

Ped

ron

iA

DF

Fis

her

Sta

t.P

rob

.W

. S

tat.

Pro

b.

Sta

t.P

rob

.

Tra

ce

Pro

b.

Ma

x-e

igen

Pro

b.

PE

TP

anel v

-1.5

60

0.9

41

-0.3

38

0.6

32

-0.6

16

0.2

69

None

92.0

70

0.0

00

78.5

40

0.0

02

P

anel rh

o1.1

72

0.8

79

-0.4

60

0.3

23

At m

ost 1

73.0

60

0.0

07

73.0

60

0.0

07

P

anel P

P0.6

91

0.7

55

-1.3

99

0.0

81

P

anel A

DF

2.0

86

0.9

82

0.2

69

0.6

06

G

roup r

ho

-1.2

67

0.1

03

G

rou

p P

P-2

.85

50

.00

2

G

roup A

DF

-1.5

21

0.0

64

LP

GP

anel v

0.1

35

0.4

46

0.5

26

0.3

00

-4.0

18

0.0

00

None

57.4

00

0.1

21

62.5

80

0.0

52

Panel rh

o-0

.260

0.3

98

-0.3

08

0.3

79

At m

ost 1

25.4

50

0.9

94

25.4

50

0.9

94

P

anel P

P-2

.135

0.0

16

-1.8

56

0.0

32

P

anel A

DF

-0.2

82

0.3

89

0.0

56

0.5

22

G

roup r

ho

1.2

78

0.8

99

G

rou

p P

P-1

.34

90

.08

9

Gro

up A

DF

1.0

86

0.8

61

HS

DP

anel v

2.8

85

0.0

02

3.0

03

0.0

01

-1.4

83

0.0

69

None

59.5

10

0.0

87

55.5

90

0.1

57

P

anel rh

o-2

.219

0.0

13

-2.4

33

0.0

08

At m

ost 1

53.1

10

0.2

19

53.1

10

0.2

19

P

anel P

P-3

.401

0.0

00

-3.5

83

0.0

00

P

anel A

DF

-0.9

44

0.1

73

-0.9

16

0.1

80

G

roup r

ho

-0.8

54

0.1

97

G

rou

p P

P-3

.09

40

.00

1

G

roup A

DF

-0.5

88

0.2

78


52

SK

OP

anel v

2.7

23

0.0

03

0.1

32

0.4

48

-1.3

31

0.0

92

None

113.3

00

0.0

00

93.1

70

0.0

00

P

anel rh

o1.2

41

0.8

93

0.4

33

0.6

68

At m

ost 1

88.9

20

0.0

00

88.9

20

0.0

00

P

anel P

P2.2

92

0.9

89

-0.6

07

0.2

72

P

anel A

DF

-1.0

07

0.1

57

-1.9

34

0.0

27

G

roup r

ho

1.7

68

0.9

61

G

roup P

P0.2

18

0.5

86

Gro

up

AD

F-1

.82

50

.03

4

FO

Panel v

1.6

07

0.0

54

0.7

28

0.2

33

1.3

27

0.0

92

None

48.4

20

0.0

81

45.3

60

0.1

36

Panel rh

o-1

.905

0.0

28

-0.6

30

0.2

65

At m

ost 1

43.1

60

0.1

92

43.1

60

0.1

92

P

anel P

P-1

.354

0.0

88

-0.2

78

0.3

91

P

anel A

DF

-0.2

88

0.3

87

-0.5

05

0.3

07

G

roup r

ho

0.5

70

0.7

16

G

rou

p P

P-0

.16

90

.43

3

Gro

up A

DF

-0.6

80

0.2

48

OT

HE

RS

Panel v

4.2

45

0.0

00

2.8

66

0.0

02

-0.4

19

0.3

38

None

79.9

60

0.0

01

76.4

60

0.0

03

Panel rh

o-7

.587

0.0

00

-6.7

39

0.0

00

At m

ost 1

56.1

50

0.1

45

56.1

50

0.1

45

P

anel P

P-6

.478

0.0

00

-5.9

36

0.0

00

P

anel A

DF

-3.7

05

0.0

00

-4.1

06

0.0

00

G

roup r

ho

-3.2

92

0.0

01

G

rou

p P

P-4

.35

40

.00

0

Gro

up A

DF

-2.9

90

0.0

01

No

tes:

AD

F,

augm

ente

d D

ickey-F

ulle

r (D

ickey a

nd F

ulle

r, 1

979);

PP,

Phill

ips-P

err

on (

Phill

ips a

nd P

err

on,

1988).

P

ET,

petr

ol; L

PG

, liq

ue

fie

d p

etr

ole

um

ga

s;

HS

D,

die

sel/hig

h s

peed d

iesel; S

KO

, superior

ke

rosene o

il; F

O,

furn

ace o

il; O

TH

ER

S,

napth

a,

avia

tion t

urb

ine f

uel, lig

ht

die

sel oil,

low

sulp

hur

heavy s

tock h

ot

heavy s

tock, lu

bes a

nd g

reases, itum

en, and o

thers

.

Ta

ble

A.3

. (c

on

tin

ue

d)

Va

ria

ble

Ped

ron

iA

DF

Fis

her

Sta

t.P

rob

.W

. S

tat.

Pro

b.

Sta

t.P

rob

.

Tra

ce

Pro

b.

Ma

x-e

igen

Pro

b.


53

Table A.4. Cointegration test results: disaggregate energy data

(high-income group)

The table presents results from two cointegration tests: Pedroni, and Kao. For the

Pedroni test, the first eight statistics refer to homogenous test – the alternative

hypothesis: common autoregressive (AR) coefficients (within-dimension) while the last

three statistics refer to heterogeneous test with the alternative hypothesis: individual AR

coefficients (between-dimension).

Variable Pedroni ADF

Stat. Prob. W. Stat. Prob. Stat. Prob.

PET Panel v 0.016 0.494 0.016 0.494 0.117 0.453

Panel rho -0.002 0.499 -0.002 0.499

Panel PP 0.259 0.602 0.259 0.602

Panel ADF 0.990 0.839 0.990 0.839

Group rho 0.504 0.693

Group PP 0.676 0.751

Group ADF 1.545 0.939

LPG Panel v 0.338 0.368 0.338 0.368 -1.064 0.144

Panel rho -1.212 0.113 -1.212 0.113

Panel PP -1.096 0.137 -1.096 0.137

Panel ADF 0.073 0.529 0.073 0.529

Group rho -0.624 0.266

Group PP -0.932 0.176

Group ADF 0.455 0.676

HSD Panel v 1.521 0.064 1.521 0.064 -0.187 0.426

Panel rho -0.917 0.180 -0.917 0.180

Panel PP -0.807 0.210 -0.807 0.210

Panel ADF -0.070 0.472 -0.070 0.472

Group rho -0.350 0.363

Group PP -0.589 0.278

Group ADF 0.286 0.613

SKO Panel v -0.695 0.757 -0.695 0.757 2.252 0.012

Panel rho 0.749 0.773 0.749 0.773

Panel PP 0.922 0.822 0.922 0.822

Panel ADF 0.999 0.841 0.999 0.841

Group rho 1.203 0.886

Group PP 1.464 0.928

Group ADF 1.555 0.940


54

FO Panel v -0.055 0.522 -0.055 0.522 0.014 0.495

Panel rho -0.257 0.398 -0.257 0.398

Panel PP -0.728 0.233 -0.728 0.233

Panel ADF 0.649 0.742 0.649 0.742

Group rho 0.265 0.605

Group PP -0.495 0.310

Group ADF 1.139 0.873

OTHERS Panel v 1.310 0.095 1.310 0.095 0.214 0.415

Panel rho -1.135 0.128 -1.135 0.128

Panel PP -0.888 0.187 -0.888 0.187

Panel ADF -1.103 0.135 -1.103 0.135

Group rho -0.553 0.290

Group PP -0.685 0.247

Group ADF -0.940 0.174

Notes: ADF, augmented Dickey-Fuller (Dickey and Fuller, 1979); PP, Phillips-Perron (Phillips and Perron, 1988). PET,

petrol; LPG, liquefied petroleum gas; HSD, diesel/high speed diesel; SKO, superior kerosene oil; FO, furnace

oil; OTHERS, naptha, aviation turbine fuel, light diesel oil, low sulphur heavy stock hot heavy stock, lubes and

greases, itumen, and others.


Variable Pedroni ADF

Stat. Prob. W. Stat. Prob. Stat. Prob.


55

Ta

ble

A.5

. C

oin

teg

rati

on

te

st

res

ult

s:

dis

ag

gre

ga

te e

ne

rgy

da

ta (

mid

dle

-in

co

me

gro

up

)

Th

e t

ab

le p

rese

nts

re

su

lts f

rom

th

ree

co

inte

gra

tio

n t

ests

: P

ed

ron

i, K

ao

, a

nd

Fis

he

r . F

or

the

Pe

dro

ni

test,

th

e f

irst

eig

ht

sta

tistics r

efe

r to

ho

mo

ge

ne

ou

s t

est

– t

he

alte

rna

tive

hyp

oth

esis

: co

mm

on

au

tore

gre

ssiv

e (

AR

) co

effic

ien

ts (

with

in-

dim

en

sio

n)

wh

ile th

e la

st

thre

e sta

tistics re

fer

to h

ete

rog

en

eo

us te

st

with

th

e a

lte

rna

tive

h

yp

oth

esis

: in

div

idu

al

AR

co

effi c

ien

ts (

be

twe

en

-dim

en

sio

n).

Vari

ab

leP

ed

ron

iA

DF

Fis

he

r

Sta

t.P

rob

.W

. S

tat.

Pro

b.

t-S

tat.

Pro

b.

Tra

ce

Pro

b.

Max-e

igen

Pro

b.

PE

TP

anel v

1.7

32

0.0

42

2.6

79

0.0

04

-0.8

22

0.2

06

None

38.4

50

0.0

00

33.8

60

0.0

01

Panel rh

o-1

.053

0.1

46

-1.8

80

0.0

30

At m

ost 1

23.5

00

0.0

24

23.5

00

0.0

24

P

anel P

P-1

.783

0.0

37

-2.4

03

0.0

08

P

anel A

DF

-2.2

18

0.0

13

-3.0

53

0.0

01

G

roup r

ho

-1.0

92

0.1

37

G

rou

p P

P-2

.33

20

.01

0

Gro

up A

DF

-3.6

07

0.0

00

LP

GP

anel v

0.2

20

0.4

13

0.4

64

0.3

22

-3.0

67

0.0

01

None

23.7

80

0.0

22

26.2

40

0.0

10

P

anel rh

o-0

.111

0.4

56

-0.0

72

0.4

71

At m

ost 1

3.8

65

0.9

86

3.8

65

0.9

86

P

anel P

P-1

.093

0.1

37

-0.6

78

0.2

49

P

anel A

DF

-0.9

37

0.1

74

-0.4

23

0.3

36

G

roup r

ho

0.7

19

0.7

64

G

rou

p P

P-0

.47

90

.31

6

Gro

up A

DF

-0.3

34

0.3

69

HS

DP

anel v

1.3

88

0.0

83

1.5

46

0.0

61

-0.3

86

0.3

50

None

23.6

30

0.0

23

24.0

80

0.0

20

P

anel rh

o-0

.898

0.1

85

-0.9

21

0.1

79

At m

ost 1

9.2

03

0.6

86

9.2

03

0.6

86

P

anel P

P-1

.804

0.0

36

-1.6

91

0.0

45

P

anel A

DF

-0.8

70

0.1

92

-0.6

95

0.2

44

G

roup r

ho

-0.7

81

0.2

17

G

rou

p P

P-2

.11

90

.01

7

Gro

up A

DF

-1.1

33

0.1

29


56

Ta

ble

A.5

. (c

on

tin

ue

d)

Va

ria

ble

Pe

dro

ni

AD

FF

ish

er

Sta

t.P

rob

.W

. S

tat.

Pro

b.

t-S

tat.

Pro

b.

Tra

ce

Pro

b.

Max-e

igen

Pro

b.

SK

OP

anel v

0.0

95

0.4

62

-0.1

27

0.5

50

-1.1

12

0.1

33

None

18.1

50.1

113

16.1

20.1

858

P

anel rh

o0.1

63

0.5

65

0.2

65

0.6

05

At m

ost 1

15.6

40.2

081

15.6

40.2

081

P

anel P

P-0

.853

0.1

97

-0.7

84

0.2

17

P

anel A

DF

-1.1

74

0.1

20

-0.7

31

0.2

32

G

roup r

ho

1.1

40

0.8

73

G

rou

p P

P-0

.39

10

.34

8

Gro

up A

DF

-0.4

97

0.3

10

FO

Panel v

3.1

76

0.0

01

2.3

95

0.0

08

-0.0

05

0.4

98

None

5.7

95

0.8

322

4.5

30.9

203

P

anel rh

o-1

.286

0.0

99

-0.2

46

0.4

03

At m

ost 1

12.2

40.2

69

12.2

40.2

69

P

anel P

P-0

.768

0.2

21

0.2

35

0.5

93

P

anel A

DF

-0.8

66

0.1

93

-0.7

36

0.2

31

G

roup r

ho

0.7

55

0.7

75

G

roup P

P0.9

07

0.8

18

Gro

up A

DF

-0.0

80

0.4

68

OT

HE

RS

Panel v

3.2

75

0.0

01

2.1

51

0.0

16

-0.9

02

0.1

84

None

26.0

60.0

105

24.0

50.0

2

P

anel rh

o-3

.074

0.0

01

-3.3

48

0.0

00

At m

ost 1

15.9

20.1

947

15.9

20.1

947

P

anel P

P-2

.590

0.0

05

-2.8

28

0.0

02

P

anel A

DF

-1.6

70

0.0

47

-2.3

45

0.0

10

G

roup r

ho

-1.9

22

0.0

27

G

rou

p P

P-2

.44

90

.00

7

Gro

up A

DF

-2.0

16

0.0

22

No

tes:

AD

F, A

ugm

ente

d D

ickey-F

ulle

r (D

ickey a

nd F

ulle

r, 1

979);

PP,

Phill

ips-P

err

on (

Phill

ips a

nd P

err

on,

1988).

P

ET,

petr

ol; L

PG

, liq

ue

fie

d p

etr

ole

um

ga

s;

HS

D,

die

sel/hig

h s

peed d

iesel; S

KO

, superior

ke

rosene o

il; F

O,

furn

ace o

il; O

TH

ER

S,

napth

a,

avia

tion t

urb

ine f

uel, lig

ht

die

sel oil,

lo

w s

ulp

hu

r h

ea

vy s

tock h

ot

heavy s

tock, lu

bes a

nd g

reases, itum

en, and o

thers

.


57

Ta

ble

A.6

. C

oin

teg

rati

on

te

st

res

ult

s:

dis

ag

gre

ga

te e

ne

rgy

da

ta (

low

-in

co

me

gro

up

)

Th

e t

ab

le p

rese

nts

re

su

lts f

rom

th

ree

co

inte

gra

tio

n t

ests

: P

ed

ron

i, K

ao

, a

nd

Fis

he

r . F

or

the

Pe

dro

ni

test,

th

e f

irst

eig

ht

sta

tistics r

efe

r to

ho

mo

ge

ne

ou

s t

est

– t

he

alte

rna

tive

hyp

oth

esis

: co

mm

on

au

tore

gre

ssiv

e (

AR

) co

effic

ien

ts (

with

in-

dim

en

sio

n)

wh

ile th

e la

st

thre

e sta

tistics re

fer

to h

ete

rog

en

eo

us te

st

with

th

e a

lte

rna

tive

h

yp

oth

esis

: in

div

idu

al

AR

co

effi c

ien

ts (

be

twe

en

-dim

en

sio

n).

Va

ria

ble

Pe

dro

ni

AD

FF

ish

er

Sta

t.P

rob

.W

. S

tat.

Pro

b.

t-S

tat.

Pro

b.

Tra

ce

Pro

b.

Max-e

igen

Pro

b.

PE

TP

anel v

2.8

29

0.0

02

2.8

60

0.0

02

-0.5

31

0.2

98

None

24.6

60

0.0

76

16.5

80

0.4

13

P

anel rh

o-2

.520

0.0

06

-3.1

64

0.0

01

At m

ost 1

29.8

10

0.0

19

29.8

10

0.0

19

P

anel P

P-3

.046

0.0

01

-3.9

19

0.0

00

P

anel A

DF

-1.0

68

0.1

43

-1.1

19

0.1

32

G

roup r

ho

-2.5

97

0.0

05

G

rou

p P

P-4

.27

70

.00

0

G

roup A

DF

-1.0

53

0.1

46

LP

GP

anel v

-0.1

78

0.5

71

-0.0

74

0.5

29

-2.0

09

0.0

22

None

16.0

90

0.4

47

17.4

60

0.3

57

Panel rh

o-0

.263

0.3

96

-0.3

27

0.3

72

At m

ost 1

10.9

10

0.8

15

10.9

10

0.8

15

P

anel P

P-1

.601

0.0

55

-1.4

80

0.0

69

P

anel A

DF

0.5

27

0.7

01

0.8

93

0.8

14

G

roup r

ho

0.2

88

0.6

13

G

rou

p P

P-1

.33

90

.09

0

Gro

up A

DF

1.8

65

0.9

69

HS

DP

anel v

1.2

73

0.1

01

1.3

18

0.0

94

-0.8

91

0.1

87

None

18.0

80

0.3

19

15.4

40

0.4

93

P

anel rh

o-2

.080

0.0

19

-2.1

77

0.0

15

At m

ost 1

22.2

90

0.1

34

22.2

90

0.1

34

P

anel P

P-3

.077

0.0

01

-3.1

37

0.0

01

P

anel A

DF

-0.4

55

0.3

24

-0.3

12

0.3

77

G

roup r

ho

-1.1

31

0.1

29

G

rou

p P

P-2

.85

10

.00

2

Gro

up A

DF

0.0

54

0.5

22


58

SK

OP

anel v

0.8

29

0.2

04

0.2

80

0.3

90

-0.0

46

0.4

82

None

28.9

40

0.0

24

23.9

90

0.0

90

Panel rh

o-0

.627

0.2

66

-0.3

38

0.3

68

At m

ost 1

27.0

90

0.0

41

27.0

90

0.0

41

Panel P

P-1

.440

0.0

75

-1.2

63

0.1

03

P

anel A

DF

0.0

57

0.5

23

-0.5

04

0.3

07

G

roup r

ho

-0.3

09

0.3

79

G

rou

p P

P-1

.28

50

.09

9

Gro

up A

DF

-0.6

46

0.2

59

FO

Panel v

3.0

43

0.0

01

1.8

16

0.0

35

-0.0

90

0.4

64

None

17.7

20

0.1

25

16.9

10

0.1

53

Panel rh

o-2

.980

0.0

01

-1.4

52

0.0

73

At m

ost 1

14.0

10

0.3

00

14.0

10

0.3

00

P

anel P

P-3

.105

0.0

01

-1.8

51

0.0

32

P

anel A

DF

-1.0

20

0.1

54

-1.0

110.1

56

G

roup r

ho

-0.7

60

0.2

24

G

rou

p P

P-1

.64

10

.05

0

Gro

up A

DF

-0.7

77

0.2

19

OT

HE

RS

Panel v

2.4

32

0.0

08

1.9

78

0.0

24

-0.4

44

0.3

28

None

23.4

70

0.1

02

22.6

40

0.1

24

P

anel rh

o-4

.425

0.0

00

-3.3

83

0.0

00

At m

ost 1

18.9

60

0.2

71

18.9

60

0.2

71

P

anel P

P-4

.107

0.0

00

-3.4

38

0.0

00

P

anel A

DF

-2.5

91

0.0

05

-2.1

69

0.0

15

G

roup r

ho

-2.0

50

0.0

20

G

rou

p P

P-3

.06

60

.00

1

Gro

up A

DF

-1.6

96

0.0

45

No

tes:

AD

F, A

ugm

ente

d D

ickey-F

ulle

r (D

ickey a

nd

Fulle

r, 1

979);

PP,

Phill

ips-P

err

on (

Phill

ips a

nd P

err

on,

1988).

P

ET,

petr

ol; L

PG

, liq

ue

fie

d p

etr

ole

um

ga

s;

HS

D,

die

sel/hig

h s

peed d

iesel; S

KO

, superior

ke

rosene o

il; F

O,

furn

ace o

il; O

TH

ER

S,

napth

a,

avia

tion t

urb

ine f

uel, lig

ht

die

sel oil,

lo

w s

ulp

hu

r h

ea

vy s

tock h

ot

heavy s

tock, lu

bes a

nd g

reases, itum

en, and o

thers

.

Ta

ble

A.6

. (c

on

tin

ue

d)

Va

ria

ble

Pe

dro

ni

AD

FF

ish

er

Sta

t.P

rob

.W

. S

tat.

Pro

b.

t-S

tat.

Pro

b.

Tra

ce

Pro

b.

Max-e

igen

Pro

b.


59

Ta

ble

A.7

. S

tate

-wis

e e

co

no

mic

gro

wth

an

d d

isa

gg

reg

ate

en

erg

y c

on

su

mp

tio

n:

fea

sib

le g

en

era

lize

d l

ea

st

sq

ua

res

re

su

lts

Usin

g t

he

fe

asib

le g

en

era

lize

d l

ea

st

sq

ua

res (

FG

LS

) m

eth

od

olo

gy ,

we

estim

ate

th

e s

ho

rt-r

un

re

latio

nsh

ip b

etw

ee

n

diff

ere

nt

typ

es o

f p

etr

ole

um

co

nsu

mp

tio

n a

nd

eco

no

mic

gro

wth

th

at

we

fo

un

d e

vid

en

ce

of

co

inte

gra

tio

n.

No

te t

ha

t w

e

fou

nd

no

evid

en

ce

of

co

inte

gra

tio

n f

or

the

gro

up

of

hig

h-i

nco

me

sta

tes.

La

g l

en

gth

se

lectio

n f

or

ea

ch

pa

ne

l i s

se

lecte

d

ba

se

d o

n t

he

Aka

ike

in

form

atio

n c

rite

rio

n (

AIC

) a

nd

Ba

ye

sia

n in

form

atio

n c

rite

rio

n (

BIC

).

