Industrial clustering, firm performance and employee welfare evidence from the shoe and flower cluster in ethiopia

© Olaf Gloeckler, Atelier Platen, Friedberg75 DEVELOPMENT ECONOMICS AND POLICYSeries edited by Joachim von Braun, Ulrike Grote and Manfred Zeller Industrial Clustering, Firm Performance

© Olaf Gloeckler, Atelier Platen, Friedberg

75

DEVELOPMENT ECONOMICS

AND POLICYSeries edited by Joachim von Braun, Ulrike Grote and Manfred Zeller

Industrial Clustering, Firm Performance

and Employee Welfare Evidence from the Shoe and Flower Cluster in Ethiopia

Tigabu Degu Getahun

75

The author examines the productivity, profitability and welfare effects of trial clustering and a public policy promoting industrial clusters in Ethiopia He uses reliable counterfactuals as well as original enterprise and worker level data By investigating the effect of firm, time, entrepreneur and site specific factors as well as endogenous location choice issues, the author finds strong evidence for the existence of significant agglomeration economies in the Ethio- pia leather footwear cluster Using primary survey data collected from firms which benefited from the cluster policy and those that did not, both before and after the implementation of the policy, the author shows the unintended negative impact of a cluster prompting policy in Ethiopia The book is essential reading for those who are interested in the gender and welfare impact of female full time labor force participation in industrial jobs.

indus-Tigabu Degu Getahun studied economics at the University of Copenhagen and the University of Bonn He is a Senior Researcher at the University of Bonn and a Research Fellow at the Ethiopian Development Research Institute (EDRI) in Ethiopia.

Industrial Clustering, Firm Performance and Employee Welfare

DEVELOPMENT ECONOMICS

AND POLICY

Series edited by Franz Heidhues †, Joachim von Braun,

Ulrike Grote and Manfred Zeller

Vol 75

Industrial Clustering, Firm Performance

and Employee Welfare

Evidence from the Shoe and Flower Cluster in Ethiopia

Tigabu Degu Getahun

Bibliographic Information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche

Nationalbibliografie; detailed bibliographic data is available in the internet at http://dnb.d-nb.de

Zugl.: Bonn, Univ., Diss., 2015

Library of Congress Cataloging-in-Publication Data

Names: Getahun, Tigabu Degu, 1980-

Title: Industrial clustering, firm performance and employee welfare : evidence from the shoe and flower cluster in Ethiopia / Tigabu Degu Getahun

Other titles: Development economics and policy ; v 75

Description: New York : Peter Lang, 2016 | Series: Development economics and policy ; vol 75 | doctoral Universität Bonn 2015

Identifiers: LCCN 2016006143 | ISBN 9783631667446

Subjects: LCSH: Industrial clusters—Ethiopia | Shoe industry—Ethiopia | Floriculture—Ethiopia | Women—Employment—Ethiopia

Classification: LCC HC845.Z9 D54 2016 | DDC 338.70963—dc23 LC record available at http://lccn.loc.gov/2016006143

D 98 ISSN 0948-1338 ISBN 978-3-631-66744-6 (Print) E-ISBN 978-3-653-06378-3 (E-Book) DOI 10.3726/978-3-653-06378-3

© Peter Lang GmbH Internationaler Verlag der Wissenschaften

Frankfurt am Main 2016 All rights reserved

PL Academic Research is an Imprint of Peter Lang GmbH

Peter Lang – Frankfurt am Main · Bern · Bruxelles · New York ·

Oxford · Warszawa · Wien All parts of this publication are protected by copyright Any

utilisation outside the strict limits of the copyright law, without

the permission of the publisher, is forbidden and liable to

prosecution This applies in particular to reproductions,

translations, microfilming, and storage and processing in

electronic retrieval systems.

This publication has been peer reviewed

www.peterlang.com

in other areas.

The study employs appropriate estimation strategies to disentangle the effect

of industrial clustering from firm heterogeneities and other cofounders The timation results from both the random effect model and the Abadie and Imbens (2011) bias corrected nearest neighbor matching model reveal the productivity and profitability increasing effect of industrial clustering, after controlling for the effects of site-, enterprise-, entrepreneur-, and time- specific factors The study also accounts for selection bias and endogenous location choice problem The results from the two way fixed effect impact evaluation model suggests that the implemented government cluster development program in Ethiopia has adversely impacted the productivity, profitability, growth, and innovation per-formance of the treated firms Due to the short span of the program, however, these findings only reflect the short-term impacts of the program

es-The regression result from the Mincerian earning function indicates that tered firms paid higher mean wages compared to non-clustered counterparts; implying a welfare increasing effect of industrial clustering In addition, the re-sults of a correlation matrix analysis disclose a positive and significant correla-tion between firm productivity and the number of employees, again implying positive welfare effects of industrial clustering The results of a gender disaggre-gated analysis of employment effects also reveals a positive relationship between gender equity and industrial clustering

clus-To further explore the welfare and gender impacts of industrial clustering, the study empirically investigates the intra-household welfare impacts of formal salaried employment of women in the most female dominated cluster in Ethio-pia, the cut flower industry cluster To this end, a unique quantitative survey was conducted with a random sample of 670 women working in the cut flower cluster

and a control group of 182 women who applied for work in the cluster but were unsuccessful

The relevance tests of the hypotheses and results of a special maximum hood estimation model, an endogenous binary treatment model, and difference

likeli-in difference models comblikeli-ined with likeli-instrumental variable estimators all suggest that, compared to employment elsewhere, salaried employment of women in the cut flower cluster improved the income and consumption welfare of the work-ing women and their household It also (i) has negative relationships with the incidence and depth of poverty; (ii) has reduced the food insecurity and hunger status of working women and their households; (iii) has improved the bargain-ing and decision making power of the flower working women (iv) has trans-formed the traditional gendered patterns of intra household time use and (v) has negative relationships with the leisure demand of the flower working women and their close substitutes

The investigation of the transmission mechanisms further suggests that creases in women’s earnings derived from employment in the cut flower cluster has effected the consumption welfare of the flower working women’s household not only through the usual Marshallian income and substitution effects,but also through the distinguishing bargaining effect The qualitative findings support the quantitative findings, but unveil additional intangible benefits and costs of female labor force participation

Trotz des Anstiegs von empirischen Cluster Fallstudien seit der wegweisenden Arbeit von Marshall (1920), haben nur wenige dieser Studien versucht, Produk-tivitäts-, Profitabilitäts-, und Wohlfahrtseffekte des industriellen Clusterings zu quantifizieren Die vorliegende Studie analysiert deshalb empirisch Produktivi-täts- und Profitablitätseffekte des industriellen Clusterings sowie einer staatlichen Politik zur Föderung von industriellen Clustern in Äthiopien unter Verwendung von neu erhobenen Umfragedaten auf Firmenebene in 196 Lederschuhe produ-zierenden Firmen, die in einem spontan entstandenen Lederschuhcluster in Äthi-opien operieren; außerdem 86 Firmen, die in einem von der Regierung initiierten Cluster operieren sowie 72 Firmen, die außerhalb von Clustern operieren

