United Nations - Trade Statistics in Policymaking
United Nations - Trade Statistics in Policymaking
United Nations - Trade Statistics in Policymaking
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 of its authorities, or
concerning the delimitation of its frontiers or boundaries.
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.
All material in this publication may be freely quoted or reprinted, but acknowledgment is required, together with a copy of the
publication containing the quotation or reprint.
The use of this publication for any commercial purpose, including resale, is prohibited unless permission is first obtained from the
secretary of the Publication Board, United Nations, New York. Requests for permission should state the purpose and the extent of
reproduction.
ii
Acronyms and abbreviations
ADB Asian Development Bank
AFTA ASEAN Free Trade Area
AMAD Agricultural Market Access Database
ANZCERTA Australia-New Zealand Closer Economic Relations Trade Agreement
APEC Asia-Pacific Economic Cooperation
APTA Asia-Pacific Trade Agreement
APTIAD Asia-Pacific Preferential Trade and Investment Agreements Database
ARCA additive revealed comparative advantage
ARIC Asian Regional Integration and Cooperation
ARTNeT Asia-Pacific Research and Training Network on Trade
ASEAN Association of Southeast Asian Nations
BIMSTEC Bay of Bengal Initiative for Multi-Sectoral Technical and Economic Cooperation
BTAs bilateral trade agreements
CAMAD Common Analytical Market Access Database
CGE computable general equilibrium
COMTRADE United Nations Commodity Trade Statistics Database
CTS Consolidated Tariff Schedule
DDA Doha development agenda
DOTS Direction of Trade Statistics (IMF)
ESCAP Economic and Social Commission for Asia and the Pacific
EV equivalent variation
FDI foreign direct investment
FTAs free trade agreements
GATT General Agreements on Tariffs and Trade
GDP gross domestic product
iii
GE general equilibrium
GTAP Global Trade Analysis Project
HHI Hirschmann-Herfindahl index
HS1996 Harmonised System 1996
IDB Integrated Data Base
IIT intra-industry trade
IMF International Monetary Fund
INDLKBTA India-Sri Lanka Bilateral Trade Agreement
ITC International Trade Centre
LDCs least developed countries
MacMaps Market Access Maps
MDG millennium development goal
MFN most favoured nation
MPM marginal propensity to import
NAFTA North American Free Trade Agreement
NTMs non-tariff measures
OECD Organisation for Economic Co-operation and Economic Development
OLS ordinary least squares
PAFTA Pan-Asian Free Trade Area
PE partial equilibrium
PTAs preferential trade agreements
QRs quantitative restrictions
RCA revealed comparative advantage
RO regional orientation
ROW the rest of the world
iv
TRAINS Trade Analysis and Information System
RTAs regional trade agreements
SITC Standard International Trade Classification
TDI trade dependence index
TID Trade and Investment Division (ESCAP)
UNCTAD United Nations Conference on Trade and Development
UNDP United Nations Development Programme
UNSD United Nations Statistics Division
US United States
WB World Bank
WDI World Development Indicators
WITS World Integrated Trade Solution
WTD World Trade Database
WTO World Trade Organization
v
Symbols
vi
Acknowledgments
This reference material is a compilation of work associated with tracking and analyzing regional trade flows and regionalism
processes. This work has been pursued by the Trade Policy Section of the Trade and Investment Division within the mandate
given by the ESCAP members. It provides comprehensive explanations of most of the indicators featured in the Asia-Pacific
Trade and Investment Agreements Database (APTIAD).
The publication was prepared by Ms. Mia Mikic, ESCAP and Mr. John Gilbert, Utah State University, United States, under
the overall supervision of Mr. Xuan Zengpei, Director, Trade and Investment Division, and Ms. Tiziana Bonapace, Chief,
Trade Policy Section of ESCAP.
vii
This page intentionally left blank
Contents
Acronyms and abbreviations ................................................................................................................................ iii
Symbols ................................................................................................................................................................... vi
Acknowledgements ................................................................................................................................................. vii
I. Introduction .......................................................................................................................................... 1
II. Notation ................................................................................................................................................. 11
III. Trade and economy .............................................................................................................................. 17
• Trade dependence ............................................................................................................................. 18
• Import penetration ............................................................................................................................ 20
• Export propensity ............................................................................................................................. 22
• Marginal propensity to import ......................................................................................................... 24
IV. Trade Performance .............................................................................................................................. 27
• Growth rate of exports ..................................................................................................................... 28
• Normalized trade balance ................................................................................................................. 30
• Export/import coverage .................................................................................................................... 32
V. Direction of Trade ................................................................................................................................ 35
• Export share ..................................................................................................................................... 36
ix
• Import share ..................................................................................................................................... 38
• Trade share ....................................................................................................................................... 40
• Regional market share ...................................................................................................................... 42
• Trade intensity .................................................................................................................................. 44
• Size adjusted regional export share .................................................................................................. 46
• Regional Hirschmann ....................................................................................................................... 48
• Trade entropy ................................................................................................................................... 50
VI. Sectoral Structure of Trade ................................................................................................................. 53
• Competiveness ................................................................................................................................. 56
• Major export category ...................................................................................................................... 58
• Sectoral Hirschmann ........................................................................................................................ 60
• Export diversification ....................................................................................................................... 62
• Revealed comparative advantage ..................................................................................................... 64
• Additive revealed comparative advantage ....................................................................................... 66
• Michelaye index ............................................................................................................................... 68
• Regional orientation ......................................................................................................................... 70
x
• Complementarity .............................................................................................................................. 72
• Export similarity ............................................................................................................................... 74
• Sectoral intra-industry trade ............................................................................................................. 76
• Aggregate intra-industry trade ......................................................................................................... 78
• Marginal intra-industry trade ........................................................................................................... 80
• Aggregate marginal intra-industry trade .......................................................................................... 82
• Trade overlap ................................................................................................................................... 84
VII. Protection .............................................................................................................................................. 87
• Average tariff ................................................................................................................................... 90
• Weighted average tariff .................................................................................................................... 92
• Tariff dispersion ............................................................................................................................... 94
VIII. Beyond Indicators ................................................................................................................................ 97
xi
This page intentionally left blank
I. Introduction
1. BACKGROUND ON TRADE MONITORING WORK IN ESCAP1
In March 2005, the Asia-Pacific Research and Training Network on Trade (ARTNeT) held its First Capacity Building
Workshop on Trade Research. The main objective of the workshop was to provide a menu of easy-to-use analytical methods
and tools for trade research which deliver findings of use to the policymakers in the Asia-Pacific region. The Trade and
Investment Division (TID) of ESCAP serves as a secretariat for ARTNeT and was tasked with organizing the workshop.
Decisions on the workshop topics had to take into account the characteristics of the target group of countries in which the
analysis was going to be most used: developing countries and the least developed countries (LDCs) of the ESCAP region.
Several constraints are linked to this target: limited availability of trade and other economic time series data and their
imperfect reliability; lack of technical and/or human resources for appropriate analyses in some of the countries, and under-
established practice by policymakers to use the findings of complex empirical research in their policymaking. Given these
constraints, as well the duration of the workshop itself, there was little option to include more sophisticated methods of
empirical research such as the gravity model and computable general equilibrium (CGE) models even though they are typical
methods of choice in the analysis of trade flows and trade policy impacts. The choice therefore fell on statistical analysis
based on a set of trade indices and indicators to assess trade patterns and characteristics and changes in trade patterns of
individual economies and their various groupings in this region. While initially this choice was perceived as a “second-best”
(as compared with the “first-best” modeling tools), further developments in the network activities and the ESCAP-TID work
had linked to and built on this approach. Moreover, more recent criticisms on shortcomings of some modeling approaches and
interpretations of their results allowed for these more straightforward and less pretentious methods to be unwrapped.
1
An earlier version under the title 'Commonly Used Trade Indicators: A Note' was first presented as part of training material for the First ARTNeT
Capacity Building Workshop on Trade Research, Bangkok, March 2005. The authors would like to thank participants at that workshop for useful
comments. Assistance was provided by Zaw Win Aung in preparing a demonstration database in Access format with the calculation of several
indicators for a CD-ROM. Yaomin Wen, Dandan Jian, Christopher Kuonqui and Xingxing Yao all contributed in different ways during their
internships with TID/ESCAP.
1
Meantime, work on mapping, tracking and monitoring of the preferential trade agreements in the ESCAP region for the Asia-
Pacific Trade and Investment Agreements Database (APTIAD) progressed to a level where tools for tracking and monitoring
had to be chosen. Again, given technical and other constraints, a natural choice fell on trade performance indicators. A
concept paper was developed to explain which indicators should be focused on in this particular work and to outline the
progression from this approach to more challenging methods based on modeling.2
The early harvest of findings on mapping and monitoring of the preferential trade agreements (PTAs) in the region is already
available online on APTIAD (http://www.unescap.org/tid/aptiad). At present APTIAD covers over 130 preferential trade
agreements linking at least one ESCAP economy with the rest of the world. Approximately 2/3 of those have been
implemented while the number of agreements under negotiation or preparation increases daily. APTIAD tracks trade
dependence, trade balance and trade shares indices over 2001-2005 for all the agreements in the database. It also has an
interactive platform for calculation of another set of trade performance indicators for various membership configurations.
In parallel with the development of APTIAD, other international organizations and research institutions also gave more
attention to using trade performance indicators in assessment of regional trade integration in the Asia-Pacific. For example,
the Asian Development Bank (ADB) developed a sub-site Asian Regional Integration and Cooperation (ARIC) which
includes information on preferential agreements in trade and investment and trade indicators.
Economists use statistical data and indicators in all branches of economics. Sometimes indicators are used as leading
indicators (when they have predictive power), sometimes to describe what has happened in the past (lagging indicators). Often
one distinguishes between business and macroeconomic indicators. Simple and composite indices are established as
recognized approaches in monitoring progress in achieving various policy goals or in benchmarking various policy options.
2
John Gilbert (2006) 'A Framework for Assessing the Role of PTAs in Trade and Development in Asia and the Pacific,' ESCAP, TID, mimeograph.
2
With better computing power and the development of exciting econometric and general equilibrium models, the use of
indicators went into a somewhat quieter phase in the 1990s. Nevertheless we have many examples of contemporary use of
economic indicators for important areas of national, regional and global policymaking, as illustrated below:
a) Indicators to monitor market access indicators for the Millennium Development Goal 8 (http://www.mdg-trade.org)
b) Indicators to track progress towards achieving the European Union internal market
(http://ec.europa.eu/internal_market/score/docs/score15/score15_en.pdf)
c) OECD’s International Trade Indicators database provides cross-disciplinary information for globalization analysis
purposes with particular emphasis on trade aspects
(http://stats.oecd.org/wbos/default.aspx?DatasetCode=TRADEINDMACRO)
d) The World Bank combines a number of indicators to provide data on population, environment, economy, states,
markets and similar (http://www.worldbank.org)
e) The Statistics Division of ESCAP offers an online database with core annual indicators including economic,
demographic, social and composite indices (http://www.unescap.org/stat/data/main/goalindicatorarea.aspx) and
another one on short-term indicators by individual country
(http://www.unescap.org/stat/data/statind/areaSectorIndicators.aspx)
f) Freedom House and the Heritage Foundation use composite indices to rank countries in terms of economic freedom
and status of corruption and bribery (http://www.freedomhouse.org)
g) “Doing Business” at the World Bank benchmarks business regulation and enforcement by tracking selected indicators
(http://www.doingbusiness.org/)
h) The World Economic Forum and World Competitiveness Center are behind now widely recognized national
competitiveness indices (http://www.weforum.org/en/initiatives/gcp/Global%20Competitiveness%20Report/index.htm
and http://www.imd.ch/research/centers/wcc/index.cfm).
i) UNCTAD offers the trade and development index “to capture the complex interaction between trade and development
and, in the process, to monitor the trade and development performance of countries.” Rankings of countries are
available online (http://www.unctad.org/Templates/Page.asp?intItemID=3582&lang=1).
3
A broad definition of a trade indicator is that it is an index or a ratio that can be used to describe and assess the state of trade
flows and trade patterns of a particular economy or economies and can be used to monitor these flows and patterns over time
or across economies/regions. Indicators are the result of using statistical data and as such are the most available, or often the
only available, input for so-called evidence-based policymaking (Scott, 2005).
Evidence-based policymaking consists of using statistical and other techniques in obtaining sound and transparent data to be
used in the consultative process between government and other stakeholders in any area of public policymaking, including
trade policy. From the examples given above it appears that evidence-based policymaking in practice has focused on tracking,
monitoring and evaluation of the results of policies that have been put in place. However, there is significant scope to use data
and empirical research in the process of making decisions in the public area (including on issues recognition, policy choice
and sequencing, or forecasting future developments). This approach to policymaking increases transparency and makes
policymakers more accountable. Moreover, evidence-based policymaking also builds the platform for democratization of
policymaking, particularly in area like trade policy where impacts always cause some income-distributional effects. As
recently reported, stakeholders in policymaking agree that “lack of reliable information leads to the dissemination of ideology
instead of knowledge, making people misperceive actions of the governments and undermining a democratic decision
process” (CESifo, 2007, p.22).
Trade indicators are used in the analysis of international or foreign trade, at the national, regional or global level. By using
these indicators one can provide insights into the following type of questions (this list is only illustrative and the ordering is
random):
4
• Are (regional) trading partners’ exports becoming more similar (more competitive) or more complementary?
• Is there a geographic “re-orientation” of exports after some external shock (such as financial crisis) or enforcement of
a trade agreement?
Asking these questions is a first step towards starting policy oriented research in trade. This is so because each of the above
can be followed by “why?” To answer “why” types of questions, one has to engage in further research to identify the most
relevant factors that are driving the changes. More often than not, this will require using some econometric testing and should
be thus the more technically challenging stage of research.
This handbook does not provide a comprehensive analysis of use of indicators. In particular, it does not contribute to a further
development of trade indices from the perspective of statistical theory. It mostly uses indices that have already being
constructed in the literature, groups them with respect to the type of information they best provide, and demonstrates their
calculation and interpretation in a way that is applicable to any of the ESCAP economies. The handbook, however,
contributes to the existing body of empirical literature as it systematizes the available trade indices in terms of the information
content they best provide, it clarifies their interpretations from the perspective of a single economy or more economies linked
by a preferential trade agreement, it provides precise formulae for calculation, and it offers some policy applications.
Often data for an individual country is available from national statistics, but for comparative purposes it is better to use data
sets available from international organizations, such as (in alphabetical order): the International Monetary Fund (IMF), the
International Trade Centre (ITC), the United Nations Statistical Division (UNSD), the United Nations Conference on Trade
and Development (UNCTAD), the World Bank (WB) and the World Trade Organization (WTO). The United Nations
Commodity Trade (COMTRADE) statistical database (http://unstats.un.org/unsd/comtrade/) is used most often. The same
commodity trade data is used in the WITS (World Integrated Trade Solution) by the World Bank and to an extent in the
5
UNCTAD Handbook of Statistics which is available on CDs and at http://www.unctad.org/statistics/handbook. There is a
downloadable (but not free) dataset World Trade Database (WTD) maintained by Statistics Canada at http://www.statcan.ca.
There are also separate databases on tariffs (and some non-tariff barriers). They are mostly sourced from the published Trade
policy reviews (WTO) and countries’ notifications on tariff schedules and are accessible through WITS or Trade Analysis
Information System (TRAINS) or Integrated Data Base (IDB).3
Although trade data is widely available, often it is not complete and it is definitely not fully reliable. In each of the sources of
trade data listed above, there are numerous remarks and notes on the weaknesses of data sets.4 ITC recently published notes
on the reliability of trade statistics and reasons for discrepancies in the trade data between trading partners (details available
on the ITC website and http://www.trademap.org/reliability_data.php). Users of trade data must be aware of potential
problems arising from aggregation of data, currency conversion, incomplete reporting of data and other potential problems
when constructing their own trade series.
Without doubt trade in goods is very limited area of economic activity even for a developing country. An increasing share of
trade, employment, and income is sourced from tradable services. However, national statistical systems, particularly in
developing countries, still have difficulty in providing reliable and timely data on services integrating available information
on exports and imports, income, employment and investment. Several agencies (UNSD, European Commission, IMF, OECD,
UNCTAD and WTO) are collectively working on devising a better system of collection of services statistics and currently the
Manual on Statistics in International Trade in Services, 2002 (http://unstats.un.org/unsd/tradeserv/Papers/m86_english.pdf)
and Balance of Payments Manual, 5th edition (http://www.imf.org/external/np/sta/bop/BOPman.pdf) offers best the practical
advice. Having a good and comprehensive database including not only trade information but also flows of capital, exchange
3
There are also other useful datasets such as Consolidated Tariff Schedule (CTS), Common Analytical Market Access Database (CAMAD), or AMAD
(Agricultural Market Access Database). See more at http://www.unescap.org/tid/projects/artnetbk05.asp.
4
The international workshop on Country practices in compilation of international merchanidise trade statistics held at ESCAP in December 2006
revealed a number of problems – more details at http://unstats.un.org/unsd/trade/WS%20Bangkok06/Bangkok_workshop_imts_trade.htm.
6
rates and data on domestic production and prices, is an important pre-requisite for informed analysis and policymaking. The
international community works on standards for the exchange of statistical information from different sources to improve the
usability of such data for all users (details on http://www.sdmx.org).
4. READY-TO-USE INDICATORS
It is not always necessary to calculate indicators. There are various online databases which provide a number of trade
performance indicators of value for use by policymakers. Below is a short-list of these “over-the-counter” trade indicators for
illustrative purposes only.
• ITC – http://www.intracen.org/menus/countries.htm. Country-level statistics provides trade and market profiles and
benchmarks for national trade performance and indicators on export supply and import demand. More detailed and
specialized datasets are available from the ITC on a subscription basis (www.trademap.org and www.macmap.org).
• WORLD BANK – http://www.worldbank.org/globaloutlook
• UNCTAD – Handbook of Statistics – Part VIII on Special studies
http://stats.unctad.org/handbook/ReportFolders/ReportFolders.aspx includes some relevant indices on trade
performance and competitiveness.
• OECD - http://stats.oecd.org/wbos/default.aspx?DatasetCode=TRADEINDMACRO
• APTIAD – http://www.unescap.org/tid/aptiad
• ARIC - http://aric.adb.org/indicator.php
Moreover, there is now a long list of empirical literature that provides a good source for various trade performance indicator
results. For economies in Asia and the Pacific of special interest is the literature available on the World Bank website:
http://lnweb18.worldbank.org/eap/eap.nsf/Sectors/PREM/7E6DC202606E30C485256DFF0062B362?OpenDocument .
This site includes trade data on East Asia.
7
Alternatively, and hopefully, one can calculate desired indicators by using formulae from this handbook and available
datasets. Furthermore (and probably more rewarding!) one can work on constructing new indices and composite indicators to
answer some of the questions one might be interested in. In constructing an index or an indicator, one needs to observe certain
criteria. Most importantly, indicators must be simple, measurable, consistent and comparable. Their coverage, of course,
may vary as they can be constructed to capture flows of trade at one or more of different levels: sectoral, country, regional, or
global. Furthermore, a database should meet these criteria: (1) present a simplified but reliable view of the country, region or
whatever the subject of the database would be; (2) contribute to shared knowledge among users, and (3) be useful in making
policymaking more transparent and policymakers more accountable.