Pan

el 1:

Dep

en

den

t vari

ab

le:

GD

P p

er

cap

ita (

DL

PG

DP

)P

an

el 2:

Dep

en

den

t vari

ab

le:

petr

ole

um

co

nsu

mp

tio

n (

dif

fere

nt

typ

es)

Vari

ab

les

(1)

(2)

(3)

(4)

(5)

(6)

(1)

(2)

(3)

(4)

(5)

(6)

Mid

dle

-M

idd

le-

Mid

dle

-L

ow

-M

idd

le-

Mid

dle

-M

idd

le-

Lo

w-

All

sta

tes

All

sta

tes

inc

om

ein

co

me

inc

om

ein

co

me

All

sta

tes

All

sta

tes

inc

om

ein

co

me

inc

om

ein

co

me

sta

tes

sta

tes

sta

tes

sta

tes

sta

tes

sta

tes

sta

tes

sta

tes

SK

OO

TH

ER

SO

TH

ER

SP

ET

LP

GL

PG

SK

OO

TH

ER

SO

TH

ER

SP

ET

LP

GL

PG

DLP

GD

Pt–

1-0

.118**

*-0

.094**

-0.0

03

-0.0

30

-0.0

04

-0.2

09**

*-0

.078

-0.2

87

0.5

18

0.5

98**

*-0

.143

-0.2

09**

*

(0.0

45)

(0.0

45)

(0.0

67)

(0.0

67)

(0.0

65)

(0.0

70)

(0.0

67)

(0.2

30)

(0.4

58)

(0.1

53)

(0.1

97)

(0.0

81)

DLP

GD

Pt–

20.1

19**

*0.1

35**

*0.0

50

-0.1

64**

-0.1

87

-0.2

19**

*

(0.0

45)

(0.0

45)

(0.0

74)

(0.0

67)

(0.2

27)

(0.0

85)

DLP

GD

Pt–

30.0

16

0.0

12

-0.0

54

0.0

06

0.4

00*

0.1

01

(0.0

43)

(0.0

43)

(0.0

72)

(0.0

63)

(0.2

16)

(0.0

83)

DLP

GD

Pt–

40.1

48**

*0.1

52**

*-0

.039

0.2

31

(0.0

42)

(0.0

42)

(0.0

62)

(0.2

13)

DLP

GD

Pt–

5-0

.070*

-0.0

61

-0.0

67

0.3

80*

(0.0

41)

(0.0

41)

(0.0

62)

(0.2

13)

DLP

EC

t–1

0.0

36

0.0

16*

0.1

00.0

46*

-0.0

21

-0.0

34

-0.0

14

-0.2

94**

*-0

.349**

*0.1

82**

*-0

.178**

*-0

.211

**

(0.0

31)

(0.0

09)

(0.0

09)

(0.0

27)

(0.0

22)

(0.0

64)

(0.0

50)

(0.0

49)

(0.0

63)

(0.0

64)

(0.0

65)

(0.0

83)

DLP

EC

t–2

0.0

12

0.0

15

-0.1

26*

0.1

62**

*-0

.115**

0.2

28**

*

(0.0

34)

(0.0

10)

(0.0

67)

(0.0

49)

(0.0

50)

(0.0

81)

DLP

EC

t–3

-0.0

68**

0.0

08

-0.1

26*

-0.0

38

-0.0

45

-0.0

37

(0.0

33)

(0.0

10)

(0.0

64)

(0.0

49)

(0.0

54)

(0.0

75)

DLP

EC

t–4

-0.0

69**

0.0

10

-0.0

18

-0.0

52

(0.0

32)

(0.0

10)

(0.0

47)

(0.0

54)

DLP

EC

t–5

-0.0

27

0.0

01

0.1

58**

*-0

.195**

*

(0.0

32)

(0.0

10)

(0.0

47)

(0.0

51)


60

Ta

ble

A.7

. (c

on

tin

ue

d)

Pan

el 1:

Dep

en

den

t vari

ab

le:

GD

P p

er

cap

ita (

DL

PG

DP

)P

an

el 2:

Dep

en

den

t vari

ab

le:

petr

ole

um

co

nsu

mp

tio

n (

dif

fere

nt

typ

es)

(1)

(2)

(3)

(4)

(5)

(6)

(1)

(2)

(3)

(4)

(5)

(6)

Va

ria

ble

sM

idd

le-

Mid

dle

-M

idd

le-

Lo

w-

Mid

dle

-M

idd

le-

Mid

dle

-L

ow

-

All

sta

tes

All

sta

tes

inc

om

ein

co

me

inc

om

ein

co

me

All

sta

tes

All

sta

tes

inc

om

ein

co

me

inc

om

ein

co

me

sta

tes

sta

tes

sta

tes

sta

tes

sta

tes

sta

tes

sta

tes

sta

tes

SK

OO

TH

ER

SO

TH

ER

SP

ET

LP

GL

PG

SK

OO

TH

ER

SO

TH

ER

SP

ET

LP

GL

PG

EC

Mt–

10.0

04

-0.0

002

-0.0

03

0.0

10

-0.0

30**

-0.0

38**

*-0

.045**

*-0

.055**

*-0

.087**

*-0

.026*

-0.0

58**

*-0

.041**

*

(0.0

06)

(0.0

06)

(0.0

08)

(0.0

11)

(0.0

04)

(0.0

14)

(0.0

13)

(0.0

18)

(0.0

31)

(0.0

15)

(0.0

19)

(0.0

13)

Observ

ations

478

476

221

223

221

177

455

453

212

214

212

168

No. of cro

ssid

23

23

99

98

23

23

99

98

No

tes:

SK

O,

superior

kero

sene o

il; P

ET,

petr

ol; L

PG

, liq

uefied p

etr

ole

um

gas;

OT

HE

RS

, napth

a,

avia

tion t

urb

ine f

uel, lig

ht

die

sel oil,

lo

w s

ulp

hu

r h

ea

vy s

tock h

ot

he

avy

sto

ck,

lub

es

an

d

gre

ase

s,

itu

me

n,

an

d

oth

ers

. **

*,

**

an

d

* in

dic

ate

re

jectio

n

of

the

n

ull

hyp

oth

esis

a

t 1

p

er

ce

nt,

5

p

er

ce

nt

an

d

10 p

er

cent sig

nific

ance levels

. S

tandard

err

ors

are

report

ed in p

are

nth

eses.


61

Table A.8. Long-run models with disaggregate energy sources

Using the feasible generalized least squares (FGLS) methodology, we estimate the

long-run relationship between different types of petroleum consumption and economic

growth that we found evidence of cointegration. Note that we found no evidence of

cointegration for the group of high-income states.

(1) (2) (3) (4)

SKO OTHERS PET LPG

PEC = f(PGDP)

All states 0.051* 0.944*

(0.026) (0.056)

Observations 641 640

Number of crossid 23 23

Middle-income states 0.756*** 1.259*** 1.820***

(0.100) (0.066) (0.072)

Observations 250 251 249

Number of crossid 9 9 9

Low-income states 0.425***

(0.033)

Observations 221

Number of crossid 8

PGDP = f(PEC)

All states 0.114* 0.328*

(0.059) (0.019)

Observations 641 640

Number of crossid 23 23

Middle-income states 0.247*** 0.469*** 0.395***

(0.033) (0.025) (0.016)

Observations 250 251 249

Number of crossid 9 9 9

Low-income states 0.425***

(0.033)

Observations 221

Number of crossid 8

Notes: PEC, per capita energy consumption; PGDP, per capita gross domestic product. SKO, superior kerosene oil;

PET, petrol; LPG, liquefied petroleum gas; OTHERS, naptha, aviation turbine fuel, light diesel oil, low sulphur

heavy stock hot heavy stock, lubes and greases, itumen, and others. ***, ** and * indicate rejection of the null

hypothesis at 1 per cent, 5 per cent and 10 per cent significance levels. Standard errors are reported in

parentheses.


62

REFERENCES

Abbas, F., and N. Choudhury (2013). Electricity consumption-economic growth nexus: an aggregatedand disaggregated causality analysis in India and Pakistan. Journal of Policy Modeling,vol. 35, No. 4, pp. 538-553.

Ahmad, A., and others (2016). Carbon emissions, energy consumption and economic growth:an aggregate and disaggregate analysis of the Indian economy. Energy Policy, No. 96,pp. 131-143.

Akarca, A.T., and T.V. Long (1980). On the relationship between energy and GNP: a re-examination. The

Journal of Energy Development, vol. 5, No. 2, pp. 326-331.

Akhmat, G., and K. Zaman (2013). Nuclear energy consumption, commercial energy consumption andeconomic growth in South Asia: bootstrap panel causality test. Renewable and Sustainable

Energy Reviews, vol. 25, pp. 552-558.

Apergis, N., and J.E. Payne (2009a). Energy consumption and economic growth: evidence from theCommonwealth of Independent States. Energy Economics, vol. 31, No. 5, pp. 641-647.

(2009b). Energy consumption and economic growth in Central America: evidence from a panelcointegration and error correction model. Energy Economics, vol. 31, No. 2, pp. 211-216.

Asafu-Adjaye, J. (2000). The relationship between energy consumption, energy prices and economicgrowth: time series evidence from Asian developing countries. Energy Economics, vol. 22,No. 6, pp. 615-625.

Bayer, Christian, and Christoph Hanck (2009). Combining non-cointegration tests. ResearchMemorandum 012. Maastricht University, Maastricht Research School of Economics ofTechnology and Organization (METEOR).

Bhattacharyya, S.C. (2015). Influence of India’s transformation on residential energy demand. Applied

Energy, vol. 143, pp. 228-237.

Bildirici, Melike E., and Tahsin Bakirtas (2014). The relationship among oil, natural gas and coalconsumption and economic growth in BRICTS (Brazil, Russian, India, China, Turkey andSouth Africa) countries. Energy, vol. 65, pp. 134-144.

British Petroleum (2019). BP Statistical Review of World Energy 2019. Available at www.bp.com/content/dam/bp/business-sites/en/global/corporate/pdfs/energy-economics/statistical-review/bp-stats-review-2019-full-report.pdf.

Cowan, W., and others (2014). The nexus of electricity consumption, economic growth and CO2emissions in the BRICS countries. Energy Policy, vol. 66, pp. 359-368.

Di Iorio, F., and S. Fachin (2007). Testing for breaks in cointegrated panels-with an application to theFeldstein-Horioka puzzle. Economics: The Open-Access, Open-Assessment E-Journal, vol. 1.

Dickey, D., and W. Fuller (1979). Distribution of the estimators for autoregressive time series with a unitroot. Journal of the American Statistical Association, vol. 74, No. 377, pp. 427-431.

Dutta, M., and S. Mukherjee (2010). An outlook into energy consumption in large scale industriesin India: the cases of steel, aluminium and cement. Energy Policy, vol. 38, No. 11,pp. 7286-7298.

Engle, Robert F., and C.W.J. Granger (1987). Co-integration and error correction: representation,estimation, and testing. Econometrica, vol. 55, No. 2 pp. 251-276.


63

Fang, Z., and Thai-Ha Le (forthcoming). Cointegrating relationship and Granger causal analysis inenergy economics – a practical guidance. In Handbook of Energy Finance: Theories,

Practices and Simulations, S. Goutte and and N. Khuong, eds. World Scientific Publishing,Forthcoming in December 2019.

Fisher, Ronald Aylmer (1932). Statistical Methods for Research Workers, 4th ed. Edinburgh: Oliver &Boyd.

Ghosh, S. (2002). Electricity consumption and economic growth in India. Energy Policy, vol. 30, No. 2,pp. 125-129.

Govindaraju, V.G.R. Chandran, and Chor Foon Tang (2013). The dynamic links between CO2emissions, economic growth and coal consumption in China and India. Applied Energy,vol. 104, pp. 310-318.

Greene, W.H. (2008). Econometric Analysis, 6th ed. Upper Saddle River, New Jersey: Prentice Hall.

Im, K.S., M.H. Pesaran, and Y. Shin (2003). Testing for unit roots in heterogeneous panels. Journal of

Econometrics, vol. 115, No. 1, pp. 53-74.

International Energy Agency (IEA) (2015). India Energy Outlook, World Energy Outlook Special Report.Paris: IEA Publications.

Johansen, S. (1988). Statistical analysis of cointegration vectors. Journal of Economic Dynamics and

Control, vol. 12, No. 2-3, pp. 231-254.

Kao, C. (1999). Spurious regression and residual-based tests for cointegration in panel data. Journal of

Econometrics, vol. 90, No. 1, pp. 1-44.

Kónya, L. (2006). Exports and growth: Granger causality analysis on OECD countries with a panel dataapproach. Economic Modelling, vol. 23, No. 6, pp. 978-982.

Kraft, J., and A. Kraft (1978). On the relationship between energy and GNP. The Journal of Energy and

Development, vol. 3, No. 2, pp. 401-403.

Le, T.H. (2016). Dynamics between energy, output, openness and financial development in sub SaharanAfrican countries. Applied Economics, vol. 48, No. 10, pp. 914-933.

Le, T.H., and C.P. Nguyen (2019). Is energy security a driver for economic growth? Evidence froma global sample. Energy Policy, vol. 129, pp. 436-451.

Le, T.H., and E. Quah (2018). Income level and the emissions, energy, and growth nexus: evidence fromAsia and the Pacific. International Economics, vol. 156, pp. 193-205.

Lee, C.C., and C.P. Chang (2005). Structural breaks, energy consumption, and economic growthrevisited: evidence from Taiwan. Energy Economics, vol. 27, No. 6, pp. 857-872.

Levin, A., C.F. Lin, and C.S.J. Chu (2002). Unit root tests in panel data: asymptotic and finite-sampleproperties. Journal of Econometrics, vol. 108, No. 1, pp. 1-24.

Maddala, G.S., and S. Wu (1999). A comparative study of unit root tests with panel data and a newsimple test. Oxford Bulletin of Economics and Statistics, vol. 61, No. S1, pp. 631-652.

Mallick, H., M.K. Mahalik, and M. Sahoo. (2018). Is crude oil price detrimental to domestic privateinvestment for an emerging economy? The role of public sector investment and financialsector development in an era of globalization. Energy Economics, vol. 69, pp. 307-324.

Nain, Z., W. Ahmad, and B. Kamaiah (2015). Economic growth, energy consumption and CO2emissions in India: a disaggregated causal analysis. International Journal of Sustainable

Energy, vol. 36, No. 8, pp. 807-824.


64

Nain, M., S.S. Bharatam, and B. Kamaiah (2017). Electricity consumption and NSDP nexus inIndian states: a panel analysis with structural breaks. Economics Bulletin, vol. 37, No. 3,pp. 1581-1601.

Narayan, P.K., S. Narayan, and S. Popp (2010a) Does electricity consumption panel Granger causeGDP? A new global evidence. Applied Energy, vol. 87, No. 10, pp. 3294-3298.

(2010b). A note on the long-run elasticities from the energy consumption-GDP relationship.Applied Energy, vol. 87, No. 3, pp. 1054-1057.

Narayan, S. (2016). Predictability within the energy consumption-economic growth nexus: someevidence from income and regional groups. Economic Modelling, vol. 54, pp. 515-521.

Narayan, S., B.N. Rath, and P.K. Narayan (2012). Evidence of Wagner’s law from Indian states.Economic Modelling, vol. 29, No. 5, pp. 1548-1557.

Naser, H. (2015). Analysing the long-run relationship among oil market, nuclear energy consumption,and economic growth: an evidence from emerging economies. Energy, vol. 89, pp. 421-434.

Nasreen, S., and S. Anwar (2014). Causal relationship between trade openness, economic growthand energy consumption: a panel data analysis of Asian countries. Energy Policy, No. 69,pp. 82-91.

Oh, W., and K. Lee (2004). Causal relationship between energy consumption and GDP revisited: thecase of Korea 1970-1999. Energy Economics, vol. 26, No. 1, pp. 51-59.

Pao, H.-T., and C.-M. Tsai (2010). CO2 emissions, energy consumption and economic growth in BRICcountries. Energy Policy, vol. 38, No. 12, pp. 7850-7860.

Parikh, J., and K. Parikh (2011). India’s energy needs and low carbon options. Energy, vol. 36, No. 6,pp. 3650-3658.

Parikh, J., and others (2009). CO2 emissions structure of Indian economy. Energy, vol. 34, No. 8,pp. 1024-1031.

Paul, S., and R.N. Bhattacharya (2004). Causality between energy consumption and economic growth inIndia: a note on conflicting results. Energy Economics, vol. 26, No. 6, pp. 977-983.

Pedroni, P. (1999). Critical values for cointegration tests in heterogeneous panels with multipleregressors. Oxford Bulletin of Economics and Statistics, vol. 61, No. S1, pp. 653-670.

(2004). Panel cointegration: asymptotic and finite sample properties of pooled time series withan application to the PPP hypothesis. Econometric Theory, vol. 20, No. 3, pp. 597-625.

Pesaran, M. (2004). General diagnostic tests for cross section dependence in panels. CESifo WorkingPaper Series, No. 1229; IZA Discussion Paper, No. 1240. Available at https://pdfs.semanticscholar.org/7f14/e40e9ff7e57a8db34178ba001be3cac0720b.pdf?_ga=2.101567751.764281370.1564432569-282743693.1564432569.

Pesaran, M.H., A. Ullah, and T. Yamagata (2008). A bias–adjusted LM test of error cross–sectionindependence. The Econometrics Journal, vol. 11, No. 1, pp.105-127.

Phillips, P., and P. Perron (1988). Testing for unit root in time series regression. Biometrica, vol. 5, No. 2,pp. 335-346.

Proops, J.L.R. (1984). Modelling the energy-output ratio. Energy Economics, vol. 6, No. 1, pp. 47-51.

Rafiq, S., and R.A. Salim (2009). Temporal causality between energy consumption and income in sixAsian emerging countries. Applied Economics Quarterly, vol. 55, No. 4, pp. 335-350.


65

Saxena, V., and P.C. Bhattacharya (2018). Inequalities in LPG and electricity consumption in India: therole of caste, tribe, and religion. Energy for Sustainable Development, vol. 42, No. C,pp. 44-53.

Sen, A., and A. Sen (2016). India’s oil demand: on the verge of ‘take off’? OIES Paper, WPM 65. OxfordInstitute for Energy Studies, University of Oxford.

Shahbaz, M., and others (2016). The role of globalization on the recent evolution of energy demand inIndia: implications for sustainable development. Energy Economics, vol. 55, pp. 52-68.

Singh, R. (2018). Energy sufficiency aspirations of India and the role of renewable resources: scenariosfor future. Renewable and Sustainable Energy Reviews, vol. 81, part 2, pp. 2783-2795.

Stern, D.I. (1993). Energy use and economic growth in the USA: a multivariate approach. Energy

Economics, vol. 15, No. 2, pp.137-150.

Tiwari, A. (2011). Energy consumption, CO2 emissions and economic growth: evidence from India.Journal of International Business and Economy, vol. 12, No. 1, pp. 85-122.

Tiwari, A.K., M. Shahbaz, and Q.M.A. Hye (2013). The environmental Kuznet’s curve and the role of coalconsumption in India: cointegration and causality analysis in an open economy. Renewable

and Sustainable Energy Reviews, vol. 18, pp. 519-527.

Toda, H.Y., and T. Yamamoto (1995). Statistical inference in vector autoregressions with possiblyintegrated processes. Journal of Econometrics, vol. 66, No. 1-2, pp. 225-250.

Verma, Nidhi (2018). India’s oil imports in 2017 surged to a record 4.4 million bpd. Reuters, 16 January.Available at www.reuters.com/article/india-oil/indias-oil-imports-in-2017-surged-to-a-record-4-4-million-bpd-idINKBN1F5234.

Vidyarthi, H. (2013). Energy consumption, carbon emissions and economic growth in India. World

Journal of Science, Technology and Sustainable Development, vol. 10, No. 4, pp. 278-287.

Wolde-Rufael, Y. (2010). Bounds test approach to cointegration and causality between nuclear energyconsumption and economic growth in India. Energy Policy, vol. 38, No. 1, pp. 52-58.

Yang, H.Y. (2000). A note of the causal relationship between energy and GDP in Taiwan. Energy

Economics, vol. 22, No. 3, pp. 309-317.

Yang, Z., and Y. Zhao (2014). Energy consumption, carbon emissions, and economic growth in India:evidence from directed acyclic graphs. Economic Modelling, vol. 38, pp. 533-540.

CURRENT TRENDS IN PRIVATE FINANCING OF WATERAND SANITATION IN ASIA AND THE PACIFIC

Hongjoo Hahm*

The present paper shows the current trends in private sector investment inthe water and sanitation sector. After peaking in 2007, private investment inthe water and sanitation sector has been volatile. The decline in privateinvestment has also been accompanied by a shift in the type and size ofinvestments taking place. Post-2007, private investment is increasinglyconcentrated in a few large and wealthy countries and municipalities; andare bankrolled and developed by smaller, regional-based investors. This isespecially worrying for low-income countries, which stand to benefit themost from private investment, but have been receiving less than 1 per centof the total project allocations in the sector. The huge financing gaprequires more innovative financing that can only come by attracting privatesector capital to improve water and sanitation services in the Asia-Pacificregion, especially for the least developed and low-income emergingeconomies.

JEL classification: C30, G23, G28, H44, H54, H72, H81

Keywords: private financing, regional financing, diaspora financing, water and sanitation,

Asia and the Pacific, innovative finance, blended finance

* The author is presently serving as Deputy Executive Secretary of the Economic and Social Commissionfor Asia and the Pacific. The views expressed in this paper are his solely and do not reflect the views ofthe United Nations.

67


68

I. INTRODUCTION

Achieving universal access to safe and affordable drinking water along with

adequate sanitation (Sustainable Development Goal 6) are essential to improving

people’s livelihoods and the environment in Asia and the Pacific. It remains, however,

a huge challenge across the region. Although 94 per cent of the population has access

to improved drinking water, only 65 per cent of it has access to basic sanitation facilities;

that means that almost 1.5 billion people lack sanitation services. The “access to water”

coverage for Asia and the Pacific is comparable to other regions (table 1), but “access to

improved sanitation” in the region lags significantly (table 2). Moreover, wastewater is

Table 1. Access to improved water source

Region National Rural Urban

Latin America and the Caribbean 94.63 83.90 97.38

Middle East and North Africa 93.50 89.32 95.67

North America 99.26 98.28 99.46

Sub-Saharan Africa 67.54 55.82 86.75

East Asia and the Pacific 94.14 90.19 97.34

South Asia 92.37 90.88 95.33

East South Asia and the Pacific 93.25 90.53 96.33

Europe and Central Asia 98.49 96.06 99.41

World 90.95 84.58 96.45

Source: World Bank, Private Participation in Infrastructure dataset. Available at https://

datacatalog.worldbank.org/dataset/private-participation-infrastructure.

Table 2. Access to improved sanitation facilities

Region National Rural Urban

Latin America and the Caribbean 83.15 64.12 87.95

Middle East and North Africa 91.14 86.89 93.31

North America 99.98 99.90 100.00

Sub-Saharan Africa 29.79 23.30 40.37

East Asia and the Pacific 77.21 64.33 87.10

South Asia 44.77 35.07 64.60

East South Asia and the Pacific 60.99 49.70 75.85

Europe Central Asia 93.08 89.22 94.67

World 67.53 50.33 82.19

Source: World Bank, Private Participation in Infrastructure dataset. Available at https://

datacatalog.worldbank.org/dataset/private-participation-infrastructure.