Die Studie nutzt verschiedene Schätzstrategien, um die Effekte des Clusterings von Firmen-Heterogenitäten und anderen Faktoren zu trennen Schätzergeb-nisse aus dem Random Effect Modell und dem von Abadie und Imbens (2011) Bias-korrigierten Nearest Neighbor Matching Schätzer zeigen einen zunehmen-den Effekt des industriellen Clusterings auf Produktivität und Profitabilität, nachdem für Effekte von natürlichen Gegebenheiten sowie Firma-, Unterneh-mer- und zeitabhängigen Faktoren kontrolliert wird Die Studie berücksichtigt sowohl Selektionsverzerrungen als auch endogene Standortwahlprobleme.Schätzergebnisse des am weitesten verbreiteten Impaktevaluationsmodells, dem Difference in Difference-Modell, zeigen an, dass das implementierte Clus-ter-Entwicklungsprogramm der Regierung adverse Effekte bezogen auf Profita-bilität, Wachstum, Innovation und Produktivität hat Jedoch bezieht sich dieses Resultat, aufgrund der kurzen Zeitspanne des Programms, nur auf den unmit-telbaren kurzfristigen Effekt des Programms

Regressionsresultate aus Mincer-Typ-Einkommensfunktionen zeigen, dass Cluster Firmen im Durchschnitt höhere Löhne für ihre Arbeiter zahlen als entsprechende Nicht-Cluster Firmen, was auf einen positiven Wohlfahrtsef-fekt des industriellen Clusterings hinweist Außerdem zeigt die Analyse der Korrelationsmatrix eine positive und signifikante Korrelation zwischen der Produktivität der Firma und der Höhe der Beschäftigung an, was wiederum positive Wohlfahrtseffekte des industriellen Clusterings widerspiegelt Die Di-saggregation des Beschäftigungseffekts nach Geschlecht deutet auf eine Redu-zierung der Geschlechterdisparität hin

Um den Effekt auf Wohlfahrt und Geschlecht weiter zu erkunden, sucht die Studie den Intra-Haushaltseffekt der Lohnbeschäftigung von Frauen

unter-in eunter-inem der von Frauen am meisten domunter-inierten Cluster unter-in Äthiopien – dem Blumencluster Dafür wurden einzigartige quantitative Umfragedaten mittels Zufallsstichprobe erhoben von (i) 670 Frauen, die im Blumencluster eine Be-schäftigung aufgenommen haben und (ii) einer Kontrollgruppe von 182 Frauen, die sich beworben hatten, aber keine Beschäftigung aufgenommen haben Der Relevanztest der Hypothese und die Schätzergebnisse des Maximum Likelihood-Schätzmodells, des Treatment Effect-Modells, des mit Matching kombinierten Difference in Differences-Modells und des Instrumentalvariab-lenschätzers zeigen alle an, dass Lohnbeschäftigung von Frauen im Blumenclus-ter, verglichen mit sonstiger Beschäftigung, den arbeitenden Frauen und deren Haushalten dazu verhilft (i) die Inzidenz und den Schweregrad der Armut zu reduzieren, (ii) Hunger und Nahrungsmittelunsicherheit bei den beschäftigten Frauen und deren Haushalten zu verringern, (iii) die Verhandlungsposition und Entscheidungsbefugnisse der Frauen zu verbessern and (iv) die Nachfrage der

im Blumensektor beschäftigten Frauen nach Freizeit zu verringern

Interessanterweise deutet die Analyse zusätzlich darauf hin, dass die Erhöhung des Einkommens der Frauen aufgrund der Beschäftigung im Blumencluster signi-fikant die kollektive Haushaltsnachfrage beeinflusst und zwar nicht nur durch ei-nen gewöhnlichen Marshall’schen Einkommens- und Substitutionseffekt, sondern auch durch einen differentiellen Verhandlungseffekt Während die qualitativen Resultate die quantitativen Resultate unterstützen, decken sie zusätzlich weitere, schwieriger zu greifende Nutzen und Kosten der weiblichen Teilnahme an der Er-werbstätigkeit auf

Acknowledgement

It is a great pleasure to thank God, the many people and institutes who made this book possible First and foremost glory to the Almighty for enabling me to accom-plish my writing successfully My boundless gratitude goes to my first Professor

Dr. Joachim Von Braun Throughout my thesis-writing period, he provided couragement, sound advice and lots of good ideas I would have been lost without him It has been an honor to be his student I am also very grateful for my second Professor Dr Ulrich Hiemenz and my Dr. Marc Müller who took time to review

en-my work I am also greatly indebted to en-my research partner and friend Dr Espen Villanger for his continued encouragement, support and inspiration

I am heavily indebted to H.E Ato Newai Gebre-ab, EDRI director and chief economic advisor to the Ethiopia government prime minister, for his invaluable advice, encouragement, moral support and easy yes for all my difficult requests I would also like to wholeheartedly thank Mrs Rosemary Zabel, Mrs Maike Retat Amin and Dr Gunther Manske for their excellent admin and logistic assistance

I would also like to take this opportunity to gratefully acknowledge the ing sources that made my doctoral study possible The first three years of my study was primarily funded by the German Academic Exchange Service while part of my last year study was funded by BMZ through ZEF My field research was mainly funded by the Dr Hermann Eiselen grant program of the Fiat Panis foun-dation Part of my work was also financially supported by IDRC through EDRI

fund-My time at Bonn was made enjoyable due to the many friends, fellow dents and colleagues Daniel Ayalew, Lukas Kornher, Million, Negash, Christine Husmann and Robert Poppe thank you so much for making my time in Germany more enjoyable Christine and Robert thank you so much for the German trans-lation of my abstract Nuru Yasin, Hussein Ahmed, Abdurrahman Ali, Ibrahim Worku, Feiruz Yimer, Biruk Teklie, Dr. Girum Abebe, Dr Seid Nuru, Ato Mezgebe Mihretu, Bedilsh, Desta Solomon, Laura Kim and friends and colleagues at EDRI, please accept my sincerest appreciation for being always with me in all my difficult times; and for providing all the emotional supports and comraderies

stu-Dr Seid and stu-Dr Assefa, many thanks for believing on me and recommending

me to study in the University of Bonn, Center for Development I hope I have not failed you

Last, but not the least, I am forever indebted to my entire extended families Words cannot express how grateful I am to my mom Tsehaynesh Kebede and late grand mom Nurit Nigatu for all of the sacrifices they made on my behalf My sweet kids Nebil Tigabu and Jasmin Tigabu, you are my real source of pleasure and call to action Thank you so much for always cheering me up.

List of Tables 15

List of Figures 19

List of Acronyms 21

1 General Introduction 23

1.1 Background 23

1.2 Literature Review and Experimental Hypotheses 26

1.3 Research Problem and Significance of the Study 35

1.4 Research Objectives and Questions 39

1.5 Methods 40

1.6 Structure of the Book 42

2 Industrial Clustering and Firm Performance: The Ethiopian Leather Shoe Industry 45

2.1 Introduction 45

2.2 Overview of the Ethiopian Leather Shoe Industry 48

2.3 Conceptual Framework 51

2.4 Survey Design, Data and Measurement 55

2.5 Characteristics of Sample Enterprises 61

2.6 Firm Location Choice 68

2.7 Industrial Clustering, Firm Performance and Employee Welfare 70

2.8 Estimation Strategy and Results 79

2.9 Transmission Mechanisms 88

2.9.1 The Leather Footwear Cluster and its Supply Chain 88

2.9.2 Industrial Clustering and Small Firm Growth Barriers 97

2.10 Concluding Remarks 100

3 Impacts of the Cluster Development Program in Ethiopia 103

3.1 Introduction 103

3.2 Data and Sampling Method 106

3.3 Pre-Intervention Characteristics of Sample Firms 108

3.4 Potential Impacts of the MSME Cluster Development Program 112

3.5 Evaluation Methods, Results and Discussions 114

3.5.1 Evaluation Methods 114

3.5.2 Estimation Results and Discussions 117

3.6 Concluding Remarks 132

4 Welfare and Gender Impacts of Female Employment in an Industry Cluster: Evidence from the Flower Cluster in Ethiopia 135