The objective of this handbook is to be used as a reference in preparation of analysis of already available merchandise trade
statistical information for assessment of various trade issues, discussion on negotiating positions and ultimately for conducting
consultations. Indicators are grouped in the following categories5:
• Trade and economy (trade dependence index, marginal propensity to import, import penetration, export propensity,
trade per capita).
• Trade performance (growth rate of trade (exports/imports), normalized trade balance, export/import coverage).
• Direction of trade (trade intensity, intra-regional trade shares, trade entropy).
5
Some of the indicators could be used to reflect on more than one of the changes in trade patterns so it is possible that other classifications would
group indicators somewhat diferently. Most of the indicators listed can be computed for total levels of trade (exports plus imports), or separately for
exports and imports for a given economy or a group of economies.
8
• Sectoral structure of trade (major export category, index of export diversification, revealed comparative advantage,
intra-industry trade, trade overlap, complementarity index, export similarity index, competitiveness index).
• Protection (average applied/bound tariffs, weighted average tariffs, dispersion of tariffs).
The handbook consists of seven chapters. Chapter II covers some basic data and notation issues that are common to the
following chapters. Chapters III to VI cover definitions, formulae and some calculations for illustrative purposes for selected
ESCAP economies of the trade indicators. While it cannot be claimed that all existing indices and indicators are covered in
this handbook, considerable effort was put into making sure that most commonly used indicators related to the real side of
economy are covered. In each case, the indicators are presented in a simple two page format. The first page has a basic
definition, notes on usage, the range of values, interpretation and limitations, and a simple application to give an idea of how
the indicator might be used. The second page contains more technical information, including a formal mathematical
definition, potential data sources, and a numerical example of the calculation. This material is intended to be helpful when it is
necessary not just to understand the indices, but to apply them to a new problem. Chapter VII discusses measures of openness
(or protection) at national, regional or global level and offers some indicators commonly used in this area. Chapter VIII
concludes with a forward-looking discussion of tools which allow further analysis, enabling more solid answers to “why?”
type questions.
9
This page intentionally left blank
II. Notation
Throughout the following chapters we have tried to adopt a common notation so that the trade flow indices are clearly and
consistently defined. Understanding the notation requires a small investment in learning about the structure of trade data, the
elements of sets, and summation notation. In this chapter we set out the basic ideas.
First consider the basic data element in trade analysis. In any given year, country A sells a given volume of product B to
country C. By the same token and from country C’s perspective, country C buys this volume of product B from country A in
the same year. In trade terminology, country A exports product B to country C, and country C imports product B from country
A. The value of this transaction is our basic data element – the product trade flow, in value terms. (Strictly, while the volume
should be the same for each country, barring loss and/or measurement error, the value of the transactions will differ slightly
because of transportation/insurance. In practice, because the export and import data is often from different sources, it can
differ substantially.) We denote this particular data element , which means simply “exports of product B from A to C.”
More generally, the basic data element in the text that follows will be where i is the good that is being traded, s is the
country that is selling (source of exports) and d is the country that is buying (destination for imports) – and the labels chosen
are entirely arbitrary. Because we are referring to single flow, it follows that = , where m denotes imports.
There are different ways in which the two parties to the transaction are referred to in the various data sources. In
COMTRADE, for example, the data is classified by who supplied it. Thus, if the data is exports, COMTRADE would refer to
the exporting country (A) as the “reporter” and the importing country (B) as the “partner.” On the other hand, in import data
from COMTRADE, the “reporter” is the importing country and the “partner” is the exporting country. This can be a little
confusing. The Global Trade Analysis Project (GTAP), on the other hand, adopts a convention that always identifies the
direction of the flow. In this terminology the exporting country is the “source” and the importing country the “destination.”
Any given trade flow is an export from the perspective of the source and an import from the perspective of the destination.
We adopt this convention here.
The label i can be thought of as representing a single product category in a larger product set. We can let I be the set of all
product groups in the data. The size of the set and its exact composition will depend on the data being used (e.g., HS1996 or
11
SITC, at various digit levels). So for example, it might contain three elements: I={Agriculture, Textiles, Manufactures}.
Similarly, the labels s and d can be thought of as representing a single element of a larger set of countries, and we can let S
and D, respectively, represent those sets. It may be that each set contains only a single country, or multiple countries, or all
countries, depending on the context. So, for example, we might have S={Australia}, and D={Japan, the Republic of Korea,
China}. Often the destination might be the entire world.
Once we have defined our sets, we can define summations (adding up) over those sets. The summation notation is a very
convenient way of expressing the various trade flow totals that we will be interested in. Its usage can be made clear with a few
examples. Suppose we want to know the total trade between two countries. Then we are interested in the following sum:
∀∈
This is the total over all products of exports from country s to country d. The sigma (∑) is the summation operator, and the
expression underneath tells us what we are adding up over. It is read “for all i that are a member of the set I.” In words, the
expression says to add up all the individual product trade flows from country s to country d. We can of course add up over
more than one index, so for example the expression:
∀∈ ∀∈
is the total exports of all products from country s to all destinations in the set D. If D is the set of all possible destinations, then
the expression is the total exports of country s. Since addition is commutative (i.e., A+B is equal to B+A), the order of the
addition does not matter, and we may write the same expression in a more compact form:
12
∀∈ ,∀∈
However, even this is a bit cumbersome. At the risk of upsetting the mathematical purists, we adopt a further simplification
and drop the sets from the summation expression, instead letting the index itself indicate which set we need, which is always
clear from context. Hence, we can write the above expression as follows:
This means that we want to add up the export flows of country s over all products and destinations.
Because of the frequency with which total trade statistics are used, we adopt a further simplification for total trade flows. That
is, we use an upper case X to denote the total trade flow of all products from one country to another. So is the total value
of exports from A to B. More generally:
=
This avoids repeatedly summing over the i index when it is clear from context that this is required.
We finish with some numerical examples that should make the link between notation and data clear. The data we use here and
throughout many of the examples in the handbook has been drawn from the GTAP project (Dimaranan, 2006). It is presented
in a highly aggregated form that is not really suitable for real world calculations, but is more than adequate to illustrate the
computations involved.
13
The trade data can be arranged in many ways, and different organizations will suit different purposes. A common method is to
use a matrix, such as the following for world trade in wheat in 2002:
In this representation the exporting countries/regions are in the rows, while the importing countries are in the columns. Hence,
for example, in 2002 Australia exported $US168 million in wheat to the Republic of Korea (highlighted in blue). On the other
hand, China imported $150 million in wheat from the rest of the world in 2002 (measured net of insurance and freight,
highlighted in red).
Note that the diagonal elements for single economies are zero. This is because, by definition, a country does not trade with
itself. If the matrix contains aggregate regions (e.g., ASEAN), however, there will be non-zero elements on the diagonal.
These represent the intra-regional exports of the group. Note also that for convenience, it is common to put the row and
6
We use the term ROW (rest of world) in this and in all subsequent tables in the handbook as shorthand for the sum over all countries/regions that do
not appear in the table expressly.
14
column sums in the table. Row sums are total exports of the product by the country/region in the row heading, while column
sums are total imports of the product by the country/region in the column heading.
For example, total exports of wheat by Australia in 2002 were worth $1393 million (highlighted in green). In terms of our
mathematical notation this would be:
Where i={wheat}, s={Australia} and d={Australia, New Zealand, China, Republic of Korea, ASEAN, rest of world}.
Similarly, total imports of wheat by the ASEAN economies in 2002 were $1433 million (highlighted in purple). In terms of
our notation this would be:
Where i={wheat}, s={Australia, New Zealand, China, Republic of Korea, ASEAN, rest of world}, and d={ASEAN}.
We can form a total trade matrix by adding across all industries first, such as in the example below. This is read in exactly the
same way as the product matrix, the rows are the exporters (sources) and the columns the importers (destinations). Again, it is
common to supply the row and column totals, since these are used so frequently. They represent the total exports of the
countries/regions in the row headings and the total imports of the countries/regions in the column headings, respectively.
15
For example, total exports from New Zealand to ASEAN in 2002 were $1187 million (highlighted in blue). On the other
hand, total exports by ASEAN to all countries in 2002 were $385344 (highlighted in red). Mathematically this is:
Where s={ASEAN}, and d={Australia, New Zealand, China, Republic of Korea, ASEAN, rest of world}.
Australia New Zealand China Rep. of Korea ASEAN ROW Total
Australia 0 3348 4878 5213 7070 41062 61571
New Zealand 2564 0 777 659 1187 9633 14820
China 6058 867 0 16243 23301 398969 445438
Rep. of Korea 2516 301 24232 0 17521 115477 160047
ASEAN 9850 1130 24832 15950 80320 253262 385344
ROW 43045 7169 156819 103738 174227 4341696 4762074
Total 64033 12815 211538 141803 303626 5160099 5893914
As a final note, tables are generally limited to two dimensions, and this limits the amount of useful data that we can present in
them. The trade data has three dimensions (i.e., product, source and destination, and four if we include time). Depending on
context, it will be useful to present only the column/row sums for various products or countries or years. Hence, for example,
if our interest is in the total trade by country of all commodities, we would drop the destinations from the table and present
only the sum over all destinations. This would free up a dimension in the table to use for sectors. Nonetheless, we continue to
keep the underlying summation in our presentation of index calculations, for preciseness, to avoid any potential ambiguity,
and because when working with original data carrying out the summations may be required.
16
III. Trade and Economy
This chapter includes information on the following indicators, which provide information on the significance of international
trade relative to the overall economy:
The first indicator, the trade dependence index, is one of the most widely used trade statistics. Also termed the openness
index, it measures the ratio of international trade to the total value of net output (gross domestic product or GDP). A high
index value is often interpreted as indicating a more open economy (hence the second terminology) although the index is
biased by other factors, including economic size.
The next two indicators are variations on trade dependence indicators that may be more useful in understanding an economy’s
vulnerability to certain types of external shocks (e.g., exchange rate movements). The import penetration index measures the
proportion of domestic demand that is satisfied by imports. It is also termed an import dependency index and an aggregate
self-sufficiency index. The second is the export propensity index, which measures the share of exports in GDP.
Finally, the marginal propensity to import index is an approximation to a commonly used macroeconomic variable. It tells us
how much we expect imports to rise for a given rise in the value of income (GDP). Again, this may be useful in evaluating the
possible response of an economy to external or internal (policy) shocks.
17
Trade Dependence Index
What does it tell us? The trade dependence index
(also often called the openness index) is a measure
Indonesia
of the importance of international trade in the
overall economy. It can give an indication of the Malaysia
degree to which an economy is open to trade
(subject to some serious limitations). Philippines
18
Trade Dependence – Technical Notes
Mathematical definition:
∑X s ds + ∑s M sd
× 100
GDPd GDP Exports Imports TDI
2001 160446 55884 28845 53
Where d is the country under study, s is 2002 195660 56857 30664 45
the set of all other countries, X is total 2003 234773 60777 52518 48
bilateral exports, M is total bilateral 2004 256835 64241 61809 49
imports and GDP is gross domestic 2005 286961 85660 71368 55
product (of country d). In words, the
numerator is total exports from d plus Trade and Production Data for Indonesia (2001-2005)
total imports to d, and the denominator
is the GDP of d. Sample calculation: The table above presents GDP and trade data for Indonesia for
the period 2001-2005. The GDP data is from the World Development Indicators
Data sources: Trade data can be (WDI), while the trade data is from COMTRADE. All of the data is in compatible
obtained from the United Nations units (millions of current US dollars). To calculate the trade dependence index
Commodity Trade database (TDI) we add the total exports to total imports to obtain total trade. In 2005 exports
(COMTRADE), the International were $85660 (in red) while imports were $71368 (in green). Hence total trade was
Monetary Fund (IMF) Direction of $158028. Dividing this number by the GDP of $286961 (in blue) and multiplying
Trade statistics, the World Trade by 100 to convert to a percentage, we obtain 55 per cent (in purple). This is the
Database (WTD) maintained by measure of trade dependence or openness for Indonesia in 2005. As we can see
Statistics Canada, and the GTAP from the table, the indicator has remained fairly constant over the five year period.
database from Purdue University. GDP
data can be obtained from the World
Development Indicators, the Penn
World Tables, and national sources.
19
Import Penetration
What does it tell us? The import penetration rate
shows to what degree domestic demand (the
difference between GDP and net exports) is satisfied Indonesia
by imports. Calculated at the sectoral level it is
termed the self-sufficiency ratio. The index may be Malaysia
used as the basis of specific policy objectives
targeting self-sufficiency. It may provide an Philippines
indication of the degree of vulnerability to certain
types of external shocks. Singapore
7
In practice the index can exceed 100 per cent if re-exports are not accounted for, see example.
20
Import Penetration – Technical Notes
Domestic Import
Mathematical definition: GDP Exports Imports
Demand Penetration
2001 71216 32077 35702 74841 48
∑M s sd
× 100
2002 76814 35116 35157 76855 46
GDPd −∑ X +∑ M
s ds s sd
2003 79634 36095 40836 84374 48
2004 86703 39610 45417 92510 49
2005 98366 41198 46310 103478 45
Where d is the country under study, s is the
set of all other countries, X is total bilateral
Trade and Production Data for the Philippines (2001-2005)
exports, M is total bilateral imports and
GDP is gross domestic product (of country
Sample calculation: The table above presents GDP and trade data for the
d). In words, the numerator is total imports
Philippines for the period 2001-2005. The GDP data is from the World
to d, the denominator is the GDP of d less
Development Indicators (WDI), while the trade data is from COMTRADE. All
total exports and plus total imports (i.e.,
of the data is in compatible units (millions of current US dollars). The fourth
total domestic demand).
column in the table is domestic demand, calculated as GDP less exports and
plus imports, this is the demominator of the import penetration ratio. We
Data sources: Trade data can be obtained
illustrate the required calculation with 2005. The value of GDP was $98366 (in
from the United Nations Commodity Trade
blue). Subtracting exports of $41198 (red) and adding imports of $46310
database (COMTRADE), the International
(green) yields domestic demand of $103478 (purple). Taking the ratio of
Monetary Fund (IMF) Direction of Trade
imports to domestic demand ($46310/$103478) and multiplying by 100 to put
statistics, the World Trade Database (WTD)
the expression in percentage terms yields 45 per cent. In words, imports (in
maintained by Statistics Canada, and the
aggregate) satisfied 45 per cent of the aggregate demand in the Philippines in
GTAP database from Purdue University.
2005, with the remaining 55 per cent satisfied by domestic production.
GDP data can be obtained from the World
Development Indicators, the Penn World
Note: It is common to calculate this indicator on a sectoral basis, using sectoral
Tables, and national sources.
imports and value added. This is usually termed a self-sufficiency ratio.
21
Export Propensity
What does it tell us? The index shows the
overall degree of reliance of domestic
producers on foreign markets. It is similar to Indonesia
the trade dependence index, but may provide a
better indicator of vulnerability to certain types Malaysia
of external shocks (e.g., falls in export prices
or changes in exchange rates). It may be a Philippines
policy target.
Singapore
Definition: The ratio of exports to GDP,
defined as a percentage. Thailand
Limitations: The export propensity index is Export Propensity Index for ASEAN-6 Economies (2005)
biased upward by the re-exports (can be
corrected for in principle). Will tend to be Example: The above figure presents the export propensity index for the
negatively correlated with economic size. A ASEAN-6 economies in 2005. Note again the very high figures for
high export propensity may be an inappropriate Singapore and Malaysia – this in part represents the presence of re-exports
policy target from an efficiency perspective. in the raw data.
8
As with the import penetration index, in practice the index can exceed 100 per cent if re-exports are not accounted for. See example.
22
Export Propensity – Technical Notes
Mathematical definition:
∑ s
X ds
× 100
GDPd
Export
GDP Exports
Propensity
Where d is the country under study, s is 2001 115536 64877 56
the set of all other countries, X is total 2002 126877 67845 53
bilateral exports, and GDP is gross 2003 142640 79892 56
domestic product (of country d). In 2004 161349 96069 60
words, the numerator is total exports 2005 176222 110108 62
from d, and the denominator is the GDP
of d. Trade and Production Data for Thailand (2001-2005)
Data sources: Trade data can be Sample calculation: The table above presents GDP and trade data for Thailand for
obtained from the United Nations the period 2001-2005. The GDP data is from the World Development Indicators
Commodity Trade database (WDI), while the trade data is from COMTRADE. All of the data is in compatible
(COMTRADE), the International units (millions of current US dollars). The third column of the table presents the
Monetary Fund (IMF) Direction of export propensity values. For 2005, the value of exports was $110108 (in blue),
Trade statistics, the World Trade while the (net) value of production was $176222 (in red). Taking the ratio of
Database (WTD) maintained by exports to GDP and multiplying by 100 to convert to percentage terms we have 62
Statistics Canada, and the GTAP per cent.
database from Purdue University. GDP
data can be obtained from the World
Development Indicators, the Penn World
Tables, and national sources.
23
Marginal Propensity to Import
What does it tell us? The marginal propensity to
import (MPM) is measure of the extent to which 3.00
imports are induced by a change in incomes. The
relevance for policymakers depends on the cycle 2.50
of the economy. With higher MPM, in an
economic downturn with a fall in GDP, there 2.00
will also be a significant fall in imports as Indonesia
compared with lower a MPM. More generally, a 1.50 Malaysia
higher MPM reduces the multiplier effect of an Philippines
1.00
increase in GDP.
Singapore
0.50
Definition: The ratio of the change in total Thailand
imports to the change in GDP over a defined 0.00
period (typically one year).
-0.50
Range of values: In macroeconomic theory 2002 2003 2004 2005
ranges between 0 (with no part of extra GDP
spent on additional imports) to 1 when the whole
extra GDP created is spent on imports. MPM for Selected ASEAN Economies (2002-2005)
Limitations: The MPM is not a constant and can Example: In the figure above we have calculated the MPM over time for
vary over time, so care should be taken in using selected economies in ASEAN. These are calculated year on year. We
it as an input to policy decisions. Calculations observe a substantial increase in the MPM in all the economies in 2003.
based on annual data only approximate the true Note that the calculated values can in fact be negative or exceed unity, in
value, and may lie outside of the theoretically particular in countries with high levels of re-exports.
sensible range (see example).
24
Marginal Propensity to Import – Technical Notes
Mathematical definition:
25
This page intentionally left blank
IV. Trade Performance
This chapter includes information on the following indicators, which provide information on the trade performance of an
economy or region:
The first indicator, the compound growth rate, directly tracks how the value of exports, imports or trade is changing over time.
This is one of the most common indicators used when assessing the progress of an economy in any area of economic activity.
Often the rate is calculated at level of product groups to identify ‘dynamic sectors’ or at the regional level to indicate
‘dynamic regions.’
The next two indicators are normalized versions of the trade balance, the difference between the value of exports and the
value of imports. The normalized trade balance index expresses the trade balance as a fraction of total trade. This adjustment
allows for meaningful comparisons across countries or time periods by eliminating problems with the units of measurement.
