Collaborative Approach to Trade

0 downloads 0 Views 9MB Size Report
Dec 11, 2017 - grounded on the concept that trade flows are made up of integrated layered net- ... Statistical data analysis are used as a descriptive tool in this ..... increase the percentage of farmed land and they link directly with the ...... its own set of 'independent variables,' hence the multilayering. ...... Papeete (PPT).
Advances in Spatial Science

Francesca Romana Medda Francesco Caravelli Simone Caschili Alan Wilson

Collaborative Approach to Trade Enhancing Connectivity in Sea- and Land-Locked Countries

Advances in Spatial Science The Regional Science Series

Series editors Manfred M. Fischer Jean-Claude Thill Jouke van Dijk Hans Westlund Advisory editors Geoffrey J.D. Hewings Peter Nijkamp Folke Snickars

More information about this series at http://www.springer.com/series/3302

Francesca Romana Medda • Francesco Caravelli • Simone Caschili • Alan Wilson

Collaborative Approach to Trade Enhancing Connectivity in Sea- and Land-Locked Countries

Francesca Romana Medda QASER Laboratory University College London London, United Kingdom

Francesco Caravelli QASER Laboratory University College London London, United Kingdom

Simone Caschili QASER Laboratory University College London London, United Kingdom

Alan Wilson The Bartlett Centre for Advanced Spatial Analysis University College London London, United Kingdom

ISSN 1430-9602 ISSN 2197-9375 (electronic) Advances in Spatial Science ISBN 978-3-319-47038-2 ISBN 978-3-319-47039-9 (eBook) DOI 10.1007/978-3-319-47039-9 Library of Congress Control Number: 2017937905 © Springer International Publishing AG 2017 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Printed on acid-free paper This Springer imprint is published by Springer Nature The registered company is Springer International Publishing AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Preface

The study on which this book is based was originally commissioned by the World Bank in order to build better foundations for future investment decisions, particularly in Uganda and the South Pacific Islands. When we started this work, our research group, the UCL QASER Lab, was relatively new. But, we all share the idea that in order to develop a strategy for more inclusive and balanced economic development, we would have to not only leverage the trade comparative advantages of the regions under investigation but also begin with the premise that joint efforts to create connections and collaborations among the main players in trade are complementary to competition. In keeping with our starting point, we agreed that we might require methodologies and models not generally within the orthodox trade and logistics studies toolkit. In the end, we proposed to the World Bank a vision and methodological means of arriving to the objectives which were challenging and innovative at the same time. We are grateful to the World Bank and particularly to Charles Kunaka and Daniel Saslavsky for having accepted this challenge and for their support in the completion of the study. Their experience, knowledge and insights, along with those of the World Bank team, have greatly helped us to refine the initial ideas, to improve the analyses and above all to interpret the obtained results. The World Bank has published two reports based on the analyses in this volume. We have approached the study and in particular the two regions, Uganda and the South Pacific Islands, with a sense of reverence. When we began we did not know much about either of the two regions; we were just in London, in an office. So with great commitment and energy we asked questions, compiled data and information and listened to those ‘in the know’. Thereafter, we developed and tested the models and then applied them to the proposed challenges. Over the course of these months, we have learned and drawn conclusions from our analyses, showing how these regions, identified as lagging, indeed have concrete potentialities for emerging from an idled economic growth. Some of the policy recommendations proposed here we think will boost development in these so-called lagging regions, enhance trade activity and ultimately improve the welfare of their citizens. As global connections v

vi

Preface

have become paramount for economic growth more so now than in the past, these connections are also becoming extremely complex due to the fact that they intertwine many interests and trends which are cultural, environmental, demographic and political, to mention a few. We therefore argue in this book that the dilemmas occur when decisions are made on short-term single interests, whereas we need to facilitate collaborative actions and approaches in order to capture and create value from the interdependency of these different interests. In sum, the book does not aim merely to be a showcase for advanced regional science analyses applied to trade; the models and techniques used in this volume were selected in relation to the defined objectives. These methodologies were best suited for us to tackle the complexity of trade connections and to delve deeply and bring to the surface the types of problems whose solutions are effectively reached through a collaborative approach to trade. For this reason, Uganda and the South Pacific Islands are exemplars of the analytical approach we are proposing in this book. We are certainly indebted to Raul Oviedo Martinez, Luca Cocconcelli, Minette Dasigi and Alessandra Coda for their direct contributions to the book. We would also like to acknowledge the many reviewers and anonymous referees who provided us with useful comments on the manuscript. And finally, special thanks to Susan Davis for her invaluable editing work. London

Francesca Romana Medda

Contents

Part I

The Challenge and the Science

1

Enhancing Trade in Sea- and Land-Locked Countries . . . . . . . . . . 1.1 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 The Current Situation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Possibilities for Development . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 3 5 6 9

2

Two Approaches to Modelling Trade . . . . . . . . . . . . . . . . . . . . . . 2.1 The Multilayer and Agent Based Model (ABM) . . . . . . . . . . . 2.2 The Multilayer Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 The Agent Based Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4 Multiplier Attachment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . .

11 11 12 17 23 24

3

The Current Situation and Future Challenges . . . . . . . . . . . . . . . . 3.1 Core Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Future Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . .

27 27 33 34

. . . . . . . . .

37 37 39 39 44 47 50 50 52

Part II 4

The South Pacific Island Countries (SPICs)

The Multilayer Model for Sea-Locked Countries . . . . . . . . . . . . . 4.1 Methodology and Study Structure . . . . . . . . . . . . . . . . . . . . . . 4.2 Multilayer Model: Specifications and Hypotheses . . . . . . . . . . 4.3 Variables and Multilayer Configuration . . . . . . . . . . . . . . . . . 4.4 Horizontal Network Model . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5 Vertical Network Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.6 Concluding Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix: The Dijkstra Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vii

viii

Contents

5

Port Attractiveness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4 Conclusions and Policy Recommendations . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

53 53 53 56 59 63

6

Scenario Analysis of Shipping Networks: Consolidation . . . . . . . . . 6.1 Introduction: Towards Consolidation of Cargo and Reduced Transport Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Scenario 1: Existing Conditions Network . . . . . . . . . . . . . . . . . 6.3 Scenario 2: Fully Connected Network . . . . . . . . . . . . . . . . . . . . 6.4 Scenario 3: Multiplier Attachment Network . . . . . . . . . . . . . . . 6.5 Conclusions and Policy Recommendations . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

65

7

Trade Coordination Agreements . . . . . . . . . . . . . . . . . . . . . . . . . 7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2 Scenario 1: Existing Conditions Network with Bilateral Agreement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.3 Scenario 2: Fully Connected Network with Bilateral Agreement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.4 Scenario 3: Multiplier Attachment Network with Free Trade Agreement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.5 Conclusions and Policy Recommendations . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

65 68 73 77 80 81

.. ..

83 83

..

85

..

89

.. .. ..

93 97 99

. . . . . . .

101 101 101 104 107 110 111

Appendix to Part II: Data Requirements for the SPICS . . . . . . . . . . . . . II.1 Data Sets for South Pacific Island Countries—SPICs . . . . . . . . II.2 The Shipping Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Port Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Container Shipping Services . . . . . . . . . . . . . . . . . . . . . . . . . . Trade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II.3 Cultural Links and Common Language . . . . . . . . . . . . . . . . . . Trade Agreements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II.4 Transport Costs for SPICS . . . . . . . . . . . . . . . . . . . . . . . . . . . . Voyage Cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Port Fees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

113 113 116 116 120 123 125 125 126 126 129 134

8

Synthesis: The Integrated Multilayer Model . . . . . . . . . . . . . . . . . 8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2 The Vertical Network Model . . . . . . . . . . . . . . . . . . . . . . . . . 8.3 Evaluation of External Shocks on Trade . . . . . . . . . . . . . . . . . 8.4 Interactions Between the SPICs . . . . . . . . . . . . . . . . . . . . . . . 8.5 Conclusions and Policy Recommendations . . . . . . . . . . . . . . . Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Contents

Part III 9

ix

Uganda

The Agriculture Supply Chain in Uganda: The Design of an Agent Based Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.2 The Conceptual Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.3 Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.4 Agents and Agent Interactions . . . . . . . . . . . . . . . . . . . . . . . . . 9.5 Farmer-Trader Interaction Through Four Channels . . . . . . . . . . CHANNEL 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CHANNEL 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CHANNEL 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CHANNEL 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

137 137 138 139 140 144 144 145 146 146 147

10

The Implementation of the Uganda Agent Based Model . . . . . . . . . 10.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.2 Multiplier Attachment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.3 Production of Crops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.4 Price of Crop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.5 Trading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feasibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Reward . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Probability of Deal Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . Utility Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.6 Simulation of the Channels in the Model . . . . . . . . . . . . . . . . . Itinerant Trader . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conditions for Market Exit . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.7 Model Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Product Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Transport Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Market Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Agent Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

149 149 153 154 155 156 158 159 159 159 160 162 162 162 163 163 164 164 165 168

11

Transport Cost and Infrastructure Investments . . . . . . . . . . . . . . 11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Scenario 1: Existing Conditions . . . . . . . . . . . . . . . . . . . . . . . 11.3 Scenario 2: Improvement of Road Infrastructure . . . . . . . . . . . 11.4 Scenario 3: Improvement of Airport Infrastructure . . . . . . . . . 11.5 Results and Policy Recommendations . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

169 169 171 174 176 178 180

. . . . . . .

x

Contents

12

Consolidation and Reform: Towards a Collaborative Approach . . 12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12.2 Scenario 1: Existing Conditions . . . . . . . . . . . . . . . . . . . . . . . 12.3 Scenario 2: Consolidation Around Ruralpolis . . . . . . . . . . . . . 12.4 Scenario 3: Reform of the Logistics Market . . . . . . . . . . . . . . 12.5 Concluding Remarks and Policy Recommendations . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

13

Information Exchange: Collaboration and Coordination Standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13.2 Scenario 1: Existing Conditions . . . . . . . . . . . . . . . . . . . . . . . 13.3 Scenario 2: Farmer Coordination-Collaboration Strategies . . . . 13.4 Scenario 3: Standardisation and Harmonisation of Product Quality in the Supply Chain . . . . . . . . . . . . . . . . . . . . . . . . . . 13.5 Conclusions and Policy Recommendations . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Part IV 14

. . . . . . .

183 183 184 187 192 193 194

. . . .

197 197 199 200

. 206 . 211 . 212

Conclusions

Collaborative Approach to Trade . . . . . . . . . . . . . . . . . . . . . . . . . 14.1 Methods to Study Interdependency . . . . . . . . . . . . . . . . . . . . . 14.2 The Collaborative Approach to Trade . . . . . . . . . . . . . . . . . . . 14.3 Trade Interdependency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . .

217 217 219 221 222

Authors

Francesca Romana Medda is the Director of the UCL QASER (Quantitative and Applied Spatial Economics Research) Laboratory which specialises in applied economics and finance. She is Professor in Transport and Infrastructure at the University College London, Department of Civil, Environmental and Geomatic Engineering. Since 2012, she serves as economic adviser to the UK Ministry of Environment and Agriculture (Defra) and in 2014 at the Ministry of Finance (Her Majesty’s Treasury) in the Infrastructure Interdependency Committee. Her research focuses on critical and social infrastructure investments with particular interest on transport, financial innovation related to new technologies, impact finance and urban/regional investments. She is Vice-President of the UK Parliamentary and Scientific Committee, House of Commons, and Fellow of the Institution of Civil Engineers. Francesco Caravelli is a JR Oppenheimer Distinguished fellow at Los Alamos National Laboratory (LANL). He is interested in classical and quantum systems out of equilibrium. In particular, his interests focus on the study of complex systems and the application of techniques of statistical physics to other disciplines such as economics and engineering. Francesco Caravelli studied Physics at Universita´ di Pisa (Italy) and has a PhD in Theoretical Physics from the University of Waterloo (Canada). He was a student at Perimeter Institute (Waterloo, CA) and then visiting at Albert Einstein (Potsdam, GER). Before arriving at LANL, he then worked at the OCIAM Mathematical Institute, University College London (QASER Lab) and Invenia Labs (Cambridge). Simone Caschili, environmental engineer and PhD in urban planning, holds the position of associate at LaSalle Investment Management where Simone is a member of the Portfolio Asset Risk group within the Research and Strategy team. His key responsibilities include risk analysis and property market modelling, to investigate real estate markets and yield socio-economic indicators and to monitor capital market imbalances and fund/asset return performance. Simone is a senior xi

xii

Authors

research fellow of the UCL QASER Lab since 2010 where he directed his research interests on a number of topics, including the modelling of urban and regional systems, spatial and temporal modelling of transport networks and policy evaluation for planning in both urban and environmental governance. Alan Wilson, is Chief Executive of the Alan Turing Institute and Professor of Urban and Regional Systems in the Centre for Advanced Spatial Analysis at University College London. He is Chair of the Home Office Science Advisory Council. He was responsible for the introduction of a number of model building techniques which are now in common use internationally—such as the use of ‘entropy’ in building spatial interaction models—summarised in Entropy in urban and regional modelling. His current research is on the evolution of cities and global dynamics. He writes the Quaestio blog on research and interdisciplinarity. He is a Member of Academia Europaea, a Fellow of the British Academy, a Fellow of the Academy of Social Sciences and a Fellow of the Royal Society.

List of Figures

GDP annual growth: average over 10 years. Data Source: World Bank data set . .. . . . . . .. . . . . . .. . . . . .. . . . . . .. . . . . . .. . . . . .. . . . . . .

5

Fig. 2.1 Fig. 2.2 Fig. 2.3 Fig. 2.4

Multilayer networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Outline of the applied methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Functional relationships between layers . . . . . . . . . . . . . . . . . . . . . . . . . . . Schema of a supply chain . .. . . . . .. . . . . .. . . . . .. . . . . .. . . . . .. . . . .. . . . . .

13 14 15 20

Fig. 3.1

Employment to population ratio. Data Sources: World Bank data set, year 2012 . . .. . . . . .. . . . . .. . . . . .. . . . . .. . . . . .. . . . . .. . . . . .. . . . . . Agriculture value added to GDP ratio. Data Source: World Bank data set, year average (2000–2012) . . . . . . . . . . . . . . . . . Export per sector of the economy (to GDP ratio %). Data Source: World Bank data set, year 2012 . . . . . . . . . . . . . . . . . . . . Main international trading partners of SPICs and Uganda. Data Source: World Bank data set, year 2012 . . . . . . . . . . . . . . . . . . . . Exports of goods and services (% of GDP). Data Source: World Bank data set; year 2012 . . . . . .. . . . . .. . . . . .. . . . . .. . . . . . .. . . . . Export indexes: volumes and values. Data Source: World Bank data set, year 2012 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Merchandise trade to GDP ratio (%). Data Source: World Bank data set, year 2012 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Foreign Direct Investment (% of GDP). Data Source: World Bank data set, year 2012 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Net bilateral aid flows to GDP ratio. Data Source: World Bank data set. QASER Lab elaboration . . . . . . . . . . . . . . . . . . . . . . . . . . .

Fig. 1.1

Fig. 3.2 Fig. 3.3 Fig. 3.4 Fig. 3.5 Fig. 3.6 Fig. 3.7 Fig. 3.8 Fig. 3.9 Fig. 4.1

Outline of the applied methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

28 28 29 29 30 30 31 32 33 38

xiii

xiv

Fig. 4.2

Fig. 4.3 Fig. 5.1

Fig. 5.2 Fig. 5.3 Fig. 5.4 Fig. 6.1

Fig. 6.2

Fig. 6.3

Fig. 6.4 Fig. 6.5

Fig. 6.6 Fig. 6.7

Fig. 6.8 Fig. 6.9

List of Figures

A schematic representation of the trade network in sea-locked countries. h and i are supply locations. j is a demand location. o, u, v, and r represent ports. Links represent trade flows between nodes. Dashed lines show flows from different locations passing through a link . . .. . . . . . . . .. . . . . . . .. . . . . . . . .. . . . . . . Functional relationships between layers . . . . . . . . . . . . . . . . . . . . . . . . . . . Structural equation model of causal relationships between factors in port attractiveness. Source: Caschili and Medda (2015) . .. . . . . . . . . . .. . . . . . . . . . .. . . . . . . . . .. . . . . . . . . . .. . . . . . . . . .. . . . . . . . . . Structural equation modelling diagram for the estimation of port attractiveness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Geo-refereed visualisation of Port Attractiveness Index for our case study . . .. .. . .. .. . .. .. . .. .. . .. .. . .. .. . .. .. . .. . .. .. . .. .. . .. .. . .. .. . . Data summary for Lae, Port Moresby, Suva, and Busan . . . . . . . . Existing Conditions shipping network in Pacific RIM countries. SPICs countries are shown in red. Linked colours identify origin to destination trips. Node size is proportional to total number of connections . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . . .. . . . . . .. . . . . . .. . . . . . Degree distribution (dots in the small frame), cumulative degree distribution (squares in the small frame) and log-log scale (dots in the large frame) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trade flows in the Existing Conditions shipping network. Node size is proportional to the total flow passing through each port. Link width is proportional to the total volumes moved between two ports .. . . .. . . .. . .. . . .. . . .. . . .. . . .. . . .. . . .. . .. . . .. . . .. . . .. . . .. . . .. . . Trade flow values in the 109 links of the shipping line network. Values are expressed in thousands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trade flows in the Existing Conditions shipping network with economies of scale. Node size is proportional to the total flow passing through each port. Link width is proportional to the total volumes moved between two ports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fully Connected shipping network in Pacific RIM countries. SPICs countries are shown in blue . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . Trade flows in the Fully Connected shipping network. Node size is proportional to the total flow passing through each port. Link width is proportional to the total volumes moved between two ports . . . . .. . . . .. . . . .. . . . . .. . . . .. . . . .. . . . .. . . . . .. . . . .. . . . .. . . . .. . . . . .. . . . Trade flow values for the Fully Connected shipping network. Values are expressed in thousands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trade flows in the Fully Connected shipping network with economies of scale. Node size is proportional to the total flow passing through each port. Link width is proportional to the total volumes moved between two ports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

41 47

55 58 60 63

68

69

70 71

72 73

74 75

76

List of Figures

Fig. 6.10

Fig. 6.11

Fig. 6.12 Fig. 6.13

Fig. 7.1

Fig. 7.2

Fig. 7.3

Fig. 7.4

Fig. 7.5

Fig. 7.6

xv

Multiplier Attachment network. SPICs countries are visualised in blue. Node size is proportional to the total number of connections of each port . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trade flows in the Multiplier Attachment network. Node size is proportional to the total flow passing through each port. Link width is proportional to the total volumes moved between two ports . . . . .. . . . .. . . . .. . . . . .. . . . .. . . . .. . . . .. . . . . .. . . . .. . . . .. . . . .. . . . . .. . . . Trade flow values in Multiplier Attachment network. Values are expressed in thousands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trade flows in the Multiplier Attachment network with economies of scale. Node size is proportional to total flow passing through each port. Link width is proportional to total volumes moved between two ports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trade flows in the Existing Conditions network with agreement among PNG, Solomon Islands, Vanuatu, New Caledonia and Australia. The acronyms of the ports in countries part of the agreement are: NOU, NOR, VLI, BNE, HIR, MAG, KIM, ORO, RAB, PMB, LAE. Node size is proportional to total flow passing through each port. Link width is proportional to total volumes moved between two ports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trade flow values in the Existing Conditions network with agreement among PNG, Solomon Islands, Vanuatu, and New Caledonia. Values are expressed in thousands . . . . . . . . . . . . . . . . . . . . Trade flows in the Existing Conditions network with economies of scale and agreement among PNG, Solomon Islands, Vanuatu, and New Caledonia. The acronyms of the ports in countries part of the agreement are: NOU, NOR, VLI, BNE, HIR, MAG, KIM, ORO, RAB, PMB, LAE. Node size is proportional to total flow passing through each port. Link width is proportional to total volumes moved between two ports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trade flow values in the Existing Conditions network with economies of scale and agreement among PNG, Solomon Islands, Vanuatu, and New Caledonia. Values are expressed in thousands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trade flows in the Fully Connected network with agreement among PNG, Solomon Islands, Vanuatu, New Caledonia, and Australia. The acronyms of the ports in countries part of the agreement are: NOU, NOR, VLI, BNE, HIR, MAG, KIM, ORO, RAB, PMB, LAE. Node size is proportional to total flow passing through each port. Link width is proportional to total volumes moved between two ports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trade flow values in the Fully Connected network with economies of scale and agreement among PNG, Solomon Islands, Vanuatu, New Caledonia, and Australia. Values are expressed in thousands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

77

78 79

80

85

86

87

88

89

90

xvi

Fig. 7.7

Fig. 7.8

Fig. 7.9

Fig. 7.10 Fig. 7.11

Fig. 7.12

Fig. 8.1

Fig. 8.2 Fig. 8.3

Fig. 8.4

Fig. II.1 Fig. II.2 Fig. II.3 Fig. II.4

List of Figures

Trade flows in the Fully Connected network with economies of scale and agreement among PNG, Solomon Islands, Vanuatu, New Caledonia, and Australia. The acronyms of the ports in countries part of the agreement are: NOU, NOR, VLI, BNE, HIR, MAG, KIM, ORO, RAB, PMB, LAE. Node size is proportional to total flow passing through each port. Link width is proportional to total volumes moved between two ports . . . . . . Trade flow values in the Fully Connected network with economies of scale and agreement among PNG, Solomon Islands, Vanuatu, New Caledonia, and Australia. Values are expressed in thousands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trade flows in the Multiplier Attachment network with free trade agreement in the SPICs. Node size is proportional to total flow passing through each port. Link width is proportional to total volumes moved between two ports . . . . . . . . . . . . . . . . . . . . . . . . . . Trade flow values in the Multiplier Attachment network with agreement in the SPICs. Values are expressed in thousands . . . . Trade flows in the Multiplier Attachment network with free trade agreement in the SPICs and economies of scale. Node size is proportional to total flow passing through each port. Link width is proportional to total volumes moved between two ports . . . . .. . . . .. . . . .. . . . . .. . . . .. . . . .. . . . .. . . . . .. . . . .. . . . .. . . . .. . . . . .. . . . Trade flow values in the Multiplier Attachment network with free trade agreement in the SPICs and economies of scale. Values are expressed in thousands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Plot of randomly removed links versus total travelled distance and linear fitting trend line between 16% and 52% of the removed links (R2 ¼ 0.99) and after 52% (R2 ¼ 0.99) . . . . . . . . . . Potential Transhipment of ports in the Existing Conditions network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Plot of removed links of ports in Suva, Brisbane, Tauranga, and Lae versus total travelled distance, and linear fitting trend line between 1% and 25% removed links (R2 ¼ 0.99) and after 25.5% (R2 ¼ 0.98) . . .. . .. . .. . .. . . .. . .. . .. . .. . . .. . .. . .. . .. . .. . . .. . .. . . Geo-visualisation of port accessibility for the Existing Conditions network (blue) and Multiplier Attachment network (purple) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Visualisation of the geographic location of the container ports considered in the study . . . .. . . . . . .. . . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . Geo-referred visualisation of Port Throughput . . . . . . . . . . . . . . . . . . . Geo-referred visualisation of Maximum allowed vessel Draft .. . . .. . .. . .. . . .. . .. . .. . . .. . .. . . .. . .. . .. . . .. . .. . . .. . .. . .. . . . Geo-referred visualisation of national GDP in 2010 . . . . . . . . . . . . .

91

92

94 95

95

96

105 106

107

111 113 118 118 119

List of Figures

Fig. II.5

xvii

Fig. II.11

Geo-referred visualisation of national Ease of Doing Business. Small values are associated with countries where it is easier to start up a business . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Geo-referred visualisation of national percentages of Internet Users . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Geo-referred visualisation of a sample of 22 container services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A simplified visualisation of the Vertical Interaction model. Depicted from top to bottom: bilateral trade flows, trade agreements, cultural links, common language, and container shipping network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Exports (left) and imports (right) in the SPICs by HS1 classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trade flows between countries in the period 2010–2012. SPICs countries are visualised in red. Linked colours identify country of origin, and link width is proportional to trade value . . . . . . . . . . Imports and exports for SPICs between 2010 and 2012 . . . . . . . . .

Fig. 9.1 Fig. 9.2 Fig. 9.3 Fig. 9.4

CHANNEL 1 schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CHANNEL 2 schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CHANNEL 3 schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CHANNEL 4 schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Fig. 10.1

Uganda paved road infrastructure. Source: GIS data provided by the World Bank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Uganda national road network. Source: Lawyer Glenn (2012) http://dx.doi.org/10.6084/m9.figshare.94407 . . . . . . . . . . . . . Dirt and gravel/paved roads in Uganda. (Red crosses indicate locations of intersections, blue lines are dirt roads and green lines are paved roads) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Location of farmers/outgrowers in the seven districts in the model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Simulation of daily temperatures on an hourly scale .. . .. . .. . .. . . Supply-demand schematic for Trader A . . . .. . . . .. . . .. . . . .. . . . .. . . . Depiction of the decision tree process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Wealth utility function U(C’) for B ¼ 100 . . . . . . .. . . . . .. . . . . .. . . . . CHANNEL 1 . . .. . . .. . .. . .. . .. . .. . .. . . .. . .. . .. . .. . .. . .. . . .. . .. . .. . .. . . The simulated chain can accommodate many agents, as shown by the inclusion of the firm SULMA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Simulation of itinerant trader in agent based model . . . . . . . . . . . . . Log-normal distribution of land . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Fig. II.6 Fig. II.7 Fig. II.8

Fig. II.9 Fig. II.10

Fig. 10.2 Fig. 10.3

Fig. 10.4 Fig. 10.5 Fig. 10.6 Fig. 10.7 Fig. 10.8 Fig. 10.9 Fig. 10.10 Fig. 10.11 Fig. 10.12 Fig. 11.1 Fig. 11.2

119 120 121

122 123

124 125 144 145 146 147 150 151

151 152 152 156 157 160 161 161 163 167

Uganda’s road infrastructure (www.roughton.com) . . . . . . . . . . . . . . 172 Exports by type expressed in tonnes through Entebbe International Airport. Source: BAA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173

xviii

Fig. 11.3 Fig. 11.4 Fig. 11.5 Fig. 11.6 Fig. 12.1

Fig. 12.2

Fig. 12.3 Fig. 12.4 Fig. 12.5

Fig. 12.6 Fig. 12.7 Fig. 13.1 Fig. 13.2 Fig. 13.3 Fig. 13.4 Fig. 13.5 Fig. 13.6 Fig. 13.7 Fig. 13.8 Fig. 13.9 Fig. 13.10 Fig. 13.11

List of Figures

Uganda’s road infrastructure embedded into the model . . . . . . . . . Road improvements at 100% level of RAI for districts Luwero, Mpigi, Masaka, Iganga, Mitiyana, Kamuli, and Mukono . . . . . . . . Improvements of Uganda Entebbe International Airport . . . . . . . . Comparison of the three scenarios in relation to the agents exiting the market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Map showing the districts of our analysis in dark colour: Luwero, Mpigi, Masaka, Iganga, Mitiyana, Kamuli, and Mukono . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . .. . . . . . . .. . . . . . .. . . . . . . Map of the control points indicating warehouse locations in the considered districts: Luwero, Mpigi, Masaka, Iganga, Mitiyana, Kamuli, and Mukono . .. . . . . . . .. . . . . . . .. . . . . . . .. . . . . . . .. . . . . . . .. . . . . . Study areas depicting ABM-generated locations of farmers (red), warehouses (green) and traders (dark green) . . . . . . . . . . . . . . Configured positions of 40 warehouses from the Existing Conditions scenario . .. . .. .. . .. . .. .. . .. . .. .. . .. . .. .. . .. . .. .. . .. . .. .. . . Warehouse use probability of 40 warehouses over a simulation period of 6 months. We show the five warehouses (W1–5) with the highest probability of being used . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Five optimal positions of warehouses obtained from the model . . . . . . Transport logistics cost impact on farmer inactivity . . . . . . . . . . . . . Subscriber and tele-density data, 2002–2012. Source: Uganda Communication Commission (2014) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fixed and mobile internet subscriptions, June 2013. Source: Uganda Communication Commission (2014) . . .. . . . .. . . .. . . . .. . . . Number of farmers/agents exiting the market in relation to Farmer Association and No Farmer Association . . . . . . . . . . . . . . . . . Total transport cost—Association vs. No Association . . . . . . . . . . . Average farm gate prices for Hot Peppers—Association vs. No Association . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Average farm gate prices for Chillies—Association vs. No Association . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Average farm gate prices for Matooke—Association vs. No Association . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Average farm gate prices for Okra—Association vs. No Association . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Average farm gate prices for Sweet Potatoes—Association vs. No Association .. . . .. . .. . . .. . . .. . .. . . .. . .. . . .. . .. . . .. . . .. . .. . . .. . . Amount of product sold to traders by farmers/outgrowers— Association vs. No Association . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Number of inactive traders/agents in relation to the implementation of GAP policy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

174 175 177 178

185

187 188 190

191 192 194 198 199 201 202 202 203 203 204 204 205 207

List of Figures

Fig. 13.12 Fig. 13.13 Fig. 13.14 Fig. 13.15 Fig. 13.16 Fig. 13.17

xix

Total transport costs in Uganda Shilling . . . . . . . . . . . . . . . . . . . . . . . . . . Average farm gate prices for Hot Peppers . . . . . . . . . . . . . . . . . . . . . . . . Average farm gate prices for Chillies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Average farm gate prices for Matooke . . . . . . . . . . . . . . . . . . . . . . . . . . . . Average farm gate prices for Okra .. . . . . .. . . . . . .. . . . . . .. . . . . .. . . . . . Average farm gate prices for Sweet Potatoes . . . . . . . . . . . . . . . . . . . . .

208 208 209 209 210 210

List of Tables

Table 2.1

Agents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

19

Table 3.1

Main merchandise trade indicators . .. . .. . . .. . . .. . .. . . .. . .. . . .. . .. .

32

Table 4.1

Summary of background context of trade issues in sea-locked countries, modelling approach (concepts and assumptions) and main outputs of the Multilayer model . . . . . . . . . . . . . . . . . . . . . . . . . . . .

40

Table 5.1 Table 5.2 Table 6.1 Table 6.2 Table 6.3 Table 6.4 Table 6.5 Table 6.6 Table 6.7 Table 7.1 Table 7.2 Table 7.3 Table 7.4

Goodness-of-fit indices with recommended values . . . . . . . . . . . . . . Ranking of ports by Port Attractiveness Index for SPICs and RIM ports . . .. . . .. . . . .. . . .. . . .. . . .. . . .. . . .. . . . .. . . .. . . .. . . .. . . .. . . .. . . Relevant topological properties of Existing Conditions shipping network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Major statistics of the Existing Conditions shipping network with economies of scale . . . . . . .. . . . . . .. . . . . . .. . . . . .. . . . . . .. . . . . . .. . . Relevant topological properties of Fully Connected shipping network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Major statistics of the Fully Connected shipping network with economies of scale .. . .. . .. . .. . .. . . .. . .. . .. . .. . .. . . .. . .. . .. . .. . .. . .. . Relevant statistics of the Multiplier Attachment network configuration using distance as an impedance factor to trade . . .. Major statistics of the Multiplier Attachment network with economies of scale .. . .. . .. . .. . .. . . .. . .. . .. . .. . .. . . .. . .. . .. . .. . .. . .. . Major traffic statistics and total transport cost for the three network configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Configurations of the Existing Conditions network . . . . . . . . . . . . . Throughput for SPICs ports in the Existing Conditions network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Configurations of the Fully Connected network . . . . . . . . . . . . . . . . . Throughput for SPICs ports in the Fully Connected shipping network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

59 61 69 72 74 76 78 80 81 88 88 92 93 xxi

xxii

Table 7.5 Table 7.6 Table 7.7 Table 8.1

Table 8.2

Table 8.3

Table 8.4

Table 8.5 Table 8.6 Table 8.7 Table II.1 Table II.2 Table II.3 Table II.4 Table II.5 Table II.6 Table II.7 Table II.8

List of Tables

Configuration of the Multiplier Attachment network . . . . . . . . . . . . Throughput for SPICs ports in the Multiplier Attachment network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Major statistics on the effect of trade liberalisation . . . . . . . . . . . . . Parameters estimates (β coefficient) and Pearson coefficient of correlation (in parentheses) using distance as impedance transport factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Parameters estimates (β coefficient) and Pearson coefficient of correlation (in parentheses) using transport cost (calculated using the Existing Conditions network with economies of scale) as impedance factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Parameters estimates (β coefficient) and Pearson coefficient of correlation (in parentheses) using transport cost (calculated using the Multiplier Attachment network with economies of scale) as impedance factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . OLS of Trade Model with Bilateral Trade (dependent variable) and independent variables: Total Export of exporting country (E), Total Import of importing country (I), distance (Dist), Cultural Linkage (Cult), Trade Agreements (Agreem) and Common Language (Lang) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model summary . .. . . .. . . .. . . .. . . .. . . .. . . .. . . .. . . .. . . .. . . .. . . .. . . .. . . Ranking of accessibility at national level (countries in the SPICs reported in bold) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ranking of accessibility at port level (ports in SPICs are reported in bold) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . List of model variables and collected data sets . . . . . . . . . . . . . . . . . . Container ports and countries examined in our case study . . . . . Relevant statistics of port variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . List of 14 container shipping companies in operation in the SPICs and Pacific Rim countries included in our study . . . . .. . . . HS1 and HS2 commodity classification . . . . . . . . . . . . . . . . . . . . . . . . . . Containerships deployed in the SPICS . . . . . . . . . . . . . . . . . . . . . . . . . . . Main technical characteristics of sample vessel used in our study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Estimated port fees per entry for a vessel with characteristics . . . .

96 96 98

102

103

103

103 104 108 109 114 117 117 121 123 127 129 130

Table 9.1

Summary of background context of trade issues in Uganda, hypotheses, and output and policy interventions . . . . . . . . . . . . . . . . 141

Table 10.1 Table 10.2

Product types and their estimations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Delivery transport modes and variables related to the commercial movement of our selected crops . . . . . . . . . . . . . . . . . . . . Market parameters showing price differentials between local markets, traders and exporters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Agents, variables and their values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 10.3 Table 10.4

164 165 166 166

Part I

The Challenge and the Science

Chapter 1

Enhancing Trade in Sea- and Land-Locked Countries

1.1

Objectives

Regional science offers a rich toolkit for analysis and policy development for substantive applications. In this book, the science is applied to the challenge for global investment agencies seeking to enhance trade in lagging regions. The challenge is particularly acute for sea-locked and land-locked countries and striking examples are provided by the widely scattered island countries of the South Pacific on the one hand, and Uganda on the other. What is needed is an in-depth analysis of trade routes as a basis for evolving policies to increase efficiency and reduce costs. The two case studies presented here provide an excellent illustration of the power of regional science, from assembling data bases in difficult situations, through to developing and applying models of the trade system. We apply network analysis, spatial interaction modelling and agent based models supplemented by appropriate statistical techniques. Our final results for both countries allow for substantive policy suggestions to be made. The South Pacific Islands are separated from each other and from larger mainland countries by vast distances, and as a result are served by a small number of shipping routes that are both expensive and too slow for the efficient export of, for example, agriculture products. What is shown here is that it is possible to model the trade patterns of these countries with a rich multi-layered spatial interaction model combined with network analysis. Our results offer up a tool with which to explore the possibility of a major hub port investment—and the optimal location for this— combined with a consolidation of shipping routes. The Uganda situation, however, demands a different approach. The challenge here is to improve the efficiency of the supply chain and to accelerate the speed of export flows as they leave their agriculture source en route, for example, to Entebbe Airport. Investigation shows a very complex supply chain, and an effective tool in this case is agent based modelling, which represents the key players in the system. In this case, a combination of transport infrastructure investment and effective © Springer International Publishing AG 2017 F.R. Medda et al., Collaborative Approach to Trade, Advances in Spatial Science, DOI 10.1007/978-3-319-47039-9_1

3

4

1 Enhancing Trade in Sea- and Land-Locked Countries

information exchange in the local economy are possible solutions for enhancing trade. A large number of elements tend to both impede and impact on trade and these elements often diffuse into ripple effects on the overall growth of a country. For instance, obstructions to trade may create country vulnerability, not only from an economic point of view but they can also contribute to food insecurity, worsen poverty, and exacerbate population displacement and conflict. If we examine with a wider angle of analysis the interdependency between economic development and long-distance trade, we can acknowledge that this is indeed a complex nexus entangled by many strands. Numerous theoretical models and applied analyses have contributed to the debate about the directionality, and importantly, the validity of the hypothesis that increased trade fosters growth, but results of these models are still inconclusive. The hypothesis of trade-led-growth has been tested extensively through sophisticated econometric approaches as with the use of co-integration and causality models (Sannassee et al. 2014; Wacziarg and Welsh 2008; Greenaway et al. 2002; Frankel and Romer 1999), suggesting that trade may contribute positively to growth, but research must also account for contingent relationships in trade and economic development (Rodriguez and Rodrik 1999, 2001). In this respect, and following a similar argument, Grossman and Helpman (1991) observe that many channels should be examined in order to determine the relationships between trade and economic growth; these include market size and access to markets, intensive and extensive margins for trade, sectoral composition, investment levels, and technological endowment, to name a few. Given this background, the main objective of the present work is to examine how some of these channels interact and connect in the search for new insights into how to spur growth by enhancing trade. The potential gains likely to arise through a better understanding of trade also motivate us to introduce new methodological paradigms in our trade analysis. In Part I the model dedicated to the South Pacific Islands is grounded in network analysis; the second model for Uganda is carried out in an application of agent based modelling. These methodologies have been selected because they are best able to examine the interdependency and the interactions present among the numerous trade elements. In both applications we ask the following questions 1. How can we improve collaboration in relation to national, regional and international trade markets? 2. How can we reduce delays and costs in the connection of the trade network? These two questions constitute the overarching framework of the case studies. However, the two case studies examine the issues from two different perspectives. In the case of the SPICs, we take an outward looking perspective where we analyse how to improve connectivity, particularly through a wider dissemination of information and lower price asymmetry in logistics. In the case of Uganda, we focus on an inward looking perspective by studying agent behaviour in the operations, trade and logistics networks in order to decrease costs through greater interactions among

1.2 The Current Situation

5

8 7 % GDP Growth

6 5 4 3 2 1 0 Fiji Fiji

Kiribati

Kiribati

Solomon Islands

Solomon Islands

Samoa Samoa

Papua New Guinea

Papua New Guinea

Tuvalu Tuvalu

Tonga Tonga

Uganda Uganda

Fig. 1.1 GDP annual growth: average over 10 years. Data Source: World Bank data set

agents, and testing how policy initiatives can impact on trade and ultimately on the growth of the region. In this introduction we review the economic and trade contexts of the South Pacific Islands and Uganda as they currently are, and then discuss in detail the common theoretical elements that tie together the two case studies.

1.2

The Current Situation

We begin by reviewing the main economic parameters of the two regions: the South Pacific Islands (SPICs) and Uganda. When looking at both regions’ GDP we can observe steady growth over the last 10 years (2004–2014) (Fig. 1.1). Over the 10 year period Uganda witnessed GDP growth, on average, of 6.7%. Whereas in the SPICs, only Papua New Guinea (PNG) achieved an average 4% GDP growth over the 10 year period. All of the other islands witnessed low economic growth: in Samoa, GDP was 3%, the Solomon Islands 2.7%, and Kiribati 2.1%. Tuvalu, Tonga and Fiji experienced the lowest rate of GDP growth over the same period among the SPICs. For instance, in Tuvalu one of the reasons for poor GDP performance is due to its being heavily dependent on food imports and wholly dependent on fuel imports. Tuvaluans work mainly in traditional agriculture and as fishermen; it is noteworthy that Tuvaluans especially are employed abroad as sailors and their remittances are major sources of income for their families back home. On the other hand, the highest economic performance of PNG is attributed to a relatively small formal economy consisting of workers engaged in mineral production and the manufacturing sector, and a large informal sector where subsistence farming accounts for the bulk of economic activity. There are clearly many possibilities for development which we outline in the next section.

6

1.3

1 Enhancing Trade in Sea- and Land-Locked Countries

Possibilities for Development

At first glance, the differences between the South Pacific Islands and Uganda loom larger than their similarities, but it is precisely the similarities which are the strongest threads in the present work. The first similarity in the two cases is their small status. In SPICs we are considering the trade flows among the relatively small islands in the South Pacific and the rest of the world, and in Uganda the main actors in the agriculture trade and supply chain (the focus of our analysis) are smallholder farmers. Beginning with the capacity to export, theory suggests that firms which export goods are usually larger and more productive than non-exporters, since fixed costs are determinant factors for export (Helpman et al. 2008; Bernard and Jensen 1995, 1999). In both of our case studies the entities under analysis are small, and although they are not firms, they nevertheless have difficulty gaining access to trade markets. One of the major obstacles to trade is the asymmetric information present in both case studies (World Bank 2013). Their small status as an exogenous attribute, along with information bias, are often difficult to overcome in the face of (prevailing) intense rivalry (ADB 2013) and cutthroat competition, which certainly do not foster growth or development. In this respect, although the Ugandan agriculture market and the SPICs shipping market may be considered as moderately contestable, collaborative activities are nonetheless not yet fully developed in either region. Collaborative actions are often studied as formal and informal agreements among trade entities in order to achieve better common resource management (Olstrom 2012). As we are considering demand-driven market processes rather than a supply-based framework, we focus on collaborative approach of trade as the primary way to overcome information asymmetry. Therefore, in our analyses we follow the marketing literature (Rushton et al. 2010), where common actions are seen as strategies for sharing know-how and best practice and as a way to pool resources, improve productivity and economies of scale, enhance bulk marketing of product due to stronger bargaining power, achieve penetration of new markets, and gain access to larger markets. We can identify through the presence of six Cs what we interpret as a collaborative approach to trade: 1) Cooperation. Cooperative organisations and collective actions, as in agreements, can be considered as the backbone for increasing trust among trade partners based on responsibilities, and also for reducing trade costs. 2) Competition. Trade liberalisation is seen here as a way to improve efficiency of local production and to expand market share. 3) Consolidation. From the point of view of the optimisation of the logistics chains and increase of economies of scale. 4) Coordination. Single small trade entities, such as the farmers in Uganda or the islands in the South Pacific, can pool their resources, thereby amplifying their results and decreasing transaction costs.

1.3 Possibilities for Development

7

5) Communication/Connection. Not only through the physical networks such as the transport network, but also through social media to achieve greater accessibility to markets, and to reduce asymmetry of price information. 6) Co-creation and co-sharing. Limited ownership can be detrimental to the improvement of trade; therefore, by innovating and adapting to context, it is possible to achieve vertical and horizontal integration of the supply chains. The literature provides a plethora of examples of successful collaborative trade actions, where for instance coordination improves competition in the market and stimulates trade; and in many cases connection satisfies both market objectives and social aims, thereby increasing cohesion and trust, and ultimately, welfare. However, the challenge of how to organise trade partners towards a collaborative approach is not simple when it comes to the establishment of rules, adjustment costs, compliance, and free-rider behaviours. These aspects are particularly significant when we consider competition and trade liberalisation and, for instance, the consequent reduction of trade tariffs. In this context it is worth mentioning that, while some modellers examine the complex problem of trade liberalisation from static frameworks, thus demonstrating the effects of increases in welfare, recent models of Stiglitz and Charlton (2010) indicate that under the endogenous paradigm it is possible to verify the relationship between free trade and economic growth. We examine collaboration in both case studies in order to overcome the aforementioned intrinsic ‘small status’: in Uganda we consider farmer associations, commercial farmers and traders, whereas in the SPICs study we analyse the impact of bilateral agreements and regional agreements for trade liberalisation. Pursued in practice, the collaborative actions of SPICs and Uganda involve investments and an array of structural reforms, processes requiring close attention to their implementation and outcomes over the short- and long-term. Our aim is to test how a collaborative approach can reduce asymmetry of information and augment the interactions, thereby leading to increased trade. Since in our case studies we examine the single entity, i.e. the island for the South Pacific and the farmer of Uganda as heterogeneous entities acting according to their maximisation of overall trade flow, the trade model thus becomes an economic model where macrobehaviours emerge from micro-motives (Schelling 1978). The second similarity between Uganda and the South Pacific Islands is the lagging character of both regions. We observe that for both cases, access to markets, particularly large international markets, is problematic (Kunaka 2010). From the access perspective, there is nearly universal agreement among scholars that transport cost and lack of logistics infrastructures act as trade impediments. The impact of distance (i.o.d) on trade is especially evident in the intensive margin of trade, where i.o.d tends to be largest in countries with low income per capita (Helpman et al. 2008). We examine the transport cost in the regions under consideration because domestic transport costs (i.e. within Uganda and SPICs, respectively) have been shown to represent a significant share of total trade cost.

8

1 Enhancing Trade in Sea- and Land-Locked Countries

The reduction of transport costs is achieved in our cases through different strategies with various policy implications. The first strategy involves the improvement of transport infrastructure, such as roads in Uganda, which determines a reduction of travel time and lower costs associated with time as it pertains to product delivery/logistics. Interestingly, Blyde and Iberti (2014) demonstrate for the case of Chilean farmers that by improving the condition of the road network, given that only 4% of the roads are in very poor condition, on average it would be possible to obtain an increase of 1.8% in exports. This achievement is significant when we consider that the average Chilean export growth rate is about 8%. The second strategies examined are the communication and the consolidation of trade flows, which respectively, aim to facilitate trade through the improvement of connectivity among the different members of the supply chain, and through the achievement of economies of scale in the forms of density and scope. We consider consolidation through the implementation of two sets of solutions, the first of which is exemplified in the Uganda case where for instance, the development of a network of warehouses can achieve an increase in consolidation and a decrease in information asymmetry (World Bank 2010, 2013). In our studies, communication, i.e. the ‘critical mass’ of the spread of information and of logistics facilities, is undoubtedly an important leverage to decrease the unit cost of production and thus stimulate trade. Moreover, we assume that the consolidation solution will initiate agglomeration effects where location is selected on the basis of minimised transport cost. Given that the trade entities (i.e. farmers/islands) are connected through interdependent relationships, they represent the (fundamental) backward and forward linkages to local agglomeration processes (Venables 1996; Krugman and Venables 1995). Agglomeration is seen in our cases as a reduction in logistics costs. The second set of solutions is tested in the South Pacific Island Countries where we are able to achieve consolidation through specific topological structures of the logistics chain. The different network structures are examined to verify how to stimulate consolidation and ultimately to increase trade. An important finding of our study is that certain types of structures in the trade networks can increase travel frequency and traffic (and consequently) improve connectivity among the trade entities. The hub-and-spoke network structure is given particular attention because although it requires ‘circuitous routing’ (Taaffe et al. 1996), it nevertheless meets our critical objective of implementing a collaborative approach to trade. However, it must be said that the hub-and-spoke structure cannot be recommended as a panacea for trade facilitation because as Wonnacott (1996) observed, rather than to increase trade, hub-and-spoke networks under bilateral agreements can inhibit the potential for regional growth, decrease the capacity to exploit the comparative advantages of trade partners, and hamper competition as well as economies of scale. The implications of this type of trade structure are that a hub will usually capture the largest percentage share of income and trade of a region, and due to high tariffs and inefficiencies, the spokes will incur losses that enlarge over time. Both cases are therefore used to also show how solutions to similar problems will need to be

References

9

adapted and contextualised to the situation at hand, and that there is no ‘one size fits all’ solution for implementing consolidation in trade. Our two common threads, small status and lagging economy, have allowed us to carry out robust analyses of the South Pacific Islands and Uganda. We have also strived to maintain comparability so that we may draw general lessons and extract policy implications from these studies for use in other contexts.