4.1 Introduction 136

4.2 Context 141

4.3 Theoretical Model 143

4.4 Econometric Model 145

4.5 Sampling Method, Data and Measurements 147

4.6 Initial Demographic and Socio-Economic Characteristics 154

4.7 Estimation Strategy and Results 159

4.7.1 Impact on the Monetary Dimension of Wellbeing 160

4.7.2 Impact on the Non-Monetary Dimension of Wellbeing 181

4.7.3 Gender Roles 186

4.7.4 Transmission Mechanism: Drivers of the Observed Welfare Changes 194

4.8 Concluding Remarks 199

5 Conclusion 203

5.1 Synopsis 203

5.2 Limitations of the Study and Suggestions for Future Research 206

References 209

Annexes 225

List of Tables

Table 2.1: Years of Operation of the leather footwear manufacturers 62Table 2.2: Initial numbers of workers among leather footwear

manufacturers 62Table 2.3: Sources of Initial Investment and Working Capital 63Table 2.4: Percentages of formally registered leather shoe

manufacturers in Ethiopia 64Table 2.5: Characteristics of leather shoe manufacturer entrepreneurs

in Ethiopia, 2013 65Table 2.6: Site-specific factors of leather manufacturing firms in

Ethiopia, 2013 68Table 2.7: Major reasons for firm location choice among leather shoe

manufacturers, 2013 69Table 2.8: The performance of clustered and non-clustered leather

shoe industry firms 71Table 2.9: Leather shoe manufacturer marketing channels

by location, 2013 74Table 2.10: Wage rates, Number of employees and Growth among

Leather shoe manufacturers 76Table 2.11: Work experience among employees of clustered and

non-clustered leather shoe manufacturers in Ethiopia, 2013 77Table 2.12: Percentages of female workers in the leather shoe industry

in Ethiopia, 2013 78Table 2.13: Nearest neighbor matching estimates of industrial clustering

effects on the leather shoe manufacturers’ performance 80Table 2.14: Random effect estimates of the industrial clustering effects

on the leather shoe manufacturers Performance 82Table 2.15: Regression estimates of the earning function of cluster

impacts on the leather shoe Manufacturers, 2013 88Table 2.16: Percentages of Shoe manufacturers that collaborate

frequently with other Manufacturers, 2013 92Table 2.17: Relationship Between firm performance and horizontal

collaboration in the leather shoe industry in Ethiopia, 2013 95

Table 2.18: Relationship between firm performance and downstream

collaboration 95Table 2.19: Relationship between firm performance and upstream

collaboration 96Table 2.20: Mean scores of small firm growth constraints in the leather

shoe industry, 2013 98Table 3.1: Pre-Intervention Entrepreneur characteristics 109Table 3.2: The Characteristics of the Control and treatment Firms

before the implementation of the cluster policy 110Table 3.3: Mean monthly performance indicator values among control

and treatment leather shoe manufacturers before the

implementation of the cluster policy 112Table 3.4: The performance of treatment and control firms in 2010 and

2013 113Table 3.5: The DID Estimates of the impacts of the cluster

development program 118Table 3.6: Business network effects of the cluster development policy:

DID model results 122Table 3.7: Impacts of cluster policy on information and experience

exchange collaboration 124Table 3.8: Percentage of Shoe manufacturers that collaborated

frequently with similar firms in the leather shoe industry in Ethiopia 125Table 3.9: The mean value of Small firm growth constraint indicators

in the leather shoe industry in Ethiopia 130Table 4.1: Education and Experience of the Women and

Their Spouse 155Table 4.2: Demographic Characteristics 157Table 4.3: Initial Economic Condition of the Respondent by

Participation Status 158Table 4.4: The Initial Characteristics of the respondent Parent 159Table 4.5: Potential Impact of female Flower Job employment 160Table 4.6: DID Estimation of Female Employment Impact on Wage

and Non-wage Income 163Table 4.7: The FIML Estimates of Selection & Consumption Welfare

Equation 167

Table 4.8: The Computed ATE and ATET values based on the

Consumption Function Estimates 169Table 4.9: DID, FE, DID_GMM and DID-3SLS Estimate of

Consumption Welfare 171

Table 4.10: Annual expenditure on clothing, cloth, tailoring and

footwear (Birr), 2013 172Table 4.11: Impact on Poverty Incidence 174Table 4.12: The Probit Model estimate of the Poverty Impact of Female

Employment 175Table 4.13: The FIML Estimates of log of per Adult Food Consumption

Equations 176Table 4.14: The Computed ATE and ATET valuesbased on the food

Consumption Function Estimates 177Table 4.15: Impact on Food consumption 178Table 4.16: Percentage of children and adults who ate one, two three

and four times per day 179Table 4.17: Percentage of household members who ever sleep hungry

last week,2013 179Table 4.18: Food Insecurity and hunger Status Level-Categorical 180Table 4.19: Impact on Food Insecurity and Hunger

scale-Continuous 180Table 4.20: Women’s Own Happiness Assessment, 2013 181

Table 4.21: Impact on Intra-Household Leisure Time

Allocation, 2013 183Table 4.22: Percentage of Member women and

Average Social Capital Score, 2013 185Table 4.23: Access to Emergency Fund 185Table 4.24: Average Monthly hour the women spend on

domestic work 188Table 4.25: Intra-Household Earning Difference 189Table 4.26: Determinants of Women’s Control over

Household Finance 190Table 4.27: Women Contribution to household expenditure 191Table 4.28: Percentage of Women Who Disagree with the Following

Gender Inequitable Statements 192

Table 4.29: Self-confidence: Percentage of women who

agree with the following statement 193Table 4.30: Determinants of Consumption Welfare and

Food Insecurity 195Table 4.31: Determinants of Leisure Demand 198

List of Figures

Figure 2.1: Initial start-up working and fixed capita reported by leather

shoe manufacturers 72

Figure 2.2: Sources of competition among leather shoe manufacturers according to firm owners own rating, 2013 73

Figure 2.3: Engagement in innovative activities among leather shoe manufacturers, 2013 75

Figure 2.4: Primary firm-level innovation motivation among leather shoe manufacturers, 2013 75

Figure 2.5: Percentages of sales value by mode of payment in the leather shoe industry, 2013 90

Figure 2.6: Percentages of Leather shoe manufacturers that collaborated frequently with input suppliers, 2013 93

Figure 2.7: Percentages of leather shoe manufacturers that collaborate frequently with shoe traders, 2013 94

Figure 3.1: Percentages of shoe manufacturers that collaborated frequently with input suppliers in the leather shoe industry in Ethiopia 126

Figure 3.2: Percentages of shoe manufacturers that collaborated frequently with shoe traders in the leather shoe industry in Ethiopia 127

Figure 4.1: Previous Occupation of the Flower Job Participating Women 154

Figure 4.2: Previous Occupation of the Non-participating/control women 155

Figure 4.3: Potential Income Impact of flower Job Participation 162

Figure 4.4: Summary of Changes in Intra-household Wellbeing 173

Figure 4.5: Intra-household allocation of domestic responsibilities 186

Figure 4.6: Intra-household domestic work burden for the Participating women 187

Figure 4.7: Percentage of the women who were involved in the final household decisions 194

List of Acronyms

ADLI Agricultural Development Led Industrialization

ATET Average Treatment Effect on the Treated

BMZ German Federal Ministry for Economic

Coopera-tion and Development (BMZ)CSA Central Statistical Agency

DAAD German Academic Exchange Service

EIFCCOS Ethiopia International Foot wear Cooperative SocietyEDRI Ethiopian Development Research Institute