Similarly, the export/import coverage index expresses the trade balance in terms of a ratio of its components rather than a
difference, thereby eliminating the units of measurement and making comparisons across time/regions easier. Although often
a variable of considerable interest to policymakers, the trade balance indices should be interpreted carefully. A trade surplus is
not necessarily better than a deficit, and the trade balance can reflect many factors outside of the realm of trade policy. This is
particularly true for balances calculated at the bilateral level.
27
Growth Rate of Exports
What does it tell us? The growth rate is one of
the most common indicators used when 30
assessing the progress of an economy in any 25
area of economic activity. Often the rate is 20
calculated at level of product groups to identify 15
‘dynamic sectors.’ Comparison of such 10 China
indicators over many countries might be of India
5
interest to producer or exporter associations,
investors, policymakers and trade negotiators. 0 ASEAN
-5 ROW
Definition: The annual compound percentage -10
change in the value of exports between two -15
periods.
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
Range of values: The growth rate is a
percentage. It can take a value between -100 per Selected Export Growth Rates (1993-2002)
cent (if trade ceases) and +∞. A value of zero
indicates that the value of trade has remained Example: Calculating the growth rates for major economies or regional
constant. groupings, and ranking them relative to the growth of world exports, can
identify ‘dynamic’ regions in world trade. The above figure depicts the
Limitations: Does not explain the source of annual growth rate in exports for China, India and ASEAN relative to the
growth. When evaluating long periods need to rest of world average. The dip in the period 1997-1998, particularly strong
be careful of changes in measurement and for ASEAN, presumably reflects the Asian financial crisis. Overall,
methods. Growth rates assessed on nominal however, the region is very dynamic, with average growth rates over the
trade figures may be distorted by exchange rate decade of 14 per cent for China, 10 per cent for India, and 8 per cent for
movements and other factors in the short run. ASEAN. This compares to only 5 per cent for the rest of the world.
28
Growth Rate of Exports – Technical Notes
Mathematical definition:
Compound
Annual
China India Growth
⎛ 1
⎞ Growth
⎜ ⎛ ∑sw X 1sw ⎞n ⎟ from 1998
⎜⎜ ⎟ − 1⎟ × 100 2002 445438 25.2 53497 8.8
⎜ ⎜⎝ ∑sw X sw ⎟
0 2001 355696 -5.6 47772 7.8
⎠ ⎟ 2000 376706 22.6 46305 10.1
⎝ ⎠ 1999 307277 13.2 41859 9.7
1998 271536 38167
Where s is the set of countries in the
source, w is the set of countries in the Total Value of Exports for China and India (1998-2002)
world, X0 is the bilateral total export
flow in the start period, X1 is the Sample calculation: The table above presents the raw export figures for China and
bilateral total export flow in the end India over a five year span (in US$ millions). To calculate the growth of China’s
period, and n is the number of periods exports between 2001 and 2002 (in blue) we take (445438/355696-1)×100=25 per
(not including the start). cent. To calculate the compound growth rate of India over the full period we take the
ratio of the starting and ending period levels (in red): 53497/38167, and raise to the
Data sources: The United Nations power of 1/4 (since there are four periods of growth). The result is 1.088. Subtracting
Commodity Trade database one and multiplying by 100 gives us 8.8 per cent.
(COMTRADE), the International
Monetary Fund (IMF) Direction of Notes: The compound growth rate is the annual growth rate required to generate a
Trade statistics, the World Trade given level of total growth over a period of length n. If n is equal to 1, this is simply
Database (WTD) maintained by the percentage change in exports. The growth rate may also be calculated for a subset
Statistics Canada, and the GTAP of destinations, or for a subset of products, with appropriate modification. The growth
database from Purdue University. rate can also be calculated for imports and/or trade (imports plus exports).
29
Normalized Trade Balance
What does it tell us? The normalized trade
balance represents a record of a country's trade Australia
transactions with the rest of the world Bangladesh
normalized on its own total trade. In general, China
India
economists expect that the trade balance will be
Indonesia
zero in the long run, thus imports are financed
Rep. of Korea
by exports, but it may vary considerably over
Malaysia
shorter periods.
New Zealand
Philippines
Definition: The trade balance (total exports less Singapore
total imports) as fraction of total trade (exports Sri Lanka
plus imports). Thailand
Viet Nam
Range of values: The index range is between
-1 and +1, which allows unbiased comparisons -0.1 0.0 0.1 0.2 0.3 0.4
across time, countries and sectors. A value of
zero indicates trade balance.
Normalized Trade Balance for Selected Economies (2002)
Limitations: The economic reasons for a trade
surplus/deficit are complex, and the index Example: The above figure depicts the normalized trade balance for
cannot directly help shed light on them. selected economies in 2002. Negative figures indicate a deficit, observed in
Potential for misuse high, especially with Australia, Singapore and Sri Lanka. Substantial surpluses are observed in
respect to bilateral balances. China and Indonesia. However, the balance depicted is calculated on
merchandise trade. The balance including services trade may differ.
30
Normalized Trade Balance – Technical Notes
Mathematical definition: Normalized
Imports Exports
Trade Balance
∑ sw
X sw − ∑ ws M ws Bangladesh
China
5919
211538
6340
445438
0.03
0.36
∑ sw
X sw + ∑ ws M ws India 45356 53497 0.08
Indonesia 36284 67001 0.30
Malaysia 69829 103150 0.19
Where s is the set of countries in New Zealand 12815 14820 0.07
the source, w is the set of Philippines 30554 39294 0.13
countries in the world, X is the Rep. of Korea 141803 160047 0.06
bilateral total export flow, and M Singapore 104726 95076 -0.05
is the bilateral total import flow in Sri Lanka 5415 4783 -0.06
the end period. In words, we take Thailand 47609 65488 0.16
Viet Nam 14624 15335 0.02
total exports from the source
region less total imports to the
source region, and divide by the Total Value of Exports and Imports for Selected Economies (2002)
total trade of the source region.
Sample calculation: The table above presents the raw export and import figures for
Data sources: The United selected economies in 2002 (in US$ millions). To calculate the normalized trade balance
Nations Commodity Trade for Bangladesh we would take the difference between total exports (in red) and total
database (COMTRADE), the imports (blue). This is $421. We divide this by total trade, $12259. The result is +0.03.
International Monetary Fund
(IMF) Direction of Trade Notes: Calculated at the product level, the normalized trade balance corresponds to the
statistics, the World Trade intra-industry trade index. The normalized trade balance can also be calculated against
Database (WTD) maintained by a sub-region or country, but economic logic does not give us any reason to expect
Statistics Canada, and the GTAP bilateral balance. Hence the results of such an analysis may be informative on the
database from Purdue University. structure of trade, but have little economic significance.
31
Export/Import Coverage
What does it tell us? This is an alternative to Australia
the normalized trade balance. It tells us whether Bangladesh
or not a country’s imports are fully paid for by China
exports in a given year. In general, economists India
expect that the trade balance will be zero in the Indonesia
long run, thus imports are financed by exports, Rep. of Korea
but it may vary considerably over shorter Malaysia
periods. New Zealand
Philippines
Definition: The ratio of total exports to total Singapore
imports. Sri Lanka
Thailand
Range of values: The values for this index Viet Nam
range from 0 when there are no exports to +∞
0.0 0.5 1.0 1.5 2.0 2.5
when there are no imports. A ratio of 1 signals
full coverage of imports with exports (trade
balance). Export/Import Coverage for Selected Economies (2002)
Limitations: Same as for the normalized trade Example: We use the same subset of countries and year as for the
balance. The economic reasons for a trade normalized trade balance, for comparative purposes. The above figure
surplus/deficit are complex, and the index depicts the export/import coverage for selected economies in 2002. Now
cannot directly help shed light on them. figures less than one indicate a deficit, observed in Australia, Singapore
Potential for misuse high, especially with and Sri Lanka. Substantial surpluses are observed in China and Indonesia.
respect to bilateral balances. Again, the balance depicted is calculated on merchandise trade. The
balance including services trade may differ.
32
Export/Import Coverage – Technical Notes
Mathematical definition: Export/Import
Imports Exports
Coverage
∑ sw
X sw Australia
Bangladesh
64033
5919
61571
6340
0.96
1.07
∑ ws
M ws China 211538 445438 2.11
India 45356 53497 1.18
Indonesia 36284 67001 1.85
Where s is the set of countries in Malaysia 69829 103150 1.48
the source, w is the set of New Zealand 12815 14820 1.16
countries in the world, X is the Philippines 30554 39294 1.29
bilateral total export flow, and M Rep. of Korea 141803 160047 1.13
is the bilateral total import flow Singapore 104726 95076 0.91
in the end period. In words, we Sri Lanka 5415 4783 0.88
take total exports from the source Thailand 47609 65488 1.38
Viet Nam 14624 15335 1.05
region, and divide by the total
imports of the source region.
Total Value of Exports and Imports for Selected Economies (2002)
Data sources: The United
Nations Commodity Trade Sample calculation: The table above presents the raw export and import figures for
database (COMTRADE), the selected economies in 2002 (in US$ millions). To calculate the export/import coverage for
International Monetary Fund Bangladesh we would take total exports (in red) and divide by total imports (blue), i.e.,
(IMF) Direction of Trade $6340/5919=1.07.
statistics, the World Trade
Database (WTD) maintained by Notes: The export/import coverage can also be calculated against a sub-region or country,
Statistics Canada, and the GTAP but economic logic does not give us any reason to expect bilateral balance. Hence the
database from Purdue results of such an analysis may be informative on the structure of trade, but have little
University. economic significance.
33
This page intentionally left blank
V. Direction of Trade
This chapter includes information on the following indicators which can inform policymakers on the level of and changes in
the regional pattern or direction of trade flows:
• Export/import/trade shares
• Regional market share
• Trade intensity
• Size adjusted regional export share
• Regional Hirschmann
• Trade entropy
Trade share statistics have many uses. Hence, for example, if we want to know which economies are the most important
export destinations for India, we could calculate the shares of India’s exports to different economies in India’s total exports,
and rank them in order of size. If we wish to know whether the introduction of AFTA increased trade among ASEAN
economies, we can measure changes over time in the proportion of ASEAN trade is that is with other ASEAN economies. The
regional market share is a special case that is relevant for assessing regional trading agreements (or trade blocs).
Trade share statistics are also a basic building block for other, more sophisticated, indices of the pattern of trade. The trade
intensity index can be thought of as a normalized export share. It tells us whether or not the observed share of trade is greater
than the world average, or ‘intense’ relative to what we might expect. The size adjusted regional export share statistic is
similar, and is a special case designed to look for at whether the intra-regional trade of a given regional trade group is high
relative to what might be expected given the size of the membership.
The last two indicators are different measures of the geographical concentration or diversification of a country (or region’s)
export profile. The regional Hirschmann and trade entropy indices address the question of whether or not an economy is
heavily reliant on a small number of export markets, or sells to a diverse range of economies. They are useful to assess
integration into global markets, or the vulnerability of an economy to shifts in economic conditions in particular markets.
35
Export Share
What does it tell us? The export share tells us how
important a particular export partner is in terms of the 22.5
overall export profile of an economy. Changes in the 22.0
export share over time may indicate that the economies in 21.5
question are becoming more integrated. In the case of 21.0
intra-regional export shares, increases in the value over 20.5
time are sometimes interpreted as an indicator of the
20.0
significance of a regional trading bloc if one exists, or as
19.5
a measure of potential if one is proposed. The latter
assumes that groups with high shares are in some sense 19.0
‘natural’ trading partners. 18.5
18.0
Definition: The export share is the percentage of exports 17.5
from the region under study (the source) to the region of
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
interest (the destination) in the total exports of the source
region.
Range of values: Takes a value between 0 and 100 per Intra-Regional Export Shares (per cent) for ASEAN (1992-
cent, with higher values indicating greater importance of 2002)
selected trading partner.
Example: Suppose that we wish to find out the proportion of
Limitations: The intra-regional export share is ASEAN’s exports that go to other ASEAN members. This
increasing in the size of the bloc considered by definition, particular export share is called the ASEAN intra-regional
so comparing the shares across different blocs may be export share. The values are depicted in the figure above. As we
misleading. High or low export shares and changes over can see, the share has increased marginally since the formation of
time may reflect factors other than trade policy. the ASEAN Free Trade Area (AFTA) in 1993.
36
Export Share – Technical Notes
Mathematical definition: The
export share is defined as:
Indonesia Malaysia Philippines Singapore Thailand Viet Nam ROW Total
Indonesia 0 2029 797 5347 1229 389 57210 67001
∑ sd
X sd
× 100 Malaysia 1550 0 1289 18471 3923 657 77260 103150
∑ sw
X sw Philippines 145 2289 0 2353 516 90 33901 39294
Singapore 4104 14355 2252 0 5370 1443 67552 95076
Where s is the set of countries in Thailand 1192 3074 802 4870 0 0 55550 65488
the source, d is the set of countries Viet Nam 257 321 272 934 0 0 13551 15335
in the destination, w is the set of ROW 29036 47761 25142 72751 36571 12045 5285264 5508570
countries in the world, and X is the Total 36284 69829 30554 104726 47609 14624 5590288 5893914
bilateral total export flow. The
numerator is thus exports from the Simplified Total Trade Matrix (2002)
source to the destination, the
denominator total exports from the Sample calculation: Exports from the ASEAN-6 economies to other members of
source. ASEAN-6 are highlighted in blue. Total exports from ASEAN-6 economies to all
countries are highlighted in red (the row sums). Adding together the numbers in blue we
Data sources: The United Nations get $80320. This is total intra-ASEAN exports. Adding the numbers in red we get
Commodity Trade database $385344. This is total ASEAN exports to all countries. Taking the ratio and multiplying
(COMTRADE), the International by 100 we get 20.8 per cent, the intra-regional export share for ASEAN in 2002.
Monetary Fund (IMF) Direction of
Trade statistics, the World Trade
Database (WTD) maintained by
Statistics Canada, and the GTAP
database from Purdue University.
37
Import Share
What does it tell us? The import share tells us how
important a particular trade partner is in terms of the 30.0
overall import profile of an economy. Changes in the
import share over time may indicate that the 25.0
economies in question are becoming more integrated.
20.0
In the case of intra-regional import shares, increases in
the value over time are sometimes interpreted as an 15.0
indicator of the significance of a regional trading bloc
if one exists, or as a measure of potential if one is 10.0
proposed.
5.0
Definition: The import share is the percentage of
imports from the region of interest (the source) to the 0.0
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
region under study (the destination) in the total
imports of the destination.
Range of values: Takes a value between 0 and 100 Intra-Regional Import Shares (per cent) for ASEAN (1992-2002)
per cent, with higher values indicating greater
importance of selected trading partner. Example: Staying with the previous example, now suppose that we
wish to find out the proportion of ASEAN’s imports that come from
Limitations: The intra-regional import share is other ASEAN members. This particular import share is called the
increasing in the size of the bloc considered by ASEAN intra-regional import share. The values are depicted in the
definition, so comparing the shares across different figure above. As we can see, the share has increased more
blocs may be misleading. High or low shares and substantially than the export share since the formation of the ASEAN
changes over time may reflect numerous factors other Free Trade Area (AFTA) in 1993.
than trade policy.
38
Import Share – Technical Notes
Indonesia Malaysia Philippines Singapore Thailand Viet Nam ROW Total
Mathematical definition:
Indonesia 0 2029 797 5347 1229 389 57210 67001
∑
Malaysia 1550 0 1289 18471 3923 657 77260 103150
M sd
sd
× 100 Philippines 145 2289 0 2353 516 90 33901 39294
∑ wd
M wd Singapore 4104 14355 2252 0 5370 1443 67552 95076
Thailand 1192 3074 802 4870 0 0 55550 65488
Where s is the set of countries in Viet Nam 257 321 272 934 0 0 13551 15335
the source, d is the set of countries ROW 29036 47761 25142 72751 36571 12045 5285264 5508570
in the destination, w is the set of Total 36284 69829 30554 104726 47609 14624 5590288 5893914
countries in the world, and M is the
bilateral import flow. The Simplified Total Trade Matrix (2002)
numerator is thus imports from the
source to the destination, the Sample calculation: Exports from the ASEAN-6 economies to other members of
denominator total imports to the ASEAN-6 are also imports of ASEAN-6 economies from other ASEAN-6 economies.
destination. These are again highlighted in blue. Total imports by ASEAN-6 economies from all
countries are highlighted in green (the column sums). Adding the numbers in blue we
Data sources: The United Nations get $80320. This is total intra-ASEAN imports. Adding the numbers in green we get
Commodity Trade database $303626. This is total ASEAN imports. Taking the ratio and multiplying by 100 we get
(COMTRADE), the International 26.5 per cent, the intra-regional import share for ASEAN in 2002.
Monetary Fund (IMF) Direction of
Trade statistics, the World Trade Note: The intra-regional import share may differ from the intra-regional export share
Database (WTD) maintained by when total trade is not balanced (i.e., the value of exports does not equal the value of
Statistics Canada, and the GTAP imports). Import and export shares differ in general because the economies from which
database from Purdue University. an economy buys goods are not necessarily the same as those to which it sells goods.
39
Trade Share
What does it tell us? The trade share tells us how
important a particular trade partner is in terms of 25.0
the overall trade profile of an economy. Changes
in the trade share over time may indicate that the
20.0
economies in question are becoming more
integrated. In the case of intra-regional shares,
15.0
increases in the value over time are sometimes
interpreted as an indicator of the significance of a
10.0
regional trading bloc if one exists, or as a measure
of potential if one is proposed.
5.0
Definition: The trade share is the percentage of
the region under study’s trade (imports plus 0.0
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
exports) with another region of interest, in the
total trade of the region under study.
Range of values: Takes a value between 0 and Intra-Regional Trade Shares (per cent) for ASEAN (1992-2002)
100 per cent, with higher values indicating greater
importance of selected trading partner. Example: The figure above presents the ASEAN intra-regional trade
shares for the years 1992-2002. Given the two preceding examples, it is
Limitations: The intra-regional trade share is not surprising that we again observe an increase in the relative
increasing in the size of the bloc considered by importance of intra-ASEAN trade over time, from around 20 per cent in
definition, so comparing the shares across 1993 to 23 per cent in 2002. In fact, the intra-regional trade share is a
different blocs may be misleading. High or low weighted average of the intra-regional export and import shares, and so
shares and changes over time may reflect must lie between the two (see technical notes).
numerous factors other than trade policy.