References Asian Development Bank (ADB) (2013) Key indicators for Asia and the Pacific 2013. Asian Development Bank, Mandaluyong City, Philippines Bernard AB, Jensen JB (1995) Exporters, jobs and wages in U.S. manufacturing, 1976–1987. Brookings Papers on Economic Activity, Microeconomics, Washington, DC Bernard AB, Jensen JB (1999) Exceptional exporter performance: cause, effect or both? J Int Econ 47:1–25 Blyde J, Iberti GA (2014) A better pathway to export: how the quality of road infrastructure affects export performance. Int Trade J 28(1):3–22 Frankel JA, Romer D (1999) Does trade cause growth? Am Econ Rev 89(3):379–399 Greenaway D, Morgan CW, Wright PW (2002) Trade liberalisation and growth: new methods, new evidence. J Dev Econ 67:229–244 Grossman GM, Helpman E (1991) Quality ladders in the theory of growth. Rev Econ Stud 58 (1):43–61 Helpman E, Melitz M, Rubinstein Y (2008) Estimating trade flows: trading partners and trading volumes. Q J Econ 123:441–487 Krugman P, Venables AJ (1995) Globalization and the inequality of nations. Q J Econ 110 (4):857–880 Kunaka C (2010) Logistics in lagging regions. World Bank, Washington, DC Olstrom E (2012) The future of the commons beyond market failure and government regulation. The Institute of Economic Affairs, London Rodriguez E, Rodrik D (1999) Trade policy and economic growth: a skeptic’s guide to the crossnational literature. NBER Working Paper no. 7081. National Bureau of Economic Research, Cambridge, MA Rodriguez E, Rodrik D (2001) Trade policy and economic growth: a skeptic’s guide to the crossnational evidence. In: Bernanke BS, Rogoff K (eds) NBER Macroeconomics Annual 2000, 15. MIT Press, Cambridge, MA Rushton A, Croucher P, Baker P (2010) The handbook of logistics & distribution management. Kogan, London Sannassee RV, Seetanah B, Lamport MJ (2014) Export diversification and economic growth: the case of Mauritius. WTO Publications, World Trade Organization Schelling R (1978) Micromotives and macrobehavior. Norton, New York, NY Stiglitz JE, Charlton A (2010) Fair trade for all: how trade can promote development. Oxford University Press, Oxford Taaffe E, Gauthier H, O’Kelly M (1996) Geography of transportation. Prentice-Hall, Englewood Cliffs, NJ Venables AJ (1996) Trade policy, cumulative causation and industrial development. J Dev Econ 49(1):179–197 Wacziarg R, Welsh KH (2008) Trade liberalization and growth: new evidence. World Bank Econ Rev 22(2):187–231 Wonnacott RJ (1996) Free-trade agreements: for better or worse? Am Econ Rev 86(2):62–66

10

1 Enhancing Trade in Sea- and Land-Locked Countries

World Bank (2010) Africa’s infrastructure: a time for transformation. World Bank, Washington, DC World Bank (2013) Uganda Report 77079-UG Diagnostic trade integration study. World Bank, Washington, DC

Chapter 2

Two Approaches to Modelling Trade

2.1

The Multilayer and Agent Based Model (ABM)

As mentioned in Chap. 1, our two case studies have some similarities as we will see, but they differ in scale. The SPICs case calls for an outward looking perspective and the scale is extensive; whereas Uganda is inward looking and demands a microscale of analysis. For the outward looking perspective of the South Pacific Islands we adopt network theory and a spatial interaction approach (Jackson 2008; Wilson 2008) through which we examine how trade is influenced by cumulative network interactions. The twofold interpretation of the cumulative network concept involves (1) representing the shipping network system; in this sense, elements are aggregated in categories whose combinations help to describe the complete network system; and (2) accounting for the vertical and horizontal interactions within the system. The inward looking perspective of Uganda uses agent based modelling (ABM) (Epstein 2006). ABMs are a class of model applied widely in agriculture economics (Happe et al. 2006; Berger 2001; Balmann 1997, 1999). The advantages of this methodology are that it allows us to employ data from different sources and different groups of heterogeneous agents who interact; exchange of information ensues and the adaptive behaviour responses to accommodate changes in the environment are taken into the model. Both of our selected modelling approaches, network theory and spatial interaction on the one hand, and agent based modelling on the other, can handle positive and negative feedback. Given that trade theory warns us of the importance of information asymmetry as an established factor that can hinder trade and limit growth, we also introduce a new concept based on information exchange. This new concept is operationalised through the multiplier attachment factor. In the next sections we set out the details of our two approaches.

© Springer International Publishing AG 2017 F.R. Medda et al., Collaborative Approach to Trade, Advances in Spatial Science, DOI 10.1007/978-3-319-47039-9_2

11

12

2.2

2 Two Approaches to Modelling Trade

The Multilayer Model

There are two objectives to be delivered by the Multilayer model: 1. To understand the connectivity and economy of scale in logistics services. We investigate dependencies and causalities among networks/determinants that inhibit trade facilitation in sea locked countries. This part of the analysis will accomplish the aforementioned objective by focussing on the connections of sea locked countries’ supply chains to international long-distance trade, international short-distance trade, and short sea shipping activities. In the case of sea locked countries, one of the Islands’ major problems is its isolated geographic condition. Towards this objective we analyse how these three major trade networks are connected and integrated, and study how to improve the trade capability and trade facilitation of the South Pacific Island Countries. 2. To identify performance of the trade market. SPICs are experiencing rapid population growth over the past decade (especially Solomon Islands and Papua New Guinea); nevertheless, transport services encounter different obstacles. Among the main hindrances are the high pricing structures for exports and imports, fragmentation of the different logistics services, and the low participation of the private sector in trade services and operations. We examine how the implementation of the optimization of the logistics system and the introduction of a new institutional system and trade policies (spatially-blind policies versus clustering and agglomeration of remote areas) can increase the diversification of export earnings. In keeping with the two objectives, our study of trade in sea-locked countries is grounded on the concept that trade flows are made up of integrated layered networks. Our model has three categories of layers: physical, economic, and sociological. Due to the interactive nature of the layers, and in order to analyse their interrelationships and impacts on international and national trade, we construct the Multilayer model based on two distinct perspectives: firstly, horizontal layers are examined through the use of complex network theory (Barabasi and Albert 1999; Watts and Strogatz 1998; Erdo˝s and Re´nyi 1959). Secondly, vertical integration is analysed by considering the dynamic spatial interaction approach (Wilson 1970, 2008). We are then able to investigate horizontal and vertical interdependency among layers, and the impact of shocks (natural and man-made), such as the introduction of a specific policy for the reduction of trade cost, maintenance of the infrastructure, and the provision of inter-island shipping services. The horizontal interrelations within each layer define the architecture of each network; (for example, in the physical network layer we examine the links between seaports which are generated by shipping liner services), and we study these through the application of network theory. The vertical interrelations are established according to the spatial interactions between the nodes belonging to various networks (i.e. the port of Honiara, a node in the physical network layer, is

2.2 The Multilayer Model

13

Fig. 2.1 Multilayer networks

vertically connected with the node representing the Solomon Islands in the economic network layer of trade agreements). This approach allows us to investigate the possible impacts/shocks transmitted across the considered layers. In the Multilayer model trade flow depends on the physical layers (e.g., quality and endowment of infrastructure, geographic position), economic layers (e.g., GDP, trade agreements, common currency) and sociological layers (e.g., Customs procedures and migration movement). Figure 2.1 gives the illustrative example of the European zone. A similar framework is adapted for the SPICs trade. Our study is structured under four main components, as depicted in Fig. 2.2: I. II. III. IV.

Data collection and analysis Horizontal Network analysis Vertical Interaction model Optimization of shipping trade cost

The data collected in component I represents the input to the Multilayer model. Statistical data analysis are used as a descriptive tool in this component. The Multilayer model consists of components II, III and IV. In component II, horizontal network analysis is applied to examine the topological structure of the layers (i.e. the existence of hub-and-spoke network architecture). The existing network architectures are investigated in relation to Barabasi-Albert model, Small-World model, and Random Network models in order to evaluate which configuration best fits each existing network. In component III we analyse the vertical interlinking between the layers. And lastly in component IV, we optimize trade shipping costs by taking into consideration the results of the previous components.

14

2 Two Approaches to Modelling Trade

Fig. 2.2 Outline of the applied methodology

The Multilayer model is built through three main components corresponding to: horizontal network analysis (component II), the vertical interaction model (component III), and the optimization of the shipping trade costs (component IV). In component II, we consider both static and dynamic networks. Static networks do not evolve over time but rather they maintain a fixed architecture. In this type of network we test the effects of introducing specific policies by adding topological modifications to the networks and thereafter evaluating the vertical consequences generated on the trade layer of the Multilayer model. Conversely, dynamic networks evolve and change their topology over time and the nodes in this type of network follow various possible patterns, such as preferential attachment (Barabasi and Albert model), clustering and agglomeration (Small World model), and spatially-driven solutions (Erdo˝s and Re´nyi class of models). The different dynamic models are used in order to study the topological configurations which best describe the existing networks applied to the SPICs. We are interested in exploring the effect of hubs and volume consolidation at different locations (nodes) in order to minimise trade costs and evaluate the impacts of hubs in different locations (inside and outside the SPICs). In component III we examine the vertical interactions between layers. Figure 2.3 visualises the vertical integration among layers; these are represented as functional relationships, where the upper layer level (trade) is the “dependent variable” and the next layer levels represent the set of associated “independent variables.” Any of these layers can in turn be considered as a “dependent variable” from its own set of “independent variables,” hence the term multilayering. All of the layers are potentially interactive. For example, if we consider the trade layer, this layer interacts with the underlying physical layer where a change (disruption or improvement) in the physical layer may generate impacts on the trade costs.

2.2 The Multilayer Model

15

Fig. 2.3 Functional relationships between layers

The vertical interaction model is able to investigate dependencies and causalities among layers/determinants that may inhibit trade, and evaluate the transmission of effects across layers. We later calibrate the Vertical Interaction model by using the analytical specification of the parameters, as in Wilson (1970). The rationale of the vertical interaction model is based on the entropy maximising theory (Wilson 1970). In the case of our study, we focus on aggregated trade flows rather than on the movement of individual goods. The entropy maximising theory assumes a stochastic formulation for trade problems in which the probability of a flow of trade pattern to occur is proportional to the number of ways in which goods can be arranged in aggregated flows to produce that pattern (Wilson 2000). The maximisation problem is subject to two constraints in order to restrict the number of assignments giving rise to a distribution. Thus, total origin and destination flows have to be equal to the sum of outward and inward flows of each spatial unit (i.e. regions, countries or intra-country units). In the case of trade, the constraints on trade flow Tij between unit i and j—Exports (E) and Imports (I)—can be written as X T ¼ Ei ð2:1Þ j ij X T ¼ Ij ð2:2Þ i ij We adapt the solution for the maximisation problem proposed by Wilson (1967, 2000) for the trade case as follows   Tij ¼ Ai Bj Ei Ij f β; costij

ð2:3Þ

where f(β, costij) is a general measure of impedance between i to j, costij is the shipping trade cost and β is a factor which considers the effect of trade costs on trade. Ai and Bj are balancing factors.

16

2 Two Approaches to Modelling Trade

The multilayer network introduced in Fig. 2.1 (vertical connection between layers) will be developed by considering physical, economic and sociological layers. The interactive nature of the model allows us to construct key performance indicators in order to measure accessibility and attractiveness among the SPICs. According to literature on spatial interaction models, balancing factors Ai and Bj can be interpreted as accessibility measures. Accessibility indicators assume the following form for the model in Eq. (2.3)   1 Ai ¼ P I j f β; costij B j j

ð2:4Þ

  1 Bj ¼ P Ei f β; costij i Ai

ð2:5Þ

As these indicators are dependent and estimated in an interactive procedure, they incorporate the competition on supplied opportunities and the competition on demand. We use these measures to benchmark the effects of new policies in the system. In component IV we optimize shipping trade costs subject to the conditions of the horizontal and vertical models (components II and III). The optimization of trade costs is central to the increase of trade volumes and economic growth in the SPICs. The optimization component of the multilayer method will be developed according to standard Origin-Destination models (Hillier and Lieberman 2005). We take into account that, as shown in the literature, trade costs are mainly influenced by transport costs (freight rates), consolidation processes to increase economies of scale, infrastructure quality, location, and vertical and horizontal organisation (Immaculada and Celestino 2005). For the case of the SPICs, the distance between ports is one of the major proxies to identify trade costs due to the impact of fuel charges on the transport cost (ADB 2007). Geographical distance is used to approximate the trade cost for the aggregated trade flow; and in the case of disaggregated goods, we need to adopt additional specifications based on available data. In relation to the trade flow cost optimization, we assume that differentiated products (i.e. products with a distinct identity and which the manufacturer has some control over pricing), have higher trade costs than homogeneous products due to the search cost buyer-supplier (Rauch 1999). We can conclude this section by observing that the Multilayer model is the culmination of three interdependent components which add flexibility and robustness to the overall model. The framework allows us to evaluate policy impacts and link the effects of physical, economic and sociological variables to trade. Horizontal and vertical analysis of the networks and layers (topological architectures) enables us to test integration policies among SPICs as well as the effect of new agreements, joint-ventures and associations. The optimization of trade costs clarifies the interlinking between the topological architectures.

2.3 The Agent Based Model

2.3

17

The Agent Based Model

Agriculture and agribusiness is seen as an opportunity to alleviate poverty and food security for the majority of Uganda’s people, while also capitalising on the untapped potential of the agriculture sector inside the country (World Bank, Growing Africa: Unlocking the Potential of Agribusiness 2013). To pursue this prospect, among other factors, the agriculture sector will need to compensate for the lower competitiveness caused by the landlocked geography and an inadequate inland market network. An inland market network can be described as a nexus of goods production, flows, and points of consolidation and exchange. We can therefore comprehend an inland market network as a spatial structure with components and processes that move goods and products from origin to destination. The main actors operating in the inland market network of our case study are farmers, traders, wholesalers, and exporters. According to Freund and Rocha (2010), in general in Africa, a 1-day increase in inland network trade delay reduces export values by about 17%. On the Uganda inland market network we can highlight three principal reasons for delay 1. Contracting, documentation, cross-border Customs. This is sometimes the longest delay and is a major barrier to private participation in trade; 2. Transit time. This represents the generalised transport cost; 3. Logistics chains and consolidation catchment areas. These delays are especially restrictive if they are associated with financial obstructions, and as such, there may be high rent costs for inventory, storage and security. Enhancing trade facilitation is therefore a primary concern in Uganda. An improvement in the inland market network is likely to boost exports and generate broad economic effects in terms of regional and national competitiveness, reduced poverty, and increases in production and supply chain performance. By addressing inland market network inadequacies we will be able to leverage the agriculture sector’s likelihood of attracting investment, and in so doing, stimulate economic growth (Bazinzi et al. 2013). Our main objective is to study how to reduce trade costs, and in particular transaction costs, in order to achieve trade facilitation and improve connectivity and coordination among farmers in Uganda. Many African countries, including Uganda, have preferential access to international corridors destined for the US and the EU; we will therefore focus on the factors that mainly impede Uganda trade and increase farmers’ transaction costs: the inland networks. Our study is structured under four specific steps, as follows 1. Identify main operating agents and geographic linkages and integration within the inland market network. Understanding of the market (production and trade) arises from our analyses of the inland market in its geographical context. In other words, knowing where farmers are located, the typology of production (e.g., large or small farms), and how consolidation centres and wholesale markets are distributed and linked.

18

2 Two Approaches to Modelling Trade

2. Understand behaviour of the different market agents. The next step is to work out how the different components and actors relate and interface with each other. We analyse the functions of individual markets (production, consolidation and wholesale) and the movement of fresh product through the inland market network. 3. Show direct and indirect interdependence between agents in relation to market performance. After the spatial links of the inland market network have been determined, we assess market performance. Market performance provides insights into the extent of goods distribution and its implications. To clarify market performance and connectivity, we examine market integration and in particular analyse how prices are transmitted among different market agents. 4. Evaluate resilience and adaptability to shocks or hazards. The three aforementioned steps allow us to clarify current and expected future conditions so as to provide a spatial baseline for decision makers. The outcomes stemming from the previous analyses dictate the structure for the final step of our study: the creation of a platform for decision makers to elaborate specific policy directives (e.g., set price ceilings to avoid speculative behaviour, incentivise cooperative production to increase economies of scale) from which we test potential scenarios and demonstrate the possibility of mitigating exogenous shocks and risks. With these four steps in mind, we present our methodological approach in the next section. Given the structure of our analysis, we propose the application of agent based modelling (ABM) techniques for this part of the work. ABMs are a class of computational models and simulations based on a large number of acting and interacting agents. The outcome of a simulation is principally determined by the agents at the micro level who interact among themselves and within their own environments. Hence, ABM provides a way to study the behaviour of agents as a result of a set of different scenarios, and thereafter to test for possible policies. One of the advantages of AB models is that they are robust and less demanding than econometric models in regard to availability of aggregate data; this makes them especially attractive for policy analysis in transition and in studies of developing countries. AB models also differ from other conventional simulation models grounded in mathematical programming because they can capture the interactions between actors explicitly (individual farms, for example), thereby allowing for the study of transaction and information costs. Second, these models can fully account for the spatial dimension: agricultural activities for instance, and hence the role of internal transport costs and the physical features of land. Consequently, the explanatory power of these models is in line with our research questions, since we aim to examine agent interactions in relation to the diffusion of market information, productivity performance and policy evaluation, and where space plays a decisive role. In order to study inland trade flow estimations and subsequently logistics cost assessment, we tailor a spectrum of agents who represent the main actors in the

2.3 The Agent Based Model

19

inland market network in the present study: (farmers, traders, wholesalers, exporters, governments, financial institutions, banks—plus other contractual arrangements that offer funds to farmers). Inclusion of the main actors in the supply chain network allows us to explore underlying dynamics. These diverse actors operate in the same context but have different goals, resources and available information. They also operate under different constraints; therefore, if the actors are exposed to identical shocks/incentives, given their differences and locations in various parts of the region, and their unique contexts, we expect them to respond differently from one another. In the next section we fully describe the agents who interact in the Uganda fresh product supply chain. Actors selected for inclusion in the ABM simulation enter at the moment when agriculture goods (in our case fresh product) leave the farmers, up until these goods reach the international market, consequently mapping part of the supply chain. Actors supporting and/or affecting the value chain without being an actual part of it, such as government and financial institutions, are also included as agents. The agent-actors are listed in Table 2.1. Whereas in Fig. 2.4, the entire schema is shown, starting with farmers at the base of the chain, moving upwards to wholesalers, and finally to exporters. The majority of agents facilitate the flow between the two endpoints, farmers and exporters. The population of farmer agents is assumed to produce within a certain time, and within a given statistical distribution of product volume. Also in the model there will be the possibility to select different types of product, and different lifetimes for each product, thus affecting its quality and selling price. In the model we distinguish between two groups: cooperative farmers and independent farmers. In the first group, the farmers are organised under cooperative agreements in order to increase the percentage of farmed land and they link directly with the traders, thus achieving low transaction costs. The second group is represented by the independent farmers who farm small quantities of land and link with different intermediary levels of traders in order to sell their product (See Fig. 2.4). Their product is available for sale to interfacing agents, e.g., Farm-level traders and Market traders, with whom the farmers negotiate the prices. Next, the resulting volume distribution of the fresh product within the market trader population is Table 2.1 Agents

Government Financial institutions Logistics Transport Service providers (LTSs) Exporters Processors Large suppliers Wholesalers Market traders Farm-level traders Farmer cooperatives Farmers

20

2 Two Approaches to Modelling Trade

Government

Financial Institutions

Exporters

Large Suppliers

Farm-level Traders

Farmer Cooperatives

Logistic Transport Services (LTS)

Wholesalers

Farmers

Fig. 2.4 Schema of a supply chain

shifted by the same negotiating process to other interfacing agents: wholesaler and large supplier agents, who in turn sell to exporter agents. The model assumes that interfacing agents conduct trade via a preferential attachment process. The flow of fresh product is depicted in Fig. 2.4, moving upwards from Farmers to Exporters (dark lines). The price of the product in the model is set by the exporter agents in a top-down fashion, following the same paths as shown in Fig. 2.4, but in the reverse direction (light lines). Exporters have an acceptable price range within which they buy fresh product. The Exporter’s accepted range of price is communicated via the negotiation process to the interfacing agents, who in turn adjust their price margins as the chain is traversed downwards to the farmer’s price negotiations. The aim of the model is the optimization of the supply chain; we therefore assume constant demand and constant supply in the first instance. Different types of demand are involved: local demand, distribution within the country and international demand. Of course, international demand requires shipping through the airport, which in turn affects the supply chain. Interaction between layers also takes place across the layers. Policies are implemented by the Government agent, while financial institutions may give loans to farmers and other agents. As shown in Fig. 2.4, we identify three macro layers of agents, which can be aggregated; represented from the top of the chain to the

2.3 The Agent Based Model

21

bottom they are Main, Middle and Bottom, respectively. All the agents, with the exception of exporters, Government and financial institutions, belong to the Main set; wholesalers, large suppliers and farmer cooperatives belong to the Middle set, which identifies the interfacing agents cooperating or having agreements; and farmers, market traders and farm-level-traders belong to the Bottom set, which represents the single interfacing agents. Processors and Logistics Service Providers, who constitute the Logistics Transport Service (LTS) can be contacted by any of the actors. Warehouses are located at different places around the country in accordance with the density of farmers. We assume that farmers can stock a finite amount of product. At each stage from top to bottom a logistics service cost is introduced which reduces the price range between each layer. Thus, results have to be averaged in a Monte Carlo simulation. A price is established from the top to the bottom, but when the price at the bottom is not above the production cost, the price is rejected and the negotiation process starts over again in a movement from top to bottom, with a biased new price proposal at the top. All the prices in the model at any stage depend on a supplydemand function. The wealth of the agents will be generated according to the distribution, location, and quality of fresh product at the start of the simulation. Each agent in the supply chain can be described briefly. Farmers and farmer cooperatives As producers, these are the essential agents. Farmers are divided in two groups: individual farmers who represent the small farmers, and farmer cooperatives operating under agreements and/or with large amounts of arable land. Farmer and farmer cooperatives have a production cost that is a direct function of the volume. Their price depends on the quality of the fresh product and this will be taken into account in the model. Farm-level traders and market traders These agents act as aggregator of supply and as traders. These pay fast, and buy small amounts of fresh product in order to sell it again to the next layer for a higher price. Large suppliers and wholesalers Large suppliers are in direct contact with exporters. Their role is to aggregate product from the supply chain for an agreed price. These agents pay with a delay. Logistics Transport Service (LTS) The Logistics Transport Services (LTSs) move fresh product from place to place. Included in the model will be a function which gives an approximate time and cost of shipping. For the sake of simplicity, processors and logistics services are twinned and we consider them as a unique agent; they are able to interact with any other agent in the system and the timing and cost depend on the distance, for a higher price. Financial institutions Financial institutions have the ability to provide loans to agents. Since any agent has a certain amount of funds, by applying to financial institutions, agents can skip steps in the supply chain in order to maximise the selling price of fresh product.

22

2 Two Approaches to Modelling Trade

Government The government will implement policies. Possible policies include the implementation (or the reduction) of tariffs in some steps of the supply chain. Government interacts with financial institutions. Exporters Exporters fix the demand and propose prices to the top actors in the supply chain. Their function is to adapt the price to the outcome of the negotiation process in the supply chain (top-to-bottom approach). The interactions between the different agents and their behaviour are what matters most in our case study context. Starting from the government and financial institutions, policy implementations are top-down implemented: they can affect parameters and structural properties of the ABM during the simulations (by reflecting the impact of a reduction or increase of Customs tariffs). The Logistics Transport Service (LTS) agent introduces prices to ship tonnes of product between different points in the country, and thus its behaviour is only limited to establishing a price (although Government could affect this). Agents will have a position and an associated distance D(i,j) between agent i and j. The LTS agent will be a function (price) pt(i,j,t,k), where i and j are the agents, t the tonnes to be moved and k the type of transport (boat, air, rail, and road). This approach, rather simplified, is different from other ABMs in the literature (Holmgren et al. 2012; Roorda et al. 2010). Changes in transport infrastructures, such as the construction of a new road, will also affect the function pt. This function pt can be affected by economic and policy initiatives and its details will be built on throughout the work. Intermediate layers between demand and supply can be modelled to some extent by Walras markets (Gee 2004). Farm level traders act as Walras auctioneers between market traders and farmers (Gee 2004). By virtue of the model’s construction, only a few farmers are in contact with market traders, and this is established through a probability distribution that is flexibly chosen, in that the Walras equilibrium is formed between market traders and farmers, where the price change is determined by the excess demand dp ¼ ED dt

ð2:6Þ

with ED being the excess demand. Such a model is along the lines of minimal agent based models for the origin of self-organisation in markets (Alfi et al. 2009). However, price information asymmetry takes place in the following sense. We assume that the farmers are connected through informal social networks. We model the spreading of information (i.e. the product price) as an epidemic spreading of information on this social network (Daley and Kendall 1965; Moreno et al. 2004). However, farmers tend to sell at the best price that they are aware of, to buyers whom they know, and for this reason price information is asymmetric. The informal social network is modelled as a biased Barabasi-Albert model, where connections are established on the basis of the distance D(i,j) between two agents. The propagation of information is modelled in two ways in order to test the robustness of the

2.4 Multiplier Attachment

23

model (Daley and Kendall 1965; Moreno et al. 2004); first, by using a DaleyKendal model (1965) and then as a Markov process (Moreno et al. 2004).1 The model includes the exit/entry of agents in Eq. (2.6) and, market traders will also act as auctioneers between farmers and market traders. By following the diagram in Fig. 2.4 any intermediate agent may be considered as an auctioneer between two layers. Each layer has a market dependent price depending on the excess demand, as from Eq. (2.6). The effect of shocks on the infrastructure, and then on price and warehousing, is then evaluated accordingly. The whole infrastructure is modelled, for simplicity, as a single agent. This macro-agent is providing a service, which in turn is subject to the Walrasian supplydemand law. The variation in the price of the service provided by the Logistics agent affects the market process directly (growth or decline). As several factors exogenous to the market under consideration might impact the price of the service, part of the model is able to account for such factors while keeping the overall model as simple as possible.

2.4

Multiplier Attachment

In order to take into account in both models the role of information and online communications, we introduce the concept of multiplier attachment. The Barabasi and Albert (1999) preferential attachment concept is grounded in the assumption that an agent will select another agent with whom to connect based on the number of connections. In other words, Agent A will connect with B rather than C, if B has more connections than C, and this will be the mechanism for building the network structure (shipping network and farmer social network). The implications of this process are that 1) Agent B gets more and more connected and thus increases its influence among agents and subsequently becomes a hub; 2) Agent A may acquire higher utility by connecting with a highly connected agent, since it can decrease its transaction costs (Schelling’s assumption) and implement economies of scale and density; 3) At this point we are not certain whether we have obtained a solution with a higher level of competition in the market, i.e. if we have a Pareto efficient 1 If A is the adjacency matrix of the informal social network of farmers, the Markov matrix of the spreading of information on the social network is given by the row normalised matrix N P Aij ¼ Aij = Aij . If information is withheld with a probability q, the Markov matrix is:

j¼1

Mij ¼ qI ij þ ð1  qÞAij Layers above the Farmers are randomly connected with a few of the farmers. In order to find the demand of each, we connect higher layers to lower layers at a probation price pp vector.

24

2 Two Approaches to Modelling Trade

allocation; we may instead be moving towards an oligopolistic structure of the trade market. The concept that we propose here is a new type of attachment between agents, which we call multiplier attachment; it differs from the Barabasi and Albert preferential attachment. In our case, Agent A at time T1 has specific knowledge (e.g., prices, trade flows, etc.). His or her information is partial and asymmetric in relation to the main principal (exporter, shipping companies). Agent A will connect with Agent B based on the expectation to acquire more knowledge than in his present knowledge state. We assume that the acquired information is still partial but is nevertheless additional (i.e. Agent B does not have complete information) and above all, the information is not corrupted (i.e. every connection diminishes the asymmetry and thus reduces the noise of cheap talk). However, the acquired information occurs at intervals over time and it is this final assumption that matters most for policy implications. We can obtain a lagged measurement adjustment of the information to give us an idea of how government can intervene to reduce the lag period (implementation of cell phone networks, facilitation of trade agreement, etc.). The concept of multiplier attachment defined above is applied in the two case studies and with the two methodologies in order to verify its robustness and its capacity as an explicative factor for better understanding of trade growth. In both cases the concept of multiplier attachment is reinterpreted and tailored to the specification of the data and the given objectives.

References Alfi V, Cristelli M, Pietronero L, Zaccaria A (2009) Mechanisms of self-organization and finite size effects in a minimal agent based model. J Stat Mech. arXiv:0811.4256 Asian Development Bank (ADB) (2007) Ocean voyages: shipping in the Pacific. Asian Development Bank, Mandaluyong City, Philippines Balmann A (1997) Farm-based modelling of regional structural change: a cellular automata approach. Eur Rev Agric Econ 24(1–2):85–108 Balmann A (1999) The dynamics of structural change and farmers’ income in a maturing sector. Working Paper. Humboldt University of Berlin Barabasi AL, Albert R (1999) Emergence of scaling in random networks. Science 286(5439):509–512 Bazinzi N, Mangeni P, Nakabuye Z, Brendah A, Agasha E (2013) Transaction costs and outreach of microfinance institutions in Uganda. Issues Bus Manag Econ 1(6):125–132 Berger T (2001) Agent based spatial models applied to agriculture: a simulation tool for technology diffusion, resource use changes and policy analysis. Agric Econ 25(2–3):245–260 Daley DJ, Kendall DG (1965) Stochastic rumours. IMA J Appl Math 1(1):42–55 Epstein JM (2006) Generative social science: studies in agent based computational modeling. Princeton University Press, Princeton Erdo˝s P, Re´nyi A (1959) On random graphs. Publ Math 6:290–297 Freund CL, Rocha N (2010) What constrains Africa’s exports? World Bank Policy Research Working Paper, no. 5184. Social Science Research Network, Washington, DC Gee EJD (2004) Agent based modeling of non-Walrasian markets with entrance and exit of agents. Williams College, Williamstown, MA

References

25

Happe K, Kellermann K, Balmann A (2006) Agent based analysis of agricultural policies: an illustration of the agricultural policy simulator Agripolis, its adaptation and behavior. Ecol Soc 11(1):49 Hillier FS, Lieberman GJ (2005) Introduction to operations research, 8th edn. McGraw Hill Education, New York Holmgren J, Davidsson P, Persson JA, Ramstedt L (2012) TAPAS: a multi-agent based model for simulation of transport chains. Simul Model Pract Theory 23:1–18 Immaculada MZ, Celestino SB (2005) Transport costs and trade: empirical evidence for Latin American imports from the European Union. J Int Trade Econ 14(3):353–371 Jackson MO (2008) Social and economic networks. Princeton University Press, Princeton, NJ Moreno Y, Nekovee M, Pacheco AF (2004) Dynamics of rumor-spreading in complex networks. Phys Rev 69:66–130 Rauch JE (1999) Networks versus markets in international trade. J Int Econ 48:7–35 Roorda MJ, Cavalcante R, McCabe S, Kwan H (2010) A conceptual framework for agent based modelling of logistics services. Transp Res E 46:18–31 Watts D, Strogatz S (1998) Collective dynamics of ‘small-world’ networks. Nature 393(6684): 440–442 Wilson AG (1967) A statistical theory of spatial distribution models. Transp Res 1:253–269 Wilson AG (1970) Entropy in urban and regional modelling. Pion, London Wilson AG (2000) Complex spatial systems: the modelling foundations of urban and regional analysis. Pearson Education, Singapore Wilson AG (2008) Boltzmann, Lotka and Volterra and spatial structural evolution: an integrated methodology for some dynamical systems. J R Soc Interface 5:865–871 World Bank (2013) Uganda Report 77079-UG Diagnostic trade integration study. World Bank, Washington, DC

Chapter 3

The Current Situation and Future Challenges

3.1

Core Statistics

In Chap. 1 we offered a preliminary review of the main economic parameters of the two regions under examination by focusing on GDP. Here, we extend the analysis beginning with the employment rate in Fig. 3.1. We note that, in relation to the South Pacific Island countries, Fiji has the lowest employment rate compared to other SPICs and also to Uganda. The Uganda employment rate is significantly high, and according to World Bank data, nearly 75% of the population is employed (Fig. 3.1). Uganda’s high employment rate is followed by PNG, with an employment to population ratio above 70%. This data confirms the trend in GDP growth experienced by PNG in the last 10 years. The Solomon Islands reports a lower (65%) employment ratio, which is mainly the result of low GDP growth rates and a large subsistence agriculture sector. In both regions the employment and GDP trends are linked with the structure of the economies and their core sectors. Agriculture is the main sector in the economies of both the SPICs and Uganda (Fig. 3.2). For instance, in Uganda 36% of the total value added of the GDP comes from the agriculture sector. Similarly, in the South Pacific region, the GDPs of the Solomon Islands and PNG in particular rely largely on agriculture production with at least a 35% share of the total value added created between them. In the two economies the main crop productions, ranked by value, are: coffee, oil, cocoa, copra, tea, rubber, and sugar. Whereas New Caledonia has the lowest ratio of agriculture value added to GDP because the economy of the island revolves around the mining industry, mainly nickel extraction. We next focus on the export and trade capacities of our two regions. In the SPICs, Samoa and Tuvalu, exports are predominantly represented by manufactured goods (Fig. 3.3). The main exported goods are derived from coconut and account for the largest share of the export of the two countries. Kiribati, Vanuatu and Tonga exports are mostly agriculture and fishing products. Due to their limited exports and the rising cost of imports, the aforementioned countries run deep trade deficits. © Springer International Publishing AG 2017 F.R. Medda et al., Collaborative Approach to Trade, Advances in Spatial Science, DOI 10.1007/978-3-319-47039-9_3

27

Employment to Population ratio %

28

3 The Current Situation and Future Challenges 85 80 75 70 65 60 55 50 45 2000

2001

2002

Fiji

2003

2004

2005

2006

Solomon Islands

2007

2008

2009

Papua New Guinea

2010

2011

2012

Uganda

% of GDP

Fig. 3.1 Employment to population ratio. Data Sources: World Bank data set, year 2012 45 40 35 30 25 20 15 10 5 0 Fiji

Kiribati

New Solomon Caledonia Islands

Samoa

Papua New Guinea

Tuvalu

Tonga

Uganda

Fiji

Kiribati

New Caledonia

Solomon Islands

Samoa

Papua New Guinea

Tuvalu

Tonga

Uganda

Fig. 3.2 Agriculture value added to GDP ratio. Data Source: World Bank data set, year average (2000–2012)

Their high dependence on imported staple foods (such as rice and flour), coupled with soaring global food prices poses a threat to food and nutrition security. Despite its significant contribution to agribusiness, PNG is nevertheless regarded as a mining country. But this is to be expected, since more than 50% of GDP and 30% of its total export revenue comes from minerals. Figure 3.3 shows that, in a short period of time, the mining and petroleum sectors in PNG have added even more to GDP growth, peaking in 2012 at a level close to 30% of GDP. Papua New Guinea and Samoa export more manufactured and fuel and mining products than agriculture goods. Over the last 8 years, PNG in particular has witnessed a steady increase in mining and extracting activities; projects under development at present involve Liquid Natural Gas (LNG) extraction, and nickel and gold mining projects (Hidden Valley gold and silver mine and Ramu mine for nickel-cobalt). Other long-term projects include petroleum fields (Kutubu, Moran and Gobe in particular) and copper mining (Ok Tedi Mine).

3.1 Core Statistics

29

100

Percentage of Total Exports

90 80 70 60 50 40 30 20 10 0

Agricultural products

Fuels and mining products

Manufactures

Fig. 3.3 Export per sector of the economy (to GDP ratio %). Data Source: World Bank data set, year 2012

90 80 70 60 50 40 30 20 10 0

Samoa to Australia

Vanuatu to Malaysia

Papua New Guinea to Australia

Fiji to Australia

Tonga to New Zealand

Tuvalu to Fiji

Kiribati to Morocco

Uganda to EU

Fig. 3.4 Main international trading partners of SPICs and Uganda. Data Source: World Bank data set, year 2012

Uganda maintains a strong agriculture sector and, as Fig. 3.3 also shows, 52% of Uganda’s total exports consist of agriculture goods. Traditional food commodities dominate Uganda, in 2008 exports of coffee accounted for over one-fourth (26%) of the total value of its exports. Uganda’s increase in manufactured exports is due mainly to its export of electrical equipment to the United Arab Emirates. Figure 3.4 depicts the main international trade partners for our two study regions. For instance, thanks to newly found liquid natural gas (LNG) resources,

30

3 The Current Situation and Future Challenges 60 50

% of GDP

40 30 20 10 0 Fiji

Kiribati Fiji

Kiribati

Solomon Islands Solomon Islands

Samoa Samoa

Tonga Tonga

Uganda

Uganda

Fig. 3.5 Exports of goods and services (% of GDP). Data Source: World Bank data set; year 2012 600 500 400 300 200 100 0 Fiji

Kiribati

New Caledonia

Solomon Islands

Export Volume Index

Samoa

Papua New Guinea

Tonga

Uganda

Export Value Index

Fig. 3.6 Export indexes: volumes and values. Data Source: World Bank data set, year 2012

PNG has established a flourishing business relationship with Australia and boosted the level of international trade between the two countries. As PNG’s main trading partner, Australia receives 35% of PNG’s exported goods. To develop further economic collaboration, Australia has invested A$19 billion in PNG in gold mining, oil and LNG. Key Australian companies in the mining and petroleum sector include Santos, Oil Search Ltd. and Highlands Pacific Ltd. When we look at export levels in relation to GDP (Fig. 3.5), among the SPICs, Fiji has the major revenues from goods exported and also has a very high level of export value to GDP ratio (over 50%). In second and third place, respectively, Solomon Islands and Samoa share roughly the same pattern by exporting more than 30% of goods produced. Conversely, when we examine Uganda, its export to GDP ratio is 20%, indicating a relatively low level of international trade volumes. In Fig. 3.6 Samoa takes the lead in exports, both in volumes traded and values. Only Uganda shares a similar propensity to export goods, and is shown in Fig. 3.6 to have high value compared to volume; as such, Fiji, Tonga and New Caledonia are less able to export their goods and services, their exports are mostly agriculture products with low value added (sugar, banana, vanilla, timber, fish, molasses, mineral water, and coconut oil).

3.1 Core Statistics

31

140

% to GDP ratio

120 100 80 60 40 20 0 2000

2001 Fiji

2002

2003

2004

2005

Papua New Guinea

2006 Tonga

2007

2008

2009

Uganda

2010

2011

2012

Samoa

Fig. 3.7 Merchandise trade to GDP ratio (%). Data Source: World Bank data set, year 2012

In particular for the SPICs, merchandise trade is a vital activity. Many small countries rely heavily on trading goods and raw materials. The main example is seen in the case of Papua New Guinea and Fiji, where merchandise trade plays a pivotal role in the local economies that are highly dependent on external trade (Fig. 3.7). The Samoa Islands also depend on merchandise trade, although their share of exports decreased markedly between 2008 and 2010 due to the economic downturn. Moreover, Samoa has relatively good indicators for trade (Table 3.1). In comparison with the other countries, Samoa requires only 22 days to export goods and 28 days to import (Table 3.1). This puts the country top in the table in terms of effective trade operations in the SPICs and in comparison with Uganda. Table 3.1 does however point up the low performance of Uganda’s economy in international trade. The cost to import/export goods is very high compared to the other countries in the SPICs region (US$2800 per container); the sea-locked characteristic of Uganda also impacts negatively on the cost for it to export goods (the average for the countries investigated is US$1000). The difficult merchandise trade situation is confirmed in the available data on export/import times. Uganda requires 30 days to export its products (against an average of 21 days) whereas Uganda takes 33 days to import foreign goods (the average for other countries is 25). A constraint affecting the speed of merchandise trade in Uganda is the number of documents needed to carry out merchandise trade: Uganda must have ten documents to export or import goods, whereas Samoa only requires six. Furthermore, Uganda ranks 164th in the World Bank trading across border index, one of the lowest positions. On the other hand, Tonga is able to over-perform in the trading across country ranking in 63rd position, and has the cheapest cost to import and export goods. The cost to Tonga to import goods is US$490 against an average of US$1141 and the cost to export goods is US$505 compared to an average of US $1047. Let us next review the context of the SPICs and Uganda regions by comparing the attractiveness of the local economies using data on Foreign Direct Investment

32

3 The Current Situation and Future Challenges

Table 3.1 Main merchandise trade indicators Indicator Trading across borders (rank) Documents to export (number) Time to export (days) Cost to export (US$ per container) Documents to import (number) Time to import (days) Cost to import (US$ per container)

Uganda 164 7 30 2800 10 33 3375

Samoa 58 5 22 490 6 28 575

Kiribati 77 6 20 870 6 21 870

Palau 96 5 26 720 9 31 680

PNG 134 7 23 1149 9 32 1250

Tonga 63 6 22 505 6 25 490

Data Source: World Bank—Doing Business, year 2012 12

% FDI to GDP ratio

10 8 6 4 2 0 Fiji

Fiji

Kiribati

Kiribati

Solomon Islands

Solomon Islands

Samoa

Samoa

PNG

PNG

Palau

Palau

Tonga

Tonga

Uganda

Uganda

Fig. 3.8 Foreign Direct Investment (% of GDP). Data Source: World Bank data set, year 2012

(Fig. 3.8). Foreign Direct Investments (FDI) are the net inflows of investment to acquire a lasting management interest (10% or more of voting stock) in an enterprise operating in an economy other than that of the investor. It is the sum of equity capital, reinvestment of earnings, other long-term capital, and short-term capital as shown in the balance of payments. This series shows total net, that is, net FDI in the reporting economy from foreign sources, minus net FDI by the reporting economy to the rest of the world. Data are in current US dollars. In recent years, Fiji and the Solomon Islands have successfully attracted net positive investment inflows. In particular, they have forged long-term business collaborations with New Zealand and Australia, thus propelling the level of investments in SPICs countries and boosting their economic growth. According to the statistics, Uganda performed well over the last 10 years and has also attracted a noteworthy level of investments from abroad. Of the East Africa countries, Uganda and Tanzania attracted the most FDI in 2012: “East Africa attracted foreign direct investment (FDI) inflows of $3.9b in 2012, a $1.8b increase from $2.6b in 2011. With a combined total inflow of $3.4 billion, the two main energy-rich countries of

3.2 Future Challenges

33

60%

% to GDP Ratio

50% 40% 30% 20% 10% 0% Fiji

Kiribati

Solomon 1990-1999

Samoa 2000-2006

PNG

Tuvalu

Tonga

Uganda

2007-2013

Fig. 3.9 Net bilateral aid flows to GDP ratio. Data Source: World Bank data set. QASER Lab elaboration

Uganda and Tanzania received 90% of the investment inflows into the region” (State of East Africa Report 2013). The final comparison of the two regions under scrutiny is the examination of amount of aid received by the SPICs and Uganda. We have collected data for net bilateral aid flows (from DAC donors, total in current US$) and Gross Domestic Product (current US$) from World Bank data sets. The calculation to create a homogeneous time series was carried out by dividing net bilateral aid flows by GDP. Three time periods are involved: the first period evaluates the situation of aid between 1990 and 2000; the second time window ranges between 2000 and 2006 (pre-crisis); and the third analyses the situation during the recent financial and economic crisis. As shown in Fig. 3.9, the situation has maintained a stable level over the three periods; in one case we recognise an upward trend in the aid inflows (Solomon), whereas in only one case a downward trend is found (PNG). Over the three periods under examination Tuvalu has received the majority of aid, followed by Kiribati and the Solomon Islands. Uganda, Fiji and PNG have benefitted least from foreign aid, and Fiji in particular has the lowest aid to GDP ratio. During the period 1990–1999, Kiribati and Samoa received one of the highest ratios of aid to GDP, but this number diminishes after 2000.

3.2

Future Challenges

The challenge in both cases is straightforward: to improve the logistics of the trade chains in ways that accelerate exports and hence stimulate economic development. In the SPICs, we explore transport cost reductions through port investment: huband-spoke models, the consolidation of shipping routes and trade agreements. In the

34

3 The Current Situation and Future Challenges

Uganda case, we focus on institutional consolidation and improved information flows. Although the cases are quite distinct and need different kinds of models as analytical bases, the underlying challenges are similar. In Part II, we tackle the SPIC’s challenges and in Part III, Uganda’s. Next in Chap. 4 we thoroughly describe the Multilayer model for the SPICs analysis.