Current trends in private financing of water and sanitation in Asia and the Pacific

69

often discharged into rivers or seas without treatment. Between 80 and 90 per cent of

all wastewater produced in the Asia-Pacific region is released untreated (ESCAP,

UN-Habitat and Asian Institute of Technology, 2015). The situation is particularly

alarming in the coastal zones of South, South-West and South-East Asia. For instance,

an estimated 77 per cent of the wastewater released in Thailand is untreated, 82 per

cent in Pakistan, 84 per cent in Armenia, and 81 per cent in Viet Nam.1

Incomplete and ageing infrastructure systems face stress from increasing

consumption, leakages, theft and extreme weather events, which affect the quantity and

quality of water, and the distribution networks that supply it. The dilemma is how to pay

for water and sanitation investments, and whether local, municipal or national

governments will have the financing to invest and achieve Sustainable Development

Goal 6. More than 80 per cent of the financial investments in water and sanitation come

from public sources. Public sector funds are mainly from local or municipal governments

and have not sufficiently covered the needs of growing populations and improved the

performance of existing water utilities. Developing country governments are constrained

by the total amount of funds they can raise through taxes, and budgetary resources

compete for many worthwhile programmes in all sectors. Governments complement

their own resources with official development assistance (ODA), and, when available,

from domestic and international private sector resources. While commercial finance is

generally far more abundant than public finance, it is also substantially more risk averse.

Commercial finance will not be channelled if the risk-return criteria of private investors

and lenders are not met. While there has been much discussion about accessing the

private sector through public-private partnerships (PPP), experiences of these

collaborations in the water and sanitation sectors have significantly fallen short of

expectations.

In the present paper, the current trends in private sector financing for the water and

sanitation sectors in the Asian and Pacific region are examined and the motivation

behind private investments, their destination and source countries, and the size of the

investments are analysed. The decrease in private investments in water and sanitation

over the past several decades is highlighted. Investments by global water multilateral

corporations have steadily decreased in the sector. This slack has been taken up by

smaller private regional and local water investors, but it has not been enough to cover

the gap. This paper is organized as follows. Section II presents the total investment

needs of the water and sanitation sector in Asia and the Pacific. Section III contains an

outline of the water and sanitation sector financing architecture, and highlights of some

of the key private sector investment trends. Section IV then provides a simple model to

explore the determinants for private investment in water and sanitation in the region.

Section V concludes and includes policy recommendations on mobilizing private

financing in the water and sanitation sector in Asia and the Pacific.

1 See https://sdgasiapacific.net/.


70

II. ASIA-PACIFIC WATER AND SANITATION FINANCING NEEDS

Physical capital investments in water and sanitation globally, excluding operation

and maintenance costs, need to triple to $1.7 trillion to meet Sustainable Development

Goal 6 by 2030. This is three times the amount that has been historically invested in the

sector (Hutton and Varughese, 2016). The Asian and Pacific region requires $800 billion,

or $53 billion annually, in investment over the period 2016–2030 to meet its water and

sanitation infrastructure needs (ADB, 2017). However, according to the most recent

UN-Water Global Analysis and Assessment of Sanitation and Drinking Water report of

2017, 80 per cent of countries globally reported insufficient financing to meet national

water, sanitation and hygiene access targets, and many countries in the Asia-Pacific

region are lagging significantly in this regard (WHO, 2017). To attract investments and

make progress towards achieving water and sanitation-related Sustainable

Development Goals, countries in the region must focus on policy and actions, and define

strategic frameworks to improve the financial sustainability and resilience of water

systems and infrastructure. This requires governments to strategically mobilize public

resources and expand opportunities for private investment.

Water and sanitation infrastructure is ultimately paid for by one, or a combination,

of the following parties: users through tariffs or charges; taxpayers through local and

national taxes; or aid donors through ODA. Traditionally, the largest funding sources are

from local sources, primarily local and municipal government coffers, and, to some

extent, from user fees. Local domestic banks also play a significant role, although

related accurate data are scarce. Local banks mostly provide financing for working

capital or operational or current expenditures of short maturity. Rarely do local banks

finance longer-term capital investments. International aid, foreign banks and foreign

private companies also extend funds for water and sanitation infrastructure, but the total

amount is a much smaller share of total expenditure (Winpenny, 2003).

In the mid-1990s, the public sector provided 65 to 70 per cent of the sector’s

resources, while the domestic private sector covered 5 per cent, ODA was responsible

for the 10 to 15 per cent and the international private sector investors, consisting of

banks and multinational water companies, covered the remaining 10 per cent (Hahm,

1996). Two decades later in the mid-2010s, the breakdown skewed even more to local

and public sources, with a notable reduction in private international funding.

Private investment in water and sanitation has tended to occur in spurts, not as

a steady flow. It takes place when there has been strong demand for investing in water

and sanitation services infrastructure (from the public, government officials, and the

water utilities themselves) and when there is ample financing. Given the vast water and

sanitation investment requirements, existing sources of funding do not cover the needs.

Countries must not only tap into new sources of finance to meet growing demand, but

they also need to fund adequate operations and maintenance required for more

sustainable services.


71

III. WATER AND SANITATION FINANCING ARCHITECTURE

Physical investments in water and sanitation, such as major network expansions

and new wastewater treatment facilities, are capital intensive with high upfront costs and

long payback periods, repaid in local currencies. Investments require long-term

financing (long maturities), preferably with ample grace periods to accommodate long

construction schedules, and in local currency to minimize foreign exchange risks. The

water and sanitation sector has historically relied on public funding to meet its

investment needs. Municipal or local governments, regional or provincial, or national

Government – through taxes, transfers, tariffs, and user fees – are the main funders of

water and sanitation infrastructure. Accordingly, in most countries, government-owned

agencies or organizations are responsible for drinking water supply and wastewater

treatment.

The current financing architecture relies on public, private and ODA investments.

With a few exceptions, developing Asia and Pacific countries have made substantial

progress towards the delivery of water services provision, but many have not fared as

well on the treating wastewater and the delivery of sanitation. Population growth and

rural to urban migration will continue to pose significant challenges for the provision of

water, and especially sanitation services. Figure 1 shows the financing framework for the

Figure 1. Financing framework for water and sanitation investments

Source: OECD (2010).

COSTS REVENUES

Repayments

REPAYABLEFINANCE

Equity

Bonds

Commercialloans

Bridge thefinancing gap

MARKET BASED

REPAYABLE

FINANCE

Private funds

Public funds

Transfers

Taxes

Investmentcosts

(rehabilitationand new)

Maintenancecosts

Operating costs

Financing gap

WATER SERVICE PROVIDERS’ FINANCES

Tariffs

Concessionary (incl.grant element)


72

water and sanitation sector. The demand for water and sanitation investment exceeds

the supply of funding sources. Taxes and user fees cannot cover the investment needs.

The financing gap needs to be filled from private sector sources.

In many developing countries, the water sector’s cash flow does not meet the mark

of financial sustainability for either service provision or sector development. Accordingly,

ODA for water and sanitation is sought. However, ODA totals are approximately $4

billion – a far cry from the $50 billion in investments required by the sector (figure 2).

ODA includes bilateral support for water and sanitation. Australia and Dutch aid have

played significant roles, but regional donors, such as Japan and the Republic of Korea,

have also become major players in recent years. For example, K-Water, a specialized

water company of the Government of the Republic of Korea, invests in overseas water

projects by taking equity stakes and financing water and sanitation projects; many of the

projects are co-financed with multilateral development banks.

Source: World Bank, Private Participation in Infrastructure dataset. Available at https://datacatalog.worldbank.org/

dataset/private-participation-infrastructure.

Figure 2. Total overseas development assistance for water and sanitation

in Asia and the Pacific

4 000

3 500

3 000

2 500

2 000

1 500

1 000

500

0

2015 m

illio

ns o

f U

S d

olla

rs

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015

Public and ODA funding combined do not adequately cover the needs related to

water and especially to sanitation. Countries need to tap into new sources of finance to

meet the growing demand by expanding opportunities for private investment. In the

1990s, the greatest influx of private spending occurred under a situation in which

governments, frustrated with poorly performing public sector monopolies, sought private

participation in infrastructure as a way of reducing the drain on government budgets.

Even though the investment of international private investment in water and sanitation


73

infrastructure increased during that period, such projects still constituted only 5.4 per

cent of all private commitments to infrastructure, including financing for energy, transport

and telecommunications.

The Private Participation in Infrastructure database of the World Bank shows that

private investments in water and sanitation peaked in 1997 and have since dropped.

The Asian financial crisis of 1997 contributed to the decline of investment flows in the

following years. The boom in the 1990s was largely replaced with pessimism, as

projects were renegotiated, cancelled or renationalized, further subduing private

investments in water and sanitation. The largest recorded amount of private investment

in water and sanitation, $10.2 billion globally and more than $8 billion in the Asia-Pacific

region alone, continues to be in 1997. In terms of projects financed, the number of

privately financed water and sanitation projects peaked in 2007 when 90 projects were

implemented around the world. This growth was driven by Asia and the Pacific, where

73 projects were implemented during that year (figure 3).

Figure 3. Private participation in water and sanitation projects, 1991-2016



While the Asia-Pacific region slowly recovered from the 1997 Asian financial crisis,

the 2008 global financial crisis marks a significant turning point for private investment in

water and sanitation. In the post-2007 environment, new water and sanitation

investments (structured as project finance or PPPs) are primarily taking place only in

developed economies that have developed capital markets capable of issuing long

12

10

8

6

4

2

0

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

Total investment (billions of US dollars) Number of projects (Asia and the Pacific)

80

70

60

50

40

30

20

10

0


74

Sanitation-related projects Water supply-related projects

Num

ber

of pro

jects

50

40

30

20

10

0

2000 2002 2004 2006 2008 2010 2012 2014 2016

Investment year

maturity transactions. The provision of drinking water and wastewater services by

private companies in 2017 was approximately 14 per cent globally. Some 47 per cent of

the population is served by the private service providers in Western Europe,

approximately 23 per cent in North America, and only 20 per cent in Asia. In the

Asia-Pacific region, since 2002, there has been more sanitation-related projects with

private investments than projects related to water, as most Asia-Pacific countries have

met their water supply-related investments (figure 4).

Figure 4. Trends of private water and sanitation investments

in Asia and the Pacific

A total of 1,072 privately financed water and sanitation projects (partly or wholly)

were undertaken between 1993 and 2017 globally. Out of this, 654 projects were carried

out in the Asia-Pacific region, with China being the greatest recipient, at 499 projects. In

comparison, Latin America and the Caribbean received 335 projects, and the Middle

East along with Africa received only 63 projects. Countries belonging to the upper-

middle-income category were the major recipients of private investment. Out of the total

1,072 projects, 955 were in upper-middle-income countries whereas only 10 were in

low-income countries. The same pattern can be observed in the Asia-Pacific region,

Source: World Bank, Private Participation in Infrastructure dataset, 2017. Available at https://datacatalog.

worldbank.org/dataset/private-participation-infrastructure.


75

where 577 out of 654 projects were implemented in upper-middle-income countries and

only one was implemented in a low-income country.

Close to 68 per cent of all the projects undertaken between 1993 and 2017 were

granted at the local or municipal level, compared to 10 per cent at the national level. In

terms of investment value, during the period, approximately 40 per cent to the projects

were sanctioned at the local or municipal level, compared to 33 per cent at the national

level. This shows that even though a lower number of projects were undertaken at the

national level, their investment value (project size) has generally been higher. The same

pattern can be seen in the Asia-Pacific region, with approximately 85 per cent of the

projects and 55 per cent of the project value being sanctioned at the local or municipal

level.

More than 26 projects were taken up each year in the Asia-Pacific region between

1993 and 2017. The predominate number of projects were carried out in China,

averaging 20 projects per year. The country’s share of the total number of projects

undertaken in the region accounted for approximately 33 per cent between 1993 and

2017. Moreover, China held a 28 per cent share in terms of the value of the privately

financed portion of these investments.

There were three notable changes in the size, type, and character of private

investments in water and sanitation pre- and post-2007. First, the average size of

private investment in a water and sanitation project fell from $91 million between 1993

and 2007 to $82 million between 2007 and 2016. In addition, the number of projects per

year has declined. The year 2007 marked the peak, with 73 water and sanitation

projects in Asia and the Pacific; subsequently, since 2007, the number of projects has

not exceeded 30 in any given year.

Second, upper-middle-income countries remain the key beneficiaries of private

investments in water and sanitation projects. Pre-2007, upper-middle-income countries

had a share of 87 per cent of total number of private investments, while lower-middle-

income and low-income countries had a share of 12 per cent and 1 per cent,

respectively. In terms of value, upper-middle-income countries had a share of 74 per

cent as compared to lower-middle-income countries and low-income countries shares of

26 per cent and 0.08 per cent, respectively. The trend has continued post-2007, with

upper-middle-income countries accounting for a share of 93.5 per cent of total

investments and lower-middle-income and low-income countries having shares of 6 per

cent and 0.5 per cent, respectively. In terms of value, upper-middle-income countries

had a share of 91 per cent compared to lower-middle-income and low-income countries

shares of 9 per cent and 0.003 per cent, respectively.

Finally, there has been a notable shift away from global multinational corporations

to smaller, regional water companies post-2007. Pre-2007, the share of domestic and

foreign investments in water and sanitation facilities was almost equal in size in which


76

80

60

40

20

0

Domestic/regionalfirms

Foreign multinational corporations

Joint investments

Pre-2007 Post-2007

Perc

enta

ge

domestic firms constituted 44 per cent of total private investments versus 43 per cent for

foreign investments. The share of joint investments from domestic and foreign firms

accounted for the remainder. The situation changed in post-2007, with a surge in

domestic investments. Domestic investments constituted 72 per cent of all the private

investments, while the share of foreign investments has declined to 22 per cent

(figure 5).

Figure 5. Domestic versus foreign investments

The post-2007 decline of private sector funding in the water and sanitation sector

reflects the changing landscape of private investors (OECD, 2009). Many large

multilateral water companies have pulled back their investments in Asia and throughout

the world, and only a handful of global water investors remain. Despite the availability of

foreign exchange risk mitigating tools, most international water companies have simply

moved out of emerging markets to focus on developed high-income emerging markets.

The few private investments in developing economies are highly concentrated in large

urban centres and municipalities, typically in economies with developed capital markets.

During the period 1990–1997, five international multinational water operators

(Suez, Veolia, Thames, Agbar and Saur) accounted for 53 per cent of all water and

sanitation projects awarded. Starting five years later, over the period 2003–2005, their

share dropped to 23 per cent. Suez, a French-owned international water concession

company, and Thames, a British firm, largely withdrew their investments in developing

countries. On the other hand, Veolia and Agbar have focused on investing through local

partners or through joint ownership with local governments. Severn Trent, another

international player, has redirected its operations to concentrate on management and

service contracts, with little to no capital investment (table 4).




77

Table 4. Top five companies incurring foreign investment in 2016

CompanyHome

Host economyNumber of foreign

economy projects undertaken

Suez France Argentina, Brazil, Bolivia, 65

Colombia, others

Veolia France China, Armenia, Romania, 58

Environment Argentina, Colombia, others

NWS Holdings Hong Kong, China China 23

Berlinwasser Germany China, Albania, Azerbaijan, 15

International Armenia, others

Manila Water Philippines Indonesia, Philippines, 13

Company Viet Nam

The retreat of multinational companies from the water and sanitation sector in

emerging markets has been partially filled by new emerging regional and niche players.

The new private investment players come from diverse backgrounds, including, among

them, water construction or engineering companies; industrial conglomerates; and local

companies seeking to expand or diversify beyond their borders. There has been a rise in

joint ventures between local and regional water operators with international operators.

Notably, regional investors have been investing in countries of their ethnic origin. The

new regional investors are focusing on water and sanitation projects within and from

their “own countries”, as diaspora investments. Among them are Manila Water, NWS

Holding, and other smaller investors from Singapore and Malaysia. The future of private

investment in water and sanitation in Asia and the Pacific may depend on these new,

regional, small private investors.

IV. REGRESSION ANALYSIS

The withdrawal of multinational water operators and the focus of private

investment in high-income countries provides the context for a regression analysis to

determine if a causal relationship exists between these factors and the level of private

investment in water and sanitation in Asia and the Pacific. The regression model

presents the main factors that determine the level of private investments in the water

and sanitation sector in the Asia-Pacific region between 1993 and 2017.

Data sources

Country-level data on “water and sewerage”, “treatment plant” and “water utility”

were obtained from the World Bank Private Participation in Infrastructure database. The


78

time-series data spans the period 1993–2017 (first half). The database provides

information on total investments in the sector along with private share in these

investments, making it possible to calculate the total private investments in projects

across countries.

Data on “access to improved water source” and “access to improved sanitation”

are from the WHO/UNICEF Joint Monitoring Programme for Water Supply, Sanitation

and Hygiene (WASH) database. The database covers the period 1993–2017. Data on

population, gross domestic product (GDP), gross domestic product per capita and

growth rate for Asia and the Pacific countries are from the World Development Indicators

database for the period 1993–2017.

As no comprehensive data are available on the diaspora, a dummy variable is

used by analysing the management board structures and other regional characteristics

of the investing private sector firm. For instance, if the management board of a country

incudes several members having host country origins, the investments are classified as

a diaspora investment, and the dummy variable is assigned a coefficient of 1. Similarly,

other subjective criteria (such as CEO birth country) was used to decide if an investment

falls in the category of diasporic investment.

The data were cleaned and checked for consistency. Once all the data were

collected, suitable fields were used to build a comprehensive dataset, which could be

used for statistical testing. The primary model was made based on panel data. Although

all possible efforts are made to complete the missing data, the information for total

investments and share of private players in those investments were not available for

many projects.

Regression equation

Using panel data, the relationship between total private investment and GDP,

population, and a diaspora proxy is examined, using the behavioral equation:

PI = f (GDPPC, population, diaspora)

where, PI = private investment in natural logarithm

GDPPC = gross domestic product per capita in natural logarithm

Population = population in natural logarithm

Diaspora = dummy for diaspora in investing country


79

Regression results

The regression provides the following result:2

PI = 2.297 + 0.385 (GDPPC) – 0.334 (Population) – 0.563 (Diaspora)

(3.662) (5.487) (–9.433) (–3.394)

Notes: (1) The figures in brackets represent the respective t-statistics.

(2) All the coefficients are significant at the 5 per cent level.

(3) Multiple R2 is equal to 0.399 and R2 is equal to 0.159.

The regressions yield a relatively low R2. This is mainly because private

investments in the water and sanitation sector are affected by many factors, above and

beyond the three independent variables modelled here. Such factors as governance,

ability to set prices, local government support, water supply availability and conditions of

physical infrastructure could affect the decision to invest far more than income and

population alone.

The regression shows that an increase in gross domestic product per capita of

a country affects private investment in water and sanitation in that country positively.

Every 1 per cent increase in GDPPC increases private investment (PI) in water and

sanitation by roughly one third in that country. The positive correlation between gross

domestic product and private investment is consistent with the findings that

upper-middle-income countries are the major beneficiaries of private investments.

Upper-middle-income countries also have a higher share in the number and the total

value of private investments. Countries with higher GDP per capita provide a better risk

return profile to private investors – they rank higher in terms of “ability to pay”.

Accordingly, private investors can better recover their investment and earn a return.

These countries also usually have better financial and judicial institutions, making their

respective markets more reliable for private investors. This raises serious challenges for

low- and middle-income countries. Low-income countries are especially a concern, as

the data show that less than 1 per cent of the total projects were directed to these

countries between 1993 and 2017.

2 Using time-series data, and deleting the dummy variable for diaspora, the regression produces a similarresult:

TotPI = –3.772 + 0.658 (GDPPC) + 0.173 (Population)(–2.199) (4.954) (2.406)

The time-series regression shows that GDP per capita and population positively affects privateinvestment in the country. All the variables in the above equation are significant at the 5 per cent level.The model has a multiple R squared of 0.372 and R squared of 0.139.


80

The regression also shows that there is a negative correlation between population

size and the level of private investment in water and sanitation in that country. Private

investments in countries with a large population are likely to be of a smaller size than

private investments going to a country with a smaller population. This does not mean

that countries with large populations receive smaller total private investments. Running

the regression using country-level time-series data show that the population is in fact

positively related to the amount of private investment. Thus, the negative sign in panel

data regression results show that private investors prefer to make smaller, more spread

out investments in countries with larger populations, such as China and India. This is

verified by data that indicate private investments in highly populated countries are more

likely to be smaller, but are more inclined to be targeting local governments and

municipalities.

The regression shows that diaspora investments have a negative relationship with

private investment in the host country. Private investment received from a country in

which its diaspora is present will likely be smaller as compared to investments received

from countries in which no diaspora is present. This result reflects that most of these

diaspora investments are regional in nature, and unlike other foreign investments, these

are small regional players making smaller investments in their “home” countries. During

the period 1993–2017, the median size of private diaspora investment was $16 million,

about half the size of non-diaspora investments, which averaged $31 million in value

terms.

V. CONCLUSION – THE WAY FORWARD

Meeting the water and sanitation needs of Asia and the Pacific and achieving the

2030 Sustainable Development Goal of universal access to safe and affordable drinking

water along with adequate sanitation requires urgently development and strengthening

of mechanisms to finance infrastructure and services in the sector. This is especially true

for sanitation. To date, private financing has not lived up to its potential in catalysing

universal access and connection to water and sanitation services. Despite the low

interest rate environment, water and sanitation projects have not been very successful in

mobilizing private capital. Public water service providers have typically low, if any,

private finance mobilization capacity. Even for the few existing corporate water supply

and sanitation providers, it is rare for them to borrow from commercial lenders because

of weak incentives or poor creditworthiness or both. Only 15 per cent of water utilities in

developing countries are currently commercially viable, meaning that they can cover

their operation cost and generate a surplus that can be used for other financial needs

(Kolker and others, 2016). Local commercial banks have limited experience in financing

water and sanitation infrastructure, perceiving it to be extremely risky. Moreover, the

financial sector in many developing countries is very thin and lacks the capacity to offer

long-term loans at affordable rates. As such, the recourse for governments has typically


81

been to source debt financing from international financial institutions, which offer

long-term loans at low concessional rates. Unfortunately, this approach neglects the

development of the private sector, which could play a significant role in financing water

and other infrastructure, in particular because commercial loans can be made timelier

and in local currencies, foregoing foreign currency risks. Many developing economies,

however, still lack domestic financial markets that have the capacity to offer affordable

long-term local currency financing. Emerging market countries in East and South Asia

are gaining ground by developing more robust financial sectors, but exposure to the

water and sanitation sector is still low and needs to improve.

Private investment in water and sanitation has dramatically shifted over the past

10 years. Since peaking in 2007, it has been volatile in terms of size and number. While

73 new projects in Asia and the Pacific were undertaken in the sector in 2007, no more

than 30 projects per year had been initiated in each of the preceding years. The decline

in private investments has also been accompanied by a shift in the type and size of

investments taking place. The “new” private investments are increasingly concentrated

in a few select wealthy countries and municipalities; and the investments are being

bankrolled by smaller, regional investors, including diaspora investors.

Bearing in mind these post-2007 features, the regression analysis was intended to

analyse factors that affected private investments in water and sanitation in Asia and the

Pacific between 1993 and 2017. The regression showed that GDP and population size

positively affected investment in water and sanitation in country, while diaspora had

a negative effect. This is worrying for low-income countries, which stand to benefit the

most from private investment, but have been on the receiving end of less than 1 per cent

of the total projects in the sector. The regression analysis confirmed that investors

preferred to diversify their investments across many smaller valued projects in countries

with larger populations. It also showed that diaspora investment tended to be smaller

than otherwise. Since 2007, investments from smaller, regional diaspora investors have

been significantly higher than non-diaspora investments.