ESSP Ethiopian Strategy Support Program

FDRE Federal Democratic Republic of Ethiopia

FeMSEDA Federal Micro and small Enterprises Development

Agency

GPRG Global Poverty Research Group

GRIPS Japan National Graduate Institute for Policy Studies

ICT Information communication Technology

IDS Industrial Development Strategy

IFPRI International Food Policy Research Institute

LDI Leather Development Institute

LLPTI Leather and Leather Product Technology instituteMoFED Ministry of Finance and Economic Development MSME Micro Small and Medium scale Enterprises

OECD Organization for Economic Cooperation and

Devel-opment

PASDEP Policy - Plan for Accelerated and Sustained

Devel-opment to End PovertyROSCAS Rotating Saving and Credit Association

R&D Research and Development

SDRP Sustainable Development and Poverty Reduction

Program

UNDP United Nation Development Program

UNIDO United Nations Industrial Development Organization

USDA United State Department of Agriculture

1 General Introduction

1.1 Background

Ethiopia has remained a predominately agrarian subsistence economy despite its long history of civilization Industry in the modern sense only emerged in the country during the 1920s.1 Yet, a cognizant effort to stimulate industrial growth began with the formation of the federal government’s first five-year develop-ment plan in 1957 During the 1960s and early 1970s the country enjoyed rela-tively better industrial performance, although most manufacturers were foreign owned (Eshetu, 1985; Shiferaw, 1995)

In 1974 a military government came to power that instituted command ented policies and nationalized virtually all medium- and large-scale industrial companies owned by both foreigners and Ethiopians Private ownership was only permitted for cottage industries and small-scale enterprises during this period The strict leftist economic and political ideology pursued by the military government suppressed private economic activity considerably and subdued entrepreneurial spirit (Eshetu, 1985; World Bank, 1985)

ori-In 1991 the present civilian government replaced military authorities and undertook a series of reforms to reorient the economy from a command to a market economy Consequently, the private sector began to increase and a re-markable number of micro-, small - and large-scale manufacturing enterprises emerged spontaneously, mainly in the capital city (Addis Ababa) and its suburbs Some of these enterprises are spatially concentrated in few areas, while others are spatially distributed throughout the country For example, the vast majority

of micro-, small- and medium-scale leather footwear manufacturers are trated in the western part of the capital, while a few are located in other parts of the capital and surrounding suburbs

concen-To further accelerate the expansion of manufacturing industries and bring about structural economic transformation, the Ethiopian federal government initiated a comprehensive Industrial Development Strategy (IDS) under the umbrella of the overall national development policy called Agricultural Devel-opment Led Industrialization (ADLI)2 in 2002 This strategy was intended to

1 Before the 1920s the production of hand-crafted goods was the major manufacturing activity in Ethiopia.

2 The ADLI strategy, which was adopted in the mid-1990s as a national economic policy, presumes that agriculture will play a leading role in the industrialization process by

stimulate industrial development by strengthening the linkages between culture and industrial sectors (IDS 2002).

agri-In pursuit of this strategy the government started providing various tives, not only for the agricultural sector, but also for agro-processing and labor intensive manufacturing industries that have strong linkages to the agricultural sector (and hence to the rural economy) These supports include tax holidays, preferential financing such as longer grace and repayment periods, below-market interest rates, preferential land lease terms (low rates and long-term contracts), and duty free importation of machinery (GTP 2011) In addition to these gen-eral support measures, the Ethiopian federal government also provides special support to a few priority sectors including the leather goods and cut flower in-dustries Various special programs that provide support such as benchmarking, institutional twining, and market research have been implemented since 2002/3.Subsequently, the Ethiopian economy has registered continuous double-digit annual growth since 2003/4, placing it among the top performing economies in sub-Saharan Africa Over this period the industrial sector has also grown at the same rate and a remarkable number of privately owned manufacturing enterpris-

incen-es have emerged Dincen-espite thincen-ese positive achievements, structural transformation

of the economy has not yet been achieved (CSA 2013) The industrial sector has lingered, however, contributing the least to employment generation and overall GDP growth, even compared with other sub-Saharan African economies (World Development Report 2013) In addition, manufacturing has been largely domi-nated by small- and micro-scale enterprises.3 Strikingly, despite their superior performance in terms of employment generation, which is surpassed only by agriculture, the growth and competiveness of micro- and small-scale manufac-turing enterprises have also remained limited and the ‘medium-scale’ enterprises have been overlooked, as is the case in most of other sub-Sahara African econo-mies (Bigsten and Södernbom 2006, Ali and Peerlings 2011)

In the beginning the policy focus of the current government exclusively geted export-oriented large-scale manufacturers Even though the small- and micro-scale enterprises represent the second largest source of employment after

tar-setting up conditions for full-fledged industrialization The rationale for this strategy is that, given the predominantly agrarian nature of the economy, industrial development should rely on and serve agriculture by providing agricultural inputs and consumer goods (IDS 2002).

3 Small- and medium-scale enterprises in Ethiopia cover more than 95% of the total number of manufacturing enterprises, and account for the greatest share of off-farm employment (CSA 2007).

agriculture, only recently are they able to attract the interest of policy makers

In the late 1990s and early 2000s a number of empirical studies such as Collier and Gunning (1999), Schreiner and Woller (2003), Hernández, Pegan and Pax-ton (2005), Bigsten and Södernbom (2006), Collier and Venables (2008), sought

to explain the major factors impeding growth of manufacturing sectors in Saharan Africa, including Ethiopia The major explanatory factors for the dismal performance of manufacturing sectors of emerging economies in general, and particularly for the micro- and small-scale manufacturing firms, were identified as: high transaction costs (due to imperfect information availability and contract enforcement), high transportation costs, poor infrastructure (particularly a lack

sub-of access to distant input and output markets), a lack sub-of access to credit, poor managerial and technical skills, a lack of trust, a lack of start-up and working capital, inconsistent demand, and difficulties achieving economies of scale.During this period industrial clustering proliferated globally as an alternative industrial development strategy to ease major growth impediments of micro- and small-scale manufacturing enterprises, particularly in emerging economies such

as in East Asia (Chaudhry 2005) Mainstream development, business, and graphical economists have recommended cluster-based industrial development strategies to overcome most of the growth constraints of small-scale manufactur-ing enterprises (Wallis and North 1987, Porter 1990, Krugman 1991, Schmitz and Nadvi 1999, 2005, Brenner 2001, 2006, Sonobe and Otsuka 2006, 2011, Ruan and Zhang 2009, Venables 2010, Altenburg 2011)

geo-The successful development of micro-, small-, and medium-scale enterprises (MSMEs) has also been attributed to industrial clustering in the USA (Krugman 1991), Germany (Brenner 2001, 2006), Italy (Parrilli 2009), Indonesia (Weijiland 1998), Peru (Visser 1994), China (Ruan and Zhang 2007), and East Asia (Sonobe and Otsuka 2006) Inspired by these countries experiences and partly owing to the growing optimism in development literature regarding the growth, export and employment prospects of small and micro enterprises via clustering, the Ethio-pian government has recently shifted its policy emphasis towards a cluster-based industrial development strategy In this regard, the handloom, and the micro- and small-scale leather shoe manufacturers stand to benefit considerably