40
Trade Share – Technical Notes
Mathematical definition:
Indonesia Malaysia Philippines Singapore Thailand Viet Nam ROW Total
Indonesia 0 2029 797 5347 1229 389 57210 67001
∑ sd
X sd + ∑ds M ds
× 100 Malaysia 1550 0 1289 18471 3923 657 77260 103150
∑ sw
X sw + ∑ws M ws Philippines 145 2289 0 2353 516 90 33901 39294
Singapore 4104 14355 2252 0 5370 1443 67552 95076
Where s is the set of countries in Thailand 1192 3074 802 4870 0 0 55550 65488
the source, d is the destination, w Viet Nam 257 321 272 934 0 0 13551 15335
is the set of countries in the world, ROW 29036 47761 25142 72751 36571 12045 5285264 5508570
X is the bilateral flow of exports Total 36284 69829 30554 104726 47609 14624 5590288 5893914
from the source and M is the
bilateral import flow to the source. Simplified Total Trade Matrix (2002)
Note the reversal of the usual
notation on the import side – we Sample calculation: Exports from the ASEAN-6 economies to other members of
want imports to and exports from ASEAN-6 are highlighted in blue. Total imports by ASEAN-6 economies from all
the same region when we calculate countries are highlighted in green. Total exports of ASEAN-6 economies to all countries
total trade. are highlighted in red. Adding the numbers in blue we get $80320. This represents both
intra-ASEAN imports and exports, so we need to multiply by two. Adding the numbers
Data sources: The United Nations in green we get $303626. This is total ASEAN imports. Adding the numbers in red we
Commodity Trade database get $385344. This is total ASEAN exports. Hence the intra-regional import share for
(COMTRADE), the International ASEAN in 2002 is $160640/($303626+$385344)×100, or 23.3 per cent.
Monetary Fund (IMF) Direction of
Trade statistics, the World Trade Note: The trade share is a weighted average of the import and export shares, with the
Database (WTD) maintained by weights being the share of total imports in total trade and the share of total exports in
Statistics Canada, and the GTAP total trade. It must therefore lie between the values for the export and import shares.
database from Purdue University.
41
Regional Market Share
What does it tell us? The regional market
Philippines Viet Nam
share statistic tells us the relative importance of 7% 2%
the members of a trade bloc in the intra-
regional trade of the bloc. It is a variation on Indonesia
the export share. The larger the value, the more 12% Singapore
the economy in question dominates the exports 34%
of the bloc in question.
Limitations: The usual limitations of shares Example: The above graph describes the regional markets shares for the
apply. A high (or low) regional market share ASEAN-6 economies in 2002. As we can see, intra-ASEAN exports are
may simply reflect the size of the economy in very heavily dominated by Singapore and Malaysia, which together account
world trade – i.e., the statistic is not for over 65 per cent of intra-ASEAN trade.
normalized.
42
Regional Market Share – Technical Notes
Mathematical definition:
∑ sb
X sb
× 100
Indonesia Malaysia Philippines Singapore Thailand Viet Nam ROW Total
∑
Indonesia 0 2029 797 5347 1229 389 57210 67001
db
X db Malaysia 1550 0 1289 18471 3923 657 77260 103150
Philippines 145 2289 0 2353 516 90 33901 39294
Where s is the set of source Singapore 4104 14355 2252 0 5370 1443 67552 95076
countries under study, b and d are Thailand 1192 3074 802 4870 0 0 55550 65488
the set of members of the trade bloc
Viet Nam 257 321 272 934 0 0 13551 15335
under study (the destinations), and
ROW 29036 47761 25142 72751 36571 12045 5285264 5508570
X is the bilateral flow of exports
from the source to the destination. Total 36284 69829 30554 104726 47609 14624 5590288 5893914
43
Trade Intensity
What does it tell us? We can think of the
trade intensity index as a uniform export 8.0
share. In other words, the statistic tells us 7.0
whether or not a region exports more (as a
6.0
percentage) to a given destination than the
world does on average. It is interpreted in 5.0
much the same way as an export share. It does 4.0
not suffer from any ‘size’ bias, so we can
compare the statistic across regions, and over 3.0
time when exports are growing rapidly. 2.0
1.0
Definition: The trade intensity statistic is the
ratio of two export shares. The numerator is 0.0
the share of the destination of interest in the
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
exports of the region under study. The
denominator is the share of the destination of
interest in the exports of the world as a whole.
Trade Intensity Index for ANZCERTA (1992-2002)
Range of values: Takes a value between 0
and +∞. Values greater than 1 indicate an Example: Suppose that we wish to assess the ‘intensity’ of trade among the
‘intense’ trade relationship. economies of ANZCERTA. The results are in the figure above. Because the
index in greater than one, trade within ANZCERTA would be regarded as
Limitations: As with trade shares, high or highly intense. Is this is consequence of the agreement? It is true that the
low intensity indices and changes over time intensity seems to have increased since the agreement was signed in 1993.
may reflect numerous factors other than trade On the other hand, the index was very high even before then. This probably
policy. reflects geographic proximity, and relative isolation from other markets.
44
Trade Intensity – Technical Notes
Mathematical definition: New Rep. of
Australia China ASEAN BIMSTEC ROW Total
Zealand Korea
∑ sd
X sd ∑
sw
X sw Australia
New Zealand
0
2564
3348
0
4878
777
5213
659
7070
1187
1723
171
39339
9462
61571
14820
∑ wd
X wd ∑ wy
X wy China 6058 867 0 16243 23301 4317 394652 445438
Rep. of Korea 2516 301 24232 0 17521 2297 113180 160047
Where s is the set of countries in ASEAN 9850 1130 24832 15950 80320 7617 245645 385344
the source, d is the destination, w BIMSTEC 590 88 2150 981 4181 2115 54515 64620
and y represent the countries in ROW 42455 7081 154669 102757 170046 38450 4246616 4762074
the world, and X is the bilateral
Total 64033 12815 211538 141803 303626 56690 5103409 5893914
flow of total exports. In words,
the numerator is the export share
Simplified Total Trade Matrix (2002)
of the source region to the
destination, the denominator is
Sample calculation: Intra-ANZCERTA exports are highlighted in blue, while total
export share of the world to the
exports from ANZCERTA are highlighted in red. Calculating the export share as before
destination.
we have $5912/$76391×100 or 7.7 per cent. World exports to ANZCERTA are
highlighted green, while total world exports are highlighted in purple. So the world
Data sources: The United
export share to ANZCERTA is $76848/$5893914 or 1.3 per cent. Taking the ratio we
Nations Commodity Trade
have 5.9, this is the trade-intensity index for ANZCERTA in 2002.
database (COMTRADE), the
International Monetary Fund
Note: In some cases the trade intensity index is adjusted by removing the intra-regional
(IMF) Direction of Trade
exports from the world total in the denominator. In the example, the denominator in the
statistics, the World Trade
world export share would be $5893914-$5912. Unless the region under study is very
Database (WTD) maintained by
large, this adjustment does not make a large difference. Trade intensity can also be
Statistics Canada, and the GTAP
calculated using trade shares instead of export shares.
database from Purdue University.
45
Size Adjusted Regional Export Share
46
Size Adjusted Regional Export Share – Technical Notes
Mathematical definition:
∑ sd
X sd ∑
sw
X sw Australia
New
Zealand
China
Rep. of
Korea
ASEAN BIMSTEC ROW Total
∑ sw
X sw ∑ wy
X wy Australia 0 3348 4878 5213 7070 1723 39339 61571
New Zealand 2564 0 777 659 1187 171 9462 14820
Where s is the set of countries in China 6058 867 0 16243 23301 4317 394652 445438
the source, d is the destination, w Rep. of Korea 2516 301 24232 0 17521 2297 113180 160047
and y the set of countries in the ASEAN 9850 1130 24832 15950 80320 7617 245645 385344
world, and X is the bilateral flow BIMSTEC 590 88 2150 981 4181 2115 54515 64620
of exports from the source. The ROW 42455 7081 154669 102757 170046 38450 4246616 4762074
numerator is the intra-regional Total 64033 12815 211538 141803 303626 56690 5103409 5893914
export share of group s. The
denominator is the share of group
s in world exports. Simplified Total Trade Matrix (2002)
Data sources: The United Sample calculation: Intra-ANZCERTA exports are highlighted in blue, while total
Nations Commodity Trade exports from ANZCERTA are highlighted in red. Total world exports are highlighted in
database (COMTRADE), the purple. First we form the intra-regional export share. This is intra-ANZCERTA exports
International Monetary Fund $(2564+3348) divided by total ANZCERTA exports $(61571+14820), or 0.077 (7.7 per
(IMF) Direction of Trade cent). Next we obtain ANZCERTA exports as a fraction of world exports, which is
statistics, the World Trade $(61571+14820)/5893914=0.013. Dividing the former share by the latter we obtain 5.9.
Database (WTD) maintained by This is the measure size adjusted export share for ANZCERTA.
Statistics Canada, and the GTAP
database from Purdue University.
47
Regional Hirschmann
What does it tell us? The Hirschmann index is 0.82
a measure of the geographical concentration of
exports. It tells us the degree to which a region 0.80
or country’s exports are dispersed across 0.78
different destinations. High concentration
levels are sometimes interpreted as an 0.76
indication of vulnerability to economic 0.74
changes in a small number of export markets.
An alternative measure is the trade entropy 0.72
index. 0.70
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
the region under study to all destinations.
9
Strictly, the lower bound is 1/n where n is the number of countries to which the economy under study exports. This will be close to zero in most
applications.
48
Regional Hirschmann – Technical Notes
Mathematical definition: New Rep. of
Australia China ASEAN BIMSTEC ROW Total
Zealand Korea
∑X
2
⎛ ⎞ Australia 0 3348 4878 5213 7070 1723 39339 61571
∑ ⎜⎜ ⎟
s sd New Zealand 2564 0 777 659 1187 171 9462 14820
⎝∑ X
d ⎟ China 6058 867 0 16243 23301 4317 394652 445438
sw sw ⎠
Rep. of Korea 2516 301 24232 0 17521 2297 113180 160047
ASEAN 9850 1130 24832 15950 80320 7617 245645 385344
Where s is the set of source BIMSTEC 590 88 2150 981 4181 2115 54515 64620
countries under study, d is the ROW 42455 7081 154669 102757 170046 38450 4246616 4762074
set of destinations, w is the set Total 64033 12815 211538 141803 303626 56690 5103409 5893914
of countries in the world, and
X is the bilateral flow of Simplified Total Trade Matrix (2002)
exports from the source to the
destination. We want to sum Sample calculation: The Republic of Korea’s bilateral exports are highlighted in blue, the
over all destinations, so the total exports in red. We can calculate export shares as a fraction in the usual manner. This
sets d and w contain the same gives us the following values: 0.016, 0.002, 0.151, 0, 0.109, 0.014, 0.707. Squaring each
elements. value and totalling gives us 0.54. The square root of this value is 0.73 – this is the regional
Hirschmann index for the Republic of Korea in 2002.
Data sources: The United Notes: This calculation illustrates the problem with aggregation bias. In this simplified trade
Nations Commodity Trade matrix the rest of world is a single share. This type of aggregation will push the calculated
database (COMTRADE), the Hirschmann index up. A Hirschmann index can also be calculated using import or trade
International Monetary Fund shares. The Hirschmann index is sometimes called the Hirschmann-Herfindahl index (HHI),
(IMF) Direction of Trade and is used in other contexts (see the sectoral version later in this volume). It is also
statistics, the World Trade calculated in several variants. It may be seen without the final square root operation, or using
Database (WTD) maintained percentages instead of fractions. It may also be normalized using the number of destinations.
by Statistics Canada, and the The latter adjustment turns the index into a measure of ‘evenness’ in the export share pattern.
GTAP database.
49
Trade Entropy
What does it tell us? The trade entropy index
1.20
is another measure of the geographical
concentration or dispersion of exports. High
1.00
values indicate that exports are geographically
diversified. This can be interpreted as a 0.80
measure of the degree to which the country
under study is integrated with the world 0.60
economy, or vulnerable to shocks in a limited
number of partners. 0.40
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
of the export shares (a weight that decreases
with the size of the share) of the country under
study across all destinations.
Trade Entropy Index for Republic of Korea (1992-2002)
Range of values: Takes a value between 0 and
+∞. Higher values indicate greater uniformity Example: We compare the trade entropy index with the regional
in the geographical dispersion of exports. The Hirschmann index by using the same example. Each export share for the
value of the index is maximized when the Republic of Korea is weighted by the ‘importance factor’ calculated by the
export share to every market is the same. natural log of the reciprocal of the trade share. The results are shown above,
again we can see that the Republic of Korea has diversified its export
Limitations: The trade entropy index is markets over the period considered, this time because the value of the index
subject to an aggregation bias. increases over time.
50
Trade Entropy – Technical Notes
Mathematical definition: New Rep. of
Australia China ASEAN BIMSTEC ROW Total
Zealand Korea
∑d ⎜⎜ s X ⎟⎟ ln⎜⎜ X ⎟ New Zealand 2564 0 777 659 1187 171 9462 14820
⎝ ∑sw sw ⎠ ⎝ ∑s sd ∑sw X sw ⎠
⎟
China 6058 867 0 16243 23301 4317 394652 445438
Rep. of Korea 2516 301 24232 0 17521 2297 113180 160047
ASEAN 9850 1130 24832 15950 80320 7617 245645 385344
Where s is the set of source countries BIMSTEC 590 88 2150 981 4181 2115 54515 64620
under study, d is the set of ROW 42455 7081 154669 102757 170046 38450 4246616 4762074
destinations, w is the set of countries Total 64033 12815 211538 141803 303626 56690 5103409 5893914
in the world, and X is the bilateral
flow of exports from the source to
Simplified Total Trade Matrix (2002)
the destination. We want to sum over
all destinations, so the sets d and w
Sample calculation: The Republic of Korea’s bilateral exports are highlighted in
contain the same elements. An
blue, the total exports in red. We can calculate export shares in the usual manner
entropy index can also be calculated
(except for the expression in percentage terms). This gives us the following values:
using import or trade shares.
0.016, 0.002, 0.151, 0, 0.109, 0.014, 0.707. Next we calculate the weights by taking
the natural logs of the reciprocal of the export shares: 4.13, and so on. Multiplying
Data sources: The United Nations
each share by the weight and summing over the destinations gives us 0.91 – this is the
Commodity Trade database
entropy index for the Republic of Korea in 2002.
(COMTRADE), the International
Monetary Fund (IMF) Direction of
Note: Again, note the potential for bias when we use aggregate statistics. Also, the
Trade statistics, the World Trade
definition of the index requires that the export shares lie in the range 0 to 1. Zero
Database (WTD) maintained by
shares are not allowed because division by zero is undefined, so these must be
Statistics Canada, and the GTAP
excluded from the calculation.
database from Purdue University.
51
This page intentionally left blank
VI. Sectoral Structure of Trade
This chapter includes indicators which reveal changes in the commodity structure of trade and thus are most useful for the
preparation of negotiating positions in trade negotiations. They also are very relevant for formulation of development
strategies, as they reflect directly or indirectly the competitive ability of a country’s or region’s economic sectors or activities.
The indices covered in this chapter are:
• Competitiveness
• Major export category (sectoral export share)
• Sectoral Hirschmann
• Export diversification
• Revealed comparative advantage
• Additive revealed comparative advantage
• Michelaye index
• Regional orientation
• Complementarity
• Export similarity
• Sectoral/aggregate intra-industry trade
• Sectoral/aggregate marginal intra-industry trade
• Trade overlap
The first two measures are based on sectoral export shares. The first, competitiveness, measures the share of an economy in
the world market for a particular good. It is a basic measure of world market power. The second, major export category,
addresses the trade profile of an economy, by measuring the proportion of total exports of the economy that are accounted for
53
by particular product categories. This is a useful starting point for evaluating the sectoral structure of an economy’s export
profile (e.g., is the export profile dominated by a few products, which ones?)
The sectoral export share is also a basic building block of some of the more complex indices. The sectoral Hirschmann index
and the export diversification index both use sector export shares to provide an overall measure of an economy’s export
diversification. This may be important in assessing the vulnerability of an economy to external trade shocks.
The next group of statistics is designed to look at the sectoral pattern of trade in great detail. International trade theory
describes how inter-industry trade (trade in different products) is driven by comparative advantage, or lower opportunity cost.
Revealed comparative advantage indices attempt to use the trade data to tell us the products in which economies have a
comparative advantage. We present three alternative measures, the revealed comparative advantage index, the additive
revealed comparative advantage index, and the Michelaye index. These measures are sometime used with the regional
orientation index, which highlights regional bias in the sectoral export pattern, to identify possible cases of trade diversion
with a regional trading agreement (where there is strong regional bias in the absence of comparative advantage).
The final group of indices might be collectively termed ‘overlap’ indices, as they share a common mathematical foundation.
Their purpose is to measure the degree to which categories match, or overlap. The first, complementarity, can be thought of as
an overall measures of the degree to which what one country has to sell matches what another wants to buy. It is often used as
an indicator of the potential for trade expansion following the formation of a regional trading bloc. Export similarity, by
contrast, provides an overall measure of the degree to which what two countries export matches, and thus is an indicator of the
degree to which countries are rivals on international markets.
The intra-industry trade indices measure overlap in the import/export profile of a single economy. They provide us with a
measure of the importance of scale economies and variety in the trade of an economy. This may be important in assessing
54
issues of trade fragmentation, or the potential size of adjustment costs under a proposed regional trade agreement, for
example. The trade overlap index is an alternative to the aggregate intra-industry trade measure, with both providing a single
economy-wide measure of the importance of intra- vs. inter-industry trade.
55
Competitiveness
What does it tell us? Competitiveness in
trade is broadly defined as the capacity of an Indonesia, 2.3%
India, 3.4%
industry to increase its share in international
Rep. of Korea,
markets at the expense of its rivals. The 3.6%
competitiveness index is an indirect measure
of international market power, evaluated
through a country’s share of world markets China, 24.1%
in selected export categories.
Limitations: The index will vary with the Example: Suppose we want to determine the most important economies in the
level of data aggregation. Somewhat limited world trade in textiles/apparel. We can calculate the share of each economy in
measure of market power, which may the world trade in textiles/apparel, and rank them. The results are shown
depend critically on market structure. above. The most important economy is China, followed by the Republic of
Korea and India.
56
Competitiveness – Technical Notes
Mathematical definition:
∑x d isd
× 100
∑ x
wd iwd
Rep. of
Australia China India Sri Lanka ROW Total
Korea
Where s is the country of Primary Products 25474 12544 670 4569 936 593865 638058
interest, d and w are the set Food Products 9422 9977 2055 4554 116 289570 315694
of all countries in the Textiles and Apparel 844 107040 16284 15103 2661 300870 442802
world, i is the sector of Other Manufactures 25831 315877 141038 29271 1070 3984273 4497360
interest, and x is the Total 61571 445438 160047 53497 4783 5168578 5893914
commodity export flow. In
words, it is the share of Total Exports by Category for Various Countries (2002)
country s’s exports of good
i in the total world exports Sample calculation: The table above presents the export flow data in a slightly different form.
of good i. We have totaled exports across destinations first, and presented total exports by category for the
selected regions. Suppose we want to calculate China’s share of the textiles and apparel
Data sources: The United (including leather) market. The total Chinese exports of textiles and apparel are highlighted in
Nations Commodity Trade blue. Total world exports of textiles and apparel are highlighted in red. The competitiveness
database (COMTRADE), index for China in textiles/apparel is therefore $107040/$442802×100=24.2 per cent.
the World Trade Database
(WTD) maintained by
Statistics Canada, and the
GTAP database from
Purdue University.