This paper has illustrated the present trends in private sector investment in the

water and sanitation sector in Asia and the Pacific. The need for greater focus on

improving sanitation services in the region, especially for the least developed and

low-income emerging economies, is highlighted. The huge financing gap requires more

innovative financing that can only come by attracting private sector capital to support

development objectives and by repurposing public sector financing instruments to

address persistent development deficits. The Asian and Pacific region must focus on

developing and strengthening mechanisms to finance sanitation infrastructure that will

enable it to reach the Sustainable Development Goal of universal access to safe and

affordable drinking water and adequate sanitation by 2030. A new financing paradigm

needs to be built around partnerships between governments and the public and private

sectors by mobilizing commercial lenders, raising credit-worthiness of service providers


82

and blending public and private financial resources to invest in sanitation infrastructure.

Greater attention is needed to empower regional and local governments to develop

policies and norms for financial frameworks and investments in decentralized projects.

The leadership of local governments and municipalities in framing policies and in

attracting increasing investment for water and sanitation infrastructure is crucial in this

regard. Additional mechanisms to attract and mobilize regional investments in the water

and sanitation sector should be considered.


83

REFERENCES

Asian Development Bank (ADB) (2017). Meeting Asia’s Infrastructure Needs. Manila.

Hahm, Hongjoo (1996). Private sector financing of water and sanitation. Draft working paper (Mimeo).Washington, D.C.: World Bank.

Hutton, G., and M. Varughese (2016). The costs of meeting the 2030 Sustainable Development Goaltargets on drinking water, sanitation, and hygiene. Water and Sanitation Program TechnicalPaper. Washington, D.C.: World Bank. Available at http://documents.worldbank.org/curated/en/415441467988938343/The-costs-of-meeting-the-2030-sustainable-development-goal-targets-on-drinking-water-sanitation-and-hygiene.

Kolker, J., and others (2016). Financing options for the 2030 water agenda. World Global PracticeKnowledge Brief, November. Available at https://openknowledge.worldbank.org/bitstream/handle/10986/25495/W16011.pdf?sequence=4&isAllowed=y.

Organization for Economic Cooperation and Development (OECD) (2009). Private Sector Participation

in Water Infrastructure: OECD Checklist for Public Action. Paris. Available at www.oecd.org/env/resources/42350657.pdf.

(2010). Innovative Finance Mechanisms for the Water Sector. Paris.

United Nations, Economic and Social Commission for Asia and the Pacific (ESCAP), United NationsHuman Settlement Programme (UN-Habitat), and Asian Institute of Technology (2015). Policy

Guidance Manual on Wastewater Management with a Special Emphasis on Decentralized

Wastewater Treatment Systems. Bangkok. Available at www.unescap.org/resources/policy-guidance-manual-wastewater-management.

Winpenny, J. (2003). Financing Water for All. Report of the World Panel on Financing Water

Infrastructure. Available at www.oecd.org/greengrowth/21556665.pdf.

World Health Organization (WHO) (2017). UN-Water Global Analysis and Assessment of Sanitation and

Drinking-Water (GLAAS) 2017 Report: Financing Universal Water, Sanitation and Hygiene

under the Sustainable Development Goals. Geneva. Available at http://apps.who.int/iris/bitstream/10665/254999/1/9789241512190-eng.pdf?ua=1.

IMPACT OF FOOD INFLATION ONHEADLINE INFLATION IN INDIA

Anuradha Patnaik*

A commonly held belief in the 1970s was that price indices rise because oftemporary noise, and then revert after a short interval (Cecchetti andMoessner, 2008). Accordingly, policy should not respond to the inflationbecause of these volatile components of the price indices. This led to thedevelopment of the concept of core inflation (Gordon, 1975), which isheadline inflation excluding food and fuel inflation. It was strongly believedthat in the long run, headline inflation converges to core inflation and thatthere are no second round effects (that is an absence of core inflationconverging to headline inflation). In recent years, however, majorfluctuations in food inflation have occurred. This has become a majorproblem in developing countries, such as India, where a large portion of theconsumption basket of the people are food items. Against this backdrop, inthe present paper, an attempt is made to measure the second round effectsstemming from food inflation in India using the measure of Grangercausality in the frequency domain of Lemmens, Croux and Dekimpe (2008).The results of empirical analysis show significant causality running fromheadline inflation to core inflation in India and as a result, the prevalence ofthe second round effects. They also show that food inflation in India is notvolatile, and that it feeds into the expected inflation of the households,causing the second round effects. This calls for the Reserve Bank of Indiato put greater effort in anchoring inflation expectations through effectivecommunication and greater credibility.

JEL classification: E31, E50

Keywords: core inflation, monetary policy, food inflation, second round effects, inflation

expectations

85

* Anuradha Patnaik, PhD, Associate Professor, Mumbai School of Economics and Public Policy, Universityof Mumbai, India (email: [email protected]).


86

I. INTRODUCTION

A commonly held belief in the 1970s was that price indices rise because of

temporary noise, resulting from volatile food or fuel prices, and then reverted after

a short interval (Cecchetti and Moessner, 2008). This led to the development of the

concept of core inflation or baseline inflation (Gordon, 1975), which is primarily defined

as the aggregate inflation or the headline inflation excluding the food and fuel inflation

(Eckstein, 1981; Blinder, 1982; Thornton, 2007; Wynne, 2008; among others). The

emphasis on core inflation was motivated by the fact that historically food and fuel

inflation have been correcting themselves in the short run. This means that there are no

second round effects of food and fuel inflation such that a disruption in the headline

inflation caused by food inflation dies out in the short run, and the core inflation remains

unaffected. As a result, headline inflation is expected to eventually converge to the core

inflation or the underlying trend inflation (Clark, 2001) and that policy should not respond

to the inflation because of the volatile components of the price indices. It is important to

note that a “core measure of inflation is not an end in itself, but rather a means to

achieve low and stable inflation by serving as a short-term operational guide for

monetary policy” (Raj and Misra, 2011). Many economists also believe that as core

inflation is a measure of the underlying trend in inflation, it may also be important in

projecting inflation (Freeman, 1998; Goyal and Pujari, 2005).

Contrary to the above viewpoint, however, research in recent years has indicated

that in low-income countries where food comprises a major portion of the consumption

basket, food prices have become more steadfast. Similar problems were apparent even

in developed countries following the rise in food prices over the period 2003-2007,

during which food price shocks were transmitted into the non-food prices also as a result

of the beginning of the breakdown of the relationship between core inflation and

headline inflation (Walsh, 2010). Both started to diverge, implying that the so-called

volatile component, food inflation, is no longer volatile. This not only obstructed the

smooth functioning of monetary policy, but it also resulted in distortions in inflation

forecasts of central banks and consequently, the inflation expectations. Needless to

mention, it is essential that monetary policy should be aimed at preventing the second

round effects of higher food prices on inflation expectations and wages, and thereby

control future inflation (Cecchetti and Moessner, 2008). Alternatively, if monetary policy

is unsuccessful in blocking the second round effects because of food inflation, the

expectations of future inflation by households and firms would be underestimates or

overestimates of future inflation. This would create a wedge between the actual inflation

and expected inflation and eventually lead to ineffective inflation targeting and loss of

credibility of the central bank.

The above analysis is extremely relevant in the case of India where the weight of

food items taken together is 47.51 in the consumer price index combined (CPI-C,

hereafter), which is the official measure of inflation. The significance of the issue is

Impact of food inflation on headline inflation in India

87

amplified further as monetary policy in India has been altered by adopting flexible

inflation targeting as the new monetary framework. Under the inflation targeting

framework, the Reserve Bank of India must be forward looking and able to predict future

inflation accurately so as to achieve the targeted inflation by aligning the inflation

expectation to its projected inflation rate. When a central bank, such as the Reserve

Bank of India, communicates the future inflation forecast to the public, expectations are

formed based on these inflation forecasts only if the institution is credible. A central bank

earns credibility over a period of time if the projected inflation is close to the actual

inflation. Thus, the success of inflation targeting lies in how accurately central bank

forecasts future inflation (Blinder, 1999). Central banks use relevant models to forecast

future inflation (Benes and others, 2016), which usually do not incorporate the second

round effects of component inflation measures. As a result, the projected inflation does

not coincide with the actual inflation (figure 9, section III); this becomes detrimental for

the central bank, which loses its credibility in the long run and the inflation targeting

framework eventually collapses. As this phenomenon of the second round effects of

food inflation has serious policy implications in terms of formulating expectations in line

with the prediction of future inflation, the objective of the present study is to explore

empirically, in the frequency domain, the second round effects of food inflation, and then

the changing dynamics between the headline and core inflation because of persistent

food inflation in India. The rest of the paper is organized as follows: section II contains

a review of the literature. The definitional aspects of the inflation measures used, and

the trends in inflation are discussed in section III. The methodological details and data

used are discussed in section IV. The results of empirical analysis are reported in

section V. Finally, section VI includes a discussion of the results and concludes.

II. REVIEW OF THE LITERATURE

The concept of core inflation was developed in the 1970s when the Organization of

Petroleum Exporting Countries (OPEC) was at its height; it was realized that the

underlying trend in inflation had to be tracked for policy purposes, rather than the

headline or aggregate inflation. Since then, several studies on the relevance of core

inflation have been conducted. Among them are the following: Sprinkel (1975); Tobin

(1981); Eckstein (1981); Blinder (1982); Rich and Sendel (2005). The authors of these

studies are among the pioneers to propose that measured inflation can be split into

three parts: the core inflation; the demand inflation; and the shock inflation.

Since then, headline inflation is constructed by assigning different weights, for the

commodities entering the consumer price index/wholesale price index (CPI/WPI) basket,

which also includes food and fuel prices, and deriving the core inflation from the

headline inflation has become a matter of extensive research. Clark (2001) compared

five different measures of core inflation. Of the five measures, three measures (CPI

excluding food and energy, trimmed mean, and median CPI) were previously developed,


88

and the other two (CPI excluding energy, and CPI excluding eight components) were

developed by Clark. The study showed that the trimmed mean and CPI excluding fuel

were the superior measures of core inflation. While giving a detailed account of the

different measures of core inflation used by different central banks, Shiratsuka (2006)

tried to identify the core inflation for Japan by capturing the nature and size of

idiosyncratic disturbances behind CPI of Japan. A review of various approaches to

measure core inflation can be found in Wynne (2008), who linked these approaches in

a single theoretical framework called the stochastic approach to index numbers. He

concluded by saying that the different measures lack a well-formed theory of what these

measures of inflation need to capture. Similar research using the stochastic approach

was also conducted by Clements and Izan (1981; 1987), Bryan and Pike (1991), and

Bryan and Cecchetti (1993; 1994).

In the past decade, recurrent and persistent spikes in food prices were

experienced globally (Bhattacharya and Gupta, 2015), resulting in questioning the

validity of core inflation as the short-term guide for monetary policy. Many studies, such

as Thornton (2007), Walsh (2010), and Cecchetti and Moessner (2008), dealt with the

issue of whether core inflation can predict future headline inflation. While Cecchetti and

Moessner (2008) concluded that in a majority of the countries considered by them,

headline inflation converged to core inflation. Other studies, however, concluded that

whether headline inflation converges to core inflation depends on the measure of core

inflation. As Thornton (2007) explains, “alternative is to consider different measures of

‘core’ inflation and not rely on the measures that exclude simply food and energy

prices”.

Similar studies related to India are also available in which the average food

inflation was the highest among the emerging market economics during the period

2006-2014 (Bhattacharya and Gupta, 2015). Literature on food inflation in India and its

causes can be found in, among others, the works of Nair and Eapen (2012), Guha and

Tripathi (2014), Bhattacharya and Gupta (2015). Estimation of the second round effects

of the food inflation, using different variants of the core inflation of India are available in

research conducted by Raj and Misra (2011), who attempted to analyse seven

exclusion-based measures of core inflation for India with regard to volatility, persistence

and predictive power for headline inflation, using time-series techniques. The study

indicated that there were the second round effects in six out of the seven measures of

core inflation considered. It was, therefore, concluded that headline inflation, not core

inflation, should be the focus of monetary policy in countries such as India where food

and fuel comprise a major portion of the consumer basket. Bhattacharya and Gupta

(2015) analysed the causes and determinants of food inflation in India using time-series

tools. Significant pass through effects from food to non-food inflation was found, clearly

implying the presence of the second round effects. Anand, Ding and Tulin (2014)

estimated the second round effects using reduced form general equilibrium models.

They clearly showed the presence of the strong second round effects. Goyal and Baikar


89

(2015) concluded that the causality from headline inflation to core inflation occurred only

when the food inflation crosses double digits. Dholakia and Kadiyala (2018) concluded

that the second round effects were weak in the case of India beginning in 2012. It can

be seen that many researchers have attempted to estimate the second round effects of

food inflation on core inflation and have arrived at different conclusions.

“There is an ongoing debate on the direction of convergence between headline

inflation and core inflation and its probable impact on future course of monetary policy”

(Goyal and Parab, 2019, p.1). Inflation targeting framework requires the central bank to

be able to forecast the future inflation rates accurately (Blinder, 1999). Central banks set

their inflation target by announcing the projected inflation rate for the next quarter. The

projected inflation rate is then taken by the firms as the expected future inflation rate for

price and wage setting for the next time period, and by households for their consumption

and savings decisions (Goyal and Parab, 2019), only if the central bank is credible.

Thus, the central bank anchors inflation expectations through its inflation forecast and

the expected inflation, and holds the key to the success of inflation targeting by aborting

the second round effects of transitory shocks that could lead to persistent inflation

(Goyal and Parab, 2019). Against this backdrop, in many studies, there have been

attempts to explore if the core inflation can be used to forecast the headline inflation

correctly in the presence of high food inflation (Thornton, 2007; Walsh, 2010; Cecchetti

and Moessner, 2008). Misati and Munene (2015) estimated the second round effects of

food inflation on non-food inflation of Kenya and stressed the importance of

communication there by anchoring inflation expectations in order to mitigate the impact

of the second round effects on actual inflation. Thus, second round effects of food shock

affect the inflation forecast through the inflation expectations of households and firms.

As a result, accurate knowledge of the second round effects is very crucial to being able

to forecast inflation accurately.

The review of literature highlights the crucial importance of the knowledge of the

second round of food shocks (or any transitory shock) for the success of inflation

targeting monetary policy. Though a number of studies have been conducted on the

second round effects of food shocks for India in the time domain, in the present paper

the second round effects of food inflation, food inflation causing headline inflation, and

headline inflation causing core inflation are revisited in the frequency domain using the

methodology of Lemmens, Croux and Dekimpe (2008). The novelty and relevance of the

present study is that the magnitude of causality and the time taken for transmission of

shocks from food inflation to headline and core inflation will also be derived. To the best

of the author’s knowledge, similar studies in the frequency domain, under Indian

conditions have not been carried out. An attempt is also made to measure the

responsiveness of the inflation measures to monetary policy. These results throw light

on the crucial importance of communication by the central bank in the success of

monetary policy in an inflation targeting framework.


90

III. INFLATION: DATA SOURCES, DEFINITIONS AND TRENDS

Variables used and database

The variables used in the empirical analysis in section V are as follows:

(a) Monthly headline inflation measured as the year-on-year growth in the

consumer price index combined, CPI-C, sourced from the Ministry of Statistics

and Programme Implementation website, for the period January 2012 – June

2019.

(b) Monthly core inflation measured as the year-on-year growth in CPI-C (core),

estimated from CPI-C data using equation 1, for the period January 2012 –

June 2019 (details given in the next section on Definitions).

(c) Food inflation measured as the year-on-year growth in the CPI-C (food),

sourced from the Ministry of Statistics and Programme Implementation

website, for the period January 2012 – June 2019.

(d) Weighted monthly average call money rate (proxy for repo rate), sourced from

the Reserve Bank of India Handbook of Statistics on Indian Economy, for the

period January 2012 – June 2019.

(e) Inflations expectation of households, sourced from the Reserve Bank of India,

Inflation Expectations Survey of Households, June 2019.

A brief note on the variables used

The consumer price index combined (CPI-C) has been adopted as the official

measure of inflation by the Reserve Bank of India, as per the recommendations of the

Expert Committee to Revise and Strengthen the Monetary Policy Framework (Reserve

Bank of India, 2014). The Ministry of Statistics and Programme Implementation

publishes monthly figures of CPI-C and its components (for details see table 1). The CPI

measures change over time in the general level of prices of goods and services

acquired by the households for consumption, and is therefore used to represent the

retail price index of the country. The monthly price data for the items included in the CPI

consumption basket are collected from 1,114 markets in 310 selected towns and 1,181

selected villages by the National Sample Survey of India and the Department of Posts,

respectively, through web portals and then three indices, CPI-Rural, CPI-Urban, and

CPI-C are calculated and published by the Ministry of Statistics and Programme

Implementation. As CPI-C is the official measure of inflation in India for the present

study, it is also the measure of inflation for the empirical analysis. The broad

components and their respective weights in CPI-C, CPI-Rural and CPI-Urban are

reported in table 1. It is interesting to note that food and beverages are assigned the

highest weight in each of the price indices. This is followed by the miscellaneous item.

The weights are constantly changing with respect to the base year used.


91

Table 1. Components and their weights in the CPI-C, CPI-Rural and CPI-Urban

(base year, 2012)

Component/itemWeight in Weight in Weight in

CPI-Rural CPI-Urban CPI-C

Food and beverages 54.18 36.29 45.49

Pan, tobacco and intoxicants 3.26 1.36 2.02

Fuel and light 7.49 5.58 6.84

Housing – 21.67 21.67

Clothing 7.36 5.57 6.28

Miscellaneous 27.26 29.53 28.31

Source: India, Ministry of Statistics and Programme Implementation. Available at www.mospi.gov.in/.

The repo rate is the official instrument used for conducting monetary policy by the

Reserve Bank of India. For empirical purposes, however, the repo rate is very difficult to

handle (does not become stationary easily). Accordingly, the weighted average monthly

call money rate, which is the operating instrument of monetary policy of India and is

closely aligned with the repo rate has been used as a proxy for monetary policy in India.

Definitions

The present study

Changes in CPI-C for all commodities are treated as the headline inflation for

policy articulation, and within CPI-C, the “non-food items” inflation is considered the

core inflation in India (Mohanty, 2011). This implies that when prices of food and

beverages, pan, tobacco and intoxicants and fuel and power prices are excluded from

CPI-C, the results in the prices of core items in CPI-C are attained. The core inflation

represents the underlying trend of inflation as shaped by the pressure of aggregate

demand against the existing capacity. The non-core part of the headline inflation

constitutes the food inflation, and is usually considered to reflect the price movements

caused by temporary shocks or relative price changes (Laflèche and Armour, 2006). It is

important to note that the Ministry of Statistics and Programme Implementation also

publishes the food price index data for CPI-C; however, the core price index is not

readily available. Consequently, for the present study, the exclusion-based measure, as

suggested by Bhattacharya and Gupta (2015), is used to estimate the core inflation as

given in equation 1 below. The exclusion based core price index can be derived from

CPI-C as follows:

(1)


92

Where, w(fa) is weight of food and beverages in CPI-C, w(fp) is weight of pan, tobacco

and intoxicants in CPI-C, w(fu) is weight of fuel and light in CPI-C, and CPIC is

CPI combined.

For the present study, the headline inflation is measured as the year-on-year

difference in CPI-C; food inflation is measured as the year-on-year difference in the

CPI-food price index, and the core inflation is derived as the year-on-year difference in

the core price index derived from equation 1 above. Because of the year-on-year

differencing of the price data to deriving the inflation data, 12 observations were lost

from the original sample of 102 observations and the sample size was reduced to

90 observations.

The trends in inflation

Source: Author’s own calculations using CPI-C data derived from the Ministry of Statistics and Programme

Implementation. Available at www.mospi.gov.in/.

Figure 1 depicts the behaviour of the mean and standard deviation of the headline,

core and food inflation rates for January 2012 to June 2019. It can be clearly seen that

while the mean values of each of the inflation measures for the period is approximately

6 per cent (5.93 headline inflation, 5.96 core inflation and 6.03 food inflation), the

standard deviation varies across the three measures. It can be used to measure the

long-term volatility in a particular time series such that the higher the standard deviation

the higher the long-term volatility. It is interesting to note that while food inflation is

normally considered to be the most volatile component, its volatility as measured by the

standard deviation is the least at 1.7675, followed by headline inflation, at 2.69, in the

Figure 1. Mean and standard deviation of headline, core and food inflation

(measured as the year-on-year growth rate of the respectve CPI combined

measures) for the period January 2012 – June 2019

7

6

5

4

3

2

1

0

Headline inflation Core inflation Food inflation

Mean Standard deviation


93

Source: Author’s own calculations using CPI-C data derived from the Ministry of Statistics and Programme

Implementation. Available at www.mospi.gov.in/.

present sample. Core inflation is usually not considered to be volatile; however, in the

present case it emerges as the most volatile measure of inflation with a standard

deviation of 3.81. The 12-month rolling standard deviations of the three inflation

measures also depict a similar picture. As indicated in figure 2, the 12-month rolling

standard deviation of core inflation is more than the 12-month rolling standard deviation

of the headline inflation and the food inflation for the entire time period, December 2012

to June 2019. It is also interesting to note that the rolling standard deviation of food

inflation appears to be the least volatile component with the lowest standard deviation

throughout the rolling window.

Ideally, the mean of the headline inflation should be around the mean of core

inflation, but the standard deviation of the core inflation should be lower than the

standard deviation of the headline inflation (Reserve Bank of India, 2019). Accordingly, it

can be concluded that the core inflation appears to be the most volatile component of

headline inflation from the preliminary analysis.

From figure 3, it can be observed that the headline inflation and the core inflation

are more or less moving together. For a major part of the study sample, the core

inflation appears to be higher than the headline inflation. Post-November 2017, however,

the core inflation fell below the headline inflation. Figures 4 and 5 clearly show that the

food inflation does not meander either with the headline inflation or the core inflation. It

Figure 2. Twelve-month rolling standard deviation of headline, core and food

inflation rates (January 2012 to June 2019)

Headline inflation Core inflation Food inflation

3

2.5

2

1.5

1

0.5

0

Ja

n-2

01

3

Ap

r-2

01

3

Ju

l-2

01

3

Oct-

20

13

Ja

n-2

01

4

Ap

r-2

01

4

Ju

l-2

01

4

Oct-

20

14

Ja

n-2

01

5

Ju

l-2

01

5

Ap

r-2

01

6

Oct-

20

17

Ja

n-2

01

7

Ap

r-2

01

7

Ju

l-2

01

7

Ja

n-2

01

8

Ap

r-2

01

5

Oct-

20

15

Ja

n-2

01

6

Ju

l-2

01

6

Oct-

20

16

Ap

r-2

01

8

Ju

l-2

01

8

Oct-

20

18

Ja

n-2

01

9

Ap

r-2

01

9


94

Headline inflation (combined)

Food inflation (combined)

Ja

n-2

01

2

Au

g-2

01

2

Ma

r-2

01

3

Ma

y-2

01

4

Ju

l-2

01

5

Ap

r-2

01

7

No

v-2

01

7

Ju

n-2

01

8

Ja

n-2

01

9

Oct-

20

13

De

c-2

01

4

Fe

b-2

01

6

Se

p-2

01

6

20

15

10

5

0

-5

Perc

enta

ge

Source: Ministry of Statistics and Programme Implementation. Available at www.mospi.gov.in/.

