african statistical journal · 2019. 6. 29. · la valeur de la revue a été prouvée par le...

African S

tatistical Journal / Journal statistique africain

FONDS AFRICAIN DE DEVELOPPEMEN

T

AFR

ICAN DEVELOPMENT FUND

BA

NQUE

A

FRICAINE DE DEVELOPPEMENT

© ADB/BAD, 2012 – Statistics Department • Département des statistiquesTemporary Relocation Agency (TRA) – Agence Temporaire de Relocalisation (ATR)13 Rue du GhanaBP. 323, 1002 Tunis Belvédère Tunis, Tunisia / TunisieTel: (+216) 71 102 342 Fax: (+216) 71 103 743 Internet: http://www.afdb.org Email: [email protected]

ISSN : 2233-2820

Volume 14 – M

ay / mai 2012

African Statistical Journal Journal statistique africain

Volume 14 – May / mai 2012African Development Bank Group

Groupe de la Banque africaine de développement2233 2820

Special Issue on the International Comparison Program (ICP)

Volume Spécial sur le Programme de Comparaison Internationale (PCI)

Designations employed in this publication do not imply the expression of any opinion on the part of the African Development Bank or the Editorial Board concerning the legal status of any country or territory, or the delimitation of its frontiers. The African Development Bank accepts no responsibility whatsoever for any consequences of its use.

African Development Bank GroupTemporary Relocation Agency (TRA)

13 rue de GhanaBP 323, 1002 Tunis Belvédère

Tunis, TunisiaTel: (216) 71 102 342

Fax: (+ 216) 71 103 743Email: [email protected]

Internet: http://www.afdb.org

Les dénominations employées dans cette publication n’impliquent, de la part de la Banque africaine de développment ou du comité de rédaction, aucune prise de position quant au statut juridique ou au tracé des frontières des pays. La Banque africaine de développment se dégage de toute responsabilité de l’utilisation qui pourra être faite de ces données.

Groupe de la Banque africaine de développementAgence temporaire de relocalisation (ATR)

13 rue de GhanaBP 323, 1002 Tunis Belvédère

Tunis, TunisieTel: (216) 71 102 342Fax: (216) 71 103 743

Courriel: [email protected]: http://www.afdb.org

@ADB/BAD, 2012 – Statistics Department / Département des statistiques

ISSN: 2233 2820

Design/layout by Phoenix Design Aid A/S, Denmark. ISO 14001/ISO9000 certified and approved CO2 neutral company. Website: www.phoenixdesignaid.dk.

African Statistical JournalJournal statistique africain

Volume 14May / mai 2012

Journal statistique africain, numéro 14, mai 20122

Contents

Editorial ................................................................................................4

Editorial Board .....................................................................................8

Acknowledgments ...............................................................................10

1. Cluster Formation, Data Validation and Outlier Detection in the International Comparisons Program: An Application to the Africa Region Statistics Department, African Development Bank ....................................... 12

2. Comparative analysis of costs of some selected infrastructure com-ponents across Africa: Results from the 2005 International Comparison Program for Africa (ICP-Africa) Oliver Chinganya, Abdoulaye Adam and Marc Kouakou ............................. 60

3. An alternative approach for determining air transportation costs in the International Comparison Program Abdoulaye Adam ....................................................................................... 83

4. Potential scalability of a methodologically harmonized consumer price index in AfricaRees Mpofu and Themba Munalula .......................................................... 105

Call for Papers ..................................................................................... 119

Editorial Policy .................................................................................... 122

Guidelines for Manuscript Submission and Preparation ................ 124

The African Statistical Journal, Volume 14, May 2012 3

Table des matières

Éditorial ................................................................................................6

Le comité de rédaction .........................................................................9

Remerciements ....................................................................................11

1. Formation de grappes, Validation des données et la détection des valeurs aberrantes dans le Programme de comparaison internationale: Une applica-tion à la Région Afrique Statistics Department, African Development Bank ....................................... 12

2. Analyse comparative des coûts de certaines composantes choisies d’infrastructure à travers l’Afrique: résultats du Programme de comparaison international 2005 pour l’Afrique (PCI-Afrique)Oliver Chinganya, Abdoulaye Adam et Marc Kouakou ................................ 60

3. Une approche alternative pour la détermination des couts de transport aérien dans le Programme de comparaison internationale. Abdoulaye Adam ....................................................................................... 83

4. Évolution potentielle d’un indice de prix à la consommation harmonisé méthodologiquement en AfriqueRees Mpofu et Themba Munalula ............................................................. 105

Demande de soumission d’articles..................................................... 120

Ligne éditoriale .................................................................................... 123

Instructions pour la préparation et la soumission de manuscits .... 127


Editorial

We welcome readers to Volume 14 of the African Statistical Journal. The value of the Journal has been proven by the number and quality of papers that continue to be submitted for publication and the comments we receive on the published papers.

This volume is dedicated to the International Comparison Program (ICP) and covers aspects related to the economic statistics that are used in the ICP. The results of the 2005 ICP demonstrated the importance of complete national accounts, particularly, expenditure-based estimates of GDP that cover all the activities in a country’s economy. The major effort in that ICP round was the collection and validation of prices. Two techniques provided a systematic means of editing the prices, but no similar editing system was available to be applied to the national accounts data.

The first paper in this volume stresses the importance of validating national accounts data. The authors note that because of the diversity of regional economies, identifying anomalies in national accounts data in African countries is especially difficult. They propose methods, based on cluster analysis, of grouping countries so that potential errors are more readily apparent. Correcting such errors in the ICP data also has the important spin-off benefit of improving the quality of national accounts time series.

A second paper explains how the results of the 2005 ICP round can be used to compare the costs of a selection of infrastructure components in African countries. Analysts are interested in the estimates of the “real” levels of GDP and GDP per capita of participating countries, all expressed in a common currency. However, another important set of data produced as part of the ICP consists of the price level indexes (PLIs). PLIs are derived by dividing the Purchasing Power Parity (PPP) for an expenditure category by the country’s exchange rate and they are usually expressed in terms of the base country, or sometimes the regional average, being equal to 1.00 or 100. The paper compares PLIs for six categories of infrastructure (housing, water, electric-ity, transport, communication and construction) and analyses variations among price levels in 48 African countries. The conclusions are important for policies related to regional and sub-regional integration in Africa

Analysis of the 2005 ICP results presented in the third paper showed air-fares to be one of the products that was poorly priced but had a significant effect on price comparisons between countries. A problem with airfares is the difficulty of describing flights in such a way that they are comparable across countries. For example, some countries have little or no domestic air sectors but substantial international operations. In addition, domestic flights


Editorial

in large countries may be longer than some international flights commonly flown by residents of other countries. This study systematically identifies potential price-determining characteristics of airfares and proposes a number of changes that would improve the comparability and consistency of airfare data collected for the ICP.

The final paper examines the benefits of harmonizing the consumer price indexes (CPIs) of countries in the Southern African Development Com-munity (SADC) and the Common Market for Eastern and Southern Africa (COMESA). The underlying motivation is the possibility of these regions creating monetary unions and therefore needing comparable CPI data from each member country in order to monitor price stability. The propos-als draw on the experiences of the European Union, which developed its Harmonized Index of Consumer Prices in the late 1990s. Harmonizing CPIs involves a range of methodological, coverage and practical data collection issues, which are described in the paper.

We trust that you will find this volume of the Journal interesting and useful in your day-to-day work. As usual, we welcome comments or debate on the content of the articles published.

Dr. Charles Leyeka LufumpaCo-Chair, Editorial Board

Professor Ben KiregyeraCo-Chair, Editorial Board


Éditorial

Nous nous félicitons de l’intérêt que les lecteurs portent au volume 14 du Journal statistique africain. La valeur de la Revue a été prouvée par le nombre et la qualité des articles qui continuent à être soumis pour publication et les commentaires que nous recevons sur les articles publiés.

Ce volume est consacré au Programme de comparaison internationale (PCI) et couvre les aspects liés aux statistiques économiques qui sont utilisées dans le PCI. Les résultats du PCI 2005 ont démontré l’importance de l’état complet des comptes nationaux, en particulier des estimations du PIB sur la base des dépenses couvrant toutes les activités dans l’économie d’un pays. Le principal effort dans ce cycle du PCI a été la collecte et la validation des données de prix. Deux techniques ont fourni un moyen systématique de faire l’apurement des données de prix, mais il n’ y avait pas de système similaire d’apurement pour les données des comptes nationaux.

Le premier article de ce volume met l’accent sur l’importance de la valida-tion des données des comptes nationaux. Les auteurs notent qu’en raison de la diversité des économies régionales, l’identification des anomalies dans les données des comptes nationaux dans les pays africains est particulière-ment difficile. Ils proposent des méthodes, basées sur l’analyse en grappes, de regrouper les pays de telle sorte que les erreurs potentielles soient plus facilement apparentes. La correction des erreurs dans les données du PCI a aussi l’avantage important d’améliorer la qualité des séries chronologiques des comptes nationaux.

Un deuxième document explique comment les résultats du cycle 2005 du PCI peuvent être utilisés pour comparer les coûts d’une sélection des com-posantes d’infrastructure dans les pays africains. Les analystes sont intéressés par les estimations des niveaux réels du PIB et le PIB par habitant des pays participants, le tout exprimé dans une monnaie commune. Cependant, un autre ensemble important de données produites dans le cadre du PCI comprend les indices de niveau de prix (INP). Les INP sont calculés en divisant la parité de pouvoir d’achat (PPA) pour une catégorie de dépenses par le taux de change du pays et ils sont généralement exprimés en termes du pays de base, ou parfois à la moyenne régionale, étant égal à 1,00 ou 100. Le document compare les INP pour six catégories d’infrastructures (logement, eau, électricité, transport, communication et construction) et analyse les variations entre les niveaux de prix dans 48 pays africains. Les conclusions sont importantes pour les politiques liées à l’intégration régio-nale et sous-régionales en Afrique.


Éditorial

L’analyse des résultats du PCI 2005 présentées dans le troisième document a montré que le billet d’avion est l’un des produits dont le prix a été mal déterminé, mais a un effet significatif sur les comparaisons de prix entre pays. Un des problèmes avec les billets d’avion est la difficulté de décrire les vols de telle manière que les tarifs aériens soient comparables entre les pays. Par exemple, certains pays ont peu ou pas de secteurs aériens intéri-eurs, mais d’importantes opérations internationales. En outre, les vols do-mestiques dans les grands pays peuvent durer plus longtemps que certains vols internationaux couramment pris par des résidents d’autres pays. Cette étude identifie systématiquement les caractéristiques potentielles de déter-mination des tarifs aériens et propose un certain nombre de changements qui permettraient d’améliorer la comparabilité et la cohérence des données des tarifs aériens collectées pour le PCI.

Le document final examine les avantages de l’harmonisation des indices des prix à la consommation (IPC) des pays de la Communauté de développement de l’Afrique australe (SADC) et le Marché commun de l’Afrique orientale et australe (COMESA). La motivation sous-jacente est la possibilité pour ces régions de créer des unions monétaires et ont donc besoin de données comparables de l’IPC de chaque pays membre afin de surveiller la stabilité des prix. Les propositions s’appuient sur les expériences de l’Union euro-péenne, qui a développé son indice harmonisé des prix à la consommation dans les années 1990. L’harmonisation des IPC implique toute une série de problèmes méthodologiques, de couverture et des problèmes pratiques de collecte de données, qui sont décrits dans le document.

Nous espérons que vous trouverez cette édition du Journal intéressante et utile dans votre travail quotidien. Comme d’habitude, nous recevrons avec plaisir des commentaires ou des discussions au sujet du contenu des articles publiés.

Dr Charles Leyeka LufumpaVice-président, comité de rédaction

Professeur Ben KiregyeraVice-président, comité de rédaction


Editorial Board

Chairpersons

1. Charles Leyeka Lufumpa, Director, Statistics Department The African Development Bank Group, Tunis, Tunisia2. Ben Kiregyera, International Statistical Consultant, Kampala, Uganda

Editorial Board Members

3. Dimitri Sanga, Director, African Centre for Statistics (ACS), UN Economic Commission for Africa (UNECA), Addis Ababa, Ethiopia

4. Victor Murinde, Director, African Development Institute, The African Development Bank Group, Tunis, Tunisia5. Martin Balépa, Directeur General, Observatoire économique et statistique

d’Afrique subsaharienne (AFRISTAT), Bamako, Mali6. James P.M Ntozi, Professor, Department of Population Studies, Institute

of Statistics and Applied Economics (ISAE), Makerere University, Kampala, Uganda

7. Rosalia Katapa, Professor of Statistics, University of Dar es Salaam, Tanzania

8. Parin Kurji, Head of Biometry Unit, Agriculture Faculty, College of Agriculture & Vet Sciences, University of Nairobi, Kenya

9. Herbert Borbor Kandeh, UNFPA Chief Census Advisor, Gaborone, Botswana

10. Naman Keita, Statistics Department, Food and Agricultural Organization (FAO), Rome, Italy

11. Koffi N’Guessan, Directeur, Ecole Nationale Supérieure de Statistique et d’Economie Appliquée (ENSEA), Abidjan, Cote d’Ivoire

Production Editor

12. Alice Nabalamba, Statistics Department, African Development Bank Group, Tunis, Tunisia


Le comité de rédaction

Coprésidents

1. Charles Leyeka Lufumpa, directeur, Département des statistiques Groupe de la Banque africaine de développement, Tunis, Tunisie2. Ben Kiregyera, consultant international en statistiques, Kampala,

Ouganda

Membres du comité de rédaction

3. Dimitri Sanga, directeur, Centre africain pour la statistique (CAS), Commission économique des Nations Unies pour l’Afrique (CEA), Addis-Abeba, Ethiopie

4. Victor Murinde, directeur, Institut africain de développement, Groupe de la Banque africaine de développement, Tunis, Tunisia5. Martin Balépa, Directeur General, Observatoire économique et statistique

d’Afrique subsaharienne (AFRISTAT), Bamako, Mali6. James P.M Ntozi, professeur, Département des études sur la population,

Institut de statistique et d’économie appliquée (ISAE), Université de Makerere, Kampala, Ouganda

7. Rosalia Katapa, professeur du Département des statistiques, Université de Dar es Salaam, Tanzanie

8. Parin Kurji, chef de l’Unité de biométrie, Faculté d’agronomie, Institut d’agronomie et de sciences vétérinaires, Université de Nairobi, Kenya

9. Herbert Borbor Kandeh, Conseiller en matière de recensement auprès du FNUAP, Gaborone, Botswana

10. Naman Keita, Département des statistiques, Organisation des Nations Unies pour l’alimentation et l’agriculture (FAO), Rome, Italie

11. Koffi N’Guessan, Directeur, Ecole Nationale Supérieure de Statistique et d’Economie Appliquée (ENSEA), Abidjan, Cote d’Ivoire

Directeur de la production

12. Alice Nabalamba, Département des statistiques, Groupe de la Banque africaine de développement, Tunis, Tunisie


Acknowledgments

The Editorial Board would like to express its appreciation to all authors who submitted papers for publication in the journal and to the following reviewers for their expert assistance:

Paul McCarthy International Comparison Program (ICP) Global Office, the World Bank

Rosalia Katapa Professor of Statistics, University of Dar es Salaam, Tanzania

Georges Bresson Professeur, Université Paris II / Sorbonne Uni-versité, Paris, France

Régis Bourbonnais Maitre de Conférences, Université de Paris-Dauphine, Paris, France

Michel Mouyelo-Katoula Global Manager, International Comparison Program, Development Data Group, the World Bank


Remerciements

Le comité de rédaction voudrait exprimer sa gratitude à tous les auteurs qui ont soumis des articles pour la publication dans ce journal et aux personnes suivantes qui composent le comité de lecture des articles pour cette édition:

Paul McCarthy Programme de Comparaison Internationale (PCI) Bureau Mondiale, Banque Mondiale

Rosalia Katapa Professeur en Statistiques, Université de Dar es Salaam, Tanzanie

Georges Bresson Professeur, Université Paris II / Sorbonne Uni-versité, Paris, France

Régis Bourbonnais Maitre de Conférences, Université de Paris-Dauphine

Michel Mouyelo-Katoula Coordonnateur Mondial, Programme de Com-paraison Internationale, Development Data Group, Banque Mondiale


1. Cluster formation, data validation and outlier detection in the International Comparison Program: An application to the Africa region

Robert Hill1

AbstractThe International Comparisons Program (ICP), an undertaking coordinated by the World Bank, compares the purchasing power of currencies and real output of almost all countries in the world. ICP is divided into six regions, each of which is required to provide prices and expenditures for a common list of basic head-ings that sum to GDP. A potential benefit of participation in ICP is that it can help countries improve their price collection and national accounts, because a fundamental part of ICP is validation of the data they supply. This validation process, to the extent that it focuses on differences across countries, can detect errors in the data that countries might otherwise have difficulty finding. A weakness of the current ICP data validation methodology, however, is that it does not pay enough attention to expenditure data. This paper proposes some new methods that are well suited to detecting anomalies in expenditure data. In addition to potentially improving the quality of the overall ICP-Africa comparison, these methods could help participating countries improve the quality of their national accounts, and thus, contribute to the capacity-building process.

Key words: cluster analysis, purchasing power parities, national accounts.

Résumé Le Programme de comparaison internationale (PCI), une initiative coordon-née par la Banque mondiale, compare le pouvoir d’achat des monnaies et la production réelle de presque tous les pays dans le monde. Le PCI est divisé en six régions, dont chacune est tenue de fournir des prix et des dépenses pour une liste commune de positions élémentaires qui constituent le PIB. Un avantage potentiel de la participation au PCI est qu’il peut aider les pays à améliorer leur collecte des prix et des comptes nationaux, car une partie fondamentale du PCI est la validation des données qu’ils fournissent. Ce processus de validation, dans la mesure où il met l’accent sur les différences entre les pays, peut détecter des erreurs dans les données que les pays pourraient autrement avoir de difficultés à trouver. Cependant une faiblesse de la méthode actuelle de validation des don-nées du PCI, est qu’elle ne prête pas suffisamment d’attention aux données de

1 Prof. Robert Hill is an independent consultant in the Statistics Department of the African Development Bank, Temporary Relocation Agency, Tunis, Tunisia. Email: [email protected]. This paper is a shortened version of the AfDB’s review of the 2005 results of the ICP-Africa.


1. Cluster Formation, Data Validation and Outlier Detection in the International Comparisons Program: An Application to the Africa Region

dépenses. Cet article propose quelques nouvelles méthodes qui sont bien adaptés à la détection d’anomalies dans les données des dépenses. En plus de potentielle-ment améliorer la qualité de la comparaison de l’ensemble du PCI-Afrique, ces méthodes pourraient aider les pays participants à améliorer la qualité de leurs comptes nationaux, et donc, contribuer au processus de renforcement des capacités.

Mots clés : analyse en grappes, anomalies, comptes nationaux

1. INTRODUCTION

The International Comparisons Program (ICP), an undertaking coordinated by the World Bank, compares the purchasing power of currencies and real output of almost all countries in the world. Its results underpin the Penn World Table (probably the most widely used data set in economics). The most recent comparison was made in 2005; the next is scheduled for 2011.

ICP is divided up into six regions, each of which is required to provide prices and expenditures for a list of basic headings (129 in ICP 2005) that sum to GDP. This list is common to all regions. A basic heading is the lowest level of aggregation for which expenditure weights are available. The basic heading prices are constructed from region-specific product lists. Each country supplies prices for a sample of products within each basic heading. In ICP 2005, these were aggregated to obtain the basic heading prices using the country-product-dummy (CPD) method (or a variant thereof) in all regions except the OECD-Eurostat region, which used the Jevons-S method (see for example Hill and Hill 2009 or Diewert 2010 for an explanation of these methods). The expenditure weights, for the most part, were obtained from the national accounts.

A potential benefit of participation in ICP is that it can help countries im-prove their price collection and national accounts, because a fundamental part of ICP is validation of the data supplied by the countries. This valida-tion process, to the extent that it focuses on differences across countries, can detect errors in the data that countries might otherwise have difficulty finding. Correction of these errors should improve the quality of the mac-roeconomic statistics in these countries.

The African Development Bank (AfDB) manages the ICP comparison in the Africa region. Its role in ICP, however, extends to serving as a capacity-building platform for price statistics and national accounts in participating countries. Innovations in data validation methodology, therefore, serve the


Statistics Department, African Development Bank

dual AfDB purposes of potentially improving the quality of the ICP-Africa comparisons and simultaneously strengthening the capacity-building platform.

The ICP data validation tool set consists primarily of the Dikhanov Table and Quaranta Table. The Quaranta Table is used only to validate the price quotes within each basic heading. The Dikhanov Table is more versatile and can be used to validate both the price quotes within each basic heading and the prices at higher levels of aggregation, including at the basic heading level. However, neither method is used to validate the expenditure weights at the basic heading level, which is an important weakness in ICP’s existing data validation methodology.

This study shows how the ICP data validation process can be improved using dissimilarity measures and cluster analysis methods. This approach can be used equally well to validate the price or the expenditure data. Based on this approach, by far, the greatest anomalies in the ICP 2005 data for Africa are in the expenditure weights. As noted above, identification of these anomalies could enable participating countries in the Africa region to improve their national accounts, thereby contributing to the capacity-building platform as well as improving the quality of the overall ICP-Africa comparison.

The starting point for this study was the observation that in Africa, and some other regions, identification of anomalous data points is made more difficult by the diversity of countries in the region. In an attempt to amel-iorate this problem, ICP-Africa is divided into subregions. However, these subregions are determined primarily by geographical location rather than economic similarity. An alternative is to use cluster analysis methods to identify more natural economic groupings of countries that can then be more easily compared.

This study considers some ways in which this can be done. The approach begins by defining a dissimilarity measure that can be applied to either the basic heading prices or quantities (which are derived implicitly from the expenditure data) or to a combination of the two. This dissimilarity measure is then used to make bilateral comparisons between all possible pairings of countries in a region. The outcome is a K by K dissimilarity matrix (where K is the number of countries in the region).

This dissimilarity matrix is a useful data validation tool, in that unusually large elements within it signal problems with the underlying data. In fact, the dissimilarity matrices are at least as useful as the resulting dendrograms and clusters for identifying anomalies in the data.



Returning to the problem of cluster formation, the dissimilarity matrix is used as the input into a cluster analysis algorithm that generates a dendro-gram as its output. A dendrogram is a tree-like structure that begins with each country as its own cluster, and then merges in a sequential and hier-archical manner until the countries are all in one large cluster. By choosing appropriate thresholds within the dendrogram, it is possible to obtain a list of country clusters that are exhaustive and non-overlapping.

A number of alternative dissimilarity measure formulas are considered in this report. Dissimilarity measures defined on the price relatives and quan-tity relatives, and the dendrograms derived from them, may be particularly useful for detecting anomalies in the data. Dissimilarity measures defined on per capita GDP, however, may be more useful for constructing stable clusters of economically similar countries that can be used to refine the search for data anomalies.

The new methods proposed in this study complement the existing ICP data validation framework. These methods could prove particularly useful for detecting anomalies in the expenditure data. Applying these methods to the ICP 2005 basic heading data for the Africa region, this study finds that the expenditure data contain far more outliers than the price data. More gener-ally, it is important, therefore, that more attention is paid to validation of the expenditure data in the Africa region (and other regions) in ICP 2011.

2. DIssIMIlARITy MEAsUREs

2.1 Description

Dissimilarity measures are useful for data validation and detecting outliers. They also provide the input for the cluster analysis methods described in section 3. The theory of dissimilarity measures in a price and quantity in-dex context has been developed by Diewert (2001, 2009). A dissimilarity measure, denoted here by djk, is inherently bilateral. In this context, it is used to compare the dissimilarity between the price and/or quantity vec-tors of a pair of countries j and k. Four axioms that djk could be required to satisfy are the following:

A1: djk = djk ; A2: djk ≥ 0; A3: djk = 0 when pkn = λpjn for all n; A4: djk = 0 when qkn = µqjn for all n;



where n in A3 and A4 indexes the basic headings over which the comparison is being made.

If there are K countries in the comparison, a K × K dissimilarity matrix can be calculated with element djk in the jth row and kth column. A1 implies that the dissimilarity matrix is symmetric. A2 implies that each element of the matrix is nonnegative. A3 implies that an element djk = 0 if the Hicks (1946) aggregation condition (i.e., pkn = λpjn for all n) is satisfied. When the Hicks condition is satisfied, all bilateral price index formulas (worth consid-ering) give the same answer. The corresponding quantity indexes can then be derived implicitly via the factor reversal test. Hence, in this case, there is no index number problem (in the sense that the answer obtained does not depend on the choice of bilateral index number formula). Conversely, A4 implies that an element djk = 0 if the Leontief (1936) aggregation con-dition (i.e., qkn = µqjn for all n) is satisfied. When the Leontief condition is satisfied, all bilateral quantity index formulas (worth considering) yield the same answer. Given that corresponding prices indexes can then be derived implicitly via the factor reversal test, again, in this case, there is no index number problem.

More generally, a dissimilarity measure in this context can be interpreted as measuring the confidence we have in our estimated bilateral price and quantity indexes, with a lower dissimilarity score implying greater confi-dence. Maximum confidence is achieved when djk = 0. A direct implication of each of A3 and A4 is that the dissimilarity matrix has zeroes on its lead diagonal. It is unlikely in practice that zero terms will be observed anywhere else in the matrix.

The first dissimilarity measure considered here is defined as follows:

djkPLS = [max(Pjk

P; PjkL) / min(Pjk

P; PjkL)] - 1

where PjkP and Pjk

L are Paasche and Laspeyres price indexes.2 These and their corresponding quantity indexes are defined as follows:

Paasche: PjkP = (Σn=1

N pknqkn)/ (Σn=1N pjnqkn),

QjkP = (Σn=1

N pknqkn)/ (Σn=1N pknqjn); (1)

Laspeyres: PjkL = (Σn=1

N pknqjn)/ (Σn=1N pjnqjn),

QjkP = (Σn=1

N pjnqkn)/ (Σn=1N pjnqjn). (2)

2 The superscript PLS in djkPLS stands for Paasche-Laspeyres spread.



djkPLS could equally well be defined as a ratio of Paasche and Laspeyres

quantity indexes,

since, by construction, PjkP/Pjk

L = QjkP/Qjk

L.

It can be verified that djkPLS satisfies all four axioms. One weakness of djk

PLS, however, is that djk

PLS = 0 is necessary but not sufficient for there to be no index number problem. This is because djk

PLS may attain the value of zero even when neither A3 or A4 is satisfied.

Our second and third dissimilarity measures are defined as follows:

djkP = Σn=1

N{[(sjn + skn)/2][pkn/(PjkFpjn) + (PjkFpjn/pkn) – 2]};

djkQ = Σn=1

N{[(sjn + skn)/2][qkn/(QjkFqjn) + (QjkFqjn/qkn) – 2]}.

The terms sjn and skn denote expenditure shares defined as follows:

sjn = (pjnqjn)/(Σm=1N pjmqjm); skn = (pknqkn)/(Σm=1

N pkmqkm).