57
Major Export Category
What does it tell us? Major export category
is a simple measure of the extent
diversification of exports across sectoral
categories. If no single category accounts for Other, 21.9%
50 per cent or more of total exports, the
economy is classified as diversified.
Wearing
Identification of dominating products in Apparel, 40.6%
country’s trade is valuable for both trade
policy and adjustment management. Manufactures
NEC, 7.7%
Definition: The index is the value of the
largest sectoral export share in total exports
of a given economy.
Textiles, 13.1%
Range of values: Takes a value between 0 Other Crops,
and 100 per cent, with higher values 16.8%
indicating greater importance of the product
in the export profile of the economy in
Top Export Categories for Sri Lanka (2002)
question.
Example: Suppose we wish to identify the major export sectors of Sri Lanka.
Limitations: The index will vary with the
We take the value of exports of each commodity and express them as a
level of data aggregation. As an indicator of
percentage of total Sri Lankan exports. We then rank the shares in order of
diversification it is limited, one of the others
magnitude. The chart above depicts the results. We find that Sri Lanka is very
listed below is preferable.
heavily dependent on exports of wearing apparel.
58
Major Export Category – Technical Notes
Mathematical definition:
∑x d isd
× 100
∑X d sd Australia China
Rep. of
Korea
India Sri Lanka ROW Total
Primary Products 25474 12544 670 4569 936 593865 638058
Where s is the country of Food Products 9422 9977 2055 4554 116 289570 315694
interest, d is the set of all Textiles and Apparel 844 107040 16284 15103 2661 300870 442802
countries in the world, i is Other Manufactures 25831 315877 141038 29271 1070 3984273 4497360
the sector of interest, x is Total 61571 445438 160047 53497 4783 5168578 5893914
the commodity export flow
and X is the total export Total Exports by Category for Various Countries (2002)
flow. In words, it is the
share of good i in the total Sample calculation: Suppose we want to calculate the share of textiles and apparel (including
exports of country s. leather) in the total exports of Sri Lanka. The total Sri Lankan exports of textiles and apparel are
highlighted in blue. Total Sri Lankan exports are highlighted in red. The sectoral export share is
Data sources: The United therefore $2661/$4783×100=55.6 per cent.
Nations Commodity Trade
database (COMTRADE), Notes: Sectoral export shares of this type form the basic building block of many of the other
the World Trade Database indices in this group. It is also possible to form sectoral import and sectoral trade shares.
(WTD) maintained by
Statistics Canada, and the
GTAP database from
Purdue University.
59
Sectoral Hirschmann
What does it tell us? The sectoral
Hirschmann index is a measure of the sectoral ROW
concentration of a region’s exports. It tells us Viet Nam
the degree to which a region or country’s Thailand
exports are dispersed across different Sri Lanka
Singapore
economic activities. High concentration levels
Philippines
are sometimes interpreted as an indication of New Zealand
vulnerability to economic changes in a small Malaysia
number of product markets. Over time, Rep. of Korea
decreases in the index may be used to indicate Indonesia
broadening of the export base. An alternative India
measure is the export diversification index. China
Bangladesh
Australia
Definition: The sectoral Hirschmann index is
defined as the square root of the sum of the 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
squared shares of exports of each industry in
total exports for the region under study.
Sectoral Hirschmann Index for Various Countries (2002)
Range of values: Takes a value between 0
and 1.10 Higher values indicate that exports Example: Suppose we wish to compare the degree of export diversification
are concentrated in fewer sectors. across of group of economies at some point in time. The Hirschmann index
results are presented above. In this sample, Australia, New Zealand,
Limitations: The Hirschmann index is Indonesia, India and China are relatively diversified. By contrast, the
subject to an aggregation bias. Philippines, Singapore, Malaysia and Bangladesh are not.
10
Strictly, the lower bound is 1/n where n is the number of countries to which the economy under study exports. This will be close to zero in most
applications.
60
Sectoral Hirschmann – Technical Notes
Mathematical definition:
Rep. of
Australia China India Sri Lanka ROW Total
⎛ ∑ x isd
2
⎞ Korea
∑i ⎜⎜ d X ⎟ Primary Products 25474 12544 670 4569 936 593865 638058
⎝ ∑d sd
⎟
⎠ Food Products 9422 9977 2055 4554 116 289570 315694
Textiles and Apparel 844 107040 16284 15103 2661 300870 442802
Other Manufactures 25831 315877 141038 29271 1070 3984273 4497360
Where s is the country of
Total 61571 445438 160047 53497 4783 5168578 5893914
interest, d is the set of all
countries in the world, i is
the sectors of interest, x is Total Exports by Category for Various Countries (2002)
the commodity export flow
and X is the total export Sample calculation: The sectoral exports of Australia are in 2002 are highlighted in blue. Total
flow. Each of the bracketed exports are highlighted in red. Calculating the export shares (as a fraction) we have: 0.41, 0.15,
0.01, and 0.42 (in row order). Squaring each term and summing yields 0.37. Taking the square
terms is the share of good i
root we have 0.6. This is the sectoral Hirschmann for Australia in 2002. Note the overstatement
in the exports of country s
of the index when calculated on aggregate data relative to the more dissaggregated data used in
(see major export
the calculations underlying the figure on the previous page.
category).
Notes: A Hirschmann index can also be calculated using import or trade shares, if one is
Data sources: The United
interested in import or overall trade dependence issues. The Hirschmann index is sometimes
Nations Commodity Trade
called the Hirschmann-Herfindahl index (HHI), and is used in many other contexts (see the
database (COMTRADE),
regional version earlier in this volume). It is also calculated in several variants. It may be seen
the World Trade Database
without the final square root operation, or using percentages instead of fractions. It may also be
(WTD) maintained by
normalized using the number of destinations. The latter adjustment turns the index into a
Statistics Canada, and the
measure of the ‘evenness’ in the sectoral export share pattern.
GTAP database from
Purdue University.
61
Export Diversification
What does it tell us? The export
diversification index is another measure of the
sectoral concentration of a region’s exports. It ROW
tells us the degree to which a region or Viet Nam
Thailand
country’s exports are dispersed across different
Sri Lanka
economic activities. Unlike the Hirschmann Singapore
index, it normalizes the export diversification Philippines
pattern by comparing it to the world average. New Zealand
Malaysia
Definition: The sum of the absolute value of Rep. of Korea
the difference between the export category Indonesia
shares of the country under study and the world India
as a whole, divided by two. China
Bangladesh
Australia
Range of values: Values range from 0 to 1. A
value of zero indicates that the export pattern 0.0 0.2 0.4 0.6 0.8 1.0
exactly matches the world average. Higher
values indicate greater dependence on a small
number of products. Export Diversification Index for Various Countries (2002)
Limitations: This measure is subject to an Example: We calculate the index for the same group of countries as for the
aggregation bias and should be calculated on sectoral Hirschmann. The results are somewhat different to the sectoral
disaggregated data. An aggregate measure Hirschmann (e.g., by this measure Australia and New Zealand appear less
cannot tell us which commodities dominate the diversified). The reason is that this index adjusts for what is ‘normal’ for the
export profile, for that we need to go back to world as a whole.
the individual shares.
62
Export Diversification – Technical Notes
Mathematical definition:
Rep. of
Australia China India Sri Lanka ROW Total
Korea
⎛
⎜ ∑d x isd − ∑wd x iwd ⎞
⎟÷2
∑
Primary Products 25474 12544 670 4569 936 593865 638058
⎜ ∑ X sd ∑ X wd
i ⎟ Food Products 9422 9977 2055 4554 116 289570 315694
⎝ d wd ⎠ Textiles and Apparel 844 107040 16284 15103 2661 300870 442802
Other Manufactures 25831 315877 141038 29271 1070 3984273 4497360
Where s is the country of Total 61571 445438 160047 53497 4783 5168578 5893914
interest, d and w are the set of
all countries in the world, i is Total Exports by Category for Various Countries (2002)
the sector of interest, x is the
commodity export flow and XSample calculation: We use the same example as before for comparison purposes. The
is the total export flow. sectoral exports of Australia are in 2002 are highlighted in blue. Total exports are highlighted
in red. Calculating the export shares (as a fraction) we have: 0.41, 0.15, 0.01, and 0.42 (in
Data sources: The United row order). World total exports by sector are highlighted in green, and world total exports in
Nations Commodity Trade purple. The shares are (in order): 0.11, 0.05, 0.08, and 0.76. We take the absolute value of the
database (COMTRADE), the difference between each pair. So, for primary products we have 0.41-0.11=0.3. Summing all
World Trade Database (WTD) the terms and dividing by two we have 0.4. This is the export diversification index for
maintained by Statistics Australia in 2002. Again, note the bias in the statistic when calculated on aggregate data
Canada, and the GTAP relative to the more dissaggregated data used in the calculations underlying the figure on the
database from Purdue previous page.
University.
63
Revealed Comparative Advantage (RCA)
What does it tell us? Comparative advantage
underlies economists’ explanations for the
observed pattern of inter-industry trade. In ROW
theoretical models, comparative advantage is Viet Nam
expressed in terms of relative prices evaluated in Thailand
the absence of trade. Since these are not observed, Sri Lanka
in practice we measure comparative advantage Singapore
indirectly. Revealed comparative advantage Philippines
New Zealand
indices (RCA) use the trade pattern to identify the
Malaysia
sectors in which an economy has a comparative Rep. of Korea
advantage, by comparing the country of interests’ Indonesia
trade profile with the world average. India
China
Definition: The RCA index is defined as the ratio Bangladesh
of two shares. The numerator is the share of a Australia
country’s total exports of the commodity of 0 1 2 3 4 5 6 7 8 9
interest in its total exports. The denominator is
share of world exports of the same commodity in
total world exports. RCA Index for Wheat (2002)
Range of values: Takes a value between 0 and Example: Suppose we are interested in the wheat market, and need to
+∞. A country is said to have a revealed determine which economies have a comparative advantage in wheat. The
comparative advantage if the value exceeds unity. RCA calculation results are presented above for 2002. We are looking for
values exceeding unity. In this sample, Australia, India and the rest of world
Limitations: The index is affected by anything are revealed to have a comparative advantage in wheat.
that distorts the trade pattern, e.g., trade barriers.
64
RCA – Technical Notes
Mathematical definition: Wheat Australia New Zealand China Rep. of Korea ASEAN ROW Total
Australia 0 4 3 168 441 777 1393
∑x ∑X
d isd d sd New Zealand 0 0 0 0 7 0 7
∑ x ∑ X
wd iwd wd wd
China
Rep. of Korea
0
0
0
0
0
0
54
0
18
1
2
0
74
1
ASEAN 0 0 0 0 6 3 9
Where s is the country of
ROW 5 3 150 339 960 12836 14293
interest, d and w are the set
Total 5 7 153 561 1433 13618 15777
of all countries in the world,
i is the sector of interest, x Total Australia New Zealand China Rep. of Korea ASEAN ROW Total
is the commodity export Australia 0 3348 4878 5213 7070 41062 61571
flow and X is the total New Zealand 2564 0 777 659 1187 9633 14820
export flow. The numerator China 6058 867 0 16243 23301 398969 445438
is the share of good i in the Rep. of Korea 2516 301 24232 0 17521 115477 160047
exports of country s, while
ASEAN 9850 1130 24832 15950 80320 253262 385344
the denominator is the share
ROW 43045 7169 156819 103738 174227 4341696 4762074
of good i in the exports of
Total 64033 12815 211538 141803 303626 5160099 5893914
the world.
Sample calculation: Above we have two simplified trade matrices, representing trade flows for
Data sources: The United
wheat and total trade flows, respectively. Total exports of wheat from Australia are highlighted in
Nations Commodity Trade
blue. Total exports from Australia are highlighted in red. Calculating the share of Australia’s
database (COMTRADE),
wheat exports to its total trade we have: $1393/$61571. World trade in wheat is highlighted in
the World Trade Database
green, and total world trade in purple. The proportion of wheat in world trade is:
(WTD) maintained by
$15777/$5893914. Taking the ratio of these two shares we have 8.45. This is the RCA index for
Statistics Canada, and the
Australia in wheat in 2002.
GTAP database.
65
Additive RCA
What does it tell us? The additive RCA
(ARCA) index is an alternative to the RCA
ROW
index. Again, it is used to identify the sectors
Viet Nam
in which an economy has a comparative Thailand
advantage, and to track changes over time. Sri Lanka
Unlike the RCA index, it is symmetric (around Singapore
zero). Philippines
New Zealand
Definition: The ARCA index is defined as the Malaysia
difference of two shares: The share of a Rep. of Korea
Indonesia
country’s total exports of the commodity of
India
interest in its total exports and the share of China
world exports of the same commodity in total Bangladesh
world exports. Australia
Range of values: Takes a value between −1 -0.01 0.00 0.01 0.02 0.03
and +1. A country is said to have a revealed
comparative advantage if the value exceeds
zero. Additive RCA Index for Wheat (2002)
Limitations: As with RCA, the index is Example: We use the same example as above for comparison purposes. The
affected by anything that distorts the trade results of the ARCA calculations for wheat are presented above. Now we are
pattern, e.g., trade barriers. It does not identify looking for values exceeding zero. Again, in this sample, Australia, India
the source of comparative advantage. and the rest of world are revealed to have a comparative advantage in wheat.
66
Additive RCA – Technical Notes
Mathematical definition: Rep. of
Wheat Australia New Zealand China
Korea
ASEAN ROW Total
∑x d isd
−
∑ wd
x iwd Australia
New Zealand
0
0
4
0
3
0
168
0
441
7
777
0
1393
7
∑X d sd ∑
wd
X wd China 0 0 0 54 18 2 74
Rep. of Korea 0 0 0 0 1 0 1
Where s is the country of ASEAN 0 0 0 0 6 3 9
interest, d and w are the set ROW 5 3 150 339 960 12836 14293
of all countries in the world, i Total 5 7 153 561 1433 13618 15777
is the sector of interest, x is
Rep. of
the commodity export flow Total Australia New Zealand China
Korea
ASEAN ROW Total
and X is the total export Australia 0 3348 4878 5213 7070 41062 61571
flow. The first term is the New Zealand 2564 0 777 659 1187 9633 14820
share of good i in the exports China 6058 867 0 16243 23301 398969 445438
of country s, while the Rep. of Korea 2516 301 24232 0 17521 115477 160047
second term is the share of ASEAN 9850 1130 24832 15950 80320 253262 385344
good i in the exports of the
ROW 43045 7169 156819 103738 174227 4341696 4762074
world.
Total 64033 12815 211538 141803 303626 5160099 5893914
67
Michelaye
What does it tell us? The Michelaye index is a
second alternative to the RCA index. Again, it is
used to identify the sectors in which an economy ROW
has a comparative advantage. Like ARCA, it is Viet Nam
symmetric around zero. The difference between Thailand
Michelaye and ARCA is that the former compares Sri Lanka
the export pattern of the country under study to Singapore
that export pattern of the world, while the latter Philippines
compares the export pattern of the country under New Zealand
Malaysia
study to its own import pattern.
Rep. of Korea
Indonesia
Definition: The Michelaye index is defined as the India
difference of two shares: The share of a country’s China
total exports of the commodity of interest in its Bangladesh
total exports and the share of the same country’s Australia
imports of the same commodity in its total
-0.03 -0.02 -0.01 0.00 0.01 0.02 0.03
imports.
Range of values: Takes a value between −1 and Michelaye Index for Wheat (2002)
+1. A country is said to have a revealed
comparative advantage if the value exceeds zero. Example: Again we use the same example for comparison purposes. The
results are presented above. As with ARCA we are looking for values
Limitations: As with RCA and ARCA, the index exceeding zero. Again, i this sample, Australia, India and the ROW are
is affected by anything that distorts the trade revealed to have a comparative advantage in wheat.
pattern, e.g., trade barriers. It does not identify the
source of comparative advantage.
68
Michelaye – Technical Notes
Mathematical definition:
Rep. of
Wheat Australia New Zealand China
Korea
ASEAN ROW Total
∑w x isw −
∑w miws Australia 0 4 3 168 441 777 1393
total export flow, m the Australia 0 3348 4878 5213 7070 41062 61571
New Zealand 2564 0 777 659 1187 9633 14820
commodity import flow, and M
China 6058 867 0 16243 23301 398969 445438
the total import flow. The first
Rep. of Korea 2516 301 24232 0 17521 115477 160047
term is the share of good i in the ASEAN 9850 1130 24832 15950 80320 253262 385344
exports of country s, while the ROW 43045 7169 156819 103738 174227 4341696 4762074
second term is the share of good Total 64033 12815 211538 141803 303626 5160099 5893914
i in the imports of country s.
Sample calculation: Above we have the same two simplified trade matrices. Total
Data sources: The United exports of wheat from Australia are highlighted in blue. Total exports from Australia are
Nations Commodity Trade highlighted in red. Calculating the share of Australia’s wheat exports to its total trade we
database (COMTRADE), the have: $1393/$61571. Total imports of wheat to Australia are highlighted in green, and
World Trade Database (WTD) total imports to Australia in purple. The proportion of Australia’s wheat imports to its total
maintained by Statistics Canada, imports is: $5/$64033. Taking the difference between these two shares we have 0.022.
and the GTAP database from This is the Michelaye index for Australia in wheat in 2002.
Purdue University
69
Regional Orientation
What does it tell us? The regional
orientation index tells us whether exports 2002 1992
Sugar
of a particular product from the region
under study to a given destination are Processed Rice Sugar
greater than exports of the same product to Fisheries Processed Rice
other destinations. In other words, it
Other Crops Fisheries
measures the importance of intra-regional
exports relative to extra-regional exports. Other Crops
Other Minerals
Other Minerals
Definition: The index is the ratio of two Plant Fibers
Plant Fibers
shares. The numerator is the share of a Oil Seeds Oil Seeds
country’s exports of a given product to the
region of interest in total exports to the Other Grains Other Grains
region. The denominator is the share of
0 5 10 15 20 0 10 20 30
exports of the product to other countries in
total exports to other countries.
RCA Index for Thailand (2002) RO Index for Thailand/ASEAN
Range of values: Takes a value between 0
Example: One application of the regional orientation index is the identification
and +∞. A value greater than unity implies
of possible cases of trade diversion. In the figure on the right we have calculated
a regional bias in exports.
the regional orientation of selected Thailand exports to ASEAN in 1992 (prior to
AFTA) and in 2002. We see a strong and increasing bias in several categories,
Limitations: The index may be affected
notably other grains and plant fibers. On the left, we have the RCA index. In
by many factors, including geographical
processed rice, fisheries and sugar, Thailand has a comparative advantage. On
ones. Because it is based on relative
the other hand it has no advantage in other grains or plant fibers. This pattern is
shares, a strong regional orientation may
suggestive of trade diversion in those categories.
be of little economic significance.