Note: Inflation measured as year-on-year growth in the respective CPI combined (for the period January 2012 –

June 2019).

Figure 3. Headline versus core inflation, January 2012 to June 2019



June 2019).

Figure 4. Headline verses food inflation, January 2012 to June 2019

Headline inflation (combined)

Core inflation (combined)

Jan-2

012

Oct-

2012

Jul-2013

Apr-

2014

Jan-2

015

Oct-

2015

July

-2016

Apr-

2017

Jan-2

018

20

15

10

5

0

-5

Oct-

2018

Perc

enta

ge


95



June 2019).

Figure 5. Core versus food inflation, January 2012 to June 2019

is also evident that the food inflation has been very high in the past few years, especially

since 2016, however, it has been declining steadily post-October 2018.

The Ministry of Statistics and Programme Implementation publishes the price

indices for rural areas, urban areas, and also the rural and urban combined. It is

interesting to note from figure 6 that for a major part of the study period, the headline

inflation in the rural areas has been higher than the headline inflation in the urban areas.

Since 2018, however, the headline inflation in the urban areas seems to be higher than

the headline inflation in the rural areas. In case of core inflation (figure 7), except for

brief periods, there is no major rural-urban divergence. Figure 8 indicates that the food

inflation has been higher in rural areas than in urban areas for three consecutive years,

from January 2015 to January 2018. Accordingly, there seems to be considerable

divergence in the behaviour of the inflation rates between the rural and the urban areas.

As the present study attempts to test the second round effects of rising food

inflation against the backdrop of inflation targeting framework in India, a peek into the

inflation projection or forecasts of the Reserve Bank of India and the actual inflation of

India is warranted. Figure 9 depicts the one quarter ahead inflation forecast of the

Reserve Bank of India (announced by the Monetary Policy Committee and available in

its reports) and the actual inflation measured as the year-on-year growth in the CPI-C

for the period January 2015 to December 2018.

Food inflation (combined)

Core inflation (combined)

20

-10

10

0

Jan-2

012

Oct-

2012

Jul-2013

Apr-

2014

Jan-2

015

Oct-

2015

Jan-2

018

Oct-

2018

Jul-2016

Apr-

2017

Perc

enta

ge


96

Core inflation (rural) Core inflation (urban)

18

14

10

8

0

-2

Jan-2

012

Jun-2

012

Apr-

2013

Nov-2

012

Sep-2

013

Feb-2

014

Jul-2014

Dec-2

014

May-2

015

Oct-

2015

Mar-

2016

Aug-2

016

Jan-2

017

Jun-2

017

Nov-2

017

Apr-

2018

Sep-2

018

Feb-2

019

16

12

6

4

2

Perc

enta

ge

Headline inflation (rural) Headline inflation (urban)

14

10

6

0

Jan-2

012

12

8

4

2

Jun-2

012

Apr-

2013

Nov-2

012

Sep-2

013

Feb-2

014

Jul-2014

Dec-2

014

May-2

015

Oct-

2015

Mar-

2016

Aug-2

016

Jan-2

017

Jun-2

017

Nov-2

017

Apr-

2018

Sep-2

018

Feb-2

019

Perc

enta

ge

Figure 6. Headline inflation (rural versus urban)


Note: Inflation measured as the year-on-year growth in the respective CPI.

Figure 7. Core inflation (rural versus urban)




97

6

5

4

3

2

1

0

Ap

r-2

01

5

Ju

n-2

01

5

Au

g-2

01

5

Oct-

20

15

De

c-2

01

5

Fe

b-2

01

6

Ap

r-2

01

6

Ju

n-2

01

6

Au

g-2

01

6

Oct-

20

16

De

c-2

01

6

Fe

b-2

01

7

Ap

r-2

01

7

Ju

n-2

01

7

Au

g-2

01

7

Oct-

20

17

De

c-2

01

7

Fe

b-2

01

8

Ap

r-2

01

8

Ju

n-2

01

8

Au

g-2

01

8

Oct-

20

18

De

c-2

01

8

Bi-monthly policy reviews

Actual CPI inflation One-quarter ahead projections

Perc

enta

ge

Food inflation (rural) Food inflation (urban)

14

10

8

0

Jan-2

012

May-2

012

Jan-2

013

Sep-2

012

May-2

013

Jan-2

014

May-2

014

Sep-2

014

Jan-2

015

May-2

015

Jan-2

016

May-2

016

Jan-2

017

May-2

017

Sep-2

017

Jan-2

018

May-2

018

Jan-2

019

12

6

4

2

Sep-2

013

Sep-2

015

Sep-2

016

Sep-2

018

May-2

019

Perc

enta

ge

Figure 9. Actual versus projected inflation of the Reserve Bank of India

for the period quarter 1, 2015 – quarter 3, 2018

Figure 8. Food inflation (rural versus urban)



Source: Reserve Bank of India, “Inflation forecasts: recent experience in India and a cross country experience”, Mint

Street Memo, No. 19. Available at https://rbidocs.rbi.org.in/rdocs/MintStreetMemos/19MSM02052019.pdf.


98

It can be clearly seen that the actual inflation diverges widely from the projected

inflation only when there is a food price shock (positive or negative). The food price

shock is positive when food prices are rising, and negative when food prices are falling.

It is also evident that even if the food price shock continues for a prolonged period, the

projected inflation (headline inflation) is either underestimated or overestimated

systematically. The phases of food price shocks and the divergence between actual and

projected inflation are highlighted in the figure 9. This clearly implies that the second

round effects of the food price shock (food prices feeding into headline price index,

which in turn gets transmitted to the core inflation), are the reason behind the widened

forecast error of the inflation forecast of the Reserve Bank of India.

IV. METHODOLOGY

As the paper purports to identify the second round effects of rising food inflation in

India the following research questions are dealt with:

(a) What are the implications of food inflation for headline and core inflation? Are

there the second round effects?

(b) Is food inflation volatile?

(c) Are inflation expectations anchored in India?

(d) Do the inflation rates respond to monetary policy?

The first and fourth questions will be tested by estimating the Granger causality in

the frequency domain (using the methodology of Lemmens, Croux and Dekimpe

(2008)). The detailed methodology is as follows:

Granger causality is a commonly used technique to measure the causal

relationship between variables. The present study employs a spectral density-based

Granger causality test as given by Lemmens, Croux and Dekimpe (2008). The merit of

this approach is that a more complete picture of the causal flow is attained by

decomposing Granger causality over different time horizons. This facilitates the

understanding of variations in the strength of causal flow between the two variables over

the spectrum (Lemmens, Croux and Dekimpe, 2008). The spectrum can be interpreted

as a decomposition of the series variance by frequency. Suppose, Xt and Yt are the two

time series. Then to test for Granger causality between these time series, the white

noise innovations series ut and vt derived after applying autoregressive moving average

(ARMA) filters to Xt and Yt become the main building block. Let Su(λ) and Sv(λ) be

the spectrum of the innovation series of Xt and Yt, respectively at frequency λ ε [-π, π ]

given as

and (2)


99

Where γu = cov(ut ut–k ) and γv = cov(vt vt–k ) (3)

are the autocovariances of ut and vt at lag k. It is important to note that as the

innovations series are the white noise process (WNP), their spectra are constant

functions represented as Su(λ) = Var(ut)/2π and Sv(λ) = Var(vt)/2π, respectively. The

cross spectrum between the two innovation series is the covariogram of the two series

in the frequency domain. It is a complex number, defined as

(4)

Where Cuv(λ) is the cospectrum or the real part of the cross spectrum and the

quadrature spectrum or the imaginary part is given by Quv(λ).γuv = cov(utvt ), gives

the cross covariance between ut and vt at lag k. The cross spectrum can be

non-parametrically estimated as follows:

(5)

Where γuv = cov(utvt ), the empirical cross covariance with, wk , the window weights for

k = –M to +M. The weights are assigned according to the Barlett weighting scheme,

where wk = 1 – —, and M is the maximum lag order, which is often chosen equal to the

square root of the number of observations following Diebold (2001). Having derived the

cross spectrum the coefficient of coherence huv (λ) can be computed. It is defined as

(6)

Lemmens, Croux and Dekimpe (2008) have shown that under the null hypothesis

that huv (λ) = 0, the estimated squared coefficient of coherence at frequency λ with 0(λ)<π when appropriately rescaled, converges to a chi-squared distribution with two

degrees of freedom. This coefficient of coherence, however, is only a symmetric

measure of association between the two time series and does not indicate anything

about the direction of relationship between the two processes. For the directional

relationship, Lemmens, Croux and Dekimpe (2008) have decomposed the cross

spectrum into three parts: (1) Su⇔

v the instantaneous relation between ut and vt, (2) Su⇒

v

the directional relationship between vt and lagged values of ut, and (3) Sv⇒

u the

directional relationship between ut and lagged values of vt, i.e.

(7)

(8)

|k|

M

Λ Λ


100

Lemmens, Croux and Dekimpe (2008) have proposed the spectral measure of

Granger causality based on the key null that Xt does not Granger cause Yt if and only if

γuv (k) = 0 for k < 0, hence only the second part of the equation 8 becomes important, i.e.

(9)

Therefore, the Granger coefficient of coherence will be

(10)

with the Su⇒

v given by equation 10. In the absence of Granger causality hu⇒

v (λ)= 0, for every frequency between 0 and π. A natural estimator for the Granger coefficient

of coherence at frequency λ is

(11)

with weights wk for k ≥ 0 put equal to zero in Su⇒

v (λ) (Lemmens, Croux and Dekimpe,

2008). The distribution of the estimator of the Granger coefficient of coherence can be

derived from the distribution of the coefficient of coherence. Under the null hypothesis

that hu⇒

v (λ) = 0, for the squared estimated Granger coefficient of coherence at

frequency λ, with 0 < λ < π

(12)

where n’=T/∑–1

w2 and →d implies convergence in distribution. As the weights wk with

a positive index k are set equal to zero when computing Su ⇒

v (λ), only the wk with

negative indices are in effect taken into account. Thus, the null hypothesis of no Granger

causality at frequency λ versus hu⇒

v(λ) > 0, is then rejected if

(13)

with X 2 being the 1-α quantile of the chi squared distribution with two degrees of

freedom (Hatekar and Patnaik, 2016).

The causality results of the inflation measures of the present study are helpful in

understanding the first round and second round effects of these measures of inflation.

This implies that for the first round effects to exist, there should be a causal flow from

Λ

k =–M k

2,(1–α)

Λ

Λ

Λ


101

food inflation to headline inflation, and for the second round effects to exist, there should

be a causal flow from headline inflation to core inflation and headline inflation should not

converge to core inflation.

For the second question, the above-mentioned inflation measures are tested for

presence of autoregressive conditional heteroskedasticity (ARCH) or generalized

autoregressive conditional heteroskedasticity (GARCH) effects using the ARCH-LM test.

ARCH/GARCH models are models of volatility in which the conditional volatility of the

residuals of a mean equation (which can be either of the following process: an

autoregressive (AR) process/moving average process/autoregressive moving average

(ARMA) process/OLS equation) is modelled as an AR or an ARMA process.

V. EMPIRICAL RESULTS

Steps of empirical analysis

Step I: Stationarity test of the variables used in the empirical analysis

All the variables (inflation measures and the weighted average call money rate)

used in the empirical analysis were found to be stationary in level (results not reported).

Step II: ARMA filtering of the inflation measures and the weighted average call money

rate to derive the innovations series for each variable

Table 2 gives the relevant ARMA models for each of the variables used in the

empirical analysis derived using the Box Jenkins methodology. The innovation series for

each variable is then derived as the residual series derived by subtracting the fitted

values of the variables from the actual values. The residual series had become a white

noise process as authenticated by the Box Pierce Test (results not reported here).

Table 2. The autoregressive moving average (ARMA) models

of the variables used in empirical analysis

Variable name ARMA model

Headline inflation Moving average(1)

Food inflation Autoregressive(1)

Core inflation Moving average(1)

Weighted average call money rate Autoregressive moving average(1,1)


102

Step III: Estimating the Granger causality in the frequency domain

Using Granger causality in the frequency domain, an attempt is made to

investigate the first round effects and second round effects of shocks attributable to food

inflation. The first round effects imply that there is a direct effect or causal flow of food

inflation shock to headline inflation. The second round effects imply that from the

headline inflation, there is a causal flow of the shock from to the core inflation (Portillo

and others, 2016). As a result, in the present study, an attempt is made to estimate the

Granger causality between the following inflation measures:

1) CPI-C (food inflation) to CPI-C (headline inflation)

2) CPI-C (headline inflation) to CPI-C (core inflation)

A statistically significant causality from headline inflation to core inflation

establishes the prevalence of the second round effects.

After the ARMA filtering, the number of observations of each series changed. In

order to maintain uniformity, 88 observations are used to construct the relevant Granger

coefficient of coherence. Hence, M, the maximum lag till which covariances have been

estimated, is (the square root of the nearest perfect square of the number of

observations) 9. It is important to mention here that based on the frequency of cycles,

short term is defined as cycles in the frequency range of 2 to 3.14, medium term as

cycles with frequency range of 1 to 2, and long term as cycles with frequency less than 1.

The Granger coefficient of coherences has been estimated in the frequency

domain. Therefore, a plot of the coefficient of coherence across various frequencies is

intuitive. In each of the plots on the Granger causality in the frequency domain, the

Granger coefficient of coherence has been plotted on the y-axis and the frequency has

been plotted on the x-axis. Figure 10 depicts the Granger causality from the food

inflation to headline inflation. The straight line parallel to x-axis is the relevant Granger

coefficient of coherence at the relevant significance level.

It can be observed that the Granger causality from the food inflation to headline

inflation lies above the 5 per cent significance level. This implies that the Granger

causality from food inflation to headline inflation is statistically significant at all

frequencies. The maximum causality of 0.67 is in the long run with cycles of

frequency 1. Thus, when food inflation rises, headline inflation also depicts an upward

trend.

It is interesting to note from figure 11 that even the Granger causality from headline

inflation to core inflation is statistically significant at all frequencies as the plot of the

coefficient of coherences lies above the 1 per cent significance level. The maximum

causality 0.94 occurs at a frequency of 2, namely cycles spanning 28 months or within

two and a half years of the occurrence of the shock. This result establishes the

prevalence of the second round effects of the food shocks.


103

Source: Author’s own calculations using data retrieved from the Ministry of Statistics and Programme Implementation.

Available at www.mospi.gov.in/.

Note: GC, Granger causality.

Figure 10. Granger causality in the frequency domain

from food inflation to headline inflation

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

GC(f-hl) 5% significance level

om

eg

a 0

0.2 0.4

0.6

0.8 1

1.2

1.4

1.6

1.8 2

2.2 2.4

2.6 2.8 3





from headline inflation to core inflation

1

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

0.9

GC(hl-c) 1% significance level

om

ega 0

0.2

0.4

0.6 0.8 1

1.2

1.4

1.6 1.8 2

2.2

2.4

2.6

2.8 3


104

Step IV: Testing the inflation measures for presence of volatility

The inflation measures were tested for presence of volatility using the ARCH-LM

test, the null of no ARCH effects for all the three inflation measures was not rejected

(table 3).

Table 3. ARCH-LM test results of the inflation measures

Inflation measure ARCH-LM statistic P-value

Food inflation 7.8499 0.6434

Core inflation 4.1677 0.9394

Headline inflation 5.7124 0.8388

Step V: Quantifying the gap between the actual inflation and the households’ inflation

expectations

The Reserve Bank of India conducts and publishes the Inflation Expectations

Survey of Households on a quarterly basis. This survey is conducted in eighteen cities

of the country and derives qualitative and quantitative responses from the households

on current, three months ahead, and one year ahead inflation rate. It is important to note

that the inflation expectations influence the wage bargaining process and the future

inflation. Under an inflation targeting framework, the Reserve Bank of India has to

anchor inflation expectations of the households to achieve the targeted inflation with

a minimum cost of disinflation. The Inflation Expectations Survey of Households

(figure 12) reveals that the inflation expectations of the households for the current, three

months ahead, and one year ahead periods are considerably higher than the actual

inflation, especially since 2013. While the average actual inflation for the entire sample

period, March 2012 to March 2019, was 6 per cent, the mean expected inflation for the

current, three months ahead and one year ahead were 9.77 per cent, 10.22 per cent,

and 10.87 per cent, respectively.

Further insights into the gap between the actual and expected inflation of the

households can be derived by estimating the mean error (ME) and root mean square

error (RMSE) of the inflation expectations of the households with respect to the actual

inflation. ME and RMSE are estimated as given in equations 14 and 15 below:

(14)

(15)


105

Source: Reserve Bank of India, “Inflation Expectation Survey of Household, June 2019”.

Figure 12. Actual inflation versus household inflation expectations

(for mean current, mean three months ahead and mean one year ahead)

Table 4. Mean error and root mean square error

of the inflation expectations of households

Mean current Mean three months Mean one year

inflation expectation ahead inflation ahead inflation

expectation expectation

Mean error -3.77913 -4.22051 -4.87224

Root mean square error 4.180188 4.532671 5.091446

Source: Author’s own calculation using data on inflation expectations of households, derived from the Reserve Bank of

India, “Inflation Expectation Survey of Household, June 2019”.

From table 4, it can be clearly seen that the mean error in all the three cases of

expected inflation for the entire sample is very high. As the value of mean error is

negative, the households have been overestimating inflation. The root mean square

error also clearly highlights a similar picture and clearly reveals that, on an average, the

expected inflation is 3 to 4 per cent above the actual inflation. It can also be seen that as

the forecast horizon increases, the error is also increasing.

Actual inflation

Mean three months ahead IE Mean one year ahead IE

Mean current IE

16

14

12

0

Infla

tio

n

10

8

6

4

2

Quarter

Ma

rch

-20

12

Ju

ly-2

01

2

No

ve

mb

er-

20

12

Ma

rch

-20

13

Ju

ly-2

01

3

No

ve

mb

er-

20

13

Ma

rch

-20

14

Ju

ly-2

01

4

No

ve

mb

er-

20

14

Ma

rch

-20

15

Ju

ly-2

01

5

No

ve

mb

er-

20

15

Ma

rch

-20

16

Ju

ly-2

01

6

No

ve

mb

er-

20

16

Ma

rch

-20

17

Ju

ly-2

01

7

No

ve

mb

er-

20

17

Ma

rch

-20

18

Ju

ly-2

01

8

No

ve

mb

er-

20

18

Ma

rch

-20

19


106

The gap between the actual and expected inflation may be because “the food and

fuel shocks have high persistence on households’ inflation expectations, which impart

stickiness to core inflation” (Dholakia and Kadiyala, 2018). The second round effects

found in step four are the outcome of the unanchored inflation expectations; it is clear

that the Reserve Bank of India is failing to anchor inflation expectations of the

households.

Step VI: Estimating the Granger causality from the monetary policy to the inflation

measures

Against the backdrop of unanchored inflation expectations and the prevalence of

the second round effects of food inflation, it would be intuitive to test if monetary policy is

able to influence these inflation measures. As a result, the Granger causality in the

frequency domain was estimated from the call money rate (proxy for repo rate, which is

the policy rate of the Reserve Bank of India) to all the three inflation measures. The

results of the causal flow are depicted in figures 13, 14 and 15.

Figure 13 shows that the Granger causality from weighted average call money rate

to food inflation is statistically significant from the frequency 2.2 to 3.14 and 0.6 to 1.8,

which are cycles of short-term frequency and medium-term frequency. The maximum

causality is at cycles with a frequency of 1.2 in the given sample. This implies that





from call rate to food inflation

1

0.8

0.6

0.4

0.2

0

Om

ega

0.2

0.6 1

1.4 1.8

2.2

2.6 3

GC(CR-f) 5% significance level


107


from call rate to core inflation







GC(CR-c) 1% significance level

0.6

0.5

0.4

0.3

0.2

0.1

0

Om

ega

0.2

0.6

1.4 1.8

2.2

2.61 3


from call rate to headline inflation

GC(CR-hl) 5% significance level

0.5

0.4

0.3

0.2

0.1

0

1 4 7 10 13 16


108

monetary policy in India is able to influence the food inflation in the short term and the

medium term. This result is contrary to conventional wisdom that monetary policy cannot

influence food inflation. This may be because the CPI-C food index comprises a number

of manufactured items, which might respond to a policy impulse. Figure 14, however,

shows that monetary policy influences the core inflation only in the long run with cycles

of 0 frequency. Figure 15 again shows that the Granger coefficient is not significant

across all frequencies at 5 per cent significance level, except at zero frequency, only in

the long run. This clearly implies that monetary policy is ineffective in the short run and

medium run in India.

VI. DISCUSSION OF RESULTS AND CONCLUSION

Results

(a) The causality tests reveal the presence of the first round effects, namely the

presence of causality from food inflation to headline inflation, which is

expected. They also show that a significant causality from headline inflation

to core inflation exists. Causality from headline inflation to core inflation

implies the presence of the second round effects. Rising food inflation feeds

into the headline inflation, which further feeds into core inflation because of

rising inflation expectations, giving rise to an upward push to the underlying

trend in inflation. As a result, the headline inflation and core inflation diverge.

(b) The volatility results of the inflation measures clearly reveal that none of the

inflation measures are volatile. Accordingly, food inflation in India cannot be

treated as transitory.

(c) Mean error and root mean square error of the inflation expectations for the

given sample clearly reveals that households are overestimating future

inflation as the Reserve Bank of India is failing to anchor inflation

expectations. This is the reason behind the second round effects.

(d) The Granger causality from the call rate to the inflation measures clearly

reveals that policy is able to influence only the food inflation in the short and

medium run. It influences the core inflation and headline inflation only in the

long run.


109

Conclusion

It can, therefore, be concluded that second round effects of food inflation are highly

significant in the case of India. These second round effects occur as inflation

expectations are not anchored. This calls for a renewed and vital role of the Reserve

Bank of India in anchoring inflation expectations of the households through effective

communication and transparency.

In addition, failure of monetary policy in influencing the headline inflation in the

short and medium run, warrants the need to revitalize the transmission mechanism of

monetary policy in India.


110

REFERENCES

Anand, R., D. Ding, and V. Tulin (2014). Food inflation in India: the role for monetary policy. IMF WorkingPapers, WP/14/178. Washington, D.C.: International Monetary Fund.

Benes, J., and others (2016). Quarterly projection model for India: key elements and properties. IMFWorking Paper, WP/117/33. Washington, D.C.: International Monetary Fund.

Bhattacharya, R., and A. Gupta (2015). Food inflation in India: causes and consequences. NationalInstitute of Public Finance and Policy One Pager, No. 14. New Delhi: NIPFP publications.

Blinder, A. (1982). Review of core inflation by Otto Eckstein. Journal of Political Economy, No. 90,pp. 1306-1309.