The government’s cluster-based industrial development policy is not only cused on small- and micro-scale manufacturing industries The government also develops industrial zones and clusters in different parts of the country including basic infrastructure and facilities such as roads, telecommunications systems, water, and electricity to enhance the competitiveness of export-oriented indus-tries and to attract foreign direct investment In this regard, the export-oriented

fo-textile, leather shoe, and cut flower industries receive the most government tention (GTP 2011) For example, in 2003 the Ethiopian government allocated 1,000 ha for flower production in the vicinity of the airport in the capital, leading

at-to the formation of cut flower industry clusters (IDS 2002) The development bank of Ethiopia also provides long-term credit (10 years) at an affordable inter-est rate (7.5%) with a three-year grace period and up to a 70:30 debt to equity ratio with no collateral requirements for foreign and domestic investors engaged

in flower production and export Consequently a large number of flower farms emerged in the Addis Ababa area, particularly in two districts, Walmera and Adaa, located on opposite sides of the capital Significant changes occurred in

2007 (see Annex 3.1), and subsequently 85,000 direct jobs and a large number

of indirect jobs were created for the neighboring rural communities, mostly for marginalized unskilled women who typically lack opportunities to find formal salaried jobs, which is expected to transform local rural livelihoods

1.2 Literature Review and Experimental Hypotheses

As indicated by Martin and Sunley, 2003, there is no consensus about the tion, the characteristics and constituting elements of industrial cluster In the present study, industrial cluster is defined as a spatial concentration of interre-lated firms that are interlinked by input, output, knowledge and other flows that may give rise to agglomerative advantages (Lublinski, 2003)4

defini-Since the industrial revolution the role of cluster-based industrial ment has been discussed in the literature, even though only recently has interest

in industrial cluster analyses proliferated as an alternative industrial ment strategy Von Thünen (1826) developed the first location model to ana-lyze the relationships among markets, production and distance, however, his model was framed to examine agricultural landscapes.5 In the second half of the

develop-19th century when industrial sectors became increasingly important to Germany, Alfred Weber, who is considered the founder of the classical theory of location, asserted that firms choose to co-locate with many other firms to exploit the ben-efits of agglomeration or external economies of scale (Heijman and Schipper 2010) According to Weber (1909), agglomeration benefits occur in industrial

4 Firms producing similar products within small geographical areas (Sonobe and Otsuka 2006) Schmitz and Nadvi (1999) define industrial cluster as a sectorial and spatial concentration of firms Porter (1990) defines cluster as a geographical concentration

of interconnected companies and institutions in a particular industry.

5 See http: //people.hofstra.edu/geotrans/eng/ch6en/conc6en/vonthunen.html.

clusters due to common use of physical infrastructure and technical knowledge Weber also identified two other factors that have important roles in firm loca-tion choice: transportation and labor costs Weber indicated that firms choose

to locate closer to larger labor pools to minimize labor costs and closer to their customers and material input providers to minimize transportation costs Besides Weber, Marshall (1920) addressed the determinants of firm location de-cisions and the economic benefits of industrial clustering Marshall (1920) argued that the transportation costs of goods, labor and ideas is cheaper when similar firms are spatially concentrated in a small geographical area According to Mar-shall clustered firms outperform dispersed firms because the former benefits from: (i) input market externality, (ii) goods market externality, (iii) labor pooling, and (iv) intra- cluster knowledge/information spillover by virtue of being in an indus-trial cluster

According to Marshall’s theory of input market externality, the co-location

of many firms that demand similar raw materials and specialized inputs for their production could easily attract a large quantity of raw material and specialized input suppliers This helps clustered firms, not only to have access to raw materi-als and intermediate input suppliers, but also to minimize transportation costs

of inputs procured More importantly, it helps them reduce the transaction costs of: (i) searching for and reaching input suppliers, (ii) formulating and enforcing contracts, (iii) getting relevant information about the reliability and reputations of their transaction partners, and (iv) verifying the prices and quality of raw materi-als and/or intermediate inputs This is because, as also discussed in McCormick (1999), Schmitz and Nadvi (1999, 2005), and Sonobe and Otsuka (2006), asym-metric information about the reputation of potential trading partners is less per-vasive inside of clusters than outside of them because clustered firms quite often exchange experience and information informally These studies specifically found that clustered firms usually share information about the credit worthiness, per-sonalities, and conduct of trading partners This minimizes adverse selection and moral hazard problems, provided that dishonest behavior can be easily detected and sanctioned within the cluster rather than externally Becker and Murphy (1992) also argued that transaction costs due to adverse selection, moral hazard, and imperfect contract enforcement are relatively lower within industrial clusters where transaction parties are in close proximity to each other

The reduction of transaction costs between traders and producers, in turn, promotes market transactions and facilitates intra-cluster specialization and di-vision of labor Consequently, for firms operating inside an industrial cluster,

the external transaction cost of procuring intermediate inputs is much lower

than the internal transaction cost of producing the input According to Coase

(1937), Williamson (1985), and Wallis and North (1987),6 if the external action costs of procuring (intermediate) inputs are lower than the transaction costs of producing them internally, manufacturers will choose to specialize in the production of the final good, leaving the production of the intermediate inputs for others

trans-The integration of the Coase (1937), Williamson (1985), and Wallis and North

(1987) transaction cost theory with Marshall’s (1920) input market externality

theory states that the cluster induced reduction in external transaction costs sters firms to specialize in particular aspects of production, leaving the auxiliary aspects of production for other firms Today there seems to be a consensus that industrial clustering reduces market transaction costs and hence enhances the

bol-division of labor and specialization, which in turn helps small enterprises to

grow in small steps at acceptable risk (Schmidth and Nadvi 1999) and to increase their productivity and profitability (Smith 1776)

Ruan and Zhang (2009) also empirically showed that the technical, institutional and financial costs of entry into a new manufacturing business is much lower with-

in a cluster than externally because manufacturing firms within industrial clusters would not be obliged to acquire the equipment or technology required for an en-tire production process Instead, they can fully contemplate particular production stages, leaving complementary stages to other firms in the cluster Firms also do not need to hold a large stock of raw materials due to the greater availability of raw materials within clusters at shorter distances Many cluster studies such as (Sonobe and Otsuka 2006) also found that the availability of used machinery and equip-ment markets are more developed in cluster scenarios, further easing financial constraints on cluster firms

In his theory of goods market externality, Marshall (1920) argued that due to

their geographical proximity to output traders, clustered firms have lower portation and transaction costs of goods sold The concentration of a large number

trans-of producers in small geographic area will also attract output traders This again helps the clustered firms to build reputations and trust with output traders, which would help them reduce the transaction costs of searching for and reaching cus-tomers This, in turn eases the financial constraints of clustered firms, which is fre-quently mentioned by firms as a major impediment of growth Clustered firms can

6 For Coase (1937, 1998), Williamson (1985), and Wallis and North (1987) the reduction

of external transaction costs not only ensures the existence of firms, but also determines their growth prospects.

only specialize only in the production of goods, leaving the distribution and sale

of their final products to the intra cluster output traders Industrial clustering also promotes trade based on credit and expedites the exchange of technical, manage-rial and marketing information between the traders and manufacturing firms

Similarly, in his theory of labor market pooling, Marshall (1920) emphasized

the human capital mobilization role of industrial clustering (also see Ellison and Glaeser 2007) Marshall argued that industrial clusters attract workers with cluster specific skills Clustering facilitates matching workers with specific skills to firms that search for workers with such skills, because in a cluster there is a larger pool of workers with diversified skills and a larger number of employers who seek diverse skills This also reduces the likelihood of mismatches between skilled workers and skilled workers needed by firms