By construction, Σn=1N sjn = Σn=1

N skn = 1. Also, PjkF and Qjk

F denote Fisher price and quantity indexes, respectively. These indexes are defined as follows:

Fisher: PjkF = (Pjk

P × PjkL)1/2, Qjk

F = (QjkP × Qjk

L)1/2. (3)

The dissimilarity measures djkP and djk

Q are considered by Diewert (2001, 2009). He refers to them as weighted asymptotically linear indexes of relative dissimilarity. Diewert also considers a number of other relative dissimilar-ity formulas. Focusing on the price index case, these formulas differ in the influence exerted on the overall measure by each of the n = 1,…, N elements

[pkn/(PjkF pjn) + (Pjk

F pjn)/pkn]. djkP and djk

Q are linear indexes in the sense that influence is a linear function of the size of each element. Diewert also considers measures that are concave or convex functions of the size of each



element. Convex functions may be overly sensitive to outlier elements, while concave functions may not be sensitive enough to outliers.

djkP satisfies only axioms A1, A2 and A3, while djk

Q satisfies only A1, A2 and A4. A weakness of both djk

P = 0 and djkQ = 0 is that each is sufficient but

not necessary for there to be no index number problem. More specifically, when the Leontief condition (i.e., qkn = µqjn for all n) is satisfied, in general, djk

P > 0. Conversely, when the Hicks condition (i.e., pkn = λpjn for all n) is satisfied, in general, djk

Q > 0.

Consider an additional axiom A5 defined as follows:

A5: djk = 0 if and only if either pkn = λpjn for all n or qkn = μqjn for all n.

All three dissimilarity measures considered thus far violate A5. Suppose now we combine djk

P and djkQ as follows:

djkPQ = (djk

P × djkQ)1/2.

It can be verified that the dissimilarity measure, djkPQ satisfies all five axi-

oms. That is, it attains the value of zero if and only if either the Hicks or Leontief aggregation conditions are satisfied (and hence there is no index number problem).

Despite the attractive axiomatic properties of djkPQ, from a data validation

perspective, the dissimilarity measures djkP and djk

Q are probably more use-ful. This is because djk

P is probably best for detecting anomalies in the price data, and dQ jk, for detecting anomalies in the quantity (expenditure) data. Given that the price and quantity data are obtained from largely independ-ent sources (the former from ICP surveys and the latter from the national accounts) the holistic approaches provided by djk

PLS and djkPQ may not be as

effective at detecting errors and anomalies in either data set. However, they are useful for detecting inconsistencies between the price and quantity data.



An alternative criterion for cluster formation is differences in per capita GDP. A dissimilarity measure derived from this criterion that has the same fundamental structure as those considered above is the following:

djkGDP = yk/yj + yj/yk - 2; (4)

where yj and yk denote the per capita GDP of countries j and k expressed in units of a common currency. In the empirical analysis that follows, this method is implemented using per capita incomes calculated using both the GEKS and Iklé methods.3 Alternatively, per capita GDP can be replaced by the price level, defined for country j as the ratio of its purchasing power parity Pj to its corresponding market exchange rate MERj - as follows:

djkPLev = (Pk/MERk)/(Pj/MERj) + (Pj/MERj)/(Pk/MERk) – 2. (5)

These dissimilarity measures are simpler than those derived from price or quantity relatives in that they are derived from only a single input from each country (i.e., per capita GDP or price level) that, moreover, is always positive. Hence, no expenditure share weights are needed, or adjustments to allow for negative or zero elements. We turn to this latter issue next.

2.2 The problem of zero or negative quantities

Two problems can arise in the construction of the price relative and quantity relative dissimilarity measures. First, the expenditure estimates for a few basic headings in some countries are zero. This may be because the commodities in that basic heading are not available, as is the case for 1102111 Spirits, 1102121 Wine, and 1102131 Beer in some countries. Alternatively, it may be because expenditure in that heading is not measured in a country, as is the case for 111220 Prostitution and 111261 Financial Services Indirectly Measured (FISIM). In ICP 2005, the quantity data are derived implicitly by dividing expenditure by price for each basic heading in each country. A zero expenditure, therefore, implies a zero quantity. In the presence of zero quantities, neither djk

Q nor djkPQ is defined.

3 The GEKS method was used by all regions except Africa to compute the aggregate level results in ICP 2005. Africa uses Iklé. Descriptions of these methods can be found, for example, in Dikhanov (1997), Hill (1997), and Diewert (2011).



The second problem is that in some cases, expenditures (and hence, quanti-ties) are negative. This can occur for the balancing item basic headings (i.e., 111300 Net purchases abroad, 130225 Receipts from sales, 130425 Receipts from sales, 140115 Receipts from sales, 160000 Change in inventories and valuables, and 180000 Balance of exports and imports). Negative quantity relatives (i.e., qkn/(Qjk

F qjn) < 0) are not allowed in the djkQ formula. This

means that when both QjkF qjn and qkn are negative, there is no problem,

because in this case, the quantity relative will still be positive. Likewise, negative expenditure shares are not allowed in either the djk

P or djkQ formulas.

To avoid negative expenditure shares, the shares used in the djkP and djk

Q are modified as follows:

sjn* = |sjn|/(Σm=1N |sjm|) ; skn* = |skn|/(Σm=1

N |skm|).

The modified djkP formula takes the following form:

djkP = Σn=1

N {[(sjn* + skn*)/2][pkn/(PjkF pjn) + (Pjk

F pjn)/pkn – 2]}. (6)

Further modifications to the djkQ formula are required to allow for the pos-

sibility of zero quantities and negative quantity relatives.

djkQ = Σn=1

N [(sjn* + skn*)/2] djk,nQ, (7)

where

djk,nQ = min[20, qkn/(Qjk

F qjn) + (QjkF qjn)/qkn – 2], for qkn/(Qjk

F qjn) > 0;

djk,nQ = 20 Σn=1

N [(sjn* + skn*)/2], for qkn/(QjkF qjn) < 0; qjn = 0; or qkn = 0.

Here an upper limit of 20 is imposed on each element djk,nQ in (7). This

ensures that djk,nQ is defined when either qjn or qkn equals zero, and that it

is well behaved when qkn/(QjkF qjn) is negative. The inclusion of an upper

limit on djk,nQ also ensures that the dendrograms derived from the quantity

dissimilarity matrices are not oversensitive to outliers among the quantity relatives. In a price context, none of the elements [pkn/(Pjk

F pjn) + (PjkF pjn)/



pkn – 2] exceeds 20 in the ICP 2005 data for Africa. Hence there is no need to impose such an upper limit in (6).

3. PRICE AND qUANTITy RElATIvEs AND DETECTION Of OUTlIERs

Price and quantity relatives are useful diagnostic tools for detecting outliers (and perhaps errors) in the basic heading data. Here, an upper price relative UPRjn for basic heading n in country j is defined as the geometric mean of the price relatives for heading n between country j and each other country in the comparison, with the country j price in the denominator. A lower price relative LPRjn for basic heading n in country j is simply the reciprocal of UPRjn. That is, now the country j price is in the numerator. The upper quantity relative UQRjn and lower quantity relative LQRjn for basic heading n in country j is defined in an analogous way. Again, for quantity relatives, the complication arises that either qjn or qkn may equal zero, or the ratio qkn/qjn may be negative. In such cases, this particular element is set to 1 in the UQRjn and LQRjn formulas. This ensures that UQRjn and LQRjn are both defined, while at the same time not drawing unnecessary attention to these observations. A quantity of zero is clearly problematic and should be checked anyway. Negative quantity relatives are only observed for balancing items. These should also always be carefully scrutinized in each round of ICP.

Lower price relative: UPRjn = Πk=1K [pkn/(Pjk

F pjn)]1/K;

Upper price relative: LPRjn = Πk=1K [(Pjk

F pjn)/pkn]1/K;

Lower quantity relative: UQRjn = Πk=1K [qkn*/(Qjk

F qjn*)]1/K;

Upper quantity relative: LQRjn = Πk=1K [(Qjk

F qjn*)/qkn*]1/K;

where qkn*/qjn* = qkn/qjn when qkn/qjn > 0;

qkn*/qjn* = QjkF otherwise.

Also, PjkF and Qjk

F again denote Fisher price and quantity indexes, respec-tively, as defined in (3).



An unusually large upper price relative UPRjn suggests that pjn may be too high, while an unusually large lower price relative LPRjn suggests that pjn may be too low. For example, suppose UPRjn = 10. This implies that the price of heading n is, on average, 10 times higher in country j than in the other countries in the comparison.

Such a large difference is probably attributable to an error in the calculation of pjn. Similarly, an unusually large upper quantity relative UQRjn suggests that qjn may be too high, while an unusually large lower quantity relative LQRjn suggests that qjn may be too low. Inclusion of bilateral quantity indexes in UQRjn and LQRjn implies that they adjust for differences in the economic size of countries. Hence, when UQRjn = 10, this means that the per capita quantity consumed of heading n in country j is, on average, 10 times larger than in the other countries in the comparison (thus, again strongly suggesting an error in the calculation of qjn).

In the empirical analysis that follows, this methodology is applied to the ICP 2005 data for Africa. The country-basic headings jn with the largest upper and lower price and quantity relatives are identified. A number of these clearly warrant closer scrutiny.

4. DENDROgRAMs AND ClUsTERs

The literature on cluster analysis methods is extensive (for example, Everitt, Lander, Leese and Stahl 2011). Much of this literature assumes that the observations over which the clusters are to be formed can be represented as points in an N dimensional space (which is often Euclidean). Dissimilarity matrices and dendrograms can be constructed for such data sets if desired. However, this is just one of a number of approaches that can be used to construct clusters on such data sets.

Our starting point, by contrast, is a dissimilarity matrix, without any corre-sponding representation in an N dimensional space. This narrows the range of options somewhat. The nature of the data lends itself to an agglomerative approach to cluster formation using dendrograms. Agglomerative methods start with each observation (in this case, a country) as an individual cluster. At each step, the closest pair of clusters is merged until all the observations have been merged into a single cluster. This requires a criterion for measur-ing cluster proximity. Two widely used criteria are the nearest neighbor and group average. The former, at each step, seeks out the pair of observations in different clusters with the smallest dissimilarity measure, and then merges



them. The latter compares all possible pairs of observations in two clusters and then calculates the average of their dissimilarity measures. This average distance ADJK between a pair of clusters J and K, is calculated as follows:

ADJK = (1/NJNK) Σj-1N

J Σk-1N

K djk;

where NJ and NK denote the number of observations in clusters J and K, and j and k index the observations in each cluster. These averages are then compared across all possible pairs of clusters. The pair with the smallest average is then merged. In our context, the group average criterion is more appropriate, since the objective is to identify groups of countries with col-lectively similar relative price and/or quantity structures.

The resulting pattern of cluster formation can be described by a dendro-gram. A dendrogram is a tree-like diagram which shows the order in which the clusters were merged. The lengths of the branches in the dendrogram measure the distance between clusters when they were linked. A simple example consisting of 12 countries from the Africa region in ICP 2005 is provided below in Figure 1. The dissimilarity matrix in this example is computed over the full set of basic headings using the Paasche-Laspeyres spread dissimilarity measure djk

PLS, and the dendrogram is calculated using the group average criterion.

Figure 1. Clustering strategy and dendrogram for 12-country exampleClustering strategy

Cluster 1st item 2nd item Distance

1 Tanzania Kenya 0.011

2 South Africa Nigeria 0.026

3 Cluster 2 Namibia 0.042

4 Cluster 1 Madagascar 0.089

5 Cluster 4 Cameroon 0.120

6 Cluster 3 Morocco 0.136

7 Cluster 6 Tunisia 0.257

8 Cluster 5 Liberia 0.277

9 Cluster 8 Cluster 7 0.354

10 Rwanda Ethiopia 0.376

11 Cluster 10 Cluster 9 0.506



The clustering strategy table in Figure 1 gives the order in the which the clusters are merged, starting from an initial situation in which each coun-try is its own cluster. The table also provides the average distance between elements of a cluster. A list of clusters can be derived from a dendrogram by specifying a critical threshold for the average distance between elements of a cluster. Once this threshold is reached, the algorithm terminates. For example, if the threshold in Figure 1 is set to 0.01, there are 12 clusters (i.e., each country is its own cluster). This is because even if we try to form a cluster between the two countries with the smallest dissimilarity meas-ure—in this case, Tanzania and Kenya—the average distance of 0.011 will surpass the allowed critical threshold of 0.01.

By contrast, if the threshold is set to 0.51, there is only one cluster containing all 12 countries. Clearly these two extreme cases defeat the whole objective of the exercise. For the results to be of any use, it is necessary to obtain at least 2 clusters and fewer than 12. Suppose the average distance threshold is set to 0.25; now we obtain 6 clusters. The constituent countries in each cluster are as follows:

Cluster 1: Rwanda;Cluster 2: Ethiopia;Cluster 3: Tanzania, Kenya, Madagascar, Cameroon;Cluster 4: Liberia;Cluster 5: South Africa, Nigeria, Namibia, Morocco;Cluster 6: Tunisia.

If the threshold is raised to 0.3, the number of clusters falls to 4, with the constituent countries being as follows:

Figure 1. Clustering Strategy and Dendrogram for 12 Country Example

Clustering Strategy

Cluster 1st Item 2nd Item Distance 1 Tanzania Kenya 0.011 2 South Africa Nigeria 0.026 3 Cluster 2 Namibia 0.042 4 Cluster 1 Madagascar 0.089 5 Cluster 4 Cameroon 0.120 6 Cluster 3 Morocco 0.136 7 Cluster 6 Tunisia 0.257 8 Cluster 5 Liberia 0.277 9 Cluster 8 Cluster 7 0.354

10 Rwanda Ethiopia 0.376 11 Cluster 10 Cluster 9 0.506

Tanzania

Kenya

South Africa

Nigeria

Namibia

Madagascar

Cameroon

Morocco

Tunisia

Liberia

Rwanda

Ethiopia

0.00 0.56



Cluster 1: Rwanda;Cluster 2: Ethiopia;Cluster 3: Tanzania, Kenya, Madagascar, Cameroon, Liberia;Cluster 4: South Africa, Nigeria, Namibia, Morocco, Tunisia.

An important implication of this example, therefore, is that a dendrogram itself does not indicate how many clusters there are. The process of cluster formation involves a subjective element, the choice of the critical threshold, which should be decided only after the dendrogram has been constructed.

In an ICP context, a rigid threshold should probably not be applied across the whole dendrogram, because such an approach is too inflexible. Rather, the critical threshold should be allowed to vary in different parts of the dendrogram, subject to the constraint that the resulting set of clusters is non-overlapping and exhaustive (i.e., each country is included in only one cluster).

To see why, again consider Figure 1. The ultimate objective is to obtain a list of clusters that are useful in an ICP context. A problem with applying a rigid threshold to the whole dendrogram is that it may generate too many singleton clusters. Singleton clusters are not helpful if the goal is to identify groups of countries that may usefully be compared as part of the data valida-tion process. Visual inspection of Figure 1 suggests that in an ICP context it might be best to group the countries into 3 clusters:

Cluster 1: Rwanda, Ethiopia;Cluster 2: Tanzania, Kenya, Madagascar, Cameroon, Liberia;Cluster 3: South Africa, Nigeria, Namibia, Morocco, Tunisia.

It is not possible to obtain this arrangement of clusters using a single aver-age distance threshold. It can, however, be obtained if, starting from the top of the dendrogram, the critical threshold for cluster 1 is, say, 0.4, and thereafter, the critical threshold is, say, 0.3.

In the empirical examples that follow, dendrograms are calculated using the full sample of 48 countries participating in Africa in ICP 2005. In some cases, suggested groupings of countries derived from these dendrograms using a flexible group average distance threshold are also provided.



5. AN APPlICATION TO ICP 2005 DATA fOR ThE AfRICA REgION

5.1 The ICP 2005 data set for the Africa region

A total of 48 countries from the Africa region participated in ICP 2005. The ICP 2005 global comparison was made over 129 basic headings. A basic heading is the lowest level of aggregation at which expenditure data are available. A basic heading consists of a group of similar products defined within a general product classification. In the overall global comparison, food and non-alcoholic beverages account for 29 headings, alcoholic bever-ages and tobacco for 4 headings, clothing and footwear for 5 headings, etc. (Blades 2007). The basic heading price indexes, which in ICP 2005 in the Africa region were calculated using the Country-Product-Dummy (CPD) method, together with their corresponding expenditure data, provide the building blocks from which the overall comparison is constructed.

The prices of all headings in ICP 2005 are expressed relative to those in the United States. That is, for all headings, the US price is normalized to 1. The price data have no gaps; that is, a price is available for all 129 headings for every country). Furthermore, all the prices are strictly positive.

The expenditure data for each basic heading in each country are expressed in units of domestic currency. In the Africa region, the expenditure level is often set to zero for 14 of the 129 basic headings. Typically, this is because of a failure of measurement rather than because the products in that head-ing are not actually available. Also, sometimes expenditure is negative. In all but one case, this is because the basic heading in question is a balancing item (e.g., 111300 Net purchases abroad, 130225 Receipts from sales, 130425 Receipts from sales, 140115 Receipts from sales, 160000 Change in inventories and valuables, and 180000 Balance of exports and imports). Expenditure on balancing items can legitimately be negative. The exception is 111261 FISIM, which is negative in Niger even though FISIM is not a balancing item.

As explained above, the presence of basic headings with zero or negative expenditures complicates computation of the dissimilarity measures used in the empirical analysis that follows. The inputs to the dissimilarity measures are prices and quantities. The quantities are obtained by dividing expendi-ture by price. A zero expenditure implies a zero quantity, while a negative expenditure implies a negative quantity and a negative expenditure share.



As explained in section 2.2, the dissimilarity measure formula must be modified to accommodate such cases.

5.2 Outliers identified by the upper and lower price and quantity relatives

The 25 largest outliers among the country-basic headings jn in terms of the upper and lower price relatives UPRjn and LPRjn are shown in Table 1. The largest UPRjn of 6.27 is for the basic heading “Education” for Mauritius. This implies that the price of “Education” is, on average, 6.27 times greater in Mauritius than in the other 47 countries in the comparison. Such a large difference is probably attributable to an error in the calculation of the price of “Education” in Mauritius. Certainly this price, and indeed, all the prices identified in Table 1, should be checked. A good rule of thumb might be that all country-basic heading prices pjn with either upper or lower price relatives UPRjn and LPRjn that exceed 2 require checking.

The lower price relatives LPRjn are generally larger than the upper price relatives UPRjn. This may be because erroneously low prices are less con-spicuous than erroneously high prices, and hence, the former are less likely to be detected during data validation.

Two countries stand out in Table 1. Djibouti appears four times in the list of the 25 largest upper price relatives UPRjn. Even more significantly, three of the five largest lower price relatives LPRjn relate to “compensation of employees” in Chad, and need checking.



Table 1. Extreme Upper and Lower Price Relatives25 Most Extreme Upper Price Relatives

Mauritius 111000 Education 6.27

Sudan 110621 Medical Services 5.93

Botswana 110724 Other services in respect of personal transport equip-ment

5.37

Zimbabwe 110613 Therapeutical appliances and equipment 4.77

Gabon 110915 Repair of audio-visual, photographic and information processing equipment

4.62

Guinea 110623 Paramedical services 4.36

Comoros 1102121 Wine 3.97

Ethiopia 111000 Education 3.95

Djibouti 110622 Dental services 3.86

Egypt 111220 Prostitution 3.81

Zambia 110623 Paramedical services 3.80

Malawi 110723 Maintenance and repair of personal transport equip-ment

3.74

Djibouti 111211 Hairdressing salons and personal grooming establish-ments

3.73

Rwanda 111231 Jewellery, clocks and watches 3.71

Djibouti 111220 Prostitution 3.70

Benin 110613 Therapeutical appliances and equipment 3.66

Cape Verde 110915 Repair of audio-visual, photographic and information processing equipment

3.65

Angola 110622 Dental services 3.61

Liberia 140111 Compensation of employees 3.56

Niger 110960 Package holidays 3.50

Djibouti 110621 Medical Services 3.48

Namibia 110512 Carpets and other floor coverings 3.48

Burundi 1101183 Confectionery, chocolate and ice cream 3.36

Burundi 110512 Carpets and other floor coverings 3.36

Congo, Rep.

110613 Therapeutical appliances and equipment 3.35

25 Most Extreme lower Price Relatives

Chad 140111 Compensation of employees 12.29




Lesotho 110623 Paramedical services 10.50

Ghana 110410 Actual and imputed rentals for housing 7.98


Zimbabwe 1105621 Domestic services 6.31

Mauritius 110810 Postal services 5.39

Burundi 110810 Postal services 5.16

Swaziland 110623 Paramedical services 4.96

Gambia, The

110622 Dental services 4.95

Ethiopia 1105621 Domestic services 4.56

Central African Republic

110623 Paramedical services 4.56

Tunisia 110440 Water supply and miscellaneous services relating to the dwelling

4.55

Guinea 1105621 Domestic services 4.30

Swaziland 110621 Medical Services 4.26

Zambia 110440 Water supply and miscellaneous services relating to the dwelling

4.24

Equatorial Guinea

110623 Paramedical services 4.16

Liberia 110941 Recreational and sporting services 4.14

Morocco 110452 Gas 4.05

Madagascar 111000 Education 4.01


110941 Recreational and sporting services 3.99

Comoros 110613 Therapeutical appliances and equipment 3.92

Zimbabwe 130421 Compensation of employees 3.89

Mauritius 110820 Telephone and telefax equipment 3.80

Egypt 110736 Other purchased transport services 3.79

The 25 largest outliers among the country-basic headings jn in terms of the upper and lower quantity relatives UQRjn and LQRjn are shown in Ta-ble 2. The most striking aspect of Table 2 is how much larger the quantity relatives are than their corresponding price relatives in Table 1. The upper quantity relatives are larger than the upper price relatives by a factor of about 10, and the lower quantity relatives are larger than the lower price



relatives by a factor of about 200. These large quantity relatives cannot be attributed to differences in the economic size of countries, since as noted above, UQRjn and LQRjn include bilateral quantity indexes that adjust for such differences. A value of LQRjn of 1,000 (the ten largest lower quantity relatives in Table 2 exceed this value) implies that the per capita quantity consumed of this heading n is, on average, a 1,000 times lower in country j than in the other countries in the comparison. Except for balancing items (see below), or perhaps exceptionally, for alcoholic beverages such as “Beer,” such results seem implausible.

Table 2. Extreme Upper and Lower Quantity Relatives25 Most Extreme Upper quantity Relatives

Lesotho 110623 Paramedical services 89.28

Malawi 110915 Repair of audio-visual, photographic and informa-tion processing equipment

69.76

Zambia 110935 Veterinary and other services for pets 54.12

Zambia 110915 Repair of audio-visual, photographic and informa-tion processing equipment

50.10

Egypt 1101143 Cheese 44.37

Chad 1101143 Cheese 43.29

Zambia 110941 Recreational and sporting services 36.01

Niger 110933 Gardens and pets 35.27

Cape Verde 110452 Gas 33.41

Kenya 110935 Veterinary and other services for pets 33.20

Malawi 110724 Other services in respect of personal transport equipment

32.08

Ethiopia 1101173 Frozen or preserved vegetables 31.05

South Africa 110921 Major durables for outdoor and indoor recreation 30.08

Zambia 110442 Miscellaneous services relating to the dwelling 29.45

Zambia 110931 Other recreational items and equipment 29.07

Guinea-Bissau

1101151 Butter and margarine 28.75

South Africa 1101182 Jams, marmalades and honey 28.74

Liberia 110533 Repair of household appliances 28.65

Morocco 110452 Gas 23.88

Guinea-Bissau

110915 Repair of audio-visual, photographic and informa-tion processing equipment

23.71



São Tomé and Principe

1101143 Cheese 23.22

Kenya 110915 Repair of audio-visual, photographic and informa-tion processing equipment

23.21

Tunisia 111120 Accommodation services 22.77

Rwanda 1102131 Beer 22.43

Ethiopia 110513 Repair of furniture, furnishings and floor cover-ings

22.40

25 Most Extreme lower quantity Relatives

Ghana 160000 Change in inventories and valuables 76241.77

Congo, Dem. Rep.

110935 Veterinary and other services for pets 2297.18

Senegal 110532 Small electric household appliances 2087.73

Botswana 110921 Major durables for outdoor and indoor recreation 1564.47

Egypt 1102121 Wine 1431.78

Morocco 1101122 Pork 1339.61

Mozambique 111250 Insurance 1288.77

Tanzania 111250 Insurance 1176.71

Ethiopia 1101115 Pasta products 1057.09

Senegal 110531 Major household appliances whether electric or not

1026.43

Sudan 1102131 Beer 863.41

Chad 110611 Pharmaceutical products 801.17

Chad 110453 Other fuels 794.61

Malawi 111270 Other services n.e.c. 725.50

Egypt 1102131 Beer 635.80

Angola 1103111 Clothing materials and accessories 567.49

Congo, Dem. Rep.

110810 Postal services 507.39

Ethiopia 110712 Motor cycles 411.57

Angola 1101183 Confectionery, chocolate and ice cream 402.82

Angola 110512 Carpets and other floor coverings 294.51

Gambia 110551 Major tools and equipment 258.67

Botswana 111262 Other financial services n.e.c 242.11

Niger 1101122 Pork 233.77

Chad 110451 Electricity 208.66

Sudan 1101132 Preserved fish and seafood 197.40



As with the price data, the fact that the lower quantity relatives LQRjn are larger than the upper quantity relatives UQRjn is probably attributable to erroneously low quantities being less conspicuous than erroneously high quantities. From Table 2 it is clear that it is especially important that data validation focus on implausibly low quantities. Ironically, the largest of the lower quantity relatives in Table 2—”Change in inventories and valuables” for Ghana—is not necessarily an error, since balancing items can legitimately equal zero. However, this entry is the only balancing item heading in the list of the 25 largest lower quantity relatives in Table 2. It is highly likely that the other 24 quantities listed in the lower half of Table 2 (with the possible exception of “Beer”) are too low by a factor of at least 100.

For the lower quantity relatives, countries that stand out are Angola and Chad (each of which appears three times in the list of the top 25). For the upper quantity relatives, the country that stands out is Zambia (which ap-pears five times in the list of the top 25).

In summary, Table 2 reveals that the outliers in the quantity data are far more pronounced than in the price data. It is, therefore, important that more attention be focused on the validation of the quantity (i.e., expenditure) data in ICP 2011, particularly, low quantities.

5.3 Dendrograms and clusters for Africa in ICP 2005

Dendrograms are constructed here based on the price relative, quantity relative and Paasche-Laspeyres spread dissimilarity measures discussed in sections 2.1 and 2.2. These dendrograms are calculated over different com-binations of basic headings. The combinations considered are as follows:

All headings (GDP)Group 1 headings (Final consumption expenditure by households)Group 2 headings (Individual consumption expenditure by government)Group 3 headings (Collective consumption expenditure by government)Group 4 headings (Expenditure on gross fixed capital formation)Group 5 headings (Changes in inventories and acquisitions, less disposals of valuables)Group 6 headings (Balance of exports and imports)

An immediate consideration when computing djkP, djk

Q and djkPQ dissimilarity

measures for subcomponents of GDP is whether the Fisher price and quantity indexes, Pjk

F and QjkF, within these formulas should be defined on GDP or

on the same subcomponents of GDP as the dissimilarity measure itself. It



is probably preferable to use price and quantity indexes defined on GDP regardless of which headings the dissimilarity measure is computed over.

In particular, in the limiting scenario where there is only one heading, as is the case in Groups 5 and 6 in ICP 2005, the dissimilarity measures djk

P, djk

Q and djkPQ all collapse to zero. This is because when Pjk

F and QjkF are

calculated over just this one heading, by construction, PjkF = pkn/pjn and

QjkF = qkn/qjn. The same happens to djk

PLS, because, likewise, PjkP = Pjk

L = pkn/pjn in this case. This problem does not arise, however, in Groups 5 and 6 when the price and quantity indexes Pjk

F and QjkF in djk

P and djkQ are always

defined on GDP.