(1999). Central bank credibility: Why do we care? How do we build it? NBER Working Paper,No. 7161. Cambridge, MA: Nation Bureau of Economic Research.

Bryan, M., and S. Cecchetti (1993). Monitoring core inflation. Working Papers (old series), 9304. FederalReserve Bank of Cleveland.

(1994). Measuring core inflation. In Monetary Policy, G. Mankiw, ed. Cambridge: NBER BookSeries.

Bryan, M., and C. Pike (1991) Median price changes: an alternative approach to measuring currentmonetary inflation. Federal Reserve Bank of Cleveland Economic Commentary Economic

Commentary. Cleveland, Ohio: Federal Reserve Bank of Cleveland.

Cecchetti, S., and R. Moessner (2008). Commodity prices and inflation dynamics. Bank for International

Settlements Quarterly Review, December, pp. 55-66.

Clark, T. (2001). Comparing measures of core inflation. Federal Reserve Bank of Kansas City Economic

Review, vol. 86, No. 2 (second quarter), pp. 5-31.

Clements, K., and H. Izan (1981). A note on estimating divisia index numbers. International Economic

Review, vol. 22, No. 3, pp. 745-747.

(1987). The measurement of inflation: a stochastic approach. Journal of Business and

Economic Statistics, vol. 5, No. 2, pp. 339-350.

Dholakia, R., and V. Kadiyala (2018). Changing dynamics of inflation in India. Economic and Political

Weekly, vol. 53, No. 9.

Diebold, F. (2001). Elements of Forecasting (second edition). Mason, Ohio: South-Western.

Eckstein, O. (1981). Core Inflation. Englewood Cliffs, New Jersey: Prentice-Hall.

Freeman, D. (1998). Do core inflation measures help forecast inflation? Economic Letters, vol. 58,pp. 143-147.

Gordon, R. (1975). Alternative responses of policy to external supply shocks. Brookings Papers on

Economic Activity, vol. 61, No. 1. pp. 183-206.

Goyal, A., and A. Baikar (2015). Psychology, cyclicality or social programs: rural wage inflation dynamicsin India. Economic and Political Weekly, vol. 50, No. 23, pp. 116-125.

Goyal, A., and P. Parab (2019). Inflation convergence and anchoring of inflation expectations. IGIDRWorking Paper, No. WP-2019-023.

Goyal, A., and A. Pujari (2005). Identifying long-run supply curve of India. MPRA Paper 24021.University Library of Munich.


111

Guha, A., and A. Tripathi (2014). Link between food price inflation and rural wage dynamics. Economic

and Political Weekly, vol. 49, Nos. 26-27.

Hatekar, N., and A. Patnaik (2016). CPI to WPI causation: empirical analysis of price indices in India.Economic and Political Weekly, vol. 51, No.1.

Lemmens, A., C. Croux, and M. Dekimpe (2008). Measuring and testing Granger causality over thespectrum: an application to European production expectation surveys. International Journal of

Forecasting, vol. 24, No. 3, pp. 414-431.

Laflèche, T., and J. Armour (2006). Evaluating measures of core inflation. Bank of Canada Review,pp. 19-29.

Misati, R., and O. Munene (2015). Second round effects and pass through of food prices to inflation inKenya. International Journal of Food Agriculture Economics, vol. 3, No. 2, pp. 75-87.

Mohanty, D. (2011). Changing inflation dynamics in India. Speech delivered at the Indian Institute ofTechnology Guwahati, Allahabad, India, 12 August. Available at www.bis.org/review/r110818d.pdf.

Nair, S., and L. Eapen (2012). Food price inflation in India (2008 to 2010). Economic and Political

Weekly, vol. 47, No. 20, pp. 46-54.

Portillo, R., and others (2016). Implications of food subsistence for monetary policy and inflation.International Monetary Fund Working Paper, No. 16/70. International Monetary Fund.Available at www.imf.org/external/pubs/ft/wp/2016/wp1670.pdf.

Raj, J., and S. Misra (2011). Measures of core inflation in India: an empirical evaluation. Reserve Bankof India Occasional Papers, vol. 32, No. 3.

Reserve Bank of India (2014). Report of the Expert Committee to Revise and Strengthen the MonetaryPolicy Framework (Chairman: Dr. Urjit Patel). Available at www.rbi.org.in/Scripts/PublicationReportDetails.aspx?ID=743.

(2019). Monetary Policy Report, April. Available at www.rbi.org.in/scripts/BS_ViewBulletin.aspx?Id=18168\.

Rich, R., and C. Sendel (2005). A review of core inflation and an evaluation of its measures. Federal

Reserve Bank of New York Economic Policy Review Staff Reports, No. 236, pp. 19-38.

Shiratsuka, S. (2006). Core indicators of Japan’s consumer price index. Bank of Japan Review, E-7.

Sprinkel, B. (1975). 1975: a year of recession, recovery and decelerating inflation. Journal of Business,vol. 48, No. 1, pp. 1-4.

Thornton, D. (2007). Measure for measure: headline versus core inflation. Economic Synopses, No. 21.Federal Reserve Bank of St. Louis.

Tobin, J. (1981). The monetarist counter-revolution today — an appraisal. Economic Journal, vol. 91,No. 361, pp. 29-42.

Walsh, J.P. (2010). Food prices and inflation. IMF Manuscript.

Wynne, M.A. (2008). Core inflation: a review of some conceptual issues. Federal Reserve Bank of St.

Louis Review, vol. 90, No. 3, pp. 205-228.

113

TAPPING CAPITAL MARKETS AND INSTITUTIONALINVESTORS FOR INFRASTRUCTURE DEVELOPMENT

Mathieu Verougstraete and Alper Aras*

The present paper is focused on using capital markets in the Asia-Pacificregion to channel more resources for infrastructure development, whilemobilizing assets managed by institutional investors, such as pensionfunds and insurance companies. To this end, the paper is structured asfollows. First, an analysis of the level of capital market development in theregion is conducted, which indicates that markets remain at a nascentstage in many economies. Banks continue to dominate private financing inthe region. Second, a review is carried out on the size of institutionalinvestors from which it is suggested that prudential regulation might needto be adjusted to enable greater infrastructure investment. Third, differentmodalities for investors seeking infrastructure exposure are highlighted andinitiatives launched by different countries to support the development ofinfrastructure-related instruments are presented. Fourth, a review is madeon the actions to support capital market development, which is critical forgreater involvement of institutional investors. Fifth, ways to addressconstraints hindering infrastructure investments are presented. Finally, thepaper concludes with proposals of strategies that are adapted to eachcountry’s circumstances and designed to further tap this source offinancing for infrastructure development.

JEL classification: G23, F21, G15

Keywords: capital markets, infrastructure development, institutional investors, financing,

Asia and the Pacific

* Mathieu Verougstraete (email: [email protected]) and Alper Aras (email: [email protected]) areformer staff members of the Macroeconomic and Financing for Development Division of the Economicand Social Commission for Asia and the Pacific.


114

INTRODUCTION

Countries in Asia and the Pacific need to spend trillions of dollars on infrastructure

development in the coming years (infrastructure is defined here as transport, power,

telecommunications, and water supply and sanitation).

While the banking sector has traditionally played a major financing role, stricter

capital adequacy requirements and maturity mismatches may constrain infrastructure

lending in the future. Capital markets may complement bank financing and provide an

alternative intermediation mechanism between investors and project developers. These

markets could connect investors seeking higher yield investments to infrastructure

projects in emerging countries. Similarly, capital markets can be used to help channel

the abundant savings available within the region as an alternative to them flowing to

more mature economies.1

Against this backdrop, the objective of this paper is to examine how more

resources from institutional investors could be channelled through capital markets for

infrastructure development in the Asia-Pacific region. In doing so, previous research

related to capital market development, such as Genberg (2015), is used as the basis.

While the latter mainly considered how institutional investors contributed to local capital

market development, the focus of this paper is on the role that capital markets and

institutional investors can play in financing infrastructure projects in the region and what

modalities can be used for that purpose.

The paper is structured as follows: the first section contains a review of the state of

financial market development in the region; the second section includes an assessment

of the potential of institutional investors as a source of finance; the third section presents

different investment modalities; the fourth section gives suggested actions to develop

capital market in the region; the fifth section contains highlights of options to address

constraints to infrastructure investment; and the sixth section concludes by identifying

strategies tailored to country situations.

I. ASIA-PACIFIC FINANCIAL MARKETS

Diverse stage

While Asia and the Pacific is home to international financial hubs, such as Hong

Kong, China; and Singapore, the region also has low-income economies where capital

markets remain at an early stage of development. Financial systems in the region differ

in terms of market size as well as from an institutional and regulatory point of view.

1 While the domestic savings rate of the emerging and developing countries in Asia was 42.8 per cent in2015, it was only 18 per cent for emerging and developing countries in Latin America and theCaribbean.

Tapping capital markets and institutional investors for infrastructure development

115

Table 1 provides a snapshot of the region’s financial market development based on

an index conceived by the International Monetary Fund (IMF), which comprises the

following indicators: stock market capitalization to gross domestic product (GDP); stock

market total value traded to GDP; international debt securities of government to GDP;

total debt securities of financial corporation to GDP; and total debt securities of non-

financial corporation to GDP (Svirydzenka, 2016).

The diversity in the region is shown in table 1, which suggests that capital markets

need to be further developed in some countries before they can contribute significantly

to infrastructure development. For instance, Central Asian countries and those in the

Pacific have underdeveloped capital markets. Many of them have neither a bond market

nor a stock exchange.

Table 1. Financial market development index (2016)

Advanced Nascent

>0.676 0.5 to 0.676 0.35 to 0.5 0.046 to 0.124 <0.046

Republic of Korea China Kazakhstan Lao People’s Bangladesh

Australia Malaysia Indonesia Democratic Bhutan

Japan Turkey Iran Republic Armenia

Thailand New Zealand (Islamic Azerbaijan Nepal

Hong Kong, China India Republic of) Uzbekistan Turkmenistan

Singapore Russian Federation Viet Nam Georgia Kyrgyzstan

Philippines Papua New Fiji

Guinea Cambodia

Brunei Darussalam Myanmar

Mongolia Micronesia

Sri Lanka (Federated

Pakistan States of)

Kiribati

Maldives

Marshall Islands

French Polynesia

Solomon Islands

Timor-Leste

Tonga

Vanuatu

Samoa

Source: IMF, Financial Development Index database. Available at https://data.imf.org/?sk=F8032E80-B36C-43B1-

AC26-493C5B1CD33B.

Note: Countries are sorted by financial market development index scores.

------------------------------------------------------------------------<


116

Bank domination

Another key feature of the financial system in the region is the dominant role

played by banks. Loans represent more than 80 per cent of total debt funding for most

Asian economies (see figure 1). This is different from the market in the United States of

America where corporate bonds are a major source of financing.

Nonetheless, the balance between loan and bond has slightly evolved over time. In

some countries, capital market financing has increased, while in others, banks have

consolidated their dominance. For example, corporate bonds in China have increased

more rapidly than bank lending, thereby pushing down the ratio of bank loans as of total

debt funding to 86.6 per cent in 2015 from 91.7 per cent in 2005. The overall size of the

financial sector has also grown exponentially from 140.3 per cent to 243.6 per cent (total

funding as of GDP). In a similar fashion, bank lending in the United States has

somewhat been substituted by corporate bonds with the same ratio declining from

69.7 per cent to 33 per cent over this ten-year period.

Figure 1. Funding structure in selected countries (2015)

Source: Authors’ calculation, based on data from the Asian Development Bank, the International Monetary Fund, the

Bank for International Settlements, and the World Bank.

Notes: Total funding is the sum of bank loans, corporate bond funding and equity funding. Bubble sizes are

proportional to GDP.

Bank loans as of total debt funding (per cent)

0 20 40 60 80 100

450

400

350

300

250

200

150

100

50

0

To

tal fu

nd

ing

as o

f G

DP

(p

er

ce

nt)

Singapore

MalaysiaChina

Japan

Thailand

India

Germany

Indonesia

Philippines

UnitedStates

Australia

Republic ofKorea Canada

France


117

On the other hand, the opposite has occurred in Japan, with the ratio of bank loans

as of total debt increasing from 66 per cent in 2005 to 90.7 per cent in 2015. This may

be explained by the prolonged period of monetary easing and low interest rates in the

country, which has made bank lending cheaper and abundant. For instance, in the

project finance industry, major Japanese banks, such as Mizuho, Mitsubishi UFJ

Financial Group (MUFG) and Sumitomo Mitsui Banking Corporation (SMBC), have been

able to provide competitive pricing for project finance, while keeping those loans on their

balance sheet. This has limited the need for capital market financing for infrastructure

projects. Similarly, the high liquidity of Filipino banks has made it possible to finance

public-private partnerships (PPP) projects domestically.

Bond versus bank loans

The overall bank domination in the region is not an issue per se; however, this may

create limitations for infrastructure project financing, notably with regard to the following:

(a) Maturity: Infrastructure projects require long-term loans to avoid refinancing

risks.2 Banks, however, generally having short-term liabilities (such as

deposits) and holding long-term assets on their balance sheets such as

infrastructure loans, generate maturity mismatches.3 Capital markets can,

instead, mobilize investors seeking a long-term horizon, such as pension

funds, insurance companies and sovereign wealth funds.

(b) Credit limit: Banks typically set single borrower limits to avoid the

concentration of risks on a few counterparts. This limits their capacity to

extend loans to the few large private companies capable of embarking on

infrastructure projects. On the contrary, bonds spread credit risks over a large

pool of investors. In addition, bonds, unlike loans, are tradable, so the credit

risk may be transferred to other parties before maturity.

(c) Pricing: Banking regulations, such as those of Basel III, tend to make loans

relatively more expensive through stricter rules in terms of provisions, capital

adequacy and liquidity ratios.4 These limitations and tighter banking

regulations create opportunities for bonds to complement loans for

infrastructure financing.

2 Banks rarely grant loans for the kind of 20-year spans required for an infrastructure project, as loanstypically reach maturity after 5 to 10 years depending on the market condition. The refinancing risk is thepossibility that the project sponsor will not be able to repay its existing debt by borrowing (or borrowingat less favourable conditions) at the time its loan reaches maturity.

3 Maturity mismatch occurs when a bank funds long-term assets (such as fixed rate mortgages) throughits short-term liabilities (such as deposits).

4 For example, the Tier-1 ratio will increase to 6 per cent in 2015 compared to 4.5 per cent in 2013 (BaselCommittee on Banking Supervision, 2010, annex 4).


118

Nevertheless, bank financing is likely to continue to play a key role, especially in

the initial phase of an infrastructure project during which the risk is typically higher.

Banks are better equipped to manage construction risk and have specialized teams that

closely monitor projects during their early days of implementation. In addition, loans

allow for gradual disbursement of funds in line with the needs of an infrastructure

project, thereby avoiding negative carry forward for the project owner.5 During the

design and construction phase, it is also common for project developers to request

waivers to debt covenants or restructure the debt structure in the light of unexpected

events. While such renegotiation can be done with banks, it is more complex with bond

financing. The latter could require negotiating with a multitude of bondholders. To have

the best of both worlds, the ideal scheme is to finance projects initially through loans

and then refinance them through bonds after the construction phase is over. An example

of such structure is the $2 billion project bond issued by the Indonesian power producer

Paiton Energy, the proceeds of which were used to replace existing project debt and

freed up capital for new projects (Stanton, 2017).

II. INSTITUTIONAL INVESTORS

Mobilizing institutional investors’ resources can be a “game changer” for

infrastructure development. The long-term nature of infrastructure projects matches the

long-term liabilities of institutional investors, such as pension funds, insurance

companies and sovereign wealth funds. Infrastructure assets are, therefore, appealing

to them, as they offer opportunities in terms of return, inflation protection and portfolio

diversification because of their low correlation to other asset classes.

The Organization for Economic Cooperation and Development (OECD) estimates

that institutional investors managed approximately $70 trillion of assets, as of 2013,

which were mainly concentrated in government debt instruments. If only a small fraction

of these resources were to be allocated to infrastructure projects, the impact would be

significant. For instance, a shift of 5 per cent in Asian institutional investors’ allocation in

favour of infrastructure over the next ten years would create an additional annual flow of

approximately $80 billion. This would, however, require enough investable infrastructure

opportunities in the region and a structural change in investors’ behaviours.

5 A negative carry occurs if an investor borrowed from a bank to purchase a bond and its cost ofborrowing is higher than the bond’s yield.


119

Status

Estimates from a study submitted to the World Economic Forum indicate that

approximately 24 per cent of the world’s total asset under management is from the

Asia-Pacific region with the following distribution: insurance (54 per cent); pensions

(25 per cent); and sovereign wealth funds and other funds (21 per cent).6

The size of institutional investors among economies differs widely. Hong Kong,

China and Singapore have the largest asset size given their position as regional

financial centres (more than 50 per cent of their assets are derived from foreign capital

inflows) (figure 2). Meanwhile, the asset size of institutional investors in Indonesia and

the Philippines, for example, is only about 6 per cent and 13 per cent of GDP,

respectively (World Bank, 2014). Obviously, countries with strong local institutional

investors have more potential to tap these investors for infrastructure development.

6 Global Asset Model from World Economic Forum and Oliver Wyman (2014).

Figure 2. Institutional investors structure (GDP per cent – 2014)

Source: World Bank, Global Financial Development database, 2016. Available at www.worldbank.org/en/publication/

gfdr/data/global-financial-development-database; Bank for International Settlements; IMF database and

Securities Industry and Financial Markets Statistics. Available at www.sifma.org/research/statistics.aspx.

United States of America

Germany

Hong Kong, China

Singapore

Japan

Republic of Korea

Australia

China

India

Indonesia

Thailand

Malaysia

PhilippinesEm

erg

ing

co

un

trie

sG

row

ing

gia

nts

Ma

ture

ma

rke

tsH

ub

sA

dvanced

econom

ies

0 100 200 300 400 500 600

Pension fundsInsuranceMutual funds


120

Prudential regulation

Institutional investors are restricted by regulatory limits on the level of risk they

may take in order to protect the savings they manage and ensure their solvency. They

need to consider the security, quality and liquidity of their portfolio and avoid

concentration. For instance, investments with any single counterpart may not exceed

5 per cent of their total assets invested in some countries. Limits can also be based on

the following:

• Asset class characteristic (such as unlisted securities);

• Currency denomination (for example, a certain percentage of assets must be

denominated in the same currency as the liabilities);

• Credit rating (for example, non-investment grade securities are usually

prohibited or limited more strictly) (OECD, 2015b).

In the context of infrastructure, rule-based investment regulations may prescribe

investment in unlisted infrastructure companies (as in Japan and the Republic of Korea),

direct investment in projects (as in Thailand), and infrastructure funds (as in China)

(Inderst, 2016).

As mentioned above, prudential regulation typically excludes non-investment grade

(generally a rating lower than BBB) often found in emerging countries, thereby limiting

significantly their potential to attract foreign institutional investors. In Asia and the

Pacific, approximately 30 countries have been assigned sovereign ratings by major

global rating agencies, such as Moody’s, Standard and Poor’s and Fitch Ratings, and

only 13 of them have been given an “investment grade” (figure 3).7 As global rating

agencies consider the country rating as a cap for any individual company rating, an

infrastructure project cannot be rated higher than the country where the project is being

carried out.8 Unfortunately (and quite logically), countries with lower ratings have the

greatest demand for infrastructure development.

The shift in developed countries from rule-based regulations to principle-based

regulations offers more flexibility. In contrast to strict investment limits, principle-based

requirements tend not to put detailed restrictions on investments. Instead, they impose

broad principles that create disincentives for riskier investments, but do not forbid them

(OECD, 2015b). In addition, domestic investors tend to have more leeway to invest in

7 The Islamic Republic of Iran is the thirtieth country in the region to be given a sovereign rating by Fitch.Its grade is B+, which is non-investment grade.

8 This makes sense as governments significantly affect infrastructure projects through regulation in termsof quality and pricing of outputs and therefore are an important source of risk (Ehlers, Packer andRemolona, 2014).


121

local infrastructure projects. This can be the case if they opt to follow credit ratings

provided by local agencies, which have a different approach regarding the country risk.

In addition, as their liabilities are in local currency, investing in domestic-currency-

denominated assets provide them with a natural hedge. Yet, this does not mean that

domestic investors should not invest abroad in order to benefit from international

diversification and limit their exposure to the local economy.

Source: Authors, based on the Tradingeconomics database. Available at https://tradingeconomics.com/country-

list/rating.

Note: The designation employed and the presentation of material on this map do not imply the expression of any

opinion whatsoever on the part of the Secretariat of the United Nations concerning the legal status of any

country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries.

III. INVESTMENT MODALITIES

Investors have four options to channel funds to infrastructure development through

capital markets. One option is to invest in infrastructure companies as a proxy to

infrastructure projects. The other three are to finance infrastructure projects directly, go

through listed infrastructure funds or purchase municipal bonds that have a large

infrastructure component. Institutional investors can also finance projects directly

through unlisted instruments, such as private equity funds, but this is outside the scope

of the present paper. The different modalities are illustrated in figure 4.

Figure 3. Sovereign rating of selected countries (as of 2019)


122

Infrastructure companies

Infrastructure companies can raise equity and debt on capital markets to finance

their activities.

Equity

By issuing equity on capital markets, infrastructure companies mobilize financial

resources, which may be used to participate in infrastructure projects. This is only

possible if these companies have access to a developed stock market. Thirty-five

economies in the Asia-Pacific region have a stock exchange, though the level of

development among them varies.9 In addition, with the exception of a few countries,

market capitalization is relatively limited in the region (see figure 5) and the liquidity in

some Asian equity markets tends to be low (table 2), which reduces their attractiveness

for investors seeking the possibility of rapid exits at a stable price.

In countries that have developed stock markets, infrastructure companies have

typically been large issuers. For instance, it is estimated that listed infrastructure and

utility companies represent 5 to 6 per cent of the equity market universe globally

(Inderst, 2016). In the Asia-Pacific region, 30 of the largest publicly listed infrastructure

companies – the companies that constitute the Standard and Poor’s Asia Infrastructure

Index – have a total market capitalization of approximately $260 billion. Table 3 shows

the geographical distribution of the index. The companies are mainly in the utilities

sector (39.5 per cent) followed by the industrials and energy sectors with 38.7 per cent

and 21.8 per cent, respectively.

9 As of March 2016, the economies in Asia and the Pacific that do not have a stock exchange areAfghanistan; Brunei Darussalam; Democratic People’s Republic of Korea; French Polynesia; Kiribati;Macao, China; Marshall Islands; Micronesia (Federal States of), New Caledonia; Palau; Samoa; SolomonIslands; Tajikistan; Timor-Leste; Tonga; Turkmenistan; Tuvalu and Vanuatu.

Figure 4. Type of capital market investments in infrastructure

Infrastructurecompanies

Stockmarket

Corporatebond

Infrastructureprojects

SPVlisting

Projectbond

Infrastructurefunds

Municipalbonds


123

Figure 5. Stock market capitalization

of listed companies to GDP

(per cent – 2017)


Malaysia

Japan

Thailand

Australia

Republic of Korea

India

Philippines

Euro area

Nepal

China

New Zealand

Indonesia

Russian Federation

Viet Nam

Kazakhstan

Iran (Islamic Republic of)

Turkey

Sri Lanka

Bangladesh

0 50 100 150 200

Source: https://data.worldbank.org/indicator/ and

CEIC. Available at www.ceicdata.com/en.