Industrial clustering also has positive impacts on the performance of clustered firms, not only by reducing the transaction costs of procuring inputs, product sales, and finding skilled labor, but also by facilitating the diffusion of administra-

tive, technical and marketing knowledge In his theory of localized knowledge

spillover, Marshall (1920) remarked that the “mysteries of trade becomes no

mys-tery in a cluster, rather are as it was in the air.” According to Marshall, intra-cluster intellectual/technological/knowledge spillover is a powerful incentive for firms to locate themselves in proximity to many similar firms Marshall argued that “if one man starts a new idea, it is taken up by others and combined with suggestions of their own; and thus it becomes the source of new ideas.” For Marshall, physical proximity facilitates the exchange of knowledge, leading to innovation

Marshall’s theory of agglomeration externalities was rediscovered by

main-stream economists such as Kenneth Arrow (1962) and Paul Romer (1986) The theory has been refined and further expanded on by other scholars such as Krug-man (1991), Porter (1991), McCormick (1999), Schmitz and Nadvi (1999, 2005), Fujita and Thisse (2002), and Sonobe and Otsuka (2006) These authors share the view that the effects of agglomeration spillover are localized, even though knowl-edge flows freely out of the clustered firms This is because: (i) knowledge spillover within clusters is likely to occur through informal exchange of information, labor turnover, and often from working relationships with traders Hence, the diffusion

of knowledge outside of the cluster is limited (Baptista and Swann 1999); (ii) formation externality diminishes with physical distance (Fujita and Ogawa 1982, Jaffe et al 1993); and (iii) some knowledge can’t be patented or excluded by labor contracts as it is simply embodied in the workforce

in-Schmitz (1989, 1995) and Nadvi (1996) extended Marshall’s (1920) theory of external economies to explicitly account for deliberate joint action by clustered

firms They argued that the cluster advantage is attributed, not only to the

inci-dental external economies (Marshall’s spillover effects), which they call passive

collective efficiency gain, but also to the deliberate pursuit of joint action by

clustered firms, which they call active collective efficiency gain According to

their ‘cluster induced collective efficiency gain hypothesis,’ which has served as

a work horse in cluster analysis since the 1990s, firms operating in an industrial cluster perform much better than firms operating in isolation from non-cluster locations, not only because industrial clusters attract traders, workers with sec-tor/cluster-specific skills, raw material and specialized input suppliers, but also because they creates venues for collaboration For Schmitz and Nadvi (1999) joint action could be initiated by clustered firms themselves or by government and nongovernmental cluster development agents Such joint action could in-clude joint training efforts, joint procurement of raw material, and joint product sales McCormick (1999) augmented the institutional context into the collective efficiency hypothesis

Porter (1991) and Sonobe and Otsuka (2011) emphasized the role of

intra-cluster competition among intra-clustered firms to explain performance gaps among

clustered firms According to these authors intra-cluster supply of a product would exceed its demand over time and hence the price of the product and the profitabil-ity of producing that product begins to decline at some point Consequently, clus-tered firms will be forced to seek new designs and products as well as to upgrade their organization, production methods, and the quality of their products, all of which requires the use of higher quality raw materials, the purchase of additional equipment and machinery, and the employment of more experienced and skilled workers Consequently, the profitability of clustered firms declines until consum-ers recognize and respond to the quality improvement At the later stage, those firms that successfully undertake multifaceted innovations improve the quality of their products and may be able to connect to new markets and thus increase their productivity and profitability

Based on these considerations, the following hypothesis attempts to explain the impacts of clustering among firms

Hypothesis 1: Clustered firms are more innovative, productive and profitable

than non-clustered firms

There is growing recognition, however, in related literature that industrial tering might not nurture all of the aforementioned benefits automatically For example, Schmitz and Nadvi (1999) compared the cases study of McCormick (1998) on industrial clusters in Africa with the case study of Wiejland (1994)

clus-on incipient clusters in Indclus-onesia and found that collective efficiency gains clus-only

emerged under certain circumstances, such as effective sanctions, trust among actors, and when there are strong external trade linkages McCormick (1998) specifically indicated that cluster induced specialization and division of labor may not occur if elements like effective sanctions and trust fail to materialize In the absence of external trade networks, effective sanctions and trust among ac-tors, clustered firms are less likely to outperform non-clustered firms

Sonobe et al (2013) postulated that although industrial clustering facilitates technological spillover through extensive imitation, this property does not con-tribute to the development of clusters when there are no innovations regarding profitable products, markets, and production processes to spillover They assert-

ed that industrial clustering does not encourage Schumpeterian-type innovation, rather it simply facilitates rampant imitation In earlier literature, Chamberlin (1933) asserted that there is a tradeoff between the benefits of agglomeration externalities and the costs of congestion and intense competition among similar spatially concentrated firms Weber (1909) and Marshall (1920) also reached the conclusion that the benefits of agglomeration may cease or fail to materialize and that the forces of de-agglomeration may set in over time As more firms become concentrated within a small geographic area, the cost of strategic inputs such

as land prices/rent and wages are likely to rise, and that firms eventually leave clusters and relocate elsewhere

Similarly, Swann et al (1998) indicated that at the later stage of cluster life cycles, congestion and intense competition might set in and reduce firm prof-itability To put it differently, at the initial stages of cluster life cycles the per unit production cost among clustered firms declines with the entry of addition-

al firms.7 However, at later stages in the life cycle clustered firms begin to face greater competition for strategic inputs such as land (working premises) and raw materials, and hence their production costs will rise with the entry of additional firms to the cluster.8 According to the cluster life cycle hypothesis, the demand (customer proximity, consumer search, information costs, reputation, and infor-mation externalities) and supply conditions (knowledge spillover, information externalities, infrastructural benefits, and greater availability of used equipment and machinery, specialized inputs, and skilled labor) are only superior in a clus-ter relative to isolation at the early stages of the cluster life cycle

More recently, Combes and Duranton (2006) indicated that when firms are spatially concentrated and thus share local labor markets, they face a trade-off

7 This phenomenon is called external economy of scale.

8 This phenomenon is called external diseconomy of scale.

between the benefits of labor pooling and the costs of labor poaching.9 Firms poach workers with specialized skills from rival firms in the cluster by offering better compensation This enables firms to imitate and poach the production practices and marketing strategies of their rivals while simultaneously increas-ing their own productivity, however, it also increases production costs Clustered firms may also be forced to pay higher wages to maintain an experienced work-force and protect their production processes According to Combes and Duran-ton (2006), when firms perceive that the costs of labor poaching are greater than the benefits of labor pooling, they may choose to strategically locate outside of the cluster That is, even if industrial clustering brings collective efficiency gains, benefits may not be an equilibrium outcome.