70
Regional Orientation – Technical Notes
Mathematical definition:
Rep. of
Wheat Australia New Zealand China
Korea
ASEAN ROW Total
∑ w
x isw ∑ w
X sw
New Zealand
China
0
0
0
0
0
0
0
54
7
18
0
2
7
74
Rep. of Korea 0 0 0 0 1 0 1
Where s is the country of interest, ASEAN 0 0 0 0 6 3 9
d is the set of countries in the ROW 5 3 150 339 960 12836 14293
regional bloc, w is the set of all Total 5 7 153 561 1433 13618 15777
71
Complementarity
What does it tell us? The complementarity
index is a type of overlap index. It measures the 70
degree to which the export pattern of one country
matches the import pattern of another. A high 60
degree of complementarity is assumed to indicate
more favorable prospects for a successful trade 50
arrangement. Changes over time may tell us 40
whether the trade profiles are becoming more or
India
less compatible. 30
Sri Lanka
Definition: The sum of the absolute value of the 20
difference between the import category shares
10
and the export shares of the countries under
study, divided by two. The index is converted to 0
percentage form.
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
Range of values: Takes a value between 0 and
100, with zero indicating no overlap and 100
indicating a perfect match in the import/export Complementarity Indices between India/Sri Lanka (1992-2002)
pattern.
Example: Suppose we want to know whether or not the exports of India
Limitations: High complementarity indices may and Sri Lanka have grown more complementary over time. We can
be misleading if the countries are geographically construct the complementarity index over several time periods and look
distant, or if the size difference in the economies for changes. The above figure presents the results. The exports of India
is large (i.e., a match in percentage terms does are quite complementary with the imports of Sri Lanka, but the converse
not imply a match in levels). Aggregation bias. is not true, although the degree has been rising over time.
72
Complementarity – Technical Notes
Mathematical definition:
India Sri Lanka ROW Total
Primary Products 4569 936 632553 638058
⎛ ⎛ ∑ m iwd ∑ x isw ⎞ ⎞
⎜1 − ⎜ ⎟ ÷ 2 ⎟ × 100
⎜ ⎜ ∑i ∑ M wd ∑ X sw
Food Products 4554 116 311024 315694
w
− w
⎟ ⎟ Textiles and Apparel 15103 2661 425038 442802
⎝ ⎝ w w ⎠ ⎠ Other Manufactures 29271 1070 4467019 4497360
Total 53497 4783 5835634 5893914
Where d is the importing country of interest, s is the
exporting country of interest, w is the set of all Exports by Category for Selected Economies (2002)
countries in the world, i is the set of industries, x is
the commodity export flow, X is the total export India Sri Lanka ROW Total
flow, m the commodity import flow, and M the total Primary Products 6057 729 631272 638058
import flow. In words, we take the sum of the Food Products 2203 513 312978 315694
Textiles and Apparel 1282 1178 440342 442802
absolute value of the difference between the sectoral
Other Manufactures 35814 2995 4458551 4497360
import shares of one country and the sectoral export
Total 45356 5415 5843143 5893914
shares of the other. Dividing by 2 coverts this to a
number between 0 and 1, with zero indicating all
Imports by Category for Selected Economies (2002)
shares matched and 1 indicating none did. Subtracting
from one reverses the sign, and multiplying by 100 Sample calculation: Start with the share of primary products in Sri
puts the measure in percentage terms. Lanka’s imports. This is $729/5415=0.134. India’s export share in
this category is $4569/53497=0.085. Taking the absolute value of the
Data sources: The United Nations Commodity Trade difference we have 0.049. Repeating this for the other categories and
database (COMTRADE), the World Trade Database summing the results we have 0.13. Halving and subtracting from one
(WTD) maintained by Statistics Canada, and the we have 0.935, or 93.5 per cent. This is India’s complementarity with
GTAP database from Purdue University. Sri Lanka in 2002. Note the aggregation bias.
73
Export Similarity
What does it tell us? The export similarity
index is another overlap index. It is designed 80
to measure the degree of similarity between
the export profiles of two economies. The 70
more similar the export profiles are, the more 60
likely that economies are competitors in global
markets. High similarity indices may also 50
indicate limited potential for inter-industry 40
trade with a regional trading arrangement.
30
Definition: The export similarity index is the 20
sum over export categories of the smaller of
the sectoral export share (as a percentage) of 10
each country under study. 0
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
Range of values: Takes a value between 0 and
100 per cent. A value of zero indicates no
overlap in the export profiles (the countries are
not competitors), a value of 100 indicates Export Similarity Indices between India/China (1992-2002)
perfect overlap.
Example: China and India are two major developing economies. To what
Limitations: Does not consider the level of extent are they competitors on the world market? We construct the export
exports, only the structure, so may be similarity index for these two economies. The results are shown above. The
misleading when the size of the economies export profiles of these two economies remain quite similar, but have been
considered is very different. Subject to diverging over time. In large part this reflects major shift by China into the
aggregation bias. electronics and machinery categories, a move not matched by India.
74
Export Similarity – Technical Notes
Mathematical definition:
Rep. of
China India Sri Lanka ROW Total
⎛ ∑ x isw ∑ x idw ⎞ Korea
∑i min⎜⎜ w X , w X ⎟⎟ ×100 Primary Products 12544 670 4569 936 619339 638058
⎝ ∑w sw ∑w dw ⎠ Food Products 9977 2055 4554 116 298992 315694
Textiles and Apparel 107040 16284 15103 2661 301714 442802
Other Manufactures 315877 141038 29271 1070 4010104 4497360
Where d and s are the countries of
Total 445438 160047 53497 4783 5230149 5893914
interest, w is the set of all countries
in the world, i is the set of industries,
x is the commodity export flow, and Exports by Category for Selected Economies (2002)
X is the total export flow. In words,
we take the smaller of the sectoral Sample calculation: Exports by sector from China are highlighted in blue, while total
export share (as a percentage) in exports are in red. Exports by sector for China are highlighed in green, with the total in
each product category, and add them purple. We start by calculating the export share for each category. In primary products
together. we have $12544/445438=0.028. For India we have $4569/53497=0.085. The former is
smaller, so we retain that value (multiplied by 100) and move to the next category.
Data sources: The United Nations Summing across all categories we obtain: 83.8, the export similarity index between
Commodity Trade database China and India.
(COMTRADE), the World Trade
Notes: The index is biased upward as the commodity data is aggregated (note the
Database (WTD) maintained by
values in the figure on the previous page, which were calculated from dissaggregated
Statistics Canada, and the GTAP
data). The export similarity index may also be calculated using an alternative formula
database from Purdue University.
similar to the complementarity index.
75
Sectoral IIT
What does it tell us? The sectoral intra-
industry trade (IIT) is a measure of the degree to
which trade in a particular sector represents Other Grains
intra-industry trade (based on scale economies Wheat
and/or market structure). By engaging in IIT, a Wearing Apparel
country can reduce the number of similar goods Other Food
it produces, and benefit from scale economies. Sugar
Higher IIT ratios suggest that these sources of Textiles
gains are being exploited. May also indicate that Electronic Equipment
Other Transport Equipment
adjustment costs would be lower with trade
Beverages and Tobacco
expansion. Paper Products
Other Machinery
Definition: One minus the ratio of the absolute Iron and Steel
value of exports in a given product category less Chemical, Rubber, Plastics
imports in the same category to the sum of Motor Vehicles
exports and imports in the category.
0.0 0.2 0.4 0.6 0.8 1.0
76
Sectoral IIT – Technical Notes
Mathematical definition:
India Sri Lanka ROW Total
Primary Products 4569 936 632553 638058
1−
∑ d
x isd − ∑d m ids Food Products 4554 116 311024 315694
∑ x isd + ∑d m ids
Textiles and Apparel 15103 2661 425038 442802
d Other Manufactures 29271 1070 4467019 4497360
Total 53497 4783 5835634 5893914
Where s is the country of interest,
d is the set of all other countries in Exports by Category for Selected Economies (2002)
the world, i is the sector of interest,
x is the commodity export flow, India Sri Lanka ROW Total
and m the commodity import flow. Primary Products 6057 729 631272 638058
In the ratio, the numerator is the Food Products 2203 513 312978 315694
Textiles and Apparel 1282 1178 440342 442802
absolute value of the difference
Other Manufactures 35814 2995 4458551 4497360
between total exports and total
Total 45356 5415 5843143 5893914
imports in sector i, the
denominator is the sum of the total
Imports by Category for Selected Economies (2002)
exports and imports in sector i.
Sample calculation: In the textiles and apparel category, India’s exports and $15103,
Data sources: The United Nations and its imports are $1282. Hence the numerator is the absolute value of $15103-
Commodity Trade database 1282=$13821.The sum of imports and exports in this category is $15103+1282=$16385.
(COMTRADE), the World Trade Taking the ratio we have $13821/16385=0.84. The result should lie between 0 and 1.
Database (WTD) maintained by Substracting from 1 yields the IIT index: 1-0.84=0.16. IIT is relatively low in this sector.
Statistics Canada, and the GTAP Note again the potential for aggregation bias.
database from Purdue University.
77
Aggregate IIT
What does it tell us? The aggregate
intra-industry trade index provides us ROW
with an overall measure of the relative Viet Nam
importance of intra-industry trade in an Thailand
economy’s trade profile. As with sectoral Sri Lanka
Singapore
IIT, higher ratios suggest that the
Philippines
economies of scale and variety sources of New Zealand
gains are being exploited. Alternatives Malaysia
are the trade overlap index, and the Rep. of Korea
complementarity index (when applied Indonesia
to a single country). India
China
Definition: The trade weighted average Bangladesh
Australia
of the sectoral IIT indices.
0.0 0.2 0.4 0.6 0.8 1.0
Range of values: The index ranges from
0 to 1, with zero indicating pure inter-
industry trade, and one indicating pure Aggregate IIT Index for Selected Economies (2002)
intra-industry trade.
Example: The above figure compares the aggregate IIT indices in selected
Limitations: As for sectoral IIT, the economies in 2002. While the ‘new’ trade theory that was first used to motivate
index is subject to aggregation bias trade flows between developed economies, IIT levels are high in many
(toward unity), both in terms of sectors developing economies also, notably Malaysia and Thailand. On the other hand,
and regions. It is not appropriate for IIT is relatively low in developed economies like New Zealand and Australia,
measuring changes in intra-industry trade which have a comparative advantage in primary products. Note the high index for
(the marginal IIT index should be used). the rest of world category, a product of aggregation bias.
78
Aggregate IIT – Technical Notes
India Sri Lanka ROW Total
Primary Products 4569 936 632553 638058
Mathematical definition:
Food Products 4554 116 311024 315694
Textiles and Apparel 15103 2661 425038 442802
⎛⎛ ∑ x isd − ∑d m ids ⎞⎟ ∑ x isd + ∑ m ids ⎞
∑ ⎜⎜⎜ ⎜⎜1 − ⎟
d d d
Other Manufactures 29271 1070 4467019 4497360
⎟⎟
i
⎝⎝ ∑
d
x isd + ∑d
m ids
⎟ ∑ X sd + ∑ M ds
⎠ d d ⎠
Total 53497 4783 5835634 5893914
79
Marginal IIT
What does it tell us? Evaluating changes
in IIT over time needs to be done Beverages and Tobacco
carefully, because the IIT index can Electronic Equipment
remain constant even as the volume of Other Food
intra-industry trade expands. The Wearing Apparel
Other Grains
marginal IIT index tells us how much of
Wheat
the change in the volume of trade Other Transport Equipment
between two periods is intra-industry Chemical, Rubber, Plastics
trade. Sugar
Iron and Steel
Definition: One minus the ratio of the Paper Products
absolute value of the change in exports in Other Machinery
a given product category less the change Textiles
Motor Vehicles
imports in the same category to the sum
of the absolute values of the changes in 0.0 0.2 0.4 0.6 0.8 1.0
exports and imports in the category.
Range of values: The index ranges from Marginal IIT Index for Selected Sectors in India (2001-2002)
0 to 1, with zero indicating pure inter-
industry trade, and one indicating pure Example: In the sectoral IIT index example we considered how much of India’s
intra-industry trade. trade in various sectors can be classified as intra-industry. Suppose we want to
know how the pattern of IIT is changing. We calculate the marginal IIT index for
Limitations: As with the other IIT the same set of sectors between 2001 and 2002. This tells us how much of the
indices, the index is subject to new trade in that period was intra-industry. The figure above shows the results.
aggregation bias (toward unity), both in While some 70 per cent of the new trade between 2001 and 2002 in motor
terms of sectors and regions. vehicles was intra-industry, none of the trade in electronic equipment was.
80
Marginal IIT – Technical Notes
Mathematical definition:
1−
∑ Δx
d isd − ∑d Δm ids
∑ Δx
d isd + ∑ d
Δm ids Exports Imports
2001 2002 2001 2002
Where s is the country of interest, w is the set of Primary Products 4280 4569 5945 6057
all countries in the world, i is the sector of Food Products 3666 4554 1982 2203
interest, x is the commodity export flow, and m Textiles and Apparel 14491 15103 1172 1282
the commodity import flow, and Δ is the change. Other Manufactures 25335 29271 29922 35814
In words, the numerator of the ratio is the Total 47772 53497 39021 45356
absolute value of the change in exports in
category i less the change in imports in the same. International Trade by Category for India in 2001, 2002
The denominator is the sum of the absolute
values of the changes in export and imports. The Sample calculation: The table above presents imports and exports by
ratio lies between 0 and 1. Subtracting from one aggregate sector for India in the consecutive years 2001 and 2002. Total
gives us a more natural interpretation, with trade in textiles and apparel expanded by $722 (million) over the period
increasing values indicating more intra-industry ($612 in exports, in blue, and $110 in imports, in red). Hence the ratio
trade. component of the marginal IIT index for this sector is ($612-
$110)/($612+110)=0.7. Subtracting from one we have 0.3. In other words
Data sources: The United Nations Commodity most of the trade expansion was inter-industry rather than intra-industry.
Trade database (COMTRADE), the World Trade
Database (WTD) maintained by Statistics
Canada, and the GTAP database from Purdue
University.
81
Aggregate Marginal IIT
What does it tell us? The
aggregate marginal intra-industry
trade index is the marginal ROW
Viet Nam
equivalent of the aggregate IIT
Thailand
index. Gives us an overall measure Sri Lanka
of how much of the trade expansion Singapore
between two periods was intra- Philippines
industry, and how much was intra- New Zealand
industry. Malaysia
Rep. of Korea
Definition: The marginal trade Indonesia
India
share weighted average of the
China
marginal IIT indices for each sector Bangladesh
of the economy. Australia
Range of values: The index ranges 0.0 0.2 0.4 0.6 0.8 1.0
from 0 to 1, with zero indicating
pure marginal inter-industry trade,
and one indicating pure marginal Aggregate Marginal IIT Index for Selected Economies (2001-2002)
intra-industry trade.
Example: The above figure depicts the overall significance of intra-industry trade in
Limitations: As with the other IIT the new trade between 2001 and 2002 for various economies. Read in conjunction with
indices, the index is subject to the aggregate IIT indices, the results suggest that IIT is of growing importance in the
aggregation bias (toward unity), trade of several countries, including India and Australia. Nonetheless, most of the new
both in terms of sectors and regions. trade over the period was inter-industry in the selected economies.
82
Aggregate Marginal IIT – Technical Notes
Mathematical definition:
Exports Imports
2001 2002 2001 2002
⎛⎛ ∑ Δx − ∑d Δm ids ⎞⎟ ∑ Δx + ∑ Δm ⎞
∑i ⎜⎜⎜ ⎜⎜1 − ⎟
isd isd ids
d d d Primary Products 4280 4569 5945 6057
+ ∑d Δm ids ⎟ ⎟⎟
⎝⎝ ∑ Δx
d isd ⎠ ∑ ΔX
d isd + ∑ ΔX
d ids
⎠
Food Products 3666 4554 1982 2203
Textiles and Apparel 14491 15103 1172 1282
Other Manufactures 25335 29271 29922 35814
Total 47772 53497 39021 45356
Where s is the country of interest, d is the set of all other
countries in the world, i is the sector of interest, x is the
commodity export flow, m the commodity import flow, International Trade by Category for India in 2001, 2002
X the total export flow, M the total import flow, and Δ
the change operator. Again, the expression is less Sample calculation: In the previous example we calculated the
complicated than it appears. In words, we take the marginal IIT index for textiles and apparel to be 0.30. The other
marginal IIT index for each sector and multiply it by the indices (in order) are 0.56, 0.40 and 0.80. Next we need to
share of that sector in the marginal trade of country s calculate the weights. The absolute values of the changes in trade
(i.e., the total new trade between the periods considered). by category (in order) are: $401, $1109, $722 and $9828, which
We then sum the weighted IIT indices across all total to $12060. The weights are therefore: 0.03, 0.09, 0.06 and
commodities. 0.81. Multiplying each marginal IIT index by the appropriate
weight we obtain the aggregate marginal IIT index for India of
Data sources: The United Nations Commodity Trade 0.73. Note again the bias caused by the high level of aggregation
database (COMTRADE), the World Trade Database used in the simplified example.
(WTD) maintained by Statistics Canada, and the GTAP
database from Purdue University.
83
Trade Overlap
What does it tell us? The trade overlap
index is an alternative to the aggregate IIT
index. Again, it can tell us about the overall
significance of intra-industry trade relative to Trade Overlap
inter-industry trade in the trade profile of an
economy at a point in time. It is marginally
easier to calculate than the aggregate IIT
Aggregate IIT
index.
84
Trade Overlap – Technical Notes
Mathematical definition:
India Sri Lanka ROW Total
Primary Products 4569 936 632553 638058
2 × ∑i min (∑ x , ∑ m )
d isd d ids
Food Products 4554 116 311024 315694
∑X + ∑d M ds
Textiles and Apparel 15103 2661 425038 442802
d sd Other Manufactures 29271 1070 4467019 4497360
Total 53497 4783 5835634 5893914
Where s is the country of interest, w is
the set of all countries in the world, i is Exports by Category for Selected Economies (2002)
the sector of interest, x is the commodity
export flow, X is the total export flow, m India Sri Lanka ROW Total
the commodity import flow, and M the Primary Products 6057 729 631272 638058
Food Products 2203 513 312978 315694
total import flow. The first term is the
Textiles and Apparel 1282 1178 440342 442802
share of good i in the exports of country
Other Manufactures 35814 2995 4458551 4497360
s, while the second term is the share of
Total 45356 5415 5843143 5893914
good i in the imports of the world. In
words, the numerator measures the
Imports by Category for Selected Economies (2002)
overlap of imports and exports, the
denominator expresses the overlap as a Sample calculation: We use the same case as used in the aggregate IIT index
fraction of total trade. above (i.e., India’s trade in 2002). First we take the smaller of exports/imports for
each category (comparing the elements in blue with the corresponding elements in
Data sources: The United Nations green). These are: $4569, $2203, $1282 and $29271 (in order). The sum is
Commodity Trade database $37325, which we multiply by two to get $74650. This is the numerator. The
(COMTRADE), the World Trade denominator is total trade, the sum of the red and purple cells:
Database (WTD) maintained by Statistics $53497+$45356=$98853. Hence the trade overlap index is $74650/$98853=0.75.
Canada, and the GTAP database from This is almost identical to the aggregate IIT index calculated above.
Purdue University.
85
This page intentionally left blank
VII. Protection
Numerous measures of the degree of openness of a country to trade have been utilized in the literature. David (2004) is an
extensive review, covering some 70 different measures, and discussing much of the debate that has surrounded their use.