Table 2. Stock market turnover ratio

(2018)

CountriesStock market

turnover ratioa

Kazakhstan 3.4

Sri Lanka 7.0

Philippines 11.3

New Zealand 14.1

Iran (Islamic Republic of) 18.3

Indonesia 21.5

Russian Federation 25.5

Malaysia 34.0

Viet Nam 39.8

India 58.1

Thailand 77.2

Republic of Korea 112.3

Australia 61.3

Japan 119.0

China 206.7

Turkey 247.8

World 104.7

Note: a Total value of shares traded during the period

divided by the average market capitalization for

the period.

Table 3. Standard and Poor’s Asia Infrastructure Index geographical distribution

Economy Number of constituentsTotal market capitalization

(billions of US dollars)

Japan 8 48

China 6 28

Hong Kong, China 4 92

Malaysia 5 37

Singapore 3 8.2

Thailand 1 15.6

Republic of Korea 1 27.6

Philippines 1 3.3

Indonesia 1 2.3

Total 30 262


124

Corporate bond

Regarding corporate bonds, companies from emerging markets represent

approximately 20 per cent of the total global outstanding amount as of 2018. This is

significantly higher than the pre-2008 Global Financial Crisis level. Growth has been

particularly strong in China. The number of issuers in emerging markets increased from

347 issuers in 2007 to 1,917 in 2016 at its peak (more than a 5.5-fold increase) (OECD,

2019). Despite this, only a limited number of the economies in the Asia-Pacific region

have a large local currency corporate bond market. Among them are China, Malaysia,

the Republic of Korea, Malaysia, Singapore and China (figure 6). Significant progress

related to corporate bonds has been achieved in other markets, such as in the

Philippines and Thailand.

Figure 6. Size of local currency corporate bond market in selected economies

from 2000 to 2018 (per cent of GDP)

Source: AsianBondsOnline. Available at https://asianbondsonline.adb.org.

As indicated above, in countries that have a developed corporate bond market,

infrastructure-related companies have been key issuers (figure 6). For example, in

China, infrastructure-related entities, such as State-owned enterprises, are among the

largest corporate bond issuers. Similarly, the corporate bond landscape in Indonesia is

dominated by mining and utilities firms, which issued more than 50 per cent of all bonds

during the period 2009–2013 (Levinger and Li, 2014).

80

70

60

50

40

30

20

10

0

29 28

3

14

73

49

8

32

21

2

China Hong Kong, China

Indonesia Japan Republic ofKorea

Malaysia Philippines Singapore Thailand Viet Nam

Dec-00 Dec-08 Dec-18


125

Infrastructure project

Investors can also invest directly in infrastructure projects by acquiring equity in the

special purpose vehicle (SPV) created for these projects or through project bonds.

Special purpose vehicle listing

Project sponsors wishing to realize an infrastructure project often establish

a dedicated project company known as a “special purpose vehicle” – or SPV – to attain

financing and implement project activities. This legally isolates the parent organization

from direct exposure to the financial risks associated with a project. If SPV is listed on

the stock exchange, investors can invest directly in the project. To facilitate SPV listings,

the Philippine Stock Exchange changed its listing rules in 2016. Under the revised rules,

a company without the required three-year track record may still apply for listing on the

stock exchange if they comply “with the rest of the general listing requirements set forth

in the Philippine Stock Exchange Main Board.” The project needs, however, to have

completed the construction phase (Dela Paz, 2016). The same types of criteria apply on

the Thai stock exchange (Stock Exchange of Thailand, n.d., b).

Infrastructure companies may also create “yieldcos” for projects producing

predictable cash flows, for instance through long-term contracts, such as those in the

energy sector. A yieldco is a company formed to own operating assets. These assets are

placed in a new subsidiary to separate them from other more volatile activities of the

parent company, such as project development, and research and development. Part of

the subsidiary shares are then listed on a stock exchange through an initial public

offering. This type of structure has yet to take off in Asia (Chua, 2015). It is, however,

well developed in North America, although the collapse of one such company,

SunEdison, in 2016, has raised questions on the viability of the model.

Project bond

Project bonds are debt instruments used for financing stand-alone infrastructure

projects, for which SPV is formed. SPV issues a project bond, the creditworthiness of

which depends on the cash flow of the underlying infrastructure project. This is quite

different from corporate bonds, which rely on the balance sheet of the issuing entity

(OECD, 2015a).

Globally, project bonds accounted for about 10 per cent of global project debt from

1994 to 2012 and are more commonly issued in North America. Project bond financing

declined during the 2008 Global Financial Crisis, but markets have rebounded since

then, although the overall volumes have remained small ($36 billion in 2013, which was

less than 0.1 per cent of global GDP). In Asia and the Pacific, the volume of project

bonds has ranged between $1 billion and $3 billion in recent years (Inderst, 2016).

Maturities also tend to be shorter in Asia and the Pacific than in other markets. While in


126

advanced economies the average maturity of issued infrastructure-related bonds is

approximately 15 years, in emerging Asian economies it is only about eight years

(figure 7).

Some countries have, nevertheless, managed to use project bonds quite

extensively. Malaysia, for example, has been successful in financing its infrastructure

development through the issuance of sukuk (Islamic bonds structured to generate

returns for investors without contravening Islamic law). The largest national highway

concessionaire, PLUS Expressways Berhad, issued sukuk worth several billion dollars

in 2012, notably for acquiring the rights for five toll concessions (Raghu and Kaiser,

2012). Project bonds can also be used to refinance infrastructure projects. They have,

for instance, been used to refinance the Mersin International Port project in Turkey, for

which a seven-year bond was issued for $450 million in 2013.10

10 See www.ebrd.com/work-with-us/projects/psd/mersin-international-port-bond.html.

Figure 7. Average maturities of infrastructure-related debt securities bonds

Infrastructure fund

Infrastructure funds are another intermediary mechanism between investors and

infrastructure projects. They serve as a vehicle to pool resources, skills and experiences

from different investors while achieving economies of scale. Specialized skills are

required for structuring and assessing infrastructure investments. It may not be efficient

for every investor to develop such expertise internally. In 2015, seven Asian-focused

infrastructure funds reached financial close, securing a combined $5.3 billion (nearly

double the capital raised in 2014) (Preqin, 2016).

Australia


Japan

Other advanced economies

Latin America

China

Emerging Asian markets

Other emerging markets

0 2 4 6 8 10 12 14 16 18 20


127

Although a large chunk of infrastructure funds is private equity, listed instruments

have also been used. For example, the listed Macquarie Korea Infrastructure Fund, set

up in 2002, has contributed to one port and eleven road projects through equity,

subordinated debt and senior debts.11 Listed infrastructure funds have also been active

in Australia and Singapore for several years.

In Thailand, infrastructure funds were established to raise capital from individual

and institutional investors. The largest one to date is the BTS Rail Mass Transit Growth

Infrastructure Fund, which raised through an initial public offering approximately

$2 billion in April 2013. Proceeds from the offering were used to buy the rights to the

future net farebox revenues (= farebox revenues – operating costs and capital

expenditure) of the Bangkok mass rapid transport system, the Bangkok skytrain, for the

remaining concession years, until 2039) (InfraPPP, 2013). This type of structure also

allows State-owned enterprises to recycle their operating assets in order to generate

cash flow for new projects.

India has also been active on this front with infrastructure debt funds launched in

2013 as an intermediary vehicle capable of refinancing PPP project loans after they are

operational through the issuance of bonds. Following three-years of operation, the level

of refinancing has been limited, but it is expected to increase in the coming years

(Rebello, 2016). Similarly, infrastructure investment trusts were established to refinance

PPP project equity investment, which can be an interesting concept for other countries

to consider (see box 1).

At the regional level, the ASEAN Infrastructure Fund was launched in 2012 to deal

with the region’s infrastructure needs. While the fund is initially providing loans from its

own resources, it is expected to issue debt to increase the resources available for

infrastructure financing. These debts will be able to be purchased by investors seeking

exposure to infrastructure projects.

Municipal bonds

With the expansion of urbanization, municipalities are under strong pressure to

deliver infrastructure services, such as public transport systems. To finance such

development, local governments can issue bonds. For example, municipal bonds are

particularly popular in the United States where tax exemption has made them attractive

to investors. In Asia and the Pacific, this type of instrument is flourishing in some

countries, such as China. In 2016, for example, local Chinese governments were

scheduled to issue about 6.2 trillion Chinese yuan (¥) ($860 billion) of securities in 2016,

compared with ¥3.8 trillion in 2015 (Bloomberg News, 2016).

11 See www.macquarie.com/mgl/mkif/en/about-mkif.


128

Municipal bonds usually attract funding at a low cost given the implicit guarantee

they enjoy from the central government (although assessing their credit worthiness is

difficult). They are also generally subject to a less stringent level of oversight than the

corporate bond market. The corollary risk is that municipalities might pile up debt,

thereby creating fiscal risks in the long run.

Box 1. Infrastructure investment trust in India

The Securities and Exchange Board of India issued infrastructure investment

trust (InvIT) regulations in 2014.

For sponsors, InvIT is a way to unlock tied up capital in infrastructure projects

by transferring operating and revenue-generating infrastructure assets to a trust.

They have to keep a minimum percentage and the capital raised has to be used for

repaying at least 50 per cent of the debt. For institutional investors, InvIT creates

investment opportunities in infrastructure projects.

While to date, the success of InvIT has been limited, several Indian companies

have initiated an approval process for this type of instrument, such as IRB Infrastructure

Developers Ltd., IL and FS Transportation Networks Ltd., Sterlite Power Transmission

Ltd., Reliance Infrastructure Ltd. and MEP Infrastructure Developers Ltd.

Source: www.ey.com/in/en/industries/real-estate/ey-real-estate-and-infrastructure-investment-trust.

Note: SPV, special purpose vehicle.

Not morethan 3

InvIT

Listing is

mandatory

Management fee

Investmentmanager

> 50% > 50%

SPV – 1 SPV – 2 SPV – 3

Asset Asset AssetProject manager

for each infra asset

SponsorInstitutionalinvestors


129

IV. CAPITAL MARKET DEVELOPMENT

While the previous section provides a comprehensive overview of the different

investment modalities to finance infrastructure through capital markets, this section is

focused on actions to support capital market development in the Asia-Pacific region,

which is a precondition for greater involvement of institutional investors.

Building domestic bond markets

Few countries in the region have a developed corporate bond market, as illustrated

in figure 8. Liquidity and maturity are also restraining the possibility of using bonds for

long-term infrastructure projects (maturities of corporate bonds issued in Viet Nam are,

for instance, relatively short). Accordingly, local capital markets need to be deeper for

them to play a greater role in infrastructure financing.

Figure 8. Size of government and local corporate bonds in selected countries

(as of 2018)

Source: AsianBondsOnline. Available at https://asianbondsonline.adb.org/.

250

200

150

100

50

0China Hong Kong,

ChinaIndonesia Japan Republic of

KoreaMalaysia Philippines Singapore Thailand Viet Nam

Government (% GDP) Corporate (% GDP)

An incremental process

To develop local capital markets, countries need to follow an incremental process,

such as the one described in box 2. This process has already been implemented in

countries that had established a government bond market prior to setting up a corporate

bond market, such as Indonesia, the Philippines and Viet Nam. Similarly, project bonds

only emerge when there is a developed corporate bond market. This incremental

process means that each country should follow a strategy based on its current market

development stage.


130

Box 2. Sequencing approach to financial market development

Typically, the money market, short-term debt securities usually issued by

governments and financial institutions, precedes the other segments because of its

central role in price discovery and interest rate setting.a Money markets are the

medium through which central banks intervene and financial institutions manage

their liquidity by lending and borrowing to and from each other. The foreign exchange

market shares a lot of similarities with the money market except that in the former

each transaction involves the exchange of local and foreign currency. The different

market segments are, however, interrelated. For instance, a liquid money market

relies on adequate depth in government bonds, as bonds are typically used as

collateral in interbank lending (for repurchasing agreements). A well-developed

government bond market also works as a catalyst for establishing appropriate bond

market infrastructure (with expected positive spillovers for other fixed income markets)

and the government bond yield curve serves as a price reference for corporate

bonds. In addition, the development of derivative markets requires well-developed

bond and equity markets, as they constitute the underlying assets of derivative

instruments.

The hierarchical order of financial markets

Source: Cem Karacadag, V. Sundararajan and Jennifer Elliott, “Managing risk in financial market development:

the role of sequencing”, IMF Working Paper, No. 03/116 (Washington, D.C., IMF, 2003).

a The money market plays central role in price discovery because long-term nominal interest rates

should be an average of current and expected nominal short-term rates.

Money market

Treasury bills market & foreign exchange market

Government bond market

Corporate bond &equity markets

Derivatives


131

Bond market determinants

Researchers have tried to determine the key factors supporting bond market

development. Studies have shown that high inflation volatility can be a constraint to

such development, as it creates uncertainty regarding real returns for investors (Burger,

Warnock and Warnock, 2015). In this respect, an increase in the issuance of

inflation-indexed bonds could signal government commitment to inflation control. For

example, the Reserve Bank of India allowed inflation indexed bonds in 2013 and 2014.

This financial product provides hedging opportunities for investors (Shenoy, 2013).

The importance of credit right protection has also been stressed in different studies

(Burger, Warnock and Warnock, 2015). Debtholders need to be confident that

governments will adhere to the rule of law and contracts will be enforced. In Asia and

the Pacific, trust seems to be lacking in several countries. This was indicated by the

2016 rule of law index in which four Asia-Pacific countries ranked in the last ten out of

the 113 countries surveyed (World Justice Project, 2016).12 In particular, treatment of

bankruptcy is important for investors and need to be predictable.

To successfully develop bond markets, the low liquidity level, a persistent issue in

many markets throughout the region, must be addressed. It is also worth noting that

a government bond market does not automatically result in the development of

a corporate bond market. For instance, one factor attributed to an underdeveloped

corporate bond market is the higher cost of issuing corporate bonds because of the

greater volume of documentation required in comparison to bank lending. Regulators

should investigate ways to lessen transaction costs without compromising the needs of

investors for transparency and security.

International support

To further develop bond markets in the Asia-Pacific region, countries need to

exploit the opportunity to work with multilateral development banks, which can issue

bonds in local markets for this purpose. For instance, the Asian Development Bank

(ADB) was the first foreign issuer in the domestic capital markets of China (co-issuer

with the International Finance Cooperation), India, Malaysia, the Philippines, the

Republic of Korea and Thailand. These issuances serve as a benchmark for lower-rated

issuers, while also attracting investors unfamiliar with a specific currency. To be

successful, such issuances should contribute towards increasing market liquidity, lead to

longer tenors and result in subsequent issuers. Once markets are developed,

infrastructure companies and projects may more easily tap long-term financing and

access a wider pool of financiers.

12 They are Afghanistan (111); Bangladesh (103); Cambodia (112); and Pakistan (106).


132

Facilitating foreign investment

Countries must also address such issues as capital controls and the lack of foreign

exchange hedging instruments in order to attract foreign investors into infrastructure

investments.

Capital controls

Progressive capital account liberalization has eased market access to foreign

investors, although there are still limits on non-residents holding and trading domestic

securities in several countries. For example, India has put restrictions on foreign

investment in rupee-denominated bonds (Patnaik and others, 2013). In the same vein,

Thailand only grants approvals to foreign entities to issue baht bonds on the condition

that they keep the proceeds in baht and use them in the country (Thaichareon and

Sriring, 2016). Most of the countries in the region have in place foreign exchange

restrictions to mitigate vulnerabilities stemming from capital outflows. These restrictions

limit investments by non-resident institutional investors, which adversely affects market

development in these countries. A balance needs to be struck between the negative and

positive effects of capital control policies.

Hedging instruments

To enable larger international allocations from institutional investors, hedging

instruments, such as interest and currency swaps, are needed. Derivative markets are

relatively underdeveloped in Asia and the Pacific, as compared to other regions. The

derivative market value represents 15 per cent of the underlying market in the region, as

compared with 35 per cent in the United States of America and 50 per cent in Europe

(as of 2012) (Deloitte, 2015).

Initiatives have been launched to overcome this issue. For example, the Reserve

Bank of India is working with the Securities and Exchange Board of India to allow non-

resident institutional investors to hedge currency risk with exchange-traded currency

futures.13 At the international level, the Currency Exchange Fund was created to provide

hedging against currency and interest rate mismatches in frontier and less liquid

emerging markets. Its services cover approximately 70 currencies, including 17

currencies of Asian countries.14 The price of these hedging instruments, however, are

often prohibitive, especially for illiquid and underdeveloped markets. Given the

importance of hedging instruments, efforts should be pursued to further develop these

instruments in the region.

13 Currency futures specifies the price at which a currency can be bought or sold at a future date.14 www.tcxfund.com/.


133

Promoting financial integration

For small-scale economies, the viability of a domestic liquid capital market that

provides a large amount of resources appears to be uncertain. In such circumstances,

countries may need to leverage offshore markets although this creates currency risks.

In the view of the amount needed for infrastructure projects and the desirability

of long maturities, tapping the United States and Eurobond offshore markets may

offer alternative sources for infrastructure investments. For example, during the

period 2009–2013, 551 infrastructure bond deals were signed with a total value of

$167.5 billion in emerging Asian countries (Ehlers, Packer and Remolona, 2014);

$2.3 billion of that total value was issued in United States market and $200 million was

issued in the Eurobond market in Asia (Ehlers, Packer and Remolona, 2014). Offshore

markets open to infrastructure companies provide a greater pool of savings to tap, but

these companies are tasked with managing the currency mismatch resulting from

issuing securities in foreign currency.15

By strengthening ties between the region’s financial markets, countries can

also diversify their sources of financing and attract foreign investment. This requires

reducing cross-border transaction costs, among other things. For example, the cost of

cross-border transactions in the ASEAN+3 region16 is three times higher than those in

the United States and the European Union (ADB, n.d.).

To facilitate cross-border investments, countries need to harmonize regulations,

corporate governance and financial products with the objective of achieving mutual

recognition of trading transactions. For instance, different standards and requirements

may prevent investors to credibly assess investment opportunities across multiple

countries. Harmonizing these standards and requirements with the international

standards and requirements go a long way in addressing this issue. Although impetus

has grown in this regard in Asia and the Pacific since 2013 when Japan and China

started working on International Financial Reporting Standards, significant discrepancies

across the region persist.

Key market infrastructure for securities, including payment systems, cross-border

clearing and settlement systems, central securities depositories and custodians are also

needed in order to strengthen financial integration. For example, most of the local

central securities do not have links with international central securities, with the

exception of a few countries, such as Malaysia and Singapore.

15 Currency mismatch means having assets that are denominated in a different currency than liabilities, sothat a change in the exchange rate between those currencies can have a large positive or negativeeffect on the balance sheet.

16 This region comprises the ASEAN member states plus China, Japan and the Republic of Korea.


134

Against this backdrop, it is important to further support regional initiatives that

promote financial integration, such as the ASEAN+3 Bond Market Forum and the

ASEAN Trading Link, which was launched in 2012. These initiatives should facilitate the

mobilization of financing beyond domestic resources for infrastructure projects. For

regional initiatives to be successfully carried out, it is also essential to further educate

investors in order to make them comfortable in investing in financial instruments abroad.

while making regional market development more demand-driven.

Supporting domestic investors

There is a high correlation between the size of the institutional investor base and

the size of capital markets (figure 9). This confirms the importance of developing

a critical mass of long-term institutional investors to support the deepening of the

financial markets, as these investors play a catalytic role in capital market development.

In addition, local institutional investors have liabilities in local currency and accordingly

are willing to invest in local currency.

Figure 9. Asian institutional investor base and capital market development

(per cent of GDP)

Source: Kang, Jeasakul and Lim (2015).

Note: AUS, Australia; CHN, China; IDN, Indonesia; IND, India; JPN, Japan; MYS, Malaysia; NZL, New Zealand;

PHL, Philippines; SGP, Singapore; THA, Thailand; KOR, Republic of Korea.

350

300

250

200

150

100

50

0

Siz

e o

f ca

pita

l m

ark

ets

Size of institutional base

0 50 100 150 200

JPN

MYS

SGP

KOR AUSTHA

PHL

IND

CHN

NZLIDN


135

Unfortunately, the size of domestic institutional investors is relatively limited in the

region despite the existence of sizeable social security and public pension schemes in

some countries.17 OECD estimates that the largest Asian funded pensions systems are

well below the OECD average of 84 per cent of GDP, with developing Asia at less than

5 per cent (OECD, 2014b). Additional efforts should, therefore, be made to support the

emergence of a larger base of domestic investors. This can be done, for instance, by

encouraging funded pension schemes.

V. CONSTRAINTS TO INFRASTRUCTURE INVESTMENTS

The focus of this section is on various constraints that impede investment in

infrastructure development by institutional investors. Options to address them are given

in the section.

Enhancing risk profile

Achieving the necessary rating to make infrastructure project bonds attractive to

institutional investors requires reducing the risk of the debt component of an

infrastructure project. This can be done through various mechanisms, such as

increasing the equity share in a project, introducing subordinated debts and providing

guarantees. For instance, by providing a corporate or rolling guarantee, the sponsors,

the parent companies, can enhance the credit rating of a project bond. Similarly, an

external guarantee can be used for the same purpose. For example, in Malaysia the

West Coast road project was granted an AAA rating because it was guaranteed by

a solid bank.

Providing a credit guarantee was the business of “monoline” insurance companies

before the 2008 Global Financial Crisis. This market has yet to recover. Consequently,

alternatives need to be found. Multilateral institutions have tried to fill the gap. For

instance, the Credit Guarantee and Investment Facility was established in 2010 to

improve the risk profile of local currency bond issuance in the ASEAN+3 region.18 In

addition to providing a credit guarantee, in 2016, a new instrument was added to its

portfolio to offer a construction period guarantee, thereby significantly improving the risk

profile of greenfield projects. Similarly, ADB and the India Infrastructure Finance

17 The Pension Investment Fund (about $1.2 trillion) of the Government of Japan, the Republic of KoreaNational Pension Service ($400 billion), the National Social Security Fund of Singapore ($200 billion),Central Provident Fund of Singapore ($190 billion), the Employees Provident Fund of Malaysia ($180billion), and the Employee Provident Fund of India ($116 billion) (OECD, 2014a).

18 See www.cgif-abmi.org/about/overview.


136

Company Ltd. set up in 2012 a project bond guarantee facility to attract more

institutional investors; to date, it has enjoyed only limited success.19

Subordinated debts have also been used to improve the rating of senior tranches;

an example of this is the Europe 2020 Project Bond Initiative implemented by the

European Investment Bank (see box 3). A similar mechanism exists with commercial

banks, namely the Pan European Bank to Bond Loan Equitisation (PEBBLE) developed

by ING Bank, and Allen and Overy in 2012.20

19 Two projects have benefited from this guarantee facility for respectively $68 million (2015) and$19.6 million (2016) (Lambert, 2016).