In their “East Asian model of cluster based industrial development,” Sonobe and Otsuka (2006, 2011) indicated that when more firms enter a cluster to benefit from localization economies, the supply of products in the cluster rises and even-tually exceeds demand This triggers more intense competition among clustered firms and drives less innovative firms out of business despite the advantages of ag-glomeration externalities Lall et al (2003) described the potential cost-increasing effect of clustering through increased competition among clustered firms and con-gestion Rocha and Sternberg (2005) evaluated Germany’s 97 planning regions and found no significant impacts of clusters on firm performance at the regional level Kennedy (1994) studied an Indian tannery cluster and indicated the potential dan-ger of external diseconomies and collective efficiency failures in industrial clusters The net effect of industrial clustering is, therefore, theoretically ambiguous and hence worthy of additional empirical evaluation Likewise the impact of industrial clustering on the welfare of the employees of clustered firms is also complicated and theoretically undetermined

Intra-cluster competition among firms and the fact that much of a firm’s cific production and marketing knowledge (tacit knowledge) are embodied in the firms’ employees motivates clustered firms to poach experienced workers of rival firms by offering better compensation Firms that poach skilled workers do

spe-so because it offers an opportunity to replicate the designs, production ods, and marketing advantages of rival firms Likewise, clustered firms have a strong motivation to retain their experienced workers and therefore must offer relatively better compensation in order to protect their own production and

meth-9 Knowledge flow is associated with labor flow, as some knowledge is embodied in the worker.

marketing advantages This raises intra-cluster equilibrium wages, thus ing the income of the workers in the cluster

improv-More importantly, in an industrial cluster workers who lose their job at an unsuccessful firm can more easily obtain another job from other firms in the cluster This is due to the boom-bust nature of demand for a given product quite often occurs at the firm level and not at the industry level This benefits workers

by mitigating unemployment risk and/or reducing the duration of ment

unemploy-Workers employed by clustered firms also have better opportunities to learn from high performing workers because they can more easily observe and imitate the most productive workers in the cluster (Glaeser and Maré 1994) The high return of cluster-specific skills and the intra-cluster visibility of worker perfor-mance further motivates workers to invest more effort into acquiring additional skills, which in turn provides greater benefits over the long term According to this argument, workers within in an industrial cluster are expected to be more produc-tive as they accumulate more skills that should translate into better compensation relative to workers who work outside of clusters Consequently, employees of clus-tered firms are expected to be better off than those employed by isolated counter-parts Some empirical studies have found a positive and significant relationship between firm productivity and employee welfare For example, Glaeser and Maré (1994) asserted that faster rates of human capital growth in industrial clusters are the key factor in explaining higher labor productivity and higher wages within

a cluster than outside Hence, industrial clustering is believed to boost the ductivity of firms and employee compensation These considerations lead to the following hypothesis:

pro-Hypothesis 2: Clustered firms create more jobs, have greater labor productivity,

and offer relatively better compensation than non-clustered firms

On the other hand, clustered firms have better access to larger pools of workers with specialized skills and therefore can hire experienced workers at a relatively lower wage than non-clustered counterparts Consequently, employees of clus-tered firms (or farms) could receive less compensation than employees of exter-nal firms In addition, firms operating in an industrial cluster can easily terminate poor performing workers and hire higher performing workers because informa-tion about the performance of workers is easily available within clusters Similarly,

to minimize production costs and thereby maximize profit clustered firms can minate workers during the seasons of low production and more easily hire as many workers as they need during peak production seasons owing to the availability of larger pools of workers This raises the risk of unemployment among clustered

ter-workers during low production periods To the contrary, it is relatively very risky for isolated firms to terminate experienced workers during unproductive periods,

as it would be relatively very difficult for them to recruit experienced workers ing peak production periods As a result employees of non-clustered firms may face less risk of being terminated Hence, the impact of industrial clustering on employees’ welfare is theoretically complicated and unpredictable

dur-The complicated nature of the effects of industrial clustering on employees welfare might be much more intense in cases of industrial clusters that hire signif-icant quantities of women, because participation in the labor force by women will not only influence the welfare of female workers and their households through the usual Slutsky effect, but also through the distinguished bargaining effect (Chiappori et al 2011) Within the framework of their cooperative bargaining household model, an improvement in bargaining power among women improves the wellbeing of working women and their household members by weighing re-source allocation more favorably towards goods that appeal to women (Chiapori, 1988; Von Braun et al 1989; Sen, 1989; Thomas 1994; Hoddinott and Haddad, 1995) The additional earned income from employment improves the wellbeing

of working women within clusters and their households, not only by raising the overall resource base of households, but also by enhancing the relative bargain-ing power of women within households In addition, the skills women develop while navigating work environments also help them negotiate more effectively with their husbands regarding decisions about health care, children’s education, household purchases, domestic responsibilities, and time allocation (Ilahi, 2000; Doss 2011)

These considerations lead to the postulation of the following hypothesis:

Hypothesis 3: Female participation in the labor force within an industrial cluster

significantly improves the wellbeing of the working women’s households, not only by raising the overall resource base of households, but also by enhancing the relative bargaining power of women within households

However, this may not be always an equilibrium outcome Due to the fact that ing woman spend time away from home each workday, the quantity of household goods and services may be reduced More importantly the quality of household goods and services, especially child care and food utilization (meal preparation) could decline, especially in countries such as Ethiopia where the gender division

work-of household labor is ubiquitous In addition, increases in the earned income work-of women may also tempt their husbands to divert part of that income from house-hold consumption These factors could deteriorate the welfare of working women

and their households Hence, the net effect of female participation in the labor force is theoretically undetermined and deserves further empirical investigation

1.3 Research Problem and Significance of the Study

The theoretically complicated and ambiguous effects of industrial clustering indicate the significance of the need for conducting rigorous country- and sector-specific case studies, especially in sub-Saharan Africa where empirical micro-econometric evidence regarding the productivity and profitability effects

of industrial clustering is limited The few existing examples of empirical dence regarding the performance of African industrial clusters are incongruous For example, McCormick (1998) reported on the dearth of successful clusters in East Africa, while Sonobe and Otsuka (2006) described the existence of well per-forming clusters in Africa Such meager and mixed evidence from sub-Saharan Africa highlights the need to undertake additional cluster case studies in the countries of the region such as Ethiopia

evi-More strikingly, despite the increase in cluster studies since the seminal work of Marshall (1920), only very few have attempted to quantify the effects of industrial clustering In part, this is due to the fact that it is uncommon to find an opportu-nity to directly compare and contrast the performance of firms operating within

an industrial cluster with those operating outside of a cluster (Visser 1999, Ter Wal and Boschma, 2011, Mano and Suzuki 2013) The coexistence of clustered and non-clustered firms in Ethiopia offers a unique opportunity to quantify the impacts of industrial clustering Following the reorientation of the economy from

a command to a market economy in 1991, a remarkable number of micro- and small-scale manufacturers spontaneously emerged, mostly in major towns Some

of these enterprises were spatially concentrated in a few areas, while others were dispersed elsewhere For example, a large number of micro- and small-scale shoe manufacturers are concentrated in the sub-city of Addis Ketema, while other shoe manufacturers were spread out in other areas This provides a rare opportunity to directly compare and contrast the performance of firms that operate in an indus-trial cluster with similar firms operating outside a cluster

However, a simple comparison of the performance of clustered and tered firms might lead to biased conclusions due to self-selection and endogenous locations of the manufacturing firms sampled The most productive and profitable firms might choose to locate in an industry cluster rather than their performance being the result of the impacts of clustering The study employed appropriate econometric techniques to address selection bias and endogenous location is-sues This approach may help to distinguish the impacts of clustering from other

non-clus-confounding effects and hence makes a significant contribution to the limited isting quantitative evidence regarding the causal relationships with productivity and profitability of industrial clustering in Ethiopia Greater understanding of this phenomenon may yield policy lessons that could also be applicable to similar con-texts; because industrial clusters are ubiquitous in many emerging economies (see Sonobe and Otsuka 2011)

ex-Despite the theoretically complicated and apparently ambiguous effects of dustrial clustering on firm performance, there recently seems to be a growing con-sensus that many small and micro enterprises in some emerging economies such

in-as China have benefited from industrial clustering Thus, the question of whether emerging economies such as Ethiopia could also benefit from industrial cluster-ing requires empirical investigation As indicated in a review of empirical cluster related literature, the magnitude of intra-cluster collective efficiency gains vary by country (Chaudhry 2005) In particular, active collective efficiency gains do not materialize in some countries Hence, it is relevant to investigate the presence or absence and magnitude of cluster induced collective efficiency gains in countries such as Ethiopia, where cluster-based industrial development policies are broadly implemented using limited government resources In other words, quantitatively scrutinizing the true impacts of industrial clustering in poor countries such as Ethiopia is not only of academic interest, rather the insights drawn from this study could serve as inputs for the discussion and formulation of future cluster related policies