Reference State of East Africa Report (2013) One people one destiny: the future of inequality in East Africa. Society for International Development

Part II

The South Pacific Island Countries (SPICs)

Chapter 4

The Multilayer Model for Sea-Locked Countries

4.1

Methodology and Study Structure

Our model of trade in sea locked countries is grounded on the concept that the flow of trade is characterised by integrated layers of networks. The Multilayer model is composed of three categories of layers: physical, economic and sociological, which will determine the main outputs of the analyses. These layers are interactive, so to comprehensively study their interrelationships and impacts on international and national trade, we have to construct the model based on two distinct approaches. First, horizontal layers are studied through the use of complex network theory (see inter alia, the works of Barabasi and Albert 1999; Watts and Strogatz 1998; Erdo˝s and Re´nyi 1959). Thereafter, vertical integration is analysed through the spatial interaction approach of Wilson (1967, 1970). In this way we are able to investigate both horizontal and vertical interdependency among layers and determine the impacts of natural and man-made shocks, such as (for the SPICs case study) the introduction of a specific policy for the reduction of transport cost, the maintenance of infrastructure and the provision of inter-island shipping services. Horizontal Interrelations within each layer define the architecture of each network; (for example, in the physical network layer we examine the links between seaports that are generated by shipping liner services). We study these links in an innovative way by extending the preferential attachment concept of Barabasi and Albert (1999) using multiplier attachment where a node connects to other nodes based on the expectation of acquiring more knowledge/information than in its state of play. Succinctly put, in the case of trade networks, multiplier attachment involves three sets of variables in establishing connections between nodes: connectivity, geographic distance and the expectation of gaining an improved strategic position. Vertical Interrelations are established according to observed spatial interactions between nodes belonging to different networks (i.e. the port of Honiara, a node in the physical network layer, is vertically connected to the node representing © Springer International Publishing AG 2017 F.R. Medda et al., Collaborative Approach to Trade, Advances in Spatial Science, DOI 10.1007/978-3-319-47039-9_4

37

38

4 The Multilayer Model for Sea-Locked Countries

Fig. 4.1 Outline of the applied methodology

the Solomon Islands in the economic network layer of trade agreements). This approach allows us to study the possible impacts/shocks transmitted across the considered layers. In the overall structure of the Multilayer model, the flow of trade depends on physical layers (e.g., quality and endowment of infrastructure, geographic position), economic layers (e.g., GDP, trade agreements, common currency) and sociological layers (e.g., customs procedures and migration). We have configured four main components of our study of trade in the South Pacific Island Countries (SPICs), shown in Fig. 4.1 as I. II. III. IV.

Data Collection and Analysis Port Attractiveness Analysis Horizontal Network Analysis Vertical Interaction Model

Component I represents the input of the Multilayer model. Statistical data analysis is used as a descriptive tool in this component. Component II explores the determinants of port attractiveness in the South Pacific region. Horizontal Network Analysis (Component III) is applied to examine and optimize the topological structure of the SPICs trade network (i.e. the presence of hub-and-spoke network architecture). The state-of-play shipping network is studied using complex network analysis; we propose a new architecture based on multiplier attachment to optimize transport cost. Thereafter in Component IV, we analyse vertical

4.3 Variables and Multilayer Configuration

39

interlinking between layers in order to verify our overall aim of trade facilitation in the SPICs. The elements of the Multilayer model, which are necessary to test it and formulate policy recommendations, are presented in Sects. 4.3–4.5 and the data requirements are presented in the Appendix to Part II. In the next section we discuss the specifications and hypotheses we test using our model.

4.2

Multilayer Model: Specifications and Hypotheses

As discussed above, the model is developed through three interlinked components corresponding to Port Attractiveness analysis (component II), Horizontal Network analysis and modelling (component III), and Vertical Interaction model (component IV). The model aims to verify the following hypotheses: H1 Can the structure of the shipping network stimulate the consolidation of cargo and decrease transport cost? (Chap. 6) H2 Do trade coordination initiatives provide greater long-term economic growth for the SPICs? (Chap. 7) H3 How can trade be facilitated and investment and growth take place by leveraging transport logistics accessibility as well as the economic and sociological factors of the SPICs? (Chaps. 5 and 8) In Table 4.1 we summarise our study in relation to our background context (challenge) by applying the Multilayer model to the sea-locked area of the SPICs. Each background challenge corresponds to a concept we employ and a hypothesis we test. Included in the table is a list of outputs of the models with different concrete policy options.

4.3

Variables and Multilayer Configuration

In this section we describe a configuration for a system which represents an appropriate general case for the sea-locked countries’ study. In Fig. 4.2 we visualise for a general product k the schematic case of two supply locations, h and i, and one demand location j, which are interconnected to ports o, r, u, and v. Port u represents a transhipment point in the shipping network of product k. Thus, supply and demand locations as well as ports are modelled as nodes interconnected through the transport network (links). We define the system at time t through the following variables

40

4 The Multilayer Model for Sea-Locked Countries

Table 4.1 Summary of background context of trade issues in sea-locked countries, modelling approach (concepts and assumptions) and main outputs of the Multilayer model Background context Sea-locked countries suffer from poor freight consolidation of trade volumes

Sea-locked countries experience thin volumes of trade due to isolated geographic conditions

Sea-locked countries have low access to opportunities (goods, services, investments, new technologies)

Concept Barabasi-Albert (B-A) network model Logistics nodes connect with each other according to potential benefits that a node may receive from connecting to another particular node (Multiplier Attachment) Economies of scale related to transport cost Multiplier Attachment Economies of scale related to transport cost Economies of agglomeration

Examine accessibility and attractiveness of ports in regard to logistics infrastructure and accessibility to socio-economic opportunities

Hypotheses Can the structure of the shipping network stimulate the consolidation of cargo and decrease transport cost? (H1)

Output (a) Architectures of the shipping network (b) Comparison analysis of multiple architectures (c) Hub-andspoke locations

Do trade coordination initiatives provide greater long-term economic growth for the SPICs? (H2)

(a) Comparison of different trade agreement regimes (b) Trade agreement selection for increase in trade

How can trade be facilitated? How can investment and growth take place by leveraging transport logistics accessibility? Can economic and sociological factors of the SPICs leverage investments, growth and transport accessibility? (H3)

Port Attractiveness indicator and Accessibility indicator

Concrete policy options (a) Create incentives at the national and regional level to support the implementation of hub-andspoke logistics (b) Develop consolidation network based on products in order to implement economies of scale

(a) Introduce free trade agreement through a stepwise approach (b) Institute technical assistance to facilitate the development of physical and institutional infrastructures (a) Assess the effects of socioeconomic measures/investments on trade (b) Estimate the accessibility of regional hubs in relation to considered shipping network architectures

4.3 Variables and Multilayer Configuration

41

Fig. 4.2 A schematic representation of the trade network in sea-locked countries. h and i are supply locations. j is a demand location. o, u, v, and r represent ports. Links represent trade flows between nodes. Dashed lines show flows from different locations passing through a link

k ¼ 1, 2, . . . :, K

number of products

n ¼ 1, 2, . . . :, N

nodes in the logistics network ðports; supply and demandÞ

m ¼ 1, 2, . . . :, M

links in the logistics network ðshipping routes; roads; etc:Þ

T ij, k

flow of product k between location ðsupplyÞ i and location ðdemand Þ j

Rij , k

set of links with minimum cost between i and j for product k

Suv, k

set of T ij, k that uses ðu; vÞ link on their shortest cost route

d uv

distance between node u and node v

θuv

link discount factor to increase economies of scale between port u and port v, 0 < θuv, k < 1

Ei, k

supply ðexportsÞ of product k from location i

I j, k

demand ðimportsÞ of product k in location j

huv , k

handling cost per unit volume in link u , v for product k

42

4 The Multilayer Model for Sea-Locked Countries

b c

fuel cost

ε

vessel fuel efficiency

βuv

discount factor due to coordination among ports βuv ¼ 0, if there is coordination between ports u  v, otherwise βuv ¼1

wuv, k ¼

P i, j2Suv, k

T ij, k

total traffic in the edge ðu; vÞfor product k.

The determination of the set of links Rij , k of product k, between location i and location j, is calculated through algorithms for shortest path problems. In our case, we use Dijkstra’s Algorithm (1959), which allows us to find the path with the lowest cost to move goods between two nodes (ports) in a shipping network. In the Appendix we report a detailed description of Dijkstra’s Algorithm and its implementation. Transport cost comprises the total costs associated with shipping one unit of product between supply and demand locations, and here we seek to estimate it through a model. In our case study we describe transport cost for maritime shipping, but the same concept can be generalised for all transport modes. Maritime transport cost in our study is classified into five main categories: vessel operating cost, maintenance cost, capital cost, voyage cost (fuel cost and port cost), and port fees (Bell et al. 2013; Park and Lee 2010). Model Assumptions: 1. In regional coordinated ports, port fees are assumed to be lower than port fees in non-regional coordinated ports; 2. Economies of scale are achieved through link discount factors that are endogenously modelled as a function of link flow volume; 3. Vessel type is homogeneous; 4. Vessel capital cost, operating cost, and maintenance cost are assumed to be constant among shipping companies operating in SPICs. The transport cost between ports u and v for product k can be expressed as a function of the five transport cost categories mentioned above, as TCuv, k ¼ f ðOC; MC; CC; VC; PFÞ,

ð4:1Þ

where OC ¼ operating cost MC ¼ maintenance cost CC ¼ capital cost plus depreciation VC ¼ voyage cost PF ¼ port fees Based on the assumptions of the study, since operating cost, capital cost and maintenance cost of the vessel are assumed to be constant, they do not influence our model. The voyage cost refers to the cost of moving a vessel from port u to port v,

4.3 Variables and Multilayer Configuration

43

and is mainly determined by the cost of fuel consumed in the journey, subject to fuel cost (b c ), distance between ports (duv), and the vessel fuel efficiency (ε). Port fees include handling cost (huv , k) and port tariff (τuv). Handling cost consists of all costs associated with handling cargo, and port tariff refers to the charges associated with port authorities, including port dues, anchorage charges, dockage dues, wharfage, and demurrage charges. Two discount factors are included in the transport cost to increase economies of scale (θuv) and introduce cooperation among ports (βuv). Link discount factor is modelled following Horner and O’Kelly’s (2001) formulation θuv ¼ ð1  α∗Suv γ Þ,

ð4:2Þ

where α and γ are two parameters, imposed exogenously, that determine the impact of the link discount factor on transport cost. Horner and O’Kelly (2001) use values of α and γ in the interval [0, 1]. The port tariff is multiplied by discount factor βuv to capture the regional coordination among ports. When there is coordination among ports this factor takes the value zero, otherwise it takes the value 1. The transport cost per unit is thus determined by TCuv, k ¼ VC þ PF

ð4:3Þ

VC ¼ θuv ∗ðb c ∗duv ∗εÞ=φ

ð4:4aÞ

PF ¼ huv, k þ τuv ∗βuv

ð4:4bÞ

where

and

φ represents number of units of a specific product k being transported in the vessel. This parameter is introduced to normalise the voyage cost in cost per unit (tonnage—or TEU—in our case study). Hence TCuv, k ¼ θuv ∗

b c ∗d uv ∗ε þ huv, k þ τuv ∗βuv φ

ð4:5Þ

The total transport cost between origin i and destination j for product k is given by the summation of transport costs between the links in the path connecting i and j, represented as X TCuv, k ð4:6Þ TCij, k ¼ u, v2R ij, k

44

4.4

4 The Multilayer Model for Sea-Locked Countries

Horizontal Network Model

In component III we examine both static and dynamic networks. Static networks do not evolve over time, but instead are fixed architectures (state-of-play/existence of shipping network, hub-and-spoke architecture, etc.). In the horizontal network we can test the effects of introducing, for example, specific policies by adding topological modifications to the networks and evaluating the vertical consequences generated on the trade layer of the Multilayer model. On the other hand, dynamic networks evolve and change their topologies over time; the nodes in this type of network follow various possible patterns which may include preferential attachment (Barabasi and Albert 1999), clustering and agglomeration (Small World model; Watts and Strogatz 1998), and spatially-driven solutions (Erdo˝s and Re´nyi class of models; Barthelemy 2011). We use a modification of the Barabasi-Albert model for spatial networks based on multiplier attachment. Our main aim here is to analyse the effect of hub allocation and volume consolidation so that transport cost(s) across the network can be minimised. The cargo shipping network is an example of real-world transportation networks in which nodes can be identified as ports situated in a particular space. Locations of the nodes are not distributed uniformly in space and are determined by external factors. Links between nodes correspond to shipping routes between ports. The degree of a node is defined as the number of shipping routes linking a node with other nodes in the network; and the shortest path between two (nodes) represents the number of stops a ship takes to travel between two ports. Empirical studies indicate that shipping networks can display both small-world properties (high local interconnectivity) and scale-free properties (presence of hubs), and thus can be identified using the Barabasi-Albert model (Hu and Zhu 2009; Xu et al. 2007; Barrat et al. 2005). Horizontal Network Model Assumptions: 1. Assumption of spatial preferential attachment: networks exhibit a spatial preferential attachment in which the likelihood of connecting to a node not only depends on distance but also on a connectivity gain. 2. Assumption of multiplier attachment: the gain is assumed to be given by the potential benefits a node may receive from connecting to another particular node. This may include economic benefits, cultural ties and market opportunities. Under assumption 1 of spatial preferential attachment, in the traditional Barabasi-Albert model a new node u is connected to node v, with a probability given by π u!v / yv

ð4:7Þ

where yv is the degree of node v (Albert and Barabasi 2002). Distance plays a significant role in real-world transportation networks, and long-range links usually

4.4 Horizontal Network Model

45

interconnect with hubs. Thus, when considering the effect of distance, the probability of attachment can be rewritten as π u!v / yt f ½d uv ,

ð4:8Þ

where f is a function depending on euclidian distance duv between nodes u and v (Barthelemy 2011). When f is a decreasing function, this implies that new links will preferentially connect to hubs unless the hub is too far away, in which case it could connect to a less-connected but closer node. Typically, f can take the form of an exponential function f ðd Þ ¼ e r c , d

ð4:9Þ

where rc represents the interaction range. This parameter regulates the significance of distance in the model. When the interaction range is of the order of the system size (or larger), distance becomes irrelevant. Conversely, when the interaction range is small compared to the system size, distance becomes more influential in the probability of attachment. In addition to distance, real-world transportation networks have several other relevant factors that can influence the probability of attachment (Hu and Zhu 2009), which in our specific case of shipping networks are: traffic flow of products between ports, levels of supply and demand and market opportunities, to name only three. These factors altogether comprise the benefits or gains that make a particular node more attractive to connect to, increasing the probability of attachment. The effect of these factors is introduced into our model through node strength, sw, leading us to rewrite the probability of attachment as s w eduv =rc πu!v ¼ Pv w d =r , sm e uv c

ð4:10Þ

m

where svw is given by the summation of the weight of links connecting node v. Since a node can have inward links (inbound traffic) and outward links (outbound traffic), node strength can also be divided into in-strength and out-strength. For the case of the shipping networks, the in-strength and out-strength for node v for product k are given by the sum of inward and outward weights respectively, and denoted by sinw, k w w ðuÞ and sout , k ðuÞ. The total strength of a node is represented by stot, k ðuÞ. Node strength can be calculated through the following X w ð4:11Þ sinw, k ðuÞ ¼ v6¼u vu, k X w w ð4:12Þ sout , k ð uÞ ¼ u6¼v uv, k X w stot ðwvu, k þ wuv, k Þ, ð4:13Þ , k ð uÞ ¼ u6¼v

46

4 The Multilayer Model for Sea-Locked Countries

w where stot , k ðuÞ represents, for our shipping networks case, the throughput of port u for product k, given by the total traffic flow of product k in that port. Nodes in shipping networks have a specific value in terms of economic competitiveness; these are in turn driven by endogenous (geographic position and infrastructure endowment) and exogenous characteristics (national economic development and reputation). Determining whether a node will connect to another node depends on the expectation of gaining greater economic benefit when compared to the state-of-play. We include parameter, χv, multiplier attachment, which represents the gain node u will receive when it attaches to node v. The probability of node u attaching to node v for product k in our model is the result arising in our modified version of the traditional Barabasi-Albert model, defined as

π u!v, k ¼ P

w TCuv , k=r c χ v stot , k ðvÞ e

m

w ðmÞ eTCum, k =rc χ m stot ,k

ð4:14Þ

To further refine the analysis, we use the Port Attractiveness Index (Caschili and Medda 2012) as a potential benefit parameter χ. Port Attractiveness is defined as “the combination of the productive capacity of a port and its level of international competitiveness which provides direct and indirect economic benefits.” A port generates freight traffic by means of its interconnectedness with inland trade routes and with other regional and international ports. Therefore, in order to be attractive and competitive, ports often need to be integrated both vertically, i.e. secure maritime routes and landside operations, and horizontally, i.e. highly specialised with a wide geographical market share. The implication here is that a port must be equipped with effective facilities, it must provide reliable services at the lowest price, and it needs to have an efficient productivity level. These characteristics altogether comprise the reputation of a port as an intricate network of operators, investors and maritime brokers (Caschili and Medda 2015). In Chap. 5 we present a thorough description of the Port Attractiveness Index and explain how it is calculated. The next step is to evaluate the transhipment potential of ports in order to determine which ports should play the transhipment role. We measure transhipment potential through the frequency of transhipment of port i for product k TPi, k ¼

X u6¼v

Suv, k ðiÞ , Suv, k

ð4:15Þ

where Suv , k(i) represents number of trips transiting from node u to node v and passing through node i, and Suv , k represents number of trips from u to v. Port i has a higher transhipment potential if more trade flows pass through it, compared to the total traded volumes in the region. We use the transhipment potential to rank ports in terms of importance in the network: a transhipment port plays the role of

4.5 Vertical Network Model

47

connecting increasing numbers of ports and thus is more important to the resilience and/or vulnerability of the trade system.

4.5

Vertical Network Model

In component IV we examine the vertical interactions between layers. In Fig. 4.3 we visualise the vertical integration among layers which are represented as functional relationships where the upper layer level (trade) is the ‘dependent variable’ and the next layer level is represented by the set of associated ‘independent variables’. Any of these layers can in turn be considered as a ‘dependent variable’ from its own set of ‘independent variables,’ hence the multilayering. All the layers are potentially interactive. For example, we can consider the trade layer, which interacts with the underlying physical layer, where a change (disruption or improvement) in the physical layer can potentially generate impacts on trade costs. The Vertical Interaction model of component IV investigates dependencies and causalities among layers/determinants that may inhibit trade and also evaluates the transmission of their effects across layers. We calibrate the Vertical Interaction model using the analytical specification of the parameters, as in Wilson (1967). Assumption: Trade is not an isolated network; it depends on logistics and economic and sociological interactions/networks between locations. Given product k, we subject our model to the following conditions: the supply of location i (Ei,k) and demand of location j (Ij,k) are equal to the sum of outward and inward flows of product k of each location X T ¼ Ei , k ð4:16Þ j ij, k X T ¼ Ij, k ð4:17Þ i ij, k

Fig. 4.3 Functional relationships between layers

48

4 The Multilayer Model for Sea-Locked Countries

Following the standard notation for spatial interaction models, we can write the trade between two locations i and j for product k, as   Tij, k ¼ Ai, k Bj, k Ei, k Ij, k f β; TCij, k

ð4:18Þ

where f(β, cij , k) is a general measure of impedance between i to j; TCij , k is the transport cost per unit of product k between location i and location j; β is a parameter which provides the interaction scale. The impedance function f (β, TCij , k) can now take an exponential form     f β, cij, k ¼ exp β; TCij, k

ð4:19Þ

  f β, cij, k ¼ TCijβ, k

ð4:20Þ

or a power decay form

Interactions in short-to-medium distance are better approximated by an exponential decay function where transport cost has a higher impact, while the power decay function better fits interactions between large distances. Our next step is to calibrate the model with both functions to assess which type of decay function best fits the data. Ai,k and Bj , k are balancing factors, written as follows 1   j Bj, k Ij, k f β; TCij, k

ð4:21Þ

1   i Ai, k Ei, k f β; TCij, k

ð4:22Þ

Ai, k ¼ P Bj, k ¼ P

We estimate the model in Eq. (4.18) from observed data and develop the multilayer framework for bilateral trade by considering physical, economic and cultural layers. We embed the economic and cultural layers as networks that are mathematicallyrepresented by matrices and include: exchange rate {ERij}, a variable to consider whether the countries of locations i and j have stipulated an economic agreement for product k {TAij,k}; cultural links between location i and j {CLij}; and migration {Mij} between i and j. The physical layer is embedded in transport cost, TCij , k. Evidence of economic and cultural links arises, for example, in relation to specific product k, as in the case of an economic agreement for a certain category of product, or may refer to interaction patterns that include every product (exchange rate, common language, etc.). The spatial interaction model of Eq. (4.18) can now be written as   Tij, k ¼ Ai, k Bj, k Ei, k Ij, k f β, TCij, k ERij EAij, k CLij Mij ,

ð4:23Þ

4.5 Vertical Network Model

49

where balancing factors are expressed as A i, k ¼ P Bj, k ¼ P

j Bj, k Ij, k f



1 

β; TCij, k ERij EAij, k CLij Mij

1   A E f β; TC ij, k ERij EAij, k CLij Mij i i, k i, k

ð4:24Þ ð4:25Þ

When we examine the functional relationships between the networks depicted in Fig. 4.3, matrices {Tij,k}, {TCij,k}, {ERij}, {TAij,k}, {CLij} and {Mij} can all be seen as networks or layers of the Multilayer model. The networks {ERij}, {TAij,k}, {CLij} and {Mij} represent a possible set of layers of the Multilayer model in our case study, but they are merely illustrative. In theory, many new layers can be added if they improve model fitness. The ability to construct a multilayer model depends on type of case study, personal knowledge of the analyst derived from exploratory statistical analyses, and results of the model calibration. The calibration of component IV is grounded in the entropy maximisation framework of Wilson (1967), which is fundamental in building key performance indicators to measure the accessibility of nodes. The analysis of accessibility considers where competition effects occur at both origin and destination locations. In the maritime shipping industry, ports compete with each other to attract new shipping lines, and shipping lines compete with each other for new customers. By measuring the accessibility of places, we can quantitatively evaluate, using an indicator, whether the introduction of a new policy, for example, is impacting on trade. Thereafter we can record the effect it has generated in terms of accessibility. Nonetheless, for sea-locked countries it is also important to evaluate the ease for a port to access regional and international markets. For this reason we define outbound accessibility Accout i as the opportunity potential for interaction of node i with other nodes, j, of our domain. Accout i is a measure of the propensity to reach certain economic activities/destinations, j, and is expressed in the following form X   Accout I f β, TCij, k ð4:26Þ i ¼ j j, k If Accout measures the opportunity potential for interaction of node i, the inbound i accessibility Accin j measures the propensity of a node j to be reached by other nodes. in We define Accj as follows Accin j ¼

X

E f i i, k



β, TCij, k



ð4:27Þ

In Eqs. (4.26) and (4.27), β is the cost sensitivity parameter calibrated in the Vertical Interaction model (Eq. 4.23). There is a slight, but significant, difference between the two indicators: Accout i measures the ease to export from node i, whereas

50

4 The Multilayer Model for Sea-Locked Countries

Accin j estimates the ease to import of node j. Both inbound and outbound accessibility are important for sea-locked countries. In order to construct a synthetic indicator of accessibility, we linearly combine Accin and Accout out Acci ¼ Accin i þ Acci

ð4:28Þ

Equation (4.28) represents the accessibility level of node i of the network.

4.6

Concluding Comments

In this chapter we have outlined the Multilayer model which can be deployed in order to optimize trade flows between the South Pacific Islands and beyond. Moreover, the use of the multiplier attachment model for network building has allowed us to identify the most efficient routes. The Port Attractiveness Index is also pivotal to this research, and in Chap. 5 we elaborate on how to construct this index. Our aim in this chapter has been to develop a spatial interaction model able to effectively integrate the variables in each of the three vertical layers, Physical, Economic and Sociological, and allow us to calculate the trade layer which identifies the trade flows along the most efficient network of routes. The fully integrated Multilayer model will be applied in subsequent chapters in order to explore different policy options in the context of various scenarios. Policy recommendations arising from our analysis are set out in Chaps. 6 and 7, and in Chap. 8 we present a synthesis using the fully-integrated model. However, next in the Appendix we discuss the Dijkstra Algorithm in detail.

Appendix: The Dijkstra Algorithm This Appendix formalises the Dijkstra shortest path algorithm (1959). In describing the algorithm we keep a consistent notation with that used throughout this book (readers are asked to return to Sect. 4.3 for a definition of the variables). The set Sij , k of links with minimum cost between source node i and target node j is said to be the shortest path if its transport cost is minimum among all i-to-j paths. Dijkstra’s Algorithm is based on the following assumptions • • • •

All link costs cuv , k are non-negative; The number of vertices is finite; The source is a single node, but the target may be all other nodes; We assume we have no loops or parallel edges.

Appendix: The Dijkstra Algorithm

51

Dijkstra’s Algorithm marks the vertices as permanent or temporary vertices. The label of a node j is denoted β(j) and we define  γðjÞ ¼

1 if the label is permanent 0 if the label is temporary

A permanent label β( j) expresses the weight of the shortest directed i-to-j path. A temporary label β( j) gives an upper limit to this weight (can be 1). Furthermore, we denote  πð j Þ ¼

the predecessor of node p on the shortest path i  j, 0 otherwise

In this way we can construct the directed path with the lowest weights. The Dijkstra Algorithm can be formalised in the following steps Set β(i) ← 0 and γ(i) ← 1. For all other vertices j:

1.

set

β(j) ← ∞ and γ(j) ← 0.

For all vertices j, set π(j) ← 0. Furthermore, set

2.

For every link (u, j), where γ(j) = 0 and β(j) > β(w) +  set and

3.

u ← i.

β(j) ← β(u) +  π(j) ← u

,

,

Find a node j* for which γ(j*) = 0, ß(j*) < ∞ and ß(  ) = minȚ(j)=0 ß(  Set

γ(j* ) ← 1 and w ←j*

If there is no such vertex j*, a directed i–v path does not exist and we stop.

4. 5.

If u ≠ v, then go to step #2. Stop

We see that the algorithm is correct as follows. We denote (for every step):   V1 ¼ permanently labelled vertices   V2 ¼ temporarily labelled vertices (V1, V2) is a cut with the completely scanned vertices on one side and other vertices on the other side. In the next chapter we provide a thorough discussion of the Port Attractiveness Index, which is a critical component in the analysis of the shipping network.

52

4 The Multilayer Model for Sea-Locked Countries

References Albert R, Barabasi AL (2002) Statistical mechanics of complex networks. Rev Mod Phys 74: 47–97 Barabasi AL, Albert R (1999) Emergence of scaling in random networks. Science 286(5439): 509–512 Barrat A, Barthelemy M, Vespignani A (2005) The effects of spatial constraints on the evolution of weighted complex networks. J Stat Mech Theory Exp May:P05003 Barthelemy M (2011) Spatial networks. Phys Rep 499(1–3):1–101 Bell MGH, Liu X, Rioult J, Angeloudis P (2013) A cost-based maritime container assignment model. Transp Res B 58:58–70 Caschili S, Medda F (2012) A review of the maritime container shipping industry as a complex adaptive system. INDECS 10(1):1–15 Caschili S, Medda F (2015) Port Attractiveness Index: an application on African ports. Region et Developpement 41:47–82 Dijkstra EW (1959) A note on two problems in connexion with graphs. Numer Math 1:269–271 Erdo˝s P, Re´nyi A (1959) On random graphs. Publ Math 6:290–297 Horner MW, O’Kelly ME (2001) Embedding economies of scale concepts for hub network design. J Transp Geogr 9:255–265 Hu Y, Zhu D (2009) Empirical analysis of the worldwide maritime transportation network. Physica A 388:2061–2071 Park Y, Lee S (2010) Port tariff setting model for Pacific Island countries. Korea Maritime Institute, December Watts D, Strogatz S (1998) Collective dynamics of ‘small-world’ networks. Nature 393(6684): 440–442 Wilson AG (1967) A statistical theory of spatial distribution models. Transp Res 1:253–269 Wilson AG (1970) Entropy in urban and regional modelling. Pion, London Xu X, Hu J, Liu F (2007) Empirical analysis of the ship transport network of China. Chaos 17: 123–129

Chapter 5

Port Attractiveness

5.1

Introduction

This chapter is based on the main concepts that Caschili and Medda (2015) have developed to construct the Port Attractiveness Index, which is used in multiplier attachment to construct the shipping network in module II. Caschili and Medda (2015) define port attractiveness as the combination of the productive capacity of a port and its level of international competitiveness which together provide direct and indirect economic benefits. A port generates freight traffic through its connections with inland trade routes and with other regional and international ports. Thus, in order to be attractive and competitive, ports often need to be integrated vertically, i.e. have secure maritime routes and landside operations, and integrated horizontally, i.e. be highly specialised with a wide geographical market share. The implication here is that a port must be equipped with effective facilities, it must provide reliable services at the lowest price, and it needs to have an efficient productivity level. These characteristics combined comprise the reputation of a port as an intricate network of operators, investors and maritime brokers.

5.2

Methodology

Port attractiveness determinants can generally be grouped into three categories: endogenous, exogenous and subjective. Endogenous factors regard the port directly: port infrastructure endowment (Murphy and Daley 1994; Slack 1985; Tiwari et al. 2003; Tongzon 2002; Ha 2003), monetary costs (Murphy and Daley 1994; Ha 2003; Foster 1978; Tongzon 2002; Tiwari et al. 2003; Lirn et al. 2003), logistics efficiency (Murphy and Daley 1994, Slack 1985, Tiwari et al. 2003; Tongzon 2002, 2009; Ha 2003), and port accessibility (Huybrechts et al. 2002).

© Springer International Publishing AG 2017 F.R. Medda et al., Collaborative Approach to Trade, Advances in Spatial Science, DOI 10.1007/978-3-319-47039-9_5

53

54

5 Port Attractiveness

On the other hand, exogenous factors include a number of external determinants that influence port throughput indirectly: national and local economic competitiveness (Przybyłowski 2008; OECD 2008), geographic location (Tiwari et al. 2003; Tongzon 2002; Lirn et al. 2003; Ha 2003), and shipping line characteristics (Noteboom 2009; Slack 1985; Tiwari et al. 2003). The third category, subjective factors, leverage port attractiveness and above all refers to the reputation of a port among sector operators (Lirn et al. 2003; Ng 2006; Tongzon 2002; Daya et al. 2006; Bird and Bland 1988). Against this background, Caschili and Medda (2015) contribute to the existing literature by setting out a new approach for assessing port attractiveness. They construct the Port Attractiveness Index by exploiting a bottom-up statistical approach (structural equation modelling) in order to combine and analyse causal relationships among exogenous, endogenous and subjective determinants and thereafter to measure their significance. Port attractiveness is defined as the combination of the productive capacity of a port and its level of international competitiveness which provides direct and indirect economic benefits. A port generates freight traffic through its interconnectedness with inland trade routes and with other regional and international ports. Thus, in order to be attractive and competitive, ports often need to be integrated vertically, i.e. secure maritime routes and landside operations; and integrated horizontally, i.e. highly specialised with a wide geographical market share. The implication here is that a port must be equipped with effective facilities, it must provide reliable services at the lowest price, and it needs to have an efficient productivity level. These characteristics altogether comprise the reputation of a port as an intricate network of operators, investors and maritime brokers. Endogenous, exogenous and subjective variables, and in particular the subjective variables, are often collected via surveys (Sequeira 2012). However, survey methodologies are expensive as well as time consuming to carry out, therefore as an alternative method Caschili and Medda (2015) consider the copious data collected from third-party organisations (i.e. World Bank, Containerisation International, UNCTAD, internet flows, crowdsourcing data, etc.) in order to increase the scale and volume of the examined data as well as the variety of the data. After having collected a significant volume of multivariate data, Caschili and Medda (2015) use Structural Equation Modeling (SEM) to define and assess variables that are not directly observable (latent variables) and examine their causal relationships. SEM is a robust statistical methodology perfectly suited to our calculation of the causal relationships between the variables influencing port attractiveness (Ullman and Bentler 2012). The Port Attractiveness Index assumes that the higher the value of endogenous, exogenous and subjective variables (hereafter called key constructs), the higher the Port Attractiveness Index (Fig. 5.1). The three key constructs are latent variables that determine the attractiveness of a port (A). The exogenous latent variables D are meant to represent the socioeconomic level of port hinterlands and the Quality of their Governance. D can be dependent on several variables such as economic development (Mazumdar 1996),

5.2 Methodology

55

Fig. 5.1 Structural equation model of causal relationships between factors in port attractiveness. Source: Caschili and Medda (2015)

quality of telecommunication infrastructure (Oyelaran-Oyeyinka and Kaushalesh 2005), and integrity level (i.e. level of corruption, accountability in governance, etc.) (Montinola and Jackman 2002). Key construct F refers to the infrastructural and operational level of the port (endogenous variables). F is usually dependent on variables such as port facilities (Slack 1985; Tongzon 2002), logistics efficiency (Murphy and Daley 1994; Ha 2003; Foster 1978; Tongzon 2009) and port productivity (the higher the port throughput, the higher the infrastructure level of a port). The use of SEM allows us to quantitatively evaluate port reputation represented by the subjective key construct R. Here, R is dependent on variables such as port quality (from shipper’s point of view), centrality in the international shipping network (the higher the interconnectivity of a port in the global shipping network, the higher its reputation in the industry), and level of reliability. The causal relationships obtained from the SEM are linearly combined to build the Port Attractiveness Index Φ. The index Φi for port i can be written in mathematical terms as Φi ¼ αAR ∗Ri þ αAF ∗Fj þ αAD ∗Di ,

ð5:1Þ

where Ri ¼

Xn k¼1

αrk ∗ri, k

ð5:2Þ

56

5 Port Attractiveness

Fi ¼ Di ¼

Xn k¼1

αf k ∗f i, k

ð5:3Þ

k¼1

αdk ∗di, k

ð5:4Þ

Xn

αAR , αAF , αAD , αrk, αfk and αdk are the path loadings obtained from the SEM and represent the relative importance of each key construct and measured variable; ri , k, fi , k and di , k are the kth observed variables for port i in the jth year. SEM consists of two processes: validation of the model measurements (factor analysis) and fitting the structural model (path analysis with latent variables). The core of the SEM methodology is parameter estimation, consisting of the comparison of the covariance matrices of observed variables with the estimated covariance matrices of the best fitting model. Readers who are unfamiliar with SEM methodology can find appropriate references in Kline (2011) and in articles published in specialised scientific media, particularly Structural Equation Modeling: A Multidisciplinary Journal.

5.3

Results

Through the use of Structural Equation modelling (SEM) we validate and compare the factors that affect port attractiveness in the SPICs in a three-step goodness-of-fit test. By combining the collected factors at port level, we first examined several models and either rejected or confirmed them based on the results of the goodnessof-fit tests. In the second step we developed alternative models based on the following model hypotheses: Hypothesis 1 The infrastructural and operational level of ports can be modelled as a latent variable determined by Max Vessel Draft, Annual Port Throughput and Customs Efficiency. Max Vessel Draft is very important because it impacts on whether large vessels can serve a port. Peter (2001) has observed that insufficient water depth limits the possibility for some ports to grow and to act as transhipment points. Port Throughput was used as an indicator to measure port performance, as it measures the number of containers moved over a certain time frame (usually during 1 year, semester or month). De Langen et al. (2007) have argued that although “throughput volume is the most widely used performance indicator in the port industry, it does not provide information on the (regional) economic impact of the port and the attractiveness of the port as a location for port-related industries.” For this reason we have also introduced throughput as a component of port attractiveness. Customs Efficiency also affects port attractiveness. Bird and Bland (1988) have demonstrated that choice of routes is influenced by port customs efficiency. We

5.3 Results

57

have also used the Logistics Performance Index for customs (Arvis et al. 2010) in order to consider the impact of customs on port performance. The information has been extracted from surveys of academics, international institutions and private companies. The efficiency of customs procedures is evaluated for speed, simplicity and predictability of formalities (Arvis et al. 2010). Hypothesis 2 Socio-economic level of hinterlands can be modelled as a latent variable determined by Gross Domestic Product, ICT and Ease of Doing Business at national level. Due to lack of information at the local scale, we have assumed that the exogenous variables are distributed evenly across a country. We are aware that this is a disputable assumption but in our case study this assumption is supported by the following conditions: 80% of our study countries are represented by only one port, usually the largest national container port. Three countries in our data sets (Fiji, Vanuatu and Solomon Islands) have two ports, one of which is the national hub, while PNG has five ports. In these cases, ports might be used as gateways by different regions/hinterlands. We have used Gross Domestic Product (GDP) to measure the market activity (goods and services) of a region, given that Ducruet (2009) finds a significant correlation between trade and GDP in a sample of 116 maritime countries. Furthermore, several studies have shown that the Internet has opened new trade opportunities between countries (Freund and Weinhold 2004) and also decreased information costs (Hagiu and Yoffie 2013). A 10% increase in the growth of web users in a country leads to an approximate 0.2–0.4% increase in export growth (Freund and Weinhold 2004; Lin 2015). We therefore find a relationship between trade growth and Internet usage, which we include in the endogenous latent variables of our model in order to evaluate a region’s capacity to reach foreign markets. Lastly, we have accounted for excessive regulations in a country as impediments to economic growth and trade activity (Bolaky and Freund 2004) through our use of the Ease of Doing Business Index. This index ranks countries’ regulatory environments in so far as they are conducive to the start-up and operation of local firms (www.doingbusiness.org). Hypothesis 3 Port Reputation can be modelled as a latent variable determined by level of connectivity to the global shipping network and logistics efficiency. At the shipping company level, port choice is influenced by several factors, including port efficiency, interconnectivity and reputation of ports (Wiegmans et al. 2008). Port Reputation is higher for ports with higher logistics efficiency and interconnectivity to the global trade network. Thus we have modelled Port Reputation as determined by these latter two variables. We used the Shipping Liner Connectivity Index (http://unctadstat.unctad.org) to capture a country’s level of connectedness to the global shipping network. And lastly, we used the Logistics Performance Index (http://lpi.worldbank.org), which estimates the logistics ‘friendliness’ of countries based on a worldwide survey of operators. Changes in the path loading were introduced if they improved the goodness-offit tests, that is, if the hypotheses could be verified. We have chosen the best model

58

5 Port Attractiveness

Fig. 5.2 Structural equation modelling diagram for the estimation of port attractiveness

among those that passed the goodness-of-fit tests and verified the hypotheses of our model. Several models were tested before finding the optimal configuration depicted in Fig. 5.2. The figure illustrates the path loading diagram along with estimated un-standardised coefficients for the estimation of port attractiveness. Error co-variances were established through an exploratory approach. The goodness-of-fit indices in Table 5.1 confirm that the chosen model is consistent with the data. Most of the fitting indices have surpassed their recommended values. In terms of hypothesised links between the measured and latent variables and their statistical significance, all links show significance paths at p-value 0.9 >0.9 >0.9

Obtained value 11.4 7 1.629 0.136 0.967 0.868 0.987 0.986 0.945

Conclusions and Policy Recommendations

The Port Attractiveness Index (Φ) is rooted in the causality relationships among the determinants of port attractiveness that we have analysed through SEM (Fig. 5.2). The Country Development Level is the major factor for determining port attractiveness in the SPICs (path coefficient 5.37). The level of GDP of the country is an important factor as is Ease of Business, which negatively affects economic development by hampering business competitiveness, entrepreneurship and overall growth. The implication here is that the development of the SPICs ports, and thus possible intervention by private investment, are hindered by setbacks in economic development (Peters 2001). Investors would therefore consider country development in the SPICs as a discriminant factor in relation to possible port investments. In Fig. 5.2 Port Features are as important as Port Reputation (with path coefficients of 1.00 and 1.67, respectively). An interesting result is that for the key determinant, Port Reputation, the capacity of a port to be integrated in the international shipping network (LSCI) is almost four times more important than the Logistics Performance Index (LPI). Thus, in order to increase port attractiveness, port operators need to develop a wide network of commercial relationships with a number of shipping companies, particularly to increase the access and connectivity with the international markets (i.e. number of destinations). Moreover, by providing effective services (LPI ¼ 1.00), ports also benefit from the positive word-of-mouth effect: ports become more attractive when they function as hubs (i.e. carriers can exploit cooperative schemes in those ports), they benefit from tacitly being promoted in the industry through a multiplier networking mechanism. Port Features (represented by Max Draft) count as much as Customs. This result is not surprising; ports with good infrastructure assets but inefficient operations are less productive, so they are less able to handle increases in container traffic. The Port Attractiveness Index is depicted in Fig. 5.3 and Table 5.2 for the 35 ports of our case study. In Table 5.2 we rank ports by Port Attractiveness Index. Each circle in Fig. 5.3 is proportional to the value of the Index, and as expected, ports in the SPICs have very low Port Attractiveness compared to the rest of the ports shown. SPICs ports with the highest Port Attractiveness are in Lae in PNG, Suva in Fiji, and Noumea in New Caledonia.

60

5 Port Attractiveness

Fig. 5.3 Geo-refereed visualisation of Port Attractiveness Index for our case study

In order to more closely examine our obtained results, we use the examples of Suva in Fiji, and Lae and Port Moresby in the PNG, and compare their port attractiveness values with a major international port: Busan in South Korea (Fig. 5.4). The three ports in the SPICs have some of the best Port Attractiveness Index values in the region (over 35); they are obviously lower in relation to Busan (52). However, if we consider the port of Busan as a benchmark, it is noteworthy that although both Port Lae and Port Moresby in PNG have higher Throughput and GDP than Suva in Fiji, they have significantly lower values for Internet users (2, whereas Suva has 20) and high values for Ease of Business (108 respectively, whereas Suva has 58). When we compare these data with values for the port of Busan, the port of Suva appears to be following a growth trend which targets important economic factors, such as the simplification and transparency of doing business and a greater distribution of income; this is evident from the level of Internet Users variable. Internet Users is a satisfactory proxy to account for the level of the middle class and also entrepreneurship activities present in the country in order to achieve the development of the port and to attract private investment. These conclusions are also supported by the greater amount of FDI that Fiji has received compared with the level received by PNG (UNCTAD 2011). Several important policy implications can be raised in order to support the growth of the SPICs maritime and trade industry. In regard to the function of the ports and in general business activity, SPIC governments should streamline

Port Noumea Papeete Suva Lae Port Moresby Lautoka Kimbe Rabaul Pago Pago Oro Bay Madang Apia Majuro Honiara Nukualofa Port Vila Noro Santo Wallys Koror Tarawa Betio Funafuti Los Angeles Singapore

Throughput (TEU) 90,974 71,865 32,169 142,146 87,889

20,958 13,836 14,343 175,000a 5311 4752 22,000 30,000a 12,500 8530 3000 2500a 2000 10,000a 5000a 6600 3000a 7,831,902 28,400,000

Port Attract 37.91 37.48 35.53 35.51 35.24

35.12 33.81 33.69 33.48 32.82 32.80 31.76 31.57 31.42 31.39 30.32 30.05 29.96 29.50 29.44 28.65 27.77 55.75 52.98

9.75 13 10.2 11 9 10.1 10 15 9.14 10 10.7 8 10 7 6.6 7 6.4 22 22

Max Draft (mt) 10.3 10.37 10.97 10 12 1.95 2.02 2.02 3a 2.02 2.02 3a 2.5a 2.08 2.9a 2.1a 2.08 2.1a 3a 2a 2.5a 2.5a 3.68 4.02

LPI— Customs 2.3a 2.3a 1.95 2.02 2.02

Table 5.2 Ranking of ports by Port Attractiveness Index for SPICs and RIM ports

3.23E + 09 9.48E + 09 9.48E + 09 6E + 08a 9.48E + 09 9.48E + 09 6E + 08a 1.8E + 08a 6.79E + 08 3.69E + 08 7.01E + 08 6.79E + 08 7.01E + 08 1E + 08a 1.97E + 08 1.5E + 08 31,824,701 1.5E + 13 2.17E + 11

GDP (US $) 1.08E + 10 8.14E + 09 3.23E + 09 9.48E + 09 9.48E + 09 20 2 2 5.5 2 2 5.5 9.7 5 16 8 5 8 8.4 28.5 9.1 25 74 71

Internet Users 42 49 20 2 2 9.4 6.4 6.4 4.8 6.4 6.4 4.8 6.5a 5.6 3.7 3.7 5.6 3.7 6a 3.4 2.9 6a 83.8 103.8

LSCI 9.4 8.9 9.4 6.4 6.4 1.98 1.91 1.91 1.9a 1.91 1.91 1.9a 1.92a 2.23 2.05a 1.93a 2.23 1.93a 1.95a 1.92a 1.85a 1.95a 4.15 4.22

LPI— Tot 1.98a 2.02a 1.98 1.91 1.91 58 108 108 100a 108 108 100a 90a 92 60 78 92 78 95a 114 117 80a 4 1

(continued)

Ease of Business 88a 89a 58 108 108

5.4 Conclusions and Policy Recommendations 61

Port Attract 52.59 52.45 51.81 50.09 49.23 47.33 46.97 46.04 45.94 44.24 37.05

Throughput (TEU) 23,532,000 14,157,291 2,550,000 8,870,000 1,000,000 818,000 1,497,000 6,500,000 1,000,000 3,160,000 96,952 Max Draft (mt) 16.5 12 11 13.4 14 13.2 14 11 13.4 13.4 16.2

LPI— Customs 3.16 3.33 3.79 3.11 3.68 3.64 3.02 2.43 2.7 2.67 3a

GDP (US $) 5.93E + 12 1.01E + 12 5.5E + 12 2.48E + 11 1.14E + 12 1.43E + 11 3.19E + 11 7.09E + 11 1.71E + 12 2E + 11 3E + 09a

The table also includes national economic variables and port variables used in the SEM of Fig. 5.2 a Estimated value

Port Hong Kong Busan Nagoya Port Kelang Brisbane Tauranga Bangkok Jakarta Calcutta Manila Guam

Table 5.2 (continued) Internet Users 34.3 83.7 78.2 56.3 76 83 22.4 10.9 7.5 25 56.3 LSCI 143.6 82.6 67.4 88.1 28.1 18.4 43.6 25.6 41.4 15.2 8.8

LPI— Tot 3.54 3.62 4.19 3.5 3.78 3.54 3.16 2.54 2.91 2.57 2.01a

Ease of Business 99 6 23 8 10 3 18 116 131 133 105a

62 5 Port Attractiveness

References

63

Fig. 5.4 Data summary for Lae, Port Moresby, Suva, and Busan

bureaucratic procedures and take steps to reduce regulation that reduce regulation that may limit competition and protect incumbent operators. In relation to the interconnections of ports, the SPICs need to ensure adequate access provision for the domestic markets, but above all for international markets, by implementing consolidation of traffic and coordination between ports. The SPICs are in the large majority too small individually to meet the challenges of the global shipping and trade industry, therefore only by pooling resources will they be able to grow and emerge from their lagging economic status. In Chaps. 6 and 7 we examine three scenarios which correspond to the consolidation and coordination strategies in the SPICs.