Measures of the trade policy stance of regions within the economy, flawed as they may be, clearly have a place in interpreting
the implications of PTAs, not least because theory indicates a critical role for both pre- and post-PTA tariff levels.
There are two broad categories of openness measure, those based on incidence (direct measures of policy instruments), and
those based on outcomes (indirect measures of the degree of distortion using prices or quantities traded). Rose (2004) further
breaks the measures down into seven categories:
1. Openness: This outcome based measure is what most trade economists have in mind when referencing ‘openness’. It is
defined as the ratio of total trade (imports plus exports) to GDP (real or nominal). This measure, discussed in Chapter
III (as the trade dependence index), is readily available and widely understood, but may be limited as a measure of
protection. For example, large economies tend to have low index values, but this can simply reflect the broader
industrial structure supported by such economies rather than actual protection.
2. Adjusted trade flows: These measures, also outcome based, use deviations of actual trade pattern from predicted
norms, most frequently the estimated via the gravity model (see Chapter VIII). These indices therefore capture implicit
protection effects well to the extent that the gravity equation is well-specified, but only at an aggregate level.
3. Price outcome measures: The classic outcome based index measure is based on unit price differentials, which compare
domestic sales prices with landed goods prices. This measure, while theoretically sound, is very data-intensive. As a
consequence it is available for relatively few economies. Other measures based on black market currency premia are
more readily available, but may reflect macroeconomic structure as much as trade distortion.
4. Tariffs: Tariff data is widely available, and incidence indices may be formed as simple averages (which will tend to be
biased upward) and weighted average tariffs (which tend to be biased downward). Sectoral tariff data is also available,
and useful in the case of PTA analysis, where identifying heavily protected sectors is critical. This chapter focuses on
measures that can be derived from tariff information. An obvious limitation is that not all sectors are protected by tariff
barriers.
87
5. NTBs: Incidence based measures of non-tariff barriers (NTBs) are usually based on frequency data. The presence of
NTBs may also be imputed indirectly through the use of adjusted trade flow or price outcome measures.
6. Informal/qualitative measures: These are based on qualitative assessments (e.g., the outward orientation index from
the World Bank, and the Index of Economic Freedom from the Heritage Foundation). These are inherently subjective.
7. Composite indices: Composite indices take other indices from the above categories, and combine them into a single
index. This approach, common in the development context, can be used when various sub-components represent
different mechanisms through which policymakers can affect the degree of openness. They can be constructed in
various ways. The Sachs and Warner (1995) index is a dummy variable that takes a value of 1 when any one of several
different conditions is met. More generally an index can be formed from a weighted average of underlying indices
(appropriately normalized). The weights can be decided on the basis of some prior on relative importance, or using a
method such as principal components (which in essence lets the variance in the data select the weights optimally).
In this chapter we focus our attention rather narrowly on the tariff data, and the basic indices that are constructed from the raw
data. Before turning to the indices, however, we need to set the stage in the same way as we did with the trade flow indices,
by very briefly setting out the sources of tariff data, the terminology used in the datasets, and the notation that we will use to
describe the basic indices.
Like trade flow data, tariff data is available from a variety of sources. By far the most widely used source is the UNCTAD
TRAINS database, which is available through WITS. This database has time series on tariff schedules at the HS 6 digit level
specified at the bilateral level, although there are numerous gaps. Many other databases (e.g., GTAP and MacMaps) are based
in part on the TRAINS data. The WTO also maintains a database of protection, but at a more aggregate level.
The TRAINS database contains both bound and applied tariffs. Bound tariffs are the maximum level of tariff that a country
has agreed to apply. Applied tariffs are the actual tariffs used, which must be less than or equal to the bound tariff. The
applied tariff is distinguished by the MFN tariff, applied to all WTO members, and the effective applied tariff, which includes
88
preferential rates (e.g., available through an RTA, where these are recorded). The effective applied rate should be less than or
equal to the MFN applied rate. The distinction between the classifications is important as it directly impacts on the type of
questions for which the tariff data can provide useful information. For example, tariff reductions are almost always negotiated
under the WTO in terms of the bound rates, not the applied rates. Hence, in cases where the bound rate substantially exceeds
the applied rate (a situation called binding overhang) tariff cuts to the bound tariff may have little effect on the actual applied
tariff. On the other hand, if one is interested in tracking what has happened to actual protection levels applied within a free
trade agreement like AFTA, then it is the effective applied tariff that is of interest. The bound tariffs need not have changed at
all.
The basic data element in the tariff schedule is very similar to the data element in the trade flow data. Consider again the flow
of exports of product i from country s to country d. We label this data element, . From the importing country perspective,
once transportation and insurance have been factored in, this is the value of imports, . A tariff is simply a tax on the
import flow, which we can label . In words, is the percentage (ad valorem) tariff rate applied by country d on imports
of product i from country s. With this notation in hand, we can use summation notation in the same way as before, and
construct aggregate indices of tariff protection. We concentrate on three simple measures:
The (simple) average tariff is a measure of the overall degree of protection in the tariff schedule. It is a useful overall measure,
but it can disguise very high protection levels in some sectors. The weighted average tariff takes into account the importance
of each product in the import profile of the economy in question. It tends to understate the degree of protection because highly
protected sectors tend to be given a low weight. The tariff dispersion index is a simple measure of how much variation there is
in the tariff schedule. All of these measures are automatically calculated in TRAINS, so our calculations are illustrative.
89
Average Tariff
What does it tell us? The (simple) Viet Nam Viet Nam
average tariff tells us how much protection
is applied by an economy or region, on Thailand Thailand
average. Higher values indicate a more
protected economy, lower values a less Malaysia Malaysia
protected economy. In general, lower
protection levels indicate a greater degree Indonesia Indonesia
of integration with the global economy.
Philippines Philippines
Can be calculated for a subset of regions
or products. Singapore Singapore
Range of values: The tariff is defined as a Effective Applied Tariffs in ASEAN Bound Tariffs in ASEAN (2006)
percentage, so the average can range from
0 to +∞ (import ban). Example: The above figures describe the simple average tariff for the ASEAN-6
economies in 2006 (from TRAINS). In the figures we show the difference
Limitations: The simple average tariff between the effective applied rates and the bound rates. We have calculated the
does not adjust for the significance of average across all trading partners, and the average across the other ASEAN-6
different products in the trade profile, so a members. Note how Singapore has a zero average applied tariff, but still
high tariff on an insignificant product may maintains a positive bound tariff. Viet Nam, having only acceded to the WTO in
overstate the degree of protection. Does 2007, had no bound tariffs in 2006. The difference between the effective applied
not provide information on tariff peaks. tariffs to the world and to ASEAN reflect tariff preferences under AFTA.
90
Average Tariff – Technical Notes
Mathematical definition: 790111 Containing by weight 99.99 % or more of zinc 5.50
790112 Containing by weight less than 99.99 % of zinc 5.50
∑ is
t isd 790120
790200
Zinc alloys
Zinc waste and scrap.
5.50
1.00
n×p 790310 Zinc dust 1.00
790390 Other 1.00
Where d is the importing country, s is the set 790400 Zinc bars, rods, profiles and wire. 5.00
of source countries, i is the set of products of 790500 Zinc plates, sheets, strip and foil. 5.00
interest, t is the tariff of interest (e.g., bound 790600 Zinc tubes, pipes and tube or pipe fittings 5.00
or applied) defined as a percentage, n is the 790700 (1996-) Other articles of zinc. 20.00
number of products in the product set, and p is Total 54.50
the number of countries in the source. In Number of Lines 10.00
words, we take the sum of all the tariffs in the Simple Average 5.45
lines of interest, and divide it by the number
of elements in those lines.11 Thailand MFN Tariff Schedule (2006) – HS2002 Category 79
Data sources: The UNCTAD Trade Analysis Sample calculation: In the table above we present a subset of the MFN tariff
and Information System (TRAINS), the ITC schedule of Thailand in 2006 (applied rates), drawn from TRAINS. The
Market Access Maps (MacMaps), the WTO information is on the HS2002 category 79 (zinc and articles thereof). To
Integrated Database (IDB) (applied rates), the obtain the simple average we sum the tariffs (all expressed as percentages) in
WTO Consolidated Tariff Schedule (CTS) the lines of interest, this gives us 54.5. We then divide by the number of
(bound rates), and the GTAP database from lines, in this case 10. The average level of protection is 5.45 per cent,
Purdue University (applied rates). relatively low.
11
The elements of the summation and division must be adjusted accordingly if the group of interest is not the full bilateral tariff schedule. If, for
example, we are interested only in the average MFN tariff, then we sum the MFN tariffs only over product categories for a given economy, and divide
by the number of product categories.
91
Weighted Average Tariff
What does it tell us? Like the (simple)
average, the weight average tariff tells us how
much protection is applied by an economy or Thailand
region, on average. Higher values indicate a
more protected economy, lower values a less Viet Nam
protected economy. The difference is that the
weighted average tariff takes into account the Singapore
volume of imports in each product category.
Philippines
92
Weighted Average Tariff – Technical Notes
Mathematical definition: Weighted
Tariff Imports Weights
Tariffs
m isd 790111 5.50 67730 0.539 2.96
∑ t isd
∑k M kd
is 790112 5.50 2508 0.020 0.11
790120 5.50 32642 0.260 1.43
790200 1.00 374 0.003 0.00
Where d is the importing country, s (k) is 790310 1.00 7436 0.059 0.06
the set of source countries, i is the set of 790390 1.00 686 0.005 0.01
products of interest, t is the tariff of interest 790400 5.00 3167 0.025 0.13
790500 5.00 541 0.004 0.02
(e.g., bound or applied) defined as a 790600 5.00 146 0.001 0.01
percentage, m is the product level imports, 790700 20.00 10533 0.084 1.68
and M is total imports by category. In 54.50 125762 1.000 6.40
words, we take each bilateral tariff and 10.00
multiply it by the share of the
corresponding bilateral import flow in total Thailand MFN Tariff Schedule (2006) – HS2002 Category 79
imports. We then sum the weighted tariffs
across all sources/product categories.12 Sample calculation: We use the same example as above for the simple average.
The value of imports (in thousands of US$) is presented in the second column.
Data sources: The UNCTAD Trade In the third column we have calculated the import shares in this category (as a
Analysis and Information System fraction). Multiplying the elements of the first column by those of the third
(TRAINS), the ITC Market Access Maps gives us the fourth column, the weighted tariffs. Summing the weighted tariffs
(MacMaps), the WTO Integrated Database yields the weighted average, 6.40 per cent. The simple average is the same as
(IDB) (applied rates), the WTO the weighted average if the import shares are all equal (to 1/n where n is the
Consolidated Tariff Schedule (CTS). number of lines under consideration).
12
As with the simple average, the elements of the summation and division must be adjusted accordingly if the group of interest is not the full bilateral
tariff schedule.
93
Tariff Dispersion
What does it tell us? The tariff dispersion
index is a single number that measures 1600
how widely spread out are the tariffs in a 1400
schedule or part thereof. In other words, a
high tariff dispersion index indicates that 1200
there is a lot of variation in the tariff 1000
schedule. Economists generally believe
800
that a uniform tariff (with low dispersion)
is more economically efficient. An 600
alternative measure is to consider the 400
difference between the maximum tariff
and the minimum tariff. 200
0
Definition: The tariff dispersion index is 0 20 40 60 80 100
the standard deviation of the selected tariff
line items.
Frequency of Viet Nam Tariffs (2006)
Range of values: The tariff is defined as a
percentage, so dispersion index is Example: The above graphic represents the frequency of Viet Nam’s MFN
measured in the same units. It can range applied tariff in 2006. The horizontal axis is the tariff value (in per cent), the
from 0 (if there is a uniform tariff) to +∞. vertical axis the frequency with which that value appears in the schedule (in this
case of roughly 5000 tariff lines). The vast majority of products have zero tariffs
Limitations: The measure should be used applied, but a few reach up 100 per cent. The red line marks the simple average –
in conjunction with the average tariff. It around 13 per cent. The green line marks one standard deviation from the
can be distorted by a small number of average. The wider the range of tariff values, the greater the distance between the
exceptional tariffs. red and green lines will be.
94
Tariff Dispersion – Technical Notes
Mathematical definition: Squared
Tariff
Deviation
∑ is
( t isd − t isd ) 2 790111
790112
Containing by weight 99.99 % or more of zinc
Containing by weight less than 99.99 % of zinc
5.50
5.50
0.002
0.002
n×p 790120 Zinc alloys 5.50 0.002
790200 Zinc waste and scrap. 1.00 19.803
790310 Zinc dust 1.00 19.803
Where d is the importing country, s is the set 790390 Other 1.00 19.803
of source countries, i is the set of products of 790400 Zinc bars, rods, profiles and wire. 5.00 0.203
interest, t is the tariff of interest (e.g., bound 790500 Zinc plates, sheets, strip and foil. 5.00 0.203
or applied) defined as a percentage, n is the 790600 Zinc tubes, pipes and tube or pipe fittings 5.00 0.203
number of products in the product set, and p 790700 (1996-) Other articles of zinc. 20.00 211.703
is the number of countries in the source. In Total 54.50 271.725
words, we take the sum of all the tariffs in Number of Lines 10.00
the lines of interest, and divide it by the Simple Average 5.45
number of elements in those lines.13
Thailand MFN Tariff Schedule (2006) – HS2002 Category 79
Data sources: The UNCTAD Trade
Analysis and Information System (TRAINS), Sample calculation: Again consider the Thai tariff schedule in category 79.
the ITC Market Access Maps (MacMaps), We have already calculated the simple average tariff to be 5.45 per cent. In
the WTO Integrated Database (IDB) (applied the second column in the table we have calculated the difference between
rates), the WTO Consolidated Tariff each tariff and the simple average, and squared the value (the squared
Schedule (CTS) (bound rates), and the GTAP deviations). The sum of the squared deviations is 271.725. We divide this by
database from Purdue University (applied the number of lines to get 27.17. Taking the square root of this number yields
rates). the tariff dispersion index, 5.21 per cent.
13
Again, the elements of the summation and division must be adjusted accordingly if the group of interest is not the full bilateral tariff schedule.
95
This page intentionally left blank
VIII. Beyond Indices
While direct examination of trade flows and summary trade indices derived from the underlying flows can yield numerous
insights on the effects of trade policy changes, such as regional trade liberalization, other more advanced methods are also of
interest and warrant careful consideration. Two approaches have proved particularly useful in the international trade policy
literature: the estimation of the gravity equation and simulation of policy changes using partial or general equilibrium models.
These methods are advanced, and a full description is far beyond the scope of this handbook. Nonetheless, it is very likely that
policymakers will encounter the results of studies using these methods, and will need to accurately and critically assess them
in order to make effective use of the results in policymaking. Hence, in this chapter we briefly consider the role that gravity
models and computable general equilibrium simulation methods can play in informing trade policy debates, with a focus on
assisting policymakers in becoming informed “consumers” of the results.
Evaluating the potential impact of trade policy changes ex ante (i.e., before they are actually put in place) is usually
accomplished using numerical simulation techniques. In essence, this just means that a (symbolic) model from economic
theory is built (usually on as a computer program) using specific and numerically defined functional forms instead of abstract
ones, and the model then perturbed in a way that represents the policy in question. The results of the model are numeric as
opposed symbolic. The two most commonly utilized simulation techniques for numerically evaluating trade liberalization
proposals are partial equilibrium (PE) and general equilibrium (GE) modeling. PE refers to an incomplete system where
various ceteris paribus assumptions are in place, generally an analysis of a single sector. PE models cannot account for the
interaction between the sector or sectors under consideration and the rest of the economy. Balanced against this limitation is
simplicity, and the fact that reduced-form specification sometimes makes econometric estimation feasible. PE models are
well-suited to shocks to a single sector that is sufficiently small for any interaction with the rest of the economy to be ignored.
Computable general equilibrium (CGE) models are numerical models based on general equilibrium theory. Their objective to
turn abstract models of general equilibrium theory into a practical tool for policy analysis. A number of features distinguish
97
GE models. They are multi-sectoral, and in many cases multi-regional, and the behavior of economic agents is modeled
explicitly through utility and profit maximizing assumptions. In addition, economy-wide constraints are rigorously enforced.
In other words, the markets in a CGE model are all linked together. Distortions in an economic system will often have
repercussions far beyond the sector in which they occur. By linking markets, CGE techniques are effective at capturing the
relevant feedback and flow-through effects.
CGE models have been widely adopted in the trade policy literature. Recent surveys of their application (including numerous
examples using GTAP) see Scollay and Gilbert (2000), Gilbert and Wahl (2002), Robinson and Thierfelder (2002), Lloyd and
MacLaren (2004) and Hertel and Winters (2005). Part of their popularity can be explained by the significant advantages that
the CGE approach has over other methods. The most important of these are:
• Theoretical consistency.
• The ability to highlight the importance of linkages between sectors.
• The ability to incorporate unique features of an economic system.
• The ability to predict values for many economic variables in the system.
Balanced against these advantages, however, are some important limitations. In particular:
98
Of course, many of these limitations can be alleviated by appropriate modeling and study design. CGE is a highly specialized
area of research, and it is not possible to provide a guide to undertaking the method. Rather, our purpose here is to provide
notes helpful for consumers of CGE studies as opposed to producers. We begin by considering the design of a CGE study, and
outlining key question that a policymaker should ask a provider of CGE modeling services. The figure below outlines the
general process, which begins with a policy question.
Shocks
Data
99
Sample policy questions might be, how would growth in country x affect trade patterns in country y? What would happen in
country x if they lowered tariffs in sector z? Would a regional trading agreement between countries x and y be a good idea,
and in what sense? And so on. In most cases the policy question is external to the modeling process, it comes from various
scenarios under consideration by policymakers.
The next step is selection of an appropriate methodology. This depends critically on the nature of the policy question being
asked. As a broad guide, computable general equilibrium will be an appropriate choice if the following conditions are met:
• Appropriate data/human capital is available for the exercise. This is an obvious precondition for doing any CGE work.
• The policy question involves large changes that are well outside of historical experiences. This suggests the need to
use simulation techniques of some kind.
• The policy question involves multiple countries and/or multiple sectors. For example, the DDA involves many
countries simultaneously liberalizing many different sectors. This suggests that we need general equilibrium rather
than partial equilibrium techniques.
• Or, the policy question involves only one sector directly, but that sector is large enough to have an impact on the
overall economy. For example, for many LDCs the textile industry is so large relative to the overall economy that
general equilibrium may be justified even if the policy scenario involves only that sector.
• Answering the policy question requires detailed information on the economic system and not broad economic
aggregates. For example, if trade flows are the only variable of interest, extrapolation from a gravity model may be
preferred. If, however, information on how sectoral employment patterns will change when a policy is implemented is
needed, CGE may be preferred.
100
Now, suppose that it has been determined that CGE techniques are an appropriate methodological choice for the policy
question at hand. What does an exercise in CGE actually look like? A CGE analysis will consist of three components: data,
the model (structure), and shocks. These three elements combine together to produce the simulation results. We consider each
in turn.