20 This mechanism was first used in the N33 widening road project in the Netherlands, with a capital valueof 120 million euros (€) ($130 million) and a 20-year concession period (Allen and Overy, 2012).

Box 3. Europe 2020 Project Bond Initiative

Note: EIB, European Investment Bank.

The pilot phase of the Europe 2020 Project Bond Initiative was launched in

2012 and implemented by the European Investment Bank. The objective of the

project is to provide an additional source of financing for transport, energy and

information technology infrastructure projects through debt capital markets.

By enhancing the credit quality of project bonds issued, the initiative is aimed at

attracting institutional investors. The Project Bond Credit Enhancement works either

as a funded subordinated debt or guarantee; its principles are described in the

figure above.

Project

bond

(target

rating

min A–)

EIBsub-debt

Equity

Project bondinvestor

Project

bond

(targetrating min

A–)

EIBunfundedsub-debt

Equity

Special

purpose

vehicle

(SPV)

project

costs

Project bondinvestor

EuropeanInvestmentBank

EuropeanCommission

EuropeanCommission

EuropeanInvestmentBank


137

These kinds of credit enhancement mechanisms are critical to support

infrastructure financing through capital markets. Lessons learned from international

experiences should help governments in designing appropriate mechanisms for their

respective countries

Addressing capacity constraints

Identifying and assessing infrastructure investment opportunities is complex. It

requires skills and local expertise that institutional investors may lack. To address this

issue, investors can use an external fund manager or partner with experienced

international investors. For instance, development finance institutions can play an

important role in this area and reduce transactions costs for institutional investors. For

decades, multilateral development banks have operated syndicated-loan programmes

that allow financiers, such as international banks, to participate in multilateral

development bank loans while benefiting from the banks’ preferred creditor advantage.

The International Finance Corporation, more recently, has created the Managed

Co-Lending Portfolio Programme to serve as a syndication platform that creates

diversified portfolios of emerging market private sector loans (instead of participating in

individual deals), allowing investors to gain exposure in these markets by co-lending

alongside the International Finance Cooperation on commercial terms. As of 2018, the

Managed Co-Lending Portfolio Programme raised $7 billion from eight global investors.

Creating investment opportunities

The lack of infrastructure investment opportunities can be an obstacle for

channelling institutional investors to this asset class. The securitization of infrastructure

project finance loans offers a way to create more opportunities (see box 4).

As of 31 July 2015, seven transactions have been supported with a total Project

Bond Credit Enhancement of 612 million euros, which enabled the issuance of more

than 3.7 billion euros worth of bonds. Based on this track record, an independent

evaluation published in March 2016 concluded that the Project Bond Credit

Enhancement solution should continue to be deployed in the future, because it has

demonstrated the ability to provide long-term competitive solutions to finance crucial

infrastructure projects.

Source: European Commission, “The ad-hoc audit report of the pilot phase of the Europe 2020 Project Bond

Initiative. Executive summary”, Commission Staff Working Document (2016). Available at http://

ec.europa.eu/dgs/economy_finance/evaluation/pdf/eval_pbi_pilot_phase_executive_en.pdf.

Box 3. (continued)


138

Box 4. How does securitization work?

Securitization is the process in which certain types of assets are pooled so that

they can be repackaged into interest-bearing securities. The interest and principal

payments from the assets are passed on to the purchasers of the securities. Basically,

the process consists of two steps (see the chart below). In step one, a company

with loans or other income-producing assets — the originator — identifies the assets

it wants to remove from its balance sheet, such as a portfolio of loans, and pools

them into what is called the reference portfolio. It then sells this asset pool to an

issuer, such as a special purpose vehicle, which is an entity set up, usually by

a financial institution, specifically to purchase the assets. In the second step, the

issuer finances the acquisition of the pooled assets by issuing tradable, interest-

bearing securities that are sold to capital market investors. The investors receive

fixed or floating rate payments from a trustee account funded by the cash flows

generated by the reference portfolio. In most cases, the originator services the

loans in the portfolio, collects payments from the original borrowers, and passes

them on – minus a servicing fee – directly to the special purpose vehicle or the

trustee.

Source: Andreas Jobst, “What is securitization?”, Finance & Development (September 2008), pp. 48-49.

Available at www.imf.org/external/pubs/ft/fandd/2008/09/pdf/basics.pdf; and Jennifer Romero-Torres,

“Enabling monetization of infrastructure assets”, presentation prepared for the UNESCAP PPP

Workshop, Kuala Lumpur, Malaysia, November 2015. Available at www.unescap.org/sites/default/files/

3a%20-%20ADB%20-%20Enabling%20Monetization%20of%20Infra%20Assets.pdf.

Transfer of assets(e.g. loans) from

the originator to issuingvehicle

Asset originator(such as bank)

Assets onbalance sheet

(e.g. portfolio ofinfrastructure

loans)

Issuing vehicle(such as SPV)

Capital marketinvestors

1 2

Originatorretains no legal

interests inassets

Debt structureinto various

tranches ratedby ratingagencies

Asset-backedsecurities traded

on capitalmarkets

Special purposevehicle (SPV) issues

debt securities(asset-backed) to

investors


139

For banks, securitization allows them to move long-term assets off their balance

sheets and relieve pressure resulting from tighter capital requirement regulations.

For example, banks may be presented with the opportunity to sell their infrastructure

loans when projects are in their operational phase and risk is much reduced, thereby

creating relative safe long-term products sought by institutional investors. To develop

a securitization market, there must be a guarantee that lenders keep some “skin in the

game” to avoid the issues faced by the “subprime” market, which triggered the 2008

Global Financial Crisis.

Examples of this kind of structure are already found in Asia and the Pacific. For

instance, the Japanese bank SMBC issued in 2016 its first project finance loan

securitization note to be sold to institutional investors (the loan portfolio was related to

large-scale solar power plants) (Nikkei Asian Review, 2016). Similarly, in Australia and

China, banks are issuing green bonds based on their green loan portfolio. Approximately

50 per cent of the labelled green bond market is issued by development and commercial

banks (Climate Bonds Initiative, 2016). For securitization to function properly, banks

need to have an incentive to sell these loans either for reasons linked to capital

adequacy ratio or because they are reaching their single borrower limits. Otherwise,

they might be reluctant to cede performing assets.

Tailoring credit rating assessment

Infrastructure assets tend to show a robust risk profile. A study conducted by

Standard and Poor Global Ratings confirmed that infrastructure credits have lower

default rates and ratings than companies active in other sectors (Standard and Poor

Global, 2018).

Credit agencies may need to develop methodologies that take into account the

specificities of infrastructure projects, such as their lower default and good recovery

rates (Moody’s Investors Service, 2014). In India, rating agencies have recently

launched a specific credit rating for infrastructure assets. By introducing credit rating

systems that reflect the unique nature of the infrastructure sector, countries may open

up more long-term funding.

Standardizing contractual provisions

Infrastructure projects are inherently complex. This makes it time consuming and

costly for investors to assess risks and allocate more resources to this sector.

Further standardization of contractual arrangements could go a long way in

facilitating risk assessment, while also simplifying project development by public

authorities. The World Bank has produced Guidance on PPP Contractual Provisions to

this effect. Scaling up private infrastructure investment requires more standardized

investment opportunities.


140

Reviewing tax policy

Tax treatment can promote or deter the use of capital markets for infrastructure

financing and may favour loans over bonds. For example, a stamp duty might create

a distortionary effect. When enforced, this tax is placed on the transfer of securities. In

particular, a stamp duty significantly hinders the development of securitization, as the

transfer of receivables from the originator to the special purpose vehicle is subject to

such payment, which can make the structure commercially unviable. To address this

issue, stamp duty exemptions have been granted in Thailand if the special purpose

vehicle arranges to transfer the infrastructure asset back to its originator or to any other

public sector (Stock Exchange of Thailand, n.d., a). Similarly, a transfer tax, lease and

mortgage register fees have been reduced to the minimum in order to make

infrastructure funds viable.

Governments can also attract investors by granting favorable tax treatments to

infrastructure-linked investment. For example, to steer investment towards infrastructure

development, the Securities and Exchange Commission of Thailand adopted rules on

infrastructure finance in February 2012. Under these rules, investors are exempt from

personal income tax on dividends for 10 years. Similarly, Malaysia and Singapore do not

impose a withholding tax on interest earned from local bonds by foreign investors

(Sahay and others, 2015). Meanwhile, the currently level of development of municipal

bonds in the United States can be attributed to tax exemption. These examples show

how tax policies can affect capital market development. The tax policies, however, need

to be balanced against the foregone revenue they create.

VI. POLICY RECOMMENDATIONS ADAPTED

TO THE LOCAL CONTEXT

With rapidly growing assets under management, institutional investors in Asia and

the Pacific can potentially play a greater role in infrastructure financing provided that

governments develop viable pipelines of infrastructure projects. The extent of this role

depends on country circumstances. Some countries have a well-developed institutional

investor base and functioning capital markets, while others are at a more preliminary

stage. High-risk country ratings also prevent deeper involvement of institutional

investors in some markets. While there is no “one size fits all” strategy for the region, it

is possible to recommend different strategies for different groups of countries. This is

shown in table 4, although the segregation among the different groups is more blurry in

reality.


141

Table 4. Strategies for mobilizing capital market for infrastructure development

Phase III Phase II Phase I

Rating High-rating Medium-rating Highly speculative

(investment grade) (just above or below grade or no rating

investment grade)

Stock market Developed and liquid Emerging No/limited market

Bond market Developed and liquid Relatively developed No/limited government

government and government and bond market

corporate bond emerging local

markets currency corporate

bond markets

Project bond Emerging Infancy N/A

Possible strategies Consider securitization Strengthen capital Strengthen the

to increase the size market development government bond

of infrastructure assets. in particular corporate market (as a price

Examine the bond market reference) and

possibility to develop (notably by improving investment

infrastructure credit information environment by

funds/special purpose services and finding reinforcing regulatory

vehicle listing and use ways to increase frameworks and

capital markets for liquidity). ensuring stable

asset recycling. Review collaboration macroeconomic

Support the opportunities with environment.

development of development banks Focus on developing

project bonds through regarding local an investor base and

credit enhancement currency issuances. seek optimal ways to

mechanisms where Expand the investor access already

appropriate. base and reinforce developed markets in

Review the prudential the legal environment. the region.

framework of Tap institutional

institutional investors investors through

related to investment direct lending to

limits. infrastructure projects.

Countr

y c

hara

cte

ristics


142

REFERENCE

Allen and Overy (2012). Allen and Overy, along with ING, has developed PEBBLE, an open-sourcefunding format for greenfield project financings, 3 December. Available at www.allenovery.com/publications/en-gb/Pages/Allen-and-Overy-along-with-ING-has-developed-PEBBLE-an-open-source-funding-format-for-greenfield-project-financings.aspx.

Asian Development Bank (ADB) (n.d.). Background on ASEAN+3 Bond Market Forum (ABMF).Available at www.adb.org/sites/default/files/publication/29702/abmf-background.pdf.

Basel Committee on Banking Supervision (2010). Basel III: A Global Regulatory Framework for More

Resilient Banks and Banking Systems. Basel, Switzerland: Bank for International Settlements.

Bloomberg News (2016). China’s muni bond boom sees turnover swell to record $34 billion, 27 July.Available at www.bloomberg.com/news/articles/2016-07-26/china-s-muni-bond-boom-sees-turnover-swell-to-record-34-billion.

Burger, J., F. Warnock, and V. Warnock (2015). Bond market development in developing Asia. ADBEconomics Working Paper Series, No. 448. Manila: ADB.

Chua, J. (2015). The emerging renewable financing landscape, 1 July. Available at www.eco-business.com/news/the-emerging-renewable-financing-landscape/.

Climate Bonds Initiative (2016). Bonds and climate change: state of the market in 2016. Available atwww.climatebonds.net/resources/publications/bonds-climate-change-2016.

Dela Paz, C. (2016). SEC approves PPP listing rules, 11 November. Available at https://ppp.gov.ph/?in_the_news=sec-approves-ppp-listing-rules.

Deloitte (2015). Decisions at the crossroads capital market trends and their implications on Asia,February. Available at www2.deloitte.com/content/dam/Deloitte/sg/Documents/financial-services/sea-fsi-capital-market-trends-and-their-implications-on-Asia-noexp.pdf.

Ehlers, T., F. Packer, and E. Remolona (2014). Infrastructure and corporate bond markets in Asia. InFinancial Flows and Infrastructure Financing Reserve Bank of Australia, RBA AnnualConference Volume, Alexandra Heath and Matthew Read, eds. Reserve Bank of Australia.

Genberg, H. (2015). Capital market development and emergence of institutional investors in theAsia-Pacific region. Asia-Pacific Development Journal, vol. 22, No. 2 (December), pp. 1–26.

Inderst, G. (2016). Infrastructure investment, private finance, and institutional investors: Asia froma global perspective. ADBI Working Paper Series, No. 555. Tokyo: Asian Development BankInstitute.

InfraPPP (2013). Bangkok’s skytrain operator raises $2.13 billion in IPO, 10 April. Available atwww.infrapppworld.com/news/bangkok-s-skytrain-operator-raises-2-13-billion-in-ipo.

Kang, H., P. Jeasakul, and C. Lim (2015). A bird’s eye view of finance in Asia. In The Future of Asian

Finance, R. Sahay and others, eds. Washington, D.C.: International Monetary Fund.

Lambert, D. (2016). Yes, we need a dedicated bond guarantee fund for Indian infrastructure. AsianDevelopment Blog, January. Available at https://blogs.adb.org/blog/yes-we-need-dedicated-bond-guarantee-fund-indian-infrastructure.

Levinger, H., and C. Li (2014). What’s behind recent trends in Asian corporate bond markets? DeutscheBank Research, 31 January. Available at www.dbresearch.com/PROD/RPS_EN-PROD/PROD0000000000451971/What%E2%80%99s_behind_recent_trends_in_Asian_corporate_bon.pdf.


143

Moody’s Investors Service (2014). Moody’s: default rates rise for infrastructure projects but remain betterthan average for project finance loans, 6 March. Available at www.moodys.com/research/Moodys-Default-rates-rise-for-infrastructure-projects-but-remain-better—PR_294378.

Nikkei Asian Review (2016). SMBC to debut project finance debt securities, 12 March. Available at http://asia.nikkei.com/Markets/Tokyo-Market/SMBC-to-debut-project-finance-debt-securities.

Organization for Economic Cooperation and Development (OECD) (2014a). Annual Survey of Large

Pension Funds and Public Pension Reserve Funds. Paris. Available at www.oecd.org/daf/fin/private-pensions/2014_Large_Pension_Funds_Survey.pdf.

(2014b). Pension Markets in Focus. Paris. Available at www.oecd.org/daf/fin/private-pensions/Pension-Markets-in-Focus-2014.pdf.

(2015a). Infrastructure Financing Instruments and Incentives. Paris. Available at www.oecd.org/finance/private-pensions/Infrastructure-Financing-Instruments-and-Incentives.pdf.

(2015b). Regulation of insurance company and pension fund Investment: OECD report to G20Finance Ministers and Central Bank Governors. Available at www.oecd.org/g20/topics/financing-for-investment/Regulation-of-Insurance-Company-and-Pension-Fund-Investment.pdf.

(2019). Risks rising in corporate debt market, 25 February. Available at www.oecd.org/finance/risks-rising-in-corporate-debt-market.htm.

Patnaik, I., and others (2013). Foreign investment in the Indian Government bond market. WorkingPapers, No. 13/126. National Institute of Public Finance and Policy.

Preqin (2016). 2016 Preqin Global Infrastructure Report. Available at www.preqin.com/docs/samples/2016-Preqin-Global-Infrastructure-Report-Sample-Pages.pdf.

Raghu, A., and E. Kaiser (2012). Malaysian toll expressway firm to sell $9.7 bln sukuk. Reuters,9 January. Available at www.reuters.com/article/malaysia-plus-sukuk-idUSL3E8C90Y120120109.

Rebello, J. (2016). Three years on infra debt funds have very few takers, ET Markets, 31 August.Available at http://economictimes.indiatimes.com/markets/stocks/news/three-years-on-infra-debt-funds-have-very-few-takers/articleshow/53939327.cms.

Sahay, R., and others (eds.) (2015). The Future of Asian Finance. Washington, D.C.: InternationalMonetary Fund.

Shenoy, D. (2013). Inflation indexed bonds in India, 16 May. Available at http://capitalmind.in/2013/05/inflation-indexed-bonds-in-india/.

Standard and Poor Global (2018). 25 years of rated infrastructure corporates and project finance.Available at www.spratings.com/documents/20184/774196/25+Years+of+Rated+Infrastructure+Corporates+and+Project+Finance.pdf.

Stanton, Daniel (2017). Asia Bond: Paiton Energy’s US$2bn dual-tranche project bond. InternationalFinancing Review. Available at www.ifre.com/asia-bond-paiton-energys-us2bn-dual-tranche-project-bond/21310072.fullarticle.

Stock Exchange of Thailand (n.d., a). Infrastructure business. Available at www.set.or.th/en/products/listing2/set_iff_p3.html.

(n.d., b). Infrastructure Fund (IFF). Available at www.set.or.th/en/products/listing2/set_infra_p1.html.


144

Svirydzenka, K. (2016). Introducing a new broad-based index of financial development. IMF WorkingPaper, No. 16/5. Washington, D.C.: IMF.

Thaichareon, K., and O. Sriring (2016). Thai finmin says restriction on foreign bond issues not a capitalcontrol. Reuters, 15 February. Available at www.reuters.com/article/thailand-bonds/thai-finmin-says-restriction-on-foreign-bond-issues-not-a-capital-control-idUSL3N15U1Y3.

World Bank (2014). Global Financial Development Dataset. Available at www.worldbank.org/en/publication/gfdr/data/global-financial-development-database.

World Economic Forum, and Oliver Wyman (2014). Direct Investing by Institutional Investors:

Implications for Investors and Policy-Makers. Geneva: World Economic Forum. Available atwww3.weforum.org/docs/WEFUSA_DirectInvestingInstitutionalInvestors.pdf.

World Justice Project (2016). World Justice Project Rule of Law Index 2016. Washington, D.C.

145

ASIA-PACIFIC SUSTAINABLE DEVELOPMENT JOURNALSUBMISSION GUIDELINES

The Asia-Pacific Sustainable Development Journal (APSDJ) is based on recognition of the

interconnected and multidisciplinary nature of sustainable development. Published biannually by

the United Nations Economic and Social Commission for Asia and the Pacific, it is intended

to stimulate debate and assist in the formulation of evidence-based policymaking in the

Asia-Pacific region towards the implementation of the 2030 Agenda for Sustainable

Development.

APSDJ welcomes the submission of original contributions on themes and issues related to

sustainable development that are policy-oriented and relevant to Asia and the Pacific. The

articles should be centred on discussing challenges pertinent to one or more dimensions of

sustainable development, policy options and implications and/or policy experiences that may be

of benefit to the region.

Authors familiar with the Asia-Pacific region are encouraged to submit their work. All

manuscripts will undergo a rigorous double-blind peer-review process.

1. Manuscripts

All materials submitted for the consideration of the Editorial Board should be in English. The

manuscripts should be typed, double-spaced, on one side of white A4 paper. They should not

exceed 8,000 words, including tables, figures, references and other materials and include a

short abstract (200 words) of the issues addressed and the most important policy-related

findings.

Manuscripts are accepted on the understanding that they may be edited. Contributors are

required to sign a consent form verifying that their articles are original and have not been

published in any other publications or refereed journal(s).

All manuscripts will be submitted to double-blind peer review by professionals in the field.

The name(s) of the author(s) or other identifying information should, therefore, be placed only on

the title page in order to preserve anonymity. The title page should also include such information

as institutional affiliation(s), complete mailing address, telephone number and the email of the

author(s).

Manuscripts should be sent by email to the chief editors of APSDJ at [email protected].

2. Tables and figures

All tables and figures should be numbered consecutively with Arabic numbers. Each table

should be typed double-spaced. Tables and figures should be planned to fit the proportions of

the printed page. Full information on the source(s) should appear below the table or figure,

followed by notes, if any, in lower-case letters.

Once a manuscript is accepted for publication, the author(s) will be asked to supply figures,

tables and charts, preferably in Microsoft Excel or any major spreadsheet programme.


146

3. Footnotes and quotations

Footnotes, if any, should be numbered consecutively with superscript Arabic numerals. They

should be typed single-spaced and placed at the bottom of each page. They should not be used

solely for citing references.

Quotations should be double-spaced. A copy of the page(s) of the original source of the

quotation, as well as a copy of the cover page of that source should be provided.

4. References

Authors should ensure that there is a complete reference for every citation in the text.

References in the text should follow the author-date format, followed, if necessary, by page

numbers, for example, Becker (1964, pp. 13-24). List only references that are cited in the text or

footnotes. References, listed alphabetically, should be typed double-spaced on a separate page

in the following style:

Sen, Amartya (2009). The Idea of Justice. Cambridge, Mass: Harvard University Press.

Husseini, Rana (2007). Women leaders attempt to bridge East-West cultural divide. Jordan

Times, 9 May.

Krueger, Alan B., and Lawrence H. Summers (1987). Reflections on the inter-industry wage

structure. In Unemployment and the Structure of Labour Markets, Kevin Lang and

Jonathan S. Leonard, eds. London: Basis Blackwell.

United Nations publications may be obtained from bookstores anddistributors throughout the world.Please consult your bookstore or write to any of the following:

Customers in: America, Asia and the Pacific

Email: [email protected]: shop.un.orgTel: +17036611571Fax: +1 703 996 1010

Mail orders to:United Nations PublicationsPO Box 960Herndon, Virginia 20172United States of America

Customers in: Europe, Africa and the Middle East

United Nations Publicationsc/o Eurospan GroupEmail: [email protected]: un.org/publicationsTel: +44 (0) 1767 604972Fax: +44 (0) 1767 601640

Mail orders to:United Nations PublicationsPegasus Drive, Stratton Business ParkBigglewade, Bedfordshire SG 18 8TOUnited Kingdom

For further information on this publication, please address your enquiries to:

Chief EditorsAsia-Pacific Sustainable Development Journal

Social Development Division andMacroeconomic Policy and Financing for Development DivisionUnited Nations Economic and Social Commission for Asia and the Pacific

United Nations Building, Rajadamnern Nok AvenueBangkok 10200, ThailandEmail: [email protected]

For more details, please visit www.unescap.org/publication-series/APSDJ.

19

-00

91

3

ISBN 978-92-1-120797-2

The Asia-Pacific Sustainable Development Journal (APSDJ) is a rebrandedpublication issued by the Economic and Social Commission for Asia andthe Pacific (ESCAP). It builds on the success of two former ESCAP journals– the Asia-Pacific Population Journal, launched in 1986, and the Asia-PacificDevelopment Journal, launched in 1994.

APSDJ is based on recognition of the interconnected and multidisciplinarynature of sustainable development. Published biannually, it aims to stimulatedebate and assist in the formulation of evidence-based policymaking in theAsia-Pacific region towards the implementation of the 2030 Agenda forSustainable Development.