References Arvis JF, Mustra MA, Panzer J, Ojala L, Naula T (2010) Connecting to compete: trade logistics in the global economy. World Bank, Washington, DC Bird J, Bland G (1988) Freight forwarders speak: the perception of route competition via seaports in the European Community’s research project. Marit Policy Manag 15(1):35–55 Bolaky B, Freund CL (2004) Trade, regulations and growth. World Bank Policy Research Working Paper, Washington, DC Caschili S, Medda F (2015) Port Attractiveness Index: an application on African ports. Region et Developpement 41:47–82 Daya Y, Ranoto TR, Letsoalo MA (2006) Intra-Africa agricultural trade: a South African perspective. Department of Agriculture, Pretoria, South Africa De Langen P, Nijdam M, van der Horst M (2007) New indicators to measure port performance. J Marit Res 4(1):23–36 Ducruet C (2009) Port regions and globalization. In: Ports in proximity: competition and coordination among adjacent seaports. Ashgate, Farnham Foster T (1978) What’s important in a port. Distribution Worldwide 78:33–36 Freund CL, Weinhold D (2004) The effect of the internet on international trade. J Int Econ 62(1): 171–189

64

5 Port Attractiveness

Ha MS (2003) A comparison of service quality at major container ports: implications for Korean ports. J Transp Geogr 11:131–137 Hagiu A, Yoffie DB (2013) The new patent intermediaries: platforms, defensive aggregators and super-aggregators. J Econ Perspect 27(1):45–65 Huybrechts M, Meersman H, Van de Voorde E, Van Hooydonk E, Verbeke A, Winkelmans W (2002) Port competitiveness: an economic and legal analysis of the factors determining the competitiveness of seaports. De Boeck, Antwerp Kline RB (2011) Principles and practice of Structural Equation Modeling, 3rd edn. The Guilford Press, New York Lin F (2015) Estimating the effect of the Internet on international trade. J Int Trade Econ Dev 24(3):409–428 Lirn TC, Thanapoulou H, Beresford A, Anthony KC (2003) Transhipment port selection and decision-making behaviour: analysing the Taiwanese case. Int J Log Res Appl 6(4):229–244 Mazumdar J (1996) Do static gains from trade lead to medium-run growth? J Polit Econ 104(6): 1328–1337 Montinola GR, Jackman R (2002) Sources of corruption: a cross-country study. Br J Polit Sci 32:142–170 Murphy P, Daley J (1994) A comparative analysis of port selection factors. Transp J 34(1):15–21 Ng AKY (2006) Assessing the attractiveness of ports in the North European container transhipment market: an agenda for future research in port competition. Marit Econ Logist 8(3): 234–250 Notteboom TE (2009) The effect of high fuel costs on liner service configuration in container shipping. J Transp Geogr 17:325–337 OECD (2008) A review of local economic and employment development policy approaches in OECD countries. OECD Centre for Entrepreneurship, SMEs and Local Development, Paris Oyelaran-Oyeyinka B, Kaushalesh L (2005) Internet diffusion in sub-Saharan Africa: a crosscountry analysis. Telecommun Policy 29(7):507–527 Peters HJ (2001) Developments in global sea trade and container shipping markets: their effects of the port industry and private sector involvement. Int J Marit Econ 3:3–26 Przybyłowski A (2008) Attractiveness goes far beyond. Baltic Transport J 5:20–21 Sequeira S (2012) Advances in measuring corruption in the field. Working Paper of London School of Economics, London Slack B (1985) Containerisation and inter-port competition. Marit Policy Manag 12(4):293–304 Tiwari P, Itoh H, Dio M (2003) Shippers’ port and carrier selection behaviour in China: a discrete choice analysis. Marit Econ Logist 5:23–39 Tongzon J (2002) Port choice determinants in a competitive environment. Proceedings of Annual IAME Meeting and Conference, Panama Tongzon J (2009) Port choice and freight forwarders. Transp Res Part E Logist Transp Rev 45(1): 186–195 Ullman JB, Bentler PM (2012) Structural equation modeling. In: Weiner IB (ed) Handbook of psychology, 2nd edn. Wiley, New York, NY UNCTAD (2011) Best practices in investment for development: case studies in FDI. United Nations Conference on Trade and Development, New York Wiegmans BW, Hoest AVD, Notteboom TE (2008) Port and terminal selection by deep-sea container operators. Marit Policy Manag 35(6):517–534

Chapter 6

Scenario Analysis of Shipping Networks: Consolidation

6.1

Introduction: Towards Consolidation of Cargo and Reduced Transport Costs

According to the Pacific Islands Forum Secretariat (2012), the shipping market in the SPICs is relatively contestable. Nonetheless, the trade flows are extremely imbalanced since imports usually far outweigh exports, and the freight rates are often significantly higher than in other parts of the world (ADB 2013). These inefficiencies in the trade markets certainly stem from the intrinsic spatial dispersion of the islands, their physical scale, thin trade flows, chronic infrastructure difficulties (above all relating to maintenance and operations) and limited trade connectivity in domestic and international markets. However, the available empirical evidence does not clearly address the entirety of the interrelated aspects which hamper trade, and for this reason we develop a scenario analysis in this chapter based on the topological structure of shipping networks in the SPICs. The objectives are to explore how to stimulate the consolidation of cargo and decrease transport cost by achieving economies of scale through the network structure. These objectives relate to our general hypothesis (1) in Chap. 4: Can the structure of the shipping network stimulate the consolidation of cargo and decrease transport cost? One of the main problems in the logistics chain of the SPICs, as we have observed, is the discrepancy between export-import cargo loading, that is, vessels travelling not fully loaded and as a result witnessing economic inefficiency in their operations. Shipping service revenues, by being generally set at a low margin, necessitate that vessels be fully loaded in order to sustain profitable operations. Therefore, in order to achieve economies of scope and density (Button 2010), and to overcome cargo imbalances and revenue loss, a possible solution is a focalised shipping routing through a hub-and-spoke trade network. The intuition behind the notion of the hub-and-spoke shipping network is that through its implementation it can be an attractive means of fostering efficient trade dynamics (Daganzo 2005; Beuth and Kreutzberger 2001). A hub-and-spoke structure is often able to © Springer International Publishing AG 2017 F.R. Medda et al., Collaborative Approach to Trade, Advances in Spatial Science, DOI 10.1007/978-3-319-47039-9_6

65

66

6 Scenario Analysis of Shipping Networks: Consolidation

effectively consolidate cargo between ports and shipping services and thus achieve high connectivity of operations through economies of scale and scope. Although it requires circuitous routings, in the hub-and-spoke network it is possible to achieve transport cost reductions and increases in trade flows (Taaffe et al. 1996). However, in our context it would be politically arduous for SPICs governments to implement the hub-and-spoke structure of trade exogenously, since we are considering different countries with their own jurisdictions and numerous shipping companies operating in the area. But such a strategy would indeed be feasible, for example, for shipping companies (Jourquin and Beuthe 1996). In our approach we assume the implementation of a policy of cargo consolidation in order to achieve economies of scale of the trade flows. This is a policy that can indeed be incentivised at national and regional levels and put in operation by port authorities. Given this assumption, we consider three networks under scenario 1, scenario 2 and scenario 3. We then explore which scenario stimulates the emergence of the hub-and-spoke structure for shipping trade in the SPICs. In this process of determining which scenario best supports the hub-and-spoke network structure we are able to develop and examine the policy implications for the SPICs region as a whole. It is important again to stress at this point that, in all three cases, the position of the hub will therefore be endogenously determined. In the first scenario, Existing Conditions, we consider the actual shipping network operating in the islands. In scenario 2, Fully Connected, we analyse a shipping network in which each port in the South Pacific region is directly connected with every port. In the third scenario, Multiplier Attachment, we consider the network built when imposing the multiplier attachment algorithm to all the ports in the region. In this third case, by using the multiplier attachment, we incorporate in the shipping network structure the connectivity between ports. In order to consider the effects of economies of scale of the trade flows, we adopt the idea of endogenous cargo consolidation (Horner and O’Kelly 2001; Jaillet et al. 1996), which is to introduce a discount factor in the transport cost. The discount factor is a function of the trade flows in each port and link, and thus accounts for the presence of economies of scale because the more trade transiting through the port or the link, the lower will be the discount factor related to the transport cost. By expressing the transport cost with the discount factor, we examine how the network can consolidate the cargo flow; that is, by means of an increase of average flow per link and with a reduction in the number of shipping links in the region. In our analysis of the three structures through hub-and-spoke, we estimate the impacts, in terms of total transport cost of the network, of the consolidation process in the logistics trade chain in the SPICs. The discount factor is measured as a non-linear function of the flow volume and thus follows our overarching hypothesis in this work: that non-linearity is an important factor for our trade analysis. We also assume that each port in the three scenarios is a possible hub, and that the total volume of trade is fixed. Moreover, each trade flow is assigned to a path that minimises the transport cost. The minimum path cost has been calculated according to the following steps: (1) we assume that each container vessel has the same technical characteristics; (2) we

6.1 Introduction: Towards Consolidation of Cargo and Reduced Transport Costs

67

split total trade Tij , k between two countries in trips of volume V; (3) we select a port of departure in origin country i and an arrival port in destination country j (if our sample origin and/or destination countries have more than one port, we randomly choose one of the ports at our disposal); (4) according to the Dijkstra Algorithm, each trip is assigned the path with the minimum transport cost; (5) we calculate the total traffic between ports as the sum of trade flows passing through two ports; (6) we iterate the process from (1) to (5) 1000 times and consider the average of the values from the 1000 iterations. The goal of step (6) is to eliminate the randomness introduced in step (3). In fact, the ‘Law of Large Numbers’ guarantees that the average of the results obtained from a large number of trials should be close to the expected value and will tend to become closer as more trials are performed. Another assumption that we need to consider relates to the transport cost. In a hub-and-spoke network every ship calling in each port in their indirect service not only must pay the port fee, but also other port dues such as loading and unloading, customs tariff, stevedoring fee, etc. Due to the significant lack of data, and in particular, the insufficient data on shipping fees and cargo movements in each port of the SPICs that we were able to collect, we assume that the total transport cost of each network for the three scenarios is equal to the average transport cost per link as obtained in the Fully Connected network, multiplied by the total number of links in each scenario. The Fully Connected network is the ‘ideal’ shipping network where we consider only direct routes. In this case the shipping companies do not incur the costs for each port call but will need to account for economic losses incurred when ships travel not fully loaded. These two costs are inversely correlated, in other words, the more calls that a vessel makes, the higher will be the total costs related to the port fees, but the lower will be the loss of revenue due to the cargo imbalance. We assume therefore that these costs are equally weighted and in the indirect routes the decision for a ship to call in a port is always made at breakeven point, therefore the indirect route is seen by the shipping companies in relation to costs as a direct route. This decision, given the aggregate level of analysis, will ease the comparability of the obtained results. Moreover, we assume that the port fee accounts for any period in port not exceeding 3 days and for vessels with constant tonnage. Finally, another important aspect of our study is that, by assuming that the hub location (similarly for multiple hubs) emerges from the definition of the network structure, we capture two notions of the hub port. The first is the notion of the port as a trade gateway, that is, as a main node for transhipment activity. The second notion is the port as origin and destination of significant trade flows to and from the regional economy of the port’s hinterland. This twofold definition of hub port allows us to define possible extensions of the analysis that investigate the processes of regional agglomeration economies stemming from the port’s economic development (Krugman 1991).

68

6.2

6 Scenario Analysis of Shipping Networks: Consolidation

Scenario 1: Existing Conditions Network

The Existing Conditions container shipping network has been constructed by merging different data sets on the container services that operate in the SPICs. In the network, each port represents a node and a direct link between two nodes/ports is generated if the two ports are linked by at least one of the existing container services. The links in the network are weighted in order to take into account the distance in kilometres between two ports. Moreover, we consider in building the network the information relative to the container services that operate between ports because we want to establish the set of transhipment points based on origin-todestination trips. The resultant network is shown in Fig. 6.1. In order to appreciate the topology of the Existing Conditions shipping network it is important to keep in mind for Fig. 6.1 that node positions do not convey the actual geography.

Fig. 6.1 Existing Conditions shipping network in Pacific RIM countries. SPICs countries are shown in red. Linked colours identify origin to destination trips. Node size is proportional to total number of connections

6.2 Scenario 1: Existing Conditions Network

69

Table 6.1 Relevant topological properties of Existing Conditions shipping network

Existing Conditions shipping network

Nodes 35

Density 0.092

Mean degree 3.11

Max degree 14

Average clustering 0.23

Fig. 6.2 Degree distribution (dots in the small frame), cumulative degree distribution (squares in the small frame) and log-log scale (dots in the large frame)

Five ports outside the SPICs have a higher number of connections than the rest of the ports: Brisbane (BNE), Tauranga (TRG), Singapore (SIN), Hong Kong (HKG), and Jakarta (JKT). Of the SPICs only Suva in Fiji (SUV), Lae in PNG (LAE) and Noumea in New Caledonia (NOU) have a high number of connections (9) compared to the average (mean) value equal to 3. In Table 6.1 we report some topological properties of the Existing Conditions shipping network. (The complete list of port acronyms for the SPICs expressed through 3-digit code is available in the Appendix to Part II). To evaluate the architecture of this network, we plot the distribution of number of connections per node (degree) in Fig. 6.2. The degree distribution is wellapproximated by an exponential distribution. The average topological distance between nodes in the network is 1.95. We have compared the topological features of the Existing Conditions network with those of a random network with an equal number of nodes and links: average clustering coefficient, 0.15, and average topological distance, 2.57, of a random architecture are very similar to the existing network. These results, together with the degree

70

6 Scenario Analysis of Shipping Networks: Consolidation

distribution analysis (Fig. 6.2), allow us to identify the Existing Conditions network as a Random network. In scenario 1 there are 86 links which are crossed by at least one trade flow. The maximum value is reached between Brisbane (Australia) and Lae (PNG). In the Existing Conditions network configuration (Fig. 6.3), the port of Lae in the PNG, due to its high throughput value, which is facilitated by the interconnections with other ports in the SPICs and with international ports such as Singapore, Tauranga (New Zealand) and Brisbane, has an advantage compared to the rest of the ports in the PNG. In view of its proximity to Brisbane, Port Moresby has gained significant traffic in recent years. We can therefore observe in the network the emergence of a composite hub forged by two nodes: Brisbane and Lae. In relation to the SPICs, Lae not only acts as a national hub for Papua New Guinea connecting

Fig. 6.3 Trade flows in the Existing Conditions shipping network. Node size is proportional to the total flow passing through each port. Link width is proportional to the total volumes moved between two ports

6.2 Scenario 1: Existing Conditions Network

71

Fig. 6.4 Trade flow values in the 109 links of the shipping line network. Values are expressed in thousands

the ports of Kimbe (KIM), Oro Bay (ORO) and Port Moresby (PMB), but Lae also constitutes a powerful hub for the SPICs region. In Fig. 6.4 we depict the histogram of frequency of trade flows for each link. The distribution is positively skewed being concentrated around the average value of US$240 billion. The distribution of the trade flows comprises 75% of the flow in the first class (0–200,000) and around 16% in the second class (200,000–400,000). We can now test the effect of economies of scale (using the method outlined in Chap. 4) on the Existing Conditions network (Fig. 6.5). We verify that the busiest link is between the ports of Brisbane and Lae in the PNG, and due to economies of scale, an additional 2.5% of trade traffic is assigned to this link. Figure 6.5 shows more plainly than Fig. 6.4 the strong influence of the PNG ports in the SPICs region, especially the significant dominant network position between the two ports of Lae in PNG, and the port of Brisbane in Australia. This influence, given the particular growth of the PNG, could ostensibly act as a magnet in relation to the economics of the entire SPICs region. We can deduce that the region may be developed under two economic speeds, either a faster double-digit growth by attracting major investment and trade as for PNG, or gradual development with slower, punctuated periods of growth on the other islands. Table 6.2 summarises the main economic characteristics of the shipping network in scenario 1. Total transport cost in the network amounts to US$84.7 million, which includes voyage costs and port fees for trade in the SPICs under the

72

6 Scenario Analysis of Shipping Networks: Consolidation

Fig. 6.5 Trade flows in the Existing Conditions shipping network with economies of scale. Node size is proportional to the total flow passing through each port. Link width is proportional to the total volumes moved between two ports Table 6.2 Major statistics of the Existing Conditions shipping network with economies of scale

S-1

Network configuration EXISTING CONDITIONS NETWORK

Max flow (106 US$) 1964

Average flow (106 US$) 137

No. of links 86

Total transport cost (106 US$) 84.7

hypothesis that all trade flows are containerised. Scenario 1 will be our baseline for the other two scenarios in the next sections in order to verify economies of scale and consolidation through potentially improved network topologies.

6.3 Scenario 2: Fully Connected Network

73

Fig. 6.6 Fully Connected shipping network in Pacific RIM countries. SPICs countries are shown in blue

6.3

Scenario 2: Fully Connected Network

In this scenario we build the Fully Connected shipping network where each port is directly connected with any other port (Fig. 6.6). This is an ‘ideal’ network because it is practically impossible to have a real-life situation where all the ports in a region are directly connected with each other. Nonetheless, this is an interesting case from which to examine policies towards greater connectivity for the ports through direct links. This network configuration allows us to remove the influence of network topology architecture. We can immediately note the obvious point that, compared to the scenario 1 network, Fully Connected has a higher number of links for the same number of ports/nodes. Table 6.3 summarises the major topological properties of the Fully Connected shipping network. Let us next examine the results of our analysis of the distribution

74

6 Scenario Analysis of Shipping Networks: Consolidation

Table 6.3 Relevant topological properties of Fully Connected shipping network

Fully Connected shipping network

Nodes 35

Density 0.5

Mean degree 35

Max degree 35

Average clustering 0.5

Fig. 6.7 Trade flows in the Fully Connected shipping network. Node size is proportional to the total flow passing through each port. Link width is proportional to the total volumes moved between two ports

of trade flows in the Fully Connected network. In Fig. 6.7, the visualised results, the average trade flow is about US$43 million per link, while the maximum value is US $490 million (Brisbane to Lae). As in the previous case, in this scenario we observe the emergence of a hub-and-spoke network structure, where the main hub is the port of Brisbane. This hub connects with two spoke systems, one with greater volumes and trade flow passing through it, and is comprised by all the ports in PNG: Kimbe

6.3 Scenario 2: Fully Connected Network

75

Fig. 6.8 Trade flow values for the Fully Connected shipping network. Values are expressed in thousands

(KIM), Oro Bay (ORO), Rabaul (RAB), Madang (MAG), Lae (LAE), and Port Moresby (PMB) plus the port of Noumea in New Caledonia. The second spoke system includes the other ports in the SPICs which have less influence due to their low levels of shipping traffic with the main hub. The histogram representing the frequency of trade flows in the shipping network (Fig. 6.8) has a similar shape as the case of the network in scenario 1. The majority of flows are depicted in the first two classes while the rest is spread thinly over the other classes. We can now discuss the effect of economies of scale on the Fully Connected shipping network (Fig. 6.9). We record that the total transport cost in this case is US $259 million, higher than the transport cost in scenario 1; this result was expected, since we have a higher number of links in this network. Moreover, we have a 27% decrease of traffic on the busiest link. The network shows a strong hub-and-spoke configuration around the port of Brisbane. In Fig. 6.9, more so than in Fig. 6.7, we can clearly see two sets of spokes, one with higher traffic composed in the SPICs by all the ports in PNG (Kimbe, Lae, Madang, Oro Bay, Port Moresby, Rabaul) and Noumea in New Caledonia, and the second set of spokes with lower traffic, consisting of the rest of the ports. Table 6.4 expresses in number what we can elicit from Fig. 6.9. We show an increase of total transport cost due to the increase in the number of links; both results are derived from our imposing direct links in this network. The nearly doubled increase in links from 86 to 171, compared to Existing Conditions, sends a clear signal that this type of network does not favour trade consolidation, which is one of our objectives. We observe that the trade flow is dispersed across the

76

6 Scenario Analysis of Shipping Networks: Consolidation

Fig. 6.9 Trade flows in the Fully Connected shipping network with economies of scale. Node size is proportional to the total flow passing through each port. Link width is proportional to the total volumes moved between two ports Table 6.4 Major statistics of the Fully Connected shipping network with economies of scale

S-2

Network configuration FULLY CONNECTED NETWORK

Max flow (106 US$) 542

Average flow (106 US$) 42

No. of links 171

Total transport cost (106 US$) 259

different links, for instance, the maximum flow in this network is US$542 million, whereas the scenario 1 network maximum flow is US$1964 million. However, like scenario 1, in scenario 2 PNG establishes with its ports the predominance in the region as major spokes in this network structure, thereby producing an imbalance in the trade flow, which is heavily directed towards the PNG ports. In our final scenario 3 we introduce the Multiplier Attachment to consider connectivity among ports.

6.4 Scenario 3: Multiplier Attachment Network

77

Fig. 6.10 Multiplier Attachment network. SPICs countries are visualised in blue. Node size is proportional to the total number of connections of each port

6.4

Scenario 3: Multiplier Attachment Network

Multiplier Attachment is the basis for the definition of the new architecture for the shipping network in the SPICs region. In this case, the gain in connectivity of a port is assumed to be given by the potential benefits that a node may receive from connecting to another particular node which increases its market opportunities or economic benefits. Since fuel consumption is one of the major costs in the transport cost in the region, we construct the Multiplier Attachment shipping network using distance as a proxy for transport cost. In this way we are able to account for the impact of space (i.e. distance) in the generation of the network. We apply an interaction range rc equal to the average distance between ports. Figure 6.10 depicts a hub-and-spoke Multiplier Attachment network structure constituted by two main hubs. One hub is situated in the SPICs, the port of Suva (SUV) in Fiji, and one hub is located outside the SPICs at the Port of Busan in South Korea. The other ports are linked to the two hubs through a series of routing links connecting the various ports. We observe that, in the spokes, the ports are relatively equally important in network connectivity.

78

6 Scenario Analysis of Shipping Networks: Consolidation

Table 6.5 Relevant statistics of the Multiplier Attachment network configuration using distance as an impedance factor to trade

Multiplier Attachment shipping network

Nodes 35

Density 0.096

Mean degree 3.1

Max degree 24

Average clustering 0.49

Fig. 6.11 Trade flows in the Multiplier Attachment network. Node size is proportional to the total flow passing through each port. Link width is proportional to the total volumes moved between two ports

We have not considered directionality of links (i.e. each link can be travelled in both directions) in order to allow for interconnectivity between any two ports. The network architecture in scenario 3 generated by the Multiplier Attachment allows for the creation of a less dense network (Table 6.5), which does not compromise the average topological distance between nodes; we can conclude that it improves the local connectedness of the network (average clustering coefficient).

6.4 Scenario 3: Multiplier Attachment Network

79

Fig. 6.12 Trade flow values in Multiplier Attachment network. Values are expressed in thousands

As noted, in the Multiplier Attachment network a hub clearly emerges within the SPICs region (Fig. 6.11), the port of Suva in Fiji connects with the other major port in the region, and the port of Brisbane. In this network the port of Suva exploits its spatial advantage of being central and of having high Port Attractiveness. The emergence of this hub may determine as a consequence the consolidation of cargo and traffic within the entire shipping network. This process is evident when we observe that a lower number of links (80) are involved in the trade flows. Flows are distributed in three classes: the first two classes include, respectively, 73% and 24% of the trade flows (Fig. 6.12). When we introduce the effect of the economies of scale in the Multiplier Attachment network (Fig. 6.13 and Table 6.6), we recognise the significant impact of port connectivity imposed in the network in terms of a more balanced distribution of shipping traffic flows. The maximum flow is shown between the hub of Suva and Brisbane with US $3.58 billion and average flow per link is US$3.68 trillion, which respectively, represents an increase of about 54% and 35% in relation to scenario 1, Existing Conditions. The port of Suva in the Multiplier Attachment network is not only connected with the ports in the SPICs but with the major international ports in the region such as Nagoya, Hong Kong, Singapore, and Brisbane. The total transport cost is the lowest among the three scenarios, with US$32 million. In this case we have been able through the construction of the network and the endogenous emergence of the hub-and-spoke structure to counteract the economic drawing power of PNG and achieve a more effective distributed level of traffic flow. In this scenario it is possible to trade so as to ultimately determine a more balanced economic growth for the SPICs.

80

6 Scenario Analysis of Shipping Networks: Consolidation

Fig. 6.13 Trade flows in the Multiplier Attachment network with economies of scale. Node size is proportional to total flow passing through each port. Link width is proportional to total volumes moved between two ports Table 6.6 Major statistics of the Multiplier Attachment network with economies of scale Max flow (106 US$) 3580

Average flow (106 US$) 368

No. of links 80

S-3

Network configuration MULTIPLIER ATTACHMENT NETWORK

6.5

Conclusions and Policy Recommendations

Total transport cost (106 US$) 32

In this chapter we have analysed three different scenarios focussing on how the shipping trade in the SPICs can be structured in order to improve consolidation of cargo and traffic, and in so doing, decrease the transport costs. We have considered three scenarios with three corresponding network structures and allowed for the emergence of the hub-and-spoke structure for the logistics chain, because this

References

81

Table 6.7 Major traffic statistics and total transport cost for the three network configurations

S-1 S-2 S-3

Network configurations Existing Conditions shipping network Fully Connected shipping network Multiplier Attachment shipping network

Max flow (106 US$) 1964

Average flow (106 US$) 137

No. of links 86

Total transport cost (106 US$) 84.7

542

42

171

259

3580

368

80

32

structure can more effectively achieve the objectives through economies of scale and foster agglomeration economy processes in the region. Table 6.7 summarises the obtained results. In all three scenarios we can observe that the hub-and-spoke structure emerges when considering the trade traffic and the economies of scale through a discount factor. In scenarios 1 and 2, the hub in the two networks is significantly located in PNG. PNG is one of the most promising and fastest-growing economies in the SPICs (ADB 2012). However, this structure may play in favour of an unbalanced growth trend for the entire region in which the ports of other South Pacific Island countries receive less shipping traffic, and end up carrying out less trade. In the third scenario we have considered a network where connectivity is the central objective. Scenario 3 shows the presence of the hub-and-spoke structure for shipping trade, but in this scenario the hub is located in the port of Suva in Fiji. Suva is located centrally within the ports of SPICs, thus significantly reducing the number of links necessary for trade to develop. We have demonstrated that consolidation in the SPICs ports is effectively obtained with a main trade link between the SPICs hub and the Australian port of Brisbane. The average traffic in the link is also increased notably in relation to the other scenarios, thus supporting the network structure as well as the hub location for fostering cargo bulking and consolidation of traffic. The governments of the SPICs could choose to incentivise the development of a comprehensive shipping hub-and-spoke structure for the whole SPICs region, for example, through the support of the Regional Maritime Programme based at the Secretariat of the Pacific Community. Implementation may, for instance, take the shape of a step-wise process using pilot initiatives such as the proposed hub-andspoke structure for shipment of specific agriculture goods. By pooling resources and know-how, the SPICs will be able to identify different priorities for hub-andspoke implementation which are consistent and coherent for the overall growth of the region and for the development of a more efficient logistics chain.

References Asian Development Bank (ADB) (2012) ADB Annual Report 2011. Asian Development Bank, Mandaluyong City, Philippines

82

6 Scenario Analysis of Shipping Networks: Consolidation

Asian Development Bank (ADB) (2013) Key indicators for Asia and the Pacific 2013. Asian Development Bank, Mandaluyong City, Philippines Beuth M, Kreutzberger E (2001) Consolidation and transshipment. In: Brewer AM, Button KJ, Hensher DA (eds) Handbook of logistics and supply chain management. Elsevier, Pergamon, pp 239–252 Button KJ (2010) Transport economics. Edward Elgar Publishing, Cheltenham Daganzo CF (2005) A variational formulation of kinematic waves: basic theory and complex boundary conditions. Transp Res B Methodol 39(2):187–196 Horner MW, O’Kelly ME (2001) Embedding economies of scale concepts for hub network design. J Transp Geogr 9:255–265 Jaillet P, Song G, Yu G (1996) Airline network design and hub location problems. Locat Sci 4:195–212 Jourquin B, Beuthe M (1996) Transportation policy analysis with a geographic information system: the virtual network of freight transportation in Europe. Transp Res C Emerg Technol 4(6):359–371 Krugman PR (1991) Geography and trade. MIT Press, Cambridge, MA Pacific Islands Forum Secretariat (2012) Review of the Pacific Islands Forum Secretariat. Draft Report. Suva, Fiji Taaffe E, Gauthier H, O’Kelly M (1996) Geography of transportation. Prentice-Hall, Englewood Cliffs, NJ

Chapter 7

Trade Coordination Agreements

7.1

Introduction

In Chap. 6 we analysed how different shipping networks can stimulate trade flows and facilitate the emergence of a hub-and-spoke structure. The hub-and-spoke offers significant gains in trade, because by implementing consolidation, it increases competition and economies of scale, reduces transport costs, and therefore fosters a collaborative approach to trade. In this chapter we extend the argument still further and test how different shipping network structures can increase trade flows when we implement trade coordination agreements. In other words, our objective here is based on hypothesis (2): Do trade coordination initiatives provide greater long-term economic growth for the SPICs? In testing hypothesis 2, we revisit the three scenarios of Chap. 6 and introduce two types of trade coordination agreements. In general, trade agreements offer numerous gains in the development of trade and growth within a region; in fact, not only may a region capitalise on economies of scale and comparative advantage, but also with increases in foreign and private capital and access to larger trade markets a region can realise more efficient and effective production. Being a participant in a trade agreement certainly generates positive ripple effects in education and infrastructure endowments and the overall income of a region increases. In this chapter we therefore combine two main concepts: trade agreement and network structure, in particular the hub-and-spoke structure. The roots of this approach are found in the work of Wonnacott (1996) and Enders and Wonnacott (1996) who demonstrate that the gains from hub-and-spoke logistics networks are reduced according to type of trade agreement implemented. Under bilateral agreements in particular, Wonnacott (1996) finds that “the hub gets a larger percentage share of the region’s total income, whereas the spokes get a smaller share.” We have reviewed the main trade coordination initiatives and agreements in the SPICs in the Appendix to Part II, and before we continue, it is important to underscore the complexity of policy on trade coordination agreements, particularly © Springer International Publishing AG 2017 F.R. Medda et al., Collaborative Approach to Trade, Advances in Spatial Science, DOI 10.1007/978-3-319-47039-9_7

83

84

7 Trade Coordination Agreements

for small countries. In general, trade coordination agreements can impose negative impacts and adjustment costs, especially over the short-run. Ebrill et al. (1999) demonstrate that the negative impacts of trade agreements such as trade liberalisation are greater in small countries because they rely largely on trade tariff revenues; for example, in the SPICs, trade tariffs in the Solomon Islands account for more than 54% of tax revenues. According to Stiglitz and Charlton (2010), emerging and developing countries are affected by trade liberalisation due to their vulnerability to policy shocks, their lack of diversification in trade portfolios, in some cases limited access to credit and low GDP per capita; and because they must comply with standards and international regulations, particularly in industries such as agriculture, where the structure of world trade sees the highest distortion in terms of tariffs. Against this background we reconsider the three cargo shipping scenarios previously examined: (1) Existing Conditions network, (2) Fully Connected network and (3) Multiplier Attachment network. In the first two cases we assume the implementation of bilateral agreements between a set of countries (PNG, Solomon Islands, Vanuatu, New Caledonia and Australia), and in the third case we assume a free trade agreement across all the countries in the SPICs. Due to the complexity of the topic and the lack of data on specific trade agreements, we introduce two simplifications in the analysis which nevertheless allow us to test our hypothesis and reach some interesting policy conclusions as well. First, we examine a distinct set of countries in the agreements; this set has been determined on the basis of the distance between the members. The agreements we implement in our scenarios have no connection to actual alternative agreements, which in some cases are still under deliberation among the SPICs. Our objective in this chapter is to focus on general policy. We are setting out to verify that, rather than to foster economic growth across the region, bilateral agreements tend to give advantages to the hub members of the agreement. Another reason for choosing a general objective is because we have scant information about the SPIC’s present and future agreements; it would therefore be incorrect to test an actual agreement and reach specific policy conclusions from this position. The second simplification applied to the three scenarios relates to tariffs. Tariffs are government-imposed taxes on trade, but also in this case we have little information on trade tariffs in the SPICs. However, we have observed that the ports in the SPICs are in the large majority under ownership and operation of the public sector; we can therefore satisfactorily consider port fees as a proxy for the trade tariffs. We will assume that trade agreements will forgo the port fees among the agreement members. As in the previous section, we examine the three scenarios and reach a number of specific policy recommendations for each one, which is then summarised.

7.2 Scenario 1: Existing Conditions Network with Bilateral Agreement

7.2

85

Scenario 1: Existing Conditions Network with Bilateral Agreement

In the first case we consider the Existing Conditions of the shipping trade in the SPICs and the emergence of the hub-and-spoke structure mainly centred in the PNG. We test the effect of a bilateral agreement among the ports in PNG, Solomon Islands, Vanuatu, New Caledonia, and Australia. In this scenario we assume that some but not all trade barriers and port fees are removed. As in the previous section, scenario 1 is the baseline scenario for the other two scenarios, in this case we analyse the state-of-play situation in terms of shipping logistics networks and aggregate level of trade volume, and simply impose a regime of bilateral trade. The results show that the busiest link is Brisbane–Lae with an annual value of trade goods at US$1750 million. The average value of trade flow in the 88 links is on the order of US$123 million (Fig. 7.1).

Fig. 7.1 Trade flows in the Existing Conditions network with agreement among PNG, Solomon Islands, Vanuatu, New Caledonia and Australia. The acronyms of the ports in countries part of the agreement are: NOU, NOR, VLI, BNE, HIR, MAG, KIM, ORO, RAB, PMB, LAE. Node size is proportional to total flow passing through each port. Link width is proportional to total volumes moved between two ports

86

7 Trade Coordination Agreements

Fig. 7.2 Trade flow values in the Existing Conditions network with agreement among PNG, Solomon Islands, Vanuatu, and New Caledonia. Values are expressed in thousands

The distribution of flows is positively skewed and concentrated in the first class (0–200,000). The other seven classes absorb a total of 25% of flows (Fig. 7.2). Given the presence of the hub-and-spoke structure, we examine the impacts of the economies of scale in the shipping network (Fig. 7.3). In this way we can more appropriately test the hypothesis of Wonnacott (1996), who demonstrates that the total collective income in hub-and-spoke with bilateral agreements is restricted in relation to a free trade agreement regime, because in the case of hub-and-spoke, different inefficiencies and complications such as excess tariffs are inherent in the trade operations. By considering the removal of trade barriers only for a part of the islands, we expect fewer opportunities to occur for the SPICs region, and this may be attributed to a decrease in competition and economies of scale (Dasgupta and Stiglitz 1980). The results of the Existing Conditions scenario are depicted in Figs. 7.3 and 7.4, and set out in Tables 7.1 and 7.2. In Table 7.1 we show the results aggregated at country level to verify the major benefits acquired under this bilateral trade agreement. This scenario, as in the remainder of the chapter, is presented in two configurations: (a) and (b). Configuration (a) represents the network with trade coordination agreement and economies of scale; configuration (b) represents the network with economies of scale but without trade coordination agreement. We also consider two countries in Table 7.2, Fiji and Tonga, which, although not part of the examined agreement, we nevertheless verify and record the negative or positive impacts imposed by the bilateral agreement on the excluded countries. When we look at Table 7.2 we can verify the economic trend in this agreement regime and logistics network structure. In the hub-and-spoke structure the main

7.2 Scenario 1: Existing Conditions Network with Bilateral Agreement

87

Fig. 7.3 Trade flows in the Existing Conditions network with economies of scale and agreement among PNG, Solomon Islands, Vanuatu, and New Caledonia. The acronyms of the ports in countries part of the agreement are: NOU, NOR, VLI, BNE, HIR, MAG, KIM, ORO, RAB, PMB, LAE. Node size is proportional to total flow passing through each port. Link width is proportional to total volumes moved between two ports

benefits in terms of trade volume, and thus economic growth, will be accrued to the hub, and the expenses will devolve to its spoke partners. In our case, as we have observed in Chap. 6, PNG is acquiring a predominant position among the SPICs and in the overall structure of the SPICs logistics chain. Moreover, the implementation of these aforementioned bilateral agreements will merely accelerate the trend already in motion, that is, a preferential trading position for PNG (the hub) and declines in trading traffic for the other islands in the SPICs. We can note from the results in Table 7.2, as expected, that the worst position in terms of trade volume is thrust upon the countries outside the trade agreements, i.e. Fiji and Tonga. The greatest total variation in trade flow is accrued by PNG, with relatively positive variations for the Solomon Islands and New Caledonia (spokes). In the next section we examine the ‘ideal’ network structure of a Fully Connected network and verify the resulting changes in the case of the implementation of a trade agreement regime, as we have here for scenario 1.

88

7 Trade Coordination Agreements

Fig. 7.4 Trade flow values in the Existing Conditions network with economies of scale and agreement among PNG, Solomon Islands, Vanuatu, and New Caledonia. Values are expressed in thousands Table 7.1 Configurations of the Existing Conditions network Network configurations (a) Existing Conditions network with bilateral agreement and economies of scale (b) Existing Conditions network with economies of scale

Max flow (106 US$) 1727

Average flow (106 US$) 137

No. of links 88

1964

137

86

Table 7.2 Throughput for SPICs ports in the Existing Conditions network

Country PNG

Solomon Islands New Caledonia Fiji Tonga

Port Kimbe Lae Madang Oro Bay Moresby Rabaul Honiara Noro Noumea

Throughput configuration (a) (1000 US$) 635 3799 697 593 2453 665 240 80 1052

Throughput configuration (b) (1000 US$) 488 3806 608 531 2303 689 303 77 1229

Total variation (a) (b) 5%

Suva Lautoka Nukuofola

1105 296 151

1076 263 151

5%

16% 15%

0%

7.3 Scenario 2: Fully Connected Network with Bilateral Agreement

7.3

89

Scenario 2: Fully Connected Network with Bilateral Agreement

In scenario 2, as in Chap. 6, we have examined the network where each port is connected with any other port by a direct route. For this reason we have observed that this network topology may be considered as an ‘ideal’ network. But in this analysis we will examine the trade traffic impacts produced by the implementation of a specific bilateral agreement. We assume the same bilateral trade agreement, as in the Existing Conditions network, to include PNG, Solomon Islands, Vanuatu, New Caledonia, and Australia (Fig. 7.5). Also here as in the previous case, we note that the network is characterised by a hub-and-spoke shipping logistics structure where the hub is located in the port of Brisbane, and the main spokes are ports under the bilateral agreement. The

Fig. 7.5 Trade flows in the Fully Connected network with agreement among PNG, Solomon Islands, Vanuatu, New Caledonia, and Australia. The acronyms of the ports in countries part of the agreement are: NOU, NOR, VLI, BNE, HIR, MAG, KIM, ORO, RAB, PMB, LAE. Node size is proportional to total flow passing through each port. Link width is proportional to total volumes moved between two ports

90

7 Trade Coordination Agreements

Fig. 7.6 Trade flow values in the Fully Connected network with economies of scale and agreement among PNG, Solomon Islands, Vanuatu, New Caledonia, and Australia. Values are expressed in thousands

histogram of the frequency of the trade flows concentrates mainly in the first class and the rest of the traffic is spread across the rest of the classes, as depicted in Fig. 7.6. Given the structure of the shipping logistics networks, we examine the impact of the economies of scale through Figs. 7.7 and 7.8. Both figures show that the bilateral trade agreement increases the frequency of traffic flow, mainly into the first class, i.e. the ports under bilateral agreement, thus sending a signal about the hub’s predominance in the flow of trade, the port of Brisbane. Let us next discuss the results relative to the traffic flows. With the bilateral agreement a larger share of the trade traffic passes through the members of the agreement, but the maximum and average flows are still lower than the values in the baseline scenario (Table 7.3). This result is due to the effect of dispersion that the Fully Connected network has imposed on the flow. In scenario 2 we observe the two sets of spokes: the first set consists mostly of the members of the bilateral trade agreement. The second set, which is less important when we consider volume of traffic, corresponds to all the countries not belonging to the agreement. In the SPICs, PNG acts as an attractor of trade traffic and in this way secures its strongly dominant position in the whole region. To support this observation, we can examine the results relative to the total variation of trade for PNG before and after agreement, which is at 8%. However, results for the countries participating in the agreement are not always consistent with an increment, particularly in the case of the Solomon Islands and Fiji (Table 7.4).

7.3 Scenario 2: Fully Connected Network with Bilateral Agreement

91

Fig. 7.7 Trade flows in the Fully Connected network with economies of scale and agreement among PNG, Solomon Islands, Vanuatu, New Caledonia, and Australia. The acronyms of the ports in countries part of the agreement are: NOU, NOR, VLI, BNE, HIR, MAG, KIM, ORO, RAB, PMB, LAE. Node size is proportional to total flow passing through each port. Link width is proportional to total volumes moved between two ports

Once again in scenario 2, with the introduction of the bilateral trade agreement, PNG gains a larger percentage share of the region’s total trade flow and consequently of the economic growth. Conversely, the other spokes in the agreement, receive a smaller share than the PNG spokes. However, if the port/country is not part of the trade agreement, it belongs to the second set of spokes where the share of trade traffic is even smaller, and is in fact, the smallest of the SPICs traffic share. The final outcome of scenario 2 is that the two sets of spokes in the SPICs, with and without trade agreement, create competing trade blocs. The spokes in the bilateral agreement are always negatively impacted by hub predominance and discrimination arising from the preferential advantages of the hub as core and main recipient of trade traffic.

92

7 Trade Coordination Agreements

Fig. 7.8 Trade flow values in the Fully Connected network with economies of scale and agreement among PNG, Solomon Islands, Vanuatu, New Caledonia, and Australia. Values are expressed in thousands

Table 7.3 Configurations of the Fully Connected network Network configurations (a) Fully Connected network with bilateral agreement and economies of scale (b) Fully Connected network with economies of scale

Max flow (106 US$) 508

Average flow (106 US$) 40,5

No. of links 175

542

42

172

However, it is important to notice that the hub in this network configuration with bilateral agreement also has fewer opportunities in terms of comparative advantage and increased production efficiency, evidenced by the decrease in the overall income into the region due to the dispersion of traffic. The combination of huband-spoke structure and bilateral trade agreement nevertheless determines a preferential trade pattern for the hub, and consequently an unbalanced trade development for the SPICs with a distorted distribution of income.

7.4 Scenario 3: Multiplier Attachment Network with Free Trade Agreement

93

Table 7.4 Throughput for SPICs ports in the Fully Connected shipping network

Country PNG

Solomon Islands New Caledonia Fiji Tonga

7.4

Port Kimbe Lae Madang Oro Bay Moresby Rabaul Honiara Noro Noumea

Throughput configuration (a) (1000 US$) 613 953 817 922 669 770 177 179 843

Throughput configuration (b) (1000 US$) 599 809 816 917 490 767 296 152 842

Suva Lautoka Nukuofola

273 478 107

404 432 90

Total variation (a) (b) 8%

25% 0% 11% 16%

Scenario 3: Multiplier Attachment Network with Free Trade Agreement

In our final third scenario of the analysis, we consider the Multiplier Attachment network whose hub is in the port of Suva (Fiji) and where we impose a free trade agreement across all the islands in the SPICs (Figs. 7.9 and 7.10). The coordination trade agreement implementation in this case of scenario 3 occurs by removing port fees. The port fees, as in the previous two cases, are assumed as a proxy of the trade tariff. Using this scenario which we have earlier demonstrated in Chap. 6, we can observe that, by introducing a network based on Multiplier Attachment, the huband-spoke logistics shipping network is endogenously determined around the port of Suma, thus creating a balancing mechanism in the economic development of the entire region. By imposing a free trade status and therefore removing all trade barriers, we aim to induce a greater collective trade income for all the SPICs that adhere to the collaborative trade approach. In Figs. 7.11 and 7.12 we depict the situation of the Multiplier Attachment network with free trade agreement and introducing economies of scale. The effects of the economies of scale enhanced the previous observation where the consequences of the topological structure of the logistics trade are shown in the histogram: one could easily observe a hierarchical frequency of trade flows across the three classes.

94

7 Trade Coordination Agreements

Fig. 7.9 Trade flows in the Multiplier Attachment network with free trade agreement in the SPICs. Node size is proportional to total flow passing through each port. Link width is proportional to total volumes moved between two ports

But clearly in Fig. 7.11 below a hierarchy network is no longer visible. Tables 7.5 and 7.6 summarise the results of the trade flow and increment of trade, and thus of income, after the implementation of the free trade agreement in the SPICs. A striking observation from Table 7.5 is that, compared to the other scenarios, the max traffic flow and average traffic flow indicate an increase, but the most significant impact in terms of flow is indeed attributed to the structure of the shipping trade network favouring economies of scale. This result is significant in that with the given assumptions we can conclude that the structure of the network is extremely influential in fostering economies of scale and efficient shipping operations. Importantly, the introduction of the agreement does not have a relevant impact in this context in terms of increasing flows for the hub, but rather it has more redistributive effects. At the aggregate level we therefore achieve the expected increase in trade and thus an overall increase in trade income for the SPICs.