However, another problem arises for the dissimilarity measure djkQ for

Groups 5 and 6. As a result of these headings both being balancing items, many quantities are negative. This limits the usefulness of computing djk

Q (and likewise, djk

PQ). Hence, the only dendrograms considered here for Groups 5 and 6 are calculated using the dissimilarity measure djkP. For the headings in GDP and Groups 1 to 4, however, dendrograms are presented using all four dissimilarity measures (i.e., djk

P, djkQ, djk

PQ and djkPLS). The

results (dendrograms and resulting clusters) for the headings in GDP are shown in Figure 2a. Dendrograms have also been calculated for Groups 1, 2, 3, 4, 5, and 6. To save space, these dendrograms are not presented here.



Country clusters obtained from price relative dissimilarity dendrogram (GDP) in Figure 2aCluster 1

Zimbabwe

Chad

Cluster 2

Rwanda

Burundi

Madagascar

Zambia

Ethiopia

Egypt, Arab Rep.

Cluster 3

Mali

Benin

Niger

Burkina Faso

Cameroon

Uganda

Tanzania

Togo

Kenya


Mozambique

Côte d’Ivoire

Sudan

Lesotho

Botswana

Angola

Senegal

Djibouti

Cluster 4

Sierra Leone

Nigeria

Mauritania

Guinea-Bissau

Ghana

Gambia, The

Guinea

Malawi

Cluster 5

Gabon

Comoros

Congo, Dem. Rep.

Equatorial Guinea

Congo, Rep.


Cluster 6

Liberia

Cluster 7

Tunisia

Mauritius

South Africa

Morocco

Cape Verde

Swaziland

Namibia



Cluster 4

Sierra Leone

Nigeria

Mauritania

Guinea-Bissau

Ghana

Gambia, The

Guinea

Malawi

Cluster 5

Gabon

Comoros

Congo, Dem. Rep.

Equatorial Guinea

Congo, Rep.


Cluster 6

Liberia

Cluster 7

Tunisia

Mauritius

South Africa

Morocco

Cape Verde

Swaziland

Namibia

Figure 2a. Dendrogram for Price Dissimilarity Matrix (GDP)

Niger

Burkina Faso

Mali

Benin

Uganda

Tanzania

Cameroon

Equatorial Guinea

Congo, Rep.

Mauritania

Guinea-Bissau

Botswana

Angola

Swaziland

Namibia

Sierra Leone

Nigeria

Togo

Gabon

Comoros

Rwanda

Burundi

Senegal


Kenya

Congo, Dem. Rep.

South Africa

Morocco

Ghana

Ethiopia

Egypt, Arab Rep.

Madagascar

Djibouti

Mozambique

Côte d'Ivoire

Gambia, The

Sudan

Lesotho


Guinea

Cape Verde

Zambia

Tunisia

Mauritius

Malawi

Liberia

Zimbabwe

Chad

0.00 0.81

Figure 2a. Dendogram for Price Dissimilarity Matrix (GDP)



Figure 2(b) Dendrogram for quantity dissimilarity matrix (GDP)Country clusters obtained from quantity relative dissimilarity dendrogram (gDP) in figure 2b

Cluster 1

Equatorial Guinea

Gabon

Congo, Rep.

Angola

Botswana

Chad

Cluster 2

Zambia

Malawi

Cluster 3

Tanzania


Cluster 4

Benin

Burundi

Togo

Kenya

Niger

Burkina Faso

Swaziland

Mozambique

Uganda

Rwanda

Cluster 5


Cape Verde

Senegal

Madagascar

Mauritania

Cluster 6

Tunisia

Egypt, Arab Rep.

Namibia

Morocco

Mauritius

South Africa

Djibouti

Cluster 7

Zimbabwe

Lesotho

Cluster 8

Sudan

Cluster 9

Sierra Leone

Gambia, The

Ghana

Liberia

Guinea-Bissau

Mali

Guinea

Comoros

Cluster 10

Ethiopia

Cluster 11

Cameroon

Côte d’Ivoire

Nigeria

Congo, Dem. Rep.



Figure 2(b). Dendrogram for quantity dissimilarity matrix (GDP) Equatorial Guinea

Gabon

Benin

Burundi

Namibia

Morocco

Togo


Cape Verde

Mauritius

Uganda

Rwanda

Congo, Rep.

Swaziland

Mozambique

Niger

Burkina Faso

South Africa

Angola

Kenya

Senegal

Sierra Leone

Gambia, The

Tunisia

Egypt, Arab Rep.

Mali

Guinea

Cameroon

Côte d'Ivoire

Madagascar

Mauritania

Ghana

Nigeria

Djibouti

Liberia

Zimbabwe

Lesotho

Tanzania


Comoros

Guinea-Bissau

Botswana

Sudan

Congo, Dem. Rep.

Ethiopia

Zambia

Malawi

Chad

0.0 8.8

Figure 2(b) Dendogram for quantity dissimilarity matrix (GDP)



Country clusters obtained from geometric mean of price relatives and quantity relatives dissimilarity dendrogram (GDP) in Figure 2cCluster 1

Liberia

Côte d’Ivoire

Cluster 2

Niger

Burkina Faso

Mali

Togo

Benin

Cameroon

Tanzania

Madagascar

Burundi

Uganda

Rwanda

Guinea


Cluster 3

Senegal

Mauritania


Guinea-Bissau

Sierra Leone

Ghana

Gambia, The

Cluster 4

Mozambique

Kenya

Lesotho

Sudan

Cluster 5

Congo, Dem. Rep.

Comoros

Cluster 6

Djibouti

Cluster 7

Ethiopia

Egypt, Arab Rep.

Cluster 8

Tunisia

Morocco

Mauritius

Swaziland

Namibia

South Africa

Cape Verde

Cluster 9

Zambia

Cluster 10

Equatorial Guinea

Gabon

Congo, Rep.

Nigeria

Botswana

Angola

Cluster 11

Malawi

Cluster 12

Zimbabwe

Cluster 13

Chad



Cluster 6

Djibouti

Cluster 7

Ethiopia

Egypt, Arab Rep.

Cluster 8

Tunisia

Morocco

Mauritius

Swaziland

Namibia

South Africa

Cape Verde

Cluster 9

Zambia

Cluster 10

Equatorial Guinea

Gabon

Congo, Rep.

Nigeria

Botswana

Angola

Cluster 11

Malawi

Cluster 12

Zimbabwe

Cluster 13

Chad

Figure 2(c) Dendogram for price-quantity dissimilarity matrix (GDP)Figure 2c. Dendrogram for price-quantity dissimilarity matrix (GDP)

Niger

Burkina Faso

Equatorial Guinea

Gabon

Togo

Benin

Uganda

Rwanda

Swaziland

Namibia

Mali

Congo, Rep.

Madagascar

Burundi


Guinea-Bissau

Cameroon

Tunisia

Morocco

Tanzania

Botswana

Angola

Sierra Leone

Ghana

Senegal

Mauritania

South Africa

Mauritius

Mozambique

Kenya

Gambia, The

Lesotho

Guinea


Nigeria

Sudan

Congo, Dem. Rep.

Comoros

Ethiopia

Egypt, Arab Rep.

Cape Verde

Djibouti

Liberia

Côte d'Ivoire

Zambia

Malawi

Zimbabwe

Chad

0.0 2.5




Figure 2d. Dendrogram for Paasche-Laspeyres spread matrix (GDP)Country clusters obtained from Paasche-laspeyres spread dissimilar-ity dendrogram (gDP) in figure 2d

Cluster 1

Tunisia

Mauritius

Ghana

Cape Verde

Senegal

Sudan

Morocco

South Africa

Namibia

Cluster 2

Zambia

Nigeria

Mali

Mozambique

Niger

Burkina Faso

Congo, Dem. Rep.

Madagascar

Benin

Togo

Guinea

Uganda


Cameroon


Cluster 3

Tanzania

Kenya

Lesotho

Mauritania

Egypt, Arab Rep.

Cluster 4

Rwanda

Malawi

Burundi

Cluster 5

Ethiopia

Cluster 6

Swaziland

Cluster 7

Equatorial Guinea

Congo, Rep.

Comoros

Chad

Gabon

Angola

Zimbabwe

Botswana

Djibouti

Cluster 8

Liberia

Côte d’Ivoire

Cluster 9

Sierra Leone

Guinea-Bissau

Gambia, The



Lesotho

Mauritania

Egypt, Arab Rep.

Cluster 4

Rwanda

Malawi

Burundi

Cluster 5

Ethiopia

Cluster 6

Swaziland

Cluster 7

Equatorial Guinea

Congo, Rep.

Comoros

Chad

Gabon

Angola

Zimbabwe

Botswana

Djibouti

Cluster 8

Liberia

Côte d’Ivoire

Cluster 9

Sierra Leone

Guinea-Bissau

Gambia, The

Figure 2d. Dendrogram for Paasche-Laspeyres Spread Matrix (GDP)

Togo

Guinea

Mauitania

Egypt, Arab Rep.

Tanzania

Kenya

Zimbabwe

Botswana

Zambia

Nigeria

Uganda


Lesotho

Madagascar

Benin

Mali

Rwanda

Malawi

Gabon

Angola

South Africa

Namibia

Niger

Burkina Faso

Sudan

Morocco

Ghana

Cape Verde

Congo, Dem. Rep.

Equatorial Guinea

Congo, Rep.

Mozambique

Burundi

Cameroon

Comoros

Senegal

Sierra Leone

Guinea-Bissau

Chad

Tunisia

Mauritius


Gambia, The

Djibouti

Liberia

Côte d'Ivoire

Ethiopia

Swaziland

0.00 0.59

Figure 2d. Dendogram for Paasche-Laspeyres Spread Matrix (GDP)



A comparison of dendrograms across the price relative, quantity relative and Paasche-Laspeyres spread dissimilarity measures and the various groups reveals that they are not robust to changes in the dissimilarity measure formula or data.

Also presented in Figure 2 is a possible configuration of clusters derived from each dendrogram. As noted previously, the choice of clusters for a given dendrogram has a subjective element. Here, configurations have been chosen consisting of 6 to 13 clusters, depending on how the dendrogram is structured. The clusters can be made broader or narrower to suit the needs of users. For example, on the dendrogram for djk

P in Figure 2a, some pos-sible configurations of clusters are listed below:

Three clusters:

Cluster 1: Zimbabwe, Chad;Cluster 2: Rwanda, Burundi, Madagascar, Zambia, Ethiopia, Egypt Arab

Rep., Mali, Benin, Niger, Burkina Faso, Cameroon, Uganda, Tanzania, Togo, Kenya, Central African Republic, Mozambique, Côte d’Ivoire, Sudan, Lesotho, Botswana, Angola, Senegal, Djibouti, Sierra Leone, Nigeria, Mauritania, Guinea-Bissau, Ghana, The Gambia, Guinea, Malawi, Gabon, Comoros, Congo Dem. Rep., Equatorial Guinea, Congo Rep., São Tomé and Principe, Liberia;

Cluster 3: Tunisia, Mauritius, South Africa, Morocco, Cape Verde, Swa-ziland, Namibia.

Five clusters:


Rep.;Cluster 3: Mali, Benin, Niger, Burkina Faso, Cameroon, Uganda, Tanzania,

Togo, Kenya, Central African Republic, Mozambique, Côte d’Ivoire, Sudan, Lesotho, Botswana, Angola, Senegal, Djibouti, Sierra Leone, Nigeria, Mauritania, Guinea-Bissau, Ghana, The Gambia, Guinea, Malawi, Gabon, Comoros, Congo Dem. Rep., Equatorial Guinea, Congo Rep., São Tomé and Principe;

Cluster 4: Liberia;Cluster 5: Tunisia, Mauritius, South Africa, Morocco, Cape Verde, Swa-

ziland, Namibia.



Six clusters:


Rep.;Cluster 3: Mali, Benin, Niger, Burkina Faso, Cameroon, Uganda, Tan-

zania, Togo, Kenya, Central African Republic, Mozambique, Côte d’Ivoire, Sudan, Lesotho, Botswana, Angola, Senegal, Djibouti, Sierra Leone, Nigeria, Mauritania, Guinea-Bissau, Ghana, The Gambia, Guinea, Malawi;

Cluster 4: Gabon, Comoros, Congo Dem. Rep., Equatorial Guinea, Congo Rep., São Tomé and Principe;

Cluster 5: Liberia;Cluster 6: Tunisia, Mauritius, South Africa, Morocco, Cape Verde, Swa-

ziland, Namibia.

Nine clusters:

Cluster 1: Zimbabwe, Chad;Cluster 2: Rwanda, Burundi, Madagascar, Zambia;Cluster 3: Ethiopia, Egypt Arab Rep.;Cluster 4: Mali, Benin, Niger, Burkina Faso, Cameroon, Uganda, Tan-

zania, Togo, Kenya, Central African Republic, Mozambique, Côte d’Ivoire, Sudan, Lesotho;

Cluster 5: Botswana, Angola, Senegal, Djibouti;Cluster 6: Sierra Leone, Nigeria, Mauritania, Guinea-Bissau, Ghana, The

Gambia, Guinea, Malawi;Cluster 7: Gabon, Comoros, Congo Dem. Rep., Equatorial Guinea,

Congo Rep., São Tomé and Principe;Cluster 8: Liberia;Cluster 9: Tunisia, Mauritius, South Africa, Morocco, Cape Verde, Swa-

ziland, Namibia.

Thirteen clusters:

Cluster 1: Zimbabwe, Chad;Cluster 2: Rwanda, Burundi, Madagascar, Zambia;Cluster 3: Ethiopia, Egypt Arab Rep.;Cluster 4: Mali, Benin, Niger, Burkina Faso, Cameroon, Uganda, Tanzania,

Togo, Kenya, Central African Republic;Cluster 5: Mozambique, Côte d’Ivoire;Cluster 6: Sudan, Lesotho;



Cluster 7: Botswana, Angola, Senegal, Djibouti;Cluster 8: Sierra Leone, Nigeria, Mauritania, Guinea-Bissau, Ghana,

Gambia The, Guinea, Malawi;Cluster 9: Gabon, Comoros, Congo Dem. Rep.;Cluster 10: Equatorial Guinea, Congo Rep., São Tomé and Principe;Cluster 11: Liberia;Cluster 12: Tunisia, Mauritius;Cluster 13: South Africa, Morocco, Cape Verde, Swaziland, Namibia.

Some implications for data validation of the dendrograms and clusters identified in Figure 2 are discussed in the next section.

For dissimilarity measures derived from differences in per capita GDP, den-drograms are constructed for the dissimilarity matrix computed using the dissimilarity measure in (4), with per capita GDP computed using GEKS and Iklé. Also considered here is the dendrogram derived from the dissimilarity measure in (5) that measures differences in the price level across countries. These three dendrograms and associated clusters are shown in Figure 3.4

4 The price-level dendrogram in Figure 9(c) is calculated from the official ICP 2005 results, which, as noted above, used the Iklé method in the Africa region. Zimbabwe is excluded because no market exchange rate is available for 2005.



Figure 3a. Dendrogram for per Capita GDP calculated using GEKSCountry clusters obtained from gEKs per capita gDP dissimilarity dendrogram in figure 3a

Cluster 1

Liberia

Burundi

Congo, Dem. Rep.

Cluster 2

Gambia, The


Malawi

Niger

Sierra Leone

Rwanda

Mozambique

Guinea-Bissau

Ethiopia

Zimbabwe

Cluster 3


Lesotho

Chad

Senegal

Mauritania

Côte d’Ivoire

Benin

Kenya

Sudan

Cameroon

Nigeria

Djibouti

Cluster 4

Togo

Guinea

Uganda

Madagascar

Tanzania

Mali

Comoros

Zambia

Ghana

Burkina Faso

Cluster 5

Swaziland

Namibia

Egypt, Arab Rep.

Tunisia

Cape Verde

Angola

Congo, Rep.

Morocco

Cluster 6

Mauritius

Equatorial Guinea

Gabon

Botswana

South Africa



Figure 3a. Dendogram for Per Capita GDP Calculated using GEKSFigure 3a. Dendrogram for Per Capita GDP Calculated using GEKS

Uganda

Madagascar

Gambia, The


Chad

Senegal

Nigeria

Djibouti

Zambia

Ghana

Mali

Comoros

Malawi

Tanzania

Cape Verde

Angola

Sierra Leone

Rwanda

Togo

Guinea

Mauritania


Lesotho

Côte d'Ivoire

Benin

Gabon

Botswana

Guinea-Bissau

Ethiopia

Congo, Rep.

Mauritius

Equatorial Guinea

Kenya

Burkina Faso

Mozambique

Sudan

Cameroon

Swaziland

Namibia

Zimbabwe

Niger

Liberia

Burundi

Morocco

Egypt, Arab Rep.

South Africa

Tunisia

Congo, Dem. Rep.

0.0 6.2



Figure 3b. Dendrogram for per capita GDP calculated using IkleCountry clusters obtained from Ikle per capita gDP dissimilarity dendro-gram in figure 3b

Cluster 1

Zimbabwe

Guinea-Bissau

Niger

Ethiopia

Burundi

Cluster 2

Malawi


Mozambique

Gambia, The

Sierra Leone

Rwanda

Cluster 3

Togo

Guinea

Tanzania

Mali

Uganda

Madagascar

Comoros

Zambia

Burkina Faso

Ghana

Cluster 4

Senegal

Mauritania

Chad

Côte d’Ivoire

Lesotho

Benin

Kenya


Djibouti

Cameroon

Nigeria

Sudan

Cluster 5

Congo, Dem. Rep.

Liberia

Cluster 6

Swaziland

Namibia

Egypt, Arab Rep.

Tunisia

Morocco

Angola

Congo, Rep.

Cape Verde

Cluster 7

South Africa

Mauritius

Equatorial Guinea

Botswana

Gabon



Figure 3b. Dendogram for Per Capita GDP Calculated using IkleFigure 3b. Dendrogram for Per Capita GDP Calculated using Ikle

Uganda

Madagascar

Morocco

Angola

Equatorial Guinea

Botswana

Tanzania

Mali

Senegal

Mauritania

Djibouti

Cameroon

Lesotho

Benin

Congo, Rep.

Mozambique

Gambia, The

Malawi


Sierra Leone

Rwanda

Zambia

Burkina Faso

Kenya

Swaziland

Namibia

Niger

Ethiopia

Chad

Nigeria


Zimbabwe

Guinea-Bissau

Gabon

Comoros

Ghana

Togo

Guinea

Côte d'Ivoire

Egypt, Arab Rep.

Sudan

Burundi

South Africa

Mauritius

Cape Verde

Tunisia

Congo, Dem. Rep.

Liberia

0.0 6.8



Figure 3c. Dendrogram for country price levelsCountry clusters obtained from price level dissimilarity dendrogram in figure 3c

Cluster 1

Sierra Leone

Mauritania

Burkina Faso

Chad

Kenya

Cluster 2

Guinea-Bissau

Ghana

Benin

Niger

Cluster 3

Madagascar

Burundi

Rwanda

Malawi

Guinea

Uganda

Tanzania

Cluster 4

Liberia

Gabon

Mauritius


Senegal

Djibouti

Cameroon

Botswana

Mozambique

Cluster 5

Togo

Mali

Congo, Dem. Rep.

Nigeria

Tunisia

Sudan

Cluster 6

Equatorial Guinea

Côte d’Ivoire

Zambia

Morocco

Angola

Lesotho

Comoros

Swaziland

Congo, Rep.


Cluster 7

South Africa

Namibia

Cape Verde

Cluster 8

Gambia, The

Ethiopia

Egypt, Arab Rep.



Figure 3c. Dendogram for Country Price LevelsFigure 3c. Dendrogram for Country Price Levels

Equatorial Guinea

Côte d'Ivoire

Morocco

Angola

Togo

Mali

Senegal

Djibouti

Rwanda

Malawi

Cameroon

Lesotho

Guinea

Guinea-Bissau

Ghana

Mauritius


Botswana

Zambia

Congo, Dem. Rep.

Sierra Leone

Mauritania

Liberia

Gabon

Chad

Kenya

Tunisia

Sudan

Nigeria

Benin

Swaziland

Congo, Rep.

Uganda

Tanzania

Mozambique

Gambia, The

Ethiopia

Burkina Faso

Madagascar

Burundi


Niger

Comoros

Egypt, Arab Rep.

South Africa

Namibia

Cape Verde

0.00 0.37



When applied to the dissimilarity measure in (4) derived from per capita GDP, the group-average dendrogram begins by ranking the countries by per capita GDP and then divides them into clusters that preserve this order. In

other words, if countries A and B are in a cluster, a country C with a per capita income between that of countries A and B must be in the same cluster.

A comparison of the GEKS and Iklé per capita GDP dendrograms and clusters reveals considerable similarity (the price level dendrogram and clusters, by contrast, are quite different). In particular, cluster 3 under GEKS is identical to cluster 4 under Iklé; cluster 4-GEKS is identical to cluster 3- Iklé; cluster 5-GEKS is identical to cluster 6- Iklé; and cluster 6-GEKS is identical to cluster 7- Iklé. Differences arise only for the lowest per capita GDP countries. Given that the differences between the GEKS and Iklé per capita GDP estimates are nontrivial and alter the ranking of countries substantially (particularly at the lower end), this suggests that dendrograms constructed on per capita GDP attain a certain degree of stability. By implication, the per capita GDP clusters identified using ICP 2005 should still be broadly applicable to ICP 2011 data.

The actual subregions in ICP-Africa are shown in Table 3. These subregions are largely geo-political and do not necessarily make sense from an economic perspective. There is not much overlap between the official ICP-Africa subregions in Table 3 and the clusters in Figures 3a and 3b. The use of the existing ICP-Africa subregions, therefore, may make it harder to validate the ICP data. It would be preferable if these geo-political subregions were replaced in ICP 2011 by clusters derived from a measure of economic similarity such as those in Figures 3a and 3b.

Table 3. Sub-Regional Organization of ICP-AfricaAfRIsTAT group COMEsA group ECOWAs group

Benin Burundi Gambia

Burkina Faso Djibouti Ghana

Cameroun Egypt Liberia

Cap Vert Eritrea Nigeria

Central African Republic Ethiopia Sierra Leone

Chad Kenya

Comoros Madagascar

Congo Rwanda

Cote d’Ivoire Sudan



Congo, Dem. Rep. Uganda

Equatorial Guinea

Gabon

Guinea Bissau

Guinea Conakry

Mali

Mauritania

Niger


Senegal

Togo

sADC group UMA group Other African countries

Angola Libya Somalia

Botswana Morocco

Lesotho Tunisia

Malawi Algeria

Mauritius Mauritania

Mozambique

Namibia

Seychelles

South Africa

Swaziland

Tanzania

Zambia

Zimbabwe

5.4 how cluster analysis can improve data validation

The price and quantity relatives discussed in section 5.2 provide a way of detecting anomalies in the individual basic heading price and quantity data of each country. The dendrograms, by contrast, provide an indication of the reliability of the complete set of headings over which the dendrogram has been constructed. Focusing first on the price data and the dissimilarity measure djk

P, it can be seen from the dendrogram in Figure 2a that Zimbabwe and Chad are clear outliers. This could be either because their price data are of poor quality or because the relative price structures in these two countries are very different from those in the rest of Africa. From a data validation



perspective, the key point is that the ICP regional office for Africa should carefully check the price data of Zimbabwe and Chad for errors.

Some progress can be made in identifying the sources of discrepancies in the price data of Zimbabwe and Chad by looking at the dendrograms generated by djk

P for each of the six constituent groups of GDP. In the dendrogram (not included here) covering only the price data for the headings in Group 1 (i.e. final consumption expenditure by households), Zimbabwe and Chad are no longer outliers. This indicates that the problems lie elsewhere. The dendrogram (also not included) covering the price data for headings in Group 2 (i.e., individual consumption expenditure by government) is revealing in that it now identifies a cluster of four countries with price structures that are very different from those of the other 44 countries in the comparison. The other two countries in this cluster, in addition to Zimbabwe and Chad, are Congo Dem. Rep. and Equatorial Guinea.

The dendrogram (not included) focusing on the price data for headings in Group 3 (i.e., collective consumption expenditure by government) shows that Chad is a huge outlier. This strongly suggests a problem with Chad’s price data for these headings.

The dendrogram (not included) focusing on the price data for headings in Group 4 (i.e., expenditure on gross fixed capital formation) is also striking in that Zimbabwe is the main outlier. Malawi, too, is somewhat of an out-lier. This dendrogram strongly suggests a problem with Zimbabwe’s price data for these headings.

The dendrogram (not included) focusing on the price data in Group 5 (i.e., changes in inventories and acquisitions, less disposals of valuables) does not contain any strong outliers. Finally, Zimbabwe is again a strong outlier in the dendrogram (not included) focusing on the price data in Group 6 (i.e., balance of exports and imports).

In the quantity data at the level of GDP and the dissimilarity measure dQ jk, no clear outliers appear in the dendrogram in Figure 2b. In the dendrogram (not included) focusing on the quantity data for the headings in Group 1 (i.e., final consumption expenditure by households), Malawi, Chad, Ethiopia, Zambia, Zimbabwe and Djibouti are all outliers, although none are extreme outliers.

The dendrogram (not included) focusing on the quantity data for headings in Group 2 (i.e., individual consumption expenditure by government) identifies



Tanzania and Liberia as outliers. In the dendrogram (not included) focusing on the quantity data for headings in Group 3 (i.e., collective consumption expenditure by government), Tanzania and Zambia are the main outliers. In the dendrogram (not included) focusing on the quantity data for head-ings in Group 4 (i.e., expenditure on gross fixed capital formation), Congo. Dem. Rep., Congo Rep., and Chad are the major outliers. The relevant headings for these countries should probably be checked.

It is also useful to compare the magnitude of the final group average link in the dendrograms (i.e., the link that joins all the countries into a single cluster). The value of the group average at which this link occurs in the dendrograms calculated using the djk

P metric defined over GDP, Group 1, 2, 3, 4, 5 and 6 is, respectively, 0.74, 0.51, 3.23, 4.25, 1.17, 0.07, 1.95. From this it can be deduced that the biggest problems with the price data occur in Groups 2 and 3 (i.e., individual and collective consumption expenditure by government), since this is where the group average is highest when all countries are included in a single cluster. Similarly, the value of the group average in the dendrograms calculated using the djk

Q metric defined over GDP, Group 1, 2, 3, 4, 5 and 6 is, respectively, 8.01, 8.01, 13.4, 5.9, and 6.6, suggesting that the biggest problems in the quantity data arise in Group 2 (i.e., individual consumption expenditure by government). The much larger group averages at the final link in the quantity data dendrograms, compared with the price data dendrograms, are striking. This suggests much greater variability in the quantity relatives across countries than in the price relatives.

The dissimilarity measures djkPQ and djk

PLS, while potentially useful for detecting outliers, do not indicate whether the problems lie with the price or quantity data. Their findings can also differ substantially. For example, Figure 2c, which constructs a dendrogram using the dissimilarity measure djk

PQ over GDP, identifies Chad and Zimbabwe as outliers, but Figure 2d, which constructs a dendrogram using the dissimilarity measure djk

PLS over GDP, does not. Zimbabwe and Chad, however, do emerge as strong outliers according to both djk

PQ and djkPLS for Group 2 (i.e., individual consumption

expenditure by government).

An alternative use for dendrograms in data validation is to identify clusters of countries with reasonably similar price relative or quantity relative struc-tures. It may then be easier to detect data errors if prices and quantities are compared across countries belonging to the same cluster. For example, in Figure 2b, cluster 4 consists of Benin, Burundi, Togo, Kenya, Niger, Burkina Faso, Swaziland, Mozambique, Uganda, and Rwanda.