Despite the enthusiasm of policy makers and development practitioners for cluster based industrial development, very few studies have attempted to quanti-tatively measure the effects of industrial clustering on firm performance in Ethi-opia One such study compares and contrasts the performance of clustered and non-clustered firms in the export oriented cut flower industry using firm-level panel data (Mano and Suzuki 2013) Another compares the performance of clus-tered and non-clustered handloom weavers using firm-level cross-sectional data (Ayele et al 2009) Both studies found positive associations between industrial clustering and firm performance

Two related studies examined the impacts of industrial clustering in Ethiopia using Central Statistical Agency (CSA) data that were collected for other purposes (Ali and Peerlings 2011, Siba et al 2013) Although these studies provided impor-tant insight into the effects of industrial clustering, important entrepreneur- and firm-specific heterogeneities were omitted from their models They also used a highly contested measure of industrial clustering, local density of employment, because according to such measures a ward with very few large firms could be

counted as an industrial cluster, while a ward with a large number of small firms may not be considered a cluster Specifically, Ali and Peelings (2011) exclusively used profit to measure firm performance This approach is likely to underestimate the positive effects of clustering on firm performance, because clustering can make benefit both firms and workers by increasing productivity and compensation with-out increasing profit They also used cross-sectional data to evaluate the effects of industrial clustering and hence fail to capture the effects of both temporal variation and fixed confounding factors This casts doubt on their findings with respect to the actual effects of industrial clustering Most of these problems could be attrib-uted to the fact that the CSA data used for these studies was collected for other purposes

This study attempted to fill this gap and contribute to the limited quantitative literature on clusters by providing empirical evidence based on recently and pur-posefully collected firm-level data on clustered and non-clustered leather shoe manufacturing MSMEs in Ethiopia This study, unlike most previous meso-level studies, did not ignore intra-cluster firm heterogeneity Instead it explicitly con-trolled for firm- and entrepreneur-specific heterogeneity This study also controlled for the effects of site-specific advantages using firm-level panel data to disentangle the agglomeration gains from the site-specific and urbanization effects

This study also appears to be the first attempt to elucidate the impacts of cluster targeted policy in Ethiopia Following the growing optimism expressed in devel-opment literature concerning the growth, export and employment prospects of MSMEs, the Ethiopian government has also begun to design and broadly imple-ment MSME cluster development programs Many countries such as Germany, Brazil, Japan, South Korea, and France have also implemented cluster-oriented policies to foster industrial development (Martin et al 2010) In many of these countries, however, the impacts of industrial clustering have not been exten-sively evaluated quantitatively Despite optimism among the government and its development partners10 about cluster policy, rigorous empirical evidence regard-ing the effects of cluster development initiatives on firm growth, productivity, and profitability is also lacking in Ethiopia This study, therefore, was intended

to contribute to the limited cluster literature by providing firm-level evidence

on the performance of clustered firms that have benefited from cluster-oriented policy and on a control sample of firms that did not benefit from such policies, both before and after the implementation of the cluster development program

10 Including organizations such as UNIDO and GIZ.

in Ethiopia, controlling for pre-intervention differences and the effects of firm-, site- and entrepreneur-specific factors.

The study also sought to contribute to the meager body of quantitative ter studies regarding the welfare and gender effects of industrial clustering The existing evidence regarding the quantitative effects of industrial clustering on employee welfare is extremely limited, although there seems to be a consensus that the vast majority of manufacturing employment in developed and emerging economies, which is the direct link between industrial clustering and employee welfare, is currently attributed to industrial clusters (Henderson 1997, Combes

clus-2000, Blien et al 2006, Sonobe et al 2013) The theoretically ambiguous welfare effects of industrial clustering also highlight the relevance of undertaking such empirical studies Therefore, to assess whether or not employees of clustered firms are better off (in terms of wages) than those of non-clustered firms, the study compared and contrasted the mean monthly wages of employees while controlling for the effects of worker experience and education According to Mincer’s (1974) model of earnings, education and experience are the major de-terminants of worker wage levels The study also accounted for the effects of the distance of firms from main roads and city centers because firms located farther from city centers and improved roads may be forced to pay a location additional allowances to cover employee transportation to and from work

Given the longer history of the traditional leather shoe cluster, however, getting worker-level baseline information (i.e., for the period prior to the emergence of the leather shoe cluster) based on retrospective surveys is less reliable For exam-ple, Nicola and Gine (2011) indicated that the reliability of retrospective survey data could be highly compromised if respondents are asked about events that are imperfectly recalled or that occurred in the distant past In addition, given the relatively shorter history of workers among present shoe manufacturers, there was

no apparent means of collecting reliable worker-level data from three years earlier Consequently, investigation of the welfare effects of industrial clustering based on cross-sectional data that were collected from leather shoe manufacturers may not

be the most preferable option Because the shoe data do not allow the study to trol for initial welfare difference between the two groups of workers

con-To account for the initial differences between the two groups of workers and to clarify the effects of industrial clustering based on trends over time and other con-founding factors, the study conducted a second cluster case study The relevance

of undertaking the second case study was justified on the grounds that: (i) very little effort has been made to understand the implications of female employment

in an industrial cluster in poor countries such as Ethiopia (Hallward-Driemeier

2011), (ii) the theoretical intra-household welfare effects of female employment

is much more complicated and ambiguous than for male employment (Chiapori

1988, Von Braun et al 1989 Thomas 1990, 1994, Hoddinott and Haddad 1995, Duflo 2003), and (iii) studying the implications of a transition from traditional household-based work to salaried employment is particularly interesting because formal jobs for women are often considered to be important for improving gen-der equity and female empowerment (UN 2013)

The Ethiopian cut flower industry cluster appears to be among the few clusters that best satisfies the aforementioned criteria and therefore is a suitable subject

to elucidate the welfare and gender effects of female employment in an industrial cluster This is because: (i) the cut flower cluster creates significant work op-portunities, mostly for poor marginalized women for whom there are few for-mal employment alternatives, (ii) the cut flower cluster emerged very recently in Ethiopia, (iii) for the overwhelming majority of the employees, this is their first formal job experience, and (iv) the mean tenure of cut flower industry employees

is less than four years because many flower farms only began to recruit workers

in 2008 The fact that flower farms only chose to employ some of the women from among the larger pool of job seekers also provides a unique opportunity to construct a representative counterfactual control sample

Some development literature (von Braun et al 1995) has noted that the tion of employment opportunities can be a pathway out of poverty, however, as discussed in the section above the intra-household welfare effects of female em-ployment is theoretically complicated Thus, predicting the direction of the over-all welfare impact of formal labor market participation in industrial clusters that predominantly employs unskilled poor women is extremely difficult Accord-ingly, this study investigated the welfare and gender impacts of female employ-ment in the Ethiopian cut flower cluster within the collective choice framework, where bargaining plays a key role

crea-1.4 Research Objectives and Questions

The general objectives of the thesis were threefold First, was to quantitatively evaluate the impacts of industrial clustering on the performance of clustered firms using firm-level evidence from Ethiopia among leather shoe manufactur-ing MSMEs Under this general objective, the study sought to address the follow-ing specific research questions

i Why did some firms choose to co-locate among many other rival firms, while others chose to locate elsewhere?