7.4 Scenario 3: Multiplier Attachment Network with Free Trade Agreement

95

Fig. 7.10 Trade flow values in the Multiplier Attachment network with agreement in the SPICs. Values are expressed in thousands

Fig. 7.11 Trade flows in the Multiplier Attachment network with free trade agreement in the SPICs and economies of scale. Node size is proportional to total flow passing through each port. Link width is proportional to total volumes moved between two ports

96

7 Trade Coordination Agreements

Fig. 7.12 Trade flow values in the Multiplier Attachment network with free trade agreement in the SPICs and economies of scale. Values are expressed in thousands

Table 7.5 Configuration of the Multiplier Attachment network Network configurations (a) Multiplier Attachment network with free trade agreement and economies of scale (b) Multiplier Attachment network with economies of scale

Max flow (106 US$) 3580

Average flow (106 US$) 360

No. of links 82

3580

368

82

Table 7.6 Throughput for ports SPICs in the Multiplier Attachment network

Country Fiji PNG

Solomon Islands Tonga

Port Lautoka Suva Kimbe Lae Madang Oro Bay Port Moresby Rabaul Honiara Noro Nukualofa

Throughput configuration (a) (1000 US$) 398 11,713 889 1073 752 633 509

Throughput configuration (b) (1000 US$) 490 12,053 846 966 955 645 457

890 265 224 86

984 275 214 86

Total variation (a) (b) 3.5% 2%

0% 0%

7.5 Conclusions and Policy Recommendations

97

When we examine the trade traffic differential with and without the free trade agreement, two interesting patterns appear. First, this hub-and-spoke structure, together with total trade liberalisation, creates an increase in total traffic for the SPICs, which corresponds to improved access to larger and international markets. These noteworthy increases in trade flow are capitalised in new trade opportunities and thus in fostering economic growth and agglomeration economies based on comparative advantage across the SPICs. But above all, given the greater dispersion of the traffic flow among all the ports in the SPICs, we are able to counterbalance the potentially negative impacts of the hub as the dominant attractor of the trade flow. In this solution, the hub is an effective operational node of the network that eases the traffic flow and facilitates trade. We can conclude that in scenario 3 we are able to achieve an overall more balanced distribution of income and economic growth than in scenarios 1 and 2 (Stolper and Samuelson 1941). The second pattern appearing in this scenario is that PNG still achieves the greatest trade growth in the SPICs. In this solution although PNG does not maintain hub status as in previous scenarios 1 and 2, it is nevertheless the leader of trade growth in the region. The reason for this second pattern is found in the greater collective trade income accrued to all of the SPICs. The increased income is spread in the various SPICs, but given the positive economic growth of the PNG, the additional income is still captured by PNG with more standard trade gains related to comparative advantage and more non-standard gains associated with the introduction of new technologies to improve production and trade efficiency.

7.5

Conclusions and Policy Recommendations

We can conclude this analysis on trade coordination agreements by observing that the results obtained here echo our results in Chap. 6, which is that, overall when we compare scenario 3 in relation to the ‘state-of-play’ (scenario 1: Existing Conditions) and the ‘ideal’ Fully Connected network (scenario 2), scenario 3: Multiplier Attachment network, is able to achieve a more balanced growth of trade traffic. In all three scenarios the hub-and-spoke logistics shipping structure in general offers greater advantages for trade (Table 7.7). Other things being equal, investors and firms will prefer to operate in a similar shipping network structure because of the positive economies of scale that can be accrued. The introduction of a bilateral trade agreement augments this effect because investors will look at the SPICs as a coordinated system. However, in scenarios 1 and 2, the hub, by being a preferential location from which to have access to all the SPICs, receives a greater share of traffic trade. In the case of scenario 3, the dominant position of the hub and the strong pulling forces toward investment and agglomeration economies are less evident, because by implementing a free trade agreement any location in the

98

7 Trade Coordination Agreements

Table 7.7 Major statistics on the effect of trade liberalisation

S1 S1 S2 S2 S3 S3

Network configurations Existing Conditions network with economies of scale Existing Conditions network with bilateral agreement and economies of scale Fully Connected network with economies of scale Fully Connected network with bilateral agreement and economies of scale Multiplier Attachment network with economies of scale Multiplier Attachment network with free trade agreement and economies of scale

Max flow (106 US$) 1727

Average flow (106 US$) 137

No. of links 88

1964

137

86

508

40,5

175

542

42

172

3580

360

82

3580

368

82

SPICs is given the opportunity to expand its trade through its free access to input and markets. Not only do we observe an increase in total trade income but we also derive a more balanced distribution of income, because here, agglomeration economies will not be driven by a preferential position in the logistics chain (hub), but rather by lowest cost and most attractive location within the entire SPICs region. It is nevertheless of immense importance to reiterate that trade agreements towards trade liberalisation will produce significant adjustment costs in the SPICs, particularly over the short-term. Better understanding of the costs of trade liberalisation is crucial in counteracting the negative consequences for growth and to gain political support for trade reforms. The negative effects related to adjustment costs can be mitigated by effective national and international policies to reduce the costs and facilitate adjustments, in for example, the creation of agencies to ease credit in order to support the growth of enterprise. Technical assistance would be beneficial for developing the required physical and institutional infrastructure and for providing compensation to decrease the costs impacting on some groups affected by the trade reforms, for example, those with vested interests in trade. The SPICs should therefore negotiate trade agreements that cover the entire region. Promoting such a policy would prevent bilateral agreements that cater to the special needs of each country. Hub-and-spoke logistics with bilateral agreements will offer greater advantages to the hub and will raise transaction costs for the spokes and countries outside the bilateral agreements (legal and administrative costs, for instance). But in particular we may witness an increase in the incentives for rent seeking waste, “especially by the firms in the hub seeking to influence the selection of countries or sequence of countries that will provide them with the greatest preferential benefits” (Wonnacott 1996). Governments in the SPICs should loosen regulations and foster port privatisation in the aim to improve efficiency and reduce domestic monopoly positions, and thus allow for greater competition. For this reason, a stepwise approach in relation to trade coordination agreements for specific goods may be more advisable than a complete free trade implementation. This type of policy will determine an incremental type of competition in

References

99

response to the incremental pattern of liberalisation of the trade domain, thus preventing unfettered competition in all markets and lowering the high adjustment costs.

References Dasgupta P, Stiglitz J (1980) Uncertainty, industrial structure and the speed of R&D. Bell J Econ 11(1):1–28 Ebrill L, Stotsky J, Gropp R (1999) Revenue implications of trade liberalization. IMF Occasional Paper, No. 180. International Monetary Fund, Washington, DC Enders A, Wonnacott RJ (1996) The liberalisation of East-West European trade: hubs, spokes and further complications. World Econ 19(3):253–272 Stiglitz JE, Charlton A (2010) Fair trade for all: how trade can promote development. Oxford University Press, Oxford Stolper W, Samuelson PA (1941) Protection and real wages. Rev Econ Stud 9(1):58–73 Wonnacott RJ (1996) Free-trade agreements: for better or worse? Am Econ Rev 86(2):62–66

Chapter 8

Synthesis: The Integrated Multilayer Model

8.1

Introduction

The purpose of this chapter is to integrate the previous chapters in relation to policy analysis. We start by examining the vertical interactions within the Multilayer model framework, and in doing so, we return to hypothesis 3: How can trade be facilitated and investment and growth take place by leveraging transport logistics accessibility as well as the economic and sociological factors of the SPICs? In Chap. 5, we discussed on the basis of hypothesis 3, how to evaluate Port Attractiveness in order to foster economic growth and investment. Thereafter in Chaps. 6 and 7, we analysed the topological structure of the networks and identified the impacts of economies of scale and trade agreements. Our next endeavour is to integrate these different layers by connecting the factors already highlighted in the Port Attractiveness Index, and above all to test the analyses and results obtained in Chaps. 5, 6 and 7 for the networks.

8.2

The Vertical Network Model

The proposed model investigates interdependencies among layers in order to evaluate their effect on trade, thus we calibrate the model using the analytical specification, described in Chap. 4, of the parameters (Eq. 4.23), as in Wilson (1970). We calculate the bilateral transport costs between two islands in the SPICs using both exponential (Eq. 4.19) and power decay functions (Eq. 4.20). For bilateral distance we use the formulation provided by the CEPII dataset, which is calculated following the great circle formula, of latitudes and longitudes of the most important cities/ agglomerations (in terms of population). In the model we aim to show that the impact of transport between countries, and therefore trade impediment, has a major

© Springer International Publishing AG 2017 F.R. Medda et al., Collaborative Approach to Trade, Advances in Spatial Science, DOI 10.1007/978-3-319-47039-9_8

101

102

8 Synthesis: The Integrated Multilayer Model

effect on the intensive margin of trade. Moreover, since our countries have low income per capita, we would expect a significant impact of distance on trade flows. Within the framework of the Multilayer model in the Appendix to Part II, we introduced the various economic and social layers as the network of bilateral Trade Agreements in the region (TA). We use Cultural Links (Cult) and Common Language (Lang) as a proxy for the sociological linkages among countries. Since we assume that a change in the shipping network (physical layer) determines a change in transport cost, we calculate the model by examining different impedance factors: geographical distance, transport costs calculated in the Existing Conditions network with economies of scale and transport cost calculated in the Multiplier Attachment network with economies of scale. The results depicted in the tables below demonstrate that both exponential and power decay functions fit our data sample well, all having very high Pearson coefficients of correlation (>0.91). These results are explained by the geographic scale of our case study. Previous research has shown that, while exponential decay function has proven to be suitable for short-medium range interactions, as cost is perceived linearly, power decay functions provide better fits with long range interactions because in that case, cost is perceived logarithmically. In our case we are dealing with an interregional scale (range of travel distance: about 2500 km), making both exponential and power decay functions suitable for the study of trade in SPICs and RIM countries. When we introduce economic, cultural and physical layers we show how, by introducing Trade Agreements (TA), Common Language (Lang) and Cultural Links (Cult) into the model, the statistics improve. The simultaneous inclusion of the three layers provides the best fit with models based on distance (Table 8.1) or transport cost (Tables 8.2 and 8.3) as impedance functions. Due to the redundancy of the results between exponential and power decay functions, we have calibrated models with transport cost using mainly exponential decay function, and power decay function in just a few cases as control tests (Tables 8.2 and 8.3). Finally, it is worth noting that the calibration of the model shows that it is statistically indifferent if we use distance or transport cost calculations. This result is explained by the fact that we are estimating bilateral trade for aggregated products in an interregional market where distance is highly correlated with the cost of transport. Finally, we use OLS regression analysis to test the effect of independent layers on the trade layer in the Vertical Network model. In Tables 8.4 and 8.5 we report the results of Table 8.1 Parameters estimates (β coefficient) and Pearson coefficient of correlation (in parentheses) using distance as impedance transport factor

TA Lang Cult Cult, Lang Cult, TA Lang, TA TA, Lang, Cult

Exponential (dij) 0.000202 (0.914) 0.000205 (0.978) 0.000202 (0.994) 0.000197 (0.983) 0.000197 (0.983) 0.000199 (0.983) 0.000205 (0.989) 0.000199 (0.995)

Power decay (dij) 1.344 (0.926) 1.358 (0.945) 1.368 (0.968) 1.383 (0.986) 1.40 (0.987) 1.386 (0.978) 1.37 (0.972) 1.399 (0.982)

8.2 The Vertical Network Model

103

Table 8.2 Parameters estimates (β coefficient) and Pearson coefficient of correlation (in parentheses) using transport cost (calculated using the Existing Conditions network with economies of scale) as impedance factor

TA Lang Cult Cult, Lang Cult, TA Lang, TA TA, Lang, Cult

Exponential (TCij) Existing Conditions network 5.19  106 (0.93) 5.26  106 (0.95) 5.18  106 (0.97) 5.06  106 (0.97) 5.07  106 (0.97) 5.13  106 (0.98) 5.26  106 (0.97) 5.12  106 (0.98)

Power decay (TCij) Existing Conditions network 0.91 (0.95) – – – – – – 0.95 (0.97)

Table 8.3 Parameters estimates (β coefficient) and Pearson coefficient of correlation (in parentheses) using transport cost (calculated using the Multiplier Attachment network with economies of scale) as impedance factor

TA Lang Cult TA, Lang, Cult

Exponential (TCij) Multiplier Attachment network 3.26  106 (0.98) 3.67  106 (0.97) 3.4  106 (0.982) 3.4  106 (0.988) 3.38  106 (0.99)

Power decay (TCij) Multiplier Attachment network 0.88 (0.97) – – – 0.93 (0.98)

Table 8.4 OLS of Trade Model with Bilateral Trade (dependent variable) and independent variables: Total Export of exporting country (E), Total Import of importing country (I), distance (Dist), Cultural Linkage (Cult), Trade Agreements (Agreem) and Common Language (Lang) Model indep. variablesa,b E I Dist Cult Agreem Lang

B (unstandardised) .837 .483 1.024 4.369 1.667 .424

Std. error (unstandardised) .074 .096 .191 1.382 .487 .496

B (standardised) 1.204 .712 .915 .061 .094 .020

T 11.334 5.021 5.357 3.161 3.423 .855

Sig. .000 .000 .000 .002 .001 .394

a

Dependent variable: bilateral trade Linear regression through the origin

b

the OLS regression analysis. The model has high values of adjusted R square and low standard error estimate (Table 8.5). The coefficients of the layers are significant, with the exception of Common Language. Coefficient values and signs of Total Imports, Total Exports of trading countries, as well as Cultural Links and

104

8 Synthesis: The Integrated Multilayer Model

Table 8.5 Model summary

R .966a

R2 .933

Adjusted R2 .931

Std. error of the estimate 2.2715

a

Predictors: export, imports, distance, language, cultural link, trade agreement

Common Language are all positive. This means that, as expected, distance impacts negatively on trade. High values of trade are highly probable between countries with high values of Total Imports and Total Exports. Cultural Links and Common Language also positively influence bilateral trade, and this result is in line with the analyses of Chap. 7, where we showed that a trade liberalisation regime rather than a bilateral trade agreement can contribute to increased economic growth and trade, particularly with the Multiplier Attachment network. In the next section we consider another important aspect worthy of scrutiny in the Multilayer model: the effects of exogenous shocks. We will verify the behaviours of the integrated Multilayer model for the SPICs, in particular its vulnerability and thus its level of resilience.

8.3

Evaluation of External Shocks on Trade

Between 1980 and 2009 the Asia-Pacific region accounted for 45% of worldwide natural disasters. As a result of these catastrophes, the region incurred 42% of economic losses and suffered 86% of disaster-related deaths, gaining a reputation as the world’s most disaster-prone region. More recently, the 2011 earthquake in Christchurch (New Zealand) was followed by the devastating earthquake and tsunami in Japan that same year. And, in 2016 Japan recorded yet another earthquake with serious aftershocks. In our SPICs countries, severe flooding occurred in 2008 in Papua New Guinea and in Fiji in 2012. By the end of 2011, damage and losses recorded for the whole of the Asia-Pacific region had spiralled into the hundreds of billions of dollars (approximately US$267 billion). When natural disasters occur, they typically hit the poor and vulnerable the hardest, since they tend to live in the most exposed areas. Natural disasters do not respect national borders and thus often affect many countries. But even when physical damage is limited to a single country, disruption of the global supply chain can transmit economic impacts to other countries across the region. For example, the 2011 earthquake and tsunami in Tohoku, Japan, impacted strongly on auto and electronics industries across the Asian region due to scarcity of certain critical parts. Similarly, floods in Thailand shut down a major producer of computer hard-drive components, impacting on both regional and global computer industries. Droughts and floods also often result in crop losses, potentially increasing regional and global food prices and heightening food insecurity.

8.3 Evaluation of External Shocks on Trade

105

Fig. 8.1 Plot of randomly removed links versus total travelled distance and linear fitting trend line between 16% and 52% of the removed links (R2 ¼ 0.99) and after 52% (R2 ¼ 0.99)

We think it is also necessary to study effects of natural disasters on trade by simulating the impacts that external shocks (natural disaster, illicit attacks, etc.) would generate on the Existing Conditions shipping network. External shocks are simulated by impacting on links; we ‘attack’ the shipping network by deleting links and then evaluating the generated impacts by calculating the total travelled distance in the network. We employ two approaches, first we randomly attack the network, and second, we attack the connections (by removing the links) of important nodes and then evaluating the impact on total travelled distance. Important nodes are defined as those with the highest transhipment potential. External shocks are simulated by randomly removing links in the shipping network; we thereafter estimate the impact that removed links have created on the total travelled distance. Random removal of nodes is carried out in the aim to simulate the real-life introduction of local cumulative shocks in the shipping network. Figure 8.1 shows the results of our simulation of three regimes: (1) between 1% and 15% of removed links we record no high variations in the total travelled distance. This means that the Existing Conditions network with its current architecture is likely to be able to sustain a loss of about 15% of its connections without incurring major costs. This result informs us that the network has a redundancy of links, a sign of a non-optimized network. In theory, around 15% of links could be removed without generating an increase of transport cost and would likely reduce shipping companies’ operational and logistics costs. Regime (2) emerges between 16% and 52%, and regime (3) appears after 52% of links are removed. We have fit both regimes with linear regressions (R2 ¼ 0.99 in both cases), but with a slope four times greater after 52% of links are removed. In other words, after the threshold of

106

8 Synthesis: The Integrated Multilayer Model

Fig. 8.2 Potential Transhipment of ports in the Existing Conditions network

52%, we record an increase four times bigger per each increase of percentage of removed links. In the second part of our resilience analysis we evaluate the effects of external shocks on targeted ports. To do so, we have already calculated the Potential Transhipment as discussed in Chap. 4 (Eq. 4.15). The Potential Transhipment provides us with a measure of the importance of a port for transhipment. Not surprisingly, ports Suva, Brisbane, Tauranga, Singapore, and Lae have high Potential Transhipment (Fig. 8.2). We apply the results of the Potential Transhipment analysis by testing the effects of shocks on important ports for transhipment. By progressively removing links of ports Suva, Brisbane, Tauranga, and Lae, we are able to simulate the effect of an external shock on targeted parts of the network. Results show that when we remove links from important nodes of our network, an increase in total travelled distance in the network can be observed immediately (Fig. 8.3). We have found two different regimes which both follow a linear regression. The second regime starts at 26% of removed links and its slope is three times greater than the first regime. We can conclude that the Existing Conditions network shows some redundancy in the level of connections which may allow the system to sustain external shocks of up to 15% of the entire system. This finding is valid in general but does not apply to every part of the network. In fact, when we attack the network selectively, we record an increase of total travelled distance, even after a 3% removal of links. The

8.4 Interactions Between the SPICs

107

Fig. 8.3 Plot of removed links of ports in Suva, Brisbane, Tauranga, and Lae versus total travelled distance, and linear fitting trend line between 1% and 25% removed links (R2 ¼ 0.99) and after 25.5% (R2 ¼ 0.98)

implication here is that the network has concentrated areas of traffic, therefore a loss of these important links could disrupt the network and generate higher trade costs across the entire region.

8.4

Interactions Between the SPICs

There is consensus in the literature and among practitioners that improving market access leads to an increase in trade and thus in economic growth. However, as we have observed in this part of the book, trade in lagging regions such as the SPICs suffers from numerous constraints that hinder market opportunities. To get closer to resolving our challenge, we have proposed the method we call Multiplier Attachment to examine interactions between nodes in a network. In this section we want to verify how the Multiplier Attachment network, which we have confirmed is characterised by the hub in Fiji, behaves in terms of trade flow when compared with the Existing Conditions network centred in PNG. In order to achieve this objective, we calculate the accessibility among ports and islands of the SPICs. In our definition of accessibility, we focus on the level of interaction among ports which are trading (export and import) in the SPICs. Accessibility is therefore formulated in terms of greater connectivity and ease of trade among the trade partners in the SPICs.

108

8 Synthesis: The Integrated Multilayer Model

To estimate accessibility for the Existing Conditions network (S-1) and the Multiplier Attachment network (S-3), it was necessary first to calculate the transport cost using the best models from those calibrated above in Chap. 7. We have used an exponential decay function and constructed the index of accessibility, as discussed in Chap. 4. The physical infrastructure (shipping network) is considered, in the usual way, in the transport cost. Accessibility incorporates both the competition on supplied opportunities and the competition on demand, as the Vertical Network model is calibrated using a maximising entropy approach, (double) constrained on the supply and demand. Furthermore, it is important to mention that our analysis is SPICs-centred. Our trade data set regards bilateral flows with and within SPICs. This means that accessibility of non-SPICs is estimated based on trade relationships between the selected countries and the SPICs, but cannot be considered in absolute terms. In Table 8.6 we report the results of our analysis for the Existing Conditions and Multiplier Attachment networks. We can notice that accessibility at country level improves for SPICs in the Multiplier Attachment network architecture. Papua New Guinea is emblematic of the general improvement in the area, as it jumps to first position in the table. Not surprisingly, in the Existing Conditions network

Table 8.6 Ranking of accessibility at national level (countries in the SPICs reported in bold)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

Accessibility: Existing Conditions network New Zealand United States Papua New Guinea Fiji American Samoa Solomon Islands Australia Vanuatu Japan Philippines Palau Tonga China Korea New Caledonia Tuvalu Kiribati Indonesia Thailand Malaysia India Singapore

Normalised value 1 0.801 0.564 0.349 0.315 0.301 0.295 0.289 0.284 0.272 0.263 0.216 0.188 0.180 0.145 0.128 0.096 0.092 0.087 0.076 0.028 0.0038

Accessibility: Multiplier Attachment network Papua New Guinea Fiji American Samoa United States Vanuatu Solomon Islands Japan Palau Australia Tonga New Zealand China New Caledonia Korea Kiribati Tuvalu Philippines Thailand Malaysia Indonesia India Singapore

Normalised value 1 0.471 0.453 0.442 0.420 0.417 0.373 0.355 0.346 0.310 0.288 0.284 0.221 0.202 0.170 0.135 0.124 0.123 0.109 0.108 0.024 0.011

8.4 Interactions Between the SPICs

109

Table 8.7 Ranking of accessibility at port level (ports in SPICs are reported in bold)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

Accessibility: Existing Conditions network SUV LAX PPG PMB JKT KOR LAE HKG HIR NOR NGO TBU VLI PKL FUN BKK BNE NOU TRG TRW BNU LTA MNL GBS ORO MAG KIM RAB SIN CCU APW

Normalised value 1 0.919 0.739 0.653 0.601 0.549 0.528 0.504 0.420 0.414 0.405 0.386 0.345 0.342 0.336 0.321 0.314 0.310 0.307 0.307 0.283 0.247 0.246 0.164 0.128 0.089 0.077 0.070 0.033 0.025 0.0018

Accessibility: Multiplier Attachment network SUV LAE PPG PMB LAX NGO VLI PKL HKG TBU HIR KOR NOR NOU BNE KIM TRG ORO TRW BNU JKT FUN RAB LTA MNL BKK MAG GBS SIN CCU APW

Normalised value 1 0.952 0.811 0.742 0.662 0.606 0.606 0.597 0.590 0.587 0.523 0.501 0.482 0.460 0.429 0.418 0.376 0.370 0.346 0.338 0.330 0.312 0.289 0.286 0.282 0.242 0.210 0.074 0.060 0.041 0.0014

New Zealand has better accessibility than the United States, but this is probably due to the larger number of liner services conducted between New Zealand and the South Pacific Islands. In general, however, we can conclude that the Multiplier Attachment network increases overall the accessibility of the SPICs region. Given its flows, PNG has a significant level of accessibility, but so do the other major countries in the region. The coordination of trade activity through the definition of a hub-and-spoke shipping network centred in Fiji does not decrease, but rather it increases the potential gains for PNG. It is also important to mention that noteworthy increases in total regional accessibility are gained by every country in the

110

8 Synthesis: The Integrated Multilayer Model

SPICs, thus setting the stage for balanced and sustainable growth potential for the region as a whole. To next examine the impacts of accessibility at port level, we construct a new Vertical Network model. In this new configuration, ports (not countries) are nodes and flows between ports have been estimated proportionally to a fraction of the total trade between origin and destination countries and the throughput of each port. In this way we are able to calibrate the Vertical Network model such that trade interactions are estimated at the port level. We have calculated an exponential decay function and included Trade Agreements, Cultural Links and Common Language layers. The two network scenarios (Existing Conditions and Multiplier Attachment) reach convergence in the calibration process with an exponent of the decay function of 0.00028 for Existing Conditions and 0.00042 for Multiplier Attachment. The level of correlation between the data and model are still very high in both cases (R2 ¼ 0.75 for the Existing Conditions network; R2 ¼ 0.95 for the Multiplier Attachment network) (See Table 8.7).

8.5

Conclusions and Policy Recommendations

Figure 8.4 illustrates the results of our analysis. In order to appreciate the difference between the values of accessibility for the Multiplier Attachment and Existing Conditions networks, we have plotted the representation of the Multiplier Attachment (purple) on top of the Existing Conditions (in blue). Ports with concentric colours represent a decrease of relative accessibility between Existing Conditions and Multiplier Attachment networks. Ports in PNG benefit from the multilayer configuration and always increase their accessibility. Suva is the most accessible port in the area in both configurations: this result confirms that Suva is the best solution for the creation of a regional hub. In general, however, ports in SPICs improve their accessibility (only Funafuti and Santo have decreased their accessibility). We can conclude that the analysis of the Multilayer model confirms our results obtained in the previous chapters. The more balanced solution of the hub-and-spoke shipping network, emerging from the Multiplier Attachment network, is able to develop as a nexus for trade in the port of Suva in Fiji for the entire region of the SPICs. Furthermore, the choice of Suva as a hub for trade neither compromises the growth of PNG nor detracts from the other countries in the SPICs. This is a significant result in support of development and increased trade in the region because it introduces a new solution which is not based on vested interest and is more likely to garner political support throughout the SPICs. The Multilayer model has also drawn attention to the importance of such variables as cultural links and language to successfully demonstrate trade growth. The interdependence among these variables and trade, as we have noted, are complex and in particular the effects are not unidirectional. We have shown through the implementation of Multiplier Attachment shipping network and trade liberalisation

Reference

111

Fig. 8.4 Geo-visualisation of port accessibility for the Existing Conditions network (blue) and Multiplier Attachment network (purple)

that the increase in trade may foster sustained growth rates in the SPICs. Our argument in this case focusses on the central concept of the Multiplier Attachment factor, that is, increasing access to larger markets for the overall SPICs region will lead to larger returns on investments and consequently increases in economic growth.

Reference Wilson AG (1970) Entropy in urban and regional modelling. Pion, London

Appendix to Part II: Data Requirements for the SPICS

II.1 Data Sets for South Pacific Island Countries—SPICs In this section we describe the data sets collected for the case study of the South Pacific Island Countries. We also include statistical analyses to describe the status quo of these countries. To begin, the collected variables are summarised in Table II.1 and fully described in the next sections. One of the limitations of the study, however, has been the lack of robust and comprehensive data sets for the

Fig. II.1 Visualisation of the geographic location of the container ports considered in the study © Springer International Publishing AG 2017 F.R. Medda et al., Collaborative Approach to Trade, Advances in Spatial Science, DOI 10.1007/978-3-319-47039-9

113

Max Draft

Max vessel size

Container Shipping Services (CSS)

Max_draft

Vessel_size

Rij,k

Liner Shipping Connectivity Index

Logistics Performance Index: Overall

LSCI

LPI

Ease_business Ease of Doing Business

Efficiency of customs: Logistics Performance Index (LPI) Internet_users Internet Users

Throughput

sw (u)

LPI_EC

Variable Ports (35)

Notation n ¼ 1, 2, . . ., N

2011

Various

Year –

2012

Non2010 dimensional

Non2010 dimensional

Ranking

Percentage

Population with access to the worldwide network Measures if a country’s regulatory environment is conducive to business Captures countries’ connectivity to global shipping networks Reflects perceptions of a country’s logistics operations

2010

Non2010 dimensional

mLOA 2011 (or dwt) NonVarious dimensional

Mt

TEU (or TON)

Unit Latitude, Longitude

Efficiency of customs clearance process

Size of the largest vessel that has visited each port Container services that operate in SPICs ports

Max Draft

Information Container ports in SPICS and relevant Pacific RIM countries Volumes handled per year

Table II.1 List of model variables and collected data sets

62%

91%

77%

100%

62%

54 shipping services

80%

100%

69%

Completeness (% or no.) 100%

World Bank

UNCTAD

World Bank www.internetworldstats.com World Bank

Source www.sea-rates.com Containerisation International – Port Authorities website – Secretariat of the Pacific Community (2011) Secretariat of the Pacific Community (2011) Secretariat of the Pacific Community (2011) – Asian Development Bank Report (2004) – shipping lines World Bank

Model component Horizontal

114 Appendix to Part II: Data Requirements for the SPICS

Supply/demand

GDP

Economic Agreement

M Cultural ties

Common Language

Transport cost Handling cost Port tariff Travel time

Fuel cost

Vessel Fuel Efficiency

Ei,k; Ij,k

GDP

TA

Migration Cult

Lang

TCuv , k huv , k τuv b d uv

e cp

ε

IFO 380 prices for the Asian-Pacific market in Nov 2013 Vessel fuel consumption according to vessel size and travel speed

Travel time between ports

Handling cost per unit

Common language

Free trade agreements between SPICs and between SPICs and major Pacific Rim countries Major demographic figures Cultural links

Import/export at country level by product GDP Until 2012

tons/day

US$/ton

2009

2013

Various Until 2012 Non2005 dimensional Non2005 dimensional US$/TEU US$/TEU US$/TEU Days

– World Bank – IMF WTO

UN Trade Dataset

100%

100%

100% 100% 76% 100%

100%

T. Notteboom and P. Cariou (2009)

Own elaboration on GIS map (distance) and vessel speed T. Notteboom and P. Cariou (2009)

Korea Maritime Institute

CEPII

4 Agreements (SPARTECA, PICTA, PACER, APEC) 100% World Bank 100% CEPII

100%

2010–2012 100%

NonVarious dimensional

US$

US$

Transport cost

Vertical

II.1 Data Sets for South Pacific Island Countries—SPICs 115

116

Appendix to Part II: Data Requirements for the SPICS

region under consideration. For this reason, the data have been drawn from different sources and merged together in order to produce a satisfactory and robust data set. Below in Fig. II.1 is the geo-referred visualisation of the 35 ports under examination.

II.2 The Shipping Network Maritime shipping is the dominant mode of transport for international trade for the SPICs and is the main focus of the model. We reconstruct the container trade shipping network within SPICs and between SPICs and major Pacific RIM countries. The remainder of this section is dedicated to the description of the shipping network we have reconstructed from the available sources.

Port Selection We have collected information on 35 container ports in 25 countries in the region: 12 belong to major Pacific RIM countries and 23 to the SPICs. The collection has been performed such that we have selected any container port in the SPICs which is served by at least one international liner service. For modelling purposes, we have chosen one port for each Pacific RIM country that has trade relationships with SPICs. In Table II.2 we report the list of container ports included in our case study. As reported in Table II.3 below for each port (or at the national level for economic variables), we introduce the following information: Throughput, Max Draft, Logistics Performance Index (Efficiency of Customs) (LPI), Internet Users, Ease of Doing Business, Liner Shipping Connectivity Index (LSCI), Logistics Performance Index (LPI) (Overall), and Port Fees. Most of the selected variables in Table II.3 have a high variance since our data set comprises both international hubs (e.g., Hong Kong, Singapore, Busan, etc.) and under-developed ports in the SPICs. As expected, ports in the SPICs show low traffic volumes and poor port characteristics (Table II.3). In Fig. II.2 we visualise throughput and Fig. II.3 shows the maximum vessel draft allowed for ports in our case study. We use vessel maximum draft as a proxy for ports’ technical characteristics. It is noteworthy that, although ports in the SPICs are on average poorly equipped, some ports (e.g., in PNG, Fiji and Majuro Atoll) can accommodate fairly large vessels. In Figs. II.4, II.5 and II.6 we visualise GDP, Ease of Doing Business, and percentage of Internet Users, respectively, in our case study countries. As expected, SPIC countries all have very low GDP, and new business activity has relatively high difficulty in getting started. Although the SPICs are better positioned than some of the strong emergent economies (e.g., India, China and Indonesia), a less rosy picture appears when we consider the percentage of people with Internet access. Apart

II.2

The Shipping Network

117

Table II.2 Container ports and countries examined in our case study Country American Samoa (S) Australia (R) China (R) Fiji (S) Guam (S) Indonesia (R) India (R) Japan (R) Kiribati (S) South Korea (R) Majuro Atoll (S) New Caledonia (S) Malaysia (R) New Zealand (R) Philippines (R) Palau (S) Papua New Guinea (S) French Polynesia (S) Singapore (R) Solomon Islands (S) Thailand (R) Tonga (S) Tuvalu (S) USA (R) Vanuatu (S)

List of ports (3-digit code) Apia (APW), Pago Pago (PPG) Brisbane (BNE) Hong Kong (HKG) Lautoka (LTA), Suva (SUV) Port of Guam (GUM) Jakarta (JKT) Calcutta (CCU) Nagoya (NGO) Tarawa Betio (TRW) Busan (BNU) Majuro Atoll (MAJ) Noumea (NOU) Port Kelang (PKL) Tauranga (TRG) Manila (MNL) Koror (KOR) Kimbe (KIM), Lae (LAE), Madang (MAG), Oro Bay (ORO), Port Moresby (PMB), Rabaul (RAB) Papeete (PPT) Singapore (SIN) Honiara (HIR), Noro (NOR) Bangkok (BKK) Nukualofa (TBU) Funafuti (FUN) Los Angeles (LAX) Port Vila (VLI), Santo (GBS)

(S)—SPICs and (R)—Pacific RIM countries

Table II.3 Relevant statistics of port variables Throughput Max Draft Efficiency customs (LPI) Internet users Ease of Doing Business LSCI Total LPI Port fees

N 29 35 22

Min 2000 6.4 1.95

Max 2.8E+7 22 4.02

Mean 3,446,542 11.68 2.7

Std. dev. 6,997,488 3.56 0.7185

Variance 4.9E+13 12.64 0.5162

Skew. 2.59 1.25 0.46

Kurtos. 6.41 2.24 1.42

35 27

2 1

83.7 133

29.01 72.11

28.57 45.08

816.19 2032.54

0.86 0.47

0.82 1.39

31 22 35

2.9 1.91 489.9

143.6 4.22 10,421

27.0 2.8 4056

35.6527 0.85 2412

1271.2 0.73 5,819,234

1.82 0.39 0.74

2.61 1.49 0.16

118

Appendix to Part II: Data Requirements for the SPICS

Fig. II.2 Geo-referred visualisation of Port Throughput

Fig. II.3 Geo-referred visualisation of Maximum allowed vessel Draft

II.2

The Shipping Network

119

Fig. II.4 Geo-referred visualisation of national GDP in 2010

Fig. II.5 Geo-referred visualisation of national Ease of Doing Business. Small values are associated with countries where it is easier to start up a business

120

Appendix to Part II: Data Requirements for the SPICS

Fig. II.6 Geo-referred visualisation of national percentages of Internet Users

from the more dynamic examples of PNG and the Solomon Islands, most of the SPICs have low to medium values of Internet usage.

Container Shipping Services We construct the shipping network in the study region by collecting information from 14 shipping companies. We have surveyed a total of 54 container services in the region and included every port in the SPICs that is served by at least one shipping service; in other words, we have collected and analysed the data for 23 ports in the SPICs (shown below in Table II.4). Moreover, we have considered one port for each country outside the SPICs that has trade relationships with the SPICs (12 ports). The reasons for selecting only one port for each neighbour country are: (i) we suppose that the transport cost from a port in the SPICs to any port in a selected neighbour country does not fluctuate substantially (for example, we assume transport cost between Suva and Brisbane or Suva and Melbourne to be similar); and (ii) we do not have detailed information on supply and demand locations within each country. We do, however, have information on trade flows (Imports and Exports) between countries, but no data on ports of origin and ports of destination of these trade flows. In Fig. II.7 we visualise 22 out of 54 container

II.2

The Shipping Network

Table II.4 List of 14 container shipping companies in operation in the SPICs and Pacific Rim countries included in our study

121

Shipping company ANL CMA CGM CONTI7 GRETER BALI HAI HAMBURG SUD MARIANNA SHIP. LTD MATSON OOCL PACIFIC DIRECT LINE PACIFIC FORUM LINE PIL SEA TRADE SWIRE ZIM Total

No. of container services included in our data set 10 1 1 1 1 2 5 1 1 7 6 1 15 2 54

Fig. II.7 Geo-referred visualisation of a sample of 22 container services

122

Appendix to Part II: Data Requirements for the SPICS

services operating in the SPICs: different colours represent different container services. We report our analysis of the complete list of the 14 examined shipping companies and the number of services that each company operates in SPICs below in Table II.4. In the vertical model component we examine the complete set of interactions between layers of the Multilayer model. Shown in Fig. II.8 are the envisioned layers: Trade, Cultural Links (Cult) and Common Language (Lang), Trade Agreements (TA), and Container Shipping Network (CSN). The Container Shipping Network used in the study has been described in the previous section; we focus on the remaining variables next.

Fig. II.8 A simplified visualisation of the Vertical Interaction model. Depicted from top to bottom: bilateral trade flows, trade agreements, cultural links, common language, and container shipping network

II.2

The Shipping Network

123

Trade In order to reconstruct the trade flows between countries in our case study we have utilised the UN COMTRADE data set. Data on the bilateral trade of SPICs with its major trading neighbours in the period 2010–2012 have been collected from www. trademap.org and covers around 95% of bilateral trade between SPICs and trade partners in the region. Some SPICs also have trade relationships with other countries outside the Pacific RIM and neighbouring areas (in Southern Africa, South America and Europe). We have excluded these countries from our model due to scant information at our disposal on the international container shipping network. Trade volumes are given in values (US$) divided into 100 commodity categories, according to the 2-digit Harmonised System (HS2). In order to ease the analysis and interpretation of trade patterns, we aggregate the 100 HS2 categories into 15 categories HS1, shown in Table II.5. Immediately after, in Fig. II.9, we Table II.5 HS1 and HS2 commodity classification HS1 Animals and animal products Vegetable products Foodstuffs Mineral products Chemical and allied industries Plastics/rubber Raw hide, leather, etc. Wood and wood products

HS2 codes 01–05 06–15 16–24 25–27 28–38 39–40 41–43 44–49

HS1 Textiles Footwear Stone and glass Metals Machinery/electrical Transportation Miscellaneous

Fig. II.9 Exports (left) and imports (right) in the SPICs by HS1 classification

HS2 codes 50–63 64–67 68–71 72–83 84–85 86–89 90–99

124

Appendix to Part II: Data Requirements for the SPICS

visualise total Exports and Imports by HS1 commodity classification. In general, the South Pacific Island Countries (SPICs) export raw materials and import manufactured goods and processed products (machinery, chemicals, plastics, transportation, food, etc.). Next in Fig. II.10 we represent the network generated by trade flows. Papua New Guinea (PNG) is by far the country with the highest total trade. Within countries outside SPICs, Australia is the leading trading partner. When we examine the graph in Fig. II.11 of total Imports and Exports in SPICs, the imbalance between trade flows is striking. Imports are shown to be twice as much as SPICs Exports, with the exception of the Solomon Islands.

Fig. II.10 Trade flows between countries in the period 2010–2012. SPICs countries are visualised in red. Linked colours identify country of origin, and link width is proportional to trade value

II.3

Cultural Links and Common Language

125

Fig. II.11 Imports and exports for SPICs between 2010 and 2012

II.3 Cultural Links and Common Language We use the CEPII data set1 in order to account for cultural and language similarities between countries. Language similarities are considered through a dummy variable which is assigned to each couple of countries in our data set: value 1 represents the existence of a cultural link when one or more languages are spoken by at least 9% of the population in two countries. Cultural linkages are taken into consideration through the existence of colonial links between countries.

Trade Agreements South Pacific Island Countries participate in four trade agreements: SPARTECA, PICTA, PACER, and APEC. A brief description of each trade agreement is given in the remainder of this section. South Pacific Regional Trade and Economic Cooperation Agreement (SPARTECA) SPARTECA was signed in 1981 between Australia, New Zealand and 14 countries of the South Pacific Forum. It allows duty free access to the products of Forum 1

http://www.cepii.fr/anglaisgraph/workpap/pdf/2011/wp2011-25.pdf

126

Appendix to Part II: Data Requirements for the SPICS

Island Countries (FICs) to the markets of Australia and New Zealand, subject to ‘Rules of Origin’ regulations. The aim is to redress the unequal trade relationships between the two groups. The Pacific Island Countries Trade Agreement (PICTA) PICTA is a free trade agreement among the 14 Forum Island Countries (FICs). Under this agreement virtually all barriers (import tariffs and quotas) to merchandise trade between FIC countries is to be removed. The initial PICTA, which covers only Trade in Goods (TIG), was ratified by (10) FICs. The agreement was signed in 2001 and came into force in 2006. Fiji ratified the agreement in 2002. The Pacific Agreement on Closer Economic Relations (PACER) The PACER is a framework agreement setting out the basis for the future development of trade relations among all 14 Forum members. The PACER provides for free trade to be established gradually among Forum members, in line with the ‘stepping stone’ approach preferred by the FICs—i.e. as the name suggests, the Agreement sets a framework for free trade in the region to be established at the appropriate ‘pace’, reflecting the different development status of the members. Asia-Pacific Economic Cooperation (APEC) The APEC is a Forum for 21 Pacific Rim member economies seeking to promote free trade and economic cooperation throughout the Asia-Pacific region.

II.4 Transport Costs for SPICS Another important variable collected and estimated for the present study is transport cost. The transport cost for a route linking two ports is calculated by measuring the information on port fees and the voyage costs due to vessel fuel consumption.

Voyage Cost Voyage cost includes the fuel cost of a vessel during its journey from one port to another. To calculate voyage cost we include distance between ports, fuel price, and vessel technical specifications that determine its fuel consumption. The fuel price in our study is based on the price for the IFO 380 bunker in the Asian and South Pacific market (Singapore) in November 2012, which was reported at US$616 per tonne (PDVSA 2013). Total fuel consumption in a vessel depends, among several factors, on vessel size and average speed. In order to estimate a typical vessel size for our study, we have conducted a survey among shipping companies that publishes schedules and deployed vessels in their websites (Table II.6). Results show that a typical container/cargo ship has Gross Tonnage (GT) of around 12,200 tonnes and deadweight tonnage (DWT) of 15,600 tonnes.

II.4

Transport Costs for SPICS

127

Table II.6 Containerships deployed in the SPICS Companya Pacific Forum Line (& PDL & PIL)

Vessel namea CAPITAINE WALLIS

GTb 9422

DWTb 13,064

Pacific Forum Line (& PDL) Pacific Forum Line (& PIL) Pacific Forum Line (& PDL) Pacific Forum Line (& PDL) Swire Shipping

CAPITAINE TASMAN TIARE MOANA SOUTHERN LILY FORUM SAMOA III NINGPO

9422

12,814

3972

4152

9684

12,502

Nukuafola, Apia, Pago Pago

9422

13,058

16,801

22,900

Swire Shipping

SHANTUNG

25,483

30,814

Swire Shipping

NGANKIN

16,801

22,900

Swire Shipping

SOOCHOW

25,483

30,721

Swire Shipping

SHANSI

25,483

30,700

Swire Shipping

ISLAND CHIEF KWEILIN PAPUAN CHIEF

10,352

10,553

18,468 10,350

23,586 13,550

Swire Shipping

HIGHLAND CHIEF

10,357

10,775

Swire Shipping

SHAOSHING

25,483

30,814

Matson

LILOA

5234

5650

Matson

OLOMANA

5025

5642

Matson

IMUA

4246

6288

Suva, Apia, Nukuafola, Pago Pago, Noumea, Port Vila, Lautoka Noumea, Lautoka, Suva, Port Moresby, Wallys and Futuna, Lae, Rabaul, Madang, Santo, Honiaria, Port Vila, Noro Lautoka, Suva, Lae, Honiaria, Noro, Port Vila, Noumea, Rabaul, Kimbe, Oro Bay, Madang, Honiara, Honiara, Noumea, Lautoka, Suva, Wallis, Port Moresby, Lae Lae, Honiara, Noumea, Noro, Port Vila, Majuro, Tarawa Betio, Lautoka, Suva, Rabaul, Kimbe, Oro Bay, Madang Lautoka, Suva, Honiara, Noro, Lae, Rabaul, Oro Bay, Kimbe, Madang, Port Vila, Noumea, Majuro, Tarawa Betio Lae, Oro Bay, Kimbe, Rabaul, Noro, Honiara, Port Moresby Noro, Honiara, Lae, Kimbe Lae, Oro Bay, Rabaul, Madang, Santo, Suva, Lautoka, Apia, Pago Pago, Majuro, Tarawa Betio, Oro Bay, Lae, Rabaul, Madang, Santo, Lautoka, Suva, Majuro, Tarawa Betio, Apia, Pago Pago Lautoka, Suva, Lae, Honiara, Noumea, Port Vila, Majuro, Tarawa Betio, Rabaul, Kimbe, Oro Bay, Madang Lautoka, Suva, Apia, Pago Pago, Nukualofa Lautoka, Suva, Apia, Pago Pago, Nukualofa Honiara, Noro, Suva, Port Vila, Santo (continued)

Swire Shipping Swire Shipping

Visited SPICs portsa Suva, Lautoka, Tawara Betio, Wallis, Futuna, Funafuti, Tarawa Betio, Majuro Suva, Apia, Nukuafola,Noumea, Port Vila, Lautoka, Pago Pago Nukuafola

128

Appendix to Part II: Data Requirements for the SPICS

Table II.6 (continued) Companya Matson (& New Pacific Line) Matson (& New Pacific Line) Pacific Direct Line (& PFL) Pacific Direct Line Pacific Direct Line Pacific Direct Line (& PIL) Pacific Direct Line

Vessel namea CAPE NATI

GTb 18,326

DWTb 23,400

Visited SPICs portsa Honiara, Port Moresby, Lae, Rabaul

CAPE NASSAU

18,326

23,327

Honiara, Port Moresby, Lae, Rabaul

9422

13,064

9935

12,343

Port Vila, Santo, Suva, Noumea, Lautoka Noumea

9684

12,502

Noumea, Honiara

9725

12,985

Papeete

16,772

24,650

Papeete

5234

6030

9684

12,502

Suva, Wallys, Futuna, Tarawa Betio, Fiji, Funafuti Fiji, Lautoka, Apia

9420

13,058

Nukualofa, Apia

5994

9606

Fiji, Honiara

5316

7708

Honiara

10,350

13,550

12,248 9,978

15,652 13,029

SOUTHERN MOANA SOFRANA SURVILLE SOFRANA TOURVILLE SOUTHERN TRADER KOTA WARUNA SOUTHERN PEARL SOUTHERN FLEUR SOUTHERN LILY 3 DANNY ROSE SOUTHERN PHOENIX PAPUA CHIEF

Pacific Direct Line (& PIL) Pacific International Lines Pacific International Lines Pacific International Lines Pacific International Lines Pacific International Lines Average total Average (vessels with GT < 20,000 tonne)

Fiji, Tarawa Betio, Majuro

a

Source: shipping company websites: www.pacificforumline.com, www.swireshipping.com, www.matson.co.nz, www.pdl123.co.nz, and www.pilship.com b Source: www.marinetraffic.com

We have taken a conservative approach to estimate the average vessel size and consider vessels with GT under 20,000 tonnes. In this case a typical vessel (average value) operating in the SPICs has GT of 10,000 tonnes and DWT of 13,000 tonnes. In the study we have selected as an archetypal vessel size that of Capitaine Wallis for two main reasons. First, if we consider the size of the average vessel operating in the SPICS, the value is lower than the size of Capitaine Wallis. However, since this is a strong hypothesis and we are also assuming in our model that vessels always travel fully loaded, we prefer to maintain a cautious attitude.