Upper and lower quantity relatives for the 10 countries in this cluster are compared in Table 4. With regard to the upper quantity relatives, Kenya (with eight entries) and Niger (with six entries) in the top half of Table 4 stand out as countries whose quantity (i.e., expenditure) data should be checked. No country stands out in the lower half of Table 4 with regard to the lower quantity relatives. Nevertheless, as in Table 2, the lower quantity relatives are larger than the upper quantity relatives.

Table 4. Extreme Upper and Lower Quantity Relatives for Cluster 4 in Figure 2b

25 Most Extreme Upper quantity Relatives

Niger 110933 Gardens and pets 27,07

Kenya 110533 Repair of household appliances 24,28

Rwanda 1101173 Frozen or preserved vegetables 18,14

Kenya 110512 Carpets and other floor coverings 17,96

Kenya 110915 Repair of audio-visual, photographic and information processing equipment

17,28

Kenya 110921 Major durables for outdoor and indoor recreation 16,03

Togo 110513 Repair of furniture, furnishings and floor coverings 14,24

Kenya 110935 Veterinary and other services for pets 14,05

Togo 1102111 Spirits 13,46

Kenya 111250 Insurance 12,66

Swazi-land

110621 Medical Services 12,50

Mozam-bique

1101173 Frozen or preserved vegetables 12,06

Mozam-bique

1101143 Cheese 11,28

Niger 110921 Major durables for outdoor and indoor recreation 11,27

Niger 111232 Other personal effects 10,53

Niger 110513 Repair of furniture, furnishings and floor coverings 10,29

Swazi-land

110520 Household textiles 9,53

Kenya 110513 Repair of furniture, furnishings and floor coverings 9,31

Niger 1101115 Pasta products 9,16

Niger 1103141 Cleaning and repair of clothing 8,97

Burkina Faso

1101122 Pork 8,84



Kenya 110612 Other medical products 8,77

Togo 1101115 Pasta products 8,56

Mozam-bique

110960 Package holidays 8,48

Burkina Faso

110533 Repair of household appliances 8,28

25 Most Extreme lower quantity Relatives

Mozam-bique

111250 Insurance 1315,42

Niger 1101122 Pork 663,52

Niger 110735 Combined passenger transport 149,30

Uganda 1101142 Preserved milk and milk products 55,13

Swazi-land

110533 Repair of household appliances 51,34

Rwanda 1103141 Cleaning and repair of clothing 51,03

Swazi-land

110733 Passenger transport by air 46,70

Togo 110452 Gas 41,52

Niger 1102131 Beer 34,58

Benin 140115 Receipts from sales 25,48

Niger 1101173 Frozen or preserved vegetables 24,41

Rwanda 1102111 Spirits 23,36

Mozam-bique

110943 Games of chance 22,87

Uganda 1101173 Frozen or preserved vegetables 19,89

Mozam-bique

110915 Repair of audio-visual, photographic and information processing equipment

17,67

Swazi-land

110921 Major durables for outdoor and indoor recreation 16,61

Uganda 1101115 Pasta products 15,77

Togo 1101151 Butter and margarine 15,66

Togo 1101143 Cheese 15,64

Kenya 1101115 Pasta products 14,71

Uganda 111232 Other personal effects 14,55

Burkina Faso

110820 Telephone and telefax equipment 14,13

Rwanda 110551 Major tools and equipment 13,33



Rwanda 111120 Accommodation services 12,81

Uganda 1101162 Frozen, preserved or processed fruits 12,79

More generally, applying such an approach to a subset of countries belonging to the same cluster may reveal anomalies in the data that might otherwise be missed if the approach was applied only to the full set of 48 countries.

In this context, it would be useful if appropriate clusters for ICP 2011 could be identified based on the ICP 2005 data. Data validation then could begin before all countries have submitted their data. As well, if the clusters are formed based on unvalidated ICP 2011 data, errors in the data could cause some countries to be placed in the wrong clusters.

A drawback, however, to using the ICP 2005 data to construct country clusters that are then used to validate the ICP 2011 data is that most of the dissimilarity measures considered here do not generate robust dendrograms and clusters. The exception is the dissimilarity measure in (4) that meas-ures differences in per capita GDP, which seems to be reasonably robust to changes in the data. Hence, it is perhaps worth considering for the ICP 2011 data validation process.

Finally, the dendrogram approach to cluster formation often generates one or more singleton clusters (although this does not happen in either the per capita GDP-GEKS or per capita GDP-Iklé dendrograms in Figures 3a and 3b). For example, both Sudan and Ethiopia end up in singleton clusters in the suggested configuration arising out the dendrogram in Figure 2b. Therefore, the clustering approach does not provide guidance about which other countries should be used as points of reference when validating the quantity data of these two countries. A solution is to refer back to the dis-similarity matrix to find the countries closest to Sudan and Ethiopia in their dissimilarity measures. The two countries closest to Sudan, based on djkQ applied at the level of GDP, are Burundi and Benin; the two countries clos-est to Ethiopia are Mali and Burundi. Hence Sudan’s quantity data can be validated in comparison to the quantity data of Burundi and Benin, and Ethiopia’s quantity data can be validated in comparison to the quantity data of Mali and Burundi. This process, however, should not be reversed. Burundi’s quantity data should be validated using the countries in its cluster, and the same is true for Benin and Mali. With slight modifications, this general approach can be applied to clusters that contain only two countries if it is felt that these clusters are too small.



6. CONClUsION

A few main themes emerge from this study. First, anomalies in the data can be identified from the dissimilarity matrices themselves, before the ap-plication of cluster analysis methods. Second, dissimilarity matrices can be constructed in a number of ways, each of which may help identify different types of outliers. Third, the dendrograms and resulting clusters in general are not very robust to the way the dissimilarity measure is defined or the sample of data over which they are calculated. Fourth, cluster analysis may prove useful in data validation in at least two ways. The construction of dendrograms over the full set of countries helps to highlight those with anomalous data. In addition, the division of countries into clusters before the application of data validation methods may increase the power of these methods. This might lead to the detection of anomalies that would other-wise have been masked by the inherent heterogeneity of the countries in the region. However, the latter ideally requires a dissimilarity measure that generates clusters that are reasonably stable from ICP 2005 to ICP 2011. A possible candidate in this regard is the dissimilarity measure derived from the difference in per capita GDP. Fifth, it seems that not enough attention was devoted to validation of the basic heading expenditure data in ICP 2005. This may be partly because neither of the two main data validation tools that were used (the Quaranta and Dikhanov Tables) is designed for this purpose. Dissimilarity measures and the dendrograms derived from them may prove useful in ICP 2011. Finally, it is recommended that the main empirical calculations in this report be repeated on the ICP 2011 data for Africa as part of data validation. The dendrograms in Figures 3a and 3b could also be used as a guide when forming subregions in Africa, thereby allowing data validation at a more refined level.

REfERENCEs

Blades, D. (2007), “GDP and Main Expenditure Aggregates,” Chapter 3, ICP 2003-2006 Handbook. Washington D.C.: The World Bank.

Diewert, W.E. (2001), “Similarity and Dissimilarity Indexes: An Axiomatic Approach,” Discussion Paper No. 02-10, Revised March 2006. Vancouver, British Columbia: Department of Economics, University of British Columbia.

Diewert, W.E. (2009), “Similarity Indexes and Criteria for Spatial Linking,” in Rao D.D. (ed.), Purchasing Power Parities of Currencies: Recent Advances in Methods and Applications. Cheltenham United Kingdom: Edward Elgar.



Diewert, W. Erwin (2010), “New Methodological Developments for the International Comparison Program,” Review of Income and Wealth 56, Special Issue 1, S11-S31.

Diewert, W.E. (2011), “Methods of Aggregation above the Basic Heading Level within Regions,” Chapter 7 in Rao D.S. and F. Vogel (eds.), Measuring the Size of the World Economy: A Framework, Methodology and Results from the International Comparison Program (ICP), forthcoming.

Dikhanov, Y. (1997), “Sensitivity of PPP-Based Income Estimates to Choice of Aggregation Procedures.” Washington D.C. Development Data Group, International Economics Department, The World Bank.

Everitt, B.S., S. Lander, M. Leese and D. Stahl (2011), Cluster Analysis, Fifth Edition, Wiley Series in Probability and Statistics. Chichester: John Wiley & Sons.

Heston, A. (1994), “A Brief Review of Some Problems in Using National Accounts Data in Level of Output Comparisons and Growth Studies,” Journal of Development Economics, vol. 44, no. 1, pp. 29-52.

Hicks, J.(1946), Value and Capital, Second Edition. Oxford: Clarendon Press.

Hill, R.J. (1997), “A Taxonomy of Multilateral Methods for Making In-ternational Comparisons of Prices and Quantities,” Review of Income and Wealth, vol. 43, no. 1, pp. 49-69.

Hill, R.J. and T. P. Hill (2009), “Recent Developments in the International Comparison of Prices and Real Output,” Macroeconomic Dynamics, vol. 13, supplement 2, pp. 194-217.

Leontief, W. (1936), “Composite Commodities and the Problem of Index Numbers,” Econometrica, vol. 4, no. 1, pp. 39-59.


2. Comparative analysis of costs of some selected infrastructure components across Africa: Results from the 2005 International Comparison Program for Africa (ICP-Africa)

Oliver Chinganya1, Abdoulaye Adam2 and Marc Kouakou3

Abstract The economic growth and development of a country depend on a solid infrastruc-ture and the robustness of systems that have been put in place. Together, these constitute a nation’s “engine of growth” and include housing, water, electricity, transportation, communication, and construction. It is postulated that the cost of doing business in Africa is much higher than in other regions, largely because of the poor quality of its infrastructure and to accessibility constraints. This paper analyzes the distribution of price levels of these economic drivers, which contribute to the cost of doing business in Africa. Price level indices (PLIs) have been calculated to provide a comparison of the cost of selected infrastructure components across African countries. The data were collected from the 2005 round of the International Comparison Program (ICP) in Africa, covering 48 out of a total of 52 countries and 22 major aggregates of the national accounts.

Key words: GDP per capita, real levels of GDP, purchasing power parities, infrastructure

Analyse comparative des coûts de certaines composantes choisies d’infrastructure à travers l’Afrique: résultats du Programme de comparaison international 2005 pour l’Afrique (PCI-Afrique)

RésuméLa croissance économique et le développement d’un pays dépendent d’une in-frastructure solide et la robustesse des systèmes mis en place. Ensemble, ils consti-tuent le “moteur de croissance’’ d’une nation et comprennent le logement, l’eau, l’électricité, le transport, la communication, et la construction. Il est admis que le coût de faire des affaires en Afrique est beaucoup plus élevé que dans d’autres régions, principalement en raison de la mauvaise qualité de son infrastructure et à des contraintes d’accessibilité. Cet article analyse la distribution des niveaux

1 Statistics Department, African Development Bank, Temporary Relocation Agency, Tunis, Tunisia2 Independent Consultant, ICP, African Development Bank, Temporary Relocation Agency, Tunis, Tunisia3 Statistics Department, African Development Bank, Temporary Relocation Agency, Tunis, Tunisia



de prix de ces facteurs économiques, qui contribuent au coût de faire des affaires en Afrique.

Des indices de niveau de prix (INP) ont été calculés pour faire une comparaison du coût des composantes d’infrastructure choisies à travers les pays africains. Les données ont été recueillies à partir du cycle 2005 du Programme de comparaison internationale pour l’Afrique (PCI-Afrique), qui a couvert 48 sur un total de 52 pays et 22 grands agrégats des comptes nationaux.

Mots clés: PIB par habitant, niveaux réels du PIB, parités de pouvoir d’achat (PPA)

1. INTRODUCTION

The incentive to invest in an economic activity is affected by factors such as the cost of labor, the available infrastructure (e.g., transportation and ICT), and the regulatory and fiscal environment. A potential investor will be drawn to regions or countries that promise to deliver the greatest economic gains. A measure for making such comparisons is the price level index (PLI),4 which is derived by dividing the purchasing power parity (PPP) index by the corresponding exchange rate. A PLI represents the average percentage by which the prices of goods and services in country X, when converted into country Z’s currency at the current exchange rate, exceed or fall below the prices of the same goods and services in country Z. Because the PLI is usually measured in percentages, a PLI of 100 denotes that the price levels in both countries are the same. A higher or lower PLI indicates higher or lower costs, respectively. When currencies are converted using market exchange rates, they provide a comparison at a single point in time of relative purchasing power of one currency over another. However, this is a somewhat distorted picture, since exchange rates are volatile and the comparison does not take the price levels into account. Price level indices are better determinants. They can be used to make investment decisions, for example, whether to transfer capital from one country to another, or whether to alter the composition of an investment portfolio by switching

4 PPPs and PLIs can also be used by international organizations and companies as equal-izing factors for wages and salaries in order to compensate for purchasing power losses or gains experienced by employees working abroad (“correction coefficients”). The IMF, for instance, has used purchasing power parity (PPP) adjusted Gross Domestic Product (GDP) measures in their World Economic Outlook (WEO) since 1993, and, more recently, in the formula used to guide decisions on the distribution of members’ quotas. PLIs are not intended to rank countries in a strict hierarchy. Rather, they indicate the order of magnitude of the price level in one country in relation to others.


Oliver Chinganya, Abdoulaye Adam and Marc Kouakou

economic activities, depending on the comparative advantage and economies of scale in one country over another.

According to the paper, “Overhauling the Engine of Growth: Infrastructure in Africa” (Foster, 2008), African countries devote 6% to 8% of their GDP to infrastructure. Calderón and Servén (2008) argue that across Africa, infrastructure contributed 99 basis points to per capita economic growth over the 1990-to-2005 period, whereas the contribution of structural poli-cies represented 68 basis points. This infrastructural contribution is almost entirely attributable to advances in the telecommunications sector. Foster notes that deterioration in the energy infrastructure over the same period has had a significant lagging effect on economic growth in a number of African countries.

The paper by Olu Ajakaiye and Mthuli Ncube, “Infrastructure and Eco-nomic Development in Africa: An overview,” argues that the cost of doing business in Sub-Saharan Africa is higher than in any other global region, with infrastructure services making up a disproportionately large part of production and trade costs. This viewpoint is supported by the World Eco-nomic Forum’s Global Competitiveness Report 2010-2011, which points out that although some African countries such as South Africa and Mauritius have made great strides, Sub-Saharan Africa as a whole lags behind the rest of the world in terms of competitiveness. This is largely attributable to a severe deficit in the quality, quantity, and ease of access to infrastructure services. Although some schools of thought suggest that the relationship between infrastructure development and economic development is far from clear-cut, evidence from other studies indicates that good infrastructure is making a major contribution to reducing inequality and improving growth in all regions of the world except Africa. The the poor quality and low level of accessibility of Africa’s infrastructure negatively impacts the continent’s productivity and growth, and acts as a major disincentive to FDI and do-mestic investment, as well as curtailing international trade.

A working paper by the African Development Bank, “Infrastructure Deficit and Opportunities in Africa” (AfDB, 2010), highlights the role of infrastruc-ture in improving a nations’ competitiveness and in facilitating domestic and international trade. Poor infrastructure means higher delivery costs, which, in turn, increase the price of goods in domestic and export markets. Moreover, this must be set against the background of the recent fuel and food crises, which led to hikes and volatility in commodity prices, render-ing Africa more vulnerable to exogenous shocks and further weakening its competitiveness.



2. METhODOlOgy

This paper provides a cost comparison for selected infrastructure compo-nents across countries using PLIs. It defines the PLI as the ratio of PPP to a corresponding market exchange rate. The PPPs were calculated using the African average as the base, i.e., they were normalized with the average for Africa = 1.0. The PLI is not designed to measure inflation from one year to the next; rather, it is used to compare the cost of six selected infrastructure components—housing, water, electricity, transportation, communication, and construction—among 48 African countries, using price data collected from the International Comparison Program (ICP) round of 2005. The infrastructure components selected represent major drivers of an economy’s development.

A descriptive analysis of each of the six infrastructure components is pro-vided. A Principal Component Analysis (PCA) is performed to explain the variation of a few uncorrelated linear combinations of the original vari-ables. This was undertaken to shed light on the multivariate nature of the infrastructure components and to identify similarities among countries. However, because PCA is an exploratory method, the question of whether these objectives can be achieved through the use of principal components cannot be ascertained in advance of the analysis of the numerical results. The paper concludes with a summary of results.

3. PRICE lEvEl INDICEs (PlI) REsUlTs

The price level indices for the six selected infrastructure components are presented in Table 1.



Table 1. Price level indices for infrastructure components by country, Africa region (Average = 1.00)

COUNTRy hous-ing

Water Elec-tricity

gas Other fuels

Trans-port

Com-muni-cation

Con-struc-tion

Angola 0.66 4.44 1.20 0.82 2.43 1.47 1.72 1.24

Benin 0.76 0.97 0.59 0.78 0.61 0.87 1.62 0.70

Botswana 1.17 1.51 1.27 1.67 1.34 1.07 0.82 0.94

Burkina Faso 0.70 0.82 0.62 0.69 0.76 1.12 1.25 0.89

Burundi 0.74 0.38 1.05 2.49 0.61 0.89 0.38 0.60

Cameroon 1.02 0.89 0.82 0.90 1.03 0.87 1.54 1.13

Cape Verde 4.47 4.19 1.13 1.24 1.42 1.12 1.02 1.30

Central Afri-can Republic

0.35 0.86 1.01 2.17 0.65 1.35 1.30 1.19

Chad 0.34 1.24 1.00 1.51 0.91 1.10 1.60 1.30

Comoros 1.73 1.22 1.73 2.22 … 1.37 1.20 0.82

Congo 1.31 1.61 0.83 0.88 1.08 1.24 1.63 1.98

Congo, Democratic Rep

1.08 1.08 2.03 3.64 1.56 1.34 1.70 0.74

Côte d’Ivoire 1.08 1.17 0.70 0.65 1.04 1.17 1.50 3.03

Djibouti 1.13 0.83 1.51 1.69 1.85 1.16 0.94 0.91

Egypt 0.92 0.66 2.44 3.19 2.57 0.40 0.65 0.54

Equatorial Guinea

1.69 … 1.42 1.22 2.30 1.29 2.06 2.24

Ethiopia 0.65 0.36 0.93 1.11 1.08 0.49 0.50 0.57

Gabon 2.02 1.99 1.21 1.10 1.84 1.27 1.62 1.04

Gambia 0.19 0.68 1.03 1.77 0.83 0.82 0.57 0.83

Ghana 0.15 1.19 0.66 0.59 1.01 0.86 0.99 0.83

Guinea 0.47 0.77 0.87 2.56 0.41 0.81 0.79 0.78

Guinea-Bissau 0.58 1.26 1.17 1.78 1.06 1.06 2.34 0.78

Kenya 0.45 1.39 0.98 1.63 0.81 0.94 1.34 0.93

Lesotho 0.78 1.47 0.95 1.44 0.87 1.01 1.34 1.53

Liberia 1.53 3.68 1.20 2.52 0.79 1.24 1.24 1.22

Madagascar 1.02 0.36 0.69 1.43 0.46 0.83 0.71 0.67

Malawi 0.53 2.05 1.03 2.57 0.57 1.10 1.30 0.35



COUNTRy hous-ing

Water Elec-tricity

gas Other fuels

Trans-port

Com-muni-cation

Con-struc-tion

Mali 1.08 0.81 0.82 1.24 0.76 0.95 1.17 1.09

Mauritania 0.32 1.68 0.95 0.78 1.59 0.88 1.15 0.84

Mauritius 2.06 0.67 0.54 0.70 0.57 1.19 0.49 1.09

Morocco 1.63 2.35 0.73 0.41 1.78 1.00 0.97 1.32

Mozambique 0.49 0.92 0.67 1.03 0.61 1.16 1.28 1.54

Namibia 1.95 2.27 1.99 2.27 2.40 1.13 1.21 1.51

Niger 0.66 1.08 0.91 1.26 0.90 1.02 1.20 0.84

Nigeria 0.52 1.15 1.26 1.78 1.23 0.79 1.14 1.10

Rwanda 1.31 0.59 0.72 0.94 0.76 0.88 0.95 0.80

Sao Tome and Principe

0.87 1.74 1.51 2.26 1.40 0.96 1.22 1.18

Senegal 0.73 1.52 1.20 1.42 1.40 0.99 0.74 0.92

Sierra Leone 0.34 1.97 0.86 1.97 0.52 0.89 1.47 0.61

South Africa 1.78 1.21 1.57 2.06 1.66 1.08 0.99 1.56

Sudan 0.78 2.45 0.90 0.54 2.04 0.79 1.03 1.33

Swaziland 1.74 1.80 1.95 1.58 3.32 1.00 1.13 1.34

Tanzania 0.76 1.75 1.10 2.05 0.81 0.82 1.21 0.76

Togo 0.45 1.19 0.52 0.55 0.68 0.99 1.62 1.26

Tunisia 1.57 0.40 0.82 0.37 2.52 0.98 0.69 0.90

Uganda 0.90 1.36 0.84 1.95 0.50 0.93 1.37 0.88

Zambia 2.65 0.42 1.13 1.93 0.91 1.30 2.28 1.09

Zimbabwe 4.85 … 5.54 … 5.83 3.56 2.28 1.28

Average 1,00 1,00 1,00 1,00 1,00 1,00 1,00 1,00

Standard deviation

0,93 0,91 0,77 0,75 0,95 0,42 0,46 0,46

Coefficient of Variation (%)

93 91 77 75 95 42 46 46

Note: Ellipsis (...) indicates that the data were not submitted.



3.1 Infrastructure components

Transportation This component includes passenger transportation by railroad, road, air, sea and inland waterways, and other purchased transport services. Figure 1 shows a ranking of countries from the most to the least expensive trans-portation price levels. The distribution of PLI varies widely, resulting in a relative variation (coefficient of variation)5 of 39.4%. This could be at-tributed to the high cost of air travel on the continent. Poor road and rail networks in many parts of Africa, and between ports and the hinterland, also contribute to high transport costs, and therefore, high PLIs. In 22 of 48 countries (nearly half ), the transport PLIs are less than 1.0, ranging from 0.4 (Egypt) to 0.99 (Senegal).

Figure 1: Transport, most and least expensive countries (2005, Africa = 1.0)

Source: Results of the International Comparison Program-Africa, 2005

Costs were highest in Zimbabwe, followed by Angola, Comoros, and the Central African Republic. The countries with the lowest costs were Egypt, Ethiopia, and Nigeria. The transport PLI for Zimbabwe—more than 250% above the African average—is an outlier, and, in part, reflects the massive inflation the country was experiencing as a result of its economic crisis. As

5 The coefficient of variation is the standard deviation of the PLIs of countries as a per-centage of their average PLI. The higher the coefficient, the higher the price dispersion across countries.

0,0

0,5

1,0

1,5

2,0

2,5

3,0

3,5

4,0

Zim

babw

e A

ngol

a C

omor

os

Cen

tral A

frica

n R

epub

lic

Con

go, D

emoc

ratic

Zam

bia

Equ

ator

ial G

uine

a G

abon

C

ongo

Li

beria

M

aurit

ius

Côt

e d'

Ivoi

re

Moz

ambi

que

Djib

outi

Nam

ibia

B

urki

na F

aso

Cap

e V

erde

M

alaw

i C

had

Sou

th A

frica

B

otsw

ana

Gui

nea-

Bis

sau

Nig

er

Leso

tho

Mor

occo

S

waz

iland

A

frica

Reg

ion

Sen

egal

To

go

Tuni

sia

Sao

Tom

e an

d P

rinci

pe

Mal

i K

enya

U

gand

a S

ierra

Leo

ne

Bur

undi

R

wan

da

Mau

ritan

ia

Cam

eroo

n B

enin

G

hana

M

adag

asca

r G

ambi

a Ta

nzan

ia

Gui

nea

Sud

an

Nig

eria

E

thio

pia

Egy

pt



well, the economic embargo imposed on the country would likely have contributed to deterioration in its infrastructure.

As indicated earlier, PLIs are not intended to rank countries in a strict hi-erarchy, but rather, to indicate the magnitude of price levels in one country relative to others. A counter-intuitive finding is that price levels of neighbor-ing countries are not always of the same order of magnitude.

Communication Communication includes postal services, telephone (cell phones and land-lines), internet, etc. These services facilitate communication for public and private enterprises within and between countries. Figure 2 shows that Guinea-Bissau, Zimbabwe, and Zambia have the highest communication PLIs, and Burundi, Mauritius, Ethiopia, The Gambia and Egypt, the low-est. Guinea-Bissau, the most expensive, has a communication PLI of 2.34, while Burundi is the cheapest at 0.38. The coefficient of variation is about 37%, indicating relatively high price variation among countries.

Figure 2: Communication, most and least expensive countries (2005, Africa = 1.0)


Thirty-four countries (71%) record PLIs for communication above the African average, ranging from 1.02 in Cape Verde to 2.34 in Guinea-Bissau.

0,0

0,5

1,0

1,5

2,0

2,5

Gui

nea-

Bis

sau

Zim

babw

e Za

mbi

a E

quat

oria

l Gui

nea

Ang

ola

Con

go, D

emoc

ratic

…

Con

go

Ben

in

Togo

G

abon

C

had

Cam

eroo

n C

ôte

d'Iv

oire

S

ierra

Leo

ne

Uga

nda

Leso

tho

Ken

ya

Mal

awi

Cen

tral A

frica

n

Moz

ambi

que

Bur

kina

Fas

o Li

beria

S

ao T

ome

and

N

amib

ia

Tanz

ania

C

omor

os

Nig

er

Mal

i M

aurit

ania

N

iger

ia

Sw

azila

nd

Sud

an

Cap

e V

erde

A

frica

Reg

ion

Sou

th A

frica

G

hana

M

oroc

co

Rw

anda

D

jibou

ti B

otsw

ana

Gui

nea

Sen

egal

M

adag

asca

r Tu

nisi

a E

gypt

G

ambi

a E

thio

pia

Mau

ritiu

s B

urun

di



Intuitively, the cost of telecommunication services between neighboring countries or those located in regional economic communities (RECs) might be expected to be roughly equivalent. However, based on the price data collected from the ICP 2005 round, this is not the case. Provision of com-munication services, particularly for telephones, seems to be differentiated and fragmented across countries. And even when countries share the same service provider, the cost of the service often varies widely, for instance, for mobile telephony.

ConstructionThis includes construction of residential buildings, non-residential build-ings, and civil engineering works. PLIs vary widely among countries, with a coefficient of variation of 42.5%. PLIs for construction were lowest in Malawi, Egypt, Ethiopia, and Burundi—Malawi was the cheapest at 0.35. The highest PLIs were in Côte d’Ivoire, Equatorial Guinea, Congo Republic, South Africa, and Lesotho. Côte d’Ivoire recorded the highest costs, at 50% above the African average.