The data used in a CGE model consists of two types of information. The first is structural information on the economic system
under study. This takes the form of ‘flows’ – the values of economic transactions among the various players in the economy,
and ‘distortions’ – the policy instruments and other factors that impact the economy in its current state. The data requirements
are actually fairly substantial. Flow data will typically include information on the flows of primary factors (labor, capital and
so on) into productive activities, the flows of production into consumption activities by households, the government, other
firms and investment, and of course international trade flows. Distortion data will often include production/consumption taxes
and subsidies, and taxes and subsidies on international trade (i.e., tariffs).
The second type of information is ‘parametric’ – it describes the behavior of the players in the economic system. Exactly what
parametric data is required will depend on the model structure, but typically it includes elasticities of substitution in
production, demand elasticities, and so on. Key questions that should be asked about the data underlying a CGE simulation
exercise include:
• What is the base year? The base year is the point in time represented by the flow/distortion data. One of the most
common sources of data is the GTAP database. The current version of this database (version 6) has a base year of
2001. The next version (to be released in 2008) will have a base year of 2004. It is important to note that because of
the immense data requirements of CGE models, it is very common for the base year data to be several years (or more)
old. Moreover, in many cases the underlying data is actually older, and has been adjusted to the new base year
(especially in the case of input-output tables, which may be produced only every ten years or so). This matters if
significant changes in the economy have occurred since the base year.
101
• What adjustments (if any) were made to the base data? In some cases researchers may be able to compensate for older
data by updating some key elements, for example the tariff rates, or by projecting the model forward to a new base
year.
• What are the data gaps? All data is incomplete in some regard, and it is important to know what the gaps are in order
to make sense of the results. For example, the GTAP dataset currently has relatively poor information on services trade
and protection. It is important to know this before contemplating using the data for a study of services (unless the
researcher is willing and able to make appropriate adjustments). Similarly, the input-output data for some countries is
not available, and is instead constructed based on other regions. This should affect how seriously output results for
such regions are taken.
The second component of an exercise in CGE is the model structure, the theory of the model. This is the set of assumptions
about how the economic system in question actually works. Because CGE models are essentially theoretical models, there is
considerable scope for variation in model structure. Certain key elements will be critical for all CGE models, however.
Critical questions include:
• What is the model closure? Closure is the selection of which variables in the model are to be exogenous (given).
There are both micro and macroeconomic elements to the choice. Microeconomic elements relate to the function of
markets: factor specificity, price fixing, rationing, unemployment, etc. Macroeconomic closure relates to the balances
identity (the relationship between savings, investment and the current account), and the choice of which of the
components of this identity are exogenous, or determined by some functional rule. Closures rule issues been discussed
at length in Dewatripont and Michel (1987). The key point is that selecting different closures reflects different
economic realities, and can therefore have a significant impact on the model results. There is no ‘right’ closure,
different choices must be made to match the economic reality on the ground as closely as possible. The key is to
question whether the closure makes sense in context, and how the particular closure chosen impacts the model results.
102
• What are the assumptions regarding market structure? Some models do not maintain the assumption of perfect
competition in all markets. A variety of specifications have been adopted to introduce imperfect competition into CGE
models (see Harris, 1984). This can have a significant effect on model results. Again, there is no right choice, a model
that incorporates imperfect competition may be better suited to some types of questions, but may involve trade-offs
that are not worth the benefits in other circumstances (e.g., in a study of primary agricultural markets, the benefit of
introducing imperfect competition to the automobile sector in a model may not be worth the extra time and uncertainty
over specification involved, time which might be better devoted to getting the protection mechanisms in agriculture
modeled correctly).
• What are the assumptions on the treatment of time? Simulation in most CGE models involves comparative statics.
That is, most models consider the role that changes in exogenous parameters (‘shocks’) have on the allocation of
goods amongst consumers and of resources amongst productive activities, and the consequences for economic
efficiency. These models have no explicit time dimension, instead representing different time frames by altering
microeconomic elements of the closure (e.g., allowing capital to move across activities or not). Some CGE models run
static simulations from a projected future equilibrium. A time element can also be introduced by solving the model
sequentially, updating the capital stock to simulate investment and depreciation, the labor stock to simulate population
growth, and productivity parameters to simulate advances in technology. Such models are known as recursive
dynamic. They generate a base growth path to which the experimental growth path can be compared. In these models,
however, the inter-temporal allocation of goods and resources will not be optimal in general. Truly dynamic models
attempt to overcome this deficiency by explicitly modeling inter-temporal behavior (Manne, 1985, describes the
approaches as ‘myopic’ and ‘clairvoyant’). This substantially increases the computational complexity. Once again, the
best approach will depend on the nature of the questions being asked and the trade-offs involved, a recursive dynamic
model may not be better than a comparative static model at addressing issues of resource allocation, and may
introduce more sources of uncertainty in the growth path, for example. An inter-temporal model may give up
substantial sectoral detail in order to get the consumption/investment behavior time-consistent.
103
The third and final element of a CGE simulation is the choice of ‘shocks.’ A shock is a change is any exogenous value in the
model. Most often, shocks take the form of removing or adding distortions to the economic system (e.g., removing a tariff) or
establishing a growth path through changes in the stocks of productive factors and the level of technology. Other possibilities
include changes in preferences or the terms-of-trade (depending on the model). In general, the shocks are chosen to represent
as fully as possible a proposed policy change underlying the policy question under study. There is really only one key
questions to ask:
• What exactly do the assumed shocks cover and how were they determined? For example, in simulating a preferential
agreement, do the shocks cover all trade or only merchandise trade? Is the simulation covering export subsidies and
domestic support, or tariffs only? Does the study assume that other taxes remain in place or do they adjust to make up
revenue? (Strictly, this is an element of closure). Are tariffs reduced to zero, or do reductions follow a known
schedule? Are sensitive products included? And so on. Basically, the objective is to ascertain whether or not the
shocks chosen are a good representation of the likely policy scenario. Again, we should note that not including some
features (e.g., services) may reflect data or theoretical limitations, and does not necessarily make the results useless,
but it does affect how we interpret them.
These three elements, data, model and shocks, combine to produce the results of a CGE analysis. We emphasize a few critical
points again. CGE is not an appropriate choice for all types of policy questions, and the question should ultimately determine
the methodology used. If CGE methods are chosen, the details of the three elements outlined above should also be carefully
planned to address the policy question, subject to the constraints imposed by data, theory and time. A modeler should be able
to articulate clearly to policymakers exactly what choices were made, why the choices were made, and what the consequences
of those choices are. A modeler should also be able to clearly articulate to policymakers the limitations of the approach taken.
At the same time, it needs to be recognized that all models are inherently limited by their very nature, and different models
will have different limitations. An honest assessment of the limitations of any framework simply helps us to evaluate the
results of a CGE study more objectively.
104
Once the results are obtained, they have to be interpreted. The meaning of the results and the way they are interpreted depends
critically on how the simulation was designed. In a comparative static framework, the results represent the way the economic
system would look (or how much it would change) once all of the adjustments allowed under the chosen closure have taken
place. Changes in economic variables (e.g., output or trade of a particular sector, employment or GDP) will be presented
either in levels (i.e., monetary units) or in percentage changes from the base. Comparative static results are sometimes called a
“one-off change”, but the terminology is a bit misleading. A comparative static model is representing the economic system at
rest, so the results are better interpreted as a “once and for all” change. On the other hand, the results of a dynamic model will
usually be presented either as the path of a particular economic variable of the economic system over time, or the difference
between the value of an economic variable at a particular point in time (e.g., at the end of the simulation period) relative to a
business as usual scenario or baseline.
Since CGE models track large numbers of economic variables (often in the tens or even hundreds of thousands), in principle a
large number of different results can be presented. Some of the more common are:
• Economic welfare. This is an aggregate measure of the benefit of a policy change to the economy. By far the most
commonly used welfare measure is the equivalent variation (EV). This is the monetary value of the increment in
income that would have to be given to (or taken away from) a household at today's prices to make them as well off
today as they would be under the proposed policy change.
• Output changes. These may be at the aggregate level (e.g., nominal or real GDP) or at the sectoral level.
• Trade changes. Again, these may be at the aggregate level or at the sectoral level.
Other measures may also be available, depending on the closure and model structure. For example, many recent exercises
feature disaggregated households, and can provide insights into the effect of reform on poverty. The results presented should
be chosen with the policy goals in mind, but beware of placing undue emphasis on the numbers. Key questions to ask:
105
• Are the results reliable? It is important to understand whether or not a particular result is robust. This is based on the
idea that the underlying parameters of the model are not known with certainty, and therefore the results are not known
with certainty either. Using sensitivity analysis it is usually possible to measure which results are robust to changes in
the parameters and which are not. For example, a model might predict a large welfare gain from a particular policy
reform. However, a moderate change in the underlying parameters (keeping them within a reasonable range) might
generate a predicted welfare loss. Such a result is not robust, and would be a poor basis for policy decisions.
• Are the results economically significant? A result may be robust but of little economic significance. This can be
evaluated by comparing the magnitude of predicted changes to the initial levels. For example, a welfare gain of $100
million may seem large in the context of the economy of Bangladesh, but is miniscule relative to the economy of
Japan. Similarly, a model may predict a 200 per cent rise in imports in a particular sector, but this may be of no
interest if the imports were negligible to begin with. Context is always important when evaluating CGE results.
• Why does this result happen? In some cases a CGE simulation may generate unexpected results. In a sense this is a
good thing, one of the main purposes of modeling using CGE methods is to bring interactions that might not have been
considered to the forefront of policy discussions. Nonetheless, unusual or important results need to be fully explained.
Because a CGE model is a theoretical model it is possible to track down the reasons why certain results occur, and this
is a critical part of the analysis. For example, a simulation exercise generates a welfare loss in a scenario where a
welfare gain was anticipated. Is this because of adverse terms-of-trade movements? Or is it because of a distortion
elsewhere in the system, and if so where? If a good explanation is found, that is useful policy information. If an
important result comes down to a dubious piece of data or shock assumption, that is cause for concern.
• How might the results change under different structural assumptions? It is reasonable to question a modeler on how
the results of their exercise might differ had constraints allowed a different approach. This is part of putting the results
in context and assessing the limitations. Some differences are well-known from surveys of the CGE literature, for
106
• example, CGE models incorporating imperfect competition and/or dynamics tend to predict larger welfare gains,
because they capture more potential sources of gains than static/competitive models.14
• How do the results of this study compare to others? This is perhaps the single most useful question that can be asked,
if other studies exist. If other studies by independent researchers using different models generate similar results, the
confidence in policy recommendations is enhanced. Moreover, if the assumptions of other studies vary, then
comparison can provide useful information on the importance of those assumptions.
Once the results are analyzed it is possible to draw policy conclusions. Is the proposed policy a good one? In what sense?
What issues will we need to be aware of as we implement the policy? Are there any potential problems? How can these be
overcome? How does the proposed policy compare with other possible choices? Properly designed and implemented, and
judiciously interpreted, CGE models can be excellent tools in framing the policy debate.
14
Beware of pushing this too far, however, to push a desired outcome, since the analysis is speculative, if you truly believe that imperfect competition
is going to change the results dramatically in a way that favors a particular outcome, then imperfect competition should be a part of the modeling
framework to begin with (assuming it can be justified in context). Policymakers need to be vigilant about chasing a particular result. Any modeling
process can be manipulated to generate a desired outcome, but to do so violates the entire premise and purpose of the modeling exercise, which is to
inform the policy debate, not justify a pre-determined stance.
107
2. GRAVITY MODELS
The gravity model is perhaps the most widely used econometric model of international trade patterns. An econometric model
uses historical data to try to estimate (and test the robustness of) a hypothesized economic relationship. Once estimated, an
econometric model may also be used to try and extrapolate to cases outside of what has been experienced, i.e., as a predictive
policy device.
The gravity model is based on the idea that the volume of bilateral trade between any pair of countries is an increasing
function of the combined ‘mass’ of the economies, and a decreasing function of distance between the two. While in the past
the approach was viewed as lacking a strong microeconomic foundation, this is no longer the dominant view. The
development of the ‘new’ trade theory helped to provide stronger theoretical foundations for the specification, and it is now
recognized that a reduced form gravity equation can be derived from most models of international trade that incorporate
transportation costs (see Deardorff, 1998). An extensive review of the gravity model in theory and practice is provided by
Anderson and van Wincoop (2004). Again, our objective here is only to provide an overview of the issues faced when
considering the gravity approach. Key questions:
• What are the advantages/disadvantages of a gravity approach? The main advantage is that it is an easy model to
implement (although there is still much refinement at the margin going on in the literature), the data for which is
readily available. Properly specified, it has also proved to be a very reliable model for explaining existing trade
patterns. The main disadvantages are that it may be less useful for predicting changes outside of the range of historical
experience, and that it is only capable of explaining trade flows, not more detailed aspects of an economic system.
• How is the model specified? In order to estimate the gravity model a double logarithmic specification is usually used,
relating the bilateral trade flows of each country pair (the dependent variable) to the product of their GDPs and the
distance between them (the independent variables), plus an error term to capture the random component in the data. In
most applications, additional independent variables are also often included in the model to improve the fit. These may
108
include measures of openness, remoteness, common language or currency, a common border, and of course the
presence or absence of a regional trading agreement. The variables may be continuous (e.g., some measures of
openness) or qualitative measures represented by dummy variables (e.g., a country pair is assigned a 1 if they share a
common language, a 0 otherwise).
• What data is needed? This depends on the exact specification. The primary data requirement is one or more cross-
sections of trade flow data (i.e., a trade matrix) that includes as many economies as possible, including those of
interest. Some studies may use only a single year of data (in effect cross-sectional studies). Others may use multiple
years (i.e., a panel). Trade flow data is available from numerous sources as outlined in the preceding chapter (e.g.,
COMTRADE or DOTS). The other requirements are measures of size (usually GDP but sometimes population, both
available through WDI) and distance (available from various sources, including the World Distance Tables), and
indicators or dummies for any other independent variables used.
• How is the model estimated? As a single equation linear model, the gravity model can be estimated easily using
ordinary least squares (OLS), although other methods (e.g., generalized least squares) may be helpful if the data
exhibits heteroscedasticity (see Frankel, 1997). If the data from which the model is estimated is a panel (i.e., is
composed of both time series and cross-sectional elements) then pooled OLS, fixed effects or random effects models
may also be used (see Rose, 2004).
• How are results interpreted? We can think of the gravity model as estimating a ‘normal’ trade flow, from which
deviations can be investigated. Hence, the approach accomplishes econometrically what indices such as the trade
intensity index attempt to do using simple computation – and in fact we can think of a well-specified gravity model as
a kind of ‘super’ trade intensity index. Results will be presented in a table with numerical estimates of the coefficients,
measures of their statistical robustness, and overall measures of how well the model fits the data (perhaps for several
alternative specifications and estimation methods). In view of the double logarithmic specification, the coefficients on
continuous variables are interpreted as elasticities. In other words, suppose that the estimated coefficient on distance is
-0.1. This means that if the distance between two countries increases by 1 per cent, we expect to see bilateral trade
109
decrease by 0.1 per cent. Coefficients on dummy variables are interpreted differently. Suppose that the estimated
coefficient on a common border is 0.3. This means that having a common border increases trade by (exp(0.3) -1)
×100=35 per cent.
• How can the results be used? The approach also allows us to draw on the theory of statistical inference. Hence,
statistically significant coefficients on a PTA dummy variable may be used to infer the presence of net trade
diversion/or creation (assuming an appropriate panel is available), in much the same way as changes in the trade
intensity index are used. A gravity model can also be used to predict changes in trade flows following a change in
some independent variable. For example, given the characteristics of a particular group and the fitted gravity equation,
it is possible to calculate the predicted trade flows. One can then change the characteristics in a way that represents the
policy proposal (e.g., decrease protection) and see how much the predicted trade flows change.
As with CGE models, properly designed and implemented, and judiciously interpreted, gravity models can be excellent tools
in framing trade policy debates, both by rigorously analyzing current trade patterns and helping to evaluate possible changes.
110
References
Anderson, J.E. and E. van Wincoop (2004). “Trade Costs,” Journal of Economic Literature, vol. 42, pp. 691-751.
David, H.L. (2004) “So You Want to Use a Measure of Openness? A (Sceptics) Guide to Measures of Trade Openness and
Policy,” Department of Economics, Claremont Graduate University, mimeographed.
CESifo, (2007). “ICC Special Question: Statistical information in public and political decision making,” CESifo World
Economic Survey, vol. 6, no 1, February.
Deardorff, A.V. (1998) “Determinants of Bilateral Trade: Does Gravity Work in a Neoclassical World?” in J.A. Frankel (ed.)
The Regionalization of the World Economy. (Chicago, University of Chicago Press).
Dewatripont, M. and G. Michel (1987). “On Closure Rules, Homogeneity and Dynamics in Applied General Equilibrium
Models,” Journal of Development Economics, vol. 26, no. 1, pp. 65-76.
Dimaranan, B.V. (2006). Global Trade, Assistance, and Production: The GTAP 6 Data Base. (West Lafayette, Center for
Global Trade Analysis).
Frankel, J.A. (1997). Regional Trading Blocs in the World Economic System. (Washington D.C., Institute for International
Economics).
Gilbert, J. (2006) “A Framework for Assessing the Role of PTAs in Trade and Development in Asia and the Pacific,” ESCAP,
TID, mimeograph.
Gilbert, J. and T. Wahl (2002). “Applied General Equilibrium Assessments of Trade Liberalisation in China,” World
Economy, vol. 25, no. 5, pp. 697-731.
111
Harris, R.G. (1984). “Applied General Equilibrium Analysis of Small Open Economies with Scale Economies and Imperfect
Competition,” American Economic Review, vol. 74, no. 5, pp. 1016-32.
Hertel, T. and L.A. Winters (2005). “Estimating the Poverty Impacts of a Prospective Doha Development Agenda,” World
Economy, vol. 28, no. 8, pp. 1057-71.
Lloyd, P.J. and D. MacLaren (2004). “Gains and Losses from Regional Trading Agreements: A Survey,” Economic Record,
vol. 80, no. 251, pp. 445-97.
Manne, A.S. (1985). Mathematical Programming Study 23: Economic Equilibrium Model Formulation and Solution.
(Amsterdam, North Holland).
Robinson, S. and K. Thierfelder (2002). “Trade Liberalisation and Regional Integration: The Search for Large Numbers,”
Australian Journal of Agricultural and Resource Economics, vol. 46, no. 4, pp. 585-604.
Rose, A.K. (2004). “Do We Really Know that the WTO Increases Trade?” American Economic Review, vol. 94, no. 1, pp. 98-
114.
Sachs, J. and A. Warner (1995). “Economic Convergence and Economic Policies,” Brookings Papers in Economic Activity,
vol 1, 1-95.
Scollay, R. and J. Gilbert (2000). “Measuring the Gains from APEC Trade Liberalisation: An Overview of CGE
Assessments,” World Economy, vol. 23, no. 2, pp. 175-93.
Scott, C., 2005. “Measuring Up to the Measurement Problem – The Role of Statistics in Evidence-based Policy-making,”
Paris 21, downloaded from the www.worldbank.org.
112