II.4

Transport Costs for SPICS

Table II.7 Main technical characteristics of sample vessel used in our study

129 Factor Vessel size (TEU) Gross tonnage DWT LOA Fuel efficiency

Value 700 9422 13,064 145.9 m 50 tonnes/day

Second, we have observed that Capitaine Wallis is largely used in SPICs by three shipping companies that operate various services in cooperation using this vessel. Table II.7 shows the technical specifications of the vessels assumed in our study. According to the study by Notteboom and Cariou (2009), vessels smaller than 1100 TEUs traveling at an average speed of 19 knots (35.2 km/h), consume approximately 50 tonnes of fuel per day of travel. Number of days of journey depends on the distance between two ports and determines the total fuel consumed. Voyage cost plays a significant role in the transport cost by being its principal component. Initial results obtained from simulations have revealed that the voyage cost per unit is between 51% and 907% higher than the port fee per unit. The long distance between ports, as well as economies of scale in the number of TEUs transported in a vessel, determine the difference between the two cost components of the transport cost. When examining the transport cost, the highest levels of this cost can be attributed to the links connecting ports in the SPICs with major ports outside the SPICs region (e.g., Singapore, Hong Kong, Manila, and Los Angeles) despite the fact that the latter ports tend to have relatively low port fees. In these cases the condition of the SPICs as a remote area was confirmed to have the significant impact of increasing the transport cost when linking to foreign markets, mainly due to the voyage cost. When analysing the transport cost within the SPICs region, we found that for the particular cases of Honiara and Pago Pago, links connecting these ports show a higher transport cost compared with other links among the SPICs as a consequence of the high port fees attributed to these ports. Conversely, lower transport costs were observed in the links connecting Oro Bay, Lautoka, and Funafuti ports due to their lower port fees, their convenient location in relative proximity to other ports in the region, and the lack of links connecting to ports outside the SPICs.

Port Fees For each port in our case study the associated port fee is the cost for each vessel entering the port at any period in the year and for a time span in the port not exceeding 3 days. The port fee includes entrance fees, dockage and anchorage charges, pilotage fee, and tug service charges. In other words, we consider all of the

130

Appendix to Part II: Data Requirements for the SPICS

port fees that are not related to container movements. Data on port fees was obtained from different Port Authority websites (P.A.W.), direct contact with Port Authorities by email or phone (P.A.D.), and from the Korean Maritime Industry (KMI). Table II.8 summarises the port fees for the ports in our case study. Table II.8 Estimated port fees per entry for a vessel with characteristics Port Apia Pago Pago Brisbane Hong Kong Lautoka Suva Port of Guam Jakarta Calcutta Nagoya Tarawa Betio Busan Majuro Atoll Noumea Port Kelang Tauranga

Entrance fee US$ 486.3 486.3 177 417 1020 1020 312.6 – 1022 256.46 530 593 – – – 10,421.04

Manila Koror Kimbe Lae Madang Oro Bay Port Moresby Rabaul Papeete Singapore Honiara Noro Bangkok Nukualofa Funafuti Los Angeles Port Vila Santo

787 – 103 207 103 103 207 103 – 712.5 870 870 – 516 – na – –

a

Dockage and anchorage 2530 2530 1138 1496 188.7 188.7 177.3 – 2677.2 1116 290 292 – – – (incl. in entr. fee) 379 – 2976 2976 2976 2976 2976 2976 – 502.5 381.2 381.2 – 0 – 1322 – –

Pilotage Tug service 1264.3 (incl. in Pilot.) 1264.3 (incl. in Pilot.) 2071 (incl. in Pilot.) 645 Na 1443 (incl. in Pilot.) 1443 (incl. in Pilot.) na Na – – 1610 (incl. in Pilot.) na Na 400 Na 204 1401 – – – – – – (incl. in (incl. in entr. fee) entr. fee) na na – – 1023 2600 1023 2600 1023 2600 1023 2600 1023 2600 1023 2600 – – 360 420 1450 na 1450 na – – 600 2400 – – 2641 (incl. in Pilot.) – – – –

Total 4280.6 4280.6 3386 2365 2651.7 2651.7 489.9 2000a 5309.2 1372.5 1220 2490 1300a 6800a 2500a 10,421.04 1166 6800a 6702 6806 6702 6702 6806 6702 2000a 1995 2681.2 2681.2 2000a 3516 1300a 3963 1300a 1300a

Reference KMI KMI P.A.D. P.A.W. KMI KMI P.A.W. P.A.W. P.A.W. KMI P.A.D.

P.A.D. P.A.W. KMI KMI KMI KMI KMI KMI P.A.W. KMI KMI P.A.W. P.A.W.

Due to a lack of available information, the fee for these ports was approximated based on the figures for ports with similar characteristics

II.4

Transport Costs for SPICS

131

In the remainder of this section we report the calculations we have implemented to estimate each charge included in the port fee. Values are reported by countries since in our set of ports the few countries with more than one port apply the same tariffs for all of their ports. Samoa Entrance fee: US$0.05 per GT Dockage: US$0.07 per GT per day Anchorage: US$78 per hour Pilotage and Tug Service: US$0.13 per GT

Australia The Port of Brisbane charges fees by moved container (for entrance and dockage/ anchorage rates). The website of the port of Brisbane provides information of containers moved each quarter aggregated by country of arrival/destination. We have estimated the average number of containers moved in the port of Brisbane from/to SPICs as follows: we consider New Caledonia as representative country of SPICs where port of Noumea handles the largest share of container traffic of New Caledonia. Thus, the national figures of New Caledonia well approximate the traffic between ports of Brisbane and Noumea. We calculate the average number of containers moved in each quarter and divide it by an estimated frequency of service of 15 days. We have found that on average in the port of Brisbane 40 containers are moved during each visit of a vessel from/to SPICs and we assume that this is the value charged due to entrance and dockage/anchorage fees. Entrance fee: US$4.4 per container Dockage: US$28.5 per container Anchorage: included in Dockage Pilotage: US$2071 per service Tug Service: included in pilotage

China Entrance fee: US$43 per 100 tonnes or part thereof on each occasion of entry Dockage: US$60 for each 8 m in length of a vessel to be berthed alongside the seawall Anchorage: US$0.02 per tonne Pilotage: US$0.065 per tonne of the ship’s gross registered tonnage Tug Service: n.a.

Fiji Entrance fee: US$2.2 per 100 GRT plus US$806.25 for each entry Dockage: included in anchorage

132

Appendix to Part II: Data Requirements for the SPICS

Anchorage: US$4.826 for each 100 GRT Pilotage: US$0.1075 per GRT plus fixed charge of US$397.75 Tug Service: included in pilotage

Guam Entrance fee: US$54.81 plus US$40.84 for each additional 2000 gross tonnes Dockage: US$59.11 per day Anchorage: US$253.44 per day Pilotage: n.a. Tug Service: n.a.

India Entrance fee: US$0.105 per GRT Dockage: included in anchorage Anchorage: US$0.138 per GRT per hour Pilotage: fixed rate Tug Service: included in pilotage

Japan Entrance fee: US$0.03 per GRT Dockage: US$0.10 per GRT Anchorage: included in dockage Pilotage: no rates found for pilotage Tug Service: no rates found for tug service

Kiribati Entrance fee: US$0.30 per metre length Dockage: US$2 per metre length Anchorage: included in dockage Pilotage: US$0.05 per GRT with a max of US$400 Tug Service: no rates found for tug service

New Zealand Entrance fee: Port of Tauranga apply a fixed fee which includes all port tariffs Dockage: included in entrance fee Anchorage: included in entrance fee

II.4

Transport Costs for SPICS

133

Pilotage: included in entrance fee Tug Service: included in entrance fee

Philippines Entrance fee: US$0.081 per GRT Dockage: US$0.020 per GRT Anchorage: included in dockage Pilotage: no rates found for pilotage Tug Service: no rates found for tug service

PNG Entrance fee: value per entrance Dockage: US$0.69 per LOA per hour Anchorage: US$65.61 per LOA per year Pilotage: US$0.3 per GT + US$86 (pilot fee) + US$251.9 (pilot boat charge) Tug Service: US$2600

Singapore The rates reported are the best approximation to be found in the Port Authority website. As Singapore is a privately owned port, PSA Corporation adopts a marketbased approach, in that shipping companies are offered special offers according to their long-term contracts. Entrance fee: US$7.5 per 100 GRT Dockage: US$0.05 per GRT Anchorage: included in dockage Pilotage: vessel above 6000 GT and up to 12,000 GT Tug Service: vessel above 5000 GT and up to 10,000 GT

Solomon Islands Entrance fee: US$6 per metre of length Dockage: US$1.94 per 100 GT Anchorage: US$4.3 per metre of length Pilotage: US$10 per metre of length Tug Service: no rates found for tug service

134

Appendix to Part II: Data Requirements for the SPICS

South Korea Entrance fee: US$0.10 per GT Dockage: US$0.03 per GT Anchorage: US$0.02 per GT Pilotage: US$204 per call Tug Service: US$1401 per call

Tonga Entrance fee: US$0.05 per GT Dockage: n.a. Anchorage: n.a. Pilotage: launch hire, per hour running Tug Service: tug hire, per hour running

USA Entrance fee: n.a. Dockage: fixed rate (class: between 135 and 150 metres of length) Anchorage: included in dockage Pilotage: US$0.0034 per GT + US$591 Tug Service: included in pilotage

Reference Notteboom TE, Cariou P (2009) Fuel surcharge practices of container shipping lines: is it about cost recovery or revenue-making? Proceedings of the International Association of Maritime Economists (IAME) 2009 Conference, Copenhagen, June

Part III

Uganda

Chapter 9

The Agriculture Supply Chain in Uganda: The Design of an Agent Based Model

9.1

Introduction

Different studies demonstrate how market inland country delays often hinder trade more than the effect of foreign tariffs (Hummels 2001; Portugal-Perez and Wilson 2009; Djankow et al. 2010; Freund and Rocha 2010). This is especially evident when we consider Africa’s exports, and in particular, highly time-sensitive goods such as agriculture products. Within this context the study and implementation of trade facilitation solutions for Uganda, and in general East Africa, plays an important role. Uganda belongs to one of the largest clusters of 10 landlocked countries spanning Central and East Africa. From this perspective, the objective to increase and facilitate trade in landlocked African countries may also, like a ‘domino effect’, improve trade across the African continent. Uganda has experienced remarkable economic growth within the last decade, with a GDP of US$16.81 billion in 2011, compared to US$6.34 billion in 2003 (World Bank 2012). Nonetheless, with an ever growing population and a sustained annual growth of 3.2% (MDG report 2011), a large proportion of Uganda’s 34.5 million inhabitants live in extreme poverty. Estimates of the share of people living in households below the poverty line have been as high as 39% over the last decade (UNHS 2013). Ongoing government efforts to tackle the poverty phenomenon (Poverty Eradication Action Plan 2009) have proved to be relatively effective, but an uneven distribution of the merits of such approaches has also been shown (Millennium Development Goals Report 2011). Measured poverty in rural areas is generally two times higher than in urban areas, while regions in the north and east of the country have had a noteworthy lack of improvement in comparison to the rest of the country according to 2010 data (Millennium Development Goals Report 2011). The National Development Policy (NDP 2010) has renewed interest in, and emphasis on, agriculture as a driver for development, economic growth and poverty reduction through private investment and enhanced competitiveness. A new Development Strategy Investment Plan (DSIP 2011) was drawn up to complement the © Springer International Publishing AG 2017 F.R. Medda et al., Collaborative Approach to Trade, Advances in Spatial Science, DOI 10.1007/978-3-319-47039-9_9

137

138

9 The Agriculture Supply Chain in Uganda: The Design of an Agent Based Model

recommendations of the NDP, and among other objectives, DSIP is designed to aid the transformation of subsistence farming into commercial agriculture. Agriculture and agribusiness is thus seen as an opportunity to alleviate poverty and food security for the majority of Uganda’s people, while also capitalising on the untapped potential of the agriculture sector within the country (World Bank, Growing Africa: Unlocking the Potential of Agribusiness 2013). To pursue this prospect, among other factors, the agriculture sector will need to compensate for the lack of competitiveness caused by landlocked geography and an inadequate inland market network. We can describe an inland market network as the nexus of goods production, flows and points of consolidation and exchange, and thus see it as a spatial structure with components and processes that move goods and products from origin to destination. The main actors who operate in this market are farmers, traders, wholesalers, and exporters. According to Freund and Rocha (2010), in general in Africa a 1-day increase in inland network trade delay reduces export values by about 17%. On the Uganda inland market network we can highlight three principal reasons for delays. 1. Contracting, documentation, cross-border Customs. This is sometimes the longest delay and is a major barrier to private participation in trade. 2. Transit time. This can be represented as a generalised transport cost. 3. Logistics chains and consolidation catchment areas. These delays are especially restrictive if associated with financial obstructions: there may be high rent costs for inventory, storage and security. Enhancing trade facilitation is therefore a primary concern since an improvement in the inland market network is likely to boost exports and yield broad economic effects in terms of regional and national competitiveness, reduction of poverty, and increases in production and supply chain performance. By addressing inland market network inadequacies we will be able to leverage the agriculture sector’s likelihood of attracting investment, and in so doing, stimulate economic growth.

9.2

The Conceptual Framework

Against the background of the Uganda agriculture sector described above, we aim to develop an agent based (AB) model (Gollin and Rogerson 2010; Happe et al. 2006; Balmann 1995, 1997) to simulate the supply chain of agriculture produce from farm gate to regional and export markets. In our context we develop a system to model the flow of goods from farmers to markets, and the flow of information from markets to farmers. The AB system is decentralised with no designated command and control, and its efficiency depends on the coordination between the actors/agents in the system and the robustness of the network. Agent based modelling is an appropriate technique for modelling the challenges posed, as it is able to identify the actors, activities, interdependencies, and objectives in the agriculture supply chain, and to better represent individual entities and their interactions. Moreover, AB modelling can capture

9.3 Hypotheses

139

individual decision making, negotiation between supply chain partners, and incorporate GIS data for geographical placement of individual entities. Lastly, agent based modelling is also useful in the representation of emergent societal-level behaviour, starting with individual actions and leading to results of individuals sharing information, adapting, and learning. In the agriculture supply chain we model farmers/ outgrowers, traders (these include farmers, large exporting firms and itinerant traders), markets, logistics service providers, financial institutions, and governments. The vast majority of farmers are small farmers or outgrowers. Outgrowers are farmers who sell their produce to large farmers or firms. Farmers/outgrowers can choose from different options when they deliver their produce to market. Farmers/ outgrowers may, for instance, either sell their produce directly to the domestic market, or sell to a trader higher up the supply chain. Traders in the next level of the chain may be itinerant (they travel from farm-gate to farm-gate making informal agreements with farmer/outgrowers); they may also be large farmers who seek to supplement their own production by using outgrowers to meet market demand; and lastly, they may be large producing and exporting firms that employ outgrowers. Another available option to farmers is to coordinate by forming farmer associations that can benefit from economies of scale. The farmer’s decision is made on the basis of logistics and processing costs, the expected market price (confirmed from farmers’ social network), and the risk-aversion utility function, which is central to the farmer’s wealth. Traders can either sell to other traders or to domestic, regional or export markets. Logistics service providers offer transport services to farmers and traders; processing and packing of agriculture produce is handled by the processors; and governments introduce policies which impact on the activities and behaviour of the other agents in the supply chain. And lastly, financial institutions provide credit.

9.3

Hypotheses

In this study we have selected seven districts on the basis of their proximity to Kampala and Entebbe and because large farmers and exporting firms are situated in these areas. Our selected districts are Luwero, Mpigi, Masaka, Iganga, Mitiyana, Kamuli, and Mukono. In regard to the products moving along the supply chain, we consider the movement of five crops: hot peppers, matooke, okra, chillies, and sweet potatoes. Through the application of agent based modelling we aim to explore the following hypotheses. H1 Which transport improvement in Uganda can more effectively enhance the growth of trade and support small farmers/outgrowers to move out of the subsistence status of production? (Chap. 10). H2 In order to provide incentives for the consolidation of trade flow, where should warehousing activities be located and how (concentrated in few locations and dispersed in the region)? Which other policy, in addition to the consolidation of the trade flow, can facilitate the logistics chain? (Chap. 11).

140

9 The Agriculture Supply Chain in Uganda: The Design of an Agent Based Model

H3 When coordination takes place among farmers/outgrowers, and agriculture standards (GAP) are introduced, do such actions generate effective benefits to farmers/outgrowers? (Chap. 12). The three hypotheses will be tested through different scenarios in order to verify possible policy strategies and long-term outcomes (Table 9.1).

9.4

Agents and Agent Interactions

Agents and their interactions represent the core of our model. In the sections below we thoroughly describe the features of the different agents. I. Farmers A farmer agent has a certain amount of land and wealth that is assigned by using probability distributions. The probability distribution is parameterised using data on farmers in the districts. We assume that all farmers adopt multi-cropping. Land is randomly assigned to cultivate the five crops that we use for our study, namely: hot peppers, chillies, matooke, okra and sweet potatoes. We assume a certain rate per acre of expenses for running a farm, and predict the yield of the farm using data on average yields for each type of crop. The assets of a farmer agent are thus his bank balance, the value of his land, and the value of his harvested crops. A farmer consumes a certain fixed amount of his produce and sells the remaining share. In our agent based model we also build a social network of farmers who know each other. The connections between farmers develop through the Multiplier Attachment method. Farmer A attaches to farmer B based on his distance from farmer B and the degree of connectivity/communication of farmer B. Information from B diffuses to farmer A at a rate that is proportional to his degree of connectivity. Therefore, a farmer spreads information at a faster rate if he is highly connected, and farmers who connect to him benefit from this rapid dissemination of information. We assume that the social network and social media network provide a twofold purpose: (1) to facilitate the spread of price information and (2) to provide a platform for two trading parties to be introduced via a common friend in the network. While a farmer agent is busy in the cultivation of his crop, he is also engaged in other marketing activities, namely forming agreements. We identify two kinds of farmers: a few large farmers, and commonly occurring small farmers (owning, on average, 1 acre of land). The activities of large farmers are: producing crop(s), soliciting agreements with small farmers (in their social network) and buying product at a fixed price, and to sell (through the use of a decision tree). The activities of small farmers are to produce a crop(s), to receive proposals for agreements from large farmers and traders, and to sell (using a decision tree). In general, after the production cycle is complete, a farmer agent makes a decision on the sale of his product using a decision tree process. He uses his social network and social media network to form an opinion of the expected price of his

When coordination takes place among farmers/outgrowers and agriculture standards (GAP) are introduced, do such actions generate

Small scale farmers/outgrowers can coordinate their activities in associations or through agriculture firms to improve bargaining power and

Logistics chain is often a weak element in the trade and export flows of Uganda due to lack of—and low quality of—storage facilities, and the cartelisation of the logistic services.

Hypotheses Which transport improvement in Uganda can more effectively enhance the growth of trade and support small farmers/outgrowers to move out of the subsistence status of production? Travel cost (comprised of transport costs and logistics/storage handling costs) can decrease by (1) improving the quality of road networks (from gravel to paved, 100% RAI); (2) improving the quality of other infrastructures that are fundamental to the success of export activity. (H1) To provide incentives for the consolidation of trade flow, where should warehousing activities be located and how (concentrated in few locations or dispersed in the region)? Which other policy, in addition to the consolidation of the trade flow, can facilitate the logistics chain? (H2)

Background context The quality and performance of the transport infrastructures are not consistent in the considered Districts. This produces high transport costs in the supply chain.

(a) Determine at the District level the impact of organised farmer associations on farm gate prices (according to the main produced crops).

(a) Identify at the District level the type and location of other transport infrastructure storage areas in order to lower transaction costs, reduce loss of post-harvest production, and increase consolidation/economies of scale.

Output (a) Identify at District level for each District type and location of road infrastructure intervention (possible interventions: paved or non-paved road and new road construction) in order to improve accessibility and further trade.

(a) Improvement of existing storage areas such as transforming part of storage areas into refrigerated warehouses to keep multiple crops. (b) Construction of new storage facility. (c) Introduce cap pricing mechanisms and regulatory regimes to increase competition of the logistics services. (a) Support for creation of coordinated farmer associations in relation to crops and District. (b) Develop guidelines for adoption (continued)

Concrete policy options (a) Upgrade of existing infrastructure such as improvement of roads from dirt road to paved/gravel. (b) Construction of new or expansion of transport infrastructure.

Table 9.1 Summary of background context of trade issues in Uganda, hypotheses, and output and policy interventions

9.4 Agents and Agent Interactions 141

Hypotheses

effective benefits to farmers/ outgrowers? (H3)

Background context

achieve economies of scale. Moreover, agriculture policies such as GAP improve standardisation of quality, traceability, contract enforcement, and finance for trade and agribusiness. However, the implementation of these procedures is still patchy in the agriculture sector in Uganda, especially for non-traditional agriculture products.

Table 9.1 (continued)

(b) At District level and for selected crops, measure the impact of standardisation of and quality control of products, and traceability on prices and transaction costs.

Output

of global GAP standards. Implement affordable certification schemes. (c) Campaign to raise awareness of commercial benefits such as best practice and know-how of participating in associations and implementation of GAP.

Concrete policy options

142 9 The Agriculture Supply Chain in Uganda: The Design of an Agent Based Model

9.4 Agents and Agent Interactions

143

product. He decides on the basis of the price and quality information of other farmer agents in his social network. The farmer then uses this expected price in the decision tree process, which involves evaluating multiple options based on logistics and processing costs, expected market price (developed through his social network), and the risk-aversion utility function (a function of the wealth of the farmer). II. Traders We assume that a trader agent has a certain amount of wealth which is assigned by using a flat probability distribution. Traders buy their goods from farmers and other traders. They are randomly assigned a minimum volume criterion; the implication is that they are able to buy only goods which satisfy the volume and standard criteria and traceability (GAP). Trader agents are engaged in three kinds of activities: they negotiate agreements (similar to large farmers), they buy and sell (using the decision tree). Traders set their prices relative to supply and demand. We assume that they can trade volumes that are less than 10% in value of their total budget; if they cannot afford them, they will buy them and pay as soon they have the budget capability. Trader agents also form social network connections based on distance and degree of connectivity. However, we do not impose the multiplier attachment principle on the trader network because traders compete with one another and are consequently reluctant to share information. We do, however, establish vertical links between the farmer network and the trader network by allowing farmer agents to connect to trader agents using the multiplier attachment principle. Connections are similarly formed on the basis of distance and degree of connectivity, i.e. information diffuses from traders to farmers in relation to degree of connectivity. A more highly connected trader spreads information at a faster rate than a trader with fewer connections. III. Exporters As in the case of farmers and traders, we assign exporters a certain amount of wealth using a probability distribution function. Also in this case exporter agent social networks are based on distance and degree of connectivity/communication. The farmer social network and trader social network connect to the exporter network using the principle of multiplier attachment. We assume exporter agents have a fixed demand, and hence do not have to engage in selling. They are engaged in only two kinds of activities, negotiating agreements and buying. Exporters buy products in bulk and have a minimum volume criterion. They also have a minimum quality criterion. The prices of the exporters are set based on supply in relation to the fixed demand. IV. Logistics Service Providers The logistics service provider (LSP) agent moves products between two locations. The transport costs incurred are assumed to be proportional to distance only. We assume that the LSP has two types of service, one with refrigerators (at a higher price and with higher reliability, thus implying a lower probability of failure) and one without refrigerators. We also assume that the path taken by the LSP is always the shortest path between two points. The time taken to deliver the product depends on path taken, and is contingent on whether it is transported by dirt or gravel/paved

144

9 The Agriculture Supply Chain in Uganda: The Design of an Agent Based Model

road(s). The quality of the transported goods depends on weather conditions in the case of no refrigeration, which is the current temperature. The logistics service provider aims to optimize the route on the basis of costs. V. Markets Domestic markets are located within each village. We assume a fixed demand for each market and the supply to the market consists of the product being sold by all farmers. In the next section we elaborate our study through a discussion of four distinct channels in the aim to visualise the agent based system of farmer-trader interactions.

9.5

Farmer-Trader Interaction Through Four Channels

The Uganda agriculture production and logistics markets can be synthesised in terms of four main channels in order to succinctly represent interactions among the different agents (farmers, exporters, etc.) and their activities (consolidation and trade, etc.).

CHANNEL 1 Large farmers sell directly to domestic, regional, and international markets. They have their own product supplemented by that of outgrowers; thereafter they export the total consolidated volume (Fig. 9.1).

Regional markets

FarmerOutgrowers

Consolidate

Large Farmer

Domesc markets

Trade

Internaonal markets

Fig. 9.1 CHANNEL 1 schematic

9.5 Farmer-Trader Interaction Through Four Channels

145

The agents are comprised of 1. Farmers-Outgrowers 2. Large Farmers/Traders 3. Logistics Service Providers We distinguish large farmers from outgrowers on the basis of their ability to buy products. Since the budget assigned to every farmer is assumed, from the beginning, to be proportional to the amount of land, we assume that only large farmers are able to buy from other farmers. Furthermore, we assume that only farmers who have on average 1 acre of land can also be traders. Since the distribution of land is such that only a few farmers have over 1 acre of land, a relatively low number of farmers will be able to buy product. If this condition is met, we assume that on average only half of these are trading as well. Since we do not have information on the distribution of trading Large Farmers, we will decide who trades and who does not, by tossing a coin.

CHANNEL 2 Farmers organise into associations and export produce to domestic, regional, and international markets. 20–30 farmers organise into groups, producer associations, cooperatives, and marketing organisations in order to consolidate volumes and export them (Fig. 9.2).

Regional markets

Trade Consolidate Farmers

Farmer organisaons, PMOs Trade

Fig. 9.2 CHANNEL 2 schematic

Domesc markets

Trade

Internaonal markets

146

9 The Agriculture Supply Chain in Uganda: The Design of an Agent Based Model

Regional markets Consolidate Farmers

Inerant traders

Trade

Domesc markets

Fig. 9.3 CHANNEL 3 schematic

The agents are comprised of 1. Farmers 2. Farmer associations-organisations/Traders at domestic, regional and international markets 3. Logistics Service Providers

CHANNEL 3 Traders consolidate and export product. Itinerant traders purchase product from farmers and either export it or act as local agents for regional buyers (Fig. 9.3). The agents are comprised of 1. Farmers 2. Traders at regional markets and domestic markets 3. Logistics Service Providers Itinerant traders consolidate their product in situ.

CHANNEL 4 In this channel a firm does not grow product but instead consolidates product from farmers, carries out quality control and packaging, and exports to domestic markets, regional markets (South Sudan, Kenya, Congo, Rwanda, and Tanzania), and international markets. The firm sources product when there is demand, then consolidates and exports it, typically to international wholesalers and supermarkets (Fig. 9.4). The agents are comprised of 1. Farmers-Outgrowers 2. Trading firms (e.g., SULMA) to domestic, regional and international markets 3. Logistics Service Providers

References

147

Regional

Farmers-

markets

Trade

Consolidate

Domesc markets

Trade

Firm (e.g. SULMA)

Outgrowers Trade

Internaonal markets

Fig. 9.4 CHANNEL 4 schematic

To allow for the consolidation of products, whenever a product is sold and it reaches a destination, we assume that the product remains in one place for 2 days. As similar products arrive at the same location, these are consolidated and then shipped together. In order to represent the agriculture supply chain in Uganda, we have in this chapter outlined the methodological framework and hypotheses at the core of our analysis, and then described in detail the agents and their interactions of our proposed agent based model. In the next chapter we metaphorically place the agents and their actions within the Uganda context. In other words, we consider how to model all the different facets which represent the environmental factors of our model. We also demonstrate how the multiplier attachment introduced in Chap. 2 is implemented, and show how to model the decision process of the different agents. Finally, a calibration of the model is carried out before tackling our specific study challenges in subsequent chapters.

References Balmann A (1995) Path dependences in agricultural structure developments: term, causes and consequences. Working Paper. Humboldt University of Berlin Balmann A (1997) Farm-based modelling of regional structural change: a cellular automata approach. Eur Rev Agric Econ 24(1–2):85–108 Development Strategy and Investment Plan (DSIP) of Uganda (2011) THE REPUBLIC OF UGANDA. Ministry of Finance planning and economic development, in conjunction with Ministry of Agriculture, Animal Industry and Fisheries, Ministry of Education and Sports, and Ministry of Health Djankow S, Sequeira S, Djankow S (2010) An empirical study of corruption in ports. MPRA Paper 21791. University of Munich Freund CL, Rocha N (2010) What constrains Africa’s exports? World Bank Policy Research Working Paper, no. 5184. Social Science Research Network, Washington, DC

148

9 The Agriculture Supply Chain in Uganda: The Design of an Agent Based Model

Gollin D, Rogerson R (2010) Agriculture, roads and economic development in Uganda. NBER Working Paper, no. 15863. National Bureau of Economic Research Happe K, Kellermann K, Balmann A (2006) Agent based analysis of agricultural policies: an illustration of the agricultural policy simulator Agripolis, its adaptation and behavior. Ecol Soc 11(1):49 Hummels D (2001) Time as a trade barrier. GTAP Working Paper, no.18 Millennium Development Goals Report (MDG Report) (2011) United Nations Report. UN, New York, June 2001 National Development Policy (NDP) (2010) A transformed Ugandan society from a peasant to a modern and prosperous country within 30 years. The Republic of Uganda Portugal-Perez A, Wilson JS (2009) Export performance and trade facilitation reform: hard and soft infrastructure. World Bank Policy Research Working Paper, no. 5261. Social Science Research Network. Washington, DC Poverty Eradication Action Plan (2009) Uganda Ministry of Finance, Planning and Economics Uganda National Household Survey (UNHS) (2013) Uganda Bureau of Statistics. UNHS, Kampala World Bank (2012) Connecting to compete: trade logistics in global economics. World Bank, Washington, DC World Bank (2013) Uganda Report 77079-UG Diagnostic trade integration study. World Bank, Washington, DC

Chapter 10

The Implementation of the Uganda Agent Based Model

10.1

Data

We focus on the seven districts mentioned in Chap. 9, located in the Central and Eastern regions of Uganda—Luwero, Mpigi, Masaka, Iganga, Mitiyana, Kamuli, and Mukono. We use available data to define the spatial, geographical and demographic environment of the agent based system. The spatial and geographical environment is captured using a network structure. The nodes of the network represent the GIS locations of villages in Uganda. The links between the village nodes correspond to the available road infrastructures which are primarily dirt road or paved/gravel roads (Figs. 10.1, 10.2 and 10.3). We also use available data of districts and a population estimation tool in order to define the demographic environment of the villages (Fig. 10.4). The population estimation tool identifies the important markets in a district and distributes the population in radially reducing density. In other words, market towns are densely populated at the centre and thinly populated as we move farther away from the centre. Geographic data is supplemented with a weather class for predicting temperature variations over the course of 1 day, which is useful for monitoring the loss in quality of products in very hot temperatures (Fig. 10.5). GIS data is examined to construct the spatial data set of Uganda consisting of geographical and road infrastructure data. The geographical data consists of weather information and locations in terms of x and y coordinates. We have mapped the Uganda road infrastructure, i.e. pinpointing location and type (paved/dirt) of road in Figs. 10.1, 10.2 and 10.3. This information is crucial, as the quality of the road impacts on travel time. Using our road information and GIS location data, it is possible to estimate the distance between any two locations, the shortest route between the locations, and the average time taken to travel, depending on type of road and mode of transport. We calculate the travel time between any two points in the following manner:

© Springer International Publishing AG 2017 F.R. Medda et al., Collaborative Approach to Trade, Advances in Spatial Science, DOI 10.1007/978-3-319-47039-9_10

149

150

10

The Implementation of the Uganda Agent Based Model

Fig. 10.1 Uganda paved road infrastructure. Source: GIS data provided by the World Bank

Nðroute; p1 ; p2 Þ ¼

distanceðp1 p2 Þ τfrouteg

ð10:1Þ

where τ{route} is the average speed of travel of a specific mode of transport and distance(p1 p2) is the weighted Dijkstra distance on the infrastructure graph. The speed of travel depends on the type of road the transport device is moving on. The main modes of transport in our model are walking, bicycle, donkey, small truck, small truck with fridge, large truck, and large truck with refrigeration (Mwebaze 2006). We estimate maximum load, average speed and cost of each mode of transport from available web resources in addition to the survey conducted by the World Bank in Uganda for this study. Figure 10.5 depicts the simulated estimation of weather information using data sets of average precipitation and seasonal temperatures. Daily temperature variation affects the quality of the product during transportation if the mode of transport does not include refrigeration.

10.1

Data

151

Fig. 10.2 The Uganda national road network. Source: Lawyer Glenn (2012) http://dx.doi.org/10. 6084/m9.figshare.94407

Fig. 10.3 Dirt and gravel/paved roads in Uganda. (Red crosses indicate locations of intersections, blue lines are dirt roads and green lines are paved roads)

152

10

The Implementation of the Uganda Agent Based Model

Fig. 10.4 Location of farmers/outgrowers in the seven districts in the model

Fig. 10.5 Simulation of daily temperatures on an hourly scale

10.2

10.2

Multiplier Attachment

153

Multiplier Attachment

In the model our next step is to build a social network (including social media network) of farmers and traders that enables us to test the hypotheses and measure the related outputs. In other words, we aim to evaluate the importance of information propagation among farmers. A social network in our analysis is illustrated by using a graph consisting of nodes representing agents, and links between nodes representing connections between agents. We assume that farmers and traders have imperfect information: agents do not know in advance the value of linking to other agents. In our simulation, Multiplier Attachment is developed through two processes that evolve simultaneously: the network formation process and the information diffusion process. We describe each in turn. In network formation we assume N agents in the system {T1, T . . . TN}. Each agent Ti has an attribute pi which describes its location. We keep track of the social connections using an adjacency matrix Aij, such that Aij ¼ 1 if and only if Ti and Tj know each other, and zero otherwise. At each time step new connections are formed between agents Ti and Tj depending on two factors: the degree of connectivity, Dj, of the agent Tj, and the spatial distance, di , j between agents Ti and Tj (Daley and Kendall 1965; Albert and Barabasi 2002). We therefore assume that the probability of agent Ti knowing agent Tj, is proportional to d i, j

Pi , j e D j e  κ :

ð10:2Þ

The additional element introduced with the multiplier attachment is that we assume that a farmer will connect and coordinate his actions with another farmer who is relatively nearby, since this reflects the cumulative effect of experience and observations that both can share (Berger 2001). We therefore consider the standard Cochrane (1979) treadmill model in agriculture where farmers connect because they achieve lower unit cost of production. In the information diffusion process, at each time step T, we assume that each agent has knowledge of the price and quality of all the products owned by all other agents in his social network. Therefore, agent Ti has an attribute that is a price vector PriceTT i indexed by all trading agents Tj, all products owned by Tj, and the quality of the product   PriceTT i ¼ PriceTT i T j ; product; quality

ð10:3Þ

Information spreads in the following manner: at each time step T, price information owned by agent Ti is withheld from his social network with probability p and diffused with probability 1  p, where p depends on the degree of node Ti. In other words, if at time step T + 1, agent Tk knows agent Ti, then PriceTT k ¼ PriceTT i

ð10:4Þ

154

10

The Implementation of the Uganda Agent Based Model

The diffusion process of the network occurs so that agents who are highly connected spread information much faster, since the speed of diffusion is proportional to the number of farmers with whom he is connected. Each agent thus has a price list known to him. Furthermore, we note that the effective network for dissemination of market information is actually bigger that the network described by the adjacency matrix. If agent Tk knows agent Ti, then the price attribute PriceTT k of agent Tk is updated with the price information of all the social connections of agent Ti. This is known as the effective network of information, Eik, such that Eik ¼ 1 if and only if there exists an agent Tj such that Ti knows Tj (i.e. Aij ¼ 1) and Tj knows Tk (i.e. Ajk ¼ 1), and zero otherwise. There are two main advantages of being part of a social network: access to a larger market through social and social media connections of traders, and awareness of market price information. We examine how an agent uses his social network and social media connections to expand his business options in our discussion on decision trees in this chapter. We capture awareness of market price information for each agent using an attribute for expected price for a fixed product in the following manner PN PriceExpected TT i ¼

 Eij PN

j¼1

j¼1

Pricej Qualityj

Eij

 Qualityi

ð10:5Þ

where Pricej is the price at which trading agent Tj is selling the product with quality Qualityj.

10.3

Production of Crops

Farmer agents produce crops on the basis of a simple model to estimate the output and assets of a farmer. This enables us to control the parameters in the testing phase of the model. The production model depends on the amount of land the farmer owns and the amount of crop yields (Ronner and Giller 2013). The model is set up in three steps (a–c). a. Initial distribution of land We assume small holders to be the most common scenario in the country and that few farmers hold large amounts of farm land. Moreover, the probability of having a farm of a certain size should not be monotonic decreasing everywhere. Analytically, we choose a lognormal (LN) distribution for the size of farmers’ holdings. Explicitly   ! 1 1 ln ðyÞ  v 2 f ðyÞ≔pffiffiffiffiffiffiffi exp  2 λ 2 π λy

ð10:6Þ

10.4

Price of Crop

155

We select parameters v and λ in order to match the empirical distribution from the data. Thus, to allocate a share of land to a farmer, F, we draw a uniform random number u, and we model farmer F as having a share Φf1g ðuÞ, where Φ is the CDF of the LN distribution above. b. Yield of land The amount of crop C in a harvest from x acres of land of type L is modelled as Yield ¼ x λðC; LÞ

ð10:7Þ

Where λðC; LÞ is the yield constant in kg/acre of crop C on land type, L. c. Farmer assets The wealth of farmer F is denoted by WF and equals W F ðtÞ ¼ BF ðtÞ þ Market value of LF ðtÞ þ Market value of CF ðtÞ

ð10:8Þ

The market value of LF can be estimated using the value of land in Uganda. We assume that most farmers market the surplus of their output, with the fraction varying by crop, by region and by distance from markets. Through the use of this model we are able to estimate the crop production of each farmer and his wealth at the end of each production cycle. We calibrate the value of the farmer assets using data from Sect. 9.2. As Uganda is a tropical country, temperature affects the quality of harvested product. We model the quality deterioration rate of a product using an exponential decay function QðtÞ ¼ q0 eðdðT ÞtÞ ,

ð10:9Þ

where d(T ) ¼ a0 + a1T is the decay rate that depends on the temperature.

10.4

Price of Crop

To calculate price in relation to supply and demand, at each time step in the trade process, price P of crop C traded by trader, T, is adjusted to reflect the imbalance of supply and demand. The price varies according to the Walras Law dP ¼ γ ðD  SÞ, dt

ð10:10Þ

where γ is a constant to incorporate the action of the Walras Law, which is the price reactivity; D and S are demand and supply for crop C from trader T, and ED is the excess demand. For Trader A, supply S is calculated as the sum of the yields of all

156

10

The Implementation of the Uganda Agent Based Model

Exporter social network (minimum volume criterion is less than Trader A’s volume) Farmer social network Yield Farmer 1 Yield Farmer 2 … … … Yield Farmer n

Demand

Trader A

Supply Demand

Exporter 1 Exporter 2 … … … Exporter n

Trader in social network (minimum volume criterion is less than Trader A’s volume) Trader 1 Trader 2 … … … Trader n

Fig. 10.6 Supply-demand schematic for Trader A

the farmers in his network. The demand D is calculated by summing the minimum trade volumes of exporters and traders in the social network of Trader A, such that Trader A is able to meet the minimum volume criterion. At each time step, demand will be mediated with previous time step demand (Fig. 10.6). This can be achieved by mediating the current time step demand with those from the previous time step Dt ¼ ð1  kÞ Dt þ k Dt1 ,

ð10:11Þ

where we impose k to be 0.25 and we assume current demand counts as one-fourth of the trend.

10.5

Trading

Decision trees form the core of our trading strategy. A diagram of a decision tree, as illustrated in Fig. 10.7 is read from left to right. The left-most node is called the root or decision node. The branching lines to the right from a decision node represent the set of available decision alternatives. One, and only one, of these alternatives can be selected. The small circles in the tree are called chance nodes. The right end of each

10.5

Trading

Fig. 10.7 Depiction of the decision tree process

157

158

10

The Implementation of the Uganda Agent Based Model

path through the tree is an endpoint, and each endpoint represents the final outcome of following a path from the root node of the decision tree to that particular endpoint. Every trading agent makes trading decisions using a decision tree. Through such a rule-based decision making process we can evaluate the outputs related to hypothesis (3). The decision tree used in our simulation has sufficient flexibility so it can be used by any type of trading agent, whether small farmer with crop to sell or large farmer or farmer cooperation with a consolidated products to sell. We can describe the decision making process in detail using Fig. 10.7. At the first stage, the trader must decide if he will sell his product or store it and sell later. If he decides to store it, there are two options available to him: he can either store his produce (in the decision tree, we assume that the farmer will consider 1 week) in a warehouse or on his farm. The former option incurs storage costs while the latter incurs costs in the loss of quality of the product, and hence a reduced expected price. If he decides to trade he checks his contract agreement and will either honour it (the outcome is the fixed price agreed in the contract) or incur a fee to break the agreement. If he breaks the contract, he then evaluates all the available trading options, based on logistics costs, risks involved with trade, and his own inclination to risk (or not risk) his wealth. The decision at the end of each branch can be described using the following function Decision ¼ ð1  U ðC0 ÞÞðPR  CÞF,

ð10:12Þ

where F is the feasibility of the trade, P is the probability of the risk involved in the trade, C is the total cost, R is the reward, U(C0 ) is the utility function Eq. (10.17) and C0 is the implied cost.

Feasibility In our simulated environment any two traders can in theory trade with each other. A trader agent Ti considers a trade with another trader agent, Tj, provided he considers it to be a ‘feasible’ option. We incorporate this notion of feasibility using a function   F Qualityi ; Quantityi ; minQualityj ; minQualityj 8 minQualityj and ¼ Quantityi > minQuantityj : 0, otherwise

ð10:13Þ

10.5

Trading

159

where Qualityi and Quantityi are the quality and quantity of the product owned by trader agent Ti, and minQualityj and minQuantityj represent the quality and quantity of the product demanded by trader agent Tj.

Costs The costs are the logistics costs of transporting goods between two locations p1 and p2. Cost ¼ Distanceðp1 ; p2 Þ ∗ TCostmode þ Storage þ Processing þ BreakAgreementFee

ð10:14Þ

where Distance( p1, p2) is the weighted Dijkstra distance on the infrastructure graph in kilometres; TCostmode is the transport cost in per unit of distance travelled for the mode of transport; Storage is the storage costs; Processing are the processing costs for the product; BreakAgreementFee is the fee incurred for ending the agreement.

Reward When trader agent Ti initiates a trade with agent Tj, Reward is the expected value of the product at the expected time of delivery Reward ¼ Quantity ∗ Pricej ,

ð10:15Þ

where Pricej is the price at which trader agent Tj is willing to buy a unit of produce. The expected time of delivery is calculated as Nðroute; p1 ; p2 Þ ¼

Distanceðp1 p2 Þ , τfrouteg

ð10:16Þ

where τ{route} is the average speed of travel of a specific mode of transport.

Probability of Deal Failure Each trader agent associates a probability of trading risk to all other trader agents in the system. This probability is initially set to 1 for all agents, which suggests that, a priori, the trader expects all trader agents to be trustworthy and to honour their commitments. However, each time the trader initiates a trade and does not receive the expected payment, either due to delays and loss in quality of the product or

160

10

The Implementation of the Uganda Agent Based Model

because of a dishonest counterparty, the probability of trading risk associated to the counterparty trader reduces by a constant value (of 0.2) until it reaches zero.

Utility Function We associate a utility curve for each trader agent. The utility curve represents the risk averseness of a trader to pay for logistics costs relative to his private wealth. He is less risk averse when costs stay low but becomes increasingly risk averse as logistics costs rise (Fig. 10.8). We use a utility function of the following form 8 < logð1 þ scÞ , 0 B Parameter s determines the slope of the curve and hence the speed with which the function saturates. At this point in our analysis, we can draw from our earlier discussion of the channels because they are now reflected in our model. Let us use CHANNEL 1 as an example (Fig. 10.9):

Fig. 10.8 Wealth utility function U(C’) for B ¼ 100

10.5

Trading

161

Regional markets Consolidate

FarmerOutgrowers

Domesc markets

Trade

Large Farmers

Internaonal markets

Fig. 10.9 CHANNEL 1

Regional markets

FarmersOutgrowers

Consolidate

Large Farmers

Trade

Firm (SULMA)

Domesc markets

Trade

Internaonal markets

Fig. 10.10 The simulated chain can accommodate many agents, as shown by the inclusion of the firm SULMA

At some period in the past, the farmers-outgrowers either contacted or were contacted by a large farmer and signed an agreement for the sale of his crop product. When the crop is harvested and ready for sale, the farmer assumes the role of trader and is at the root node of the decision tree process. At this stage, the farmer has a number of options, based on his social network and the feasibility of pursuing those options. In CHANNEL 1 he decides to honour his agreement with the large farmer because it offers the best outcome. In the next step it is now the large farmer who is at the root node of the decision tree process. Once again, the large farmer has choices based on his social network. In CHANNEL 1 he limits these choices to three categories: regional market, domestic market and international market (Fig. 10.10).