Figure 3: Construction, most and least expensive countries (2005, Africa = 1.0)


0,0

0,5

1,0

1,5

2,0

2,5

3,0

3,5

Côt

e d'

Ivoi

re

Equ

ator

ial G

uine

a C

ongo

S

outh

Afri

ca

Moz

ambi

que

Leso

tho

Nam

ibia

S

waz

iland

S

udan

M

oroc

co

Cap

e V

erde

C

had

Zim

babw

e To

go

Ang

ola

Libe

ria

Cen

tral A

frica

n R

epub

lic

Sao

Tom

e an

d P

rinci

pe

Cam

eroo

n N

iger

ia

Zam

bia

Mau

ritiu

s M

ali

Gab

on

Afri

ca R

egio

n B

otsw

ana

Ken

ya

Sen

egal

D

jibou

ti Tu

nisi

a B

urki

na F

aso

Uga

nda

Nig

er

Mau

ritan

ia

Gam

bia

Gha

na

Com

oros

R

wan

da

Gui

nea

Gui

nea-

Bis

sau

Tanz

ania

C

ongo

, Dem

ocra

tic …

B

enin

M

adag

asca

r S

ierra

Leo

ne

Bur

undi

E

thio

pia

Egy

pt

Mal

awi



HousingThis component includes actual and imputed rentals for housing and main-tenance, plus the cost of repair for dwellings. The distribution of the PLIs for housing in Figure 4 shows that Zimbabwe is, by far, the most expensive country, followed by Cape Verde, Zambia, and Mauritius. At the other end of the scale, Ghana enjoys the lowest housing costs, followed by The Gambia and Mauritania. The coefficient of variation for the housing PLI is 81.4%, revealing significantly high variation among countries.

Figure 4: Housing, most and least expensive countries (2005, Africa = 1.0)


Water This includes water supply and miscellaneous services, such as sanitation and sewage. Also included are associated costs such as the hire of meters, the reading of meters, and standing charges. The cost excludes drinking water sold in bottles or containers, and hot water or steam supplied by distinct heating plants. The distribution of water PLIs in Figure 5 shows high price dispersion across countries, indicated by a coefficient of variation of 64.7%. Water is most expensive in Angola, followed by Cape Verde and Liberia, while it is cheapest in Ethiopia, Madagascar, Burundi, Tunisia, and Zambia.

0,0

1,0

2,0

3,0

4,0

5,0

6,0

Zim

babw

e C

ape

Ver

de

Zam

bia

Mau

ritiu

s G

abon

N

amib

ia

Sou

th A

frica

S

waz

iland

C

omor

os

Equ

ator

ial G

uine

a M

oroc

co

Tuni

sia

Libe

ria

Rw

anda

C

ongo

B

otsw

ana

Djib

outi

Mal

i C

ongo

, Dem

ocra

tic …

C

ôte

d'Iv

oire

C

amer

oon

Mad

agas

car

Afri

ca R

egio

n E

gypt

U

gand

a S

ao T

ome

and

Prin

cipe

S

udan

Le

soth

o Ta

nzan

ia

Ben

in

Bur

undi

S

eneg

al

Bur

kina

Fas

o N

iger

A

ngol

a E

thio

pia

Gui

nea-

Bis

sau

Mal

awi

Nig

eria

M

ozam

biqu

e G

uine

a K

enya

To

go

Cen

tral A

frica

n R

epub

lic

Sie

rra L

eone

C

had

Mau

ritan

ia

Gam

bia

Gha

na



Figure 5: Water, most and least expensive countries (2005, Africa = 1.0)


ElectricityThe PLI for electricity includes associated costs, such as the hire and reading of meters and standing charges. Figure 6 shows substantial price variation across countries, indicated by a coefficient variation of 64.9%. The cost of electricity in Zimbabwe is, by, far the highest, with prices 454% above the African average. The two countries with the next highest costs (though far below Zimbabwe) are Egypt and the Democratic Republic of Congo. Prices in the latter might be expected to be lower, given the country’s production potential, but the 2005 ICP data refute this. The lowest electricity costs are in Togo, Mauritius, and Benin.

0,0

0,5

1,0

1,5

2,0

2,5

3,0

3,5

4,0

4,5

5,0

Ang

ola

Cap

e V

erde

Li

beria

S

udan

M

oroc

co

Nam

ibia

M

alaw

i G

abon

S

ierra

Leo

ne

Sw

azila

nd

Tanz

ania

S

ao T

ome …

M

aurit

ania

C

ongo

S

eneg

al

Bot

swan

a Le

soth

o K

enya

U

gand

a G

uine

a-…

C

had

Com

oros

S

outh

Afri

ca

Gha

na

Togo

C

ôte

d'Iv

oire

N

iger

ia

Con

go, …

N

iger

A

frica

Reg

ion

Ben

in

Moz

ambi

que

Cam

eroo

n C

entra

l …

Djib

outi

Bur

kina

Fas

o M

ali

Gui

nea

Gam

bia

Mau

ritiu

s E

gypt

R

wan

da

Zam

bia

Tuni

sia

Bur

undi

M

adag

asca

r E

thio

pia



Figure 6: Electricity, most and least expensive countries (2005, Africa = 1.0)


Principal Component Analysis and scatter plot A Principal Component Analysis (PCA) for all the infrastructure components under study (housing, water, electricity, transportation, communication, and construction) was performed to explain the total variation with a few uncorrelated linear combinations of the original variables, called principal components. The number of principal components in the analysis is less than or equal to the number of original variables. This transformation is defined so that the first principal component has the highest possible vari-ance among all linear combinations of the original variables, while each succeeding component has the next highest variance possible under the constraint that it be uncorrelated with the preceding components.

The PCA in Table 2 shows that most of the total variation is explained by the first four principal components (83%), with the first two accounting for 57%. The correlation coefficients of these components with variables used in the analysis are presented in Table 3. The first number is the correlation coefficient and the second number in parentheses is the observed significance level (OSL) of the null hypothesis of a zero correlation coefficient.

0,0

1,0

2,0

3,0

4,0

5,0

6,0

Zim

babw

e E

gypt

C

ongo

, Dem

ocra

tic …

N

amib

ia

Sw

azila

nd

Com

oros

S

outh

Afri

ca

Sao

Tom

e an

d P

rinci

pe

Djib

outi

Equ

ator

ial G

uine

a B

otsw

ana

Nig

eria

G

abon

S

eneg

al

Ang

ola

Libe

ria

Gui

nea-

Bis

sau

Cap

e V

erde

Za

mbi

a Ta

nzan

ia

Bur

undi

M

alaw

i G

ambi

a C

entra

l Afri

can

Rep

ublic

A

frica

Reg

ion

Cha

d K

enya

Le

soth

o M

aurit

ania

E

thio

pia

Nig

er

Sud

an

Gui

nea

Sie

rra L

eone

U

gand

a C

ongo

M

ali

Cam

eroo

n Tu

nisi

a M

oroc

co

Rw

anda

C

ôte

d'Iv

oire

M

adag

asca

r M

ozam

biqu

e G

hana

B

urki

na F

aso

Ben

in

Mau

ritiu

s To

go



Table 2: Proportion of variation explained by first four componentsComponent Eigen value Difference Proportion Cumulative

Component 1 2.182 0.942 36% 36%

Component 2 1.240 0.351 21% 57%

Component 3 0.889 0.208 15% 72%

Component 4 0.684 --- 11% 83%


Table 3: Correlation coefficients of the three principal components with variables

Infra-structure components

Component 1 Component 2 Component 3 Component 4

Housing 0.57 (<0.0001) 0.52 (0.0002) -0.39 (0.006) 0.09 (0.527)

Water 0.66 (<0.0001) 0.24 (0.097) -0.005 (0.973) -0.66 (<0.0001)

Electricity 0.17 (0.247) 0.76 (<0.0001) 0.45 (0.002) 0.33 (0.023)

Transporta-tion

0.81 (<0.0001) -0.17 (0.254) 0.13 (0.380) 0.11 (0.468)

Communi-cation

0.59 (<0.0001) -0.45 (0.001) 0.55 (<0.0001) 0.045 (0.767)

Construc-tion

0.61 (<0.0001) -0.302 (0.041) -0.46 (0.0013) 0.34 (0.022)


The first component, which accounts for about 36% of the total variation, is correlated with housing (0.57), water (0.66), transportation (0.81), communication (0.59), and construction (0.61). It may be interpreted as a measure of price levels on all infrastructure components except electricity. Countries with relatively high costs for housing, water, transport, com-munication, and construction will have large values for this component.

The second component is positively correlated with housing (0.52) and electricity (0.76), and negatively correlated with communication (-0.45) and construction (-0.30). Countries that have a high value for this component are characterized by high costs for electricity and housing and low costs for communication and construction.



The third component is positively correlated with electricity (0.45) and communication (0.55), and negatively correlated with housing (-0.39) and construction (-0.46). Countries with high values for these components will have high PLIs for electricity and communication, and low PLIs for construction and housing.

The fourth component is positively correlated with electricity (0.33) and construction (0.34), and negatively correlated with water (-0.66). Countries with high values for these components will have relatively high PLIs for electricity and construction and low PLIs for water.

A scatter plot of countries in the plane of the first two principal components is presented in Figure 7. From left to right, the plot shows the least expensive to the most expensive countries in terms of all infrastructure components except electricity. From the top down, it presents the most expensive coun-tries for electricity and the cheapest for communication to the cheapest for electricity and the most expensive for communication. When the two dimensions are cross-tabulated, a possible grouping of countries into 15 clusters emerges. Some countries like Egypt, Cape Verde, Angola, and Zambia stand out and constitute single-element clusters. Within clusters, countries might be expected to exhibit similarities with respect to all components, but in some clusters, countries differ for one or two components. Country codes are presented in Annex 1.



Figure 7: Countries Projections in the plan of principal components 1 and 2


The 15 possible clusters are:

Cluster 1: Egypt has the lowest costs for communication, construction and transportation; the third lowest cost for water; a low cost for housing; and the highest cost for electricity.

Cluster 2: Cape Verde is the most expensive for housing; the second most expensive for water; 30% above the African average for construction; 13% above the average for electricity; 12% above the average for transportation; and an average cost for communication.

Cluster 3: Angola has the highest costs for water and transportation; 20% above the average cost for electricity; 23% above the average for construc-tion; 72% above the average for communication; and a low cost of housing (35% below the average).

AGO

BEN

BWA

BFA

BDI

CMR

CPV

CAF TCD

COM

COG

ZAR

CIV

DJI

EGY

ETH GAB

GMB

GHA

GIN

GNB KEN

LSO

LBR

MDG MWI MLI MRT

MUS MAR

MOZ

NAM

NER NGA

RWA

STP SEN

SLE

ZAF

SDN

SWZ

TZA

TGO

TUN

UGA ZMB

-2

-1

0

1

2

3

PC2: (Housing, Electricity) Minus Communication

-4 -2 0 2 4 PC1: Housing, Water, Transport, Communication and Construction



Cluster 4: Zambia has the highest cost for communication; the second highest costs for transportation and housing; an above-average cost for electricity, the lowest cost for water; and an average cost for construction.

Cluster 5: Namibia and Swaziland have 84%, 103%, 96% and 42.5% above-average costs for housing, water, electricity and construction, re-spectively; an above-average cost for communication; and an average cost for transportation.

Cluster 6: Côte d’Ivoire and Congo Republic have the highest cost for construction (150% above average); the second lowest cost for electricity; and above-average costs for housing (19%), water (39%), transportation (21%), and communication (56%).

Cluster 7: Gabon and Liberia have above-average costs for all components, varying from 13% above average for construction to 183% above average for water. This cluster has 20%, 25%, 43% and 78% above-average costs for electricity, transportation, communication, and housing, respectively. The costs of water are higher in Liberia.

Cluster 8: Comoros, the Democratic Republic of Congo, and South Africa have average costs for construction, above-average costs for water (16%), and high to very high costs for transportation (26%), communication (29%), housing (52%), and electricity (77%).

Cluster 9: The cluster made up of Botswana, Djibouti, and São Tomé and Príncipe is characterized by average costs for housing, transportation, com-munication, and construction, but above-average costs for water (35%) and electricity (43%). Djibouti differs from the other cluster members in the cost of water.

Cluster 10: Burundi, Ethiopia, The Gambia, Guinea, and Madagascar make up this cluster, which has the lowest cost of communication, the second-lowest costs for housing, water, transportation, and construction, and a cost of electricity about 10% below average. The cost of housing in The Gambia is far below the cluster average, while the costs of water in The Gambia and Guinea are above the average.

Cluster 11: The cluster consisting of Central African Republic, Chad, Guinea-Bissau, Lesotho, Mozambique, and Togo has the lowest cost of housing; a below-average cost of electricity; and above-average costs of transportation (11%), water (15%), construction (26%), and communication (58%).



Cluster 12: The cluster made up of Kenya, Mali, Mauritania, Niger, Rwanda, Sierra Leone, and Uganda has below-average costs of housing (28%), elec-tricity (14%) construction (15%), and transport (8%). It has above-average costs of water (26%) and communication (23%). In Rwanda, the cost of housing is far above the cluster average, and in Mauritania and Sierra Leone, the cost of water is also above the cluster average.

Cluster 13: Benin, Burkina Faso, Cameroon, and Ghana constitute this cluster, which has the lowest cost of electricity, belowthe regional -average costs for housing (35%) and construction (12%); average regional costs for water and transportation; and an above-average cost for communication (35%). Within the cluster, the price of housing in Cameroon is 37% above the cluster average, and in Ghana, it is 50% below the average.

Cluster 14: Nigeria, Senegal, Tanzania, and Tunisia make up this cluster which has below the regional average costs for housing and transportation (11%), communication (6%) and construction (8%). It has above the regional average costs for electricity (9%) and water (20%). Tunisia’s cost of housing is far above (68%) the cluster average, while the cost of water is far below (80%) the cluster average.

Cluster 15: The cluster consisting of Malawi, Morocco, Mauritius, and Sudan is characterized by below-the regional average costs for electricity (21%) and communication (6%); average costs for transportation and construction, and above-the regional average costs of housing (24%) and water (87%). The cost of housing in Mauritius and Morocco is above the cluster average, while the cost of water in Mauritius is 120% below the cluster average.

4. CONClUsIONs

The costs of the infrastructure components examined in this study (hous-ing, water, electricity, transport, communication, and construction) vary substantially among countries. And while it might be expected that the price levels of these infrastructure components in neighboring countries, or countries within the same regional economic communities, would be roughly similar, the price data collected from the ICP 2005 round show that this is not the case.

The variation in price levels suggests that policy frameworks in different countries, even those within the same regional economic communities, are not fully integrated, and so seem to be out of alignment with the prevailing



climate of support for integration at subregional and regional levels. Policy frameworks should aim to channel investment toward economic drivers that will accelerate economic transformation, and thus, productivity. According to the World Economic Forum,6 the level of productivity determines the rates of return to investments in an economy. By extension, an improvement in productivity should increase trade and foster subregional and regional integration. Some studies have shown that infrastructure is key to creating an environment that attracts foreign direct investment (FDI), which should translate into sustainable economic development. This is more likely to oc-cur in countries with policies support infrastructure development.

Country projections of the costs of infrastructure components indicate possible clustering on the basis of similarities. The results suggest that the costs of some components are unexpectedly cheaper in some clusters than in others. Further research is needed to better understand the dynamics that could improve the restructuring and to formulate evidenced-based policy and economic decisions.

The variation in the price levels of items such energy, communication, and transportation—essential to a country’s competitiveness—should prompt African governments to direct more investment toward infrastructure. International and multilateral development agencies such as the African Development Bank should continue to prioritize and increase development funding for subregional and regional infrastructure projects.

There is also a need to define strategies and mechanisms for mobilizing resources and financing infrastructure. Various instruments for financ-ing infrastructure within the framework of public–private partnerships (PPPs) have been considered by Mthuli Ncube in his paper on financing and managing infrastructure in Africa (Ncube, 2010). He analyzes various options, which include build-operate-transfer (BOT), build-own-operate-transfer (BOOT), design-bid-build (DBB), design-build-operate-maintain (DBOM), and design-build-finance-operate (DBFO). These proposals should be reviewed and implemented at national, subregional, and regional levels.

The present study reveals differences in price levels among countries, which, in part, may reflect current policies. The paper also argues for a better clustering or grouping approach for countries with similar attributes, to allow comparison on the basis of economic similarities within and outside the regional economic communities. It is recommended that this study be repeated by calculating price level indices for the 2006-to-2011 period, and

6 World Economic Forum, Global Competitiveness Report 2010–2011, p. 4.



be expanded to include other economic variables. The objective is to gain a clearer understanding of the relationships between the variations in price levels of infrastructure services. Such an analysis would shed light not only on the relationship between the costs of infrastructure services in differ-ent countries, but also on the distortion in prices. Further, it would be an indication of the pace and direction of regional integration efforts aimed at enhancing investment and expanding trade.

REfERENCEs

African Development Bank (2010), “Infrastructure Deficit and Opportuni-ties in Africa,” Economic Brief, Vol. 1, Sept. 2010, pp. 2-15.

Ajakaiye, O. and M. Ncube (2010), “Infrastructure and Economic De-velopment in Africa: An overview,” Journal of African Economies, vol. 19, suppl.1, pp. i3-i12. AERC Supplement 1. Available at SSRN: http://ssrn.com/abstract=1601765 or doi:ejq003.

Bureau of Economic and Business Research, University of Florida (2010), The 2009 Florida Price Level Index. Gainesville, Florida: University of Florida.

Calderón, C. and L. Servén (2008), Infrastructure and Economic Development in Sub-Saharan Africa, World Bank Policy Report Working Paper Series 4712. Washington, DC: World Bank.

Foster, V. (2008), Overhauling the Engine of Growth: Infrastructure in Africa, Africa Infrastructure Country Diagnostic. Washington, DC: World Bank.

Ma, D., K. Fukao, and T. Yuan (2004), “Price Level and GDP in Pre-War East Asia: 1934-36 Benchmark Consumption Purchasing Power Parity Analysis for China, Japan, Korea and Taiwan.” Paper prepared for the Conference Towards a Global History of Prices and Wages, Utrecht, Holland, August 19-21, 2004.

Ncube, M. (2010), “Financing and Managing Infrastructure in Africa,” Journal of African Economies, vol. 19, suppl.1, pp. i114-i164. doi:10.1093/jae/ejp020.

Silver, M. (2010), IMF Applications of Purchasing Power Parities Estimates, IMF Working Paper # WP/08/253. Washington, DC: International Monetary Fund.

World Economic Forum (2010), Global Competitiveness Report 2010–2011. Geneva: World Economic Forum.



ANNEX 1: COUNTRy CODEs

Country Code

Algeria DZA

Angola AGO

Benin BEN

Botswana BWA

Burkina Faso BFA

Burundi BDI

Cameroon CMR

Cape Verde CPV

Central African Republic CAF

Chad TCD

Comoros COM

Congo COG

Congo, Democratic Re-public

ZAR

Côte d’Ivoire CIV

Djibouti DJI

Egypt EGY

Equatorial Guinea GNQ

Eritrea ERI

Ethiopia ETH

Gabon GAB

Gambia GMB

Ghana GHA

Guinea GIN

Guinea-Bissau GNB

Kenya KEN

Lesotho LSO

Liberia LBR

Country Code

Libya LBY

Madagascar MDG

Malawi MWI

Mali MLI

Mauritania MRT

Mauritius MUS

Morocco MAR

Mozambique MOZ

Namibia NAM

Niger NER

Nigeria NGA

Rwanda RWA

São Tomé and Príncipe STP

Senegal SEN

Seychelles SYC

Sierra Leone SLE

Somalia SOM

South Africa ZAF

Sudan SDN

Swaziland SWZ

Tanzania TZA

Togo TGO

Tunisia TUN

Uganda UGA

Zambia ZMB

Zimbabwe ZWE

Source: Results of the International Com-parison Program-Africa, 2005



ANNEX 2: COUNTRIEs PROjECTIONs IN ThE PlAN Of fIRsT fOUR PRINCIPAl COMPONENTs, WhICh EXPlAIN 83% Of TOTAl vARIATION



AGO

BEN

BWA

BFA

BDI

CMR

CPV

CAF TCD

COM

COG

ZAR

CIV

D

JI

EGY

ETH GAB

GMB

GHA

GIN

GNB

KEN LSO

LBR

MDG I

MLI MRT MUS MAR

MOZ

NAM

NER

NGA RWA

STP SEN

LE

ZAF

SN

SWZ

TGO

TUN

ZMB

-2

-1

0

1

2

3

PC

2: (H

ousi

ng, E

lect

ricity

) Min

us C

omm

unic

atio

n


AGO

BEN

BWA BFA

BDI

CMR

CPV

CAF TCD COM

COG

ZAR

CIV

DJI

EGY

ETH

GAB

GMB GHA

GIN

GNB

KEN

LSO LBR

MDG

MWI

MLI

MRT

MUS

MAR

MOZ

NAM NER NGA

RWA

STP

SEN

SLE

ZAF SDN

SWZ TZA

TGO

TUN

UGA

ZMB

-2

-1

0

1

2

PC3: (Electricity, Communicatioon) Minus (Housinng, Construction)

-2 -1 0 1 2 3 PC2: Housing, Electricity minus Communication

Countries Projections in the plan of principal components 2 and 3

AGO

BEN

BWA BFA BDI CMR

CPV

CAF TCD

COM COG

ZAR

CIV

DJI EGY

ETH GAB GMB

GHA

GIN GNB

KEN

LSO

LBR

MDG

MWI

MLI

MRT

MUS

MAR

MOZ NAM

NER NGA

RWA STP

SEN

SLE

ZAF

SDN

SWZ

TZA

TGO

TUN

UGA

ZMB

-2

-1

0

1

2

PC

4: (E

lect

ricity

, Con

stru

ctio

n) M

inus

Wat

er







AGO

BEN

BWA BFA

BDI

CMR

CPV

CAF TCD COM

COG

ZAR

CIV

DJI

EGY

ETH

GAB

GMB GHA

GIN

GNB

KEN

LSO LBR

MDG

MWI

MLI

MRT

MUS

MAR

MOZ

NAM NER NGA

RWA

STP

SEN

SLE

ZAF SDN

SWZ TZA

TGO

TUN

UGA

ZMB

-2

-1

0

1

2

PC3: (Electricity, Communication) Minus (Housing, Construction)

-2 -1 0 1 2 3 PC2: Housing, Electricity minus Communication


AGO

BEN

BWA BFA BDI CMR

CPV

CAF TCD

COM COG

ZAR

CIV

DJI EGY

ETH GAB GMB

GHA

GIN GNB

KEN

LSO

LBR

MDG

MWI

MLI

MRT

MUS

MAR

MOZ NAM

NER NGA

RWA STP

SEN

SLE

ZAF

SDN

SWZ

TZA

TGO

TUN

UGA

ZMB

-2

-1

0

1

2

PC

4: (E

lect

ricity

, Wat

er) M

inus

Con

stru

ctio

n

-2 -1 0 1 2 3 PC2: Housing, Electricity, Minus Communication





AGO

BEN

BWA BFA BDI CMR

CPV

CAF TCD

COM COG

ZAR

CIV

DJI EGY

ETH GAB GMB

GHA

GIN GNB

KEN

LSO

LBR

MDG

MWI

MLI

MRT

MUS

MAR

MOZ NAM

NER NGA

RWA STP

SEN

SLE

ZAF

SDN

SWZ

TZA

TGO

TUN

UGA

ZMB

-2

-1

0

1

2

PC

4: (E

lect

ricity

, Con

stru

ctio

n) M

inus

Wat

er

-2 -1 0 1 2 PC3: (Electricity, Communication) Minus (Housing, Construction)



3. An alternative approach for determining air transportation costs in the International Comparison Program

Abdoulaye Adam1

AbstractIn the International Comparison Program (ICP) the cost of air transportation is compared across countries. The cost of air transportation between two cities varies depending on the quality of service and the air carrier. The quality of service (aircraft type, meals and refreshments, on-board entertainment, flight schedule, etc) is reflected in the price of the ticket. The date of issuing the ticket before travel is also a price-determining factor. Precisely capturing all the price-determining characteristics in the product description is difficult and may not be feasible. Therefore, airfare prices constitute a challenge for ICP price collection if like is to be compared like with like.

This paper examines air transportation structured product descriptions (SPDs) from the 2005 and in 2011 ICP rounds and notes their deficiencies for purposes of comparison. It proposes a number of improvements: (i) using a geographic coordinate system (latitude and longitude) to determine orthodromic distances between departure and destination points; and (ii) basing comparisons on unit (kilometer or mile) costs of these orthodromic distances. It would then be possible for air transportation price data to be collected directly by Regional Coordination teams or by the Global Office through travel agencies, using the major Computer Reservation System (CRS) vendors.

Key words: Airfare prices, Structured Product Descriptions, Latitude, Longitude, Orthodromic Distances, Computer Reservation System

RésuméLes coûts du transport aérien sont comparés entre les pays dans le Programme de comparaison internationale (PCI). Le coût du transport aérien entre deux villes varie en fonction de la qualité du service et de la compagnie aérienne. La qualité du service (type d’aéronef, les repas, les rafraîchissements et le divertissement à bord, l’horaire de vol, etc.) est répercutée sur le prix du billet. La date d’émission du billet avant le voyage est aussi un facteur déterminant le prix. Capter avec précision toutes les caractéristiques déterminant le prix dans la description du produit est difficile et peut ne pas être réalisable. En conséquence, les prix des billets d’avion constituent un défi pour la collecte des prix du PCI en vue de comparer des choses comparables.

1 Abdoulaye ADAM, International Statistical Consultant BP 12965 Niamey NIGER, Email:[email protected]


Abdoulaye Adam

Ce document examine les descriptions structurées de produits (DSP) du transport aérien des deux cycles 2005 et 2011 du PCI et met en exergue leurs lacunes en termes d’exhaustivité pour des fins de comparaison. Il propose un certain nombre d’améliorations et recommande de: (i) utiliser les coordonnées géographiques (latitude et longitude) pour déterminer les distances orthodromiques entre les points de départ et de destination et ; baser les comparaisons sur les coûts uni-taires (km ou mile) de ces distances orthodromiques. Alors il est possible de faire la collecte des prix du transport aérien directement par les équipes régionales de coordination ou par le Bureau mondial à travers des agences de voyages en utilisant les systèmes informatiques majeurs de réservation.

Mots clés : Tarifs aériens, Descriptions structurées de produits, Latitude, Lon-gitude, Distances orthodromiques, Système informatique de réservation

1. INTRODUCTION

The qualityof Purchasing Power Parities (PPPs), which are used for country comparisons, depends on the capacity to separate pure price levels from other factors, and to measure them accordingly. One of the most difficult problems in constructing those indices is accurately describing products so as to capture and treat quality change and/or differences across countries. A product is the smallest entity on which price measurements may be made and compared. It should be defined by a “complete description.”

In the 2005 International Comparison Program (ICP) round, Structured Product Descriptions (SPDs) were used to describe products, the prices of which were collected and compared among countries. An SPD is a coherent and structured set of price-determining characteristics or parameters aimed at describing goods and services in a rationalized fashion in order to develop an item list and harmonize item identification for price surveys in different countries. SPDs provide all the relevant characteristics of products, whereby each product represents a particular configuration of those characteristics.

Airfare prices were collected during the 2005 ICP round from two main sources: (i) airline companies and (ii) travel agencies using Computer Reser-vation Systems and/or airlines. In addition to airfare, surcharges, and taxes, the price of a ticket includes the quality of the service (aircraft type, meals and refreshments, on-board entertainment, flight schedule, etc). The date of issuing the ticket before travel is also a price-determinant.



The air transportation products in 2005 were not directly comparable. The SPDs failed to capture differences in the quality of services. As well, the distances involved differed considerably.

This paper examines the 2005 and in 2011 air transportation SPDs and notes their shortcomings for comparative purposes. To make the prices of air transportation more comparable, improvements are proposed for the 2011 ICP product list: (i) use of a geographic coordinate system (latitude and longitude) to determine orthodromic distances between the points of departure and destination, and (ii) basing the comparisons on the unit (kilometer or mile) cost of these orthodromic distances. This would make it possible for price data for air transportation to be collected directly by Regional Coordination teams or by the Global Office through travel agen-cies, using the major Computer Reservation System (CRS) vendors.

2. DEsCRIPTIONs UsED IN 2005 AND 2011 ICP-AfRICA ROUNDs

Structured Product Descriptions (SPDs) were used in 2005 to define prod-ucts in the ICP-Africa list, including transportation by air. An example is provided in Table 1.


Abdoulaye Adam

Table 1: Example of air transportation product description in 2005 ICP-Africa

Basic heading: Passenger transport by air

Position élémentaire : Transport de voyageurs par air

Product name: Airfare to johannesburg Product code: 088.01

Nom du produit : Tarif vol aérien pour johannesburg

Code du produit : 088.01

Preferred quantity and unit of meas-urement:

1 service Quantité de référence et Unité de Mesure :

1 Service

Product description:

• Quantity: 1 service

• Type : Scheduled service

• Discounts/Type of ticket : Weekend rule (night Saturday/Sunday must be spent at destination)

• Discounts/Type of ticket : Minimum stay (of _7__ days)

• Discounts/Type of ticket : Maximum stay (of __60_ days)

• Discounts/Type of ticket : Advance reservation ( _14_ days)

• Discounts/Type of ticket : Advance payment ( _14_ days)

• Taxes : Airport tax included

• Type of Trip : Round trip

• Luggage : Included up to_25_kg

• Other item features : Cheapest non-rebookable Economy class fare

• Other item features : Exclude airport or departure tax

A problem with the description in Table 1 is that a major price-determining factor—the distance between the points of departure and destination—is not considered, so the product is not directly comparable across countries. And because the data source (airline or travel agency) is not indicated, the difficulty of product comparability is compounded. Furthermore, as



pointed out earlier, the quality of the service, and thus, the cost depends on the airline. One of the characteristics of the product is “Cheapest non-rebookable Economy class fare,” but this does not cover other restrictive conditions like date fixity or date change penalties. A better way of taking conditions of ticket usage into account would be to include the type of Economy class ticket.

The descriptions of Transportation by air products have been improved in the 2011 ICP product list. An example is presented in Table 2.

Table 2: Example of air transportation product description in 2011 ICP-Africa

Basic heading 1107331 Passenger transport by air

Product 110733106 Flight, International, return ticket – Long Distance

Number of units: 1 Unit of measurement: Ticket

Product specifications:

Transportation Type: International flight; round trip

Ticket type: Return ticket; not changeable after issue; non refundable

Starting point: Major airpoirt of the country

Destination: Most popular international destination

Distance: Approximately 14,000 km

Class: Economy adult passenger

Departure date: On or about last Wedneysday of April

Issued before departure: 42 - 56 days

Payment: When issued

Price includes: All additional taxes, fees, surcharges and charges included in the ticket, e.g. departure tax, passenger service charge, fuel surcharge

Price excludes: Price reductions, e.g. discount or special offer only for best customers

Purchasing place: Travel agent

Comments: Specify destination

The 2011 SPD is an improvement over 2005, but the description still does not properly address the distance issue, which is a major price determinant. The most popular international destinations vary in different countries and can be far less than 14,000 km away. The 2011 SPD requires that the destination, distance in km, additional taxes, fees, surcharges and charges be


Abdoulaye Adam

specified. However, these data, all of which are important in determining the price of a ticket, may not be collected unless the it is clearly specified that they be included. Thus, the data collection instruments should have fields for these elements, just as there is for the price of the ticket. Also, it should be clear that distance refers to flying distance based on orthodromic distance.

3. sUggEsTED APPROACh fOR COllECTION Of AIR TRANsPORTATION PRICEs

3.1 Data sources

Airline companies and travel agencies are the two main sources of air trans-portation price data. Tariffs can be classified into two groups: (i) public domain information, which is accessible worldwide, provided by the three major global Computer Reservation System vendors—Amadeus, Sabre and Travelport (including Galileo and Worldspan); and (ii) market tariff, which is accessible in countries or regions only for the points of departure. For example, a travel agency in Tunis, Tunisia, does not have access to the U.S. market prices, and American travel agencies do not have access to price information for North Africa. To ensure comparability, it is essential that the same source be used to determine the price of air transportation products. This can be done if prices are determined globally.

3.2 Class types

In addition to the different data sources, prices vary within the same class depending on the class type:

• First Class: F, A and P• Business Class: J, C, D and Z• Economy Class: Y, L, K, Q, V, T, H, S and B

Each class type has different own conditions of ticket usage (e.g., date fixity, date change penalties, reimbursement, and validity).

3.3 geographic coordinate system

A location can be identified based on its latitude and its longitude, which Wikipedia defines as:



• The latitude of a location is the angular distance of that location south or north of the Equator. It is an angle and is usually measured in degrees. The Equator has a latitude of 0°, while the North and South Poles have, respectively, a latitude of 90° north (written 90° N or +90°), and a latitude of 90° south (written 90° S or −90°). The latitude of point A is the angle φ between the normal of the point and the Equator.

Figure 1: Definition of the latitude of a point A

• longitude specifies the east-west position of a point on the Earth’s surface. Points with the same longitude lie in lines running from the North to the South Pole, known as meridians. By convention, the position of zero degrees longitude is the position of the Prime Meridian, which passes through the Royal Observatory, Greenwich, England. The longitude of other places is measured as an angle east or west of the Prime Meridian. Specifically, it is the angle between a plane containing the Prime Meridian and a plane containing the North Pole, South Pole and the location in question. It ranges from 0° at the Prime Meridian to +180° eastward and −180° westward.

To precisely locate points on Earth, latitude and longitude are expressed in degrees, minutes, and seconds. A one-degree variation in latitude is about 69 miles; a minute of latitude, approximately 1.15 miles; and a second of latitude, approximately 0.02 miles. The magnitude of a difference induced by a longitude variation depends on the latitude. At the Equator, it is ap-proximately 69 miles, the same as a degree of latitude. At latitude of 45 degrees, a degree of longitude is approximately 49 miles. The size gradually decreases to zero as the meridians converge at the poles.

North Pole

φ Equator

Normal

Tangent

A


Abdoulaye Adam

3.4 Central angle

A central angle is an angle whose vertex is the center of a circle and whose sides pass through a pair of points on the circle. The sides subtend an arc between the two points whose angle is (by definition) equal to the central angle itself.

Figure 2: Definition of a central angle

The angle AOB forms a central angle, with center O and sides OA and OB.

Let ffss λφλφ ,;, be the geographical latitude and longitude of the two points A and B (a point of departure and a point of destination of a journey), and let λφ ΔΔ , be, respectively, their differences; then , the central angle be-tween them, is given by the spherical law of cosines defined in Wikipedia as:

)coscoscossincos(sinˆ λφφφφσ Δ+=Δ fsfsar (1)

When the distance is small (the two points are not far apart), the above formula can have large rounding errors. In that case, the harversine formula below is more numerically stable.

⎟⎟⎠

⎞⎜⎜⎝

⎛⎟⎠

⎞⎜⎝

⎛ Δ+⎟⎠

⎞⎜⎝

⎛ Δ=Δ2

sincoscos2

sinarcsin2ˆ 22 λφφ

φσ fs (2)

If the points are on opposite ends of the sphere (antipodal points), the following more complicated formula, which is accurate for all distances, should be used:

O

B A



⎟⎟⎟

⎠

⎞

⎜⎜⎜

⎝

⎛

Δ+

Δ−+Δ=Δ

λφφφφ

λφφφφλφσ

coscoscossinsin

)coscossinsin(cos)sin(cosarctanˆ

22

fsfs

fsfsf

(3)

3.5 Orthodromic distance

In Euclidean geometry, the shortest distance between two points is the straight line joining them. In spherical geometry, distances are generally measured along a path on the surface of a sphere. The orthodromic distance, or the great-circle distance, between two points is the shortest distance between the two points on the surface of a sphere.

Geodesics/Great circles are defined as circles on the sphere whose centers are coincident with the center of the sphere. Geodesics/Great circles possess two essential features of Euclidean geometry:

• A point that moves along the surface of a sphere without turning will follow a great circle.

• The shortest distance between two points on a sphere lies along a great circle.

The distance d between two points on a sphere of radius r, and central angle σ̂Δ (expressed in radians) is given by:

σ̂Δ= rd (4)

The shape of the earth resembles a flattened sphere. The average radius for that approximation is 6371.01 km, while the quadratic mean or root mean square approximation of the average great circle circumference gives a radius of about 6372.8 km

Example of the computation of the orthodromic distanceTo calculate the orthodromic distance between Niamey, Niger and Paris, their latitudes and longitudes are converted into decimal degrees:

Niamey (Lat: 13.5167 / Long: 2.1167)

Paris (Lat: 48.8667 / Long: 2.3333)


Abdoulaye Adam

Based on the above notation and the first equation:

φN = 13.5167 λN = 2.1167 φP = 48.8667 λN = 2.33333 Δφ = - 35.35 Δλ = - 0.2166 Cos φN = 0.9723 Cos φP = 0.6580 Cos Δλ = 0.9999 Sin φN = 0.0368 Sin φP = 0.0407

σ̂Δ = arcos{(0.2337)(0.7532) + (0.97237)(0.65780)(0.9999)} = 0.616981789

Orthodromic Distance Niamey-Paris = (0.616981789)(6372.8) = 3,931.9 km

Other orthodromic distances calculated using the above three expressions are presented in Annex 1. The same values of σ̂Δ , and thus, of distances were found with all three expressions.

An easier way to find the orthodromic distance between two cities is through Internet websites:

www.travelmath.com www.mapcrow.info www.infoplease.com www.globefeed.com www.worldatlas.com

However, because the latitude and longitude and the formulae used differ slightly from one site to another, these sites do not give exactly the same distances. For the ICP, it is important to use the same source for distances on which prices will be determined in order to control variations that are not due to price differences.

4. PROCEDURE TO DETERMINE ORThODROMIC AND CRs-BAsED AIRfARE PRICE

The price of air transportation products can be determined by ICP regional or global coordinates using orthodromic distances between points of depar-ture and destination and prices obtained from one of the main Computer Reservation System (CRS) vendors—Amadeus, Sabre and Travelport (in-cluding Galileo and Worldspan). The procedure is:

1. For each participating country, determine the geographical coordinates of the departure and destination points of the air transportation products and calculate orthodromic distances between them using one of the



central angle formulae. These distances can be obtained from one of websites listed above. However, it is important to use the same method to determine all the distances.

2. Obtain the cost of such journeys from one of the main CRS vendors. These prices are more comparable than those from airline companies, which differ globally and across categories like airfare, surcharges and taxes.

3. Calculate the price per kilometer (km) of each product. These prices will then be used to compute PPPs for the basic headings after converting them in local currencies if necessary.

For ICP-Africa, the air transportation PPPs could be calculated based on the price per km of round-trip tickets from the capital cities of participat-ing African countries to selected destinations—London, Washington DC, New York, Johannesburg, Nairobi, Cairo and Dakar—and to cities where the implementing agencies for other ICP regions are located—Tunis (ICP-Africa), Manila (ICP-Asia), Moscow (CIS-Rosstat), Santiago, Chile (ICP-LAC), Beirut, Lebanon (ICP-Western Asia), and Paris (OECD). The final list of destinations for Africa may be decided based on the information in Amadeus, which is the most common CRS in Africa.

Orthodromic distances between each capital city and the above destina-tions, obtained from www.worldatlas.com, are presented in Annex 2. Prices of an Economy class Y, one-way ticket from African capital cities to those destinations were obtained by the World Bank Travel Agency (American Express) from the Amadeus database. These prices were converted into costs per km, as presented in Annex 3. The prices per km of all products will be converted into local currencies and used to compute the PPPs.

5. CONClUsION

Air transportation between any two cities is a product whose quality (aircraft type, meals and refreshments, on-board entertainment, flight schedule, etc.) depends on the services the air carrier offers. The quality of the serv-ice is reflected in the price of the ticket and is not entirely captured in the product description. Consequently, products of different qualities are being compared. In addition, the distance between the departure and destination points should be considered in the comparison. This study demonstrates that it is possible to collect airfare prices globally, or at least regionally, us-


Abdoulaye Adam

ing one of the major CRS vendors, and to take the distance of the trip into consideration. Ideally, distance should be orthodromic, and comparisons should be based on prices per unit distance (kilometer or mile).

REfERENCE

African Development Bank (2005), Product Catalogue for ICP Price Survey in African Countries Vol 2.

Gade, K (2004), NavLab, a Generic Simulation and Post-processing Tool for Navigation- European,” Journal of Navigation 2 51-59

Longley, PA Goodchild, M.F, Maguire,D.J and Rhind,D.W (2003). Geo-graphic Information Systems and Science, John Wiley and Sons 2nd Edition

The World Bank, ICP Global Office, DECDG (2011). Product Catalogue for ICP Price Surveys Global Core List 2011

Wikipedia, the free Encyclopedia. Available at: http://en.wikipedia.org/wiki/Great-circle_distance



AN

NE

X 1

: DE

TE

RM

INAT

ION

Of

OR

Th

OD

RO

MIC

DIs

TA

NC

Es

UsI

Ng

CE

NT

RA

l A

Ng

lE

fO

RM

Ul

AE

Nia

mey

-Lon

don

13.5

251

.50

2.12

-0.1

22.

23-1

8.99

1.12

0.66

370.

6637

0.66

37N

iam

ey-N

ew Y

ork

13.5

240

.71

2.12

-74.

0176

.12

-13.

6038

.06

1.23

531.

2353

1.23

53N

iam

ey-P

aris

13.5

248

.87

2.12

2.33

-0.2

2-1

7.68

-0.1

10.

6170

0.61

700.

6170

Nia

mey

-W

ash

DC

13.5

238

.90

2.12

-77.

0479

.15

-12.

6939

.58

1.27

741.

2774

1.27

74T

unis

-Lon

don

34.0

051

.50

9.00

-0.1

29.

12-8

.75

4.56

0.32

640.

3264

0.32

64T

unis

-New

Yor

k34

.00

40.7

19.

00-7

4.01

83.0

1-3

.36

41.5

01.

1138

1.11

381.

1138

Tun

is-P

aris

34.0

048

.87

9.00

2.33

6.67

-7.4

33.

330.

2735

0.27

350.

2735

Tun

is-

Was

h D

C34

.00

38.9

09.

00-7

7.04

86.0

4-2

.45

43.0

21.

1640

1.16

401.

1640

Nai

robi

-Lon

don

-1.2

851

.50

36.8

2-0

.12

36.9

3-2

6.39

18.4

71.

0702

1.07

021.

0702

Nai

robi

-New

Yor

k-1

.28

40.7

136

.82

-74.

0111

0.82

-21.

0055

.41

1 .85

871.

8587

1.85

87N

airo

bi -

Par

is-1

.28

48.8

736

.82

2.33

34.4

8-2

5.08

17.2

41.

0178

1.01

781.

0178

Nai

robi

- W

ash

DC

-1.2

838

.90

36.8

2-7

7.04

113.

85-2

0.09

56.9

31.

9057

1.90

571.

9057

Jobu

rg-L

ondo

n-2

6.20

51.5

028

.08

-0.1

228

.20

-38.

8514

.10

1.42

351.

4235

1.42

35Jo

burg

-New

Yor

k-2

6.20

40.7

128

.08

-74.

0110

2.09

-33.

4651

.04

2.01

582.

0158

2.01

58Jo

burg

-Par

is-2

6.20

48.8

728

.08

2.33

25.7

5-3

7.53

12.8

81.

3704

1.37

041.

3704

Jobu

rg-W

ash

DC

-26.

2038

.90

28.0

8-7

7.04

105.

12-3

2.55

52.5

62.

0481

2.04

812.

0481

7097

.948

317

42.8

094

7417

.643

5

1305

2.04

62

1184

5.42

3864

86.3

451

1214

4.96

3490

71.8

909

1284

6.06

7887

33.1

164

6820

.195

2

4229

.502

378

72.4

912

3931

.901

6

Est

imat

ed F

lyin

g D

ista

nce

(km

)

8140

.910

020

80.1

855

Δλ

ΔØ

/2Δλ/

2D

ista

nce

Øs

Øf

λ sλ f

ΔσE1

ΔσE2

ΔσE3


Abdoulaye Adam

AN

NE

X 2

: OR

Th

OD

RO

MIC

DIs

TA

NC

Es

BE

TW

EE

N A

fRIC

AN

CA

PIT

Al

CIT

IEs

AN

D s

El

EC

TE

D D

EsT

INAT

ION

s

Des

tina

tion

Cai

roD

akar

jobu

rgN

airo

bi l

ondo

nW

ash-

ingt

on

DC

New

yo

rkTu

nis

Man

ila M

osco

wsa

ntia

goB

eiru

tPa

ris

Dep

artu

re P

oint

Alg

iers

- A

lger

ia27

03.3

831

79.0

7376

.653

97.2

1687

.068

77.9

6547

.968

2.1

1145

6.1

3367

.110

829.

529

78.9

1335

.6

Luan

da -

Ang

ola

4643

.25

4166

.725

41.6

2725

.966

85.1

1057

7.7

1034

5.0

4575

.812

125.

373

62.6

8948

.451

31.3

6318

.1

Cot

onou

- B

enin

4026

.59

2337

.345

05.9

3896

.550

09.4

8674

.784

24.5

3091

.112

880.

262

83.6

8808

.845

47.5

4670

.2

Gab

oron

e -

Bot

swan

a60

31.5

363

17.7

309.

728

14.4

8721

.612

710.

712

504.

365

71.4

1113

8.3

8855

.191

44.0

6419

.183

44.6

Oua

gado

ugou

-

Bur

kina

Fas

o39

25.2

317

25.8

5300

.544

74.5

4338

.579

25.9

7666

.025

93.0

1305

0.4

5864

.888

52.5

4427

.040

19.3

Buj

umbu

ra -

B

urun

di36

74.9

354

83.0

2561

.385

8.0

6624

.611

646.

011

358.

645

13.3

1026

1.4

6493

.810

747.

440

54.7

6248

.0

Yaou

ndé

- C

amer

oon

3651

.26

3394

.836

82.9

2847

.254

47.4

9675

.594

11.9

3342

.212

015.

662

41.9

9453

.541

65.7

5080

.2

Prai

s - C

ape

Ver

de58

61.5

666

6.7

7132

.468

46.8

4619

.258

80.0

5678

.839

50.4

1501

7.1

6912

.872

27.4

6291

.444

68.0

Ban

gui -

Cen

tral

A

fric

an R

epub

lic31

84.2

140

88.6

3489

.121

03.9

5508

.610

222.

099

39.5

3358

.111

232.

959

24.1

1016

8.0

3673

.551

31.2

Ndj

amen

a - C

had

2607

.63

4173

.344

53.5

2805

.445

49.4

9369

.390

73.7

2401

.511

355.

951

70.4

1032

9.1

3128

.041

73.3

Mor

oni -

C

omor

os47

36.6

772

34.8

2309

.013

19.7

8061

.613

399.

613

103.

660

58.8

9006

.273

57.8

1147

6.7

4961

.376

97.7

Kin

shas

a -

Con

-go

, Dem

. Rep

.41

44.4

4118

.127

98.3

2413

.362

90.0

1048

5.7

1023

4.2

4158

.711

804.

268

74.6

9380

.246

31.5

5918

.2

Bra

zzav

ille

-Con

go41

45.2

241

16.2

627

99.3

2415

.36

6289

.510

484

1023

2.57

4158

.45

1180

6.21

6875

.193

78.3

946

32.4

259

17.8

3



Des

tina

tion

Cai

roD

akar

jobu

rgN

airo

bi l

ondo

nW

ash-

ingt

on

DC

New

yo

rkTu

nis

Man

ila M

osco

wsa

ntia

goB

eiru

tPa

ris

Dep

artu

re P

oint

Abi

djan

-

Cot

e D

’ivoi

re46

24.8

117

68.1

4881

.92

4585

.38

5127

.881

79.2

679

50.5

534

16.2

813

590.

7966

83.5

581

65.3

151

39.0

148

19.7

9

Djib

outi

-

Djib

outi

2444

.95

6557

.12

4426

.57

1506

.66

5923

1175

8.67

1142

442

09.2

8416

.86

4916

.54

1287

4.6

2568

.02

5590

.64

Cai

ro -

Egy

pt0

5270

.86

6236

.25

3497

.24

3487

.494

09.6

490

68.6

720

81.5

192

08.3

828

25.7

212

776.

3652

2.72

3167

.98

Mal

abo

- E

quat

o-ri

al G

uine

a38

22.2

131

16.8

838

33.5

531

47.6

354

02.1

9436

.03

9179

.14

3336

.77

1230

7.36

6338

.06

9200

.86

4341

.750

41.5

6

Asm

ara

- E

ritr

ea18

43.4

760

40.6

947

08.5

1794

.18

5302

.411

120.

9810

786.

435

73.1

487

74.1

4449

.23

1270

7.94

2031

.23

4965

.04

Add

is A

baba

-E

thio

pia

2461

.82

6123

.04

4060

.111

32.3

258

6411

566.

2811

238.

8740

11.7

8934

.84

5115

.94

1234

8.05

2689

.49

5514

.34

Libr

evill

e - G

abon

4055

.84

3333

.98

3511

.89

3044

.557

42.6

9700

.06

9450

.93

3669

.44

1232

5.26

6629

.94

9078

.27

4570

.553

81.2

4

Ban

jul -

Gam

bia

5252

.25

138

6488

.56

6088

.99

4478

.865

45.4

463

24.5

234

58.1

414

458.

0565

88.7

957

07.9

157

07.9

142

67.8

1

Acc

ra -

Gha

na43

11.6

821

27.2

345

99.0

741

64.9

451

1585

13.3

882

73.6

632

72.9

413

190.

7365

02.6

685

07.9

948

31.0

347

87.5

3

Con

akry

-

Gui

nea

5210

.36

683.

8159

55.6

957

02.3

148

32.7

7067

.15

6853

.34

3577

.16

1439

5.8

6800

.55

7611

.34

5692

.04

4590

.55

Bis

sau

- G

uine

a-B

issa

u52

71.0

338

3.08

6260

.66

5945

.05

4665

.367

64.7

6549

3544

.07

1447

7.74

6719

.78

7602

.157

39.1

4441

.62

Nai

robi

- K

enya

3497

.24

6190

.21

2934

.25

067

49.9

1214

5.82

1183

8.75

4739

.91

9403

.56

6227

.35

1155

8.34

3783

.17

6383

.72

Mas

eru

- Le

soth

o65

72.6

567

83.6

133

8.19

3267

.192

99.1

1313

412

946.

2771

48.9

411

088.

4693

90.4

989

75.2

6945

.64

8922

.12

Mon

rovi

a -

Libe

ria

5139

.78

1141

.35

5496

.42

5342

.72

5100

.675

23.3

373

1236

65.2

714

247.

4169

31.7

776

40.7

256

38.6

148

33.9

7

Trip

oli -

Lib

ya17

33.7

736

88.4

6639

.07

4432

.55

2377

.579

07.8

375

76.1

735

9.88

1079

3.55

3193

.84

1136

3.53

2091

.11

1999

.73


Abdoulaye Adam

Des

tina

tion

Cai

roD

akar

jobu

rgN

airo

bi l

ondo

nW

ash-

ingt

on

DC

New

yo

rkTu

nis

Man

ila M

osco

wsa

ntia

goB

eiru

tPa

ris

Dep

artu

re P

oint

Ant

anan

ariv

o -

Mad

agas

car

5615

.19

7951

.09

2193

.46

2223

.389

64.5

1423

1.41

1394

7.17

6960

.288

00.4

581

71.4

311

338.

3158

13.4

186

01.3

9

Lilo

ngw

e -

Mal

awi

4789

.08

6380

.41

1564

.89

1375

.54

7837

.412

708.

7612

440.

3157

26.0

310

058.

2775

69.8

510

517.

8551

18.2

7461

.1

Bam

ako

- M

ali

4507

.88

1035

.27

5749

.58

5154

.95

4395

7359

.52

7116

.69

2939

.28

1369

1.55

6199

.783

00.7

249

92.6

741

17.0

7

Nou

akch

ott -

M

auri

tani

a49

42.9

246

7.75

6823

.94

6146

.66

3927

.762

74.9

760

31.0

830

42.3

814

103.

660

90.0

481

07.1

753

71.6

637

26.7

3

Port

Lou

is -

M

auri

tius

6229

.390

10.4

930

80.0

830

85.7

396

83.3

1522

014

919.

5977

87.3

179

01.9

385

50.5

511

946.

0863

39.7

493

36.2

4

Cas

abla

nca

- M

oroc

co36

76.1

123

31.8

275

65.9

759

90.1

420

91.6

6164

.67

5851

.37

1594

.85

1245

9.09

4255

.49

9861

.79

3988

.19

1863

.84

Map

uto

- M

ozam

biqu

e61

17.3

769

33.5

246

8.6

2707

.14

9038

.813

343.

3513

123.

9168

95.6

310

501.

3589

08.5

696

42.5

764

56.6

886

60.3

1

Win

dhoe

k -

Nam

ibia

5977

.27

5486

.59

1198

.75

3164

.81

8290

.611

836.

2411

645.

5861

67.9

311

992.

2287

82.1

384

90.1

764

20.3

679

21.3

8

Nia

mey

- N

iger

3540

.44

2104

.89

5165

.33

4131

.66

4230

.381

79.3

7907

.62

2348

.21

1264

2.01

5589

.28

9236

.19

4047

.77

3891

.89

Lago

s -

Nig

eria

3951

.31

2435

.844

42.4

637

91.0

250

14.4

8759

.68

8506

.98

3068

.76

1277

8.49

6241

.47

8896

.55

4472

.98

4671

.24

Kig

ali -

Rw

anda

3449

.36

5478

.15

2788

.21

751.

3364

46.1

1156

7.1

1127

3.18

4349

.95

1013

2.39

6265

.14

1093

5.31

3823

.39

6070

.77

Sao

Tom

e -

Sa

o To

me

And

Pr

inci

pe

4226

.41

3071

.95

3685

.58

3347

.457

07.7

9461

.55

9219

.67

3681

.21

1261

7.83

6728

.888

25.5

847

45.7

153

54.0

7

Dak

ar -

Sen

egal

5270

.86

066

23.7

543

95.4

6190

.264

13.5

561

90.8

3441

1446

8.88

6540

.11

7693

.21

5719

.07

4193

.37



Des

tina

tion

Cai

roD

akar

jobu

rgN

airo

bi l

ondo

nW

ash-

ingt

on

DC

New

yo

rkTu

nis

Man

ila M

osco

wsa

ntia

goB

eiru

tPa

ris

Dep

artu

re P

oint

Vic

tori

a -

Seyc

helle

s45

74.4

782

68.2

838

08.5

520

96.4

180

61.8

1390

1.76

1356

9.86

6342

.23

7494

.76

6762

.92

1300

9.59

4619

.86

7736

.59

Free

tow

n -

Sier

ra

Leon

e52

30.5

380

3.52

5844

.67

5637

.38

4931

.171

74.8

769

63.6

336

34.9

714

400.

3268

71.2

275

76.3

157

17.1

4683

.94

Joha

nnes

burg

So

uth

Afr

ica

6236

.25

5844

.67

5844

.67

5844

.67

4931

.171

74.8

769

63.6

336

34.9

714

400.

3268

71.2

275

76.3

157

17.1

4683

.94

Kha

rtou

m

- Sud

an16

73.0

853

57.8

345

84.7

518

40.4

749

51.2

1061

7.51

1028

9.63

3062

.32

9444

.76

4469

.37

1211

3.41

2023

.72

4594

.38

Mba

bane

-

Swaz

iland

6226

.93

6874

.41

308.