162

10.6

10

The Implementation of the Uganda Agent Based Model

Simulation of the Channels in the Model

In the model we simulate the behaviour of farmers and traders; behaviour of exporters is not included in the study. Farmers grow and sell their crops following the decision tree as traders; however, traders behave differently than farmers, in that they search for products in the market in order to re-sell them at a higher price. We have assumed that large farmers in Uganda buy from other farmers and this assumption is addressed in the model by introducing agents who are traders and farmers at the same time: they cultivate crops, search for buyers and sell product. Itinerant traders are an important component of our model. They have another feature compared to the other traders, the ability in the model, to move using the infrastructure, and to buy at farm gate. Itinerant traders are assumed to have lower transport costs in comparison with other traders. Itinerant traders are thus simulated in the model as traders who travel on the infrastructure network.

Itinerant Trader When the Itinerant trader’s mode of transport reaches capacity, they sell their product and begin the search once again. One important difference between itinerant trader and regular trader is related to the quality control policy (GAP). Itinerant traders collect products at the farm gate, and in this way the farmer’s transport cost is lower. However, although the transport cost is lower for the farmer than if he were to deliver the product himself, his opportunity cost becomes higher because the Itinerant trader picks up his product at farm gate, thus incurring a different cost. However, we assume that when an Itinerant trader collects a product, the product is mixed with the other products already collected, thus losing the ability to track a product back to its origin. This is related to the GAP policies we will consider in Chap. 13, as these traders will have the disadvantage of not being able to sell to exporters when GAP policies are applied (Fig. 10.11).

Conditions for Market Exit Agents can decide to exit the market if their debt is higher than their assets. For farmers, this means that they will no longer produce a surplus of their crop and will go into subsistence. For traders, it means that they will stop trading. For each agent, his asset is given by the sum of their trading volume and the property they own. Land has a market value, while trading asset is based on the trading volume. We assume in the evaluation of the trading volume that only the last 3 months of the year are considered.

10.7

Model Calibration

163

Fig. 10.11 Simulation of itinerant trader in agent based model

10.7

Model Calibration

Product Parameters In the model we consider five different products grown in Uganda: hot peppers, chillies, matooke, okra, and sweet potatoes. In Table 10.1 the growing rate for each crop is expressed in percentage rate per hour in column two, and this value has been estimated by taking into account that these crops grow to harvest maturity over periods of, respectively, 4, 3, 5, 2 and 4 months. We also calculate the spoilage rate parameters in column three under the assumption that the functional form of the spoilage is exponential. The parameter in column four is a calculation based on the impact of temperature; these parameters are all estimated according to the realistic spoilage rate of the products. In so far as the cost of maintenance per acre/per hour is concerned, we have made a strong assumption. In the absence of data on the specific costs for each product, and instead by having only the aggregated values, we assume that for each product the cost of growing per hour is the same as that which has been estimated from the survey conducted by the World Bank in Uganda for this study, resulting in approximately 200 USh per hour/per acre. Seeding costs were also considered and are included in

164

10

The Implementation of the Uganda Agent Based Model

Table 10.1 Product types and their estimations Product type Hot peppers Chillies Matooke Okra Sweet potatoes

Growing rate (%/h) 0.000347222

Spoilage rate 0.128642

Spoilage rate/ temp 0.00047352

Tonnes/ acres 6.5

Cost/ acres/h 200 USh

Seeding cost 0

0.000462963 0.000277778 0.000694444 0.000347222

0.0965432 0.259104 0.9828852 0.211632

0.00035493 0.00093370 0.00029814 0.00007986

6.5 4 2.6 4

200 USh 200 USh 200 USh 200 USh

0 0 0 0

the aggregated cost. The model is however able to incorporate the costs at the beginning of production, although these are set to zero. In the absence of data, we can only give an approximation of the scale of production cost for each crop, as shown here, resulting from our estimation. The model has nevertheless been developed so that with the inclusion of detailed data, it can simulate single product costs for each crop.

Transport Parameters Some of the parameters used for the calculation of transport are estimated from the survey conducted by the World Bank in Uganda and others are assumed. The cost of transport for each mode has been estimated directly from the survey. It is important to keep in mind that we calculate these costs as dependent only on number of kilometres, not on the time taken to move product from one place to another. Here again we have a simplification of the actual situation for our case studies with respect to delivery of product; but we can nevertheless estimate the scale of the costs related to transport (Table 10.2). Travel speeds are estimated by considering the conditions on the ground, that is, transport mode, road type, different weights of product, and distance.

Market Parameters The market parameters were the most complicated to estimate due to the lack of information on prices for the current Existing Conditions. However, we have aimed to recreate in the model the actual conditions of difficulty for the farmers consistent with farmer expenses. Thus, we introduce an atypical initial condition for the prices of the traders (which are randomised at around 10% of the initial price): we take the average initial condition for the prices of goods offered by the traders to be the actual market clearing prices. We consider these selling prices on the basis of the cost of producing 1 tonne of product. The prices at local market are thereafter

10.7

Model Calibration

165

Table 10.2 Delivery transport modes and variables related to the commercial movement of our selected crops Transport mode Bicycle Motorcycle Pickup Pickup Lorry

Speed: dirt roads (km/h) 15 35 50 50 50

Speed: paved roads (km/h) 20 45 90 90 90

Weight (tonnes) 0.06 0.11 1 1 7.5

Cost (per km) USh 300 614 4649 4849 20,478

Fridge 0 0 0 1 0

estimated at a slightly lower value (i.e. 10%) on average. In the case of the exporters we can clearly notice that the prices offered to them are nearly double the price offered to traders (i.e. 80% higher). Despite the fact that these prices are not drawn from price information data, we are nevertheless in keeping with the actual conditions faced by farmers in Uganda. And, as we will see, many farmers leave the market under these conditions, which happens whenever a farmer has more debt than assets. Thus, with the actual prices obtained from the survey, the simulations have shown that all the farmers left the market within the year (0% survival). As such, we found empirically the market prices in which at least a non-zero percentage of farmers survived 1 year (10% survival). This was the equilibrium condition we considered in the model. These prices are obtained in Table 10.3. The price reactivity parameter for traders shown in Table 10.3, used in the Walras Law equation, has been found empirically, such that in the Existing Conditions scenario prices are approximately stable over the year (net increase of 2% in 1 year). The two association parameters, Distance and Probability, have been chosen in accordance with the literature, where it is shown that association occurs within 50 km. We took a more conservative line and fixed this parameter to 30 km. The association probability in the Existing Conditions scenario is set to zero (we do not allow association), but later when we examine our simulations, we set the probability to 0.1, which implies that only 1/10 of the farmers enter into an association. For us it is critical to show the importance of coordination for farmers, and thus for us to set this parameter to a non-zero value is sufficient.

Agent Parameters We have empirically chosen the parameters associated to the agents. On the basis of the given data, i.e. number of agents, we simulate a period of 1 year at 6 h time steps, and run 30 simulations during 1 day. The ratio between farmers and traders, and traders and exporters has been kept to 1:6 and 1:10, respectively (Table 10.4). The average amount of cash for outgrowers/farmers is chosen according to the actual cost of farming in Uganda. We assume that farmers have a budget such that without selling a product, they are able to pay living expenses for 1 month. The

166 Table 10.3 Market parameters showing price differentials between local markets, traders and exporters

Table 10.4 Agents, variables and their values

10

The Implementation of the Uganda Agent Based Model

Parameter Exporters Average price exporters Hot peppers/tonnes Chillies/tonnes Matooke/tonnes Okra/tonnes Sweet potatoes/tonnes Traders Initial average price traders Hot peppers/tonnes Chillies/tonnes Matooke/tonnes Okra/tonnes Sweet potatoes/tonnes Local markets Average price local markets Hot peppers/tonnes Chillies/tonnes Matooke/tonnes Okra/tonnes Sweet potatoes/tonnes Price reactivity traders Association parameter: distance Association parameter: probability

Variable Total farmers Total large traders Total itinerant traders Total exporters Total warehouses Bank asset ratio Taxation factor Average cash farmers/acre Average cash traders/vol Average cash itinerant/vol Average minimum accepted tonnes Traders Exporters Itinerant traders Information spreading probability Number of districts Land distribution parameter 1 Land distribution parameter 2 Land value

Value

160,564.1 USh 120,923.08 USh 300,000 USh 180,205.1 USh 240,666.7 USh

88,615.3846 USh 66,461.5385 USh 180,000 USh 110,769.231 USh 144,000 USh

78,564.1 USh 60,923.08 160,000 100,205.1 130,006.7 0.0001 30 km 0

Value 600 70 100 10 10 0.1 0.2 288,000 USh 20,495,000 USh 4,099,000 0.04 2 0.01 0.05 7 1 0.5 10,000 USh/acre

10.7

Model Calibration

167

average amount of cash for traders is selected based on the minimum accepted volume of product, which needs to be enough to provide economic stability; for traders we have assumed the minimum accepted volume to be on the order of 400 kg, while for itinerant traders we assume a budget that is approximately 1/5 of the value of traders, on average. The number of warehouses is set to 10, with at least one warehouse per district, but warehouses are thereafter positioned at random as a function of the density of farmers. Although we do not have information on the tax rate for Uganda farmers, we nevertheless include a taxation value, 0.2, which corresponds to a value between the lowest value in the tax table of Uganda (10%) as well as the next increment (30%). We also assume that, should a farmer need it, a bank will provide a loan of up to 10% of the owned assets of the farmer at 0% interest, to be repaid within 12 months. The value of farm land is estimated from Uganda property value websites (http:// ugandaproperty.org/), at 10,000 USh/acre. The distribution of land is adjusted such that the average amount of land held by farmers is 3 acres; this value is calculated based on the two land distribution parameters and is depicted graphically in Fig. 10.12. The spread of information probability has been found empirically, in that all the information is spread over a 1 year time period (i.e. all the contacts from one agent are transferred to another agent within 1 year).

Fig. 10.12 Log-normal distribution of land

168

10

The Implementation of the Uganda Agent Based Model

Now that we have fully calibrated the model, our final step is towards examining specific cases and drawing policy conclusions. In Chap. 11 we explore transport costs and infrastructure investment in further detail in order to suggest feasible policy.

References Albert R, Barabasi AL (2002) Statistical mechanics of complex networks. Rev Mod Phys 74: 47–97 Berger T (2001) Agent based spatial models applied to agriculture: a simulation tool for technology diffusion, resource use changes and policy analysis. Agric Econ 25(2–3):245–260 Cochrane WW (1979) The development of American agriculture. A historical analysis, Chapter 19. University of Minnesota Press, Minneapolis Daley DJ, Kendall DG (1965) Stochastic rumours. IMA J Appl Math 1(1):42–55 Mwebaze SMN (2006) Country pasture/forage resource profiles UGANDA. FAO Ronner E, Giller KE (2013) Background information on agronomy, farming systems and ongoing projects on grain legumes in Uganda. (N2Africa)

Chapter 11

Transport Cost and Infrastructure Investments

11.1

Introduction

The trade reforms initiated in Uganda during the 1990s have gradually led to trade liberalisation aiming to improve competitiveness and reduce distortion in the markets. For instance, the granting of preferential treatment in the form of duty reductions to COMISA countries, and the East African Community Custom Union treaty signed in 2004 are two pro-trade initiatives that have helped lower the barriers to trade, such as restrictions and tariff controls, and have significantly increased the volume of traffic of non-traditional agriculture exports.1 The amount of non-traditional agriculture exports has risen strongly from 14% in 1990 to over 68% in 2012, in line with Uganda’s annual average real GDP growth rate of 7.5% over the last 25 years (Uganda Government Statistics 2013). Strong export rates convincingly show us that agriculture production is pivotal in the economic growth of Uganda. Furthermore, according to UNCTAD (2004), much of Uganda’s export trade is destined for Europe and Asia (in the last decade Asia has tripled its share of imports), thus indicating the significant role of transport infrastructures and transport costs in the development and growth of trade. For this reason, the trade deficit in Uganda is often linked to the substantial lack of infrastructures, and additional spending on capital improvements and maintenance has been highly recommended (AICD 2011). Within this context it is noteworthy that the level and quality of infrastructure is considered by many experts (World Bank 2001; Hummels 2001; Venables 1999) to be a discriminative factor for the growth of trade flows. Under this perspective, transport infrastructures and transport services become essential intermediary inputs for the many sectors in the economy. Uganda is a landlocked country and most of the trade flows consist of, often perishable, agriculture products which are subjected to stringent standards. Therefore, the development of an 1

Whereas traditional agriculture products include coffee, cotton, tobacco, and tea.

© Springer International Publishing AG 2017 F.R. Medda et al., Collaborative Approach to Trade, Advances in Spatial Science, DOI 10.1007/978-3-319-47039-9_11

169

170

11 Transport Cost and Infrastructure Investments

efficient transport system in terms of speed, reliability and affordability is critical to trade development and growth in the country (Rudaheranwa 2009). Transport costs in Uganda at all levels are very high. Svensson and Yanaizawa (2009) show that transportation costs between farm gate and district markets, and costs between wholesale markets in Uganda are on average seven times the unit cost in the USA (assuming an average distance of 25 km). On the basis of this finding the authors conclude that transport costs in Uganda are about two orders of magnitude greater than in the USA. Air freight costs for perishable products are often 50% higher than the unit price of the products. In a study carried out by ADC (1999) on the cost of a shipment of passion fruits from Uganda to the UK, the results show that air freight costs represented 49% of the total cost, insurance and freight (CIF) in 1996. As a consequence, the cost burden imposed by transportation costs strongly impacts on export prices, which then lowers production efficiency and erodes returns to capital, and ultimately decreases the possible share of investments in the country. However, to simply invest in transport, particularly the roads, through significant investment policies may not reach the expected results in terms of increases in trade flows. In the last decade the Uganda government has invested significantly in road construction and road improvements, and although the RAI (rural access index) has reached only 30%, many experts (Gachassin et al. 2010; Raballand et al. 2009; Lall et al. 2004) have argued that, given the characteristics of the agriculture sector in Uganda, i.e. most are subsistence farmers, “reaching an RAI at 100% should not be a government objective for rural roads, given the existing transport patterns, because the expected benefits will be minimal—whereas the investment required is unaffordable” (Raballand et al. 2009). And in relation to the affordability and effectiveness of road investment to reach 100% RAI in Uganda, Carruthers et al. (2008) have estimated that, for an investment programme to reach 75% of RAI, an allocation of 3.6% of the yearly Ugandan GDP for a period of 10 years would be needed. In the trade supply chain for Uganda there are numerous determinants of transport cost, of which road transport cost is just one. Importantly, the export supply chain is strongly anchored on its airport and port connections. Interestingly, Uganda’s major international airport in Entebbe has shown increasing demand over the last decade with, for example, an increase of 9.6% in its passenger flow between 2012 and 2013 (CAA 2013). Moreover, Uganda’s trade exports may also pass through either of two ports (Mombasa and Dar-es-Salaam) for containerised cargo via several routes using different transport modes (rail, boat and truck). In this chapter we examine whether investment in the transport system can result in a reduction of time and cost of logistics and, subsequently lead to an increase in trade flows. We observe, however, that the transport logistics chain, particularly in our Uganda case, is constituted by different elements that interact and which have different impacts on our overall objective, and for this reason the study of only one transport intervention would not be particularly useful. Using the benchmark Existing Conditions, our analysis examines two possible transport intervention scenarios. In the first transport intervention we analyse the impacts of a possible

11.2

Scenario 1: Existing Conditions

171

investment to pave the roads of the entire Central and Eastern region, which encompasses seven considered districts (Luwero, Mpigi, Masaka, Iganga, Mitiyana, Kamuli, and Mukono). In our second intervention we assume an improvement only in the area around Entebbe Airport. We aim in this chapter to investigate which of the two transport interventions can best achieve an increase in welfare for Uganda farmers. To carry out our analysis we focus on the agents, that is, the subsistence farmers, who may be more predisposed to exiting the market if they can no longer afford to sell their product; these farmers are the most vulnerable agents in our agent based model because they are more significantly affected, i.e. they may exit when small fluctuations occur in the market and in their environment. Subsistence farmers are often characterised by a mind-set of short-term survival, as they are confronted with very small revenue margins, and are consequently the most responsive to even slight changes in the supply chain. Our objective is to test which of the two transport interventions reduces the number of farmers exiting the market otherwise. The Existing Conditions scenario and the two scenarios with transport intervention are compared in the final part of the chapter in order to reach and discuss policy recommendations.

11.2

Scenario 1: Existing Conditions

Uganda is dependent on its road network for 90% of its volume of freight and human movement. The road infrastructure constitutes approximately 70,500 km of national, district, urban and community access roads. The national road infrastructure accounts for about 30% of the road sector network, amounting to about 20,000 km in length, of which 29% is paved and 71% is unpaved gravel. These roads carry approximately 85% of the total road traffic and provide vital transport corridors linking Rwanda, Burundi, Eastern DRC, and Southern Sudan to the sea. Urban roads account for 5% of the road infrastructure and are mostly paved. District roads constitute 24% and community access roads about 43% of the road infrastructure; these are mostly unpaved or dirt roads (Fig. 11.1). Although Uganda has adequate road density and high traffic volume when compared to its East Africa neighbours, the poor condition of its district and rural feeder roads provides low accessibility to rural areas (Ranganathan and Foster 2012). The poor quality of district and community access roads and the resulting high logistics costs are a debilitating factor in the development of the agriculture sector in Uganda (Natamba et al. 2013; Gollin and Rogerson 2010; Collinson et al. 2005). According to the Civil Aviation Authority in Uganda (CAA 2013), given Uganda’s geographical features, a large share of its perishable exports is airlifted to destinations from the Entebbe International Airport (EIA). As shown in Fig. 11.2, non-traditional agriculture exports moving through EIA grew steadily from 1367 tonnes in 1991 to 33,820 tonnes in 2012. The major exports transiting through the

172

11 Transport Cost and Infrastructure Investments

Fig. 11.1 Uganda’s road infrastructure (www. roughton.com)

airport are fish products (14,882 tonnes in 2012), followed by fresh produce (7900 tonnes in 2012) and flowers (6723 tonnes in 2012). However, as we have mentioned above in regard to passion fruits, air freight costs in EIA for perishable goods (i.e. landing costs, parking fees, aviation fuel, aeronautical charges, etc.) are steep, and may be as high as 50% of the unit price. Entebbe Airport fees are one of the highest in relation to other airports in Africa, including Nairobi, Kigali and Addis Ababa. From data collected by Rudaheranwa (2009), the charge for a B747 aircraft with a capacity of 395 tonnes in Entebbe is equal to US$1975; whereas in Nairobi Airport the fee is US$1750 and in Dar es Salaam, US$1430. In order to simulate the two policy scenarios, we first map the roads and their existing conditions based on GIS data from the Africa Infrastructure Country Diagnostic map. In Fig. 11.3 we depict the starting point for our simulations the current road infrastructure. Green lines represent paved roads (gravel or bitumen paved); blue lines represent the majority of the roads in the country, which are dirt; and red Xs indicate the location points where we have taken measurements. In our model we assume that the farmers are in a subsistence status (farmers/ outgrowers), that is, they produce agriculture goods for their own consumption and only a fraction of their production is sold in the market. In the Existing Conditions context, farmers/trader agents leave the market when they have no production surplus to sell. In the next sections we describe the two intervention scenarios and provide a comparison of the results. Although we have allowed for a simplified analysis, one of our first observations after applying our ABM approach was to notice that the improvement of infrastructures does indeed impact on the micro-economic behaviour of the farmers. Agents/ farmers are affected by the transport road improvement because they can travel at

11.2

Scenario 1: Existing Conditions

173

n/ 2500

EXPORTS BY TYPE (TONNES) THROUGH ENTEBBE INTERNATIONAL

2000

1500

1000

2007

FISH

FLOWERS

JUL

NOV

SEPT

JAN

MAY

MAR

NOV

JUL

2011

SEPT

JAN

MAY

MAR

NOV

JUL

2010

SEPT

JAN

MAY

MAR

JUL

NOV

MAY

2009

SEPT

MAR

JUL

2008

JAN

NOV

SEPT

JAN

MAY

MAR

JUL

NOV

SEPT

JAN

2006

MAY

MAR

JUL

NOV

MAY

SEPT

MAR

JAN

NOV

JUL

2005

SEPT

JAN

MAY

MAR

NOV

JUL

2004

SEPT

JAN

MAY

MAR

NOV

JUL

2003

SEPT

JAN

MAY

0

MAR

500

2012

FRESH PRODUCE

Fig. 11.2 Exports by type expressed in tonnes through Entebbe International Airport. Source: BAA

higher speeds, thereby lowering travel time and reaching the market more quickly (time and market accessibility). In so doing, they are able to reduce their postharvest loss (transport cost takes the form of an iceberg cost and is symmetric in relation to origin and destination). Furthermore, given the increase in market accessibility, farmers/agents will augment their capital liquidity because trade exchanges will be more frequent and farmers and traders will become more connected. Also, by following the model of Gollin and Rogerson (2010), the transport interventions can lead to a decrease of transaction costs and to higher financial stability. As a consequence, farmers will be less likely to rely solely on their agriculture product for their subsistence and instead be able to accrue the advantage of their surplus production. Let us next discuss scenario 2, the improvement of road infrastructure.

174

11 Transport Cost and Infrastructure Investments

Fig. 11.3 Uganda’s road infrastructure embedded into the model

11.3

Scenario 2: Improvement of Road Infrastructure

In scenario 2 we assume the first transport improvement: that the roads in districts Luwero, Mpigi, Masaka, Iganga, Mitiyana, Kamuli, and Mukono are improved to the standard of all-season roads in order to reach a 100% RAI level. In these districts 37% of the roads are already paved. We have taken a conservative approach rather than to assume 100% RAI in all of Uganda’s roads, since in the implementation of road infrastructure we consider the improvement of only the most agriculture-based districts (Fig. 11.4). Our choice is motivated by (Raballand et al. 2009; Fan et al. 2004) who stress the unaffordability of such massive investment in rural road infrastructure, and claim that farmers/ outgrowers generally transport their product on a weekly basis (approximately 100–200 kg) by bicycle or motorcycle. However, we want to confirm whether in such areas this type of target intervention is likely to produce spillover effects in terms of productivity and technological advancement. We have observed that the improvement of rural road infrastructure certainly leads to a domino-effect situation where farmers’ access to markets is expanded,

11.3

Scenario 2: Improvement of Road Infrastructure

175

Fig. 11.4 Road improvements at 100% level of RAI for districts Luwero, Mpigi, Masaka, Iganga, Mitiyana, Kamuli, and Mukono

increased market participation and competitiveness takes place, and better prices and increased household incomes are the end result (Barrett 2008; Walle Van De 2002). Another way to measure the impact of improving road quality is through export trade flows. Blyde and Iberti (2014) show that, in Chile, an improvement in the quality of all its roads could generate an average reduction of transport costs equal to 16%, and an increase of average exports by around 2%. Lastly, improvements in rural infrastructure enhance farmers’ ability to deliver crops to market quickly and minimise post-harvest losses (Hodges et al. 2011). We will verify the results of our scenarios in relation to the number of farmers exiting the market in the final section of this chapter. In the meantime, let us next examine the scenario of the second transport improvement, the Entebbe International Airport.

176

11.4

11 Transport Cost and Infrastructure Investments

Scenario 3: Improvement of Airport Infrastructure

In scenario 3 we consider the second transport intervention in our ABM simulation: the introduction of an intervention at Entebbe International Airport to improve export trade flows. Entebbe Airport, as observed by the African Development Bank, is notorious for its high charges in relation to other airports in East Africa. Moreover, the airports in Kenya, Tanzania and Rwanda are planning investment in order to refurbish their facilities. In particular, Kenya will invest US$650 million in the terminals at Jomo Kenyatta International Airport; Tanzania will receive US $164.3 million from the Dutch government to expand the Julius Nyerere International Airport in Dar-es-Salaam, while Rwanda plans to build a new airport at an expected cost of over US$650 million (The East African 2014). It is important to mention that high transport costs always feed into trade export flows, that competitiveness is very high, and foreign importers will always regard the freight costs as a main discriminative factor; so they are likely to choose to move their goods from other airports rather than Entebbe. At present, according to different experts, Entebbe International Airport seems to be playing the role of bottleneck in the trade supply chain of agriculture products due to its low cargo flight utilisation and higher costs in relation to its competitors (CAA Uganda 2013). Possible improvements at the Entebbe Airport should aim to achieve a reduction of travel time and travel cost by tackling the possible inadequate infrastructure (for instance, according to ENHAS (2013), Entebbe Airport has storage facilities of 120 tonnes with a total capacity of 230 m3 and is planning the construction of an additional cargo storage and refrigeration storage for up to 110 tonnes). Given the aforementioned context and the critical importance of intermodality in the supply chain, we can safely say that the improvement of nodes in the supply chain, particularly the nodes with high levels of traffic flow, will determine decreases in time and cost of transport. Against this background, it is important to test a possible improvement in the Entebbe Airport; in our particular case, however, we have insufficient data to introduce into the simulation on the investment and performance of the airport. Nevertheless, we can observe and verify that an improvement in one element of the chain is directly related to the other elements. This important observation allows us to surmount the data constraint. To achieve this, we assume that a possible intervention at Entebbe Airport is equivalent—in terms of the reduction of transport cost—to an improvement of the roads around the airport. We calculate the road improvement at the standard of all-season roads, i.e. 100% RAI on the roads around Entebbe Airport within a radius of 30 km (Fig. 11.5). At present, only 21% of the roads surrounding the airport of Entebbe within a radius of 30 km are paved, indicating that the large majority of the roads in the area around the airport do not reach the 100% RAI standard. Therefore, an improvement of the roads around the airport, to reduce transport cost and time, can be linked with possible improvements of the airport in order to produce the same effects in terms of transport cost and travel time reduction. Additionally, in our assumption, the

11.4

Scenario 3: Improvement of Airport Infrastructure

177

Fig. 11.5 Improvements of Uganda Entebbe International Airport

choice of the 30 km radius has been imposed for a specific reason. We know that the relationship between transport cost and distance is not always linear, particularly for long distances, since transport cost decreases per-unit-distance travelled. In our case, within the 30 km distance from Entebbe Airport, we are able to confirm a linear relation between distance and transport cost. Similar analyses based on indirect impacts have been developed on port efficiency and investments (Clark et al. 2004; Limao and Venables 2001; World Bank 2001). In these studies possible improvements in ports are estimated as equivalent to a reduction of travel distance to/from the market and are estimated as such. The assumptions used in this scenario are purposely conservative in order to be cautious about the results. The goals of these scenarios have been twofold. We have tested indirectly whether possible improvements at Entebbe Airport, which reduce transport time and thus transport cost, will significantly impact on trade and especially on the welfare of subsistence farmers. This objective can also be

178

11 Transport Cost and Infrastructure Investments

interpreted by evaluating the actual road improvement around the airport, and to ascertain whether it is more efficient to intervene on the roads around the airport or to improve the roads in the seven examined districts, as in scenario 2. And secondly, through the use of agent based modelling, we have been able to demonstrate the flexibility and adaptability of the method in our analysis of three different scenarios. In testing the assumptions we have considered different scenarios and different travel time reductions. The results that we compare in the next section involve the most cautious simulations of our tests. We have assumed a fixed speed, based on an improvement of the road around the airport, and we achieve with 100% RAI a 60% decrease in travel time.

11.5

Results and Policy Recommendations

Our study in this chapter has focussed on the impacts of transport costs and improvements in transport infrastructure in the trade chain of non-standard agriculture products in Uganda. The agent based model used here has provided us with a representation of the complexity of real trade exchanges and interactions, but it can also help us identify the constraints of different policy initiatives. The results of the two simulations with the Existing Conditions scenario are depicted in Fig. 11.6. In the horizontal axis the time span is expressed in days, while the vertical axis indicates the number of farmer/outgrower agents who will leave Investment normalized Bankrupt Agent

300

Normalized Bankrupt Agents

250

200

150

100

50

0

Scenario 2:road intervention Scenario 1:airport intervention Current Conditions

0

50

100

150

200

250

300

350

Days

Fig. 11.6 Comparison of the three scenarios in relation to the agents exiting the market

11.5

Results and Policy Recommendations

179

the market (the value is a normalised function of the Km of road being improved). The three scenarios are identified using three different colours. Over the long-run (1 year), we can conclude that when we target the reduction of transport time and cost, both transport interventions (road or airport) have long-lasting benefits in terms of trade growth and increases in overall productivity and welfare. The two transport interventions do not converge in the long-run toward the Existing Conditions trend, which always remains above the two transport interventions; the implication here is that transport intervention can effectively target subsistence farmers who, due to the improvements, will now be less exposed to economic risks and will thus remain in the market. Although the number of agents who leave the market increases in the short-run (i.e. within 6 months), in the long-run (i.e. over 1 year) the number of agents who exit the market tends to be stable but different among the three scenarios: Existing Conditions shows the highest number of farmers exiting the market, whereas scenario 3, Airport Improvement (intervention) has the smallest number of farmers exiting the market. We interpret this as the possibility that the improvement of infrastructure can indeed provide economic relief to the supply chain and may be seen as a feasible solution. Therefore, in our comparison of the two transport interventions we can observe that scenario 3, Airport Improvement, appears to be more effective in terms of market exit rate than scenario 2, Improvement of Road Infrastructure (road intervention) in the seven districts. Not only is this result in line with the various studies arguing that Uganda should not focus on road improvements of the complete national network but should instead invest in other transport modes as well, to achieve a more affordable and effective investment policy, but it also dovetails with the present intervention of the Uganda government and CAA Uganda. In January 2014, the debate in Uganda on on the construction of new airports tilted in favour of expanding Entebbe International Airport instead (The East African 2014). However, already by October 2013 the Minister of Transport had announced the expansion of the airport with the acquisition of 66 ha of land dedicated to the construction of cargo facilities, aprons, runways, fuel farm, and passenger terminal facilities. The African Development Bank is supporting the project by having approved the disbursement of US$27.1 million. In general we can observe that a combination of transport and economic interventions are most likely to enhance the opportunities that trade and export offer to growth. Therefore, to merely tackle problems associated with agriculture trade in Uganda from the perspective of decreasing transport costs and achieving the 100% RAI level is seen as a reductionist approach. But most importantly, these types of policy may not achieve the given objectives of yielding positive impacts to the farmers in terms of production, price dispersion and market distortion and (ultimately) poverty reduction, because these policies do not address the underlying structure of the agriculture sector in Uganda within which 80% of Uganda’s population receives its main source of income from subsistence agriculture. We can conclude by observing that transport infrastructure improvements are a fundamental investment in the challenge to decrease transport costs and foster trade

180

11 Transport Cost and Infrastructure Investments

and economic growth. According to our analysis, transport interventions may produce long-term impacts under the present structure of the agriculture sector in Uganda. However, an integrated policy intervention to tackle the different facets of transport as well as the logistics chain may be more advisable. The improvement of Entebbe International Airport will lead to stronger support for the farmers/ outgrowers, who will subsequently gain higher liquidity in the market and therefore be less likely to fall into subsistence production (loss of agriculture surplus/market exit). In Uganda, waterways and railways are also used as transport modes in the flow of trade; it is possible in future to extend our analysis to evaluate the opportunities for improvement in the railway systems and waterway networks and thereafter compare the different initiatives. Any transport intervention should nevertheless be carefully designed to promote the most effective and affordable ways to develop trade. By doing so, the interventions may achieve greater political consensus and thereby become integral to the Uganda National Vision 2040 Strategy. In Chap. 12 we will examine in detail the logistics flows of the movement of agriculture products in Uganda, and identify possible strategies for improving trade.

References ADC (1999) Competitive handling of fresh fruits, vegetables and flowers at Entebbe Airport, Uganda’s investment in developing export agriculture (IDEA) project. Paper presented for the Horticultural Association of Uganda, the Uganda Flowers Association and the Civil Aviation Authority, USAID Funded Project African Infrastructure Country Diagnostic (AICD) (2011) Uganda’s infrastructure: a continental perspective. World Bank, Washington, DC Barrett C (2008) Poverty traps and resource dynamics in smallholder agrarian systems. Springer 25(2):17–40 Blyde J, Iberti GA (2014) A better pathway to export: how the quality of road infrastructure affects export performance. Int Trade J 28(1):3–22 Carruthers R, Krishnamam RR, Murray S (2008) Improving connectivity: investing in transport infrastructure in Sub-Saharan Africa. AICD Background Paper. World Bank, Washington, DC Civil Aviation Authority Uganda (CAA) (2013) Aviation Forum 3(2):3–23. Kampala Clark C, Rosenzweig W, Long D, Olsen S (2004) Double bottom line project report: assessing social impact in double bottom line ventures. Methods Catalog. Community Foundations of Canada Collinson C, Kleih U, Burnett D, Muganga A, Jagwe J, Ferris RSB (2005) Transaction cost analysis for selected crops with export potential in Uganda. Prepared for the Plan for the Modernization of Agriculture by the National Resources Institute, UK and the International Institute for Tropical Agriculture, Nigeria. ASARECA/IITA Monograph 6, IITA, Ibadan, Nigeria Entebbe Handling Services (ENHAS) (2013) The global compact, Entebbe handling services. Activities and engagements. ENHAS, Entebbe Fan S (2004) Infrastructure and pro-poor growth. Paper prepared for the OECD DACT POVNET. Helsinki, 17–18 June, 2004

References

181

Gachassin M, Najman B, Raballand G (2010) Roads impact on poverty reduction: a Cameroon case study. World Bank Policy Research Working Paper, no. 5209. World Bank, Washington, DC Gollin D, Rogerson R (2010) Agriculture, roads and economic development in Uganda. NBER Working Paper, no. 15863. National Bureau of Economic Research Hodges RJ, Buzby JC, Bennett B (2011) Post-harvest losses and waste in developed and less developed countries: opportunities to improve resource use. J Agric Sci 149:37–45 Hummels D (2001) Time as a trade barrier. GTAP Working Paper, no.18 Lall SV, Shalizi Z, Deichmann U (2004) Agglomeration economies and productivity in Indian industry. J Dev Econ 73(2):643–673 Limao N, Venables AJ (2001) Infrastructure, geographical disadvantage, transport costs and trade. World Bank Econ Rev 15:451–479 Natamba B, Mangeni P, Nakabuye Z, Brendah A, Agasha E (2013) Transaction costs and outreach of microfinance institutions in Uganda. Bus Manag Econ 1(6):125–132 Raballand G, Macchi P, Merotto D, Petracco C (2009) Revising the roads investment strategy in rural areas: an application for Uganda. World Bank Policy Research Working Paper, no. 5036. Social Science Research Network, Washington, DC Ranganathan R, Foster V (2012) Uganda’s infrastructure: a continental perspective. World Bank Policy Research Working Paper, no. 5963. Africa Region Sustainable Development Department, Washington, DC Rudaheranwa N (2009) Trade policy and transport costs in Uganda. CREDIT Research Paper. Centre for Research in Economic Development and International Trade, University of Nottingham, UK Svensson J, Yanaizawa D (2009) Getting prices right: the impact of the market information service in Uganda. J Eur Econ Assoc 7(2–3):435–445 The East African (2014) Editorial. 1 Feb 2014 Uganda National Household Survey (UNHS) (2013) Uganda Bureau of Statistics. UNHS, Kampala UNCTAD (2004) An investment guide to Uganda: opportunities and conditions. International Chamber of Commerce, The World Business Organization, United Nations, Geneva Venables AJ (1999) But why does geography matter, and which geography matters? Int Reg Sci Rev 22(2):238–241 Walle van de D (2002) Choosing rural road investments to help reduce poverty. World Dev 30(4): 152–178 World Bank (2001) Trade policies, developing countries and globalization. World Bank, Washington, DC

Chapter 12

Consolidation and Reform: Towards a Collaborative Approach

12.1

Introduction

We have seen in the previous chapter how high transport costs and lack of transport infrastructures act as impediments to trade, and how these conditions largely shape the structure of the agriculture sector and negatively impact on trade flows in Uganda. There is consensus among experts and scholars (USAID 2008) that improving access to markets and to trade can lead to a shift away from a supplybased situation of the agriculture sector where farmers/outgrowers merely sell their crop surplus, to a demand-driven agriculture where farmers produce for the markets. However, this shift is complex and beset by numerous constraints that still prevent Uganda’s farmers from fully participating in the market. One main problem is that the non-traditional export agriculture goods market is still very patchy and, depending on the product, it can fluctuate significantly in terms of pricing, production volume, export volume, and export values (Uganda Bureau of Statistics 2013). Agriculture products are highly vulnerable to weather conditions and to changes in the demand, and concerted efforts are therefore necessary to increase critical mass in production, to develop efficient supply chains, and to establish market quality and standards in international markets. The specific example of vanilla can shed some light on the difficulties currently existing in the market. In 2003, 2004 and 2005 cyclones impacted heavily on the production of the world’s main and largest producer and supplier of vanilla, Madagascar. An immediate upsurge was observed in the demand for vanilla along with a consequent dramatic rise in price. We have observed that Uganda has the right climatic conditions for achieving two vanilla harvests per year. Moreover, given the high level of unemployment in the unskilled labour force, labour-intensive pollination of the vanilla plants, which must be done by hand, could have been accomplished at a low labour cost. Uganda was certainly in pole position for capturing a large share of the global market. But in aiming to take hold of the market following the shocks, © Springer International Publishing AG 2017 F.R. Medda et al., Collaborative Approach to Trade, Advances in Spatial Science, DOI 10.1007/978-3-319-47039-9_12

183

184

12 Consolidation and Reform: Towards a Collaborative Approach

and thus in a rush for quick profit, many Uganda farmers and traders flooded the market with unripe, green and ultimately inferior, vanilla. In effect, this get-richquick strategy did not bode well for the reputation of Uganda as a major supplier of high quality vanilla. Nevertheless, although vanilla is subjected to price shocks, it is regarded by many experts (Komarek 2010; Nyapendi 2009) and Uganda farmers themselves, who call vanilla the green gold, as a good crop towards achieving diversification in the Uganda agriculture market and for boosting economic growth. In order to draw from the lesson above, the objective of our analysis in this chapter is to examine how to limit the impacts of production fluctuations and price shocks of agriculture crops through the development of consolidation and logistics policies. One of the major barriers to trade growth is high transaction costs; the potential gains from consolidation are therefore accrued through the reduction of transaction costs by increasing economies of scale and competitiveness. In this chapter, as in several others, we examine three scenarios. Scenario 1 describes the Existing Conditions and identifies the model for the optimization of the warehouse locations to be implemented in the ABM. In scenario 2 we consider the implementation of the consolidation of the supply chain; and finally in scenario 3 we examine the introduction of regulation and reform in the logistics market. The three scenarios should not be construed as three different, albeit comparable policies, but instead should be seen as a possible stepwise policy strategy of reform of the logistics supply chain in Uganda.

12.2

Scenario 1: Existing Conditions

Gollin and Rogerson (2010) find that the low levels of physical infrastructure (energy, water, transport) and services intended to foster trade can be cited as the main culprits for the high transaction costs present in Uganda’s agriculture sector. High transaction costs are certainly one of the causes of the significant dispersion of crop prices. In fact, the price of crops not only has a large spatial variation, but price is also observable within the value-added chain, as we noticed from our Uganda data. The spatial variation of price is observable from region to region, and above all from farm gate price to urban wholesale price. Urban wholesale price is considered to be one of the underlying reasons for the high population density living in rural areas of Uganda, and a reflection of the condition of low urbanisation. The price of agriculture products also has high variations vertically, where we observe low prices for bulk products which reach very high prices as processed and manufactured agriculture products (an interesting example is given by the cassava or matooke price variations from bulk price to dried products; the highest price variation, and thus added value, is achieved with the production of beer from cassava or matooke). The map in Fig. 12.1 shows the seven districts of our study. For many crops, the logistics costs (movements, loading and unloading, storage, bagging, packaging,

12.2

Scenario 1: Existing Conditions

185

Fig. 12.1 Map showing the districts of our analysis in dark colour: Luwero, Mpigi, Masaka, Iganga, Mitiyana, Kamuli, and Mukono

controls, etc.) can surpass the farm gate price by a large margin. Raballand et al. (2009) argue that logistics services suffer from significant market imperfections as well as collusion, and even though many logistics firms present in Uganda have low operational costs, they are nevertheless able to obtain very substantial mark-ups for their services. We interpret this as a signal of possible cartelisation in transport logistics. In this context Romanik (2007) demonstrates that logistics transport services operate within a monopolistic regime with significant marketing margins. Our next step is to analyse the logistics transport optimization. We assume that the agriculture trader needs to select the best shipping route between an origin and a destination to minimise his transport cost while also satisfying demand. We can formulate the problem as a constrained linear optimization in the following form Minimise cT x,

subject to Ax ¼ b; x  0,

ð12:1Þ

where x represents the routing strategy, b the demand, x the supply, and c the transport costs.

186

12 Consolidation and Reform: Towards a Collaborative Approach

We are aware that any linear optimization problem, as defined above, has a Luenberge and Yinye dual formulation given by Maximise γ T b,

subject to γ T A  cT ; γ  0

ð12:2Þ

In order to keep transport costs low, and in so doing ship agriculture product more efficiently, the trader will offer to buy the farmer’s crop at the farm gate (origin) and then transport it to warehouses (destinations). In our case, γ represents the warehouses that the trader should use. The product price to be used in these transactions varies from point to point and the trader decides in advance. The trader must choose these prices at each warehouse destination point in order for his offer to be acceptable to the farmer. The dual formulation case is suitable for an agent based model adaptation. In our model the trader chooses the most suitable price and tries to maximise his profit. However, the trader offers a price that also includes shipment to a local warehouse and the effective price offered to the farmer strongly depends on the location of the warehouses. Therefore, having a warehouse in the best position will affect the average transport costs to the farmers, especially if they have to pay these costs. In our ABM simulations the trader decides whether the farmer must pay for transport in relation to his (the trader’s) own budget. But the cost of transport will significantly affect the farmer’s decision on whether or not to sell his crop. Our simulation findings show that the number of farmers/outgrowers who do decide to sell to a trader—who has chosen a specific warehouse—is statistically relevant, even if transport costs for shipment of crops vary between traders. Therefore, in order to account for the variations observed in pricing, we calculate the average number of warehouses over simulations by keeping the positioning of the warehouses fixed (Fig. 12.2). It is noteworthy that, in scenario 2, the approach we apply in relation to warehouses, differs slightly from the simulations where we obtain effects of other policies. Since detailed information on the positioning of existing warehouses was not available, we have run each simulation by fixing the number of warehouses; they are situated in relation to the density of farmer locations. Figure 12.3 illustrates the context of our main agents in our seven districts with farmers depicted in red, warehouses in green, and traders in dark green. However, as we will see in the next section on consolidation, in order to obtain information on the best locations for warehouses, it has been necessary to change our simulation strategy to account for a possible increase in the number of warehouses.

12.3

Scenario 2: Consolidation Around Ruralpolis

187

Fig. 12.2 Map of the control points indicating warehouse locations in the considered districts: Luwero, Mpigi, Masaka, Iganga, Mitiyana, Kamuli, and Mukono

12.3

Scenario 2: Consolidation Around Ruralpolis

According to Raballand et al. (2009), transport by truck per kilometre in Uganda is 10 times more expensive than movement by bicycle or motorcycle. Farmers produce on average between 40 kg to 3 tonnes of agriculture goods per year and transport their product mainly by bicycle or motorcycle. Transport by truck only becomes more effective and profitable if the consolidation of products is implemented to reach a critical load, i.e. at least 500 kg of product per trip, which is transported no less than 50 km (DFID 2014). At the current production level a consolidation centre (warehouse) should be able to collect the production of at least 600 farmers. The three encircled areas in Fig. 12.3 of our ABM-generated consolidation of warehouses can be defined as a Ruralpolis, since we have in these areas high population density where agriculture is the main economic activity. The large majority of farmers/outgrowers work as independent farmers and the land is broken

188

12 Consolidation and Reform: Towards a Collaborative Approach

Fig. 12.3 Study areas depicting ABM-generated locations of farmers (red), warehouses (green) and traders (dark green)

down into holdings of approximately 3 acres. Some farmers also store their production for short periods either in their own homes or in old storage areas that are generally in poor condition; when this is the case, farmers tend to sell their product fairly quickly so as not to incur losses due to deterioration. An advanced warehouse system with improved facilities can therefore help to resolve the problem of post-harvest handling wherever storage is deemed to be an effective improvement. In our analysis using agent based modelling we assert that consolidation within a supply chain can reduce unit produce costs. The creation of logistics poles that act as hubs for the flows of goods around the three areas is supported by the idea that agglomeration and consolidation of activities can accrue higher traffic, better use of trucks and large vehicles for main routes, a reduction in the cost of operations

12.3

Scenario 2: Consolidation Around Ruralpolis

189

(control, standard tests, packaging, etc.), and fewer post-harvest losses. In order to identify the locations and number of warehouse that optimize the logistics transport cost, we assume that the farmer minimises his total transport cost. However, we need to mention that the farmer, as simulated in the ABM model, makes his decision according to the decision tree described in Chaps. 9 and 10, and to the maximisation of his revenue. Since every farmer is trying to optimize his revenue, as such, the use probability reflects the optimality of the location of the warehouse. During the simulation, each farmer chooses an offer from a trader which includes the cost of shipping to a local warehouse. In the Monte Carlo simulation we change both the location of the warehouses (we restrict the analysis to locations in the vicinity of farmers) and the location of farmers, while keeping fixed the location of the airport. The decision-making of the farmer formula can be expressed as follows Di ¼ Ri  Ci ,

ð12:3Þ

where Ri is the amount offered to the farmer for the product. The cost term can be written as Ci ¼ Č i þ CW i ,

ð12:4Þ

where CWi is the cost of shipping to the warehouse. Thus, we can now see that maxDi ¼ maxðRi  Ci Þ ¼ maxðRi Þ  minðCi Þ:

ð12:5Þ

The minimisation of the shipping costs averaged in the Monte Carlo simulation selects the locations where the warehouses serve as many farmers as possible. A helpful device for farmer inclusion is to calculate the usage probability, i.e. how many times a warehouse is used, by performing many simulations, changing the conditions and averaging over them. This is equivalent to the minimisation of the costs using a Monte Carlo technique by fixing the location of the warehouses while also changing the environment in each simulation (location of each farmer and trader, prices, etc.). This allows only for certain effects to arise which are truly due to locations within the district; in this way other peripheral districts do not exert influence. A comparable result would be reached if we were to simulate an infinite number of agents, but since this is not a feasible option we are able to draw some conclusions from results obtained after the Monte Carlo simulations. We have pre-selected approximately 40 control warehouse locations in accordance with the geographical location of the seven districts of our study shown in Fig. 12.4. In the simulation farmers are then positioned around these initial locations within a distance of 10 km. We calculate the average probability of a warehouse to be used by considering the average use frequency and dividing it by the total number of times a consolidation has occurred. Overall, we simulate a

190

12 Consolidation and Reform: Towards a Collaborative Approach

Fig. 12.4 Configured positions of 40 warehouses from the Existing Conditions scenario

6-month period, averaged over 30 simulation runs for each location of the farmers, and 10 times for each location of the warehouse. The results shown in Fig. 12.4 are based on 300+ simulation runs. In our calculation, W is is the number of times warehouse i is used in simulation s. From Fig. 12.4 we plot the quantity Pn Ws fi ¼ P n s¼1 Pn i W i¼1

s¼1

W is

,

ð12:6Þ

Pn fi ¼ 1 and W where n represents the number of simulations. We observe that i¼1 fi can be interpreted as the warehouse probability of being used. thus W The results of the warehouse simulations are depicted graphically in Fig. 12.5, in which we plot the averaged and normalised number of times each of the 40 warehouses is used. We have averaged the results over 30 simulation runs.

12.3

Scenario 2: Consolidation Around Ruralpolis

191

Probability of Warehouse to be used 0.07

W4

W5

0.06 W2 W1

Use probability

0.05

W3

0.04

0.03

0.02

0.01

0 0

5

10

15

20

25

30

35

40

Warehouse identification number

Fig. 12.5 Warehouse use probability of 40 warehouses over a simulation period of 6 months. We show the five warehouses (W1–5) with the highest probability of being used

In our simulations for scenario 2 we reach the result that, in order to optimize the use of the warehouses in the three given areas and thus increase farmers’ revenue, five warehouses are needed as consolidation centres, as depicted in Fig. 12.6. This result is in line with other studies, such as Taaffe et al. (1996), who demonstrate with a few well-localised warehouses, the possibility to reduce transport costs and promote economies of scale. Given that the movement of products and the distribution of freight are interrelated, we would maintain that the choice of the five locations for warehouses shown in the figure is able to achieve the best accessibility to and from origin and destination in our case study supply chain. This solution is certainly the best fit in regard to attracting new investment and business, since firms will move to areas where there is competitive advantage in terms of resources and established activity (Porter 2002). The solution proposed here fosters the consolidation of agriculture production not only by exploiting economies of scale but also by taking the opportunity to strengthen intermodal solutions as part of an integrated and efficient transport system in the supply chain.

192

12 Consolidation and Reform: Towards a Collaborative Approach

Fig. 12.6 Five optimal positions of warehouses obtained from the model

We are able to observe another interesting impact of the consolidation strategy associated with the development and growth of the Ruralpolis. In these areas ‘family farm’ and ‘family business’ will spend their profits in the local community and contribute largely to the welfare of the rural community as a whole. Let us next discuss our final scenario 3, which examines a 10% reduction in transport costs due to the implementation of reform(s) in the logistics market.

12.4

Scenario 3: Reform of the Logistics Market

We have discussed the transport cost in supply chains as one of the major costs. In the case of the Uganda market, transport costs have risen sharply in recent years. Our analysis of the survey reports a doubling of logistics costs compared to year

12.5

Concluding Remarks and Policy Recommendations

193

2008. This stark rise impacts most strongly on farmers/outgrowers, since they must often pay up front to transport their goods. Wide price distortions in the Uganda agriculture market relate largely to the logistics costs, and although the market is contestable, limited access to the market is still the case for farmers. In the meantime, adjustment costs have pushed logistics companies to behave like a cartel. Logistics companies often use their market power to set prices through the use of a mark-up strategy. According to Lall et al. (2004), as France deregulated its logistics markets between 1978 and 1998, the government witnessed a 33% decrease in trucking costs. The reduction of logistics costs through deregulation of the market, or cap pricing, makes conditions ripe for exploiting economies of scale and increasing efficiencies in production. In scenario 3 we impose a cap on the logistics transport price and thus reduce the transport price by 10%. We then simulate the model and the results are expressed in relation to the number of subsistence farmers who will exit the market (hereafter referred to as inactive agents). The model certainly tests other percentage changes of the transport cost, but in this case we take a conservative approach and assume the introduction of a regulatory regime that can impose a 10% reduction of the logistics transport cost. Using our agent based model we have been able to observe the impact of a 10% transport cost reduction. We compared the number of inactive farmers/outgrowers when we imposed the cap on transport cost equal to the cost reduction of 10% against the Existing Conditions scenario, as shown in Fig. 12.7. Most significantly, we found that the 10% transport cost reduction has a direct effect on the number of inactive agents; this effect is equal to a net reduction of farmers exiting the market above 10% in relation to the number of farmers who will exit the market in the Existing Conditions scenario. This finding clearly shows the importance of a regulatory regime that supports reform as a necessary mechanism in the Uganda logistics supply chain towards fostering trade growth, and ultimately protecting the agents most exposed to risk and cartelisation of transport: subsistence farmers.

12.5

Concluding Remarks and Policy Recommendations

In this chapter we have demonstrated how the agent based (AB) model can analyse policy alternatives in relation to the logistics of the agriculture trade supply chain. First we examined the introduction of a logistics consolidation policy of a warehouse system. The results showed that in order to achieve a decrease in farmers leaving the market and hence to facilitate trade and improve production efficiency, five main warehouses, rather than a large number of warehouses (as is presently the case) should be situated in the three cluster areas identified in Fig. 12.3 as Ruralpolis. By locating a warehouse in each Ruralpolis, the warehouse effectively becomes the focal point for consolidation and traffic of trade, and thus leverages

194

12 Consolidation and Reform: Towards a Collaborative Approach 350 Existing Conditions Reduced Transport Costs

300

Inactive Agents

250 200 150

100 50 0 0

50

100

150

200

250

300

350

Days

Fig. 12.7 Transport logistics cost impact on farmer inactivity

market risk, for example, in relation to post-harvest losses or the management of sustainable volumes of standardised products. A second policy intervention relates to the regulatory intervention in the logistics services. In this case, the AB model has shown that, by introducing a cap price system for the transport cost, we can achieve a lower rate of farmers leaving the market and as a result, increase liquidity in farmers’ markets. In support of these two different but integrated policy recommendations we can also acknowledge that, already by year 2000, the Uganda Ministry of Trade and Tourism had established a programme based on four points, and above all has encouraged the implementation of a private sector-run warehouse system and the establishment of a collateralised warehouse receipt system inventory. In relation to the latter point, certainly a warehouse receipt scheme can foster consolidation and help farmers raise enough cash and receive adequate credit when needed. However, in summing up we need to emphasise that, in order to be effective, our recommendations and various schemes and policies must always be supported by Uganda’s government institutions.

References Department for International Development UK (DFID) (2014) Operational Plan: 2011–2016 DFID Uganda. Department for International Development, London Gollin D, Rogerson R (2010) Agriculture, roads and economic development in Uganda. NBER Working Paper, no. 15863. National Bureau of Economic Research

References

195

Komarek AK (2010) Crop diversification decisions: the case of vanilla in Uganda. Q J Int Agric 49 (3):227–242 Lall SV, Shalizi Z, Deichmann U (2004) Agglomeration economies and productivity in Indian industry. J Dev Econ 73(2):643–673 Nyapendi M (2009) Uganda: vanilla gate prices hit by international speculation. http://allafrica. com/stories/200901080066.html Porter G (2002) Living in a walking world: rural mobility and social equity issues in Sub-Saharan Africa. World Dev 30(2):285–300 Raballand G, Macchi P, Merotto D, Petracco C (2009) Revising the roads investment strategy in rural areas: an application for Uganda. World Bank Policy Research Working Paper, no. 5036. Social Science Research Network, Washington, DC Romanik C (2007) An urban rural focus on food markets in Africa. The Urban Institute, Washington, DC Taaffe E, Gauthier H, O’Kelly M (1996) Geography of transportation. Prentice-Hall, Englewood Cliffs, NJ Uganda National Household Survey (UNHS) (2013) Uganda Bureau of Statistics. Kampala USAID (2008) Uganda Agricultural Productivity Enhancement Program (APEP). United States Agency for International Development

Chapter 13

Information Exchange: Collaboration and Coordination Standards

13.1

Introduction

Crop price dispersion across farmers/outgrowers is one of the main problems arising when we examine price asymmetric information in the agriculture sector in Uganda (Gollin and Rogerson 2010; Calderon 2009; Minten and Stifel 2008; Mandl and Mukhebi 2002). Farm gate prices are often 50% less than the urban wholesale price, and a large portion of this increment is comprised by transport cost. However, it is precisely transport cost reduction and logistics chain consolidation which are two of the most important initiatives in the goal to reduce price asymmetry; and we would assert that the core of the problem—lack of information—needs to be tackled. Farmers and traders often have imperfect or no knowledge whatsoever of their market positions in relation to production completion, demand, standards, best practice, and exposure to risk. Moreover, the agriculture sector is heavily subjected to price and weather variations, and since farmers/ outgrowers are significantly affected by the lack of information, they would prefer to reduce risk wherever possible. But as they reduce risk, so do farmers/outgrowers also shy away from new technologies and instead remain within the production segment of unprocessed products, which are generally bulky and of low value. Therefore, in this chapter our objective is to address the hypothesis that, if coordination were to take place among farmers/outgrowers and agriculture standards (GAP) were to be introduced, these actions would generate effective benefits to farmers/outgrowers, the agents with the highest risk exposure in the supply chain (hypothesis 3). In Chap. 9 we described the three channels where we examined coordination in our agent based model, and we also introduced in Chap. 10 the multiplier attachment network. The multiplier attachment algorithm takes into consideration the capacity of farmers/outgrowers to exchange information between themselves, but also to venture beyond the boundaries of neighbourhood and village. The multiplier attachment network is applied in our Uganda agriculture case to account for the exchange of information developed through information and © Springer International Publishing AG 2017 F.R. Medda et al., Collaborative Approach to Trade, Advances in Spatial Science, DOI 10.1007/978-3-319-47039-9_13

197

198

13

Information Exchange: Collaboration and Coordination Standards

20 18 16 14 12

60

10 8 6 4

30

50 40

20

Tele density

Subscriptions in Millions

communication technologies (ICT), e.g., social media such as the mobile telephone network. The benefits of social media in the agriculture sector have been studied widely (World Bank 2011) and ICT technologies have been applied in numerous cases with great success in agriculture (Venables 2001). It is particularly interesting in our context how, across continents and time (nearly 20 years), ICT has proved to be an effective tool for: reducing transport and logistics costs with just-in-time information on consolidation and market trends; improving yield and quality of products through information about weather forecasting and new agriculture practices; increasing productivity and access to markets and value creation through risk reduction and by supporting private sector interventions; fostering financial systems for farmers, such as microfinance, facilities for insurance and credit; and last but not least, for improving food standards and safety through traceability and product monitoring. Uganda’s ICT sector in particular has been growing quickly, and according to the Uganda Communication Commission, the number of telephone subscribers in March 2009 was 10 million, of which 9.8 million were mobile phone users. Uganda launched its third-generation mobile broadband service in March 2008, boosted over 2500 km of fibre-optic network. The Uganda government has also established the Rural Communication Development Fund (RCDF) in order to further promote social media. In Figs. 13.1 and 13.2 the ICT development in Uganda shows rapid growth in subscriptions over a 10-year period to 2012. In the next sections we study how coordination through access to information can support farmers/outgrowers in order to capitalise on economies of scale and agglomeration economies. From this perspective, Grabowski (2012) examines how the lack of incentive and information can hinder investment in agriculture and thus development. In fact, as previously observed, the risk posed to farmers/outgrowers for investing more in their production and on new crops or new technologies is directly related to insufficient knowledge of the market and ultimately to the fear of

10

2 -

0 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Subscribers

2012

Tele density

Fig. 13.1 Subscriber and tele-density data, 2002–2012. Source: Uganda Communication Commission (2014)

Scenario 1: Existing Conditions

199 120

40 35 30

100 80

25 20

60

15

40

10 5

20 -

Dec–10

Jun–11

Dec–11

Jun–12

Dec–10

Jun–11

Dec–11

Mobile internet subscriptions

610,000

850,200

977,500

Fixed internet subscriptions

35,000

84,558

88,786

Dec–12 Jun–12

Fixed internet Thousands

Mobile internet x 100000

13.2

Jun–13 Dec–12

Jun–13

1,586,325 2,692,705 3,458,351 92,934

96,000

95,434

Fig. 13.2 Fixed and mobile internet subscriptions, June 2013. Source: Uganda Communication Commission (2014)

not finding a buyer (Kydd and Dorward 2004; Kindness and Gordon 2001). An instructive example of the obstacles to investment encountered by farmers/ outgrowers corresponds directly to farming practices. The fee a farmer must pay in order to obtain formal certification to farm organically in Uganda can cost up to US$1500 (Booth and Golooba-Mutebi 2014). Such a high fee for a knowledge investment makes this option virtually impossible for small farmers/outgrowers. Against this background, we next briefly discuss Existing Conditions (without coordination) in scenario 1. Thereafter in Sect. 13.3, we consider scenario 2 where the introduction of farming associations and coordination activities takes place. Following on in Sect. 13.4, we study the implementation of standards in the supply chain in scenario 3 in order to improve growth. In Sect. 13.5 we reach conclusions and offer some policy recommendations.

13.2

Scenario 1: Existing Conditions

In scenario 1, Existing Conditions, we assume that no coordination is taking place. In the agent based model, farmers/outgrowers are myopic agents or bounded rational agents, and they make decisions as a function of the information available to them. When we consider association through multiplier attachment, we assume that each pair of farmers not in debt, i.e. who are financially solvent, start to coordinate with each other: they fix the price of their products, share the resources for paying the shipping cost, and pass information to one another. We assume that association is realised between farmers no farther away than 30 km from each other in order to maintain the concept of Ruralpolis and thus agglomeration economies; within this distance moreover, we can keep the direct relation between distance and transport cost. The distance parameter under the ABM may be changed however in

200

13

Information Exchange: Collaboration and Coordination Standards

order to verify the sensitivity of a specific policy. Each farmer may have at the start a network of farmers/outgrowers known to him and this network is then shared every time a new multiplier attachment/coordination with a new farmer/outgrower is developed. We set the model so that sharing of the network of known farmers/ outgrowers occurs within one calendar year; thus, the spread of information is over 1 year to allow for adjustment and clearing of noise. Since the association process is time-consuming from the computational point of view, we assume that the coordination process takes place every 3 months.

13.3

Scenario 2: Farmer Coordination-Collaboration Strategies

There are already many producer associations present in Uganda, including Upland Rice Millers, Kabarole Integrated Women’s Effort in Development, Buy-ambe Saving, and Credit Cooperative Society (SACCO) among others, and the USAID Uganda Agriculture Productivity Enhancement Program (2008) has developed a successful network of farmers/outgrowers based on the concepts of farming activity coordination and adding value to the agriculture chain. We take a similar approach. In our context coordination initiatives/association among farmers/outgrowers— through the spread of information—aims to lessen the long chain of transactions and address the lack of competition in the agriculture sector in Uganda. In our study we consider any type of association (formal and informal) that fosters coordination1 among farmers/outgrowers based on the exchange of information. We have in scenario 1 established the assumptions for the multiplier attachment of the coordination among farmers/outgrowers; next, by using the agent based model, we examine four different impacts: (1) farmers/outgrowers leaving the market; (2) transport cost; (3) crop price; and (4) volume sold to traders, all of which relate to increased association. At regular time steps we ‘allow’ farmers to associate. For two farmers this implies that they • share information; • produce and sell together at the same price; • consolidate and transport together in the same warehouse and at the same logistics cost; • share revenues in proportion to the amount produced. We introduce a control parameter into the model which monitors how farmers associate. At each 3 months in the model, we allow the farmers to consider association. Farmers who satisfy the conditions described above can associate and begin a partnership with probability p: we fix p ¼ 0.1. We thus compare the 1

In this analysis the terms association and coordination may be used interchangeably for the same activity.

13.3

Scenario 2: Farmer Coordination-Collaboration Strategies

201

Number of bankrupt agents as function of time

300 Association No Association

250

Inactive Agents

200

150

Association shock

100

50

0

0

50

100

150

200

250

300

350

Days

Fig. 13.3 Number of farmers/agents exiting the market in relation to Farmer Association and No Farmer Association

two simulations, with p ¼ 0.1 (farmers can associate) against p ¼ 0 (farmers do not associate). Each plot is obtained after a Monte Carlo simulation, with an average of 20+ simulation runs. In Fig. 13.3 we show the number of farmers/agents who exit the market under the two scenarios. In scenario 1, Existing Conditions, the farmers have no form of coordination and in scenario 2, the farmers enjoy association. We consider the trends across a 1-year time span in both scenarios. In the case of no association, approximately one-half of the agents leave the market (approximately 280 agents of the 550 in total). However, when the farmers associate, the value (number) of inactive agents decreases sharply (to approximately 70 agents of the total of 550). This result suggests that association among farmers is highly significant, particularly over the long-term. In Fig. 13.4 below we consider the total transport cost in scenarios 1 and 2. As we can observe in Fig. 13.4, in the case of the transport cost, the curve related to Association among farmers/outgrowers lies conspicuously below the curve representing the case of No Association. The third impact we examine in relation to farmers/outgrowers association is crop price at farm gate in Figs. 13.5, 13.6, 13.7, 13.8 and 13.9. In the horizontal axis we represent the time lag as equal to 1 year and in the vertical axis the average price is shown for the considered products: hot peppers, chillies, matooke, okra, and sweet potatoes. In the study of this impact, and due to the lack of data on prices, the results must be interpreted as the likely trend were ‘Association’ to be implemented

202

13

12

Information Exchange: Collaboration and Coordination Standards Transport Costs

x 106

Association No Association

Total Transport Costs Ugsh

10

8

6

4

2

0

0

50

100

150

200 Days

250

300

350

Fig. 13.4 Total transport cost—Association vs. No Association

Average Hot Pepper

5 4 x 10

No Association Association

3.5

Ugsh /tons

3

2.5

2

1.5

1

0

50

100

150

200

250

300

350

Days

Fig. 13.5 Average farm gate prices for Hot Peppers—Association vs. No Association

13.3

Scenario 2: Farmer Coordination-Collaboration Strategies Average Prices Chillies

x 105

2.8

203

No Association Association

2.6 2.4

Ugsh/tons

2.2 2 1.8 1.6 1.4 1.2 1 0.8

0

50

100

150

200

250

300

350

Days

Fig. 13.6 Average farm gate prices for Chillies—Association vs. No Association

6

Average Prices Matooke

x 105 No Association Association

5.5

Ugsh/tons

5

4.5

4

3.5

3

2.5

0

50

100

150

200

250

300

Days

Fig. 13.7 Average farm gate prices for Matooke—Association vs. No Association

350

204

13

2.8

Information Exchange: Collaboration and Coordination Standards Average Price Okra

x 105 No Association Association

2.6

Ugsh/tons

2.4

2.2

2

1.8

1.6

1.4

0

50

100

150

200

250

300

350

Days

Fig. 13.8 Average farm gate prices for Okra—Association vs. No Association

5

Average Price Sweet Potatoes

x 105 No Association Association

4.5

Ugsh / tons

4

3.5

3 2.5

2

1.5

0

50

100

150

200

250

300

350

Days

Fig. 13.9 Average farm gate prices for Sweet Potatoes—Association vs. No Association

13.3

Scenario 2: Farmer Coordination-Collaboration Strategies

205

rather than the actual price of the crops. The results indicate that the prices farmers can receive are always higher under Association than prices of crops with No Association. For any crop, the supply chain is composed of many intermediaries, each taking a margin at every stage of the chain. However, when farmers/ outgrowers form an association, they are able to coordinate their actions and avoid some of the intermediary stages by selling directly to the major wholesaler/ trader. The bell shape of the price function is due to interdependency between supply and demand of the product, and the highest price corresponds to price of the product in the market when farmers are able to avoid lower stage intermediaries. Therefore, this maximum price corresponds to the trader’s willingness to pay. In other words, the trader is ready to pay for the specific product at time t*, the highest price, due to the availability of the product in the market, which relates to the crop production time. After t*, the price decreases because the supply of the crop increases once again. Nevertheless, we can cautiously conclude with our observation that the price trend under farmer Association is always higher than with No Association. Also linked with the results on crop price is the fourth and final impact relating to number/volume of product sold to traders. In the crop price mechanisms depicted in Figs. 13.5–13.10 we can observe a reduction of product sold to the traders, a consequence of the reduction of intermediary traders in the supply chain when farmers/outgrowers form associations and sell and negotiate price in coordination. We notice that in the case where farmer Number of product sold to traders 30 No Association Association

Product sold to Traders

25

20

15

10

5

0

0

50

100

150

200

250

300

350

Days

Fig. 13.10 Amount of product sold to traders by farmers/outgrowers—Association vs. No Association

206

13

Information Exchange: Collaboration and Coordination Standards

associations are taking place, number of products sold to traders is, sensibly, less compared to the case where no association is allowed; this happens because farmers sell directly to exporters, and whenever crop supply to farmers decreases, prices start to rise. In the next discussion of scenario 3 on the spread of information, we examine an essential component of agriculture: standardisation and harmonisation of product quality.

13.4

Scenario 3: Standardisation and Harmonisation of Product Quality in the Supply Chain

We have observed in previous chapters how Uganda’s export market has risen steadily over the past two decades but that its non-traditional crop market is still poorly structured. Given Uganda’s high transaction costs, the non-traditional agriculture market also suffers insufficient implementation of standards and quality control mechanisms. This is a major hindrance to the development of trade, particularly for major international importers like the European Union Community. We have already mentioned that formal organic farming certification is beyond the reach of most farmers/outgrowers and would require significant investment; but as the vanilla market example has shown, in order for Uganda to foster growth in its agriculture sector, eradicate poverty and thus improve food security (FAO 2011), it is of utmost importance that the country become a reliable supplier of quality products. In this scenario we assume the implementation of GAP products in the supply chain. Good Agricultural Practices (GAP) “address environmental, economic and social sustainability for on-farm processes, and result in safe and quality food and non-food agricultural products” (FAO 2003). In our case the introduction of GAP is a reference policy tool used to verify how information can create greater diffusion of standards and technologies in conjunction with an overall increase in agriculture production. Using the agent based model we examine the effects of implementing GAP policies at the exporter level. In the model we distinguish between traders who apply GAP procedures and traders who do not. We assume for instance, that itinerant traders buy mixed products and as a result lose the traceability of the products. Itinerant traders will therefore not be able to sell to the exporters who require GAP standards. GAP policies are introduced in a top-down approach. Suddenly, the exporters no longer accept any product whose source is not known. As such, only the product not handled by the itinerant farmers is then exported. As a result, the prices of the itinerant traders, who are now losing the demand, must decrease, so they therefore cease to be competitive. However, the shock changes the equilibrium in the model. So we think it necessary to also analyse the effect on the average prices.

13.4

Scenario 3: Standardisation and Harmonisation of Product Quality in the. . .

207

300 GAP No Association and Gap Association

250

Inactive Agents

200

150

100

50

0

0

50

100

150

200

250

300

350

Days

Fig. 13.11 Number of inactive traders/agents in relation to the implementation of GAP policy

The introduction of GAP procedures, as we can observe in Figs. 13.11 and 13.12, has a positive impact in terms of decreased inactivity and reduction of transport cost. However, in relation to transport cost, scenarios without GAP (Existing Conditions, scenario 1) and with the introduction of GAP (Standardisation, scenario 3) show that the difference between the two scenarios is not as apparent as in the cases with and without Association. Nevertheless, the introduction of standards and controls is conducive to the spread of information and hence the spread of the benefits associated with GAP. We assume here that information is not only spread, but it is also operationalised across the districts and crops under consideration. In the comparison, the information and knowledge about procedures and standards (GAP) constitutes the effort to improve production quality and yield. The total transport costs in the case of GAP policy implementation are more similar than in the case with no policy, as depicted in Fig. 13.12. We can observe from the graph that over the long-run, transport cost under the GAP implementation is lower than the transport cost in the no GAP case. The implication here is that the GAP in general has a positive impact on the logistics systems towards a reduction in cartelisation and lower transport costs. We can now examine the impacts of the introduction of the GAP policy on the prices that farmers can receive for five products (Hot Peppers, Chillies, Matooke, Okra, and Sweet Potatoes). Also in this case we compare introduction of the GAP policy with status of no implementation of GAP (Figs. 13.13, 13.14, 13.15, 13.16 and 13.17). For all the products, farm gate prices over the long-run are always higher in the case of implementation of GAP than in the case of no implementation. We can

208

13

3

Information Exchange: Collaboration and Coordination Standards Total Transport Costs

x 107

GAP No GAP

1.22 Billions with GAP policy 1.55 Billions without GAP policy

2.5

Ugsh

2

1.5

1

0.5

0

0

50

100

150

200

250

300

350

400

Days

Fig. 13.12 Total transport costs in Uganda Shilling

1.55

Average Price Hot Pepper

x 105 GAP NO GAP

Ugsh / tons

1.5

1.45

1.4

1.35

1.3 0

50

100

150

200 Days

Fig. 13.13 Average farm gate prices for Hot Peppers

250

300

350

13.4

Scenario 3: Standardisation and Harmonisation of Product Quality in the. . .

10

209

Average Price Chillies

x 104

GAP No GAP

9.8 9.6

Ugsh / tons

9.4 9.2 9 8.8 8.6 8.4 8.2 8

0

50

100

150

200

250

300

350

250

300

350

Days

Fig. 13.14 Average farm gate prices for Chillies

2.35

Average Price Matooke

x 105 GAP No GAP

2.3 2.25

Ugsh / tons

2.2 2.15 2.1 2.05 2 1.95 1.9

0

50

100

150

200 Days

Fig. 13.15 Average farm gate prices for Matooke

210

13

2.35

Information Exchange: Collaboration and Coordination Standards Average Price Okra

x 105 GAP No GAP

2.3 2.25

Ugsh / tons

2.2 2.15 2.1 2.05 2 1.95 1.9 0

50

100

150

200

250

300

350

300

350

Days

Fig. 13.16 Average farm gate prices for Okra

2.35

Average Price Sweet Potatoes

x 105 GAP No GAP

2.3 2.25

Ugsh / tons

2.2 2.15 2.1 2.05 2 1.95 1.9

0

50

100

150

200 Days

Fig. 13.17 Average farm gate prices for Sweet Potatoes

250

13.5

Conclusions and Policy Recommendations

211

conclude this section with the observation that the effects of applying Good Agricultural Practices (GAP) policy indicate clear results which support the introduction of GAP to strengthen the real-life agriculture sector and thus improve trade.

13.5

Conclusions and Policy Recommendations

The organisation of agriculture production and trade is best examined as a dynamic structure where the different agents interact continuously with each other and where access to appropriate information and its diffusion has a fundamental role. The main aims are to shorten the chain of transaction, attend to the lack of competition, and improve the quality of crop production. According to Booth and GoloobaMutebi (2014), agriculture production in Africa is still patchy and remains in the hands of small farmers/outgrowers who operate mainly as subsistence farmers. Their observations have provided the springboard into this analysis, which has focussed on the application of the multiplier attachment between farmers and traders in order to spread information and by doing so, to facilitate association and GAP policy for agriculture production. Our results show that coordination among farmers and the establishment of GAP procedures can achieve better support in terms of lower transport costs and higher farm gate prices to the farmers who in general under both policies will be less likely to exit the market. The implementations of both policies can foster the development of a formal credit system, an important step in the stabilisation of the farmers’ cycle of production, and also lower risk for farmers. At present, formal banks do not lend easily to farmers because of the risks associated with agriculture in addition to banks’ collateral requirements. Credit is therefore mainly accrued from informal networks such as family and friends who may loan up to two-thirds of the capital requirements. However, by implementing coordination and farmer associations, as well as GAP as a control system for products, it is possible to advance from an informal system to a more formal system. Okurut et al. (2004) have demonstrated that co-operative and non-government organisations can shift credit away from family and friends by acquiring the 13% and 7% of market share, respectively. Furthermore, given the structure of consolidation systems around the three ruralpoli, it is also possible to envisage a Ruralpolis banking scheme along the lines of the programme established by FINCA (McIntosh 2008). Access to credit and formal local or farming lending institutions would need to be guaranteed by the Uganda government in order to leverage and reduce financial risk. As risk is reduced more opportunities are also likely to open up for public and private partnership involvement in Uganda’s agriculture sector. Access to credit is an important long-term strategy because it can, for example, improve production through crop diversification. By taking a proactive approach, farmers’ application of crop diversification can deliver increased income as well as more stable returns, and can thus help to ‘isolate’ price, at least partially, from external shocks (Coelli and Fleming 2004). For instance, when we return to the case

212

13

Information Exchange: Collaboration and Coordination Standards

of vanilla described above in the introduction to Chap. 12, Komarek (2010) demonstrates that the introduction of vanilla production, together with the production of matooke, has improved the household welfare of farming communities by 16% in the south of Uganda. The government of Uganda can play another pivotal role in the provision of services for the management of protocols and standards of the products. Also in this case, opportunities can be created for businesses to enter into the value chain production and add value to agriculture production, in areas of food processing and specialised products. The aims for the value chain, which are embedded in the horizontal and vertical agriculture production, should be to expand quality production and reach critical mass for consolidation. The farmers/outgrowers associations and the introduction of GAP lie at the heart of all these different policies, because at local level they can minimise input prices and bargain collectively to obtain the highest market prices through reduced transaction costs. From this perspective, we have reached two main conclusions, firstly, that it is not possible to find a single ‘one size fits all’ solution, i.e. due to the complexity of the agriculture sector in Uganda. Secondly, rather than take an ad hoc approach, which merely addresses issues as they emerge, we see the necessity for a retroactive approach, a willingness to design a comprehensive strategy that can be adapted locally but can above all proactively improve the production of agriculture in Uganda for national prosperity and development of trade. In conclusion, the strength and success of associations and coordination activity is based on trust which must be nurtured across farmers, traders, exporters, organisations, and government departments. The obvious vehicle for fostering trust is social media, through which the diffusion of information, best practice and innovation can allow for transparent and fair transactions, and ultimately economic growth.

References Booth D, Golooba-Mutebi F (2014) How the international system hinders the consolidation of developmental regimes in Africa. Developmental Regimes in Africa Project: The Overseas Development Institute, London Calderon C (2009) Infrastructure and growth in Africa. World Bank Policy Research Working Paper, 4914, Washington, DC Coelli T, Fleming E (2004) Diversification economies and specialisation efficiencies in a mixed food and coffee smallholder farming system in Papua New Guinea. Agric Econ 31 (2–3):229–239 Food and Agriculture Organization of the United Nations (FAO) (2003) Good agricultural practices: a working concept. FAO GAP Working Paper Series, Rome Food and Agriculture Organization of the United Nations (FAO) (2011) Uganda Nutrition Action Plan 2011–2016. LEX-FAOC144970, Rome Gollin D, Rogerson R (2010) Agriculture, roads and economic development in Uganda. NBER Working Paper, no. 15863. National Bureau of Economic Research Grabowski R (2012) Implicit taxation of agriculture: the cause of development failure in Egypt. Afr Dev Rev 24(3):183–193

References

213

Kindness H, Gordon A (2001) Agricultural marketing in developing countries. Natural Resources Institute Policy Series, 13. University of Greenwich, UK Komarek AK (2010) Crop diversification decisions: the case of vanilla in Uganda. Q J Int Agric 49 (3):227–242 Kydd J, Dorward A (2004) Implications of market and coordination failures for rural development in least-developed countries. World Dev 16:951–970 Mandl P, Mukhebi A (2002) Commodity market information and risk management: the case for a commodity exchange and warehouse receipt system for Uganda. Prepared for the Uganda Delegation of the European Community, February McIntosh C (2008) Estimating treatment effects from spatial policy experiments: an application to Ugandan microfinance. Rev Econ Stat 90(1):15–28 Minten B, Stifel D (2008) Isolation and agricultural productivity. Agric Econ 39(1):1–15 Okurut FN, Banga M, Mukungu A (2004) Microfinance and poverty reduction in Uganda: achievements and challenges. Research Paper 41. Economic Policy Research Centre, Kampala Uganda Communication Commission (UCC) (2014) Post, broadcasting and telecommunications market & industry report. Market and Industry Report, Kampala USAID (2008) Uganda Agricultural Productivity Enhancement Program (APEP). United States Agency for International Development Venables AJ (2001) Geography and international inequalities: the impact of new technologies. J Ind Compet Trade 1(2):135–159 World Bank (2011) Ugandan coffee supply chain risk assessment. World Bank, Washington, DC

Part IV

Conclusions

Chapter 14

Collaborative Approach to Trade

14.1

Methods to Study Interdependency

The task to define a single, consistent and accurate method for our analysis has posed a serious initial challenge, since our aim was to capture the complexities of multiple trade variables which continuously and simultaneously interact, change and adapt to specific contexts and trade chains. Our endeavours throughout this book have certainly led us to the need for a change in paradigm (Dubois and Prade 2004; Rotmans and Van Asselt 2001). The most recent advancement in global systems science and interactive agent economics (for a detailed review, see Arthur 2013) hinges on a pluralist method/ model approach. A pluralistic approach makes the modelling exercise more flexible, that is, the model is created through a sequential framework in which its different tiers can be analysed on the basis of different perspectives. In so doing we can ensure that the methods are evidence-based on available information but are also extensible; in other words, they are able to embrace new (or different) data, new spatial and temporal definitions, new indicators, and new modelling tools, ideally in a seamless fashion. Given the overarching and specific objectives addressed in this book, our first step has been to define the level of scales resolution, model structure, variables, and processes which needed to be included (Loehle 1990). The method we chose would have to provide a better understanding of the trade in the case study regions, it would have to create testable predictions and define strategies to lower trade cost, increase accessibility to markets, and visualise connections among trade entities. However, Helbing (2010) observed that there are many challenges which prevent the development of a model to capture some of the features of real economic and trade systems. Since our study required us to examine process-oriented policy initiatives (Hertel 1997) such as trade liberalisation, the main available method, based on partial and general equilibrium, suffers different drawbacks, including the scant number of studies on heterogeneity of trade entities, interaction behaviour, © Springer International Publishing AG 2017 F.R. Medda et al., Collaborative Approach to Trade, Advances in Spatial Science, DOI 10.1007/978-3-319-47039-9_14

217

218

14

Collaborative Approach to Trade

and dynamic and adaptive processes. Moreover, the specificity of our study context raised yet another constraint; we lacked consistent and available aggregate data and had to compile both quantitative as well as qualitative data from different sources. It was important therefore that we postulated our analysis by stating that the economic trade processes are driven by the local environments under scrutiny. As we have stated all along, our main objective in this book was to develop different strategies on how to increase trade and economic growth by focussing on lagging regions; and given the various signposts in the above discussion, we decided to implement two advanced methods of Regional Science: network analysis and agent based modelling. The far-reaching idea beyond the use of these two methods was to propose an analytical platform which would enhance interrelations among different variables and thus show an integration of the problem. And in considering this platform we have indeed structured our work to allow comparability between different lagging regions, both spatially and temporally, under trade capabilities. Furthermore, we have created modularity and a sequential framework in which the interactions between the different tiers of the models can be analysed from different perspectives. The models are thus grounded on data and information that is evidence-based and these models have proven to be extensible by seamlessly including new (or different) data, new spatial and temporal definitions, new indicators, and new modelling tools, as mentioned above. From our platform of analysis we have examined how trade interdependent dynamics constitutes different components (layers) in the methods, and have shown how these layers interact with each other. We have also considered two main study perspectives, the outward-looking, where we examined how to improve connectivity between sea-and land-locked regions and outside markets (long-distance international trade, short-distance international trade, regional trade, and accessibility). The second perspective was inward-looking, where we studied the behaviour of agents operating in the trade and logistics networks of sea-and land-locked regions in order to decrease costs and delays; we also tested how decision making and other behaviours of these agents are impacted by incentives and policies. For the outward-looking perspective we adopted a network theory approach (Jackson 2008) to examine how trade is influenced by cumulative network interactions. The concept of cumulative networks was interpreted in a twofold manner. The first facet involves how the modeller decides to represent a network system. In this sense, elements are aggregated in categories whose combinations help to describe the complete network system. The second facet can be ascribed to the emergence of global behaviours caused by local interactions which spread or percolate vertically and horizontally within the network system. Conversely, the inward-looking perspective was studied through the use of agent based modelling (ABM) (Epstein 2006). ABMs are a class of model applied widely in agricultural economics (Happe et al. 2006; Berger 2001; Balmann 1997, 1999). The advantages of this model are that it allows us to estimate data from different sources, and consider a set of the heterogeneous agents who interact so as to allow the exchange of information and who have adaptive behaviours to accommodate changes in the

14.2

The Collaborative Approach to Trade

219

environment. Both methodologies can incorporate positive and negative feedback, as described by Arthur (1994), who discusses in these cases how increasing and diminishing returns of the trade process can spur growth and lead to welfare improvements. In our case studies the variables we consider interact in a non-linear fashion, in that, due to their interdependencies, the directionality of their relation is difficult to specify. Non-stationary solutions were also possible outcomes. By taking these possibilities into account, we have focussed on the scale definition of the two perspectives (outward-looking for the SPICs and inward-looking for Uganda). Effectively for the case of the South Pacific Islands, the method was grounded in network analysis, whereas an application of agent based modelling was most suitable for the Uganda case study. Both methodologies were selected not only because they best address the given objectives, but also because they unearth deep insights into the interdependency and interrelations between trade variables without necessarily imposing causality in the relationships among the variables. The study of interdependency and interaction of trade factors, as interpreted in our context, is the study of the spread and adoption of new information and changes arising through connections. Keeping in mind that trade theory warns us of the impact of information asymmetry as a well-established factor hindering trade and undermining growth, to our immense satisfaction we have introduced in this case study analysis a new concept based on information exchange. Another important concept, like a ribbon running through the narrative of our study, is a collaborative approach to trade.

14.2

The Collaborative Approach to Trade

Economic and trade imbalances have grown steadily in recent years around the world despite globalisation trends, effectively creating country vulnerability, not only from an economic point of view but also through exacerbating food and resource insecurity, worsening poverty and increasing population displacement and conflict. Although recent attempts have been made by national governments and international organisations to reduce the negative effects of trade imbalances, it continues still, and affects many countries, especially emerging economies and medium and small-sized countries. Trade and consequently transport, the backbone of trade, are nowadays facing unprecedented challenges, particularly when competition dynamics are confronted by the new collaborative approaches to trade based on coordinated and collective action. The bedrock of this tension between competition and coordination in trade can be potentially ameliorated with the rise of new communication and Internet of Things (IoT) technologies. As observed by Miller (2013), coordination and collective action not only can support resource and information sharing but can also foster self-organising solutions, such as tailoring global solutions to local contexts. These are indeed the main findings and implications of our study.

220

14

Collaborative Approach to Trade

The revolution that the information technology sector has generated, and will continue to produce, is evident and demonstrated in both of our case studies. Growth and trade advantages can be gained through connectivity and interdependency within countries and between countries, and from this stand point physically insular countries and lagging regions, like the SPICs and Uganda, can act as hubs of services, operations, and transport (to mention only three) by being trade connectors through not only their physical infrastructures, but above all through information technology infrastructures. We began our analysis in Chap. 1 by asking two fundamental questions, how do we increase trade collaboration and how do we improve connections? By applying different perspectives from chapter to chapter and developing specific solutions, we have constructed the collaborative approach to trade concept. The 6Cs, the constituent factors of our analytical approach throughout this book, are as follows: • Cooperation. In Chap. 12 we tested how an integrated policy based on cooperation in exchanging know-how and best practice can reduce the number of Ugandan farmers leaving the market, thereby improving agriculture production and thus trade. • Competition. We demonstrated how trade competition is obtained if the transport systems allow for speed, reliability, and affordable access to markets. Port attractiveness (Chap. 5) and investment in the transport system (Chap. 11) are two elements which can produce long-term impacts by promoting new economic activity and ultimately trade competition. • Consolidation. In Chaps. 6 and 12 we explored how an increase in consolidation can actually facilitate trade and inclusive regional economic growth. • Coordination. Trade coordination can sometimes be detrimental for small countries as we observed in Chap. 7. However, when with coordination we aim to increase total trade income in a framework of a more balanced distribution of income among regions (e.g., the SPICs) we are in the position of reducing domestic monopolies and gaining political support for trade reforms. • Communication/Connection. Chapter 13 explored how access to information can help farmers capitalise on economies of scale and thus raise levels of investment. By connecting people and diffusing information about best practice, know-how and innovation, communication therefore helps to organise transparent and fair trade practices and leads to more inclusive economic growth. • Co-creation and co-sharing. The integrated multilayer model (Chaps. 2, 4 and 8) and the multiplier attachment (Chaps. 2 and 10) have been proposed in this study as tools for modelling and understanding how interdependency is the core element for trade value creation. All 6C factors represent significant and essential steps towards improving trade. As observed above, the common thread tying these factors together is trade interdependency: from information on goods and services distribution, to supply chain networks that leverage financial resources and investments, which catalyse and scale possible solutions, and eventually overcome trade ‘insularity’. The motivation to extract values from trade interdependency goes well beyond the

14.3

Trade Interdependency

221

mere increase in trade but also looks at the economic development of a country from a different perspective, one in which the collaborative trade approach connects both market and social objectives and increases cohesion and inclusion, and ultimately welfare.

14.3

Trade Interdependency

In our context, the collaborative approach to trade is inextricably linked with interdependency and trade interactions and is interpreted as the spread and adoption of new information and improvements arising through connections. Coordination mechanisms and access are two of the main strategies tested in our models of trade growth in relation to lagging regions. The results gave insights for several policy recommendations. Following the standard assumption in trade, that an improvement in transport infrastructure will increase trade and thus economic growth, we have demonstrated that, in so far as fostering economic growth is concerned, it would be more effective and financially efficient to increase access to information and value chain coordination rather than to address only physical connectivity. And certainly in line with this conclusion is the idea that a successful country needs to be connected and accessible to the wider world by leveraging the Internet of Things. The collaborative trade approach matters most when building new transport infrastructures such as logistics structures, because the net effect of a new infrastructure is strongly related to existing ones. This is especially relevant if we consider that IT infrastructure is gravitating towards the paradigm of Big Data where, hopefully, most of our infrastructures will be monitored continuously, allowing for fast responses, for instance, to changes and economic losses. New advances in technologies for transport information, renewable energy sources, and sharing economics are just a few of the trends discussed in this study. An opportunity to recognise and take advantage of the high level of interdependency between infrastructures and lagging regions is now ripe for the picking. The values of infrastructure interdependency, as in a supply chain, can be the source of an additionality, compared to the single infrastructure benefit baselines because interdependent infrastructures can be linked to opportunity of the investments, e.g. value creation and value capture of investment costs. The effects of the economic and financial crisis of 2008 are still being felt across the world. A significant global tightening of credit has altered the roles of governments and the private sector and has also greatly impacted on international trade. In October 2013 Chinese President Xi Jinping proposed the construction of “One Belt and One Road” as a new paradigm for international and regional trade where the strategy is one of “balancing convergence and divergence” and focussing on “inclusive development and upholding the basic concept of open cooperation and harmonious inclusiveness, market operations and mutual benefit.” In this book we have discussed that the solutions for a collaborative approach to trade are in fact possible, especially in sea- and land-locked countries, if trade channels are widened

222

14

Collaborative Approach to Trade

and flexible connections are achieved. New trends such as information economics and other technologies have spurred the redefinition of participation in international trade through, e.g., internet platforms, crowdsourcing and peer-to-peer finance. Is an improved paradigm for trade emerging just above the horizon? We think so, and in this book we have demonstrated how interdependency and the collaborative approach can reshape the foundations of established institutions and networks of international trade. Some of these trade institutions have already faded, giving rise to our view that now is the right time to create a new paradigm, the collaborative approach to trade. And in this context, we can observe that even though a large number of obstacles and elements impact on trade, certainly ‘physical insularity’ can never be a justification for a lagging economy and above all, for trade insularity.

References Arthur WB (1994) Increasing returns and path dependence in the economy. University of Michigan Press, Ann Arbor, MI Arthur WB (2013) Complexity economics. Oxford University Press, Oxford Balmann A (1997) Farm-based modelling of regional structural change: a cellular automata approach. Eur Rev Agric Econ 24(1–2):85–108 Balmann A (1999) The dynamics of structural change and farmers’ income in a maturing sector. Working Paper. Humboldt University of Berlin Berger T (2001) Agent based spatial models applied to agriculture: a simulation tool for technology diffusion, resource use changes and policy analysis. Agric Econ 25(2–3):245–260 Dubois D, Prade H (2004) Possibilistic logic: a retrospective and prospective view. Fuzzy Sets Syst 144:3–23 Epstein JM (2006) Generative social science: studies in agent based computational modeling. Princeton University Press, Princeton Happe K, Kellermann K, Balmann A (2006) Agent-based analysis of agricultural policies: an illustration of the agricultural policy simulator AgriPoliS, its adaptation and behavior. Ecol Soc 11(1):49 Helbing D (2010) Quantitative sociodynamics, 2nd edn. Springer, Zurich Hertel, T.W. (1997). Global trade analysis: modeling and applications. Cambridge: Cambridge University Press Jackson MO (2008) Social and economic networks. Princeton University Press, Princeton, NJ Loehle G (1990) A guide to increased creativity in research: inspiration or perspiration? Bioscience 40(2):123–129 Miller HJ (2013) Beyond sharing: cultivating cooperative transportation systems through geographic information science. J Transp Geogr 31:296–308 Rotmans J, Kemp R, van Asselt M (2001) More evolution than revolution: transition management in public policy. Foresight 3(1):15–31