4128

45.7

490

94.4

1327

5.52

1306

4.32

6946

.98

1066

8.22

9028

.06

9459

.62

6576

.72

8716

Dar

Es

Sala

am

-Tan

zani

a40

93.4

566

28.4

2523

.08

619.

0273

68.9

1271

0.6

1241

0.13

5352

.67

9279

.26

6784

.62

1146

0.58

4354

.88

7002

.69

Lom

é - T

ogo

4122

.23

2215

.25

4587

.28

4028

.53

5005

.385

68.4

783

21.5

131

23.3

513

007.

163

37.3

986

99.5

146

41.9

946

71.4

4

Tuni

s - T

unis

ia20

71.2

3671

.571

44.9

749

53.2

118

58.9

7434

.39

7099

.59

338.

5210

880.

3329

70.3

611

361.

123

38.3

1480

.53

Kam

pala

-

Uga

nda

3338

.83

5707

.89

2931

.68

483.

964

52.2

1171

3.4

1141

2.13

4390

.94

9833

.36

6132

.93

1123

0.4

3682

.160

80.1

9

Lusa

ka-

Zam

bia

5013

.84

5964

.212

23.4

1816

.82

7841

.312

357.

1712

109.

0756

96.9

710

686.

8978

33.2

799

03.8

153

92.0

474

62.7

8

Har

are

- Z

imba

bwe

5226

.263

31.7

310

47.8

418

87.5

981

48.5

1272

6.97

1247

9.07

6011

.52

1044

9.52

8030

.110

022.

2855

79.7

577

70.2

9

* D

ista

nce

betw

een

citi

es in

kilo

met

ers.


Abdoulaye Adam

AN

NE

X 3

: PE

R K

IlO

ME

TE

R P

RIC

Es

IN U

s$ f

OR

AN

EC

ON

OM

y C

lA

ss y

, ON

E-W

Ay T

ICK

ET

fR

OM

AfR

ICA

N

CA

PIT

Al

CIT

IEs

TO

sE

lE

CT

ED

Ch

OsE

N D

EsT

INAT

ION

s2

Des

tina

tion

Cai

roD

akar

jobu

rgN

airo

bi l

ondo

nW

ash-

ingt

on

DC

New

yo

rkTu

nis

Man

ila M

osco

wsa

ntia

goB

eiru

tPa

ris

Dep

artu

re

Alg

iers

- A

lger

ia0.

344

0.72

00.

385

0.17

40.

615

0.16

90.

168

0.28

30.

114

0.19

20.

199

0.45

90.

362

Luan

da-A

ngol

a0.

229

0.53

40.

279

1.43

90.

502

0.46

70.

364

0.25

80.

107

0.17

10.

405

0.19

10.

404

Cot

onou

-Ben

in0.

168

1.33

50.

457

0.38

70.

604

0.38

90.

400

0.78

10.

247

0.43

70.

823

0.61

70.

527

Gab

oron

e -

Bot

swan

a0.

103

0.20

00.

478

0.22

40.

236

0.22

60.

227

0.34

90.

251

0.27

20.

216

0.23

80.

173

Oua

gado

ugou

-

B. F

aso

0.11

70.

285

0.44

50.

492

0.69

70.

126

0.07

90.

129

0.20

50.

468

0.43

20.

127

0.61

3

Buj

umbu

ra -

B

urun

di0.

277

0.18

20.

309

0.52

30.

245

0.08

60.

143

0.59

20.

213

0.18

50.

609

0.16

80.

188

Yaou

ndé

- C

amer

oon

0.53

70.

906

0.48

40.

515

0.27

70.

284

0.12

50.

479

0.31

10.

164

0.84

30.

145

0.46

3

Prai

s -

Cap

e V

erde

0.65

50.

336

0.17

60.

283

0.65

10.

289

0.17

30.

379

0.46

70.

113

0.22

30.

131

0.16

0

Ban

gui -

C A

r1.

409

0.18

20.

200

1.73

00.

511

0.40

30.

415

0.13

10.

493

0.55

20.

436

1.06

80.

562

Ndj

amen

a - C

had

0.23

90.

348

0.18

20.

451

0.66

00.

394

0.40

71.

312

0.24

50.

609

0.41

90.

926

0.63

0

Mor

oni -

C

omor

os0.

365

0.27

20.

487

0.52

10.

274

0.42

40.

175

0.38

30.

808

0.44

20.

310

0.49

30.

125

2 A

ir F

ares

wer

e ob

tain

ed fr

om th

e IC

P G

loba

l Off

ice

thro

ugh

the

Wor

ld B

ank

Trav

el A

genc

y



Des

tina

tion

Cai

roD

akar

jobu

rgN

airo

bi l

ondo

nW

ash-

ingt

on

DC

New

yo

rkTu

nis

Man

ila M

osco

wsa

ntia

goB

eiru

tPa

ris

Dep

artu

re

Kin

shas

a -

Con

go, D

em.

Rep

.

0.18

10.

349

0.86

90.

646

0.56

70.

114

0.32

60.

399

0.13

00.

383

0.43

10.

961

0.42

9

Bra

zzav

ille

- C

ongo

0.18

10.

345

0.76

60.

523

0.19

10.

095

0.35

30.

387

0.48

60.

529

0.45

70.

641

0.51

1

Abi

djan

- C

ote

D’iv

oire

0.32

90.

492

0.46

10.

389

0.21

50.

413

0.42

40.

706

0.13

00.

172

0.63

10.

217

0.65

3

Djib

outi

-

Djib

outi

0.16

30.

198

0.12

50.

200

0.11

20.

085

0.12

40.

352

0.28

40.

389

0.29

20.

144

0.12

3

Cai

ro -

Egy

pt0.

000

0.44

30.

109

0.12

00.

137

0.07

40.

053

0.09

20.

153

0.82

80.

382

0.41

70.

306

Mal

abo

-

E. G

uine

a0.

221

0.46

10.

384

0.46

70.

689

0.10

60.

237

0.88

40.

352

0.62

00.

535

1.02

30.

797

Asm

ara

- E

ritr

ea1.

323

0.15

60.

266

2.69

10.

330

0.20

10.

207

0.58

70.

203

0.51

60.

435

0.35

90.

219

Add

is A

baba

-

Eth

iopi

a0.

195

0.23

20.

151

0.40

60.

102

0.07

80.

101

0.17

40.

140

0.29

00.

350

0.23

40.

109

Libr

evill

e -

Gab

on0.

315

0.27

00.

615

0.42

00.

623

0.33

10.

340

0.42

40.

363

0.56

10.

864

0.91

80.

633

Ban

jul -

Gam

bia

0.17

81.

014

0.16

40.

305

0.17

20.

356

0.29

80.

287

0.24

70.

361

0.65

00.

166

0.42

0

Acc

ra -

Gha

na0.

181

0.40

00.

306

0.30

00.

205

0.27

30.

285

0.21

10.

121

0.34

40.

502

0.10

50.

380


Abdoulaye Adam

Des

tina

tion

Cai

roD

akar

jobu

rgN

airo

bi l

ondo

nW

ash-

ingt

on

DC

New

yo

rkTu

nis

Man

ila M

osco

wsa

ntia

goB

eiru

tPa

ris

Dep

artu

re

Con

akry

-

Gui

nea

0.32

10.

781

0.21

40.

184

0.16

60.

392

0.40

80.

641

0.22

50.

494

0.47

10.

583

0.62

3

Bis

sau

- G

uine

a-B

issa

u0.

434

0.44

90.

429

0.40

20.

354

0.53

00.

545

0.27

60.

077

0.37

60.

390

0.54

10.

375

Nai

robi

- K

enya

0.24

40.

204

0.34

70.

000

0.20

30.

232

0.09

60.

197

0.12

70.

157

0.23

40.

219

0.14

4

Mas

eru

- Les

otho

0.15

20.

256

0.63

00.

306

0.29

50.

348

0.35

30.

425

0.34

40.

324

0.40

10.

094

0.22

9

Mon

rovi

a -

Libe

ria

0.22

11.

165

0.14

60.

129

0.35

50.

281

0.29

70.

752

0.26

40.

201

1.06

90.

223

0.27

1

Trip

oli -

Lib

ya0.

174

0.13

50.

136

0.19

70.

314

0.41

20.

569

0.35

30.

125

0.29

40.

333

0.25

00.

355

Ant

anan

ariv

o -

Mad

agas

car

0.33

20.

252

0.22

30.

416

0.33

60.

389

0.39

70.

176

0.74

10.

331

0.45

30.

446

0.20

4

Lilo

ngw

e -

Mal

awi

0.30

50.

161

0.52

70.

536

0.42

70.

151

0.15

50.

318

0.31

20.

275

0.35

40.

159

0.16

7

Bam

ako

- M

ali

0.17

90.

239

0.33

90.

353

0.65

10.

469

0.48

50.

230

0.43

00.

509

0.45

00.

749

0.68

9

Nou

akch

ott -

M

auri

tani

a0.

166

0.62

90.

274

0.22

30.

661

0.39

60.

412

0.07

50.

272

0.44

60.

709

0.17

30.

514

Port

Lou

is -

M

auri

tius

0.24

20.

069

0.13

60.

488

0.18

70.

238

0.18

00.

172

0.14

40.

176

0.28

50.

242

0.14

7

Cas

abla

nca

- M

oroc

co0.

113

0.48

00.

457

0.13

80.

341

0.12

50.

110

0.10

00.

083

0.33

50.

414

0.29

20.

500

Map

uto

- M

ozam

biqu

e0.

113

0.05

90.

361

0.09

00.

355

0.25

80.

153

0.09

00.

309

0.25

30.

437

0.11

60.

170



Des

tina

tion

Cai

roD

akar

jobu

rgN

airo

bi l

ondo

nW

ash-

ingt

on

DC

New

yo

rkTu

nis

Man

ila M

osco

wsa

ntia

goB

eiru

tPa

ris

Dep

artu

re

Win

dhoe

k -

Nam

ibia

0.17

40.

096

0.21

30.

306

0.21

00.

330

0.31

90.

636

0.29

50.

521

0.33

90.

717

0.18

7

Nia

mey

- N

iger

0.32

80.

301

0.34

30.

559

0.75

30.

436

0.43

80.

131

0.24

90.

527

0.40

40.

319

0.59

9

Lago

s -

Nig

eria

0.20

20.

263

0.19

60.

360

0.31

00.

181

0.18

60.

269

0.08

60.

148

0.19

50.

112

0.25

3

Kig

ali -

Rw

anda

0.16

80.

305

0.34

80.

261

0.18

60.

086

0.14

40.

485

0.75

70.

180

0.06

40.

177

0.35

4

Sao

Tom

e -

Stp

0.51

80.

952

0.38

20.

488

0.41

30.

593

0.43

01.

212

0.32

00.

145

0.51

30.

320

0.75

8

Dak

ar -

Sen

egal

0.21

20.

000

0.25

70.

480

0.41

90.

155

0.16

00.

103

0.07

00.

312

0.83

30.

432

0.46

5

Vic

tori

a -

Seyc

helle

s0.

247

0.15

70.

205

0.25

20.

161

0.11

00.

113

0.28

40.

096

0.08

50.

161

0.11

30.

190

Free

tow

n -

Sier

ra L

eone

0.32

00.

266

0.18

60.

396

0.23

70.

583

0.58

50.

172

0.21

90.

298

0.96

40.

159

0.36

7

Joha

nnes

burg

So

uth

Afr

ica

0.12

60.

260

0.00

00.

100

0.26

70.

605

0.62

30.

274

0.24

90.

152

0.29

60.

316

0.34

3

Kha

rtou

m

-Sud

an0.

130

0.27

50.

089

0.32

60.

147

0.07

30.

076

0.20

30.

926

0.17

10.

260

0.24

60.

242

Mba

bane

-

Swaz

iland

0.15

70.

249

0.61

90.

332

0.28

30.

279

0.26

90.

077

0.34

90.

302

0.25

80.

292

0.21

0

Dar

Es

Sala

am

-Tan

zani

a0.

161

0.16

50.

331

0.52

30.

183

0.07

10.

067

0.09

10.

141

0.21

50.

262

0.08

20.

070

Lom

é - T

ogo

0.28

20.

349

0.26

40.

486

0.18

20.

117

0.40

50.

774

0.25

30.

434

0.60

40.

170

0.52

8

Tuni

s Tun

isia

0.09

50.

162

0.30

80.

181

0.19

80.

302

0.27

40.

000

0.06

40.

233

0.25

40.

071

0.06

7


Abdoulaye Adam

Des

tina

tion

Cai

roD

akar

jobu

rgN

airo

bi l

ondo

nW

ash-

ingt

on

DC

New

yo

rkTu

nis

Man

ila M

osco

wsa

ntia

goB

eiru

tPa

ris

Dep

artu

re

Kam

pala

-

Uga

nda

0.16

10.

169

0.29

60.

558

0.11

20.

071

0.07

30.

213

0.13

40.

233

0.59

10.

177

0.06

8

Lusa

ka-

Zam

bia

0.35

20.

103

0.31

50.

455

0.14

50.

119

0.11

40.

471

0.36

30.

133

0.26

00.

136

0.11

1

Har

are

- Z

imba

bwe

0.08

60.

082

0.30

30.

352

0.15

10.

278

0.28

30.

422

0.27

30.

340

0.24

30.

116

0.09

6


4. Potential scalability of a methodologically harmonized consumer price index in Africa

Rees Mpofu1 and Themba Munalula23

AbstractBuilding on the work of the COMESA and SADC Secretariats in implementing interim harmonized consumer price indices (HCPIs), this paper demonstrates the value of methodological harmonization in potentially achieving a continen-tal HCPI. The paper argues that the methodological basis for harmonization strengthens existing CPIs, because they then conform to the International Labour Organization ( ILO) Consumer Price Index (CPI) Handbook’s recommenda-tions. As a consequence, the national HCPIs, benefit from the harmonization inherent in CPIs compiled according to ILO recommendations. The result of combining the HCPIs of COMESA and SADC is an HCPI encompassing 21 countries that represent about 39% of Africa. The scalability of these HCPIs holds promise for a continental index and is premised on methodological har-monization and participation in the International Comparison Program (ICP), which yields the aggregation weights.

Key words: harmonized consumer price index, scalability, ICP

RésuméS’appuyant sur les travaux des secrétariats du COMESA et de la SADC dans la mise en œuvre des indices harmonisés de prix à la consommation (IHPC) provisoires, ce document démontre la valeur de l’harmonisation méthodologique dans l’obtention potentielle d’un IHPC continentale. Le document fait valoir que la base méthodologique pour l’harmonisation renforce les IPC existants, car ils se conforment alors aux recommandations du Manuel de l’Indice des prix à la consommation (IPC) de l’Organisation internationale du Travail (OIT). En conséquence, les IHPC nationaux bénéficient de l’harmonisation inhérente à l’IPC compilé selon les recommandations de l’OIT. Le résultat de la combinaison des IHPC du COMESA et de la SADC est un IHPC englobant 21 pays qui représentent environ 39% de l’Afrique. L’évolution de ces IHPC est très prometteuse pour un indice continental et repose sur l’harmonisation méthodologique et la participation au Programme de comparaison internationale (PCI), qui produit les pondérations pour l’agrégation.

1 Price Statistician, COMESA Secretariat, [email protected] Senior Statistician, COMESA Secretariat, [email protected]. 3 Acknowledgements: We are grateful to John Astin and Caesar Cheelo for their insight-ful comments and we note that we are solely responsible for any errors there may be in this article.


Rees Mpofu and Themba Munalula

Mots clés : Indice harmonises de prix à la consommation, évolution, PCI

1. INTRODUCTION

Regional and global analyses of economic phenomena have made cross-country comparisons of economic indicators increasingly necessary. The consumer price index (CPI), which measures price inflation, is such an indicator. Harmonization of these indices has been dominated by European efforts to obtain a harmonized index of consumer prices (HICP) for the European Union (European Union, 2000). The EU HICP has become the de facto inflation index for the Euro Zone. Using a similar basket ap-proach, the UEMOA countries in Africa have developed a regional CPI (Brilleau, 2001).

Creation of monetary unions in the Southern African Development Com-munity (SADC) and the Common Market for Eastern and Southern Africa (COMESA) is an idea rooted in Heads of State declarations and embedded in both organization’s Treaties (Article 21 and 22 of the SADC Treaty and the SADC Finance and Development Protocol for SADC; and Articles 72, 76 and 78 of the COMESA Treaty). This policy context has implications for a wide array of statistics (Belle, 2001). A regional HCPI is imperative for monitoring convergence criteria related to price stability (COMESA, 2011). Fenwick (2008) argues that development of a globally harmonized index would, first, respond to user need and demand (such as the subre-gional efforts highlighted above), and second, foster improvements in CPI methodologies.

How such an index can be attained is the subject of this paper. This study builds on current work by the COMESA and SADC Secretariats to har-monize their consumer price indices to show the potential for extension to a continental index. This is particularly policy relevant in that it provides a framework for merging various efforts whose basis is methodological harmo-nization. The paper is organized as follows: a discussion of methodological issues related to implementation of an HCPI; a brief review of current work related to the COMESA and SADC HCPIs; a discussion of the potential for achieving greater aggregation of the index within the region and in Africa; stage 2 harmonization issues; and a conclusion.



2. METhODOlOgICAl IssUEs

Although the HCPI is derived from data used to compile national CPIs, specific products are excluded so that it is a purely inflation measure (rather than a cost-of-living index, a role that the CPI also plays). Development of the COMESA and SADC HCPIs followed the EU approach of methodo-logically based harmonization. A set of implementation Regulations were defined, which, in turn, specify the steps of HCPI methodology that are consistent with the ILO recommendations for measuring the CPI (ILO, 2004).

Strengthening national CPIs is fundamental to creating a robust HCPI. Harmonization is premised on adoption of the ILO recommendations with regard to issues related to classification, formulae, etc. The pre-imple-mentation assessment in both the COMESA and SADC regions examined CPI compilation practices. The resulting roadmaps for achieving an interim HCPI were based on a comparison of current CPI compilation practice and ILO recommendations, and where they diverged, provision of technical assistance to conform. Thus, a benefit of the work related to the HCPI has been strengthening the methodological soundness of national CPIs.

For both COMESA and SADC, this is a two-stage process. The first stage, which is the current interim HCPIs of the respective regions, focuses on implementing 11 harmonization topics (based on implementation Regula-tions). The second stage, implementation of which is to start in 2013, adds 10 more harmonization topics. For COMESA, these steps were formulated as Implementation Regulations defining the scope of harmonization and are legally binding on Member states.

The implementation Regulations that guide stage 1 harmonization pertain to: the framework; product coverage; formulae; domestic concept; outlets; weights and the treatment of missing items and the introduction of new products; seasonal goods; second-hand goods; and data transmission and publication standards. A brief description of each follows.

The Framework Regulation defines the statistical requirements of the HCPI. It is a Laspeyres type index, calculated monthly, and the process of developing the indices follows well-defined timelines. The Product Coverage Regulation refers to the Classification of Individual Consumption according to Purpose (COICOP) and lists the initial product exclusions and some products permanently excluded from the HCPI. The Formula Regulation specifies that the Jevons (the geometric mean) is the default formula for



averaging unweighted price data at an elementary level. The Domestic Concept Regulation defines the economic territory of the HCPI. It specifies that, unlike national CPIs, HCPIs include expenditures by foreign visitors.

To construct a comparable HCPI for each member state, the Outlets Regu-lation attempts to establish a harmonized approach to the coverage and sampling of outlet types and individual outlets. The objective is to achieve a representative sample.

The Regulation on Weights, Product Sample and Item Substitution aims at establishing a common approach to:

Expenditure weights and product samples representative of the consump-tion patterns of the reference populations.

Replacement of missing products.

Inclusion of newly significant goods and services.

The Seasonal Goods and Second-hand Goods Regulations attempt to ensure a common approach to the treatment of such goods. The Data Transmis-sion and Publication Standards Regulation defines the standards for data transmission to COMESA and for the publication of country-specific and COMESA-wide HCPIs.

The Product Coverage Regulation proposes common product coverage across countries. The harmonization process deliberately and consistently excludes certain products. This sets the HCPI apart from some national CPIs. Be-cause inflation is a monetary phenomenon, the HCPI must be confined to actual monetary transactions. However, African economies are characterized by substantial agricultural activity at the household level, which results in considerable own-consumption of household produce. But consistent with the use of the HCPI as an inflation index, own-consumption is excluded because no cash exchanges hands. Similarly, for housing, imputed rents are excluded. Interest payments are also excluded because they are neither a service nor a good.

Exclusion of certain goods and services and certain types of transactions from the HCPI is explained by the concept of Household Final Monetary Consumption Expenditure (HFMCE) (Hill, 1999). HFMCE consists of spending by private households, and these are the only expenditures that are taken into account for the construction of an inflation index. For in-



stance, in many countries, education and health services are subsidized. If a household initially pays 100% of the price of a health product or service, but is subsequently reimbursed by the government or any other source, the HCPI should include what the private household actually paid net of any assistance. For instance, if the household ultimately pays 80% of the total cost, only that component is taken into account, and the subsidized 20% is removed from HFMCE. Thus, a distinction is made between what is actually paid for and what is consumed, with the HCPI including only what is paid for.

The HCPI also includes net expenditures on second-hand goods, primarily, clothing, footwear and motor vehicles. However, expenditures on second-hand goods in a particular country must be sizeable before they can be considered for inclusion in the HCPI.

Given that the methodological harmonization of the COMESA and SADC HCPIs followed EU practices in implementing their HICP, differences be-tween the European and the COMESA/SADC experiences require elaboration. These differences derive from the regions’ divergent economic structures, which affect the composition of HFMCE. African households allocate a higher percentage of their expenditures to food than do EU households. Some types of expenditure, for example, on second-hand clothing, are more common in Africa, and therefore, specific guidelines on how to treat them in, for instance, the COMESA HCPI as opposed to the EU’s HICP, are needed. The nature of outlet types also illustrates differences in economic structures. The EU has more sophisticated and largely homogeneous outlet types. By contrast, many African and COMESA outlets are mobile vendors and informal markets. In recent decades, online shopping has grown phenomenally in the EU, but is still non-existent or nascent in Africa. Hence, such outlets are not used for African CPIs, although they are important in the EU.

3. CURRENT WORK ON COMEsA AND sADC hCPIs

As described earlier, harmonization in the COMESA and SADC regions is guided by a set of implementation Regulations underpinned by a Framework Regulation. The current COMESA and SADC interim HCPIs are based on first-stage regulations.

The statistics for computing regional estimates come from the Member States and are based on initially computed national HCPIs using the full methodological scope of stage 1. The national HCPIs are aggregated using



weights based on the volumes of individual consumption expenditure by households (ICEH). The volumes are derived from values in national cur-rencies converted using purchasing power parities (PPPs) obtained from the International Comparison Program

The COMESA and SADC HCPIs are aggregated according to the number of countries that participate in the two regions. Hence, each country’s share determines its relative importance.

Table 1 shows the aggregate indices derived from the stage 1 harmonization process. The EAC HCPI is also shown, because it was easily derived from the COMESA and SADC indices.

Table 1: Harmonized consumer price index, COMESA, SADC and EAC, December 2010 to July 2011

Month-year

Region Dec-10 Jan-11 Feb-11 Mar-11 Apr-11 May-11 Jun-11 Jul-11

COMESA 100.00 101.63 101.74 104.20 105.46 106.00 106.49 108.06

EAC 100.00 100.11 102.41 104.86 107.44 108.40 109.29 110.99

SADC 100.00 100.90 101.51 102.61 103.04 103.66 104.22 105.32

Source: COMESA, SADC and EAC member states’ consumer price aggregates, 2010 and 2011.

The COMESA, SADC and EAC HCPIs are compared at the COICOP division level (Annex 1). The contributions of the food and non-alcoholic beverages group to the total HCPI basket of each Regional Economic Community (REC) are high, accounting for 53.1%, 40.3% and 31.9% for COMESA, EAC and SADC, respectively. The housing, water, electricity, gas and other fuels group ranks second for COMESA and EAC, account-ing for 10.3% and 15.7%, respectively, of the total basket. For SADC, the transport group (15.5%) is the second largest group. From December, 2010 to July, 2011, the all-items indices for COMESA, EAC and SADC rose by 8.7%, 11.0% and 5.3%, respectively. Over the same period, the food and non-alcoholic beverages group indices rose more quickly: 11.3%, 15.2%, and 6.4%, respectively.

4. POTENTIAl sCAlABIlITy

Scalability refers to the ability to aggregate data. For the HCPI, scalability is not simply additive, but rather, weighted averaging over several countries



or regions. Key to the scalability of the HCPI is that much of the methodo-logical harmonization has a priori been achieved. This can be illustrated by the use of the domestic or national concepts, the former of which considers expenditures of all residents in the economy, while the latter concerns only nationals of the economy. Without a harmonized approach before aggrega-tion, the use of domestic concepts by one country and the use of national concepts by another would lead to double-counting. However, the joint technical consultations on methodology between COMESA and SADC ensured harmonization. The derivation of weights from the International Comparison Program’s (ICP) PPP-adjusted volumes of ICEH shows that it is feasible to aggregate and achieve scalability of the HCPI for all ICP-participating countries, provided that other minimum conditions are met, particularly, use of the same classification and a common reference period. SADC and COMESA Member States use the same COICOP classification and December 2010=100 as the index and the price reference period. Because countries may have conducted household budget surveys at different times, it is necessary to realign or price update the weights to a common price and index reference period so that the indices can be aggregated. From country HCPIs, it is possible to construct groupings such as a regional economic community like COMESA and SADC, and potentially, further aggregate until a continental HCPI is achieved (Figure 1).

Figure 1: Scalability potential of HCPIs



Scalability can be illustrated by combining SADC and COMESA HCPIs.4 Based on the same principle of aggregation using ICEH weights, a com-bined SADC-COMESA-EAC HCPI (Tripartite HCPI) can be obtained. Aggregation of data from the 21 participating countries yields a Tripartite HCPI for the region, whose evolution in the first seven months of 2011 is shown in Figure 2, along with the individual HCPIs for COMESA, SADC and EAC. Between January and July 2011, the Tripartite HCPI rose 7.1 % overall, and its food and non-alcoholic beverages group (which accounts for 43.3% of the total Tripartite HCPI basket) registered an increase of 9.5%. (Annex 2 displays the HCPI indices by COICOP divisions.) Figure 3 shows the all-item indices of the four HCPIs.

Figure 2: Regional month on month inflation


4 SADC, COMESA and EAC have a total of 26 member countries. However, some overlapping membership exists, with 8 common members between SADC and COMESA, 4 common members between EAC and COMESA, and 1 common member between EAC and SADC. Five countries are not included in the aggregate HCPIs.

0,0

0,5

1,0

1,5

2,0

2,5

3,0

January 2011 February 2011 March 2011 April 2011 May 2011 June 2011 July 2011

%

COMESA EAC Region SADC Region Tripartite



Figure 3: Regional harmonized indices


5. sTAgE 2 hARMONIZATION

Given the amount of work needed to harmonize inflation measurement, it was decided to divide the process into at least two stages.5 Stage 2 involves further harmonization. Some of the implementation Regulations expected from this stage are: the common treatment of tariff prices, discounts and reduced prices, data validation and editing, education, health, sampling, data collection procedures, quality adjustment, and derived statistics. Imple-mentation of Stage 2 Regulations will mean greater harmonization, which is expected to be achieved in 2013.

6. CONClUsION

This paper demonstrates the potential scalability of the COMESA and SADC harmonized consumer price indices. Combining the SADC, COMESA

5 Harmonization should be regarded as a continuous process because of the time it can take to resolve issues and implement decisions. Therefore, attainment of stage 2 does not mark the end of the process, but rather, an advance from the interim HCPIs that are currently being implemented.

90,0

95,0

100,0

105,0

110,0

115,0

120,0

Dec

embe

r 201

0

Janu

ary

2011

Febr

uary

201

1

Mar

ch 2

011

Apr

il 20

11

May

201

1

June

201

1

July

201

1

Units

COMESA EAC Region SADC Region Tripartite



and EAC HCPIs shows that the methodology-based harmonization has the potential to yield larger grouping of HCPIs. This, of course, is contingent on the number of countries participating in the ICP that provide weights for aggregation. In the 2005 ICP round, 48 African countries participated. By following the recommendations of the ILO manual on consumer price indices for national CPIs, a foundation for compilation of HCPIs is achieved. If methodology-based based harmonization efforts are pursued, a continental index is a realistic possibility.

REfERENCEs

Brilleau, A. (2001), “The Methodology of the Harmonized Consumer Price Index in UEMOA Countries,” INTER-STAT, no. 23, pp. 17-40.

Mshiyeni, B. (2010), Regional Economic Integration in SADC: Progress, Prospects and Statistical Issues for Monetary Union, Proceedings of the South African Reserve Bank (SARB)/Irvin Fisher Committee Seminar, pp. 85-95.

COMESA (2011), Report on the Study on Facilitating Multilateral Fiscal Surveillance in Monetary Union Context with a Focus on COMESA Region. Lusaka, Zambia: COMESA.

European Union (2000). Report from the Commission to the Council on Harmonisation of Consumer Price Indices in the European Union.

Fenwick, D. (2008), A Globally Harmonized Consumer Price Index: A Review of the Need for such an Index and the Extent to which It Is and Can Be Met and the Implications for the ILO Resolution on Consumer Price Indices, Pro-ceedings of the 2008 World Congress on National Accounts and Economic Performance Measures for Nations. Washington DC.

ILO, IMF, OECD, UNECE, Eurostat and the World Bank (2004), Con-sumer Price Index Manual: Theory and Practice. Geneva: International Labour Organization.

Hill, P. (1999), Inflation, the Cost of Living and the Domain of a Consumer Price Index, Paper submitted to the Joint UN ECE/ILO Meeting on Con-sumer Price Indices. Geneva.



AN

NE

X

CO

ME

SA r

egio

n H

CP

Is, b

y cl

assi

fica

tion

of i

ndiv

idua

l con

sum

ptio

n ac

cord

ing

to p

urpo

se (

CO

ICO

P)

cate

gory

an

d m

onth

, Dec

embe

r 20

10 t

o Ju

ly 2

011

CO

ICO

PW

eigh

tsD

ec-1

0ja

n-11

feb-

11M

ar-1

1A

pr-1

1M

ay-1

1ju

n-11

jul-

11

Food

and

non

- al

coho

lic b

ever

ages

531.

8510

010

2.33

102.

5310

710

9.08

109.

0110

9.17

111.

3

Alc

ohol

ic b

ever

ages

and

toba

cco

27.6

110

010

0.65

100.

6710

0.88

101.

3510

4.58

110.

4511

5.9

Clo

thin

g an

d fo

otw

ear

7310

010

0.83

99.6

799

.99

100.

1510

1.66

102.

1710

2.6

Hou

sing

, wat

er, e

lect

rici

ty, g

as a

nd o

ther

fu

els

102.

7510

010

0.52

101.

4410

1.59

102.

0810

4.65

105

105

Furn

ishi

ngs,

hou

seho

ld e

quip

men

t and

ro

utin

e m

anta

inan

ce55

.76

100

101.

3410

1.62

102.

3310

2.77

103.

7210

4.28

105.

3

Hea

lth

Tran

spor

t68

.01

100

101.

8610

1.91

102.

9910

3.49

103.

8710

3.86

104.

6

Com

mun

icat

ion

24.2

310

010

0.85

101.

8999

.29

99.3

100.

0199

.92

103.

2

Rec

reat

ion

and

cultu

re19

.23

100

101.

5810

1.05

101.

210

1.87

102.

1110

2.39

103.

2

Edu

cati

on

Res

taur

ants

and

hot

els

58.7

110

010

1.33

101.

1510

1.76

101.

7710

1.91

102.

3910

2.7

Mis

cella

neou

s go

ods

and

serv

ices

38.8

610

010

1.12

101.

1410

1.88

102.

1510

2.51

103.

5110

3.7

All-

item

inte

rim

HC

PI10

010

010

1.63

101.

7410

4.2

105.

4610

610

6.49

108.

1

Sour

ce: C

OM

ESA

mem

ber

stat

es’ H

CPI

wei

ghts

, 201

0 an

d 20

11.



EAC

reg

ion

HC

PIs

, by

clas

sifi

cati

on o

f ind

ivid

ual c

onsu

mpt

ion

acco

rdin

g to

pur

pose

(C

OIC

OP

) ca

tego

ry a

nd

mon

th, D

ecem

ber

2010

to

July

201

1C

OIC

OP

Wei

ghts

Dec

-10

jan-

11fe

b-11

Mar

-11

Apr

-11

May

-11

jun-

11ju

l-11

Food

and

non

-alc

ohol

ic b

ever

ages

403.

3510

010

010

3.3

108.

2111

1.89

112.

6811

2.95

115.

2

Alc

ohol

ic b

ever

ages

and

toba

cco

27.8

410

010

0.09

101.

2210

2.08

100.

7810

1.25

103.

4910

4.6

Clo

thin

g an

d fo

otw

ear

65.9

210

010

0.00

101.

2210

2.08

103.

2610

5.03

106.

2510

7.9

Hou

sing

, wat

er, e

lect

rici

ty, g

as a

nd o

ther

fu

els

157.

2410

010

0.05

102.

6810

4.74

107.

2710

8.38

109.

2311

1.4

Furn

ishi

ngs,

hou

seho

ld e

quip

men

t and

ro

utin

e m

anta

inan

ce55

.72

100

100.

1310

4.84

106.

1810

8.17

109.

1911

0.45

111.

8

Hea

lth

Tran

spor

t10

7.83

100

100.

8710

1.1

103.

5810

7.41

108.

7811

0.67

111.

8

Com

mun

icat

ion

19.3

610

010

0.00

98.9

590

.83

90.7

691

.04

91.0

591

.24

Rec

reat

ion

and

cultu

re23

.79

100

100.

0010

2.01

102.

310

3.29

104.

2910

5.3

108.

4

Edu

cati

on

Res

taur

ants

and

hot

els

86.9

110

010

0.oo

102.

1910

3.22

104.

6510

5.78

106.

6410

7.5

Mis

cella

neou

s go

ods

and

serv

ices

52.0

410

099

.63

100

100.

8810

1.87

102.

6210

3.86

105.

5

All-

item

inte

rim

HC

PI10

0010

010

0.11

102.

4110

4.86

107.

4410

8.4

109.

2911

1

Sour

ce: E

AC

mem

ber

stat

es’ H

CPI

wei

ghts

, 201

0 an

d 20

11.



SAD

C R

egio

n H

CP

Is, b

y cl

assi

fica

tion

of i

ndiv

idua

l con

sum

ptio

n ac

cord

ing

to p

urpo

se (

CO

ICO

P)

cate

gory

and

m

onth

, Dec

embe

r 20

10 t

o Ju

ly 2

011

CO

ICO

PW

eigh

tsD

ec-1

0ja

n-11

feb-

11M

ar-1

1A

pr-1

1M

ay-1

1ju

n-11

jul-

11

Food

and

non

-alc

ohol

ic b

ever

ages

318.

6310

010

0.00

102.

7210

3.82

103.

9110

5.1

105.

5810

6.4

Alc

ohol

ic b

ever

ages

and

toba

cco

72.8

810

010

0.09

100.

810

1.07

104.

0510

4.23

104.

3210

4.6

Clo

thin

g an

d fo

otw

ear

60.4

310

010

0.00

100.

810

1.07

101.

4110

1.84

102.

1710

2.5

Hou

sing

, wat

er, e

lect

rici

ty, g

as a

nd o

ther

fu

els

136.

6310

010

0.05

100.

8710

1.87

102.

2410

2.57

103.

2210

8.3

Furn

ishi

ngs,

hou

seho

ld e

quip

men

t and

ro

utin

e m

aint

enan

ce74

.99

100

100.

1310

0.75

101.

4910

1.4

101.

8910

2.69

102.

9

Hea

lth

Tran

spor

t15

5.49

100

100.

8710

2.16

103.

9310

5.44

106.

5110

7.68

107.

7

Com

mun

icat

ion

31.6

510

010

0.00

99.6

799

.81

99.7

499

.66

99.7

899

.95

Rec

reat

ion

and

Cul

ture

41.6

110

010

0.00

99.7

510

0.52

100.

9410

1.28

101.

0810

0.9

Edu

cati

on

Res

taur

ants

and

hot

els

50.0

810

010

0.00

101.

1110

2.29

102.

6810

3.16

103.

710

3.9

Mis

cella

neou

s go

ods

and

serv

ices

57.6

210

099

.63

101.

0410

1.39

101.

910

2.5

102.

6510

3.1

All-

item

inte

rim

HC

PI10

0010

010

0.11

101.

5110

2.61

103.

0410

3.66

104.

2210

5.3

Sour

ce: S

AD

C m

embe

r st

ates

’ HC

PI w

eigh

ts, 2

010

and

2011

.



CO

ICO

PW

eigh

tsD

ec-1

0ja

n-11

feb-

11M

ar-1

1A

pr-1

1M

ay-1

1ju

n-11

jul-

11

Food

and

non

-alc

ohol

ic b

ever

ages

433.

1810

010

0.00

102.

4710

5.63

106.

9410

7.52

107.

8310

9.45

Alc

ohol

ic b

ever

ages

and

toba

cco

47.7

510

010

0.09

100.

0710

0.39

102.

6110

4.66

108.

1811

1.57

Clo

thin

g an

d fo

otw

ear

67.7

910

010

0.00

100.

0710

0.39

100.

6310

1.66

102.

1010

2.60

Hou

sing

, wat

er, e

lect

rici

ty, g

as a

nd o

ther

fu

els

119.

4910

010

0.05

101.

1610

1.69

102.

1710

3.85

104.

2910

6.71

Furn

ishi

ngs,

Hou

seho

ld e

quip

men

t and

ro

utin

e m

anta

inan

ce63

.73

100

100.

1310

1.27

102.

0010

2.22

103.

0010

3.63

104.

35

Hea

lth

Tran

spor

t10

8.54

100

100.

8710

1.91

103.

2110

4.23

104.

9310

5.46

105.

90

Com

mun

icat

ion

28.4

410

010

0.00

100.

9699

.45

99.4

799

.71

99.7

210

1.83

Rec

reat

ion

and

cultu

re29

.04

100

100.

0010

0.50

100.

9110

1.48

105.

2210

1.79

102.

16

Edu

cati

on

Res

taur

ants

and

hot

els

53.8

510

010

0.00

101.

2210

2.06

102.

2610

2.59

103.

0610

3.33

Mis

cella

neou

s go

ods

and

serv

ices

48.1

910

099

.63

101.

0610

1.67

102.

0510

2.55

103.

1810

3.49

All-

item

-nt

erim

HC

PI10

0010

010

0.11

101.

5910

3.55

104.

5310

5.17

105.

6910

7.14

Sour

ce: T

ripa

rtit

e re

gion

mem

ber

stat

es’ H

CPI

wei

ghts

, 201

0 an

d 20

11.

Trip

arti

te r

egio

n H

CP

Is, b

y cl

assi

fica

tion

of i

ndiv

idua

l con

sum

ptio

n ac

cord

ing

to p

urpo

se (

CO

ICO

P)

cate

gory

an

d m

onth


African statistical journalCall for Papers

The African Statistical Journal (ASJ) is currently accepting manuscripts for publication in French or/and English. The ASJ was established to promote the understanding of statistical development in the African region. It fo-cuses on issues related to official statistics as well as application of statistical methodologies to solve practical problems of general interest to applied statisticians.

In addition to individual academic and practicing statisticians, the Jour-nal should be of great interest to a number of institutions in the region including National Statistical Offices, Central Banks, research and train-ing institutions and sub-regional economic groupings, and international development agencies.

The Journal serves as a research outlet and information sharing publication among statisticians and users of statistical information mainly in the Africa region. It publishes, among other things:

• articles of an expository or review nature that demonstrate the vital role of statistics to society rather than present technical materials,

• articles on statistical methodologies with special emphasis on applications, • articles about good practices and lessons learned in statistical development

in the region, • opinions on issues of general interest to the statistical community and

users of statistical information in the African region, • notices and announcements on upcoming events, conferences, calls for

papers, and recent statistical developments and anything that may be of interest to the statistical community in the region.

All manuscripts are reviewed and evaluated on content, language and presentation. The ASJ is fully committed to providing free access to all articles as soon as they are published. We ask you to support this initiative by publishing your papers in this journal. Prospective authors should send their manuscript(s) to [email protected]

The ASJ is also looking for qualified reviewers. Please contact us if you are interested in serving as a reviewer.

For instructions for authors and other details, please visit our website – http://www.afdb.org/en/knowledge/publications/african-statistical-journal/


T

AFR

ICAN DEVELOPMENT FUND B

ANQ

UE A



journal statistique africain Demande de soumission d’articles

Le journal statistique africain (JSA) accepte actuellement des manuscrits pour la publication en anglais ou en français. Le JSA a été établi pour favoriser la compréhension du développement statistique dans la région africaine. Il se concentre sur des questions liées aux statistiques officielles aussi bien que l’application des méthodologies statistiques pour résoudre des problèmes pratiques d’intérêt général pour les praticiens de la statistique.

En plus des universitaires et des statisticiens de métier, le Journal devrait revêtir un grand intérêt pour les institutions de la région, notamment les offices nationaux de statistiques, les banques centrales, les instituts de recherche et les organisations économiques sous-régionaux et les agences internationales de développement.

Le Journal constitue un document de recherche et d’information entre les statisticiens et les utilisateurs de l’information statistique, principalement dans la région africaine. Il publie entre autres:

• des articles sur le plaidoyer en matière de statistique qui démontrent le rôle essentiel des statistiques dans la société, plutôt que de présenter le matériel technique,

• des articles sur les méthodologies statistiques, avec un accent particulier sur les applications,

• des articles sur les meilleures pratiques et les leçons tirées sur le développement de la statistique dans la région,

• des avis sur des questions d’intérêt général pour la communauté statistique et les utilisateurs de l’information statistique dans la région africaine,

• des informations et des annonces sur les prochains événements, les conférences, les appels à contribution pour des papiers, et

• les développements statistiques récents et tout autre aspect susceptible d’intéresser la communauté statistique dans la région.

Tous les manuscrits sont passés en revue et évalués sur le contenu, la langue et la présentation. Le JSA s’engage entièrement à fournir le libre accès à


T

AFR

ICAN DEVELOPMENT FUND B

ANQ

UE A



tous les articles dès qu’ils sont publiés. Nous vous demandons de soutenir cette initiative en publiant vos articles dans ce journal. Les auteurs éventuels devraient envoyer leur manuscrit(s) à [email protected]

Le JSA cherche également les critiques qualifiés. Veuillez nous contacter si vous êtes intéressé à contribuer en tant que critique.

Veuillez visiter notre site Web http://www.afdb.org/en/knowledge/pub-lications/african-statistical-journal/ pour les instructions aux auteurs et autres détails.


Editorial policy

The African Statistical Journal was established to promote the understanding of statistical development in the African region. It focuses on issues related to official statistics as well as application of statistical methodologies to solve practical problems of general interest to applied statisticians. Of particular interest will be exposition of: how statistics can help to illuminate develop-ment and public policy issues like poverty, gender, environment, energy, HIV/AIDS, etc.; development of statistical literacy; tracking national and regional development agenda; development of statistical capacities and ef-fective national statistical systems; and the development of sectoral statistics e.g. educational statistics, health statistics, agricultural statistics, etc.

In addition to individual academic and practicing statisticians, the Jour-nal should be of great interest to a number of institutions in the region including National Statistical Offices, Central Banks, research and train-ing institutions and sub-regional economic groupings, and international development agencies.

The Journal serves as a research outlet and information sharing publica-tion among statisticians and users of statistical information mainly in the African region. It publishes, among other things: articles of an expository or review nature that demonstrate the vital role of statistics to society rather than present technical materials, articles on statistical methodologies with special emphasis on applications, articles about good practices and lessons learned in statistical development in the region, opinions on issues of general interest to the statistical community and users of statistical information in the African region, notices and announcements on upcoming events, con-ferences, calls for papers, and recent statistical developments and anything that may be of interest to the statistical community in the region.

The papers which need not contain original material, should be of general interest to a wide section of professional statisticians in the region.

All manuscripts are peer reviewed and evaluated on content, language and presentation.


ligne éditoriale

Le Journal statistique africain a été établi pour favoriser la compréhension du développement statistique dans la région africaine. Il se concentre sur des questions liées aux statistiques officielles aussi bien que l’application des méthodologies statistiques pour résoudre des problèmes pratiques d’intérêt général pour les statisticiens de métier. L’intérêt particulier est de montrer comment les statistiques peuvent aider à mettre en exergue les problèmes de développement et de politique publique tels que la pauvreté, le genre, l’envi-ronnement, l’énergie, le VIH/ SIDA, etc.; le développe- ment de la culture statistique ; la prise en compte des questions de développement régional et national; le développement des capacités statistiques et des systèmes statis-tiques nationaux efficaces; et le développement des statistiques sectorielles comme les statistiques d’éducation, de santé, des statistiques agricoles, etc.

En plus des universitaires et des statisticiens de métier, le Journal devrait revêtir un grand intérêt pour les institutions de la région, notamment les offices nationaux de statistiques, les banques centrales, les instituts de recherche et les organisations économiques sous-régionaux et les agences internationales de développement.

Le Journal constitue un document de recherche et d’information entre les statisticiens et les utilisateurs de l’information statistique, principalement dans la région africaine. Il publie entre autres: des articles sur le plaidoyer en matière de statistique qui démontrent le rôle essentiel des statistiques dans la société plutôt que la présentation des outils techniques, des articles sur les méthodologies statistiques, avec un accent particulier sur les applications, des articles sur les meilleures pratiques et les leçons tirées de la région, des avis sur des questions d’intérêt général pour la communauté statistique et les utilisateurs de l’information statistique dans la région africaine, des in-formations et des annonces sur les prochains événements, les conférences, les appels à contribution pour des papiers, et les développements statistiques récents et tout autre aspect susceptible d’intéresser la communauté statis-tique dans la région.

Les articles, qui n’ont pas besoin de contenir du matériel original, devraient intéresser une grande partie des statisticiens professionnels dans la région.

Tous les manuscrits seront passés en revue et évalués sur le contenu, la langue et la présentation.


guidelines for Manuscript submission and Preparation

submissions

Manuscripts in English or French should be sent by email to [email protected]

Title

The title should be brief and specific. The title page should include the title, the author’s name, affiliation and address. The affiliation and address should be given as a footnote on the title page. If the manuscript is co-authored, the same information should be given for the co-author(s).

Abstract, Key Words, and Acknowledgments

A short abstract of about 150 words must be included at the beginning of the manuscript, together with up to 6 key words used in the manuscript. These key words should not repeat words used in the title. Acknowledge-ments, if any, should be inserted at the bottom of the title page.

sections and Numbering

Major headings in the text should be numbered (e.g. “1. INTRODUC-TION”). Numbered subheadings (e.g. “1.1 The establishment of the NsDs”) may be used but thereafter sub-subheadings should be unnum-bered. Main body text in the form of paragraphs should not be numbered.

formatting

Please use minimal formatting as this will facilitate harmonization of all the papers. As your default, keep to “normal” (12 pt. Times New Roman) for main text with a single line space between paragraphs. Do not apply “body text” as an inbuilt style. The levels of heading need to be easily identifiable. We recommend all capitals bold for the first level of heading in the main text (e.g. “1. INTRODUCTION”); thereafter bold upper and lower case for subheadings (e.g. “1.1 The establishment of the NsDs”) and unnumbered bold italic (e.g. “Creating a culture of cooperation”) thereafter. Please refer to the latest volume of the AJS as a guide.

house style

The Bank’s house style is US rather than British spellings (e.g. “organiza-tion” not “organisation”; “program” rather than “programme”, “analyse”


etc.). Use % rather than “percent” or “per cent” and double rather than single quotation marks. Dates should be US style (e.g. December 11, 1985 not 11 December 1985).

Tables and figures

Tables and figures should be numbered and given a title. These should be referred to in the text by number (e.g. “See Table 1”), not by page or indica-tions such as “below” or “above”.

Equations

Any equations in the paper should be numbered. The numbers should be placed to the right of the equation.

References

A list of references should be given at the end of the paper (to precede the Annexes, if included). The references should be arranged alphabetically by surname/name of organization. Where there is more than one publication listed for an author, order these chronologically (starting with the earliest). The references should give the author’s name, year of publication, title of the essay/book, name of journal if applicable. Use a, b, c, etc. to separate publications of the same author in the same year. Titles of journals and books should be in italic; titles of working papers and unpublished reports should be set in double quotation marks and not italicized.

Examples:Fantom, N. and N. Watanabe (2008). “Improving the World Bank’s Da-tabase of Statistical Capacity,” African Statistical Newsletter, vol. 2, no. 3, pp. 21–22.

Herzog, A. R. and L. Dielman (1985). “Age Differences in Response Accuracy for Factual Survey Questions,” Journal of Gerontology, vol. 40, pp. 350–367.

Kish, L. (1988a). “Multipurpose Sample Designs,” Survey Methodology, vol. 14, no. 3, pp. 19–32.

Kish, L. (1988b). A Taxonomy of Elusive Populations, Proceedings of the Section on Survey Research Methods, American Statistical Association, pp. 44–46.


World Bank (2006). Statistical Capacity Improvement in IDA Countries – Progress Report. Washington, DC: World Bank.

Cross References

In the main body of the article, cross-references should be Harvard-style, e.g. (Kish 1988a; Herzog and Dielman 1985: 351). For cross-references to three or more authors, only the first surname should be given, followed by et al., although the names of all the authors must be provided in the References entry itself. Abbreviations ibid. and op. cit. should not be used in the text or in footnotes.


Instructions pour la préparation et la soumission de manuscrits

soumission

Les manuscrits en anglais ou en français doivent être envoyés à [email protected]

Titre

Le titre devrait être bref et détaillé. La page de titre doit inclure le titre du papier, le nom de l’auteur, l’affiliation et l’adresse. L’affiliation et l’adresse doivent figurer comme note de bas de page. Si le manuscrit est produit par des coauteurs, la même information doit être donnée pour les coauteurs.

Résumé, mots clés et remerciements

Un résumé court d’environ 150 mots doit être inclus au début du manus-crit ainsi qu’environ 6 mots clés utilisés dans le manuscrit. Les mots clés ne doivent pas répéter les mots utilisés dans le titre. Les remerciements, s’il y en a, doivent être insérés en bas de la page titre.

section et numérotation

Les principaux titres doivent être numérotés (par exemple “1. INTRO-DUCTION “). Les sous-titres numérotés (par exemple “1.1 l’élaboration de sNDs”) peuvent être employés mais par la suite les sous sous-titres ne devraient pas être numérotés. Le corps principal du texte sous forme de paragraphes ne devrait pas être numéroté.

formatage

Veuillez utiliser le formatage minimal car ceci facilitera l’harmonisation de tous les articles. Garder par défaut le format “normal” (12 pt. Times New Roman) pour le texte principal avec l’espace d’une seule ligne entre les paragraphes. Ne pas appliquer le “corps de texte “ en tant que modèle intégré. Les niveaux du titre doivent être facilement identifiables. Nous recommandons les majuscules en gras pour le premier niveau titre dans le texte principal (par exemple “1. INTRODUCTION“) ; ensuite les lettres minuscules en gras pour les sous-sections (par exemple “1.1 l’élaboration de la sNDs”) et ensuite l’italique en gras sans numérotation (par exemple “créant une culture de coopération“).Veuillez vous référer au dernier vo-lume du JSA comme guide.


Tables and figures

Les tableaux et les graphiques doivent être numérotés et comporter un titre. Ceux-ci devraient être mentionnés (par exemple “voir Tableau 1” ) dans le texte par le nombre correspondant, et non par une indication de page ou par d’autres indications telles que “ci-dessous” ou “au-dessus de”.

Équations

Toutes les équations dans le papier doivent être numérotées. Les nombres doivent être placés à la droite de l’équation.

Bibliographie

Une liste de références doit être fournie à la fin de l’article (avant les annexes, le cas échéant). Les références doivent être classées par ordre alphabétique selon le nom de l’auteur ou de l’organisation. Là où il y’a plus d’une publi-cation listée pour un auteur, elles doivent être classées chronologiquement (en commençant par les plus récents). Les références doivent donner le nom de l’auteur et l’année de publication, le titre du livre, le nom du journal le cas échéant. Utiliser a, b, c, etc. pour séparer les publications du même auteur au cours der la même année. Les titres des journaux et des livres devraient être en italique ; les titres des documents de travail et des rapports non publiés devraient être placés dans de doubles guillemets et ne pas être imprimés en italique.

Exemples:Fantom, N. et N. Watanabe (2008). « Improving the World Bank’s Database of Statistical Capacity », African Statistical Newsletter, vol. 2, no. 3, pp. 21–22.

Kish, L. (1988a). « Multipurpose Sample Designs », Survey Methodology, vol. 14, no. 3, pp. 19–32.

Kish, L. (1988b). A Taxonomy of Elusive Populations, Proceedings of the Conference on Survey Research Methods, American Statistical Association, pp. 44–46.

Herzog, A. R. et L. Dielman (1985). « Age Differences in Response Accuracy for Factual Survey Questions », Journal of Gerontology, vol. 40, pp. 350–367.

World Bank (2006). Statistical Capacity Improvement in IDA Countries – Progress Report. Washington DC: The World Bank.


Renvois

Dans le corps principal de l’article, les renvois devraient suivre le modèle de Harvard, par exemple (Kish 1988a ; Herzog et Dielman 1985 :351). Pour des renvois à trois auteurs ou plus, seulement le premier nom de famille devrait être donné, suivi par et al., bien que les noms de tous les auteurs doivent être fournis dans la Bibliographie elle-même. Les abréviations ibid. et op. cit. ne devraient pas être employées dans le texte ou dans les notes de bas de page.

African S

tatistical Journal / Journal statistique africain


T

AFR

ICAN DEVELOPMENT FUND

BA

NQUE

A


© ADB/BAD, 2012 – Statistics Department • Département des statistiquesTemporary Relocation Agency (TRA) – Agence Temporaire de Relocalisation (ATR)13 Rue du GhanaBP. 323, 1002 Tunis Belvédère Tunis, Tunisia / TunisieTel: (+216) 71 102 342 Fax: (+216) 71 103 743 Internet: http://www.afdb.org Email: [email protected]

ISSN : 2233-2820

Volume 14 – M

ay / mai 2012

African Statistical Journal Journal statistique africain

Volume 14 – May / mai 2012African Development Bank Group

Groupe de la Banque africaine de développement2233 2820

Special Issue on the International Comparison Program (ICP)

Volume Spécial sur le Programme de Comparaison Internationale (PCI)

african statistical journal · 2019. 6. 29. · la valeur de la revue a été prouvée par le...

Documents